Engineering Vibration Fourth Edition DaniEl J. inman University of Michigan Boston Columbus Indianapolis New York San Francisco Upper Saddle River Amsterdam Cape Town Dubai London Madrid Milan Munich Paris Montréal Toronto Delhi Mexico City São Paulo Sydney Hong Kong Seoul Singapore Taipei Tokyo www.TechnicalBooksPDF.com Editorial Director, Computer Science and Engineering: Marcia J. Horton Acquisitions Editor: Norrin Dias Editorial Assistant: Sandia Rodriguez Senior Managing Editor: Scott Disanno Art Director: Jayne Conte Cover Designer: Bruce Kenselaar Project Manager: Greg Dulles Manufacturing Buyer: Lisa McDowell Senior Marketing Manager: Tim Galligan © 2014, 2008, 2001 by Pearson Education, Inc. Upper Saddle River, New Jersey 07458 All rights reserved. No part of this book may be reproduced, in any form or by any means, without permission in writing from the publisher. The author and publisher have used their best efforts in preparing this book. These efforts include the development, research, and testing of the theories and programs to determine their effectiveness. The author and publisher shall not be liable in any event for incidental or consequential damages in connection with, or arising out of, the furnishing, performance, or use of these theories and programs. Printed in the United States of America 10 9 8 7 6 5 4 3 2 1 Library of Congress Cataloging-in-Publication Data on File ISBN-13: 978-0-13-287169-3 ISBN-10: 0-13-287169-6 www.TechnicalBooksPDF.com Contents PREFACE 1 viii INTRODUCTION TO VIBRATION AND THE FREE RESPONSE 1.1 Introduction to Free Vibration 2 1.2 Harmonic Motion 1.3 Viscous Damping 21 1.4 Modeling and Energy Methods 1.5 Stiffness 46 1.6 Measurement 58 1.7 Design Considerations 1.8 Stability 68 1.9 Numerical Simulation of the Time Response 1 13 31 63 72 1.10 Coulomb Friction and the Pendulum 81 Problems 95 MATLAB Engineering Vibration Toolbox Toolbox Problems 2 115 116 RESPONSE TO HARMONIC EXCITATION 117 2.1 Harmonic Excitation of Undamped Systems 118 2.2 Harmonic Excitation of Damped Systems 130 2.3 Alternative Representations 144 2.4 Base Excitation 2.5 Rotating Unbalance 160 2.6 Measurement Devices 166 151 iii www.TechnicalBooksPDF.com iv Contents 2.7 Other Forms of Damping 170 2.8 Numerical Simulation and Design 2.9 Nonlinear Response Properties 180 188 Problems 197 MATLAB Engineering Vibration Toolbox Toolbox Problems 3 214 214 GENERAL FORCED RESPONSE 216 3.1 Impulse Response Function 217 3.2 Response to an Arbitrary Input 226 3.3 Response to an Arbitrary Periodic Input 235 3.4 Transform Methods 242 3.5 Response to Random Inputs 247 3.6 Shock Spectrum 3.7 Measurement via Transfer Functions 260 3.8 Stability 262 3.9 Numerical Simulation of the Response 267 255 3.10 Nonlinear Response Properties 279 Problems 287 MATLAB Engineering Vibration Toolbox Toolbox Problems 4 301 301 MULTIPLE-DEGREE-OF-FREEDOM SYSTEMS 303 4.1 Two-Degree-of-Freedom Model (Undamped) 4.2 Eigenvalues and Natural Frequencies 318 4.3 Modal Analysis 332 4.4 More Than Two Degrees of Freedom 4.5 Systems with Viscous Damping 356 4.6 Modal Analysis of the Forced Response 340 www.TechnicalBooksPDF.com 362 304 v Contents 4.7 Lagrange’s Equations 369 4.8 Examples 4.9 Computational Eigenvalue Problems for Vibration 389 377 4.10 Numerical Simulation of the Time Response 407 Problems 415 MATLAB Engineering Vibration Toolbox Toolbox Problems 5 433 433 DESIGN FOR VIBRATION SUPPRESSION 5.1 Acceptable Levels of Vibration 436 5.2 Vibration Isolation 442 5.3 Vibration Absorbers 455 5.4 Damping in Vibration Absorption 463 5.5 Optimization 5.6 Viscoelastic Damping Treatments 5.7 Critical Speeds of Rotating Disks 485 435 471 479 Problems 491 MATLAB Engineering Vibration Toolbox Toolbox Problems 6 501 501 DISTRIBUTED-PARAMETER SYSTEMS 502 6.1 Vibration of a String or Cable 504 6.2 Modes and Natural Frequencies 6.3 Vibration of Rods and Bars 519 6.4 Torsional Vibration 525 6.5 Bending Vibration of a Beam 532 6.6 Vibration of Membranes and Plates 544 6.7 Models of Damping 6.8 Modal Analysis of the Forced Response 508 550 www.TechnicalBooksPDF.com 556 vi Contents Problems 566 MATLAB Engineering Vibration Toolbox Toolbox Problems 7 572 572 VIBRATION TESTING AND EXPERIMENTAL MODAL ANALYSIS 7.1 Measurement Hardware 7.2 Digital Signal Processing 579 7.3 Random Signal Analysis in Testing 584 7.4 Modal Data Extraction 7.5 Modal Parameters by Circle Fitting 591 7.6 Mode Shape Measurement 7.7 Vibration Testing for Endurance and Diagnostics 606 7.8 Operational Deflection Shape Measurement 609 573 575 588 596 Problems 611 MATLAB Engineering Vibration Toolbox Toolbox Problems 8 615 616 FINITE ELEMENT METHOD 8.1 Example: The Bar 8.2 Three-Element Bar 8.3 Beam Elements 8.4 Lumped-Mass Matrices 8.5 Trusses 641 8.6 Model Reduction 617 619 625 630 638 646 Problems 649 MATLAB Engineering Vibration Toolbox Toolbox Problems 656 656 APPENDIX A COMPLEX NUMBERS AND FUNCTIONS 657 APPENDIX B LAPLACE TRANSFORMS 663 www.TechnicalBooksPDF.com vii Contents APPENDIX C MATRIX BASICS 668 APPENDIX D THE VIBRATION LITERATURE 680 APPENDIX E LIST OF SYMBOLS 682 APPENDIX F CODES AND WEB SITES 687 APPENDIX G ENGINEERING VIBRATION TOOLBOX AND WEB SUPPORT 688 REFERENCES 690 ANSWERS TO SELECTED PROBLEMS 692 INDEX 699 www.TechnicalBooksPDF.com Preface This book is intended for use in a first course in vibrations or structural dynamics for undergraduates in mechanical, civil, and aerospace engineering or engineering mechanics. The text contains the topics normally found in such courses in accredited engineering departments as set out initially by Den Hartog and refined by Thompson. In addition, topics on design, measurement, and computation are addressed. Pedagogy Originally, a major difference between the pedagogy of this text and competing texts is the use of high level computing codes. Since then, the other authors of vibrations texts have started to embrace use of these codes. While the book is written so that the codes do not have to be used, I strongly encourage their use. These codes (Mathcad®, MATLAB®, and Mathematica®) are very easy to use, at the level of a programmable calculator, and hence do not require any prerequisite courses or training. Of course, it is easier if the students have used one or the other of the codes before, but it is not necessary. In fact, the MATLAB® codes can be copied directly and will run as listed. The use of these codes greatly enhances the student’s understanding of the fundamentals of vibration. Just as a picture is worth a thousand words, a numerical simulation or plot can enable a completely dynamic understanding of vibration phenomena. Computer calculations and simulations are presented at the end of each of the first four chapters. After that, many of the problems assume that codes are second nature in solving vibration problems. Another unique feature of this text is the use of “windows,” which are distributed throughout the book and provide reminders of essential information pertinent to the text material at hand. The windows are placed in the text at points where such prior information is required. The windows are also used to summarize essential information. The book attempts to make strong connections to previous course work in a typical engineering curriculum. In particular, reference is made to calculus, differential equations, statics, dynamics, and strength of materials course work. viii www.TechnicalBooksPDF.com ix Preface WHAT’S NEW IN THIS EDITION Most of the changes made in this edition are the result of comments sent to me by students and faculty who have used the 3rd edition. These changes consist of improved clarity in explanations, the addition of some new examples that clarify concepts, and enhanced problem statements. In addition, some text material deemed outdated and not useful has been removed. The computer codes have also been updated. However, software companies update their codes much faster than the publishers can update their texts, so users should consult the web for updates in syntax, commands, etc. One consistent request from students has been not to reference data appearing previously in other examples or problems. This has been addressed by providing all of the relevant data in the problem statements. Three undergraduate engineering students (one in Engineering Mechanics, one in Biological Systems Engineering, and one in Mechanical Engineering) who had the prerequisite courses, but had not yet had courses in vibrations, read the manuscript for clarity. Their suggestions prompted us to make the following changes in order to improve readability from the student’s perspective: Improved clarity in explanations added in 47 different passages in the text. In addition, two new windows have been added. Twelve new examples that clarify concepts and enhanced problem statements have been added, and ten examples have been modified to improve clarity. Text material deemed outdated and not useful has been removed. Two sections have been dropped and two sections have been completely rewritten. All computer codes have been updated to agree with the latest syntax changes made in MATLAB, Mathematica, and Mathcad. Fifty-four new problems have been added and 94 problems have been modified for clarity and numerical changes. Eight new figures have been added and three previous figures have been modified. Four new equations have been added. Chapter 1: Changes include new examples, equations, and problems. New textual explanations have been added and/or modified to improve clarity based on student suggestions. Modifications have been made to problems to make the problem statement clear by not referring to data from previous problems or examples. All of the codes have been updated to current syntax, and older, obsolete commands have been replaced. Chapter 2: New examples and figures have been added, while previous examples and figures have been modified for clarity. New textual explanations have also been added and/or modified. New problems have been added and older problems modified to make the problem statement clear by not referring to data from previous problems or examples. All of the codes have been updated to current syntax, and older, obsolete commands have been replaced. www.TechnicalBooksPDF.com x Preface Chapter 3: New examples and equations have been added, as well as new problems. In particular, the explanation of impulse has been expanded. In addition, previous problems have been rewritten for clarity and precision. All examples and problems that referred to prior information in the text have been modified to present a more self-contained statement. All of the codes have been updated to current syntax, and older, obsolete commands have been replaced. Chapter 4: Along with the addition of an entirely new example, many of the examples have been changed and modified for clarity and to include improved information. A new window has been added to clarify matrix information. A figure has been removed and a new figure added. New problems have been added and older problems have been modified with the goal of making all problems and examples more self-contained. All of the codes have been updated to current syntax, and older, obsolete commands have been replaced. Several new plots intermixed in the codes have been redone to reflect issues with Mathematica and MATLAB’s automated time step which proves to be inaccurate when using singularity functions. Several explanations have been modified according to students’ suggestions. Chapter 5: Section 5.1 has been changed, the figure replaced, and the example changed for clarity. The problems are largely the same but many have been changed or modified with different details and to make the problems more self-contained. Section 5.8 (Active Vibration Suppression) and Section 5.9 (Practical Isolation Design) have been removed, along with the associated problems, to make room for added material in the earlier chapters without lengthening the book. According to user surveys, these sections are not usually covered. Chapter 6: Section 6.8 has been rewritten for clarity and a window has been added to summarize modal analysis of the forced response. New problems have been added and many older problems restated for clarity. Further details have been added to several examples. A number of small additions have been made to the to the text for clarity. Chapters 7 and 8: These chapters were not changed, except to make minor corrections and additions as suggested by users. Units This book uses SI units. The 1st edition used a mixture of US Customary and SI, but at the insistence of the editor all units were changed to SI. I have stayed with SI in this edition because of the increasing international arena that our engineering graduates compete in. The engineering community is now completely global. For instance, GE Corporate Research has more engineers in its research center in India than it does in the US. Engineering in the US is in danger of becoming the ‘garment’ workers of the next decade if we do not recognize the global work place. Our engineers need to work in SI to be competitive in this increasingly international work place. www.TechnicalBooksPDF.com xi Preface Instructor Support This text comes with a bit of support. In particular, MS PowerPoint presentations are available for each chapter along with some instructive movies. The solutions manual is available in both MS Word and PDF format (sorry, instructors only). Sample tests are available. The MS Word solutions manual can be cut and pasted into presentation slides, tests, or other class enhancements. These resources can be found at www.pearsonhighered.com and will be updated often. Please also email me at daninman@umich.edu with corrections, typos, questions, and suggestions. The book is reprinted often, and at each reprint I have the option to fix typos, so please report any you find to me, as others as well as I will appreciate it. Student Support The best place to get help in studying this material is from your instructor, as there is nothing more educational than a verbal exchange. However, the book was written as much as possible from a student’s perspective. Many students critiqued the original manuscript, and many of the changes in text have been the result of suggestions from students trying to learn from the material, so please feel free to email me (daninman@umich.edu) should you have questions about explanations. Also I would appreciate knowing about any corrections or typos and, in particular, if you find an explanation hard to follow. My goal in writing this was to provide a useful resource for students learning vibration for the first time. ACKNOWLEDGEMENTS The cover photo of the unmanned air vehicle is provided courtesy of General Atomics Aeronautical Systems, Inc., all rights reserved. Each chapter starts with two photos of different systems that vibrate to remind the reader that the material in this text has broad application across numerous sectors of human activity. These photographs were taken by friends, students, colleagues, relatives, and some by me. I am greatly appreciative of Robert Hargreaves (guitar), P. Timothy Wade (wind mill, Presidential helicopter), General Atomics (Predator), Roy Trifilio (bridge), Catherine Little (damper), Alex Pankonien (FEM graphic), and Jochen Faber of Liebherr Aerospace (landing gear). Alan Giles of General Atomics gave me an informative tour of their facilities which resulted in the photos of their products. Many colleagues and students have contributed to the revision of this text through suggestions and questions. In particular, Daniel J. Inman, II; Kaitlyn DeLisi; Kevin Crowely; and Emily Armentrout provided many useful comments from the perspective of students reading the material for the first time. Kaitlyn and Kevin checked all the computer codes by copying them out of the book to www.TechnicalBooksPDF.com xii Preface make sure they ran. My former PhD students Ya Wang, Mana Afshari, and Amin Karami checked many of the new problems and examples. Dr. Scott Larwood and the students in his vibrations class at the University of the Pacific sent many suggestions and corrections that helped give the book the perspective of a nonresearch insitution. I have implemented many of their suggestions, and I believe the book’s explanations are much clearer due to their input. Other professors using the book, Cetin Cetinkaya of Clarkson University, Mike Anderson of the University of Idaho, Joe Slater of Wright State University, Ronnie Pendersen of Aalborg University Esbjerg, Sondi Adhikari of the Universty of Wales, David Che of Geneva College, Tim Crippen of the University of Texas at Tyler, and Nejat Olgac of the University of Conneticut, have provided discussions via email that have led to improvements in the text, all of which are greatly appreciated. I would like to thank the reviewers: Cetin Cetinkaya, Clarkson University; Dr. Nesrin Sarigul-Klijn, University of California–Davis; and David Che, Geneva College. Many of my former PhD students who are now academics cotaught this course with me and also offered many suggestions. Alper Erturk (Georgia Tech), Henry Sodano (University of Florida), Pablo Tarazaga (Virginia Tech), Onur Bilgen (Old Dominian University), Mike Seigler (University of Kentucky), and Armaghan Salehian (University of Waterloo) all contributed to clarity in this text for which I am grateful. I have been lucky to have wonderful PhD students to work with. I learned much from them. I would also like to thank Prof. Joseph Slater of Wright State for reviewing some of the new materials, for writing and managing the associated toolbox, and constantly sending suggestions. Several colleagues from government labs and companies have also written with suggestions which have been very helpful from that perspective of practice. I have also had the good fortune of being sponsored by numerous companies and federal agencies over the last 32 years to study, design, test, and analyze a large variety of vibrating structures and machines. Without these projects, I would not have been able to write this book nor revise it with the appreciation for the practice of vibration, which I hope permeates the text. Last, I wish to thank my family for moral support, a sense of purpose, and for putting up with my absence while writing. DANIEL J. INMAN Ann Arbor, Michigan www.TechnicalBooksPDF.com 1 introduction to Vibration and the Free Response Vibration is the subdiscipline of dynamics that deals with repetitive motion. most of the examples in this text are mechanical or structural elements. However, vibration is prevalent in biological systems and is in fact at the source of communication (the ear vibrates to hear and the tongue and vocal cords vibrate to speak). in the case of music, vibrations, say of a stringed instrument such as a guitar, are desired. On the other hand, in most mechanical systems and structures, vibration is unwanted and even destructive. For example, vibration in an aircraft frame causes fatigue and can eventually lead to failure. an example of fatigue crack is illustrated in the circle in the photo on the bottom left. Everyday experiences are full of vibration and usually ways of mitigating vibration. automobiles, trains, and even some bicycles have devices to reduce the vibration induced by motion and transmitted to the driver. The task of this text is to teach the reader how to analyze vibration using principles of dynamics. This requires the use of mathematics. in fact, the sine function provides the fundamental means of analyzing vibration phenomena. The basic concepts of understanding vibration, analyzing vibration, and predicting the behavior of vibrating systems form the topics of this text. The concepts and formulations presented in the following chapters are intended to provide the skills needed for designing vibrating systems with desired properties that enhance vibration when it is wanted and reduce vibration when it is not. This first chapter examines vibration in its simplest form in which no external force is present (free vibration). This chapter introduces both the important concept of natural frequency and how to model vibration mathematically. The internet is a great source for examples of vibration, and the reader is encouraged to search for movies of vibrating systems and other examples that can be found there. 1 www.TechnicalBooksPDF.com 2 Introduction to Vibration and the Free Response Chap. 1 1.1 INTRODUCTION TO FREE VIBRATION Vibration is the study of the repetitive motion of objects relative to a stationary frame of reference or nominal position (usually equilibrium). Vibration is evident everywhere and in many cases greatly affects the nature of engineering designs. The vibrational properties of engineering devices are often limiting factors in their performance. When harmful, vibration should be avoided, but it can also be extremely useful. In either case, knowledge about vibration—how to analyze, measure, and control it—is beneficial and forms the topic of this book. Typical examples of vibration familiar to most include the motion of a guitar string, the ride quality of an automobile or motorcycle, the motion of an airplane’s wings, and the swaying of a large building due to wind or an earthquake. In the chapters that follow, vibration is modeled mathematically based on fundamental principles, such as Newton’s laws, and analyzed using results from calculus and differential equations. Techniques used to measure the vibration of a system are then developed. In addition, information and methods are given that are useful for designing particular systems to have specific vibrational responses. The physical explanation of the phenomena of vibration concerns the interplay between potential energy and kinetic energy. A vibrating system must have a component that stores potential energy and releases it as kinetic energy in the form of motion (vibration) of a mass. The motion of the mass then gives up kinetic energy to the potential-energy storing device. Engineering is built on a foundation of previous knowledge and the subject of vibration is no exception. In particular, the topic of vibration builds on previous courses in dynamics, system dynamics, strength of materials, differential equations, and some matrix analysis. In most accredited engineering programs, these courses are prerequisites for a course in vibration. Thus, the material that follows draws information and methods from these courses. Vibration analysis is based on a coalescence of mathematics and physical observation. For example, consider a simple pendulum. You may have seen one in a science museum, in a grandfather clock, or you might make a simple one with a string and a marble. As the pendulum swings back and forth, observe that its motion as a function of time can be described very nicely by the sine function from trigonometry. Even more interesting, if you make a free-body diagram of the pendulum and apply Newtonian mechanics to get the equation of motion (summing moments in this case), the resulting equation of motion has the sine function as its solution. Further, the equation of motion predicts the time it takes for the pendulum to repeat its motion. In this example, dynamics, observation, and mathematics all come into agreement to produce a predictive model of the motion of a pendulum, which is easily verified by experiment (physical observation). www.TechnicalBooksPDF.com Sec. 1.1 3 Introduction to Free Vibration This pendulum example tells the story of this text. We propose a series of steps to build on the modeling skills developed in your first courses in statics, dynamics, and strength of materials combined with system dynamics to find equations of motion of successively more complicated systems. Then we will use the techniques of differential equations and numerical integration to solve these equations of motion to predict how various mechanical systems and structures vibrate. The following example illustrates the importance of recalling the methods learned in the first course in dynamics. Example 1.1.1 Derive the equation of motion of the pendulum in Figure 1.1. Fy Fx O O g l l  m m mg (a) (b) Figure 1.1 (a) A schematic of a pendulum. (b) The free-body diagram of (a). Solution Consider the schematic of a pendulum in Figure 1.1(a). In this case, the mass of the rod will be ignored as well as any friction in the hinge. Typically, one starts with a photograph or sketch of the part or structure of interest and is immediately faced with having to make assumptions. This is the “art” or experience side of vibration analysis and modeling. The general philosophy is to start with the simplest model possible (hence, here we ignore friction and the mass of the rod and assume the motion remains in a plane) and try to answer the relevant engineering questions. If the simple model doesn’t agree with the experiment, then make it more complex by relaxing the assumptions until the model successfully predicts physical observation. With the assumptions in mind, the next step is to create a free-body diagram of the system, as indicated in Figure 1.1(b), in order to identify all of the relevant forces. With all the modeled forces identified, Newton’s second law and Euler’s second law are used to derive the equations of motion. In this example Euler’s second law takes the form of summing moments about point O. This yields ΣMO = J𝛂 www.TechnicalBooksPDF.com 4 Introduction to Vibration and the Free Response Chap. 1 where MO denotes moments about the point O, J = ml 2 is the mass moment of inertia of the mass m about the point O, l is the length of the massless rod, and 𝛂 is the angular acceleration vector. Since the problem is really in one dimension, the vector sum of moments equation becomes the single scalar equation $ Jα(t) = - mgl sin θ(t) or ml 2θ(t) + mgl sin θ(t) = 0 Here the moment arm for the force mg is the horizontal distance l sin θ, and the two overdots indicate two differentiations with respect to the time, t. This is a second-order ordinary differential equation, which governs the time response of the pendulum. This is exactly the procedure used in the first course in dynamics to obtain equations of motion. The equation of motion is nonlinear because of the appearance of the sin(θ) and hence difficult to solve. The nonlinear term can be made linear by approximating the sine for small values of θ(t) as sin θ ≈ θ. Then the equation of motion becomes g $ θ(t) + θ(t) = 0 l This is a linear, second-order ordinary differential equation with constant coefficients and is commonly solved in the first course of differential equations (usually the third course in the calculus sequence). As we will see later in this chapter, this linear equation of motion and its solution predict the period of oscillation for a simple pendulum quite accurately. The last section of this chapter revisits the nonlinear version of the pendulum equation. n Since Newton’s second law for a constant mass system is stated in terms of force, which is equated to the mass multiplied by acceleration, an equation of motion with two time derivatives will always result. Such equations require two constants of integration to solve. Euler’s second law for constant mass systems also # yields two time derivatives. Hence the initial position for θ(0) and velocity of θ(0) must be specified in order to solve for θ(t) in Example 1.1.1. The term mgl sin θ is called the restoring force. In Example 1.1.1, the restoring force is gravity, which provides a potential-energy storing mechanism. However, in most structures and machine parts the restoring force is elastic. This establishes the need for background in strength of materials when studying vibrations of structures and machines. As mentioned in the example, when modeling a structure or machine it is best to start with the simplest possible model. In this chapter, we model only systems that can be described by a single degree of freedom, that is, systems for which Newtonian mechanics result in a single scalar equation with one displacement coordinate. The degree of freedom of a system is the minimum number of displacement coordinates needed to represent the position of the system’s mass at any instant of time. For instance, if the mass of the pendulum in Example 1.1.1 were a rigid body, free to rotate about the end of the pendulum as the pendulum swings, the angle of rotation of the mass would define an additional degree of freedom. The problem would then require two coordinates to determine the position of the mass in space, hence two degrees of freedom. On the other hand, if the rod in Figure 1.1 is flexible, www.TechnicalBooksPDF.com Sec. 1.1 5 Introduction to Free Vibration its distributed mass must be considered, effectively resulting in an infinite number of degrees of freedom. Systems with more than one degree of freedom are discussed in Chapter 4, and systems with distributed mass and flexibility are discussed in Chapter 6. The next important classification of vibration problems after degree of freedom is the nature of the input or stimulus to the system. In this chapter, only the free response of the system is considered. Free response refers to analyzing the vibration of a system resulting from a nonzero initial displacement and/or velocity of the system with no external force or moment applied. In Chapter 2, the response of a single-degree-of-freedom system to a harmonic input (i.e., a sinusoidal applied force) is discussed. Chapter 3 examines the response of a system to a general forcing function (impulse or shock loads, step functions, random inputs, etc.), building on information learned in a course in system dynamics. In the remaining chapters, the models of vibration and methods of analysis become more complex. The following sections analyze equations similar to the linear version of the pendulum equation given in Example 1.1.1. In addition, energy dissipation is introduced, and details of elastic restoring forces are presented. Introductions to design, measurement, and simulation are also presented. The chapter ends with the introduction of high-level computer codes (MATLAB®, Mathematica, and Mathcad) as a means to visualize the response of a vibrating system and for making the calculations required to solve vibration problems more efficiently. In addition, numerical simulation is introduced in order to solve nonlinear vibration problems. 1.1.1 The Spring–Mass Model From introductory physics and dynamics, the fundamental kinematical quantities used to describe the motion of a particle are displacement, velocity, and acceleration vectors. In addition, the laws of physics state that the motion of a mass with changing velocity is determined by the net force acting on the mass. An easy device to use in thinking about vibration is a spring (such as the one used to pull a storm door shut, or an automobile spring) with one end attached to a fixed object and a mass attached to the other end. A schematic of this arrangement is given in Figure 1.2. fk m 0 m mg (a) (b) Figure 1.2 A schematic of (a) a single-degree-of-freedom spring–mass oscillator and (b) its free-body diagram. www.TechnicalBooksPDF.com 6 Introduction to Vibration and the Free Response Chap. 1 Ignoring the mass of the spring itself, the forces acting on the mass consist of the force of gravity pulling down (mg) and the elastic-restoring force of the spring pulling back up (fk). Note that in this case the force vectors are collinear, reducing the static equilibrium equation to one dimension easily treated as a scalar. The nature of the spring force can be deduced by performing a simple static experiment. With no mass attached, the spring stretches to the position labeled x0 = 0 in Figure 1.3. As successively more mass is attached to the spring, the force of gravity causes the spring to stretch further. If the value of the mass is recorded, along with the value of the displacement of the end of the spring each time more mass is added, the plot of the force (mass, denoted by m, times the acceleration due to gravity, denoted by g) versus this displacement, denoted by x, yields a curve similar to that illustrated in Figure 1.4. Note that in the region of values for x between 0 and about 20 mm (millimeters), the curve is a straight line. This indicates that for deflections less than 20 mm and forces less than 1000 N (newtons), the force that is applied by the spring to the mass is proportional to the stretch of the spring. The constant of proportionality is the slope of the straight line between 0 and 20 mm. For the particular spring of Figure 1.4, the constant is 50 N>mm, or 5 * 104 N>m. Thus, the equation that describes the force applied by the spring, denoted by fk, to the mass is the linear relationship (1.1) fk = kx The value of the slope, denoted by k, is called the stiffness of the spring and is a property that characterizes the spring for all situations for which the displacement is less than 20 mm. From strength-of-materials considerations, a linear spring of stiffness k stores potential energy of the amount 12 kx2. x0 x1 x2 x3 g Figure 1.3 A schematic of a massless spring with no mass attached showing its static equilibrium position, followed by increments of increasing added mass illustrating the corresponding deflections. fk 103 N 0 20 mm www.TechnicalBooksPDF.com x Figure 1.4 The static deflection curve for the spring of Figure 1.3. Sec. 1.1 7 Introduction to Free Vibration Note that the relationship between fk and x of equation (1.1) is linear (i.e., the curve is linear and fk depends linearly on x). If the displacement of the spring is larger than 20 mm, the relationship between fk and x becomes nonlinear, as indicated in Figure 1.4. Nonlinear systems are much more difficult to analyze and form the topic of Section 1.10. In this and all other chapters, it is assumed that displacements (and forces) are limited to be in the linear range unless specified otherwise. Next, consider a free-body diagram of the mass in Figure 1.5, with the massless spring elongated from its rest (equilibrium or unstretched) position. As in the earlier figures, the mass of the object is taken to be m and the stiffness of the spring is taken to be k. Assuming that the mass moves on a frictionless surface along the x direction, the only force acting on the mass in the x direction is the spring force. As long as the motion of the spring does not exceed its linear range, the sum of the forces in the x direction must equal the product of mass and acceleration. Summing the forces on the free-body diagram in Figure 1.5 along the x direction yields $ $ mx(t) = -kx(t) or mx(t) + kx(t) = 0 (1.2) $ where x(t) denotes the second time derivative of the displacement (i.e., the acceleration). Note that the direction of the spring force is opposite that of the deflection ( + is marked to the right in the figure). As in Example 1.1.1, the displacement vector and acceleration vector are reduced to scalars, since the net force in the y direction is zero (N = mg) and the force in the x direction is collinear with the inertial force. Both the displacement and acceleration are functions of the elapsed time t, as denoted in equation (1.2). Window 1.1 illustrates three types of mechanical systems, which for small oscillations can be described by equation (1.2): a spring–mass system, a rotating shaft, and a swinging pendulum (Example 1.1.1). Other examples are given in Section 1.4 and throughout the book. One of the goals of vibration analysis is to be able to predict the response, or motion, of a vibrating system. Thus it is desirable to calculate the solution to equation (1.2). Fortunately, the differential equation of (1.2) is well known and is covered extensively in introductory calculus and physics texts, as well as in texts on differential equations. In fact, there are a variety of ways to calculate this solution. These are all discussed in some detail in the next section. For now, it is sufficient to present a solution based on physical observation. From experience k y x0 0   m x Friction-free surface kx mg N Rest position (a) (b) Figure 1.5 (a) A single spring–mass system given an initial displacement of x0 from its rest, or equilibrium, position and zero initial velocity. (b) The system’s freebody diagram. www.TechnicalBooksPDF.com 8 Introduction to Vibration and the Free Response Chap. 1 Window 1.1 Examples of Single-Degree-of-Freedom Systems (for small displacements) k k Torsional stiffness Gravity l  length  J (t) m g m x(t) Spring–mass mx  kx  0 (a) Shaft and disk J  k  0 (b) Simple pendulum   (g/l)  0 (c) watching a spring, such as the one in Figure 1.5 (or a pendulum), it is guessed that the motion is periodic, of the form x(t) = A sin(ωnt + ϕ) (1.3) This choice is made because the sine function describes oscillation. Equation (1.3) is the sine function in its most general form, where the constant A is the amplitude, or maximum value, of the displacement; ωn, the angular natural frequency, determines the interval in time during which the function repeats itself; and ϕ, called the phase, determines the initial value of the sine function. As will be discussed in the following sections, the phase and amplitude are determined by the initial state of the system (see Figure 1.7). It is standard to measure the time t in seconds (s). The phase is measured in radians (rad), and the frequency is measured in radians per second (rad >s). As derived in the following equation, the frequency ωn is determined by the physical properties of mass and stiffness (m and k), and the constants A and ϕ are determined by the initial position and velocity as well as the frequency. To see if equation (1.3) is in fact a solution of the equation of motion, it is substituted into equation (1.2). Successive differentiation of the displacement, x(t) # in the form of equation (1.3), yields the velocity, x(t), given by # x(t) = ωnA cos(ωnt + ϕ) (1.4) $ and the acceleration, x(t), given by $ (1.5) x(t) = -ω2nA sin(ωnt + ϕ) www.TechnicalBooksPDF.com Sec. 1.1 9 Introduction to Free Vibration Substitution of equations (1.5) and (1.3) into (1.2) yields -mω2nA sin(ωnt + ϕ) = -kA sin(ωnt + ϕ) Dividing by A and m yields the fact that this last equation is satisfied if ω2n = k , m ωn = or k Am (1.6) Hence, equation (1.3) is a solution of the equation of motion. The constant ωn characterizes the spring–mass system, as well as the frequency at which the motion repeats itself, and hence is called the system’s natural frequency. A plot of the solution x(t) versus time t is given in Figure 1.6. It remains to interpret the constants A and ϕ. The units associated with the notation ωn are rad>s and in older texts natural frequency in these units is often referred to as the circular natural frequency or circular frequency to emphasize that the units are consistent with trigonometric functions and to distinguish this from frequency stated in units of hertz (Hz) or cycles per second, denoted by fn, and commonly used in discussing frequency. The two are related by fn = ωn >2π as discussed in Section 1.2. In practice, the phrase natural frequency is used to refer to either fn or ωn, and the units are stated explicitly to avoid confusion. For example, a common statement is: the natural frequency is 10 Hz, or the natural frequency is 20π rad>s. Recall from differential equations that because the equation of motion is of second order, solving equation (1.2) involves integrating twice. Thus there are two constants of integration to evaluate. These are the constants A and ϕ. The physical significance, or interpretation, of these constants is that they are determined by the initial state of motion of the spring–mass system. Again, recall Newton’s laws, if no force is imparted to the mass, it will stay at rest. If, however, the mass is displaced to a position of x0 at time t = 0, the force kx0 in the spring will result in motion. Also, if the mass is given an initial velocity of v0 at time t = 0, motion will result because x(t) (mm) A 1.5 1 0.5 Time (s) 0 0.5 2 4 6 8 10 12 1 A 1.5 T  2  n www.TechnicalBooksPDF.com Figure 1.6 The response of a simple spring–mass system to an initial displacement of x0 = 0.5 mm and an initial velocity of v0 = 2 12 mm >s. The natural frequency is 2 rad >s and the amplitude is 1.5 mm. The period is T = 2π>ωn = 2π>2 = πs. 10 Introduction to Vibration and the Free Response Chap. 1 of the induced change in momentum. These are called initial conditions and when substituted into the solution (1.3) yield x0 = x(0) = A sin(ωn0 + ϕ) = A sin ϕ (1.7) # v0 = x(0) = ωnA cos(ωn0 + ϕ) = ωnA cos ϕ (1.8) and Solving these two simultaneous equations for the two unknowns A and ϕ yields A = 2ω2n x20 + v20 ωn and ϕ = tan-1 ωn x0 v0 (1.9) as illustrated in Figure 1.7. Here the phase ϕ must lie in the proper quadrant, so care must be taken in evaluating the arc tangent. Thus, the solution of the equation of motion for the spring–mass system is given by x(t) = 2ω2n x20 + v20 ωn x0 b sin aωn t + tan-1 ωn v0 (1.10) and is plotted in Figure 1.6. This solution is called the free response of the system, because no force external to the system is applied after t = 0. The motion of the spring– mass system is called simple harmonic motion or oscillatory motion and is discussed in detail in the following section. The spring–mass system is also referred to as a simple harmonic oscillator, as well as an undamped single-degree-of-freedom system. x0 A x02  v0 n x0 2 v0 n x0 90º  90º v0 v0 n  x0 n A x02  v0 n 2 (b) (a) Figure 1.7 The trigonometric relationships between the phase, natural frequency, and initial conditions. Note that the initial conditions determine the proper quadrant for the phase: (a) for a positive initial position and velocity, (b) for a negative initial position and a positive initial velocity. www.TechnicalBooksPDF.com Sec. 1.1 11 Introduction to Free Vibration Example 1.1.2 The phase angle ϕ describes the relative shift in the sinusoidal vibration of the spring– mass system resulting from the initial displacement, x0. Verify that equation (1.10) satisfies the initial condition x(0) = x0. Solution Substitution of t = 0 in equation (1.10) yields 2ω2n x20 + v20 ωn x0 sin atan-1 b ωn v0 x(0) = A sin ϕ = Figure 1.7 illustrates the phase angle ϕ defined by equation (1.9). This right triangle is used to define the sine and tangent of the angle ϕ. From the geometry of a right triangle, and the definitions of the sine and tangent functions, the value of x(0) is computed to be x(0) = ωn x0 2ω2n x20 + v20 = x0 ωn 2ω2n x20 + v20 which verifies that the solution given by equation (1.10) is consistent with the initial displacement condition. n Example 1.1.3 A vehicle wheel, tire, and suspension assembly can be modeled crudely as a singledegree-of-freedom spring–mass system. The (unsprung) mass of the assembly is measured to be about 30 kilograms (kg). Its frequency of oscillation is observed to be 10 Hz. What is the approximate stiffness of the suspension assembly? Solution so that The relationship between frequency, mass, and stiffness is ωn = 2k>m, k = mω2n = (30 kg) a10 cycle 2π rad 2 # b = 1.184 * 105 N>m s cycle This provides one simple way to estimate the stiffness of a complicated device. This stiffness could also be estimated by using a static deflection experiment similar to that suggested by Figures 1.3 and 1.4. n Example 1.1.4 Compute the amplitude and phase of the response of a system with a mass of 2 kg and a stiffness of 200 N>m, to the following initial conditions: a) x0 = 2 mm and v0 = 1 mm>s b) x0 = -2 mm and v0 = 1 mm>s c) x0 = 2 mm and v0 = - 1 mm>s Compare the results of these calculations. www.TechnicalBooksPDF.com 12 Introduction to Vibration and the Free Response Chap. 1 Solution First, compute the natural frequency, as this does not depend on the initial conditions and will be the same in each case. From equation (1.6): ωn = 200 N>m k = = 10 rad>s Bm B 2 kg Next, compute the amplitude, as it depends on the squares of the initial conditions and will be the same in each case. From equation (1.9): A = 2ω2n x20 + v20 2102 # 22 + 12 = = 2.0025 mm ωn 10 Thus the difference between the three responses in this example is determined only by the phase. Using equation (1.9) and referring to Figure 1.7 to determine the proper quadrant, the following yields the phase information for each case: a) ϕ = tan-1 a (10 rad>s) (2 mm) ωn x0 b = tan-1 a b = 1.521 rad (or 87.147°) v0 1 mm>s which is in the first quadrant. b) ϕ = tan-1 a (10 rad>s) ( -2 mm) ωn x0 b = tan-1 a b = - 1.521 rad (or -87.147°) v0 1 mm>s which is in the fourth quadrant. c) ϕ = tan-1 a (10 rad>s)(2 mm) ωn x0 b = tan-1 a b = ( - 1.521 + π) rad (or 92.85°) v0 - 1 mm>s which is in the second quadrant (position positive, velocity negative places the angle in the second quadrant in Figure 1.7 requiring that the raw calculation be shifted 180°). Note that if equation (1.9) is used without regard to Figure 1.7, parts b and c would result in the same answer (which makes no sense physically as the responses each have different starting points). Thus in computing the phase it is important to consider which quadrant the angle should lie in. Fortunately, some calculators and some codes use an arc tangent function, which corrects for the quadrant (for instance, MATLAB uses the atan2(w0*x0, v0) command). The tan(ϕ) can be positive or negative. If the tangent is positive, the phase angle is in the first or third quadrant. If the sign of the initial displacement is positive, the phase angle is in the first quadrant. If the sign is negative or the initial displacement is negative, the phase angle is in the third quadrant. If on the other hand the tangent is negative, the phase angle is in the second or fourth quadrant. As in the previous case, by examining the sign of the initial displacement, the proper quadrant can be determined. That is, if the sign is positive, the phase angle is in the second quadrant, and if the sign is negative, the phase angle is in the fourth quadrant. The remaining possibility is that the tangent is equal to zero. In this case, the phase angle is either zero or 180°. The initial velocity determines which quadrant is correct. If the initial displacement is zero and if the initial velocity is zero, then the phase angle is zero. If on the other hand the initial velocity is negative, the phase angle is 180°. n www.TechnicalBooksPDF.com Sec. 1.2 13 Harmonic Motion The main point of this section is summarized in Window 1.2. This illustrates harmonic motion and how the initial conditions determine the response of such a system. Window 1.2 Summary of the Description of Simple Harmonic Motion Displacement, x(t) 2 T  n Period  Slope here is v0 Amplitude 1 A  n x0 Initial displacement n2 x02  v02 Time, t  n Maximum velocity  1 x(t)   n   tan1 n2 x02 + v02 nx0 v0 sin(nt  ) v0  initial velocity 1.2 HARMONIC MOTION The fundamental kinematic properties of a particle moving in one dimension are displacement, velocity, and acceleration. For the harmonic motion of a simple spring–mass system, these are given by equations (1.3), (1.4), and (1.5), respectively. These equations reveal the different relative amplitudes of each quantity. For systems with natural frequency larger than 1 rad >s, the relative amplitude of the velocity response is larger than that of the displacement response by a multiple of ωn, and the acceleration response is larger by a multiple of ω2n. For systems with frequency less than 1, the velocity and acceleration have smaller relative amplitudes than the displacement. Also note that the velocity is 90° (or π >2 radians) out of phase with the position [i.e., sin(ωnt + π >2 + ϕ) = cos(ωnt + ϕ)], while the acceleration is 180° out of phase with the position and 90° out of phase with the velocity. This is summarized and illustrated in Window 1.3. www.TechnicalBooksPDF.com 14 Introduction to Vibration and the Free Response Chap. 1 Window 1.3 The Relationship between Displacement, Velocity, and Acceleration for Simple Harmonic Motion A Displacement x(t)  A sin (nt  ) 0 t A A Velocity . x(t)  nA cos (nt  ) 0 t A 2A Acceleration .. x(t)  2n A sin (nt  ) 0 2A www.TechnicalBooksPDF.com t Sec. 1.2 15 Harmonic Motion The angular natural frequency, ωn, used in equations (1.3) and (1.10), is measured in radians per second and describes the repetitiveness of the oscillation. As indicated in Window 1.2, the time the cycle takes to repeat itself is the period, T, which is related to the natural frequency by T = 2π rad 2π = s ωn ωn rad>s (1.11) This results from the elementary definition of the period of a sine function. The frequency in hertz (Hz), denoted by fn, is related to the frequency in radians per second, denoted by ωn: fn = ωn rad>s ωn cycles ωn ωn = = = (Hz) 2π 2π rad>cycle 2π s 2π (1.12) Equation (1.2) is exactly the same form of differential equation as the linear pendulum equation of Example 1.1.1 and of the shaft and disk of Window 1.1(b). As such, the pendulum will have exactly the same form of solution as equation (1.3), with frequency ωn = g rad>s Al The solution of the pendulum equation thus predicts that the period of oscillation of the pendulum is T = l 2π = 2π s ωn Ag where the non-italic s denotes seconds. This analytical value of the period can be checked by measuring the period of oscillation of a pendulum with a simple stopwatch. The period of the disk and shaft system of Window 1.1 will have a frequency and period of ωn = k rad>s AJ and T = 2π J s Ak respectively. The concept of frequency of vibration of a mechanical system is the single most important physical concept (and number) in vibration analysis. Measurement of either the period or the frequency allows validation of the analytical model. (If you made a 1-meter pendulum, the period would be about 2 s. This is something you could try at home.) As long as the only disturbance to these systems is a set of nonzero initial conditions, the system will respond by oscillating with frequency ωn and period T. For www.TechnicalBooksPDF.com 16 Introduction to Vibration and the Free Response Chap. 1 the case of the pendulum, the longer the pendulum, the smaller the frequency and the longer the period. That’s why in museum demonstrations of a pendulum, the length is usually very large so that T is large and one can easily see the period (also a pendulum is usually used to illustrate the earth’s precession; Google the phrase Foucault Pendulum). Example 1.2.1 Consider a small spring about 30 mm (or 1.18 in) long, welded to a stationary table (ground) so that it is fixed at the point of contact, with a 12-mm (or 0.47-in) bolt welded to the other end, which is free to move. The mass of this system is about 49.2 * 10−3 kg (equivalent to about 1.73 ounces). The spring stiffness can be measured using the method suggested in Figure 1.4 and yields a spring constant of k = 857.8 N>m. Calculate the natural frequency and period. Also determine the maximum amplitude of the response if the spring is initially deflected 10 mm. Assume that the spring is oriented along the direction of gravity as in Window 1.1. (Ignore the effect of gravity; see below.) Solution From equation (1.6) the natural frequency is ωn = 857.8 N>m k = = 132 rad>s Bm B 49.2 * 10-3 kg In hertz, this becomes fn = ωn = 21 Hz 2π The period is T = 2π 1 = = 0.0476 s ωn fn To determine the maximum value of the displacement response, note from Figure 1.6 that this corresponds to the value of the constant A. Assuming that no initial velocity is given to the spring (v0 = 0), equation (1.9) yields x(t)max = A = 2ω2n x20 + v20 = x0 = 10 mm ωn Note that the maximum value of the velocity response is ωnA or ωnx0 = 1320 mm>s and the acceleration response has maximum value Since v0 = 0, the phase is ϕ = tan−1(ωnx0 >0) = π>2, or 90°. Hence, in this case, the response is x(t) = 10 sin(132t + π>2) = 10 cos(132t) mm. n ω2nA = ω2nx0 = 174.24 * 103 mm>s2 Does gravity matter in spring problems? The answer is no, if the system oscillates in the linear region. Consider the spring of Figure 1.3 and let a mass of value m extend www.TechnicalBooksPDF.com Sec. 1.2 17 Harmonic Motion the spring. Let Δ denote the distance deflected in this static experiment (Δis called the static deflection); then the force acting upon the mass is kΔ. From static equilibrium the forces acting on the mass must be zero so that (taking positive down in the figure) mg - kΔ = 0 Next, sum the forces along the vertical for the mass at some point x and apply Newton’s law to get $ mx(t) = -k(x + Δ) + mg = -kx + mg - Δk Note the sign on the spring term is negative because the spring force opposes the motion, which is taken here as positive down. The last two terms add to zero (mg - kΔ = 0) because of the static equilibrium condition, and the equation of motion becomes $ mx(t) + kx(t) = 0 Thus gravity does not affect the dynamic response. Note x(t) is measured from the elongated (or compressed if upside down) position of the spring–mass system, that is, from its rest position. This is discussed again using energy methods in Figure 1.14. Example 1.2.2 (a) A pendulum in Brussels swings with a period of 3 seconds. Compute the length of the pendulum. (b) At another location, assume the length of the pendulum is known to be 2 meters and suppose the period is measured to be 2.839 seconds. What is the acceleration due to gravity at that location? Solution The relationship between period and natural frequency is given in equation (1.11). (a) Substitution of the value of natural frequency for a pendulum and solving for the length of the pendulum yields T = (9.811 m>s2 )(3)2s2 gT 2 g 2π 4π2 = = 2.237 m 1 ω2n = = 2 1 l = 2 ωn l T 4π 4π2 Here the value of g = 9.811 m>s2 is used, as that is the value it has in Brussels (at 51° latitude and an altitude of 102 m). (b) Next, manipulate the pendulum period equation to solve for g. This yields g l = 4π2 4π2 4π2 l = 1 g = (2)m = 9.796 m>s2 T2 T2 (2.839)2 s2 This is the value of the acceleration due to gravity in Denver, Colorado, United States (at an altitude 1638 m and latitude 40°). These sorts of calculations are usually done in high school science classes but are repeated here to underscore the usefulness of the concept of natural frequency and period in terms of providing information about the vibration system’s physical properties. In addition, this example serves to remind the reader of a familiar vibration phenomenon. n www.TechnicalBooksPDF.com 18 Introduction to Vibration and the Free Response Chap. 1 The solution given by equation (1.10) was developed assuming that the response should be harmonic based on physical observation. The form of the response can also be derived by a more analytical approach following the theory of elementary differential equations (see, e.g., Boyce and DiPrima, 2009). This approach is reviewed here and will be generalized in later sections and chapters to solve for the response of more complicated systems. Assume that the solution x(t) is of the form x(t) = ae λt (1.13) where a and λ are nonzero constants to be determined. Upon successive differen# $ tiation, equation (1.13) becomes x(t) = λae λt and x(t) = λ2ae λt. Substitution of the assumed exponential form into equation (1.2) yields mλ2ae λt + kae λt = 0 λt (1.14) λt Since the term ae is never zero, expression (1.14) can be divided by ae to yield mλ2 + k = 0 (1.15) Solving this algebraically results in λ = { k k - = { j = {ωn j A m Am (1.16) where j = 1-1 is the imaginary number and ωn = 1k>m is the natural frequency as before. Note that there are two values for λ, λ = + ωnj and λ = - ωn j, because the equation for λ is of second order. This implies that there must be two solutions of equation (1.2) as well. Substitution of equation (1.16) into equation (1.13) yields that the two solutions for x(t) are x(t) = a1e +jωnt and x(t) = a2e -jωnt (1.17) Since equation (1.2) is linear, the sum of two solutions is also a solution; hence, the response x(t) is of the form x(t) = a1e +jωnt + a2e -jωnt (1.18) where a1 and a2 are complex-valued constants of integration. The Euler relations for trigonometric functions state that 2j sin θ = (eθj - e–θj) and 2 cos θ = (eθj + e–θj), where j = 1-1. [See Appendix A, equations (A.18), (A.19), and (A.20), as well as Window 1.5.] Using the Euler relations, equation (1.18) can be written as x(t) = A sin(ωnt + ϕ) (1.19) where A and ϕ are real-valued constants of integration. Note that equation (1.19) is in agreement with the physically intuitive solution given by equation (1.3). The relationships among the various constants in equations (1.18) and (1.19) are given in Window 1.4. Window 1.5 illustrates the use of Euler relations for deriving harmonic functions from exponentials for the underdamped case. www.TechnicalBooksPDF.com Sec. 1.2 19 Harmonic Motion Window 1.4 Three Equivalent Representations of Harmonic Motion $ The solution of mx + kx = 0 subject to nonzero initial conditions can be written in three equivalent ways. First, the solution can be written as x(t) = a1e jωnt + a2e -jωnt, ωn = k , j = 1-1 Am where a1 and a2 are complex-valued constants. Second, the solution can be written as x(t) = A sin(ωnt + ϕ) where A and ϕ are real-valued constants. Last, the solution can be written as x(t) = A1 sin ωnt + A2 cos ωnt where A1 and A2 are real-valued constants. Each set of two constants is determined by the initial conditions, x0 and v0. The various constants are related by the following: A = 2A21 + A22 ϕ = tan-1 a A1 = (a1 - a2)j a1 = A2 - A1j 2 A2 b A1 A2 = a1 + a2 a2 = A2 + A1j 2 all of which follow from trigonometric identities and Euler’s formulas. Note that a1 and a2 are a complex conjugate pair, so that A1 and A2 are both real numbers provided that the initial conditions are real valued, as is normally the case. Often when computing frequencies from equation (1.16) such as λ2 = -4, there is a temptation to write that the frequency is ωn = {2. This is incorrect because the { sign is used up when the Euler relation is used to obtain the function sin ωnt from the exponential form. The concept of frequency is not defined until it appears in the argument of the sine function and, as such, is always positive. Precise terminology is useful in discussing an engineering problem, and the subject of vibration is no exception. Since the position, velocity, and acceleration change continually with time, several other quantities are used to discuss vibration. The peak value, defined as the maximum displacement, or magnitude A of equation (1.9), is www.TechnicalBooksPDF.com 20 Introduction to Vibration and the Free Response Chap. 1 often used to indicate the region in space in which the object vibrates. Another quantity useful in describing vibration is the average value, denoted by x‾, and defined by T 1 x‾ = lim x(t) dt T S ∞ T L0 (1.20) Note that the average value of x(t) = A sin ωnt over one period of oscillation is zero. Since the square of displacement is associated with a system’s potential energy, the average of the displacement squared is sometimes a useful vibration property to discuss. The mean-square value (or variance) of the displacement x(t), denoted by x‾2, is defined by T 1 x2(t) dt x‾2 = lim T S ∞ T L0 (1.21) The square root of this value, called the root mean-square (rms) value, is commonly used in specifying vibration. Because the peak value of the velocity and acceleration are multiples of the natural frequency times the displacement amplitude [i.e., equations (1.3)–(1.5)], these three peak values often differ in value by an order of magnitude or more. Hence, logarithmic scales are often used. A common unit of measurement for vibration amplitudes and rms values is the decibel (dB). The decibel was originally defined in terms of the base 10 logarithm of the power ratio of two electrical signals, or as the ratio of the square of the amplitudes of two signals. Following this idea, the decibel is defined as dB K 10 log 10 a x1 x1 2 b = 20 log 10 x0 x0 (1.22) Here the signal x0 is a reference signal. The decibel is used to quantify how far the measured signal x1 is above the reference signal x0. Note that if the measured signal is equal to the reference signal, then this corresponds to 0 dB. The decibel is used extensively in acoustics to compare sound levels. Using a dB scale expands or compresses vibration response information for convenience in graphical representation. Example 1.2.3 Consider a 2-meter long pendulum placed on the moon and given an initial angular displacement of 0.2 rad and zero initial velocity. Calculate the maximum angular velocity and the maximum angular acceleration of the swinging pendulum (note that gravity on the earth’s moon is gm = g>6, where g is the acceleration due to gravity on earth). Solution From Example 1.1.1 the equation of motion of a pendulum is gm $ θ(t) + θ(t) = 0 l www.TechnicalBooksPDF.com Sec. 1.3 21 Viscous Damping This equation is of the same form as equation (1.2) and hence has a solution of the form θ(t) = A sin(ωnt + ϕ), ωn = gm A l From equation (1.9) the amplitude is given by ω2nx20 + v20 A = C ω2n = x0 = 0.2 rad From Window 1.3 the maximum velocity is just ωnA or vmax = ωn A = 9.8>6 gm (0.2) = (0.2) = 0.18 rad>s B 2 B l The maximum acceleration is amax = ω2nA = gm l A = 9.8>6 2 (0.2) = 0.163 rad>s2 n Frequencies of concern in mechanical vibration range from fractions of a hertz to several thousand hertz. Amplitudes range from micrometers up to meters (for systems such as tall buildings). According to Mansfield (2005), human beings are more sensitive to acceleration than displacement and easily perceive vibration around 5 Hz at about 0.01 m>s2 (about 0.01 mm). Horizontal vibration is easy to experience near 2 Hz. Work attempting to characterize comfort levels for human vibrations is still ongoing. 1.3 VISCOUS DAMPING The response of the spring–mass model (Section 1.1) predicts that the system will oscillate indefinitely. However, everyday observation indicates that freely oscillating systems eventually die out and reduce to zero motion. This observation suggests that the model sketched in Figure 1.5 and the corresponding mathematical model given by equation (1.2) need to be modified to account for this decaying motion. The choice of a representative model for the observed decay in an oscillating system is based partially on physical observation and partially on mathematical convenience. The theory of differential equations suggests that adding a term to equation (1.2) # of the form cx(t), where c is a constant, will result in a solution x(t) that dies out. Physical observation agrees fairly well with this model and is used successfully to model the damping, or decay, in a variety of mechanical systems. This type of damping, called viscous damping, is described in detail in this section. www.TechnicalBooksPDF.com 22 Introduction to Vibration and the Free Response Seal Mounting point Mounting point Case Piston Chap. 1 Oil Orifice x(t) # Figure 1.8 A schematic of a dashpot that produces a damping force fc(t) = cx(t), where x(t) is the motion of the case relative to the piston. While the spring forms a physical model for storing potential energy and hence causing vibration, the dashpot, or damper, forms the physical model for dissipating energy and thus damping the response of a mechanical system. An example dashpot consists of a piston fit into a cylinder filled with oil as indicated in Figure 1.8. This piston is perforated with holes so that motion of the piston in the oil is possible. The laminar flow of the oil through the perforations as the piston moves causes a damping force on this piston. The force is proportional to the velocity of the piston in a direction opposite that of the piston motion. This damping force, denoted by fc, has the form # fc = cx(t) (1.23) where c is a constant of proportionality related to the oil viscosity. The constant c, called the damping coefficient, has units of force per velocity, or N s>m, as it is customarily written. However, following the strict rules of SI units, the units on damping can be reduced to kg>s, which states the units on damping in terms of the fundamental (also called basic) SI units (mass, time, and length). In the case of the oil-filled dashpot, the constant c can be determined by fluid principles. However, in most cases, fc is provided by equivalent effects occurring in the material forming the device. A good example is a block of rubber (which also provides stiffness fk) such as an automobile motor mount, or the effects of air flowing around an oscillating mass. In all cases in which the damping force fc is proportional to velocity, the schematic of a dashpot is used to indicate the presence of this force. The schematic is illustrated in Figure 1.9. Unfortunately, the damping coefficient of a system cannot be measured as simply as the mass or stiffness of a system can be. This is pointed out in Section 1.6. Using a simple force balance on the mass of Figure 1.9 in the x direction, the equation of motion for x(t) becomes $ mx = -fc - fk (1.24) or $ # mx(t) + cx(t) + kx(t) = 0 www.TechnicalBooksPDF.com (1.25) Sec. 1.3 23 Viscous Damping y x(t) k x fk m c Friction-free surface mg fc N (a) (b) Figure 1.9 (a) The schematic of a single-degree-of-freedom system with viscous damping indicated by a dashpot and (b) the corresponding free-body diagram. # subject to the initial conditions x(0) = x0 and x(0) = v0. The forces fc and fk are negative in equation (1.24) because they oppose the motion (positive to the right). Equation (1.25) and Figure 1.9, referred to as a damped single-degree-of-freedom system, form the topic of Chapters 1 through 3. To solve the damped system of equation (1.25), the same method used for solving equation (1.2) is used. In fact, this provides an additional reason to choose # fc to be of the form cx. Let x(t) have the form given in equation (1.13), x(t) = ae λt. Substitution of this form into equation (1.25) yields (mλ2 + cλ + k) ae λt = 0 Again, ae λt (1.26) ≠ 0, so that this reduces to a quadratic equation in λ of the form mλ2 + cλ + k = 0 (1.27) called the characteristic equation. This is solved using the quadratic formula to yield the two solutions λ1, 2 = - 1 c { 2c 2 - 4km 2m 2m (1.28) Examination of this expression indicates that the roots λ will be real or complex, depending on the value of the discriminant, c 2 - 4km. As long as m, c, and k are positive real numbers, λ1 and λ2 will be distinct negative real numbers if c 2 - 4km 7 0. On the other hand, if this discriminant is negative, the roots will be a complex conjugate pair with a negative real part. If the discriminant is zero, the two roots λ1 and λ2 are equal negative real numbers. Note that equation (1.15) represents the characteristic equation for the special undamped case (i.e., c = 0). In examining these three cases, it is both convenient and useful to define the critical damping coefficient, ccr, by ccr = 2mωn = 21km www.TechnicalBooksPDF.com (1.29) 24 Introduction to Vibration and the Free Response Chap. 1 where ωn is the undamped natural frequency in rad>s. Furthermore, the nondimensional number ζ, called the damping ratio, defined by ζ = c c c = = ccr 2mωn 21km (1.30) can be used to characterize the three types of solutions to the characteristic equation. Rewriting the roots given by equation (1.28) yields λ1, 2 = -ζωn { ωn 2ζ2 - 1 (1.31) where it is now clear that the damping ratio ζ determines whether the roots are complex or real. This in turn determines the nature of the response of the damped single-degree-of-freedom system. For positive mass, damping, and stiffness coefficients, there are three cases, which are delineated next. 1.3.1 Underdamped Motion In this case, the damping ratio ζ is less than 1 (0 6 ζ 6 1) and the discriminant of equation (1.31) is negative, resulting in a complex conjugate pair of roots. Factoring out ( -1) from the discriminant in order to clearly distinguish that the second term is imaginary yields 2ζ2 - 1 = 2(1 - ζ2)( -1) = 21 - ζ2 j (1.32) where j = 1-1. Thus the two roots become λ1 = -ζωn - ωn 21 - ζ2 j (1.33) λ2 = -ζωn + ωn 21 - ζ2 j (1.34) and Following the same argument as that made for the undamped response of equation (1.18), the solution of (1.25) is then of the form x(t) = e -ζωnt ( a1e j21 - ζ ωnt + a2e -j21 - ζ ωnt ) 2 2 (1.35) where a1 and a2 are arbitrary complex-valued constants of integration to be determined by the initial conditions. Using the Euler relations (see Window 1.5), this can be written as x(t) = Ae -ζωnt sin(ωdt + ϕ) (1.36) where A and ϕ are constants of integration and ωd, called the damped natural frequency, is given by ωd = ωn 21 - ζ2 in units of rad>s. www.TechnicalBooksPDF.com (1.37) Sec. 1.3 25 Viscous Damping Window 1.5 Euler Relations and the Underdamped Solution $ # An underdamped solution of mx + cx + kx = 0 to nonzero initial conditions is of the form x(t) = a1e λ1t + a2e λ2t where λ1 and λ2 are complex numbers of the form where ωn = 2k>m, ζ = c >(2mωn), ωd = ωn 21 - ζ2, and j = 1-1. The two constants a1 and a2 are complex numbers and hence represent four unknown constants rather than the two constants of integration required to solve a second-order differential equation. This demands that the two complex numbers a1 and a2 be conjugate pairs so that x(t) depends only on two undetermined constants. Substitution of the foregoing values of λi into the solution x(t) yields λ1 = -ζωn + ωdj and λ2 = -ζωn - ωdj x(t) = e -ζωnt ( a1e ωd jt + a2e -ωd jt ) Using the Euler relations eϕj = cos ϕ + j sin ϕ and e–ϕj = cos ϕ - j sin ϕ, x(t) becomes x(t) = e -ζωnt 3 (a1 + a2) cos ωdt + j(a1 - a2) sin ωdt4 Choosing the real numbers A2 = a1 + a2 and A1 = (a1 - a2)j, this becomes x(t) = e -ζωnt(A1 sin ωdt + A2 cos ωdt) which is real valued. Defining the constant A = 2A21 + A22 and the angle ϕ = tan−1(A2 >A1) so that A1 = A cos ϕ and A2 = A sin ϕ, the form of x(t) becomes [recall that sin a cos b + cos a sin b = sin(a + b)] x(t) = Ae -ζωnt sin (ωdt + ϕ) where A and ϕ are the constants of integration to be determined from the initial conditions. Complex numbers are reviewed in Appendix A. The constants A and ϕ are evaluated using the initial conditions in exactly the same fashion as they were for the undamped system as indicated in equations (1.7) and (1.8). Set t = 0 in equation (1.36) to get x0 = A sin ϕ. Differentiating (1.36) yields # x(t) = -ζωn Ae -ζωnt sin(ωdt + ϕ) + ωd Ae -ζωnt cos(ωd t + ϕ) Let t = 0 and A = x0 >sin ϕ in this last expression to get # x(0) = v0 = -ζωn x0 + x0ωd cot ϕ www.TechnicalBooksPDF.com 26 Introduction to Vibration and the Free Response Chap. 1 Solving this last expression for ϕ yields tan ϕ = x0ωd v0 + ζωn x0 With this value of ϕ, the sine becomes sin ϕ = x0ωd 2 ( v0 + ζωn x0 ) 2 + ( x0ωd ) 2 Thus the value of A and ϕ are determined to be A = (v0 + ζωnx0)2 + (x0ωd)2 B ω2d , ϕ = tan-1 x0ωd v0 + ζωnx0 (1.38) where x0 and v0 are the initial displacement and velocity. A plot of x(t) versus t for this underdamped case is given in Figure 1.10. Note that the motion is oscillatory with exponentially decaying amplitude. The damping ratio ζ determines the rate of decay. The response illustrated in Figure 1.10 is exhibited in many mechanical systems and constitutes the most common case. As a check to see that equation (1.38) is reasonable, note that if ζ = 0 in the expressions for A and ϕ, the undamped relations of equation (1.9) result. Displacement (mm) 1.0 Time (s) 0.0 1.0 Figure 1.10 The response of an underdamped system: 0 6 ζ 6 1. 1.3.2 Overdamped Motion In this case, the damping ratio is greater than 1 (ζ 7 1). The discriminant of equation (1.31) is positive, resulting in a pair of distinct real roots. These are λ1 = -ζωn - ωn 2ζ2 - 1 (1.39) λ2 = -ζωn + ωn 2ζ2 - 1 (1.40) and www.TechnicalBooksPDF.com Sec. 1.3 27 Viscous Damping The solution of equation (1.25) then becomes x(t) = e -ζωnt ( a1e -ωn2ζ 2 - 1t + a2e +ωn2ζ 2 - 1t ) (1.41) which represents a nonoscillatory response. Again, the constants of integration a1 and a2 are determined by the initial conditions indicated in equations (1.7) and (1.8). In this nonoscillatory case, the constants of integration are real valued and are given by a1 = -v0 + ( -ζ + 2ζ2 - 1 ) ωnx0 2ωn 2ζ2 -1 (1.42) and a2 = v0 + ( ζ + 2ζ2 - 1 ) ωnx0 2ωn 2ζ2 -1 (1.43) Typical responses are plotted in Figure 1.11, where it is clear that motion does not involve oscillation. An overdamped system does not oscillate but rather returns to its rest position exponentially. Displacement (mm) 0.4 1. x0 = 0.3, v0 = 0 2. x0 = 0, v0 = 1 3. x0 = 0.3, v0 = 0 1 0.2 2 0.0 0.2 3 Time (s) 0.4 0 1 2 3 4 5 6 Figure 1.11 The response of an overdamped system, ζ 7 1, for two different values of initial displacement (in mm) both with the initial velocity set to zero and one case with x0 = 0 and v0 = 1 mm>s. 1.3.3 Critically Damped Motion In this last case, the damping ratio is exactly one (ζ = 1) and the discriminant of equation (1.31) is equal to zero. This corresponds to the value of ζ that separates oscillatory motion from nonoscillatory motion. Since the roots are repeated, they have the value λ1 = λ2 = -ωn (1.44) x(t) = (a1 + a2t)e -ωnt (1.45) The solution takes the form www.TechnicalBooksPDF.com 28 Introduction to Vibration and the Free Response Chap. 1 where, again, the constants a1 and a2 are determined by the initial conditions. Substituting the initial displacement into equation (1.45) and the initial velocity into the derivative of equation (1.45) yields a1 = x0, a2 = v0 + ωnx0 (1.46) Critically damped motion is plotted in Figure 1.12 for two different values of initial conditions. It should be noted that critically damped systems can be thought of in several ways. They represent systems with the smallest value of damping rate that yields nonoscillatory motion. Critical damping can also be thought of as the case that separates nonoscillation from oscillation, or the value of damping that provides the fastest return to zero without oscillation. Displacement (mm) 0.6 1 0.4 0.2 1. x0  0.4 mm, v0   1mm/s 2. x0  0.4 mm, v0  0 mm/s 3. x0  0.4 mm, v0  1mm/s 2 3 Time (s) 0 0.1 0.5 1 1.5 2 2.5 3 3.5 Figure 1.12 The response of a critically damped system for three different initial velocities. The physical properties are m = 100 kg, k = 225 N >m, and ζ = 1. Example 1.3.1 Recall the small spring of Example 1.2.1 (i.e., ωn = 132 rad>s). The damping rate of the spring is measured to be 0.11 kg>s. Calculate the damping ratio and determine if the free motion of the spring–bolt system is overdamped, underdamped, or critically damped. Solution From Example 1.2.1, m = 49.2 * 10−3 kg and k = 857.8 N>m. Using the definition of the critical damping coefficient of equation (1.29) and these values for m and k yields ccr = 21km = 22 ( 857.8 N>m )( 49.2 * 10-3 kg ) = 12.993 kg>s If c is measured to be 0.11 kg>s, the critical damping ratio becomes ζ = 0.11(kg>s) c = = 0.0085 ccr 12.993(kg>s) or 0.85% damping. Since ζ is less than 1, the system is underdamped. The motion resulting from giving the spring–bolt system a small displacement will be oscillatory. n www.TechnicalBooksPDF.com Sec. 1.3 29 Viscous Damping The single-degree-of-freedom damped system of equation (1.25) is often written in a standard form. This is obtained by dividing equation (1.25) by the mass, m. This yields k c # $ x + x = 0 x + m m (1.47) The coefficient of x(t) is ω2n, the undamped natural frequency squared. A little # manipulation illustrates that the coefficient of the velocity x is 2ζωn. Thus equation (1.47) can be written as $ # (1.48) x(t) + 2ζωnx(t) + ω2nx(t) = 0 In this standard form, the values of the natural frequency and the damping ratio are clear. In differential equations, equation (1.48) is said to be in monic form, meaning that the leading coefficient (coefficient of the highest derivative) is one. Example 1.3.2 The human leg has a measured natural frequency of around 20 Hz when in its rigid (knee-locked) position in the longitudinal direction (i.e., along the length of the bone) with a damping ratio of ζ = 0.224. Calculate the response of the tip of the leg bone to an initial velocity of v0 = 0.6 m>s and zero initial displacement (this would correspond to the vibration induced while landing on your feet, with your knees locked from a height of 18 mm) and plot the response. Last, calculate the maximum acceleration experienced by the leg assuming no damping. Solution The damping ratio is ζ = 0.224 6 1, so the system is clearly underdamped. The natural frequency is ωn = 20 cycles 2π rad = 125.66 rad>s. The damped natural 1 s cycles frequency is ωd = 125.6621 - (0.224)2 = 122.467 rad>s. Using equation (1.38) with v0 = 0.6 m>s and x0 = 0 yields A = 230.6 + (0.224)(125.66)(0) 4 2 + 3(0)(122.467)4 2 122.467 ϕ = tan-1 a (0)(ωd) v0 + ζωn(0) b = 0 = 0.005 m The response as given by equation (1.36) is x(t) = 0.005e -28.148t sin (122.467t) This is plotted in Figure 1.13. To find the maximum acceleration rate that the leg experiences for zero damping, use the undamped case of equation (1.9): A = B x20 + a v0 2 b , ωn = 125.66, v0 = 0.6, x0 = 0 ωn www.TechnicalBooksPDF.com 30 Introduction to Vibration and the Free Response Chap. 1 Displacement (mm) 5 4 3 2 1 0 1 2 3 4 Time (s) 5 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Figure 1.13 A plot of displacement versus time for the leg bone of Example 1.3.2. A = v0 0.6 m = m ωn ωn 0.6 $ max (x) = 0 -ω2n A 0 = ` - ω2n a b ` = (0.6)(125.66 m>s2) = 75.396 m>s2 ωn In terms of g = 9.81 m>s2, this becomes maximum acceleration = 75.396 m>s2 9.81 m>s2 g = 7.69 g>s n Example 1.3.3 Compute the form of the response of an underdamped system using the Cartesian form of the solution given in Window 1.5. Solution From basic trigonometry sin(x + y) = sin x cos y + cos x sin y. Applying this to equation (1.36) with x = ωdt and y = ϕ yields x(t) = Ae -ζωnt sin (ωdt + ϕ) = e -ζωnt(A1 sin ωdt + A2 cosωdt) where A1 = A cos ϕ and A2 = A sin ϕ, as indicated in Window 1.5. Evaluating the initial conditions yields x(0) = x0 = e 0(A1 sin 0 + A2 cos 0) www.TechnicalBooksPDF.com Sec. 1.4 31 Modeling and Energy Methods Solving yields A2 = x0. Next, differentiate x(t) to get # x = - ζωne -ζωnt (A1 sinωdt + A2 cos ωdt) + ωde -ζωnt (A1 cos ωdt - A2 sin ωdt) Applying the initial velocity condition yields # v0 = x(0) = - ζωn (A1 sin 0 + x0 cos 0) + ωd (A1 cos 0 - x0 sin 0) Solving this last expression yields A1 = v0 + ζωnx0 ωd Thus the free response in Cartesian form becomes x(t) = e -ζωnt a v0 + ζωnx0 sin ωdt + x0 cos ωdtb ωd n 1.4 MODELING AND ENERGY METHODS Modeling is the art or process of writing down an equation, or system of equations, to describe the motion of a physical device. For example, equation (1.2) was obtained by modeling the spring–mass system of Figure 1.5. By summing the forces acting on the mass along the x direction and employing the experimental evidence of the mathematical model of the force in a spring given by Figure 1.4, equation (1.2) can be obtained. The success of this model is determined by how well the solution of equation (1.2) predicts the observed and measured behavior of the actual system. This comparison between the vibration response of a device and the response predicted by the analytical model is discussed in Section 1.6. The majority of this book is devoted to the analysis of vibration models. However, two methods of modeling—force balance and energy methods—are presented in this section. Newton’s three laws form the basis of dynamics. Fifty years after Newton, Euler published his laws of motion. Newton’s second law states: the sum of forces acting on a body is equal to the body’s mass times its acceleration, and Euler’s second law states: the rate of change of angular momentum is equal to the sum of external moments acting on the mass. Euler’s second law can be manipulated to reveal that the sum of moments acting on a mass is equal to its rotational inertia times its angular acceleration. These two laws require the use of free-body diagrams and the proper identification of forces and moments acting on a body, forming most of the activity in the study of dynamics. An alternative approach, studied in dynamics, is to examine the energy in the system, giving rise to what is referred to as energy methods for determining the equations of motion. The energy methods do not require free-body diagrams www.TechnicalBooksPDF.com 32 Introduction to Vibration and the Free Response Chap. 1 but rather require an understanding of the energy in a system, providing a useful alternative when forces are not easy to determine. More comprehensive treatments of modeling can be found in Doebelin (1980), Shames (1980, 1989), and Cannon (1967), for example. The best reference for modeling is the text you used to study dynamics. There are also many excellent descriptions on the Internet which can be found using a search engine such as Google. The force summation method is used in the previous sections and should be familiar to the reader from introductory dynamics. For systems with constant mass (such as those considered here) moving in only one direction, the rate of change of momentum becomes the scalar relation d # $ (mx) = mx dt which is often called the inertia force. The physical device of interest is examined by noting the forces acting on the device. The forces are then summed (as vectors) to produce a dynamic equation following Newton’s second law. For motion along the x direction only, this becomes the scalar equation $ a fxi = mx (1.49) i where fxi denotes the ith force acting on the mass m along the x direction and the summation is over the number of such forces. In the first three chapters, only singledegree-of-freedom systems moving in one direction are considered; thus, Newton’s law takes on a scalar nature. In more practical problems with many degrees of freedom, energy considerations can be combined with the concepts of virtual work to produce Lagrange’s equations, as discussed in Section 4.7. Lagrange’s equations also provide an energy-based alternative to summing forces to derive equations of motion. For rigid bodies in plane motion (i.e., rigid bodies for which all the forces acting on them are coplanar in a plane perpendicular to a principal axis) and free to rotate, Euler’s second law states that the sum of the applied torques is equal to the rate of change of angular momentum of the mass. This is expressed as $ a M0i = Jθ (1.50) i where M0i are the torques acting on the object about the point 0, J is the moment of inertia (also denoted by I0) about the rotation axis, and θ is the angle of rotation. The sum of moments method was used in Example 1.1.1 to find the equation of motion of a pendulum and is discussed in more detail in Example 1.5.1. If the forces or torques acting on an object or mechanical part are difficult to determine, an energy approach may be more efficient. In this method, the differential equation of motion is established by using the principle of energy conservation. This principle is equivalent to Newton’s law for conservative systems and states that www.TechnicalBooksPDF.com Sec. 1.4 33 Modeling and Energy Methods the sum of the potential energy and kinetic energy of a particle remains constant at each instant of time throughout the particle’s motion: T + U = constant (1.51) where T and U denote the total kinetic and potential energy, respectively. Conservation of energy also implies that the change in kinetic energy must equal the change in potential energy: U1 - U2 = T2 - T1 (1.52) where U1 and U2 represent the particle’s potential energy at the times t1 and t2, respectively, and T1 and T2 represent the particle’s kinetic energy at times t1 and t2, respectively. For periodic motion, energy conservation also implies that Tmax = Umax (1.53) Since energy is a scalar quantity, using the conservation of energy principle yields a possibility of obtaining the equation of motion of a system without using force or moment summations. Equations (1.51), (1.52), and (1.53) are three statements of the conservation of energy. Each of these can be used to determine the equation of motion of a spring– mass system. As an illustration, consider the energy of the spring–mass system of Figure 1.14 hanging in a gravitational field of strength g. The effect of adding the mass m to the massless spring of stiffness k is to stretch the spring from its rest position at 0 to the static equilibrium position Δ. The total potential energy of the spring–mass system is the sum of the potential energy of the spring (or strain energy; see, e.g., Shames, 1989) and the gravitational potential energy. The potential energy of the spring is given by Uspring = 1 2 k(Δ + x)2 (1.54) The gravitational potential energy is Ugray = -mgx (1.55) where the negative sign indicates that x = 0 is the reference for zero potential energy. The kinetic energy of the system is # (1.56) T = 21 mx2 k k 0 m  g m mg x(t) (a) (b) Figure 1.14 (a) A spring–mass system hanging in a gravitational field. Here Δ is the static equilibrium position and x is the displacement from equilibrium. (b) The free-body diagram for static equilibrium. www.TechnicalBooksPDF.com 34 Introduction to Vibration and the Free Response Chap. 1 Substituting these energy expressions into equation (1.51) yields #2 1 1 2 2 mx - mgx + 2 k(Δ + x) = constant (1.57) Differentiating this expression with respect to time yields # $ # x(mx + kx) + x(kΔ - mg) = 0 (1.58) Since the static force balance on the mass from Figure 1.14(b) yields the fact that kΔ = mg, equation (1.58) becomes # $ x(mx + kx) = 0 (1.59) # The velocity x cannot be zero for all time; otherwise, x(t) = constant and no vibration would be possible. Hence equation (1.59) yields the standard equation of motion $ mx + kx = 0 (1.60) This procedure is called the energy method of obtaining the equation of motion. The energy method can also be used to obtain the frequency of vibration directly for conservative systems that are oscillatory. The maximum value of sine (and cosine) is one. Hence, from equations (1.3) and (1.4), the maximum displacement is A and the maximum velocity is ωnA (recall Window 1.3). Substitution of these maximum values into the expression for Umax and Tmax and using the energy equation (1.53) yields 1 2 m(ωnA)2 = 1 2 kA2 (1.61) Solving equation (1.61) for ωn yields the standard natural frequency relation ωn = 1k>m. Example 1.4.1 Figure 1.15 is a simple single-degree-of-freedom model of a wheel mounted on a spring. The friction in the system is such that the wheel rolls without slipping. Calculate the natural frequency of oscillation using the energy method. Assume that no energy is lost during the contact. x(t) r m, J  k Figure 1.15 The rotational displacement of the wheel of radius r is given by θ(t) and the linear displacement is denoted by x(t). The wheel has a mass m and a moment of inertia J. The spring has a stiffness k. Solution #From introductory dynamics, the rotational kinetic energy of the wheel is T rot = 12 Jθ2, where J is the mass moment of inertia of the wheel and θ = θ(t) is the angle of rotation of the wheel. This assumes that the wheel moves relative to the surface without slipping (so that no energy is lost at contact). The translational kinetic energy # of the wheel is TT = 12 mx2. www.TechnicalBooksPDF.com Sec. 1.4 35 Modeling and Energy Methods T rot # # The rotation θ and the translation x are related by x = rθ. Thus x = rθ and # 1 #2 2 = 2 Jx >r . At maximum energy x = A and x = ωn A, so that Tmax = 1 #2 1 1 J #2 xmax = (m + J>r 2)ω2nA2 mxmax + 2 2 r2 2 and Umax = 1 2 kx2max = 1 2 kA2 Using conservation of energy in the form of equation (1.53) yields Tmax = Umax, or 1 J 1 am + 2 bω2n = k 2 2 r Solving this last expression for ωn yields ωn = k A m + J>r 2 the desired frequency of oscillation of the suspension system. The denominator in the frequency expression derived in this example is called the effective mass because the term (m + J>r2) has the same effect on the natural frequency as does a mass of value (m + J>r2). n Example 1.4.2 Use the energy method to determine the equation of motion of the simple pendulum (the rod l is assumed massless) shown in Example 1.1.1 and repeated in Figure 1.16. O l l cos θ l g θ m h Figure 1.16 The geometry of the pendulum for Example 1.4.2. Solution Several assumptions must first be made to ensure simple behavior (a more complicated version is considered in Example 1.4.6). Using the same assumptions given in Example 1.1.1 (massless rod, no friction in the hinge), the mass moment of inertia about point 0 is J = ml 2 www.TechnicalBooksPDF.com 36 Introduction to Vibration and the Free Response Chap. 1 The angular displacement θ(t) is measured from the static equilibrium or rest position of the pendulum. The kinetic energy of the system is T = # 1 1 #2 Jθ = ml 2θ2 2 2 The potential energy of the system is determined by the distance h in the figure so that U = mgl (1 - cos θ) since h = l(1 - cos θ) is the geometric change in elevation of the pendulum mass. Substitution of these expressions for the kinetic and potential energy into equation (1.51) and differentiating yields d 1 2#2 3 ml θ + mgl(1 - cos θ)4 = 0 dt 2 or #$ # ml 2θθ + mgl(sin θ)θ = 0 # Factoring out θ yields # $ θ(ml 2θ + mgl sin θ) = 0 # Since θ(t) cannot be zero for all time, this becomes $ ml 2θ + mgl sin θ = 0 or g $ θ + sin θ = 0 l This is a nonlinear equation in θ and is discussed in Section 1.10 and is derived from summing moments on a free-body diagram in Example 1.1.1. However, since sin θ can be approximated by θ for small angles, the linear equation of motion for the pendulum becomes g $ θ + θ = 0 l This corresponds to an oscillation with natural frequency ωn = 1g>l for initial conditions such that θ remains small, as defined by the approximation sin θ ≈ θ, as discussed in Example 1.1.1. In Example 1.4.2, it is important to not invoke the small-angle approximation before the final equation of motion is derived. For instance, if the small-angle approximation is used in the potential energy term, then U = mgl(1 - cos θ) = 0, since the small-angle approximation for cos θ is 1. This would yield an incorrect equation of motion. n www.TechnicalBooksPDF.com Sec. 1.4 37 Modeling and Energy Methods Example 1.4.3 Determine the equation of motion of the shaft and disk illustrated in Window 1.1 using the energy method. Solution The shaft and disk of Window 1.1 are modeled as a rod stiffness in twisting, resulting in torsional motion. The shaft, or rod, exhibits a torque in twisting proportional to the angle of twist θ(t). The potential energy associated with the torsional spring stiffness is U = 21 kθ2, where the stiffness coefficient k is determined much like the method used to determine the spring stiffness in translation, as discussed in Section 1.1. The angle θ(t) is measured from the static equilibrium, or rest, position. The kinetic energy # associated with the disk of mass moment of inertia J is T = 12 Jθ2. This assumes that the inertia of the rod is much smaller than that of the disk and can be neglected. Substitution of these expressions for the kinetic and potential energy into equation (1.51) and differentiating yields d dt $ # # ( 12 Jθ2 + 12 kθ2 ) = ( Jθ + kθ ) θ = 0 # so that the equation of motion becomes (because θ ≠ 0) $ Jθ + kθ = 0 This is the equation of motion for torsional vibration of a disk on a shaft. The natural frequency of vibration is ωn = 1k>J. n Example 1.4.4 Model the mass of the spring in the system shown in Figure 1.17 and determine the effect of including the mass of the spring on the value of the natural frequency. y ydy ms , k l Figure 1.17 A spring–mass system with a spring of mass ms that is too large to neglect. m x(t) Solution One approach to considering the mass of the spring in analyzing the system vibration response is to calculate the kinetic energy of the spring. Consider the kinetic enm ergy of the element dy of the spring. If ms is the total mass of the spring, then l s dy, is the mass of the element dy. The velocity of this element, denoted by vdy, may be approximated by assuming that the velocity at any point varies linearly over the length of the spring: vdy = y # x(t) l www.TechnicalBooksPDF.com 38 Introduction to Vibration and the Free Response Chap. 1 The total kinetic energy of the spring is the kinetic energy of the element dy integrated over the length of the spring: l ms y # 2 1 c x d dy 2 L0 l l Tspring = 1 ms # 2 a bx 2 3 = m From the form of this expression, the effective mass of the spring is 3s , or one-third of that of the spring. Following the energy method, the maximum kinetic energy of the system is thus T max = ms 1 am + b ω2n A2 2 3 Equating this to the maximum potential energy, 12kA2 yields the fact that the natural frequency of the system is ωn = k A m + ms >3 Thus, including the effects of the mass of the spring in the system decreases the natural frequency. Note that if the mass of the spring is much smaller than the system mass m, the effect of the spring’s mass on the natural frequency is negligible. n Example 1.4.5 Fluid systems, as well as solid systems, exhibit vibration. Calculate the natural frequency of oscillation of the fluid in the U-shaped manometer illustrated in Figure 1.18 using the energy method. x(t) x(t) l   weight density (volume) A  cross-sectional area l  length of fluid www.TechnicalBooksPDF.com Figure 1.18 A U-shaped manometer consisting of a fluid moving in a tube. Sec. 1.4 39 Modeling and Energy Methods Solution The fluid has weight density γ (i.e., the specific weight). The restoring force is provided by gravity. The potential energy of the fluid [(weight)(displacement of c.g.)] is 0.5(γAx)x in each column, so that the total change in potential energy is U = U2 - U1 = 1 2 γAx2 - (- 21 γAx2 ) = γAx2 The change in kinetic energy is T = 1 Alγ # 2 1 Alγ # 2 (x - 0) = x 2 g 2 g Equating the change in potential energy to the change in kinetic energy yields 1 Alγ # 2 x = γAx2 2 g Assuming an oscillating motion of the form x(t) = X sin(ωnt + ϕ) and evaluating this expression for maximum velocity and position yields 1 l 2 2 ωn X = X 2 2g where X is used to denote the amplitude of vibration. Solving for ωn yields ωn = 2g B l which is the natural frequency of oscillation of the fluid in the tube. Note that it depends only on the acceleration due to gravity and the length of the fluid. Vibration of fluids inside mechanical containers (called sloshing) occurs in gas tanks in both automobiles and airplanes and forms an important application of vibration analysis. n Example 1.4.6 Consider the compound pendulum of Figure 1.19 pinned to rotate around point O. Derive the equation of motion using Euler’s second law (sum of moments as in Example 1.1.1). A compound pendulum is any rigid body pinned at a point other than its center of mass. If the only force acting on the system is gravity, then it will oscillate around that point and behave like a pendulum. The purpose of this example is to determine the equation of motion and to introduce the interesting dynamic property of the center of percussion. Solution A compound pendulum results from a simple pendulum configuration (Examples 1.1.1 and 1.4.2) if there is a significant mass distribution along its length. In the figure, G is the center of mass, O is the pivot point, and θ(t) is the angular displacement of the centerline of the pendulum of mass m and moment of inertia J measured about the z-axis at point O. Point C is the center of percussion, which is defined as the distance q0 along the centerline such that a simple pendulum (a massless rod pivoted at zero with mass m at its tip, as in Example 1.4.2) of radius q0 has the same period. Hence q0 = J mr www.TechnicalBooksPDF.com 40 Introduction to Vibration and the Free Response Fy O O Chap. 1 r q0 G Fx G mgr sin(t) mgr cos(t) C mg (t) (a) (b) Figure 1.19 (a) A compound pendulum pivoted to swing about point O under the influence of gravity (pointing down). (b) A free-body diagram of the pendulum. where r is the distance from the pivot point to the center of mass. Note that the pivot point O and the center of percussion C can be interchanged to produce a pendulum with the same frequency. The radius of gyration, k0, is the radius of a ring that has the same moment of inertia as the rigid body does. The radius of gyration and center of percussion are related by q0r = k 20 Consider the equation of motion of the compound pendulum. Taking moments about its pivot point O yields $ ΣM0 = Jθ(t) = - mgr sin θ(t) For small θ(t) this nonlinear equation becomes (sin ~ θ) $ Jθ(t) + mgr θ(t) = 0 The natural frequency of oscillation becomes ωn = mgr A J This frequency can be expressed in terms of the center of percussion as ωn = g A q0 which is just the frequency of a simple pendulum of length q0. This can be seen by examining the forces acting on the simple (massless rod) pendulum of Examples 1.1.1, 1.4.2, and Figure 1.20(a) or recalling the result obtained in these examples. www.TechnicalBooksPDF.com Sec. 1.4 41 Modeling and Energy Methods f0 f0 O O k0 l G = center of mass   q0 Figure 1.20 (a) A simple pendulum consisting of a massless rod pivoted at point O with a mass attached to its tip. (b) A compound pendulum consisting of a shaft with center of mass at point G. Here f0 is the pin reaction force. m mg (a) mg (b) Summing moments about O yields $ ml 2θ = - mgl sin θ or after approximating sin θ with θ, g $ θ + θ = 0 l This yields the simple pendulum frequency of ωn = 2g>l, which is equivalent to that obtained previously for the compound pendulum using l = q0. Next, consider the uniformly shaped compound pendulum of Figure 1.20(b) of length l. Here it is desired to calculate the center of percussion and radius of gyration. The mass moment of inertia about point O is J, so that summing moments about O yields $ l Jθ = -mg sin θ 2 since the mass is assumed to be evenly distributed and the center of mass is at r = l>2. The moment of inertia for a slender rod about O is J = 31 ml 2; hence, the equation of motion is ml 2 $ l θ + mg θ = 0 3 2 where sin θ has again been approximated by θ, assuming small motion. This becomes $ 3g θ = 0 θ + 2 l so that the natural frequency is ωn = 3g A2 l www.TechnicalBooksPDF.com 42 Introduction to Vibration and the Free Response Chap. 1 The center of percussion becomes q0 = J 2 = l mr 3 and the radius of gyration becomes k0 = 1q0r = l 13 These positions are marked on Figure 1.20(b). The center of percussion and pivot point play a significant role in designing an automobile. The center of percussion is the point on an object where it may be struck (impacted) producing forces that cancel, causing no motion at the point of support. The axle of the front wheels of an automobile is considered as the pivot point of a compound pendulum parallel to the road. If the back wheels hit a bump, the frequency of oscillation of the center of percussion will annoy passengers. Hence automobiles are designed such that the center of percussion falls over the axle and suspension system, away from passengers. The concept of center of percussion is used in many swinging, or pendulum-like, situations. This notion is sometimes used to define the “sweet spot” in a tennis racket or baseball bat and defines where the ball should be hit. If the hammer is shaped so that the impact point is at the center of percussion (i.e., the hammer’s head), then ideally no force is felt if it is held at the “end” of the pendulum. n The energy method can be used in two ways. The first is to equate the maximum kinetic energy to the maximum potential energy [equation (1.53)] while assuming harmonic motion. This yields the natural frequency without writing out the equation of motion, as illustrated in equation (1.61). Beyond the simple calculation of frequency, this approach has limited use. However, the second use of the energy method involves deriving the equation of motion from the conservation of energy by differentiating equation (1.51) with respect to time. This concept is more useful and is illustrated in Examples 1.4.2 and 1.4.3. The concept of using energy quantities to derive the equations of motion can be extended to more complicated systems with many degrees of freedom, such as those discussed in Chapters 4 (multiple-degreeof-freedom systems) and 6 (distributed-parameter systems). The method is called Lagrange’s method and is simply stated here to introduce the concept. Lagrange’s method is introduced more formally in Chapter 4, where multiple-degree-of-freedom systems make the power of Lagrange’s method obvious. Lagrange’s method for conservative systems consists of defining the Lagrangian, L, of the system defined by L = T - U. Here T is the total kinetic energy of the system and U is the total potential energy in the system, both stated in terms of “generalized” coordinates. Generalized coordinates are denoted by “qi(t)” and will be formally defined later. Here it is sufficient to state that qi would be x in Example 1.4.4 and θ in Example 1.4.3. Then Lagrange’s method for conservative www.TechnicalBooksPDF.com Sec. 1.4 43 Modeling and Energy Methods systems states that the equations of motion for the free response of an undamped system result from d 0L 0L a # b = 0 dt 0qi 0qi (1.62) Substitution of the expression for L into equation (1.62) yields d 0T 0T 0U a # b + = 0 dt 0qi 0qi 0qi (1.63) Here one equation results for each subscript, i. In the case of the single-degree-offreedom systems considered in this chapter, there is only one coordinate (i = 1) and only one equation of motion will result. The following example illustrates the use of the Lagrange approach to derive the equation of motion of a simple spring– mass system. Example 1.4.7 Use Lagrange’s method to derive the equation of motion of the simple spring–mass system of Figure 1.5. Compare this derivation to using the energy method described in Examples 1.4.2 and 1.4.3. Solution In the case of the simple spring–mass system, the kinetic and potential energy are, respectively, T = 1 1 #2 mx and U = kx2 2 2 Here the generalized coordinate qi(t) is just the displacement x(t). Following the Lagrange approach, the Lagrange equation (1.63) becomes d 0T 0U 0T a # b + = 0 dt 0x 0x 0x 1 d 0 1 2 # $ (mx) + a kx b = mx + kx = 0 dt 0x 2 This, of course, agrees with the approach of Newton’s sum of forces. Next, consider the energy method, which starts with T + U = constant. Taking the total derivative of this expression with respect to time yields 1 d 1 #2 #$ # # $ $ a mx + kx2 b = mxx + kxx = x(mx + kx) = 0 1 mx + kx = 0 dt 2 2 since the velocity cannot be zero for all time. Thus the two energy-based approaches yield the same result and that result is equivalent to that obtained by Newton’s sum of forces. Note that in order to follow the above calculations, it is important to remember the difference between total derivatives and partial derivatives and their respective rules of calculation from calculus. n www.TechnicalBooksPDF.com 44 Introduction to Vibration and the Free Response Chap. 1 Example 1.4.8 Use Lagrange’s method to derive the equation of motion of the simple spring–mass pendulum system of Figure 1.21 and compute the system’s natural frequency. O θ g k m Figure 1.21 A pendulum attached to a spring. Assume that the pendulum swings through only small angles so that the spring has negligible deflection in the vertical direction and assume that the mass of the pendulum rod is negligible. Solution In approaching a problem where there are several choices of variables, as in this case, it is a good idea to first write down the energy expressions in easy choices of velocities and displacements and then use a diagram to identify kinematic relationships and geometry as indicated in Figure 1.22. Referring to the figure, the kinetic energy of the mass is T = # 1 #2 1 Jθ = ml 2 θ2 2 2 O l l cos θ l g θ k m h x Figure 1.22 The geometry of the pendulum attached to a spring for small angles showing the kinematic relationships needed to formulate the energies in terms of a single generalized coordinate, θ. The potential energy in the system consists of two parts, one due to the spring and one due to gravity. Thus the total potential energy is U = 1 2 kx + mgh 2 www.TechnicalBooksPDF.com Sec. 1.4 45 Modeling and Energy Methods Next, use the figure to write each energy expression in terms of the single variable θ. From the figure, the mass moves up a distance h = l - l cos θ, and the distance the spring compresses is x = l sin θ. Thus the potential energy becomes U = 1 2 2 l sin θ + mgl(1 - cos θ) 2 With the energies all stated in terms of the single generalized coordinate θ, the derivatives required in the Lagrange formulation become # $ d 0T d a # b = (ml 2θ) = ml 2θ dt 0θ dt 0 1 2 2 0U = a l sin θ + mgl(1 - cos θ) b = kl 2 sin θ cos θ + mgl sin θ 0θ 0θ 2 Combining these expressions, the Lagrange equation (1.63) yields $ 0U 0T d 0T + a # b = ml 2θ + kl 2 sin θ cos θ + mgl sin θ = 0 0θ dt 0θ 0θ " 0 For small θ the equation of motion becomes $ ml 2θ + (kl 2 + mgl)θ = 0 Thus the natural frequency is ωn = B kl + mg ml Note that the equation of motion reduces to that of the pendulum given in Example 1.4.2 without the spring (k = 0). n The Lagrange approach presented here is for the free response of undamped systems (conservative systems) and has only been applied to a single-degree-offreedom system. However, the method is general and can be expanded to include the forced response and damping. So far, three basic systems have been modeled: rectilinear or translational motion of a spring–mass system, torsional motion of a disk–shaft system, and the pendulum motion of a suspended mass system. Each of these motions commonly experiences energy dissipation of some form. The viscous-damping model of Section 1.3 developed for translational motion can be applied directly to both torsional and pendulum motion. In the case of torsional motion of the shaft, the energy dissipation is assumed to come from heating of the material and/or air resistance. Sometimes, as in the case of using the rod and disk to model an automobile crankshaft or camshaft, the damping is assumed to come from the oil that surrounds the disk and shaft, or bearings that support the shaft. www.TechnicalBooksPDF.com 46 Introduction to Vibration and the Free Response Chap. 1 TABLE 1.1 A CoMPARISon oF RECtIlInEAR AnD RotAtIonAl SyStEMS AnD A SuMMARy oF unItS Rectilinear, x (m) Spring force Damping force Inertia force Equation of motion Stiffness units Damping units Inertia units Force/torque kx # cx $ mx $ # mx + cx + kx = 0 N>m N # s>m, kg>s Kg N = kg # m>s2 Torsional/pendulum, θ (rad) kθ# cθ$ Jθ$ # Jθ + cθ + kθ = 0 # N m>rad M # N # s>rad kg # m2 >rad N # m = kg # m2 >s2 In all three cases, the damping is modeled as proportional to velocity (i.e., # # fc = cx or fc = cθ). The equations of motion are then of the form indicated in Table 1.1. Each of these equations can be expressed as a damped linear oscillator given in the form of equation (1.48). Hence, each of these three systems is characterized by a natural frequency and a damping ratio. Each of these three systems has a solution based on the nature of the damping ratio ζ, as discussed in Section 1.3. 1.5 STIFFNESS The stiffness in a spring, introduced in Section 1.1, can be related more directly to material and geometric properties of the spring. This section introduces the relationships between stiffness, elastic modulus, and geometry of various types of springs and illustrates various situations that can lead to simple harmonic motion. A spring-like behavior results from a variety of configurations, including longitudinal motion (vibration in the direction of the length), transverse motion (vibration perpendicular to the length), and torsional motion (vibration rotating around the length). Consider again the stiffness of the spring introduced in Section 1.1. A spring is generally made of an elastic material. For a slender elastic material of length l, cross-sectional area A, and elastic modulus E (or Young’s modulus), the stiffness of the bar for vibration along its length is given by k = EA l (1.64) This describes the spring constant for the vibration problem illustrated in Figure 1.23, where the mass of the rod is ignored (or very small relative to the mass m in the figure). The modulus E has the units of pascal (denoted by Pa), which are N>m2. The modulus for several common materials is given in Table 1.2. www.TechnicalBooksPDF.com Sec. 1.5 47 Stiffness k  EA l l m x(t) E  elastic modulus A  cross-sectional area l  length of bar x(t)  deflection GJP  stiffness of rod l J  mass moment of inertia of the disk G  shear modulus of rigidity of the rod JP  polar moment of inertia of the rod l  length of rod   angular displacement k JP J Figure 1.23 The stiffness associated with the longitudinal (along the long axis) vibration of a slender prismatic bar. (0) (t) Figure 1.24 The stiffness associated with the torsional rotation (twisting) of a shaft. Next, consider a twisting motion with a similar rod of circular cross section, as illustrated in Figure 1.24. In this case, the rod possesses a polar moment of inertia, JP, and (shear) modulus of rigidity, G (see Table 1.2). For the case of a wire or shaft of diameter d, JP = πd 4 >32. The modulus of rigidity has units N>m2. The torsional stiffness is k = TABLE 1.2 GJP l (1.65) PHySICAl ConStAntS FoR SoME CoMMon MAtERIAlS Material Young’s modulus, E(N>m2) Steel Aluminum Brass Copper Concrete Rubber Plywood 2.0 7.1 10.0 6.0 3.8 2.3 5.4 * * * * * * * 1011 1010 1010 1010 109 109 109 Density, (kg>m3) 7.8 2.7 8.5 2.4 1.3 1.1 6.0 * * * * * * * 103 103 103 103 103 103 102 www.TechnicalBooksPDF.com Shear modulus, G(N>m2) 8.0 2.67 3.68 2.22 — 8.21 — * * * * 1010 1010 1010 1010 * 108 48 Introduction to Vibration and the Free Response Chap. 1 which is used to describe the vibration problem illustrated in Figure 1.24, where the mass of the shaft is ignored. In the figure, θ(t) represents the angular position of the shaft relative to its equilibrium position. The disk of radius r and rotational moment of inertia J will vibrate around the equilibrium position θ(0) with stiffness GJP >l. Example 1.5.1 Calculate the natural frequency of oscillation of the torsional system given in Figure 1.24. Solution Using the moment equation (1.50), the equation of motion for this system is $ Jθ(t) = - kθ(t) This may be written as $ k θ(t) + θ(t) = 0 J This agrees with the result obtained using the energy method as indicated in Example 1.4.3. This indicates an oscillatory motion with frequency ωn = GJP k = BJ B lJ Suppose that the shaft is made of steel and is 1 m long with a diameter of 5 cm. The polar moment of inertia of a rod of circular cross section is JP = (πd4) > 32. If the disk has mass moment of inertia J = 0.5 kg # m2 and considering that the shear modulus of steel is G = 8 * 1010 N > m2, the frequency can be calculated by GJP k = = ω2n = J lJ ( 8 * 1010 N>m 2 ) c = 9.817 * 104 ( rad2 >s2 ) π ( 1 * 10-2 m ) 4 d 32 (1 m) ( 0.5 kg # m2 ) Thus the natural frequency is ωn = 313.3 rad > s, or about 49.9 Hz. n Consider the helical spring of Figure 1.25. In this figure the deflection of the spring is along the axis of the coil. The stiffness is actually dependent on the “twist” of the metal rod forming this spring. The stiffness is a function of the shear modulus G, the diameter of the rod, the diameter of the coils, and the number of coils. The stiffness has the value k = Gd 4 64nR3 (1.66) The helical-shaped spring is very common. Some examples are the spring inside a retractable ballpoint pen and the spring contained in the front suspension of an automobile. www.TechnicalBooksPDF.com Sec. 1.5 49 Stiffness d 2R x(t) d  diameter of spring material 2R  diameter of turns n  number of turns x(t)  deflection k Gd4 64nR3 l 3EI x(0) k  l3 m x(t) E  elastic modulus l  length of beam I  moment of inertia of cross-sectional area about the neutral axis Figure 1.25 The stiffness associated with a helical spring. Figure 1.26 The stiffness associated with the transverse (perpendicular to the long axis) vibration of the tip of a beam, also called the bending stiffness (Blevins, 1987). Assumes the mass of the beam is negligible. Next, consider the transverse vibration of the end of a “leaf” spring illustrated in Figure 1.26. This type of spring behavior is similar to the rear suspension of an automobile as well as the wings of some aircraft. In the figure, l is the length of the beam, E is the elastic (Young’s) modulus of the beam, and I is the (area) moment of inertia of the cross-sectional area. The mass m at the tip of the beam will oscillate with frequency ωn = k 3EI = Am A ml 3 (1.67) in the direction perpendicular to the length of the beam x(t). Example 1.5.2 Consider an airplane wing with a fuel pod mounted at its tip as illustrated in Figure 1.27. The pod has a mass of 10 kg when it is empty and 1000 kg when it is full. Calculate the change in the natural frequency of vibration of the wing, modeled as in Figure 1.27, as the airplane uses up the fuel in the wing pod. The estimated physical parameters of the beam are I = 5.2 * 10 - 5 m4, E = 6.9 * 109 N>m2, and l = 2 m. www.TechnicalBooksPDF.com 50 Introduction to Vibration and the Free Response E, I Chap. 1 x(0) m l x(t) Vertical wing vibration Figure 1.27 A simple model of the transverse vibration of an airplane wing with a fuel pod mounted on its tip. Solution The natural frequency of the vibration of the wing modeled as a simple massless beam with a tip mass is given by equation (1.67). The natural frequency when the fuel pod is full is ωfull = (3) ( 6.9 * 109 )( 5.2 * 10-5 ) 3EI = = 11.6 rad>s C B ml 3 1000(2)3 which is about 1.8 Hz (1.8 cycles per second). The natural frequency for the wing when the fuel pod is empty becomes ωempt = (3) ( 6.9 * 109 )( 5.2 * 10-5 ) 3EI = = 116 rad>s C B ml 3 10(2)3 or 18.5 Hz. Hence the natural frequency of the airplane wing changes by a factor of 10 (i.e., becomes 10 times larger) when the fuel pod is empty. Such a drastic change may cause changes in handling and performance characteristics of the aircraft. n The above calculation ignores the mass of the beam. Clearly if the mass of the beam is significant compared to the tip mass, then the frequency calculation of equation (1.67) must be altered to account for the beam inertia. Similar to including the mass of the spring in Example 1.4.4, the kinetic energy of the beam itself must be considered. To estimate the kinetic energy, consider the static deflection of the beam due to a load at the tip (see any strength of materials text) as illustrated in Figure 1.28. l y m  x(0) u(y) u(y)  x(t) mgy 2 (3ly) 6EI mg www.TechnicalBooksPDF.com Figure 1.28 The static deflection of a beam of modulus, E, cross sectional moment of inertia, I, length, l, and mass density, ρ, with tip mass m, for those cases where the mass of the beam is significant. Sec. 1.5 51 Stiffness Figure 1.28 is roughly the same as Figure 1.25 with the static deflection indicated and the mass density of the beam, ρ, taken into consideration. The static deflection is given by u(y) = mgy2 (3l - y) 6EI (1.68) Here u is the deflection of the beam perpendicular to its length and y is the distance along the length from the fixed end (left end in Figure 1.28). At the tip, the value of u is u(l) = mgl 2 mgl 3 (3l - l) = 6EI 3EI (1.69) Using equation (1.69), equation (1.68) can be written in terms of the maximum deflection as u(y) = u(l)y2 2l 3 (3l - y) (1.70) The maximum value of u is the displacement of the tip so it must be equal to x(t). Thus the velocity of a differential element of the beam will be of the form # x(t)y2 d (u) = (3l - y) dt 2l 3 (1.71) Substitution of this velocity into the expression for kinetic energy of the beam yields Tbeam # l l l 2 x(t)y2 1 1 1 #2 2 ρv dy = ρu dy = ρc (3l y)d dy = 2 L0 2 L0 2 L0 2l 3 (1.72) Here v is velocity and ρ is the mass per unit length of the beam. Substitution of equation (1.71) into (1.72) yields Tbeam = # # ρ l x2 4 ρl x2 33 7 1 33 # 2 y (3l y) dy = l = a ρlb x2(t) 6 6 2 L0 4l 2 4l 35 2 140 (1.73) Thus the mass of the beam, M = ρl, adds this amount to the kinetic energy of the beam and tip-mass arrangement and the total kinetic energy of the system is T = Tbeam + Tm = 1 # 1 33 1 33 # # a ρlb x2(t) + mx2(t) = a M + mb x2(t) (1.74) 2 140 2 2 140 www.TechnicalBooksPDF.com 52 Introduction to Vibration and the Free Response Chap. 1 From equation (1.45), the equivalent mass of the single-degree-of-freedom system of a beam with a tip mass is meq = 33 M + m 140 (1.75) and the corresponding natural frequency is ωn = k 33 M + m H 140 (1.76) This of course reduces to equation (1.67) if the mass, m, of the tip is much larger then the mass, M, of the beam. Example 1.5.3 Referring to the airplane/wing tank of Example 1.5.2, if the wing has a mass of 500 kg, how much does this change the frequency from that calculated for a wing with a full tank (1000 kg)? Does the frequency still change by a factor of 10 when comparing full versus empty? Solution Since the wing weight is equal to half of the fuel tank when full and 50 times the fuel tank when empty, the mass of the wing is clearly a significant factor in the frequency calculation. The two frequencies using equation (1.76) are ωfull = ωempty = 33 a M + mbl 3 R 140 a 3EI = 3EI = 33 M + mbl 3 R 140 (3) ( 6.9 * 109 )( 5.2 * 10-5 ) 33 a (500) + 1000b(2)3 S 140 (3) ( 6.9 * 109 )( 5.2 * 10-5 ) S a 33 (500) + 10b(2)3 140 = 10.97 rad>s = 32.44 rad>s Note that modeling the wing without considering the mass of the wing is very inaccurate. Also note that the frequency shift with and without fuel is still significant (factor of 3 rather than 10). n If the spring of Figure 1.26 is coiled in a plane as illustrated in Figure 1.29, the stiffness of the spring is greatly affected and becomes k = EI l (1.77) Several other spring arrangements and their associated stiffness values are listed in Table 1.3. Texts on solid mechanics and strength of materials should be consulted for further details. www.TechnicalBooksPDF.com Sec. 1.5 53 Stiffness k EI l l  total length of spring E  elastic modulus of spring I  moment of inertia of cross-sectional area TABLE 1.3 Figure 1.29 The stiffness associated with a spring coiled in a plane. SAMPlE SPRIng ConStAntS Axial stiffness of a rod of length l, cross-sectional area A, and modulus E k = EA l Torsional stiffness of a rod of length l, shear modulus G, and torsion constant JP depending on the cross section (π2 r 4 for circle of radius r and 0.1406a4 for a square of side a) k = GJP l Bending stiffness of a cantilevered beam of length l, modulus E, crosssectional moment of inertia I k = 3EI l3 Axial stiffness of a tapered bar of length l, modulus E, and end diameters d1 and d2 k = πEd 1d 2 4l Torsional stiffness on a hollow uniform shaft of shear modulus G, length l, inside diameter d1, and outside diameter d2 k = Transverse stiffness of a pinned–pinned beam of modulus E, area moment of inertia I, and length l for a load applied at point a from its end k = 3EIl a (l - a)2 Transverse stiffness of a clamped–clamped beam of modulus E, area moment of inertia I, and length l for a load applied at its center k = 192EI l3 πG ( d 42 - d 41 ) 32l 2 Example 1.5.4 As another example of vibration involving fluids, consider the rolling vibration of a ship in water. Figure 1.30 illustrates a schematic of a ship rolling in water. Compute the natural frequency of the ship as it rolls back and forth about the axis through M. In the figure, G denotes the center of gravity, B denotes the center of buoyancy, M is the point of intersection of the buoyant force before and after the roll (called the metacenter), and h is the length of GM. The perpendicular line from the center of gravity to the line of action of the buoyant force is marked by the point Z. Here W denotes the weight of the ship, J denotes the mass moment of the ship about the roll axis, and θ(t) denotes the angle of roll. www.TechnicalBooksPDF.com 54 Introduction to Vibration and the Free Response Chap. 1 Centerline of ship  (t) W  weight Water level h M G Z B W  buoyant force G  center of gravity M  metacenter B  center of buoyance Solution Figure 1.30 The dynamics of a ship rolling in water. Summing moments about M yields $ Jθ(t) = - WGZ = - Wh sin θ(t) Again, for small enough values of θ, this nonlinear equation can be approximated by $ Jθ(t) + Whθ(t) = 0 Thus the natural frequency of the system is ωn = hW B J n www.TechnicalBooksPDF.com Sec. 1.5 55 Stiffness Springs in series k2 k1 a b c kac = 1 1/k1 + 1/k2 = k1k2 k1 + k2 Springs in parallel k1 kab = k1 + k2 a Figure 1.31 The rules for calculating the equivalent stiffness of parallel and series connections of springs. b k2 All of the spring types mentioned are represented schematically as indicated in Figure 1.2. If more than one spring is present in a given device, the resulting stiffness of the combined spring can be calculated by two simple rules, as given in Figure 1.31. These rules can be derived by considering the equivalent forces in the system. Example 1.5.5 Consider the spring–mass arrangement of Figure 1.32(a) and calculate the natural frequency of the system. Solution To find the equivalent single stiffness representation of the five-spring system given in Figure 1.32(a), the two simple rules of Figure 1.31 are applied. First, the parallel arrangement of k1 and k2 is replaced by the single spring, as indicated at k2 k1 k1 + k2 m k1 + k2 k m m m k3 k5 k4 (a) 1 1/k3 + 1/k4 k5 + k5 (b) (c) 1 1/k3 + 1/k4 x (d) Figure 1.32 The reduction of a five-spring, one-mass system to an equivalent single-spring–mass system having the same vibration properties. www.TechnicalBooksPDF.com 56 Introduction to Vibration and the Free Response Chap. 1 the top of Figure 1.32(b). Next, the series arrangement of k3 and k4 is replaced with a single spring of stiffness 1 1>k3 + 1>k4 as indicated in the bottom left side of Figure 1.32(b). These two parallel springs on the bottom of Figure 1.32(b) are next combined using the parallel spring formula to yield a single spring of stiffness k5 + k3k4 1 = k5 + 1>k3 + 1>k4 k3 + k4 as indicated in Figure 1.32(c). The final step is to realize that both the spring acting at the top of Figure 1.32(c) and the spring at the bottom attach the mass to ground and hence act in parallel. These two springs then combine to yield the single stiffness k = k1 + k2 + k5 + = k1 + k2 + k5 + k3k4 k3 + k4 (k1 + k2 + k5)(k3 + k4) + k3k4 k3k4 = k3 + k4 (k3 + k4) as indicated symbolically in Figure 1.32(d). Hence the natural frequency of this system is ωn = B k1k3 + k2k3 + k5k3 + k1k4 + k2k4 + k5k4 + k3k4 m(k3 + k4) Note that even though the system of Figure 1.32 contains five springs, it consists of only one mass moving in only one (rectilinear) direction and hence is a single-degreeof- freedom system. n Springs are usually manufactured in only certain increments of stiffness values depending on such things as the number of turns, material, and so on (recall Figure 1.25). Because mass production (and large sales) brings down the price of a product, the designer is often faced with a limited choice of spring constants when designing a system. It may thus be cheaper to use several “off-the-shelf” springs to create the stiffness value necessary than to order a special spring with specific stiffness. The rules of combining parallel and series springs given in Figure 1.31 can then be used to obtain the desired, or acceptable, stiffness and natural frequency. Example 1.5.6 Consider the system of Figure 1.32(a) with k5 = 0. Compare the stiffness and frequency of a 10-kg mass connected to ground, first by two parallel springs (k3 = k4 = 0, k1 = 1000 N>m, and k2 = 3000 N>m), then by two series springs (k1 = k2 = 0, k3 = 1000 N>m, and k4 = 3000 N>m). www.TechnicalBooksPDF.com Sec. 1.5 57 Stiffness Solution First, consider the case of two parallel springs so that k3 = k4 = 0, k1 = 1000 N>m, and k2 = 3000 N>m. Then the equivalent stiffness is given by Figure 1.31 to be the simple sum given by keq = 1000 N>m + 3000 N>m = 4000 N>m and the corresponding frequency is ωparallel = 4000 N>m B 10 kg = 20 rad>s In the case of a series connection (k1 = k2 = 0), the two springs (k3 = 1000 N>m, k4 = 3000 N>m) combine according to Figure 1.31 to yield keq = 3000 3000 1 = = = 750 N>m 1>1000 + 1>3000 3 + 1 4 The corresponding natural frequency becomes ωseries = 750 N>m B 10 kg = 8.66 rad>s Note that using two identical sets of springs connected to the same mass in the two different ways produces drastically different equivalent stiffness and resulting frequency. A series connection decreases the equivalent stiffness, while a parallel connection increases the equivalent stiffness. This is useful in designing systems. n Example 1.5.6 illustrates that fixed values of spring constants can be used in various combinations to produce a desired value of stiffness and corresponding frequency. It is interesting to note that an identical set of physical devices can be used to create a system with drastically different frequencies simply by changing the physical arrangement of the components. This is similar to the choice of resistors in an electric circuit. The formulas of this section are intended to be aids in designing vibration systems. In addition to understanding the effect of stiffness on the dynamics—that is, on the natural frequency—it is important not to forget static analysis when using springs. In particular, the static deflection of each spring system needs to be checked to make sure that the dynamic analysis is correctly interpreted. Recall from the discussion of Figure 1.14 that the static deflection has the value Δ = mg k where m is the mass supported by a spring of stiffness k in a gravitational field providing acceleration of gravity g. Static deflection is often ignored in introductory treatments but is used extensively in spring design and is essential in nonlinear analysis. Static deflection is denoted by a variety of symbols. The symbols δ, Δ, δs, and x0 are all used in vibration publications to denote the deflection of a spring caused by the weight of the mass attached to it. www.TechnicalBooksPDF.com 58 Introduction to Vibration and the Free Response Chap. 1 1.6 MEASUREMENT Measurements associated with vibration are used for several purposes. First, the quantities required to analyze the vibrating motion of a system all require measurement. The mathematical models proposed in previous sections all require knowledge of the mass, damping, and stiffness coefficients of the device under study. These coefficients can be measured in a variety of ways, as discussed in this section. Vibration measurements are also used to verify and improve analytical models. Other uses for vibration testing techniques include reliability and durability studies, searching for damage, and testing for acceptability of the response in terms of vibration parameters. This chapter introduces some basic ideas on measurement. Further discussion of measurement can be found throughout the book, culminating with all the various concepts on measurement summarized in Chapter 7. In many cases, the mass of an object or device is simply determined by using a scale. Mass is a relatively easy quantity to measure. However, the mass moment of inertia may require a dynamic measurement. A method of measuring the mass moment of inertia of an irregularly shaped object is to place the object on the platform of the apparatus of Figure 1.33 and measure the period of oscillation of the system, T. By using the methods of Section 1.4, it can be shown that the moment of inertia of an object, J (about a vertical axis), placed on the disk of Figure 1.33 with its mass center aligned vertically with that of the disk, is given by J = gT 2r 20 (m0 + m) 4π2l - J0 (1.78) Here m is the mass of the part being measured, m0 is the mass of the disk, r0 the radius of the disk, l the length of the wires, J0 the moment of inertia of the disk, and g the acceleration due to gravity. The stiffness of a simple spring system can be measured as suggested in Section 1.1. The elastic modulus, E, of an object can be measured in a similar l Suspension wires of length l Disk of known moment J0, mass m0, and radius r0 www.TechnicalBooksPDF.com Figure 1.33 A Trifilar suspension system for measuring the moment of inertia of irregularly shaped objects. 59 Measurement  Stress Sec. 1.6   E  Strain Figure 1.34 An example of a stress–strain curve of a test specimen used to determine the elastic modulus of a material. fashion by performing a tensile test (see, e.g., Shames, 1989). In this method, a tensile test machine is employed that uses strain gauges to measure the strain, ϵ, in the test specimen as well as the stress, σ, both in the axial direction of the specimen. This produces a curve such as the one shown in Figure 1.34. The slope of the curve in the linear region defines the Young’s modulus, or elastic modulus, for the test material. The relationship that the extension is proportional to the force is known as Hooke’s law. The elastic modulus can also be measured by using some of the formulas given in Section 1.5 and measurement of the vibratory response of a structure or part. For instance, consider the cantilevered arrangement of Figure 1.26. If the mass at the tip is given a small deflection, it will oscillate with frequency ωn = 1k>m. If ωn is measured, the modulus can be determined from equation (1.67), as illustrated in the following example. Example 1.6.1 Consider a steel beam configuration as shown in Figure 1.26. The beam has a length l = 1 m and moment of inertia I = 10−9 m4, with a mass m = 6 kg attached to the tip. If the mass is given a small initial deflection in the transverse direction and oscillates with a period of T = 0.62 s, calculate the elastic modulus of steel. Solution Since T = 2π>ωn, equation (1.67) yields T = 2π ml 3 B 3EI Solving for E yields E = 4π2(6 kg)(1 m)3 4π2ml 3 = = 205 * 109 N>m2 2 3T I 3(0.62 s)2(10-9 m4) n The period T, and hence the frequency ωn, can be measured with a stopwatch for vibrations that are large enough and last long enough to see. However, many vibrations www.TechnicalBooksPDF.com 60 Introduction to Vibration and the Free Response Chap. 1 Displacement (mm) 1.0 0.5 0.0 t1 Time (s) t2 0.5 Figure 1.35 The response of an underdamped system used to measure damping. 1.0 of interest have very small amplitudes and happen very quickly. Hence several very sophisticated devices for measuring time and frequency have been developed, requiring more sophisticated concepts presented in the chapter on measurement. The damping coefficient or, alternatively, the damping ratio is the most difficult quantity to determine. Both mass and stiffness can be determined by static tests; however, damping requires a dynamic test to measure. A record of the displacement response of an underdamped system can be used to determine the damping ratio. One approach is to note that the decay envelope, denoted by the dashed line in Figure 1.35, for an underdamped system is Ae -ζωnt. The measured points x(0), x(t1), x(t2), x(t3), and so on can then be curve fit to A, Ae -ζωnt1, Ae -ζωnt2, Ae -ζωnt3, and so on. This will yield a value for the coefficient ζωn. If m and k are known, ζ and c can be determined from ζωn. This approach also leads to the concept of logarithmic decrement, denoted by δ and defined by δ = ln x(t) x(t + T) (1.79) where T is the period of oscillation. Substitution of the analytical form of the underdamped response given by equation (1.36) yields δ = ln Ae -ζωnt sin (ωdt + ϕ) Ae -ζωn(t + T) sin(ωdt + ωdT + ϕ) (1.80) Since ωdT = 2π, the denominator becomes e -ζωn(t + T) sin(ωdt + ϕ), and the expression for the decrement reduces to δ = ln e ζωnT = ζωnT (1.81) The period T in this case is the damped period (2π>ωd) so that δ = ζωn 2π ωn 21 - ζ 2 = 2πζ 21 - ζ2 www.TechnicalBooksPDF.com (1.82) Sec. 1.6 61 Measurement Solving this expression for ζ yields δ ζ = (1.83) 24π2 + δ2 which determines the damping ratio given the value of the logarithmic decrement. Thus if the value of x(t) is measured from the plot of Figure 1.35 at any two successive peaks, say x(t1) and x(t2), equation (1.79) can be used to produce a measured value of δ, and equation (1.83) can be used to determine the damping ratio. The formula for the damping ratio [equations (1.29) and (1.30), also listed in the inside front cover] and knowledge of m and k subsequently yield the value of the damping coefficient c. Note that peak measurements can be used over any integer multiple of the period (see Problem 1.95) to increase the accuracy over measurements taken at adjacent peaks. The computation in Problem 1.95 yields δ = x(t) 1 b ln a n x(t + nT) where n is any integer number of successive (positive) peaks. While this does tend to increase the accuracy of computing δ, the majority of damping measurements performed today are based on modal analysis methods (Chapters 4 and 6) presented later in Chapter 7. Example 1.6.2 The free response of the damped single-degree-of-freedom system in Figure 1.9 with a mass of 2 kg is recorded to be of the form given in Figure 1.35. A static deflection test is performed and the stiffness is determined to be 1.5 * 103 N > m. The displacements at t1 and t2 are measured to be 9 and 1 mm, respectively. Calculate the damping coefficient. Solution From the definition of the logarithmic decrement δ = lnc From equation (1.83), ζ = x(t 1) x(t 2) d = lnc 2.1972 24π2 + 2.19722 9 mm d = 2.1972 1 mm = 0.33 or 33, Also, ccr = 22km = 22(1.5 * 103 N>m)(2 kg) = 1.095 * 102 kg>s And from equation (1.30) the damping coefficient becomes c = ccr ζ = ( 1.095 * 102 ) (0.33) = 36.15 kg>s n www.TechnicalBooksPDF.com 62 Introduction to Vibration and the Free Response Chap. 1 Example 1.6.3 Mass and stiffness are usually measured in a straightforward manner as shown in Section 1.3. However, there are certain circumstances that preclude using these simple methods. In these cases, a measurement of the frequency of oscillation both before and after a known amount of mass is added can be used to determine the mass and stiffness of the original system. Suppose then that the frequency of the system in Figure 1.36(a) is measured to be 2 rad>s and the frequency of Figure 1.36(b) with an added mass of 1 kg is known to be 1 rad>s. Calculate m and k. 1  2 rad/s k 0  1 rad/s k m m0  1 kg m Figure 1.36 A schematic of using added mass (b) and frequency measurements to determine an unknown mass, m, and stiffness, k, of the original system (a). m0 (a) Solution (b) From the definition of natural frequency k Am and ω0 = 1 = 4m = k and m + 1 = k ω1 = 2 = k Am + 1 Solving for m and k yields or m = 1 kg 3 and k = 4 N>m 3 This formulation can also be used to determine changes in mass of a system. As an example, the frequency of oscillation of low amplitude vibration of a hospital patient in bed can be used to monitor the change in the patient’s weight (mass) without having to move the patient from the bed. In this case the mass m0 is considered to be the change in mass of the original system. If the original mass and frequency are known, measurement of the frequency ω0 can be used to determine the change in mass m0. Given that the original weight is 120 lb (54.4 kg), the original frequency is 100.4 Hz, and the frequency of the patient-bed system changes to 100 Hz, determine the change in the patient’s weight. www.TechnicalBooksPDF.com Sec. 1.7 63 Design Considerations From the two frequency relations ω21m = k and ω20(m + m0) = k Thus, ω21m = ω20(m + m0). Solving for the change in mass m0 yields m0 = m a ω21 ω20 - 1b Multiplying by g and converting the frequency to hertz yields W0 = W a or W0 = 120 lb c a f 21 f 20 - 1b 100.4 Hz 2 b - 1d 100 Hz = 0.96 lb (0.4kg) Since the frequency decreased, the patient gained almost a pound. An increase in frequency would indicate a loss of weight. n Measurement of m, c, k, ωn, and ζ is used to verify the mathematical model of a system and for a variety of other reasons. Measurement of vibrating systems forms an important aspect of the activity in industry related to vibration technology. Chapter 7 is specifically devoted to measurement, however comments on vibration measurements are mentioned throughout the remaining chapters. 1.7 DESIGN CONSIDERATIONS This section introduces the idea of designing vibration systems, which forms the topic of Chapter 5. Design in vibration refers to adjusting the physical parameters of a device to cause its vibration response to meet a specified shape or performance criteria. For instance, consider the response of the single-degree-of-freedom system of Figure 1.9. The shape of the response is somewhat determined by the value of the damping ratio in the sense that the response is either overdamped, underdamped, or critically damped (ζ 7 1, ζ 6 1, ζ = 1, respectively). The damping ratio, in turn, depends on the values of m, c, and k. A designer may choose these values to produce the desired response. Section 1.5 on stiffness considerations is actually an introduction to design as well. The formulas given there for stiffness, in terms of modulus and geometric www.TechnicalBooksPDF.com 64 Introduction to Vibration and the Free Response Chap. 1 dimensions, can be used to design a system that has a given natural frequency. Example 1.5.2 points out one of the important problems in design, that often the properties that we are interested in designing for (frequency in this case) are very sensitive to operational changes. In Example 1.5.2, the frequency changes a great deal as the airplane consumes fuel. Another important issue in design often focuses on using devices that are already available. For example, the rules given in Figure 1.31 are design rules for producing a desired value of spring constant from a set of “available” springs by placing them in certain combinations, as illustrated in Example 1.5.6. Design work in engineering often involves using available products to produce configurations (or designs) that suit a particular application. In the case of spring stiffness, springs are usually mass produced, and hence inexpensive, in only certain discrete values of stiffness. The formulas given for parallel and series connections of springs are then used to produce the desired stiffness. If cost is not a restriction, then formulas such as those given in Table 1.3 may be used to design a single spring that meets the stated stiffness requirements. Of course, designing a spring–mass system to have a desired natural frequency may not produce a system with an acceptable static deflection. Thus, the design process becomes complicated. Design is one of the most active and exciting disciplines in engineering because it often involves compromise and choice with many acceptable solutions. Unfortunately, the values of m, c, and k have other constraints. The size and material of which the device is made determine these parameters. Hence, the design procedure becomes a compromise. For example, geometric limitations might cause the mass of a device to be between 2 and 3 kg, and for static displacement conditions, the stiffness may be required to be greater than 200 N>m. In this case, the natural frequency must be in the interval 8.16 rad>s … ωn … 10 rad>s (1.84) This severely limits the design of the vibration response, as illustrated in the following example. Example 1.7.1 Consider the system of Figure 1.9 with mass and stiffness properties as summarized by inequality (1.84). Suppose that the system is subject to an initial velocity that is always less than 300 mm>s, and to an initial displacement of zero (i.e., x0 = 0, v0 … 300 mm>s). For this range of mass and stiffness, choose a value of the damping coefficient such that the amplitude of vibration is always less than 25 mm. Solution This is a design-oriented example, and hence, as is typical of design calculations, there is not a nice, clean formula to follow. Rather, the solution must be obtained using theory and parameter studies. First, note that for zero initial displacement, the response may be written from equation (1.38) as x(t) = v0 -ζω t e n sin (ωdt) ωd www.TechnicalBooksPDF.com Sec. 1.7 65 Design Considerations Also note that the amplitude of this periodic function is v0 -ζω t e n ωd Thus, for small ωd, the amplitude is larger than for larger ωd. Hence for the range of frequencies of interest, it appears that the worst case (largest amplitude) will occur for the smallest value of the frequency (ωn = 8.16 rad>s). Also, the amplitude increases with v0 so that using v0 = 300 mm>s will ensure that amplitude is a large as possible. Now, v0 and ωn are fixed, so it remains to be investigated how the maximum value of x(t) varies as the damping ratio is varied. One approach is to compute the amplitude of the response at the first peak. From Figure 1.10 the largest amplitude occurs at the first time the derivative of x(t) is zero. Taking the derivative of x(t) and setting it equal to zero yields the expression for the time to the first peak: ωde -ζωnt cos (ωdt) - ζωne -ζωnt sin (ωdt) = 0 Solving this for t and denoting this value of time by Tm yields Tm = ωd 21 - ζ2 1 1 tan-1 a tan-1 a b = b ωd ωd ζωn ζ The value of the amplitude of the first (and largest) peak is calculated by substituting the value of Tm into x(t), resulting in Am(ζ) = x(Tm) = v0 ωn 21 - ζ2 Simplifying yields Am(ζ) = ζ e - 21 - ζ2 tan -1 a 21 - ζ2 b ζ sin atan-1 a v0 - ζ tan-1 a 21 - ζ ζ e 21 - ζ2 ωn 2 21 - ζ2 bb ζ b For fixed initial velocity (the largest possible) and frequency (the lowest possible), this value of Am(ζ) determines the largest value that the highest peak will have as ζ varies. The exact value of ζ that will keep this peak, and hence the response, at or below 25 mm, can be determined by numerically solving Am(ζ) = 0.025 (m) for a value of ζ. This yields ζ = 0.281. Using the upper limit of the mass values (m = 3 kg) then yields the value of the required damping coefficient: c = 2mωnζ = 2(3)(8.16)(0.281) = 13.76 kg>s For this value of damping, the response is never larger than 25 mm. Note that if there is no damping, the same initial conditions produce a response of amplitude A = v0 > ωn = 37 mm. n As another example of design, consider the problem of choosing a spring that will result in a spring–mass system having a desired or specified frequency. The formulas of Section 1.5 provide a means of designing a spring to have a specified stiffness in terms of the properties of the spring material (modulus) and its geometry. The following example illustrates this concept. www.TechnicalBooksPDF.com 66 Introduction to Vibration and the Free Response Chap. 1 Example 1.7.2 Consider designing a helical spring such that when attached to a 10-kg mass, the resulting spring–mass system has a natural frequency of 10 rad/s (about 1.6 Hz). Solution From the definition of the natural frequency, the spring is required to have a stiffness of k = ω2nm = (10)2(10) = 103 N>m The stiffness of a helical spring is given by equation (1.66) to be k = 103 N>m = Gd 4 64nR3 6.4 * 104 = or Gd 4 nR3 This expression provides the starting point for a design. The choices of variables that affect the design are: the type of material to be used (hence various values of G); the diameter of the material, d; the radius of the coils, R; and the number of turns, n. The choices of G and d are, of course, restricted by available materials, n is restricted to be an integer, and R may have restrictions dictated by the size requirements of the device. Here it is assumed that steel of 1-cm diameter is available. The shear modulus of steel is about G = 8.0 * 1010 N>m2 so that the stiffness formula becomes 6.4 * 104 N>m = ( 8.0 * 1010 N>m2 )( 10-2 m ) 4 nR3 or nR3 = 1.25 * 10-2 If the coil radius is chosen to be 10 cm, this yields that the number of turns should be n = 1.25 * 10-2 m3 = 12.5 or 13 10-3 m3 Thus, if 13 turns of 1-cm-diameter steel are coiled at a radius of 10 cm, the resulting spring will have the desired stiffness and the 10-kg mass will oscillate at approximately 10 rad>s. To get an exact answer, the modulus of steel must be modified. This can be done through the use of different alloys of steel, but would become expensive. So depending on the precision needed for a given application, modifying the type of steel used may or may not be practical. n In Example 1.7.2, several variables were chosen to produce a desired design. In each case the design variables (such as d, R, etc.) are subject to constraints. Other aspects of vibration design are presented throughout the text as appropriate. There are no set rules to follow in design work. However, some organized approaches to design are presented later in Chapter 5. The following example illustrates another difficulty in design by examining what happens when operating conditions are changed after the design is over. www.TechnicalBooksPDF.com Sec. 1.7 67 Design Considerations Example 1.7.3 As a final example, consider modeling the vertical suspension system of a small sports car, as a single-degree-of-freedom system of the form $ # mx + cx + kx = 0 where m is the mass of the automobile and c and k are the equivalent damping and stiffness of the four-shock-absorber–spring systems. The car deflects the suspension system 0.05 m under its own weight. The suspension is chosen (designed) to have a damping ratio of 0.3. a) If the car has a mass of 1361 kg (mass of a Porsche Boxster), calculate the equivalent damping and stiffness coefficients of the suspension system. b) If two passengers, a full gas tank, and luggage totaling 290 kg are in the car, how does this affect the effective damping ratio? Solution The mass is 1361 kg and the natural frequency is ωn = k A 1361 so that k = 1361 ω2n At rest, the car’s springs are compressed an amount Δ, called the static deflection, by the weight of the car. Hence, from a force balance at static equilibrium, mg = kΔ, so that k = mg Δ and ωn = g 9.8 k = = = 14 rad>s AΔ A 0.05 Am The stiffness of the suspension system is thus k = 1361(14)2 = 2.668 * 105 N>m For ζ = 0.3, equation (1.30) becomes c = 2ζmωn = 2(0.3)(1361)(14) = 1.143 * 104 kg/s Now if the passengers and luggage are added to the car, the mass increases to 1361 + 290 = 1651 kg. Since the stiffness and damping coefficient remain the same, the new static deflection becomes Δ = mg k = 1651(9.8) 2.668 * 105 ≈ 0.06 m www.TechnicalBooksPDF.com 68 Introduction to Vibration and the Free Response Chap. 1 The new frequency becomes ωn = g AΔ = 9.8 = 12.78 rad>s A 0.06 From equations (1.29) and (1.30), the damping ratio becomes ζ = 1.143 * 104 1.143 * 104 c = = = 0.27 ccr 2mωn 2(1651)(12.78) Thus the car with passengers, fuel, and luggage will exhibit less damping and hence larger amplitude vibrations in the vertical direction. The vibrations will take a little longer to die out. n Note that this illustrates a difficulty in design problems, in the sense that the car cannot be damped at exactly the same value for all passenger situations. In this case, even if ζ = 0.3 is desirable, it really cannot be achieved. Designs that do not change dramatically when one parameter changes a small amount are said to be robust. This and other design concepts are discussed in greater detail in Chapter 5, as the analytical skills developed in the next few chapters are required first. 1.8 STABILITY In the preceding sections, the physical parameters m, c, and k are all considered to be positive in equation (1.27). This allows the treatment of the solutions of equation (1.27) to be classified into three groups: overdamped, underdamped, or critically damped. The case with c = 0 provides a fourth class called undamped. These four solutions are all well behaved in the sense that they do not grow with time and their amplitudes are finite. There are many situations, however, in which the coefficients are not positive, and in these cases the motion is not well behaved. This situation refers to the stability of solutions of a system. Recalling that the solution of the undamped case (c = 0) is of the form A sin(ωnt + ϕ), it is easy to see that the undamped response is bounded. That is, if 0 x(t) 0 denotes the absolute value of x, then 0 x(t) 0 … A 0 sin(ωnt + ϕ) 0 = A = 1 3ω2nx20 + v20 ωn (1.85) for every value of t. Thus 0 x(t) 0 is always less than some finite number for all time and for all finite choices of initial conditions. In this case, the response is well behaved and said to be stable (sometimes called marginally stable). If, on the other www.TechnicalBooksPDF.com Sec. 1.8 69 Stability hand, the value of k in equation (1.2) is negative and m is positive, the solutions are of the form which increases without bound as t does. In this case 0 x(t) 0 no longer remains finite and such solutions are called divergent or unstable. Figure 1.37 illustrates a stable response and Figure 1.38 illustrates an unstable, or divergent, response. Consider the response of the damped system of equation (1.27) with positive coefficients. As illustrated in Figures 1.10, 1.11, 1.12, and 1.13, it is clear that x(t) approaches zero as t becomes large because of the exponential-decay terms. Such solutions are called asymptotically stable. Again, if c or k is negative (and m is positive), the motion grows without bound and becomes unstable, as in the undamped case. In the damped case, however, the motion may be unstable in one of two ways. Similar to overdamped solutions and underdamped solutions, the motion may grow without bound and may or may not oscillate. The nonoscillatory case is called divergent instability and the oscillatory case is called flutter instability, or sometimes just flutter. Flutter instability is sketched in Figure 1.39. The trend of growing without bound for large t continues in Figures 1.38 and 1.39, even though the figure stops. These types of instability occur in a variety of situations, often called self-excited vibrations, and require some source of energy. The following example illustrates such instabilities. x(t) = A sinh ωnt + B cosh ωnt (1.86) Displacement (mm) 1.0 0.5 Time (s) 0.0 5 10 15 20 0.5 1.0 Figure 1.37 An example response of a stable single-degree-of-freedom system. Displacement (mm) Time (s) Figure 1.38 An example response of a unstable single-degree-of-freedom system (divergence). www.TechnicalBooksPDF.com 70 Introduction to Vibration and the Free Response Chap. 1 Displacement (mm) 8 6 4 2 0 2 4 6 Time (s) 8 2 4 6 8 10 12 14 16 18 20 Figure 1.39 An example response of a unstable single-degree-of-freedom system which also oscillates, called flutter instability. Example 1.8.1 Consider the inverted pendulum connected to two equal springs, shown in Figure 1.40. + m m k k 2 kl sin  2 l 2 mg  O (a) fp O (b) Figure 1.40 (a) An inverted pendulum oscillator and (b) its free-body diagram. Here fp is the total reaction force at the pin. The pendulum has length l. Solution Assume that the springs are undeflected when in the vertical position and that the mass m of the ball at the end of the pendulum rod is substantially larger than the mass of the rod itself, so that the rod is considered to be massless. The total length of the rod is l and the springs are attached at the point l>2. Summing the moments around the pivot point (point O) yields $ ml 2θ = a M0 There are three forces acting. The spring force is the stiffness times the displacement (kx) where the displacement x is (l>2) sin θ. There are two such springs, so the total force acting on the pendulum by the springs is kl sin θ. This force acts through a moment arm of (l>2) cos θ. The gravitational force acting on the mass m is mg acting through a moment arm of l sin θ. Thus summing moments about point O yields $ kl 2 ml 2θ = - a sin θb cos θ + mgl sin θ 2 www.TechnicalBooksPDF.com Sec. 1.8 71 Stability and the equation of motion becomes $ kl 2 ml 2θ + a sin θb cos θ - mgl sin θ = 0 2 (1.87) For values of θ less than about π>20, sin θ and cos θ can be approximated by sin θ ≅ θ and cos θ ≅ 1. Applying this approximation to equation (1.87) yields $ kl 2 θ - mgl θ = 0 ml 2θ + 2 which upon rearranging becomes $ 2mlθ(t) + (kl - 2mg)θ(t) = 0 where θ is now restricted to be small (smaller than π>20). If k, l, and m are all such that the coefficient of θ, called the effective stiffness, is negative, that is, if kl - 2mg 6 0 the pendulum motion will be unstable by divergence, as illustrated in Figure 1.38. n Example 1.8.2 The vibration of an aircraft wing can be crudely modeled as $ # # mx + cx + kx = γx where m, c, and k are the mass, damping, and stiffness values of the wing, respectively, # modeled as a single-degree-of-freedom system, and where the term γx is an approximate model of the aerodynamic forces on the wing (γ 7 0 for high speed). Solution Rearranging the equation of motion yields $ # mx + (c - γ)x + kx = 0 If γ and c are such that c - γ 7 0, the system is asymptotically stable. However, if γ is such that c - γ 6 0, then ζ = (c - γ)>2mωn 6 0 and the solutions are of the form x(t) = Ae -ζωnt sin(ωdt + ϕ) where the exponent (–ζωnt) 7 0 for all t 7 0 because of the negative damping term. Such solutions increase exponentially with time, as indicated in Figure 1.39. This is an example of flutter instability and self-excited oscillation. n This brief introduction to stability applies to systems that can be treated as linear and homogenous. More complex definitions of stability are required for forced systems and for nonlinear systems. The notions of stability can be thought of in terms of changing energy: stable systems having constant energy, unstable systems having increasing energy, and asymptotically stable systems having decreasing energy. Stability can also be thought of in terms of initial conditions and this is discussed in Section 1.10 where a brief introduction to nonlinear vibrations is given. An essential difference between linear and nonlinear systems lies in their respective stability properties. www.TechnicalBooksPDF.com 72 Introduction to Vibration and the Free Response Chap. 1 1.9 NUMERICAL SIMULATION OF THE TIME RESPONSE So far, most of the vibration problems examined have all been cast as linear differential equations that have solutions that can be determined analytically. These solutions are often plotted versus time in order to visualize the physical vibration and obtain an idea of the nature of the response. However, there are many more complex and nonlinear systems that are either difficult or impossible to solve analytically (i.e., that do not have closed-form solutions for the displacement as a function of time). The nonlinear pendulum equation given in Example 1.1.1 is “linearized” by making the approximation sin(θ) = θ to provide a system which is simple to solve (having the same analytical form as a linear spring–mass system). The approximation made to linearize the pendulum equation is only valid for certain, relatively small initial conditions. The approximation of sin(θ) = θ requires that the initial displacement and velocity are such that θ(t) remains less than about 10°. For cases with larger initial conditions, a numerical integration routine may be used to compute and plot a solution of the nonlinear equation of motion. Numerical integration can be used to compute the solutions of a variety of difficult problems and is introduced here on simple problems that have known analytical solutions so that the nature of the approximation can be discussed. Later, numerical integration will be used for problems not having closed-form solutions. The free response of any single-degree-of-freedom system may easily be computed by simple numerical means such as Euler’s method or Runge–Kutta methods. This section examines the use of these common numerical methods for solving vibration problems that are difficult to solve in closed form. Runge–Kutta schemes can be found on calculators and in most common mathematical software packages such as Mathematica, Mathcad, Maple, and MATLAB. Alternately the numerical schemes may be programmed in more traditional languages, such as FORTRAN, or into spreadsheets. This section reviews the use of numerical methods for solving differential equations and then applies these methods to the solution of several vibration problems considered in the previous sections. These techniques are then used in the following section to analyze the response of nonlinear systems. Appendix F introduces the use of Mathematica, Mathcad, and MATLAB for numerical integration and plotting. Many modern curriculums introduce these methods and codes early in the engineering curriculum, in which case this section can be skipped or used as a quick review. There are many schemes for numerically solving ordinary differential equations, such as those of vibration analysis. Two numerical solution schemes are presented here. The basis of numerical solutions of ordinary differential equations is to essentially undo calculus by representing each derivative by a small but finite difference (recall the definition of a derivative from calculus given in Window 1.6). A numerical solution of an ordinary differential equation is a procedure for constructing approximate discrete values: x1, x2, c, xn, of the solution x(t) at the discrete values of time: t0 6 t1 6 t2 c 6 tn. Thus a numerical procedure produces a list of discrete www.TechnicalBooksPDF.com Sec. 1.9 numerical Simulation of the time Response 73 values xi = x(ti) that approximate the exact solution, which is the continuous function of time x(t). The initial conditions of the vibration problem of interest form the starting point of computing a numerical solution. For a single-degree-of-freedom system of the form $ # # mx + cx + kx = 0, x(0) = x0 x(0) = v0 (1.88) the initial values x0 and v0 form the first two points of the numerical solution. Let Tf be the total length of time over which the solution is of interest (i.e., the equation is to be solved for values of t between t = 0 and t = Tf). The time interval Tf - 0 is then divided up into n intervals (so that Δt = Tf >n). Then equation (1.88) is calculated at the values of t0 = 0, t1 = Δt, t2 = 2Δt,c, tn = nΔt = Tf to produce an approximate representation, or simulation, of the solution. The concept of a numerical solution is easiest to grasp by first examining the numerical solution of a first-order scalar differential equation. To this end, consider the first-order differential equation # x(0) = x0 (1.89) x(t) = ax(t) The Euler method proceeds from the definition of the slope form of the derivative given in Window 1.6, before the limit is taken: xi + 1 - xi = axi Δt (1.90) where xi denotes x(ti), xi + 1 denotes x(ti + 1), and Δt indicates the time interval between ti and ti + 1 (i.e., Δt = ti+ 1 - ti). This expression can be manipulated to yield xi + 1 = xi + Δt(axi) (1.91) This formula computes the discrete value of the response xi+ 1 from the previous value xi, the time step Δt, and the system’s parameter a. This numerical solution is called an Euler or tangent line method. The following example illustrates the use of the Euler formula for computing a solution. Window 1.6 Definition of the Derivative The definition of a derivative of x(t) at t = ti is dx(t i) = dt lim Δt S 0 x(t i + 1) - x(t i) Δt where ti+ 1 = ti + Δt and x(t) is continuous. www.TechnicalBooksPDF.com 74 Introduction to Vibration and the Free Response Chap. 1 Example 1.9.1 # Use the Euler formula to compute the numerical solution of x = - 3x, x(0) = 1 for various time increments in the time interval 0 to 4, and compare the results to the exact solution. # TABLE 1.4 CoMPARISon oF tHE ExACt SolutIon oF x = - 3x, x(0) = 1 to tHE SolutIon obtAInED by tHE EulER MEtHoD wItH lARgE tIME StEP (Δt = 0.5) FoR tHE IntERVAl t = 0 to 4 Index 0 1 2 3 4 5 6 7 8 Elapsed time Exact Euler 0 0.5000 1.0000 1.5000 2.0000 2.5000 3.0000 3.5000 4.0000 1.0000 0.2231 0.0498 0.0111 0.0025 0.0006 0.0001 0.0000 0.0000 1.0000 –0.5000 0.2500 –0.1250 0.0625 –0.0312 0.0156 –0.0078 0.0039 Absolute error 0 0.7231 0.2002 0.1361 0.0600 0.0318 0.0155 0.0078 0.0039 Solution First, the exact solution can be obtained by direct integration or by assuming a solution of the form x(t) = Aeλt. Substitution of this assumed form into the equa# tion x = - 3x yields Aλeλt = - 3Aeλt, or λ = - 3, so that the solution is of the form x(t) = Ae–3t. Applying the initial conditions x(0) = 1 yields A = 1. Hence the analytical solution is simply x(t) = e–3t. Next, consider a numerical solution using the Euler method suggested by equation (1.91). In this case the constant a = - 3, so that xi +1 = xi + Δt( - 3xi). Suppose that a very crude time step is taken (i.e., Δt = 0.5) and the solution is formed over the interval from t = 0 to t = 4. Then Table 1.4 illustrates the values obtained from equation (1.91): x0 = 1 x1 = x0 + (0.5)( - 3)(x0) = - 0.5 x2 = - 0.5 - (1.5)( - 0.5) = 0.25 f forms the column marked “Euler.” The column marked “Exact” is the value of e-3t at the indicated elapsed time for a given index. Note that while the Euler approximation gets close to the correct final value, this value oscillates around zero while the exact value does not. This points out a possible source of error in a numerical solution. On the other hand, if Δt is taken to be very small, the difference between the solution obtained by the Euler equation and the exact solution becomes hard to see, as Figure 1.41 illustrates. Figure 1.41 is a plot of x(t) obtained via the Euler formula for Δt = 0.1. Note that it looks very much like the exact solution x(t) = e-3t. n www.TechnicalBooksPDF.com Sec. 1.9 75 numerical Simulation of the time Response x(t) 1 Exact 0.5 t  0.1 s t  0.5 s 0 Time (s) 0.5 0.5 1 1.5 2 2.5 3 3.5 4 # Figure 1.41 A plot of x(t i) versus ti for x(t) = - 3x using various time steps in equation (1.91) with x(0) = 1. It is important to note from the example that two sources of error are present in computing the solution of a differential equation using a numerical scheme such as the Euler method. The first is called the truncation error, which is the difference between the exact solution and the solution obtained by the Euler approximation. This is the error indicated in the last column of Table 1.4. Note that this error accumulates as the index increases because the value at each discrete time is determined by the previous value, which is already in error. This can be somewhat controlled by the time step and the nature of the formula. The other source of error is the roundoff error due to machine arithmetic. This is, of course, controlled by the computer and its architecture. Both sources of error can be significant. The successful use of a numerical method requires an awareness of both sources of errors in interpreting the results of a computer simulation of the solution of any vibration problem. The Euler method can be improved upon by decreasing the step size, as Example 1.9.1 illustrates. Alternatively, a more accurate procedure can be used to improve the accuracy (smaller formula error) without decreasing the step size Δt. Several methods exist (such as the improved Euler method and various Taylor series methods) and are discussed in Boyce and DiPrima (2009), for instance. Only the Runge–Kutta method is discussed and used here. The Runge–Kutta method was developed by two different researchers from about 1895 to 1901 (C. Runge and M. W. Kutta). The Runge–Kutta formulas (there are several) involve a weighted average of values of the right-hand side of the differential equation taken at different points between the time intervals ti and ti + Δt. The derivations of various Runge–Kutta formulas are tedious but straightforward and are not presented here (see Boyce and DiPrima 2009). One useful formulation www.TechnicalBooksPDF.com 76 Introduction to Vibration and the Free Response Chap. 1 # can be stated for the first-order problem x = f(x, t), x(0) = x0, where f is any scalar function (linear or nonlinear) as xn + 1 = xn + Δt (k + 2kn2 + 2kn3 + kn4) 6 n1 (1.92) where kn1 = f (xn, t n) kn2 = f axn + kn3 = f axn + Δt Δt kn1, t n + b 2 2 Δt Δt k ,t + b 2 n2 n 2 kn4 = f (xn + Δtkn3, t n + Δt) The sum in parentheses in equation (1.92) represents the average of six numbers, each of which looks like a slope at a different time; for instance, the term kn1 is the slope of the function at the “left” end of the time interval. Such formulas can be enhanced by treating Δt as a variable, Δti. At each time step ti, the value of Δti is adjusted based on how rapidly the solution x(t) is changing. If the solution is not changing very rapidly, a large value of Δti is allowed without increasing the formula error. On the other hand, if x(t) is changing rapidly, a small Δti must be chosen to keep the formula error small. Such step sizes can be chosen automatically as part of the computer code for implementing the numerical solution. The Runge–Kutta and Euler formulas just listed can be applied to vibration problems by noting that the most general (damped) vibration problem can be put into a first-order form. Returning to a damped system of the form # # $ # mx(t) + cx(t) + kx(t) = 0 x(0) = x0, x(0) = x0 (1.93) the Euler method of equation (1.91) can be applied by writing this expression as two first-order equations. To this end, divide equation (1.93) by the mass m, and # define two new variables by x1 = x(t) and x2 = x(t). Then differentiate the definition of x1(t), rearrange equation (1.93), and replace x and its derivative with x1 and x2 to get # x1(t) = x2(t) k c # x (t) x2(t) = - x2(t) m m 1 (1.94) # subject to the initial conditions x1(0) = x0 and x2(0) = x0. The two coupled firstorder differential equations given in (1.94) may be written as a single expression by www.TechnicalBooksPDF.com Sec. 1.9 77 numerical Simulation of the time Response using a vector and matrix form determined by first defining the vector 2 * 1 x(t) and the 2 * 2 matrix A by 0 A = C k m 1 cS m x(t) = c x1(t) d x2(t) x(0) = c x1(0) d x2(0) (1.95) The matrix A defined in this way is called the state matrix and the vector x is called the state vector. The position x1 and the velocity x2 are called the state variables. Using these definitions (see Appendix C), the rules of vector differentiation (element by element) and multiplication of a matrix times a vector, equations (1.95) may be written as # x(t) = Ax(t) (1.96) subject to the initial condition x(0). Now the Euler method of numerical solution given in equation (1.91) can be applied directly to this vector-matrix formulation of Equation (1.96), by simply calling the scalar xi, the vector xi, and replacing the scalar a with the matrix A to produce x(t i + 1) = x(t i) + ΔtAx(t i) (1.97) This, along with the initial condition x(0), defines the Euler formula for integrating the general single-degree-of-freedom vibration problem described in equation (1.92). Equation (1.97) allows the time response to be computed and plotted. As suggested, the Euler-formula method can be greatly improved by using a Runge–Kutta program. For instance, MATLAB has two different Runge–Kutta-based simulations: ode23 and ode45. These are automatic step-size integration methods (i.e., Δt is chosen automatically). The Engineering Vibration Toolbox has one fixed-step Runge–Kutta-based method, VTB1_3, for comparison. The M-file ode23 uses a simple second- and third-order pair of formulas for medium accuracy and ode45 uses a fourthand fifth-order pair for greater accuracy. Each of these corresponds to a formulation similar to that expressed in equations (1.92) with more terms and a variable step size Δt. In general, the Runge–Kutta simulations are of a higher quality than those obtained by the Euler method. Example 1.9.2 $ # Use the ode45 function to simulate the response to 3x + x + 2x = 0 subject to the # initial conditions x(0) = 0, x(0) = 0.25 over the time interval 0 … t … 20. Solution yields The first step is to write the equation of motion in first-order form. This # x1 = x2 # x2 = - 23 x1 - 1 3 x2 www.TechnicalBooksPDF.com 78 Introduction to Vibration and the Free Response Chap. 1 Next, an M-file is created to store the equations of motion. An M-file is created by choosing a name, say, sdof.m, and entering function xdot = sdof(t,x); xdot = zeros(2,1); xdot(1) = x(2); xdot(2) = –(2/3)*x(1) – (1/3)*x(2); Next, go to the command mode and enter t0 = 0;tf = 20; x0 = [0 0.25]; [t,x] = ode45(‘sdof’,[t0 tf],x0); plot(t,x) The first line establishes the initial, (t0), and final, (tf), times. The second line creates the vector containing the initial conditions x0. The third line creates the two vectors t, containing the time history, and x, containing the response at each time increment in t, by calling ode45 applied to the equations set up in sdof. The fourth line plots the vector x versus the vector t. This is illustrated in Figure 1.42. Displacement (solid) and velocity (dashed) 0.3 0.2 0.1 0 0.1 0.2 0 5 10 15 20 Time Figure 1.42 A plot of the displacement x(t) of the single-degree-of freedom system # of Example 1.9.2 (solid line) and the corresponding velocity x(t) (dashed line). n The preceding example may also be solved using Mathematica, Mathcad, and Maple, by writing a FORTRAN routine, or by using any number of other computer codes or programmable calculators. The following example illustrates the commands required to produce the result of Example 1.9.2 using Mathematica and again using Mathcad. These approaches are then used in the next section to examine the response of certain nonlinear vibration problems. www.TechnicalBooksPDF.com Sec. 1.9 79 numerical Simulation of the time Response Example 1.9.3 Solve Example 1.9.2 using the Mathematica program. Solution The Mathematica program uses an iterative method to compute the solution and accepts the second-order form of the equation of motion. The text after the prompt In[1]:= is typed by the user and returns the solution stored in the variable x[t]. Mathematica has several equal signs for different purposes. In the argument of the NDSolve function, the user types in the differential equation to be solved, followed by the initial conditions, the name of the variable (response), and the name of the independent variable followed by the interval over which the solution is sought. NDSolve computes the solution and stores it as an interpolating function; hence the code returns the plot following the output prompt Out[1]=. The plot command requires the name of the interpolating function returned by NDSolve, x[t] in this case, the independent variable, t, and the range of values for the independent variable. In[1]:= NDSolve[{x''[t]+(1/3)*x'[t]+(2/3)*x[t]==0,x'[0]==0.25,x[0]==0}, x,{t,0,20}]; Plot[Evaluate[x[t]/.%],{t,0,20}] Out[1]={{x–>InterpolatingFunction[{{0.,20.}},<>]}} Out[2]= 0.2 0.15 0.1 0.05 5 10 15 20 0.05 0.1 n Example 1.9.4 Solve Example 1.9.2 using the Mathcad program. Solution The Mathcad program uses a fixed time step Runge–Kutta solution and returns the solution as a matrix with the first column consisting of the time step, the second column containing the response, and the third column containing the velocity response. www.TechnicalBooksPDF.com 80 Introduction to Vibration and the Free Response Chap. 1 First type in the initial condition vector: y : = y := c 0 d 0.25 Then type in the system in first-order form: D(t,y) : = D(t,y) := C y1 S 1 2 - a y1b y 3 3 0 Solve using Runge–Kutta: Z := rkfixed(y,0,20,1000,D) Name the time vector from the Runge–Kutta matrix solution: t := Z<0> Name the displacement vector from the Runge–Kutta matrix solution: x := Z<1> Name the velocity vector from the Runge–Kutta matrix solution: dxdt := Z<2> Plot the solutions. 0.2 x dxdt 0 0.2 5 10 15 20 t n The use of these computational programs to simulate the response of a vibrating system is fairly straightforward. Further information on using each of these programs can be found in Appendix F or by consulting manuals or any one of numerous books written on using these codes to solve various math and engineering problems. You are encouraged to reproduce Example 1.9.4 and then repeat the problem for various www.TechnicalBooksPDF.com Sec. 1.10 81 Coulomb Friction and the Pendulum different values of the initial conditions and coefficients. In this way, you can build some intuition and understanding of vibration phenomena and how to design a system to produce a desired response. A note about the use of the codes presented in this text is in order. At the time of printing this edition, all the codes ran as typed. However, each year, or sometimes more frequently, companies who provide these codes update them and in so doing they often change syntax. These changes can be found on the companies’ websites and should be checked if difficulty is encountered in using the codes presented here. 1.10 COULOMB FRICTION AND THE PENDULUM In the previous sections, all of the systems considered are linear (or linearized) and have solutions that can be obtained by analytical means. In this section, two common systems are analyzed that are nonlinear and do not have simple analytical solutions. The first is a spring–mass system with sliding friction (Coulomb damping), and the second is the full nonlinear pendulum equation. In each case a solution is obtained by using the numerical integration techniques introduced in Section 1.9. The ability to compute the solution to general nonlinear systems using these numerical techniques allows us to consider vibration in more complicated configurations. Nonlinear vibration problems are much more complex than linear systems. Their numerical solutions, however, are often fairly straightforward. Several new phenomena result when nonlinear terms are considered. Most notably, the idea of a single equilibrium point of a linear system is lost. In the case of Coulomb damping, a continuous region of equilibrium positions exists. In the case of the nonlinear pendulum, an infinite number of equilibrium points result. This single fact greatly complicates the analysis, measurement, and design of vibrating systems. Coulomb damping is a common damping effect, often occurring in machines, that is caused by sliding friction or dry friction. It is characterized by the relation -μN # fc = Fc(x) = c 0 μN # x 7 0 # x = 0s # x 6 0 where fc is the dissipation force, N is the normal force (see any introductory physics text), and μ is the coefficient of sliding friction (or kinetic friction). Figure 1.43 is a schematic of a mass m sliding on a surface and connected to a spring of stiffness k. The frictional force fc always opposes the direction of motion causing a system with Coulomb friction to be nonlinear. Table 1.5 lists some measured values of the coefficient of kinetic friction for several different sliding objects. Summing forces www.TechnicalBooksPDF.com 82 Introduction to Vibration and the Free Response Chap. 1 x0 x(t) k m Figure 1.43 A spring–mass system sliding on a surface of kinetic coefficient of friction μ. TABLE 1.5 APPRoxIMAtE CoEFFICIEntS oF FRICtIon FoR VARIouS objECtS SlIDIng togEtHER Material Kinetic Static Metal on metal (lubricated) Wood on wood Steel on steel (unlubricated) Rubber on steel 0.07 0.2 0.3 1.0 0.09 0.25 0.75 1.20 in part (a) of Figure 1.44 in the x direction yields that (note that the mass changes direction when the velocity passes through zero) $ # mx + kx = μ mg for x 6 0 (1.98) Here the sum of the forces in the vertical direction yields the fact that the normal force N is just the weight, mg, where g is the acceleration due to gravity (not the case if m is on an inclined plane as N is no longer along the same direction as W). In a similar fashion, summing forces in part (b) of Figure 1.44 yields # $ mx + kx = -μ mg for x 7 0 (1.99) . x(t) . x(t) W  mg W  mg kx kx   fc  N fc  N  mg  mg N (b) N (a) Figure 1.44 A free-body diagram of the forces acting on the sliding block system # of Figure 1.43: (a) mass moving to the right (x 6 0), (b) mass moving to the right # (x 7 0). From the y direction: N = mg. www.TechnicalBooksPDF.com Sec. 1.10 83 Coulomb Friction and the Pendulum # Since the sign of x determines the direction in which the opposing frictional force acts, equations (1.98) and (1.99) can be written as the single equation $ # mx + μ mg sgn(x) + kx = 0 (1.100) where sgn(τ) denotes the signum function, defined to have the value 1 for τ 7 0, - 1 for τ 6 0, and 0 for τ = 0. This equation cannot be solved directly using methods such as the variation of parameters or the method of undetermined coefficients. This is because equation (1.100) is a nonlinear differential equation. Rather, equation (1.100) can be solved by breaking the time intervals up into segments corresponding to the changes # in direction of motion (i.e., at those time intervals separated by x = 0). Alternatively, equation (1.100) can be solved numerically, as is done in the following. Because the system’s equation of motion is linear in two ranges, that is, equations (1.98) and (1.99) are linear, such systems are also called bilinear. The sliding block in Figure 1.44 requires nonzero initial conditions to set it in motion. Suppose first that the initial velocity is zero. The motion will result only if the initial position x0 is such that the spring force kx0 is large enough to overcome the static friction force μsmg (kx0 7 μsmg). Here μs is the coefficient of static friction, which is generally larger than the kinetic or dynamic coefficient of friction for sliding surfaces. If x0 is not large enough, no motion results. The range of values of x0 for which no motion results defines the equilibrium position. If, on the other hand, the initial velocity is nonzero, the object will move. One of the distinguishing features of nonlinear systems is their multiple equilibrium positions. The solution of the equation of motion for the case when motion results can be obtained by considering the following cases. With x0 to the right of any equilibrium, the mass is moving to the left, the friction force is to the right, and equation (1.98) holds. Equation (1.98) has a solution of the form x(t) = A1 cos ωnt + B1 sin ωnt + μ mg k (1.101) where ωn = 2k>m and A1 and B1 are constants to be determined by the initial conditions. Here we have dropped the distinction between static and kinetic friction. Applying the initial conditions yields x(0) = A1 + μ mg = x0 k # x(0) = ωnB1 = 0 (1.102) (1.103) Hence B1 = 0 and A1 = x0 - μmg>k specifies the constants in equation (1.101). Thus when the mass starts from rest (at x0) and moves to the left, it moves as x(t) = ax0 - μ mg μ mg b cos ωnt + k k www.TechnicalBooksPDF.com (1.104) 84 Introduction to Vibration and the Free Response Chap. 1 # This motion continues until the first time x(t) = 0. This happens when the derivative of equation (1.104) is zero, or when μ mg # x(t) = - ωn ax0 b sin ωnt 1 = 0 k (1.105) μ mg k (1.106) # Thus when t1 = π > ωn, x(t) = 0 and the mass starts to move to the right provided that the spring force, kx, is large enough to overcome the maximum frictional force μmg. Hence equation (1.99) now describes the motion. Solving equation (1.99) yields x(t) = A2 cos ωnt + B2 sin ωnt - # for π/ωn 6 t 6 t2, where t2 is the second time that x becomes zero. The initial conditions for equation (1.106) are calculated from the previous solution given by equation (1.104) at t1 xa μ mg 2μ mg μ mg π b cos π + = - x0 b = ax0 ωn k k k μmg # π x a b = -ωn ax0 b sin π = 0 ωn k (1.107) (1.108) From equation (1.106) and its derivatives it follows that A2 = x0 - 3μ mg k B2 = 0 (1.109) The solution for the second interval of time then becomes x(t) = ax0 - 3μ mg μ mg b cos ωnt k k 2π π 6 t 6 ωn ωn (1.110) This procedure is repeated until the motion stops. The motion will stop when the # velocity is zero (x = 0) and the spring force (kx) is insufficient to overcome the maximum frictional force (μ mg). The response is plotted in Figure 1.45. Several things can be noted about the free response with Coulomb friction versus the free response with viscous damping. First, with Coulomb damping the amplitude decays linearly with slope - 2μ mgωn πk (1.111) rather than exponentially as does a viscously damped system. Second, the motion under Coulomb friction comes to a complete stop, at a potentially different equilibrium position than when initially at rest, whereas a viscously damped system www.TechnicalBooksPDF.com Sec. 1.10 85 Coulomb Friction and the Pendulum x0 (x  4N k ) 0  sN k t 0 2 n 4 n ( 2N k ) Slope 2N  x0 k ( n ) Figure 1.45 A plot of the free response, x(t), of a spring–mass system with Coulomb friction. oscillates around a single equilibrium, x = 0, with infinitesimally small amplitude. Finally, the frequency of oscillation of a system with Coulomb damping is the same as the undamped frequency, whereas viscous damping alters the frequency of oscillation. Example 1.10.1 The response of a mass oscillating on a surface is measured to be of the form indicated in Figure 1.45. The initial position is measured to be 30 mm from its zero rest position, and the final position is measured to be 3.5 mm from its zero rest position after four cycles of oscillation in 1 s. Determine the coefficient of friction. Solution First, the frequency of motion is 4 Hz, or 25.13 rad>s, since four cycles were completed in 1 s. The slope of the line of decreasing peaks is - 30 + 3.5 = -26.5 mm>s 1 Therefore, from expression (1.111), - 26.5 mm>s = - 2μ g -2μ mgωn - 2μg ωn = = 2 π ωn πωn πk Solving for μ yields μ = π (25.13 rad>s) ( -26.5 mm>s) ( - 2) ( 9.81 * 103 mm>s2 ) = 0.107 This small value for μ indicates that the surface is probably very smooth or lubricated. n www.TechnicalBooksPDF.com 86 Introduction to Vibration and the Free Response Chap. 1 The response of the system of equation (1.100) can also be obtained by the numerical integration techniques of the previous section, which is substantially easier than the preceding construction of the solution. For example, VTB1_5 uses a fixed-step Runge–Kutta method to integrate equation (1.100). The second-order equation of motion can be reformulated into two first-order equations somewhat like equation (1.96) and integrated by the Euler method of equation (1.97), or standard Runge–Kutta methods may be employed. Figure 1.46 illustrates the response of a system subject to Coulomb friction for two different initial conditions using Mathcad’s fixed-time-step Runge–Kutta routine. Note in particular that the system comes to rest at a different value of xf depending on the initial conditions. Such a system has the same frequency, yet could come to rest anywhere in the region bounded by the two vertical lines (x = { μmg>k). The response will come to rest at the first time the velocity is zero and the displacement is within this region. Comparing the response of a linear spring–mass system with viscous damping (say the underdamped response of Figure 1.10) to the response of a spring–mass system with Coulomb damping given previously, an obvious and significant difference is the rest position. These multiple rest positions constitute a major feature of nonlinear systems: the existence of more than one equilibrium position. 6 Displacement (mm) 4 2 0 1 2 3 4 5 2 4 Time (s) Figure 1.46 A plot of the free response (displacement versus time) of a system subject to Coulomb friction with two different initial positions (the solid line is x0 = 5 mm and the dashed line is x0 = 4.5 mm, both with v0 = 0) for the same physical parameters (m = 1000 kg, μ = 0.3 and k = 5000 N>m). www.TechnicalBooksPDF.com 6 Sec. 1.10 87 Coulomb Friction and the Pendulum The equilibrium point of a system, or of a set of governing equations, may be defined best by first placing the equation of motion in state space, as was done in the previous section for the purpose of numerical integration. A general singledegree-of-freedom system may be written as $ # x(t) + f ( x(t), x(t) ) = 0 (1.112) where the function f can take on any form, linear or nonlinear. For example, for # # a linear spring–mass system the function f is just f (x, x) = cx (t) + kx (t), which is a linear function of the state variables of position and velocity. On the other hand, in a nonlinear system f will be some nonlinear function of the state variables. For instance, the pendulum equation derived and discussed in Example 1.1.1, $ θ + (g>l) $ sin θ = 0, can be written in the form of equation (1.112) by defining f to be f (θ, θ) = (g>l) sin (θ), where θ is the displacement variable. Using the approach following equations (1.94) and (1.95). the general statespace model of equation (1.112) is written by defining the two state variables x1 = x(t) # and x2 = x(t). Then equation (1.112) can be written as the first-order pair # x1(t) = x2(t) # x2(t) = -f(x1, x2) (1.113) This state-space form of the equation is used both for numerical integration (as before for the Coulomb friction problem) as well as for formally defining an equilibrium position by defining the state vector, x, used in equation (1.96) and a nonlinear vector function, F, as F = J x2(t) R -f(x1, x2) (1.114) Equations (1.113) may now be written in the simple form # x = F(x) (1.115) An equilibrium point of this system, denoted by xe, is defined to be any value of x for which F(x) is identically zero (called zero phase velocity). Thus the equilibrium point is any vector of constants that satisfies the relations F(xe) = 0 (1.116) A mechanical system is in equilibrium if its state does not change with time (i.e., # $ x and x are both zero). For Coulomb friction, the equilibrium position cannot be directly determined by using the signum function (see below equation 1.100) because of the discontinuity www.TechnicalBooksPDF.com 88 Introduction to Vibration and the Free Response Chap. 1 at zero velocity. To compute the equilibrium position, consider equation (1.116) for the system of Figure 1.44. This yields x2 0 k S = J R C fc x1 0 m m Solving yields the two conditions: x2 = 0 and fc - kx1 = 0 Realize that this last expression is static, so that the expression is satisfied as long as - μsmg μsmg 6 x1 6 k k As discussed earlier, the friction force is static, or in equilibrium, until the spring force kx1 is large enough to overcome the friction force as expressed by this inequality. This describes the condition that the velocity (x2) is zero and the position lies within the region defined by the force of friction. Depending on the initial conditions, the response will end up at a value of xe somewhere in this region. Usually, the equilibrium values are a discrete set of numbers, as the following example illustrates. Example 1.10.2 $ Calculate the equilibrium position for the nonlinear system defined by x + x - β2x3 = 0, or in state equation form, letting x1 = x as before, # x1 = x2 # x2 = x1 ( β2x21 - 1 ) Solution These equations represent the vibration of a “soft spring” and correspond to an approximation of the pendulum problem of Example 1.4.2, where sin x ≈ x - x3 >6. The equations for the equilibrium position are x2 = 0 x1 β2x21 ( - 1) = 0 There are three solutions to this set of algebraic equations corresponding to the three equilibrium positions of the soft spring. They are 1 0 -1 xe = J R , J β R , J β R 0 0 0 n Sec. 1.10 89 Coulomb Friction and the Pendulum The next example considers the full nonlinear pendulum equation illustrated in Figures 1.1 and 1.47. Physically the pendulum may swing all the way around its pivot point and has equilibrium positions in both the straight-up and straight-down positions, as illustrated in Figure 1.47(b) and 1.47(c). g .. g  sin   0 l Unstable equilibrium Stable equilibrium x1  . x2   x1  0, 2, 4, ... x2  0  (a) x1  , 3, 5, ... x2  0 (b) (c) Figure 1.47 (a) A pendulum consisting of a massless rod of length l and a tip mass m. (b) The straight-down equilibrium position. (c) The straight-up equilibrium position. Example 1.10.3 Calculate the equilibrium positions of the pendulum of Figure 1.47 with the equation of motion given in Example 1.1.1. Solution The pendulum equation in state-space form is given by # x1 = x2 g # x2 = - sin (x1) l so that the vector equation F(x) = 0 yields the following equilibrium solutions: x2 = 0 and x1 = 0, π, 2π, 3π, 4π, 5π p since sin(x1) is zero for any multiple of π. Note that there are an infinite number of equilibrium positions, or vectors xe. These are all either the up position corresponding to the odd values of π [Figure 1.47(c)], or the down position corresponding to even multiples of π [Figure 1.47(b)]. These positions form two distinct types of behavior. The response for initial conditions near the even values of π is a stable oscillation around the down position, just as in the linearized case, while the response to initial conditions near odd values of π moves away from the equilibrium position (called unstable) and the value of the response increases without bound. n The stability of equilibrium of a nonlinear vibration problem is very important and is based on the definitions given in Section 1.8. However, in the linear case, there is only one equilibrium value and every solution is either stable or unstable. In this case, the stability condition is said to be a global condition. In the nonlinear 90 Introduction to Vibration and the Free Response Chap. 1 case, there is more than one equilibrium point, and the concept of stability is tied to each particular equilibrium point and is therefore referred to as local stability. As in the example of the nonlinear pendulum equation, some equilibrium points are stable and some are not. Furthermore, the stability of the response of a nonlinear system depends on the initial conditions. In the linear case, the initial conditions have no influence on the stability, and the system parameters and form of the equation of motion completely determine the stability of the response. To see this, look again at the pendulum of Figure 1.47. If the initial position and velocity are near the origin, the system response will be stable and oscillate around the equilibrium point at zero. On the other hand, if the same pendulum (i.e., same l) is given initial conditions near the equilibrium point at θ = π rad, the response will grow without bound. Hence, θ = π rad is an unstable equilibrium. Even though nonlinear systems have multiple equilibria and more exotic behavior, their response may still be simulated using the numerical-integration methods of the previous section. This is illustrated for the pendulum in the following example, which compares the response to various initial conditions of both the nonlinear pendulum equation and its corresponding linearization treated in Examples 1.1.1 and 1.4.2. Example 1.10.4 Compare the responses of the nonlinear and linear pendulum equations using numerical integration and the value (g>/) = (0.1)2, for (a) the initial conditions x0 = 0.1 rad and v0 = 0.1 rad>s, and (b) the initial conditions x0 = 1 rad and v0 = 1 rad>s, by plotting the responses. Here x and v are used to denote θ and its derivative, respectively, in order to accommodate notation available in computer codes. Solution Depending on which program is used to integrate the solution numerically, the equations must first be put into first-order form, and then either Euler integration or Runge–Kutta routine may be implemented and the solutions plotted. Integrations in MATLAB, Mathematica, and Mathcad are presented. More details can be found in Appendix F. Note that the response to the linear system is fairly close to that of the full nonlinear system in case (a) with slightly different frequency, while case (b) with larger initial conditions is drastically different. The Mathcad solution follows. First, enter the initial conditions for each response: v0 := 1 x0 := 1 v10 := 0.1 x10 := 0.1 Next, define the frequency and the number of and size of the time steps: ω := 0.1 N := 2000 i := 0..N The nonlinear Euler integration is J xi + 1 vi # Δ + xi R := J R vi + 1 - ω2 # sin(xi) # Δ + vi Δ := 6#π ω#N Sec. 1.10 91 Coulomb Friction and the Pendulum The linear Euler integration is J x1i + 1 v1i # Δ + x1i R := J R 2# v1i + 1 - ω (Δ) # xi1 + v1i The plot of these two solutions yields 2 1 xi x1i 0 50 100 150 200 1 2 Here the dashed line is the linear solution. Next, compute these solutions again using initial conditions close to unstable equilibrium values: x0 := π J v0 := 0.1 x10 := π v10 := 0.1 xi + 1 vi # Δ + xi x1 v1i # Δ + x1i R := J R J i + 1 R := J R vi + 1 - ω2 # sin(xi) # Δ + vi v1i + 1 - ω2 # (Δ) # xi1 + v1i 40 20 xi x1i 0 50 100 150 200 20 The MATLAB code for running the solutions (using Runge–Kutta this time) and plotting is obtained by first creating the appropriate M-files (named lin_pend_dot.m 92 Introduction to Vibration and the Free Response Chap. 1 and NL_pend_dot.m, defining the linear and nonlinear pendulum equations, respectively). function xdot = lin_pend_dot(t,x) omega = 0.1; % define the natural frequency xdot(1,1) = x(2); xdot(2,1) = -omega^2*x(1); function xdot = NL_pend_dot(t,x) omega = 0.1; % define the natural frequency xdot(1,1) = x(2); xdot(2,1) = -omega^2*sin(x(1)); In the command mode type the following: % Overplot linear & nonlinear simulations of the free % response of a pendulum. x0 = 0.1; v0 = 0.1; ti = 0; tf = 200; % linear [time_lin,sol_lin]=ode45('lin_pend_dot',[ti tf],[x0 v0]); % nonlinear [time_NL,sol_NL]=ode45('NL_pend_dot',[ti tf],[x0 v0]); % overplot displacements figure plot(time_lin,sol_lin(:,1),'-') hold plot(time_NL,sol_NL(:,1),'—') xlabel('time (s)') ylabel('theta') title(['Linear vs. nonlinear pendulum with x0 = ' ... num2str(x0) ' and v0 = ' num2str(v0)]) legend('linear','nonlinear') Here the plots have been suppressed as they are similar to those from the Mathcad solution. Next, consider the Mathematica code to solve the same problem. First, we load the add-on package that will enable us to add a legend to our plot. In[1]:= <<PlotLegends' Sec. 1.10 93 Coulomb Friction and the Pendulum Define the natural circular frequency, ω In[2]:= 𝛚=0.1; The following cell solves the linear differential equation, then the nonlinear differential equation, and then produces a plot containing both responses. In[3] := xlin=NDSolve[{xl''[t]+𝛚2*xl[t]==0, xl[0]==.1, xl'[0]==.1}, xl[t],{t,0,200}] xnonlin=NDSolve[{xnl''[t]+𝛚2*Sin[xnl[t]]==0, xnl[0]==.1, xnl'[0]==.1}, xnl[t], {t,0,200}] Plot[{Evaluate[xl[t]/.xlin], Evaluate[xnl[t]/.xnonlin]},{t,0,200}, PlotStyle S {Dashing[{}], Dashing[{.01,.01}]}, PlotLabel S "Linear and Nonlinear Response, Stable Equilibrium", AxesLabel S {"time,s",""}, PlotLegend S {"Linear","Non-Linear"}, LegendPosition S {1,0}, LegendSize S {.7, .3}] Out[3]= {{xl[t] S InterpolatingFunction[{{0., 200.}},<>][t]}} Out[4]= {{xnl[t] S InterpolatingFunction[{{0., 200.}},<>][t]}} Out[5]= Linear and Nonlinear Response, Stable Equilibrium 1 Linear 0.5 Non-Linear time, s 50 100 150 200 0.5 1 In[6]:= Clear[xlin, xnonlin, xl, xnl] xlin=NDSolve[{xl''[t]+𝛚2*xl[t] == 0, xl[0] ==𝛑, xl'[0] ==.1}, xl[t], {t, 0, 200}] xnonlin=NDSolve[{xnl''[t]+𝛚2*Sin[xnl[t]] ==0, xnl[0] ==𝛑, xnl'[0] ==.1}, xnl[t], {t, 0, 200}] Plot[{Evaluate[xl[t]/.xlin], Evaluate[xnl[t]/.xnonlin]}, {t, 0, 200}, PlotStyle S {Dashing[{}], Dashing[{.01,.01}]}, PlotRange S {-20, 40}, PlotLabel S "Linear and Nonlinear Response, Unstable Equilibrium", AxesLabel S {"time,s", ""}, PlotLegend S {"Linear", "Nonlinear"}, LegendPosition S {1, 0}, LegendSize S {.7, .3}] Out[7]= {{xl[t] S InterpolatingFunction[{{0., 200.}},<>][t]}} Out[8]= {{xnl[t] S InterpolatingFunction[{{0., 200.}},<>][t]}} Out[9]= 94 Introduction to Vibration and the Free Response 40 Chap. 1 Linear and Nonlinear Response, Unstable Equilibrium 30 Linear 20 Nonlinear 10 50 100 150 200 time, s 10 20 n Note from the plots in Example 1.10.4 that even in the case where the initial conditions are small, the linear response is not exactly the same as the full nonlinear system response. However, if the initial velocity is changed to zero, the solutions are very similar. This is illustrated in Figure 1.48. In summary, nonlinear systems have several interesting aspects that linear systems do not. In particular, nonlinear systems have multiple equilibrium positions rather than just one, as in the linear case. Some of these extra equilibrium points may be unstable and some may be stable. The stability of a response depends on the initial conditions, which can send the solution to different equilibrium positions and hence different types of response. Thus the behavior of the Displacement (m) 1.5 1 0.5 0 50 100 150 200 0.5 1 1.5 i Time (s) Figure 1.48 Plots of the response of both the linear (dashed line) and nonlinear (solid line) systems of Example 1.10.4 with the initial velocity set to zero in each case. Chap. 1 95 Problems response depends on the initial conditions, not just the parameters and form of the equation, as is the case for the linear system. This is illustrated in Example 1.10.4. Even though the response of a nonlinear system is much more complicated and closed-form analytical solutions are not always available, the response can be simulated using numerical integration. In modeling real systems, some degree of nonlinearity is always present. Whether or not it is important to include the nonlinear part of the model in computing, the response depends on the initial conditions. If the initial conditions are such that the system’s nonlinearity comes into play, then these terms should be included. Otherwise a linear response is perfectly acceptable. The same can be said for whether or not to include damping in a system model. Which effects to include and which not to include when modeling and analyzing a vibrating system form one of the important aspects of engineering practice. This section has introduced a little bit about the vibrations of systems with nonlinearities. The important points are that nonlinear systems potentially have multiple equilibrium positions, each with potentially different stability behavior. Nonlinear systems typically do not have closed-form solutions so that the time history is often computed by numerical integration. Not addressed here, but nontheless very important, is that the principle of superposition, used extensively in Chapters 3 and 4, does not apply to nonlinear systems. The majority of this text focuses on linear vibration problems. With this brief introduction to nonlinear systems, it is important to emphasize that when solving linear problems, initial conditions must be limited such that only the linear range is excited. Students are encouraged to take a course in nonlinear systems and/or nonlinear vibrations to learn more about the analysis and behavior of nonlinear vibrations. PROBLEMS Those problems marked with an asterisk are intended to be solved using computational software. Section 1.1 (Problems 1.1 through 1.26) 1.1. Consider a simple pendulum (see Example 1.1.1) and compute the magnitude of the restoring force if the mass of the pendulum is 2 kg and the length of the pendulum is 0.5 m. Assume the pendulum is at the surface of the earth at sea level. 1.2. Compute the period of oscillation of a pendulum of length 1 m at the North Pole where the acceleration due to gravity is measured to be 9.832 m>s2. 96 Introduction to Vibration and the Free Response Chap. 1 1.3. The spring of Figure 1.2, repeated here as Figure P1.3, is loaded with mass of 15 kg and the corresponding (static) displacement is 0.01 m. Calculate the spring’s stiffness. k g m x(t) Figure P1.3 1.4. The spring of Figure P1.3 is successively loaded with mass and the corresponding (static) displacement is recorded below. Plot the data and calculate the spring’s stiffness. Note that the data contain some errors. Also calculate the standard deviation. m(kg) 10 11 12 13 14 15 16 x(m) 1.14 1.25 1.37 1.48 1.59 1.71 1.82 1.5. Consider the pendulum of Example 1.1.1 reproduced in Figure P1.5 and compute the amplitude of the restoring force if the mass of the pendulum is 2 kg and the length of the pendulum is 0.5 m if the pendulum is at the surface of the moon. Fy Fx O O g l l  m m mg (a) (b) Figure P1.5 (a) The pendulum of Example 1.1.1 and (b) its freebody diagram. 1.6. Consider the pendulum of Example 1.1.1 and compute the angular natural frequency (radians per second) of vibration for the linearized system if the mass of the pendulum is 2 kg and the length of the pendulum is 0.5 m if the pendulum is at the surface of the earth. What is the period of oscillation in seconds? $ 1.7. Derive the solution of mx + kx = 0 and plot the result for at least two periods for the case with ωn = 2 rad>s, x0 = 1 mm, and v0 = 15 mm>s. Chap. 1 97 Problems $ 1.8. Solve mx + kx = 0 for k = 4 N>m, m = 1 kg, x0 = 1 mm, and v0 = 0. Plot the solution. 1.9. The amplitude of vibration of an undamped system is measured to be 1 mm. The phase shift from t = 0 is measured to be 2 rad and the frequency is found to be 5 rad>s. Calculate the initial conditions that caused this vibration to occur. Assume the response is of the form x(t) = A sin (ωnt + ϕ). 1.10. Determine the stiffness of a single-degree-of-freedom spring–mass system with a mass of 100 kg such that the natural frequency is 10 Hz. 1.11. Find the equation of motion for the system of Figure P1.11, and find the natural frequency. In particular, using static equilibrium along with Newton’s law, determine what effect gravity has on the equation of motion and the system’s natural frequency. Assume the block slides without friction. k x(t) m g Frictionless surface  Figure P1.11 1.12. An undamped system vibrates with a frequency of 10 Hz and amplitude 1 mm. Calculate the maximum amplitude of the system’s velocity and acceleration. 1.13. Show by calculation that A sin(ωnt + ϕ) can be represented as A1 sin ωnt + A2 cos ωnt and calculate A1 and A2 in terms of A and ϕ. 1.14. Using the solution of equation (1.2) in the form x(t) = A1 sin ωnt + A2 cos ωnt, calculate the values of A1 and A2 in terms of the initial conditions x0 and v0. 1.15. Using the drawing in Figure 1.7, verify that equation (1.10) satisfies the initial velocity condition. 1.16. A 0.5 kg mass is attached to a linear spring of stiffness 0.1 N>m. (a) Determine the natural frequency of the system in hertz. (b) Repeat this calculation for a mass of 50 kg and a stiffness of 10 N>m. Compare your result to that of part (a). 1.17. Derive the solution of the single-degree-of-freedom system of Figure 1.4 by writing Newton’s law, ma = - kx, in differential form using a dx = dv and integrating twice. 1.18. Determine the natural frequency of the two systems illustrated in Figure P1.18. k1 k2 k1 m k3 m k2 (a) (b) Figure P1.18 98 Introduction to Vibration and the Free Response Chap. 1 *1.19. Plot the solution given by equation (1.10) for the case k = 1000 N>m and m = 10 kg for two complete periods for each of the following sets of initial conditions: a) x0 = 0 m, v0 = 1 m>s, b) x0 = 0.01 m, v0 = 0 m>s, and c) x0 = 0.01 m, v0 = 1 m>s. *1.20. Make a three dimensional surface plot of the amplitude A of an undamped oscillator given by equation (1.9) versus x0 and v0 for the range of initial conditions given by –0.1 … x0 … 0.1 m and –1 … v0 … 1 m >s for a system with natural frequency of 10 rad >s. 1.21. A machine part is modeled as a pendulum connected to a spring as illustrated in Figure P1.21. Ignore the mass of the pendulum’s rod and derive the equation of motion. Then following the procedure used in Example 1.1.1, linearize the equation of motion and compute the formula for the natural frequency. Assume that the rotation is small enough so that the spring only deflects horizontally. O  g k m Figure P1.21 1.22. A pendulum has length of 250 mm. What is the system’s natural frequency in hertz? 1.23. The pendulum in Example 1.1.1 (see Figure P1.5) is required to oscillate once every second. What length should it be? 1.24. The approximation of sin θ = θ, is reasonable for θ less than 10°. If a pendulum of length 0.5 m, has an initial position of θ(0) = 0, what is the maximum value of the initial angular velocity that can be given to the pendulum without violating this smallangle approximation? (Be sure to work in radians.) 1.25. A machine, modeled as a simple spring–mass system, oscillates in simple harmonic motion. Its acceleration is measured to have an amplitude of 10,000 mm>s2 with a frequency of 8 Hz. Compute the maximum displacement the machine undergoes during this oscillation. 1.26. Derive the relationships given in Window 1.4 for the constants a1 and a2, used in the exponential form of the solution, in terms of the constants A1 and A2, used in sum of sine and cosine form of the solution. Use the Euler relationships for sine and cosine in terms of exponentials as given following equation (1.18). Chap. 1 99 Problems Section 1.2 (Problems 1.27 through 1.40) 1.27. The acceleration of a machine part modeled as a spring–mass system is measured and recorded in Figure P1.27. Compute the amplitude of the displacement of the mass. Acceleration in m/s 2 1.2 0.8 0.4 0 0.4 0.8 1.2 0 0 0.4 0.8 1.2 1.6 2 2.4 2.8 3.2 3.6 4 4.4 4.8 Time, s Figure P1.27 1.28. A vibrating spring–mass system has a measured acceleration amplitude of 8 mm>s2 and measured displacement amplitude of 2 mm. Calculate the system’s natural frequency. 1.29. A spring–mass system has a measured period of 5 s and a known mass of 20 kg. Calculate the spring stiffness. *1.30. Plot the solution of a linear spring–mass system with frequency ωn = 2 rad>s, x0 = 1 mm, and v0 = 2.34 mm>s for at least two periods. *1.31. Compute the natural frequency and plot the solution of a spring–mass system with mass of 1 kg, stiffness of 4 N>m, and initial conditions of x0 = 1 mm and v0 = 0 mm>s for at least two periods. 1.32. When designing a linear spring–mass system it is often a matter of choosing a spring constant such that the resulting natural frequency has a specified value. Suppose that the mass of a system is 4 kg and the stiffness is 100 N>m. How much must the spring stiffness be changed in order to increase the natural frequency by 10%? 1.33. The pendulum in the Chicago Museum of Science and Industry has a length of 20 m, and the acceleration due to gravity at that location is known to be 9.803 m>s2. Calculate the period of this pendulum. 100 Introduction to Vibration and the Free Response Chap. 1 1.34. Calculate the RMS values of displacement, velocity, and acceleration for the undamped single-degree-of-freedom system of equation (1.19) with zero phase. 1.35. A foot pedal mechanism for a machine is crudely modeled as a pendulum connected to a spring as illustrated in Figure P1.35. The purpose of the spring is provide a return force for the pedal action. Compute the spring stiffness needed to keep the pendulum at 1° from the horizontal and then compute the corresponding natural frequency. Assume that the angular deflections are small, such that the spring deflection can be approximated by the arc length; that the pedal may be treated as a point mass; and that pendulum rod has negligible mass. The pedal is horizontal when the spring is at its free length. The values in the figure are m = 0.5 kg, g = 9.8 m>s2, l1 = 0.2 m, and l2 = 0.3 m. l2 l1 g m  k Figure P1.35 1.36. An automobile is modeled as a 1000-kg mass supported by a spring of stiffness k = 400,000 N>m. When it oscillates it does so with a maximum deflection of 10 cm. When loaded with passengers, the mass increases to as much as 1300 kg. Calculate the change in frequency, velocity amplitude, and acceleration amplitude if the maximum deflection remains 10 cm. 1.37. The front suspension of some cars contains a torsion rod as illustrated in Figure P1.37 to improve the car’s handling. (a) Compute the frequency of vibration of the wheel assembly x Wheel g  l Frame Figure P1.37 Chap. 1 Problems 101 given that the torsional stiffness is 2000 N m> rad and the wheel assembly has a mass of 38 kg. Take the distance x = 0.26 m. (b) Sometimes owners put different wheels and tires on a car to enhance the appearance or performance. Suppose a thinner tire is put on with a larger wheel raising the mass to 45 kg. What effect does this have on the frequency? 1.38. A machine oscillates in simple harmonic motion and appears to be well modeled by an undamped single-degree-of-freedom oscillation. Its acceleration is measured to have an amplitude of 10,000 mm > s2 at 8 Hz. What is the machine’s maximum displacement? 1.39. A simple undamped spring–mass system is set into motion from rest by giving it an initial velocity of 100 mm>s. It oscillates with a maximum amplitude of 10 mm. What is its natural frequency? 1.40. An automobile exhibits a vertical oscillating displacement of maximum amplitude 5 cm and a measured maximum acceleration of 2000 cm>s2. Assuming that the automobile can be modeled as a single-degree-of-freedom system in the vertical direction, calculate the natural frequency of the automobile. Section 1.3 (Problems 1.41 through 1.64) 1.41. Consider a spring–mass–damper system, like the one in Figure 1.9, with the following values: m = 10 kg, c = 3 N>s, and k = 1000 N>m. (a) Is the system overdamped, underdamped, or critically damped? (b) Compute the solution if the system is given initial conditions x0 = 0.01 m and 0 = 0. 1.42. Consider a spring–mass–damper system with equation of motion given by $ # x + 2x + 2x = 0. Compute the damping ratio and determine if the system is overdamped, underdamped, or critically damped. $ # 1.43. Consider the system x + 4x + x = 0 for x0 = 1 mm, v0 = 0 mm > s. Is this system overdamped, underdamped, or critically damped? Compute the solution and determine which root dominates as time goes on (that is, one root will die out quickly and the other will persist). $ # 1.44. Compute the solution to x + 2x + 2x = 0 for x0 = 0 mm, v0 = 1 mm>s and write down the closed-form expression for the response. 1.45. Derive the form of λ1 and λ2 given by equation (1.31) from equation (1.28) and the definition of the damping ratio. 1.46. Use the Euler formulas to derive equation (1.36) from equation (1.35) and to determine the relationships listed in Window 1.4. 1.47. Using equation (1.35) as the form of the solution of the underdamped system, calculate the values for the constants a1 and a2 in terms of the initial conditions x0 and v0. 1.48. Calculate the constants A and ϕ in terms of the initial conditions and thus verify equation (1.38) for the underdamped case. 1.49. Calculate the constants a1 and a2 in terms of the initial conditions and thus verify equations (1.42) and (1.43) for the overdamped case. 1.50. Calculate the constants a1 and a2 in terms of the initial conditions and thus verify equation (1.46) for the critically damped case. 1.51. Using the definition of the damping ratio and the undamped natural frequency, derive equation (1.48) from (1.47). 102 Introduction to Vibration and the Free Response Chap. 1 1.52. For a damped system, m, c, and k are known to be m = 1 kg, c = 2 kg>s, k = 10 N>m. Calculate the value of ζ and ωn. Is the system overdamped, underdamped, or critically damped? 1.53. Plot x(t) for a damped system of natural frequency ωn = 2 rad>s and initial conditions x0 = 1 mm, v0 = 1 mm, for the following values of the damping ratio: ζ = 0.01, ζ = 0.2, ζ = 0.1, ζ = 0.4, and ζ = 0.8. 1.54. Plot the response x(t) of an underdamped system with ωn = 2 rad>s, ζ = 0.1, and v0 = 0 for the following initial displacements: x0 = 10 mm and x0 = 100 mm. $ # 1.55. Calculate the solution to x + x + x = 0 with x0 = 1 and v0 = 0 for x(t) and sketch the response. 1.56. A spring–mass–damper system has mass of 100 kg, stiffness of 3000 N>m, and damping coefficient of 300 kg>s. Calculate the undamped natural frequency, the damping ratio, and the damped natural frequency. Does the solution oscillate? 1.57. A rough sketch of a valve-and-rocker-arm system for an internal combustion engine is give in Figure P1.57. Model the system as a pendulum attached to a spring and a mass and assume the oil provides viscous damping in the range of ζ = 0.01. Determine the equations of motion and calculate an expression for the natural frequency and the damped natural frequency. Here J is the rotational inertia of the rocker arm about its pivot point, k is the stiffness of the valve spring, and m is the mass of the valve and stem. Ignore the mass of the spring.  Rocker arm J Follower k Valve spring Oil m Valve Cam x l Figure P1.57 1.58. A spring–mass–damper system has mass of 150 kg, stiffness of 1500 N>m, and damping coefficient of 200 kg>s. Calculate the undamped natural frequency, the damping ratio, and the damped natural frequency. Is the system overdamped, underdamped, or critically damped? Does the solution oscillate? *1.59. The spring–mass system of 100 kg mass, stiffness of 3000 N>m, and damping coefficient of 300 Ns>m is given a zero initial velocity and an initial displacement of 0.1 m. Calculate the form of the response and plot it for as long as it takes to die out. Chap. 1 103 Problems *1.60. The spring–mass system of 150 kg mass, stiffness of 1500 N>m, and damping coefficient of 200 Ns>m is given an initial velocity of 10 mm>s and an initial displacement of –5 mm. Calculate the form of the response and plot it for as long as it takes to die out. How long does it take to die out? *1.61. Choose the damping coefficient of a spring–mass–damper system with mass of 150 kg and stiffness of 2000 N>m such that its response will die out after about 2 s, given a zero initial position and an initial velocity of 10 mm>s. 1.62. Derive the equation of motion of the system in Figure P1.62 and discuss the effect of gravity on the natural frequency and the damping ratio. k c g m Figure P1.62 1.63. Derive the equation of motion of the system in Figure P1.63 and discuss the effect of gravity on the natural frequency and the damping ratio. You may have to make some approximations of the cosine. Assume the bearings provide a viscous-damping force only in the vertical direction. (From A. Diaz-Jimenez, South African Mechanical Engineer, Vol. 26, pp. 65–69, 1976) (1976) 0 A g  k h m Figure P1.63 1.64. Consider the response of an underdamped system given by x(t) = e -ζωntA sin (ωdt + ϕ) where A and ϕ are given in terms of the initial conditions x0 = 0, and v0 ≠ 0. Determine the maximum value that the acceleration will experience in terms of v0. 104 Introduction to Vibration and the Free Response Chap. 1 Section 1.4 (Problems 1.65 through 1.81) 1.65. Calculate the frequency of the compound pendulum of Figure P1.65 if a mass mT is added to the tip, by using the energy method. Assume the mass of the pendulum is evenly distributed so that its center of gravity is in the middle of the pendulum of length l. l g  Figure P1.65 A compound pendulum with a tip mass. mT 1.66. Calculate the total energy in a damped system with frequency 2 rad>s and damping ratio ζ = 0.01 with mass 10 kg for the case x0 = 0.1 m and v0 = 0. Plot the total energy versus time. 1.67. Use the energy method to calculate the equation of motion and natural frequency of an airplane’s steering mechanism for the nose wheel of its landing gear. The mechanism is modeled as the single-degree-of-freedom system illustrated in Figure P1.67. (Steering wheel)  k1 J k2 r x (Tire–wheel assembly) m Figure P1.67 The steering wheel and tire assembly are modeled as being fixed at ground for this calculation. The steering rod gear system is modeled as a linear spring–mass system (m, k2) oscillating in the x direction. The shaft-gear mechanism is modeled as the disk of inertia J and torsional stiffness k2. The gear J turns through the angle θ such that the disk does not slip on the mass. Obtain an equation in the linear motion x. 1.68. Consider the pendulum-and-spring system of Figure P1.68. Here the mass of the pendulum rod is negligible. Derive the equation of motion using the energy method. Then linearize the system for small angles and determine the natural frequency. The length of the pendulum is l, the tip mass is m, and the spring stiffness is k. m g k  l Figure P1.68 A simple pendulum connected to a spring. Chap. 1 105 Problems 1.69. A control pedal of an aircraft can be modeled as the single-degree-of-freedom system of Figure P1.69. Consider the lever as a massless shaft and the pedal as a lumped mass at the end of the shaft. Use the energy method to determine the equation of motion in θ and calculate the natural frequency of the system. Assume the spring to be unstretched at θ = 0 and gravity points down. k l1  l2 m Figure P1.69 1.70. To save space, two large pipes are shipped, one stacked inside the other, as indicated in Figure P1.70. Calculate the natural frequency of vibration of the smaller pipe (of radius R1) rolling back and forth inside the larger pipe (of radius R). Use the energy method and assume that the inside pipe rolls without slipping and has a mass of m. Large pipe Small pipe  a' R TRUCKER O' b O R1 a mg Truck bed (a) (b) Figure P1.70 (a) Pipes stacked in a truck bed. (b) Vibration model of the inside pipe. 1.71. Consider the example of a simple pendulum given in Example 1.4.2. The pendulum motion is observed to decay with a damping ratio of ζ = 0.001. Determine a damping coefficient and add a viscous-damping term to the pendulum equation. 106 Introduction to Vibration and the Free Response Chap. 1 1.72. Determine a damping coefficient for the disk-rod system of Example 1.4.3. Assuming that the damping is due to the material properties of the rod, determine c for the rod if it is observed to have a damping ratio of ζ = 0.01. 1.73. The rod and disk of Window 1.1 are in torsional vibration. Calculate the damped natural frequency if J = 1000 m2 # kg, c = 20 N # m # s>rad, and k = 400 N # m>rad. 1.74. Consider the system of Figure P1.74, which represents a simple model of an aircraft landing system. Assume, x = rθ. What is the damped natural frequency? k c (t) r m, J x(t) Figure P1.74 1.75. Consider Problem 1.74 with k = 400,000 N>m, m = 1500 kg, J = 100 m2 # kg>rad, r = 25 cm, and c = 8000 kg>s. Calculate the damping ratio and the damped natural frequency. How much effect does the rotational inertia have on the undamped natural frequency? 1.76. Use Lagrange’s formulation to calculate the equation of motion and the natural frequency of the system of Figure P1.76. Model each of the brackets as a spring of stiffness k, and assume the inertia of the pulleys is negligible. k k g m Figure P1.76 1.77. Use Lagrange’s formulation to calculate the equation of motion and the natural frequency of the system of Figure P1.77. This figure represents a simplified model of a jet engine mounted to a wing through a mechanism which acts as a spring of stiffness Chap. 1 107 Problems k and mass ms. Assume the engine has inertia J and mass m and that the rotation of the engine is related to the vertical displacement of the engine, x(t) by the “radius” r0 (i.e., x = r0θ). Wing, ground Mount, k, ms l  Engine, J, m x(t) Figure P1.77 1.78. Consider the inverted simple pendulum connected to a spring of Figure P1.68. Use Lagrange’s formulation to derive the equation of motion. 1.79. Lagrange’s formulation can also be used for non-conservative systems by adding the applied non-conservative term to the right side of equation (1.63) to get 0Ri 0T 0U d 0T a # b + + # = 0 dt 0qi 0qi 0qi 0qi Here Ri is the Rayleigh dissipation function defined in the case of a viscous damper attached to ground by Ri = 1 #2 cq 2 i Use this extended Lagrange formulation to derive the equation of motion of the damped automobile suspension driven by a dynamometer illustrated in Figure P1.79. Assume here that the dynamometer drives the system such that x = rθ. k g c (t) r m, J Figure P1.79 x(t) 108 Introduction to Vibration and the Free Response Chap. 1 1.80. Consider the disk of Figure P1.80 connected to two springs. Use the energy method to calculate the system’s natural frequency of oscillation for small angles θ(t). (t) k k s a x(t) r m  mass Figure P1.80 1.81. A pendulum of negligible mass is connected to a spring of stiffness k at halfway along its length, l, as illustrated in Figure P1.81. The pendulum has two masses fixed to it: one at the connection point with the spring and one at the top. Derive the equation of motion using the Lagrange formulation, linearize the equation, and compute the system’s natural frequency. Assume that the angle remains small enough so that the spring only stretches significantly in the horizontal direction. m2 k g m1  Figure P1.81 Section 1.5 (Problems 1.82 through 1.93) 1.82. A bar of negligible mass fixed with a tip mass forms part of a machine used to punch holes in a sheet of metal as it passes the fixture as illustrated in Figure P1.82. The impact to the mass and bar fixture causes the bar to vibrate and the speed of the process demands that frequency of vibration not interfere with the process. The static design yields a mass of 50 kg and that the bar be made of steel of length 0.25 m with a cross sectional area of 0.01 m2. Compute the system’s natural frequency. v0 EA m l Figure P1.82 A bar model of a punch fixture. Chap. 1 109 Problems 1.83. Consider the punch fixture of Figure P1.82. If the system is giving an initial velocity of 10 m/s, what is the maximum displacement of the mass at the tip if the mass is 1000 kg and the bar is made of steel of length 0.25 m with a cross sectional area of 0.01 m2? 1.84. Consider the punch fixture of Figure P1.82. If the punch strikes the mass off center, it is possible that the steel bar may vibrate in torsion. The mass is 1000 kg and the bar 0.25-m long, with a square cross section of 0.1 m on a side. The mass polar moment of inertia of the tip mass is 10 kg>m2. The polar moment of inertia for a square bar is b4 >6, where b is the length of the side of the square. Compute both the torsion and longitudinal frequencies. Which is larger? 1.85. A helicopter landing gear consists of a metal framework rather than the coil spring based suspension system used in a fixed-wing aircraft. The vibration of the frame in the vertical direction can be modeled by a spring made of a slender bar as illustrated in Figure 1.23 where the helicopter is modeled as ground. Here l = 0.4 m, E = 20 * 1010 N>m2, and m = 100 kg. Calculate the cross-sectional area that should be used if the natural frequency is to be fn = 500 Hz. 1.86. The frequency of oscillation of a person on a diving board can be modeled as the transverse vibration of a beam as indicated in Figure 1.26. Let m be the mass of the diver (m = 100 kg) and l = 1.5 m. If the diver wishes to oscillate at 3 Hz, what value of EI should the diving board material have? 1.87. Consider the spring system of Figure 1.32. Let k1 = k5 = k2 = 100 N>m, k3 = 50 N>m, and k4 = 1 N>m. What is the equivalent stiffness? 1.88. Springs are available in stiffness values of 10, 100, and 1000 N>m. Design a spring system using these values only, so that a 100-kg mass is connected to ground with frequency of about 1.5 rad>s. 1.89. Calculate the natural frequency of the system in Figure 1.32(a) if k1 = k2 = 0. Choose m and nonzero values of k3, k4, and k5 so that the natural frequency is 100 Hz. *1.90. Example 1.4.4 examines the effect of the mass of a spring on the natural frequency of a simple spring–mass system. Use the relationship derived there and plot the natural frequency (normalized by the natural frequency, ωn, for a massless spring) versus the percent that the spring mass is of the oscillating mass. Determine from the plot (or by algebra) the percentage where the natural frequency changes by 1% and therefore the spring’s mass should not be neglected. 1.91. Calculate the natural frequency and damping ratio for the system in Figure P1.91 given the values m = 10 kg, c = 100 kg>s, k1 = 4000 N>m, k2 = 200 N>m, and k3 = 1000 N>m. Assume that no friction acts on the rollers. Is the system overdamped, critically damped, or underdamped? c k1 m k2 k3 Figure P1.91 110 Introduction to Vibration and the Free Response Chap. 1 1.92. Calculate the natural frequency and damping ratio for the system in Figure P1.92. Assume that no friction acts on the rollers. Is the system overdamped, critically damped, or underdamped? 2 kN/m 3 kN/m 4 kN/m 10 kN/m 10 kg 1 kg/s 1 kN/m Figure P1.92 1.93. A manufacturer makes a cantilevered leaf spring from steel (E = 2 * 1011 N>m2) and sizes the spring so that the device has a specific frequency. Later, to save weight, the spring is made of aluminum (E = 7.1 * 1010 N>m2). Assuming that the mass of the spring is much smaller than that of the device the spring is attached to, determine if the frequency increases or decreases and by how much. Section 1.6 (Problems 1.94 through 1.101) 1.94. The displacement of a vibrating spring–mass–damper system is recorded on an x - y plotter and reproduced in Figure P1.94. The y coordinate is the displacement in cm and the x coordinate is time in seconds. From the plot determine the natural frequency, the damping ratio, and the damped natural frequency. 0.75 0.5 y 0.25 0 1 2 3 4 5 6 x 7 8 0.25 0.5 Figure P1.94 A plot of displacement versus time for a vibrating system. 9 10 Chap. 1 111 Problems 1.95. Show that the logarithmic decrement is equal to δ = 1 x0 ln n xn where xn is the amplitude of vibration after n cycles have elapsed. 1.96. Derive the equation (1.78) for the trifalar suspension system. 1.97. A prototype composite material is formed and hence has an unknown modulus. An experiment is performed consisting of forming it into a cantilevered beam of length 1 m and I = 10-9 m4 with a 6-kg mass attached at its end. The system is given an initial displacement and found to oscillate with a period of 0.5 s. Calculate the modulus E. 1.98. The free response of a 1000-kg car with stiffness of k = 400,000 N>m is observed to be of the form given in Figure 1.35. Modeling the car as a single-degree-of-freedom oscillation in the vertical direction, determine the damping coefficient if the displacement at t1 is measured to be 2 cm and 0.22 cm at t2. 1.99. A pendulum decays from 10 cm to 1 cm over one period. Determine its damping ratio. 1.100. The relationship between the log decrement δ and the damping ratio ζ is often approximated as δ = 2πζ. For what values of ζ would you consider this a good approximation to equation (1.82)? 1.101. A damped system is modeled as illustrated in Figure 1.9. The mass of the system is measured to be 5 kg and its spring constant is measured to be 5000 N>m. It is observed that during free vibration the amplitude decays to 0.25 of its initial value after five cycles. Calculate the viscous-damping coefficient, c. Section 1.7 (Problems 1.102 through 1.110, also see Problem Section 1.5) 1.102. Consider the system of Example 1.7.2 consisting of a helical spring of stiffness 103 N>m attached to a 10-kg mass. Place a dashpot parallel to the spring and choose its viscousdamping value so that the resulting damped natural frequency is reduced to 9 rad>s. 1.103. For an underdamped system, x0 = 0 mm and v0 = 10 mm>s. Determine m, c, and k such that the amplitude is less than 1 mm. 1.104. Repeat Problem 1.103 if the mass is restricted to be between 10 kg 6 m 6 15 kg. 1.105. Use the formula for the torsional stiffness of a shaft from Table 1.1 to design a 1-m shaft with torsional stiffness of 105 N # m>rad. 1.106. Consider designing a helical spring made of aluminum, such that when it is attached to a 10-kg mass the resulting spring–mass system has a natural frequency of 10 rad/s. Thus repeat Example 1.7.2 which uses steel for the spring and note any difference. 1.107. Try to design a bar that has the same stiffness as the helical spring of Example 1.7.2 (i.e., k = 103 N>m). This amounts to computing the length of the bar with its cross sectional area taking up about the same space at the helical spring (R = 10 cm). Note that the bar must remain at least 10 times as long as it is wide in order to be modeled by the stiffness formula given for the bar in Figure 1.23. 1.108. Repeat Problem 1.107 using plastic (E = 1.40 * 109 N>m2) and rubber (E = 7 * 106 N>m2). Are any of these feasible? 112 Introduction to Vibration and the Free Response Chap. 1 1.109. Consider the diving board of Figure P1.109. For divers, a certain level of static deflection is desirable, denoted by Δ. Compute a design formula for the dimensions of the board (b, h, and l) in terms of the static deflection, the average diver’s mass, m, and the modulus of the board. mg  l 3 I bh 12 b h end view Figure P1.109 1.110. In designing a vehicle suspension system using a “quarter car model” consisting of a spring, mass, and damper system, studies show the desirable damping ratio is ζ = 0.25. If the model has a mass of 750 kg and a frequency of 15 Hz, what should the damping coefficient be? Section 1.8 (Problems 1.111 through 1.115) 1.111. Consider the system of Figure P1.111. (a) Write the equations of motion in terms of the angle, w, the bar makes with the vertical. Assume linear deflections of the springs and linearize the equations of motion. (b) Discuss the stability of the linear system’s solutions in terms of the physical constants, m, k, and l. Assume the mass of the rod acts at the center as indicated in the figure. k k c.g. l 2 l mg O Figure P1.111 1.112. Consider the inverted pendulum of Figure 1.40 as discussed in Example 1.8.1 and repeated in Figure P1.112. Assume that a dashpot (of damping rate c) also acts on the pendulum parallel to the two springs. How does this affect the stability properties of the pendulum? Chap. 1 113 Problems + m m k k 2 kl 2 sin  1 2 mg  O fp O Figure P1.112 The inverted pendulum of Example 1.8.1. 1.113. Replace the massless rod of the inverted pendulum of Figure P1.112 with a solid object compound pendulum of Figure 1.20(b). Calculate the equations of vibration and discuss values of the parameter relations for which the system is stable. 1.114. A simple model of a control tab for an airplane is sketched in Figure P1.114. The equation of motion for the tab about the hinge point is written in terms of the angle θ from the centerline to be $ # Jθ + (c - fd)θ + kθ = 0 Here J is the moment of inertia of the tab, k# is the rotational stiffness of the hinge, c is the rotational damping in the hinge, and fdθ is the negative damping provided by the aerodynamic forces (indicated by arrows in the figure). Discuss the stability of the solution in terms of the parameters c and fd. Figure P1.114 A simple model of an airplane control tab. *1.115. In order to understand the effect of damping in design, develop some sense of how the response changes with the damping ratio by plotting the response of a single-degreeof-freedom system for a fixed amplitude, frequency, and phase as ζ changes through the following set of values ζ = 0.01, 0.05, 0.1, 0.2, 0.3, and 0.4. That is, plot the response x(t) = e -10ζt sin (1021 - ζ2t) for each value of ζ. 114 Introduction to Vibration and the Free Response Chap. 1 Section 1.9 (Problems 1.116 through 1.123) # *1.116. Compute and plot the response to x = -3x, x(0) = 1 using Euler’s method for time steps of 0.1 and 0.5. Also plot the exact solution and hence reproduce Figure 1.41. *1.117. Use numerical integration to solve the system of Example 1.7.3 with m = 1361 kg, k = 2.688 * 105 N>m, c = 3.81 * 103 kg>s subject to the initial conditions x(0) = 0 and v(0) = 0.01 mm>s. Compare your result using numerical integration to just plotting the analytical solution (using the appropriate formula from Section 1.3) by plotting both on the same graph. *1.118. Consider again the damped system of Problem 1.117 and design a damper such that the oscillation dies out after 2 seconds. There are at least two ways to do this. Here it is intended to solve for the response numerically, following Examples 1.9.2, 1.9.3, or 1.9.4, using different values of the damping parameter c until the desired response is achieved. *1.119. Consider again the damped system of Example 1.9.2 and design a damper such that the oscillation dies out after 25 seconds. There are at least two ways to do this. Here it is intended to solve for the response numerically, following Examples 1.9.2, 1.9.3, or 1.9.4, using different values of the damping parameter c until the desired response is achieved. Is your result overdamped, underdamped, or critically damped? *1.120. Repeat Problem 1.119 for the initial conditions x(0) = 0.1 m and v(0) = 0.01 mm>s. *1.121. A spring and damper are attached to a mass of 100 kg in the arrangement given in Figure 1.8. The system is given the initial conditions x(0) = 0.1 m and v(0) = 1 mm>s. Design the spring and damper (i.e., choose k and c) such that the system will come to rest in 2 s and not oscillate more than two complete cycles. Try to keep c as small as possible. Also compute ζ. *1.122. Repeat Example 1.7.1 by using the numerical approach of the previous 5 problems. *1.123. Repeat Example 1.7.1 for the initial conditions x(0) = 0.01 m and v(0) = 1 mm>s. Section 1.10 (Problems 1.124 through 1.136) 1.124. A 2-kg mass connected to a spring of stiffness 103 N>m has a dry sliding friction force (Fc) of 3 N. As the mass oscillates, its amplitude decreases 20 cm. How long does this take? 1.125. Consider the system of Figure 1.44 with m = 5 kg and k = 9 * 103 N>m with a friction force of magnitude 6 N. If the initial amplitude is 4 cm, determine the amplitude one cycle later as well as the damped frequency. *1.126. Compute and plot the response of the system of Figure P1.126 for the case where x0 = 0.1 m, v0 = 0.1 m>s, μ = 0.05, m = 250 kg, θ = 20°, and k = 3000 N>m. How long does it take for the vibration to die out? g k m  Figure P1.126  Chap. 1 115 Problems *1.127. Compute and plot the response of a system with Coulomb damping of equation (1.100) for the case where x0 = 0.5 m, v0 = 0, μ = 0.1, m = 100 kg, and k = 1500 N>m. How long does it take for the vibration to die out? *1.128. A mass moves in a fluid against sliding friction as illustrated in Figure P1.128. Model the damping force as a slow fluid (i.e., linear-viscous damping) plus Coulomb friction because of the sliding, with the following parameters: m = 250 kg, μ = 0.01, c = 25 kg>s, and k = 3000 N>m. (a) Compute and plot the response to the initial conditions: x0 = 0.1 m, v0 = 0.1 m>s. (b) Compute and plot the response to the initial conditions: x0 = 0.1 m, v0 = 1 m>s. How long does it take for the vibration to die out in each case? Fluid m k/2 k/2 Figure P1.128 *1.129. Consider the system of Problem 1.128 part (a), and compute a new damping coefficient, c, that will cause the vibration to die out after one oscillation. $ 1.130. Compute the equilibrium positions of x + ω2nx + βx2 = 0. How many are there? $ 1.131. Compute the equilibrium positions of x + ω2nx - β2x3 + γx5 = 0. How many are there? *1.132. Consider the pendulum# of Example 1.10.3 with length of 1 m and initial conditions of θ0 = π>10 rad and θ0 = 0. Compare the difference between the response of the linear version of the pendulum equation (i.e., with sin(θ) = θ) and the response of the nonlinear version of the pendulum equation by plotting the response of both for four periods. *1.133. Repeat Problem 1.132 if the initial displacement is θ0 = π>2 rad. 1.134. If the pendulum of Example # 1.10.3 is given an initial condition near the equilibrium position of θ0 = π rad and θ0 = 0, does it oscillate around this equilibrium? *1.135. Calculate the response of the system of Problem 1.121 for the initial conditions of x0 = 0.01 m, v0 = 0, and a natural frequency of 3 rad>s and for β = 100, γ = 0. *1.136. Repeat Problem 1.135 and plot the response of the linear version of the system (β = 0) on the same plot to compare the difference between the linear and nonlinear versions of this equation of motion. MATLAB® ENGINEERING VIBRATION TOOLBOX Dr. Joseph C. Slater of Wright State University has authored a MATLAB Toolbox keyed to this text. The Engineering Vibration Toolbox (EVT) is organized by chapter and may be used to solve the Toolbox problems found at the end of each chapter. In addition, the EVT may be used to solve those homework problems suggested for computer usage in Sections 1.9 and 1.10, rather than using MATLAB directly. MATLAB and the EVT are interactive and are intended to assist in analysis, parametric studies, and design, as well as in solving homework problems. The Engineering Vibration Toolbox is licensed free of charge for educational use. For professional use, users should contact the Engineering Vibration Toolbox author directly. 116 Introduction to Vibration and the Free Response Chap. 1 The EVT is updated and improved regularly and can be downloaded for free. To download, update, or obtain information on usage or current revision, go to the Engineering Vibration Toolbox home page at http://www.cs.wright.edu/~vtoolbox This site includes links to editions that run on earlier versions of MATLAB, as well as the most recent version. An email list of instructors who use the EVT is maintained so users can receive email notification of the latest updates. The EVT is designed to run on any platform supported by MATLAB (including Macintosh and VMS) and is regularly updated to maintain compatibility with the current version of MATLAB. A brief introduction to MATLAB and UNIX is available on the home page as well. Please read the file Readme.txt to get started and type help vtoolbox after installation to obtain an overview. Once it is installed, typing vtbud will display the current revision status of your installation and attempt to download the current revision status from the anonymous FTP site. Updates can then be downloaded incrementally as desired. Please see Appendix G for further information. TOOLBOX PROBLEMS TB1.1. Fix [your choice or use the values from Example 1.3.1 with x(0) = 1 mm] the values of m, c, k, and x(0) and plot the responses x(t) for a range of values of the initial velocity # x(0) to see how the response depends on the initial velocity. Remember to use numbers with consistent units. # TB1.2. Using the values from Problem TB1.1 and x(0) = 0, plot the response x(t) for a range of values of x(0) to see how the response depends on the initial displacement. TB1.3. Reproduce Figures 1.10, 1.11, and 1.12. TB1.4. Consider solving Problem 1.53 and compare the time for each response to reach and stay below 0.01 mm. TB1.5. Solve Problems 1.121, 1.122, and 1.123 using the Engineering Vibration Toolbox. TB1.6. Solve Problems 1.126, 1.127, and 1.128 using the Engineering Vibration Toolbox. 2 Response to Harmonic Excitation This chapter focuses on the most fundamental concept in vibration analysis: the concept of resonance. Resonance occurs when a periodic external force is applied to a system having a natural frequency equal to the driving frequency. This often happens when the excitation force is derived from some rotating part, such as the helicopter shown in the top left. The rotating blade causes a harmonic force to be applied to the body of the helicopter. if the frequency of the blade rotation corresponds to the natural frequency of the body, resonance will occur as described in Section 2.1. Resonance causes large deflections, which may exceed the elastic limits and cause the structure to fail. an example familiar to most is the resonance caused by an out-of-balance tire on a car (bottom photo). The speed of tire rotation corresponds to the driving frequency. at a certain speed, the out-of-balance tire causes resonance, which is felt as shaking of the steering-wheel column. if the car is driven slower, or faster, the frequency moves away from the resonance condition and the shaking stops. 117 118 Response to Harmonic Excitation Chap. 2 This chapter considers the response to harmonic excitation of the single-degreeof-freedom spring–mass–damper system presented in Chapter 1. Harmonic excitation refers to a sinusoidal external force of a single frequency applied to the system. Recall from introductory physics that resonance is the tendency of a system to absorb more energy when the driving frequency matches the system’s natural frequency of vibration. This phenomenon commonly occurs in mechanical, acoustic, biological, and electrical systems. Examples include acoustic resonance in musical instruments, the tidal resonance in bays, basilar membranes in biological transduction of auditory input, and shaking of the front suspension of a car caused by an out-of-balance wheel. As a child, one discovers resonance when learning to swing on a playground swing (modeled as the pendulum considered in Chapter 1). Harmonic excitations are a common source of external force applied to machines and structures. Rotating machines such as fans, electric motors, and reciprocating engines transmit a sinusoidally varying force to adjacent components. In addition, the Fourier theorem indicates that many other forcing functions can be expressed as an infinite series of harmonic terms. Since the equations of motion considered here are linear, knowing the response to individual terms in the series allows the total response to be represented as the sum of the response to the individual terms. This is the principle of superposition. In this way, knowing the response to a single harmonic input allows the calculation of the response to a variety of other input disturbances of periodic nature. General periodic disturbances are discussed in Chapter 3. A harmonic input is also chosen for study because it can be solved mathematically with straightforward techniques. In addition, the response of a singledegree-of-freedom system to a harmonic input forms the foundation of vibration measurement, the design of devices intended to protect machines from unwanted oscillation, and the design of transducers used in measuring vibration. Harmonic excitations are simple to produce in laboratories, hence they are very useful in studying damping and stiffness properties. 2.1 HARMONIC EXCITATION OF UNDAMPED SYSTEMS Consider the system of Figure 2.1 for the case of negligible damping (c = 0). There are several ways to model the harmonic nature of the applied force, F(t). A harmonic function can be represented as a sine, a cosine, or a complex exponential. In the following, the driving force F(t) is chosen to be of the form F(t) = F0 cos ωt (2.1) where F0 represents the magnitude, or maximum amplitude, of the applied force and ω denotes the frequency of the applied force. The frequency ω is also called the input frequency, or driving frequency, or forcing frequency and has units of rad>s. Note that some texts use Ω rather than ω to denote the driving frequency. Sec. 2.1 119 Harmonic Excitation of undamped Systems y x c . cx F(t) m N F kx k x(t) mg Figure 2.1 A schematic of a single-degree-of-freedom system acted on by an external force, F(t), and sliding on a friction-free surface. The figure on the right is a free-body diagram of the friction-free spring–mass–damper system. Alternately, the harmonic forcing function can be represented as the sinusoid F(t) = F0 sin ωt or as the complex exponential F(t) = F0 e jω t where j is the imaginary unit. Each of these three forms of F(t) yields the same phenomenon, but in some cases one form may be easier to manipulate than others. Each is used in the following. From the right side of Figure 2.1, the sum of the forces in the y direction yields N = mg, with the result of no motion in that direction. Summing forces on the mass of Figure 2.1 in the x direction for the undamped case yields the result that the displacement x(t) must satisfy $ mx(t) + kx(t) = F0 cos ωt (2.2) where the harmonic driving force is represented by the cosine function. Note that this expression is a linear equation in the variable x(t). As in the homogeneous (unforced) case of Chapter 1, it is convenient to divide this expression by the mass, m, to yield $ x(t) + ω2n x(t) = f0 cos ωt (2.3) where f0 = F0 >m. The magnitude f0 is called the mass-normalized force with units N>kg. A variety of techniques can be used to solve this equation, which are commonly studied in a first course in differential equations. First, recall from the study of differential equations that equation (2.3) is a linear nonhomogeneous equation and that its solution is therefore the sum of the homogeneous solution (i.e., the solution for the case f0 = 0) and a particular solution. The particular solution can often be found by assuming that it has the same form as the forcing function. This is also consistent with observation. That is, the 120 Response to Harmonic Excitation Chap. 2 oscillation of a single-degree-of-freedom system excited by f0 cos ωt is observed to be of the form xp(t) = X cos ωt (2.4) where xp denotes the particular solution and X is the amplitude of the forced response. Substitution of the assumed form of the solution (2.4) into the equation of $ motion (2.3) yields (note xp = -ω2 X cos ωt) -ω2X cos ωt + ω2n X cos ωt = f0 cos ωt (2.5) Factoring out cos ωt yields ( -ω2X + ω2nX - f0) cos ωt = 0 Since cos ωt cannot be zero for all t 7 0, the coefficient of cos ωt must vanish. Setting the coefficient to zero and solving for X yields X = f0 ωn2 (2.6) - ω2 provided that ωn ∙ ω. Thus, as long as the driving frequency and natural frequency are different (i.e., as long as ωn ∙ ω), the particular solution will be of the form xp(t) = f0 ωn2 - ω2 cos ωt (2.7) This approach, of assuming that xp = X cos ωt, to determine the particular solution is called the method of undetermined coefficients in differential equations courses. Since the system is linear, the total solution x(t) is the sum of the particular solution of equation (2.7) plus the homogeneous solution given by equation (1.19). Recalling that A sin(ωnt + ϕ) can be represented as A1 sin ωnt + A2 cos ωnt (see Window 2.1), the total solution can be expressed in the form x(t) = A1 sin ωnt + A2 cos ωnt + f0 ω2n - ω2 cos ωt (2.8) where it remains to determine the values of the coefficients A1 and A2. These are determined by enforcing the initial conditions. Let the initial position and velocity be given by the constants x0 and 0 as before. Then equation (2.8) yields x(0) = A2 + f0 ωn2 - ω2 = x0 (2.9) and # x(0) = ωn A1 = v0 (2.10) Sec. 2.1 121 Harmonic Excitation of undamped Systems Window 2.1 Review of the Solution of the an Undamped Homogeneous Vibration Problem from Chapter 1 $ # mx + kx = 0 subject to x(0) = x0, x(0) = v0 has solution x(t) = A sin(ωnt + ϕ), which becomes, after evaluating the constants A and ϕ in terms of the initial conditions, x(t) = x0ωn 2x20ω2n + v20 b sin aωnt + tan-1 ωn v0 where ωn = 1k>m is the natural frequency. Via some simple trigonometry, this solution can also be written as x(t) = A1 sin ωnt + A2 cos ωnt = 0 ωn sin ωnt + x0 cos ωnt where the constants A1, A2, A, and ϕ are related by A = 2A21 + A22, tan ϕ = A2 A1 Solving equations (2.9) and (2.10) for A1 and A2 and substituting these values into equation (2.8) yields the total response x(t) = f0 f0 v0 cos ωt sin ωnt + ax0 - 2 b cos ωnt + 2 2 ωn ωn - ω ωn - ω2 (2.11) Note that the coefficients A1 and A2 for the total response given in equation (2.11) are different than those given for the free response as reviewed in Window 2.1. Also note that if the driving force is zero, f0 = 0 in equation (2.11), then A1 and A2 for the total response reduce to the values for the free response. Figure 2.2 illustrates a plot of the total response of an undamped system to a harmonic excitation and specified initial conditions. Note that both the second and third terms in equation (2.11) are not valid if the driving frequency happens to be equal to the natural frequency (i.e., if ω = ωn). Also note that as the driving frequency gets close to the natural frequency the amplitude of the resulting vibration gets very large. This large increase in amplitude defines the phenomenon of resonance, perhaps the most important concept in vibration analysis. Resonance is defined and discussed in detail in the paragraphs following the examples. 122 Response to Harmonic Excitation Chap. 2 0.1 Displacement (m) 0.05 0 0.05 0.1 0.15 0 20 40 80 60 Time (s) Figure 2.2 The response of an undamped system with ωn = 1 rad >s to harmonic excitation at ω = 2 rad >s and nonzero initial conditions of x0 = 0.01 m and 0 = 0.01 m >s and magnitude f0 = 0.1 N >kg. The motion is the sum of two sine curves of different frequencies. Example 2.1.1 Compute and plot the response of a spring–mass system modeled by equation (2.2) to a force of magnitude 23 N, driving frequency of twice the natural frequency, and initial conditions given by x0 = 0 m and v0 = 0.2 m>s. The mass of the system is 10 kg and the spring stiffness is 1000 N>m. Solution First, compute the various coefficients for the response as given in equation (2.11). The natural frequency, driving frequency, and mass-normalized force magnitude are ωn = 1000 N>m B 10 kg = 10 rad>s, ω = 2ωn = 20 rad>s, f0 = 23 N = 2.3 N>kg 10 kg The coefficients of the three terms in the response (note x0 = 0) become 0.2 m>s 2.3 N>kg f0 v0 = = -7.6667 * 10-3 m = = 0.02 m, 2 ωn 10 rad>s ωn - ω2 (102 - 202) rad2 >s2 With these values, equation (2.11) becomes x(t) = 0.02 sin 10t + 7.667 * 10-3 (cos 10t - cos 20t) m The plot of the time response is given in Figure 2.3. Any of the software packages described in Section 1.9 or Appendix G can be used to generate this time history. Sec. 2.1 123 Harmonic Excitation of undamped Systems 0.04 Displacement (m) 0.02 0 0.5 1 1.5 2 2.5 3 0.02 0.04 Time (s) Figure 2.3 The time response of the undamped system of Example 2.1.1 illustrating the effect of both initial conditions and the forcing function on the response. n Example 2.1.2 Consider the forced vibration of a mass m connected to a spring of stiffness 2000 N>m being driven by a 20-N harmonic force at 10 Hz (20π rad>s). The maximum amplitude of vibration is measured to be 0.1 m and the motion is assumed to have started from rest (x0 = 0 = 0). Calculate the mass of the system. Solution From equation (2.11) the response with x0 = x(t) = f0 ω2n - ω2 cos u - cos v = 2 sin a x(t) = 2f0 ω2n - ω2 sin a = 0 becomes (cos ωt - cos ωnt) Using the trigonometric identity equation (2.12) becomes 0 v - u v + u b sin a b 2 2 ωn - ω ωn + ω tb sin a tb 2 2 The maximum value of the total response is evident from (2.13) so that 2f0 ω2n - ω2 (2.12) = 0.1 m (2.13) 124 Response to Harmonic Excitation Solving this for m from ω2n = k>m and f0 = F0 >m yields m = (0.1 m) (2000 N>m) - 2(20 N) (0.1 m) (10 * 2π rad>s) 2 = Chap. 2 4 = 0.405 kg π2 n Two very important phenomena occur when the driving frequency becomes close to the system’s natural frequency. First, consider the case where (ωn - ω) becomes very small. For zero initial conditions, the response is given by equation (2.13), which is plotted in Figure 2.4. Since (ωn - ω) is small, (ωn + ω) is large by comparison and the term sin[(ωn - ω)>2]t oscillates with a much longer period than does sin[(ωn + ω)>2]t. Recall that the period of oscillation, T, is defined as 2π>ω, or in this case 2π >(ωn + ω)>2 = 4π>(ωn + ω). The resulting motion is a rapid oscillation with slowly varying amplitude that is called a beat. The beat frequency is based on the period of oscillation of the solid line in Figure 2.4. This is based on the time between two successive maximums, which is half the time for one complete oscillation of the dashed line in Figure 2.4, or ωbeat = |ωn - ω| To see this more clearly, note that the mathematical definition of a period, T, is the smallest time such that f(t + T) = f(t). From the solid line in Figure 2.4, this Displacement (m) 0.5 0 1 2 3 4 5 6 0.5 Time (s) Figure 2.4 The response of an undamped system of equation (2.13) for small ωn - ω illustrating the phenomenon of beats. Here f0 = 10 N, ωn = 10 rad>s, 2f0 ωn - ω tb. sin a and ω = 1.1 ωn rad >s. The dashed line is a plot of 2 2 ωn - ω2 7 125 Harmonic Excitation of undamped Systems occurs at half the period of the dashed line. This period corresponds to a frequency of |ωn - ω|. As ω becomes exactly equal to the system’s natural frequency, the solution given in equation (2.11) is no longer valid. In this case, the choice of the function X cos ωt for a particular solution fails because it is also a solution of the homogeneous equation. Therefore, the particular solution is of the form xp(t) = tX sin ωt (2.14) as explained in Boyce and DiPrima (2009). Substitution of (2.14) into equation (2.3) and solving for X yields xp(t) = f0 t sin ωt 2ω (2.15) Thus the total solution is now of the form (ω = ωn) x(t) = A1 sin ωt + A2 cos ωt + Evaluating the initial displacement x0 and velocity x(t) = f0 t sin ωt 2ω 0 (2.16) as before yields f0 v0 sin ωt + x0 cos ωt + t sin ωt ω 2ω (2.17) A plot of x(t) is given in Figure 2.5, where it can be seen that x(t) grows without bound. This defines the phenomenon of resonance (i.e., that the amplitude of vibration becomes unbounded at ω = ωn = 1k>m). This would cause the spring to break and fail. 0.15 0.1 Displacement (m) Sec. 2.1 f0 t 2n 0.05 0 0.05 f0 2n 0.1 0.15 0 2 t 4 6 8 10 Time (s) Figure 2.5 The forced response of a spring–mass system driven harmonically at its natural frequency (ω = ωn), called resonance. 126 Response to Harmonic Excitation Chap. 2 Example 2.1.3 A security camera is to be mounted on a building in an alley and will be subject to wind loads producing an applied force of F0 cos ωt, where the largest value of F0 is measured to be 15 N. This is illustrated on the left side of Figure 2.6. It is desired to design a mount such that the camera will experience a maximum deflection of 0.01 m when it vibrates under this load. The wind frequency is known to be 10 Hz and the camera mass is 3 kg. The mounting bracket is made of a solid piece of aluminum, 0.02 * 0.02 m in cross section. Compute the length of the mounting bracket that will keep the vibration amplitude less than the desired 0.01 m (ignore torsional vibration and assume the initial conditions are both zero). Note that the length must be at least 0.2 m in order to have a clear view. x(t) Wind Camera l F0 cos t m x(t) Mounting bracket ⇒ k 3EI l3 F0 cos t ⇒ k 3EI l3 .. mx(t)  kx(t)  F0 cos t Figure 2.6 Simple models of a camera and mounting bracket subject to a harmonic wind load. The sketch is on the left, the strength-of-materials model is in the middle, and the spring–mass model is on the right. Solution The sketch on the left in Figure 2.6 suggests the beam–mass, transversevibration model of Figure 1.26, repeated in the middle of Figure 2.6, for modeling this system. The beam–mass model, in turn, suggests the spring–mass system on the right in Figure 2.6. The equation of motion is then given by equation (2.2) with the beam stiffness suggested in the figure or 3EI $ mx(t) + 3 x(t) = F0 cos ωt l From strength of materials, the value of I for a rectangular beam is I = bh3 12 Thus the natural frequency of the system is given by ω2n = 3Ebh3 Ebh3 rad2 a 2 b = 3 s 12ml 4ml 3 Note that the length l is the quantity we need to solve for in the design. Sec. 2.1 127 Harmonic Excitation of undamped Systems Next, consider the expression for the maximum deflection of the response computed in Example 2.1.2. Requiring this amplitude to be less than 0.01 m yields the following two cases: ` 2f0 ω2n - ω2 ` 6 0.01 1 (a) -0.01 6 2f0 ω2n - ω2 and (b) 2f0 ω2n - ω2 6 0.01 First, consider case (a), which holds for ω2n - ω2 6 0: 2f0 - 0.01 6 ω2n - ω 2 1 2f0 6 0.01ω2 - 0.01ω2n 1 0.01ω2 - 2f0 7 0.01 1 l 3 7 0.01 Ebh3 4ml 3 Ebh3 = 0.02 1 l 7 0.272 m 4m(0.01ω2 - 2f0) Next, consider case (b), which holds for ω2n - ω2 7 0: 2f0 ω2n - ω 2 6 0.01 1 2f0 6 0.01ω2n - 0.01ω2 1 2f0 + 0.01ω2 6 0.01 1 l 3 6 0.01 Ebh3 4ml 3 Ebh3 = 0.012 1 l 6 0.229 m 4m(2f0 + 0.01ω2) Here the value of E for aluminum is taken from Table 1.2 ( 7.1 * 1010 N>m2 ) and ω is changed to rad>s. To conserve material, case (b) is chosen. Given the constraint that the length of the bracket must be at least 0.2 m, 0.2 6 l 6 0.229 so the value of l = 0.22 is chosen for the solution. Next, a couple of simple checks are performed to make sure the assumptions made in solving the problem are reasonable. First, compute the value of ωn for this value of l to see that the solution is consistent with the inequality: ω2n - ω2 = 3Ebh3 - (20π)2 = (74.543)2 - (20π)2 = 1609 7 0 12ml 3 Thus case (b) is satisfied. Note that the mass of the mounting bracket was neglected. It is always important to check assumptions. Using the density for aluminum given in Table 1.2, the mass of the bracket is m = ρlbh = ( 2.7 * 103 ) (0.22) (0.01) (0.01) = 0.149 kg This is less than the mass of the camera (about 5%), so according to Example 1.4.4 it is reasonable to ignore the spring’s mass in these calculations. n 128 Response to Harmonic Excitation Chap. 2 Example 2.1.4 In solving the security camera design problem of Example 2.1.3, torsional vibration, illustrated in Figure 2.7, was not considered. The purpose of this example is to examine if the assumption of ignoring torsion is correct or not. To decide this, determine if a wind load of 15 N vibrating at 10 Hz causes the end to move more than 0.01 m. M cos t Wind Mounting bracket l L F0 cos t Camera + + (t) k GJp ⇒ l 2r J r1 ⇒ GJp l (t) Figure 2.7 A torsional model of a camera and mounting bracket subject to a harmonic wind load. The sketch on the left is the system of Figure 2.6 showing blowing across the side of the camera causing an applied moment resulting in torsional motion. The schematic is shown in the middle and the free-body diagram is on the right. Solution Here we model the wind load as acting at a point on the tip of the camera, a distance r1 = 0.09 m from its center, creating an applied moment of M(t) = r1F0 cos 20πt Nm. Summing the moments diagram of in Figure 2.6, the equation of motion is $ Jθ(t) + kθ(t) = r1F0 cos (20πt) Here θ is the rotational displacement of the camera about the center where the bracket connects to the camera. Modeling the camera as a solid cylinder, the mass moment of inertia is (see a dynamics text or perform a Google search) J = m ( 3r 2 + L2 ) 12 where m is the mass, r = 0.05 m is the radius, and L = 2r1 = 0.18 m is the length of the cylinder. From Figure 1.24, the torsional stiffness of the mounting bracket is k = GJp l Here G is the shear modulus of aluminum (G = 2.67 * 1010 N>m2, from Table 1.2) and the polar moment of inertia of a square “rod” is Jp = 0.1406 a4, where a = 0.01 m is the length of the side of the square bracket (from Table 1.3). The value of l is taken from the solution of Example 2.1.3 to be l = 0.22 m. Substitution of the appropriate values yields k = GJp l = ( 2.67 * 1010 ) a ωn = N b ( (0.1406) ( 0.014 ) ) ( m4 ) m2 = 2.73 * 103 Nm 0.22 # m k 2.73 * 103 Nm = = 523.67 rad>s BJ B 9.975 * 10-3 kg # m2 Sec. 2.1 129 Harmonic Excitation of undamped Systems Following the expression for the maximum value of the response for zero initial conditions given in Example 2.1.2 applied to the torsional vibration problem yields θmax = † 2 ω2n r1F0 J † - ω 2 = † † 9.975 * 10-3 = 1.003 * 10-3 rad 2 2 (523.167) - (62.832) (0.09)(15) 2 The maximum linear displacement of the tip is then Xmax = r1θmax = (0.09) (1.003 * 10-3) = 9.031 * 10-5 m This is much less than the required 0.01 m so that the assumption of ignoring torsion was a reasonable one. This is largely because the torsional frequency (523 rad>s) is much higher then the bending frequency (75 rad>s), making the denominator in the maximum deflection calculation larger and hence the deflection smaller. n The previous development assumed that the harmonic forcing function was described by the cosine function. As pointed out earlier, the harmonic forcing function may also be represented in terms of the sine function. In this case, equation (2.2) becomes $ $ mx(t) + kx(t) = F0 sin ωt or x(t) + ω2n x(t) = f0 sin ωt (2.18) Proceeding as before, using the method of undetermined coefficients, the particular solution for the case of sinusoidal excitation becomes xp(t) = X sin ωt (2.19) Substitution of this assumed solution form into equation (2.18) yields -ω2X sin ωt + ω2n X sin ωt = f0 sin ωt (2.20) Factoring out sin ωt ( ∙0) and solving for X yields X = f0 ω2n (2.21) - ω2 so that the particular solution is xp(t) = f0 ωn2 - ω2 sin ωt (2.22) The total solution is the sum of the homogenous solution and the particular solution, or x(t) = A1 sin ωnt + A2 cos ωnt + f0 ωn2 - ω2 sin ωt (2.23) 130 Response to Harmonic Excitation Chap. 2 It remains to evaluate the constants A1 and A2 in terms of the given initial conditions x0 and 0. To this end set t = 0 in equation (2.23) and its first derivative to get # x(0) = x0 = A2 and x(0) = ωnA1 + 1 A1 = f0 v0 ω 2 ωn ωn ωn - ω2 ωf0 ω2n - ω2 = v0 (2.24) and A2 = x0 The total solution for a sinusoidal harmonic input is thus x(t) = x0 cos ωnt + a f0 f0 v0 ω sin ωt b sin ωnt + 2 2 2 ωn ωn ωn - ω ωn - ω2 (2.25) It is important to note the differences between equations (2.11) and (2.25). Equation (2.11) is the response due to a cosine input, and equation (2.25) is the response due to a sine input. In particular, note that the input force term modifies the initial condition response differently in each case. Note that equation (2.17) can also be obtained from equation (2.11) by taking the limit as ω S ωn using the limit theorems from calculus. 2.2 HARMONIC EXCITATION OF DAMPED SYSTEMS As noted in Chapter 1, some sort of damping or energy dissipation is always present (see Window 2.2). In this section, the response of a viscously damped single-degreeof-freedom system subjected to harmonic excitation is considered. Summing forces on the mass of Figure 2.1 in the x direction yields $ # mx + cx + kx = F0 cos ωt (2.26) Dividing by the mass m yields $ # x + 2ζωnx + ω2nx = f0 cos ωt where ωn = 1k>m, ζ = c>(2mωn), and f0 = F0 >m. The calculation of the particular solution for the damped case is similar to that of the undamped case and follows the method of undetermined coefficients. From differential equations it is known that the forced response of a damped system is of the form of a harmonic function of the same frequency as the driving force with a different amplitude and phase. The phase shift is expected because (2.27) Sec. 2.2 131 Harmonic Excitation of Damped Systems Window 2.2 Review of the Solution of the Damped Homogeneous Vibration Problem (0 * 𝛇 * 1) from Chapter 1 $ # # mx + cx + kx = 0 subject to x(0) = x0, x(0) = v0 has the solution x(t) = Ae -ζωnt sin (ωdt + ϕ) where k is the undamped natural frequency Am 2 ζ = is the damping ratio 2mωn ωn = ωd = ωn 21 - ζ2 is the damped natural frequency and the constants A and ϕ are determined by the initial conditions to be A = B v0 + ζωn x0 2 b ωd x0ωd v0 + ζωn x0 x20 + a ϕ = tan-1 Alternately, the solution can be written as x(t) = e -ζωnt c v0 + ζωnx0 sin ωdt + x0 cos ωdt d ωd of the effect of the damping force. Following the method of undetermined coefficients, the particular solution is assumed to be of the form xp(t) = X cos (ωt - θ) (2.28) To make the computations easy to follow, this is written in the equivalent form xp(t) = As cos ωt + Bs sin ωt (2.29) where the constants As = X cos θ and Bs = X sin θ satisfying X = 2A2s + B2s and θ = tan-1 are the undetermined constant coefficients. Bs As (2.30) 132 Response to Harmonic Excitation Chap. 2 Taking derivatives of the assumed form of the solution given by (2.29) yields # xp(t) = -ωAs sin ωt + ωBs cos ωt (2.31) $ xp(t) = - ω2(As cos ωt + Bs sin ωt) (2.32) and # $ Substitution of xp, xp, and xp into the equation of motion given by equation (2.27) and grouping terms as coefficients of sin ωt and cos ωt yields 1-ω2 As + 2ζωnω Bs + ω2n As - f02 cos ωt + 1-ω2 Bs - 2ζωn ωAs + ω2n Bs2 sin ωt = 0 (2.33) This equation must hold for all time, in particular for t = π >2ω, so that the coefficient of sin ωt must vanish. Similarly, for t = 0 the coefficient of cos ωt must vanish. This yields the two equations 1ωn2 - ω22 As + 1 -2ζωnω2 As + 1 2ζωnω2 Bs 1ωn2 - ω2 2 Bs = f0 (2.34) = 0 in the two undetermined coefficients As and Bs. These two linear equations may be written as the single matrix equation c 2ζωnω As f d = c 0d 2 2d c ωn - ω Bs 0 ω2n - ω2 -2ζωnω 1ωn2 - ω22 f0 - ω222 + 12ζωnω22 which has solution (compute the matrix inverse and multiply; see Appendix C) As = Bs = 1ω2n 1ω2n - ω222 + 12ζωnω22 (2.35) 2ζωnωf0 Substitution of these values into equations (2.30) and (2.28) yields that the particular solution is 21 X f0 2 1 2ζωnω 2 cos aωt - tan-1 4 xp(t) = ω2n 2 2 - ω + 2 θ 2ζωnω 4 ω2n - ω2 b (2.36) Sec. 2.2 133 Harmonic Excitation of Damped Systems The total solution is again the sum of the particular solution and the homogeneous solution obtained in Section 1.3. For the underdamped case (0 6 ζ 6 1) this becomes x(t) = Ae -ζωnt sin 1ωdt + ϕ2 + X cos 1ωt - θ2 (2.37) where X and θ are the coefficients of the particular solution as defined by equation (2.36), and A and ϕ (different from those of Window 2.2) are determined by the initial conditions. Note that for large values of t, the first term, or homogeneous solution, approaches zero and the total solution approaches the particular solution. Thus xp(t) is called the steady-state response and the first term in equation (2.37) is called the transient response. The values of the constants of integration in equation (2.37) can be found from the initial conditions by the same procedure used in the development for computing the response of an undamped system given in equation (2.11). Following the development from equation (2.29), write the transient term as A sin ωdt + C cos ωdt rather than writing the constants of integration as a magnitude and phase and the particular solution in the form given in equation (2.37). The resulting response for the underdamped case is x(t) = e -ζωnt e ax0 - + - ω22 2 + 12ζωnω22 b cos ωdt f0 1ωn - ω 2 ζωn ax0 b ωd 1ω2n - ω22 2 + 12ζωnω2 2 +a - 1ω2n f0 1ωn2 - ω22 ωd 31 1 ω2n 2 2 + 12ζωnω2 2ζωnω2f0 ω2n 2 2 - ω 2 - ω + 12ζωnω2 2 f0 2 2 2 4 2 + 31ω2n (2.38) v0 b sin ωdt f ωd - ω22 cos ωt + 2ζωnω sin ωt 4 This is an alternate form of equation (2.37) showing the direct influence of the forcing function on the transient part of the response (i.e., the coefficient of the exponential term). Note that the expression for the forced response of an underdamped system given here collapses to the forced response of an undamped system given by equation (2.11) when the damping is set to zero in equation (2.38). Problem 2.20 requires the calculation of the constants A and ϕ for equation (2.37). A summary of phase and amplitude formulas is given in Window 2.3 for both the free and forced response. Note that A and ϕ, the constants describing the transient response in equation (2.37), will be different from those calculated for the free-response case given in equation (1.38) or Window 2.2. This is because part of the transient term in 134 Response to Harmonic Excitation Chap. 2 Window 2.3 Summary of Phase and Amplitude Relationships for Undamped and Underdamped Single-Degree-of-Freedom Systems for Both the Free Response, F(t) ∙ 0, and for the Forced Response, F(t) ∙ F0 cos 𝛚t, Cases x(t) = A sin 1ωnt + ϕ2 + X cos ωt The general response for the undamped case has the form where for the free response ϕ = tan-1 ωn x0 v20 , A = x20 + 2 , X = 0 v0 C ωn and for the forced response: ϕ = tan-1 ωn ( x0 - X ) f0 v0 2 ,A = a b + ( x0 - X ) 2, X = 2 ω v0 C n ωn - ω2 x(t) = Ae -ζωnt sin 1ωdt + ϕ2 + X cos 1ωt - θ2 The general response in the underdamped case has the form x0ωd 1 ,A = 2 1 v0 + ζωnx0 2 2 + 1x0ωd2 2, X = 0 ωd v0 + ζωn x0 where the free response: ϕ = tan-1 and for the forced response θ = tan-1 ϕ = tan-1 v0 + 2ζωnω ω2n 1 x0 , X = 21 ωd 1 x0 - X cos θ 2 2 - ω 2 f0 ω2n 2 2 - ω - X cos θ 2 ζωn - ω X sin θ + 1 2ζωnω 2 2 and A = , x0 - X cos θ sin ϕ equation (2.37) is due to the magnitude of the excitation force and part is due to the initial conditions as indicated in equation (2.38). Example 2.2.1 Examine the units for computing the forced response of a damped system. Often the equation-of-motion quantities (forces) are given in Newtons, whereas the initial displacement and velocity are given in mm. It is important to write the initial conditions in the correct units. Sec. 2.2 135 Harmonic Excitation of Damped Systems Solution First, examine units in the mass-normalized equation of motion as given in equation (2.27) repeated here: $ # x + 2ζωnx + ω2nx = f0 cos ωt The units for f0 are N>kg = m>s2, the units of acceleration, agreeing with the first term in the equation. The damping ratio ζ has no units, so the units of the damping term are those of ωn or rad>s, which when multiplied by the velocity yields m>s2. Likewise, the units of the natural frequency squared are rad2 >s2 so the stiffness term also has units of m>s2. Thus equation (2.27) is consistent in terms of units. Next, consider the solution. Since the amplitude of the particular solution, X, has the units of m (the units of f0 are N>Kg or m>s2) X = 2 1 ω2n - ω22 2 f0 m 2 ∙s f0 a rad b = 2 2 2 ∙s2 + 1 2ζωnω2 2 1 ωn - ω 2 2 + 1 2ζωnω 2 2 m the initial condition x0 must also be given in m because the value of amplitude A contains the numerator term x0 - X cos(θ). The same is true for the phase angle ϕ, which also contains the initial velocity added to Xζωn which will have units of m>s. Thus in solving for the force response of a damped system it is important that the initial conditions are stated in terms of the same units that the equation of motion is expressed in. n Example 2.2.2 A damped spring–mass system with values of c = 100 kg>s, m = 100 kg, and k = 910 N>m, is subject to a force of 10 cos (3t) N. The system is also subject to initial conditions: x0 = 1 mm and 0 = 20 mm>s. Compute the total response, x(t), of the system. Solution Following equation (2.26) with the values given here, the system to be solved is $ # 100 x(t) + 100 x(t) + 910x(t) = 10 cos 3t, x0 = 0.001 m, v0 = 0.02 m>s where the units (and thus numerical values) of the initial conditions have been changed to agree with the equation of motion per the previous example. Dividing by the mass yields the vibration properties f0 = F0 10 m k 910 rad = = 0.1 2 , ωn = = = 3.017 , m 100 A m A 100 s s ζ = c 100 = = 0.166, 21mk 21100 · 910 ωd = ωn 21 - ζ2 = 3.01721 - 0.1662 = 2.975 rad>s Since ω = 3 rad>s, the system is near resonance. Computing the amplitude and phase for the particular solution from the values given in Window 2.3 yields X = 21 2 f0 ω2n - ω 2 2 + 1 2ζωnω 2 = 2 2 1 3.017 - 3 2 2 2 2 1 2 · 0.166 · 3.017 · 3 2 2 0.1 + = 0.033 m 136 Response to Harmonic Excitation θ = tan-1 2ζωnω ω2n - ω2 = tan-1 Chap. 2 2 # 0.166 # 3.017 # 3 = 1.537 rad 3.0172 - 32 ωd 1 x0 - X cos θ 2 Next, compute phase for the transient response ϕ = tan-1 = v0 + 1 x0 - X cos θ 2 ζωn - ωX sin θ 2.975(0.001 - 0.033 # 0.033) 0.02 + (0.001 - 0.033 # 0.033)0.166 # 3.017 - 3 # 0.033 # 0.999 = 4.089 * 10-3 rad The amplitude of the transient is A = x0 - X cos θ 0.001 - 0.033 # 0.0333 = -0.027 m = sin ϕ 4.089 * 10-3 x(t) = Ae -ζωnt sin 1ωdt + ϕ2 + X cos 1ωt - θ2 The total response is then written from equation (2.37) as = -0.027e -0.5t sin 1 2.975t + 4.089 * 10-3 2 + 0.033 cos 1 3t - 1.537 2 m n Example 2.2.3 Compute the constants of integration A and ϕ of equation (2.37) and compare these values to the values of A and ϕ for the unforced case given in Window 2.2 for the parameters ωn = 10 rad>s, ω = 5 rad>s, ζ = 0.01, F0 = 1000 N, m = 100 kg, and the initial conditions x0 = 0.05 m and 0 = 0. Solution First, compute the values of X and θ from equation (2.36) for the particular solution (forced response). These are X = f0 3(ωn2 - ω2)2 + (2ζωnω)2 = 0.133 m and 2ζωnω θ = tan-1 a 2 b = 0.013 rad ωn - ω2 so that the phase for the particular solution is nearly zero (0.76°). Thus the solution given by equation (2.37) is of the form x(t) = Ae -0.1t sin (9.999t + ϕ) + 0.133 cos (5t - 0.013) Differentiating this solution yields the velocity expression v(t) = - 0.1Ae -0.1t sin (9.999t + ϕ) + 9.999Ae -0.1t cos (9.999t + ϕ) - 0.665 sin (5t - 0.013) Sec. 2.2 137 Harmonic Excitation of Damped Systems Setting t = 0 and x(0) = 0.05 in the expression for x(t) and solving for A yields A = 0.005 - 0.133 cos ( - 0.013) sin (ϕ) = - 0.083 sin (ϕ) Substitution of this value of A, setting t = 0 and (0) = 0 in the expression for the velocity yields 0 = - 0.1( - 0.083) + 9.999( -0.083) cot(ϕ) + 0.665sin(0.013) Solving for the value of ϕ yields ϕ = 1.55 rad (88.8°) and thus A = - 0.083 m, the values of the amplitude and phase of the transient part of the solution (including the effects of the initial conditions and the applied force). Next, consider the coefficients A and ϕ evaluated for the homogenous case F0 = 0, but keeping the same initial conditions. Using these values (see Window 2.2), the incorrect magnitude and phase become A = B (0.05)2 + c (0.01)(10)(0.05) 1021 - (0.01)2 d = 0.05 m 2 and ϕ = tan-1 a 9.999 b = 1.521 rad (87.137°) 0.05(10) Comparing these two sets of values for the magnitude A and the two sets of values for the phase ϕ, we see that they are very different. Thus the values of the constants of integration are greatly affected by the forcing term. In particular, the amplitude of the transient is greatly increased and its phase is reduced by the influence of the driving force. n It is common to ignore the transient part of the total solution given by equation (2.37) and to focus only on the steady-state response: X cos (ωt - θ). The rationale for considering only the steady-state response is based on the value of the damping ratio ζ. If the system has relatively large damping, the term e -ζωnt causes the transient response to die out very quickly—perhaps in a fraction of a second. If, on the other hand, the system is lightly damped (ζ is very small), the transient part of the solution may last long enough to be significant and should not be ignored. The decision whether to ignore the transient part of the solution should also be based on the application. In fact, in some applications (such as earthquake analysis 138 Response to Harmonic Excitation Chap. 2 or satellite analysis) the transient response may become even more important than the steady-state response. An example is the Hubble space telescope, which originally experienced a transient vibration that lasted over 10 minutes, causing the telescope to be unusable every time it passed out of the earth’s shadow, until the system was corrected. The transient response can also be very important if it has a relatively large amplitude. Usually, devices are designed and analyzed based on the steady-state response, but the transient should always be checked to determine whether it is reasonable to ignore it or if it should be considered seriously. With this caveat in mind, it is of interest to consider the magnitude, X, and the phase, θ, of the steady-state response as a function of the driving frequency. Examining the form of equation (2.36) and comparing it to the assumed form X cos (ωdt - θ) yields the fact that the amplitude, X, and the phase, θ, are X = 2 1 ω2n - ω2 2 + f0 1 2ζωnω 2 2 , θ = tan-1 2ζωnω ω2n - ω2 (2.39) After some manipulation (i.e., factoring out ω2n and dividing the magnitude by F0 >m), these expressions for the magnitude and phase can be written as Xω2n 1 Xk = = F0 f0 2 2 311 - r 2 + 1 2ζr2 2 , θ = tan-1 2ζr 1 - r2 (2.40) Here r is the frequency ratio r = ω >ωn, a dimensionless quantity. Equations (2.40) for the magnitude and phase are plotted versus the frequency ratio r in Figure 2.8 for several values of the damping ratio ζ. Note that as the driving frequency approaches the undamped natural frequency (r S 1), the magnitude approaches a maximum value for those curves corresponding to light damping (ζ … 0.1). Also note that as the driving frequency approaches the undamped natural frequency, the phase shift crosses through 90°. The phase lies between zero and π, as discussed in Window 2.4. This defines resonance for the damped case. These two observations have important uses in both vibration design and measurement. As ω approaches zero, the amplitude approaches f0 >ω2n and as ω becomes very large, the amplitude approaches zero asymptotically. It is also important from the design point of view to note how the amplitude of steady-state vibration is affected by changing the damping ratio. This is illustrated in Figure 2.9 and Example 2.2.4. Figure 2.9 is a repeat of the magnitude curve presented in Figure 2.8, except that here the magnitude is plotted on a log scale so that the curves for both small and large values of damping may be detailed on the same graph. The use of a log scale for vibration amplitude is common in vibration analysis and measurement. Note that as the damping ratio is 139 Harmonic Excitation of Damped Systems X 2n f0 Normalized Amplitude   0.1 4   0.25 2   0.5   0.7 0 1 0.5 1.5 2 r Frequency ratio (a) (r) 4   0.1   0.25   0.5   0.7 3 Phase Sec. 2.2  2 2 1 0 0.5 1 1.5 2 r Frequency ratio Figure 2.8 A plot of the (a) normalized magnitude 1 Xω2n >f0 = Xk>F0 2 and (b) phase of the steady-state response of a damped system versus the frequency ratio for several different values of the damping ratio ζ as determined by equation (2.40). (b) 140 Response to Harmonic Excitation Chap. 2 Window 2.4 The Phase Shift in the Particular Solution for the Forced Response of an Underdamped System The particular solution for a driving force of F0 cos ωt is assumed to be of the form xp(t) = X cos (ωt - θ) as illustrated in (a). Here the phase shift θ is determined by θ = tan-1 a 2ζωnω ωn2 - ω2 b Since the numerator of the argument is always positive, the quadrant that θ lies in is determined by the sign of ω2n - ω2. F0 cos (t)  t  2n  n22 n (a) X cos (t) X cos (t) 2n n22 n (b) The phase shift is calculated from the arctangent function, which must be computed assuming that θ lies between 0 … θ … π, as the polar plot in (b) shows. The simple definition of the arctangent, however, looks for values of θ between –π >2 … θ … π >2. Thus, when using a code or calculator, use the “atan2” function, which treats the negative value of the tangent to be in the fourth quadrant rather than the second. increased, the peak in the magnitude curve decreases and eventually disappears. As the damping ratio decreases, however, the peak value increases and becomes sharper. In the limit as ζ goes to zero, the peak climbs to an infinite value in agreement with the undamped response at resonance (see Figure 2.5). Also note that the peak value of the amplitude at resonance varies a full order of magnitude as the damping changes. 141 Harmonic Excitation of Damped Systems Xk F0   0.01 4   0.02 2 Normalized magnitude Sec. 2.2 10 6 4   0.1 2   0.2 1 6 4   0.75   1.0 2 0.1 0.0 0.5 1.0 1.5 2.0 2.5 r 3.0 Frequency ratio Figure 2.9 The magnitude (log scale) of the steady-state response versus the frequency ratio for several values of the damping ratio ζ. Example 2.2.4 Consider a simple spring–mass–damper system with m = 49.2 * 10 - 3 kg, c = 0.11 kg>s, and k = 857.8 N>m. Calculate the value of the steady-state response if ω = 132 rad>s for f0 = 10 N>kg. Calculate the change in amplitude if the driving frequency changes to ω = 125 rad>s. Solution The frequency and damping ratio are determined from the given values as ωn = k c = 132 rad>s, ζ = = 0.0085 Am 22mk respectively. From equation (2.39) the magnitude of xp(t) is 0 xp(t) 0 = X = = = f0 3(ω2n - ω2) + (2ζωnω)2 53(132)2 - (132)24 2 + 10 2(0.0085)(132)2 = 0.034 m 3 2(0.0085)(132)(132)4 261>2 10 142 Response to Harmonic Excitation Chap. 2 If the driving frequency is changed to 125 rad>s, the amplitude becomes 53(132) 2 - (125) 4 2 2 3 2(0.0085)(132)(125) 4 2 6 1>2 10 + = 0.005 m So a slight change in the driving frequency from near resonance at 132 rad > s to 125 rad > s (about 5%) causes an order-of-magnitude change in the amplitude of the steady-state response. n It is important to note that resonance is defined to occur when ω = ωn (i.e., when the driving frequency becomes equal to the undamped natural frequency). This also corresponds with a phase shift of 90°(π>2). Resonance does not, however, exactly correspond with the value of ω at which the peak value of the steady-state response occurs. This can be seen by the simple calculation in the following example. Example 2.2.5 Derive equation (2.40) for the normalized magnitude and calculate the value of r = ω>ωn for which the amplitude of the steady-state response takes on its maximum value. Solution From equation (2.36) the magnitude of the steady-state response is X = 3 1ω2n - ω22 2 + 12ζωnω2 2 f0 = F0 >m 3 1ω2n - ω22 + 12ζωnω2 2 Factoring ω2n out of the denominator and recalling that ω2n = k>m yields X = ω2n B c1 - a F0 >m ω ω b d + c2ζ a b d ωn ωn 2 2 F0 >k = 2 3 ( 1 - r 2) 2 + (2ζr ) 2 where r = ω>ωn. Dividing both sides by F0 >k yields equation (2.40). The maximum value of X will occur where the first derivative of X>F0 vanishes, that is Thus d Xk d a b = 53( 1 - r 2 ) 2 + (2ζr)2 4 -1>2 6 = 0 dr F0 dr rpeak = 31 - 2ζ2 = ωp ωn (2.41) defines the value of the driving frequency, ω p at which the peak value of the magnitude occurs. This holds only for underdamped systems for which ζ 6 1> 12. Sec. 2.2 143 Harmonic Excitation of Damped Systems Otherwise, the magnitude does not have a maximum value or peak for any value of ω 7 0 because 21 - 2ζ2 becomes an imaginary number for values of ζ larger than 1> 12. Note also that this peak occurs a little to the left of, or before, resonance 1 r = 1 2 since rpeak = 21 - 2ζ2 6 1 This can be seen in both Figures 2.8 and 2.9. The value of the magnitude at rpeak is Xk 1 = F0 2ζ21 - ζ2 (2.42) which is obtained simply by substituting rpeak = 21 - 2ζ2 into the expression for the normalized magnitude Xk>F0. n Note that for damped systems resonance is usually defined, as in the undamped case, by r = 1 or ωn = ω. However, this condition does not define precisely the peak value of the magnitude of the steady-state response as defined by equation (2.40) and as plotted in Figure 2.9. This is the point of Example 2.2.5, which illustrates that the maximum value of Xk>F0 occurs at r = 21 - 2ζ2 if 0 … ζ 6 1> 12 and at r = 0 if ζ 7 1> 12. For the small damping case 1 ζ 6 1> 12 2 the value of the driving frequency corresponding to the maximum value of Xk>F0 is called the peak frequency, denoted by ωp, which has the value derived previously: ωp = ωn 21 - 2ζ2 for 0 … ζ … 1 12 (2.43) Note that as the damping decreases, ωp approaches ωn, resulting in the usual undamped resonance condition. As ζ increased from zero, the curves in Figure 2.9 have peaks that occur farther and farther away from the vertical line r = 1. Eventually, the damping ratio increases past the value 1> 12 and the largest value of Xk>F0 occurs at r = 0 In many applications ζ is small, so that the value 21 - 2ζ2 is very close to 1. Hence the undamped resonance condition ω = ωn 1 i.e., r = 1 2 is often used for resonance in the (lightly) damped case as well. As an example, for ζ = 0.1 a system with an undamped natural frequency of 200 Hz would have a peak value of 198 Hz, which is less than a 1% error (i.e., r = 0.9899 instead of r = 1). Hence, in practice, the value of the frequency corresponding to the peak is often taken to be simply the undamped natural frequency. It is interesting to examine the peak value of the magnitude response as a function of the damping ratio for underdamped systems. The bottom plot in Figure 2.10 shows how the damping affects the peak amplitude by plotting Xk>F0 as a function of ζ as given by equation (2.41).The plot in the upper right shows how the peak frequency ratio changes with the damping ratio. Note from the figure that the magnitude varies across three orders of magnitude as the damping is increased (0 … ζ … 0.707). Also note that the peak value moves to the left of r = 1 for high values of the damping ratio. 144 Response to Harmonic Excitation Chap. 2 1 rpeak 0.5 1000 Xk F0 100 0 0.2 0.4 0.6 0.8  10 0 0 0.2 0.4 0.6 0.8  Figure 2.10 The lower graph is a plot of equation (2.41) on a log scale indicating how the increased damping reduces the peak response. The plot on the upper right shows how much the peak amplitude shifts from r = 1 as the damping increases. (Note that these plots are only defined for ζ … 0.707.) To explain physically the phenomena of resonance, consider the steady-state forced response of the system where the applied force is F0 cos (ωt), the displacement # is xp(t) = X cos (ωt - θ), and the velocity is xp(t) = -ωX sin (ωt - θ). At resonance, # θ = π >2. Thus xp(t)(resonance) = ωX cos (ωt). This shows that at resonance the velocity and the force are exactly in phase but have different magnitudes. Physically, this means that the force is always pushing in the direction of the velocity and that the force changes magnitude and direction just as the velocity does. This condition will cause the vibration amplitude of the system to reach its maximum value because at resonance the external force never opposes the velocity. 2.3 ALTERNATIVE REPRESENTATIONS As you may recall from the theory of differential equations, there are a variety of methods useful for calculating solutions of a spring–mass–damper system excited by a harmonic force as described by equation (2.26). In Section 2.2, the method of undetermined coefficients was used. In this section, three other approaches to solve this problem are discussed: a geometric approach, a frequency response approach, and a transform approach. Sec. 2.3 145 Alternative Representations 2.3.1 Geometric Method The geometric approach consists of solving equation (2.26) by treating each force as # $ a vector. Recall that xp, xp, and xp will each be 90∙ out of phase with each other. Each of these are plotted in Figure 2.11 for the assumed solution xp = X cos (ωt - θ), # $ xp = ωX cos (ωt - θ + 90∙), and xp = -ω2X cos (ωt - θ). Adding these three quantities as vectors indicates that X can be solved in terms of F0 by combining the sides of the right triangle ABC to yield F 20 = ( k - mω2) 2 X2 + ( cω ) 2X2 (2.44) or X = F0 3 ( k - mω2 ) 2 + ( cω ) 2 (2.45) and θ = tan-1 cω k - mω2 (2.46) Substituting in the values for F0 = mf0 and c = 2mωnζ illustrates that this solution is identical with equation (2.36) derived by the method of undetermined coefficients. Note from the figure that at resonance ω2 = k>m causing line AB to be zero, lines CD and AE become the same length and the angle θ becomes a right angle. Thus at resonance the phase shift is 90°. The graphical method of solving equation (2.26) is more illustrative than useful, as it is difficult to extend to other forms of the forcing function or to more complicated problems. It is presented here because the method potentially helps clarify and illustrate the forced vibration of a simple single-degree-of-freedom system. In particular, Figure 2.11 makes it easy to see that at resonance (θ = 90°) the applied force and damping force are acting in the same direction and the stiffness force is equal and opposite to the inertial force. Im C D F0 m2X cX C F0 E  t A kX B Re A kX cos(t– )  (k – m2)X cX B Figure 2.11 A graphical representation of the solution of equation (2.26). 146 Response to Harmonic Excitation Chap. 2 An alternative method that is similar to the graphical approach is to treat the solution of equation (2.26) as a complex function. This leads to a frequency response description of forced harmonic motion and is more useful for complicated problems involving many degrees of freedom. Complex functions are reviewed in Appendix A. 2.3.2 Complex Response Method Euler’s formula for trigonometric functions relates the exponential function to harmonic motion by the complex relation Ae jωt = A cos ωt + (A sin ωt)j (2.47) where j = 1-1. Thus Ae is a complex function with a real part (A cos ωt) and an imaginary part (A sin ωt). Appendix A reviews complex numbers and functions. With this notation in mind, Ae jωt represents a harmonic function and can be used to discuss forced harmonic motion by rewriting the equation of motion (2.26) $ # mx + cx + kx = F0 cos ωt jωt as the complex equation $ # mx(t) + cx(t) + kx(t) = F0 e jωt (2.48) Here, the real part of the complex solution corresponds to the physical solution x(t). This representation is extremely useful in solving multiple-degree-of-freedom systems (Chapter 4), as well as in understanding vibration measurement systems (Chapter 7). This method proceeds by assuming that the complex particular solution of equation (2.48) is of the exponential form xp(t) = Xe jωt (2.49) where X is now a complex-valued constant to be determined. Substitution of this into equation (2.48) yields ( -ω2m + cjω + k)Xe jωt = F0 e jωt (2.50) Since ejωt is never zero, it can be canceled and this last expression can be rewritten as X = F0 (k - mω2) + (cω)j = H(jω)F0 (2.51) The complex quantity H(jω), defined by H(jω) = 1 (k - mω2) + (cωj) (2.52) Sec. 2.3 147 Alternative Representations is called the (complex) frequency response function. Following the rules for manipulating complex numbers (i.e., multiplying by the complex conjugate over itself and taking the modulus of the result as outlined in Appendix A) yields X = where 3( k ) + (cω)24 1>2 F0 2 2 - mω θ = tan-1 e -jθ cω (2.53) (2.54) (k - mω2) Substituting the value for X into equation (2.49) yields the solution xp(t) = 3( k ) + (cω)24 1>2 F0 2 2 - mω e -j(ωt - θ) (2.55) The real part of this expression corresponds to the solution given in equation (2.36) obtained by the method of undetermined coefficients. The complex exponential approach for obtaining the forced harmonic response corresponds to the graphical approach described in Figure 2.11 by labeling the x axis as the real part of ejωt and the y axis as the complex part. Example 2.3.1 Use the frequency response approach to compute the amplitude of the particular solution for the undamped system of equation (2.2) defined by $ mx(t) + kx(t) = F0 cos ωt Solution First, write equation (2.2) with the forcing function modeled as a complex exponential: $ mx(t) + kx(t) = F0 e jωt Dividing by the mass, m, yields the monic form $ x(t) + ω2nx(t) = f0 e jωt Assume a particular solution of the exponential form given in equation (2.49) and substitute into the last expression to get Solving for X yields 1- ω2 + ω2n 2 Xe jωt = f0 Xe jωt X = f0 ω2n - ω2 This is in perfect agreement with equation (2.7) derived using the cosine representation of the forcing function and (2.21) derived using the sine representation. This also agrees with solution given in equation (2.36) for the damped case, by setting ζ = 0 in that expression. n 148 Response to Harmonic Excitation Chap. 2 The method used here to derive the solution for the forced response and the resulting frequency response function is very similar to the eigenvalue approach to solving vibration problems. This approach, introduced in Section 1.3, is used extensively in Chapter 4 and consists of assuming solutions with exponential time dependence, as illustrated previously. 2.3.3 Transfer Function Method Next consider using the Laplace transform (see Appendix B and Section 3.4 for a review) approach to solve for the particular solution of equation (2.26). The Laplace transform method is a powerful approach that can be used for a variety of forcing functions (see Section 3.4) and can be readily applied to multipledegree-of-freedom systems. Taking the Laplace transform of the equation of motion (2.26) $ # mx(t) + cx(t) + kx(t) = F0 cos ωt assuming that the initial conditions are zero, yields ( ms2 + cs + k ) X(s) = F0 s 2 s + ω2 (2.56) where s is the complex transform variable and X(s) denotes the Laplace transform of the unknown function x(t). Solving algebraically for the unknown function X(s) yields X(s) = F0 s ( ms + cs + k )( s2 + ω2 ) 2 (2.57) which represents the transformed solution. To calculate the inverse Laplace transform of X(s), the right side of equation (2.57) can be found in a table of Laplace transform pairs, or the method of partial fractions can be used to reduce the righthand side of equation (2.57) to simpler quantities for which the inverse Laplace transform is known. The solution obtained by the inversion procedure is, of course, equivalent to the solution given in (2.36) and again in (2.55). This solution technique is discussed in more detail in Section 3.4. Of particular use is the frequency response function defined by equation (2.52). This function is related to the Laplace transform for a vibrating system. Consider equation (2.26) and its Laplace transform ( ms2 + cs + k ) X(s) = F(s) (2.58) where F(s) symbolically denotes the Laplace transform of the driving function [i.e., the right-hand side of (2.26)]. Manipulating equation (2.58) yields X(s) = F(s) 1 ( ms2 + cs + k ) = H(s) (2.59) Sec. 2.3 149 Alternative Representations which expresses the ratio of the Laplace transform of the output (response) to the Laplace transform of the input (driving force) for the case of zero initial conditions. This ratio, denoted by H(s), is called the transfer function of the system and provides an important tool for vibration analysis, design, and measurement as discussed in the remaining chapters. Recall that the Laplace transform variable s is a complex number. If the value of s is restricted to lie along the imaginary axis in the complex plane (i.e., if s = jω), the transfer function becomes H(jω) = 1 k - mω2 + cωj (2.60) which, upon comparison with equation (2.52), is the frequency response function of the system. Hence the frequency response function of the system is the transfer function of the system evaluated along s = jω. Both the transfer function and the frequency response function are used in Chapters 4 and 7. Example 2.3.2 Consider the system of Figure 2.12. Let J denote the inertia of the wheel and hub about the shaft, and let k denote the torsional stiffness of the system. The suspension system is subjected to a harmonic excitation as indicated. Compute an expression for the forced response using the Laplace transform method (assuming zero initial conditions and that the tire is not touching the ground). F0 sin t a Wheel g  l Frame Figure 2.12 A schematic of a torsional suspension. Solution Modeling the system as a torsional vibration problem and summing moments about the shaft, the equation of motion becomes $ Jθ + kθ = aF0 sin ωt 150 Response to Harmonic Excitation Chap. 2 Taking the Laplace transform (see Appendix B) of the equation of motion yields Js2X(s) + kX(s) = aF0 ω s + ω2 2 where X(s) is the Laplace transform of θ(t). Algebraically solving for X(s) yields X(s) = aωF0 1 ( s2 + ω2 )( Js2 + k ) Next use a table (see Appendix B) to compute the inverse Laplace transform to get θ(t) = L-1 ( X(s) ) = aωF0L-1 a = 1 ( s2 + ω2 )( Js2 + k ) b aωF0 -1 aωF0 1 1 1 1 sin ωntb L a 2 b = a sin ωt 2 2 ω 2 2 2 ω J J ω - ωn ( s + ω )( s + ωn ) n Here L - 1 denotes the inverse Laplace transform and the natural frequency is ωn = k AJ Note that the solution computed here using the Laplace transform agrees with the solution obtained using the method of undetermined coefficients expressed in equation (2.25) for the case of zero initial conditions. The transfer function for the system is simply X(s) F(s) = H(s) = 1 Js2 + k This result is in agreement with equation (2.59) for the case c = 0. n Example 2.3.3 As an example of using Laplace transforms to solve a homogeneous differential equation, consider the undamped single-degree-of-freedom system described by $ # x(0) = x0, x(0) = v0 x(t) + ω2nx(t) = 0, $ Solution Taking the Laplace transform of x + ω2n x = 0 for these nonzero initial conditions results in s2X(s) - sx0 - v0 + ω2nX(s) = 0 by direct application of the definition given in Appendix B and the linear nature of the Laplace transform. Algebraically solving this last expression for X(s) yields X1s2 = x0 + sv0 s2 + ω2n Sec. 2.4 151 base Excitation Using L - 1[X(s)] = x(t) and entries (6) and (5) of Table B.1 yields that the solution is x(t) = x0 cos ωnt + 0 ωn sin ωnt This is, of course, in total agreement with the solution obtained in Chapter 1. n This section presented three alternative methods for calculating the particular solution for a harmonically excited system. Each was shown to yield the same result. The concepts presented in these solution techniques are generalized and used for more complicated problems in later chapters. 2.4 BASE EXCITATION Often, machines, or parts of machines, are harmonically excited through elastic mountings, which may be modeled by springs and dashpots. For example, an automobile suspension system is excited harmonically by a road surface through a shock absorber, which may be modeled by a linear spring in parallel with a viscous damper. Other examples are the rubber motor mounts that separate an automobile engine from its frame or an airplane’s engine from its wing or tail section. Such systems can be modeled by considering the system to be excited by the motion of its support. This forms the base-excitation or support-motion problem modeled in Figure 2.13. Summing the relevant forces on the mass, m, Figure 2.13 yields (i.e., the iner$ tial force mx is equal to the sum of the two forces acting on m, and the gravitational force is balanced against the static deflection of the spring as before) $ # # mx + c(x - y) + k(x - y) = 0 (2.61) Here note that the spring deflects a distance (x - y) and the damper experiences # # a velocity of (x - y). For the base-excitation problem it is assumed that the base moves harmonically, that is, that y(t) = Y sin ωbt (2.62) where Y denotes the amplitude of the base motion and ωb represents the frequency of the base oscillation. Substitution of y(t) from equation (2.62) into the equation of motion given in (2.61) yields, after some rearrangement, x(t) .. x(t) mx(t) m m k c y(t) Base (a) . . k(x  y) c(x  y) (b) Figure 2.13 (a) Base-excitation problem models the motion of an object of mass m as being excited by a prescribed harmonic displacement acting through the spring and damper. (b) A free-body diagram of the base motion problem in (a). 152 Response to Harmonic Excitation $ # mx + cx + kx = cYωb cos ωbt + kY sin ωbt Chap. 2 (2.63) This can be thought of as a spring–mass–damper system with two harmonic inputs. The expression is very similar to the problem stated by equation (2.26) for the forced harmonic response of a damped system with F0 = cYωb and ω = ωb, except for the “extra” forcing term kY sin ωbt. The solution approach is to use the linearity of the equation of motion and realize that the particular solution of equation (2.63) will be the sum of the particular solution obtained by assuming an input force of cYωb cos ωbt, denoted by x(1) p , and the particular solution obtained by assuming an input force of kY sin ωbt, denoted by x(2) p . Calculating these particular solutions follows directly from the calculation made in Section 2.2. Dividing equation (2.63) by m and using the definitions of damping ratio and natural frequency yields $ # x + 2ζωnx + ω2nx = 2ζωnωbY cos ωbt + ω2nY sin ωbt (2.64) Thus substituting f0 = 2ζωnωbY into equation (2.36) yields that the particular solution xp(1) due to the cosine excitation is x(1) p = where 3 1 ω2n - ω2b 2 2 + 1 2ζωnωb 2 2 2ζωnωbY θ1 = tan-1 cos ( ωbt - θ1 ) 2ζωnωb (2.65) (2.66) ωn2 - ω2b To calculate xp(2), the method of undetermined coefficients is applied again with the harmonic input ω2nY sin ωbt. Following the procedures used to calculate equation (2.36) results in x(2) p = 2 1 ω2n - ω2b 2 2 + ω2nY 1 2ζωnωb 2 2 sin ( ωbt - θ1 ) (2.67) Note that equation (2.67) with ζ = 0 agrees with equation (2.22) for the undamped case, as it should. Here the particular solution is assumed to be of the form x(2) p = X sin 1ωbt - θ12. Here the angle θ1 is the same as given in equation (2.66) because the phase angle is independent of the excitation amplitude (i.e., ζ, ωn, and ωb have not changed). The phase difference between the two particular solutions is accounted for by using the sine and cosine solution. Because the arguments of the two particular solutions are the same (ωbt - θ1), they can be easily added using simple trigonometry. Sec. 2.4 153 base Excitation From the principle of linear superposition, the total particular solution is the sum of equations (2.65) and (2.67) (i.e., xp = xp(1) + x(2) p ). Adding solutions (2.65) and (2.67) yields xpt = ωnY c where 1 ω2n 1 2ζωb 2 2 ω2b 2 2 + 1 2ζωnωb2 2 ωn2 + - θ2 = tan-1 d 1>2 cos 1ωbt - θ1 - θ22 ωn 2ζωb 1>2 1 2ζr 2 2 d r 2 2 2 + 1 2ζr 2 2 (2.68) (2.69) It is convenient to denote the magnitude of the particular solution, xp(t), by X so that X = Yc 11 1 + (2.70) where the frequency ratio r = ωb >ωn. Dividing this last expression by the magnitude of base motion, Y, yields - 1>2 1 + 1 2ζr 2 2 X = c d Y 1 1 - r 2 2 2 + 1 2ζr 2 2 (2.71) which expresses the ratio of the maximum response magnitude to the input displacement magnitude. This ratio is called the displacement transmissibility and is used to describe how motion is transmitted from the base to the mass as a function of the frequency ratio ωb >ωn. This ratio is plotted in Figure 2.14. Note that near r = ωb >ωn = 1, or resonance, the maximum amount of base motion is transferred to displacement of the mass. Note from Figure 2.14 that for r 6 12 the transmissibility ratio is greater than 1, indicating that for these values of the system’s parameters (ωn) and base frequency (ωb), the motion of the mass is an amplification of the motion of the base. Notice also that for a given value of r, the value of the damping ratio ζ determines the level of amplification. Specifically, larger ζ yields smaller transmissibility ratios. For values r 7 12 the transmissibility ratio is always less than 1 and the motion of the mass will be of smaller amplitude than the amplitude of the exciting base motion. In this higher-frequency range, the effect of increasing damping is just the opposite of that in the low-frequency case. Increasing the damping actually increases the amplitude ratio in the higher-frequency range. However, the amplitude is always less than 1 for underdamped systems. The frequency range defined by r 7 12 forms the important concept of vibration isolation discussed in detail in Section 5.2. For a fixed amount of damping, say ζ = 0.01, the important aspect of base motion is that the mass experiences larger amplitude oscillations than the base excitation provides for r 6 12 and experiences smaller amplitude oscillations 154 Response to Harmonic Excitation Chap. 2   0.05 X –– Y 6 5   0.1 Displacement ratio 4 3   0.25 2   0.5   0.7 1 0 0.5   0.7   0.5   0.25   0.1   0.05 1 2 1.5 2 r Frequency ratio Figure 2.14 Displacement transmissibility as a function of the frequency ratio, illustrating how the dimensionless deflection X >Y varies as the frequency of the base motion increases for several different damping ratios. than the base excitation provides for r 7 12. Near resonance, most of the motion of the base is amplified into motion of the mass, causing it to have large amplitude oscillations. It is interesting to compare the transmissibility plot of Figure 2.14 and equation (2.71) for base excitation with the steady-state magnitude plot for harmonic excitation of the mass as given in Figure 2.9 and equation (2.40). First, note that the frequency ratio, r, is independent of the damping ratio by definition in both cases. However, the dependence of the peak value, rpeak, on the damping ratio is different in each case (recall the computation in Example 2.2.5 for the peak value). In particular, Figure 2.10 will be different for base excitation than for harmonic excitation of Sec. 2.4 155 base Excitation the mass directly. This difference is caused by the numerator having an additional term in the base-excitation problem. This term, 2ζr, comes from the load carried through the damper which is not present when the mass is excited directly, as in equation (2.40). Another quantity of interest in the base-excitation problem is the force transmitted to the mass as the result of a harmonic displacement of the base. The force transmitted to the mass is done so through the spring and damper. Hence, the force transmitted to the mass is the sum of the force in the spring and the force in the damper, or from the free-body diagram, Figure 2.13, # # F(t) = k(x - y) + c(x - y) (2.72) This force must balance the inertial force of the mass m; thus $ F(t) = -mx(t) (2.73) In the steady state, the solution for x is given by equation (2.68). Differentiating equation (2.68) twice and substituting into equation (2.73) yields F(t) = mω2bωnY c 1 ω2n - ω2b 2 2 + ( 2ζωnωb ) ωn2 + ( 2ζωb ) 2 Again using the frequency ratio r, this becomes d 2 1>2 cos ( ωb t - θ1 - θ2 ) F(t) = FT cos(ωbt - θ1 - θ2) (2.74) (2.75) where the magnitude of the transmitted force, FT, is given by FT = kYr 2 c 11 - r 2 2 2 + ( 2ζr ) 1 + ( 2ζr ) 2 d 2 1>2 (2.76) Equation (2.76) is used to define force transmissibility by forming the ratio 1>2 1 + ( 2ζr ) 2 FT = r2c d kY 1 1 - r 2 2 2 + ( 2ζr ) 2 (2.77) This force transmissibility ratio, FT >kY, expresses a dimensionless measure of how displacement in the base of amplitude Y results in a force magnitude applied to the mass. Note from equations (2.75) and (2.68) that the force transmitted to the mass is in phase with the displacement of the mass. Figure 2.15 illustrates the force transmissibility as a function of the frequency ratio for four values of the damping ratio. Note that unlike the displacement transmissibility, the force transmitted does not necessarily fall off for r 7 12. In fact, as the damping increases, the force transmitted increases dramatically for r 7 12. 156 Response to Harmonic Excitation Chap. 2 FT kY  0.01  0.1 Normalized force 10  1.0  0.2 1 0.1 0.01 r 1 0 2 2 3 Frequency ratio 4 Figure 2.15 The force transmitted to the mass as a function of the frequency ratio illustrating how the dimensionless force ratio varies as the frequency of the base motion increases for four different damping ratios (ζ = 0.01, 0.1, 0.2, and 1.0). Example 2.4.1 Consider the base-excitation problem with the following data: m = 100 kg, c = 30 kg>s, k = 2000 N>m, Y = 0.03 m, and ωb = 6 rad>s. Compute the magnitude of the transmissibility ratio and then the force transmissibility ratio. Solution First, define the usual vibration properties by dividing by the mass to get 2000 c 30 = 4.472 rad>s, ζ = = = 0.034 A 100 21mk 232 * 105 ωb 6 r = = = 1.342 ωn 4.472 ωn = Then use equation (2.71) to compute the magnitude of the particular solution: 1 + ( 2ζr ) 2 X = c d Y 1 1 - r 2 2 2 + ( 2ζr ) 2 1 >2 = c 11 The force transmissibility ratio becomes - (1.342)2 2 2 + (2 # 0.034 # 1.342)2 1 + (2 # 0.034 # 1.342)2 d = 0.557 FT X = r2 = (1.342)2 (0.557) = 1.003 kY Y Note that if the damping value is changed to c = 300 kg>s, the force transmissibility ratio is 1.203 and the transmissibility is 0.669. So increasing the damping increases both the force and the displacement transmissibility. n 157 base Excitation A comparison between force transmissibility (force applied to the mass normalized by the magnitude of the displacement of the base) and displacement transmissibility (displacement of the mass normalized by the magnitude of the displacement of the base) is given in Figure 2.16. The formulas for transmissibility of force and displacement are very useful in the design of systems to provide protection from unwanted vibration. This is discussed in detail in Section 5.2 on vibration isolation, where the transmissibility ratio is derived for a fixed base and compared to the development here in Window 5.1. The following example illustrates some practical values of transmissibility for the base-excitation problem. 100 10 Log of transmissibility Sec. 2.4 1 0.1 0.01 0 1 2 r Frequency ratio 3 4 Figure 2.16 A comparison between force transmissibility (dashed line) and displacement transmissibility (solid line) for a damping ratio of ζ = 0.05 on a semilog plot using equations (2.71) and (2.77). Example 2.4.2 A common example of base motion is the single-degree-of-freedom model of an automobile driving over a road or an airplane taxiing over a runway, indicated in Figure 2.17. The road (or runway) surface is approximated as sinusoidal in cross section providing a base motion displacement of y(t) = (0.01 m) sin ωbt 158 Response to Harmonic Excitation where ωb = v(km>h) a Chap. 2 1 hour 2π rad ba ba b = 0.2909v rad>s 0.006 km 3600 s cycle where denotes the vehicle’s velocity in km>h. Thus the vehicle’s speed determines the frequency of the base motion. Determine the effect of speed on the amplitude of displacement of the automobile as well as the effect of the value of the car’s mass. Assume that the suspension system provides an equivalent stiffness of 4 * 104 N>m and damping of 20 * 102 N # s>m. x (t) m  mass of car Velocity of car Suspension system k c Neglected unsprung mass 0.02 m 6m Road surface Figure 2.17 A simple model of a vehicle traveling with constant velocity on a wavy surface that is approximated as a sinusoid. Solution First, to determine the effect of speed on the amplitude of the vehicle’s motion, note that from the previous calculations, ωb, and hence r, vary linearly with the car’s velocity. Hence the deflection ratio versus velocity curve will be much like the curve of Figure 2.14. Some sample values can be calculated from equation (2.70). At 20 km>h, ωb = 5.818. If the car is small or a sports car, its mass might be around 1007 kg, so that the natural frequency is ωn = 4 * 104 N>m C 1007 kg = 6.303 rad>s ( ≈1 Hz) so that r = 5.818>6.303 = 0.923 and ζ = 2000 N # s>m c = = 0.158 21km 23(4 * 104 N>m)(1007 kg) Equation (2.70) then yields that the deflection experienced by the car will be X = (0.01 m) E 31 - 3 2(0.158)(0.923) 4 2 2 (0.923)2 4 2 + 3 2(0.158)(0.923) 4 1 + = 0.0319 Sec. 2.4 159 base Excitation This means that a 1-cm bump in the road is transmitted into a 3.2-cm “bump” experienced by the chassis and subsequently transmitted to the occupants. Hence the suspension system amplifies the rough road bumps in this circumstance and is not desirable. Table 2.1 lists several different values of the vehicle displacement for two different vehicles traveling at four different speeds over the same 1-cm bump. Car 1 with frequency ratio r1 is a 1007-kg sports car, while car 2 is a 1585-kg sedan with frequency ratio r2. The same suspension system was used on both cars to illustrate the need to design suspension systems based on a given vehicle’s specifications (see Chapter 5). Note that with higher speed, negligible vibration is experienced by the occupants of the car. Also, notice that the suspension system parameters chosen (k and c) work better in general for the larger car except at very low speeds. TABLE 2.1 CoMPARISon oF CAR VEloCIty, FREquEnCy, AnD DISPlACEMEnt FoR two DIFFEREnt CARS Speed (km>h) ωb r1 r2 x1 (cm) x2 (cm) 20 80 100 150 5.817 23.271 29.088 43.633 0.923 3.692 4.615 6.923 1.158 4.632 5.79 8.686 3.19 0.12 0.09 0.05 2.32 0.07 0.05 0.03 n Example 2.4.3 A large rotating machine causes the floor of a factory to oscillate sinusoidally. A punch press is to be mounted on the same floor (Figure 2.18). The displacement of the floor at the point where the punch press is to be mounted is measured to be y(t) = 0.1 sin ωbt (cm). Using the base support model of this section, calculate the maximum force transmitted to the punch press at resonance if the press is mounted on a rubber fitting of stiffness k = 40,000 N>m; damping, c = 900 N # s>m; and mass, m = 3000 kg. Excitation source: Rotating machine Punch press Support: c, k Floor Base motion y(t) Figure 2.18 A model of a machine causing support motion. Solution The force transmitted to the punch press is given by equation (2.77). At resonance, r = 1, so that equation (2.77) becomes or 1 + ( 2ζ ) 2 1>2 FT = c d kY ( 2ζ ) 2 FT = kY 1 1 + 4ζ2 2 1>2 2ζ 160 Response to Harmonic Excitation Chap. 2 From the definition of ζ and the values given previously for m, c, and k, ζ = c 900 ≅ 0.04 = 21km 2 3 (40,000)(3000) 4 1>2 (40,000 N>m)(0.001 m) kY 1 1 + 4ζ2 2 1>2 = 3 1 + 4(0.04)2 4 1>2 2ζ 2(0.04) From the measured excitation Y = 0.001 m, so that FT = = 501.6 N n The analysis presented here for base motion is very useful in design. This forms the topic of Chapter 5, which includes as Section 5.2 a more detailed analysis of base motion in the context of the vibration isolation problem for both fixed-base and moving-base models. Section 5.2 also includes a discussion of shock isolation. Some may prefer to jump to Section 5.2 at this point to examine how the concepts of transmissibility are applied to the base isolation problem. 2.5 ROTATING UNBALANCE A common source of troublesome vibration is rotating machinery. Many machines and devices have rotating components, usually driven by electric motors. Small irregularities in the distribution of the mass in the rotating component can cause substantial vibration. This is called a rotating unbalance. A schematic of such a rotating unbalance of mass, m0, a distance e from the center of rotation is given in Figure 2.19. Guide Guide x(t) Machine of total mass m m0 e k Figure 2.19 r t Rubber floor mounting modeled as a spring and a damper c Friction-free surface A model of a machine causing support motion. Sec. 2.5 161 Rotating unbalance x(t)  xr(t) Fr x(t) m0 e rt m  m0 Fr . cx kx (a) (b) Figure 2.20 A free-body diagram of the unbalance (a) and the machine (b). The frequency of rotation of the machine is denoted by ωr. Summing forces in the vertical direction (x) from the free-body diagram of the out-of-balance mass given in Figure 2.20(a) yields $ $ m0(x + xr) = -Fr (2.78) Summing forces from the free-body diagram of the machine given in Figure 2.20(b) yields $ # (m - m0)x = Fr - cx - kx (2.79) Combining equations (2.78) and (2.79) yields $ $ # mx + m0xr + cx + kx = 0 (2.80) The forces in the horizontal direction are canceled by the guides and are not considered here. Assuming that the machine rotates with a constant frequency, ωr, the x component of the motion of the mass, m0, is xr = e sin ωr t, so that $ xr = -eω2r sin ωr t (2.81) Substitution of equation (2.81) into (2.80) yields $ # mx + cx + kx = m0eω2r sin ωrt (2.82) after rearranging the terms. Equation (2.82) is similar to equation (2.26) with F0 = m0eω2r , with the exception of the phase shift of the forcing function (i.e., sin ωr t instead of cos ωt). The sine excitation is discussed as the second particular solution 162 Response to Harmonic Excitation Chap. 2 in the previous section and the solution is given in equation (2.67). The solution procedure is the same and results in a particular solution of the form Let r = ωr >ωn, as before, to get xp(t) = X sin (ωr t - θ) X = and m0e m (2.83) 3 1 1 - r 2 2 2 + ( 2ζr ) 2 r2 θ = tan-1 (2.84) 2ζr (2.85) 1 - r2 These last two expressions yield the magnitude and phase of the motion of the mass, m, due to the rotating unbalance of mass m0. Note that the mass m in equation (2.84) is the total mass of the machine and includes the unbalance mass m0. The magnitude of the steady-state displacement, X, as a function of the rotating speed (frequency) is examined by plotting the dimensionless displacement magnitude mX>m0e versus r, as indicated in Figure 2.21 for various values of the damping ratio ζ. Note that equation (2.84) is similar to the magnitude analyzed in Example 2.2.4. From the form of the denominator, which is identical to that of Example 2.2.4, it is observed that the maximum deflection is less than or equal to 1 for any system with ζ 7 1. This indicates that the increase in amplification of the amplitude caused by the unbalance can be eliminated by increasing the damping in mX m0e 5 Normalized magnitude   0.1 4 3   0.25 2   0.707 1   1.0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5  r  r n Frequency ratio Figure 2.21 Magnitude of the dimensionless displacement versus frequency ratio caused by a rotating unbalance of mass m0 and radius e. Sec. 2.5 163 Rotating unbalance the system. However, large damping is not always practical. Note from Figure 2.21 that the magnitude of the dimensionless displacement approaches unity if r is large. Hence, if the running frequency ωr is such that r W 1 the effect of the unbalance is limited. For large values of r, all the magnitude curves for each value of ζ approach unity, so that the choice of damping coefficient for large r is not important. These results can be obtained from examining the plots of Figure 2.21 or from investigating the limit of mX>m0e as r goes to infinity. These observations have important implications in the design of rotating machines. The rotating unbalanced model can also be used to explain the behavior of an automobile with an out-of-balance wheel and tire. Here ωr is determined by the speed of the car and e by the diameter of the wheel. The deflection xp can be felt through the steering mechanism as shaking of the steering wheel. This usually only happens at a certain speed (near r = 1). As the driver increases or decreases speed, the shaking effect in the steering wheel reduces. This change in speed is equivalent to operating conditions on either side of the peak in Figure 2.21. Example 2.5.1 Consider a machine with rotating unbalance as described in Figure 2.19. At resonance, the maximum deflection is measured to be 0.1 m. From a free decay of the system, the damping ratio is estimated to be ζ = 0.05. From manufacturing data, the out-ofbalance mass, m0, is estimated to be 10%. Estimate the radius e and hence the approximate location of the unbalanced mass. Also determine how much mass should be added (uniformly) to the system to reduce the deflection at resonance to 0.01 m. Solution At resonance, r = 1, so that mX 1 1 = = m0e 2ζ 2(0.05) Hence (10) (0.1 m) e = 1 1 = 10 = 2ζ 0.1 so that e = 0.1 m. Again at resonance m X a b = 10 m0 0.1 m If it is desired to change m, say by Δm, so that X = 0.01 m, the foregoing resonance expression becomes m + Δm 0.01 a b = 10 or m0 0.1 m + Δm = 100 (0.1)m which implies that Δm = 9m. Thus the total mass must be increased by a factor of 9 in order to reduce the deflection to a centimeter. n 164 Response to Harmonic Excitation Chap. 2 Example 2.5.2 Rotating unbalance is also important in rotorcraft such as helicopters and prop planes. The tail rotor of a helicopter (the small rotor rotating in a vertical plane at the back of a helicopter used to provide yaw control and torque balance) as sketched in Figure 2.22 can be modeled as a rotating-unbalance problem discussed in this section with stiffness k = 1 * 105 N>m (provided by the tail section in the vertical direction) and mass of 20 kg. The tail section providing the vertical stiffness has a mass of 60 kg. Suppose that a 500-g mass is stuck on one of the blades at a distance of 15 cm from the axis of rotation. Calculate the magnitude of the deflection of the tail section of the helicopter as the tail rotor rotates at 1500 rpm. Assume a damping ratio of 0.01. At what rotor speed is the deflection at maximum? Calculate the maximum deflection. 0.5 kg r t 60 kg 20 kg Figure 2.22 A schematic of a helicopter tail section illustrating a tail rotor. The tail rotor provides a “counterclockwise” thrust (when looking at the top of the helicopter) to counteract the “clockwise” thrust created by the main rotor, which provides lift and horizontal motion. An out-of-balance rotor can cause damaging vibrations and limit the helicopter’s performance. Solution The rotor system is modeled as a machine of mass 20.5 kg attached to a spring, as indicated in Figure 2.23. Here only the vibration of the tail section in the vertical direction is modeled and the helicopter body is modeled as ground. The spring m0 r t m m0 ms , k X m ms , k (a) (b) Figure 2.23 (a) The vertical vibration model of a tail section modeled as a spring consisting of a long, slender bar with machine mounted on it with a rotational unbalance. (b) This sketch is the equivalent spring–mass model used for unbalance problems (note that to be consistent with Figure 2.19, m includes m0). Sec. 2.5 165 Rotating unbalance used to represent the tail section has significant mass, so equation (1.76) of Section 1.5 for a heavy beam is used to find the equivalent mass of the system. Using the equivalent mass concept yields that the natural frequency is ωn = 105 N>m k = = 53.727 rad>s 33 33 m + ms 20.5 + 60 kg H 140 S 140 The frequency of rotation in rad>s is ωr = 1500 rpm = 1500 rev min 2π rad = 157 rad>s min 60 s rev Hence, the frequency ratio, r, becomes r = 157 rad>s ωr = = 2.92 ωn 53.727 rad>s With r = 2.92 rad>s and ζ = 0.01, equation (2.84) yields that the magnitude of oscillation of the tail rotor is X = = m0e r2 m 2(1 - r 2)2 + (2ζr)2 (0.5 kg)(0.15 m) 34.64 kg 2 3 1 - (2.92)2 4 2 - 3 2(0.01)(2.92) 4 2 (2.92)2 = 0.002 m Here the equivalent mass is meq = m + ms = 34.64 kg. The maximum deflection occurs at about r = 1 or ωr = ωn = 53.72 rad>s = 53.72 rad revs 60 s = 513.1 rpm s 2π rad min In this case, the (maximum) deflection becomes X = (0.5 kg)(0.15) 34.34 kg 1 = 0.108 m = 10.8 cm 2(0.01) which represents a large unacceptable deflection of the rotor. Thus the tail rotor should not be allowed to rotate at 513.1 rpm. n More on the special nature of rotating systems is discussed in Section 5.5. Additional discussion of vibration problems associated with rotating machinery is given in Section 5.7 on critical speeds. Some treatments include the discussion of critical speeds in rotating shafts immediately following the discussion of unbalance, and it is possible to skip to Section 5.7 before continuing. Vibration problems associated with rotor dynamics are both important and vast enough to study as a separate course. 166 Response to Harmonic Excitation Chap. 2 2.6 MEASUREMENT DEVICES An important application of the forced-harmonic-vibration analysis and baseexcitation problem presented in Sections 2.2 and 2.4, respectively, is in the design of devices used to measure vibration. A device that changes mechanical motion into a voltage (or vice versa) is called a transducer. Several transducers are sketched in Figures 2.24 to 2.26. Each of these devices changes mechanical vibration into a voltage proportional to acceleration. Referring to the accelerometer of Figure 2.24, a balance of forces on the seismic mass m yields $ # # mx = -c(x - y) - k(x - y) (2.86) Here it is assumed that the base that is mounted to the structure being measured undergoes a motion of y = Y cos ωbt (i.e., that the structure being measured is undergoing simple harmonic motion). The motion of the accelerometer mass relative to the base, denoted by z(t), is defined by (2.87) z(t) = x(t) - y(t) x(t) k .. mx(t) Casing Voltage m m y(t) c . . k(xy) c(xy) y(t)  motion of structure y(t) Voltage-generating strain gauges Figure 2.24 A schematic of an accelerometer mounted on a structure. The insert indicates the relevant forces acting on the mass m. The force k(x - y) is actually parallel to the damping force because they are both connected to ground. x(t) m Elastic beam (k) Mounted to vibrating structure Figure 2.25 A schematic of a seismic accelerometer made of a small beam. Sec. 2.6 167 Measurement Devices k x(t) m Piezoelectric crystal Voltage y(t) Mounted to structure Figure 2.26 A schematic of a piezoelectric accelerometer and a photograph of a commercially available version. Equation (2.86) can then be written in terms of the relative displacement z(t). Hence equation (2.86) becomes the familiar expression $ # mz + cz + kz = mω2bY cos ωb t (2.88) This expression has exactly the same form as equation (2.26). Thus the solution in steady state will be of the same form as given by equation (2.36) or z(t) = ω2bY 2(ω2n - ω2b)2 + (2ζωnωb)2 cos c ωbt + a -tan -1 2ζωnωb ω2n - ω2b bd (2.89) The difference between equations (2.36) and (2.89) is that the latter is for the relative displacement (z) and the former is for the absolute displacement (x). Further manipulation of the magnitude of equation (2.89) yields Z r2 = Y 2(1 - r 2)2 + (2ζr)2 for the amplitude ratio as a function of the frequency ratio r = ωb >ωn, and θ = tan-1 2ζr 1 - r2 (2.90) (2.91) for the phase shift. Consider the plot of the magnitude of Z>Y versus the frequency ratio as given in Figure 2.27. Note that for larger values of r (i.e., for r Ú 3) the magnitude ratio approaches unity, so that Z>Y = 1 or Z = Y and the relative displacement and the displacement of the base have the same magnitude. Hence the accelerometer of Figure 2.24 can be used to measure harmonic base displacement if the frequency of the base displacement is at least three times the accelerometer’s natural frequency. 168 Response to Harmonic Excitation Z Y Relative displacement ratio 2.5 Chap. 2 Accelerometer region 2 Seismometer region   0.25 1.5 1   0.5 0.5 0 0 1 2 3 4 r Frequency ratio Figure 2.27 The magnitude versus frequency of the relative displacement for a transducer used to measure acceleration and for seismic measurements. Next, consider the equation of the system in Figure 2.24 for the case when r is small. Factoring ω2n out of the denominator, equation (2.89) can be written as ω2nz(t) = 1 2 2 2(1 - r ) + (2ζr) 2 ω2bY cos (ωbt - θ) (2.92) $ Since y = Y cos (ωbt - θ), the last term is recognized to be -y(t) so that ω2nz(t) = -1 2 2 2(1 - r ) + (2ζr) 2 $ y(t) (2.93) This expression illustrates that, for small values of r, the quantity ω2nz(t) is propor$ tional to the base acceleration, y(t), since lim rS0 311 - r 2 1 2 2 + (2ζr)2 = 1 (2.94) In practice, this coefficient is taken as close to 1 for any value of r 6 0.5. This indicates that, for these frequencies of base motion, the relative position z(t) is proportional to the base acceleration. The effect of the accelerometer internal damping, ζ, in the constant of proportionality between the relative displacement and the base acceleration is illustrated in Figure 2.28, which consists of a plot of this constant versus the frequency ratio for a variety of values of ζ, for values of r 6 1. Note from the figure that the curve corresponding to ζ = 0.7 is closest to being constant at unity over the largest range of r 6 1. For this curve, the magnitude is relatively flat for values of r between zero and about 0.2. In fact, within this region, the curve varies 169 Measurement Devices 1.1   0.6 1 Coefficient of ÿ Sec. 2.6 0.9   0.7 0.8 1 0.7 0.6 0.5 0 0.2 0.4 0.6 0.8 1 Frequency ratio Figure 2.28 The effect of damping on the constant of proportionality between base acceleration and the relative displacement (voltage) for an accelerometer. from one by less than one percent. This defines the useful range of operation for the accelerometer: 0 6 ωb 6 0.2 ωn (2.95) where ωn is the natural frequency of the device. Multiplying this inequality by the device frequency yields 0 6 ωb 6 0.2ωn or 0 6 fm 6 0.2fn (2.96) where fm is the frequency to be measured by the accelerometer in hertz. For the mechanical accelerometer of Figure 2.24, the device frequency may be on the order of 100 Hz. Thus inequality (2.96) indicates that the highest frequency that can be effectively measured by the device would be 20 Hz (0.2 * 100). Many structures and machines vibrate at frequencies larger than 20 Hz. The piezoelectric accelerometer design indicated in Figure 2.26 provides a device with a natural frequency of about 8 * 104 Hz. In this case, the inequality predicts that vibration with frequency content up to 16,000 Hz can be measured. Vibration measurement is discussed in more detail in Chapter 7, where practical problems, such as phase and amplitude distribution of signals measured with accelerometers, are discussed. Example 2.6.1 This example illustrates how an independent measurement of acceleration can provide a measurement of a transducer’s mechanical properties. An accelerometer is used to measure the oscillation of an airplane wing caused by the plane’s engine operating at 6000 rpm (628 rad>s). At this engine speed the wing is known, from other measurements, to experience 1.0-g acceleration. The accelerometer measures an acceleration of 10 m>s2. If the accelerometer has a 0.01-kg moving mass and a damped natural frequency of 100 Hz (628 rad>s), the difference between the measured and the known 170 Response to Harmonic Excitation Chap. 2 acceleration is used to calculate the damping and stiffness parameters associated with the accelerometer. Solution From equation (2.93), the amplitude of the measured values of acceleration $ 0 ω2nz(t) 0 is related to the actual values of acceleration 0 y(t) 0 by 0 ω2nz(t) 0 10 m>s2 1 = = = 1.02 $ 0 y(t) 0 9.8 m>s2 2(1 - r 2)2 + (2ζr)2 Rewriting this expression yields one equation in ζ and r: (1 - r 2)2 + (2ζr)2 = 0.96 A second expression in ζ and r can be obtained from the definition of the damped natural frequency: 628 rad>s ωb ωb 1 1 = = r = = 1 2 ωd ωn 21 - ζ2 628 rad>s 21 - ζ Thus r = 21 - ζ2, providing a second equation in ζ and r. This can be manipulated to yield ζ2 = (1 - r 2), which when substituted with the preceding expression for r and ζ yields ζ4 + 4ζ2(1 - ζ2) = 0.96 This is a quadratic equation in ζ2: 3ζ4 - 4ζ2 + 0.96 = 0 This quadratic expression yields the two roots ζ = 0.56, 1.01. Using ζ = 0.56, the damping constant is ( 21 - ζ2 = 0.83, ωn = ωd > 21 - ζ2 = 758.0 rad>s ) c = 2mωnζ = 2(0.01)(758.0)(0.56) = 8.49 N # s>m Similarly, the stiffness in the accelerometer is k = mω2n = (0.01)(758.0)2 = 5745.6 N>m n 2.7 OTHER FORMS OF DAMPING The damping used in previous sections has been treated as linear-viscous damping, with the exception of the treatment of Coulomb damping in Section 1.10. In this section, the discussion of Coulomb damping is continued and other forms of damping are introduced. Because damping is both difficult to model mathematically and difficult to measure, choosing the correct form of damping is not an easy task. Hence damping is often approximated as a linear dependence on velocity, as done in the previous sections. Other models, though not mathematically convenient, may provide a more accurate description of the damping in a vibrating system. Coulomb, Sec. 2.7 171 other Forms of Damping or friction, damping is one example of such a form. A number of other mathematical forms of damping exist. In this section, these damping forms are all treated in the forced-response case by examining an equivalent linear system based on the energy dissipated during vibration. In Section 2.9, these systems are numerically integrated in their nonlinear form and compared to the equivalent linear response discussed here. First, consider the response of a system with Coulomb damping, introduced in Section 1.10, to a harmonic driving force. Recall the systems of Figures 1.43 and 1.44 described in the free-response case by equation (1.101). The equation of motion in the forced-response case becomes $ $ mx + μ mg sgn(x) + kx = F0 sin ωt (2.97) Rather than solving this equation directly, one can approximate the solution of equation (2.97) with the solution of a viscously damped system that dissipates an equivalent amount of energy per cycle. This is a reasonable assumption if the magnitude of the applied force is much larger than the Coulomb force (F0 W μmg). This approximation is accomplished by again assuming that the steady-state response will be of the form xss(t) = X sin ωt (2.98) The energy dissipated, ΔE, in a viscously damped system per cycle with viscousdamping coefficient c is given by ΔE = C Fd dx = L0 2π>ω # dx cx dt = dt L0 2π>ω # cx2dt (2.99) # At steady state, x = X sin ωt, x = ωX cos ωt, and equation (2.99) becomes ΔE = c L0 2π>ω (ω2X2 cos 2ωt)dt = πcωX2 (2.100) This is the energy dissipated per cycle by a viscous damper. On the other hand, the energy dissipated by the Coulomb friction on a horizontal surface per cycle is ΔE = μ mg L0 2π>ω 3 sgn(x# )x# 4 dt (2.101) Substitution of the steady-state velocity into this expression and splitting the inte# grations up into segments corresponding to the sign change in x yields ΔE = μ mgX a L0 π>2 3π>2 cos u du - Lπ>2 2π cos u du + L3π>2 cos u du b (2.102) 172 Response to Harmonic Excitation Chap. 2 where u = ωt and du = ωdt. Completing the integration yields that the energy dissipated by Coulomb friction is ΔE = 4μ mg X (2.103) To create a viscously damped system of equivalent energy loss, the energy-loss expression for viscous damping of equation (2.100) is equated to the energy loss associated with Coulomb friction, given by equation (2.103) to yield πceqωX2 = 4μ mg X (2.104) where ceq denotes the equivalent viscous-damping coefficient. Solving for ceq yields ceq = 4μ mg πωX (2.105) In terms of an equivalent damping ratio, ζeq, equation (2.105) must also equal 2ζeqωnm, so that 2μg πωnωX ζeq = (2.106) Thus the viscously damped system described by $ # x + 2ζeqωnx + ω2nx = f0 sin ωt (2.107) will dissipate as much energy as does the Coulomb system described by equation (2.97). Considering (2.107) as an approximation of equation (2.97), the approximate magnitude and phase of the steady-state response of equation (2.97) can be calculated. Substitution of the equivalent viscous-damping ratio given in equation (2.106) into the magnitude of equation (2.40) yields the result that the magnitude X of the steady-state response is X = F0 >k = 2 ( 1 - r 2 ) 2 + ( 2ζeqr ) 2 3( 1 - r Solving this expression for the amplitude X yields X = F0 >k ) + ( 4μ mg>πkX ) 24 1>2 2 2 F0 21 - (4μ mg>πF0)2 k 0 (1 - r 2) 0 (2.108) (2.109) with phase shift given by equation (2.40) as θ = tan-1 2ζeqr 1 - r 2 = tan-1 4μ mg πkX(1 - r 2) (2.110) Sec. 2.7 173 other Forms of Damping The expression for the phase can be further examined by substituting for the value of X from equation (2.109). This yields θ = tan-1 {4μ mg πF0 21 - (4μ mg>πF0)2 (2.111) where the { originates from the absolute value in equation (2.109). Thus θ is positive if r 6 1 and negative if r 7 1. Also note from equation (2.111) that θ is constant for a given F0 and μ and is independent of the driving frequency. Several differences are apparent in the behavior of the approximate phase and amplitude of the response with Coulomb friction compared with that of viscous friction. First, at resonance, r = 1, the magnitude in equation (2.109) becomes infinite, unlike the viscously damped case. Second, the phase is discontinuous at resonance, rather than passing through 90° as in the viscous case. Note also from equation (2.111) that the approximation is good only if the argument in the radical of (2.109) is positive, that is, if 4μ mg 6 πF0 (2.112) This confirms and quantifies the physical statement made at the outset (i.e., that the applied force must be larger in magnitude than the sliding friction force in order to overcome the friction to provide motion). Example 2.7.1 Consider a spring–mass system with sliding friction described by equation (2.97) with stiffness k = 1.5 * 104 N>m, driving harmonically a 10-kg mass by a force of 90 N at 25 Hz. Calculate the approximate amplitude of steady-state motion assuming that both the mass and the surface are steel (unlubricated). Solution First, look up the coefficient of friction in Table 1.5, which is μ = 0.3. Then from inequality (2.112), 4μ mg = 4(0.3)(10 kg)(9.8 m>s2) = 117.6 N 6 (90 N)(3.1415) = 282.74 = πF0 so that the approximation developed previously for the steady-state-response amplitude is valid. Converting 25 Hz to 157 rad>s and using equation (2.109) then yields X = 21 - ( 117.6 N>282.74 N ) 2 90 N = 3.53 * 10-4 m 1.5 * 104 N>m 0 1 - ( 2.467 * 104 >1.5 * 103 ) 0 Thus the amplitude of oscillation will be less than 1 mm. n Several other forms of damping are available for modeling a particular mechanical device or structure in addition to viscous and Coulomb damping. It is common to study damping mechanisms by examining the energy dissipated per cycle under a harmonic loading. Often, force versus displacement curves, or stress versus strain curves, 174 Response to Harmonic Excitation Chap. 2 are used to measure the energy lost and hence determine a measure of the damping in the system. The energy lost per cycle, given in equation (2.100) to be πcωX2, is used to define the specific damping capacity as the energy lost per cycle divided by the peak potential energy, ΔE >U. A more commonly used quantity is the energy lost per radian divided by the peak-potential, or strain, energy Umax. This is defined to be the loss factor or loss coefficient, denoted by η, and given by η = ΔE 2πUmax (2.113) where Umax is defined as the potential energy at maximum displacement of X (or strain energy). The loss factor is related to the damping ratio of a viscously damped system at resonance. To see this, substitute the value for ΔE from (2.100) into (2.113) to get η = 1 2π 1 12 kX2 2 πcωX2 (2.114) At resonance, ω = ωn = 1k>m so that (2.113) becomes η = c = 2ζ 1km (2.115) Hence, at resonance, the loss factor is twice the damping ratio. Next, consider a force-displacement curve for a system with viscous damping. The force required to displace the mass is that force required to overcome the spring and damper forces, or # F = kx + cx (2.116) At steady state, as given by equation (2.98), this becomes F = kx + cX ω cos ωt (2.117) Using a trigonometric identity (sin2ϕ + cos2ϕ = 1) on the cos ωt term yields = kx { cω 3X2 - (X sin ωt)2 4 1>2 F = kx { cωX(1 - sin2 ωt)1>2 (2.118) = kx { cω2X2 - x2 Squaring this expression yields, upon rearrangement, F 2 + (c 2ω2 + k 2)x2 - (2k)xF - c 2ω2X2 = 0 (2.119) which can be recognized as the general equation for an ellipse (c2ω2 7 0) rotated about the origin in the F-x plane (see a precalculus text). This is plotted in Figure 2.29. Sec. 2.7 175 other Forms of Damping . F(t)  cx(t)  kx(t) 100 50 x(t) –0.06 –0.04 –0.02 0 0.02 0.04 0.06 –50 –100 Figure 2.29 A plot of force versus displacement defining the hysteresis loop for a viscously damped system. The ellipse in the Figure 2.29 is called a hysteresis loop, and the area enclosed is the energy lost per cycle as calculated in equation (2.100) and is equal to πcωX2. Note that if c = 0, the ellipse of Figure 2.29 collapses to the straight line of slope k indicated by the dashed line in the figure. Materials are often tested by measuring stress (force) and strain (displacement) under carefully controlled steady-state harmonic loading. Many materials exhibit internal friction between various planes of material as the material is deformed. Such tests produce hysteresis loops of the form shown in Figure 2.30. Note that, for Stress Loading Strain Unloading Figure 2.30 An experimental stress– strain plot for one cycle of harmonically loaded material for steady state illustrating a hysteresis loop associated with internal damping. 176 Response to Harmonic Excitation Chap. 2 increasing strain (loading), the path is different than for decreasing strain (unloading). This type of energy dissipation is called hysteretic damping, solid damping, or structural damping. The area enclosed by the hysteresis loop is again equal to the energy loss. If the experiment is repeated for a number of different frequencies at constant amplitude, it is found that the area is independent of frequency and proportional to the square of the amplitude of vibration and stiffness: ΔE = πkβX2 (2.120) where k is the stiffness, X is the amplitude of vibration, and β is defined as the hysteretic damping constant. Note that some texts formulate this equation differently by defining h = kβ to be the hysteretic damping constant. Next, apply the equivalent viscous-damping concept used for Coulomb damping. If this concept is applied here, equating the energy dissipated by a viscously damped system to that of a hysteretic system is equivalent to finding the ellipse of Figure 2.29 that has the same area as the loop of Figure 2.20. Thus equating (2.120) with the energy calculated in (2.100) for the viscously damped system yields πceqωX2 = πkβX2 (2.121) Solving this expression yields that the viscously damped system dissipating the same amount of energy per cycle as the hysteretic system will have the equivalent damping constant given by ceq = kβ ω (2.122) where β is determined experimentally from the hysteresis loop. The approximate steady-state response of a system with hysteretic damping can be determined from substitution of this equivalent damping expression into the equation of motion to yield βk # $ x + kx = F0 cos ωt mx + ω (2.123) In this case, the steady-state response is approximated by assuming the response is of the form xss(t) = X cos (ωt - ϕ). Following the procedures of Section 2.2, the magnitude of the response, X, is given by equation (2.40) to be X = F0 >k where ζeq is now ceq > ( 21km ) and r = ω >ωn. Substituting for ζeq yields 2(1 - r 2)2 + (2ζeqr)2 X = F0 >k 2(1 - r 2)2 + β2 (2.124) (2.125) Sec. 2.7 177 other Forms of Damping Similarly, the phase becomes ϕ = tan-1 β (2.126) 1 - r2 These are plotted in Figure 2.31 for several values of β. Compared to a viscously damped system, the hysteretic system’s magnitude obtains a maximum of F0 >βk. This is obtained by setting r = 1 in equation (2.125) so that the maximum value is obtained at the resonant frequency rather than below it, as is the case for viscous damping. Examination of the phase shift shows that the response of a hysteretic system is never in phase with the applied force, which is not true for viscous damping. In Section 2.3, the complex exponential was used to represent a harmonic input. Using equation (2.48), the equivalent hysteretic system can be written as βk # $ mx + x + kx = F0e jωt ω (2.127) Substitution of the assumed form of the solution given by x(t) = Xejωt for just the velocity term yields $ mx + k(1 + jβ)x = F0e jωt (2.128) 3.0 Phase (rad) 5   0.2 Normalized displacement  1 2.5 Xk F0 4 2.0 1.5   0.5 1.0   0.2 0.5 0 3 1   0.5 3 2 4 Frequency ratio, r 2 1  1 0 1 2 3 Frequency ratio, r Figure 2.31 Steady-state magnitude and phase versus frequency ratio for a system with hysteretic damping coefficient β approximated by a system with viscous damping. 4 178 Response to Harmonic Excitation Chap. 2 This gives rise to the notion of complex stiffness or complex modulus. The damped problem is represented in equation (2.128) as an undamped problem with complex stiffness coefficient k(1 + jβ). This approach is very popular in the material engineering literature on damping. Example 2.7.2 An experiment is performed on a hysteretic system with known spring stiffness of k = 4 * 104 N>m. The system is driven at resonance, the area of the hysteresis loop is measured to be ΔE = 30 N # m, and the amplitude, X, is measured to be X = 0.02 m. Calculate the magnitude of the driving force and the hysteretic damping constant. Solution At resonance, equation (2.125) yields X = F0 kβ or kβ = F0 >X. The area enclosed by the hysteresis loop is equal to πkβX2, so that 30 N # m = πkβX2 = π F0 2 X X and hence F0 = 30 N # m 30 N # m = = 477.5 N πX π(0.02 m) From the resonance expression, β = F0 477.5 = 0.60 = Xk (0.02 m)(4 * 104 N>m) which is the hysteretic damping constant calculated based on the principle of equivalent viscous damping. n Several other useful models of damping mechanisms can be analyzed by using the equivalent viscous-damping approach. For instance, if an object vibrates in air (or a fluid), it often experiences a force proportional to the square of the velocity (Blevins, 1977). The equation of motion for such a vibration is $ # # mx + α sgn(x)x2 + kx = F0 cos ωt (2.129) The damping force is CρA # # # # Fd = α sgn(x)x2 = sgn(x)x2 2 which opposes the direction of motion, similar to Coulomb friction, and depends # on the square of the velocity, x; the drag coefficient of the mass, C; the density of the fluid, ρ; and the cross-sectional area, A, of the mass. This type of damping is Sec. 2.7 179 other Forms of Damping referred to as air damping, quadratic damping, or velocity-squared damping. As in the Coulomb friction case, this is a nonlinear equation that does not have a convenient closed-form solution to analyze. While it can be solved numerically (see Section 2.8), an approximation to the behavior of the solution during steady-state harmonic excitation can be made by the equivalent viscous-damping method. Assuming a solution of the form x = X sin ωt and computing the energy integrals following the steps taken in equation (2.102) yields that the energy dissipated per cycle is ΔE = 8 αX3ω2 3 (2.130) Again, equating this to the energy dissipated by a viscously damped system given in equation (2.100) yields ceq = 8 αωX 3π (2.131) This equivalent viscous-damping value can then be used in the amplitude and phase formulas for a linear viscously damped forced harmonic motion to approximate the steady-state response. Example 2.7.3 Calculate the approximate amplitude at resonance for velocity-squared damping. Solution Using the magnitude expression for viscous damping at resonance (r = 1 and ω = ωn), equation (2.40) and the expression for the damping ratio given in equation (1.30) yields X = f0 mf0 = 2ζωωn ceqω Substitution of (2.131) yields X = mf0 (8>3π)αω2X so that X = 3πmf0 B 8αω 2 = 3πf0 m2 C 8kCρA As expected, for larger values of the mass density of the fluid, the drag coefficient of the cross-sectional area produces a smaller amplitude at resonance. n If several forms of damping are present, one approach to examining the harmonic response is to calculate the energy lost per cycle of each form of damping present, add them up, and compare them to the energy loss from a single viscous damper. Then the formulas for magnitude and phase from equation (2.40) are used 180 Response to Harmonic Excitation TABLE 2.2 Chap. 2 DAMPIng MoDElS Name Damping Force Linear-viscous damping # cx Air damping # # a sgn (x)x2 Coulomb damping # β sgn x Displacement-squared damping # d sgn (x)x2 Solid, or structural, damping # b sgn (x)|x| ceq Source c 8aωX 3π 4β πωX 4dX 3πω 2b πω Slow fluid Fast fluid Sliding friction Material damping Internal damping to approximate the response. For n damping mechanisms dissipating energy per cycle of ΔEi, for the ith mechanism, the equivalent viscous-damping constant is a i = 1 ΔEi = πωX2 n ceq (2.132) A study of various damping mechanisms is presented by Bandstra (1983) and summarized in Table 2.2. 2.8 NUMERICAL SIMULATION AND DESIGN In the previous sections, great effort was put forth to derive analytical expressions for the response of various single-degree-of-freedom systems driven by a harmonic load. These analytical expressions are extremely useful for design and for understanding some of the physical phenomena. Plots of the time response and of the steady-state magnitude and phase were constructed to realize the nature and features of the response. Rather than plotting the analytical function describing the response, the time response may also be computed numerically using an Euler or Runge–Kutta integration and computational software packages as introduced in Section 1.8. While numerical solutions such as these are not exact, they do allow nonlinear terms to be considered. In addition, these packages may be used to generate all of the plots of phase and magnitude and the time responses given in the previous sections (in fact, all plots in this text are generated using MATLAB or, in some cases, Mathcad). The computational packages will also help you derive expressions, such as equation (2.38), by using symbolic algebra. Perhaps the most advantageous use of computational software is the ability to resolve quickly the time response for various values of parameters. The ability to plot the solution quickly allows engineers to examine what would happen if the damping changes or the input force level changes. Such parametric studies of the time response are useful for design and for building intuition about a given system. Sec. 2.8 181 numerical Simulation and Design In order to solve for the forced response to a harmonic input numerically, equation (1.97) needs to be modified slightly to incorporate the applied force. The first-order, or state-space, form of equation (2.27) becomes # x1(t) = x2(t) # x2(t) = -2ζωn x2(t) - ω2n x1(t) + f0 cos ωt (2.133) # where x2 denotes the velocity x(t) and x1 denotes the position x(t), as before. This is subject to the initial conditions x1(0) = x0 and x2(0) = v0. Given ω, ωn, ζ, f0, x0, and 0, the solution of equation (2.133) can be determined numerically. The matrix form of equation (2.133) becomes # x(t) = Ax(t) + f(t) (2.134) where x and A are the state vector and state matrix as defined previously in equation (1.96). The vector f is the applied force and takes the form f(t) = c The Euler form of equation (2.134) is 0 d f0 cos ωt (2.135) x(t i + 1) = x(t i) + Ax(t i)Δt + f(t i)Δt (2.136) This expression can also be adapted to the Runge–Kutta formulation, and most of the codes mentioned in Appendix G have built-in commands for a Runge–Kutta solution. The numerical integration to determine the response of a system is an approximation, whereas the plotting of the analytical solution is exact. So why bother to integrate numerically to find the solution? Because closed form solutions often do not exist, such as in the treatment of nonlinear terms as was illustrated in Section 1.10. This section discusses the solution of the forced response using numerical integration in an environment where the exact solution is available for comparison. The examples in this section introduce numerical integration to compute the forced response and to compare these to the exact solution. The following example illustrates the use of various programs to compute and plot the solution. Example 2.8.1 Numerically integrate and plot the response of an underdamped system determined by m = 100 kg, k = 2000 N>m, and c = 200 kg>s, subject to the initial conditions of x0 = 0.01 m and 0 = 0.1 m>s, and the applied force F(t) = 150 cos 10t. Then plot the exact response as computed by equation (2.38). Compare the plot of the exact solution to the numerical simulation. Solution This is first worked out in Mathcad. The equivalent commands for MATLAB and Mathematica are given at the end of this example. Start by entering the relevant numbers. x0 := 0.01 ωn := v0 := 0.1 k Am ζ := m := 100 c 21k # m c := 200 F0 := 150 k := 2000 ω := 10 182 Response to Harmonic Excitation ωd := ωn # 21 - ζ2 B := xTc(t) := e xTs(t) := e-ζ 3 x0 f0 1 ωn2 f0 := - ζ # ωn # t # # ωn # t # 2 2 - c 3x0 - 1ωn2 - ω22 # B4 xSS(t) := B # 3( ωn2 2 - ω F0 m - ω2 2 # B 4 # cos (ωd # t) (ωn - ω ) + (2 # ζ # ωn # ω)2 2 #ζ # ωn Chap. 2 Defines one coefficient - 2 # ζ # ωn # Write out equation (2.38) in parts ω2 # v0 # d sin (ωd # t) B + ωd ωd ) # cos(ω # t) + 2 # ζ # ωn # ω # sin(ωd # t)4 ωd x(t) := xTc(t) + xTs(t) + xSS(t) Next compute the same response using Runge–Kutta by setting up a state-space representation, use rkfixed to solve, and save the solution in x, t vectors X := c x0 d v0 D(t,X) := c X1 d - 2 # ζ # ωn # X1 - ωn2 # X0 + f0 # cos(ω # t) Z := rkfixed (X, 0, 6, 2000, D) t := Z 6 0 7 ¡ x := x(t) xs := Z 6 1 7 Change exact solution into a vector for plotting 0.04 0.02 xs x 0 1 2 3 4 5 6 0.02 0.04 t Figure 2.32 The exact solution (dashed line) and a Runge–Kutta solution (solid line) plotted on the same graph. From Figure 2.32, note that the numerical solution and the exact solution are the same. However, it is always important to remember that the numerical integration yields only an approximate solution. Sec. 2.8 numerical Simulation and Design 183 Next the MATLAB code for computing these plots is given. First an M-file is created with the equation of motion given in first-order form. -------------------function v=f(t,x) m=100; k=2000; c=200; Fo=150; w=10; v=[x(2); x(1).*-k/m + x(2).*-c/m + Fo/m*cos(w*t)]; ---------------------------- Then the following is entered into the command window: clear all xo=0.01; vo=0.1; m=100; c=200; k=2000; Fo=150; w=10; t=0:0.01:5; wn=sqrt(k/m); z=c/(2*sqrt(k*m)); wd=wn*sqrt(1-z^2); fo=Fo/m; % Defines one coefficient B= fo/((wn^2-w^2)^2 + (2*z*wn*w)^2); for i=1:max(length(t)) %Write out equation (2.38) in parts xTc(i)=exp(-z*wn*t(i)) * (xo-(wn^2-w^2)*B) * cos(wd*t(i)); xTs(i)=exp(-z*wn*t(i)) * ((xo-(wn^2-w^2)*B)*z*wn/wd 2*z*wn*w^2/wd*B + vo/wd) * sin(wd*t(i)); xSs(i)=B*((wn^2-w^2)*cos(w*t(i)) + 2*z*wn*w*sin(w*t(i))); x(i)=xTc(i) + xTs(i) + xSs(i); end figure(1) plot(t,x) clear all xo=[0.01; 0.1]; ts=[0 5]; [t,x]= ode45('f',ts,xo); figure (2) plot(t,x(:,1)) 184 Response to Harmonic Excitation Chap. 2 In Mathematica, the exact solution and numerical solution are computed and plotted by the following list of commands. In[1] : = m = 100; k = 2000; c = 200; x0 = .01; v0 = .1; k ωn = ; Am c ζ = ; 21k * m ω = 10 ωd = ωn * 21 - ζ2; F0 = 150; F0 ; f0 = m f0 ; 2(ωn - ω ) + (2 * ζ * ωn * ω)2 2 2 θ = ArcTan[ωn - ω , 2 * ζ * ωn * ω]; ϕ = ArcTan[v0 - X * ω * Sin[θ] + ζ * ωn * (x0 - X * Cos[θ]), ωd * (x0 - X * Cos[θ])]; x0 - X * Cos [θ] A = ; Sin [ϕ] xanal[t_] = A * Exp[-ζ * ωn * t] * Sin[ωd * t + ϕ] + X * Cos[ω * t - θ]; X = 2 2 2 numerical = NDSolve[{x"[t] + 2 * ζ * ωn * x'[t] + ωn2 * x[t] == f0 * Cos[ω * t], x[0] == x0, x'[0] == v0}, x[t], {t, 0, 5}]; Plot[{Evaluate[x[t] /. numerical], xanal[t]}, {t, 0, 5}, PlotRange S {-0.04, 0.04}] n With the ability to compute numerical solutions, either by solving a differential equation and plotting it or by plotting the analytical solution, comes the ability to perform parametric studies of the response quickly. Once a response plot is written into a code, it is a trivial matter to reproduce the plot with new values of the physical parameters (mass, damping, stiffness, initial conditions, and the driving force magnitude and frequency). Such parametric studies can be used both to understand the physical nature of the response and to design. Here design refers to choosing the physical parameters to obtain a more desirable response. Chapter 5 focuses on design. The following example illustrates how the computer may be used to determine design parameters. Sec. 2.8 185 numerical Simulation and Design Example 2.8.2 An electronics module is mounted on a machine and is modeled as a single-degreeof-freedom spring, mass, and damper. During normal operation, the module (having a mass of 100 kg) is subject to a harmonic force of 150 N at 5 rad>s. Because of material considerations and static deflection, the stiffness is fixed at 500 N>m and the natural damping in the system is 10 kg>s. The machine starts and stops during its normal operation, providing initial conditions to the module of x0 = 0.01 m and 0 = 0.5 m>s. The module must not have an amplitude of vibration larger than 0.2 m even during the transient stage. First, compute the response by numerical simulation to see if the constraint is satisfied. If the constraint is not satisfied, find the smallest value of damping that will keep the deflection less than 0.2 m. Solution This requires the numerical integration of a second-order differential equation. Codes for these are given in Example 2.8.1. Use either equation (2.136) or a Runge–Kutta equivalent to integrate numerically the equation of motion and plot the result to see if the response is larger than 0.2 m. Here Mathcad is used to generate the response plot of Figure 2.33. x0 : = 0.01 v0 : = 0.5 F0 : = 150 m : = 100 k ωn : = Am f0 : = k : = 500 ζ := c : = 10 c 21k # m F0 m ζ = 0.022 ω := 5 D(t, X) : = c ωn = 2.236 X := J x0 R v0 X1 d - 2 # ζ # ωn # X1 - ωn2 # X0 + f0 # cos(ω # t) Z : = rkfixed(X, 0, 40, 4000, D) t : = Z<0> x : = Z<1> Note from the simulated response that the transient term is larger than the steady state and has violated the constraint that x(t) … 0.2 m. Thus the damping must be increased to bring down the amplitude of the transient response. The design is performed by simply increasing the value of c in the preceding code and running it again. This is repeated until the response falls below 0.2 m. Because damping is expensive to add to a system, the increment of damping at each iteration is very small. This “design” procedure produces the plot shown in Figure 2.34 for a damping coefficient of c = 195 kg >s (ζ = 0.436). A value of the damping coefficient that is a few kg>s less than 195 kg>s will produce a response larger than the desired 0.2 m. This is a fairly large value of damping, a source of concern for the designer. Next the MATLAB code for computing these plots is given. First an M-file is created with the equation of motion given in first-order form. 186 Response to Harmonic Excitation Chap. 2 0.4 0.2 x 0 10 20 30 40 0.2 0.4 t Figure 2.33 The simulated response for c = 10 kg >s illustrating that the transient response exceeds 0.2 m. 0.2 0.1 x 0 5 10 15 0.1 t Figure 2.34 The simulated response for c = 195 kg >s illustrating that the transient does not exceed 0.2 m. --------------------------------------------------function v=f(t,x) m=100; k=500; c=10; Fo=150; w=5; v=[x(2); x(1).*-k/m+x(2).*-c/m + Fo/m*cos(w*t)]; --------------------------------------------------- 20 Sec. 2.8 187 numerical Simulation and Design Then the following is typed in the command window: clear all xo=[0.01; 0.5]; ts=[0 40]; [t,x]=ode45('f',ts,xo); figure(1) plot(t,x(:,1)) This is repeated with different values of damping until the desired amplitude is reached. In Mathematica, the exact solution and numerical solution are computed and plotted by the following list of commands: In[1] : = m = 100; k = 500; c = 10; x0 = .01; v0 = .5; k ωn = ; Am c ζ = ; 2 * 1k * m ω = 5; ωd = ωn * 21 - ζ2; F0 = 150; F0 f0 = ; m X = f0 2(ωn2 - ω2)2 + (2 * ζ * ωn * ω)2 ; θ = ArcTan[ωn2 - ω2, 2 * ζ * ωn * ω]; ϕ = ArcTan[v0 - X * ω * Sin [θ] + ζ * ωn * (x0 - X * Cos [θ]), ωd * (x0 - X * Cos[θ])]; x0 - X * Cos[θ] A = ; Sin[ϕ] xanal[t_] = A * Exp[- ζ * ωn * t] * Sin[ωd * t + ϕ] + X * Cos[ω * t - θ]; numerical = NDSolve[{x"[t] + 2 * ζ * ωn * x'[t] + ωn2 * x[t] == f0 * Cos[ω * t], x[0] == x0, x'[0] == v0}; x[t], {t,0,40}]; Plot[{Evaluate[x[t]/.numerical], xanal[t]}, {t,0,40}, PlotRange S {–4, .4}, PlotStyle S {RGBColor[1,0,0], RGBColor[0,1,0]}] n 188 Response to Harmonic Excitation Chap. 2 2.9 NONLINEAR RESPONSE PROPERTIES The use of numerical integration as introduced in the previous section allows us to consider the effects of various nonlinear terms in the equation of motion. As noted in the free-response case discussed in Section 1.10, the introduction of nonlinear terms generally results in an inability to find exact solutions, so we must rely on numerical integration and qualitative analysis to understand the response. Several important differences between linear and nonlinear systems are as follows: 1. A nonlinear system has more than one equilibrium point and each may be either stable or unstable. 2. Steady-state behavior of a nonlinear system does not always exist, and the nature of the solution is strongly dependent on the value of the initial conditions. 3. The period of oscillation of a nonlinear system depends on the initial conditions, the amplitude of excitation, and the physical parameters, unlike the linear response, which depends only on mass, damping, and stiffness values and is independent from the initial conditions. 4. Resonance in nonlinear systems may occur at excitation frequencies that are not equal to the linear system’s natural frequency. 5. We cannot use the idea of superposition, used in Section 2.4, in a nonlinear system. 6. A harmonic excitation may cause a nonlinear system to respond in a nonperiodic, or chaotic, motion. Many of these phenomena are very complex and require analysis skills beyond the scope of a first course in vibration. However, some initial understanding of nonlinear effects in vibration analysis can be observed by using the numerical solutions covered in the previous section. In this section, several simulations of the response of nonlinear systems are numerically computed and compared to their linear counterparts. Recall from Section 1.10 that, if the equations of motion are nonlinear, the general single-degree-of-freedom system may be written as $ # (2.137) x(t) + f 3x(t), x(t)4 = 0 where the function, f, can take on any form, linear or nonlinear. In the forced response case considered in this chapter, the equation of motion becomes $ # (2.138) x(t) + f 3x(t), x(t)4 = f0 cos ωt Formulating this last expression into the state-space, or first-order, equation (2.138) takes on the form # x1(t) = x2(t) # x2(t) = -f(x1, x2) + f0 cos ωt (2.139) Sec. 2.9 189 nonlinear Response Properties This state-space form of the equation is used for numerical simulation in several of the codes. By defining the state vector, x = [x1(t), x2(t)]T, used in equation (1.96) and a nonlinear vector function F as F(x) = c x2(t) d -f(x1, x2) equations (2.139) may now be written in the first-order vector form # x = F(x) + f(t) (2.140) (2.141) Equation (2.141) is the forced version of equation (1.116). Here f(t) is simply f(t) = c 0 d f0 cos ωt (2.142) Then the Euler integration method for the equations of motion in the first-order form becomes x(t i + 1) = x(t i) + F[x(t i)]Δt + f(t i)Δt (2.143) This expression forms a basic approach to numerically integrating to compute the forced response of a nonlinear system and is the nonlinear, forced-response version of equations (1.100) and (2.134). Nonlinear systems are difficult to analyze numerically as well as analytically. For this reason, the results of a numerical simulation must be examined carefully. In fact, using a more sophisticated integration method, such as Runge–Kutta, is recommended for nonlinear systems. In addition, checks on the numerical results using qualitative behavior should also be performed whenever possible. In the following we consider the single-degree-of-freedom system illustrated in Figure 2.35, with nonlinear spring or damping elements. A series of examples is presented using numerical simulation to examine the behavior of nonlinear systems and to compare them to the corresponding linear systems. Example 2.9.1 Compute the response of the system in Figure 2.35 for the case that the damping is linear and viscous and the spring is a nonlinear softening spring of the form fk(x) = kx - k1x3 and the system is subject to a harmonic excitation of 1500 N at a frequency of approximately one-third the natural frequency (ω = ωn >2.964) and initial conditions of x0 = 0.01 m and x (t) k m c F (t) Figure 2.35 A spring–mass–damper system with potentially nonlinear elements. 190 Response to Harmonic Excitation Chap. 2 0 = 0.1 m>s. The system has a mass of 100 kg, a damping coefficient of 170 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 520 N>m3. (a) Compute the solution and compare it to the linear solution (k1 = 0). (b) Examine the response for the case that the driving force is near linear resonance (ω = ωn >1.09). Solution The equation of motion becomes $ # mx(t) + cx(t) + kx(t) - k1x3(t) = 1500 cos ωt Dividing by the mass yields $ # x(t) + 2ζ ωnx(t) + ω2nx(t) - αx3(t) = 15 cos ωt Next, write this equation in state-space form to get # x1(t) = x2(t) # x2(t) = - 2ζωnx2(t) - ω2nx1(t) + αx31(t) + 150 cos ωt This last set of equations can be used in MATLAB or Mathcad to integrate numerically for the time response. Mathematica uses the second-order equation directly. Figures 2.36 and 2.37 illustrate three plots. The straight line in each is the magnitude of the linear steady-state response for the parameters given as computed by equation (2.39). The solid line in each is the response of the nonlinear system, while the dashed line is the response of the linear system. In case (a), the response of the nonlinear system exceeds that of the linear system and appears to be in resonance even though the driving frequency is almost one-third that of the natural frequency. However, in case (b) for the system near resonance (Figure 2.37), the nonlinear system response 2 Displacement 1 0 10 20 30 1 2 Time, t Figure 2.36 The solid line is a plot of the response of the nonlinear system, the dashed line is a plot of the response of the linearized system, and the straight dashed line is the magnitude of the steady-state amplitude of the linear system as given by equation (2.39) for a driving frequency near one-third of the natural frequency. 40 191 nonlinear Response Properties 4 Displacement Sec. 2.9 2 0 5 10 20 15 2 Time, t Figure 2.37 The solid line is a plot of the response of the nonlinear system, the dashed line is a plot of the response of the linearized system, and the straight dashed line is the magnitude of the steady-state amplitude of the linear system as given by equation (2.39) for a driving frequency near the natural frequency of the linear system. is lower in amplitude than the linear system response and appears to be oscillating at two frequencies. An essential difference between linear and nonlinear systems is that a harmonically excited nonlinear system may oscillate at frequencies other than the frequency of excitation. The codes for numerically simulating and plotting the curves given in Figures 2.36 and 2.37 are given next. In Mathcad the code is x0 : = 0.01 v0 : = 0.1 m : = 100 α : = 5.2 k ωn := Am ζ := D(t,X) : = J c : = 170 F0 : = 1500 c 21k # m X := J k: = 2000 f0 : = x0 R v0 F0 m ω := ωn 2.964 Y := X X1 R - 2 # ζ # ωn # X1 - ωn2 # X0 + α # (X0)3 + f0 # cos(ω # t) L(t,Y) : = J Y1 R (- 2 # ζ # ωn # Y1 - ωn2 # Y0) + f0 # cos(ω # t) Z := rkfixed(X,0,40,4000,D) t := Z xL : = W<1> <0> x : = Z<1> d(t) : = W : = rkfixed(Y,0,40,4000,L) f0 2(ωn - ω ) + (2 # ζ # ωn # ω) 2 2 2 2 ¡ F : = d(t) 192 Response to Harmonic Excitation Chap. 2 In MATLAB the code is % (a) clear all xo=[0.01; 0.1]; ts=[0 40]; [t,x]=ode45('f',ts,xo); plot(t,x(:,1)); hold on [t,x1]=ode45('f1',ts,xo); plot(t,x1(:,1),'r'); hold off %-----------------------------------------------------function v=f(t,x) m=100; k=2000; k1=0; c=170; Fo=1500; w=sqrt(k/m)/2.964; v=[x(2); x(1).*-k/m+x(2).*-c/m + x(1)^3*k1/m + Fo/m*cos(w*t)]; %-----------------------------------------------------function v=f1(t,x) m=100; k=2000; k1=520; c=170; Fo=1500; w=sqrt(k/m)/2.964; v=[x(2); x(1).*-k/m+x(2).*-c/m + x(1)^3*k1/m + Fo/m*cos(w*t)]; %(b) clear all xo=[0.01; 0.1]; ts=[0 20]; [t,x]=ode45('f',ts,xo); figure(2) plot(t,x(:,1)); hold on [t,x1]=ode45('f1',ts,xo); plot(t,x1(:,1),'r'); hold off %-----------------------------------------------------function v=f(t,x) m=100; k=2000; k1=0; c=170; Fo=1500; w=sqrt(k/m)/1.09; v=[x(2); x(1).*-k/m + x(2).*-c/m + x(1)^3*k1/m + Fo/m*cos(w*t)]; %-----------------------------------------------------function v=f1(t,x) m=100; k=2000; k1=520; c=170; Fo=1500; w=sqrt(k/m)/1.09; v=[x(2); x(1).*-k/m+x(2).*-c/m + x(1)^3*k1/m + Fo/m*cos(w*t)]; Sec. 2.9 193 nonlinear Response Properties In Mathematica the code is In[1] : = <<PlotLegends' In[2] : = m = 100; k = 2000; c = 170; F0 = 1500; F0 ; F0 = m α = 5.2; k ωn = ; Am c ζ = ; 2 * 1k * m x0 = 0.01; v0 = 0.1; ωn ω = ; 2.964 ssmagnitude = f0 2(ωn2 - ω2)2 + (2 * ζ * ωn * ω)2 ; In[14] : = nonlinear = NDSolve[{x"[t] + 2 * ζ * ωn * x'[t] + ωn2 * x[t] – α * (x[t])3 == f0 * Cos[ω * t], x[0] == x0, x'[0] == v0}, x[t], {t, 0, 40}, MaxSteps S 2000]; linear = NDSolve[{x1"[t]+2 * ζ * ωn * xl'[t]+ωn2] * x1[t] == f0 * Cos[ω * t], x1[0] == x0, x1'[0] == v0}, x1[t], {t, 0, 40}, MaxSteps S 2000]; Plot[{Evaluate[x[t] /. nonlinear], Evaluate[x1[t] /. linear], ssmagnitude}, {t, 0, 40}, PlotRange S {–2,2}, PlotStyle S {RGBColor[1,0,0], RGBColor[0, 1, 0], RGBColor[0, 0, 1]}, PlotLegend S {Nonlinear", "Linear", "Steady State Amp."}, LegendPosition S {1, 0}, LegemdSize S {1, .3}] A very important point is that the initial conditions are critical in determining the nature of the response (recall the pendulum of Example 1.10.4). If the initial position and>or velocity are increased, the nonlinear solution will grow without bound and the numerical integration will fail. On the other hand, the linear solution will still oscillate with amplitude less than the straight line in Figures 2.36 and 2.37. n Example 2.9.2 Compare the forced response of a system with velocity-squared damping as defined in equation (2.129) using numerical simulation of the nonlinear equation to that of the 194 Response to Harmonic Excitation Chap. 2 response of the linear system obtained using equivalent viscous damping as defined by equation (2.131). Solution Velocity-squared damping with a linear spring and harmonic input is described by equation (2.129), repeated here: $ # # mx + α sgn(x)x2 + kx = F0 cos ωt The equivalent viscous-damping coefficient is calculated in equation (2.131) to be ceq = 8 αωX 3π The value of the magnitude, X, can be approximated for near resonance conditions. The value is computed in Example 2.7.3 to be X = 3πmf0 B 8αω2 Combining these last two expressions yields an equivalent viscous-damping value of ceq = 8mαf0 B 3π Using this value as the damping coefficient results in a linear system of Figure 2.34 that approximates equation (2.131). Figures 2.38 and 2.39 are plots of the linear system with equivalent viscous damping and a numerical simulation of the full nonlinear equation (2.129) for two different values of the parameter α depending on the drag coefficient. Several conclusions can be made from these two plots. 40 20 x xL -- 0 5 10 15 20 t 40 Figure 2.38 The displacement of the equivalent viscous damping (dashed line) and the displacement of the nonlinear system (solid line) versus time for the case of α = 0.005. 20 Sec. 2.9 195 nonlinear Response Properties 5 x xL -- 0 5 10 5 15 20 t Figure 2.39 The displacement of the equivalent viscous damping (dashed line) and the displacement of the nonlinear system (solid line) versus time for the case of α = 0.5. First, the larger the drag coefficient, the greater the error is in using the concept of equivalent viscous damping. Second, the frequency of the response looks similar, but the amplitude of oscillation is greatly overestimated by the equivalent viscousdamping technique. The computer codes for solving and plotting both the linear and nonlinear equations follow. The code in Mathcad is m : = 10 k Am ωn : = k : = 200 f0 : = ζ := D(t,X) : = C F0 m ceq 21k # m α : = .0050 ω : = 1 # ωn X := J 1 R 0.1 F0 : = 150 ceq : = 8#α#m# f0 A 3#π Y := X X1 X1 α S 2 - ωn # X0 - # (X1) # + f0 # cos(ω # t) m |X1| L(t,Y) : = J 2 Y1 R (- 2 # ζ # ωn # Y1 - ωn2 # Y0) + f0 # cos(ω # t) Z : = rkfixed(X, 0, 40, 2000, D) <0> t: = Z x : = Z<1> W : = rkfixed(Y, 0, 40, 2000, L) xL : = W<1> 196 Response to Harmonic Excitation Chap. 2 The MATLAB code consists of the following commands and M-files: % α=0.005 clear all xo=[1; 0.1]; ts=[0 20]; [t,x]=ode45('f',ts,xo); figure(1) plot(t,x(:,1)); hold on [t,x1]=ode45('f1',ts,xo); plot(t,x1(:,1),'r'); hold off %------------------------------function v=f(t,x) m=10; k=200; alpha=0.005; Fo=150; w=sqrt(k/m); ceq=sqrt(8*m*alpha*Fo/m/3/pi); v=[x(2); x(1).*-k/m+x(2).*-ceq/m + Fo/m*cos(w*t)]; %------------------------------function v=f1(t,x) m=10; k=200; alpha=0.005; Fo=150; w=sqrt(k/m); v=[x(2); x(1).*-k/m + x(2)^2.*-alpha/m * sign(x(2)) + Fo/m*cos(w*t)]; The Mathematica code is In[1] : = <<PlotLegends' In[2] : = m = 10; k = 200; F0 = 150; F0 ; f0 = m α = .005; k ωn = ; Am 8*α*m ceq = * f0; A 3*π ceq ζ = ; 2 * 1k * m x0 = 1; v0 = 0.1; ω = ωn; α In[13] : = velsquared = NDSolve[{x''[t] + * Sign[x'[t]] * (x'[t])2 m 2 + ωn * x[t] == f0 * Cos[ω * t], x[0] == x0, x'[0] == v0}, x[t], {t, 0, 20}, MaxSteps S 2000]; Chap. 2 197 Problems equivdamping = NDSolve[{xeq''[t] + 2 * ζ * ωn * xeq'[t] + ωn2 * xeq[t] == f0 * Cos[ω * t], xeq[0] == x0, xeq’[0] == v0}, xeq[t], {t, 0, 20}, MaxSteps S 2000]; Plot[{Evaluate[x[t] /. velsquared], Evaluate[xeq[t] /. equivdamping]}, {t, 0, 20}, PlotRange S {-40, 40}, PlotStyle S {RGBColor[1, 0, 0], RGBColor[0, 1, 0], RGBColor[0, 0, 1]}, PlotLegend S {"Velocity Squared", "Equivalent Damping", LegendPosition S {1, 0}, LegendSize S {1, .3}] n PROBLEMS Those problems marked with an asterisk are intended to be solved using computational software. Section 2.1 (Problems 2.1 through 2.19) 2.1. The forced response of a single-degree-of-freedom, spring–mass system is modeled by (assume the units are Newtons) $ 3x(t) + 12x(t) = 3 cos ωt Compute the magnitude of the forced response for the two cases ω = 2.1 rad >s and ω = 2.5 rad >sec. Comment on why one value is larger then the other. 2.2. Consider the forced response of a single-degree-of-freedom, spring–mass system that is modeled by (assume the units are Newtons) $ 3x(t) + 12x(t) = 3 cos ωt Compute the total response of the system if the driving frequency is 2.5 rad>s and the initial position and velocity are both zero. 2.3. Compute the response of a spring–mass system modeled by equation (2.2) to a force of magnitude 23 N, driving frequency of twice the natural frequency, and initial conditions given by x 0 = 0 m and v0 = 0.2 m>s. The mass of the system is 10 kg, the spring stiffness is 1000 N>m, and the mass of the spring is considered and known to be 1 kg. What percent does the natural frequency change if the mass of the spring is not taken into consideration? 2.4. Show that the solution x(t) = x(t) = f0 ω2n - ω2 f0 2(ω2n - ω2) 3cos ωt sin - cos ωnt 4 can be written ωt + ωnt ωnt - ωt sin . 2 2 198 Response to Harmonic Excitation Chap. 2 2.5. A spring–mass system is driven from rest harmonically such that the displacement response exhibits a beat of period of 0.2π s. The period of oscillation is measured to be 0.02π s. Calculate the natural frequency and the driving frequency of the system. 2.6. An airplane wing modeled as a spring–mass system with natural frequency 40 Hz is driven harmonically by the rotation of its engines at 39.9 Hz. Calculate the period of the resulting beat. 2.7. Compute the total response of a spring–mass system with the following values: k 1000 N >m, m = 10 kg, subject to a harmonic force of magnitude F0 = 100 N and frequency of 8.162 rad>s, and initial conditions given by x0 = 0.01 m and v0 = 0.01 m>s. Plot the response. 2.8. Consider the system in Figure P2.8, write the equation of motion, and calculate the response assuming (a) that the system is initially at rest, and (b) that the system has an initial displacement of 0.05 m. x (t) 10 sin 10t N 100 kg Frictionfree surface 2,000 N/m Figure P2.8 2.9. Consider the system in Figure P2.9, write the equation of motion, and calculate the response assuming that the system is initially at rest for the values k1 = 100 N>m, k2 = 500 N>m, and m = 89 kg. k1 k2 10 sin 10t N m Figure P2.9 2.10. Consider the system in Figure P2.10, write the equation of motion, and calculate the response assuming that the system is initially at rest for the values θ = 30°, k = 1000 N>m, and m = 50 kg. k m 90 sin 2.5t N  Figure P2.10 Chap. 2 199 Problems 2.11. Compute the initial conditions such that the response of $ mx + kx = F0 cos ωt oscillates at only one frequency, (ω). 2.12. The natural frequency of a 65-kg person illustrated in Figure P2.12 is measured along vertical, or longitudinal, direction to be 4.5 Hz. (a) What is the effective stiffness of this person in the longitudinal direction? (b) If the person, 1.8 m in length and 0.58 m2 in crosssectional area, is modeled as a thin bar, what is the modulus of elasticity for this system? x(t) Figure P2.12 The longitudinal vibration of a person. 2.13. If the person in Problem 2.12 is standing on a floor vibrating at 4.49 Hz with an amplitude of 1 N (very small), what longitudinal displacement would the person “feel”? Assume that the initial conditions are zero. 2.14. Vibration of body parts is a significant problem in designing machines and structures. A jackhammer provides a harmonic input to the operator’s arm. To model this situation, treat the forearm as a compound pendulum subject to a harmonic excitation (say of mass 6 kg and length 44.2 cm) as illustrated in Figure P2.14. Consider point O as a fixed pivot. Compute the maximum deflection of the hand end of the arm if the jackhammer applies a force of 10 N at 2 Hz. O  F0 cos t m g Figure P2.14 A vibration model of a forearm driven by a jackhammer. 200 Response to Harmonic Excitation Chap. 2 2.15. An airfoil is mounted in a wind tunnel for the purpose of studying the aerodynamic properties of the airfoil’s shape. A simple model of this is illustrated in Figure P2.15 as a rigid inertial body mounted on a rotational spring fixed to the floor with a rigid support. Find a design relationship for the spring stiffness, k, in terms of the rotational inertia, J; the magnitude of the applied moment, M0; and the driving frequency, ω, that will keep the magnitude of the angular deflection less then 5°. Assume that the initial conditions are zero and that the driving frequency is such that ω2n - ω2 7 0. k J  u(t)  M0 cos t Figure P2.15 A vibration model of a wing in a wind tunnel. 2.16. The spar of an airplane wing is a relatively rigid beam extending along the length of the wing inside the wing to provide strength. It is typical to model a spar as a cantilever beam with the fixed end at the body of the aircraft. An example is given in Figure P2.16. Using the modeling methods given in Section 1.5, determine a single-degree-of-freedom model for the spar and compute its natural frequency. The spar here is modeled as a cantilever beam of dimensions length 560 mm, width 38 mm, and thickness 3.175 mm, and has a mass of 13.975 grams. The beam’s Young’s modulus is 10.29 GPa and its shear modulus is 1.65 GPa. Figure P2.16 A small, unmanned air vehicle with a rigid spar, modeled as a beam. 2.17. Compute the response of a shaft-and-disk system to an applied moment of M = 10 sin 312 t as indicated in Figure P2.17. Assume that the shaft is initially at rest (zero initial conditions) and J = 0.5 kg m2, the shear modulus is G = 8 * 1010 N>m2, the shaft is 1 m long of diameter 5 cm, and made of steel. 201 Problems k J Jp Fixed base Chap. 2 10 sin 312t Figure P2.17 2.18. Consider a spring–mass system with zero initial conditions described by $ # x(t) + 4x(t) = 12 cos 2t, x(0) = 0, x(0) = 0 and compute the form of the response of the system. 2.19. Consider a spring–mass system with zero initial conditions described by $ # x(t) + 4x(t) = 10 sin 5t, x(0) = 0, x(0) = 0 and compute the form of the response of the system. Section 2.2 (Problems 2.20 through 2.38) 2.20. Calculate the constants A and ϕ for arbitrary initial conditions, x0 and the forced response given by 0, in the case of x(t) = Ae -ζωnt sin (ωdt + ϕ) + X cos (ωt - θ) Compare this solution to the transient response obtained in the case of no forcing function (i.e., F0 = 0). 2.21. Consider the spring–mass–damper system defined by (use basic SI units) $ # 4x(t) + 24x(t) + 100x(t) = 16 cos 5t First, determine if the system is underdamped, critically damped, or overdamped. Then compute the magnitude and phase of the steady-state response. 2.22. Show that the following two expressions are equivalent: xp(t) = X cos (ωt - θ) and xp (t) = As cos ωt + Bs sin ωt 2.23. Calculate the total solution of $ # x + 2ζωnx + ω2nx = f0 cos ωt for the case that m = 1 kg, ζ = 0.01, ωn = 2 rad>s, f0 = 3 N>kg, and ω = 10 rad>s, with initial conditions x0 = 1 m and v0 = 1 m>s, and then plot the response. 2.24. A 100-kg mass is suspended by a spring of stiffness 30 * 103 N>m with a viscousdamping constant of 1000 Ns>m. The mass is initially at rest and in equilibrium. Calculate the steady-state displacement amplitude and phase if the mass is excited by a harmonic force of 80 N at 3 Hz. 2.25. Plot the total solution of the system of Problem 2.24 including the transient. 202 Response to Harmonic Excitation Chap. 2 2.26. A damped spring–mass system modeled by (units are Newtons) $ # 100x(t) + 10x(t) + 1700x(t) = 1000 cos 4t is also subject to initial conditions: x0 = 1 mm and response, x(t), of the system. 0 = 20 mm>s. Compute the total 2.27. Consider the pendulum mechanism of Figure P2.27 which is pivoted at point O. Calculate both the damped and undamped natural frequency of the system for small angles. Assume that the mass of the rod, spring, and damper are negligible. What driving frequency will cause resonance? k l1  0.05 m 0 l2  0.07 m c l  0.10 m m F(t) Figure P2.27 2.28. Consider the pivoted mechanism of Figure P2.27 with k = 4 * 103 N>m, l1 = 0.05 m, l2 = 0.07 m, l = 0.10 m, and m = 40 kg. The mass of the beam is 40 kg. It is pivoted at point 0 and assumed to be rigid. Design the dashpot (i.e., calculate c) so that the damping ratio of the system is 0.2. Also determine the amplitude of vibration of the steady-state response if a 10-N force is applied to the mass, as indicated in the figure, at a frequency of 10 rad>s. 2.29. Compute the response of a shaft-and-disk system to an applied moment of M = 10 sin 312 t k J Jp 10 sin 312t Figure P2.29 Fixed base as indicated in Figure P2.29. Assume that the shaft is initially at rest (zero initial conditions) and J = 0.5 kg m2, the shear modulus is G = 8 * 1010 N>m2, the shaft is 1-m long of diameter 5 cm, and made of steel. Assume the damping ratio of steel is ζ = 0.01. Chap. 2 203 Problems 2.30. Compute the forced response of a spring–mass–damper system with the following values: c = 200 kg>s, k = 2000 N>m, m = 100 kg, subject to a harmonic force of magnitude F0 = 15 N and frequency of 10 rad>s and initial conditions of x0 = 0.01 m and v0 = 0.1 m>s. Plot the response. How long does it take for the transient part to die off? 2.31. Compute a value of the damping coefficient, c, such that the steady-state response amplitude of the system in Figure P2.31 is 0.01 m. x (t) c 20 cos 6.3t N 100 kg Frictionfree surface 2,000 N/m Figure P2.31 2.32. Consider a spring–mass–damper system like the one in Figure P2.31 with the following values: m = 100 kg, c = 100 kg>s, k = 3000 N>m, F0 = 25 N, and the driving frequency ω = 5.47 rad>s. Compute the magnitude of the steady-state response and compare it to the magnitude of the forced response of an undamped system. 2.33. Compute the response of the system in Figure P2.33 if the system is initially at rest for the values k1 = 100 N>m, k2 = 500 N>m, c = 20 kg>s, and m = 89 kg. k2 k1 25 cos 3t m c Figure P2.33 2.34. Write the equation of motion for the system given in Figure P2.34 for the case that F(t) = F cos ω t and the surface is friction free. Does the angle θ affect the magnitude of oscillation? c k m  F(t) Figure P2.34 2.35. A foot pedal for a musical instrument is modeled by the sketch in Figure P2.35: k = 2000 N>m, c = 25 kg>s, m = 25 kg, and F(t) = 50 cos 2πt N. Compute the 204 Response to Harmonic Excitation Chap. 2 steady-state response assuming the system starts from rest. Also use the small-angle approximation. F(t) 0.05 m 0.05 m 0.05 m m k c Figure P2.35 2.36. Consider the system of Problem 2.15, repeated here as Figure P2.36 with the effects of damping indicated. The physical constants are J = 25 kg m2, k = 2000 Nm>rad, and the applied moment is 5 Nm at 1.432 Hz acting through the distance r = 0.5 m. Compute the magnitude of the steady-state response if the measured damping ratio of the spring system is ζ = 0.01. Compare this to the response for the case where the damping is not modeled (ζ = 0). r k c J  u(t)  M0 cos t Figure P2.36 Model of an airfoil in a wind tunnel including the effects of damping. 2.37. A machine, modeled as a linear spring–mass–damper system, is driven at resonance (ωn = ω = 2 rad >s). Design a damper (that is, choose a value of c) such that the maximum deflection at steady state is 0.05 m. The machine is modeled as having a stiffness of 2000 kg >m, and the excitation force has a magnitude of 100 N. 2.38. Derive the total response of the system to initial conditions x0 and 0 using the homogenous solution in the form xh(t) = e -ζωnt(A1 sin ωdt + A2 cos ωdt) and hence verify equation (2.38) for the forced response of an underdamped system. Section 2.3 (Problems 2.39 through 2.44) 2.39. Referring to Figure 2.11, draw the solution for the magnitude X for the case m = 100 kg, c = 4000 N s>m, and k = 10,000 N>m. Assume that the system is driven at resonance by a 10-N force. Chap. 2 205 Problems 2.40. Use the graphical method to compute the phase shift for the system with m = 100 kg, c = 4000 N s>m, k = 10,000 N>m, and F0 = 10 N, if ω = ωn >2 and again for the case ω = 2ωn. 2.41. A body of mass 100 kg is suspended by a spring of stiffness of 30 kN>m and dashpot of damping constant 1000 N s>m. Vibration is excited by a harmonic force of amplitude 80 N and a frequency of 3 Hz. Calculate the amplitude of the displacement for the vibration and the phase angle between the displacement and the excitation force using the graphical method. 2.42. Calculate the real part of equation (2.55) xp(t) = 31 k - mω22 2 + (cω)2 4 1>2 F0 e j(ωt - θ) to verify that this is consistent with the equation (2.36) Xp = 31 2 f0 ω2n - ω 2 2 + (2ζωnω)2 and hence establish the equivalence of the exponential approach to solving the damped vibration problem with method of undetermined coefficients. 1 ms2 2.43. Referring to equation (2.56) + cs + k 2 X(s) = F0s s 2 + ω2 and a table of Laplace transforms (see Appendix B), calculate the solution x(t) by using a table of Laplace transform pairs, and show that the solution obtained this way is equivalent to (2.36). 2.44. Solve the following system using the Laplace transform method and the table in Appendix B: $ # mx(t) + kx(t) = F0 cos ωt, x(0) = x0, x(0) = v0 Check your solution against equation (2.11) obtained via the method of undetermined coefficients. Section 2.4 (Problems 2.45 through 2.60) 2.45. For a base motion system described by $ # mx + cx + kx = cYωb cos ωbt + kY sin ωbt with m = 100 kg, c = 50 kg>s, k = 1000 N>m, Y = 0.03 m, and ωb = 3 rad>s, compute the magnitude of the particular solution. Last, compute the transmissibility ratio. 2.46. For a base motion system described by $ # mx + cx + kx = cYωb cos ωbt + kY sin ωbt with m = 100 kg, c = 50 N>m, Y = 0.03 m, and ωb = 3 rad>s, find largest value of the stiffness k and that makes the transmissibility ratio less than 0.75. 206 Response to Harmonic Excitation Chap. 2 2.47. A machine weighing 2000 N rests on a support as illustrated in Figure P2.47. The support deflects about 5 cm as a result of the weight of the machine. The floor under the support is somewhat flexible and moves, because of the motion of a nearby machine, harmonically near resonance (r = 1) with an amplitude of 0.2 cm. Model the floor as base motion, assume a damping ratio of ζ = 0.01, and calculate the transmitted force and the amplitude of the transmitted displacement. Rubber mount modeled as a stiffness k and a damper c Machine of mass m   Static deflection c k Flexible floor y(t) Figure P2.47 2.48. Derive equation (2.70) from (2.68) xp(t) = ωnY c X = Yc 1 2 11 2 1 + (2ζr)2 - r 2 2 + (2ζr) ω2n + (2ζωb)2 ω2n - ω2b 2 + (2ζωnωb) to see if the author has done it correctly. 2 d 2 d 1>2 1>2 cos (ωbt - θ1 - θ2) 2.49. From the equation describing Figure 2.14, show that the point ( 12, 1) corresponds to the value TR 7 1 (i.e., for all r 6 12, TR 7 1). 2.50. Consider the base-excitation problem for the configuration shown in Figure P2.50. In this case, the base motion is a displacement transmitted through a dashpot or pure damping element. Derive an expression for the force transmitted to the support in steady state. Support k x(t) m c Figure P2.50 y(t)  Y sin bt Chap. 2 207 Problems 2.51. A very common example of base motion is the single-degree-of-freedom model of an automobile driving over a rough road. The road is modeled as providing a base motion displacement of y(t) = (0.01)sin(5.818t) m. The suspension provides an equivalent stiffness of k = 3.273 * 104 N>m, a damping coefficient of c = 231 kg>s, and a mass of 1007 kg. Determine the amplitude of the absolute displacement of the automobile mass. 2.52 A vibrating mass of 300 kg mounted on a massless support by a spring of stiffness 40,000 N>m and a damper of unknown damping coefficient is observed to vibrate with a 10-mm amplitude while the support vibration has a maximum amplitude of only 2.5 mm (at resonance). Calculate the damping constant and the amplitude of the force on the base. 2.53. Referring to Example 2.4.2, at what speed does Car 1 experience resonance? At what speed does Car 2 experience resonance? Calculate the maximum deflection of both cars at resonance. 2.54. For cars of Example 2.4.2, calculate the best choice of the damping coefficient so that the transmissibility is as small as possible by comparing the magnitude of ζ = 0.01, ζ = 0.1, and ζ = 0.2 for the case r = 2. What happens if the road “frequency” changes? 2.55. A system modeled by Figure 2.13, has a mass of 225 kg with a spring stiffness of 3.5 * 104 N >m. Calculate the damping coefficient, given that the system has a deflection (X) of 0.7 cm when driven at its natural frequency while the base amplitude (Y) is measured to be 0.3 cm. 2.56. Consider Example 2.4.2 for Car 1 illustrated in Figure P2.56 if three passengers totaling 200 kg are riding in the car. Calculate the effect of the mass of the passengers on the deflection at 20, 80, 100, and 150 km>h. What is the effect of the added passenger mass on Car 2? mp m c k y(t) Figure P2.56 A model of a car suspension with the mass of the occupants, mp, included. 2.57. Consider Example 2.4.2. Choose values of c and k for the suspension system for Car 2 (the sedan) such that the amplitude transmitted to the passenger compartment is as small as possible for the 1-cm bump at 50 km>h. Also calculate the deflection at 100 km>h for your values of c and k. 208 Response to Harmonic Excitation Chap. 2 2.58. Consider the base motion problem of Figure 2.13. (a) Compute the damping ratio needed to keep the displacement magnitude transmissibility less than 0.55 for a frequency ratio of r = 1.8. (b) What is the value of the force transmissibility ratio for this system? 2.59. Consider the effect of variable mass on an aircraft landing suspension system by modeling the landing gear as a moving base problem similar to that shown in Figure P2.56 for a car suspension. The mass of a regional jet is 13,236 kg empty and its maximum takeoff mass is 21,523 kg. Compare the maximum deflection for a wheel motion of magnitude 0.50 m and frequency of 35 rad>s for these two different masses. Take the damping ratio to be ζ = 0.1 and the stiffness to be 4.22 * 106 N>m. 2.60. Consider the simple model of a building subject to ground motion suggested in Figure P2.60. The building is modeled as a single-degree-of-freedom spring–mass system where the building mass is lumped atop two beams used to model the walls of the building in bending. Assume the ground motion is modeled as having amplitude of 0.1 m at a frequency of 7.5 rad>s. Approximate the building mass by 105 kg and the stiffness of each wall by 3.519 * 106 N>m. Compute the magnitude of the deflection of the top of the building. x(t) m k 3EI l3 k 3EI l3 y(t) Figure P2.60 A simple model of a building subject to ground motion, such as an earthquake. Section 2.5 (Problems 2.61 through 2.68) 2.61. A lathe can be modeled as an electric motor mounted on a steel table. The table plus the motor have a mass of 50 kg. The rotating parts of the lathe have a mass of 5 kg at a distance 0.1 m from the center. The damping ratio of the system is measured to be ζ = 0.06 (viscous damping) and its natural frequency is 7.5 Hz. Calculate the amplitude of the steady-state displacement of the motor, assuming ωr = 30 Hz. 2.62. The system of Figure 2.19 produces a forced oscillation of varying frequency. As the frequency is changed, it is noted that at resonance the amplitude of the displacement is 10 mm. As the frequency is increased several decades past resonance, the amplitude of the displacement remains fixed at 1 mm. Estimate the damping ratio for the system. 2.63. An electric motor (Figure P2.63) has an eccentric mass of 10 kg (10% of the total mass of 100 kg) and is set on two identical springs (k = 3200>m). The motor runs at 1750 rpm, and the mass eccentricity is 100 mm from the center. The springs are mounted 250 mm apart with the motor shaft in the center. Neglect damping and determine the amplitude of the vertical vibration. Chap. 2 209 Problems m e e  0.1 m r t k  3200 N/m m  10 kg k k 125 mm Figure P2.63 A vibration model for an electric motor with an unbalance. 125 mm 2.64. Consider a system with rotating unbalance as illustrated in Figure P2.63. Suppose the deflection at 1750 rpm is measured to be 0.05 m and the damping ratio is measured to be ζ = 0.1. The out-of-balance mass is estimated to be 10%. Locate the unbalanced mass by computing e. 2.65. A fan of 45 kg has an unbalance that creates a harmonic force. A spring-damper system is designed to minimize the force transmitted to the base of the fan. A damper is used having a damping ratio of ζ = 0.2. Calculate the required spring stiffness so that only 10% of the force is transmitted to the ground when the fan is running at 10,000 rpm. 2.66. Plot the normalized displacement magnitude versus the frequency ratio for the out-ofbalance problem (i.e., repeat Figure 2.21) for the case of ζ = 0.05. 2.67. Consider a typical unbalanced machine problem as given in Figure P2.67 with a machine mass of 120 kg, a mount stiffness of 800 kN>m, and a damping value of 500 kg>s. Guide Guide x(t) Machine of total mass m m0 e k rt Rubber floor mounting modeled as a spring and a damper Figure P2.67 c A typical unbalance machine problem. Friction-free surface 210 Response to Harmonic Excitation Chap. 2 The out-of-balance force is measured to be 374 N at a running speed of 3000 rev>min. (a) Determine the amplitude of motion due to the out-of-balance force. (b) If the outof-balance mass is estimated to be 1% of the total mass, estimate the value of e. 2.68. Plot the response of the mass in Problem 2.67 assuming zero initial conditions. Section 2.6 (Problems 2.69 through 2.72) 2.69. Calculate damping and stiffness coefficients for the accelerometer of Figure 2.24 with moving mass of 0.04 kg such that the accelerometer is able to measure vibration between 0 and 50 Hz within 5%. (Hint: For an accelerometer it is desirable for Z>ω2bY = constant.) 2.70. The damping constant for a particular accelerometer of the type illustrated in Figure 2.26 is 50 N s>m. It is desired to design the accelerometer (i.e., choose m and k) for a maximum error of 3% over the frequency range 0 to 75 Hz. 2.71. The accelerometer of Figure 2.24 has a natural frequency of 120 kHz and a damping ratio of 0.2. Calculate the error in measurement of a sinusoidal vibration at 60 kHz. 2.72. Design an accelerometer (i.e., choose m, c, and k) configured as in Figure 2.24 with very small mass that will be accurate to 1% over the frequency range 0 to 50 Hz. Section 2.7 (Problems 2.73 through 2.89) 2.73. Consider a spring–mass sliding along a surface providing Coulomb friction, with stiffness 1.2 * 104 N>m and mass 10 kg, driven harmonically by a force of 50 N at 10 Hz. Calculate the approximate amplitude of steady-state motion assuming that both the mass and the surface that it slides on are made of lubricated steel. 2.74. A spring–mass system with Coulomb damping of 10 kg, stiffness of 2000 N>m, and coefficient of friction of 0.1 is driven harmonically at 10 Hz. The amplitude at steady state is 5 cm. Calculate the magnitude of the driving force. 2.75. A system of mass 10 kg and stiffness 1.5 * 104 N>m is subject to Coulomb damping. If the mass is driven harmonically by a 90-N force at 25 Hz, determine the equivalent viscous-damping coefficient if the coefficient of friction is 0.1. 2.76. a. Plot the free response of the system of Problem 2.75 to initial conditions of x(0) = 0 # and x(0) = 0 F0 >m 0 = 9 m>s using the solution in Section 1.10. b. Use the equivalent viscous-damping coefficient calculated in Problem 2.75 and plot the free response of the “equivalent” viscously damped system to the same initial conditions. 2.77. Referring to the system of Example 2.7.1; a spring–mass system with sliding friction described by equation (2.97) with stiffness k = 1.5 * 104 N>m, driving harmonically a 10-kg mass by a force of 90 N at 25 Hz, calculate how large the magnitude of the driving force must be to sustain motion if the steel is lubricated. How large must this magnitude be if the lubrication is removed? 2.78. Calculate the phase shift between the driving force and the response for the system of Problem 2.77 using the equivalent viscous-damping approximation. 2.79. Derive the equation of vibration for the system of Figure P2.79 assuming that a viscous dashpot of damping constant c is connected in parallel to the spring. Calculate the energy loss and determine the magnitude and phase relationships for the forced response of the equivalent viscous system. Chap. 2 211 Problems k x(t) m c F0 cos t Coulomb friction Figure P2.79 2.80. A system of unknown damping mechanism is driven harmonically at 10 Hz with an adjustable magnitude. The magnitude is changed, and the energy lost per cycle and amplitudes are measured for five different magnitudes. The measured quantities are ΔE(J) 0.25 0.45 0.8 1.16 3.0 X(M) 0.01 0.02 0.04 0.08 0.15 Is the damping viscous or Coulomb? 2.81. Calculate the equivalent loss factor for a system with Coulomb damping. 2.82. A spring–mass system (m = 10 kg, k = 4 * 103 N>m) vibrates horizontally on a surface with coefficient of friction μ = 0.15. When excited harmonically at 5 Hz, the steady-state displacement of the mass is 5 cm. Calculate the amplitude of the harmonic force applied. 2.83. Calculate the displacement for a system with air damping using the equivalent viscousdamping method. 2.84. Calculate the semimajor and semiminor axis of the ellipse of equation (2.119). Then calculate the area of the ellipse. Use c = 10 kg>s, ω = 2 rad>s, and X = 0.01 m. 2.85. The area of a force deflection curve of Figure 2.29 is measured to be 2.5 N # m, and the maximum deflection is measured to be 8 mm. From the “slope” of the ellipse, the stiffness is estimated to be 5 * 104 N>m. Calculate the hysteretic damping coefficient. What is the equivalent viscous damping if the system is driven at 10 Hz? 2.86. The area of the hysteresis loop of a hysterically damped system is measured to be 5 N # m and the maximum deflection is measured to be 1 cm. Calculate the equivalent viscous-damping coefficient for a 20-Hz driving force. Plot ceq,versus ω for 2π … ω … 100π rad>s. 2.87. Calculate the nonconservative energy of a system subject to both viscous and hysteretic damping. 2.88. Derive a formula for equivalent viscous damping for the damping force of the form, # Fd = c(x)n, where n is an integer. 2.89. Using the equivalent viscous-damping formulation, determine an expression for the steady-state amplitude under harmonic excitation for a system with both Coulomb and viscous damping present. Section 2.8 (Problems 2.90 through 2.96) *2.90. Numerically integrate and plot the response of an underdamped system determined by m = 100 kg, k = 20,000 N>m, and c = 200 kg>s, subject to the initial conditions of 212 Response to Harmonic Excitation Chap. 2 x0 = 0.01 m and 0 = 0.1 m>s, and the applied force F(t) = 150 cos 5t. Then plot the exact response as computed by equation (2.33). Compare the plot of the exact solution to the numerical simulation. *2.91. Numerically integrate and plot the response of an underdamped system determined by m = 150 kg and k = 4000 N>m, subject to the initial conditions of x0 = 0.01 m and v0 = 0.1 m>s, and the applied force F(t) = 15 cos 10t, for various values of the damping coefficient. Use this “program” to determine a value of damping that causes the transient term to die out within 3 seconds. Try to find the smallest such value of damping remembering that added damping is usually expensive. *2.92. Compute the total response of a spring–mass system with values k = 1000 N > m, m = 10 kg, subject to a harmonic force of magnitude F0 = 100 N and frequency of 8.162 rad > s, and initial conditions given by x0 = 0.01 m and 0 = 0.01 m > s, by numerically integrating rather than using analytical expressions, as was done in Problem 2.7. Plot the response. *2.93. A foot pedal for a musical instrument is modeled by the sketch in Figure P2.93. With k = 2000 N>m, c = 25 kg>s, m = 25 kg, and F(t) = 50 cos 2πt N, numerically simulate the response of the system assuming the system starts from rest. Use the small-angle approximation. F(t) 0.05 m 0.05 m 0.05 m m k c Figure P2.93 *2.94. Numerically integrate and plot the response of an underdamped system determined by m = 100 kg, k = 2000 N>m, and c = 200 kg>s, subject to the applied force F(t) = 150 cos 10t for the following sets of initial conditions: (a) x0 = 0.0 m and 0 = 0.1 m>s (b) x0 = 0.01 m and 0 = 0.0 m>s (c) x0 = 0.05 m and 0 = 0.0 m>s (d) x0 = 0.0 m and 0 = 0.5 m>s Plot these responses on the same graph and note the effects of the initial conditions on the transient part of the response. *2.95. A DVD drive is mounted on a chassis and is modeled as a single-degree-of-freedom spring, mass, and damper. During normal operation, the drive (having a mass of 0.4 kg) is subject to a harmonic force of 1 N at 10 rad>s. Because of material considerations and static deflection, the stiffness is fixed at 500 N>m and the natural damping in the system is 10 kg>s. The DVD player starts and stops during its normal operation providing initial conditions to the module of x0 = 0.001 m and 0 = 0.5 m>s. The DVD drive must not have an amplitude of vibration larger then 0.008 m even during the transient Chap. 2 213 Problems stage. First, compute the response by numerical simulation to see if the constraint is satisfied. If the constraint is not satisfied, find the smallest value of damping that will keep the deflection less than 0.008 m. 2.96. Use a plotting routine to examine the base motion problem (see Figure 2.13) by plotting the particular solution (for an undamped system) for the three cases k = 1500 N>m, k = 2500 N>m, and k = 700 N>m. Also note the values of the three frequency ratios and the corresponding amplitude of vibration of each case compared to the input. Use the following values: ωb = 4.4 rad>s, m = 100 kg, and Y = 0.05 m. Section 2.9 (Problems 2.97 through 2.102) *2.97. Compute the response of the system in Figure P2.93 for the case that the damping is linear viscous, the spring is a nonlinear soft spring of the form k(x) = kx - k1x3 and the system is subject to a harmonic excitation of 300 N at a frequency of approximately one third the natural frequency (ω = ωn >3) and initial conditions of x0 = 0.01 m and 0 = 0.1 m>s. The system has a mass of 100 kg, a damping coefficient of 170 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 10,000 N>m3. Compute the solution and compare it to the linear solution (k1 = 0). Which system has the largest magnitude? *2.98. Compute the response of the system in Figure P2.97 for the case that the damping is linear viscous, the spring is a nonlinear hard spring of the form k(x) = kx + k1x3 and the system is subject to a harmonic excitation of 300 N at a frequency equal to the natural frequency (ω = ωn) and initial conditions of x0 = 0.01 m and 0 = 0.1 m>s. The system has a mass of 100 kg, a damping coefficient of 170 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 10,000 N>m3. Compute the solution and compare it to the linear solution (k1 = 0). Which system has the largest magnitude? x (t) k m F (t) c Figure P2.97 *2.99. Compute the response of the system in Figure P2.97 for the case that the damping is linear viscous, the spring is a nonlinear soft spring of the form k(x) = kx - k1x3 and the system is subject to a harmonic excitation of 300 N at a frequency equal to the natural frequency (ω = ωn) and initial conditions of x0 = 0.01 m and 0 = 0.1 m>s. The system has a mass of 100 kg, a damping coefficient of 15 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 100 N>m3. Compute the solution and compare it to the hard spring solution (k(x) = kx + k1x3). 214 Response to Harmonic Excitation Chap. 2 *2.100. Compute the response of the system in Figure P2.97 for the case that the damping is linear viscous, the spring is a nonlinear soft spring of the form k(x) = kx - k1x3 and the system is subject to a harmonic excitation of 300 N at a frequency equal to the natural frequency (ω = ωn) and initial conditions of x0 = 0.01 m and 0 = 0.1 m>s. The system has a mass of 100 kg, a damping coefficient of 15 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 1000 N>m3. Compute the solution and compare it to the quadratic soft spring (k(x) = kx + k1x2). *2.101. Compare the forced response of a system with velocity-squared damping with equation of motion given by $ # # mx + α sgn(x)x2 + kx = F0 cos ωt using numerical simulation of the nonlinear equation to that of the response of the linear system obtained using equivalent viscous damping as defined by equation (2.131) ceq = 8 αωX 3π Use as initial conditions, x0 = 0.01 m and 0 = 0.1 m>s with a mass of 10 kg, stiffness of 25 N>m, applied force of 150 cos (ωnt), and drag coefficient of α = 250. *2.102. Compare the forced response of a system with structural damping (see Table 2.2) using numerical simulation of the nonlinear equation to that of the response of the linear system obtained using equivalent viscous damping as defined in Table 2.2. Use as initial conditions, x0 = 0.01 m and 0 = 0.1 m>s with a mass of 10 kg, stiffness of 25 N>m, applied force of 150 cos (ωnt), and solid damping coefficient of b = 25. MATLAB® ENGINEERING VIBRATION TOOLBOX If you did not use the Engineering Vibration Toolbox for Chapter 1, refer to that section for information regarding using MATLAB files or refer to Appendix G. The files for Chapter 2, entitled VTB2_1, VTB2_2, and so on, can be found in folder VTB2. The files in VTB2 can be used to help solve the preceding homework problems and to help gain information about the nature of the response of single-degree-of-freedom systems to harmonic inputs. The following problems are intended to help you gain some experience with the concepts in this chapter. TOOLBOX PROBLEMS TB2.1. Using file VTB2_1, reproduce Figure 2.2. TB2.2. Carefully investigate the response of an undamped system near resonance by trying several values of ω near ωn for the values of Figure 2.2. Do you get the beats of Figure 2.3? Chap. 2 Problems 215 TB2.3. Using file VTB2_3, reproduce Figure 2.9. Also plot Xk>f0 versus r for the values given in Example 2.2.3 and plot the associated time response, xp(t), for a value of r = 0.5 using VTB2_2. Do these plots again for ζ = 0.01 and ζ = 0.1 and comment on how the time response changes as the damping ratio, ζ, changes by an order of magnitude. TB2.4. Using file VTB2_5 for rotating unbalance, make a plot of x versus r for the helicopter of Example 2.4.2. TB2.5. Using file VTB2_6 for damping mechanisms, compare the time response of a system (with physical parameters of m = 10, k = 100, α = 0.05, X = 1) with air damping as given by equation (2.129) with initial conditions x0 = 1 and 0 = 0 to that of an equivalent viscously damped system using equation (2.131) for an input of 10 sin 3t. 3 General Forced Response This chapter starts out by considering the response to systems subject to a shock loading or impulse. an example of such a load occurs during landing an airplane. The aircraft landing gear, pictured at the top, provides stiffness and damping designed (see Chapter 5) to mitigate the effect of the shock on the aircraft. The input to the structure is not completely periodic, as examined in Chapter 2, but has random shock and other components, as discussed in this chapter. another source of vibration that is not at a single frequency (as in Chapter 2) is the human heart. The heart vibrates at a variety of different rates depending on the level of activity and emotional state of the person. Some hearts need regulating by using a device such as the pacemaker, pictured at the bottom. These are run by batteries, which require replacement every seven to ten years, involving major surgery. Vibration researchers have recently developed energy harvesting devices that transduce the heart-induced, chest-cavity vibrations into electrical energy. This energy is then used to recharge the pacemaker battery. Fitting such a device inside the pacemaker required a basic understanding of vibration and, in particular models, of the vibration response of a structure (the harvester in this case) to inputs that have energy at many different frequencies, as discussed in this chapter. 216 Sec. 3.1 Impulse Response Function 217 In Chapter 2 the forced response of a single-degree-of-freedom system was considered for the special case of a harmonic driving force. Harmonic excitation refers to an applied force that is sinusoidal of a single frequency. In this chapter, the response of a system to a variety of different types of forces is considered, as well as a general formulation for calculating the forced response for any type of applied force. If the system considered is linear, the principle of superposition can be used to calculate the response to various combinations of forces based on the individual response to a specific force. Superposition refers to the fact that for a linear equation of motion, say $ x + ω2nx = 0, if x1 and x2 are both solutions of the equation, then x = a1x1 + a2x2 is also a solution where a1 and a2 are any constants. This concept also implies that if $ $ x1 is a particular solution to x + ω2nx = f1 and x2 is a solution to x + ω2nx = f2, then $ 2 x1 + x2 is a solution of x + ωnx = f1 + f2. Thus this method of superposition can be used to construct the solution to a complicated forcing function by solving a series of simpler problems. Superposition was used to solve the base-excitation problem of Section 2.3. The principle of superposition in linear systems is a very powerful technique and is used extensively. A variety of forces are applied to mechanical systems that result in vibration. Earthquake forces are sometimes modeled as sums of decaying periodic or harmonic forces. High winds can be a source of impulsive or step loadings to structures. Rough roads provide a variety of forcing conditions to automobiles. The ocean waves and wind provide forces to ships at sea. Various manufacturing processes produce applied forces that are random, periodic, nonperiodic, or transient in nature. Air and relative motion provide forces to the wing of an aircraft that can cause it to oscillate. All of these forces can cause vibration. Periodic forces are those that repeat in time. An example is an applied force consisting of the sum of two harmonic forces at different frequencies. A nonperiodic force is one that does not repeat itself in time. A step function is an example of a force that is a nonperiodic excitation. A transient force is one that reduces to zero after a finite, usually small, time. An impulse or a shock are examples of transient excitations. All of the aforementioned classes of excitation are deterministic (i.e., they are known precisely as a function of time). On the other hand, a random excitation is one that is unpredictable in time and must be described in terms of probability and statistics. This chapter introduces a sample of these various classes of force excitations and how to calculate and analyze the resulting motion when applied to a single-degree-of-freedom spring–mass–damper system. 3.1 IMPULSE RESPONSE FUNCTION A very common source of vibration is the sudden application of a short-duration force called an impulse. An impulse excitation is a force that is applied for a very short, or infinitesimal, length of time and represents one example of a shock loading. 218 general Forced Response Chap. 3 An impulse is a nonperiodic force. The response of a system to an impulse is identical to the free response of the system to certain initial conditions, as shown in this section. In many useful situations the applied force F(t) is impulsive in nature (i.e., acts with large magnitude for a very short period of time). First, consider a mathematical model of an impulse excitation. A graphical time history of a model of the impulse is given in Figure 3.1. This is a rectangular pulse of very large magnitude and very small width (duration). F(t) F 2 0 t    Figure 3.1 The time history of an impulse force used to model impulsive loading consisting of a large magnitude applied over a short time interval. The rule of describing the force in Figure 3.1 is stated symbolically as t … τ - ε 0 Fn F(t) = d 2ε 0 τ - ε 6 t 6 τ + ε (3.1) t Ú τ + ε where ε is a small positive number. This simple rule, F(t), can be integrated to define the impulse. The impulse of the force F(t) is defined by the integral, denoted by I(ε), by τ+ε I(ε) = Lτ - ε F(t) dt which provides a measure of the strength of the forcing function, F(t). Since the rule, F(t), is zero outside the time interval from τ - ε to τ + ε, the limits of integration on I(ε) can be extended to yield ∞ I(ε) = L-∞ F(t) dt (3.2) which has the units of N # s. In this case, the integral of equation (3.2) is evaluated by calculating the area under the curve using equation (3.1), which becomes ∞ I(ε) = L-∞ F(t)dt = Fn 2ε = Fn 2ε (3.3) Sec. 3.1 219 Impulse Response Function independent of the value of ε as long as ε ∙ 0. In the limit as ε S 0 (but ε ∙ 0), the integral takes the value I(ε) = Fn. This is used to define the impulse function as the function F(t) with the two properties F(t - τ) = 0 t ∙ τ (3.4) and ∞ F(t - τ) dt = Fn (3.5) L-∞ If the magnitude of Fn is unity, this becomes the definition of the unit impulse function, denoted by δ(t), also called the Dirac delta function (Boyce and DiPrima, 2009). The solution for response of the single-degree-of-freedom system (see Window 3.1) $ # # mx(t) + cx(t) + kx(t) = Fn(t), x(0) = 0, x(0) = 0 to an impulsive load for the system initially at rest is calculated by recalling from physics that an impulse imparts a change in momentum to a body. For the sake Window 3.1 Review of the Free Response of the Single-Degree-of-Freedom System of Chapter 1 .. . mx  cx  kx  F(t) . x(0)  vo x(0)  x0 .. . x  2nx  2n x  f(t) x(t) c m F(t) k This system has free response [i.e., f(t)  0] in the underdamped case (i.e., 0    1) given by x(t)  (0  nx0)2  (x0d)2 n 12 ent sin (dt  ) where d  n 12 and   tan1 x0d 0  nx0 Here n  km,   c(2mn), and 0    1 must hold for the preceding solution to be valid [from equations (1.36) to (1.38)]. 220 general Forced Response Chap. 3 of simplicity, take τ = 0 in the definition of an impulse. Consider the mass to be at rest just prior to the application of an impulse force. This instant of time is denoted by 0-. Likewise the instant of time just after t = 0 is denoted by 0+. The # initial conditions are both zero, so that x(0-) = x(0-) = 0, since the system is # initially at rest. However, the velocity just after the impulse is x(0+), denoted here # + # as 0. Thus the change in momentum at impact is mx(0 ) - mx(0-) = mv0, so n that F = F∆t = mv0 - 0 = mv0, while the initial displacement remains at zero. By this physical line of thought, an impulse applied to a single-degree-of-freedom spring–mass–damper system is the same as applying the initial conditions of zero displacement and an initial velocity of 0 = F∆t>m. Referring to Window 3.1, the response of an underdamped single-degree-offreedom system (0 6 ζ 6 1) with zero initial displacement (x0 = 0) is just 0 -ζωnt x(t) = ωd e sin ωdt Substitution of v0 = Fn >m (Fn = F∆t, with units of N s) into this last expression yields x(t) = Fne -ζωnt sin ωdt mωd (3.6) as predicted by equations (1.36) and (1.38), repeated in Window 3.1. It is convenient to write this solution in the form x(t) = Fnh(t) (3.7) where h(t) is defined by h(t) = 1 -ζωnt e sin ωdt mωd (3.8) Note that the function h(t) is the response to a unit impulse applied at time t = 0. If applied at time t = τ, τ ∙ 0, this can also be written as (replace t in the foregoing with t - τ) h(t - τ) = 1 -ζωn(t - τ) e sin ωd(t - τ) mωd t 7 τ (3.9) and zero for the interval 0 6 t 6 τ. The functions h(t) and h(t - τ) are each called the impulse response function of the system. While the impulse is a mathematical abstraction of an infinite force applied over an infinitesimal time, in applications it presents an excellent model of a large force applied over a short period of time. The impulse response is physically interpreted as the response to an initial velocity with no initial displacement (hence no phase shift for τ = 0). The impulse response function, combined with the principle of superposition, is also useful for calculating the response of a system to a general applied force excitation as discussed in Section 3.2. Sec. 3.1 221 Impulse Response Function A common occurrence that causes an impulse excitation is an impact. In vibration testing, a mechanical device under test is often given an impact, and the response is measured to determine the system’s vibration properties. The impact is often created by hitting the test specimen with a hammer containing a device for measuring the force of the impact. Use of the impulse response for vibration testing is also discussed in Chapter 7. In practice, a force is considered to be an impulse if its duration (∆t) is very short compared with the period, T = 2π>ωn, associated with the structure’s undamped natural frequency. In typical vibration tests, ∆t is on the order of 10 - 3 s. Consider a spring–mass–damper system with m = 100 kg, c = 20 kg >s, and k = 2000 N > m with an impulse force applied to it of 1000 N for 0.01 s. Compute the resulting response. Example 3.1.1 Solution A 1000 N force acting over 0.01 s provides (area under the curve) a value of Fn = F∆t = 1000 · 0.01 = 10 N # s. Using the values given, the equation of motion is $ # 100x(t) + 20x(t) + 2000x(t) = 10δ(t) Thus the natural frequency, damping ratio, and damped natural frequency are ωn = 20 2000 = 4.427 rad>s, ζ = = 0.022, B 100 21100 · 2000 ωd = 4.47231 - 0.0222 = 4.471 rad>s Using equation (3.6), the response becomes x(t) = Fne -ζωnt sin ωdt = 0.022e -0.1t sin (4.471t) mωd n Example 3.1.2 Suppose a 1-kg bird flies into the 3-kg security camera of Example 2.1.3, repeated in Figure 3.2. If the bird is flying at 72 kmph, compute the maximum deflection the impact causes based on the design given in Example 2.1.3. Does the maximum deflection violate the design constraint? Ignore damping. Solution From the design solution of Example 2.1.3, the stiffness of the camera’s mounting bracket is (I = bh3 >12): k = 3Ebh3 = 12l 3 ( 7.1 * 1010 N>m2 ) (0.02 m) ( 0.02 m ) 3 = 1.707 * 104 N>m 3 4 ( 0.55 m ) The mass of the camera is mc = 3 kg, so the natural frequency is 75.43 rad>s. Combining equations (3.7) and (3.8) for ζ = 0, the response is x(t) = mbv F∆t sin ωnt = sin ωnt mcωn mcωn 222 general Forced Response x(t) Wind Camera F(t) m x(t) Mounting bracket l Chap. 3 k 3EI l3 F(t) k 3EI l3 .. mx(t)  kx(t)  F(t) Figure 3.2 A vibration model of a security camera and mount. where mb is the linear momentum of the bird, which imparts the impact force F. The impulse is thus mbv = 1 kg · 72 km 1000 m hour · · = 20 kg · m>s hour km 3600 s This has maximum amplitude of X = ` 20 kg · m>s mbv F∆t ` = ` ` = ` ` = 0.088 m mcωn mcωn 3 kg · 75.43 rad>s Thus, the design constraint of holding the vibration of the camera within 0.01 m required in Example 2.1.3 is violated under a bird strike. n Example 3.1.3 In vibration testing, an instrumented hammer is often used to hit a device to excite it and to measure the impact force simultaneously. If the device being tested is a single-degreeof-freedom system, plot the response given that m = 1 kg, c = 0.5 kg>s, k = 4 N>m, and Fn = 0.2 N # s. It is often difficult to provide a single impact with a hammer. Sometimes a “double hit” occurs, so the exciting force may have the form F(t) = 0.2δ(t) + 0.1δ(t - τ) Plot the response of the same system with a double hit and compare it with the response to a single impact. Assume that the initial conditions are zero. Solution The solution to the single unit impact and time t = 0 is given by equations (3.7) and (3.8) with ω n = 14 = 2 rad>s and ζ = c>(2mωn) = 0.125. Thus, with Fn = 0.2δ(t) x1(t) = (1) a231 - (0.125) b 0.2 e -(0.125)(2)t sin 231 - ( 0.125 ) 2t 2 = 0.1008e -0.25t sin(1.984t)m 223 Impulse Response Function Load cell signal Impact hammer f(t) Test structure m c x(t) k 0.15 0.1 Displacement (m) Sec. 3.1 0.05 0 2 4 6 8 10 0.05 0.1 Time (s) (a) f(t)(N) Area 2 Area 1 t(s) 0 0.5 (b) Figure 3.3 (a) The response of a single-degree-of-freedom system to a single impact (solid line) and a double impact for τ = 0.5 s (dashed line). Plot (b) indicates the force-versus time curve for the applied double impact. Note the larger amplitude of the double hit. 224 general Forced Response Chap. 3 which is plotted in Figure 3.3(a). Similarly, the response to 0.1δ(t - τ) is calculated from equations (3.7) and (3.9) as x2(t) = 0.0504e -0.25(t - τ) sin 1.984(t - τ)m t 7 τ and x2(t) = 0 for 0 6 t 6 τ = 0.5. The force input f(t) is indicated in Figure 3.3(b). It is important to note that no contribution from x2 occurs until time t = τ. Using the principle of superposition for linear systems, the response to the “double impact” will be the sum of the preceding two impulse responses: x(t) = x1(t) + x2(t) = c 0.1008e -0.25t sin (1.984t) 0 6 t 6 τ 0.1008e -0.25t sin (1.984t) + 0.0504e -0.25 (t - τ) sin 1.984(t - τ) t 7 τ This is plotted in Figure 3.3(a) for the value τ = 0.5 s. Note that the obvious difference between the two responses is that the “doublehit” response has a “spike” at τ = t = 0.5, causing a larger amplitude. The time, τ, represents the time delay between the two hits. n The following example illustrates the calculation of the response due to both an applied impulse and initial conditions, forming the total response of the system. The example also introduces the concept of using a Heaviside step function to represent the response. Example 3.1.4 Consider the system (mass normalized) $ # x(t) + 2x(t) + 4x(t) = δ(t) - δ(t - 4) and compute and plot the response with initial conditions x0 = 1 mm and 0 = -1 mm>s. Solution By inspection, the natural frequency is ω n = 2rad>s. Examining the velocity coefficient yields 2 = 2ζωn or ζ = 0.5 Thus, the system is underdamped and the response given in Window 3.1 applies. Computing the damped natural frequency yields ωd = ωn 31 - ζ2 = 2 1 2 1 - a b = 23 C 2 First, compute the response for the time interval 0 … t … 4 s. In this interval, only the first impulse is active. The corresponding impulse solution is, by equation (3.6), xI(t) = Fn -ζωnt 1 -t e sin ωdt = e sin 23t, mωd 23 0 … t 6 4 Sec. 3.1 225 Impulse Response Function The total solution for the first time interval is then equal to the sum of the homogeneous and impulse solutions. The homogeneous solution is xh(t) = e -t(A sin ωdt + B cos ωdt), 0 … t 6 4 where A and B are the constants of integration to be determined by the initial conditions and the subscript h denotes the solution due to the initial conditions. Differentiating the displacement yields the velocity: # xh(t) = - e -t(A sin23t + Bcos 23t) + e -t( 23A cos 13t - 23B sin 23t) Setting t = 0 in these last two expressions and using the initial conditions yields the following two equations: xh(0) = 1 = B # xh(0) = - 1 = - B + 23A Solving for A and B yields A = 0 and B = 1, so that xh(t) = e -t cos 23t. Next, compute the response due to the impulse at t = 0, which is equivalent to solving the initial # value problem for xI(0) = 0 and xI(0) = 1. Following the same procedure to compute the constants of integration for the impulse yields B = 1 and A = 1 23 so that xI(t) = e -t 23 sin 23t Adding the homogenous response and the impulse response yields x1(t) = e -t acos 23t + 1 23 sin23tb, 0 … t 6 4 Next, compute the response of the system to the second impulse, which starts at t = 4s. Using equation (3.9) with τ = 4s, the response to the second impulse is x2(t) = Fn -ζωn(t - τ) 1 -t + 4 e sin ωd(t - τ) = e sin23(t - 4), mωd 23 t 7 4 The Heaviside step function defined by Φ(t - τ) = e 0, 1, t 6 τ f t Ú τ is perfect for writing functions that “turn on” after some time has evolved. Heaviside functions are also denoted by H(t - τ). Using superposition, the total solution is 226 general Forced Response Chap. 3 Displacement, x(t) (mm) 1 0.5 0 2 4 6 8 10 0.5 Figure 3.4 at t = 4s. Time, t (s) A plot of displacement versus time for a double impact, with the second impact applied x = x1 + x2, and the Heaviside function is used to indicate that x2 “starts” after τ = 4. The solution can be written as x(t) = e -t acos 23t + sin23(t - 4) d Φ(t - 4) mm 23 23 This is plotted in Figure 3.4. Note the sharp change in the response as the second impact is applied. This is in contrast to the double hit in the previous example. In the previous example, the second impulse occurs in the same “direction” as the current response. However, in this example, the second impact occurs out of phase with the response of the first impact and causes an abrupt change in direction. n 1 sin23tb - c e -(t - 4) 3.2 RESPONSE TO AN ARBITRARY INPUT The response of a single-degree-of-freedom system to an arbitrary, general excitation is examined in this section. The response of a single-degree-of-freedom system to an arbitrary force of varying magnitude can be calculated from the concept of Sec. 3.2 227 Response to an Arbitrary Input F(t) F(t4) t  t/n t3 t1 t2 ti t t t4 Figure 3.5 An arbitrary excitation force F(t) split up into n impulse forces. the impulse response defined in Section 3.1. The procedure is to divide the exciting force up into impulses of infinitesimal area, calculate the responses to these individual impulses, and add the individual responses to calculate the total response using the concept of superposition. This is best shown in Figure 3.5, which illustrates an arbitrary applied force F(t) divided into n time intervals of length ∆t so that each time increment is defined by ∆t = t>n. At each time interval ti, the solution can be calculated by considering the response to be due to an impulse ∆t in duration and of force magnitude F(t) [i.e., an impulse of magnitude F(ti)∆t]. The part of the response due to the impulse acting during the time interval between ti and ti+1 is then given by equation (3.7) as the increment ∆x(t i) = F(t i)h(t - t i)∆t (3.10) so that the total response after n intervals is the sum x(t n) = a F(t i)h(t - t i)∆t n (3.11) i=1 This again uses the fact that the equation of motion is linear, so that the principle of superposition applies. Forming the sequence of partial sums and finding the limit as ∆t S 0 (n S ∞) yields x(t) = L0 t F(τ)h(t - τ)dτ (3.12) from the first fundamental theorem of integral calculus. The integral in equation (3.12) is called the convolution integral. A convolution integral is simply the integral of the product of two functions, one of which is shifted by the variable of integration. Convolution is used again in Section 3.4 as a useful technique in using transforms. Additional properties of the convolution integral are given in Window 3.2. For an underdamped single-degree-of-freedom system, the impulse response function h(t - τ) is given by equation (3.9). Substitution of the impulse response function of equation (3.9) into equation (3.12) then yields the result that the response of an underdamped system to an arbitrary input F(t) of the form $ # mx(t) + cx(t) + kx(t) = F(t), x0 = 0, v0 = 0 228 general Forced Response Chap. 3 Window 3.2 Useful Properties of the Convolution Integral Let α = t - τ, so that dα = -dτ for fixed t. Since τ ranges from 0 to t, α ranges from t to 0. Substitution of this change of variables into the definition x(t) = L0 t F(τ)h(t - τ)dτ yields Lt x(t) = - 0 F(t - α)h(α)dα = L0 t F(t - α)h(α)dα Thus, L0 t F(τ)h(t - τ)dτ = L0 t F(t - τ)h(τ)dτ is given by t x(t) = 1 -ζωnt e [F(τ)e ζωnτ sin ωd(t - τ)]dτ mωd L0 t = 1 F(t - τ)e -ζωnτ sin ω d τdτ mωd L0 (3.13) as long as the initial conditions are zero. The integral in equation (3.13) is a convolution integral where one of the functions is the impulse response function—hence the shift in the integral. A convolution integral used to compute a system response is called a Duhamel integral, after the French mathematician J. M. C. Duhamel (1797–1872). The Duhamel integral can be used to calculate the response to an arbitrary input as long as it satisfies certain mathematical conditions. The following example illustrates the procedure. Example 3.2.1 Consider an excitation force of the form given in Figure 3.6. The force is zero until time t0, when it jumps to a constant level, F0. This is called the step function, and when used to excite a single-degree-of-freedom system, it might model some machine operation or an automobile running over a surface that changes level (such as a curb). The step function Sec. 3.2 229 Response to an Arbitrary Input F(t) F0 t t0 Figure 3.6 Step function of magnitude F0 applied at time t = t0. of unit magnitude is called the Heaviside step function as defined in Example 3.1.4. Calculate the solution of 0, $ # mx + cx + kx = F(t) = e F0, t0 7 t 7 0 f t Ú t0 (3.14) with x0 = 0 = 0 and F(t) as described in Figure 3.6. Here it is assumed that the values of m, c, and k are such that the system is underdamped (0 6 ζ 6 1). Solution Applying the convolution integral given by equation (3.13) directly yields 0 1 -ζωnt e £ (0)e ζωnτ sin ωd(t - τ)dτ + F0e ζωnτ sin ωd(t - τ)dτ § mωd L0 Lt0 t x(t) = = t F0 -ζω t t ζω τ e n e n sin ωd(t - τ)dτ mωd Lt0 Using a table of integrals to evaluate this expression yields x(t) = F0 F0 e -ζωn(t - t0)cos[ωd(t - t 0) - θ] t Ú t 0 k k31- ζ2 (3.15) where θ = tan -1 ζ 31 - ζ2 (3.16) Note that if t0 = 0, equation (3.15) becomes just x(t) = F0 F0 e -ζωntcos(ω dt - θ) k k31 - ζ2 (3.17) and if there is no damping (ζ = 0), this expression simplifies further to x(t) = F0 (1 - cos ωnt) k (3.18) 230 general Forced Response Chap. 3 x(t) 0.06 O.S. 0.04 F0 k 0.02 0 t 0 2 4 6 8 10 12 14 ts tp Figure 3.7 The response of an underdamped system to the step excitation of Figure 3.6 for ζ = 0.1 and ωn = 3.16 rad >s (with F0 = 30 N, k = 1000 N >m, t0 = 0). Examining the damped response given in equation (3.17), it is obvious that for large time the second term of the response dies out and the steady state is just xss(t) = F0 k (3.19) In fact, the underdamped step response given by equations (3.15) and (3.17) consists of the constant function F0 >k minus a decaying oscillation, as illustrated in Figure 3.7. Often in the design of vibrating systems subject to a step input, the time it takes for the response to reach the largest value, called the time to peak and denoted tp in Figure 3.7, is used as a measure of the quality of the response. Other quantities used to measure the character of the step response are the overshoot, denoted by O.S. in Figure 3.7, which is the largest value of the response “over” the steady-state value, and the settling time, denoted by ts in Figure 3.7. The settling time is the time it takes for the response to get and stay within a certain percentage of the steady-state response. For the case of t0 = 0, tp and ts are given by tp = π >ωd and ts = 3.5>ζωn. The peak time is exact (see Problem 3.25), and the settling time is an approximation of when the response stays within 3% of the steady-state value. n Example 3.2.2 Another common excitation in vibration is a constant force that is applied for a short period of time and then removed. A rough model of such a force is given in Figure 3.8. Calculate the response of an underdamped system to this excitation. Solution This pulse-like loading can be written as a combination of step functions calculated in Example 3.2.1, as illustrated in Figure 3.8. The response of a singledegree-of-freedom system to F(t) = F1(t) + F2(t) is just the sum of the response to F1(t) and the response of F2(t), because the system is linear. First, consider the response of an underdamped system to F1(t). This response is just that calculated in Example 3.2.1 for t0 = 0 and given by equation (3.17). Next, consider the response of 231 Response to an Arbitrary Input F(t) F1(t)  F0 0 F2(t)  F0 0 t 0 t t1 0 t1 F0 Figure 3.8 Square-pulse excitation of magnitude F0 lasting for t1 seconds can be written as the sum of a step function starting at zero of magnitude F0 and a step function starting at t1 of magnitude - F0 [i.e., F(t) = F1(t) + F2(t)]. 0.04 Displacement (m) Sec. 3.2 t1 1.5 0.02 0 t1 0.1 0.02 0 2 4 6 8 10 Time (s) Figure 3.9 Response of an underdamped system to a pulse input of width t1. The dashed line is for t1 = 0.1 6 π > ωn and the solid line is for t1 = 1.5 7 π > ωn. Both plots are for the case F0 = 30 N, k = 1000 N >m, ζ = 0.1, and ωn = 3.16 rad >s. the system to F2(t). This is just the response given by equation (3.15) with F0 replaced by - F0 and t0 replaced by t1. Hence, subtracting equation (3.15) from equation (3.17) yields the result that the response to the pulse of Figure 3.8 is x(t) = F0e -ζωnt k21 - ζ2 {e ζωnt1 cos [ωd(t - t 1) - θ] - cos(ωdt - θ)}, t 7 t1 where θ is as defined in equation (3.16). A plot of this response is given in Figure 3.9 for different pulse widths t1. Note that the response is much different for t1 7 π>ωn and has a maximum magnitude of about five times the maximum magnitude of the time response for t1 6 π>ωn. Also, note that the steady-state response (i.e., the response for large time) is zero in this case. n 232 general Forced Response Chap. 3 Example 3.2.3 A load of dirt of mass md is dropped on the floor of a truck bed. The truck bed is modeled as a spring–mass–damper system (of values k, m, and c, respectively). The load is modeled as a force F(t) = md g applied to the spring–mass–damper system, as illustrated in Figure 3.10. This allows a crude analysis of the response of the truck’s suspension when the truck is being loaded. Calculate the vibration response of the truck bed, and compare the maximum deflection with the static load on the truck bed. F  md g x(t) m md x(t) k Truck being filled with dirt c g  9.8 m/s2 Vibration model (a) (b) Figure 3.10 A model of a truck being filled with a load of dirt of weight mdg and a vibration model that considers the mass of the dirt as an applied constant force. Solution In this case, the input force is just a constant [i.e., F(t) = md g], so that the equation of vibration becomes m g $ # mx + cx + kx = e d 0 t 7 0 t … 0 From equation (3.17) of Example 3.2.1, the response (let F0 = mdg) is just x(t) = mdg 1 £1 e -ζωntcos(ωdt - θ) § k 2 31 - ζ To obtain a rough idea about the nature of this expression, its undamped value is x(t) = md g (1 - cos ωnt) k which has a maximum amplitude (when t is such that cos ωnt = - 1) of xmax = 2 md g k This is twice the static displacement (i.e., twice the distance the truck would be deflected if the dirt were placed gently and slowly onto it). Thus, if the truck were designed with springs based only on the static load, with no margins of safety, the Sec. 3.2 233 Response to an Arbitrary Input springs in the truck would potentially break, or permanently deform, when subjected to the same mass applied dynamically (i.e., dropped) to the truck. Hence, it is important to consider the vibration (dynamic) response in designing structures that could be loaded dynamically. n It should be noted that the response of a single-degree-of-freedom system to an arbitrary input can be calculated numerically, even if the integral in equation (3.12) cannot be evaluated in closed form as done in the preceding examples. Such general numerical procedures, based roughly on equation (3.10), are discussed in Section 3.8, in which the numerical solutions discussed in Section 2.8 are applied. Numerical integration is often used to solve vibration problems with arbitrary forcing functions. The response calculations for a general external disturbance (input) force do not include the response that might exist because of nonzero initial conditions. The total response for an impulse disturbance with nonzero initial conditions is given in Example 3.1.4. The total response to an arbitrary input force as well as nonzero initial conditions for an undamped system is given using the convolution integral by x(t) = x0 cos ωnt + t v0 sin ωnt + h(t - τ)F(τ) dτ ωn L0 Note that the effect of the applied force on the homogeneous response is zero because the value of the convolution form of the particular solution and its derivative (velocity) are both zero at time t = 0. On the other hand, if the particular solution xp(t) or its derivative do not vanish at t = 0, the form of the total response is x(t) = A cos ωnt + B sin ωnt + xp(t) Here the constants A and B must be determined by the initial conditions and will be affected by the value of xp and its derivative at t = 0. In this last case, the constants of integration A and B become A = x0 - xp(0) and B = # v0 - xp(0) ωn A similar result holds for damped systems. The analytical calculations made with the convolution integral are not always easy to evaluate, and in many cases must be evaluated numerically. Laplace transform methods are often useful in convolution-type evaluations but, in practice, solutions are often found through numerical integration and simulation. While numerical simulation is used in practice, the concept of convolution is essential to understanding signal processing and for understanding the results of numerical simulations. 234 general Forced Response Chap. 3 Example 3.2.4 $ Solve x(t) + 16x(t) = cos 2t for the response to arbitrary initial conditions x0 and 0 using the convolution integral. Next, compare this to the result obtained by solving this problem using the method of undetermined coefficients explained in Section 2.1 and given in equation (2.11). Solution From the equation of motion, m = 1, ωn = 4, ω = 2, and F0 = f0 = 1, where the units are assumed to be consistent. Using the convolution expression, equation (3.12), the particular solution has the form t h(t - τ)F(τ)dτ L0 The impulse response function, h(t - τ), for an undamped system is found from equation (3.8) with ζ = 0. With the values given above for mass and frequency, the impulse response function is xp(t) = h(t - τ) = 1 sin (4t - 4τ) 4 Thus the convolution expression for the particular solution is t 1 sin (4t - 4τ) cos (2τ)dτ 4 L0 Integrating (using a symbolic code or repeated use of trig identities) yields xp(t) = xp(t) = cos (4t - 6τ) t 1 1 cos (4t - 2τ) a + b = ( cos 2t - cos 4t) 4 4 12 12 0 The total solution is of the form x(t) = A sin 4t + B cos 4t + 1 ( cos 2t - cos 4t) 12 Using the initial conditions to evaluate the constants of integration A and B yields x(0) = x0 = A sin (0) + B cos (0) + 1 ( cos (0) - cos (0)), 12 2 4 # sin (0) + sin (0) x(0) = v0 = 4A cos (0) + 4x0 sin (0) 12 12 Solving this set of equations for A and B yields A = 0 4 and B = x0 The total solution is then x(t) = v0 1 1 sin 4t + (x0 ) cos 4t + cos 2t 4 12 12 Sec. 3.3 235 Response to an Arbitrary Periodic Input This is in total agreement with the solution given in equation (2.11) with m = 1, ωn = 4, ω = 2, F0 = f0 = 1, and ω2n - ω2 = 12. The purpose of this is example is to show that two different methods yield the same answer, as they should. The method of undetermined coefficients is a much simpler calculation to make, but only works for harmonic forcing functions. The convolution approach is more complicated, but can be used for any forcing function and is thus a more general approach. n 3.3 RESPONSE TO AN ARBITRARY PERIODIC INPUT The specific case of periodic inputs is considered in this section. The response to periodic inputs can be calculated by the methods of Section 3.2. However, periodic disturbances that occur quite often merit special consideration. In Chapter 2, the response to a harmonic input is considered. The term harmonic input refers to a sinusoidal driving function at a single frequency. Here, the response to any periodic input is considered. A periodic function is any function that repeats itself in time [i.e., any function for which there exists a fixed time, T, called the period, such that f(t) = f(t + T) for all values of t]. A simple example is a forcing function that is the sum of two sinusoids of different frequency with a rational frequency ratio. An example of a general periodic forcing function, F(t), of period T is given in Figure 3.11. Note from the figure that the periodic force does not look periodic at all if examined in an interval less than the period T. However, the forcing function does repeat itself every T seconds. According to the theory developed by Fourier, any periodic function F(t), with period T, may be represented by an infinite series of the form F(t) = ∞ a0 + a (an cos nωTt + bn sin nωTt) 2 n=1 (3.20) F(t) T t Figure 3.11 An example of a general periodic function of period T. 236 general Forced Response Chap. 3 where ωT = 2π >T and where the coefficients a0, an, and bn for a given periodic function F(t) are calculated by the formulas a0 = T 2 F(t)dt T L0 (3.21) T an = 2 F(t)cos nωT t dt n = 1, 2, c T L0 bn = 2 F(t)sin nωT t dt n = 1, 2, c T L0 (3.22) T (3.23) Note that the first coefficient a0 is twice the average of the function F(t) over one cycle. The coefficients a0, an, and bn, are called Fourier coefficients. The series of equation (3.20) is the Fourier series. A more complete discussion of Fourier series can be found in most introductory differential equation texts (e.g., Boyce and DiPrima, 2009). The Fourier series is useful and relatively straightforward to work with because of a special property of the trigonometric functions used in the series. This special property, called orthogonality, can be stated as follows: L0 L0 and T T sin nωT t sin mωT t dt = e 0 T>2 cos nωT t cos mωT t dt = e L0 m ∙ n m = n (3.24) m ∙ n m = n (3.25) 0 T>2 T cos nωT t sin mωT t dt = 0 (3.26) The m and n here are integers. The truth of these three orthogonality conditions follows from direct integration. The orthogonality property (i.e., the integral of the product of two functions is zero) is used repeatedly in vibration analysis. In particular, orthogonality is used extensively in Chapters 4, 6, 7, and 8. Orthogonality is also used in statics and dynamics (i.e., the unit vectors are orthogonal). In Fourier analysis, the orthogonality of the sine and cosine functions on the interval 0 6 t 6 T is used to derive the formulas given in equations (3.21), (3.22), and (3.23). These coefficient values are derived as follows. The Fourier coefficients an are determined by multiplying equation (3.20) by cos mωTt and integrating over the period T. Similarly, the coefficients bn are determined by multiplying by sin mωTt and integrating. The summation on the right side of equation (3.20) vanishes except Sec. 3.3 237 Response to an Arbitrary Periodic Input for one term because of the orthogonality properties of the trigonometric function. Using the orthogonality conditions given previously determines that all terms of the T integrated product 10 F(t) sin mωT t dt will be zero except for the term containing bm. This yields equation (3.23). Likewise, all the terms in the series 10TF(t) cos mωT t dt are zero, except for the term containing am. This yields equation (3.22). Furthermore, this procedure can be repeated for each of the values of n in the summation in the Fourier series. The procedure for calculating the Fourier coefficient of a simple force is illustrated in the following example. Example 3.3.1 A triangular wave of period T is illustrated in Figure 3.12 and is described by 4 t - 1 T F(t) = µ T 4 1 - at - b T 2 0 … t … T 2 T … t … T 2 Determine the Fourier coefficients for this function. F(t) 1 T 3T 4 T 2 T 4 Figure 3.12 Plot of a triangular wave of period T. 1 Solution Straightforward integration of equation (3.21) yields 2 T L0 a0 = a T>2 4 2 4 T t - 1b dt + c1 at b d dt = 0 T T LT>2 T 2 T which is also the average value of the triangular wave over one period. Similarly, integration of equation (3.23) yields the result that bn = 0 for every n. Equation (3.22) yields an = 2 T L0 = d a T>2 4 2 4 T t - 1b cos nωTt dt + c1 at b d cos nωTt dt T T LT>2 T 2 T 0 n even -8 π2n2 n odd Thus the Fourier representation of this function becomes F(t) = - 2π 1 6π 1 10π 8 c cos t + cos t + cos tcd T 9 T 25 T π2 238 general Forced Response Chap. 3 1 F (t) 0.5 0 2 6 4 8 10 0.5 1 Time, t Figure 3.13 Plots of F(t) for one (long dashed line), two (short dashed line), and four (solid line) terms of the Fourier series indicate how close each series gets to the plot of Figure 3.12. which has frequency 2π >T. It is instructive to plot F(t) by adding one term at a time to make clear how many terms of the infinite series are needed to obtain a reasonable representation of F(t) as plotted in Figure 3.12. (Run VTB3_3 to observe this convergence.) This is done in Figure 3.13, which is a plot of F(t) for one, two, and four terms of the Fourier series. Computer codes for computing the series and plotting the results are given in Section 3.8 and VTB3_3. Toolbox file VTB3_3 can be used to obtain the coefficients of an arbitrary signal and for plotting the results. Substitution of the values an and bn into VTB3_5 will visually verify the result. n Note that when a Fourier series is used to approximate a periodic function with discontinuities, an overshoot (or ringing) of the Fourier series occurs at the discontinuity. This overshoot is called Gibbs phenomenon. Since a general periodic force can be represented as a sum of sines and cosines, and since the system under consideration is linear, the response of a singledegree-of-freedom system is calculated by computing the response to the individual terms of the Fourier series and adding the results. This is similar to the procedure used to solve the base-excitation problem of equation (2.63), where the input to a single-degree-of-freedom system consisted of the sum of a single sine term and a single cosine term. This is how superposition and Fourier series are used together to compute the solution for any periodic input. Thus, the particular solution x(t) of $ # mx(t) + cx(t) + kx(t) = F(t) (3.27) where F(t) is periodic, can be written as xp(t) = x1(t) + a [xcn(t) + xsn(t)] ∞ n=1 (3.28) Sec. 3.3 239 Response to an Arbitrary Periodic Input Here the particular solution x1(t) satisfies the equation a0 $ # mx1(t) + cx1(t) + kx1(t) = 2 (3.29) the particular solution xcn(t) satisfies the equation $ # mxcn(t) + cxcn(t) + kxcn(t) = an cos nωTt (3.30) for all values of n, and the particular solution xsn(t) satisfies the equation $ # mxsn(t) + cxsn(t) + kxsn(t) = bn sin nωTt (3.31) for all values of n. The solutions to equations (3.30) and (3.31) are calculated in Section 2.2, and the solution to equation (3.29) is calculated in Section 3.2. If the system is subject to nonzero initial conditions, this must also be taken into consideration. The particular solution to equation (3.29) is that of the step response calculated in equation (3.17) with F0 = a0 >2. This yields x1(t) = a0 2k an >m 2 2 (3.32) The particular solution of equation (3.30) is calculated in equation (2.36) to be xcn(t) = [[ω2n - (nωT) ] + (2ζωnnωT)2]1>2 cos(nωTt - θn) (3.33) where θn = tan-1 2ζωnnωT ω2n bn >m 2 2 - (nωT)2 Similarly, the particular solution of equation (3.31) is calculated to be xsn(t) = [[ωn2 - (nωT) ] + (2ζωnnωT)2]1>2 sin(nωTt - θn) (3.34) The total particular solution of equation (3.27) is then given by the sum of equations (3.32), (3.33), and (3.34) as indicated by equation (3.28). The total solution x(t) is the sum of the particular solution xp(t) calculated previously and the homogeneous solution obtained in Section 1.3. For the underdamped case (0 6 ζ 6 1), this becomes x(t) = Ae -ζωnt sin (ωdt + ϕ) + ∞ a0 + a [xcn(t) + xsn(t)] 2k n=1 (3.35) where A and ϕ are determined by the initial conditions. As in the case of a simple harmonic input as described in equation (2.37), the constants A and ϕ describing the 240 general Forced Response Chap. 3 transient response will be different than those calculated for the free-response case given in equation (1.38). This is because part of the transient term of equation (3.35) is the result of initial conditions and part is due to the excitation force F(t). Example 3.3.2 Consider the base-excitation problem of Section 2.4 (see Window 3.3), and calculate the total response of the system to initial conditions x0 = 0.01 m and 0 = 3.0 m>s. Assume that ωb = 3 rad>s, m = 1 kg, c = 10 kg>s, k = 1000 N>m, and Y = 0.05 m. Window 3.3 Review of the Base-Excitation Problem of Section 2.4 The base-excitation problem is to solve the expression .. . x  2n x  2n x  2nbY cos bt  2n Y sin bt for the motion of a mass, x(t), excited by a harmonic displacement of frequency b and amplitude Y through its spring–damper connections. This has the particular solution indicated by the second term on the right-hand side of equation (3.37). x(t) m k c y(t)Y sin bt Base Solution input of The equation of motion is given by equation (2.63), which has a periodic F(t) = cYωb cos ωbt + kY sin ωbt (3.36) Comparing coefficients with the Fourier expansion of equation (3.20) yields a0 = 0, an = bn = 0, for all n 7 1, and a1 = cYωb = (10 kg>s)(0.05 m)(3 rad>s) = 1.5 N b1 = kY = (1000 N>m)(0.05 m) = 50 N The solution for xc1(t) from equation (3.30) is given by equations (2.65) and (2.66), and the solution for xs1(t) from equation (3.31) is given by equations (2.66) and (2.67). The Sec. 3.3 241 Response to an Arbitrary Periodic Input summation of solutions indicated in equation (3.28) is then given by equation (2.68), and the total solution becomes x(t) = Ae -ζωnt sin (ωdt + ϕ) + ωnY £ ω2n + (2ζωb)2 (ω2n - ω2b)2 + (2ζωnωb)2 § 1>2 cos (ωbt - θ1 - θ2) (3.37) where A and ϕ are to be determined by the initial conditions, and θ1 and θ2 are as defined by equations (2.66) and (2.69). Since ζ = c>(21km) = 0.158 and ωn = 1k>m = 31.62 rad>s, these phase angles become θ1 = tan -1a θ2 = tan-1 a and the magnitude becomes ωn Y £ 1 ω2n + ω2b 2 2 + (2ζωnωb)2 ω2n + (2ζωb)2 § 2(0.158)(31.62)(3) (31.62)2 - (3)2 31.62 b = 1.541 rad (2)(0.158)(3) 1>2 = (31.62)(0.05) • = 0.05 m b = 0.03 rad 3 (31.62)2 - (3)2 4 2 + [2(0.158)(3)(31.62)]2 (31.62)2 + [2(0.158)(3)]2 ¶ 1>2 The solution given in equation (3.37) takes the form x(t) = Ae -5t sin (31.22t + ϕ) + 0.05 cos (3t - 1.571) (3.38) where ωd = ωn 21 - ζ2 = 31.225 rad>s. At t = 0, this becomes x(0) = A sin (ϕ) + 0.05 cos ( - 1.571) or 0.01 m = A sin ϕ + (0.05)( -0.00204) (3.39) Differentiating x(t) yields # x(t) = Ae -5t cos(31.225t + ϕ)(31.225) - 5Ae -5t sin(31.225t + ϕ) - 0.15 sin (3t - 1.571) At t = 0, this becomes 3 = (31.225) A cos (ϕ) - 5A sin (ϕ) - 0.15 sin ( -1.571) (3.40) Equations (3.39) and (3.40) represent two equations in the two unknown constants of integration, A and ϕ. Solving these yields A = - 0.0096 m and ϕ = 0.1083 rad, so that the total solution is x(t) = 0.0096e -5t sin (31.225t + 0.1083) + 0.05 cos (3t - 1.571) 242 general Forced Response Chap. 3 Displacement x(t) (mm) 0.10 0.05 0.00 0.05 0.10 0 1 2 3 4 Time t (s) Figure 3.14 The total time response of a spring–mass–damper system under base excitation as calculated in Example 3.3.1. This is plotted in Figure 3.14. Note that the transient term is not noticeable after 1 s. A comparison with a numerical solution of the same problem is given in Example 3.8.3 of Section 3.8. n The computation of the response to complicated inputs becomes tedious when using the analytical approach. With the advent of computational software, practicing engineers are more likely to use a numerical approach to compute the solution. While numerical approaches are approximations, they do allow quick calculation of the response to systems with complicated inputs consisting of step functions and long periodic disturbances. Numerical approaches are discussed in Section 3.9. 3.4 TRANSFORM METHODS The Laplace transform was introduced briefly in Section 2.3 as an alternative method of solving for the forced harmonic response of a single-degree-of-freedom system. The Laplace transform technique is even more useful for calculating the responses of systems to a variety of force excitations, both periodic and nonperiodic. The usefulness of the Laplace transform technique of solving differential equations and, in particular, solving for the forced response lies in the availability of tabulated Laplace transform pairs. Using tabulated Laplace transform pairs reduces the solution of forced vibration problems to algebraic manipulations and table “lookup.” In addition, the Laplace transform approach provides certain theoretical advantages and leads to a formulation that is very useful for experimental vibration measurements. The definition of a Laplace transform of the function of time f(t) is L[f(t)] = F(s) = L0 ∞ f(t)e -stdt (3.41) Sec. 3.4 243 transform Methods for an integrable function f(t) such that f(t) = 0 for t 6 0. The variable s is complex valued. The Laplace transform changes the domain of the function from the positive real-number line (t) to the complex-number plane (s). The integration in the Laplace transform changes differentiation into multiplication, as the following example illustrates. Example 3.4.1 # Calculate the Laplace transform of the derivative f(t). Solution # L 3 f (t) 4 = L0 ∞ # f (t)e -stdt = L0 ∞ e -st d 3 f (t) 4 dt dt Integration by parts yields ∞ # L 3 f (t) 4 = e -stf (t) ` + s 0 L0 ∞ e -stf (t) dt Recognizing that the integral in the last term of the preceding equation is the definition of F(s) yields # L 3 f (t) 4 = sF (s) - f (0) $ where F(s) denotes the Laplace transform of f(t). Repeating this procedure on f (t) yields $ # L 3 f (t) 4 = s2F(s) - sf(0) - f(0). n Example 3.4.2 Calculate the Laplace transform of the unit step function defined by the right-hand side of equation (3.41) and denoted by Φ(t) for the case t0 = 0. Solution L 3 Φ(t) 4 = L0 ∞ e -stdt = - e -st ∞ e- ∞ e -0 1 ` = + = s 0 s s s n The procedure for solving for the forced response of a mechanical system is first to take the Laplace transform of the equation of motion. Next, the transformed expression is algebraically solved for X(s), the Laplace transform of the response. The inverse transform of this expression is found by using a table of Laplace transforms to yield the desired time history of the response x(t). This is illustrated in the following example. A sample table of Laplace transform pairs is given in Table 3.1. 244 general Forced Response TABLE 3.1 CoMMon lAPlACE tRAnSFoRMS FoR ZERo InItIAl ConDItIonSa F(s) 5. f(t) δ(0) unit impulse 1, unit step Φ(t) e - at 1. 1 2. 1>s 1 3. s + a 4. Chap. 3 1 (s + a)(s + b) ωn s2 + ω2n s 6. 2 s + ωn2 1 7. 2 s(s + ω2n) 1 8. 2 s + 2ζωns + ω2n ω2n 9. 2 s(s + 2ζωns + ω2n) 10. e - as 11. F(s - a) 12. e - asF(s) 1 (e -at - e -bt) b - a sin ωnt cos ωnt 1 (1 - cos ωnt) ω2n 1 -ζωnt e sin ωdt, ζ 6 1, ωd = ωn 31 - ζ2 ωd ωn -ζω t 1 e n sin (ωdt + ϕ), ϕ = cos-1ζ , ζ 6 1 ωd δ(t - a) eatf(t) Ú 0 f(t - a)Φ(t - a) a A more complete table appears in Appendix B. Here the Heaviside step function or unit step function is denoted by Φ. Other notations for this function include μ and H. Example 3.4.3 Calculate the forced response of an undamped spring–mass system to a unit step function. Assume that both initial conditions are zero. Solution The equation of motion is $ mx(t) + kx(t) = Φ(t) Taking the Laplace transform of this equation yields (ms2 + k)X(s) = 1 s Solving algebraically for X(s) yields X(s) = 1 2 s(ms + k) = 1>m s(s + ω2n) 2 Sec. 3.4 245 transform Methods Examining the definition of the Laplace transform, note that the coefficient 1>m passes through the transform. The time function corresponding to the value of X(s) in the preceding equation can be found as entry 7 in Table 3.1. This implies that x(t) = 1>m ω2n (1 - cos ωnt) = 1 (1 - cos ωnt) k which, of course, agrees with the solution given by equation (3.18) with F0 = 1. n Example 3.4.4 Calculate the response of an underdamped spring–mass system to a unit impulse. Assume zero initial conditions. Solution The equation of motion is $ # mx + cx + kx = δ(t) Taking the Laplace transform of both sides of this expression using the results of Example 3.4.1 and entry 1 in Table 3.1 yields 1 ms2 Solving for X(s) yields + cs + k 2 X(s) = 1 X(s) = 1>m s + 2ζωns + ωn2 2 Assuming that ζ 6 1 and consulting entry 8 of Table 3.1 yields x(t) = 1>m ωn 31 - ζ 2 e -ζωnt sin 1 ωn 31 - ζ2t 2 = 1 -ζωnt e sin ωdt mωd in agreement with equation (3.6). n Example 3.4.5 Compute the solution of the spring–mass–damper system subject to an impulse at time t = π s defined by the following equation of motion: $ # x(t) + 2x(t) + 2x(t) = δ(t - π), x0 = v0 = 0 Solution From the coefficients ωn = 12 rad>s, ζ = 2 1 = and ωd = 1231 - (1/12)2 = 1 rad>s 212 12 Taking the Laplace transform of the equation of motion yields (s2 + 2s + 2)X(s) = e -πs 246 general Forced Response Chap. 3 Solving algebraically for X(s) yields X(s) = e -πs 1 b = (e -πs)a 2 s + 2s + 2 s + 2s + 2 2 The inverse Laplace transform of the last term is (from entry 8 in Table 3.1) L-1 a 1 b = e -t sin t s2 + 2s + 1 From entry 12 of Table 3.1, the inverse Laplace transform of X(s) then becomes x(t) = e -(t - π) sin (t - π)Φ(t - π) = e 0, e -(t - π) sin (t - π), t 6 π t Ú π n It should be noted that the Laplace transform may be used for problems with untabulated pairs by inverting the integration indicated in equation (3.41). The inversion integral is ∞ x(t) = 1 X(s)e stds 2πj L- ∞ (3.42) where j = 1-1. The inverse Laplace transform is discussed in greater detail in Appendix B. An often-used tool in Laplace transform analysis is the idea of convolution, introduced in Section 3.2. In fact, the convolution integral is often defined first in terms of the Laplace transform. Consider the response written as the convolution integral as given in equation (3.12), and take the Laplace transform assuming zero initial conditions. This yields X(s) = F(s)H(s) (3.43) called Borel’s theorem. Here F(s) is the Laplace transform of the driving force, f(t), and H(s) is the Laplace transform of the impulse response function h(t). Taking the transform of h(t) defined by equation (3.8) and using entry 8 of Table 3.1 yields H(s) = 1 s + 2ζωns + ω2n 2 (3.44) which is the transfer function of a single-degree-of-freedom oscillator as defined in Section 2.3, equation (2.59). A related transform is the Fourier transform, which arises from considering the Fourier series of a nonperiodic function. The Fourier transform of a function x(t) is denoted by X(ω) and is defined by ∞ X(ω) = 1 x(t)e -jωtdt 2π L- ∞ (3.45) Sec. 3.5 247 Response to Random Inputs which transforms the variable x(t) from a function of time into a function of frequency ω. The inversion of this transform is performed by the integral ∞ x(t) = L- ∞ X(ω)e jωtdω (3.46) The Fourier transform integral defined by equation (3.45) arises from the Fourier series representation of a function described by equation (3.20) by writing the series in complex form and allowing the period to go to infinity. [See Newland (1993), page 39, for details.] Note that the definitions of the Fourier transform and the Laplace transform are similar. In fact, the form of the Fourier transform pairs given in equations (3.45) and (3.46) can be obtained by substituting s = jω into the Laplace transform pair given by equations (3.41) and (3.42). Although this does not constitute a rigorous definition, it does provide a connection between the two types of transforms. Fourier transforms are not used as frequently for solving vibration problems as are Laplace transforms. However, the Fourier transform is used extensively in discussing random vibration problems and in the measurement of vibration parameters. Appendix B discusses additional details of transforms. A rigorous description of the use of various transforms, their properties, and their applications can be found in Churchill (1972). 3.5 RESPONSE TO RANDOM INPUTS So far, all the driving forces considered have been deterministic functions of time. That is, given a value of the time t, the value of F(t) is precisely known. Here the response of a system subject to a random force input F(t) is investigated. Disturbances are often characterized as random if the value of F(t) for a given value of t is known only statistically. That is, a random signal has no obvious pattern. For random signals it is not possible to focus on the details of the signal, as it is with a pure deterministic signal. Hence random signals are classified and manipulated in terms of their statistical properties. Randomness in vibration analysis can be thought of as the result of a series of experiments, all performed in an identical fashion under identical circumstances, each of which produces a different response. One record or time history is not enough to describe such a vibration; rather, a statistical description of all possible responses is required. In this case, a vibration response x(t) should not be thought of as a single signal, but rather as a collection, or ensemble, of possible time histories resulting from the same conditions (i.e., same system, same controlled environment, same length of time). A single element of such an ensemble is called a sample function (or response). Consider a random signal x(t), or sample, as pictured in Figure 3.15(d). The first distinction to be made about a random time history is whether or not the signal is stationary. A random signal is stationary if its statistical properties (usually 248 general Forced Response x(t) Chap. 3 x(t) T A Arms 0 Time Time A (a) (d) Rxx() Rxx() T A2 2 A2rms  Time shift  Time shift 0 2 A 2 (e) (b) Sxx() Sxx() Frequency (Hz) 1 T Frequency (Hz) 0 (c) (f) Figure 3.15 (a) A simple sine function; (b) its autocorrelation; (c) its power spectral density. (d) A random signal; (e) its autocorrelation; (f) its power spectral density. its average or mean square) do not change with time. The average of the random signal x(t) is defined and denoted by T 1 x∙ = lim x(t)dt (3.47) T S ∞ T L0 as introduced in Section 1.2, equation (1.20), for deterministic signals. Here, it is convenient to consider signals with a zero average or mean 3 i.e., ∙x(t) = 0 4 . This is not too restrictive an assumption, since if x∙(t) ∙ 0, a new variable x= = x - x∙ can be defined. The new variable x′ now has zero mean. The mean-square value of the random variable x(t) is denoted by x2 and is defined by x2 = lim T S∞ T 1 x2(t) dt T L0 (3.48) Sec. 3.5 249 Response to Random Inputs as introduced in Section 1.2, equation (1.21), for deterministic signals. In the case of random signals, this is also called the variance and provides a measure of the magnitude of the fluctuations in the signal x(t). A related quantity, called the root-mean-square (rms) value, is just the square root of the variance: xrms = 3x2 (3.49) This definition can be applied to the value of a single response over its time history or to an ensemble value at a fixed time. Another measure of interest in random variables is how fast the value of the variable changes. This addresses the issue of how long it takes to measure enough samples of the variable before a meaningful statistical value can be calculated. Many measured vibration signals are random and, as such, an indication of how quickly a variable changes is very useful. The autocorrelation function, denoted by Rxx(τ) and defined by Rxx(τ) = lim T S∞ T 1 x(t)x(t + τ)dt T L0 (3.50) provides a measure of how fast the signal x(t) is changing. The value τ is the time difference between the values at which the signal x(t) is sampled. The prefix auto refers to the fact that the term x(t)x(t + τ) is the product of values of the same sample at two different times. The autocorrelation is a function of the time difference τ only in the special case of stationary random signals. Figure 3.15(e) illustrates the autocorrelation of a random signal, and Figure 3.15(b) illustrates that of a sine function. The Fourier transform of the autocorrelation function defines the power spectral density (PSD). Denoting the PSD by Sxx(ω) and repeating the definition of equation (3.45) results in ∞ Sxx(ω) = 1 R (ω)e -jωτdτ 2π L- ∞ xx (3.51) Note that this integral of Rxx(τ) changes the real number τ into a frequency-domain value ω. Figure 3.15(c) illustrates the PSD of a pure sine signal, and Figure 3.15(f) illustrates the PSD of a random signal. The autocorrelation and power spectral density, defined by equations (3.50) and (3.51), respectively, can be used to examine the response of a spring–mass system to a random excitation. Recall from Section 3.2 that the response x(t) of a spring–mass–damper system to an arbitrary forcing function F(t) can be represented by using the impulse response function h(t - τ) given by equation (3.9) for underdamped systems. The Fourier transform of the function h(t - τ) can be used to relate the PSD of the random input of an underdamped system to the PSD of the system’s response. First note 250 general Forced Response Chap. 3 from equation (3.8), Example 3.4.4, and entry 8 of Table 3.1 that the Laplace transform of h(t) for a single-degree-of-freedom system is L[h(t)] = L c = 1 -ζωnt 1 e sin ωdt d = L 3e -ζωnt sin ωdt4 mωd mωd (3.52) 1 = H(s) ms + cs + k 2 where H(s) is the system transfer function as defined by equation (2.59). In this case, the Fourier transform of h(t) can be obtained from the Laplace transform by setting s = jω in equation (3.52). This yields simply H(jω) = 1 k - ω m + cωj 2 (3.53) which, upon comparison with equations (2.60) and (2.52), is also the frequency response function for the single-degree-of-freedom oscillator. Let X(ω) denote the Fourier transform of the impulse response function, h(t); then, from equations (3.45) and (3.53) ∞ X(ω) = 1 h(t)e -jωnt dt = H(ω) 2π L- ∞ (3.54) where the j is dropped from the argument of H for convenience. Thus the frequency response function of Section 2.3 can be related to the Fourier transform of the impulse response function. This becomes extremely significant in vibration measurement as discussed in Chapter 7. Next, recall the formulation of the solution of a vibration problem using the impulse response function. From equation (3.12), the response x(t) to a driving force F(t) is simply x(t) = L0 t F(τ)h(t - τ)dτ (3.55) Note that since h(t - τ) = 0 for t 6 T, the upper limit can be extended to plus infinity. Since F(t) = 0 for t 6 0, the lower limit can be extended to minus infinity. Thus, expression (3.55) can be rewritten as ∞ x(t) = L- ∞ F(τ)h(t - τ)dτ (3.56) Next, the variable of integration τ can be changed to θ by using τ = t - θ, and hence dτ = - dθ. Using this change of variables, the previous integral can be written -∞ x(t) = - L∞ ∞ F(t - θ)h(θ)d(θ) = L- ∞ F(t - θ)h(θ)dθ (3.57) Sec. 3.5 251 Response to Random Inputs which provides an alternative form of the solution of a forced vibration problem in terms of the impulse response function. Finally, consider the PSD of the response x(t) given by equation (3.51) as ∞ Sxx(ω) = 1 R (τ)e -jωτdτ 2π L- ∞ xx (3.58) Upon substitution of the definition of Rxx(τ) from equation (3.50), this becomes T 1 1 c lim Sxx(ω) = x(σ)x(σ + τ)dσ d e -jωτdτ 2π L- ∞ T S ∞ T L0 ∞ (3.59) The expressions for x(t) in the integral are evaluated next using equation (3.57), which results in Sxx(ω) = 1 1 c F(σ - θ)h(θ)dθ F(σ - θ + τ)h (θ) dθ d dσ d e -jωτdτ c lim 2π L- ∞ T S ∞ T L0 L- ∞ L- ∞ (3.60) ∞ ∞ = ∞ T 1 c F(tn)F(tn + τ) h(θ)e -jωθdθ h(θ)e jωθdθ d dσe -jωτdτ T L0 L- ∞ L- ∞ (3.61) ∞ T 1 2π L- ∞ n ∞ lim T S∞ ∞ n where e (t - t )jω = 1 has been inserted inside the inner integrals and a subsequent change of variables (tn = σ - θ) has been performed on the argument of F, which is subsequently moved outside the integral. The two integrals inside the brackets in equation (3.61) are H(ω) and its complex conjugate H(- ω), according to equation (3.54). Recognizing the frequency response functions H(ω) and H(- ω) in equation (3.61), this expression can be rewritten as or simply Sxx(ω) = 0 H(ω) 0 2 c 1 R (τ)e -jωτdτ d 2π L- ∞ ff ∞ Sxx(ω) = 0 H(ω) 0 2 Sff (ω) (3.62) Here Rff denotes the autocorrelation function for F(t) and Sff denotes the PSD of the forcing function F(t). The notation ∙ H(ω) ∙ 2 indicates the square of the magnitude of the complex frequency response function. A more rigorous derivation of the result can be found in Newland (1993). It is more important to study the result [i.e., equation (3.62)] than the derivations at this level. 252 general Forced Response Chap. 3 Equation (3.62) represents an important connection between the power spectral density of the driving force, the dynamics of the structure, and the power spectral density of the response. In the deterministic case, a solution was obtained relating the harmonic force applied to the system and the resulting response (Chapter 2). In the case where the input is a random excitation, the statement equivalent to a solution is equation (3.62), which indicates how fast the response x(t) changes in relation to how fast the (random) driving force is changing. In the deterministic case of a sinusoidal driving force, the solution allowed the conclusion that the response was also sinusoidal with a new magnitude and phase, but of the same frequency as the driving force. In a way, equation (3.62) makes the equivalent statement for a random excitation (see Window 3.4). It states that when the excitation is a stationary random process, the response will be a stationary random process and the response changes as rapidly as the driving force, but with a modified amplitude. In both the deterministic case and the random case, the amplitude of the response is related to the frequency response function of the structure. Example 3.5.1 Consider the single-degree-of-freedom system of Window 3.1 subject to a random (white noise) force input F(t). Calculate the power spectral density of the response x(t) given that the PSD of the applied force is the constant value S0. Solution The equation of motion is $ # mx + cx + kx = F(t) From equation (2.59) or equation (3.53), the frequency response function is H(ω) = Thus, 0 H(ω) 0 2 = ` 2 1 ` = 2 k - mω + cωj = 1 k - mω2 + cωj 1k 1k - mω2 2 + cωj 1 # - mω2 2 2 + c 2ω2 1 1k - mω2 2 - cωj 1 From equation (3.62), the PSD of the response becomes Sxx = 0 H(ω) 0 2Sff = 1k - mω2 2 2 + c 2ω2 S0 This states that if a single-degree-of-freedom system is excited by a stationary random force (of constant mean and rms value) that has a constant power spectral density of value S0, the response of the system will also be random with nonconstant (i.e., frequency-dependent) PSD of Sxx(ω) = S0 > 31 k - mω2 2 2 + c 2ω2 4 . n Sec. 3.5 253 Response to Random Inputs Another useful quantity in discussing the response of a system to random vibration is the expected value. The expected value (or more appropriately, the ensemble average) of x(t) is denoted by E[x] and defined by T E[x] = lim T S∞ x(t) dt L0 T (3.63) which, from equation (3.47), is also the mean value, x∙. The expected value is also related to the probability that x(t) lies in a given interval through the probability density function p(x). An example of p(x) is the familiar Gaussian distribution function (bell-shaped curve). In terms of the probability density function, the expected value is defined by ∞ E[x] = L- ∞ xp(x)dx (3.64) The average of the product of the two functions x(t) and x(t + τ) describes how the function x(t) changes with time and, for a stationary random process, is the autocorrelation function E 3 x(t)x(t + τ) 4 = lim T S∞ T 1 x(t)x(t + τ)dt = Rxx(τ) T L0 (3.65) upon comparison with equation (3.50). From equation (3.48), the mean-square value becomes x2 = Rxx(0) = E 3 x2 4 (3.66) The mean-square value can, in turn, be related to the power spectral density function by inverting equation (3.51) using the Fourier transform pair of equations (3.45) and (3.46). This yields ∞ Rxx(τ)τ = 0 = L- ∞ Sxx(ω)dω = E 3 x2 4 (3.67) 0 H(ω) 0 2Sff (ω) dω (3.68) Equation (3.62) relates the PSD of the response x(t) to the PSD of the driving force F(t) and the frequency response function. Combining equations (3.62) and (3.67) yields E 3 x2 4 = ∞ L- ∞ This expression relates the mean-square value of the response to the PSD of the (random) driving force and the dynamics of the system. Equations (3.68) and (3.62) form the basis for random vibration analysis for stationary random driving forces. These expressions represent the equivalent of using the impulse response function and frequency response functions to describe deterministic vibration excitations (see Window 3.4). 254 general Forced Response Chap. 3 Window 3.4 Comparison between Calculations for the Response of a Spring–Mass–Damper System and Deterministic and Random Excitations x(t) c f(t) m k Transfer function: G(s) 1 ms2 cs Frequency response function: G(j ) k H( ) 1 e m d Impulse response function: h(t) 1 k m 2 c j nt sin dt The Laplace transform of the impulse response function is 1 L[h(t)] ms2 cs k G(s) and the Fourier transform of the impulse response function is just the frequency response function H( ). These quantities relate the input and response by For deterministic f(t): X(s) For random f(t): Sxx( ) G(s)F(s) t x(t) h(t )f( )d E[x2] 0 2 |H( )| S ( ) |H( ) | S ( )d ff 2 ff To use equation (3.68), the integral involving ∙ H(ω) ∙ 2 must be evaluated. In 2 many useful cases, Sff (ω) is constant. Hence, values of 1 0 H(ω) 0 have been tabulated (see Newland, 1993). For example ∞ L- ∞ and ∞ ` ` 2 πB02 B0 ` dω = A0 + jωA1 A0 A1 (3.69) ` dω = (3.70) 2 L- ∞ A0 + jωA1 - ω A2 B0 + jωB1 2 π 1 A0 B21 + A2 B02 2 A0 A1 A2 Such integrals, along with equation (3.68), allow computation of the expected value, as the following example illustrates. Sec. 3.6 255 Shock Spectrum Example 3.5.2 Calculate the mean-square value of the response of the system described in Example $ # 3.5.1 with equation of motion mx + cx + kx = F(t), where the PSD of the applied force is the constant value S0. Solution Since the PSD of the forcing function is the constant S0, equation (3.68) becomes E 3 x2 4 = S0 ∞ L- ∞ ` 2 1 ` dω k - mω2n + jcω Comparison with equation (3.70) yields B0 = 1, B1 = 0, A0 = k, A1 = c, and A2 = m. Thus, E 3 x2 4 = S0 πS0 πm = kcm kc Hence, if a spring–mass–damper system is excited by a random force described by a constant PSD, S0, it will have a random response, x(t), with mean-square value πS0 >kc. n Two basic relationships used in analyzing spring–mass–damper systems excited by random inputs are illustrated in this section. The output or response of a randomly excited system is also random and, unlike deterministic systems, cannot be exactly predicted. Hence, the response is related to the driving force through the statistical quantities of power spectral density and mean-square values. See Window 3.4 for a comparison of response calculations for deterministic and random inputs. In deterministic vibrations, the concern in design is usually to compute the magnitude and phase of the response to a known deterministic forcing function. This section addressed the same problem when the forcing function has a random nature. Given a statistical property of the forcing function, say the average magnitude, then the best we can do is to compute the average value of the magnitude of the response. 3.6 SHOCK SPECTRUM Many disturbances are abrupt or sudden in nature. The impulse is an example of a force applied suddenly. Such a sudden application of a force or other form of disturbance resulting in a transient response is referred to as a shock. Because of the common occurrence of shock inputs, a special characterization of the response to a shock has developed as a standard design and analysis tool. This characterization is called the response spectrum and consists of a plot of the maximum absolute value of the system’s time response versus the natural frequency of the system. The impulse response discussed in Section 3.1 provides a mechanism for studying the response of a system to a shock input. Recall that the impulse response 256 general Forced Response Chap. 3 function, h(t), was derived from considering a force input, δ(t), of large magnitude and short duration and can be used to calculate the response of a system to any input. The impulse response function forms the basis for calculating the response spectrum introduced here. Recall from equation (3.12) that the response of a system to an arbitrary input F(t) can be written as x(t) = L0 t F(τ)h(t - τ)dτ (3.71) where h(t – τ) is the impulse response function for the system. For an underdamped system, h(t – τ) is given by equation (3.9): h(t - τ) = 1 -ζωn(t - τ) e sin ωd(t - τ) mωd t 7 τ (3.72) which becomes h(t - τ) = 1 sin ωn(t - τ) mωn (3.73) in the undamped case. The response spectrum is defined to be a plot of the peak or maximum value of the response versus frequency. For an undamped system, equations (3.71) and (3.73) can be combined to yield the maximum value of the displacement response as 1 ` F(τ) sin [ωn(t - τ)]dτ ` mωn L0 max t x(t)max = (3.74) Calculating a response spectrum then involves substitution of the appropriate F(t) into equation (3.74) and plotting x(t)max versus the undamped natural frequency. This is usually done numerically on a computer; however, the following example illustrates the procedure by hand calculation. Example 3.6.1 Calculate the response spectrum for the forcing function given in Figure 3.16 applied to the linear spring–mass system. The abruptness of the response is characterized by the time t1. F0 t1 t Figure 3.16 A step disturbance with rise time of t1 seconds. Sec. 3.6 257 Shock Spectrum Solution As in Example 3.2.2, the forcing function F(t) sketched in Figure 3.16 can be written as the sum of two other simple functions. In this case, the input is the sum of F1(t) = t F t1 0 and 0 t - t1 F2(t) = c F0 t1 0 6 t 6 t1 t Ú t1 Following the steps taken in Example 3.2.2, the response is calculated by evaluating the response to F1(t) and separately to F2(t). Linearity is then used to obtain the total response to F(t) = F1(t) + F2(t). The response to F1(t), denoted by x1(t), calculated using equations (3.71) and (3.73), becomes x1(t) = t ωn F0τ F0 t sin ωnt sin ωn(t - τ) dτ = a b k L0 t 1 k t1 ωn t 1 (3.75) Similarly, the response to F2(t), denoted by x2(t), becomes x2(t) = L0 t F2(τ) t - F0 τ - t1 1 sin ωn(t - τ)dτ = sin ωn(t - τ)dτ mωn mωn Lt1 t 1 which becomes x2(t) = - sin ωn(t - t 1) F0 t - t 1 c d k t1 ωnt 1 (3.76) so that the total response becomes the sum x(t) = x1(t) + x2(t): sin ωnt F0 t a b k t1 ωn t 1 x(t) = d F0 [ω t - sin ωnt + sin ωn(t - t 1)] kωnt 1 n 1 t 6 t1 (3.77) t Ú t1 Alternately, the Heaviside step function may be used to write this solution as x(t) = sin ωn(t - t 1) F0 t sin ωnt F0 t - t 1 a b a bΦ(t - t 1) k t1 ωn t 1 k t1 ωn t 1 (3.78) Equation (3.77) is the response of an undamped system to the excitation of Figure 3.16. To find the maximum response, the derivative of equation (3.77) is set equal to zero and solved for the time tp at which the maximum occurs. This time tp is then substituted into the response x(tp) given by equation (3.77) to yield the maximum response x(tp). # Differentiating equation (3.77) for t 7 t1 yields x(t p) = 0 or - cos ωnt p + cos ωn(t p - t 1) = 0 (3.79) 258 general Forced Response Chap. 3 2(1  cos nt 1 ) sin n t1  nt p Figure 3.17 A graphical representation of equations (3.80) and (3.81). 1  cos  n t 1 Using simple trigonometry formulas and solving for ωntp yields tan ωnt p = 1 - cos ωnt 1 sin ωnt 1 or ωnt p = tan-1 a 1 - cos ωnt 1 b sin ωnt 1 (3.80) where tp denotes the time to the first peak [i.e., the time for which the maximum value of equation (3.77) occurs]. Expression (3.80) corresponds to a right triangle of sides (1 - cos ωnt 1), and sin ωnt 1, and hypotenuse 2sin2ωnt 1 + (1 - cos ωnt 1)2 = 22(1 - cos ωnt 1) (3.81) This relationship is illustrated in Figure 3.17. Hence, sin ωntp can be calculated from sin ωnt p = - 1 (1 - cos ωnt 1) A2 (3.82) and cos ωnt p = - sin ωnt 1 22(1 - cos ωnt 1) (3.83) Substitution of this expression into solution (3.77) evaluated at tp yields, after some manipulation [here x(tp) = xmax], x maxk 1 = 1 + 22(1 - cos ωnt 1) F0 ωn t 1 (3.84) where the left side represents the dimensionless maximum displacement. It is customary to plot the response spectrum (dimensionless) versus the dimensionless frequency ωn t 1 t1 = T 2π (3.85) where T is the structure’s natural period. This provides a scale related to the characteristic time, t1, of the input. Figure 3.18 is a plot of the response spectrum for the ramp input force of Figure 3.16. Note that each point on the plot corresponds to a different rise time, t1, of the excitation. The vertical scale is an indication of the relationship between the structure and the rise time of the excitation. 259 Shock Spectrum 2.0 ( xk F ) max 0 1.0 0 1 2 3 4 t1 T Figure 3.18 Response spectrum for the input force of Figure 3.16. The vertical axis is the dimensionless maximum response, and the horizontal axis is the dimensionless frequency (or delay time). 0.2 Displacement x(t), maximum displacement, and normalized input force Sec. 3.6 0.15 0.1 0.05 0 5 10 15 Time, t Figure 3.19 A plot of the time response (solid line) of an undamped system to the input given in Figure 3.16. Also shown are the maximum magnitude (long dashed line) and the input function (short dashed line) for the parameters k = 10 N >m, m = 10 kg, F0 = 1 N, and t1 = 1 s. The response is plotted using equation (3.77) along with the maximum magnitude as given by equation (3.84) and the ramp input function in Figure 3.19. Note from these plots that the amplitude of the response is magnified, or larger than the level of the input force. If t1 is chosen to be near a period (the minimum in Figure 3.18), then the response is lower than this value and the maximum response will be equal to the input level. The effects of the various parameters form the topic of shock isolation (Section 5.2, which can be read now) and are examined numerically in Section 3.9. n 260 general Forced Response Chap. 3 3.7 MEASUREMENT VIA TRANSFER FUNCTIONS The forced response of a vibrating system is very useful in measuring the physical parameters of a system. As indicated in Section 1.6, the measurement of a system’s damping coefficient can only be made dynamically. In some systems the damping is large enough that the vibration does not last long enough for a free decay measurement to be taken. This section examines the use of the forced response, transfer functions, and random vibration analysis introduced in previous sections to measure the mass, damping, and stiffness of a system. The use of transfer functions to measure the properties of structures comes from electrical engineering. In circuit applications, a function generator is used to apply a sinusoidal voltage signal to a circuit. The output is measured for a range of input frequencies. The ratio of the Laplace transform of the two signals then yields the transfer function of the test circuit. A similar experiment may be performed on mechanical structures. A signal generator is used to drive a force-generating device (called a shaker) that drives the structure sinusoidally through a range of frequencies at a known amplitude and phase. Both the response (either acceleration, velocity, or displacement) and the input force are measured using various transducers. The transform of the input and output signal is calculated and the frequency response function for the system is determined. The physical parameters are then derived from the magnitude and phase of the frequency response function. The details of the measurement procedures are discussed in Chapter 7. The methods of extracting physical parameters from the frequency response function are introduced here. Several different transfer functions are used in vibration measurement, depending on whether displacement, velocity, or acceleration is measured. The various transfer functions are illustrated in Table 3.2. The table indicates, for example, that the accelerance transfer function and corresponding frequency response function are obtained from dividing the transform of the acceleration response by the transform of the driving force. TABLE 3.2 tRAnSFER FunCtIonS uSED In VIbRAtIon MEASuREMEnt Response Measurement Transfer Function Inverse Transfer Function Acceleration Velocity Displacement Accelerance Mobility Receptance Apparent mass Impedance Dynamic stiffness The three transfer functions given in Table 3.2 are related to each other by simple multiplications of the transform variable s, since this corresponds to differentiation. Thus, with the receptance transfer function (also called the compliance or admittance) denoted by X(s) 1 = H(s) = 2 F(s) ms + cs + k (3.86) Sec. 3.7 261 Measurement via transfer Functions the mobility transfer function becomes s X(s) s = sH(s) = 2 F(s) ms + cs + k (3.87) because sX(s) is the transform of the velocity. Similarly, s2X(s) is the transform of the acceleration, and the accelerance transfer function (inertance) becomes s2X(s) s2 = s2H(s) = F(s) ms2 + cs + k (3.88) Each of these also defines the corresponding frequency response function by substituting s = jω. Consider calculating the magnitude of the complex compliance H(jω) from equation (2.53) or (3.86). As expected from equation (2.70), this yields 0 H(jω) 0 = 1 2(k - mω2)2 + (cω)2 (3.89) Note that the largest value of this magnitude occurs near k - mω2 = 0, or when the driving frequency is equal to the undamped natural frequency, ω = ωn = 2k>m. Recall from Section 2.2 that this also corresponds to a phase shift of 90°. This argument is used in testing to determine the natural frequency of vibration of a test particle from a measured magnitude plot of the system transfer function. This is illustrated in Figure 3.20. The exact value of the peak frequency is derived in Example 2.2.5. In principle, each of the physical parameters in the transfer function can be determined from the experimental plot of the frequency response function’s magnitude. The natural frequency, ωn, is determined from the position of the Mag H( j) 5 1  c 1 k 0 0.01 0.1 1 log ()   n 1  22  n  k/m 10 Figure 3.20 A magnitude plot for a spring–mass–damper system for the compliance transfer function indicating the determination of the natural frequency and stiffness. 262 general Forced Response Chap. 3 peak. The damping constant c is approximated from the value of the frequency and a measurement of the magnitude ∙ H(jω) ∙ at ω = 2k>m, since, from equation (3.89) ` H aj k 1 1 b` = = 2 cωn Am 2(cωn) (3.90) This formulation for measuring the damping coefficient provides an alternative to the logarithmic decrement technique presented in Section 1.6. Next, the stiffness can be determined from the zero frequency point. For ω = 0, equation (3.87) yields 0 H(0) 0 = 1 2k 2 = 1 k (3.91) Since ωn = 2k>m, knowledge of ωn and k yields the value for m. In this way, m, c, k, ωn, and ζ can all be determined from measurements of ∙ H(jω) ∙ . More practical methods are discussed in Chapter 7. Of course, m and k can usually be measured by static experiments as well, for comparison. The preceding analysis all depends on the experimentally determined function ∙ H(jω) ∙ . Most experiments contain several sources of noise, so that a clean plot of ∙ H(jω) ∙ is hard to get. The common approach is to repeat the experiment several times and essentially average the data (i.e., use ensemble averages). In practice, matched sets of input force time histories, f(t), and response time histories, x(t), are averaged to produce Rxx(t) and Rff(t) using equation (3.48). The Fourier transform of these averages is then taken using equation (3.51) to get the PSD functions Sxx(ω) and Sff(ω). Equation (3.62) is then used to calculate ∙ H(jω) ∙ from the PSD values of the measured input and response. This procedure works for averaging noisy data as well as for the case of using a random excitation (zero mean) as the driving force. The transforms and computations required to calculate ∙ H(jω) ∙ are usually made digitally in a dedicated computer used for vibration testing. This is discussed in Chapter 7 in more detail. 3.8 STABILITY The concept of stability was introduced in Section 1.8 in the context of free vibration. Here the definitions of stability for free vibration are extended to include the forced-response case. Recall from equation (1.86) that the free response is stable if it stays within a finite bound for all time (see Window 3.5). This concept of a wellbehaved response can also be applied to the forced motion of a vibrating system. In fact, in a sense, the inverted pendulum of Example 1.8.1 is an analysis of a forced response if gravity is considered to be the driving force. Sec. 3.8 263 Stability Window 3.5 Review of Stability of the Free Response from Section 1.8 0 x(t) 0 6 M A solution x(t) is stable if there exists some finite number M such that for all t 7 0. If this bound cannot be satisfied, the response x(t) is said to be unstable. If a response x(t) is stable and x(t) approaches zero as t gets large, the solution x(t) is said to be asymptotically stable. An undamped spring–mass system is stable as long as m and k are positive and the value of M is just the amplitude A [i.e., M = A, where x(t) = A sin (ωn + ϕ)]. A damped system is asymptotically stable if m, c, and k are all positive [i.e., x(t) = Ae -ζωnt sin(ωdt + ϕ) goes to zero as t increases]. The stability of the forced response of a system can be defined by considering the nature of the applied force or input. The system $ # mx + cx + kx = F(t) (3.92) is defined to be bounded-input, bounded-output stable (or simply BIBO stable) if, for any bounded input, F(t), the output, or response x(t), is bounded for any arbitrary set of initial conditions. Systems that are BIBO stable are manageable at resonance and do not “blow up.” Note that the undamped version of equation (3.92) is not BIBO stable. To see this, note that if F(t) is chosen to be F(t) = sin[(k>m)1/2t] for the case c = 0, the response x(t) is clearly not bounded as indicated in Figure 2.5. Also recall that the magnitude of the forced response of an undamped system is F0/[ω2n - ω2], which approaches infinity as ω approaches ωn (see Window 3.6). However, the input force F(t) is bounded since 0 F(t) 0 = ` sina k tb ` … 1 Am (3.93) for all time. Thus, there is some bounded force for which the response is not bounded and the definition of BIBO stable is violated. This situation corresponds to resonance. Clearly, an undamped system is poorly behaved at resonance. Next, consider the damped case (c 7 0). Immediately, the preceding example of resonance is no longer unbounded. The forced-response magnitude curves given in Figure 2.8 illustrate that the response is always bounded for any bounded periodic driving force. To see that the response of an underdamped system is bounded 264 general Forced Response Chap. 3 Window 3.6 Review of the Response of a Single-Degree-of-Freedom System to Harmonic Excitation The undamped system $ mx + kx = F0 cos ωt # x(0) = v0 x(0) = x0, has the solution x(t) = f0 f0 v0 cos ωt sin ωnt + ax0 - 2 b cos ωnt + 2 2 ωn ωn - ω ωn - ω2 (2.11) where f0 = F0 >m, and ωn = 2k>m, and x0 and 0 are initial conditions. The underdamped system $ # mx + cx + kx = F0 cos ωt has the steady-state solution x(t) = f0 2(ωn2 - ω ) + (2ζωnω) 2 2 2 cos aωt - tan -1 2ζωnω ω2n - ω2 b (2.36) where the damping ratio ζ satisfies 0 6 ζ 6 1. for any bounded input, recall that the solution for an arbitrary driving force is given in terms of the impulse response in equation (3.12) to be x(t) = L0 t f(τ)h(t - τ)dτ (3.94) where f(t) = F(t)>m. Taking the absolute value of both sides of this expression yields 0 x(t) 0 = ` L0 f(τ)h(t - τ)dτ ` … t L0 t |f(τ)h(t - τ)| dτ (3.95) where the inequality results from the definition of integrals as a limit of summations. Noting that 0 hf 0 … 0 h 0 0 f 0 yields 0 x(t) 0 … L0 t 0 f(τ) 0 0 h(t - τ) 0 dτ … M L0 t 0 h(t - τ) 0 dτ (3.96) where f(t) [and hence F(t)] is assumed to be bounded by M [i.e., ∙ f(t) ∙ 6 M]. Note that the choice of the constant M is arbitrary and is always chosen as a Sec. 3.8 265 Stability matter of convenience. Next consider evaluating the integral on the right of inequality (3.96) for the underdamped case. The impulse response for an underdamped system is given by equation (3.9). Substitution of equation (3.9) into (3.96) yields 0 x(t) 0 … M 1 0 e -ζωn(t - τ) 0 0 sin ωd(t - τ) 0 dτ mω d L0 t (3.97) t … M M -ζωnt e e ζωnτdτ = (1 - e -ζωnt) … M mωd mζωnωd L0 since ∙ sin ωd(t – τ) ∙ … 1 for all t 7 0, mζωnωd 7 1 and 1 - e -ζωnt 6 1. Thus 0 x(t) 0 … M Hence as long as the input force is bounded (say, by M) the preceding calculation illustrates that the response x(t) of an underdamped system is also bounded, and the system is BIBO stable. The results of Section 2.1 clearly indicate that the response of an undamped system is well behaved, or bounded, as long as the harmonic input is not at or near the natural frequency (see Window 3.6). In fact, the response given by equation (2.11) illustrates that the maximum magnitude will be less than some constant as long as ω ∙ ωn. To see this, take the absolute value of equation (2.11), which yields 0 x(t) 0 … ` f0 f0 v0 ` + ` x0 - 2 ` + ` 2 ` 6 M ωn ωn - ω2 ωn - ω2 (3.98) where M is finite since each term is finite as long as ω ∙ ωn. Here 0 and x0 are the initial velocity and displacement, respectively. Thus the undamped forced response is sometimes well behaved and sometimes not. Such systems are said to be Lagrange stable. Specifically, a system is defined to be Lagrange stable, or bounded, with respect to a given input if the response is bounded for any set of initial conditions. Undamped systems are Lagrange stable with respect to many inputs. This definition is useful when F(t) is known completely or known to fall into a specific class of functions. Both the damped and undamped solutions given in Window 3.6 are Lagrange stable for ωn ∙ ω. In general, if the homogeneous solution is asymptotically stable, the forced response will be BIBO stable. If the homogeneous response is stable (marginally stable), the forced response will only be Lagrange stable. The forced response of an unstable homogeneous system can still be BIBO stable, as illustrated in the following example. 266 general Forced Response Chap. 3 Example 3.8.1 Consider the inverted pendulum of Example 1.8.1, illustrated in Figure 3.21, and discuss its stability properties. k m  l mg Figure 3.21 The spring-supported inverted pendulum of Example 3.8.1. Solution Summing the moments about the pivot point yields $ 2 a M0 = ml θ(t) = [ - kl sin θ(t)][l cos θ(t)] + mg[l sin θ(t)] Considering the small-angle approximation of the inverted pendulum equation results in the equation of motion $ ml 2θ(t) + kl 2θ(t) = mglθ(t) If mglθ is considered to be an applied force, the homogeneous solution is stable since m, l, and k are all positive. Writing the equation of motion as a homogeneous equation yields $ ml 2θ(t) + [kl 2 - mgl ]θ(t) = 0 The forced response, however, is not bounded unless kl 7 mg and was shown in Example 1.8.1 to be divergent (unbounded) in this case. Hence the forced response of this system is Lagrange stable for F(t) = mglθ if kl 7 mg, and unbounded (unstable) if kl 6 mg. n Example 3.8.2 Consider again the inverted pendulum of Example 3.8.1. Design an applied force F(t) such that the response is bounded for kl 6 mg. Solution The problem is to find F(t) such that θ satisfying $ ml 2θ(t) + (kl 2 - mgl)θ(t) = F(t) is bounded. As a starting point, assume that F(t) has the form $ F(t) = - aθ(t) - bθ(t) Sec. 3.9 numerical Simulation of the Response 267 where a and b are to be determined by the design for stability. This form is attractive because it changes the inhomogeneous problem into a homogeneous problem. The equation of motion then becomes $ $ ml 2θ(t) + (kl 2 - mgl)θ(t) = - aθ(t) - bθ(t) This can be written as a homogenous equation: # $ ml 2θ(t) + bθ + (kl 2 - mgl + a)θ = 0 From Section 1.8, it is known that if each of the coefficients is positive, the response is asymptotically stable, which is certainly bounded. Hence, choose b 7 0 and a such that kl 2 - mgl + a 7 0 and the forced response will be bounded. n An applied force can also cause a stable (or asymptotically stable) system response to become unstable. To see this, consider the system $ # x(t) + x(t) + 4x(t) = f(t) (3.99) # where f(t) = ax(t) + bx(t). If a is chosen to be 2 and b = 2, the equation of motion becomes $ # x(t) - x(t) + 2x(t) = 0 (3.100) which has a solution that grows exponentially and illustrates flutter instability. At first glance, this seems to violate the earlier statement that asymptotically stable homogeneous systems are BIBO stable. This example, however, does not violate the definition because the input f(t) is not bounded. The applied force is a function of the displacement and velocity, which grow without bound. 3.9 NUMERICAL SIMULATION OF THE RESPONSE Numerical simulation, as introduced in Section 1.9 for the free response and Section 2.8 for the response to harmonic inputs, can be used to compute and plot the response to any arbitrary forcing function. Numerical simulation has become the preferred method for computing the response, as it requires a minimum amount of analysis and can be applied to any type of input force, including experimental data (such as time histories of earthquakes). Numerical solutions may also be used to check analytical work, and analytical work should be used to check numerical results as often as possible. This section presents some common codes for simulating and plotting the response of systems to a general force. In the previous sections, great effort was put forth to derive analytical expressions for the response of various single-degree-of-freedom systems driven by a variety of forces. These analytical expressions are extremely useful for design and 268 general Forced Response Chap. 3 for understanding some of the physical phenomena. Plots of the time response were constructed to realize the nature and features of the response. Rather than plotting the analytical function describing the response, the time response may also be computed numerically using an Euler or Runge–Kutta integration and computational software packages as introduced in Sections 1.8 and 2.8. While numerical solutions such as these are not exact, they do allow nonlinear terms to be considered, as well as the solution of systems with complicated forcing terms that do not have analytical solutions. In order to solve for the force response to an arbitrary input numerically, equation (2.133) needs to be modified slightly to incorporate an arbitrary applied force. It is possible to generate an approximate numerical solution of $ # mx(t) + cx(t) + kx(t) = F(t) (3.101) subject to any initial conditions for any arbitrary force F(t). For most codes, equation (3.101) must be cast into the first-order, or state-space, form by renaming x as x1 and writing # x1(t) = x2(t) # x2(t) = -2ζωnx2(t) - ω2nx1(t) + f(t) (3.102) # where x2 denotes the velocity x(t) and x1 denotes the position x(t) as before and f(t) = F(t)>m. This is subject to the initial conditions x1(0) = x0 and x2(0) = 0. Given ωn, ζ, f0, x0, 0, and either the analytical or numerical form of F(t), the solution of equation (3.101) can be determined numerically. The matrix form of equation (3.102) becomes # x(t) = Ax(t) + f(t) (3.103) where x and A are the state vector and state matrix as defined previously in equation (1.96) and repeated here: x1 x = Jx d , 2 and 0 A = J -ω2 n The vector f is the applied force and takes the form 1 -2ζωn d 0 f(t) = J f(t) d (3.104) x(t i + 1) = x(t i) + Ax(t i)∆t + f(t i)∆t (3.105) The Euler form of equation (3.103) is In this case, where f(t) is an arbitrary force, either f(ti) is f evaluated at each time instant, if the analytical form of f is known, or it is a discrete time history of data points, if f is known numerically or experimentally. This expression can also be Sec. 3.9 269 numerical Simulation of the Response adapted to the Runge–Kutta formulation, and most of the codes mentioned in Appendix G have built-in commands for a Runge–Kutta solution. The numerical integration to determine the response of a system is an approximation, whereas the plotting of the analytical solution is exact. So the question arises, “Why bother to integrate numerically to find the solution?” The answer lies in the fact that many practical problems do not have exact analytical solutions, such as the treatment of nonlinear terms, as was illustrated in Section 1.10. This section discusses the solution of the forced response using numerical integration in an environment where the exact solution is available for comparison. The examples in this section introduce numerical integration to compute forced responses and to compare them with the exact solution. The following examples illustrate the use of various programs to compute and plot the solution. Note that VTB1_3 and VTB1_4 allow the use of data points as a forcing function. Example 3.9.1 Solve for the response of the system in Example 3.2.1, using the parameters given in Figure 3.7 numerically. Recall the equation of motion is 0, $ # mx + cx + kx = F(t) = e F0, t0 7 t 7 0 t Ú t0 [Use the values ζ = 0.1 and ωn = 3.16 rad>s (with F0 = 30 N, k = 1000 N>m, t0 = 0).] Compare the numerical solution with the analytical solution given in Example 3.2.1. Solution The analytical solution given in equation (3.15) with the parameter values given in Figure 3.7 becomes x(t) = (0.03 - 0.03e -0.316(t - t0)cos[3.144(t - t 0) - 0.101])Φ(t - t 0) (3.106) where Φ denotes the Heaviside step function used to indicate that x(t) is zero until t reaches t0. The state equations for the equation of motion are written as 0 # x1 c# d = C k x2 m 0 x1 S c S c x d + C F0 2 Φ(t - t 0) m m 1 (3.107) The result of plotting the analytical solution of equation (3.106) and numerically integrating and plotting equation (3.107) is given in Figure 3.22. Note that this is virtually identical to the plot of Figure 3.7, which is the analytical solution. The codes for producing the numerical solution follow. In Mathcad, the code for solving equation (3.107) is F0 : = 30 θ : = atanc ζ 1 - ζ2 d k : = 1000 F0 = 0.03 k ωn : = 3.16 ζ : = 0.1 F0 k # 21 - ζ2 = 0.03015 t0 : = 0 ζ # ωn = 0.316 270 general Forced Response Chap. 3 Displacement 0.06 0.04 0.02 0 2 4 8 6 10 12 Time, t Figure 3.22 A plot of the numerical solution of the system of Examples 3.2.1 and 3.9.1. ωd : = ωn # 21 - ζ2 xa(t) : = C £ ωd = 3.14416 θ = 0.101 m: = a k b ωn2 F0 F0 # e-ζ # ωn # (t - t0). cos[ωd # (t - t0) - θ] § S # Φ(t - t0) k k # 21 - ζ2 0 X: = J d 0 X1 D (t, X) : = C - 2 # ζ # ωn # X 1 - ωn2 # X 0 + F0 m # Φ(t - t0) S Z : = rkfixed (X, 0, 12, 2000, D) > t : = Z 6 0 7 x : = xa(t) xn : = Z 6 1 7 The MATLAB code for computing the solution and plotting equation (3.107) is clear all %% Analytical solution F0=30; k=1000; wn=3.16; zeta=0.1; theta=atan(zeta/(1-zeta^2)); wd=wn*sqrt(1-zeta^2); t=0:0.01:12; t0=0; Heaviside=stepfun(t,t0); % define Heaviside step function for 0 6 t 6 12 Sec. 3.9 numerical Simulation of the Response 271 xt=(F0/k-F0/(k*sqrt(1-zeta^2)) * exp(-zeta*wn*(t - t0)). *cos(wd*(t-t0)-theta)).*Heaviside; plot(t,xt);’Hold on’ %% Numerical Solution xo=[0; 0]; ts=[0 12]; [t, x]=ode45('f',ts,xo); plot(t,x(:,1),'r'); hold off %-------------------------------------------function v=f(t,x) Fo=30; k=1000; wn=3.16; zeta=0.1; to=0; m=k/wn^2; v=[x(2); x(2).*-2*zeta*wn + x(1).*-wn^2 + Fo/m*stepfun(t, to)]; The Mathematica code for computing the solution and plotting equation (3.107) is In[1]:= 6 6 PlotLegends' In[2]:= F0 = 30; k = 1000; ωn = 3.16; ζ = 0.1; t0 = 0; θ = ArcTan [ 21 - ζ2, ζ]; k ; m = ωn2 ωd = ωn * 21 - ζ2; In[10]:= xanal[t_] = ° F0 F0 *Exp[- ζ*ωn*(t - t0)]*Cos[ωd*(t - t0)- θ]¢ k k* 21 - ζ2 * UnitStep[t - t0]; In[11]:= xnumer=NDSolve [{x"[t] + 2 * ζ * ωn * x'[t] + ωn2 * x[t] F0 == * UnitStep[t - t0], x[0] == 0, x'[0] == 0}, x[t], m {t, 0, 12}]; Plot[{Evaluate[x[t] /. xnumer], xanal[t]}, {t, 0, 12}, PlotStyle S {RGBColor[1, 0, 0], RGBColor[0, 1, 0]}, PlotLegend S {"Numerical", "Analytical"}, LegendSize S {1, .3}, LegendPosition S {1, 0}] n 272 general Forced Response Chap. 3 Next, consider computing the response to systems where the force is applied for only a specified amount of time. Such problems are difficult to solve analytically because they require the use of a change of variables in the convolution integral. Alternately, a number of analytical responses must be calculated, as indicated in Example 3.2.2. In the next example, numerical integration is used to solve for the response to these types of forcing functions. Example 3.9.2 Numerically compute the solution to Example 3.2.2 using the data given in Figure 3.9 (F0 = 30 N, k = 1000 N>m, ζ = 0.1, and ωn = 3.16 rad>s). Plot the result for the two pulse times given in the figure (i.e., for t1 = 0.5 and for t1 = 1.5 s). Solution First, write the equation of motion using a Heaviside step function to represent the driving force. The equation of motion written with Heaviside functions to describe the applied force is $ # mx(t) + cx(t) + kx(t) = F0[1 - Φ(t - t 1)] (3.108) This can be solved numerically directly in this form using Mathematica or by putting equation (3.108) into state-space form and solving via Runge–Kutta in Mathcad or MATLAB. The response for two different values of t1 is given in Figure 3.23. The codes follow. 0.05 x x2 -- 0 2 4 6 0.05 t Figure 3.23 The response of a spring–mass–damper system to a square input of pulse duration t1 = 0.1 s (dashed line) and t1 = 1.5 s (solid line), with k = 1000 N >m, m = 100.14 kg, ζ = 0.1, and F0 = 30 N. 8 Sec. 3.9 273 numerical Simulation of the Response The Mathcad code for solving for the response is t1 := 15 ζ := 0.1 0 X: = J d 0 ωn := 3.16 k := 1000 m: = k ωn2 f(t) := F0 - F0 # Φ(t - t1) F0 := 30 m = 100.144 f2(t) := F0 - F0 Φ(t - 0.1) Y := X Y1 X1 D(t, X) := C - ωn 2 # X0 - 2ζ # ωn # X1 f(t) S + m D2(t, Y) := C Z := rkfixed (X, 0, 8, 2000, D) t := Z 607 x := Z - ωn 2 # Y0 - 2ζ # ωn # Y1 + f(t) S m Z2 :=rkfixed (Y, 0, 8, 2000, D2) 617 x2 := Z2 6 1 7 The MATLAB code for solving for the response is clear all xo=[O; 0]; ts=[O 8]; [t, x]=ode45('f', ts, xo); plot(t,x(:,1),'--'); hold on [t, x]=ode45('f1', ts, xo); plot(t,x(:,1)); hold off %--------------------------------------------function v=f(t, x) Fo=30; k=1000; wn=3.16; zeta=0.1; to=0.1; m=k/wn^2; v=[x(2); x(2).*-2*zeta*wn + x(1).*-wn^2 + Fo/m*(1-stepfun(t, to))]; %--------------------------------------------function v=f1(t,x) Fo=30; k=1000; wn=3.16; zeta=0.1; to=1.5; m=k/wn^2; v=[x(2); x(2).*-2*zeta*wn + x(1).*-wn^2 + Fo/m*(1-stepfun(t,to))]; The Mathematica code for solving for the response is In[1]:= 6 6 PlotLegends' In[2]:= F0 = 30; k = 1000; ωn = 3.16; ζ = 0.1; t1 = 1.5; 274 general Forced Response Chap. 3 t2 = 0.1; m = k ωn2 ; In[9]:= xpointone = F0 m * (1 - UnitStep[t - t1]), x1[0] == 0, x1’[0] == 0}, x[t], {t, 0, 8}] xonepointfive = NDSolve [{x1''[t] + 2 * ζ * ωn * x1'[t] + ωn2 * x1[t] == NDSolve [{x2''[t] + 2 * ζ * ωn * x2'[t] + ωn2 * x1[t] == F0 m * (1 - UnitStep[t - t2]), x2[0] == 0, x2'[0] == 0}, x2[t], {t, 0, 8}]; Plot[{Evaluate[x1[t] /. xpointone], Evaluate[x2[t] /. xonepointfive]}, {t, 0, 8}, PlotStyle S {RGBColor[1, 0, 0], RGBColor[0, 1, 0]}, PlotRange S {-.06, .06}, PlotLegend S {"t1=1.5", "t2=0.1"}, LegendSize S {1, .3}, LegendPosition S {1, 0}]; n Example 3.9.3 Consider the base-excitation problem of Example 3.3.2 and compute the response numerically. Compare the result to the analytical solution computed in Example 3.3.2 by plotting both the analytical solution and the numerical solution on the same graph. Solution The equation of motion to be solved in this example is $ # x(t) + 10x(t) + 1000x(t) = 1.5 cos 3t + 50 sin 3t (3.109) From Example 3.3.2, the analytical solution is x(t) = 0.09341e -5t sin (31.225t + 0.1074) + 0.05 cos (3t - 1.571) The equation of motion can be solved numerically in the form given by equation (3.109) directly in Mathematica or by putting equation (3.109) into state-space form and solving via Runge–Kutta in Mathcad or MATLAB. The plots of both the numerical solution and analytical solution are given in Figure 3.24. The code for solving this system numerically and for plotting the analytical solution in Mathcad follows: x0 := 0.01 k := 1000 ωb := 3 v0 := 3 ωn := k ζ := Cm ωd := ωn 21 - ζ2 m := 1 c 2 # 1k # m ωd = 31.22499 A := 0.09341 ϕ := - 0.1074 c := 10 ωn=31.623 θ := 1.53 t0 := 0 Y := 0.05 Sec. 3.9 275 numerical Simulation of the Response 0.1 0.05 xn x -- 0 1 2 4 3 5 0.05 0.1 t Figure 3.24 The numerical solution (solid line) and analytical solution (dashed line) for the system of Example 3.3.2 indicating almost perfect agreement. # # xa(t) := A # e-ζ ωn t sin(ωd # t - ϕ) + 0.05 # cos(3 # t - θ) D(t, X) := £ X := J X1 c - 2 # ζ # ωn # X1 - ωn2 # X0 + m # Y # ωb # cos(ωb # t) + k m x0 d v0 # Y # sin(ωb # t) Z := rkfixed (X, 0, 6, 2000, D) > x :xa(t) t := Z 6 0 7 xn := Z 6 1 7 The MATLAB code for solving for the response is clear all %% Analytical solution t=0:0.01:5; xt=0.09341*exp(-5*t).*sin(31.225*t+0.1074)+0.05*cos(3*t-1.571); plot(t, xt,'--');’hold on’ %% Numerical Solution xo=[0.01; 3]; ts=[0 5]; [t, x]=ode45('f', ts, xo);plot(t, x(:,1)); hold off %--------------------------------------------function v=f(t, x) m=1; c=10; k=1000; wb=3; wn=sqrt(k/m); zeta=c/2*sqrt(m*k); wd=wn*sqrt(1-zeta^2); Y=0.05; v=[x(2); x(2).*-2*zeta*wn + x(1).*-wn^2 + c/m*Y*wb*cos(wb*t) +... k/m*Y*sin(wb*t)]; § 276 general Forced Response Chap. 3 The Mathematica code for solving for the response is In[1]:= 6 6 PlotLegends' In[2]:= xanal[t_]= .09341 * Exp[-5 * t] * Sin[31.225 * t + .1074] + .05 * Cos[3 * t - 1. 571]; xnum = NDSolve[{x''[t] + 10 * x'[t] + 1000 * x[t] == 1.5 * Cos[3 * t] + 50 * Sin[3 * t], x[0] == .01, x’[0] == 3}, x[t], {t, 0, 4.5}]; Plot [{Evaluate [x[t] /. xnum], xanal[t]}, {t, 0, 4.5}, PlotStyle S {RGBColor[1, 0, 0], RGBColor[O, 1, 0]}, PlotLegend S {“Numerical”, “Analytical”}, LegendPosition S {1, 0}, LegendSize S {1, .5}]; n Example 3.9.4 Consider the problem of Example 3.6.1 and compute the response numerically. Compare the result to the analytical solution computed in Example 3.6.1 by plotting both the analytical solution and the numerical solution on the same graph. Solution The equation of motion to be solved in this example is t - t1 t $ - a bΦ(t - t 1) 10x(t) + 10x(t) = t1 t1 (3.110) sin ωn(t - t 1) F0 t sin ωnt F0 t - t 1 a b a bΦ(t - t 1) k t1 ωn t 1 k t1 ωn t 1 (3.111) where Φ is the Heaviside step function used to “turn on” the second term in the forcing function at t 7 t1. The parameter t1 is used to control how steep the disturbance is. See Figure 3.16 for a plot of the driving force. The analytical solution is given in equation (3.78) to be x(t) = Again, the equation of motion can be solved numerically in the form given by equation (3.110) directly in Mathematica or by putting equation (3.110) into state-space form and solving Runge–Kutta in Mathcad or MATLAB. The plots of both the numerical solution and analytical solution are given in Figure 3.25. The Mathcad code for this solution follows: x0 := 0 k := 10 tp := v0 := 0 ωn := m := 10 k Cm ωn = 1 1 - cos (ωn # t1) 1 # atan a b ωn sin (ωn # t1) xm(t) := F0 # 1 c1 + ωn # t1 k t1 := 1 F0 := 1 ωn # tp = 0.5 # 32 # (1 - cos (ωn # t1)) d tp = 0.5 Sec. 3.9 277 numerical Simulation of the Response 0.2 0.15 x xn 0.1 0.05 0 5 10 15 t Figure 3.25 The numerical solution (solid line) and analytical solution (dashed line) for the system of Example 3.6.1 indicating almost perfect agreement. xa(t) := sin (ωn # t) sin[ωn # (t - t1)] F0 t - t1 F0 # t a c b d Φ(t - t1) # t1 ωn t1 t1 ωn # t1 k k f(t) := t # F0 t - t1 # F0 # Φ(t - t1) t1 m t1 m D(t, X) := J X1 d - (ωn2 # X0) + f(t) 0 X := c d 0 xm(0) = 0.196 Z := rkfixed (X, 0, 15, 2000, D) t := Z 6 0 7 > x := xa(t) xn := Z 6 1 7 > F := f(t) > Xmax := xm(t) The MATLAB code for solving for the response is clear all %% analytical solution t=0:0.01:15; m=10; k=10; Fo=1; t1=1; wn=sqrt(k/m); Heaviside=stepfun(t, t1);% define Heaviside Step function for 0 6 t 6 15 278 general Forced Response Chap. 3 for i=1:max(length(t)), xt(i)=Fo/k*(t(i)/t1 - sin(wn*t(i))/wn/tl) - Fo/k*((t(i)-t1)/t1 – sin(wn*(t(i)-t1))/wn*t1)*Heaviside(i); end plot(t,xt,'- -'); hold on %% Numerical Solution xo=[0; 0]; ts=[0 15]; [t, x]=ode45('f', ts, xo); plot(t, x(:,1)); hold off %-------------------------------------------function v=f(t, x) m=10; k=10; wn=sqrt(k/m); Fo=1; t1=1; v=[x(2); x(1).*-wn^2 + t/t1*Fo/m-(t-t1)/tl*Fo/m*stepfun(t, t1)]; The Mathematica code for solving for the response is In[1]:= 6 6 PlotLegends' In[2]:= x0 = 0; v0 = 0; m = 10; k = 10; ωn = k Cm ; t1 = 1; F0 = 1; tp = 0.5; In[10]:= xanal[t_] = Sin[ωn*(t - t1)] Sin[ωn*t] t F0 t - t1 F0 *a b*a b t1 ωn*t1 t1 ωn*t1 k k * UnitStep[t - t1]; xnum = NDSolve[{10 * x''[t] + 10 * x[t] == t t1 - a t - t1 b t1 * UnitStep[t - t1], x[0] == x0, x'[0] == v0}, x[t], {t, 0, 15}]; Plot[{Evaluate[x[t] /. xnum], xanal[t]}, {t, 0, 15}, PlotStyle S {RGBColor[1, 0, 0], RBColor[0, 1, 0]}, PlotLegend S {"Numerical", "Analytical"}, LegendPosition S {1, 0}, LegendSize S {1, 0.5}]; n Sec. 3.10 nonlinear Response Properties 279 Using numerical integration, the response of a system to a variety of different forcing functions may be computed relatively easily. Furthermore, once the solution is programmed, it is a trivial matter to change parameters and solve the system again. By visualizing the response through simple plots, one can gain the ability to design and understand the system’s dynamic behavior. 3.10 NONLINEAR RESPONSE PROPERTIES The use of numerical integration as introduced in the previous section allows us to consider the effects of various nonlinear terms in the equation of motion. As noted in the free-response case discussed in Section 1.10 and the response to a harmonic load discussed in Section 2.9, the introduction of nonlinear terms results in an inability to find exact solutions, so we must rely on numerical integration and qualitative analysis to understand the response. Several important differences between linear and nonlinear systems are outlined in Section 2.9. In particular, when working with nonlinear systems, it is important to remember that a nonlinear system has more than one equilibrium point and each may be either stable or unstable. Furthermore, we cannot use the idea of superposition, used in all of the previous sections of this chapter, in a nonlinear system. Many of the nonlinear phenomena are very complex and require analysis skills beyond the scope of a first course in vibration. However, some initial understanding of nonlinear effects in vibration analysis can be observed by using the numerical solutions covered in the previous section. In this section, several simulations of the response of nonlinear systems are numerically computed and compared to their linear counterparts. Recall from Section 2.9 that if the equations of motion are nonlinear, the general single-degree-of-freedom system may be written as $ # x(t) + f(x(t), x(t)) = F(t) (3.112) where the function f can take on any form, linear or nonlinear, and the forcing term F(t) can be almost anything (periodic or not), each term of which has been divided by the mass. Formulating this last expression into the state space, or first-order, equation (3.112) takes on the form # x1(t) = x2(t) # x2(t) = -f(x1, x2) + F(t) (3.113) This state-space form of the equation is used for numerical simulation in several of the codes. By defining the state vector, x = [x1(t) x2(t)]T, used in equation (3.113) and the nonlinear vector function F as x2 (t) F(x) = J-f(x , x ) d 1 2 (2.140) 280 general Forced Response equations (3.113) may now be written in the first-order vector form # x = F(x) + f(t) Chap. 3 (3.114) Equation (3.114) is the forced version of equation (1.115). Here f(t) is simply f(t) = c 0 d F(t) (3.115) Then the Euler integration method for the equations of motion in the first-order form becomes x(t i + 1) = x(t i) + F(x(t i))∆t + f(t i)∆t (3.116) This expression forms a basic approach to integrating numerically to compute the forced response of a nonlinear system and is the nonlinear, general forced-response version of equations (1.98) and (2.134). This is basically identical to equation (2.143) except for the interpretation of the force f. Nonlinear systems are difficult to analyze numerically as well as analytically. For this reason, the results of a numerical simulation must be examined carefully. In fact, use of a more sophisticated integration method, such as Runge–Kutta, is recommended for nonlinear systems. In addition, checks on the numerical results using qualitative behavior should also be performed whenever possible. In the following examples, consider the single-degree-of-freedom system illustrated in Figure 3.26, with a nonlinear spring element subject to a general driving force. A series of examples are presented using numerical simulation to examine the behavior of nonlinear systems and to compare them to the corresponding linear systems. x(t) k m c F(t) Figure 3.26 A spring–mass–damper system with potentially nonlinear elements and general applied force. Example 3.10.1 Compute the response of the system in Figure 3.26 for the case that the damping is linear viscous, the spring is a nonlinear “hardening” spring of the form k(x) = kx + k1x3 (3.117) and the system is subject to an applied excitation of the form F(t) = 1500[Φ(t - t 1) - Φ(t - t 2)]N (3.118) and initial conditions of x0 = 0.01 m and 0 = 1 m>s. Here Φ denotes the Heaviside step function and the times t1 = 1.5 s and t2 = 5 s. This driving function is plotted in Sec. 3.10 281 nonlinear Response Properties 20 F(t) 10 0 1 2 3 4 5 6 t Figure 3.27 The pulse input function defined by equation (3.118). Figure 3.27. The system has a mass of 100 kg, a damping coefficient of 20 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 300 N>m3. Compute the solution and compare it to the linear solution (k1 = 0). Solution Summing forces in the horizontal direction, the equation of motion becomes $ # mx(t) + cx(t) + kx(t) + k1x3(t) = 1500[Φ(t - t 1) - Φ(t - t 2)] Dividing by the mass yields $ # x(t) + 2ζωnx(t) + ω2nx(t) + αx3(t) = 15[Φ(t - t 1) - Φ(t - t 2)] Next write this equation in state-space form to get # x1(t) = x2(t) # x2(t) = - 2ζωnx2(t) - ω2nx1(t) - αx31(t) + 15[Φ(t - t 1) - Φ(t - t 2)] This last set of equations can be used in MATLAB or Mathcad to integrate numerically for the time response. Mathematica uses the second-order equation directly. Figure 3.28 illustrates the response to both the linear and nonlinear system. The solid line is the response of the nonlinear system while the dashed line is the response of the linear system. The difference between linear and nonlinear systems is that, in this case, the nonlinear spring has smaller response amplitude than the linear system does. This is useful in design as it illustrates that the use of a hardening spring reduces the amplitude of vibration to a shock type of input. One possibility for designing a nonlinear isolation spring is to use the numerical codes listed later in this example to vary parameters (damping, mass, and stiffness) until a desired response is obtained. It is important to remember, however, that if designing with a nonlinear element, new equilibriums are introduced that may be unstable. Hence care must be taken to assure that additional difficulties are not introduced when using a nonlinear spring to reduce the response. The following codes can be used to generate and plot the solution just given. 282 general Forced Response Chap. 3 Displacement (m) 2 1 0 2 4 6 8 1 t(s) 2 Figure 3.28 The response of the system of Figure 3.26 to the input force given in Figure 3.27. The solid line is the response of the nonlinear system while the dashed line is the response of the linear system. In Mathcad, the code is x0 := 0.01 v0 := 1 α := 3 ωn := X := c k Cm x0 d v0 F0 := 1500 ζ := Y := X D(t, X) := J m :=100 c t1 := 1.5 f0 := 22k # m k := 2000 F0 m c := 20 t2 := 5 ζ = 0.022 f(t) := f0 # Φ(t - t1) - f0 # Φ(t - t2) X1 - 2 # ζ # ωn # X1 - ωn2 # X0 + [- α # (X0)3 + f(t)] Y1 L(t, Y) := J # # d - 2 ζ ωn # Y1 - ωn2 # Y0) + f(t) Z := rkfixed (X, 0, 10, 2000, D) t := Z 607 xs := Z d W := rkfixed (Y, 0, 10, 2000, L) 617 xL:= W 6 1 7 The MATLAB code is clear all xo=[0.01; 1]; ts=[0 8]; [t,x]=ode45('f',ts,xo); plot(t, x(:,1)); hold on % The response of nonlinear system [t,x]=ode45('f1',ts,xo); plot(t,x(:,1),'——'); hold off % The response of linear system Sec. 3.10 nonlinear Response Properties 283 %--------------------------------------------function v=f(t,x) m=100; k=2000; c=20; wn=sqrt(k/m); zeta=c/2/sqrt(m*k); Fo=1500; alpha=3; t1=1.5; t2=5; v=[x(2); x(2).*-2*zeta*wn + x(1).*-wn^2 - x(1)^3.*alpha+... Fo/m*(stepfun(t,t1)-stepfun(t,t2))]; %--------------------------------------------function v=f1(t,x) m=100; k=2000; c=20; wn=sqrt(k/m); zeta=c/2/sqrt(m*k); Fo=1500; alpha=0; t1=1.5; t2=5; v=[x(2); x(2).*-2*zeta*wn + x(1).*-wn^2 - x(1)^3.*alpha+... Fo/m*(stepfun(t,t1)-stepfun(t,t2))]; The Mathematica code is In[1]:= 6 6 PlotLegends' In[2]:= x0 = .01; v0 = 1; m = 100; k = 2000; k1 = 300; c = 20; ωn = k Cm ; k1 ; m t1 = 1.5; t2 = 5; F0 = 1500; α = f0 = ζ = F0 ; m c 2* 1k*m ; In[21]:= x1in = NDSolve [{x1''[t] + 2 * ζ * ωn * x1'[t] + ωn2 * x1[t] == 15 * (UnitStep[t - t1] - UnitStep[t - t2]), x1[0] == x0, x1'[0] == v0}, x1[t], {t, 0, 8}]; xnon1 = NDSolve [{xn1''[t] + 2 * ζ * ωn * xn1'[t] + ωn2 * xn1[t] + α * (xn1[t])^3 == 15 * (UnitStep[t - t1] - UnitStep[t - t2]), xn1[0] == x0, xn1'[0] == v0}, xn1[t], {t, 0, 8}, Method S “ExplicitRungeKutta”]; Plot[{Evaluate[x1[t] /. x1in], Evaluate[xn1[t] /. xnon1]}, {t, 0, 8}, PlotRange S {-2, 2}, 284 general Forced Response Chap. 3 PlotStyle S {RGBColor[1, 0, 0], RGBColor[0, 1, 0]}, PlotLegend S {"Linear", "NonLinear"}, LegendPosition S {1, 0}, LegendSize S {1, .5}] n Example 3.10.2 Compare the forced response of a system with velocity-squared damping, as defined in equation (2.129) using numerical simulation of the nonlinear equation, to that of the response of the linear system obtained using equivalent viscous damping, as defined by equation (2.131), where the input force is given by F(t) = 150[Φ(t - t 1) - Φ(t - t 2)]m and initial conditions are x0 = 0 and 0 = 1 m>s. Here Φ denotes the Heaviside step function and the times t1 = 0.5 s and t2 = 2 s. Solution This is essentially the same as Example 2.9.2 except that the driving force is a pulse here rather than harmonic as in Section 2.9. Velocity-squared damping with a linear spring and pulsed input is described by $ # # mx + α sgn(x)x2 + kx = 150[Φ(t - t 1) - Φ(t - t 2)] The equivalent viscous-damping coefficient is calculated in equation (2.131) to be ceq = 8 αωX 3π This value assumes that the motion is harmonic, which is somewhat violated here. The value of the magnitude, X, can be approximated for near resonance conditions. The value is computed in Example 2.9.2 to be X = 3πmf0 B 8αω2 Combining these last two expressions yields an equivalent viscous-damping value of ceq = 8mα f0 B 3π Using this value as the damping coefficient results in a linear system that approximates equation (2.131). Figure 3.29 illustrates the linear and nonlinear response for the values m = 10 kg, k = 200 N>m, and α = 5. Note that the linear response underestimates the actual response maximum by about 10%. For large times, the response of the linear system dies out much sooner than that of the nonlinear system, indicating a large error. 285 nonlinear Response Properties 1 Displacement (m) Sec. 3.10 0.5 0 2 4 6 8 0.5 t(s) Figure 3.29 The response of a nonlinear system (solid line) and a linear system (dashed line) formed by using the concept of equivalent viscous damping. The computer codes for generating the solutions and plotting Figure 3.29 are given next. The Mathcad code is x0 := 0 v0 := 1 m := 10 ωn := t1 := 0.5 X := c x0 d v0 k Y := X ζ := f0 := Cm t2 := 2 ceq := ceq - ωn L(t, Y) := J 2 F0 := 150 F0 m 8#α#m# C 3#π f0 f(t) := f0 # Φ(t - t1) - f0 # Φ(t - t2) 21k # m D(t, X) := £ α := 5 k := 200 # X0 X1 § X1 α # (X1)2 # + f(t) m 0 X1 0 1 - 2 # ζ # ωn # Y1 Y1 d - ωn2 # Y0 2 + f(t) Z := rkfixed (X, 0, 20, 2000, D) W :=rkfixed (Y, 0, 20, 2000, L) t := Z 6 0 7 x := Z 6 1 7 xL:= W 6 1 7 286 general Forced Response Chap. 3 The MATLAB code is clear all xo=[0; 1]; ts=[0 8]; [t,t]=ode45('f',ts,xo); plot(t,x(:,1)); hold on % The response of nonlinear system [t,x]=ode45('f1',ts,xo); plot(t,x(:,1),'--'); hold off % The response of linear system %--------------------------------------------function v=f(t,x) m=10; k=200; wn=sqrt(k/m); Fo=150; alpha=5; t1=0.5; t2=2; v=[x(2); x(1).*-wn^2 + x(2)^2.*-alpha/m.*sign(x(2))+... Fo/m*(stepfun(t, t1)-stepfun(t,t2))]; %--------------------------------------------function v=f1(t,x) m=10; k=200; wn=sqrt(k/m); Fo=150; alpha=5; t1=0.5; t2=2; ceq=sqrt(8*alpha*m/3/pi*Fo/m); zeta=ceq/2/sqrt(m*k); v=[x(2); x(2).*-2*zeta*wn + x(1).*-wn^2+... Fo/m*(stepfun(t,t1)-stepfun(t,t2))]; The Mathematica code is In[1]:= 6 6 PlotLegends' In[2]:= x0 = 0; v0 = 1; m = 10; k = 200; α =5; F0 = 150; f0 = ωn = ceq = ζ = k Cm ; t1 = 0.5; t2 = 2; 8*α*m C 3*π ceq 2 * 1k * m *f0; ; α * Sign[x'[t]] m * (x'[t])^2 + wn^2 * x[t] == f0 * (UnitStep[t - t1] - UnitStep[t - t2]), x[0] == 0, x’[0] == 1}, x[t], {t, 0, 20}]; In[6]:= xnonlin = NDSolve[{x''[t] + F0 ; m Chap. 3 287 Problems In[7]:= xlin = NDSolve[{x''[t] + 2 * ζ * ωn * x'[t] + ωn^2 * x[t] == f0 * (UnitStep[t - t1] - UnitStep[t - t2]), x[0] == 0, x'[0] == 1}, x[t], {t, 0, 20}]; In[8]:= Plot[{x[t]/.xnonlin, x[t]/.xlin},{t, 0, 8}, PlotStyle S {GrayLevel [0], Dashing[{.03}]}] n PROBLEMS Those problems marked with an asterisk are intended to be solved using computational software. Section 3.1 (Problems 3.1 through 3.17) 3.1. Calculate the solution to $ # 1000x(t) + 200x(t) + 2000x(t) = 100δ(t), x0 = 0, v0 = 0 3.2. Consider a spring–mass–damper system with m = 1 kg, c = 2 kg>s, and k = 2000 N>m with an impulse force applied to it of 10,000 N for 0.01 s. Compute the resulting response. 3.3. Calculate the solution to $ # x + 2x + 2x = δ(t - π) # x(0) = 1 x(0) = 0 and plot the response. 3.4. Calculate the solution to $ # x + 2x + 3x = sin t + δ(t - π) # x(0) = 0 x(0) = 1 and plot the response. 3.5. Calculate the response of a critically damped system to a unit impulse. 3.6. Calculate the response of an overdamped system to a unit impulse. 3.7. Derive equation (3.6) from equations (1.36) and (1.38). 3.8. Consider a simple model of an airplane wing given in Figure P3.8. The wing is approximated as vibrating back and forth in its plane and is massless compared to 288 general Forced Response Chap. 3 Carriage l Missile Top view with wing modeled as cantilevered (b) Jet with missiles (a) I k  3EI l3 E F (t) k c m m x(t) Simplified beam model (c) F (t) x(t) Vibration model (d) Figure P3.8 Modeling of wing vibration resulting from the release of a missile. Figure (a) is the system of interest; (b) is the simplification of the detail of interest; (c) is a crude model of the wing: a cantilevered beam section (recall Figure 1.26); and (d) is the vibration model used to calculate the response neglecting the mass of the wing. the missile carriage system (of mass m). The modulus and the moment of inertia of the wing are approximated by E and I, respectively, and l is the length of the wing. The wing is modeled as a simple cantilever for the purpose of estimating the vibration resulting from the release of the missile, which is approximated by the impulse function F δ(t). Calculate the response and plot your results for the case of an aluminum wing 2-m long with m = 1000 kg, ζ = 0.01, and I = 0.5 m4. Model F as 1000 N lasting for 10–2s. 3.9. A cam in a large machine can be modeled as applying a 10,000 N-force over an interval of 0.005 s. This can strike a valve that is modeled as having the physical parameters m = 10 kg, c = 18 N s>m, and stiffness k = 9000 N>m. The cam strikes the valve once every 1 s. Calculate the vibration response, x(t), of the valve once it has been impacted by the cam. The valve is considered to be closed if the distance between its rest position and its actual position is less than 0.0001 m. Is the valve closed the very next time it is hit by the cam? 3.10. The vibration of a package dropped from a height of h meters can be approximated by considering Figure P3.10 and modeling the point of contact as an impulse applied to the system at the time of contact. Calculate the vibration of the mass m after the system falls and hits the ground. Assume that the system is underdamped. Chap. 3 289 Problems x(t) m c k Figure P3.10 The vibration model of a package being dropped onto the ground. 3.11. Calculate the response of $ # 3x(t) + 12x(t) + 12x(t) = 3δ(t) for zero initial conditions. The units are in Newtons. Plot the response. 3.12. Compute the response of the system $ # 3x(t) + 12x(t) + 12x(t) = 3δ(t) subject to the initial conditions x(0) = 0.01 m and (0) = 0. The units are in Newtons. Plot the response. 3.13. Calculate the response of the system $ # 3x(t) + 6x(t) + 12x(t) = 3δ(t) - δ(t - 1) subject to the initial conditions x(0) = 0.01 m and (0) = 1 m>s. The units are in Newtons. Plot the response. 3.14. A chassis dynamometer is used to study the unsprung mass of an automobile as illustrated in Figure P3.14, and discussed in Example 1.4.1. Compute the maximum magnitude of the center of the wheel due to an impulse of 5000 N applied over 0.01 seconds in the x direction. Assume the wheel mass is m = 15 kg, the spring stiffness is k = 500,000 k c (t) r m, J x(t) Figure P3.14 A simple model of an automobile suspension system mounted on a chassis dynamometer. The rotation of the car’s wheel>tire assembly (of radius r) is given by θ(t) and is vertical deflection by x(t). 290 general Forced Response Chap. 3 N>m, the shock absorber provides a damping ratio of ζ = 0.3, and the rotational inertia is J = 2.323 kg m2. Assume that the dynamometer is controlled such that x = rθ. Compute and plot the response of the wheel system to an impulse of 5000 N over 0.01 s. Compare the undamped maximum amplitude to that of the maximum amplitude of the damped system (use r = 0.457 m). 3.15. Consider the effect of damping on the bird strike problem of Example 3.1.2. Recall from the example that the bird strike causes the camera to vibrate out of limits. Adding damping will cause the magnitude of the response to decrease but may not keep the camera from vibrating past the 0.01 m limit. If the damping in the aluminum is modeled as ζ = 0.05, approximately how much time will pass before the camera vibration reduces to the required limit? (Hint: Plot the time response and note the value for time after which the oscillations remain below 0.01 m.) 3.16. Consider the jet engine and mount indicated in Figure P3.16 and model it as a mass on the end of a beam, as done in Figure 1.26. The mass of the engine is usually fixed. Find an expression for the value of the transverse mount stiffness, k, as a function of the relative speed of the bird, , the bird mass, the mass of the engine, and the maximum displacement that the engine is allowed to vibrate. Wing, ground Mount, k F(t) x(t) Figure P3.16 A model of a jet engine in transverse vibration due to a bird strike. Engine, m 3.17. A machine part is regularly subject to a force of 350 N lasting 0.01 seconds, as part of a manufacturing process. Design a damper (i.e., choose a value of the damping constant, c, such that the part does not deflect more that 0.01 m), given that the part has a mass of 100 kg and a stiffness of 1250 N>m. Section 3.2 (Problems 3.18 through 3.29) 3.18. Calculate the analytical response of an overdamped single-degree-of-freedom system to an arbitrary nonperiodic excitation. 3.19. Calculate the response of an underdamped system to the excitation given in Figure P3.19 where the pulse ends at π s. f(t) F0 0  t 2 Figure P3.19 Plot of a pulse input of the form f(t) = F0 sin t. Chap. 3 291 Problems 3.20. Speed bumps are used to force drivers to slow down. Figure P3.20 is a model of a car going over a speed bump. Using the data from Example 2.4.2 and an undamped model of the suspension system (i.e., k = 4 * 105 N>m, m = 1007 kg), find an expression for the maximum relative deflection of the car’s mass versus the velocity of the car. Model the bump as a half sine of length 40 cm and height 20 cm. Note that this is a movingbase problem. x(t) m k Velocity v y(t)  Y sin bt Y  20 cm Speed bump 40 cm Figure P3.20 Model of a car driving over a speed bump. 3.21. Calculate and plot the response of an undamped system to a step function with a finite rise time of t1 for the case m = 1 kg, k = 1 N>m, t1 = 4 s, and F0 = 20 N. This function is described by F0t F(t) = c t 1 F0 0 … t … t1 t 7 t1 3.22. A wave consisting of the wake from a passing boat impacts a seawall. It is desired to calculate the resulting vibration. Figure P3.22 illustrates the situation and suggests a model. The force in Figure P3.22 can be expressed as F(t) = c F0 a1 0 t b t0 0 … t … t0 t 7 t0 Calculate the response of the seawall–dike system to such a load. 292 general Forced Response Chap. 3 Dike F(t) Concrete seawall k m k  EA l Water F m F Wake Physical setting 0 Model t Input model Figure P3.22 A wave hitting a seawall modeled as a nonperiodic force exciting an undamped single-degree-of-freedom, spring–mass system. 3.23. Determine the response of an undamped system to a ramp input of the form F(t) = F0 t, where F0 is a constant. Plot the response for three periods for the case m = 1 kg, k = 100 N>m, and F0 = 50 N. 3.24. A machine resting on an elastic support can be modeled as a single-degree-of-freedom, spring–mass system arranged in the vertical direction. The ground is subject to a motion y(t) of the form illustrated in Figure P3.24. The machine has a mass of 5000 kg, and the support has stiffness 1.5 * 103 N>m. Calculate the resulting vibration of the machine. y(t) (mm) 0.5 0.2 0.6 t (s) Figure P3.24 Triangular pulse input function. 3.25. Consider the step response described in Figure 3.7 and Example 3.2.1. Calculate the analytical value of tp by noting that it occurs at the first peak, or critical point, of the curve. 3.26. Calculate the value of the overshoot (O.S.), for the system of Example 3.2.1. Note from the example that the overshoot is defined as occurring at the peak time defined by t p = π/ωd and is the difference between the value of the response at tp and the steadystate response at tp. 3.27. It is desired to design a system so that its step response has a settling time of 3 s and a time to peak of 1 s. Calculate the appropriate natural frequency and damping ratio to use in the design. 3.28. Plot the response of a spring–mass–damper system to a square input of magnitude F0 = 30 N, illustrated in Figure 3.8 of Example 3.2.1, for the case that the pulse width is Chap. 3 293 Problems the natural period of the system (i.e., t1 = π ωn). Recall that k = 1000 N>m, ζ = 0.1, and ωn = 3.16 rad>s. 3.29. Consider the spring–mass system described by $ mx(t) + kx(t) = F0 sin ωt, x0 = 0.01 m and v0 = 0 Compute the response of this system for the values of m = 100 kg, k = 2500 N>m, ω = 10 rad>s, and F0 = 10 N, using the convolution integral approach outlined in Example 3.2.4. Check your answer using the results of equation (2.25). Section 3.3 (Problems 3.30 through 3.38) 3.30. Derive equations (3.24), (3.25), and (3.26) and hence verify the equations for the Fourier coefficient given by equations (3.21), (3.22), and (3.23). 3.31. Calculate bn from Example 3.3.1 for the triangular force given by 4 t - 1 T F(t) = d 4 T 1 at b T 2 0 … t … T 2 T … t … T 2 and show that bn = 0, n = 1,2,…,∞ . Also verify the expression an by completing the integration indicated. (Hint: Change the variable of integration from t to x = 2πnt>T.) 3.32. Determine the Fourier series for the rectangular wave illustrated in Figure P3.32. f(t) 1 2  3 t Figure P3.32 A rectangular periodic signal. 1 3.33. Determine the Fourier series representation of the sawtooth curve illustrated in Figure P3.33. f(t) 1 2 4 6 8 t Figure P3.33 A sawtooth periodic signal. 294 general Forced Response Chap. 3 3.34. Calculate and plot the response of the base-excitation problem with base motion specified by the velocity # y(t) = 3e -t>2Φ(t) m>s where Φ(t) is the unit step function and m = 10 kg, ζ = 0.01, and k = 1000 N>m. Assume that the initial conditions are both zero. 3.35. Calculate and plot the total response of the spring–mass–damper system with m = 100 kg, ζ = 0.1, and k = 1000 N>m to the signal defined by 4 t - 1 T F(t) = d 4 T 1 at b T 2 0 … t … T 2 T … t … T 2 with maximum force of 1 N. Assume that the initial conditions are zero, and let T = 2π s. 3.36. Calculate the total response of the system of Example 3.3.2 for the case of a base motion driving frequency of ωb = 3.162 rad>s with amplitude Y = 0.05 m subject to initial conditions x0 = 0.01 m and 0 = 3.0 m>s. The system is defined by m = 1 kg, c = 10 kg>s, and k = 1000 N>m. *3.37. Validate your solution to the square wave Problem 3.32 by calculating an and bn using VTB3_3 in the Vibration Toolbox. Print the function and its Fourier series approximation for 5, 20, then 100 terms. The Toolbox makes this easy. The purpose is to illustrate the Gibbs effect in approximation by Fourier series. *3.38. Validate your solution to the sawtooth wave of Problem 3.33 by calculating an and bn using VTB3_3 in the Vibration Toolbox. Print the function and its Fourier series approximation for 5, 20, and 100 terms. The Toolbox makes this easy. The purpose is to illustrate the Gibbs effect in approximation by Fourier series. Section 3.4 (Problems 3.39 through 3.43) 3.39. Calculate the response of $ # mx + cx + kx = F0Φ(t) where ϕ(t) is the unit step function for the case with x0 = 0 = 0. Use the Laplace transform method and assume that the system is underdamped. 3.40. Using the Laplace transform method, calculate the response of the system $ # mx(t) + cx(t) + kx(t) = δ(t), x0 = 0, v0 = 0 for the overdamped case (ζ 7 1). Plot the response for m = 1 kg, k = 100 N>m, and ζ = 1.5. 3.41. Calculate the response of the underdamped system given by $ # mx + cx + kx = F0e -at using the Laplace transform method. Assume a 7 0 and the initial conditions are both zero. Chap. 3 295 Problems 3.42. Solve the following system for the response x(t) using Laplace transforms: $ 100x(t) + 2000x(t) = 50δ(t) where the units are in Newtons and the initial conditions are both zero. 3.43. Use the Laplace transform approach to solve for the response of the spring–mass system with equation of motion and initial conditions given by $ x(t) + x(t) = sin 2t, x0 = 0, v0 = 1 Assume the units are consistent. (Hint: See the example in Appendix B.) Section 3.5 (Problems 3.44 through 3.48) 3.44. Calculate the mean-square response of a system to an input force of constant PSD, S0, and frequency response function H(ω) = 10>(3 + 2jω). 3.45. Consider the base-excitation problem of Section 2.4 as applied to an automobile model of Example 2.4.2 and illustrated in Figure 2.17. Recall that the model is a spring–mass– damper system with values m = 1007 kg, c = 2000 kg>s, k = 40,000 N>m. In this problem let the road have a random stationary cross section producing a PSD of S0. Calculate the PSD of the response and the mean-square value of the response. 3.46. To obtain a feel for the correlation functions, compute autocorrelation Rxx(τ) for the deterministic signal A sin ωnt. 3.47. The autocorrelation of a signal is given by Rxx(τ) = 10 + 4 3 + 2τ + 4τ2 Compute the mean-square value of the signal. 3.48. Verify that the average x - x∙ is zero by using the definition given in equation (3.47) to compute the average. Section 3.6 (Problems 3.49 through 3.50) 3.49. A power line pole with a transformer is modeled by $ $ mx + kx = -y where x and y are as indicated in Figure P3.49. Assuming the initial conditions are zero, calculate the response of the relative displacement (x – y) if the pole is subject to an earthquake-based excitation of t Aa1 b $ t0 y(t) = c 0 0 … t … 2t 0 t 7 2t 0 296 general Forced Response Chap. 3 m  Effective mass x(t) k  Effective stiffness Figure P3.49 A vibration model of a power-line pole with a transformer mounted on it. y(t) 3.50. Calculate the response spectrum of an undamped system to the forcing function F(t) = c F0 sin πt t1 0 0 … t … t1 t 7 t1 assuming the initial conditions are zero. Section 3.7 (Problems 3.51 through 3.58) 3.51. Using complex algebra, derive equation (3.89) from (3.86) with s = jω. 3.52. Using the plot in Figure P3.52, estimate the system’s parameters m, c, and k, as well as the natural frequency. Mag H( j) 5 1  c 1 k 0 0.01 1 0.1   n 1  22 log () 10  n  k/m Figure P3.52 The magnitude plot of a spring–mass–damper system. Chap. 3 297 Problems 3.53. From a compliance transfer function of a spring–mass–damper system the stiffness is determined to have a value of 0.5 N>m, a natural frequency of 0.25 rad>s, and a damping coefficient of 0.087 kg>s. Plot the inertance transfer function’s magnitude and phase for this system. 3.54. From a compliance transfer function of a spring–mass–damper system the stiffness is determined to have a value of 0.5 N>m, a natural frequency of 0.25 rad>s, and a damping coefficient of 0.087 kg>s. Plot the mobility transfer function’s magnitude and phase for the system. 3.55. Calculate the compliance transfer function for a system described by a d 4x(t) dt 4 + b d 3x(t) dt 3 + c d 2x(t) dt 2 + dx(t) dt + ex(t) = f(t) where f(t) is the input force and x(t) is a displacement. 3.56. Calculate the frequency response function for the compliance for the system defined by a d 4x(t) dt 4 + b d 3x(t) dt 3 + c d 2x(t) dt 2 + dx(t) dt + ex(t) = f(t). *3.57. Plot the magnitude of the frequency response function for the system of Problem 3.56 for a = 1, b = 4, c = 11, d = 16, and e = 8. 3.58. An experimental (compliance) magnitude plot is illustrated in Fig. P3.58. Determine ω, ζ, c, m, and k. Assume that the units correspond to m>N along the vertical axis. Mag H(j) 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 0.1 1 log () 10 Figure P3.58 An experimentally determined compliance magnitude plot. 298 general Forced Response Chap. 3 Section 3.8 (Problems 3.59 through 3.64) 3.59. Show that a critically damped system is bounded-input, bounded-output stable. 3.60. Show that an overdamped system is bounded-input, bounded-output stable. $ 3.61. Is the solution of 2x + 18x = 4 cos 2t + cos t Lagrange stable? 3.62. Calculate the response of the system described by $ # # x(t) + x(t) + 4x(t) = ax(t) + bx(t) for x0 = 0,v0 = 1 for the case that a = 4 and b = 0. Is the response bounded? 3.63. A crude model of an aircraft wing can be modeled as $ # # 100x(t) + 25x(t) + 2000x(t) = ax(t) Here the factor a is determined by the aerodynamics of the wing and is proportional to the air speed. At what value of the parameter a will the system start to flutter? 3.64. Consider the inverted pendulum of Figure P3.64 and compute the value of the stiffness k that will keep the linear system stable. Assume that the pendulum rod is massless. + m m k k 2 kl sin  2 l 2 mg  O (a) fp O (b) Figure P3.64 An inverted pendulum. Section 3.9 (Problems 3.65 through 3.72) *3.65. Numerically integrate and plot the response of an underdamped system determined by m = 100 kg, k = 1000 N>m, and c = 20 kg>s, subject to the initial conditions of x0 = 0 and 0 = 0, and the applied force F(t) = 30Φ(t - 1). Then plot the exact response as computed by equation (3.17). Compare the plot of the exact solution to the numerical simulation. *3.66. Numerically integrate and plot the response of an underdamped system determined by m = 150 kg and k = 4000 N>m subject to the initial conditions of x0 = 0.01 m and 0 = 0.1 m>s, and the applied force F(t) = Φ(t) = 15 (t - 1), for various values of the damping coefficient. Use this “program” to determine a value of damping that causes the transient term to die out within 3 seconds. Try to find the smallest such value of damping remembering that added damping is usually expensive. *3.67. Calculate the total response of the base isolation problem given in Example 3.3.2, with the parameters ωb = 3 rad>s, m = 1 kg, c = 10 kg>s, k = 1000 N>m, and Y = 0.05 m, subject to initial conditions x0 = 0.01 m and 0 = 3.0 m>s, by numerically integrating rather than using analytical expressions. Plot the response, reproduce Figure 3.14, and compare the results to see that they are the same. Chap. 3 299 Problems *3.68. Numerically simulate the response of the system of a single-degree-of-freedom spring– mass system subject to the motion y(t) given in Figure P3.68 and plot the response. The mass is 5000 kg and the stiffness is 1.5 * 103 N>m. y(t) (mm) 0.5 0.2 t (s) 0.6 Figure P3.68 The base motion for Problem 3.68. *3.69. Numerically simulate the response of an undamped system to a step function with a finite rise time of t1 for the case m = 1 kg, k = 1 N>m, t1 = 4 s, and F0 = 20 N. This function is described by F0t F(t) = c t 1 F0 0 … t … t1 t 7 t1 plot the response. *3.70. Numerically simulate the response of the system of Problem 3.22 for a 2-meter concrete wall with cross section 0.03 m2 and mass modeled as lumped at the end of 1000 kg. Use F0 = 100 N, and plot the response for the case t0 = 0.25 s. *3.71. Numerically simulate the response of an undamped system to a ramp input of the form F(t) = F0 t, where F0 is a constant. Plot the response for three periods for the case m = 1 kg, k = 100 N>m, and F0 = 50 N. *3.72. Compute and plot the response of the following system using numerical integration: $ # 10x(t) + 20x(t) + 1500x(t) = 20 sin 25t + 10 sin 15t + 20 sin 2t with initial conditions of x0 = 0.01 m and 0 = 1.0 m>s. Section 3.10 (Problems 3.73 through 3.79) *3.73. Compute the response of the system in Figure 3.26 for the case that the damping is linear viscous, the spring is a nonlinear soft spring of the form k(x) = kx - k1x3 the system is subject to an excitation of the form (t1 = 1.5 and t2 = 1.6) F(t) = 1500[Φ(t - t 1) - Φ(t - t 2)] N and initial conditions of x0 = 0.01 m and 0 = 1.0 m>s. The system has a mass of 100 kg, a damping coefficient of 30 kg>s, and a linear stiffness coefficient of 2000 N>m. 300 general Forced Response Chap. 3 The value of k1 is taken to be 300 N>m3. Compute the solution and compare it to the linear solution (k1 = 0). Which system has the largest magnitude? Compare your solution to that of Example 3.10.1. *3.74. Compute the response of a spring–mass system for the case that the damping is linear viscous, the spring is a nonlinear soft spring of the form k(x) = kx - k1x3 the system is subject to an excitation of the form (t1 = 1.5 and t2 = 1.6) F(t) = 1500[Φ(t - t 1) - Φ(t - t 2)] N and initial conditions of x0 = 0.01 m and 0 = 1.0 m>s. The system has a mass of 100 kg, a damping coefficient of 30 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 300 N>m3. Compute the solution and compare it to the linear solution (k1 = 0). How different are the linear and nonlinear responses? Repeat this for t2 = 2. What can you say regarding the effect of the time length of the pulse? *3.75. Compute the response of a spring–mass–damper system for the case that the damping is linear viscous, the spring stiffness is of the form k(x) = kx - k1x2 the system is subject to an excitation of the form (t1 = 1.5 and t2 = 2.5) F(t) = 1500[Φ(t - t 1) - Φ(t - t 2)] N and initial conditions of x0 = 0.01 m and 0 = 1 m>s. The system has a mass of 100 kg, a damping coefficient of 30 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 450 N>m3. Which system has the largest magnitude? *3.76. Compute the response of a spring–mass–damper system for the case that the damping is linear viscous, the spring stiffness is of the form k(x) = kx + k1x2 the system is subject to an excitation of the form (t1 = 1.5 and t2 = 2.5) F(t) = 1500[Φ(t - t 1) - Φ(t - t 2)] N and initial conditions of x0 = 0.01 m and 0 = 1 m>s. The system has a mass of 100 kg, a damping coefficient of 30 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 450 N>m3. Which system has the largest magnitude? *3.77. Compute the response of a spring–mass–damper system for the case that the damping is linear viscous, the spring stiffness is of the form k(x) = kx - k1x2 the system is subject to an excitation of the form (t1 = 1.5 and t2 = 2.5) F(t) = 150[Φ(t - t 1) - Φ(t - t 2)] N and initial conditions of x0 = 0.01 m and 0 = 1 m>s. The system has a mass of 100 kg, a damping coefficient of 30 kg>s, and a linear stiffness coefficient of 2000 N>m. The value of k1 is taken to be 5500 N>m3. Which system has the largest transient magnitude? Which has the largest magnitude in steady state? Chap. 3 301 Problems *3.78. Compare the forced response of a system with velocity-squared damping, as defined in equation (2.129) using numerical simulation of the nonlinear equation, to that of the response of the linear system obtained using equivalent viscous damping, as defined by equation (2.131). Use the initial conditions x0 = 0.01 m and 0 = 0.1 m>s with a mass of 10 kg, stiffness of 25 N>m, applied force of the form (t1 = 1.5 and t2 = 2.5) F(t) = 15[Φ(t - t 1) - Φ(t - t 2)] N and drag coefficient of α = 25. *3.79. Compare the forced response of a system with structural damping (see Table 2.2) using numerical simulation of the nonlinear equation to that of the response of the linear system obtained using equivalent viscous damping as defined in Table 2.2. Use the initial conditions x0 = 0.01 m and 0 = 0.1 m>s with a mass of 10 kg, stiffness of 25 N>m, applied force of the form (t1 = 1.5 and t2 = 2.5) F(t) = 15[Φ(t - t 1) - Φ(t - t 2)] N and solid damping coefficient of b = 8. Does the equivalent viscous-damping linearization overestimate the response or underestimate it? MATLAB ENGINEERING VIBRATION TOOLBOX You may use the files contained in the Engineering Vibration Toolbox, first discussed at the end of Chapter 1 (immediately following the problems) and discussed in Appendix G, to help solve many of the preceding problems. The files contained in folder VTB3 may be used to help understand the nature of the general forced response of a single-degree-of-freedom system as discussed in this chapter and the dependence of this response on various parameters. VTB1_3 and VTB1_4 must be used if an arbitrary forcing function is applied (one other than a simple function call). The following problems are suggested to help build some intuition regarding the material on general forced response and to become familiar with the various formulas. TOOLBOX PROBLEMS TB3.1. Use file VTB3_1 to solve for the response of a system with a 10-kg mass, damping c = 2.1 kg>s, and stiffness k = 2100 N>m, subject to an impulse at time t = 0 of magnitude 10 N. Next, vary the value of c, first increasing it, then decreasing it, and note the effect in the responses. TB3.2. Use file VTB3_2 to reproduce the plot of Figure 3.7. Then see what happens to the response as the damping coefficient is varied by trying an overdamped and critically damped value of ζ and examining the resulting response. 302 general Forced Response Chap. 3 TB3.3. If you are confident with MATLAB, try using the plot command to plot (say, for T = 6) - 8 2π t, cos 2 T π then - - 8 2π 1 6π acos t + cos tb 2 T 9 T π 8 2π 1 6π 1 10π acos t + cos t + cos tb T 9 T 25 T π2 and so on, until you are satisfied that the Fourier series computed in Example 3.3.1 converges to the function plotted in Figure 3.13. If you are not familiar enough with MATLAB to try this on your own, run VTB3_3, which is a demo that does this for you. TB3.4. Using VTB3_3, rework Problem 3.32 for first 5, then 10, and finally 50 terms. TB3.5. Using file VTB3_4, examine the effect of varying the system’s natural frequency on the response spectrum for the force given in Figure 3.16. Pick the frequencies f = 10 Hz, 100 Hz, and 1000 Hz and compare the various response spectrum plots. 4 multiple-Degreeof-Freedom Systems This chapter introduces the analysis needed to understand the vibration of systems with more than one degree of freedom. The number of degrees of freedom of a system is determined by the number of moving parts and the number of directions in which each part can move. more than one degree of freedom means more than one natural frequency, greatly increasing the opportunity for resonance to occur. This chapter also introduces the important concept of a mode shape, and the highly used method of modal analysis for studying the response of multiple-degree-of-freedom (mDOF) systems. most structures are modeled as mDOF systems. The all-terrain vehicle suspension shown in the photo at the top forms an example of a system that can be modeled as two or more degrees of freedom. Designers need to be able to predict the vibration response in order to improve the ride and ensure durability. The blades of a jet engine pictured at the bottom also require mDOF analysis but with a much larger number of degrees of freedom. airplanes, satellites, automobiles, and so on, all provide examples of vibrating systems well modeled by the mDOF analysis introduced in this chapter. 303 304 Multiple-Degree-of-Freedom Systems Chap. 4 In the preceding chapters, a single coordinate and single second-order differential equation sufficed to describe the vibratory motion of the mechanical device under consideration. However, many mechanical devices and structures cannot be modeled successfully by single-degree-of-freedom models. For example, the base-excitation problem of Section 2.4 requires a coordinate for the base as well as the main mass if the base motion is not prescribed, as assumed in Section 2.4. If the base motion is not prescribed and if the base has significant mass, then the coordinate, y, will also satisfy a second-order differential equation, and the system becomes a two-degree-of-freedom model. Machines with many moving parts have many degrees of freedom. In this chapter, a two-degree-of-freedom example is first used to introduce the special phenomena associated with multiple-degree-of-freedom systems. These phenomena are then extended to systems with an arbitrary but finite number of degrees of freedom. To keep a record of each coordinate of the system, vectors are introduced and used along with matrices. This is done both for the ease of notation and to enable vibration theory to take advantage of the disciplines of matrix theory, linear algebra, and computational codes. 4.1 TWO-DEGREE-OF-FREEDOM MODEL (UNDAMPED) This section introduces two-degree-of-freedom systems and how to solve for the response of each degree of freedom. The approach presented here is detailed because the goal is to provide background for solving systems with any number of degrees of freedom. In practice, computer methods are most commonly used to solve for the response of complex systems. This was not the case when most vibration texts were written. Hence, the approach here is a bit different than the approach found in the more traditional and older texts on vibration; the focus here is in setting up vibration problems in terms of matrices and vectors used in computer codes for solving practical problems. In moving from single-degree-of-freedom systems to two or more degrees of freedom, two important physical phenomena result. The first important difference is that a two-degree-of-freedom system will have two natural frequencies. The second important phenomenon is that of a mode shape, which is not present in single-degreeof-freedom systems. A mode shape is a vector that describes the relative motion between the two masses or between two degrees of freedom. These important concepts of multiple natural frequencies and mode shapes are intimately tied to the mathematical concepts of eigenvalues and eigenvectors of computational matrix theory. This establishes the need to cast the vibration problem in terms of vectors and matrices. Consider the two-mass system of Figure 4.1(a). This undamped system is similar to the system of Figure 2.13 except that the base motion is not prescribed in this case and the base now has mass. Figure 4.1(b) illustrates a single-mass system capable of moving in two directions and hence provides an example of a Sec. 4.1 305 two-Degree-of-Freedom Model (undamped) x1 k1 m1 k2 x2 x2 m2 k,k k1 m k2  J m x x1 (a) (b) (c) Figure 4.1 (a) A simple two-degree-of-freedom model consisting of two masses connected in series by two springs. (b) A single mass with two degrees of freedom (i.e., the mass moves along both the x1 and x2 directions). (c) A single mass with one translational degree of freedom and one rotational degree of freedom. two-degree-of-freedom system as well. Figure 4.1(c) illustrates a single rigid mass that is capable of moving in translation as well as rotation about its axis. In each of these three cases, more than one coordinate is required to describe the vibration of the system. Each of the three parts of Figure 4.1 constitutes a two-degree-of-freedom system. A physical example of each system might be (a) a two-story building, (b) the vibration of a drill press, or (c) the rocking motion of an automobile or aircraft. A free-body diagram illustrating the spring forces acting on each mass in Figure 4.1(a) is illustrated in Figure 4.2. The force of gravity is excluded following the reasoning used in Figure 1.14 (i.e., the static deflection balances the gravitational force and no friction is present). Summing forces on each mass in the horizontal direction yields $ m1x1 = -k1x1 + k2(x2 - x1) (4.1) $ m2 x2 = -k2(x2 - x1) Rearranging these two equations yields $ m1x1 + (k1 + k2)x1 - k2x2 = 0 $ m2x2 - k2x1 + k2x2 = 0 (4.2) Equations (4.2) consist of two coupled second-order ordinary differential equations, with constant coefficients, each of which requires two initial conditions to solve. Hence these two coupled equations are subject to the four initial conditions: # # # # x1(0) = x10 x1(0) = x10 x2(0) = x20 x2(0) = x20 (4.3) k1x1 m1 k2(x2  x1) k2(x2  x1) x1 m2 x2 Figure 4.2 Free-body diagrams of each mass in the system of Figure 4.1(a), indicating the restoring force provided by the springs. 306 Multiple-Degree-of-Freedom Systems Chap. 4 # # where the constants x10, x20 and x10, x20 represent the initial velocities and displacements of each of the two masses. These initial conditions are assumed to be known or given and provide the four constants of integration needed to solve the two second-order differential equations for the free response of each mass. There are several approaches available to solve equations (4.2) given (4.3) and the values of m1, m2, k1, and k2 for the responses x1(t) and x2(t). First, note that neither equation can be solved by itself because each equation contains both x1 and x2 (i.e., the equations are coupled). Physically, this states that the motion of x1 affects the motion of x2, and vice versa. A convenient method of solving this system is to use vectors and matrices. The vector approach to solving this simple two-degreeof-freedom problem is also readily extendable to systems with an arbitrary finite number of degrees of freedom and is compatible with computer codes. Vectors and matrices were introduced in Section 1.9 in order to enter a single-degree-of-freedom vibration problem into a numerical equation solver such as MATLAB. Vectors and matrices are reviewed here briefly and more details can be found in Appendix C. Here vectors and matrices are used to compute a solution of equation (4.1). Define the vector x(t) to be the column vector consisting of the two responses of interest: x(t) = J x1(t) R x2(t) (4.4) This is called a displacement or response vector and is a 2 * 1 array of functions. Differentiation of a vector is defined here by differentiating each element so that # x (t) # x(t) = J # 1 R x2(t) $ x (t) $ and x(t) = J $1 R x2(t) (4.5) are the velocity and the acceleration vectors, respectively. A square matrix is a square array of numbers, which could be made, for instance, by combining two 2 * 1 column vectors to produce a 2 * 2 matrix. An example of a 2 * 2 matrix is given by M = J m1 0 0 R m2 (4.6) Note here that italic capital letters are used to denote matrices and bold lowercase letters are used to denote vectors. Vectors and matrices can be multiplied together in a variety of ways. In vibration analysis the method most useful to define the product of a matrix times a vector is to define the result to be a vector with elements consisting of the dot product of the vector with each “row” of the matrix (i.e., by treating the row as a vector). The dot product of a vector is defined by xTx = [x1 x2]J x1 R = x21 + x22 x2 (4.7) Sec. 4.1 307 two-Degree-of-Freedom Model (undamped) which is a scalar. The symbol xT denotes the transpose of the vector and changes a column vector into a row vector. Equation (4.7) is also called the inner product or scalar product of the vector x with itself. A scalar, a, times a vector, x, is simply defined as ax = [ax1 ax2]T (i.e., a vector of the same dimension with each element multiplied by the scalar). (Recall that a scalar is any real or complex number.) These rules for manipulating vectors should be familiar from introductory mechanics (i.e., statics and dynamics) texts. The following example illustrates the rules for multiplying a matrix times a vector. Example 4.1.1 $ Consider the product of the matrix M of equation (4.6) and the acceleration vector x of equation (4.5). This product becomes m $ Mx = J 1 0 $ $ $ $ 0 x m x + 0x2 m1x1 R J $1 R = J $1 1 $ R = J $ R m2 x2 0x1 + m2x2 m2x2 (4.8) where the first element of the product is defined to be the dot product of the row vector $ [m1 0] with the column vector x, and the second element is the dot product of the row $ vector [0 m2] with x. Note that the product of a matrix and a vector is a vector. n Example 4.1.2 Consider the 2 * 2 matrix K defined by K = J k1 + k2 - k2 - k2 R k2 (4.9) and calculate the product Kx. Solution Again, the product is formed by considering the first element to be the inner product of the row vector [k1 + k2 - k2] and the column vector x. The second element of the product vector Kx is formed from the inner product of the row vector [ - k2 k2] and the vector x. This yields Kx = J k1 + k2 - k2 - k2 x1 (k + k2)x1 - k2x2 RJ R = J 1 R k2 x2 - k2x1 + k2x2 (4.10) n Two vectors of the same size are said to be equal if and only if each element of one vector is equal to the corresponding element in the other vector. With this in mind, consider the vector equation $ Mx + Kx = 0 (4.11) 308 Multiple-Degree-of-Freedom Systems Chap. 4 where 0 denotes the column vector of zeros: 0 0 = J R 0 Substitution of the value for M from equation (4.6) and the value for K from equation (4.9) into equation (4.11) yields J m1 0 $ 0 x k + k2 R J $1 R + J 1 m2 x2 -k2 - k2 x1 0 RJ R = J R k2 x2 0 These products can be carried out as indicated in Example 4.1.1 and equation (4.10) to yield J $ m1x1 (k1 + k2)x1 - k2 x2 0 R = J R $ R + J m2x2 - k2 x1 + k2 x2 0 Adding the two vectors on the left side of the equation, element by element, yields J $ m1x1 + (k1 + k2)x1 - k2x2 0 R = J R $ m2x2 - k2x1 + k2x2 0 (4.12) Equating the corresponding elements of the two vectors in equation (4.12) yields $ m1x1 + (k1 + k2)x1 - k2x2 = 0 (4.13) $ m2x2 - k2x1 + k2x2 = 0 which are identical to equations (4.2). Hence the system of equations (4.2) can be written as the vector equation given in (4.11), where the coefficient matrices are defined by the matrices of equations (4.6) and (4.9). The matrix M defined by equation (4.6) is called the mass matrix, and the matrix K defined by equation (4.9) is called the stiffness matrix. The preceding calculation and comparison provide an extremely important connection between vibration analysis and matrix analysis. This simple connection allows computers to be used to solve large and complicated vibration problems quickly (discussed in Section 4.10). It also forms the foundation for the rest of this chapter (as well as the rest of the book). The mass and stiffness matrices, M and K, described previously have the special property of being symmetric. A symmetric matrix is a matrix that is equal to its transpose. The transpose of a matrix, denoted by AT, is formed from interchanging the rows and columns of a matrix. The first row of AT is the first column of A and so on. The mass matrix M is also called the inertia matrix, and the $ force vector Mx corresponds to the inertial forces in the system of Figure 4.1(a). Similarly, the force Kx represents the elastic restoring forces of the system described in Figure 4.1(a). Sec. 4.1 309 two-Degree-of-Freedom Model (undamped) Example 4.1.3 Consider the matrix A defined by A = J a c b R d where a, b, c, and d are real numbers. Calculate values of these constants such that the matrix A is symmetric. Solution For A to be symmetric, A = AT or A = J a c b a R = J d b c R = AT d Comparing the elements of A and AT yields that c = b must hold if the matrix A is to be symmetric. Note that the elements in the c and b position of the matrix K given in equation (4.9) are equal so that K = KT. n It is useful to note that if x is a column vector x = J x1 R x2 then xT is a row vector (i.e., xT = [x1 x2]). This makes it convenient to write a column vector in one line. For example, the vector x can also be written as x = [x1 x2]T , a column vector. The act of forming a transpose also undoes itself, so that (AT)T = A. The initial conditions can also be written in terms of vectors as # x (0) # x0 = J # 1 R (4.14) x2(0) # Here x0 denotes the initial displacement vector and x0 denotes the initial velocity vector. Equation (4.12) can now be solved by following the procedures used for solving single-degree-of-freedom systems and incorporating a few results from matrix theory. Recall from Section 1.2 that the single-degree-of-freedom version of equation (4.11) was solved by assuming a harmonic solution and calculating values for the constants in the assumed form. The same approach is used here. Following the argument used in equations (1.13) to (1.19), a solution is assumed of the form x0 = J x1(0) R x2(0) x(t) = ue jωt (4.15) Here u is a nonzero vector of constants to be determined, ω is a constant to be determined, and j = 1-1. Recall that the scalar e jωt represents harmonic motion since e jωt = cos ωt + j sin ωt. The vector u cannot be zero; otherwise no motion results. 310 Multiple-Degree-of-Freedom Systems Chap. 4 Substitution of this assumed form of the solution into the vector equation of motion yields ( -ω2M + K)ue jωt = 0 (4.16) where the common factor uejωt has been factored to the right side. Note that the scalar ejωt ∙ 0 for any value of t and hence equation (4.16) yields the fact that ω and u must satisfy the vector equation ( -Mω2 + K)u = 0, u ∙ 0 (4.17) Note that this represents two algebraic equations in the three unknown scalars: ω, u1, and u2 where u = [u 1 u 2]T. For this homogeneous set of algebraic equations to have a nonzero solution for the vector u, the inverse of the coefficient matrix (–Mω2 + K) must not exist. To see that this is the case, suppose that the inverse of (–Mω2 + K) does exist. Then multiplying both sides of equation (4.17) by (–Mω2 + K) -1 yields u = 0, a trivial solution, as it implies no motion. Hence the solution of equation (4.11) depends in some way on the matrix inverse. Matrix inverses are reviewed in the following example. Example 4.1.4 Consider the 2 * 2 matrix A defined by A = J a c b R d and calculate its inverse. Solution The inverse of a square matrix A is a matrix of the same dimension, denoted by A-1, such that AA-1 = A-1A = I where I is the identity matrix. In this case I has the form I = J 1 0 0 R 1 The inverse matrix for a general 2 * 2 matrix is A-1 = d 1 J det A - c -b R a (4.18) provided that det A ∙ 0, where det A denotes the determinant of the matrix A. The determinant of the matrix A has the value det A = ad - bc Sec. 4.1 311 two-Degree-of-Freedom Model (undamped) To see that equation (4.18) is in fact the inverse, note that A-1A = 1 ad - bc J d -c = 1 ad - bc J ad - bc ac - ac -b a RJ a c b R d bd - bd 1 R = J ad - bc 0 0 R 1 It is important to realize that the matrix A has an inverse if and only if det A ∙ 0. Thus, requiring det A = 0 forces A not to have an inverse. Matrices that do not have an inverse are called singular matrices. Note that if the matrix A is symmetric, c = b and A–1 is also symmetric. n Applying the condition of singularity to the coefficient matrix of equation (4.17) yields the result that for a nonzero solution u to exist, det( -ω2M + K) = 0 (4.19) which yields one algebraic equation in one unknown (ω2). Substituting the values of the matrices M and K into this expression yields det J -ω2m1 + k1 + k2 -k2 -k2 R = 0 -ω m2 + k2 2 (4.20) Using the definition of the determinant yields that the unknown quantity ω2 must satisfy m1m2ω4 - (m1k2 + m2k1 + m2k2)ω2 + k1k2 = 0 (4.21) This expression is called the characteristic equation for the system and is used to determine the constants ω in the assumed form of the solution given by equation (4.15) once the values of the physical parameters m1, m2, k1, and k2 are known. Example 4.1.5 Calculate the solutions for ω of the characteristic equation given by equation (4.21) for the case that the physical parameters have the values m1 = 9 kg, m2 = 1 kg, k1 = 24 N>m, and k2 = 3 N>m. Solution For these values the characteristic equation (4.21) becomes ω4 - 6ω2 + 8 = (ω2 - 2)(ω2 - 4) = 0 so that ω21 = 2 and ω22 = 4. There are two roots and each corresponds to two values of the constant ω in the assumed form of the solution: ω1 = { 12 rad>s, ω2 = {2 rad>s n 312 Multiple-Degree-of-Freedom Systems Chap. 4 Note that in the statement of equation (4.17) ω2 appears, not ω. However, in proceeding to the solution in time, the frequency of oscillation will become ω and the plus and minus signs on ω are absorbed in changing the exponential into a trigonometric function, as described in the following pages. Once the value of ω in equation (4.15) is established, the value of the constant vector u can be found by solving equation (4.17) for u given each value of ω2. That is, for each value of ω2 (i.e., ω21 and ω22) there is a vector u satisfying equation (4.17). For ω21, the vector u1 satisfies ( -ω21 M + K)u 1 = 0 (4.22) and for ω22, the vector u2 satisfies ( -ω22 M + K)u 2 = 0 (4.23) These two expressions can be solved for the direction of the vectors u1 and u2, but not for the magnitude. To see that this is true, note that if u1 satisfies equation (4.22), so does the vector au1, where a is any nonzero number. Hence the vectors satisfying (4.22) and (4.23) are of arbitrary magnitude. The following example illustrates one way to compute u1 and u2 for the values of Example 4.1.5. Example 4.1.6 Calculate the vectors u1 and u2 of equations (4.22) and (4.23) for the values of ω, K, and M of Example 4.1.5. Solution Let u1 = [u 11 J u 21]T . Then equation (4.22) with ω2 = ω21 = 2 becomes 27 - 9(2) -3 -3 R 3 - (2) J u 11 0 R = J R 0 u 21 Performing the indicated product and enforcing the equality yields the two equations 9u 11 - 3u 21 = 0 and -3u 11 + u 21 = 0 Note that these two equations are dependent and yield the same solution; that is, u 11 1 = u 21 3 or u 11 = 1 u 3 21 Only the ratio of the elements is determined here [i.e., only the direction of the vector is determined by equation (4.17), not its magnitude]. As mentioned previously, this happens because if u satisfies equation (4.17), then so does au, where a is any nonzero number. Sec. 4.1 313 two-Degree-of-Freedom Model (undamped) A numerical value for each element of the vector u may be obtained by arbitrarily assigning one of the elements. For example, let u21 = 1; then the value of u1 is 1 u1 = C 3 S 1 This procedure is repeated using ω22 = 4 to yield that the elements of u2 must satisfy -9u 12 - 3u 22 = 0 or 1 u 12 = - u 22 3 Choosing u22 = 1 yields 1 u2 = C 3 S 1 which is the vector satisfying equation (4.23). There are several other ways of fixing the magnitude of a vector besides the one illustrated here. Some other methods are presented in Example 4.2.3 and equation (4.44). A more systematic method called normalizing will be used with larger problems and is presented in the next section. n # The solution of equation (4.11) subject to initial conditions x0 and x0 can be constructed in terms of the numbers {ω1, {ω2 and the vectors u 1 and u 2. This is similar to the construction of the solution of the single-degree-of-freedom case discussed in Section 1.2. Since the equations to be solved are linear, the sum of any two solutions is also a solution. From the preceding calculation, there are four solutions in the form of equation (4.15) made up of the four values of ω and the two vectors: x(t) = u 1e -jω1t, u 1e +jω1t, u 2e -jω2t, and u 2e +jω2t (4.24) Thus a general solution is the linear combination of these: x(t) = (ae jω1t + be -jω1t)u 1 + (ce jω2t + de -jω2t)u 2 (4.25) where a, b, c, and d are arbitrary constants of integration to be determined by the initial conditions. Applying Euler formulas for the sine function to equation (4.25) yields an alternative form of the solution (provided neither ωi is zero): x(t) = A1 sin (ω1t + ϕ1)u 1 + A2 sin (ω2t + ϕ2)u 2 (4.26) 314 Multiple-Degree-of-Freedom Systems Chap. 4 where the constants of integration are now in the form of two amplitudes, A1 and A2, and two phase shifts, ϕ1 and ϕ2. Recall that this is the same procedure used in equations (1.17), (1.18), and (1.19). These constants can be calculated from the # initial conditions x0 and x0. Equation (4.26) is the two-degree-of-freedom analog of equation (1.19) for a single-degree-of-freedom case. The form of equation (4.26) gives physical meaning to the solution. It states that each mass in general oscillates at two frequencies: ω1 and ω2. These are called the natural frequencies of the system. Furthermore, suppose that the initial conditions are chosen such that A2 = 0. With such initial conditions, each mass oscillates at only one frequency, ω1, and the relative positions of the masses at any given instant of time are determined by the elements of the vector u1. Hence u1 is called the first mode shape of the system. Similarly, if the initial conditions are chosen such that A1 is zero, both coordinates oscillate at frequency ω2, with relative positions given by the vector u2, called the second mode shape. The mode shapes and natural frequencies are clarified further in the following exercises and sections. Mode shapes have become a standard in vibration engineering and are used extensively in vibration analysis. The concepts of natural frequencies and mode shapes are extremely important and form one of the major ideas used in vibration studies. In the derivation of equation (4.26) it is assumed that neither of the values of ω is zero. One or the other may have the value zero in some applications, but then the solution takes on another form. The value of u, however, cannot be zero. A frequency can be zero, but a mode shape cannot be zero. The zero frequency case corresponds to rigid body motion and is the topic of Problem 4.12. The concept of rigid body motion is detailed in Section 4.4. Note that the positive and negative sign on ω resulting from the solution of equation (4.19) is used in going from equation (4.25) to equation (4.26) when invoking the Euler formula for trig functions. Thus in equation (4.26) ω is only a positive number. Equation (4.26) also provides the interpretation of ω as a frequency of vibration, which is now necessarily a positive number. This is similar to the explanation provided in the single-degree-of-freedom case following equation (1.19). Example 4.1.7 Calculate the solution of the system of Example 4.1.5 for the initial conditions x1(0) = # # 1 mm, x2(0) = 0, and x1(0) = x2(0) = 0. Solution To solve this, equation (4.26) is written as J x1(t) A sin (ω1t + ϕ1) R = [u1 u2] J 1 d x2(t) A2 sin (ω2t + ϕ2) 1 1 A1 sin ( 12t + ϕ1) - A2 sin (2t + ϕ2) 3 = C3 S A1 sin ( 12t + ϕ1) + A2 sin (2t + ϕ2) (4.27) Sec. 4.1 two-Degree-of-Freedom Model (undamped) 315 At t = 0 this yields 1 1 A1 sin ϕ1 - A2 sin ϕ2 1 3 J R = C3 S 0 A1 sin ϕ1 + A2 sin ϕ2 (4.28) Differentiating equation (4.27) and evaluating the resulting expression at t = 0 yields 12 2 # A1 cos ϕ1 - A2 cos ϕ2 x1(0) 0 3 S J# R = J R = C 3 x2(0) 0 12 A1 cos ϕ1 + 2 A2 cos ϕ2 (4.29) Equations (4.28) and (4.29) represent four equations in the four unknown constants of integrations A1, A2, ϕ1, and ϕ2. Writing out these four equations yields 3 = A1 sin ϕ1 - A2 sin ϕ2 (4.30) 0 = A1 sin ϕ1 + A2 sin ϕ2 (4.31) 0 = 12A1 cos ϕ1 - 2A2 cos ϕ2 (4.32) 0 = 12A1 cos ϕ1 + 2A2 cos ϕ2 (4.33) Adding equations (4.32) and (4.33) yields that 212A1 cos ϕ1 = 0 so that ϕ1 = π>2. Since ϕ1 = π>2, equation (4.33) reduces to 2A2 cos ϕ2 = 0 so that ϕ2 = π>2. Substitution of the values of ϕ1 and ϕ2 into equations (4.30) and (4.31) yields 3 = A1 - A2 and 0 = A1 + A2 which has solutions A1 = 3>2 mm, A2 = - 3>2 mm. Thus x1(t) = 0.5 sin a12t + π π b + 0.5 sin a2t + b = 0.5 ( cos 12t + cos 2t ) mm 2 2 3 π 3 π x2(t) = sin a 12t + b - sin a2t + b = 1.5 ( cos 12t - cos 2t ) mm 2 2 2 2 (4.34) These are plotted in Figure 4.3. More efficient ways to calculate the solutions are presented in later sections. The numerical aspects of calculating a solution are discussed in Section 4.10. 316 Multiple-Degree-of-Freedom Systems Chap. 4 Displacement mm 1 0.5 0 0.5 1 0 1 2.5 5 7.5 10 Time (s) 12.5 15 17.5 20 5 7.5 10 12.5 15 17.5 20 Displacement (mm) 0.5 0 2.5 0.5 1 Time (s) Figure 4.3 Plots of the responses of x1(t) on the top and x2(t) on the bottom for Example 4.1.7. Sec. 4.1 two-Degree-of-Freedom Model (undamped) 317 Note that in this case, the response of each mass contains both frequencies of the system. That is, the responses for both x1(t) and x2(t) are combinations of signals containing the two frequencies ω1 and ω2 (i.e., the sum of two harmonic signals). Note from the development of equation (4.34) that the mode shapes determine the relative magnitude of these two harmonic signals. n In the previous example, the arbitrary choice of the magnitude of the modeshape vectors u1 and u2 made in Example 4.1.6 does not affect the solution because these vectors are multiplied by the constants of integration A1 and A2, respectively. The initial conditions then scale the magnitude of these vectors, so that the solution given in equation (4.34) will be the same for any choice of fixing the vector magnitude made in Example 4.1.6. Frequencies It is interesting and important to note that the two natural frequencies ω1 and ω2 of the two-degree-of-freedom system are not equal to either of the natural frequencies of the two single-degree-of-freedom systems constructed from the same components. To see this, note that in Example 4.1.5, 1k1 >m1 = 1.63, which is not equal to ω1 or ω2 (i.e., ω1 = 12, ω2 = 2). Similarly, 1k2 >m2 = 1.732, which does not coincide with either frequency of the two-degree-of-freedom system composed of the same springs and masses, each attached to ground. Beats The beat phenomenon introduced in Example 2.1.2 for the forced response of a single-degree-of-freedom system can also exist in the free response of a two- (or more) degree-of-freedom system. If the mass and stiffness of the system of Figure 4.1(a) are such that the two frequencies are close to each other, then solutions derived in Example 4.1.7 will produce beats. In fact, a close examination of the plots of the response in Figure 4.3 shows that the response x2(t) is close to the shape of the beat illustrated in Figure 2.4. This happens because the two frequencies of Example 4.1.7 are reasonably close to each other (1.414 and 2). As the two natural frequencies become closer, the beat phenomenon will become more evident (see Problem 4.17). Thus, beats in vibrating systems can occur in two separate circumstances: first, in a forced-response case as the result of a driving frequency being close to a natural frequency (Example 2.1.2) and, second, as the result of two natural frequencies being close together in a free-response situation (Problem 4.18). Calculations The method used to compute the natural frequencies and mode shapes presented in this section is not the most efficient way to solve vibration problems. Nor is the approach presented here the most illuminating of the physical nature of vibration of two-degree-of-freedom systems. The calculation method presented in Examples 4.1.5 and 4.1.6 are instructive, but tedious. The approach of these examples also ignores the key issues of orthogonality of mode shapes and decoupling of the equations of motion, which are key concepts in understanding vibration analysis, design, and measurement. These issues are discussed in the following section, which connects the problem of computing natural frequencies and 318 Multiple-Degree-of-Freedom Systems Chap. 4 mode shapes to the symmetric eigenvalue and eigenvector calculations of mathematics. Once the natural frequency and mode-shape formulation is connected to the algebraic eigenvalue–eigenvector problem, then mathematical software packages can be used to compute the mode shapes and natural frequencies without going through the tedious computations of the preceding examples. This connection to the algebraic eigenvalue problem is also a key in understanding the topics of vibration testing discussed in Chapter 7. 4.2 EIGENVALUES AND NATURAL FREQUENCIES The method of solution indicated in Section 4.1 can be extended and formalized to take advantage of the symmetric algebraic eigenvalue problem. This allows the power of mathematics to be used in solving vibration problems, allows the use of mathematical software packages, and sets the background needed for analyzing systems with an arbitrary number of degrees of freedom. In addition, the important concepts of mode shapes and natural frequencies can be generalized by connecting the undampedvibration problem to the mathematics of the algebraic eigenvalue problem. There are many ways to connect the solution of the vibration problem with that of the algebraic eigenvalue problem. The most productive approach is to cast the vibration problem as a symmetric eigenvalue problem because of the special properties associated with symmetry. Note that the physical nature of both the mass and stiffness matrices is that they are usually symmetric. Hence preserving this symmetry is also a natural approach to solving the vibration problem. Since M is symmetric and positive definite, it may be factored into two terms: M = LLT where L is a special matrix with zeros in every position above the diagonal (called a lower triangular matrix). A matrix M is positive definite if the scalar formed from the product xTMx 7 0 for every nonzero choice of the vector x. The factorization L is called the Cholesky decomposition and is examined, along with the notion of positive definite, in more detail in Appendix C and Section 4.9. In the special case that M happens to be diagonal, as in the examples considered so far, the Cholesky decomposition just becomes the notion of a matrix square root, and the notion of positive definite just means that the diagonal elements of M are all positive, nonzero numbers. In solving a single-degree-of-freedom system it was useful to divide the equation of motion by the mass. Hence, consider resolving the system of two equations described in matrix form by equation (4.19) by making a coordinate transformation that is equivalent to dividing the equations of motion by the mass in the system. To that end, consider the matrix square root defined to be the matrix M1>2 such that M1>2M1>2 = M, the mass Sec. 4.2 319 Eigenvalues and natural Frequencies matrix. For the simple example of the mass matrix given in equation (4.6), the mass matrix is diagonal and the matrix square root becomes simply L = M1>2 = J 1m1 0 0 R 1m2 (4.35) This factors M into M = M1 >2M1 >2 (or into M = LLT) in the common case of a diagonal mass matrix. If M is not diagonal, the notion of a square root is dropped in favor of using the Cholesky decomposition L, which is discussed in Section 4.9 under dynamically coupled systems. The use of the Cholesky decomposition L is preferred because it is a single command in most codes, and hence more convenient for numerical computation. The inverse (Example 4.1.4) of the diagonal matrix M1 >2, denoted by M - 1>2, becomes simply -1 L = M -1>2 = D 1 1m1 0 0 1 1m2 T (4.36) The matrix of equation (4.36) provides a means of changing coordinate systems to one in which the vibration problem is represented by a single symmetric matrix. This allows the vibration problem to be cast as the symmetric eigenvalue problem described in Window 4.1. The symmetric eigenvalue problem has distinct advantages both computationally and analytically, as will be illustrated next. The analytical advantage of this change of coordinates is similar to the advantage gained in solving the inclined plane problem in statics by writing the coordinate system along the incline rather than along the horizontal. To accomplish this transformation, or change of coordinates, let the vector x in equation (4.11) be replaced with x(t) = M1>2q(t) (4.37) and multiply the resulting equation by M - 1>2. This yields $ M -1>2MM -1>2q(t) + M -1>2KM -1>2q(t) = 0 (4.38) Since M - 1>2MM - 1>2 = I, the identity matrix, expression (4.38) reduces to $ ∙ Iq(t) + Kq(t) = 0 (4.39) ∙ The matrix K = M -1>2KM -1>2, like the matrix K, is a symmetric matrix. The matrix ∙ K is called the mass-normalized stiffness and is analogous to the single-degree-offreedom constant, k>m. 320 Multiple-Degree-of-Freedom Systems Chap. 4 Window 4.1 Properties of the Symmetric Eigenvalue Problem The algebraic eigenvalue problem is the problem of computing the scalar λ and the nonzero vector v satisfying Av = λv Here, A is an n * n real valued, symmetric matrix, the vector v is n * 1, and there will be n values of the scalar λ, called the eigenvalues, and n values of the corresponding vector v, one for each value of λ. The vectors v are called the eigenvectors of the matrix A. The eigenvalues of A are all real numbers. The eigenvectors of A are real valued. The eigenvalues of A are positive numbers if and only if A is positive definite. The eigenvectors of A can be chosen to be orthogonal, even for repeated eigenvalues. Symmetry also implies that the set of eigenvectors are linearly independent and can be used like a Fourier series to expand any vector into a sum of eigenvectors. This forms the basis of modal expansions used in both analysis and experiments (Chapter 7). In addition, the numerical algorithms used to compute the eigenvalues and eigenvectors of A are faster and more efficient for symmetric matrices. Example 4.2.1 ∙ Show that K is a symmetric matrix if K and M are symmetric. Note that if M is symmetric so is M–1 and M -1>2. This is trivial in the cases used here where M is diagonal, but this is also true for fully populated symmetric matrices. Also show the matrix is symmetric using the Cholesky factors of M. Solution To show that a matrix is symmetric, use the rule that for any two square matrices of the same size (AB)T = BTAT. Applying this rule twice yields ∙ ∙ KT = (M -1>2KM -1>2)T = (KM -1>2)TM -1>2 = M -1>2KTM -1>2 = M -1>2KM -1>2 = K ∙ ∙ Thus K = KT and is a symmetric matrix. n Equation (4.39) is solved, as before, by assuming a solution of the form q(t) = ve jωt where v ∙ 0 is a vector of constants. Substitution of this form into equation (4.39) yields ∙ Kv = ω2v (4.40) Sec. 4.2 321 Eigenvalues and natural Frequencies upon dividing by the nonzero scalar ejωt. Here it is important to note that the constant vector v cannot be zero if motion is to result. Next let λ = ω2 in equation (4.40). This yields ∙ Kv = λv (4.41) where v ∙ 0. This is precisely the statement of the algebraic eigenvalue problem. The scalar λ satisfying equation (4.41) for nonzero vectors v is called the eigenvalue ∙ and v is called the (corresponding) eigenvector. Since the matrix K is symmetric, this is called the symmetric eigenvalue problem. The eigenvector v generalizes the concept of a mode shape u used in Section 4.1. If the system being modeled has n degrees of freedom, each free to move with ∙ a single displacement labeled xi(t), the matrices M, K, and hence K will be n * n, and the vectors x(t), q(t), and v will be n * 1 in dimension. Each subscript i denotes a single degree of freedom where i ranges from 1 to n, and the vector x(t) denotes the collection of the n degrees of freedom. It is also convenient to label the frequencies ω and eigenvectors v with subscripts i, so that ωi and vi denote the ith natural frequency and corresponding ith eigenvector, respectively. In this section, only n = 2 is considered, but the notation is useful and valid for any number of degrees of freedom. Equation (4.41) connects the problem of calculating the free vibration response of a conservative system with the mathematics of symmetric eigenvalue problems. This allows the developments of mathematics to be applied directly to vibration. The theoretical advantage of this relationship is significant and is used here. These properties are summarized in Window 4.1 and reviewed in Appendix C. The computational advantage, which is substantial, is discussed in Section 4.9. Example 4.2.2 ∙ Calculate the matrix K for M = J 9 0 0 R, 1 K = J 27 -3 - 3 R 3 as given in Example 4.1.5. Solution Matrix products are defined here for matrices of the same size by extending the idea of a matrix times a vector outlined in Example 4.1.1. The result is a third matrix of the same size. The first column of the matrix product AB is the product of the matrix A with the first column of B considered as a vector, and so on. To illustrate this, consider the product KM - 1>2, where M - 1>2 is defined by equation (4.36). KM -1>2 27 = J -3 1 -3 RC3 3 0 1 (27)a b + ( - 3)(0) 3 S = D 1 1 ( - 3)a b + 3(0) 3 0 = J 9 -1 -3 R 3 (27)(0) + ( - 3)(1) T ( -3)(1) + (3)(1) 322 Multiple-Degree-of-Freedom Systems Chap. 4 Multiplying this by M - 1>2 yields M -1>2 KM -1>2 1 = C3 0 0 SJ 1 9 -1 -3 R 3 1 a b (9) + (0)( - 1) = C 3 (0)(9) + (1)( - 1) 1 a b( - 3) + (0)(3) 3 3 S = J -1 (0)( - 3) + (1)(3) -1 R 3 ∙ Note that (M - 1>2KM - 1>2)T = M - 1>2KM - 1>2, so that K is symmetric. It is tempting to relate equation (4.11) to the algebraic eigenvalue problem by simply multiplying equation (4.17) by M -1 to get λu = M -1 Ku. However, as the following computation indicates, this does not yield a symmetric eigenvalue problem. Computing the product yields 1 M K = C9 0 -1 0 1 27 SJ -3 -3 3 R = C 3 -3 1 3 3S ∙ C 1 3 3 - -3 3 S = KM -1 so that this matrix product is not symmetric. The use of M -1K also becomes computationally more expensive, as discussed in Section 4.9. n Alternately, the Cholesky factorization may be used as described in Window 4.2. The symmetric eigenvalue problem has several advantages. A summary of properties of the symmetric eigenvalue problem is given in Window 4. 1. For example, it can readily be shown that the solutions of equation (4.41) are real numbers. Furthermore, it can be shown that the eigenvectors satisfying equation (4.41) are orthogonal and never zero just like the unit vectors (ni, nj , kn , en 1, en 2, en 3 ) used in the vector analysis of forces (regardless of whether or not the eigenvalues are repeated). Two vectors v1 and v2 are defined to be orthogonal if their dot product is zero, that is, if vT1 v2 = 0 (4.42) (Ortho comes from a Greek word meaning straight.) The eigenvectors satisfying (4.41) are of arbitrary length just like the vectors u1 and u2 of Section 4.1. Following the analogy of unit vectors from statics (introductory mechanics), the eigenvectors can be normalized so that their length is 1. The norm of a vector is denoted by 7 x 7 and defined by 7 x 7 = 2xTx = c a ( x2i ) d n 1>2 (4.43) A set of vectors that satisfies both (4.42) and 7 x 7 = 1 are called orthonormal. The unit vectors from a Cartesian coordinate system form an orthonormal set of vectors (recall that ni # ni = 1, ni # nj = 0, etc.). A summary of vector inner products is given in Appendix C. i=1 Sec. 4.2 323 Eigenvalues and natural Frequencies Window 4.2 Symmetric Eigenvalue Problem by Cholesky Factorization The Cholesky factorization of the mass matrix is M = LLT for any symmetric M, even if it is not necessarily diagonal. Consider the substitution x(t) = (LT)–1z(t) into equation (4.11) and multiply by L–1. This yields $ L-1(LLT)(LT)-1z(t) + L-1K(LT)-1z(t) = 0 The action of taking the transpose and taking the inverse are interchangeable. This, combined with the rule that (AB)T = BTAT, yields that the first coefficient is the identity matrix: L-1(LLT)(LT)-1 = (L-1L)(LT)(LT)-1 = (I)(I) = I The coefficient of z is symmetric and used as an alternate definition of the massnormalized stiffness matrix. To see that it is symmetric, calculate its transpose: ∙ ∙ KT = [L-1K(LT)-1]T = [(LT)-1]T(L-1K)T = L-1KT(L-1)T = K since K = KT. Combining all three equations and substitution of z = eλt ∙ yields the symmetric eigenvalue problem Kv = λv. Using L rather the M1>2 to form the mass-normalized stiffness matrix has numerical advantages which come into play for larger problems. To normalize the vector u1 = [1>3 1]T , or any vector for that matter, an unknown scalar α is sought such that the scaled vector αu1 has unit norm, that is, such that (αu 1)T(αu 1) = 1 Writing this expression for u1 = [1>3 1]T yields α2(1>9 + 1) = 1 or α = 3> 110. Thus the new vector αu 1 = [1> 110 3> 110]T is the normalized version of the vector u1. Remember that the eigenvalue problem determines only the direction of the eigenvector, leaving its magnitude arbitrary. The process of normalization is just a systematic way to scale each eigenvector or mode shape. In general, any vector x can be normalized simply by calculating 1 2xTx x (4.44) Equation (4.44) can be used to normalize any nonzero real vector of any length. Note again that since x here is an eigenvector, it cannot be zero so that dividing by the scalar xTx is always possible. 324 Multiple-Degree-of-Freedom Systems Chap. 4 The normalizing of the eigenvectors to remove the arbitrary choice of one element in the vector is the systematic method mentioned at the end of Example 4.1.6. Normalizing is an alternative to choosing one element of the vector to have the value 1, as was done in Example 4.1.6. The following example illustrates the eigenvalue problem, the process of normalizing vectors, and the concept of orthogonal vectors. Example 4.2.3 Solve the eigenvalue problem for the two-degree-of-freedom system of Example 4.2.2 where 3 ∙ K = J -1 -1 R 3 Normalize the eigenvectors, check if they are orthogonal, and compare them to the mode shapes of Example 4.1.6. Solution The eigenvalue problem is to calculate the eigenvalues λ and eigenvectors v that satisfy equation (4.41). Rewriting equation (4.41) yields ∙ (K - λI)v = 0 or J 3 - λ -1 -1 Rv = 0 3 - λ (4.45) where v must be nonzero. Hence the matrix coefficient must be singular and therefore its determinant must be zero. det J 3 - λ -1 -1 R = λ2 - 6λ + 8 = 0 3 - λ This last expression is the characteristic equation and has the two roots λ1 = 2 and λ1 = 4 ∙ which are the eigenvalues of the matrix K. Note that these are also the squares of the 2 natural frequencies, ωi , as calculated in Example 4.1.5. The eigenvector associated with λ 1 is calculated from equation (4.41) with λ = λ 1 = 2 and v1 = [v11 v21]T: ∙ (K - λ1I ) v1 = 0 = J 3 - 2 -1 -1 v 0 R J 11 R = J R 3 - 2 v21 0 This results in the two dependent scalar equations v11 - v21 = 0 Hence 11 = 21, and - v11 + v21 = 0 which defines the direction of the vector v1. Sec. 4.2 325 Eigenvalues and natural Frequencies To fix a value for the elements of v1, the normalization condition of equation (4.44) is used to force v1 to have a magnitude of 1. This results in (setting 11 = 21) Solving for 21 yields 1 = 7 v1 7 = 2v221 + v221 = 12v21 v21 = 1 12 so that the normalized vector v1 becomes v1 = 1 1 J R 12 1 Similarly, substitution of λ2 = 4 into (4.41), solving for the elements of v2, and normalizing the result yields 1 1 J R 12 - 1 v2 = Now note that the product vT1 v2 yields vT1 v2 = 1 1 [1 12 12 1] J 1 1 R = (1 - 1) = 0 -1 2 so that the set of vectors v1 and v2 are orthogonal as well as normal. Hence the two vectors v1 and v2 form an orthonormal set as described in Window 4.3. Next, consider the mode-shape vectors computed in Example 4.1.6 directly from equations (4.22) and (4.23): u1 = [1>3 1]T and u2 = [ - 1>3 1]T . The following calculation shows that these vectors are not orthogonal: uT1 u2 1 = J 3 1 1 8 1R C 3 S = - + 1 = ∙ 0 9 9 1 - Next, normalize u1 and u2 using equation (4.44) to get un 1 = 1 2uT1 u1 u1 = 1 1 0.31623 C3S = J R 0.94868 1 + 1 1 A9 where the “hat” is used to denote a unit vector, a notation that is usually dropped in favor of renaming the normalized version u1 as well. Following a similar calculation, the normalized version of u2 becomes u2 = [ - 0.31623 0.948681]T. Note that the normalized versions of u1 and u2 are not orthogonal either. Only the eigenvectors computed ∙ from the symmetric matrix K are orthogonal. However, the vectors ui computed in 326 Multiple-Degree-of-Freedom Systems Chap. 4 Example 4.1.6 can be normalized and made orthogonal with respect to the mass matrix M, as discussed later. The mode shapes ui and eigenvectors vi are related by the square root of the mass matrix as discussed next. n The previous example showed how to calculate the eigenvalues and eigenvectors of the symmetric eigenvalue problem related to the vibration problem. In comparing this to the mode shape and natural frequency calculation made earlier, the eigenvalues are exactly the squares of the natural frequencies. However, there is some difference between the mode shapes of Example 4.1.6 and the eigenvectors of the previous example. This difference is captured by the fact that the mode shapes as calculated in Example 4.1.6 are not orthogonal, but the eigenvectors are orthogonal. The property of orthogonality is extremely important in developing modal analysis (Section 4.3) because it allows the equations of motion to uncouple, reducing the analysis to that of solving several single-degree-of-freedom systems defined by scalar equations. The eigenvectors and the mode shapes are related through equation (4.37) by u 1 = M -1>2v1 and v1 = M1>2u 1 To see this, note that M1>2u 1 = J 3 0 1 0 1 R C 3 S = J R = v1 1 1 1 using the values from the examples. Thus the mode shapes and eigenvectors are related by a simple matrix transformation, M1 >2. The important point to remember from this series of examples is that the eigenvalues are the squares of the natural frequencies and that the mode shapes are related to the eigenvectors by a factor of the mass matrix. As indicated in Example 4.2.3, the eigenvectors of a symmetric matrix are orthogonal and can always be calculated to be normal. Such vectors are called orthonormal, as summarized in Window 4.3. This fact can be used to decouple the equations of motion of any order undamped system by making a new matrix P out of the normalized eigenvectors, such that each vector forms a column. Thus the matrix P is defined by p = [v1 v2 v3 cvn] (4.46) where n is the number of degrees of freedom in the system (n = 2 for the examples of this section). Note the matrix P has the unique property that PTP = I, which follows directly from considering the matrix product definition that the ijth element of PTP is the product of the ith row of PT with the jth column of P. Matrices that satisfy the equation PTP = I are called orthogonal matrices. Sec. 4.2 327 Eigenvalues and natural Frequencies Window 4.3 Summary of Orthonormal Vectors Two vectors x1 and x2 are normal if xT1 x1 = 1 and xT2 x2 = 1 and orthogonal if xT1 x2 = 0. If x1 and x2 are normal and orthogonal, they are said to be orthonormal. This is abbreviated xTi xj = δij i = 1, 2, j = 1, 2 where δij is the Kronecker delta, defined by If a set of n vectors δij = e 0 if i ∙ j f 1 if i = j 5 x 6 i = 1 is orthonormal, it is denoted by n xiTxj = δij i, j = 1, 2, . . . n Be careful not to confuse δij, the Kronecker delta used here, with δ = ln[x(t)> x(t + T)], the logarithmic decrement of Section 1.6, or with δ(t – tj), the Dirac delta or impulse function of Section 3.1, or with the static deflection δs of Section 5.2. Example 4.2.4 Write out the matrix P for the system of Example 4.2.3 and calculate PTP. Solution Using the values for the orthonormal vectors v1 and v2 from Example 4.2.3 yields P = [v1 so that PTP becomes P TP = a = v2] = 1 1 J 12 1 1 1 1 ba bJ 12 12 1 1 1 + 1 J 2 1 - 1 1 R -1 1 1 RJ -1 1 1 - 1 1 R = J 1 + 1 0 1 R -1 0 R = I 1 n Another interesting and useful matrix calculation is to consider the product of ∙ the three matrices PTKP. It can be shown (see Appendix C) that this product results 328 Multiple-Degree-of-Freedom Systems Chap. 4 in a diagonal matrix. Furthermore, the diagonal entries are the eigenvalues of the ma∙ trix K and the squares of the system’s natural frequencies. This is denoted by ∙ (4.47) Λ = diag(λi) = PT KP ∙ and is called the spectral matrix of K. The following example illustrates this calculation. Example 4.2.5 ∙ Calculate the matrix PT KP for the two-degree-of-freedom system of Example 4.2.2. 1 1 ∙ J PT KP = 12 1 = = 1 1 J 2 1 1 4 c 2 0 1 3 RJ -1 -1 1 2 RJ -1 2 0 2 d = c 8 0 -1 1 1 R J 3 12 1 1 R -1 4 R -4 0 d = Λ 4 Note that the diagonal elements of the spectral matrix Λ are the natural frequencies ω1 and ω2 squared. That is, from Example 4.2.3, ω21 = 2 and ω22 = 4, so that Λ = diag(ω21 ω22) = diag(2 4). n Examining the solution of Example 4.2.5 and comparing it to the natural frequencies of Example 4.1.3 suggests that in general Λ = diag(λi) = diag(ω2i ) (4.48) This expression connects the eigenvalues with the natural frequencies (i.e., λi = ω2i ). The following example illustrates the matrix methods for vibration analysis presented in this section and provides a summary. Example 4.2.6 Consider the system of Figure 4.4. Write the dynamic equations in matrix form, calcu∙ late K, its eigenvalues and eigenvectors, and hence determine the natural frequencies of the system (use m1 = 1 kg, m2 = 4 kg, k1 = k3 = 10 N>m, and k2 = 2 N>m). Also calculate the matrices P and Λ, and show that equation (4.47) is satisfied and that PTP = I. k1 k2 m1 x1 k3 m2 x2 Figure 4.4 A two-degree-offreedom model of a structure fixed at both ends. Sec. 4.2 329 Eigenvalues and natural Frequencies Solution Using free-body diagrams of each of the two masses yields the following equations of motion: $ m1x1 + (k1 + k2)x1 - k2x2 = 0 $ m2x2 - k2x1 + (k2 + k3)x2 = 0 (4.49) In matrix form this becomes J m1 0 0 $ k + k2 R x(t) + J 1 m2 - k2 - k2 R x(t) = 0 k2 + k3 (4.50) Using the numerical values for the physical parameters mi and ki yields that M = J 1 0 0 R 4 K = J 12 -2 -2 R 12 The matrix M - 1>2 becomes 1 M -1>2 = C 0 1S 2 0 so that ∙ K = M -1>2 ( KM -1>2 ) = C 1 0 1S 2 0 J 12 -2 -1 12 R = J 6 -1 -1 R 3 ∙ ∙ ∙ ∙ Note that K is symmetric (i.e., KT = K), as expected. The eigenvalues of K are calculated from 12 - λ ∙ det(K - λI) = det J -1 -1 R = λ2 - 15λ + 35 = 0 3 - λ This quadratic equation has the solution λ = 15 1 { 185 2 2 so that λ1 = 2.8902 λ2 = 12.1098 Thus ω1 = 1λ1 = 1.7 and ω2 = 1λ2 = 3.48 rad>s. The eigenvectors are calculated from (for λ1) J 12 - 2.8902 -1 -1 v 0 R J 11 R = J R 3 - 2.8902 v21 0 330 Multiple-Degree-of-Freedom Systems so that the vector v1 = [v11 21] T Chap. 4 satisfies 9.1098v11 = v21 1 = 7 v1 7 = 2v211 + v221 = 2v211 + (9.1098)2 v211 Normalizing the vector v1 yields so that v11 = 1 = 0.1091 21 + (9.1098)2 and v21 = 9.1098v11 = 0.9940 Thus the normalized eigenvector v1 = [0.1091 0.9940]T. Similarly, the vector v2 corresponding to the eigenvalue λ2 in normalized form becomes v2 = [– 0.9940 0.1091]T. Note that vT1 v2 = 0 and 2vT1 v1 = 1. The matrix of eigenvectors P becomes P = [v1 v2] = J 0.1091 0.9940 - 0.9940 R 0.1091 Thus the matrix Λ becomes 0.1091 ∙ Λ = PT KP = J - 0.9940 0.9940 12 RJ 0.1091 -1 - 1 0.1091 RJ 3 0.9940 - 0.9940 2.8402 R = J 0.1091 0 0 R 12.1098 This shows that the matrix P transforms the mass-normalized stiffness matrix into a diagonal matrix of the squares of the natural frequencies. Furthermore, P TP = J 0.1091 - 0.9940 0.9940 0.1091 RJ 0.1091 0.9940 - 0.9940 1 R = J 0.1091 0 0 R = I 1 as it should. n The computations made in the previous example can all be performed easily in most programmable calculators as well as in the mathematics software packages used in Sections 1.9, 2.8, and 3.9, in the Toolbox, and as illustrated in Section 4.9. It is good to work a few of the calculations for frequencies and eigenvectors by hand. However, larger problems require the accuracy of using a code. An alternative approach to normalizing mode shapes is often used. This method is based on equation (4.17). Each vector ui corresponding to each natural frequency ω i is normalized with respect to the mass matrix M by scaling Sec. 4.2 331 Eigenvalues and natural Frequencies the mode-shape vector, which satisfies equation (4.17) such that the vector wi = α iui satisfies (αiu i)TM(αiu i) = 1 (4.51) or α2i u Ti Mu i = wTi Mwi = 1 This yields the special choice of αi = 1> 2u Ti Mu i. The vector wi is said to be mass normalized. Multiplying equation (4.17) by the scalar αi yields (for i = 1 and 2) -ω2i Mwi + Kwi = 0 (4.52) Multiplying this by wTi yields the two scalar relations (for i = 1 and 2) ω2i = wTi Kwi i = 1, 2 (4.53) where the mass normalization wTi Mwi = 1 of equation (4.51) was used to evaluate the left side. Next, consider the vector vi = M1 >2ui, where vi is normalized so that vTi vi = 1. Then by substitution vTi vi = (M1>2u i)T M1>2u i = u Ti M1>2 M1>2u i = u Ti Mu i = 1 so that the vector ui is mass normalized. In this last argument, the property of the transpose is used [i.e., (M1 >2u)T = uT(M1 >2)T] and the fact that M1 >2 is symmetric, so that (M1 >2)T = M1 >2. As vibration problems become more complex, more degrees of freedom are needed to model the system’s behavior. Thus the mass and stiffness matrices introduced in Section 4.1 become large and the analysis becomes more complicated, even though the basic principle of eigenvalues>frequencies and eigenvectors>mode shapes remains the same. Increased understanding of the vibration problem can be obtained by borrowing from the theory of matrices and computational linear algebra. There are three different methods of casting the undamped-vibration problem in terms of the matrix theory. They are (i) ω2Mu = Ku (ii) ω2u = M -1Ku (iii) ω2v = M -1>2KM -1>2v Each of these methods results in identical natural frequencies and mode shapes to within numerical precision. Each of these three eigenvalue problems is related by a simple matrix transformation. The following summarizes their differences. (i) The Generalized Symmetric Eigenvalue Problem This is the simplest method to compute by hand. However, when using a code, u must be further normalized with respect to the mass matrix to obtain an orthonormal set. In addition, the 332 Multiple-Degree-of-Freedom Systems Chap. 4 matrix transformation from (i) to (iii) must be used to prove that the resulting ωi are real and to prove the existence of orthogonal mode shapes. This is the second most expensive computational algorithm, requiring 7n3 floating-point operations per second (flops) where n is the number of degrees of freedom. (ii) The Asymmetric Eigenvalue Problem When using a code, u must be further normalized with respect to the mass matrix to obtain an orthonormal set. This is the most expensive computational algorithm, requiring 15n3 flops. (iii) The Symmetric Eigenvalue Problem Because symmetry is preserved, this, or its Cholesky equivalent, is the best method to use for larger systems or when using a code. The resulting eigenvectors vi form an orthonormal set. Algorithms require only n3 flops, including transforming to the symmetric form and transforming the result to physical mode shapes (computed by a simple matrix multiplication). The question remains: Which approach should be used? The symmetric form (iii) is used here because computational algorithms produce orthonormal eigenvectors without additional calculation and because the algorithm for (iii) is less computationally intensive. In addition, the analytical properties of the symmetric eigenvalue problem can be used directly. For two-degree-of-freedom systems, computation should be done by hand, and the most straightforward hand calculation is to proceed with the generalized eigenvalue problem λMu = Ku, as analyzed in Section 4.1. For problems with more than two degrees of freedom, a code should be used to avoid numerical mistakes and ensure accuracy. In these practical cases, the most efficient approach is to use the symmetric eigenvalue problem (iii) presented in this section. ∙ The symmetric eigenvalue problem for the mass-normalized stiffness matrix (K) ∙ provides the transformation (P) that diagonalizes K and uncouples the equations of ∙ motion. This process of transforming K into a diagonal form is called modal analysis and forms the topic of the next section. 4.3 MODAL ANALYSIS The matrix of eigenvectors P calculated in Section 4.2 can be used to decouple the equations of vibration into two separate equations. The two separate equations are then second-order-single-degree-of-freedom equations that can be solved and analyzed using the methods of Chapters 1 through 3. The matrices P and M - 1>2 can be used again to transform the solution back to the original coordinate system. The matrices P and M - 1>2 can also be called transformations, which is appropriate in this case because they are used to transform the vibration problem between different coordinate systems. This procedure is called modal analysis, because the transformation S = M - 1>2P, often called the modal matrix, is related to the mode shapes of the vibrating system. Sec. 4.3 333 Modal Analysis Consider the matrix form of the equation of vibration $ Mx(t) + Kx(t) = 0 (4.54) subject to the initial conditions x(0) = x0 # # x(0) = x0 Here x0 = [x1(0) x2(0)]T is the vector of initial displacements, and # # # x0 = [x1(0) x2(0)]T is the vector of initial velocities. As outlined in Section 4.2, substitution of x = M - 1>2q(t) into equation (4.54) and multiplying from the left by M - 1>2 yields $ ∙ I q(t) + Kq(t) = 0 (4.55) $ $ ∙ where x = M -1>2q, since the matrix M is constant, and where K = M -1>2KM -1>2, as before. The transformation M - 1>2 simply transforms the problem from the coordinate system defined by x = [x1(t) x2(t)]T to a new coordinate system, q = [q1(t) q2(t)]T, defined by q(t) = M1>2x(t) Next, define a second coordinate system r(t) = [r1(t) (4.56) r2(t)]T by (4.57) ∙ where P is the matrix composed of the orthonormal eigenvectors of K as defined in equation (4.46). Substitution of the vector q = Pr(t) into equation (4.55) and multiplying from the left by the matrix PT yields $ ∙ PT P r(t) + PT KPr(t) = 0 (4.58) q(t) = Pr(t) Using the result PTP = I and equation (4.47), this can be reduced to $ I r(t) + Λr(t) = 0 (4.59) Equation (4.59) can be written out by performing the indicated matrix calculations as 1 J 0 $ 0 r1(t) ω2 RJ$ R + J 1 1 r2(t) 0 0 r (t) 0 RJ 1 R = J R ω22 r2(t) 0 (4.60) $ r (t) + ω21r1(t) 0 J $1 R = J R 2 r2(t) + ω2r2(t) 0 (4.61) The equality of the two vectors in this last expression implies the two decoupled equations $ r1(t) + ω21r1(t) = 0 (4.62) and $ r2(t) + ω22r2(t) = 0 (4.63) 334 Multiple-Degree-of-Freedom Systems Chap. 4 These two equations are subject to initial conditions which must also be transformed into the new coordinate system r(t) from the original coordinate system x(t). Following the preceding two transformations applied to the initial displacement yields r0 = J r10 R = PTq(0) = PTM1>2x0 r20 (4.64) where q(t) = Pr(t) was multiplied by PT to get r(t) = PTq(t), since PTP = I. Likewise the initial velocity in the decoupled coordinate system, r(t), becomes # r # # # r0 = J # 10 R = PTq(0) = PT M1>2x0 r20 (4.65) Equations (4.62) and (4.63) are called the modal equations, and the coordinate system r(t) = [r1(t) r2(t)]T is called the modal coordinate system. Equations (4.62) and (4.63) are said to be decoupled because each depends only on a single coordinate. Hence each equation can be solved independently by using the method of Sections 1.1 and 1.2 (see Window 4.4). Denoting the initial conditions individually # # by r10, r10, r20, and r20 and using equation (1.10), the solution of each of the modal equations (4.62) and (4.63) is simply # ω1r10 2ω21r 210 + r 210 r1(t) = sin aω1t + tan-1 # b ω1 r10 (4.66) r2(t) = (4.67) # ω2r20 2ω22r 220 + r 220 sin aω2t + tan-1 # b ω2 r20 provided ω1 and ω2 are nonzero. Window 4.4 Review of the Solution to a Single-Degree-of-Freedom Undamped System from Section 1.1 and Window 1.2 $ $ The solution to mx + kx = 0 or x + ω2nx = 0 subject to x(0) = x0 and # x(0) = v0 is x(t) = where ωn = 1k>m ∙ 0. C x20 + v20 ω2n sin aωnt + tan-1 ωnx0 b v0 (1.10) Sec. 4.3 335 Modal Analysis Once the modal solutions (4.66) and (4.67) are known, the transformations M1 >2 and P can be used on the vector r(t) = [r1(t) r2(t)]T to recover the solution x(t) in the physical coordinates x1(t) and x2(t). To obtain the vector x from the vector r, substitute equations (4.56) into q(t) = Pr(t) to get x(t) = M -1>2q(t) = M -1>2Pr(t) (4.68) The matrix product M - 1>2P is again a matrix, which is denoted by S = M -1>2P (4.69) and is the same size as M and P (2 * 2). The matrix S is called the matrix of mode shapes, each column of which is a mode-shape vector. This procedure, referred to as modal analysis, provides a means of calculating the solution to a two-degree-offreedom vibration problem by performing a number of matrix calculations. The usefulness of this approach is that these matrix computations can easily be automated in a computer code (even on some calculators). In addition, the modal-analysis procedure is easily extended to systems with an arbitrary number of degrees of freedom, as developed in the next section. Figure 4.5 illustrates the coordinate transformation used in modal analysis. The following summarizes the procedure of modal analysis using the matrix transformation S. This is followed by an example that re-solves Example 4.1.6 using modal methods. Figure 4.5 summarizes how computing the matrix of mode shapes S transforms the vibration problem from a coupled set of equations of motion into a set of single-degreeof-freedom problems. Effectively, the matrix S transforms multiple-degree-of-freedom problems that are complicated to solve into single-degree-of-freedom problems that are easy to solve (from Chapter 1). Furthermore, the single-degree-of-freedom problems  21 1 r1 k2 k1 m1 m2 x1 x2 22 1 r2 r  S1x x  Sr Physical coordinates (coupled) where S Modal coordinates (uncoupled) M1/2P Figure 4.5 Schematic illustration of decoupling equations of motion using modal analysis and the matrix of mode shapes S. 336 Multiple-Degree-of-Freedom Systems Chap. 4 obtained by the modal transformation S all have unit mass and each has a stiffness corresponding to one of the system’s natural frequencies squared. Not only is the modal description given on the right side of Figure 4.5 easy to solve, but it forms the basis of most vibration tests, called modal testing, as discussed in Chapter 7. The idea of modal analysis is one of the foundations of vibration analysis (the others being the concepts of natural frequency and resonance), and as such the properties of the matrix S are extremely important. Taking the transpose of equation (4.69) yields ST = (M1>2P)T = PTM -1>2 T T (4.70) T since (AB) = B A (see Appendix C). In addition, the inverse of a matrix product is given by (AB) -1 = B-1A -1 (see Appendix C), so that S -1 = (M1>2P)-1 = P-1M1>2 T (4.71) T However, the matrix P has as its inverse P since P P = I. Thus equation (4.71) yields that the inverse of the matrix of mode shapes is S -1 = PTM1>2 (4.72) These matrix results are useful for solving equation (4.54) by modal analysis. The modal analysis of equation (4.54) starts with the substitution of x(t) = Sr(t) into equation (4.54). Multiplying the result by ST yields $ STMS r(t) + STKSr(t) = 0 (4.73) Expanding the matrix S in equation (4.73) into its factors as given by equation (4.69) yields $ PTM -1>2MM -1>2P r(t) + PTM -1>2KM -1>2Pr(t) = 0 (4.74) or $ ∙ PTP r(t) + PT KPr(t) = 0 Using the properties of the matrix P, this becomes $ r(t) + Λr(t) = 0 (4.75) (4.76) which represents the two decoupled equations (4.62) and (4.63). Recall that these are called the modal equations, r(t) is called the modal coordinate system, and the diagonal matrix Λ contains the squares of the natural frequencies. The initial conditions for r(t) are calculated by solving for r(t) in equation (4.68). Multiplying equation (4.68) by S -1 yields r(t) = S -1x(t), which becomes, after using equation (4.72), r(t) = PTM1>2x(t) = S -1x(t) (4.77) The initial conditions on r(t) are thus # # r(0) = PTM1>2x0 and r(0) = PTM1>2x0 (4.78) Sec. 4.3 337 Modal Analysis as derived in equations (4.64) and (4.65). With the initial conditions transformed into modal coordinates by equations (4.78), the modal equations given by equation (4.76) yield the solution vector in modal coordinates r(t). In order to obtain the solution in the original physical coordinate system, x(t), the transformation S is again used. Multiplying r(t) = S -1x(t) by S to get x(t) = Sr(t) = M -1>2Pr(t) (4.79) yields the solution in physical coordinates. Equation (4.79) effectively takes the solution from the right side of Figure 4.5 back to the left side. The basic idea is that the matrix S -1 takes the problem from the physical coordinates, where the equations of motion are coupled and hard to solve, to modal coordinates, where the equations are uncoupled and easy to solve. Then the modal matrix S takes the solution back to the physical coordinates corresponding to the original problem. These steps are summarized in Window 4.5 and illustrated in the following examples. Window 4.5 Steps in Solving Equation (4.54) by Modal Analysis Calculate M - 1>2. ∙ Calculate K = M -1>2KM -1>2, the mass-normalized stiffness matrix. ∙ Calculate the symmetric eigenvalue problem for K to get ω2i and vi. Normalize vi and form the matrix P = [v1 v2]. Calculate S = M - 1>2P and S-1 = PTM1 >2. # # Calculate the modal initial conditions: r(0) = S-1x0, r(0) = S -1x0. # Substitute the components of r(0) and r(0) into equations (4.66) and (4.67) to get the solution in modal coordinate r(t). 8. Multiply r(t) by S to get the solution x(t) = Sr(t). 1. 2. 3. 4. 5. 6. 7. Note that S is the matrix of mode shapes and P is the matrix of eigenvectors. Example 4.3.1 Calculate the solution of the two-degree-of-freedom system given by M = J 9 0 0 R 1 K = J 27 -3 -3 R 3 1 x(0) = J R 0 # x(0) = 0 using modal analysis. Compare the result to that obtained in Example 4.1.6 for the same system and initial conditions. 338 Multiple-Degree-of-Freedom Systems Chap. 4 Solution From Examples 4.1.5, 4.2.2, 4.2.4, and 4.2.5 the following have been calculated 1 M -1>2 = C 3 0 0 1 1 1 J 12 1 P = 3 ∙ K = J -1 S 1 R -1 -1 R 3 Λ = diag (2, 4) which provides the information required in the first three steps of Window 4.5. The next step is to calculate the matrix S and its inverse. 1 1 C3 12 0 S = M -1>2P = S -1 = PT M1>2 = 1 1 J 12 1 0 SJ 1 1 1 1 3 RJ -1 0 1 1 1 R = C3 -1 12 1 1 3 S -1 0 1 3 R = J 1 12 3 1 R -1 The reader should verify that SS -1 = I, as a check. In addition, note that STMS = I. The modal initial conditions are calculated from equations (4.78): 1 3 r(0) = S-1x0 = J 12 3 3 1 1 12 RJ R = D T -1 0 3 12 # # r(0) = S-1x0 = S-10 = 0 # # so that r1(0) = r2(0) = 3> 12 and r1(0) = r2(0) = 0. Equations (4.66) and (4.67) yield that the modal solutions are r1(t) = r2(t) = 3 π 3 sin a12t + b = cos 12t 2 12 12 3 π 3 sin a2t + b = cos 2t 2 12 12 The solution in the physical coordinate system x(t) is calculated from 1 1 C3 x(t) = Sr(t) = 12 1 3 1 cos 12t (0.5)(cos 12t + cos 2t) 12 3 SD T = J R 3 (1.5)(cos 12t - cos 2t) -1 cos 2t 12 This is, of course, identical to the solution obtained in Example 4.1.7 and plotted in Figure 4.3. n Sec. 4.3 339 Modal Analysis Example 4.3.2 Calculate the response of the system J 1 0 0 $ 12 R x(t) + J -2 4 -2 R x(t) = 0 12 of Example 4.2.6, illustrated in Figure 4.4, to the initial displacement x(0) = [1 # and x(0) = 0 using modal analysis. 1]T Solution Again following the steps illustrated in Window 4.5, the matrices M - 1>2 and ∙ K become 1 M -1>2 = C 0 0 1S 2 12 ∙ K = J -1 -1 R 3 ∙ Solving the symmetric eigenvalue problem for K (this time using a computer and commercial code as outlined in Section 4.9) yields P = J - 0.1091 - 0.9940 - 0.9940 R 0.1091 Λ = diag (2.8902, 12.1098) Here the arithmetic is held to eight decimal places, but only four are shown. The matrices S and S-1 become S = J - 0.1091 - 0.9940 - 0.4970 R 0.0546 S -1 = J - 0.1091 - 1.9881 - 0.9940 R 0.2182 As a check, note that 2.8902 ∙ PT KP = J 0 0 R 12.1098 P TP = I The modal initial conditions become r(0) = S-1x0 = J -0.1091 - 0.9881 - 0.9940 1 - 2.0972 RJ R = J R 0.2182 1 - 0.7758 # r(0) = S-1x0 = 0 # # Using these values of r1(0), r2(0), r1(0), and r2(0) in equations (4.66) and (4.67) yields the modal solutions r1(t) = - 2.0972 cos (1.7001t) r2(t) = - 0.7758 cos (3.4799t) Using the transformation x = Sr(t) yields that the solution in physical coordinates is x(t) = J 0.2288 cos (1.7001t) + 0.7712 cos (3.4799t) R 1.0424 cos (1.7001t) - 0.0424 cos (3.4799t) 340 Multiple-Degree-of-Freedom Systems Chap. 4 Note that x(t) satisfies the initial conditions, as it should. A plot of the responses is given in Figure 4.6. x2(t) x1(t) 1.5 1.5 0.0 0.0 1.5 0 2 4 6 8 10 t 1.5 0 2 4 6 8 10 t Figure 4.6 Plot of the solutions given in Example 4.3.2. The plot of x2(t) in the figure illustrates that the mass m2 is not much affected by the second frequency. This is because the particular initial condition does not cause the first mass to be excited very much in the second mode, i.e., at ω2 = 3.4799 rad>s [note the coefficient of cos (3.4799t) in the equation for x2(t)]. However, the plot of x1(t) clearly indicates the presence of both frequencies, because the initial condition strongly excites both modes (i.e., both frequencies) in this coordinate. The effects of changing the initial conditions on the response can be examined by using the program VTB4_2 in the Engineering Vibration Toolbox to solve Problem TB4.3 at the end of the chapter. Changing initial conditions is also discussed in Section 4.9. n This section introduces the computations of modal analysis for a two-degreeof-freedom system. This entire approach is easily extended to any number of degrees of freedom, as discussed in the following section. Furthermore, the process of modal analysis is easily performed using any of the modern mathematical software packages, as discussed in Section 4.9, or as indicated in the Toolbox associated with this text. This section forms the foundation for the rest of this chapter and is applied to damped systems (in Section 4.5) and forced systems (in Section 4.6). The idea of modal analysis is used again in studying distributed systems (Chapter 6) and vibration testing (Chapter 7). All of this material depends on understanding the modal matrix S, how to compute it, and the physical interpretation of S given in Figure 4.5. 4.4 MORE THAN TWO DEGREES OF FREEDOM Many structures, machines, and mechanical devices require numerous coordinates to describe their vibrational motion. For instance, an automobile suspension was modeled in earlier chapters as a single degree of freedom. However, a car has four wheels; hence a more accurate model is to use four degrees of freedom or Sec. 4.4 341 More than two Degrees of Freedom coordinates. Since an automobile can roll, pitch, and yaw, it may be appropriate to use even more coordinates to describe the motion. Systems with any finite number of degrees of freedom can be analyzed by using the modal analysis procedure outlined in Window 4.5. For each mass in the system and>or for each degree of freedom, there corresponds a coordinate, xi(t), describing its motion in one dimension; this gives rise to an n * 1 vector x(t), with n * n mass matrix M and stiffness matrix K satisfying $ Mx(t) + Kx(t) = 0 (4.80) The form of equation (4.80) also holds if each mass is allowed to rotate or move in the y, z, or pitch and yaw directions. In this situation, the vector x could reflect up to six coordinates for each mass, and the mass and stiffness matrices would be modified to reflect the additional inertia and stiffness quantities. Figure 4.7 illustrates the possibilities of coordinates for a simple element. However, for the sake of simplicity of explanation, the initial discussion is confined to mass elements that are free to move in only one direction. As a generic example, consider the n masses connected by n springs in Figure 4.8. Summing the forces on each of the n masses yields n equations of the form $ mi xi + ki ( xi - xi - 1 ) - ki + 1 ( xi + 1 - xi ) = 0 i = 1, 2 c, n (4.81) where mi denotes the ith mass and ki the ith spring coefficient. In matrix form these equations take the form of equation (4.80) where M = diag ( m1, m2, c, mn ) (4.82) x3 3  x6 2  x5 x2 1  x4 x1 k1 kn1 k2 m1 x1 Figure 4.7 A single mass element illustrating all the possible degrees of freedom. The six degrees of freedom of the rigid body consist of three rotational and three translational motions. If the predominant forward motion of the body is in the x2 direction, such as in an airplane, then θ2 is called roll, θ3 is called yaw, and θ1 is called pitch. kn m2 mn1 mn x2 xn1 xn Figure 4.8 An example of an n-degree-of-freedom system. 342 Multiple-Degree-of-Freedom Systems Chap. 4 and k1 + k2 -k2 K = E 0 f 0 -k2 k2 + k3 -k3 0 -k3 k3 + k4 0 ... ... 0 f kn - 1 + kn -kn -kn kn U (4.83) The n * 1 vector x(t) becomes x1(t) x (t) x(t) = D 2 T f xn(t) (4.84) The notation of Window 4.5 can be directly used to solve n-degree-of-freedom problems. Each of the steps is exactly the same; however, the matrix computations are all n * n and the resulting modal equation becomes the n decoupled equations $ r1(t) + ω21r1(t) = 0 $ r2(t) + ω22r2(t) = 0 f (4.85) $ rn(t) + ω2nrn(t) = 0 There are now n natural frequencies, ωi, which correspond to the eigenvalues of the n * n matrix M - 1>2KM - 1>2. The n eigenvalues are determined from the characteristic equation given by ∙ det ( λI - K ) = 0 (4.86) which gives rise to an nth-order polynomial in λ. The determinant of an n * n matrix A is given by det A = a aps ∙ Aps ∙ n (4.87) s=1 for any fixed value of p between 1 and n. Here aps is the element of the matrix A at the intersection of the pth row and sth column, and ∙ Aps ∙ is the determinant of the submatrix formed from A by striking out the pth row and sth column, multiplied by ( - 1)p { s. The following example illustrates the use of equation (4.87). Sec. 4.4 343 More than two Degrees of Freedom Example 4.4.1 Expand equation (4.87) for p = 1 to calculate the following determinant: det A = det 1 C0 2 3 1 5 -2 1S 3 = 1[(1)(3) - (1)(5)] - 3[(0)(3) - (2)(1)] - 2[(0)(5) - (1)(2)] = 8 n Once the ω2i are determined from equation (4.86), the normalized eigenvectors are obtained following the methods suggested in Example 4.2.3. The key differences between using the modal approach described by Window 4.5 for multiple-degreeof-freedom systems and for a two-degree-of-freedom system is in the computation of the characteristic equation, its solution to get ω2i , and solving for the normalized eigenvectors. With the exception of calculating the matrix P and Λ, the rest of the procedure is simple matrix multiplication. The following example illustrates the procedure for a three-degree-of-freedom system. Example 4.4.2 Calculate the solution of the n-degree-of-freedom system of Figure 4.8 for n = 3 by modal analysis. Use the values m1 = m2 = m3 = 4 kg and k1 = k2 = k3 = 4 N>m, and the initial condition x1(0) = 1 m with all other initial displacements and velocities zero. Solution The mass and stiffness matrices for n = 3 for the values given become 8 K = C -4 0 M = 4I -4 8 -4 0 -4 S 4 Following the steps suggested in Window 4.5 yields 1. M -1>2 = 1 I 2 8 1 ∙ 2. K = M -1>2KM -1>2 = C - 4 4 0 ∙ 3. det ( λI - K ) = det ° C λ - 2 1 0 = (λ - 2) det a J -4 8 -4 0 2 - 4 S = C -1 4 0 1 λ - 2 1 λ - 2 1 - (1) det a J -1 2 -1 0 -1 S 1 0 1 S¢ λ - 1 1 Rb λ - 1 1 0 1 1 R b + (0) det a J λ - 1 0 λ - 2 Rb 1 344 Multiple-Degree-of-Freedom Systems Chap. 4 = (λ - 2)[(λ - 2)(λ - 1) - 1] - 1[(λ - 1) - 0] = (λ - 1)(λ - 2)2 - (λ - 2) - λ + 1 = λ3 - 5λ2 + 6λ - 1 = 0 The roots of this cubic equation are λ1 = 0.1981 λ2 = 1.5550 λ3 = 3.2470 Thus the system’s natural frequencies are ω1 = 0.4450 ω2 = 1.2470 ω3 = 1.8019 ∙ To calculate the first eigenvector, substitute λ1 = 0.1981 into (K - λI)v1 = 0 T and solve for the vector v1 = [ 11 12 13] . This yields C 2 - 0.1981 -1 0 -1 2 - 0.1981 -1 0 v11 0 -1 S C v21 S = C 0 S v31 0 1 - 0.1981 Multiplying out this last expression yields three equations, only two of which are independent: (1.8019)v11 - v21 = 0 - v11 + (1.8019)v21 - v31 = 0 - v21 + (0.8019)v31 = 0 Solving the first and third equations yields v11 = 0.4450v31 and v21 = 0.8019v31 The second equation is dependent and does not yield any new information. Substituting these values into the vector v1 yields 0.4450 v1 = v31 C 0.8019 S 1 vT1 v1 = v231 3(0.4450)2 + (0.8019)2 + 124 = 1 4. Normalizing the vector yields Solving for 31 and substituting back into the expression for v1 yields the normalized version of the eigenvector v1 as 0.3280 v1 = C 0.5910 S 0.7370 Similarly, v2 and v3 can be calculated and normalized to be -0.7370 v2 = C -0.3280 S 0.5910 - 0.5910 v3 = C 0.7370 S - 0.3280 Sec. 4.4 345 More than two Degrees of Freedom The matrix P is then 0.3280 P = C 0.5910 0.7370 - 0.7370 - 0.3280 0.5910 -0.5910 0.7370 S - 0.3280 ∙ (The reader should verify that PTP = I and PT KP = Λ.) 5. The matrix S = M -1>2P = 1 IP or 2 0.1640 S = C 0.2955 0.3685 -0.3685 -0.1640 0.2955 - 0.2955 0.3685 S - 0.1640 and S -1 T = P M 1>2 0.6560 = 2P I = C - 1.4740 - 1.1820 T 1.1820 -0.6560 1.4740 1.4740 1.1820 S - 0.6560 (Again the reader should verify that S -1S = I.) 6. The initial conditions in modal coordinates become # # r(0) = S -1x0 = S -10 = 0 and 0.6560 r(0) = S-1x0 = C - 1.4740 - 1.1820 1.1820 -0.6560 1.4740 1.4740 1 0.6560 1.1820 S C 0 S = C - 1.4740 S - 0.6560 0 - 1.1820 7. The modal solutions of equation (4.85) are each of the form given by equation (4.67) and can now be determined as r1(t) = (0.6560) sin a0.4450t + r2(t) = ( - 1.4740) sin a1.247t + π b = 0.6560 cos (0.4450t) 2 r3(t) = ( - 1.1820) sin a1.8019t + π b = -1.4740 cos (1.2470t) 2 π b = -1.1820 cos(1.8019t) 2 8. The solution in physical coordinates is next calculated from 0.1640 x = Sr(t) = C 0.2955 0.3685 -0.3685 - 0.1640 0.2955 - 0.2955 0.6560 cos (0.4450t) 0.3685 S C -1.4740 cos (1.2470t) S - 0.1640 - 1.1820 cos (1.8019t) 346 Multiple-Degree-of-Freedom Systems Chap. 4 x1(t) 0.1075 cos (0.4450t) + 0.5443 cos (1.2470t) + 0.3492 cos (1.8019t) C x2(t) S = C 0.1938 cos (0.4450t) + 0.2417 cos (1.2470t) - 0.4355 cos (1.8019t) S x3(t) 0.2417 cos (0.4450t) - 0.4355 cos (1.2470t) + 0.1935 cos (1.8019t) The calculations in this example are a bit tedious. Fortunately, they are easily made using software as done in Section 4.9 and in the Engineering Vibration Toolbox. In fact, computing the frequencies, the matrix of eigenvectors P, and subsequently the mode-shape matrix S starting with the determinant is not the recommended way to proceed. Rather, the symmetric algebraic eigenvalue problem should be solved directly, and this is best done using the software methods covered in Section 4.9. The solution in this example is plotted and compared to a numerical simulation in Section 4.10. n Mode Summation Method Another approach to modal analysis is to use the mode summation or expansion method. This procedure is based on a fact from linear algebra—that the eigenvectors of a real symmetric matrix form a complete set (see Window 4.1; i.e., that any n-dimensional vector can be represented as a linear combination of the eigenvectors of an n * n symmetric matrix). Recall the symmetric statement of the vibration problem: $ ∙ I q(t) + Kq(t) = 0 (4.88) ∙ Let vi denote the n eigenvectors of the matrix K, and let λi ∙ 0 denote the corresponding eigenvalues. According to the argument preceding equation (4.41), a solution of (4.88) is q i(t) = vi e { 1λi jt (4.89) since λi = ω2i . This represents two solutions that can be added together following the argument used for equation (1.18) to yield q i(t) = (ai e -1λi jt + bi e 1λi jt )vi (4.90) q i(t) = d i sin (ωi t + ϕi)vi (4.91) or, using Euler’s formula, where di and ϕi are constants to be determined by initial conditions. Since the set of vectors vi, i = 1, 2, c, n are eigenvectors of a symmetric matrix, a linear combination can be used to represent any n * 1 vector, and in particular, the solution vector q(t). Hence q(t) = a d i sin (ωit + ϕi)vi n i=1 (4.92) Sec. 4.4 347 More than two Degrees of Freedom The constants di and ϕi can be evaluated from the initial conditions q(0) = a d i sin ϕivi (4.93) n # q(0) = a ωid i cos ϕivi (4.94) n i=1 and i=1 Multiplying equation (4.93) by vTj and using the orthogonality (see Window 4.3) of the vector vi (i.e., vjTvi = 0 for all values of the summation index i = 1, 2, c, n except for i = j) yields vTj q(0) = d j sin ϕj (4.95) for each value of j = 1, 2, c, n. Similarly, multiplying equation (4.94) by vjT yields # vTj q(0) = ωjd j cos ϕj (4.96) for each j = 1, 2, c, n. Combining equations (4.95) and (4.96) and renaming the index yields ϕi = tan-1 ωivTi q(0) # vTi q(0) i = 1, 2, cn (4.97) and (if ϕi ∙ 0) di = vTi q(0) sin ϕi i = 1, 2, cn (4.98) Equations (4.92), (4.97), and (4.98) represent the solution in modal summation form. Equation (4.92) is sometimes called the expansion theorem and is equivalent to writing a function as a Fourier series. The constants di are sometimes referred to as expansion coefficients. Note as an immediate consequence of equation (4.97) that if the system has zero # initial velocity, q = 0, each coordinate has a phase shift of 90°. The initial conditions in the coordinate system q(t) can also be chosen such that di = 0 for all i = 2, c, n. In this case the summation in equation (4.92) reduces to the single term q(t) = d 1 sin (ω1t + ϕ1)v1 (4.99) This states that each coordinate qi(t) oscillates with the same frequency and phase. To obtain the solution in physical coordinates, recall that x = M - 1>2q so that x(t) = d 1 sin (ω1t + ϕ1)M -1>2v1 (4.100) 348 Multiple-Degree-of-Freedom Systems Chap. 4 The product of a matrix and a vector is another vector. Defining u1 = M - 1>2v1, equation (4.100) becomes x(t) = d 1 sin (ω1t + ϕ1)u 1 (4.101) This states that if the system is given a set of initial conditions such that d2 = d3 = c = dn = 0, then (4.101) is that total solution and each mass oscillates at the first natural frequency (ω1). Furthermore, the vector u1 specifies the relative magnitudes of oscillation of each mass with respect to the rest position. Hence u1 is called the first mode shape. Note that the first mode shape is related to the first ∙ eigenvector of K by u 1 = M -1>2v1 (4.102) This argument can be repeated for each of the indices i so that u2 = M - 1>2v2, u3 = M - 1>2v3, etc., which become the second, third, etc., mode shapes. The series solution x(t) = a d i sin ( ωit + ϕi ) u i n (4.103) i=1 illustrates how each mode shape contributes to forming the total response of the system. The constants of integration di represent a scaling of how each mode participates in the total response. The larger di is, the more the ith mode affects the response. Hence the di are called modal participation factors. The initial condition required to excite a system into a single mode can be determined from equation (4.98). Because of the mutual orthogonality of the eigenvectors, if q(0) is chosen to be one of the eigenvectors, vj, then each di is zero except for the index i = j. Hence to excite the structure in, say, the second mode, choose # q(0) = v2 and q(0) = 0. Then the solution for equation (4.92) becomes q(t) = d 2 sin aω2t + π bv 2 2 (4.104) and each coordinate of q oscillates with frequency ω2. To transform q(0) = v2 into physical coordinates, note that x = M - 1>2q, so that x(0) = M - 1>2q(0) = M - 1>2 v2 = u2. Hence exciting the system by imposing an initial displacement equal to the second mode shape results in each mass oscillating at the second natural frequency. The modal summation method is illustrated in the next example. Example 4.4.3 Consider a simple model of the horizontal vibration of a four-story building as illustrated in Figure 4.9, subject to a wind that gives the building an initial displacement of x(0) = [0.001 0.010 0.020 0.025]T and zero initial velocity. Sec. 4.4 349 More than two Degrees of Freedom m4 x4 m4 k4 / 2 k4(x4  x3) k4 /2 k4(x4  x3) m3 x3 m3 k3 / 2 k3 /2 k3(x3  x2) k3(x3  x2) m2 x2 k2 / 2 m2 k2 / 2 k2(x2  x1) m1 x1 k1/ 2 k2(x2  x1) k1/ 2 m1 k1x1 (a) (b) Figure 4.9 (a) A simple model of the horizontal vibration of a four-story building. Here each floor is modeled as a lumped mass, and the walls are modeled as providing horizontal stiffness. (b) The restoring forces acting on each mass (floor). Solution In modeling buildings, it is known that most of the mass is in the floor of each section and that the walls can be treated as massless columns providing lateral stiffness. From Figure 4.9 the equations of motion of each floor are $ m1x1 + (k1 + k2)x1 - k2x2 = 0 $ m2x2 - k2x1 + (k2 + k3)x2 - k3x3 = 0 $ m3x3 - k3x2 + (k3 + k4)x3 - k4x4 = 0 $ m4x4 - k4x3 + k4x4 = 0 In matrix form, these four equations can be written as m1 0 D 0 0 0 m2 0 0 0 0 m3 0 0 k1 + k2 0 $ - k2 Tx + D 0 0 m4 0 -k2 k2 + k3 - k3 0 0 -k3 k3 + k4 - k4 0 0 Tx = 0 -k4 k4 Some reasonable values for a building are m1 = m2 = m3 = m4 = 4000 kg and k1 = k2 = k3 = k4 = 5000 N>m. In this case, the numerical values of M and K become M = 4000I 10,000 - 5000 K = D 0 0 - 5000 10,000 -5000 0 0 -5000 10,000 - 5000 0 0 T - 5000 5000 350 Multiple-Degree-of-Freedom Systems Chap. 4 To simplify the calculations, each matrix is divided by 1000. Since the equation of motion is homogeneous, this corresponds to dividing both sides of the matrix equation by 1000 so that the equality is preserved. The initial conditions are 0.001 0.010 x(0) = D T 0.020 0.025 0 0 # x = D T 0 0 ∙ The matrices M - 1>2 and K become M -1>2 1 = I 2 2.5 1.25 ∙ K = D 0 0 - 1.25 2.5 -1.25 0 0 - 1.25 2.5 - 1.25 The matrix M1 >2 and the initial condition on q(t) become M 1>2 = 2I q(0) = M 0.002 0.020 x(0) = D T 0.040 0.050 1>2 0 0 T -1.25 1.25 # q(0) = M1>20 = 0 Using an eigenvalue solver (see Section 4.9 for details or use the files discussed at the ∙ end of this chapter and contained in the Toolbox), the eigenvalue problem for K yields λ1 = 0.1508 0.2280 0.4285 v1 = D T 0.5774 0.6565 λ2 = 1.2500 0.5774 0.5774 v2 = D T 0.0 - 0.5774 λ3 = 2.9341 0.6565 -0.2280 v3 = D T -0.5774 0.4285 λ4 = 4.4151 - 0.4285 0.6565 v4 = D T - 0.5774 0.2280 Converting this into natural frequencies and mode shapes (ωi = 1λi and ui = M - 1>2vi) yields ω1 = 0.3883, ω2 = 1.1180, ω3 = 1.7129, ω4 = 2.1012, and 0.1140 0.2143 u1 = D T 0.2887 0.3283 0.2887 0.2887 u2 = D T 0.0 - 0.2887 0.3283 - 0.1140 u3 = D T - 0.2887 0.2143 - 0.2143 0.3283 u4 = D T - 0.2887 0.1140 # Since q(0) = 0, equation (4.96) yields that each of the phase shifts is ϕi = π>2 and equation (4.98) becomes di = vTi q(0) sin (π>2) = vTi q(0) 351 More than two Degrees of Freedom Substituting vTi and q(0) into the expansion above yields the following values for the modal participation factors: d 1 = 0.065 d 2 = - 0.016 d 3 = - 4.9 * 10-3 d 4 = 5.8 * 10-4 The solution given by equation (4.103) then becomes (in meters) 0.007 - 4.67 * 10-3 0.014 - 4.67 * 10-3 T cos (0.3883t) + D T cos (1.1180t) x(t) = D 0.019 0 0.021 4.67 * 10-3 -1.61 5.60 + D 1.42 - 1.05 * * * * 10-3 - 1.24 10-4 1.91 T cos (1.7129t) + D 10-3 - 1.68 6.62 10-3 * * * * 10-4 10-4 T cos (2.101t) 10-4 10-5 The mode shapes u1, u2, u3, and u4 are plotted in Figure 4.10. Note that the modal participation factor d4 is much smaller than the others. Thus oscillation at 2.102 rad>s will not be too evident. The response of x3 will be dominated almost completely by the first natural frequency. 0 0 0 0 x4 Coordinate Sec. 4.4 x3 x2 x1 0 Mode number 1 2 3 4 Figure 4.10 Plot of the four mode shapes associated with the solution for the system of Figure 4.9 (not to scale). n Nodes of a Mode A node of a mode shape is simply the coordinate of a zero entry in the mode shape. For instance, the second mode shape in Example 4.4.3 has a zero value in the location of coordinate x3(t). Thus the third coordinate is a node of the second mode. This means that if the system is excited by an initial condition to vibrate only at the second natural frequency, the third coordinate will not move! Thus a node has a place of no motion for certain initial conditions. If a sensor were to be placed on the third mass, it would not be able to measure any vibration at the second natural frequency, because that mass does not have a response at ω2. Nodes also make excellent mounting points for machines. Note the word node is also used in finite elements to mean something different (see Chapter 8). 352 Multiple-Degree-of-Freedom Systems Chap. 4 Rigid-Body Modes It often happens that a vibrating system is also translating, or rotating, away from its equilibrium position in one coordinate while the other coordinates are vibrating about their equilibrium point. Such systems are said to be unrestrained and technically violate the stability conditions given in Section 1.8. An example is a train (see Problem 4.13), where the coupling between each car may be modeled as a spring, the cars themselves as lumped masses. As the train rolls down a track it is moving with rigid-body, unrestrained motion while the cars vibrate relative to each other. Figure 4.11 is an example of an unrestrained, two-degreeof-freedom system. m2g m1g x1 m1 x2 k m2 m1 N1 k(x2  x1) m2 N2 Figure 4.11 An unrestrained, two-degree-of-freedom system illustrating both rigidbody translation and vibration. The existence of the unrestrained degree of freedom in the equations of motion changes the analysis slightly. First, the motion consists of a translation plus a vibration. Second, the stiffness matrix becomes singular and the eigenvalue problem results in a value of zero for one of the natural frequencies. The zero frequency renders equation (4.66) incorrect and requires the modal participation factors given by equations (4.95) and (4.96) to be altered. The following example illustrates how to compute the response for a system with unrestrained motion and how to correct these equations for a zero natural frequency. Example 4.4.4 Compute the solution of the unrestrained system given in Figure 4.11 using both the eigenvector method and modal analysis. Let m1 = 1, m2 = 4, k = 4, x0 = [1 0]T, and v0 = 0. Assume that the units are consistent. Solution Summing forces in the horizontal direction on each of the free-body diagrams given in Figure 4.11 yields $ m1x1 = k(x2 - x1) $ m2x2 = - k(x2 - x1) Sec. 4.4 353 More than two Degrees of Freedom Bringing all the forces to the left side and writing in matrix form yields $ 1 0 x 1 -1 x1 0 J R J $1 R + 4 J RJ R = J R 0 4 x2 -1 1 x2 0 Note that the determinant of the stiffness matrix K is zero, indicating that it is singular and hence has a zero eigenvalue (see Appendix C). Following the steps of Window 4.5 and substituting the values for M and K yields the following: 1. M -1>2 = J 1 0 0 1R 2 1 ∙ 2. K = M - 1>2 KM - 1>2 = 4 J 0 0 1R 2 J 1 -1 -1 1 RJ 1 0 0 1R 2 = J 4 -2 -2 R 1 3. Calculating the eigenvalue problem for 2 yields 4 - λ ∙ det ( K - λI ) = det a J -2 -2 R b = ( λ2 - 5λ ) = 0 1 - λ This has solutions λ1 = 0 and λ2 = 5, so that ω 1 = 0 and ω 2 = 15 = 2.236 rad > s. Note the zero eigenvalue > frequency. However, the eigenvector for λ1 is not zero (eigenvectors are never zero) as the following calculation for the eigenvector for λ1 = 0 yields J 4 - 0 -2 -2 v 0 R J 11 R = J R or 4v11 - 2v21 = 0 1 - 0 v21 0 Thus 2 11 = 21, or v1 = [1 2]T. Repeating the procedure for λ2 yields v2 = [2 - 1]T. 4. Normalizing both eigenvectors yields v1 = J 0.4472 - 0.8944 R and v2 = J R 0.8944 0.4472 Note that the eigenvector v1 associated with the eigenvalue λ1 = 0 is not zero. Combining these to form the matrix of eigenvectors yields P = J As a check note that 0.4472 0.8944 - 0.8944 R 0.4472 ∙ PTP = I and PT KP = diag[0 5] 5. Calculating the matrix of mode shapes yields S = M -1>2P = J S-1 = PTM1>2 = J 1 0 0 1R 2 J 0.4472 - 0.8944 0.4472 0.8944 - 0.8944 0.4472 R = J 0.4472 0.4472 1.7889 R 0.8944 -0.8944 R 0.2236 354 Multiple-Degree-of-Freedom Systems Chap. 4 6. Calculating the modal initial conditions yields r(0) = S-1x0 = J = J 0.4472 - 0.8944 1.7889 1 RJ R 0.8944 0 0.4472 # R , r(0) = S-1v0 = 0 - 0.8944 7. Here is where the zero eigenvalue makes a difference as equations (4.66) and (4.67) only apply for the second natural frequency. The modal equation for the $ first mode becomes r1 = 0, which has solution r1(t) = a + bt. Here a and b are the constants of integration to be determined from the modal initial conditions. Applying the modal initial conditions yields the two equations r1(0) = a = 0.4472 # r1(0) = b = 0.0 Thus the first modal equation has solution r1(t) = 0.4472 a constant. The solution for the second mode follows directly from equations (4.66) and (4.67) as before and yields r2(t) = - 0.894 cos ( 15t ) Thus the modal response vector is r(t) = J 0.447 R - 0.894 cos ( 15t ) 8. Transforming back into the physical coordinates yields the solution x(t) = Sr(t) = J = J 0.4472 0.4472 -0.8944 0.447 RJ R 0.2236 - 0.894 cos ( 15t ) 0.2 + 0.8 cos ( 15t ) R 0.2 - 0.2 cos ( 15t ) Note that each of the two physical coordinates moves a constant distance 0.2 units and then oscillates at the second natural frequency. Next consider the effect of the zero frequency in using the mode summation method. In this case, equation (4.90) becomes q 1(t) = (a + bt)v1 and equations (4.93) and (4.94) become q(0) = (a + b0)v1 + a d i sin ϕivi and q (0) = bvi + a ωid i cos ϕivi n .# i=2 n i=2 Following the same steps as before and using orthogonality yields # a = vT1 q(0) and b = vT1 q(0) Sec. 4.4 355 More than two Degrees of Freedom which replace the modal constants of integration di and ϕi for the zero valued mode. Computing d2 and ϕ2 following equations (4.97) and (4.98), and combining the modes according to equation (4.103), again yields the solution x(t) = au1 + d 2 cos ( 15t)u2 = J 0.2 + 0.8 cos ( 15t ) R 0.2 - 0.2 cos ( 15t ) The solutions are plotted in Figure 4.12. 1 x1(t) x2(t) 0 –1 2 4 6 8 10 t Figure 4.12 Plots of the solution versus time (s) for Example 4.4.4 showing vibration superimposed over a rigid-body mode. n Note that in all the previous examples of more than two-degrees-of freedom, the stiffness matrix K is banded (i.e., the matrix has nonzero elements on the diagonal and one element above and below the diagonal, the other elements being zero). This is typical of structural models but is not necessarily the case for machine parts or other mechanical devices. The concept of mode shapes presented in this section and illustrated in Example 4.4.3 is extremely important. The language of modes, mode shapes, and natural frequencies forms the basis for discussing vibration phenomena of complex systems. The word mode generally refers to both the natural frequency and its corresponding mode shape. A mode shape is a mathematical description of a deflection. It forms a pattern that describes the shape of vibration if the system were to vibrate only at the corresponding natural frequency. It is neither tangible nor simple to observe; however, it provides a simple way to discuss and understand the vibration of complex objects. Its physical significance lies in the fact that every vibrational response of a system consists of some combinations of mode shapes. An entire industry has been formed around the concept of modes. 356 Multiple-Degree-of-Freedom Systems Chap. 4 4.5 SYSTEMS WITH VISCOUS DAMPING Viscous energy dissipation can be introduced to the modal analysis solution suggested previously in two ways. Again, as in modeling single-degree-of-freedom systems, viscous damping is introduced more as a mathematical convenience rather than a physical truth. However, viscous damping provides an excellent model in many physical situations and represents a significant improvement over the undamped model. The simplest method of modeling damping is to use modal damping. Modal damping places an energy dissipation term of the form # 2ζiωiri(t) (4.105) # in equations (4.85). Here ri(t) denotes the velocity of the ith modal coordinate, ωi is the ith natural frequency, and ζi is the ith modal damping ratio. The modal damping ratios, ζi, are assigned by “experience” or measurement (see Chapter 7) to be some number between 0 and 1, or by making measurements of the response and estimating ζi. Usually, ζi is small unless the structure contains viscoelastic material or a hydraulic damper is present. Common values are 0 … ζ 6 0.05 (see Section 5.6). An automobile shock absorber, which uses fluid, may yield values as high as ζ = 0.5. Once the modal damping ratios are assigned, equations (4.85) become $ # ri(t) + 2ζiωiri(t) + ω2i ri(t) = 0 i = 1, 2, c, n (4.106) which have solutions of the form (0 6 ζi 6 1) ri(t) = Aie -ζiωit sin (ωdit + ϕi) i = 1, 2, c, n (4.107) where Ai and ϕi are constants to be determined by the initial conditions and ωdi = ωi 21 - ζi2 as given in Window 4.6. Once this modal solution is established, Window 4.6 Review of a Damped Single-Degree-of-Freedom System $ # # # $ # The solution of mx + cx + kx = 0, x(0) = x0, x(0) = x0, or x + 2ζωnx + ω2nx = 0 is (for the underdamped case 0 6 ζ 6 1) x(t) = Ae -ζωnt sin (ωdt + θ) where ωn = 1k>m, ζ = c>(2m ωn), ωd = ωn 21 - ζ2, and A = c # (x0 + ζωnx0)2 + (x0ωd)2 ω2d d from equations (1.36), (1.37), and (1.38). 1>2 x0ωd θ = tan-1 # x0 + ζωnx0 Sec. 4.5 357 Systems with Viscous Damping the modal analysis method of solution suggested in Window 4.5 is used to transform the response into the physical coordinate system. The only difference is that equation (4.107) replaces step 7 where (ωi ∙ 0) # (ri0 + ζiωiri0)2 + (ri0ωdi)2 1>2 d (4.108) Ai = c 2 ωdi ri0ωdi ϕi = tan-1 # ri0 + ζiωiri0 (4.109) # # Here ri0 and ri0 are the ith elements of r(0) and r(0), respectively. Equations (4.108) and (4.109) are derived directly from equation (1.38) for a single-degree-of-freedom system of the same form as equation (4.107). These equations are correct only if each modal damping ratio is underdamped and no rigid-body modes are present. If a zero frequency exists, then the method of Example 4.4.4 must be used. The following example illustrates the solution technique for a system with assumed modal damping. Example 4.5.1 Consider again the system of Example 4.3.1, which has equation of motion J 9 1 0 $ 27 R x(t) + J -3 0 -3 1 # R x(t) = 0, x(0) = J R , x(t) = 0 3 0 and calculate the solution of the same system if modal damping of the form ζ1 = 0.05 and ζ2 = 0.1 is assumed. Solution From Example 4.3.1, ω1 = 12 and ω2 = 2. Since ωdi = ωi 21 - ζi2 for underdamped systems, then the damped natural frequencies become ωd1 = 12[1 - (0.05)2]1/2 = 1.4124 and ωd2 = 2[1 - (0.1)2]1>2 = 1.9900. The modal initial conditions calculated in Example 4.3.1 are r10 = r20 = 3> 12 and # # r10 = r20 = 0. Substitution of these values into equations (4.108) and (4.109) yields A1 = 2.1240 ϕ1 = 1.52 rad (87.13°) A2 = 2.1340 ϕ2 = 1.47 rad (84.26°) Note that compared to the constant Ai and ϕi in the magnitude and phase of the undamped system, only a small change occurs in the amplitude. The phase, however, changes 3° and 6°, respectively, because of the damping. The solution is then of the form x(t) = Sr(t) = 1 1 C3 12 1 1 2.1240e -0.0706t sin (1.4124t + 1.52) R 3S J 2.1320e -0.2t sin (1.9900t + 1.47) -1 or x(t) = J 0.5006e -0.0706t sin (1.4124t + 1.52) + 0.5025e -0.2t sin (1.9900t + 1.47) R 1.5019e -0.0706t sin (1.4124t + 1.52) - 1.5076e -0.2t sin (1.9900t + 1.47) 358 Multiple-Degree-of-Freedom Systems Chap. 4 A plot of x1(t) versus t and x2(t) versus t is given in Figure 4.13. x2(t) x1(t) 1.0 1.0 0.5 0.0 0.0 1.0 0.5 0 20 40 t 100 80 60 0 20 40 60 80 t 100 Figure 4.13 Plot of the damped response of the system of Example 4.5.1. n The modal damping ratio approach is also easily applicable to the mode summation method of Section 4.4. In this case, equation (4.91) is replaced by the damped version q i(t) = d ie -ζiωit sin (ωdit + ϕi)vi (4.110) The initial displacement condition calculation becomes q i(0) = d i sin (ϕi)vi so that equation (4.95) still holds. However, the velocity becomes # q i(t) = d i [ωdie -ζiωit cos (ωdit + ϕi) - ζiωie -ζiωit sin (ωit + ϕi)]vi (4.111) (4.112) or at t = 0, # q i(0) = d i (ωdi cos ϕi - ζiωi sin ϕi)vi (4.113) Multiplying equations (4.111) and (4.113) by vTi from the left and solving for the constants ϕi and di yields ϕi = tan-1 ωdivTi q(0) # vTi q(0) + ζiωivTi q(0) viTq(0) di = sin ϕi (4.114) Again, the values of ζi are assigned based on experience or on measurement; then the calculations of Section 4.4 are used with the initial conditions given by equations (4.114). The solution given by equation (4.103) is replaced by x(t) = a d ie -ζiωit sin (ωdit + ϕi)u i n i=1 which yields the damped response. (4.115) Sec. 4.5 359 Systems with Viscous Damping Example 4.5.2 Recall Example 4.4.3, and the model of building vibration defined by the equation of motion 10,000 -5000 $ 4000I x(t) + D 0 0 -5000 10,000 - 5000 0 0 - 5000 10,000 - 5000 0 0 T x(t) = 0 -5000 5000 subject to a wind that gives the building an initial displacement of x(0) = [0.001 0.010 0.020 0.025]T and zero initial velocity, and assume that the damping in the building is measured to be about ζ = 0.01 in each mode. Solution Each of the steps of the solution to Example 4.4.3 is the same until the # # initial conditions are calculated. From equation (4.114) with q(0) = M1>2x(0) = 0, for each i, ϕi = tan-1 21 - ζ2i ωdi = tan-1 ζiωi ζi For ζi = 0.01, this becomes ϕi = 89.42° i = 1, 2, 3, 4 (or 1.56 rad) Since 1>sin(89.42°) = 1.00005, the expansion coefficients di are taken to be vTi q(0), as calculated in Example 4.4.3 because of the small damping. The natural frequencies are ω1 = 0.3883, ω2 = 1.1180, ω3 = 1.7129, and ω4 = 2.1012 rad>s. With a damping ratio of 0.01 assigned to each mode, the damped natural frequencies become (ωdi = ωi 21 - ζ2i ) nearly the same to the third decimal place as the natural frequencies. The value of the exponent in the exponential decay terms become -ζ1ω1 = - 0.004, -ζ2ω2 = - 0.011, - ζ3ω3 = -0.017, and -ζ4ω4 = - 0.021 The mode shapes are unchanged, so the solution becomes 0.007 - 4.67 * 10-3 - 4.67 * 10-3 -0.011t 0.014 -0.004t x(t) = D Te cos (0.388t + 1.56) + D Te cos (1.118t + 1.56) 0.019 0 4.67 * 10-3 0.021 -1.61 5.60 + D 1.42 - 1.05 * * * * 10-3 - 1.24 10-4 -0.017t 1.91 Te cos (1.713t + 1.56) + D 10-3 - 1.68 10-3 6.62 * * * * 10-4 10-4 T e -0.021t cos (2.101t + 1.56) 10-4 10-5 Each coordinate of x(t) is plotted in Figure 4.14. Note that each plot shows the effects of multiple frequencies and is lightly damped. Window 4.7 summarizes this use of modal damping in the mode summation method. 360 Multiple-Degree-of-Freedom Systems Chap. 4     [ W  [ W  [ W  [ W          W Figure 4.14 A plot of the damped response of the system of Example 4.5.2, which is the building of Example 4.4.3 with damping of ζ = 0.01 in each mode. The plot was made in Mathcad so that x1(t) is denoted by x(t)0, x2(t) is x(t)1, etc. n Damping can also be modeled directly. For instance, consider the system of Figure 4.15. The equations of motion of this system can be found from summing the forces on each mass, as before. This yields the following equations of motion in matrix form: J m1 0 0 $ c + c2 Rx + J 1 -c2 m2 -c2 # k + k2 Rx + J 1 c2 -k2 -k2 Rx = 0 k2 (4.116) where x = [x1(t) x2(t)]T. Equation (4.116) yields an example of a damping matrix C, defined by C = J c1 + c2 -c2 -c2 R c2 (4.117) Here c1 and c2 refer to the damping coefficients indicated in Figure 4.15. The damping matrix C is symmetric and, in a general n-degree-of-freedom system, will be an Sec. 4.5 361 Systems with Viscous Damping Window 4.7 Modal Damping in the Mode Summation Method First transform the undamped equations of motion into the q coordinate system and calculate ωi and vi. Choose the modal damping ratios and write q(t) = a d i e -ζiωit sin (ωdit + ϕi)vi n i=1 where M -1>2KM -1>2vi = ω2i vi, ωdi = ωi 21 - ζ2i , d i = vTi q(0) sin ϕi and ϕi = tan-1 ωdivTi q(0) # vTi q(0) + ζiωivTi q(0) Recall the initial conditions are found from q(0) = M1>2x(0) and # # q(0) = M1>2x(0) Once q(t) is computed, transform back to the physical coordinate system by x(t) = M -1>2q(t) k1 k2 m1 c1 m2 c2 x1 x2 Figure 4.15 A two-degree-of-freedom system with viscous damping. n * n matrix. Thus a damped n-degree-of-freedom system is modeled by equations of the form $ # Mx + Cx + Kx = 0 (4.118) The difficulty with modeling damping in this fashion is that modal analysis cannot in general be used to solve equation (4.118) unless CM -1K = KM-1C holds. This is true because the damping provides additional coupling between the equations of motion that cannot always be decoupled by the modal transformation S (see Caughey and O’Kelly, 1965). Other methods can be used to solve equation (4.118), as discussed in Sections 4.9 and 4.10. 362 Multiple-Degree-of-Freedom Systems Chap. 4 Modal analysis can be used directly to solve equation (4.118) if the damping matrix C can be written as a linear combination of the mass and stiffness matrix, that is, if C = αM + βK (4.119) where α and β are constants. This form of damping is called proportional damping. Substitution of equation (4.119) into equation (4.118) yields $ # Mx(t) + (αM + βK)x(t) + Kx(t) = 0 (4.120) Substitution of x(t) = M - 1>2q(t) and multiplying by M - 1>2 yields $ ∙ # ∙ q(t) + (αI + βK)q(t) + Kq(t) = 0 (4.121) Continuing to follow the steps of Window 4.5, substituting q(t) = Pr(t) and premul∙ tiplying by PT where P is the matrix of eigenvectors of K, yields $ # r(t) + (αI + βΛ)r(t) + Λr(t) = 0 (4.122) This corresponds to the n decoupled modal equations $ # ri(t) + 2ζiωiri(t) + ω2i ri(t) = 0 where 2ζiωi = α + βω2i (4.123) or ζi = βωi α + 2ωi 2 i = 1, 2, c, n (4.124) Here α and β can be chosen to produce some measured (or desired, in the design case) values of the modal damping ratio ζi. On the other hand, if α and β are known, equation (4.124) determines the value of the modal damping ratios ζi. The solution of equation (4.123) for the underdamped case (0 6 ζi 6 1) is ri(t) = Aie -ζiωit sin (ωdit + ϕi) (4.125) where Ai and ϕi are determined by applying the initial conditions on r(t). The solution in physical coordinates is then calculated from x(t) = Sr(t), where S = M - 1>2 P as before. The most general case of proportional damping is if CM-1K = KM -1C holds. Note that if equation (4.119) holds, then CM-1K = KM -1C is satisfied. 4.6 MODAL ANALYSIS OF THE FORCED RESPONSE The forced response of a multiple-degree-of-freedom system can also be calculated by use of modal analysis. For example, consider the building system of Figure 4.9 with a force F4(t) applied to the fourth floor. For example, this force could be the result of an out-of-balance rotating machine on the fourth floor. The equation of motion takes the form $ # Mx + Cx + Kx = BF(t) (4.126) Sec. 4.6 363 Modal Analysis of the Forced Response where F(t) = [0 0 0 F4(t)]T and the matrix B is given by 0 0 B = D 0 0 0 0 0 0 0 0 0 0 0 0 T 0 1 On the other hand, if the different forces are applied at each degree of freedom, B and F(t) would take on the form 1 0 B = D 0 0 0 1 0 0 0 0 1 0 0 0 T, 0 1 F1(t) F (t) F(t) = D 2 T F3(t) F4(t) (4.127) Alternately, if only a single force is applied at one coordinate, the matrix B may be collapsed to the vector b and the applied force reduces to the scalar F(t). For example, the single force F4(t) applied to the fourth coordinate may also be written in equation (4.126) as bF4(t), where b = [0 0 0 1]T. The approach of modal analysis again follows Window 4.5 and uses transformations to reduce equation (4.126) to a set of n decoupled modal equations, which in this case will be inhomogeneous. Then the methods of Chapter 3 can be applied to solve for the individual forced response in the modal coordinate system. The modal solution is then transformed back into the physical coordinate system. To this end, assume that the damping matrix C is proportional of the form given by equation (4.119). Following the procedure in Window 4.5, let x(t) = M - 1>2q(t) in equation (4.126) and multiply by M - 1>2. This yields $ ∙# ∙ Iq(t) + Cq(t) + Kq(t) = M -1>2BF(t) (4.128) ∙ ∙ where C = M -1>2CM -1>2. Next, calculate the eigenvalue problem for K. Let q(t) = ∙ T Pr(t), where P is the matrix of eigenvectors of K and multiply by P . This yields $ # (4.129) r(t) + diag [2ζiωi]r(t) + Λr(t) = PTM -1>2BF(t) where the matrix diag[2ζi ωi] follows from equation (4.123). The vector PTM - 1>2 BF(t) has elements fi(t) that will be linear combinations of the forces Fi applied to each mass. Hence the decoupled modal equations take the form $ # (4.130) ri(t) + 2ζiωiri(t) + ω2i ri(t) = fi(t) Referring to Section 3.2, this has the solution (reviewed in Window 4.8 for the underdamped case) t ri(t) = d ie -ζiωit sin (ωdit + ϕi) + 1 -ζiωit e fi(τ)e ζiωiτ sin ωdi(t - τ) dτ ωdi L0 (4.131) 364 Multiple-Degree-of-Freedom Systems Chap. 4 Window 4.8 Forced Response of an Underdamped System from Section 3.2 The response of an underdamped system $ # mx(t) + cx(t) + kx(t) = F(t) (with zero initial conditions) is given by (for 0 6 ζ 6 1) t x(t) = 1 -ζωnt F(τ)e ζωnτ sin ωd(t - τ) dτ e mωd L0 where ωn = 1k>m, ζ = c>(2mωn), and ωd = ωn 21 - ζ2. With nonzero initial conditions this becomes t x(t) = Ae -ζωnt sin (ωdt + ϕ) + 1 -ζωnt e f(τ)e ζωnτ sin ωd(t - τ) dτ ωd L0 where f = F>m and A and ϕ are constants determined by the initial conditions. where di and ϕi must be determined by the modal initial conditions and ωdi = ωi 21 - ζ2i as before. Note that fi may represent a sum of forces if more than one force is applied to the system. In addition, if a force is applied to only one mass of the system, this force becomes applied to each of the modal equations (4.131) by the transformation S, as illustrated in the following example. Example 4.6.1 Consider the simple two-degree-of-freedom system with a harmonic force applied to one mass as indicated in Figure 4.16. For this example, let m1 = 9 kg, m2 = 1 kg, k1 = 24 N>m, and k2 = 3 N>m. Also assume that the damping is proportional with α = 0 and β = 0.1, so that c1 = 2.4 N # s>m and c2 = 0.3 N # s>m. Calculate the steady-state response. F2(t)  3 cos 2t k2 k1 m2 m1 c2 c1 x1 x2 Figure 4.16 A damped two-degreeof-freedom system for Example 4.6.1. Sec. 4.6 365 Modal Analysis of the Forced Response Solution The equations of motion in matrix form become J 9 0 0 $ 2.7 Rx + J - 0.3 1 - 0.3 # 27 Rx + J 0.3 -3 The matrices M1 >2 and M - 1>2 become 3 0 M1>2 = J 0 R 1 -3 0 Rx = J 3 0 1 M -1>2 = C 3 0 0 0 0 RJ R 1 F2(t) S 1 so that 0.3 ∙ C = M -1>2CM -1>2 = J - 0.1 - 0.1 R 0.3 3 ∙ and K = J -1 -1 R 3 ∙ The eigenvalue problem for K yields λ1 = 2 λ2 = 4 P = 0.7071 J 1 1 -1 R 1 Hence the natural frequencies of the system are ω1 = 12 and ω2 = 2; the matrices ∙ ∙ PT CP and PT KP become 0.2 ∙ P T CP = J 0 0 R 0.4 2 ∙ and PT KP = J 0 0 R 4 The vector f(t) = PTM - 1>2BF(t) becomes f(t) = J 0.2357 - 0.2357 0.7071 0 F (t) RJ R = 0.7071 J 2 R 0.7071 F2(t) F2(t) Hence the decoupled modal equations become $ # r1 + 0.2r1 + 2r1 = 0.7071(3) cos 2t = 2.1213 cos 2t $ # r2 + 0.4r2 + 4r2 = 0.7071(3) cos 2t = 2.1213 cos 2t # Comparing the coefficient of ri in each case to 2ζiωi yields 0.2 = 0.0707 212 0.4 = 0.1000 ζ2 = 2(2) ζ1 = Thus the damped natural frequencies become ωd1 = ω1 21 - ζ21 = 1.4106 ≈ 1.41 ωd2 = ω2 21 - ζ22 = 1.9899 ≈ 1.99 366 Multiple-Degree-of-Freedom Systems Chap. 4 Note that while the force F2 is applied only to mass m2, it becomes applied to both coordinates when transformed to modal coordinates. The modal equations for r1 and r2 can be solved by equation (4.131), or in this case of a simple harmonic excitation, the particular solution is given directly by equation (2.36) as r1p(t) = 2.1213 2 2(2 - 4) + [2(0.0707) 12(2)] 2 cos a2t - tan-1 2(0.0707) 12(2) 222 - 22 = (1.040) cos (2t + 0.1974) = 1.040 cos (2t - 2.9449) b Note that the argument of the arctangent function is negative ( 222 - 22 6 0 ) so that the fourth quadrant angle must be used (see Window 2.4), yielding 2.9449 radians. The second mode particular solution is r2p(t) = 2.1213 2 2(4 - 4) + (2(0.1)(2)(2)) = 2.6516 cos a2t - 2 cos a2t - tan-1 2(0.1)(2)(2) π b = 2.6516 sin 2t 2 22 - 22 b Here rip is used to denote the particular solution of the ith modal equation. Note that r2(t) is excited at its resonance frequency but has high damping, so that the larger but finite amplitude for r2p(t) is not unexpected. If the transient response is ignored [it dies out per equation (2.30)], the preceding solution yields the steady-state response. The solution in the physical coordinate system is xss(t) = M -1>2Pr(t) = J 0.2357 0.7071 - 0.2357 1.040 cos (2t - 2.9442) RJ R 0.7071 2.6516 sin 2t so that in the steady state x1(t) = 0.2451 cos (2t - 2.9442) - 0.6249 sin 2t x2(t) = 0.7354 cos (2t - 2.9442) + 8749 sin 2t Note that even though there is a fair amount of damping in the resonant mode, the coordinates each have a large component vibrating near the resonant frequency. n Resonance The concept of resonance in multiple-degree-of-freedom systems is similar to that introduced in Section 2.2 for single-degree-of-freedom systems. It is based on the idea that a harmonic driving force is exciting the system at its natural frequency, causing an unbounded oscillation in the undamped case and a response with a maximum amplitude in the damped case. However, in multiple-degree-of-freedom systems, there are n natural frequencies, and the concept of resonance is complicated by the effects of mode shapes. Basically, if a force is applied orthogonally to Sec. 4.6 367 Modal Analysis of the Forced Response the mode of the exciting frequency, the system will not resonate at any frequency, a fact that can be used in design. The following example illustrates resonance in a two-degree-of-freedom system. Example 4.6.2 Consider the following system and determine if the driving frequency will cause the system to experience resonance. J 4 0 0 $ 30 Rx + J 9 -5 -5 1 R x(t) = J R sin (2.757t) 5 0 If so, which mode experiences resonance? Does this cause both degrees of freedom to experience resonance? Solution First, compute the mass-normalized stiffness matrix and then the eigenvalue problem for this system 1 ∙ K = M -1>2KM -1>2 = J 2 0 0 1 3 RJ 30 -5 1 -5 R J2 0 5 0 1 3 R = J 7.5 - 0.833 - 0.833 R 0.556 Solving the eigenvalue problem for this matrix yields λ1 = 0.456956 and λ2 = 7.5986 so that ω1 = 0.676 rad>s and ω2 = 2.757 rad>s. Note that the second frequency is within round off to the driving frequency so that this is a resonant system. Next, compute the modal equations. From equation (4.129), the modal force vector is computed from PTM -1>2b = J 0.118 0.993 1 0.993 R J2 - 0.118 0 0 1 3 1 0.059 RJ R = J R 0 0.497 Thus the modal equations are $ r1(t) + (0.676)2r1(t) = sin (2.57t) $ r2(t) + (2.57)2r2(t) = sin (2.57t) Thus the second mode is clearly in resonance. Note however that once transformed back to physical coordinates, each mass will be affected by both modes. That is, both x1(t) and x2(t) are a linear combination of r1(t) and r2(t). Thus each mass will experience resonance. This is because the transformation back to physical coordinates couples the modal solutions. n Forced Response via Mode Summation First, compute the particular solution of the forced response. Consider $ Mx(t) + Kx(t) = F(t) (4.132) Here F(t) is a general force input. Let xp denote the particular solution computed for a given force input. Next, consider the free response using mode summation. First 368 Multiple-Degree-of-Freedom Systems Chap. 4 consider the transformation of coordinates x(t) = M -1>2q(t) substituted into equation (4.132), and then premultiplying (4.132) by M -1>2 yields ∙ $ (4.133) q(t) + Kq(t) = M -1>2F(t) ∙ where K = M -1>2KM -1>2 as before. From equation (4.92) the homogeneous solution in mode summation form is q H(t) = a d i sin (ωit + ϕi)vi n (4.134) i=1 ∙ where vi are the eigenvectors of the symmetric matrix K. Rewriting this last expression in orthogonal form (i.e., sine plus cosine instead of magnitude and phase) and adding in the particular solution yields the total solution: q(t) = a 3bi sin ωit + ci cos ωit4 vi + q p(t) n i=1 5 homogeneous (4.135) " particular Now it remains to find an expression for qp and to evaluate the constants of integration bi and ci in terms of the given initial conditions. From the coordinate transformation x(t) = M -1>2q(t), q p is related to xp by q p(t) = M1>2xp(t). Thus equation (4.135) becomes q(t) = a ( bi sin ωit + ci cos ωit ) vi + M1>2xp(t) n (4.136) i=1 The initial conditions can now be used to compute the constants of integration. Setting t = 0 in equation (4.136) yields q(0) = q 0 = a ( bi sin ωi 0 + ci cos ωi 0 ) vi + M1>2xp(0) n (4.137) i=1 Premultiplying this last expression by vTi yields vTi q 0 = ci + viTM1>2xp(0) (4.138) Likewise, differentiating equation (4.136), setting t = 0, and multiplying by vTi yields # # (4.139) vTi q 0 = ωibi + viTM1>2xp(0) Solving equations (4.138) and (4.139) for the constants of integration yields ci = viTq 0 - vTi M1>2xp(0) bi = 1 ( vTq# - vTi M1>2x# p(0) ) ωi i 0 (4.140) Sec. 4.7 369 lagrange’s Equations Substitution of these values into equation (4.136) yields the expression for q(t). Premultiplying by M -1>2 then yields the displacement in physical coordinates: x(t) = a ( bi sin ωit + ci cos ωit ) u i + xp(t) n (4.141) i=1 where the constants are given by equation (4.140), ui are the mode shapes, ωi are the natural frequencies, and xp is the particular solution. 4.7 LAGRANGE’S EQUATIONS Lagrange’s equation was introduced in Section 1.4 as a follow-on to the energy method of deriving equations of motion. Equations (1.62) and (1.63) along with Example 1.4.7 introduced the method for single-degree-of-freedom systems. Just as in the single-degree-of-freedom case, the Lagrange formulation can be used to model multiple-degree-of-freedom systems as an alternative to using Newton’s law (summing forces and moments) for those cases where the free-body diagram is not as obvious. Recall that the Lagrange formulation requires identification of the energy in the system, rather than the identification of forces and moments acting on the system, and requires the use of generalized coordinates. A brief working account of the Lagrange formulation is given here. A more precise and detailed account is given in Meirovitch (1995), for instance. The procedure begins by assigning a generalized coordinate to each moving part. The standard rectangular coordinate system is an example of a generalized coordinate, but any length, angle, or other coordinate that uniquely defines the position of the part at any time forms a generalized coordinate. It is usually desirable to choose coordinates that are independent. It is customary to designate each coordinate by the letter q with a subscript so that a set of n generalized coordinates is written as q1, q2, c, qn. Note that we have run out of symbols and that the qi used here are different than the qi used to denote mass-normalized coordinates in previous sections. An example of generalized coordinates is illustrated in Figure 4.17. In the figure, the location of the two masses can be described by the set of four coordinates x l1 y 1 m1 l2 (x1, y1) m2 (x2, y2) 2 Figure 4.17 An example of generalized coordinates for a double pendulum illustrating an example of constraints. 370 Multiple-Degree-of-Freedom Systems Chap. 4 x1, y1, x2, and y2 or the two coordinates θ1 and θ2. The coordinates θ1 and θ2 are taken to be generalized coordinates because they are independent. The Cartesian coordinates (x1, x2, y1, y2) are not independent and hence would not make a desirable choice of generalized coordinates. Note that x21 + y21 = l 21 and (x2 - x1)2 + (y2 - y1)2 = l 22 (4.142) express the dependence of the Cartesian coordinates on each other. The relationships in equation (4.142) are called equations of constraint. A new configuration of the double pendulum of Figure 4.17 can be obtained by changing the generalized coordinates q1 = θ1 and q2 = θ2 by an amount δq1 and δq2, respectively. Here δqi are referred to as virtual displacements, which are defined to be infinitesimal displacements that do not violate constraints and such that there is no significant change in the system’s geometry. The virtual work, denoted by δW, is the work done in causing the virtual displacement. The principle of virtual work states that if a system at rest (or at equilibrium) under the action of a set of forces is given a virtual displacement, the virtual work done by the forces is zero. The generalized force (or moment) at the ith coordinate, denoted by Qi, is related to the work done in changing qi by the amount δqi and is defined to be Qi = δW δqi (4.143) The quantity Qi will be a moment if qi is a rotational coordinate and a force if it is a translational coordinate. The Lagrange formulation follows from variational principles and states that the equations of motion of a vibrating system can be derived from d 0T 0T 0U a # b + = Qi dt 0qi 0qi 0qi i = 1, 2, c, n (4.144) d 0T 0T 0U a # b + = 0 dt 0qi 0qi 0qi i = 1, 2, c, n (4.145) # where qi = 0qi >0t is the generalized velocity, T is the kinetic energy of the system, U is the potential energy of the system, and Qi represents all the nonconservative forces corresponding to qi. Here 0 > 0qi denotes the partial derivative with respect to the coordinate qi. For conservative systems, Qi = 0 and equation (4.144) becomes Equations (4.144) and (4.145) represent one equation for each generalized coordinate. These expressions allow the equations of motion of complicated systems to be derived without using free-body diagrams and summing forces and moments. The Lagrange equation can be rewritten in a slightly simplified form by defining the Sec. 4.7 371 lagrange’s Equations Lagrangian, L, to be L = (T – U), the difference between the kinetic and potential # energies. Then if 0U>qi = 0, the Lagrange equation becomes [this is equation (1.63)] d 0L 0L a # b = 0 dt 0qi 0qi i = 1, 2, c, n (4.146) The following examples illustrate the procedure. Example 4.7.1 Derive the equations of motion of the system of Figure 4.18 using the Lagrange equation. x(t) k1 (t) k2 m r J Frictionless surface Frictionless shaft M(t) Figure 4.18 A vibration model of a simple machine part. The quantity M(t) denotes an applied moment. The disk rotates without translation. Solution The motion of this system can be described by the two coordinates x and θ, so a good choice of generalized coordinates is q1(t) = x(t) and q2(t) = θ(t). The kinetic energy becomes T = 1 #2 1 # mq1 + Jq22 2 2 The potential energy becomes U = 1 1 k q2 + k2(rq2 - q1)2 2 1 1 2 Here Q1 = 0 and Q2 = M(t). Using equation (4.145) yields, for i = 1, d # (mq1 + 0) - 0 + k1q1 + k2(rq2 - q1)( - 1) = 0 dt or $ mq1 + (k1 + k2)q1 - k2rq2 = 0 Similarly, for i = 2, equation (4.144) yields $ Jq2 + k2r 2q2 - k2rq1 = M(t) (4.147) (4.148) 372 Multiple-Degree-of-Freedom Systems Chap. 4 Combining equations (4.147) and (4.148) into matrix form yields J 0 $ k1 + k2 R x(t) + J - rk2 J m 0 - rk2 0 R x(t) = J R r 2k2 M(t) (4.149) Here the vector x(t) is x(t) = J q1(t) x(t) R = J R θ(t) q2(t) n Example 4.7.2 A machine part consists of three levers connected by lightweight linkages. A vibration model of this part is given in Figure 4.19. Use the Lagrange method to obtain the equation of vibration. Take the angles to be the generalized coordinates. Linearize the result and put it in matrix form. k1 m1 k2 1 Solution m2 2 m3 3 Figure 4.19 A vibration model of three coupled levers. The lengths of the levers are l, and the springs are attached at α units from the pivot points. The kinetic energy is T = # # # 1 1 1 m l 2θ2 + m2l 2θ22 + m3l 2θ23 2 1 1 2 2 The potential energy becomes U = m1gl(1 - cos θ1) + m2gl(1 - cos θ2) + m3gl(1 - cos θ3) + 1 1 k (αθ - αθ1)2 + k2(αθ3 - αθ2)2 2 1 2 2 Applying the Lagrange equation for i = 1 yields $ m1l 2θ1 + m1gl sin θ1 - αk1(αθ2 - αθ1) = 0 For i = 2, the Lagrange equation yields $ m2l 2θ2 + m2gl sin θ2 + αk1(αθ2 - αθ1) - αk2(αθ3 - αθ2) = 0 and for i = 3, the Lagrange equation becomes $ m3l 2θ3 + m3gl sin θ3 + αk2(αθ3 - αθ2) = 0 Sec. 4.7 373 lagrange’s Equations These three equations can be linearized by assuming θ is small so that sin θ ∼ θ. Note that this linearization occurs after the equations have been derived. In matrix form this becomes m1l 2 C 0 0 0 m2l 2 0 0 $ 0 S x(t) m3l 2 m1gl + α2k1 + C - α2k1 0 where x(t) = [q1 q2 q3]T = [θ1 -α2k1 m2gl + α2(k1 + k2) -α2k2 θ2 0 - α2k2 S x(t) = 0 m3gl + α2k2 θ3]T is the generalized set of coordinates. n Example 4.7.3 Consider the wing vibration model of Figure 4.20. Using the vertical motion of the point of attachment of the springs, x(t), and the rotation of this point, θ(t), determine the equations of motion using Lagrange’s method. Use the small-angle approximation (recall the pendulum of Example 1.4.6) and write the equations in matrix form. Note that G denotes the center of mass and e denotes the distance between the point of rotation and the center of mass. Ignore the gravitational force. k1 k2 G (t) e x(t) (a) (b) Figure 4.20 An airplane in flight (a) presents a number of different vibration models, one of which is given in part (b). In (b) a vibration model of a wing in flight is sketched which accounts for bending and torsional motion by modeling the wing as attached to ground (the aircraft body in this case) through a linear spring k1 and a torsional spring k2. Solution Let m denote the mass of the wing section and J denote the rotational inertia about point G. The kinetic energy is T = 1 #2 1 # mxG + Jθ2 2 2 374 Multiple-Degree-of-Freedom Systems Chap. 4 where xG is the displacement of the point G. This displacement is related to the coordinate, x(t), of the point of attachment of the springs by xG(t) = x(t) - e sin θ(t) # which is obtained by examining the geometry of Figure 4.20. Thus xG(t) becomes # dθ # # # xG(t) = x(t) - e cos θ(t) = x(t) - eθ cos θ dt The expression for kinetic energy in terms of the generalized coordinates q1 = x and q2 = θ then becomes T = # 1 1 # # m[x - eθ cos θ]2 + Jθ2 2 2 The expression for potential energy is U = 1 1 k x2 + k2θ2 2 1 2 which is already in terms of the generalized coordinates. The Lagrangian, L, becomes L = T - U = # 1 1 # 1 1 # m[x - eθ cos θ]2 + Jθ2 - k1x2 - k2θ2 2 2 2 2 Calculating the derivatives required by equation (4.146) for i = 1 yields # 0L 0L # # = # = m[x - eθ cos θ] 0q1 0x $ # d 0L $ a # b = mx - meθ cos θ + meθ2 sin θ dt 0x 0L 0L = = -k1x 0q1 0x so that equation (4.146) becomes $ # $ mx - meθ cos θ + emθ2 sin θ + k1x = 0 Assuming small motions# so that the approximations cos θ S 1, and sin θ S θ hold, and assuming that the term θ2θ is small enough to ignore, results in a linear equation in x(t) given by $ $ mx - meθ + k1x = 0 Sec. 4.7 375 lagrange’s Equations Calculating the derivatives of the Lagrangian required by equation (4.146) for i = 2 yields # # # # 0L 0L # # # = # = m[x - eθ cos θ]( - e cos θ) + Jθ = -me cos θx + me 2θ cos 2θ + Jθ 0q2 0θ d 0L d 0L a # b = a # b dt 0q2 dt 0θ # $ # $ $ # = -me cos θx + mex sin θθ + me 2θ cos 2θ - 2me 2θ2 sin θ cos θ + Jθ # # 0L 0L # = = m[x - eθ cos θ](eθ sin θ) - k2θ 0q2 0θ # ## = mexθ sin θ - me 2θ2 sin θ cos θ - k2θ so that equation (4.146) becomes $ $ # $ Jθ - me cos θx + me 2 cos 2 θθ - me 2θ2 sin θ cos θ + k2θ = 0 Again if the small-angle, small-motion approximation (i.e., sin θ S θ, cos θ S 1, # θ2θ S 0) is used, a linear equation in θ(t) results given by $ $ (J + me 2)θ - mex + k2θ = 0 Combining the expression for i = 1 and i = 2 into one vector equation in the generalized vector x = [q1(t) q2(t)]T = [x(t) θ(t)]T yields J m - me $ - me x(t) k $ R J R + J 1 me 2 + J θ(t) 0 0 x(t) 0 RJ R = J R k2 θ(t) 0 Note here that the two equations of motion are coupled, not through stiffness terms, as in Example 4.7.2, but rather through the inertia terms. Such systems are called dynamically coupled, meaning that the terms that couple the equation in θ(t) to the equation in x(t) are in the mass matrix (i.e., meaning that the mass matrix is not diagonal). In all previous examples, the mass matrix is diagonal and the stiffness matrix is not diagonal. Such systems are called statically coupled. Dynamically coupled systems have nondiagonal mass matrices, and hence require the use of the Cholesky decomposition for factoring the mass matrix (M = LTL) in the modal analysis steps of Window 4.5 (replacing M - 1>2 with L). This is discussed in Section 4.9. The programs in the Toolbox and the various codes given in Section 4.9 are capable of solving dynamically coupled systems as easily as those that have a diagonal mass matrix. n Example 4.7.3 not only illustrates a dynamically coupled system but also presents a system that is easier to approach using Lagrange’s method than by using a force balance to obtain the equation of motion. Several vibration texts have reported an incorrect set of equations of motion for the problem of the preceding example by using the sum of forces and moments rather than taking a Lagrangian approach. 376 Multiple-Degree-of-Freedom Systems Chap. 4 Damping Viscous damping is a nonconservative force and may be modeled by defining the Rayleigh dissipation function. This function assumes that the damping forces are proportional to the velocities. The Rayleigh dissipation function then takes the form (recall Problem 1.79) F = 1 n n # # crs qr qs a a 2 r=1s=1 (4.150) Here the damping coefficients crs = csr and n is again the number of generalized coordinates. With this form, the generalized forces for viscous damping can be derived from 0F Qj = - # , for each j = 1, 2, c, n 0qj (4.151) Then to derive equations of motion with viscous damping, substitute equation (4.150) into (4.151) and (4.151) into equation (4.144). The following example illustrates the procedure. Example 4.7.4 Consider again the system of Example 4.7.1 and assume that there is a viscous damper of coefficient c1, parallel to k1, and a damper of coefficient c2, parallel to k2. Derive the equations of motion for the system using Lagrange’s equations. Solution The dissipation function given by equation (4.150) becomes F = 1 # # # [c q2 + c2(rq2 - q1)2] 2 1 1 Substitution into equation (4.151) yields the generalized forces 0F # # # # # Q1 = - # = -c1q1 - c2(rq2 - q1)( - 1) = -(c1 + c2)q1 + c2rq2 0q1 0F # # # # Q2 = - # = -c2(rq2 - q1)(r) = -c2r 2q2 + rc2q1 0q2 Adding the moment as indicated in Example 4.7.1, the second generalized force becomes # # Q2 = M(t) - c2r 2q2 + rc2q1 Next, using T and U as given in Example 4.7.1, recalculate the equations of motion using equation (4.144) to get, for i = 1: $ # # mq1 + (k1 + k2)q1 - k2rq2 = Q1 = - (c1 + c2)q1 + c2rq2 or: $ # # mq1 + (c1 + c2)q1 - c2rq2 + (k1 + k2)q1 - k2q2 = 0 Sec. 4.8 377 Examples and for i = 2: $ # # J q2 + k2r 2q2 - k2rq1 = Q2 = M(t) - c2r 2q2 + rc2q1 or: $ # # Jq2 + c2r 2q2 - rc2q1 + k2r 2q2 - k2rq1 = M(t) Combining the expressions for i = 1 and i = 2 yields the matrix form of the equations of motion: J m 0 0 $ c + c2 R x(t) + J 1 - rc2 J - rc2 # k + k2 R x(t) + J 1 r 2c2 - rk2 - rk2 0 R x(t) = J R r 2k2 M(t) n 4.8 EXAMPLES Several examples of multiple-degree-of-freedom systems, their schematics, and equations of motion are presented in this section. The “art” in vibration analysis and design is often related to choosing an appropriate mathematical model to describe a given structure or machine. The following examples are intended to provide additional “practice” in modeling and analysis. Example 4.8.1 A drive shaft for a belt-driven machine such as a lathe is illustrated in Figure 4.21(a). The vibration model of this system is indicated in Figure 4.21(b), along with a free-body diagram of the machine. Write the equations of motion in matrix form and solve for the case J1 = J2 = J3 = 10 kg m2 >rad, k1 = k2 = 103 N # m>rad, c = 2 N # m # s>rad for zero initial conditions, and where the applied moment M(t) is a unit impulse function. Solution In Figure 4.21(a) the bearings and shaft lubricant are modeled as lumped viscous damping, and the shafts are modeled as torsional springs. The pulley and machine disks are modeled as rotational inertias. The motor is modeled simply as supplying a moment to the pulley. Figure 4.21(b) illustrates a free-body diagram for each of the three disks, where the damping is assumed to act in proportion to the relative motion of the masses and of the same value at each coordinate (other damping models may be more appropriate, but this choice yields an easy form to solve). Examining the free-body diagram of Figure 4.21(b) and summing the moments on each of the disks yields $ # # J1θ1 = k1(θ2 - θ1) + c(θ2 - θ1) $ # # # # J2θ2 = k2(θ3 - θ2) + c(θ3 - θ2) - k1(θ2 - θ1) - c(θ2 - θ1) $ # # J3θ3 = - k2(θ3 - θ2) - c(θ3 - θ2) + M(t) 378 Multiple-Degree-of-Freedom Systems 1 Chap. 4 3 2 M k1 J1 k2 J2 J3 Pulley Belt Bearings Bearings Motor (a) 1 2 J1 3 J3 J2 k1(2  1) • • c(2  1) M k2(3  2) • • c(3  2) k1(2  1) • • c(2  1) (b) k2(3  2) • • c(3  2) Figure 4.21 (a) Schematic of the moving parts of a lathe. The bearings that support the rotating shaft are modeled as providing viscous damping while the shafts provide stiffness and the belt drive provides an applied torque. (b) Freebody diagrams of the three inertias in the rotating system of part (a). The shafts are modeled as providing stiffness, or as rotational springs, and the bearings are modeled as rotational dampers. where θ1, θ2, and θ3 are the rotational coordinates as indicated in Figure 4.21. The unit for θ is radians. Rearranging these equations yields # $ # J1θ1 + cθ1 + k1θ1 - cθ2 - k1θ2 = 0 $ # # # J2θ2 + 2cθ2 - cθ1 - cθ3 + (k1 + k2)θ2 - k1θ1 - k2θ3 = 0 $ # # J3θ3 + cθ3 - cθ2 - k2θ2 + k2θ3 = M(t) In matrix form this becomes J1 C0 0 0 J2 0 0 c $ 0 S 𝛉 + C -c J3 0 -c 2c -c 0 k1 # - c S 𝛉 + C - k1 c 0 - k1 k1 + k2 - k2 0 0 - k2 S 𝛉 = C 0 S k2 M(t) Sec. 4.8 379 Examples where 𝛉(t) = [θ1(t) ously, this becomes 10 C0 0 0 10 0 θ2(t) θ3(t)]T. Using the values for the coefficients given previ- 0 1 $ 0 S 𝛉 + 2C - 1 10 0 -1 2 -1 0 1 # - 1 S 𝛉 + 103 C - 1 1 0 0 0 -1 S 𝛉 = C 0 S 1 δ(t) -1 2 -1 Note that the damping matrix is proportional to the stiffness matrix so that modal analysis can be used to calculate the solution. Also note that 1 ∙ C = 0.2C - 1 0 -1 2 -1 0 -1 S 1 1 ∙ K = 102 C - 1 0 -1 2 -1 0 -1 S 1 M -1>2F = 0 1 C 0 S 110 δ(t) ∙ Following the steps of Example 4.6.1, the eigenvalue problem for K yields λ1 = 0 λ2 = 100 λ3 = 300 ∙ Note that one of the eigenvalues is zero, thus the matrix K is singular. The physical ∙ meaning of this is interpreted in this example. The normalized eigenvectors of K yield 0.5774 P = C 0.5774 0.5774 0.7071 0 - 0.7071 0.4082 -0.8165 S 0.4082 0.5774 PT = C 0.7071 0.4082 0.5774 0 - 0.8165 Further computation yields ∙ PT CP = diag[0 0.2 0.6] ∙ PT KP = diag[0 100 300] 0.1826 PTM -1>2F(t) = C -0.2236 S δ(t) 0.1291 The decoupled modal equations are $ r1(t) = 0.1826δ(t) $ # r2(t) + 0.2r2(t) + 100r2(t) = - 0.2236δ(t) $ # r3(t) + 0.6r3(t) + 300r3(t) = 0.1291δ(t) Obviously, ω1 = 0, ω2 = 10 rad>s, and ω3 = 17.3205 rad>s. 0.5774 - 0.7071 S 0.4082 380 Multiple-Degree-of-Freedom Systems Chap. 4 # Comparing coefficients of ri with 2ζi ωi yields the three modal damping ratios ζ1 = 0 ζ2 = 0.2 = 0.01 2(10) ζ3 = 0.6 = 0.01732 2(17.3205) so that the second two modes are underdamped. Hence the two damped natural frequencies become ωd2 = ω2 21 - ζ22 = 9.9995 rad>s ωd3 = ω3 21 - ζ23 = 17.3179 rad>s As was the case in Example 4.6.1, while the moment is applied to only one physical location, it is applied to each of the three modal coordinates. The modal equations for r2 and r3 have solutions given by equation (3.6). The solution corresponding to the zero eigenvalue (ω1 = 0) can be calculated by direct integration or by using the Laplace transform method. Taking the Laplace transform yields s2r(s) = 0.1826 or r(s) = 0.1826 s2 The inverse Laplace transform of this last expression yields (see Appendix B) r1(t) = (0.1826)t Physically, this is interpreted as the unconstrained motion of the shaft (i.e., the shaft rotates or spins continuously through 360°). This is also called the rigid-body mode ∙ (see Example 4.6.2) or zero mode and results from K being singular (i.e., from the zero eigenvalue). Such systems are also called semidefinite, as explained in Appendix C. Following equation (3.6), the solution for r2(t) and r3(t) becomes r2(t) = f2 -ζ ω t e 2 2 sin ωd2t = - 0.0224e -0.1t sin 9.9995t ωd2 r3(t) = f3 -ζ ω t e 3 3 sin ωd3t = 0.0075e -0.2999t sin 17.3179t ωd3 The total solution in physical coordinates is then calculated from θ(t) = M - 1>2Pr(t) = (1> 110)Pr(t) or 0.0333t - 0.0050e -0.1t sin (9.9995t) + 0.0010e -0.2999t sin (17.3179t) 0.0333t - 0.0019e -0.2999t sin (17.3179t) S 𝛉(t) = C 0.0333t - 0.0053e -0.1t sin (9.9995t) + 0.0010e -0.2999t sin (17.3179t) 381 Examples The three solutions θ1(t), θ2(t), and θ3(t) are plotted in Figure 4.22. Figure 4.23 plots the three solutions without the rigid-body term. This represents the vibrations experienced by each disk as it rotates. 1(t) 2(t) 0.16 30  103 0.12 0.08 0.04 0.00 t 0 1 2 1 2 3 4 20 10 0 0.0 5 0.2 0.4 0.6 0.8 t 1.0 3(t) 0.16 0.12 0.08 0.04 0.00 t 0 3 4 5 Figure 4.22 The response of each of the disks of Figure 4.21 to an impulse at θ3, illustrating the effects of a rigid-body rotation. 1(t) 2(t) 1.5 1.0 2 103 103 4 0 –2 –4 t 0 2 4 6 8 10 2 4 6 8 10 0.5 0.0 –0.5 0 2 4 6 8 t 10 3(t) 4 103 Sec. 4.8 0 –4 t 0 Figure 4.23 The response of each of the disks of Figure 4.21 to an impulse at θ3 without the rigid-body mode, illustrating the vibration that occurs in each disk. n 382 Multiple-Degree-of-Freedom Systems Chap. 4 Example 4.8.2 In Figure 2.17 a vehicle is modeled as a single-degree-of-freedom system. In this example, a two-degree-of-freedom model is used for a vehicle that allows for bounce and pitch motion. This model can be determined from the schematic of Figure 4.24. Determine the equations of motion, solve them by modal analysis, and determine the response to the engine being shut off, which is modeled as an impulse moment applied to θ(t) of 103 Nm. Bounce Pitch Good Humor c.g. k1 k2 c1 c2 l1 l2 Figure 4.24 A sketch of the side section of a vehicle used to suggest a vibration model for examining its angular (pitch) and up-and-down (bounce) motion. The center of gravity is denoted by c.g. Solution The sketch of the vehicle of Figure 4.24 can be simplified by modeling the entire mass of the system as concentrated at the center of gravity (c.g.). The tire-andwheel assembly is approximated as a simple spring–dashpot arrangement as illustrated in Figure 4.25. The rotation of the vehicle in the x–y plane is described by the angle θ(t), and the up-and-down motion is modeled by x(t). The angle θ(t) is taken to be positive in the clockwise direction, and the vertical displacement is taken as positive in the downward direction. Rigid translation in the y direction is ignored for the sake of concentrating on the vibration characteristics of the vehicle (e.g., Example 4.8.1 illustrates the concept of ignoring rigid-body motion). Summing the forces in the x direction yields # # $ # # mx = - c1(x - l1θ) - c2(x + l2θ) - k1(x - l1θ) - k2(x + l2θ) (4.152) y l1 x  l1  l2 x c.g. x  l2 k1 c1 x(t) k2 c2 Figure 4.25 The vehicle of Figure 4.24 modeled as having all of its mass at its c.g. and two-degrees-of-freedom, consisting of the pitch, θ(t), about the c.g. and a translation x(t) of the c.g. Sec. 4.8 383 Examples since the spring k1 experiences a displacement x – l1θ and k2 experiences a# displace# ment x + l2θ. # Similarly, the velocity experienced by the damper c1 is x - l1θ and that # of c2 is x + l2θ. Taking moments about the center of gravity yields $ # # # # Jθ = c1l1(x - l1θ) - c2l2(x + l2θ) + k1l1(x - l1θ) - k2l2(x + l2θ) (4.153) where J = mr2. Here r is the radius of gyration of the vehicle (recall Example 1.4.6). Equations (4.152) and (4.153) can be rewritten as # $ # mx + (c1 + c2)x + (l2c2 - l1c1)θ + (k1 + k2)x + (l2k2 - l1k1)θ = 0 $ # # mr 2θ + (c2l2 - c1l1)x + (l 22c2 + l 21c1)θ + (k2l2 - k1l1)x + (l 21k1 + l 22k2)θ = 0 (4.154) In matrix form, these two coupled equations become mJ 1 0 0 $ c + c2 Rx + J 1 r2 l2c2 - l1c1 l2c2 - l1c1 # k + k2 Rx + J 1 l 22c2 + l 21c1 k2l2 - k1l1 k2l2 - k1l1 Rx = 0 l 21k1 + l 22k2 (4.155) where the vector x is defined by x = J x(t) R θ(t) Reasonable values for a truck are r 2 = 0.64 m2 m = 4000 kg c1 = c2 = 2000 N # s>m k1 = k2 = 20,000 N>m l1 = 0.9 m l2 = 1.4 m With these values, equation (4.155) becomes J 4000 0 0 4000 $ Rx + J 2560 1000 1000 # 40,000 Rx + J 5540 10,000 10,000 0 Rx = J R 55,400 0 (4.156) Note that C = (0.1)K, so that the damping is proportional. If a moment M(t) is applied to the angular coordinate θ(t) the equations of motion become J 4000 0 0 4000 $ Rx + J 2560 1000 1000 # 40,000 Rx + J 5540 10,000 10,000 0 Rx = J R 103 55,400 δ(t) Following the usual procedures of modal analysis, calculation of M - 1>2 yields M -1>2 = J 0.0158 0 0 R 0.0198 and 10.000 ∙ K = J 3.1250 Thus 1.0000 ∙ C = J 0.3125 0.3125 R 2.1641 3.1250 R 21.6406 384 Multiple-Degree-of-Freedom Systems Chap. 4 ∙ Solving the eigenvalue problem for K yields P = J 0.9698 - 0.2439 0.2439 R 0.9698 and PT = J 0.9698 0.2439 - 0.2439 R 0.9698 with eigenvalues λ1 = 9.2141 and λ2 = 22.4265, so that the natural frequencies are ω1 = 3.0355 rad>s and ω2 = 4.7357 rad>s Thus ∙ PT KP = diag [9.2141 ∙ 22.4265] and PTCP = diag [0.9214 2.2426] ∙ Comparing the elements of PTCP to ω1 and ω2 yields the modal damping ratios ζ1 = 0.9214 = 0.1518 2(3.0355) and ζ2 = 2.2426 = 0.2369 2(4.7357) Using the formula from Window 4.8 for damped natural frequencies yields ωd1 = 3.0003 rad>s and ωd2 = 4.6009 rad>s. The modal forces are calculated from PTM -1>2 J 0 15.3 R 103 = J δ(t) 3.9 - 4.8 δ 0 -4.8 δ(t) RJ R = J R 19.2 δ(t) 19.2 δ(t) The decoupled modal equations become $ # r1(t) + (0.9214)r1(t) + (9.2141)r1(t) = -4.8δ(t) $ # r2(t) + (2.2436)r2(t) + (22.4265)r2(t) = 19.2δ(t) From equation (3.6) these have solutions r1(t) = -4.8 -ζ1ω1t 1 e -(0.1518)(3.035)t sin (3.0003t) e sin ωd1t = mωd1 (1)(3.0003) = - 1.6066e -0.4607t sin (3.0003t) r2(t) = 19.2 -(0.2369)(4.7357)t e sin (4.6009t) 4.6009 = 4.1659e -1.1219t sin (4.6009t) The solution in physical coordinates is obtained from J x(t) R = M -1>2Pr(t) θ(t) which yields x(t) = - 2.41 * 10-2e -0.4607t sin (3.0003t) + 1.606 * 10-2e -1.1213t sin (4.6009t) θ(t) = 7.744 * 10-4e -0.4607t sin (3.0003t) + 7.915 * 10-2e -1.1213t sin (4.6009t) 385 Examples These coordinates are plotted in Figure 4.26. 7  102 (t) 6 5 x(t) in meters and (t) in radians Sec. 4.8 4 3 2 x(t) 1 0 1 2 3 0 0.5 1 1.5 2 2.5 Time (s) 3 3.5 4 4.5 5 Figure 4.26 Plot of the bounce and pitch vibrations of the vehicle of Figure 4.24 as the result of the engine being shut off. n Example 4.8.3 The punch press of Figure 4.27 can be modeled for vibration analysis in the x direction as indicated by the three-degree-of-freedom system of Figure 4.28. Discuss the solution for the response due to an impact at m1 using modal analysis. Solution The mass and stiffness of the various components can be easily approximated using the static methods suggested in Chapter 1. However, it is very difficult to estimate values for the damping coefficients. Hence, an educated guess is made for the modal damping ratios. Such guesses are often made based on experience or from measurements such as the logarithmic decrement. In this case, the values of various masses and stiffness coefficients are [in mks units and f(t) = 1000δ(t)] m1 = 400 kg k1 = 300,000 N>m m2 = 2000 kg k2 = 80,000 N>m m3 = 8000 kg k3 = 800,000 N>m 386 Multiple-Degree-of-Freedom Systems Chap. 4 x z Tool y Tool base Isolation mounting pad Platform Rubber mounting Floor Figure 4.27 A schematic of a punch-press machine. Foundation x1 m1 k1 Tool-base mass c? Mounting pad x2 m2 k2 Platform mass c? Rubber mounting x3 m3 Floor k3 c? Figure 4.28 A vibration model of the punch press of Figure 4.27. From free-body diagrams of each mass, the summing of forces in the x direction yields the three coupled equations $ m1x1 = - k1(x1 - x2) + f(t) $ m2x2 = k1(x1 - x2) - k2(x2 - x3) $ m3x3 = - k3x3 + k2(x2 - x3) Rewriting this set of coupled equations in matrix form yields m1 C 0 0 0 m2 0 0 k1 $ 0 S x + C -k1 m3 0 - k1 k1 + k2 - k2 0 f(t) -k2 S x = C 0 S k2 + k3 0 Sec. 4.8 387 Examples where x = [x1(t) x2(t) 0.4 0 x3(t)]T. Substituting the numerical values for mi and ki yields 0 30 $ ( 103 ) C 0 2 0 S x + ( 104 ) C -30 0 0 8 0 - 30 38 -8 0 1000δ(t) -8 S x = C 0 S 88 0 Following the modal analysis procedure for an undamped system yields 20 M1>2 = C 0 0 0 44.7214 0 0 0 S 89.4427 M - 1/2 = C 0.0500 0 0 0 0.0224 0 0 0 S 0.0112 and 750 ∙ K = C -335.4102 0 -335.4102 190 - 20 0 -20 S 110 ∙ Solving the eigenvalue problem for K yields - 0.4116 P = C - 0.8848 - 0.2185 -0.1021 -0.1935 0.9758 0.9056 -0.4239 S 0.0106 -0.4116 PT = C -0.1021 0.9056 -0.8848 -0.1935 - 0.4239 -0.2185 0.9758 S 0.0106 and λ1 = 29.0223 ω1 = 5.3872 λ2 = 113.9665 ω2 = 10.6755 λ3 = 907.0112 ω3 = 30.1166 The modal force vector becomes PTM -1>2 C 1000δ(t) - 20.5805 0 S = C - 5.1026 S δ(t) 0 45.2814 Hence, the undamped modal equations are $ r1(t) + 29.0223 r1(t) = - 20.5805δ(t) $ r2(t) + 113.9665 r2(t) = - 5.1026δ(t) $ r3(t) + 907.0112 r3(t) = 45.2814δ(t) To model the damping, note that each mode shape is dominated by one element. From examining the first column of the matrix P, the second element is larger than the other two elements. Hence if the system were vibrating only in the first mode, the motion of x2(t) would dominate. This element corresponds to the platform mass, which receives high damping from the rubber support. Hence it is given a large damping ratio of ζ1 = 0.1 (rubber provides a lot of damping). Similarly, the second mode is dominated by its third element, corresponding to the motion of x3(t). This is a predominantly metal part, so it is given a low damping ratio of ζ2 = 0.01. The third 388 Multiple-Degree-of-Freedom Systems Chap. 4 mode shape is dominated by the first element, which corresponds to the mounting pad. Hence it is given a medium damping ratio of ζ3 = 0.05. Recalling that the velocity coefficient in modal coordinates has the form 2ζiωi, the damped modal coordinates become 2ζ1ω1 = 2(0.1)(5.3872), 2ζ2ω2 = 2(0.01)(10.6755), and 2ζ3ω3 = 2(0.05) (30.1166). Therefore, the damped modal equations become $ # r1(t) + 1.0774 r1(t) + 29.0223 r1(t) = - 20.5805δ(t) $ # r2(t) + 0.2135 r2(t) + 113.9665 r2(t) = - 5.1026δ(t) $ # r3(t) + 3.0117 r3(t) + 907.0112 r3(t) = 45.2814δ(t) These have solutions given by equation (3.6) as r1(t) = - 3.8395e -0.5387t sin (5.3602t) r2(t) = - 0.4780e -0.1068t sin (10.6750t) r3(t) = 1.5054e -1.5058t sin (30.0789t) Using the transformation x(t) = M - 1>2Pr(t) yields x1(t) = 0.0790e -0.5387t sin (5.3602t) + 0.0024e -0.1068t sin (10.6750t) + 0.0682e -1.5058t sin (30.0789t) x2(t) = 0.0760e -0.5387t sin (5.3602t) + 0.0021e -0.1068t sin (10.6750t) - 0.0143e -1.5058t sin (30.0789t) x3(t) = 0.0094e -0.5387t sin (5.3602t) - 0.0052e -0.1068t sin (10.6750t) + 0.0002e -1.5058t sin (30.0789t) These solutions are plotted in Figure 4.29. x1(t) x2(t) 0.10 60  103 0.05 0.00 40 20 0 20 0.05 40 2 0 4 6 8 10 Time 0 2 4 6 Time  103 x3(t) 10 5 0 5 10 0 2 4 6 8 10 Time Figure 4.29 A numerical simulation of the vibration of the punch press of Figures 4.27 and 4.28 as the result of the machine tool impacting the tool base. 8 10 Sec. 4.9 Computational Eigenvalue Problems for Vibration 389 This example illustrates a method of assigning modal damping to an analytical model. This is a somewhat arbitrary procedure that falls in the category of an educated guess. A more sophisticated method is to measure the modal damping. This is discussed in Chapter 7. Note that the floor, x3(t), vibrates much longer than the machine parts do. This is something to consider in designing how and where the machine is mounted to the floor of a building. n 4.9 COMPUTATIONAL EIGENVALUE PROBLEMS FOR VIBRATION This section examines the various approaches to solving eigenvalue problems and how vibration problems may be solved using these eigenvalue problems. This is presented in the context of the three computer codes Mathematica, MATLAB, and Mathcad. The Engineering Vibration Toolbox also contains a variety of ways of calculating mode shapes, natural frequencies, and damping ratios based on M-files created in MATLAB. The books by Datta (1995), Golub and Van Loan (1996), and Meirovitch (1995) should be consulted for more detail. Additional detail can be found in Appendix C. Each of the matrix computations and manipulations made in the previous sections can be obtained easily by standard functions of most mathematical software packages. Hence the tedious solutions of modal analysis and the eigenvalue problem can be automated and used to solve systems with large numbers of degrees of freedom. Here we introduce the various eigenvalue problems and illustrate how to use various software packages to obtain the needed computation. We refer the reader to Appendix C and references there for details on how the algorithms actually work. Consider the undamped-vibration problem of equation (4.11) with n degrees of freedom, repeated here: $ Mx(t) + Kx(t) = 0 (4.157) Here the displacement vector x is n * 1, and the matrices M and K are n * n and symmetric. There are a number of different ways to relate equation (4.157) to the mathematics of eigenvalue problems and these are presented next along with examples and computer steps for solving them. Dynamically Coupled Systems The matrix M is positive definite (see Window 4.9) in most cases and up until Example 4.7.3 was considered to be diagonal, so that factoring M and taking its inverse amounted to scalar arithmetic on the diagonal entries. In Example 4.7.3, the mass matrix was not diagonal, in which case the system is called dynamically coupled, and more sophisticated means are needed to handle the inverse and 390 Multiple-Degree-of-Freedom Systems Chap. 4 Window 4.9 The Definition of Positive Definite A symmetric matrix M (M = MT) is positive definite if, for every nonzero vector x, the scalar xTMx 7 0. In addition, M is positive definite if and only if all of the eigenvalues of M are positive numbers. The matrix A is said to be positive semidefinite if, for every nonzero vector x, the scalar xTAx Ú 0. A matrix is positive semidefinite if and only if all of the eigenvalues of the matrix A are greater than or equal to zero. In particular, A could have one or more zero eigenvalues, as in the case of rigid-body motion. factoring of the matrix M used in vibration analysis. The method of factoring a positive definite matrix is called Cholesky decomposition, which finds a lower triangular matrix L such that M = LLT. Cholesky decomposition can also be formed from two upper triangular matrices, say U, such that M = UTU. Some codes use upper triangular and others use lower triangular, but the results applied to vibration problems are the same. Both calculating the inverse and computing the factors are simple commands in most computational software programs (and calculators). The algorithms in these codes use the most sophisticated methods. However, one way to compute any function of a matrix is by decomposing the matrix using the eigenvalue problem. Let M be a symmetric, positive definite matrix and let f be any function defined for positive numbers. Denote the eigenvalues of the matrix M (not the M, K system, just the matrix M) by μ and let R denote the matrix of normalized eigenvectors of M. Then f(μ1) 0 f(M) = R E f 0 0 f(μ2) f 0 c c f c 0 0 U RT f f(μn) 1>μ1 0 = RE f 0 0 1>μ2 f 0 c c 0 0 U RT f 1>μn In particular M -1 f c (4.158) Sec. 4.9 391 Computational Eigenvalue Problems for Vibration and M -1>2 1> 1μ1 0 = RE f 0 0 1> 1μ2 f 0 c c f c 0 0 U RT f 1> 1μn which provides one method of computing the matrices used in modal analysis. These decompositions can also be used to prove that the inverse and the square root of a symmetric matrix is symmetric. However, the codes use more numerically sophisticated techniques beyond the scope of this book. In fact, it is better not to compute the inverse directly but to use a modified form of Gaussian elimination to compute the inverse (M\I in MATlAb, for instance). Computing the matrix square root is best not done in the dynamically coupled case, but rather the Cholesky factorization should be used. These are illustrated in the following example. Example 4.9.1 Consider the nondiagonal mass matrix 5 M = C2 0 2 4 1 0 1S 3 and compute the inverse and factors. Solution The matrix inverse is computed via the following commands: In MATlAb In Mathematica In Mathcad inv(M) Inverse[M] M-1 (typed M^-1) The Cholesky factors of a matrix are computed via the following commands: In MATlAb In Mathematica In Mathcad Chol(M) or Chol(M,’lower’) CholeskyDecomposition[M] cholesky(M) Using any one of these yields the inverse of the mass matrix: 0.2558 M -1 = C -0.1395 0.0465 -0.1395 0.3488 - 0.1163 0.0465 - 0.1163 S 0.3721 392 Multiple-Degree-of-Freedom Systems Chap. 4 The factors of the mass matrix are 2.23607 L = C 0.89443 0 0 1.78885 0.55902 0 2.23607 0 S and LT = C 0 1.63936 0 0.89443 1.78885 0 0 0.55902 S 1.63936 such that M = LLT. Also note that L-1M(LT)-1 = I, the identity matrix. Note that MATlAb uses an upper triangular Cholesky decomposition. Thus the mass matrix is factored as M = chol(M)'*chol(M) and inv(chol(M)')*M*inv(chol(M)) = I. So when using MATlAb, the following code has to be modified accordingly. Alternately use the command L = chol(M,'lower') which creates the lower triangular Cholesky factor. n In order to perform modal analysis on equation (4.157) for a dynamically coupled system, replace M - 1>2 with the matrix L using Window 4.5 in the following manner: 2. Calculate the mass-normalized stiffness matrix by ∙ K = L-1K(LT)-1 (4.159) and note that this is a symmetric matrix. 5. Compute the matrix of mode shapes S from S = (LT)-1P and S-1 = PTLT (4.160) Using Codes Several examples are given next that illustrate how to use math software to compute the eigenvalues and eigenvectors of a system and then to solve for frequencies and mode shapes. Calculation of the algebraic eigenvalue problem formed the object of intensive study over a 30-year period, resulting in very sophisticated methods. Many of these studies were funded by government agencies and hence are in the public domain. As computer technology advanced, several high-level codes evolved to enable engineers to make eigenvalue calculations simply and accurately. Today almost every code and calculator contains eigensolvers. The Toolbox contains M-files for computing frequencies and mode shapes using MATlAb. The only small difficulty in using math software is that mathematicians always number eigenvalues starting with the largest first and engineers like the frequencies to be numbered with the smallest value first. So in some codes you may want to sort the eigenvalues and eigenvectors accordingly. The following examples illustrate how to use Mathcad, MATlAb, and Mathematica to solve for natural frequencies and mode shapes. Please note that the developers of these codes often update their codes and syntax, so it is wise to check their websites for updates if you have troubles with syntax errors. Sec. 4.9 393 Computational Eigenvalue Problems for Vibration Example 4.9.2 Compute the solution of Example 4.2.6 using math software. Solution First consider Mathcad. This program enters the elements of a matrix by selecting the appropriate size from the matrix pallet. Note that Mathcad starts counting elements of vectors and matrices with 0 rather than 1. The following illustrates the remaining steps: Enter the values of M and K: M := J 1 0 0 R 4 K := J 12 -2 -2 R 12 Compute the root of M, then K tilde: Ms := J 1 0 0 R 2 Kt := Ms-1 # K # Ms-1 Kt = J 12 -1 -1 R 3 Compute the eingenvalues: λ := eingenvals(Kt) λ = J 12.109772 R 2.890228 Compute the eigenvectors and reorder with the lowest first and compute frequencies: v2 := eigenvec(Kt,λ0) v1 := eigenvec(Kt,λ1) ω1 := 1λ1 ω2 := 1λ0 Display the results: v1 = J 0.109117 R 0.994029 v2 = J 0 v1 0 = 1 - 0.994029 R 0.109117 0 v2 0 = 1 ω1 = 1.7 ω2 = 3.48 Check to see the eigenvectors are orthonomal: Form the matrix P: P := augment (v1,v2) P = J 0.1091 0.994 v1 # v2 = 0 - 0.994 R 0.1091 PT = J 0.1090.994 R - 0.9940.109 Show that P is orthoganal and diagonalizes K tilde PT # P = J 1 0 0 R 1 PT # Kt # P = J 2.8902 0 0 R 12.1098 Next consider using MATlAb. MATlAb enters matrices using spaces between elements of a row and semicolons to start a new row. like Mathcad, MATlAb produces the eigenvalues from highest to lowest, so care must be taken to mind the order of eigenvalues and eigenvectors. This is handled here by using the fliplr(V) 394 Multiple-Degree-of-Freedom Systems Chap. 4 command, which is used to reorder the eigenvectors with the one corresponding to λ1 first rather than λn, as produced from the eigensolver in the code. % enter M and K, compute K tilde M=[1 0;0 4];K=[12 -2;-2 12]; Mr=sqrtm(M);Kt=inv(Mr)*K*inv(Mr); % solve eigenvalue problem puts eigenvectors in the % matrix V, eigenvalues in the diagonal matrix D [V,D]=eig(Kt); %check and reorder eigenvalues, smallest first eignvalues=V'*Kt*V eignvalues = 2.8902 -0.0000 0 12.1098 V'*V % check to see that V1 is orthogonal ans = 1.0000 0 0 1.0000 In Mathematica the code is as follows: Input mass and stiffness matrices: In[1] : = M = a K = a 1 0 12 -2 0 b; 4 -2 b; 12 ∙ Calculate inverse square root of mass matrix then find K. In[3]: = Mnegsqrt = MatrixPower[M, -0.5]; Khat = Mnegsqrt.K.Mnegsqrt MatrixForm[Mnegsqrt] MatrixForm[Khat] Out[5]//Matrix Form= a 1 0 0 b 0.5 Out[6]//MatrixForm= a 12 -1 -1 b 3 Calculate eingenvalues and eigenvectors. Note that Mathematica returns eigenvectors in rows, not columns as in Mathcad and MATlAb. In[7]: = {𝛌, v} = Eigensystem[Khat]; MatrixForm[𝛌] MatrixForm[v] Sec. 4.9 Computational Eigenvalue Problems for Vibration 395 Out[8]//MatrixForm= a 12.1098 b 2.89023 Out[9]//MatrixForm= a 0.994029 0.109117 - 0.109117 b 0.994029 Arrange eigenvectors from least to greater and find natural frequencies. Mathematica usually returns eigenvectors normalized to 1. In[10]: = v2 = v[[1, A11]] v1 = v[[2, A11]] 𝛚1 = 1𝛌[[2]] 𝛚2 = 1𝛌[[1]] Out[10]= {0.994029, -0.109117} Out[11]= {0.109117, 0.994029} Out[12]= 1.70007 Out[12]= 3.47991 In[14]: = v1.v1 v2.v2 Chop[v1.v2] Out[14]= 1. Out[15]= 1. Out[16]= 0. Since the eigenvectors are in rows in Mathematica, it is easiest to form PT first, then transpose to get P. In[17]: = PT = {v1, v2} P = Transpose[PT] MatrixForm[PT] MatrixForm[P] Out[19]//MatrixForm= 0.109117 0.994029 a b 0.994029 - 0.109117 Out[20]//MatrixForm= 0.109117 0.994029 a b 0.994029 - 0.109117 Show that P is orthogonal and that P diagonalizes k̃ In[21]: = MatrixForm[Chop[PT.P]] MatrixForm[Chop[PT.Khat.P]] 396 Multiple-Degree-of-Freedom Systems Chap. 4 Out[21]//MatrixForm= a 0 b 1 1 0 Out[22]//MatrixForm= a 2.89023 0 0 b 12.1098 Note that Mathematica produces eigenvalues for the highest to the lowest, so an alternative way to reorder them is to use the Reverse command. For example {vals, vecs} = Eigensystem[Khat] valsr = Reverse[vals] vecsr = Reverse[vecs] Transpose[vecsr] P = Transpose[vecsr] will reorder the eigenvalues and eigenvectors. n Example 4.9.3 Compute the coefficients for the modal equations for the damped, forced-response problem given in Example 4.6.1. Solution The solution in Mathcad is as follows: Enter the matrices M, C, K and the force vector b M := J 9 0 0 R 1 C := J 2.7 - 0.3 - 0.3 R 0.3 K := J 27 -3 -3 R 3 0 b := J R 3 Note that the damping is proportional, so modal analysis may be used C # M-1K - K # M-1 # C = J - 1.78 # 10-15 0 0 R effectively zero 0 L : = cholesky(M) Kt := L-1 # K # (LT)-1 Ct : = L-1 # C # (LT)-1 Kt = J Ct = J 3 -1 0.3 - 0.1 -1 R 3 - 0.1 R 0.3 Sec. 4.9 397 Computational Eigenvalue Problems for Vibration λ : = eigenvals (Kt) v1 : = eigenvec(Kt, λ1) v2 : = eigenvec(Kt, λ0) P = J P : = augment(v1, v2) PT # P = J 1 0 0 R 1 PT # Ct # P = J 0.2 0 0.707 0.707 0 R 0.4 - 0.707 R 0.707 PT # Kt # P = J 2 0 0 R 4 Next compute the modal force amplitudes: bt: = PT # L-1 # b bt = J 2.121 R 2.121 The solution in MATlAb is as follows: % enter M, C, and K M=[9 0;0 1];K=[27 -3;-3 3];C=K/10;b=[0;3]; %compute L and the mass normalized quantities L=chol(M);Kt=inv(L)*K*inv(L');Ct=inv(L)*C*inv(L'); % Compute eigensolution, reorder eigenvectors [V,D]=eig(Kt); P=V P = -0.7071 -0.7071 -0.7071 0.7071 P'*P ans = 1.0000 0 0 1.0000 P'*Kt*P ans = 2.0000 0 0 4.0000 P'*C*P ans = 1.2000 1.2000 1.2000 1.8000 P'*Ct*P ans = 0.2000 0 0 0.4000 bt=P'*inv(L)*b bt = -2.1213 2.1213 Note that in MATlAb, the first eigenvector is the negative of that produced in Mathcad. This is not a problem as both are correct. In the MATlAb version, note that the mass-normalized input vector also has a sign change. When the modal equations are written out and transformed back to physical coordinates, these signs will recombine to 398 Multiple-Degree-of-Freedom Systems Chap. 4 give the same solution as given in Mathcad. Remember that an eigenvector may always be multiplied by a scalar (in this case - 1) without changing its direction, and the value 1 does not change its magnitude. The solution in Mathematica is as follows: In[1]: = M ∙ a 9 0 c ∙ a K ∙ a 0 b; 1 2.7 ∙0.3 27 ∙3 0 b ∙ a b; 3 ∙0.3 b; 0.3 ∙3 b; 3 Note that the damping is proportional, so modal analysis may be used. In[5]: = MatrixForm[Chop[c.Inverse[M].K-K.Inverse[M].c]] Out[5]//MatrixForm= 0 0 a b 0 0 & & Calculation of K and C In[6]: = L = CholeskyDecomposition[M]; Khat = Inverse[L].K.Inverse[Transpose[L]]; Chat = Inverse[L].c.Inverse[Transpose[L]]; MatrixForm[Khat] MatrixForm[Chat] Out[10]//MatrixForm= 3 -1 a b -1 3 Out[11]//MatrixForm= 0.3 - 0.1 a b - 0.1 0.3 Calculate eigenvalues and eigenvectors. Note that Mathematica returns eigenvectors in rows, not columns as in Mathcad and Matlab. For this particular system, the eigenvalues were found in the correct order and the eigenvectors must be normalized to 1. In[11]: = {𝛌,v}=Eigensystem[Khat]; MatrixForm[𝛌] MatrixForm[v] v1 = Normalize[v[[1, All]]]; v2 = Normalize[v[[2, All]]]; N[MatrixForm][{ v1, v2}] Sec. 4.9 Computational Eigenvalue Problems for Vibration 399 Out[12]//MatrixForm= 2 a b 4 Out[13]//MatrixForm= ω2: = 2Re(λ1)2 + Im(λ1)2 Out[16]//MatrixForm= 0.707107 0.707107 a b - 0.707107 0.707107 In[17]: = PT = {v1, v2}; P = Transpose[PT]; In[19]: = MatrixForm[Chop[PT.P]] MatrixForm[Chop[PT.Khat.P]] MatrixForm[Chop[PT.Chat.P]] Out[19]//MatrixForm= 1 0 a b 0 1 Out[20]//MatrixForm= 2 0 a b 0 4 Out[21]//MatrixForm= 0.2 0 a b 0 0.4 Compute modal force amplitudes. ln[22]: = bt = PT.Inverse[L].b; N[MatrixForm[bt]] Out[23]//MatrixForm= 2.12132 a b 2.12132 n Various Eigenvalue Problems There are several ways to relate the vibration problem to the eigenvalue problem. The simplest way is unfortunately the worst in terms of computational effort. This is to use the generalized eigenvalue problem, formed from equation (4.157) by substitution of x = ejωtu, which results in Ku = λMu (4.161) 400 Multiple-Degree-of-Freedom Systems Chap. 4 Here λ = ω2, and the u are the mode shapes. This is the most direct approach but suffers from computational burden (four times as much as using the Cholesky approach for a two-degree-of-freedom example). The solution of equation (4.157) then becomes x(t) = a ci sin (ωit + ϕi)u i n (4.162) i=1 Here ci and ϕi are constants determined by initial conditions. Next multiply equation (4.161) by the matrix M -1. Again, assuming a solution of the form x(t) = ejωtu yields -ω2u + M -1Ku = 0 (4.163) (M -1K)u = λu (4.164) or This is the standard algebraic eigenvalue problem. The matrix M -1K is neither symmetric nor banded. Again there are n eigenvalues λi, which are the squares of the natural frequencies ω2i , and n eigenvectors ui. The solution of equation (4.157), x(t), is again in the form x(t) = a ci sin (ωit + ϕi)u i n (4.165) i=1 where ci and ϕi are constants to be determined by the initial conditions. Thus the eigenvectors ui are also the mode shapes. Since M -1K is not symmetric, the solution of the algebraic eigenvalue problem could yield complex values for the eigenvalues and eigenvectors. However, they are known to be real valued because of the generalized eigenvalue problem formulation, equation (4.161), which has the same eigenvalues and eigenvectors. Next, consider the vibration problem (following Window 4.5) obtained by substitution of the coordinate transformation x(t) = (LT)-1q(t) into equation (4.161) and multiplying by L-1. This yields the form $ ∙ q(t) + Kq(t) = 0 (4.166) ∙ where the matrix K is symmetric but not necessarily sparse or banded unless M is diagonal. The solution of equation (4.166) is obtained by assuming a solution of the form q(t) = ejωtv where v is a nonzero vector of constants. Substituting this form into equation (4.166) yields ∙ -ω2v + Kv = 0 (4.167) or ∙ Kv = λv (4.168) Sec. 4.9 Computational Eigenvalue Problems for Vibration 401 where again λ = ω2. This is the symmetric eigenvalue problem and again results in n eigenvalues λi, which are the squares of the natural frequencies ω2i , and n eigenvectors vi. The solution of equation (4.166) becomes q(t) = a ci sin (ωit + ϕi)vi n (4.169) i=1 where ci and ϕi are again constants of integration. The solution in the original coordinate system x is obtained from this last expression by multiplying by the matrix (LT)-1: x(t) = (LT)-1 q(t) = a ci sin (ωit + ϕi)(LT)-1vi n (4.170) i=1 Hence the mode shapes are the vectors (LT)-1vi, where vi are the eigenvectors of the ∙ symmetric matrix K. Since the eigenvalue problem here is symmetric, it is known that the eigenvalues and eigenvectors are real valued, as are the mode shapes. In addition, the orthogonality of the vi allows easy computation of the modal initial conditions. The numerical advantage here is that the eigenvalue problem is symmetric, so more efficient numerical algorithms can be used to solve it. Computationally this is the most efficient method of the four possible formulations presented here. Again consider the vibration problem defined in equation (4.161), which is $ (4.171) x(t) + M -1Kx(t) = 0 This equation can be transformed to a first-order vector differential equation by defining two new n * 1 vectors, y1(t) and y2(t), by # y1(t) = x(t), y2(t) = x(t) (4.172) Note that y1 is the vector of displacements and y2(t) is a vector of velocities. Differentiating these two vectors yields # # y1(t) = x(t) = y2(t) (4.173) # $ y2(t) = x(t) = -M -1Ky1(t) # $ where the equation for y2(t) has been expanded by solving equation (4.171) for x(t). Equations (4.173) can be recognized as the first-order vector differential equation # y(t) = Ay(t) (4.174) Here A = J 0 -M -1K I R 0 (4.175) is called the state matrix. The 0 denotes an n * n matrix of zeros, I denotes the n * n identity matrix, and the state vector y(t) is defined by the 2n * 1 vector y(t) = J y1(t) x(t) R = J# R y2(t) x(t) (4.176) 402 Multiple-Degree-of-Freedom Systems Chap. 4 The solution of equation (4.174) proceeds by assuming the exponential form y(t) = zeλt, where z is a nonzero vector of constants and λ is a scalar. Substitution into equation (4.174) yields λz = Az or Az = λz z ∙ 0 (4.177) This is again the standard algebraic eigenvalue problem. While the matrix A has many zero elements, it is now a 2n * 2n eigenvalue problem. It can be shown that the 2n eigenvalues λi again corresponds to the n natural frequencies ωi by the relation λi = ωij, where j = 1-1. The extra n eigenvalues are λi = -ωij, so that there are still only n natural frequencies, ωi. The 2n eigenvectors, z of the matrix A, however, are of the form zi = J ui R λiu i (4.178) where ui are the mode shapes of the corresponding vibration problem. The matrix A (see Window 4.10) is not symmetric and the eigenvalues λi and eigenvectors zi would therefore be complex. In fact, the eigenvalues λi in this case are imaginary numbers of the form ωij. Window 4.10 Various Uses of the Symbol A Do not be confused by the matrix A. The symbol A is used to denote any matrix. Here, A is used to denote A = M -1K A = J 0 -M -1K I R -M -1C A = J 0 M -1K I R 0 to name a few. Which matrix A is being discussed should be clear from the context. Damped Systems For large-order systems, computing the eigenvalues using equation (4.177) becomes numerically more difficult because it is of order 2n rather than n. The main advantage of the state-space form is in numerical simulations and in solving the damped multiple-degree-of-freedom vibration problem, which is discussed next. Now consider the damped vibration problem of the form $ # Mx(t) + Cx(t) + Kx(t) = 0 (4.179) Sec. 4.9 Computational Eigenvalue Problems for Vibration 403 where C represents the viscous damping in the system (see Section 4.5) and is assumed only to be symmetric and positive semidefinite. The state-matrix approach and related standard eigenvalue problem of equation (4.177) can also be used to describe the nonconservative vibration problem of equation (4.179). Multiplying equation (4.179) by the matrix M-1 yields $ # x(t) + M -1Cx(t) + M -1Kx(t) = 0 (4.180) Again it is useful to rewrite this expression in a first-order or state-space form by # defining the two n * 1 vectors, y1(t) = x(t) and y2(t) = x(t), as indicated in equation (4.172). Then equation (4.173) becomes # # y1(t) = x(t) = y2(t) # $ # y2(t) = x(t) = -M -1Kx(t) - M -1Cx(t) (4.181) $ where the expression for x(t) in equation (4.181) is taken from equation (4.180) for # the damped system by moving the terms M -1Cx(t) and M -1Kx(t) to the right of the # equal sign. Renaming x(t) = y1(t) and x(t) = y2(t) in equation (4.181) and using matrix notation yields # y (t) 0y1(t) + Iy2(t) 0 # y(t) = J # 1 R = J R = J -1 -1 y1(t) -M Ky1(t) -M Cy2(t) -M - 1K I y1(t) R = Ay(t) -1 R J -M C y2(t) (4.182) where the vector y(t) is defined as the state vector of equation (4.176), and the state matrix A for the damped case is defined as the partitioned form A = J 0 -M -1K I R -M -1C (4.183) The eigenvalue analysis for the system of equation (4.182) proceeds directly as for the undamped state-matrix system of equation (4.176). A solution of equation (4.183) is again assumed of the form y(t) = zeλt and substituted into equation (4.183) to yield the eigenvalue problem of equation (4.177) (i.e., Az = λz). This again defines the standard eigenvalue problem of dimension 2n * 2n. In this case, the solution again yields 2n values λi that may be complex valued. The 2n eigenvectors zi described in equation (4.178) may also be complex valued (if the corresponding λi is complex). This, in turn, causes the physical mode shape ui to be complex valued as well as the free-response vector x(t). Fortunately, there is a rational physical interpretation of the complex eigenvalue, modes and the resulting solution determined by the state-space formulation of the eigenvalue problem given in equation (4.183). The physical time response x(t) is simply taken to be the real part of the first n coordinates of the vector y(t) computed from x(t) = a ciu ie λit 2n i=1 (4.184) 404 Multiple-Degree-of-Freedom Systems Chap. 4 The time response is discussed in more detail in Section 4.10 and was introduced in equation (4.115). The physical interpretation of the complex eigenvalues λi is taken directly from the complex numbers arising from the solution of an underdamped singledegree-of-freedom system given in equations (1.33) and (1.34) of Section 1.3. In particular, the complex eigenvalues λi will appear in complex conjugate pairs in the form λi = -ζiωi - ωi 21 - ζi2j λi + 1 = -ζiωi + ωi 21 - ζi2j (4.185) where j = 1-1, ωi is the undamped natural frequency of the ith mode and ζi is the modal damping ratio associated with the ith mode. The solution of the eigenvalue problem for the state matrix A of equation (4.183) produces a set of complex numbers of the form λi = αi + βij, where Re(λi) = αi and Im(λi) = βi. Comparing these expressions with equations (4.185) yields ωi = 2α2i + β2i = 2Re(λi)2 + Im(λi)2 ζi = -αi 2αi2 + βi2 = -Re(λi) 2Re(λi)2 + Im(λi)2 (4.186) (4.187) which provides a connection to the physical notions of natural frequency and damping ratios for the underdamped case. (See Inman, 2006, for the overdamped and critically damped cases.) The complex-valued mode-shape vectors ui also appear in complex conjugate pairs and are referred to as complex modes. The physical interpretation of a complex mode is as follows: each element describes the relative magnitude and phase of the motion of the degree of freedom associated with that element when the system is excited at that mode only. In the undamped real-mode case, the mode-shape vector is real (recall Section 4.1) and indicates the relative positions of each mass at any given instant of time at a single frequency. The difference between the realmode case and the complex-mode case is that if the mode is complex, the relative position of each mass can also be out of phase by the amount indicated by the complex part of the mode shapes entry (recall that a complex number can be thought of as a magnitude and a phase rather than a real part and an imaginary part). The state-space formulation of the eigenvalue problem for the matrix A given by equation (4.183) is related to the most general linear vibration problem. It also forms the most difficult computational eigenvalue problem of the five problems discussed previously. The following example illustrates how to compute the natural frequencies and damping ratios using the state-matrix approach. Example 4.9.4 Consider the following system and compute the natural frequencies and damping ratios. Note that the system will not uncouple into modal equations and has complex modes. The system is given by J 2 0 0 $ 1 R x(t) + J 1 - 0.5 - 0.5 # 3 R x(t) + J 0.5 -1 -1 R x(t) = 0 1 Sec. 4.9 405 Computational Eigenvalue Problems for Vibration Solution In each of the following codes, the M, C, and K matrices are entered, put into state-space form, and solved. The frequencies and damping ratios are then extracted using equations (4.186) and (4.187). The solution in Mathcad is as follows: M := J 2 0 0 R 1 C := J O := J 0 0 1 0.5 0 R 0 0.5 R 0.5 K := J I := J 1 0 3 -1 -1 R 1 0 R 1 A : = augment(stack(O,-M-1K),stack(I,-M-1 C)) 0 0 A = D - 1.5 1 0 0 0.5 -1 1 0 - 0.5 0.5 0 - 0.417 1 - 0.417 T λ := eigenvals(A) λ = D - 0.083 0.25 - 0.5 - 0.083 ω1 := 2Re(λ3)2 + Im(λ3)2 ω2 := 2Re(λ1)2 + Im(λ1)2 ζ1 := + + - 1.345i 1.345i T 0.705i 0.705i - Re(λ3) ω1 - Re(λ1) ζ2 := ω2 ω1 = 0.71 ω2 = 1.408 ζ1 = 0.117 ζ2 = 0.296 U : = eigenvecs(A) 0.193 + 0.341i - 0.126 - 0.407i U = D - 0.539 + 0.118i 0.6 0.193 - 0.341i - 0.126 + 0.407i - 0.539 - 0.118i 0.6 0.37 + 0.042i 0.725 - 0.061 + 0.257i - 0.06 + 0.512i 0.37 - 0.042i 0.725 T - 0.061 - 0.257i - 0.06 - 0.512i The solution in MATlAb requires defining the real and imaginary parts (also see VTb4_3) and is as follows: % enter data n=2;M=[2 0;0 1];C=[1 -0.5;-0.5 0.5];K=[3 -1;-1 1]; % compute the state matrix A=[zeros(n) eye(n); -M\K -M\C]; [V,D]=eig(A);% computes eigenvalues and eigenvectors % compute the real and imaginary parts ReD=(D+D')/2;ImD=(D'-D)*i/2; W=(ReD^2+ImD^2).^.5, Zeta=-ReD/W W = 1.4078 0 0 0 0 1.4078 0 0 0 0 0.7103 0 0 0 0 0.7103 406 Multiple-Degree-of-Freedom Systems Zeta = 0.2962 0 0 0 0 0.2962 0 0 0 0 0.1169 0 Chap. 4 0 0 0 0.1169 The solution in Mathematica is as follows: In[1]: = M ∙ a 2 0 c ∙ a K ∙ a o ∙ a 0 b; 1 1 ∙.5 3 1 0 0 ∙.5 b; .5 1 b; 1 0 b; 0 i = IdentityMatrix[2]; Formation of A matrix. In[6]: = A = ArrayFlatten[{o, i}, {-Inverse[M].K, -Inverse[M].c}}]; MatrixForm[A] Out[7]//MatrixForm= 0 0 • 3 2 1 0 0 1 2 -1 1 0 0 1 - 0.5 0.25 0.5 - 0.5 µ Solution of the eigenvalue problem. In[8]: = {𝛌, v} = Eigensystem[A]; MatrixForm[λ] MatrixForm[v] Out[9]//MatrixForm= - 0.416934 - 0.416934 § - 0.0830665 - 0.0830665 + + - 1.34464i 1.34464i ¥ 0.705454i 0.705454i Out[10]//MatrixForm= 0.193156 + 0.341044i 0.193156 - 0.341044i § 0.369554 + 0.0422688i 0.369554 - 0.0422688i -0.126254 - 0.40718i -0.126254 + 0.40718i 0.725456 + 0.i 0.725454 + 0.i -0.539116 + 0.117534i -0.539116 - 0.117534i -0.0605163 + 0.257193i -0.0605163 - 0.257193i 0.600152 + 0.0i 0.600152 + 0.0i ¥ -0.060261 + 0.511776i -0.060261 - 0.511776i Sec. 4.10 407 numerical Simulation of the time Response Calculation of natural frequencies and damping ratios. In[11]: = 𝛚1 ∙ 2Re[𝛌[[3]]]2 ∙ Im[𝛌[3]]]2 𝛚2 ∙ 2Re[𝛌[1]]]2 ∙ Im[𝛌[1]]]2 ∙Re[𝛌[[3]]] 𝛇1 ∙ 𝛚1 ∙Re[𝛌[[1]]] 𝛇2 ∙ 𝛚2 Out[11]= Out[12]= Out[13]= Out[14]= 0.710328 1.4078 0.116941 0.29616 n 4.10 NUMERICAL SIMULATION OF THE TIME RESPONSE This section is a simple extension of Sections 1.9, 2.8, and 3.9, which examine the use of numerical integration to simulate and plot the response of a vibration problem. Simulation is a much easier way to obtain the system’s response when compared to computing the response by modal decomposition, which was stressed in the last nine sections. However, the modal approach is needed to perform design and to gain insight into the dynamics of the system. Important design criteria are often stated in terms of modal information, not directly available from the time response. In addition, the modal properties may be used to check numerical simulations. likewise, numerical solutions may also be used to check analytical work. Consider the forced response of a damped linear system. The most general case can be written as $ # Mx + Cx + Kx = BF(t) x(0) = x0, # # x(0) = x0 (4.188) # Following the development of equation (4.181), define y1(t) = x(t) and y2(t) = x(t) # so that y1(t) = y2(t). Then multiplying equation (4.188) by M -1 yields the coupled first-order vector equations # y1(t) = y2(t) # y2(t) = -M -1Ky1(t) - M -1Cy2(t) + M -1BF(t) (4.189) # with initial conditions y1(0) = x0 and y2(0) = x0. Equation (4.189) can be written as the single first-order equation # y(t) = Ay(t) + f(t) y(0) = y0 (4.190) 408 Multiple-Degree-of-Freedom Systems Chap. 4 where A is the state matrix given by equation (4.183): A = J 0 -M -1K I R -M -1C and y(t) = J y1(t) R y2(t) f(t) = J 0 R M BF(t) -1 y0 = J y1(0) x R = J 0R y2(0) x0 Here y(t) is the (2n * 1) state vector, where the first n * 1 elements correspond to the displacement x(t) and where the second n * 1 elements correspond to the # velocities x(t). The Euler method of the numerical solution given in Section 3.9, equation (3.105) can be applied directly to the vector (state-space) formulation given in equation (4.190), repeated here: y(t i + 1) = y(t i) + ∆tAy(t i) (4.191) which defines the Euler formula for integrating the general vibration problem described in equation (4.190) for the zero-force input case. This can be extended to the forced-response case by including the term fi = f(ti): yi + 1 = yi + ∆tAyi + fi (4.192) where yi + 1 denotes y(ti + 1), and so on, using y(0) as the initial value. As before in Sections 1.9, 2.8, and 3.9, Runge–Kutta integration methods are used in most codes to produce a more accurate approximation to the solution. The following examples illustrate how to use MATlAb, Mathcad, and Mathematica to perform numerical integration to solve vibration problems. Example 4.10.1 Consider the system given by J 4 0 $ 0 x (t) 30 R J $1 R + J 9 x2(t) -5 -5 x (t) 0.23500 RJ 1 R = J R sin (2.756556t) 5 x2(t) 2.97922 A quick computation shows that the driving frequency is also the natural frequency of the second mode. However, because of the force distribution vector b is proportional to the first mode shape, no resonance will occur. The system does not resonate because the force is distributed orthogonal to the second mode. The simulation will verify that this prediction is correct. Solution The problem is first put into state-space form and then solved numerically. In Mathcad the code becomes M := J 4 0 0 R 9 K := J 30 5 5 R 5 B := J 0.23500 R 2.97922 ω:= 2.75655 Sec. 4.10 409 numerical Simulation of the time Response f := M-1 # B 0 0 X := D T 0 0 M-1 # K = K = J 0 0 A1 := D - 7.5 0.556 0 0 1.25 - 0.556 1 0 0 0 7.5 - 0.556 0 1 T 0 0 - 1.25 R 0.556 D(t, X) := f = J A1 # X + 0.059 R 0.331 0 0 D # T f0 sin(ω # t) f1 # sin(ω # t) Z : = rkfixed (X, 0, 20, 3000, D) t : = Z<0> x1 : = Z<1> x2 : = Z<2> The MATlAb code to produce the solution is as follows: clear all xo=[0; 0; 0; 0]; ts=[0 20]; [t,x]=ode45('f',ts,xo); plot(t,x(:,1),t,x(:,2),'--') %------------------------------------------function v=f(t,x) M=[4 0; 0 9]; K=[30 -5; -5 5]; B=[0.23500; 2.97922]; w=2.75655; A1=[zeros(2) eye(2); -inv(M)*K zeros(2)]; f=inv(M)*B; v=A1*x+[0;0; f]*sin(w*t); 0.25 0.2 0.15 [1 [2 Response (m) 0.1 0.05 0 5 10 0.05 0.1 0.15 0.2 0.25 T (s) 15 20 410 Multiple-Degree-of-Freedom Systems Chap. 4 The Mathematica code to produce the solution is as follows: In[1]: = <<PlotLegends' In[2]: = m ∙ a 4 0 k ∙ a 30 ∙5 0 b; 9 ∙5 b; 5 𝛚 = 2.756556; f = a x = a .23500 * Sin[ω * t] b; 2.97922 * Sin[ω * t] x1[t] b; x2[t] xdd = a x1''[t] b; x2''[t] system = m.xdd + k.x; In[9]: = num = NDSolve [{system[[1]] == f[[1]], x1'[0] == 0, x1[0] == 0, system[[2]] == f[[2]], x2[0] == 0, x2'[0] == 0}, {x1[t], x2[t]}, {t, 0, 20}]; Plot[{Evaluate[x1[t] /. num], Evaluate[x2[t] /. num]}, { t, 0, 20}, PlotStyle S {RGBColor[1, 0, 0], RGBColor[0, 1, 0]}, PlotLegend S {"x1[t]", "x2[t]"}, LegendPosition S {1, 0}, LegendSize S {1, .3}] n Example 4.10.2 Compute and plot the time response of the system (newtons) $ # 2 0 x1(t) 3 -0.5 x1(t) 3 - 1 x1(t) 1 J R J$ R + J RJ# R + J RJ R = J R sin (ωt) 1 1 x2(t) - 0.5 0.5 x2(t) -1 1 x2(t) 1 subject to the initial conditions x0 = J 0 R m, 0.1 1 v0 = J R m/s 0 Solution This amounts to formulating the system into the state-space equations given by (4.190) and running a Runge–Kutta routine. The solution in Mathcad is M : = J 2 0 0 R 1 C : = J O : = J B : 0 0 3 - 0.5 0 R 0 = 1 J R 1 - 0.5 R 0.5 I : = J K : = J 1 0 ω : = 2 0 R 1 3 -1 -1 R 1 Sec. 4.10 411 numerical Simulation of the time Response 0 0.1 X: = D T 1 0 A : = augment(stack(O,-M-1 # K),stack(I,-M-1 C)) f : = M-1 # B D(t, X) : = A#X + f = J 0.5 R 1 0 0 D T # sin (ω # t) f0 f1 Z : = rkfixed (X, 0, 20, 3000, D) t : = Z<0> x1 : = Z<1> x2 : = Z<2> 1.5 1.0 0.5 x1 x2 5 10 15 0.5 1.0 The MATlAb code to produce the same plot is as follows: clear all xo=[0; 0.1; 1; 0]; ts=[0 20]; [t,x]=ode45('f',ts,xo); plot(t,x(:,1),t,x(:,2),'--') %-----------------------------------------function v=f(t,x) M=[2 0; 0 1]; C=[3 -0.5; -0.5 0.5]; K=[3 -1; -1 1]; B=[1; 1]; 20 412 Multiple-Degree-of-Freedom Systems Chap. 4 w=2; A1=[zeros(2) eye(2); -inv(M)*K -inv(M)*C]; f=inv(M)*B; v=A1*x+[0;0; f]*sin(w*t); The Mathematica code for simulating and plotting the response is as follows: In[1]: = <<PlotLegends' In[2]: = m = a c = a k = a 2 0 3 - .05 3 -1 ω= 2; f = a 0 b; 1 - .05 b; .05 -1 b; 1 Sin[ω * t] b; Sin[ω * t] x = J x1[t] R; x2[t] xd = J x1 ' [t] R; x2 ' [t] xdd = J x1 '' [t] R; x2 '' [t] system = m.xdd + c.xd + k.x; In[11]: = num = NDSolve [{system[[1]] == f[[1]], x1'[0] == 1, x1[0] == 0, system[[2]] == f[[2]], x2[0] == .1, x2'[0] == 0}, {x1[t], x2[t]}, { t, 0, 20} ]; Plot [{Evaluate [x1[t] /. num], Evaluate[x2[t] /. num]}, {t, 0, 20}, PlotStyle S {RGBColor[1, 0, 0], RGBColor[0, 1, 0]}, PlotLegend S {"x1[t]", "x2[t]"}, LegendPosition S {1, 0}, LegendSize S {1, .3}] n Example 4.10.3 Consider the following system excited by a pulse of duration 0.1 s (units in newtons): J 2 0 $ 0 x 0.3 R J $1 R + J 1 x2 - 0.05 # - 0.05 x 3 R J #1 R + J 0.05 x2 -1 -1 x1 RJ d 1 x2 0 = J R [Φ(t - 1) - Φ(t - 1.1)] 1 Sec. 4.10 413 numerical Simulation of the time Response and subject to the initial conditions x0 = J 0 R m, - 0.1 0 v0 = J R m/s 0 Compute and plot the response of the system. Here Φ indicates the Heaviside step function introduced in Section 3.2. Solution This again follows the same format as the previous examples of putting the equations of motion into state-matrix form and solving using one of the software programs. In Mathcad the code and solution are M : = J 2 0 0 R 1 C : = J o : = J 0 B: = J R 1 0 0 0.3 - 0.05 0 R 0 - 0.05 R 0.05 I : = J 1 0 K : = J 3 -1 -1 R 1 0 R 1 A : = augment (stack (o, -M-1 # K), stack (I, -M-1 C)) -1 f : = M #B 0 f = J R 1 0 (- 0.1) X: = D T 1 0 0 0 D(t, X) : = A # X + D T # (Φ(t - 1) - Φ(t - 1.1)) f0 f1 Z : = rkfixed (X, 0, 30, 3000, D) t : = Z<0> x1 : = Z<1> x2 : = Z<2> The MATlAb code for producing the same plot is as follows. Note that in this case it is necessary to set the tolerance for ODE45 in order to clearly define the “impulse” as the difference between two step functions. This is done using the options command listed in the following. clear all xo = [0; -0.1; 0; 0]; ts = [0 30]; options=odeset('RelTol',le-4); [t,x] = ode45('f',ts,xo); plot(t,x(:,1),t,x(:,2),'– –') %-------------------------------------------function v = f(t,x) M C K B = = = = [2 0; 0 1]; [0.3 -0.05; -0.05 0.05]; [3 -1; -1 1]; [0; 1];t1 = 1; t2 = 1.1; 414 Multiple-Degree-of-Freedom Systems Chap. 4 A1 = [zeros(2) eye(2); -inv(M)*K -inv(M)*C]; f = inv(M)*B; v = A1*x + [0;0; f]*(stepfun(t,t1)-stepfun(t,t2)); 0.2 0.1 5 10 15 20 25 30 0.1 0.2 The Mathematica code for simulating and plotting the response is as follows. As in the MATlAb case, the tolerances must be set in order to define the impulse. In this case, the command is PrecisionGoal->10 as listed below: In[1]: = <<PlotLegends' In[2]: = m = a c = a k = a f = a 0 b; 1 2 0 0.3 - 0.05 3 -1 - 0.05 b; 0.05 -1 b; 1 0 b; UnitStep[t - 1] - UnitStep[t - 1.1] x = J x1[t] R; x2[t] xd = J x1 ' [t] R; x2 ' [t] xdd = J x1 '' [t] R; x2 '' [t] system = m.xdd + c.xd + k.x; Sec. 4.10 415 numerical Simulation of the time Response In[10]: = num = NDSolve [{system[[1]] == f[[1]], x1'[0] == 0, x1[0] == 0, system[[2]] == f[[2]], x2[0] == -.1, x2'[0] == 0}, {x1[t], x2[t]}, {t, 0, 30}, PrecisionGoal->10]; Plot [{Evaluate [x1[t] /. num], Evaluate[x2[t] /. num]}, {t, 0, 30}, PlotStyle S {RGBColor[1, 0, 0], RGBColor[0, 1, 0]}, PlotRange S {-.2,.2}, PlotLegend S {"x1[t]", "x2[t]"}, LegendPosition S {1, 0}, LegendSize S {1, .3}] n The preceding examples illustrate the basic features of using math software to solve vibration problems. These examples are all simple two-degree-of-freedom systems, but the methods and routines work for any number of degrees of freedom, limited only by the array size for a particular code. The Toolbox provides additional solution possibilities. PROBLEMS Those problems marked with an asterisk are intended to be solved using computational software. Section 4.1 (Problems 4.1 through 4.19) 4.1. Consider the system of Figure P4.1. For c1 = c2 = c3 = 0, derive the equation of motion and calculate the mass and stiffness matrices. Note that setting k3 = 0 in your solution should result in the stiffness matrix given by equation (4.9). k2 k1 k3 m2 m1 c2 c1 x1 c3 x2 Figure P4.1 4.2. Calculate the characteristic equation from Problem 4.1 for the case m1 = 9 kg m2 = 1 kg k1 = 24 N>m k2 = 3 N>m k3 = 3 N>m and solve for the system’s natural frequencies. 4.3. Calculate the vectors u1 and u2 for Problem 4.2. # 4.4. For initial conditions x(0) = [1 0]T and x(0) = [0 0]T calculate the free response of the system of Problem 4.2. Plot the response x1(t) and x2(t). 416 Multiple-Degree-of-Freedom Systems Chap. 4 4.5. Calculate the response of the system J 9 0 0 $ 27 R x(t) + J 1 -3 -3 R x(t) = 0 3 # described in Example 4.1.7, to the initial condition x(0) = 0, x(0) = [1 response, and compare the result to Figure 4.3. 0]T , plot the 4.6. Write the equations of motion for the system of Figure P4.1 for the case that k1 = k3 = 0 and identify the mass and stiffness matrix for this case. 4.7. Calculate and solve the characteristic equation for the following system: J 9 0 0 $ 1 R x(t) + 10J 1 -1 -1 R x(t) = 0 1 4.8. Compute the natural frequencies of the following system: J 6 2 2 $ 3 R x(t) + J 4 -1 -1 R x(t) = 0. 1 4.9. Calculate the solution J 9 0 0 $ 27 R x(t) + J 1 -3 -3 R x(t) = 0, 3 1 x(0) = C 3 S 1 # x(0) = 0 Compare the response with that of Figure 4.3. 4.10. Calculate the solution J 9 0 0 $ 27 R x(t) + J 1 -3 -3 R x(t) = 0, 3 1 x(0) = C 3 S 1 # x(0) = 0 Compare the response with that of Figure 4.3.If you worked Problem 4.9, compare your solution to that response also. 4.11. Compute the natural frequencies and mode shapes of the following system: J 4 0 0 $ 4 R x(t) + 10J 1 -2 -2 R x(t) = 0 1 4.12. Determine the equation of motion in matrix form, then calculate the natural frequencies and mode shapes of the torsional system of Figure P4.12. Assume that the torsional stiffness values provided by the shaft are equal (k1 = k2) and that disk 1 has three times the inertia as that of disk 2 (J1 = 3J2). 1(t) k1 J1 2(t) k2 J2 Figure P4.12 Torsional system with two disks and, hence, two degrees of freedom. Chap. 4 417 Problems 4.13. Two subway cars of Figure P4.13 have 2000 kg mass each and are connected by a coupler. The coupler can be modeled as a spring of stiffness k = 280,000 N>m. Write the equation of motion and calculate the natural frequencies and (normalized) mode shapes. x2(t) x1(t) m k m Figure P4.13 Vibration model of two subway cares connected by a coupling device modeled as a massless spring. 4.14. Suppose that the subway cars of Problem 4.13 are given the initial position of x10 = 0, x20 = 0.1 m, and initial velocities of 10 = 20 = 0. Calculate the response of the cars. 4.15. A slightly more sophisticated model of a vehicle suspension system is given in Figure P4.15. Write the equations of motion in matrix form. Calculate the natural frequencies for k1 = 103 N>m, k2 = 104 N>m, m2 = 50 kg, and m1 = 2000 kg. x1(t) m1 Car mass m1 k1 k1 Car spring x2(t) k2 Tire stiffness m2 m2 Tire mass k2 Figure P4.15 A two-degreeof-freedom model of a vehicle suspension system. 4.16. Examine the effect of the initial condition of the system of Figure 4.1(a) on the responses x1(t) and x2(t) by repeating the solution of Example 4.1.7 given by 1 1 A1 sin ( 12t + ϕ1) - A2 sin (2t + ϕ2) x1(t) 3 3 C S J = x2(t) R A1 sin ( 12t + ϕ1) + A2 sin (2t + ϕ2) # # # first for x10 = 0, x20 = 1 with x10 = x20 = 0 and then for x10 = x20 = x10 = 0 and # x20 = 1. Plot the time response in each case and compare your results against Figure 4.3. 4.17. Consider the system defined by J 9 0 0 $ 24 + k2 Rx + J 1 - k2 - k2 Rx = 0 k2 # # Using the initial conditions x1(0) = 1 mm, x2(0) = 0, and x1(0) = x2(0) = 0, resolve and plot x1(t) for the cases that k2 takes on the values 0.3, 3, 30, and 300. In each case, compare the plots of x1 and x2. What can you conclude? 418 Multiple-Degree-of-Freedom Systems 4.18. Consider the system defined by $ m1 0 x k + k2 R J $1 R + J 1 J 0 m2 x2 - k2 Chap. 4 - k2 x1 0 RJ R = J R k2 x2 0 Determine the natural frequencies in terms of the parameters m1, m2, k1, and k2. How do these compare to the two single-degree-of-freedom frequencies ω1 = 1k1 >m1 and ω2 = 1k2 >m2 ? 4.19. Consider the problem of Example 4.1.7 with the first degree-of-freedom response given by x1(t) = 0.5(cos 12t + cos 2t). Use a trig identity to show the x1(t) experiences a beat. Plot the response to show the beat phenomena in the response. Section 4.2 (Problems 4.20 through 4.35) 4.20. Calculate the square root of the matrix 3Hint: let M1>2 M = J = J 13 - 10 - 10 R 8 -b R ; calculate (M1>2 ) 2 and compare to M.4 c a -b 4.21. Normalize the vectors J 1 0 -0.1 R, J R, J R -2 5 0.1 first with respect to unity (i.e., xTx = 1) and then again with respect to the matrix M (i.e., xTMx = 1), where M = J 3 - 0.1 - 0.1 R 2 4.22. Consider the vibrating system described by J 4 0 0 $ 4 R x(t) + J 1 -2 -2 R x(t) = 0 1 Compute the mass-normalized stiffness matrix, the eigenvalues, the normalized eigenvectors, the matrix P, and show that PTMP = I and PTKP is the diagonal matrix of eigenvalues Λ. ∙ 4.23. Calculate the matrix K for the system defined by J m1 0 0 $ k + k2 R x(t) + J 1 m2 - k2 - k2 R x(t) = 0 k2 + k3 and see that it is symmetric. 4.24. Consider the vibrating system described by J 4 0 0 $ 4 R x(t) + J 1 -2 -2 R x(t) = 0 1 Chap. 4 419 Problems Compute the mass-normalized stiffness matrix, the eigenvalues, the normalized eigenvectors, the matrix P, and show that PTMP = I and PTKP is the diagonal matrix of eigenvalues Λ. 4.25. Discuss the relationship or difference between a mode shape of equation (4.54) and an ∙ eigenvector of K. ∙ 4.26. Calculate the units of the elements of matrix K. 4.27. Calculate the spectral matrix Λ and the modal matrix P for the vehicle model of Problem 4.15 described by J 2000 0 0 $ 1000 R x(t) + J 50 - 1000 - 1000 R x(t) = 0 11,000 4.28. Calculate the spectral matrix Λ and the modal matrix P for the system given by J 2000 0 0 280,000 $ R x(t) + J 2000 - 280,000 -280,000 R x(t) = 0 280,000 ∙ 4.29. Calculate K for the torsional vibration problem given by J2 J 3 0 0 $ 2 R 𝛉(t) + kJ 1 -1 -1 R 𝛉(t) = 0 1 ∙ What are the units of K ? 4.30. Consider the system in Figure P4.30 for the case where m1 = 1 kg, m2 = 4 kg, k1 = 240 N>m, and k2 = 300 N>m. Write the equations of motion in vector form and compute each of the following: (a) the natural frequencies (b) the mode shapes (c) the eigenvalues (d) the eigenvectors (e) show that the mode shapes are not orthogonal (f) show that the eigenvectors are orthogonal (g) show that the mode shapes and eigenvectors are related by M -1>2 (h) write the equations of motion in modal coordinates Note that the purpose of this problem is to help you see the difference between these various quantities. x1(t) x2(t) m2 m1 k1 k2 Figure P4.30 A two-degree-of-freedom system. 4.31. Consider the following system: J 1 0 0 $ 3 R x(t) + J 4 -1 -1 R x(t) = 0 1 420 Multiple-Degree-of-Freedom Systems Chap. 4 where M is in kg and K is in N>m. (a) Calculate the eigenvalues of the system. (b) Calculate the eigenvectors and normalize them. 4.32. The torsional vibration of the wing of an airplane is modeled in Figure P4.32. Write the equation of motion in matrix form and calculate analytical forms of the natural frequencies in terms of the rotational inertia and stiffness of the wing. 1(t) k1 Airplane wing with engines 2(t) J1 k2 J2 Wing modeled as two shafts and two disks for torsional vibration Figure P4.32 A crude model of the torsional vibration of a wing consisting of a twoshaft, two-disk system similar to Problem 4.12 used to estimate the torsional natural frequencies of the wing where the engine inertias are approximated by the disks. 4.33. Calculate the value of the scalar a such that x1 = [a are orthogonal. -1 1]T and x2 = [1 0 1]T 4.34. Normalize the vectors of Problem 4.33. Are they still orthogonal? 4.35. Which of the following vectors are normal? Orthogonal? 1 12 x1 = E 0 U 1 12 0.1 x2 = C 0.2 S 0.3 0.3 x3 = C 0.4 S 0.3 Section 4.3 (Problems 4.36 through 4.46) 4.36. Decouple the following equation of motion into two decoupled equations of motion: J 4 0 0 $ 4 R x(t) + J 1 -2 -2 R x(t) = 0 2 4.37. Solve the system of Problem 4.12 given by J2 J 3 0 0 $ 2 R 𝛉 + kJ 1 -1 -1 R𝛉 = 0 1 Chap. 4 421 Problems Using modal analysis for the case where the rods have equal stiffness (i.e., k1 = k2), J1 = 3J2, # and the initial conditions are x(0) = [0 1]T and x(0) = 0. 4.38. Consider the system J 9 0 0 $ 27 R x(t) + J 1 -3 -3 R x(t) = 0 3 # of Example 4.3.1. Calculate a value of x(0) and x(0) such that both masses of the system oscillate with a single frequency of 2 rad>s. 4.39. Consider the system of Figure P4.39 consisting of two pendulums coupled by a spring. Determine the natural frequency and mode shapes. Plot the mode shapes as well as the solution to an initial condition consisting of the first mode shape for k = 20 N>m, l = 0.5 m, and m1 = m2 = 10 kg, a = 0.1 m along the pendulum. k a l 1 m1 2 m2 Figure P4.39 Two pendulums of the same length connected by a spring used to model a machine part. 4.40. Compute and plot the response of J subject to x(0) = [1 Figure 4.6. 1 0 0 $ 12 R x(t) + J -2 10 -2 R x(t) = 0 12 # 1]T and x(0) = 0. Compare your result to Example 4.3.2 and 4.41. Use modal analysis to calculate the solution of J 1 0 0 $ 3 R x(t) + J -1 4 -1 R x(t) = 0 1 # 1]T (mm) and x(0) = [0 for the initial conditions x(0) = [0 0]T (mm>s). 4.42. For the matrices 1 M -1>2 = C 12 0 0 4 S and P = 1 1 J 12 -1 1 R 1 calculate M -1>2P, 1 M -1>2P 2 T, and PTM -1>2 and hence verify that the computations in equation (4.70) make sense. 422 Multiple-Degree-of-Freedom Systems Chap. 4 4.43. Consider the two-degree-of-freedom system defined by J 9 0 0 $ 27 R x(t) + J 1 -3 -3 R x(t) = 0 3 Calculate the response of the system to the initial conditions x0 = 1 1 C3S 12 1 # x0 = 0 What is unique about your solution compared to the solution of Example 4.3.1? 4.44. Consider the two-degree-of-freedom system defined by J 9 0 0 $ 27 R x(t) + J -3 1 -3 R x(t) = 0 3 Calculate the response of the system to the initial conditions x0 = 0 and 1 1 # x0 = C 3 S 12 -1 What is unique about your solution compared to the solution of Example 4.3.1 and to Problem 4.40, if you also worked that? 4.45. Consider the system defined by J 100 0 0 $ 25,000 R x(t) + J - 15,000 100 -15,000 R x(t) = 0 25,000 Solve for the free response of this system using modal analysis and the initial conditions. 4.46. Consider the model of a vehicle given in Problem 4.15 illustrated in Figure P4.15 defined by J 2000 0 0 $ 1000 Rx + J 50 - 1000 - 1000 Rx = 0 11,000 Suppose that the tire rolls over a bump modeled as the initial conditions of # x(0) = [0 0.01]T and x(0) = 0. Use modal analysis to calculate the response of the car x1(t). Plot the response for three cycles. Section 4.4 (Problems 4.47 through 4.59) 4.47. A vibration model of the drive train of a vehicle is illustrated as the three-degree-offreedom system of Figure P4.47. Calculate the undamped free response [i.e., M(t) = # F(t) = 0, c1 = c2 = 0] for the initial condition x(0) = 0, x(0) = [0 0 1]T . Assume that the hub stiffness is 10,000 N>m and that the axle>suspension is 20,000 N>m. Assume the rotational element J is modeled as a translational mass of 75 kg. Chap. 4 423 Problems Rotational Tire damping Hub damping x1(t) M(t) Clutch torque x2(t) x3(t) Equivalent inertia of transmission m2100 kg Torque convertor inertia J75 kg•m2/rad Vehicle equivalent inertia m33000 kg F(t) Wind and road load Axle and suspension stiffness Hub stiffness Translational Figure P4.47 A simplified model of an automobile for vibration analysis of the drive train. The parameter values given are representative and should not be considered as exact. 4.48. Calculate the natural frequencies and normalized mode shapes of 4 C0 0 0 2 0 0 4 $ 0 S x + C -1 1 0 0 -1 S x = 0 1 -1 2 -1 4.49. The vibration is the vertical direction of an airplane and its wings can be modeled as a three-degree-of-freedom system, with one mass corresponding to the right wing, one x1(t) x2(t) l l m EI (a) x3(t) 4m EI m (b) x2(t) x1(t) m x3(t) m 4m k 3EI k l3 3EI l3 (c) Figure P4.49 A model of the wing vibration of an airplane: (a) vertical wing vibration; (b) lumped mass >beam deflection model; (c) spring–mass model. 424 Multiple-Degree-of-Freedom Systems Chap. 4 mass for the left wing, and one mass for the fuselage. The stiffness connecting the three masses corresponds to that of the wing and is a function of the modulus E of the wing. The equation of motion is 1 mC 0 0 $ 0 x1(t) 3 EI $ 0 S C x2(t) S + 3 C -3 $ l x3(t) 1 0 0 4 0 -3 6 -3 0 x1(t) 0 - 3 S C x2(t) S = C 0 S 3 x3(t) 0 The model is given in Figure P4.49. Calculate the natural frequencies and mode shapes. Plot the mode shapes and interpret them according to the airplane’s deflection. 4.50. Solve for the free response of the system of Problem 4.49, where E = 6.9 * 109 N>m2, l = 2 m, m = 3000 kg, and I = 5.2 * 10 - 6 m4. let the initial displacement correspond # to a gust of wind that causes an initial condition of x(0) = 0, x(0) = [0.2 0 0]T m. Discuss your solution. 4.51. Consider the two-mass system of Figure P4.51. This system is free to move in the x1 - x2 plane. Hence, each mass has two degrees of freedom. Derive the linear equations of motion, write them in matrix form, and calculate the eigenvalues and eigenvectors for m = 10 kg and k = 100 N>m. x2 3k x4 4k x1 4k x3 m m 4k k k 2k Figure P4.51 A two-mass system free to move in two directions. 4.52. Consider again the system discussed in Problem 4.51. Use modal analysis to calculate the solution if the mass on the left is raised along the x2 direction exactly 0.01 m and let go. 4.53. The vibration of a floor in a building containing heavy machine parts is modeled in Figure P4.53. Each mass is assumed to be evenly spaced and significantly larger than the mass of the floor. The equation of motion then becomes (m1 = m2 = m3 = m). Chap. 4 425 Problems l x1(t) x2(t) x3(t) m1 m2 m3 l l Figure P4.53 A lumped-mass model of boxes loaded on the floor of a building. l 9 64 1 EI $ mI x + 3 F 6 l 13 192 1 6 1 3 1 6 13 192 x1 1 V C x2 S = 0 6 x3 9 64 Calculate the natural frequencies and mode shapes. Assume that in placing box m2 on the floor (slowly) the resulting vibration is calculated by assuming that the initial displacement at m2 is 0.05 m. If l = 2 m, m = 200 kg, E = 0.6 * 109 N>m2, I = 4.17 * 10 - 5 m4, calculate the response and plot your results. 4.54. Recalculate the solution to Problem 4.53 for the case that m2 is increased in mass to 2000 kg. Compare your results to those of Problem 4.53. Do you think it makes a difference where the heavy mass is placed? 4.55. Repeat Problem 4.49 for the case that the airplane body is 10 m instead of 4 m as indicated in the figure. What effect does this have on the response, and which design (4 m or 10 m) do you think is better for reduced vibration? 4.56. Often in the design of a car, certain parts cannot be reduced in mass. For example, consider the drive train model illustrated in Figure P4.47. The mass of the torque converter and transmission are relatively the same from car to car. However, the mass of the car could change as much as 1000 kg (e.g., a two-seater sports car versus a family sedan). With this in mind, resolve Problem 4.47 for the case that the vehicle inertia is reduced to 2000 kg. Which case has the smallest amplitude of vibration? 4.57. Use the mode summation method to compute the analytical solution for the response of the two-degree-of-freedom system: J 1 0 0 $ 540 R x(t) + J 4 - 300 - 300 R x(t) = 0 300 to the initial conditions of x0 = J 0 0 # R , x0 = J R 0.01 0 426 Multiple-Degree-of-Freedom Systems Chap. 4 4.58. For a zero value of an eigenvalue and hence frequency, what is the corresponding time response? Or asked another way, the form of the modal solution for a nonzero frequency is A sin (ωnt + ϕ), what is the form of the modal solution that corresponds to a zero frequency? Evaluate the constants of integration if the modal initial conditions # are r1(0) = 0.1, and r1(0) = 0.01. 4.59. Consider the system described by J 1 0 0 $ 400 R x(t) + J 4 - 400 subject to the initial conditions x(0) = [1 versus time. - 400 R x(t) = 0 400 # 0]T, x(0) = 0. Plot the displacements Section 4.5 (Problems 4.60 through 4.72) 4.60. Consider the following two-degree-of-freedom system and compute the response assuming modal damping rations of ζ1 = 0.01 and ζ2 = 0.1 : J 5 0 0 $ 20 R x(t) + J 1 -3 -3 R x(t) = 0, 3 x0 = J 0.1 # R , x0 = 0 0.05 Plot the response. 4.61. Consider the example of the automobile drive train system discussed in Problem 4.47, modeled by 75 C0 0 0 100 0 0 1 $ 0 S x + 10,000 C - 1 3000 0 # x(0) = 0 and x(0) = [0 0 -1 3 -2 0 -2 S x = 0 2 1]T m>s Add 10% modal damping to each coordinate, calculate and plot the system response. 4.62. Consider the following two-degree-of-freedom system and compute the response assuming modal damping ratios of ζ1 = 0.05 and ζ2 = 0.01 : J 50 0 0 $ 20 R x(t) + J 1 -9 -9 R x(t) = 0, 9 x0 = J 0.05 # R , x0 = 0 0.09 Plot the response. 4.63. Consider the model of an airplane discussed in Problem 4.49, Figure P4.49 modeled by 3000 C 0 0 0 12,000 0 0 13,455 $ 0 S x + C - 13,455 3000 0 - 13,455 26,910 - 13,455 0 -13,455 S x = 0 13,455 # subject to the initial conditions x(0) = [0.02 0 0]T m and x(0) = 0. (a) Calculate the response assuming that the damping provided by the wing rotation is ζi = 0.01 in Chap. 4 427 Problems each mode. (b) If the aircraft is in flight, the damping forces may increase dramatically to ζi = 0.1. Recalculate the response and compare it to the more lightly damped case of part (a). 4.64. Repeat the floor vibration problem of Problem 4.53 given by 9 64 1 $ 200x + 3.197 * 10-4 F 6 13 192 x(0) = [0 0.05 1 6 1 3 1 6 13 192 1 Vx = 0 6 9 64 # 0]Tm and x(0) = 0 by assigning modal damping ratios of ζ1 = 0.01 ζ2 = 0.1 ζ3 = 0.2 4.65. Repeat Problem 4.64 with constant modal damping of ζ1, ζ2, ζ3 = 0.1. If you worked the previous problem, compare this solution with the solution of Problem 4.64. 4.66. Consider the damped system of Figure P4.66. Determine the damping matrix and use the formula of equation (4.119) to determine values of the damping coefficient cI for which this system would be proportionally damped. k2 k1 k3 m2 m1 c2 c1 c3 x2 x1 Figure P4.66 A damped two-degreeof-freedom system fixed at each end. 4.67. let k3 = 0 in Figure P4.66. Also let m1 = 1, m2 = 4, k1 = 2, k2 = 1, and calculate c1, c2, and c3 such that ζ1 = 0.01 and ζ2 = 0.1. 4.68. Calculate the constants α and β for the two-degree-of-freedom system given by J 1 0 0 $ 3 # R x + (αM + βK)x + J 4 -1 -1 Rx = 0 1 such that the system has modal damping of ζ1 = ζ2 = 0.3. 4.69. Equation (4.124) represents n equations in only two unknowns and hence cannot be used to specify all the modal damping ratios for a system with n 7 2. If the floor vibration system of Problem 4.53 has measured damping of ζ1 = 0.01 and ζ2 = 0.05, determine ζ3. 4.70. If you worked the previous problem, calculate the damping matrix for the system of Problem 4.69. What are the units of the elements of the damping matrix? 428 Multiple-Degree-of-Freedom Systems Chap. 4 4.71. Does the following system decouple? If so, calculate the mode shapes and write the equation in decoupled form. J 1 0 0 $ 5 Rx + J 1 -3 -3 # 5 Rx + J 3 -1 -1 Rx = 0 1 4.72. Show that if the damping matrix satisfies C = αM + βK, then the matrix CM -1K is symmetric and hence that CM -1K = KM -1C. Section 4.6 (Problems 4.73 through 4.83) 4.73. Calculate the response of the system of Figure P4.73 discussed in Example 4.6.1 if F2(t) = δ(t) and the initial conditions are set to zero. This might correspond to a twodegree-of-freedom model of a car hitting a bump. F2(t)  3 cos 2t k2 k1 m2 m1 c2 c1 Figure P4.73 A damped two-degree-of-freedom system. x2 x1 4.74. Calculate the response of the system of Figure P4.73 discussed in Example 4.6.1 if F1(t) = δ(t) and the initial conditions are set to zero. This might correspond to a twodegree-of-freedom model of a car hitting a bump. 4.75. For an undamped two-degree-of-freedom system, show that resonance occurs at one or both of the system’s natural frequencies. 4.76. Use modal analysis to calculate the response of the drive train system of Problem 4.44 given by 75 C0 0 0 100 0 0 1 $ 0 S x + 10,000C -1 3000 0 -1 3 -2 0 -2 S x = 0 2 to a unit impulse on the car body (i.e., at location x3). Use the modal damping of 10% in each mode. Calculate the solution in terms of physical coordinates, and after subtracting the rigid-body modes, compare the responses of each part. 4.77. Consider the machine tool of Figure 4.28 with a floor mass of m = 1000 kg, subject to a force of 10 sint (in newtons) so that the equation of motions is 1 103 2 C 0.4 0 0 0 2 0 0 $ 0 S x(t) + 1 1 104 2 C - 30 30 0 - 30 38 -8 0 0 - 8 S x(t) = C 0 S 88 10 sin t Calculate the response. How much does this floor vibration affect the machine’s toolhead compared to the solution given in Example 4.8.3? Chap. 4 429 Problems 4.78. Consider the airplane of Figure P4.49 with modal damping of 0.1 in each mode. Suppose that the airplane hits a gust of wind, which applies an impulse of 3δ(t) at the end of the left wing and δ(t) at the end of the right wing. Calculate the resulting vibration of the cabin [x2(t)]. 4.79. Consider again the airplane of Figure P4.49. with 10% modal damping in each mode. Suppose that this is a propeller-driven airplane with an internal combustion engine mounted in the nose. At a cruising speed, the engine mounts transmit an applied force to the cabin mass (4m at x2) which is harmonic of the form 50 sin10t. Calculate the effect of this harmonic disturbance at the nose and on the wing tips after subtracting out the translational or rigid motion. 4.80. Consider the automobile model of Problem 4.15 illustrated in Figure P4.15 with equations of motion: J 2000 0 0 $ 1000 Rx + J 50 - 1000 - 1000 Rx = 0 11000 Add modal damping to this model of ζ1 = 0.01 and ζ2 = 0.2 and calculate the response of the body [x2(t)] to a harmonic input at the second mass of 10 sin3t N. 4.81. Determine the modal equations for the following system and comment on whether or not the system will experience resonance. 2 $ x + J -1 -1 1 R x = J R sin (0.618t) 1 0 4.82. Consider the following system and compute the solution using the mode summation method. M = J 9 0 0 27 R, K = J 1 -3 -3 1 0 # R , x(0) = J R , x(0) = J R 3 0 0 4.83. Consider the following two systems, and in each case determine if a resonance response occurs. $ m1 0 x k + k2 -k2 x 0.642 R J $1 R + J 1 R J 1R = J R sin (2t) (a) J 0 m2 x2 - k2 k2 x2 0.761 $ m 0 x k + k2 -k2 x 0.23500 (b) J 1 R J $1 R + J 1 R J 1R = J R sin (2.756556t) 0 m2 x2 - k2 k2 x2 2.97922 where m1 = 4 kg, k1 = 25 N>m, m2 = 9 kg, and k2 = 5 N>m. Section 4.7 (Problems 4.84 through 4.87) 4.84. Use lagrange’s equation to derive the equations of motion of the lathe of Figure 4.21 for the undamped case. 4.85. Use lagrange’s equations to derive the equations of motion for the automobile of Example 4.8.2 illustrated in Figure 4.25 for the case c1 = c2 = 0. 4.86. Use lagrange’s equations to derive the equations of motion for the building model presented in Figure 4.9 of Example 4.4.3 for the undamped case. 430 Multiple-Degree-of-Freedom Systems Chap. 4 4.87. Consider again the model of the vibration of an automobile of Figure 4.25. In this case, include the tire dynamics as indicated in Figure P4.87. Derive the equations of motion using lagrange formulation for the undamped case. let m3 denote the mass of the car acting at c.g. } x3 l1 l2 l1 c.g.  } x3 l2 c1 k1 m1 c3 Figure P4.87 } c2 x3 k2 Suspension m2 k3 x1 x2 } c4 k4 Tire dynamics A simple car model including tire dynamics. Section 4.9 (Problems 4.88 through 4.98) *4.88. Consider the mass matrix M = J 10 -1 -1 R 1 and calculate M - 1, M - 1 >2, and the Cholesky factor of M. Show that LLT = M M -1>2 M -1>2 = I M1>2 M1>2 = M *4.89. Consider the matrix and vector A = J 1 -ε -ε R ε b = J 10 R 10 use a code to solve Ax = b for ε = 0.1, 0.01, 0.001, 10 - 6, and 1. *4.90. Calculate the natural frequencies and mode shapes of the system of Example 4.8.3. Use the undamped equation and the form given by equation (4.161). *4.91. Compute the natural frequencies and mode shapes of the undamped version of the system of Example 4.8.3 using the formulation of equation (4.164) and (4.168). Compare your answers. *4.92. Use a code to solve for the modal information of the following system J 9 0 0 $ 27 R x(t) + J 1 -3 -3 R x(t) = 0 3 Chap. 4 431 Problems *4.93. Write a program to normalize the vector x = [0.4450 0.8019 1]T . *4.94. Use a code to calculate the natural frequencies and mode shapes obtained for the system J 1 0 0 $ 12 R x(t) + J -2 4 -2 R x(t) = 0 12 *4.95. Following the modal analysis solution of Window 4.5, write a program to compute the time response of the system of Problem 4.94. *4.96. Use a code to solve the damped vibration problem J 9 0 0 $ 2.7 Rx + J 1 - 0.3 -0.3 # 27 Rx + J 0.3 -3 -3 Rx = 0 3 by calculating the natural frequencies, damping ratios, and mode shapes. *4.97. Consider the vibration of the airplane of Problems 4.46 and 4.47 as given in Figure P4.46. The mass and stiffness matrices are given as 1 M = mC 0 0 0 4 0 0 0S 1 K = 3 EI C -3 l3 0 -3 6 -3 0 -3 S 3 where m = 3000 kg, l = 2 m, I = 5.2 * 10 - 6 m4, E = 6.9 * 109 N>m2, and the damping matrix C is taken to be C = (0.002)K. Calculate the natural frequencies, normalized mode shapes, and damping ratios. *4.98. Consider the proportionally damped, dynamically coupled system given by M = J 9 -1 -1 R 1 C = J 3 -1 -1 R 1 K = J 49 -2 -2 R 2 and calculate the mode shapes, natural frequencies, and damping ratios. Section 4.10 (Problems 4.99 through 4.106) *4.99. Recall the system of Example 1.7.3 for the vertical suspension system of a car modeled by $ # mx(t) + cx(t) + kx(t) = 0, with m = 1361 kg, k = 2.668 * 105 N>m, and c = 3.81 * 4 10 kg>s subject to the initial conditions of x(0) = 0 and v(0) = 0.01 m>s2. Solve this and plot the solution using numerical integration. *4.100. Solve for the time response of Example 4.4.3 (i.e., the four-story building of Figure 4.9) modeled by 1 0 4000D 0 0 0 1 0 0 0 0 1 0 0 10,000 0 $ -5000 T x(t) + D 0 0 1 0 -5000 10,000 - 5000 0 0 - 5000 10,000 - 5000 0 0 T x(t) = 0 -5000 5000 432 Multiple-Degree-of-Freedom Systems Chap. 4 subject to an initial displacement of x(0) = [0.001 0.010 0.020 0.025]T and zero initial velocity. Compare the solutions obtained with using a modal analysis approach to a solution obtained by numerical integration. *4.101. Reproduce the plots of Figure 4.13 for the two-degree-of-freedom system of Example 4.5.1 given by J 9 1 0 $ 27 R x(t) + J 0 -3 1 # x(0) = J R , x(t) = 0 0 -3 R x(t) = 0, 3 using a numerical integration code. *4.102. Consider Example 4.8.3 and (a) using the damping ratios given, compute a damping matrix in physical coordinates, (b) use numerical integration to compute the response and plot it, and (c) use the numerical code to design the system so that all three physical coordinates die out within 5 seconds (i.e., change the damping matrix until the desired response results). *4.103. Compute and plot the time response of the system (newtons) J 5 0 $ 0 x 3 R J $1 R + J 1 x2 - 0.5 # - 0.5 x 3 R J #1R + J 0.5 x2 -1 - 1 x1 1 R J R = J R sin (4t) 1 x2 1 subject to the initial conditions x0 = J 0 R m, 0.1 1 v0 = J R m>s 0 using numerical integration. *4.104. Consider the following system excited by a pulse of duration 0.1 s (in newtons): 2 J 0 $ 0 x 0.3 R J $1 R + J 1 x2 - 0.05 # - 0.05 x 3 R J #1 R + J 0.05 x2 -1 -1 x1 RJ d x2 1 0 = J R [Φ(t - 1) - Φ(t - 1.1)] 1 and subject to the initial conditions x0 = J 0 R m, - 0.1 0 v0 = J R m>s 0 Compute and plot the response of the system using numerical integration. Here Φ indicates the Heaviside step function introduced in Section 3.2. *4.105. Use numerical integration to compute and plot the time response of the system (newtons) J 5 0 $ 0 x 3 R J $1 R + J 1 x2 - 0.5 # - 0.5 x 30 R J #1R + J 0.5 x2 -1 - 1 x1 1 R J R = J R sin (4t) 1 x2 1 Chap. 4 433 Problems subject to the initial conditions x0 = J 0 R m, 0.1 1 v0 = J R m>s 0 *4.106. Use numerical integration to compute and plot the time response of the system (newtons) 4 0 ≥ 0 0 0 3 0 0 0 0 2.5 0 $ 0 x1 4 $ 0 x2 -1 ¥≥$ ¥ + ≥ 0 x3 0 $ 6 x4 0 -1 2 -1 0 500 -100 + ≥ 0 0 0 -1 2 -1 - 100 200 - 100 0 # 0 x1 # 0 x2 ¥≥ # ¥ - 1 x3 # 1 x4 0 x1 0 0 x2 0 ¥ ≥ ¥ = ≥ ¥ sin (4t) -100 x3 0 1 100 x4 0 - 100 200 - 100 subject to the initial conditions 0 0 T m, x0 = D 0 0.01 1 0 v0 = D T m>s 0 0 MATLAB ENGINEERING VIBRATION TOOLBOX If you have not yet used the Engineering Vibration Toolbox program, return to the end of Chapter 1 or Appendix G for a brief introduction to using MATlAb files. TOOLBOX PROBLEMS TB4.1. Calculate the natural frequencies and mode shapes of the system of Example 4.1.5 using file VTb4_1. TB4.2. Recalculate Example 4.2.6 using file VTb4_1 and compare your answer with that of ∙ the example obtained with a calculator. Verify that PT KP = Λ and PTP = I. TB4.3. Consider Example 4.3.1 and investigate the effect of the initial condition on the response by using file VTb4_2 and plotting the responses to the following initial displacements (initial velocities all zero): 1 x0 = J R 1 0 x0 = J R 1 x0 = J 1 R -1 x0 = J 0.1 R 0.9 x0 = J 0.2 R 0.8 etc. Discuss the results by formulating a short paragraph summarizing what you observed. 434 Multiple-Degree-of-Freedom Systems Chap. 4 TB4.4. Check the calculation of Example 4.4.2. TB4.5. Using file VTb4_2, examine the effect of increasing the mass m4 in the building vibration problem of Example 4.4.3. Do this by doubling m4 and recalculating the solution. Notice what happens to the various responses. Try doubling m4 until the response does not change or the program fails. Discuss your observations. TB4.6. Consider the system of Example 4.7.3. Choose the values m = 10, J = 5, e = 1, and k1 = 1000 and calculate the eigenvalues as k2 varies from 10 to 10,000, in increments of 100. What can you conclude? 5 Design for Vibration Suppression This chapter introduces the techniques useful in designing structures and machines so they vibrate as little as possible. Often this happens after a product is designed, prototyped, and tested. in many cases, vibration problems are found late in the process so redesign is often needed. in this chapter the important concepts of shock and vibration isolation and vibration absorption are introduced, as they are key methods in vibration design. Optimization as a design tool, use of adding damping and the concept of critical speeds in rotating machines are also introduced. The cable spacer-damper used on power lines and pictured at the top is used to cut down the cable whistling caused by vibration induced by moderate winds and stop the conductors from hitting one another in strong winds (or resonance) as the cables vibrate. The ideas introduced in the chapter are fundamental to the design of such devices. The cables themselves are modeled in Chapter 6. The suspension system of any ground vehicle is an example a common design problem. a spring and damper system for an offroad vehicle is shown in the picture at the bottom. a vehicle’s size, mass, and road conditions all affect the ability of the suspension system to perform its function of isolating passengers from shock and vibration induced by road conditions and speed. 435 436 Design for Vibration Suppression Chap. 5 In this chapter, it is assumed that vibration is undesirable and is to be suppressed. The topics of the previous chapters present a number of techniques and methods for analyzing the vibration response of various systems subject to various inputs. Here the focus is to use the skills developed in the preceding chapters to determine ways of adjusting the physical parameters of a system or device in such a way that the vibration response meets some specified shape or performance criteria. This is called design; design was introduced in Section 1.7, and it is the focus of this chapter. Vibration can often lead to a number of undesirable circumstances. For example, vibration of an automobile or truck can lead to driver discomfort and, eventually, fatigue. Structural or mechanical failure can often result from sustained vibration (e.g., cracks in airplane wings). Electronic components used in airplanes, automobiles, machines, and so on may also fail because of vibration, shock, and>or sustained vibration input. The “fragility level” of devices, how much vibration a given device can withstand, is addressed by the International Organization of Standards (ISO) as well as by some national agencies. Almost every device manufactured for use by the military must meet certain military specifications (“mil specs”) regarding the amount of vibration it can withstand. In addition to government and international agencies, manufacturers also set desired vibration performance standards for some products. If a given device does not meet these regulations, it must be redesigned so that it does. This chapter presents several formulations that are useful for designing and redesigning various devices and structures to meet desired vibration standards. Design is a difficult subject that does not always lend itself to simple formulations. Design problems typically do not have a unique solution. Many different designs may all give acceptable results. Sometimes a design may simply consist of putting together a number of existing (off-the-shelf) devices to create a new device with desired properties. Here we focus on design as it refers to adjusting a system’s physical parameters to cause its vibration response to behave in a desired fashion. 5.1 ACCEPTABLE LEVELS OF VIBRATION To design a device in terms of its vibration response, the desired response must be clearly stated. Many different methods of measuring and describing acceptable levels of vibration have been proposed. Whether or not the criteria should be established in terms of displacement, velocity, or acceleration, and exactly how these should be measured needs to be clarified before a design can begin. These choices often depend on the specific application. For instance, in practice, it is generally accepted that the best indication of potential structural damage is the amplitude of the structure’s velocity, while acceleration amplitude is the most perceptible by humans. Some common ranges of vibration frequency and displacement are given in Table 5.1. Sec. 5.1 437 Acceptable levels of Vibration TABLE 5.1 RAngES oF FREquEnCy AnD DISPlACEMEnt oF VIbRAtIon Atomic vibration Threshold of human perception Machinery and building vibration Swaying of tall buildings Frequency (Hz) Displacement amplitude (mm) 1012 1–8 10–100 1–5 10 -7 10 -2 10 -2–1 10–1000 The ISO (International Organization for Standardization, www.iso.org) provides a published standard of acceptable levels of vibration that has the intent of providing a mechanism to facilitate communications between manufacturers and consumers. The standards are tested in terms of root mean-square (rms) values of displacement, velocity, and acceleration. Recall that the value (defined in Section 1.2) is the square root of the time average of the square of a quantity. For the displacement x(t), the rms value is given in equation (1.21) as xrms = c Tlim S∞ 1>2 1 x2(t)dt d T L0 T A convenient way to express the acceptable values of vibration allowed under ISO standards is to plot them on a nomograph, as illustrated in Figure 5.1. Several examples and further details of nomographs can be found in Niemkiewicz, J. (2002). The nomograph of Figure 5.1 is a graphical representation of the relationship among displacement, velocity, acceleration, and frequency for an undamped singledegree-of-freedom system. Figure 5.1 is representative and is based on vibration induced around machines as an example of how ISO attempts to classify vibration levels. Acceptable vibration levels are then stated in terms of all three physical responses: displacement, velocity, and acceleration, as well as the frequency. The solution for the displacement is given by equation (1.19) as x(t) = A sin ωnt (for zero phase), which has amplitude A. Differentiating the displacement solution yields the velocity # v(t) = x(t) = Aωn cosωnt which has amplitude ωnA. Differentiating again yields the acceleration $ a(t) = x(t) = -Aω2n sin ωnt which has amplitude Aω2n. These three expressions for the magnitude, along with the definition of rms value, allow the nomograph of Figure 5.1 to be constructed. 438 Design for Vibration Suppression Velocity (mm/s) (peak) 2000 100 80 60 40 Chap. 5 20 Displacement (mm) (peak) 1000 10 800 8 600 6 400 4 2 200 2 4 6 8 10 Frequency (Hz) 1000 10000 8000 2000 4000 6000 Acceleration (mm/s2) (peak) Figure 5.1 An example of a nomograph for specifying acceptable levels of vibration. From equation (1.21) the average value of x(t) = A sinωnt is 1 A2 1 A2 A2 x∙2 = lim A2 sin 2 ωntdt = lim c ωnT (sin ωnT cos ωnT) d = 2 Tωn 2 T S ∞ T L0 T S ∞ 2Tωn 2 so that x∙ = A> 12. likewise, v ∙ = Aωn > 12 and α ∙ = Aωn > 12. T Example 5.1.1 A machine part is subject to an rms amplitude of vibration of 6 mm. The mass and stiffness are measured to be 5 kg and 20,000 Nm, respectively. Use Figure 5.1 to determine if there is any concern with these values based on the acceptable region marked on the figure. Determine the velocity and acceleration experienced by the part. If the standard represented by Figure 5.1 is not met, then suggest a means of redesigning the bearing cap so that the response does meet the standard. Sec. 5.1 439 Acceptable levels of Vibration Solution The natural frequency of a simple spring–mass system is ωn = ωn k 20000 N>m = = 63.246 rad/s, fn = = 10.066 Hz Am A 2π 5 kg This frequency is outside the acceptable range and must be adjusted. Thus, the standard is not met and some redesign is required. In order to bring the frequency to an acceptable range, the mass or stiffness of the bearing must be changed to reduce the frequency. For instance, if a 45-kg mass is added to the bearing, and a 2500-N>m spring is added in series with the bearing stiffness (recall Figure 1.31), the resulting frequency is ωn = 1 1>k + 1>200 H m + 2 = 198.02 N>m A 7 kg = 37.231 rad>s, fn = ωn ≈ 6 Hz 2π This brings the design in line with the acceptable frequency range, but is quite a reduction in stiffness, which may not be acceptable for other reasons. looking at the chart, the 6 mm line crosses 6 Hz inside the box, at an acceleration of about 8000 mm>s2 and a velocity of about 200 mm>s. Using the formulas to get a more precise value yields v ∙ = ωnx∙ = (37.231 rad>s)(6 mm) = 223.4 mm>s 2 3 2 2 ∙a = ωnx∙ = (37.231 rad>s) (6 mm) = 8.317 * 10 mm>s While the redesign achieves the goal of reducing the vibration levels of the bearing, it is an order-of-magnitude change in the bearing stiffness. This could have negative effects on other parts of the machine (see critical speeds in Section 5.7, for instance). This example is what makes design so difficult and challenging. Changing the design to meet one specification could cause another specification to be violated. n The design procedure just suggested is oversimplified but helps introduce the ad hoc nature of many design problems. Analysis is used as a tool. Here the desired vibration criteria are provided by an ISO standard represented in a nomograph. This plot, together with the formula ωn = 1k>m and the series spring formula, provides the analysis tools. In using any thought process to perform a design, it is important to think through potential flaws in the procedure. In the preceding example there are several possible points of error. One important issue is how well the simple singledegree-of-freedom spring–mass model captures the dynamics of the part. It might be that a more sophisticated multiple-degree-of-freedom model is required (recall Example 4.8.2). Another possible problem with a proposed design change is that the stiffness of the part may not be allowed to be lower than a certain value because of load requirements (static deflection) or other design constraints. In this case, the mass might be changed, but that too may have other constraints on it. In some cases it just may not be possible to design a system to have this desired vibration response. That is, not all design problems have a solution. 440 Design for Vibration Suppression Chap. 5 The range of vibrations with which an engineer is concerned is usually from about 10 -4 mm at between 0.1 and 1 Hz for objects such as optical benches or sophisticated medical equipment, to a meter displacement for tall buildings in the range of 0.1 to 5 Hz. Machine vibrations can range between 10 and 1000 Hz, with deflections between fractions of a millimeter and several centimeters. As technology advances, limitations and acceptable levels of vibration change. Thus these numbers should be considered as rough indications of common values. Example 5.1.2 Calculate and compare the natural frequency, damping ratio, and damped natural frequency of the single-degree-of-freedom model of a stereo turntable and of the automobile given in Figure 5.2. Also plot and compare their frequency response functions and their impulse response functions. Discuss the similarities and differences of these two devices. Solution To calculate the undamped natural frequency, damping ratio, and damped natural frequency of the car is simple. From the definitions ωn = ζ = k 400,000 = = 20 rad/s Am A 1000 c 8000 = = 0.2 2mωn 2(1000)(20) ωd = ωn 21 - ζ2 = 2021 - (0.2)2 = 19.5959 rad>s The same calculations for the stereo turntable are ωn = ζ = k 400 = = 20 rad>s Am A 1 c 8 = 0.2 = 2mωn 2(1)(20) ωd = ωn 21 - ζ2 = 2021 - (0.2)2 = 19.5959 rad>s m  1000 kg k  400,000 N/m m  1 kg c  8000 N•s/m c  8 N•s/m Automobile model k  400 N/m Stereo turntable model Figure 5.2 A single-degree-of-freedom model of an automobile and a stereo turntable, each with the same frequency. 441 Acceptable levels of Vibration Magnitude (dB) 40  10–6 80 Car 120 160 0.1 2 h(t) (m) Turntable 4 6 2 1 (  20) 4 6 r 10 20 Car 0 t 20 0.0 Phase (deg) 0.2 0.4 0.6 0.8 1.0 h(t) (m) 0 50 100  10–3 Sec. 5.1 Car 20 Turntable 0 150 0.1 2 4 6 1 2 4 6 r 10 t 20 0.0 0.2 0.4 0.6 0.8 1.0 Figure 5.3 The frequency response function and the impulse response of that for the car and stereo turntable of Figure 5.2. This illustrates that two objects of very different size can have the same natural frequencies and damping ratios. In Figure 5.3 the transfer function of each device is plotted as well as the impulse response function for each. Note that these plots do indicate a difference in devices. The phase plots of the transfer function of both the car and the stereo are identical, while the magnitude plots differ by a constant. The impulse response function of the car has a smaller amplitude, although the responses both die out at the same time since the decay rate (ζ) and hence log decrement are the same for each device. The acceptable levels of vibration for the car will be much larger than those of the stereo. For instance, a displacement amplitude of 10 mm for the stereo would cause its needle to skip out of the groove in a record and, hence, not perform properly (a reason why phonograph records never made it in automobiles—thank goodness for bluetooth technology and mp3 players). On the other hand, a similar amplitude of vibration for the car is below the perception level on a nomograph. n An important consideration in specifying vibration response is to specify the nature of the input or driving force that causes the response. Disturbances, or inputs, are normally classified as either shock or as vibration, depending on how long the input lasts. An input is considered to be a shock if the disturbance is a sharp, aperiodic one lasting a relatively short time. In contrast, an input is considered to be a vibration if it lasts for a long time and has some oscillatory features. The distinction between shock and vibration is not always clear as the sources of shock and vibration disturbances are numerous and very difficult to place into categories. In Chapter 2 only vibration inputs (e.g., harmonic inputs) are discussed. In 442 Design for Vibration Suppression Chap. 5 Input force 200 N 10 100 Time (s  103) Figure 5.4 A sample of a shock input to an automobile, illustrating that the form is not entirely known—that only the bounds of the force’s magnitude and time history are known. Chapter 3 the response of a single-degree-of-freedom system to a variety of inputs, including an impulse, which is a shock, and general periodic inputs (vibration) are discussed. These input signals may result from bumps in the road (for cars), turbulence (for airplanes), rotating machinery, or simply from dropping something. Often, inputs are a combination of the types just discussed and those discussed in Chapter 3. In addition, inputs are often not known precisely but rather are known to be of less than a certain magnitude and lasting less than a certain time. For instance, a given shock input to an automobile due to its hitting a bump may take the form of a single-valued curve falling somewhere inside the shaded region of Figure 5.4. Vibrations that are not harmful exist in many devices. For instance, automobiles continually experience vibration without being damaged or causing harm to passengers. However, some vibrations are extremely damaging, such as severe vibration from an earthquake or a badly out-of-balance tire on a car. The difficult issue for design engineers is deciding between acceptable levels of vibration and those that will cause damage or become so annoying that consumers will not use the device. Once acceptable levels are established, several techniques can be used to limit and alter the shock and vibration response of mechanical systems and structures. These are discussed in the following section. 5.2 VIBRATION ISOLATION The most effective way to reduce unwanted vibration is to stop or modify the source of the vibration. If this cannot be done, it is sometimes possible to design a vibration isolation system to isolate the source of vibration from the system of interest or to isolate the device from the source of vibration. This can be done by using highly damped, compliant materials, such as rubber, to change the stiffness and damping between the source of vibration and the device that is to be protected from the vibrations. The problem of isolating a device from a source of vibration is analyzed in terms of reducing vibration displacement transmitted through base motion, as discussed in Section 2.3 and summarized in Window 5.1. The problem of isolating a source of vibration from its surroundings is analyzed in terms of reducing the force transmitted by the source through its mounting points, as discussed in Section 2.4 and summarized in Window 5.1. both force transmissibility and displacement transmissibility are called isolation problems. Sec. 5.2 443 Vibration Isolation Window 5.1 Summary of Vibration Isolation Formulas for Both Force and Displacement Transmissibility The moving-base model on the left is used in designing isolation to protect the device from motion of its point of attachment (base). The model on the right is used to protect the point of attachment (ground) from vibration of the mass. Displacement Transmissibility Force Transmissibility F(t)  F0 sin (t) x(t)  X sin (nt  ) Device m k Vibration isolator c x(t) Source of m vibration k Vibration isolator c y(t)  Y sin (t) Moving base (source of vibration) Vibration source modeled as base motion 1/2 2 1  (2r) 2 FT  F0 2 (1  r 2)  (2r) FT 1  (2r) 2  r 2 2 kY (1  r 2)  (2r) 1/2 2 1  (2r) 2 2 (1  r2)  (2r) defines the force transmissibility for isolating the source of vibration as derived in Section 5.2. defines the displacement transmissibility and is plotted in Figure 2.13. From equation (2.77), 2 Vibration source mounted on isolator Here F(t)  F0 sin t is the disturbance and Here y(t)  Y sin t is the disturbance and from equation (2.71) X  Y Fixed base 1/2 defines the related force transmissibility and is plotted in Figure 2.15. The analysis tool used to design isolators is the concept of force and >or displacement transmissibility introduced in Section 2.4. by way of review, consider the problem of calculating the transmissibility ratio, denoted T.R., defined as the ratio of the magnitude of the force transmitted through the spring and dashpot to the fixed base to the sinusoidal force applied (see Window 5.1) by the machine (modeled as a mass). Symbolically, T.R. = FT >F0. To calculate the value of T.R., first consider the force transmitted. The force transmitted to the fixed base in 444 Design for Vibration Suppression Chap. 5 Window 5.1 is denoted by FT(t) and is the force applied to the base acting through the spring and dashpot, that is, # FT (t) = kx(t) + cx(t) (5.1) The solution for the case that the driving force is harmonic of the form F0 cos ωt is given in Section 2.2, equation (2.37), to be of the form x(t) = Ae -ζωnt sin (ωdt + θ) + X cos (ωt - ϕ) In steady state (i.e., after some time has elapsed), the first term decays to zero and the response is modeled by x(t) = X cos (ωt - ϕ) (5.2) Differentiating equation (5.2), the velocity in steady state becomes # x(t) = -ωX sin (ωt - ϕ) (5.3) Substitution of equations (5.2) and (5.3) into equation (5.1) for the force transmitted at steady state yields FT(t) = kX cos (ωt - ϕ) - cωX sin (ωt - ϕ) (5.4) = kX cos (ωt - ϕ) + cωX cos (ωt - ϕ + π>2) The magnitude of FT(t), denoted by FT, can be calculated from equation (5.4) by noting that the two terms are 90°(π>2) out of phase with each other and hence can be thought of as two perpendicular vectors (recall Figure 2.11 of Section 2.3). Hence the magnitude of FT(t) is calculated by taking the vector sum of the two terms on the right of equation (5.4). This yields that the magnitude of the transmitted force is FT = 3(kX)2 + (cωX)2 = X 3k 2 + c 2ω2 (5.5) Window 5.2 Review of the Steady-State Response of an Underdamped System Subject to Harmonic Excitation as Discussed in Section 2.2 The steady-state response of $ # x + 2ζωnx + ω2nx = f0cos ωt where ωn = 2k>m, ζ = c>(2mωn) and f0 = F0 >m, is x(t) = X cos (ωt - ϕ). Here X = f0 2(ω2n - ω ) + (2ζωnω ) 2 2 2 , ϕ = tan-1 2ζωnω ω2n - ω2 Sec. 5.2 445 Vibration Isolation The value of X, the amplitude of vibration at steady state, is given in Window 5.2 to be X = 3( ) + ( 2ζωnω ) 4 f0 ω2n 2 2 - ω 2 1>2 = 3( 1 - r F0 >k ) + ( 2ζr ) 24 1>2 2 2 where r = ω>ωn, as before. Substituting this value of X into equation (5.5) yields FT = F0 = 3 (1 3 (1 - r F0 >k ) + ( 2ζr ) 2 4 1>2 2 2 31 + c ω >k ) + ( 2ζr ) 4 2 2 - r 2 2 2k 2 + c 2ω2 (5.6) 2 2 1>2 = F0 1 + ( 2ζr ) 2 C ( 1 - r 2 ) 2 + ( 2ζr ) 2 where c2ω2 >k2 = (2mωnζ)2ω2 >k2 = (2ζr)2. The transmissibility ratio, or transmission ratio, denoted T.R., is defined as the ratio of the magnitude of the transmitted force to the magnitude of the applied force. by a simple manipulation of equations (5.6), this becomes T.R. = 1 + ( 2ζr ) 2 FT = F0 C ( 1 - r 2 ) 2 + ( 2ζr ) 2 (5.7) A comparison of this force transmissibility expression with the displacement transmissibility given in Window 5.1 indicates that they are identical. It is important to note, however, that even though they have the same value, they come from different isolation problems and hence describe different phenomena. The displacement transmissibility ratio given in the left column of Window 5.1 describes how a steady-state displacement (Y) of the base of a device mounted on an isolator is transmitted into motion of the device (X). Figure 5.5 is a plot of the T.R. for various values of the damping ratio ζ and frequency ratio r. The larger the value of T.R., the larger the amplitude of motion of the mass. These curves are useful for designing the isolators. In particular, the design process consists of choosing ζ and r, within the available isolator’s material, such that T.R. is small. Note from Figure 5.5 that if the frequency ratio r is greater than 12, the magnitude of vibration of the device is smaller than the disturbance magnitude Y and vibration isolation occurs. For r less than 12, the amplitude X increases (i.e., X is larger than Y). The value of the damping ratio (each curve in Figure 5.5 corresponds to a different ζ) determines how much smaller the amplitude of vibration is for a given frequency ratio. Near resonance, the T.R. is determined completely by the value of ζ (i.e., by the damping in the isolator). In the isolation region, the smaller the value of ζ, the smaller the value of T.R. and the better the isolation. Also note that as r is increased for a fixed ω, the value of T.R. decreases. This corresponds to increasing the mass or decreasing the stiffness of the isolator. 446 Design for Vibration Suppression 10   0.05   0.01 8 T.R.   0.1 6 4 2 1 0 Chap. 5   0.2   0.5   1.0 0 0.5 r 1.0 1.5 2 Amplification occurs 2.0 Isolation occurs Magnification of the isolation area T.R. 1.2 1.0 0.8 0.6 0.4 0.2  increasing   1.0   0.5 1.0 1.2 1.4 2 1.6 1.8 2.0 2.2 2.4   0.2 r Figure 5.5 Plot of the transmissibility ratio, T.R., indicating the value of T.R. for a variety of choices of the damping ratio ζ and the frequency ratio r. This is a repeat of Figure 2.14 and is a plot of equation (5.7). Example 5.2.1 An electronic control system for an automobile engine is to be mounted on top of the fender inside the engine compartment of the automobile as illustrated in Figure 5.6. The control module electronically computes and controls the engine timing, fuel>air mixture, and so on, and completely controls the engine. To protect it from fatigue and breakage, it is desirable to isolate the module from the vibration induced in the car body by road and engine vibration, hence the module is mounted on an isolator. Design the isolator (i.e., pick c and k) if the mass of the module is 3 kg and the dominant vibration of the fender is approximated by y(t) = (0.01) (sin35t) m. Here it is desired to keep the displacement of the module less than 0.005 m at all times. Once the design values for isolators are chosen, calculate the magnitude of the force transmitted to the module through the isolator. Solution Since it is desired to keep the vibration of the module, x(t), less than 0.005 m, the response amplitude becomes X = 0.005 m. The amplitude of y(t) is Y = 0.01 m; hence the desired displacement transmissibility ratio becomes T.R. = X 0.005 = = 0.5 Y 0.01 Examining the transmissibility curves of Figure 5.5 yields several possible solutions for ζ and ωn. A straight horizontal line through T.R. = 0.5, illustrated in Figure 5.7, crosses at several values of ζ and r. For instance, the ζ = 0.02 curve intersects the Sec. 5.2 447 Vibration Isolation Electronic control module x(t) Module Isolator Fender Car body y(t) (b) Engine Isolation mounting x(t) m k c y(t) (a) (c) Figure 5.6 (a) A cutaway sketch of the engine compartment of an automobile illustrating the location of the car’s electronic control module. (b) A close-up of the control module mounted on the inside fender on an isolator. (c) A vibration model of the module isolator system. T.R. = 0.5 line at r = 1.73. Thus r = 1.73, ζ = 0.02 provides one possible design solution. Recalling that r = ω>ωn = 1.73 and ω = 35 rad>s, the isolation system’s natural frequency is about ωn = 35>1.73 = 20.231 rad>s. Since the mass of the module is m = 3 kg, the stiffness is (recall ωn = 1k>m) k = mω2n = (3 kg)(20.231 rad>s)2 = 1228 N>m Thus the isolation mount must be made of a material with this stiffness (or add a stiffener). The damping ratio ζ is related to the damping constant c by equation (1.30): c = 2ζmωn = 2(0.02)(3 kg)(20.231 rad>s) = 2.428 kg>s  0.01 0.05 0.1 0.2 0.5 1  1.2 1.0 r 1.73231 1.73855 1.75803 1.83580 2.35401 3.76976 0.8 T.R. 0.6 = 1 0.4 = 0.5 = 0.2 0.2 0.0 r 0 1 2 3 4 Figure 5.7 The transmissibility curve of Figure 5.6, repeated here, indicating possible design solutions for Example 5.2.1. Each point of intersection with one of the curves of constant ζ yields the desired T.R. 448 Design for Vibration Suppression Chap. 5 These values for c and k, together with the geometric size of the module and fender shape, can now be used to choose the isolation mount material. At this point, the designer would look through vendor catalogs to search for existing isolation mounts and materials that have approximately these values. If none exactly meet these values, the curves in Figure 5.7 are consulted to see if one of the other solutions corresponds more closely to an existing isolation material. Of course, equation (2.71) or (5.7) can be used to calculate solutions lying in between those illustrated in Figure 5.7. If many solutions are still available after a search of existing products is made, the choice of a mount can be “optimized” by considering other functions, such as cost, ease of assembly, temperature range, reliability of vendor, availability of product, and quality required. Eventually, a good design must consider all of these aspects. The electronic module may also have a limit on the amount of force it can withstand. One way to estimate the amount of force is to use the theory developed in Section 2.4, in particular in equation (2.77), which is reproduced in Window 5.1. This expression relates the force transmitted to the module by the motion of the fender through the isolator. Using the values just calculated in equation (2.77) yields FT = kYr 2 c 1 + (2ζr)2 2 2 (1 - r ) + (2ζr) 2 d 1>2 = kYr 2 (T.R.) = (1228 N>m)(0.01 m)(1.73)2(0.5) = 18.375 N If this force happens to be too large, the design must be redone. With the maximum force transmitted as an additional design consideration, the curves of Figure 2.15 must also be consulted when choosing the values of r and ζ to meet the required design specifications. The static deflection caused by this design is δ = mg>k = 0.024 m. The static deflection and the ratio X>Y define the rattle space or physical dimensions needed for the isolator to vibrate in. In a car, the 2.4 cm might be acceptable. If the application were to isolate a CD drive in a laptop computer, this distance would be unacceptable because it is large compared to the thickness of the laptop. Static deflection and rattle space are important design considerations and often limit the ability to design a good isolator. n Example 5.2.1 may have seemed very reasonable. However, many assumptions were made in reaching the final design, and all of these must be given careful thought. For example, the assumption that the motion of the fender is harmonic of the form y(t) = (0.01) sin 35t is very restrictive. In reality, it is probably random, or at least the value of ω varies through a range of frequencies. This is not to say that the solution presented in Example 5.2.1 is useless, just that the designer should keep in mind its limitations. Even though the real input to the system is random, the chosen amplitude of Y = 0.01 m and frequency of ω = 35 rad>s might represent a deterministic bound on all possible inputs to the system (i.e., all other disturbance amplitudes may be smaller than Y = 0.01 m, and all other driving frequencies might be larger than ω = 35 rad>s). Hence in many practical cases the designer is faced with choosing an isolator that will protect the part from, say, 5 g between 20 and 200 Hz, or the designer will be given a plot of PSD versus ωr (recall Section 3.5) Sec. 5.2 449 Vibration Isolation and try to design the isolator to service these types of inputs. Section 5.9 examines the isolation problem from the practical aspect of working with manufacturers of isolation products. The design of shock isolation systems is performed by examining the shock spectrum, as introduced in Section 3.6. To make the comparison to vibration isolation clear, the shock spectrum is reconsidered here as a plot of the ratio of the maximum motion of the response acceleration amplitude (i.e., ω2X) to the disturbance $ acceleration amplitude [i.e., the amplitude of y(t)] versus the product of the natural frequency and the time duration of the pulse, t1. Here the disturbance y(t) is modeled as a half sinusoid of the form Y sin ωpt y(t) = d 0 … t … t1 = π ωp (5.8) π t 7 t1 = ωp 0 as indicated in Figure 5.8. This type of disturbance is often called a shock pulse. The frequency ωp and the corresponding time t1 = π>ωp determine how long the shock pulse lasts. The product ωnt1 is used for plotting shock transmissibility rather than plotting the frequency ratio used to design vibration isolators. A plot of the acceleration amplitude ratio versus the product ωt1 is given in Figure 5.8. This figure is determined by calculating the acceleration amplitude of the response and comparing it to the magnitude of the acceleration of the input disturbance. Note that as the abscissa increases, corresponding to a longer pulse width, the acceleration experienced by the module is larger than the input acceleration. by examining the plots in Figure 5.8, it can be seen that reduction of the acceleration  0  0.01  0.1 .. |x|max .. |y|max  0.3 Input 1 y(t)  0.5  1.0 t t1 Response of half-sine shock pulse 0.5 1.0 np  nt1 Figure 5.8 A plot of the ratio of output acceleration magnitude to input acceleration magnitude versus a frequency ratio 1ωn >ωp2 for a single-degree-of-freedom system and a base excitation consisting of a shock pulse for different values of damping ratios. Note that a large ωp value corresponds to a short pulse width. 450 Design for Vibration Suppression Chap. 5 through isolation occurs only if the transmissibility falls below the horizontal line passing through the number 1. So, for instance, consider the line defined by ζ = 0.5. For shock isolation to occur, this would require ωnt 1 mπ2 6 1.0, or k 6 2 π t1 Thus shock isolation enforces a bound on the stiffness of the isolation material. Next consider the effects of damping on shock isolation. Examination of the plots in Figure 5.8 shows that increasing the damping greatly reduces the maximum acceleration to the point that for critical damping (ζ = 1.0) isolation occurs for any pulse duration (t1). Thus, good shock isolators require high damping. Next consider the problem of isolating a source of harmonic vibration from its surroundings. This is the fixed-base isolation problem illustrated in Window 5.1, where the right side is concerned with reducing the force transmitted through the isolator due to harmonic excitation at the mass. The common example is a rotating machine generating a harmonic force at a constant frequency (recall Figure 2.18). Examples of such machines are electric motors, steam turbines, internal combustion engines, generators, washing machines, and disk drive motors. Examination of the transmissibility curve of Figure 5.5 indicates that vibration isolation begins for isolation stiffness values such that ω>ωn 7 12. This gives transmissibility ratios of less than 1, so that the force transmitted to ground is less than the force generated by the rotating machine. Since the mass is usually fixed by the nature of the machine, the isolation mounts are generally chosen based on their stiffness, so that r 7 12 is satisfied. If this does not give an acceptable solution (for low-frequency excitation), mass can sometimes be added to the machine 1ωn = 1k>m 2 . Since r = ω> 1k>m, lower stiffness values correspond to larger values of r, which yields better isolation (lower T.R. values). As damping is increased for a fixed r, the value of T.R. increases, so that low damping is often used. However, some damping is desirable, since when the machine starts up and causes a harmonic disturbance through a range of frequencies, it generally passes through resonance (r = 1) and the presence of damping is required to reduce the transmissibility at resonance. Examination of the transmissibility curve indicates that for a large enough frequency ratio (about r 7 3) and small enough damping (ζ 6 0.2) the T.R. value is not affected by damping. Since most springs have very small internal damping (e.g., less than 0.01), the term (2ζr)2 is very small [e.g., for r = 3, (2ζr)2 = 0.0036]. Hence it is common to design a vibration isolation system by neglecting the damping in Sec. 5.2 451 Vibration Isolation equation (5.7). In this case T.R. becomes (taking the negative square root for positive values of T.R.) T.R. = 1 r - 1 (r 7 3) 2 (5.9) Equation (5.9) can be used to construct design charts for use in choosing vibration isolation pads for mounting rotating machinery. The driving frequency of a machine is usually specified in terms of its speed of rotation, or revolutions per minute (rpm). If n is the motor speed in rpm, ω = 2πn 60 (5.10) In addition, springs are often classified in terms of their static deflection defined by ∆ = W>k = mg>k, where m is the mass of the machine and g the acceleration due to gravity. It has become very common to design isolators in terms of the machine’s rotating speed n and the static deflection ∆. A third quantity, R, defined as the reduction in transmissibility, R = 1 - T.R. (5.11) is commonly used to quantify the success of the vibration isolator. Substitution of the (undamped) value of T.R. into equation (5.11) and solving for r yields r = ω 2k>m = 2 - R A1 - R (5.12) Substitution of (5.10) for ω and replacing k = mg> ∆ yields n = 30 g(2 - R) 2 - R = 29.9093 π B ∆(1 - R) B ∆(1 - R) (5.13) which relates the motor speed to the reduction factor and the static deflection of the spring. Equation (5.13) can be used to generate design curves, by taking the log of the expression. This yields 1 2 - R log n = - log ∆ + log a29.9093 b 2 A1 - R (5.14) which is a straight line on a log–log plot for each value of R. This expression is then used to provide the design chart of Figure 5.9, consisting of plots of motor speed versus static deflection. 452 Design for Vibration Suppression Chap. 5 Excitation frequency, n (rpm or cpm) 50,000 20,000 10,000 5,000 2.0 3.0 1.5 2,000 1.2 1.1 1,000 1.05 500 Percent reduction 98 97 95 93 90 85 80 70 20 50 Amplification 200 100 0.002 0.001 0.005 0.01 0.02 0.05 0.1 0.2 0.5 1 Static deflection,  (in.) Figure 5.9 Design curves consisting of plots of running speed versus static deflection (or stiffness) for various values of percent reduction in transmitted force. Example 5.2.2 Consider the computer disk drive of Figure 5.10. The disk drive motor is mounted to the computer chassis through an isolation pad (spring). The motor has a mass of 3 kg and operates at 5000 rpm. Calculate the value of the stiffness of the isolator needed to provide a 95% reduction in force transmitted to the chassis (considered as ground). How much clearance is needed between the motor and the chassis? Solution From the chart of Figure 5.9, the line corresponding to a speed of n = 5000 rpm hits the curve corresponding to 95% reduction at a static deflection of 0.03 in. or 0.0762 cm. This corresponds to a spring stiffness of k = Motor Isolator Computer chassis mg ∆ = 1 3 kg)(9.8 m>s2 2 0.000762 m = 38,582 N>m Figure 5.10 A schematic model of a personal computer illustrating the motor running the disk drive system. A small amount of out-of-balance in the motor can transmit harmonic forces to the chassis and onto circuit boards and other components if not properly isolated. Sec. 5.2 Vibration Isolation 453 The choice of clearance (i.e., distance needed between the motor and the chassis) should be more than twice the static deflection so that the spring has room to extend and compress, providing isolation. n The issue of vibration isolation against harmonic forces for large (heavy) machines quickly becomes one of static deflection. For machines requiring extreme isolation, coiled springs often must be used to provide the large static deflection required at low frequency. This can be seen by examining the 98% curves in Figure 5.9 for low values of n. In some cases, the static deflection required may be too large to be physically obtainable even in small devices. An example of a similar design constraint is the miniaturization of computers. Manufacturers of laptop computers believe sales are tied to how compact and, in particular, how thin the chassis can be. One constraint could be the isolation system required for the disk drive motor or other components. In designing isolators, it is often difficult to design an isolator that works effectively against both shock and vibration (harmonic) excitations. One reason for this can be seen by examining Figure 5.7 for vibration isolation and Figure 5.8 for shock isolation. In Figure 5.7, isolation occurs in the region r 7 12, and in this region it is clear that increased damping reduces the effective isolation. Thus, low damping is required for vibration isolation. However, examining the shock isolation plots in Figure 5.8 shows that large damping is required for effective shock isolation. These two requirements are often at odds, as the following example points out. Example 5.2.3 In this example the design goal is to develop an effective isolator for base-induced shock motion that will also provide an acceptable level of vibration isolation in terms of force transmissibility from the vibrating equipment supported by the isolation system to the base. In actual isolation applications it is often difficult to design an effective combined shock and vibration isolation system. As mentioned earlier and is clear from Figures 5.7 and 5.8, this is due to the fact that an effective vibration isolator must be very lightly damped, whereas an effective shock isolator tends to require large damping forces. Typically in design problems, a main constraint is that parts such as isolators are available only in discrete values of stiffness and damping. Here we suppose that a set of three off-the-shelf isolators (mounts) are available for use, that their natural frequencies are 5 Hz, 6 Hz, and 7 Hz, and all have 8% damping. The shock input that is being isolated is a 15-g, half-sine pulse as shown in Figure 5.8 for the case t1 = 40 ms. With this as input, it is desired to limit the mount response to 15 g’s and the mount deflection to 3 in. (76.2 mm). The vibration isolation goal is 20 db of isolation from an abovemount vibration source of 15 Hz. Figures 5.11 and 5.12 show the time simulation results to shock input in terms of mount deflection and above-mount response for the three mount possibilities. From Figures 5.11 and 5.12, it is clear that only the 7-Hz mount satisfies all the shock isolation goals, that is, 6 15 g’s above mount and 6 3 in. of deflection. 454 Design for Vibration Suppression Chap. 5 Relative Displacement 4 5 Hz Mount 6 Hz Mount 7 Hz Mount 3 Displacement (in) 2 1 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 1 2 3 4 5 Time (s) Figure 5.11 A simulated response [relative displacement, z(t)]. Absolute Acceleration 20 5 Hz Mount 6 Hz Mount 7 Hz Mount 15 Acceleration (g’s) 10 5 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 5 10 15 Time (s) $ Figure 5.12 A simulated response [absolute acceleration, x(t)]. Now consider the vibration isolation performance using the same mount. A minimum of 20 db of vibration isolation to the 15-Hz above-mount vibration source is desired. Recall that the mount has an inherent damping that can adequately be Sec. 5.3 455 Vibration Absorbers Force Transmissibility 10 1 0 0.5 1 1.5 2.5 3 3.5 4 T.R.  9.3 dB 0.1 0.01 Frequency ratio Figure 5.13 Force transmissibility (above-mount source to base). approximated as 8% viscous damping. To determine the vibration isolation performance, the force transmissibility is plotted in Figure 5.13 for an 8% damped isolator. The frequency ratio of interest is r = ω>ωn = 2 # π # 15>2 # π # 7 ≈ 2.1 for this example. Recall that a minimum of 20 db of isolation is required, but from Figure 5.13 only about 9.4 db of isolation is achieved. To achieve the required vibration isolation would require a lower damping ratio (see Figure 5.7) than is inherent to the isolator. It is not possible to lower the damping ratio, as damping is a fixed property of the isolation material. Damping can be raised by adding external dampers, adding layers of damping material, and so forth, but it cannot be decreased without significantly modifying the mount design. As such, these shock and vibration isolation design parameters cannot simultaneously be met using the devices at hand, a typical situation in design. n In designing isolation mounts, two factors are key. The first is deciding whether to design for vibration or to design for shock. The next is to check the static deflection. The following section examines the use of absorbers to reduce vibrations from harmonic disturbances. 5.3 VIBRATION ABSORBERS Another approach to protecting a device from steady-state harmonic disturbance at a constant frequency is a vibration absorber. Unlike the isolator of the previous sections, an absorber consists of a second mass–spring combination added to the 456 Design for Vibration Suppression Optical table F(t)  F0 sin(t) Primary mass x m ka Absorber system Legs Table k/2 Chap. 5 xa ma k/2 Model Figure 5.14 An optical table protected by an added vibration absorber. The table and its supporting legs are modeled as a single-degree-of-freedom system with mass m and stiffness k. primary device to protect it from vibrating. The major effect of adding the second mass–spring system is to change from a single-degree-of-freedom system to a twodegree-of-freedom system. The new system has two natural frequencies (recall Section 4.1). The added spring–mass system is called the absorber. The values of the absorber mass and stiffness are chosen such that the motion of the original mass is a minimum. This is accompanied by substantial motion of the added absorber system, as illustrated in the following: Absorbers are often used on machines that run at constant speed, such as sanders, compactors, reciprocating tools, and electric razors. Probably the most visible vibration absorbers can be seen on transmission lines and telephone lines. A dumbbell-shaped vibration absorber is often used on such wires to provide vibration suppression against wind blowing, which can cause the wire to oscillate at its natural frequency. The presence of the absorber prevents the wire from vibrating so much at resonance that it breaks (or fatigues). Figure 5.14 illustrates a simple vibration absorber attached to a spring–mass system. The equations of motion [summing forces in the vertical direction (refer to Chapter 4)] are J m 0 $ 0 x k + ka R J$ R + J ma xa -ka - ka x F sin ωt RJ R = J 0 R ka xa 0 (5.15) where x = x(t) is the displacement of the table modeled as having mass m and stiffness k, xa is the displacement of the absorber mass (of mass ma and stiffness ka), and the harmonic force F0 sin ωt is the disturbance applied to the table mass. It is desired to design the absorber (i.e., choose ma and ka) such that the displacement of the primary system is as small as possible in steady state. Here it is desired to reduce the vibration of the table, which is the primary mass. Sec. 5.3 457 Vibration Absorbers Window 5.3 Recall that the inverse of a 2 * 2 matrix A given by A = J a c b R d is defined to be A-1 = 1 d J det A -c -b R a where det A = ad - bc In contrast to the solution technique of modal analysis used in Chapter 4, here it is desired to obtain a solution in terms of parameters (m, k, ma, and ka) that can then be solved for as part of a design process. To this end, let the steady-state solution of x(t) and xa(t) be of the form x(t) = X sin ωt (5.16) xa(t) = Xa sin ωt Substitution of these steady-state forms into equation (5.15) yields (after some manipulation) J k + ka - mω2 -ka - ka X F R sin ωt = J 0 R sin ωt 2R J ka - maω Xa 0 (5.17) which is an equation in the vector [X Xa]T . Dividing by sin ωt, taking the inverse of the matrix coefficient of [X Xa]T (see Window 5.3), and multiplying from the left yields J X R = Xa = 1k 1k 2 1 ka 1 2 + ka - mω - maω 2 2 + ka - mω2 2 1 ka - maω2 2 - k 2a 1 1 ka - c k 2a c ka - maω2 ka - maω2 2 F0 d kaF0 ka F0 dc d k + ka - mω2 0 (5.18) Equating elements of the vector equality given by equation (5.18) yields the result that the magnitude of the steady-state vibration of the device (table) becomes X = 1k + ka - mω2 2 1 ka - maω2 2 - k 2a (ka - maω2)F0 (5.19) 458 Design for Vibration Suppression Chap. 5 while the magnitude of vibration of the absorber mass becomes Xa = 1k 21 ka kaF0 2 + ka - mω - maω2 2 - k 2a (5.20) Note from equation (5.19) that the absorber parameters ka and ma can be chosen such that the magnitude of steady-state vibration, X, is exactly zero. This is accomplished by equating the coefficient of F0 in equation (5.19) to zero: ω2 = ka ma (5.21) Hence if the absorber parameters are chosen to satisfy the tuning condition of equation (5.21), the steady-state motion of the primary mass is zero (i.e., X = 0). In this event the steady-state motion of the absorber mass is calculated from equations (5.20) and (5.16) with ka = ma ω2 to be xa(t) = - F0 sin ωt ka (5.22) Thus the absorber mass oscillates at the driving frequency with amplitude Xa = F0 >ka. Note that the magnitude of the force acting on the absorber mass is just kaxa = ka( - F0 >ka) = - F0. Hence when the absorber system is tuned to the driving frequency and has reached steady state, the force provided by the absorber mass is equal in magnitude and opposite in direction to the disturbance force. With zero net force acting on the primary mass, it does not move and the motion is “absorbed” by motion of the absorber mass. Note that while the applied force is completely absorbed by the motion of the absorber mass, the system is not experiencing resonance because 1ka >ma is not a natural frequency of the two-mass system. The success of the vibration absorber discussed previously depends on several factors. First, the harmonic excitation must be well known and not deviate much from its constant value. If the driving frequency drifts much, the tuning condition will not be satisfied, and the primary mass will experience some oscillation. There is also some danger that the driving frequency could shift to one of the combined systems’ resonant frequencies, in which case one or the other of the system’s coordinates would be driven to resonance and potentially fail. The analysis used to design the system assumes that it can be constructed without introducing any appreciable damping. If damping is introduced, the equations cannot necessarily be decoupled and the magnitude of the displacement of the primary mass will not be zero. In fact, damping defeats the purpose of a tuned vibration absorber and is desirable only if the frequency range of the driving force is too wide for effective operation of the absorber system. This is discussed in the next section. Another key Sec. 5.3 459 Vibration Absorbers factor in absorber design is that the absorber spring stiffness ka must be capable of withstanding the full force of the excitation and hence must be capable of the corresponding deflections. The issue of spring size and deflection, as well as the value of the absorber mass, places a geometric limitation on the design of a vibration absorber system. The issue of avoiding resonance in absorber design in case the driving frequency shifts can be quantified by examining the mass ratio μ, defined as the ratio of the absorber mass to the primary mass: μ = ma m In addition, it is convenient to define the frequencies ωp = k Am ωa = ka the natural frequency of the absorber system before it is attached to m A a the primary system the original natural frequency of the primary system without the absorber attached With these definitions, also note that ω2a ka = μ 2 = μβ2 k ωp (5.23) where the frequency ratio β is β = ωa >ωp. Substitution of the values for μ, ωp, and ωa into equation (5.19) for the amplitude of vibration of the primary mass yields (after some manipulation) Xk = F0 31 + μ 1 ωa >ωp 2 2 - 1 - ω2 >ωn2 1 ω>ωp 2 2 4 3 1 - 1 ω>ωa 2 2 - μ 1 ωa >ωp 2 2 4 (5.24) The absolute value of this expression is plotted in Figure 5.15 for the case μ = 0.25. Such plots can be used to illustrate how much drift in driving frequency can be tolerated by the absorber design. Note that if ω should drift to either 0.781 ωa or 1.28 ωa, the combined system would experience resonance and fail, since these are the natural frequencies of the combined system. In fact, if the driving frequency shifts such that 0 Xk>F0 0 7 1, the force transmitted to the primary system is amplified and the absorber system is not an improvement over the original design of the primary system. The shaded area of Figure 5.15 illustrates the values of ω>ωa such that 0 Xk>F0 0 … 1. This illustrates the useful operating range of the absorber design (i.e., 0.908 ωa 6 ω 6 1.118 ωa). Hence if the driving frequency drifts within this range, the absorber design still offers some protection to the primary system by reducing its steady-state vibration magnitude. 460 Design for Vibration Suppression Chap. 5 Xk F0 3.0 2.5 Normalized magnitude   0.25 1 2.0 1.5 1.0 0.5 0.0 0.0 0.781 0.5 1.0 0.908 1.281 1.118 1.5 2.0  a Frequency Figure 5.15 A plot of normalized magnitude of the primary mass versus the normalized driving frequency for the case μ = 0.25. The two natural frequencies of the system occur at 0.781 and 1.281. The design of an absorber can be further illuminated by examining the mass ratio μ and the frequency ratio β. These two dimensionless quantities indirectly specify both the mass and stiffness of the absorber system. The frequency equation (characteristic equation) for the two-mass system is obtained by setting the determinant of the matrix coefficient in equation (5.17) [i.e., the denominator of equation (5.18)] to zero and interpreting ω as the system natural frequency ωn. Substitution for the value of β and rearranging yields β2 a ωn2 ω2a b - 3 1 + β2(1 + μ) 4 2 ωn2 ωa2 + 1 = 0 (5.25) Sec. 5.3 which is a quadratic equation in 1 ω2n >ω2a 2 . Solving this yields Vibration Absorbers 461 a 1 + β2(1 + μ) ωn 2 1 { b = 3β4(1 + μ)2 - 2β2(1 - μ) + 1 (5.26) 2 ωa 2β 2β2 which illustrates how the system’s natural frequencies vary with the mass ratio μ and the frequency ratio β. This is plotted for β = 1 in Figure 5.16. Note that as μ is increased, the natural frequencies split farther apart, and farther from the operating point ω = ωa of the absorber. Therefore, if μ is too small, the combined system will not tolerate much fluctuation in the driving frequency before it fails. As a rule of thumb, μ is usually taken to be between 0.05 and 0.25 (i.e., 0.05 … μ … 0.25), as larger values of μ tend to indicate a poor design. Vibration absorbers can also fail because of fatigue if xa(t) and the stresses associated with this motion of the absorber are large. Hence limits are often placed on the maximum value of Xa by the designer. The following example illustrates an absorber design.  1 0.4 0.3 Figure 5.16 A plot of mass ratio versus system natural frequency (normalized to the frequency of the absorber system), illustrating that increasing the mass ratio increases the useful frequency range of a vibration absorber. Here ω1 and ω2 indicate the normalized value of the system’s natural frequencies.   0.25 0.2 0.1 1 1 2 n/a Example 5.3.1 A radial saw base has a mass of 73.16 kg and is driven harmonically by a motor that turns the saw’s blade as illustrated in Figure 5.17. The motor runs at constant speed and produces a 13-N force at 180 cycles>min because of a small unbalance in the motor. The resulting forced vibration was not detected until after the saw had been manufactured. The manufacturer wants a vibration absorber designed to drive the table oscillation to zero simply by retrofitting an absorber onto the base. Design the absorber assuming that the effective stiffness provided by the table legs is 2600 N>m. In addition, the absorber must fit inside the table base and hence has a maximum deflection of 0.2 cm. Handle Motor Blade Base Legs Figure 5.17 A schematic of a radial saw system in need of a vibration absorber. 462 Design for Vibration Suppression Chap. 5 Solution To meet the deflection requirement, the absorber stiffness is chosen first. This is calculated by assuming that X = 0, so that 0 Xaka 0 = 0 F0 0 [i.e., so that the mass ma absorbs all of the applied force, see equation (5.20) with ka = maω2]. Hence ka = F0 13 N 13 N = = = 6500 N>m Xa 0.2 cm 0.002 m Since the absorber is designed such that ω = ωa, ma = ka ω2 = 3(180>60)2π4 2 6500 N>m = 18.29 kg Note in this case that μ = 18.29>73.16 = 0.25. n Example 5.3.2 Calculate the bandwidth of operation of the absorber design of Example 5.3.1. Assume that the useful range of an absorber is defined such that 0 Xk>F0 0 6 1. For values of Xk>F0 7 1, the machine could easily drift into resonance and the amplitude of vibration actually becomes an amplification of the effective driving force amplitude. Solution From equation (5.24) with Xk>F0 = 1, 1 - a ωa 2 ωa 2 ω 2 ω 2 ω 2 b = c 1 + μa b - a b d c 1 - a b d - μa b ωa ωp ωp ωp ωp Solving this for ω>ωa yields the two solutions ω = { 11 + μ ωa For the system of Example 5.3.1, μ = 0.25, so that the second solution becomes ω = 1.1180 ωa The condition that 0 Xk>F0 0 = 1 is also satisfied for Xk>F0 = - 1. Substitution of this into equation (5.24) followed by some manipulation yields a ωa 2 ω 4 ωa 2 ω 2 b a b = c 2 + (μ + 1) a b d a b + 2 = 0 ωp ωa ωp ωa which is quadratic in (ω>ωa)2. Using the values of ω2a = 6500>18.29, ω2p = 2600>73.16, and μ = 0.25, this simplifies to 10 a ω 4 ω 2 b - 14.5 a b + 2 = 0 ωa ωa Sec. 5.4 463 Damping in Vibration Absorption Solving for ω>ωa yields a ω 2 b = 0.1544, 1.2956 or ωa ω = 0.3929, 1.1382 ωa Hence the three roots satisfying 0 Xk>F0 0 = 1 are 0.3929, 1.1180, and 1.1382. Following the example of Figure 5.15 indicates that the driving frequency may vary between 0.3929ωa and 1.1180ωa, or, since ωa = 18.857, 7.4089 6 ω 6 21.0821 (rad>s) before the response of the primary mass is amplified or the system is in danger of experiencing resonance. n The preceding discussion and examples illustrate the concept of performance robustness; that is, the examples illustrate how the design holds up as the parameter values (k, ka, etc.) drift from the values used in the original design. Example 5.3.2 illustrates that the mass ratio greatly affects the robustness of absorber designs. This is stated in the caption of Figure 5.16; up to a certain point, increasing μ increases the robustness of the absorber design. The effects of damping on absorber design are examined in the next section. 5.4 DAMPING IN VIBRATION ABSORPTION As mentioned in Section 5.3, damping is often present in devices and has the potential for destroying the ability of a vibration absorber to protect the primary system fully by driving X to zero. In addition, damping is sometimes added to vibration absorbers to prevent resonance or to improve the effective bandwidth of operation of a vibration absorber. Also, a damper by itself is often used as a vibration absorber by dissipating the energy supplied by an applied force. Such devices are called vibration dampers rather than absorbers. First consider the effect of modeling damping in the standard vibration absorber problem. A vibration absorber with damping in both the primary and absorber system is illustrated in Figure 5.18. This system is dynamically equal to the system of Figure 4.15 of Section 4.5. The equations of motion are given in matrix form by equation (4.116) as xa(t) ma ka F0 sin(t) ca x(t) Excitation force m c Absorber system k Primary system Figure 5.18 A schematic of a vibration absorber with damping in both the primary and absorber system. 464 Design for Vibration Suppression J $ 0 x(t) c + ca R J$ R + J ma xa(t) -ca m 0 +J k + ka -ka Chap. 5 # -ca x(t) RJ# R ca xa(t) -ka x(t) F RJ R = J 0 R sin ωt ka xa(t) 0 (5.27) Note, as was mentioned in Section 4.5, that these equations cannot necessarily be solved by using the modal analysis technique of Chapter 4 because the equations do not decouple (KM -1C ∙ CM -1K). The steady-state solution can be calculated, however, by using a combination of the exponential approach discussed in Section 2.3 and the matrix inverse used in previous sections for the undamped case. To this end, let F0 sin ωt be represented in exponential form by F0e jωt in equation (5.27) and assume that the steady-state solution is of the form x(t) = Xe jωt = J X jωt Re Xa (5.28) where X is the amplitude of vibration of the primary mass and Xa is the amplitude of vibration of the absorber mass. Substitution into equation (5.27) yields J 1k + ka - mω2 2 + (c + ca)ωj -ka - caωj 1 ka -ka - caωj X F d J d e jωt = J 0 d e jωt - maω2 2 + caωj Xa 0 (5.29) Note that the coefficient matrix of the vector X has complex elements. Dividing equation (5.29) by the nonzero scalar e jωt yields a complex matrix equation in the amplitudes X and Xa. Calculating the matrix inverse using the formula of Example 4.1.4, reviewed in Window 5.3, and multiplying equation (5.29) by the inverse from the right yields J X R = Xa J 1 ka - maω2 2 + caωj ka + caωj ka + caωj F R J 0R 2 k + ka - mω + (c + ca)ωj 0 det 1 K - ω2M + ωjC 2 det 1 K - ω2M + ωjC 2 = mmaω4 + 1 c ac + ma(ka + ka) + kam 2 ω2 + kak (5.30) Here the determinant in the denominator is given by (recall Example 4.1.4) + 3(kca + cka)ω - 1 ca(m + ma) + cma 2 ω34 j (5.31) Sec. 5.4 465 Damping in Vibration Absorption and the system coefficient matrices M, C, and K are given by M = J 0 R ma m 0 C = J c + ca -ca 31 ka -ca R ca K = J - maω2 2 + caωj 4 F0 k + ka -ka - ka R ka Simplifying the matrix vector product yields X = Xa = det 1K - ω2M + ωjC 2 (5.32) det 1K - ω2M + ωjC 2 ( ka + caωj ) F0 (5.33) which expresses the magnitude of the response of the primary mass and absorber mass, respectively. Note that these values are now complex numbers and are multiplied by the complex value e jωt to get the time responses. Equations (5.32) and (5.33) are the two-degree-of-freedom version of the frequency response function given for a single-degree-of-freedom system in equation (2.52). The complex nature of these values reflects a magnitude and phase. The magnitude is calculated following the rules of complex numbers and is best done with a symbolic computer code, or after substitution of numerical values for the various physical constants. It is important to note from equation (5.32) that unlike the tuned undamped absorber, the response of the primary system cannot be exactly zero even if the tuning condition is satisfied. Hence the presence of damping ruins the ability of the absorber system to exactly cancel the motion of the primary system. Equations (5.32) and (5.33) can be analyzed for several specific cases. First, consider the case for which the internal damping of the primary system is neglected (c = 0). If the primary system is made of metal, the internal damping is likely to be very low and it is reasonable to neglect it in many circumstances. In this case, the determinant of equation (5.31) reduces to the complex number = 31-mω 2 + k 2 1-maω + ka 2 - makaω 4 det(K - ω2M + ωCj) 2 2 + 31k - (m + ma)ω 2 caω 4 j (5.34) 2 The maximum deflection of the primary mass is given by equation (5.32) with the determinant in the denominator evaluated as given in equation (5.34). This is the ratio of two complex numbers and hence is a complex number representing the phase and the amplitude of the response of the primary mass. Using complex arithmetic (see Window 5.4) the amplitude of the motion of the primary mass can be written as the real number X2 = F 20 31k - mω2 2 1ka - 1ka maω2 2 - maω2 2 + ω2c 2a - makaω2 4 + 2 2 3k - (m + ma)ω2 4 c 2aω2 2 (5.35) 466 Design for Vibration Suppression Chap. 5 Window 5.4 Reminder of Complex Arithmetic The response magnitude given by equation (5.32) can be written as the ratio of two complex numbers: A1 + B1 j X = F0 A2 + B2 j where A1, A2, B1, and B2 are real numbers and j = 1-1. Multiplying this by the conjugate of the denominator divided by itself yields X = F0 ( A1 + B1j )( A2 - B2j ) ( A1A2 + B1B2 ) B1A2 - A1B2 + j = A22 + B22 A22 + B22 ( A2 + B2j )( A2 - B2j ) which indicates how X>F0 is written as a single complex number of the form X>F0 = a + bj. This is interpreted, as indicated, that the response magnitude has two components: one in phase with the applied force and one out of phase. The magnitude of X>F0 is the length of the preceding complex number (i.e., 0 X>F0 0 = 2a2 + b2). This yields ` A12 + B12 X ` = 2 F0 A2 + B22 which corresponds to the expression given in equation (5.35). (Also see Appendix A.) It is instructive to examine this amplitude in terms of the dimensionless ratios introduced in Section 5.3 for the undamped vibration absorber. The amplitude X is written in terms of the static deflection ∆ = F0 >k of the primary system. In addition, consider the mixed “damping ratio” defined by ζ = ca 2maωp (5.36) where ωp = 1k>m is the original natural frequency of the primary system without the absorber attached. Using the standard frequency ratio r = ω>ωp, the ratio of natural frequencies β = ωa >ωp (where ωa = 1ka >ma), and the mass ratio μ = ma >m, equation (5.35) can be rewritten as (2ζr)2 + 1r 2 - β22 2 X Xk = = ∆ F0 C (2ζr)2 1 r 2 - 1 + μr 2 2 2 + 3 μr 2β2 - 1 r 2 - 1 2 1 r 2 - β2 24 2 (5.37) Sec. 5.4 467 Damping in Vibration Absorption which expresses the dimensionless amplitude of the primary system. Note from examining equation (5.37) that the amplitude of the primary system response is determined by four physical parameter values: μ the ratio of the absorber mass to the primary mass β r the ratio of the decoupled natural frequencies the ratio of the driving frequency to the primary natural frequency ζ the ratio of the absorber damping and 2maωp These four numbers can be considered as design variables and are chosen to give the smallest possible value of the primary mass’s response, X, for a given application. Figure 5.19 illustrates how the damping value, as reflected in ζ, affects the response for a fixed value of μ = 0.25 and β = 1, as r varies. As mentioned at the beginning of this section, damping is often added to the absorber to improve the bandwidth of operation. This effect is illustrated in Figure 5.19. Recall that if there is no damping in the absorber (ζ = 0), the magnitude of the response of the primary mass as a function of the frequency ratio r is as illustrated in Figure 5.15 (i.e., zero at r = 1 but infinite at r = 0.781 and r = 1.281). Thus the completely undamped absorber has poor bandwidth (i.e., if r changes by a small amount, the amplitude grows). In fact, as noted in Section 5.3, the bandwidth, or useful range of operation of that undamped absorber, is 0.908 … r … 1.118. For these values of r, 0 Xk>F0 0 … 1. However, if damping is added to the absorber (ζ ∙ 0), Figure 5.19 results, and the bandwidth, or useful range of operation, is extended. The price for this increased operating region is that 0 Xk>F0 0 is never zero in the damped case (see Figure 5.19). Examination of Figure 5.19 shows that as ζ is varied, the amplification of 0 Xk>F0 0 over the range of r can be reduced. The design question now becomes: For Xk F0 60 50   0.01 40 30   0.4   0.1 20 10 0 r 0.6 0.8 1.0 1.2 1.4 Figure 5.19 The normalized amplitude of vibration of the primary mass as a function of the frequency ratio for several values of the damping in the absorber system for the case of negligible damping in the primary system [i.e., a plot of equation (5.37)]. 468 Design for Vibration Suppression Chap. 5 Xk F0 80 60   0.01 40   0.1 20 0   0.4 0.6 0.8 1.0 1.2 1.4 1.6 r Figure 5.20 Repeat of the plot of Figure 5.19 with μ = 0.25 and β = 0.9 for several values of ζ. Note that in this case, ζ = 0.4 yields a lower magnitude than does ζ = 0.1. what values of the mass ratio μ, the absorber damping ratio ζ, and the frequency ratio β is the magnitude 0 Xk>F0 0 smallest over the region 0 … r … 2? Just increasing the damping with μ and β fixed does not necessarily yield the lowest amplitude. Note from Figure 5.19 that ζ = 0.1 produces a smaller amplification over a larger region of r than does the higher ratio, ζ = 0.4. Figures 5.20 and 5.21 yield some hint of how the various parameters affect the magnitude by providing plots of 0 Xk>F0 0 for various combinations of ζ, μ, and β. A solution of the best choice of μ and ζ is discussed again in Section 5.5. Note from Figure 5.21 that μ = 0.25, β = 0.8, and ζ = 0.27 yield a minimum value of 0 Xk>F0 0 over a large range of values of r. However, amplification of the response Xk F0 40   0.01 30   1.0 20 10 0 0.4   0.1   0.27 0.6 0.8 1.0 1.2 1.4 1.6 r Figure 5.21 Repeat of the plots of Figure 5.19 with μ = 0.25, β = 0.8 for several values of ζ. In this case, ζ = 0.27 yields the lowest amplification over the largest bandwidth. Sec. 5.4 469 Damping in Vibration Absorption Primary system Viscous absorber ca k ma m Figure 5.22 Damper–mass system added to a primary mass (with no damping) to form a viscous vibration absorber. xa(t) x(t) X still occurs (i.e., 0 Xk>F0 0 7 1 for values of r 6 12), but no order-of-magnitude increase in 0 X 0 occurs as in the case of the undamped absorber. Next consider the case of an appended absorber mass connected to an undamped primary mass only by a dashpot, an arrangement illustrated in Figure 5.22. Systems of this form arise in the design of vibration reduction devices for rotating systems such as engines, where the operating speed (and hence the driving frequency) varies over a wide range. In such cases a viscous damper is added to the end of the crankshaft (or other rotating device) as indicated in Figure 5.23. The shaft spins through an angle θ1 with torsional stiffness k and inertia J1. The damping inertia J2 spins through an angle θ2 in a viscous film providing a damping force ca(θ1 - θ2). If an external harmonic torque is applied of the form M0e ωtj, the equation of motion of this system becomes $ # J1 0 θ1 ca -ca θ1 k 0 θ1 M J R J$ R + J RJ# R + J R J R = J 0 R e ωtj (5.38) 0 J2 θ2 -ca ca θ2 0 0 θ2 0 This is a rotational equivalent to the translational model given in Figure 5.22. It is easy to calculate the undamped natural frequencies of this two-degree-of-freedom system. They are ωp = k and ωa = 0 A J1 Torsional stiffness, k 1 2 Casing of rotational inertia J1 Internal disk of inertia J2 and rotational coordinate 2 Viscous oil with damping coefficient ca Figure 5.23 A viscous damper and mass added to a rotating shaft for broadband vibration absorption. Often called a Houdaille damper. 470 Design for Vibration Suppression Chap. 5 100   0.1 10   0.3 Xk M0 1   0.9 0.1 0 0.5 1 1.5 2 r Figure 5.24 The amplitude curves for a system with a viscous absorber, a plot of equation (5.39), for the case μ = 0.25 and for three different values of ζ. The solution of this set of equations is given by equations (5.32) and (5.33) with m and ma replaced by J1 and J2, respectively, c = 0, ka = 0, and F0 replaced by M0. Equation (5.32) is given in nondimensional form as equation (5.37). Hence letting β = ωa >ωp = 0 in equation (5.37) yields that amplitude of vibration of the primary inertia J1 [i.e., the amplitude of θ1(t)] is described by 4ζ2 + r 2 Xk = M0 C 4ζ2(r 2 + μr 2 - 1)2 + (r 2 - 1)2r 2 (5.39) where ζ = c>(2J2ωp), r = ω>ωp, and μ = J2 >J1. Figure 5.24 illustrates several plots of Xk>M0 for various values of ζ for a fixed μ as a function of r. Note again that the highest damping does not correspond to the largest amplitude reduction. The various absorber designs discussed previously, excluding the undamped case, result in a number of possible “good” choices for the various design parameters. When faced with a number of good choices, it is natural to ask which is the best choice. looking for the best possible choice among a number of acceptable or good choices can be made systematic by using methods of optimization introduced in the next section. Sec. 5.5 471 optimization 5.5 OPTIMIZATION In the design of vibration systems, the best selection of system parameters is often sought. In the case of the undamped vibration absorber of Section 5.3 the best selection for values of mass and stiffness of the absorber system is obvious from examining the expression for the amplitude of vibration of the primary system. In this case, the amplitude could be driven to zero by tuning the absorber mass and stiffness to the driving frequency. In the other cases, especially when damping is included, the choice of parameters to produce the best response is not obvious. In such cases, optimization methods can often be used to help select the best performance. Optimization techniques often produce results that are not obvious. An example is in the case of the undamped primary system or the damped absorber system discussed in the preceding section. In this case Figures 5.19 to 5.21 indicate that the best selection of parameters does not correspond to the highest value of the damping in the system, as intuition might dictate. These figures essentially represent an optimization by trial and error. In this section a more systematic approach to optimization is suggested by taking advantage of calculus. Recall from elementary calculus that minimums and maximums of particular functions can be obtained by examining certain derivatives. Namely, if the first derivative vanishes and the second derivative of the function is positive, the function has obtained a minimum value. This section presents a few examples where optimization procedures are used to obtain the best possible vibration reduction for various isolator and absorber systems. A major task of optimization is first deciding what quantity should be minimized to best describe the problem under study. The next question of interest is to decide which variables to allow to vary during the optimization. Optimization methods have developed over the years that allow the parameters during the optimization to satisfy constraints, for example. This approach is often used in design for vibration suppression. Recall from calculus that a function f(x) experiences a maximum (or minimum) at value of x = xm given by the solution of f =(xm) = d 3 f ( xm )4 = 0 dx (5.40) If this value of x causes the second derivative, f ==(xm), to be less than zero, the value of f(x) at x = xm is the maximum value that f(x) takes on in the region near x = xm. Similarly, if f ==(xm) is greater than zero, the value of f(xm) is the smallest or minimum value that f(x) obtains in the interval near xm. Note that if f ==(x) = 0, at x = xm, the value f(xm) is neither a minimum or maximum for f(x). The points where f′(x) vanish are called critical points. These simple rules were used in Section 2.2, Example 2.2.5, for computing the value (rpeak) where the maximum value of normalized magnitude of the steadystate response of a harmonically driven single-degree-of-freedom system occurs. The second derivative test was not checked because several plots of the function 472 Design for Vibration Suppression Chap. 5 clearly indicated that the curve contains a global maximum value rather than a minimum. In both absorber and isolator design, plots of the magnitude of the response can be used to avoid having to calculate the second derivative (second derivatives are often unpleasant to calculate). If the function f to be minimized (or maximized) is a function of two variables [i.e., f = f(x, y)], the preceding derivative tests become slightly more complicated and involve examining the various partial derivatives of the function f(x, y). In this case, the critical points are determined from the equations 0f(x, y) = 0 0x 0f(x, y) fy(x, y) = = 0 0y fx(x, y) = (5.41) Whether or not these critical points (x, y) are a maximum of the value f(x, y) or a minimum depends on the following: 1 If fxx(x, y) 7 0 and fxx(x, y)fyy(x, y) 7 f 2xy(x, y), then f(x, y) has a relative minimum value at x, y. 2 If fxx(x, y) 6 0 and fxx(x, y)fyy(x, y) 7 f 2xy(x, y), then f(x, y) has a relative maximum value at x, y. 3 If f 2xy(x, y) 7 fxx(x, y)fyy(x, y), then f(x, y) is neither a maximum nor a minimum value; the point x, y is a saddle point. 4 If f 2xy(x, y) = fxx(x, y)fyy(x, y), the test fails and the point x, y could be any or none of the above. Plots of f(x, y) can also be used to determine whether or not a given critical point is a maximum, minimum, saddle point, or none of these. These rules can be used to help solve vibration design problems in some circumstances. As an example of using these optimization formulations for designing a vibration suppression system, recall the damped absorber system of Section 5.4. In this case, the magnitude of the primary mass-normalized displacement with respect to the input force (moment) magnitude is given in equation (5.39) to be 4ζ2 + r 2 Xk = M0 C 4ζ2 1 r 2 + μr 2 - 1 2 2 + 1r2 - 12 r2 2 = f(r, ζ) (5.42) which is considered to be a function of the mixed damping ratio ζ and the frequency ratio r for a fixed mass ratio μ. In Section 5.4, values of f(r) are plotted versus r for several values of ζ in an attempt to find the value of ζ that yields the smallest maximum value of f(r, ζ). This is illustrated in Figure 5.24. Figure 5.25 illustrates the magnitude as a function of both ζ and r. From the figure it can be concluded that the derivative 0f>0r = 0 yields the maximum value of the magnitude for each fixed ζ. Sec. 5.5 473 optimization Xk M0 20 1.5 2 1  r 0.1 0 Figure 5.25 A plot of the normalized magnitude of the primary system versus both ζ and r [i.e., a two-dimensional plot of equation (5.42) for μ = 0.25]. This illustrates that the most desirable response is obtained at the saddle point. looking along the ζ axis, the partial derivative 0f>0ζ = 0 yields the minimum value of f(r, ζ) for each fixed value of r. The best design, corresponding to the smallest of the largest amplitudes, is thus illustrated in Figure 5.25. This point corresponds to a saddle point and can be calculated by evaluating the appropriate first partial derivatives. First consider 0(Xk>M0)>0ζ. From equation (5.42), the function to be differentiated is of the form f = A1>2 (5.43) B1>2 2 2 where A = 4ζ2 + r2 and B = 4ζ2 (r2 + µr2 - 1) + (r2 - 1) r2. Differentiating and equating the resulting derivatives to zero yields 0f 1 1 A-1>2dA dB - A1>2 3>2 = 0 = 1>2 0ζ 2 B 2 B (5.44) BdA = AdB (5.45) Solving this yields the form [B dA - A dB]>2B3 > 2 = 0 or dB = 8ζ 1 r 2 + μr 2 - 12 where A and B are as defined previously and dA = 8ζ 11 and - r22 = 11 - r 2 - μr 2 2 2 (5.46) Substitution of these values of A, dA, B, and dB into equation (5.45) yields 2 2 (5.47) 474 Design for Vibration Suppression Chap. 5 For μ ∙ 0, r 7 0, this has the solution r = 2 A2 + μ (5.48) Similarly, differentiating equation (5.42) with respect to r and substituting the value for r obtained previously yields ζop = 1 22(μ + 1)(μ + 2) (5.49) Equation (5.49) reveals the value of ζ that yields the smallest amplitude at the point of largest amplitude (resonance) for the response of the primary mass. The maximum value of the displacement for the optimal damping is given by a Xk 2 b = 1 + μ M0 max (5.50) which is obtained by substitution of equations (5.48) and (5.49) into equation (5.42). This last expression suggests that μ should be as large as possible. However, the practical consideration that the absorber mass should be smaller than the primary mass requires μ … 1. The value μ = 0.25 is fairly common. The second derivative conditions for the function f to have a saddle point (condition 3 in the preceding list) are too cumbersome to calculate. However, the plot of Figure 5.25 clearly illustrates that these conditions are satisfied. Furthermore, the plot indicates that f as a function of ζ is convex and f as a function of r is concave so that the saddle point condition is also the solution of minimizing the maximum value f(r, ζ), called the min–max problem in applied mathematics and optimization. Example 5.5.1 A viscous damper–mass absorber is added to the shaft of an engine. The mass moment of inertia of the shaft system is 1.5 kg # m2 >rad and has a torsional stiffness of 6 * 103 N # m>rad. The nominal running speed of the engine is 2000 rpm. Calculate the values of the added damper and mass moment of inertia such that the primary system has a magnification (Xk>M0) of less than 5 for all speeds and is as small as possible at the running speed. Solution Since ωp = 2k>J, the natural frequency of the engine system is ωp = 1.5 kg # m2 >rad 6.0 * 103 N # m>rad C = 63.24 rad>s Sec. 5.5 475 optimization The running speed of the engine is 2000 rpm or 209.4 rad>s, which is assumed to be the driving frequency (actually, it is a function of the number of cylinders). Hence the frequency ratio is r = ω 209.4 = = 3.31 ωp 63.24 so that the running speed is well away from the maximum amplification as illustrated in Figures 5.24 and 5.25 and the absorber is not needed to protect the shaft at its running speed. However, the engine spends some time getting to the running speed and often runs at lower speeds. The peak response occurs at rpeak = ω 2 = ωp A2 + μ as given by equation (5.48), and has a value of a Xk 2 b = 1 + μ M0 max as given by equation (5.50). The magnification is restricted to be 5, so that 1 + 2 … 5 or μ Ú 0.5 μ Thus μ = 0.5 is chosen for the design. Since the mass of the primary system is J1 = 1.5 kg # m2 >rad and μ = J2 >J1, the mass of the absorber is J2 = μJ1 = 1 (1.5) kg # m2 >rad = 0.75 kg # m2 # rad 2 The damping value required for equation (5.50) to hold is given by equation (5.49) or ζop = 1 22(μ + 1)(μ + 2) = 1 22(1.5)(2.5) = 0.3651 Recall from Section 5.4 [just following equation (5.39)], that ζ = c>(2J2ωp), so that the optimal damping constant becomes cop = 2ζop J2ωp = 2(0.3651)(0.75)(63.24) = 34.638 N # m # s>rad The two values of J2 and c given here form an optimal solution to the problem of designing a viscous damper–mass absorber system so that the maximum deflection of the primary shaft is satisfied 0 Xk>M0 0 6 5. This solution is optimal in terms of a choice of ζ, which corresponds to the saddle point of Figure 5.25 and yields a minimum value of all maximum amplifications. n 476 Design for Vibration Suppression Chap. 5 F0 cost m x1(t) c k1 x2(t) k2 Figure 5.26 Model of a machine mounted on an elastic foundation through an elastic damper to provide vibration isolation. Optimization methods can also be useful in the design of certain types of vibration isolation systems. For example, consider the model of a machine mounted on an elastic damper and spring system as illustrated in Figure 5.26. The equations of motion of the system of Figure 5.26 are $ # # mx1 + c(x1 - x2) + k1x1 = F0 cos ωt # # c(x1 - x2) = k2x2 (5.51) because no mass term appears in the second equation, the system given by equation (5.51) is of third order. Equation (5.51) can be solved by assuming periodic motions of the form x1(t) = X1e jωt and x2(t) = X2e jωt (5.52) and considering the exponential representation of the harmonic driving force. Substitution of equation (5.52) into (5.51) yields (k1 - mω2 + jcω)X1 - jcωX2 = F0 jcωX1 - (k2 + jcω)X2 = 0 (5.53) Solving for the amplitudes X1 and X2 yields X1 = and X2 = k2 1 k1 + mω 2 k2 1 k1 + mω 2 2 + jcω 1 k1 + k2 - mω2 2 F0(k2 + jcω) 2 + cω 1 k1 + k2 - mω2 2 j cωF0 j (5.54) (5.55) These two amplitude expressions can be simplified further by substituting the nondimensional quantities r = ω> 1k1 >m, γ = k1 >k2, and ζ = c> 121k1m 2 . The force transmitted to the base is the vector sum of the two forces k1x1 and k2x2. Using Sec. 5.5 477 optimization complex arithmetic and a vector sum (recall Sections 2.3 and 2.4) the force transmitted can be written as T.R. = 31 + 4(1 + γ)2ζ2r 2 FT = 2 F0 3 1 1 - r 2 2 + 4ζ2r 2 1 1 + γ - r 2γ2 (5.56) which describes the transmissibility ratio for the system of Figure 5.26. The force transmissibility ratio can be optimized by viewing the ratio FT >F0 as a function of r and ζ. Figure 5.27 yields a plot of FT >F0 versus r for γ = 0.333 and for several values of ζ. This illustrates that the value of the damping ratio greatly affects the transmissibility at resonance. A three-dimensional plot of FT >F0 versus r and ζ is given in Figure 5.28, which illustrates that the saddle point value of ζ and r yields the best design for the minimum transmissibility of the maximum force transmitted. 104 103 FT F0   0.01 102 101   0.1 100 101   0.4 0 0.5 1 r 1.5 2 Figure 5.27 Plot of equation (5.56) illustrating the effect of damping on the magnification of force transmitted to ground. FT F0 1 0.75 0.50 r  1.25 0.05 Figure 5.28 Plot of equation (5.56) illustrating FT >F0 versus ζ versus r. The plot shows the point where damping minimizes the maximum transmissibility. 478 Design for Vibration Suppression Chap. 5 The saddle point illustrated in Figure 5.28 can be found from the derivative of T.R. as given in equation (5.56). These partial derivatives are 0(T.R.) 22(1 + γ) = 0 yields rmax = 0ζ 11 + 2γ (5.57) 22(1 + 2γ)>γ 0(T.R.) = 0 yields ζop = 0r 4(1 + γ) (5.58) and These values of r correspond to an optimal design of this type of isolation device. At the saddle point, the value of T.R. becomes (T.R.)max = 1 + 2γ (5.59) which results from substitution of equations (5.57) and (5.58) into equation (5.56). This illustrates that as long as γ 6 1, T.R. 6 3 and the isolation system will not cause much difficulty at resonance. Example 5.5.2 An isolation system is to be designed for a machine modeled by the system of Figure 5.26 (i.e., an elasticity coupled viscous damper). The mass of the machine is m = 100 kg and the stiffness k1 = 400 N>m. The driving frequency is 10 rad>s at nominal operating conditions. Design this system (i.e., choose k2 and c) such that the maximum transmissibility ratio at any speed is 2 (i.e., design the system for “startup” or “run through”). What is the T.R. at the normal operating condition of a driving frequency of 10 rad>s? Solution For m = 100 kg and k1 = 400 N>m, ωn = 2400>100 = 2 rad>s, 2 rad>s, so that the normal operating condition is well away from resonance (i.e., r = ω>ωn = 10>2 = 5 at running conditions). Equation (5.59) yields that the maximum value for T.R. is (T.R.) max = 1 + 2γ … 2 so that γ = 0.5 and k2 = (0.5)(k1) = (0.5)(400 N>m) = 200 N>m. With γ = 0.5, the optimal choice of damping ratio is given by equation (5.58) to be ζop = 22(1 + 2γ)>γ 4(1 + γ) = 0.4714 Hence the optimal choice of damping coefficient is cop = 2ζopωnm = 2(0.4714)(2)(100) = 188.56 kg>s Sec. 5.6 479 Viscoelastic Damping treatments The T.R. value at nominal operating frequency of ω = 10 rad>s is given by equation (5.56) to be (r = 10>2 = 5) T.R. = 2 1 1 - 52 2 2 + 4(0.4714)2(5)2 3 1 + 0.5 - 52(0.5) 4 2 21 + 4(1 + 0.5)2(0.4714)2(5)2 = 0.12 Hence the design k2 = 200 N>m and c = 188.56 kg>s will protect the surroundings by a T.R. of 0.12 (i.e., only 12% of the applied force is transmitted to ground) and limits the force transmitted near resonance to a factor of 2. n 5.6 VISCOELASTIC DAMPING TREATMENTS A common and very effective way to reduce transient and steady-state vibration is to increase the amount of damping in the system so there is greater energy dissipation. This is especially useful in aerospace structures applications, where the added mass of an absorber system may not be practical. While a rigorous derivation of the equations of vibration for structures with damping treatments is beyond the scope of this book, formulas are presented that provide a sample of design calculations for using damping treatments. A damping treatment consists of adding a layer of viscoelastic material, such as rubber, to an existing structure. The combined system often has a higher damping level and thus reduces unwanted vibration. This is standard in the auto industry to reduce vibration-induced noise in the car’s interior and can be found under the flooring carpet. This procedure is described by using the complex stiffness notation. The concept of complex stiffness results from considering the harmonic response of a damped system of the form $ # mx + cx + kx = F0e jωt (5.60) Recall from Section 2.3 that the solution to equation (5.60) can be calculated by assuming the form of the solution to be x(t) = Xe jωt, where X is a constant and j = 1-1. Substitution of the assumed form into equation (5.60) and dividing by the nonzero function e jωt yields This can be written as or 3 -mω2 + (k + jωc) 4 X = F0 c -mω2 + k a1 + 3 -mω2 ωc j b d X = F0 k + k*4 X = F0 (5.61) (5.62) (5.63) 480 Design for Vibration Suppression Chap. 5 where k* = k(1 + ηj). Here η = ωc>k is called the loss factor and k* is called the complex stiffness. This illustrates that in steady state, the viscous damping in a system can be represented as an “undamped” system with a complex-valued stiffness. The imaginary part of the stiffness, η, corresponds to the energy dissipation in the system. Since the loss factor has the form η = c ω k (5.64) the loss factor depends on the driving frequency and hence is said to be frequency dependent. Thus the value of the energy dissipation term depends on the value of the driving frequency of the external force exciting the structure. The concept of complex stiffness just developed is called the Kelvin–Voigt model of a material. This corresponds to the standard spring–dashpot configuration as sketched in Figure 5.29 and used extensively in the first four chapters. The difference between the Kelvin–Voigt model used here and the viscous-damping model of the previous chapters is that the Kelvin–Voigt model used here is valid only in steady-state harmonic motion. The complex stiffness and the corresponding frequency-dependent loss factor, η = ωc>k, model the energy dissipation at steady state during harmonic excitation of frequency ω only. The viscous dashpot representation introduced in Section 1.3 models energy dissipation in free decay as well as in other transient and forced-response excitation. However, the Kelvin–Voigt representation is a more accurate, though limited, model of the internal damping in materials. The complex stiffness formulation can also be derived from the stress–strain relationship for a linear viscoelastic material. Such materials are called viscoelastic because they exhibit both elastic behavior and viscous behavior, as captured in the Kelvin–Voigt model described in Figure 5.29. Other viscoelastic models exist in addition to this one, but such models are beyond the scope of this book [see Snowden (1968)]. An alternative viscoelastic model is given in Figure 5.26, for instance. The stress–strain relationship for viscoelastic material can be summarized by extending the modulus of a material, denoted by E, to a complex modulus, denoted E*, by the relation E* = E(1 + ηj) (5.65) where j = 1-1 as before and η is the loss factor of the viscoelastic material. The complex modulus of a material, as defined in equation (5.65), can be measured, Material with viscoelastic properties k m F(t)  F0e jt c x(t) Figure 5.29 Kelvin–Voigt damping model gives rise to the complex stiffness concept of representing damping in steady-state vibration. Sec. 5.6 481 Viscoelastic Damping treatments Thiokol RD 5 20 C 1  108 5 C E (dyn/cm2) 5 2 1 0.5  107 35 C 5 1  106 1 5 10 50 102 5  102 103 0.05 5  103 4 10 Frequency (Hz) Figure 5.30 A sample plot of elastic modulus (solid lines) and loss factor (dashed lines) versus frequency for several fixed temperatures. and in general is both frequency and temperature dependent over a broad range of values. Some sample values for frequency dependence are given in Figure 5.30 for fixed temperatures. Materials that exhibit viscoelastic behavior are rubber and rubber-like substances (e.g., butyl rubber, neoprene, polyurethane) as well as plexiglass, vinyl, and nylon. A common use of these viscoelastic materials in design is as an additive damping treatment to increase the combined structure’s damping or as an isolator. layers of viscoelastic material are often added to structures composed of lightly damped material such as aluminum or steel to form a new structure that has sufficient stiffness for static loading and sufficient damping for controlling vibration. Table 5.2 lists some values of E and η for a viscoelastic material at two different temperatures and several frequencies. TABLE 5.2 SoME CoMPlEx MoDuluS DAtA (I.E., E AnD η) FoR PARACRIl-bj wItH 50 PHRCa E (psi) 3 4 7 4 6 13 a * * * * * * 103 103 103 103 103 103 η 0.21 0.28 0.55 0.25 0.5 1 T (°F) 75 75 75 50 50 50 ω (Hz) ω (rad>s) 10 100 1000 10 100 1000 62.8 628.3 6283.2 62.8 628.3 6283.2 Nitrile rubber elastomeric material made by U.S. Rubber Company. Source: Nashif, Jones, and Henderson, 1985, Data Sheet 27. E (N>m2) 2.068 2.758 4.826 2.758 4.137 8.963 * * * * * * 107 107 107 107 107 107 482 Design for Vibration Suppression Chap. 5 The loss factor η defined in terms of the complex modulus as given in equation (5.65) is related to the loss factor η defined by examining the notion of complex stiffness as defined in equation (5.64) in the same way that the stiffness and modulus of a material are related in Table 1.1 and Section 1.5. For example, if the specimen of interest is a cantilevered beam, the stiffness associated with the deflection of the tip in the transverse direction is related to the elastic modulus by k = 3EI l3 (5.66) where I is the area moment of inertia and l is the length of the beam. Hence if the beam is made of viscoelastic material, k* = 3E*I 3I = 3 E(1 + ηj) = k(1 + η ∙ j) 3 l l so that η = η ∙ and the two notions of loss factor are identical. The notion of loss factor η is related to the definition of a damping ratio ζ only at resonance (i.e., ω = ωn = 1k>m). When the driving frequency is the same as the system’s natural frequency, η = 2ζ. This simple relationship is often used to describe the free decay of a viscoelastic material (an approximation). The design of structures for reduced vibration magnitude often consists of adding a viscoelastic damping treatment to an existing structure. Many structures are made of metals and alloys that have relatively little internal damping. A viscoelastic damping material (such as rubber) is often added as a layer to the outside surface of a structure (called free-layer damping treatment or unconstrained-layer damping). A much more effective approach is to cover the free layer with another layer of metal to form a constrained-layer damping treatment. In the constrainedlayer damping treatment, the damping layer is covered with a (usually thin) layer of metal (stiff) to produce shear deformation in the viscoelastic layer. The constrained-layer approach produces higher loss factors and generally costs more. These damping treatments are manufactured as sheets, tapes, and adhesives for ease of application. A free-layer damping treatment for a pinned–pinned beam (see Table 6.4) in transverse or bending vibration is illustrated in Figure 5.31. Material 1, the bottom layer, is usually a metal providing the appropriate stiffness. The second layer, denoted as having modulus E2 and thickness H2, is the damping treatment. Using l H2 E2(1  2 j) Damping treatment H1 E1(1  1 j) Metal structure Figure 5.31 A simple supported beam with an unconstrained damping treatment illustrating the geometry and physical parameters. Sec. 5.6 483 Viscoelastic Damping treatments the notation of Figure 5.31, the combined stiffness EI is related to the original stiffness E1I1 by e2h2 EI = 1 + e2h32 + 3(1 + h2)2 E1I1 1 + e2h2 (5.67) where e2 = E2 >E1 and h2 = H2 >H1 are dimensionless. Note that since all the quantities on the left side of equation (5.67) are positive, the damping treatment increases the stiffness of the system a small amount (h2 6 1). In addition, the combined system’s loss factor, η, is given by [assuming that (e2h2)2 6 6 e2h2] η = 11 e2h2 13 + 6h2 + 4h22 + 2e2h23 + e 22h24 2 + e2h2 2 1 + 4e2h2 + 6e2h22 + 4e2h32 + e 22h42 η2 (5.68) Equation (5.68) yields a formula that can be used in the design of add-on damping treatments, as illustrated in the following example. Example 5.6.1 An electric motor that drives a cooling fan is mounted on an aluminum shelf (1 cm thick) in a cabinet holding electronic parts (perhaps a mainframe computer) as illustrated in Figure 5.32. The vibration of the motor causes the mounting platform, and hence the surrounding cabinet, to shake. The motor rotates at an effective frequency of 100 Hz. The temperature in the cabinet remains at 75°F. A damping treatment is added to reduce the vibration of the shelf. Cabinet Component shelves Cooling fan Damping treatment End view Figure 5.32 treatment. Electronic cabinet with cooling fan illustrating the use of a damping Solution The shelf is modeled as a simply supported beam so that equation (5.68) can be used to design the damping treatment. If nitrile rubber is used as the damping treatment, calculate the loss factor of the combined system at 75°F if H2 = 1 cm. Referring to Table 5.2, the modulus of the rubber at 100 Hz and 75°F is E2 = 2.758 * 107 Pa 484 Design for Vibration Suppression Chap. 5 The modulus of aluminum is E1 = 7.1 * 1010 Pa, so that e2 = 2.758 * 107 7.1 * 1010 = 0.00039 = 3.885 * 10-4 The thickness of both the shelf and the damping treatment are taken to be the same, so that h2 = 1. From equation (5.68) the combined loss factor becomes η = (0.00039) 3 3 + 6 + 4 + 2(0.00039) + (0.00039)2 4 (1.00039) 31 + 4(0.00039) + 6(0.00039) + 4(0.00039) + (0.00039)2 4 η2 = 5.021 * 10-3η2 From Table 5.2, η2 = 0.28 at 100 Hz and 75°C, so that η = 0.00141 which is about 50% higher than the loss factor given by pure aluminum. The formula given in equation (5.68) is a bit cumbersome for design work. Often it is approximated by η = 14 1e2h22 2 η2 (5.69) which is reasonable for many situations. The values of e2 and η2 are fixed by the choice of materials and the operating temperature. Once these parameters are fixed, the parameter h2 = H2 >H1 is the only remaining design choice. Since H1 is usually determined by stiffness considerations, the remaining design choice is the thickness of the damping layer, H2. n Example 5.6.2 An aluminum shelf is to be given a damping treatment to raise the system loss factor to η = 0.03. A rubber material is used with modulus at room temperature of 1% of that of aluminum (i.e., e2 = 0.01). What should the thickness of the damping material be if its loss factor is η2 = 0.261 and the aluminum shelf is 1 cm thick? Solution From the approximation given by equation (5.69), η = 14η2e2h22 Using the values given, this becomes 0.03 = 14(0.261)(0.01) H22 (1 cm)2 Solving this for H22 yields H22 = 0.03 (cm)2 = 0.82 cm2 14(0.01)(0.261) so that H2 = 0.91 cm will provide the desired loss factor. n Sec. 5.7 485 Critical Speeds of Rotating Disks 5.7 CRITICAL SPEEDS OF ROTATING DISKS Of primary concern in the design of rotating machinery is the vibration phenomenon of critical speeds. This phenomenon occurs when a rotating shaft with a disk, such as a jet engine turbine blade rotating about its shaft mounted between two bearings, rotates at a speed that excites the natural bending frequency of the shaft– disk system. This defines a resonance condition that causes large deflection of the shaft, which in turn causes the system to fail violently (i.e., the engine blows apart). The nature of the resonance and the factors that control the resonance values need to be known and calculated by designers so that they can ensure that a given design is safe for production. The analytical formulation of the critical speed problem also provides some insight into how to avoid such resonance, or critical speeds. If the rotating mass modeled by the disk is not quite homogeneous or symmetric due to some imperfection, its geometric center and center of gravity will be some distance apart (say, a). This is illustrated in Figure 5.33, which presents a simplified model of a large electric motor’s shaft-and-rotor system (or a bladed turbine engine, for example). The shaft is constrained from moving in the radial direction by two bearings. As the shaft rotates about its long axis with angular velocity ω, the offset center of gravity pulls the shaft away from the centerline, causing it to bow as it rotates. This is called whirling. The forces acting on the center of mass are the inertial force, any damping force (internal or external), and the elastic force of the shaft. In vector form, the force balance yields $ # # mr = -kx ni - kyjn - cxni - cy nj (5.70) where ni and nj are unit vectors, r the position vector defined by the line OG, m the mass of the disk, c the damping coefficient of the shaft system, and k the stiffness yj Center of mass t  Shaft t G ya sin t a y E O Support bearing Disk (a) Support bearing a  line EG y xi x xa cos t (b) Figure 5.33 A schematic of a model of a disk rotating on a shaft and the corresponding geometry of the center of mass, G, of the disk relative to the neutral axis of the shaft, O, and the center of the rotating shaft, E: (a) side view; (b) end view. This diagram is useful in modeling the whirling of rotating machines (engines, turbine compressors, etc.), which are not perfectly balanced (i.e., a ∙ 0). 486 Design for Vibration Suppression Chap. 5 coefficient provided by the shaft system. From examining the end view of Figure 5.33, the vector r can also be written in terms of the unit vectors ni and nj as r = (x + a cosωt) in + (y + a sin ωt) jn (5.71) Taking two derivatives yields that the acceleration vector of the center of mass is $ $ $ r = (x - aω2cosωt) ni + (y - aω2 sin ωt)jn Substituting equation (5.72) into equation (5.70) yields 1mx$ - maω2 cosωt + cx# + kx2 ni + 1my$ - maω2 sin ωt + cy# + ky2 nj = 0 Since this is a vector equation, it is equivalent to the two scalar equations $ # mx + cx + kx = maω2 cos ωt $ # my + cy + ky = maω2 sin ωt (5.72) (5.73) (5.74) (5.75) These two equations are exactly the form of equation (2.82) for the response of a spring–mass system to a rotating unbalance discussed in Section 2.5. In this case, the x and y motion corresponds to the bending vibration of the shaft instead of the translational motion of a machine in the vertical direction discussed in Section 2.5. Window 5.5 Solution of the Rotating Unbalance Equation from Section 2.5 The steady-state solution to $ # mx + cx + kx = m0eω2 sin ωt where ω is the driving frequency of the unbalanced mass, m0 the mass of the unbalance, and e the distance from m0 to the center of rotation, is X sin (ω t - ϕ). Here X = And m0e m 3 11 - r 2 2 + 12ζr 2 ϕ = tan-1 r2 2 2ζr 1 - r2 where r = ω> 2k>m and ζ = c>(2mωn). 2 (2.84) (2.85) Sec. 5.7 487 Critical Speeds of Rotating Disks Referring to Window 5.5, equation (5.75) has steady-state response magnitude given by equation (2.84); that is, equation (5.75) has the steady-state solution (since m = m0 and e = a in this case) y(t) = 3 11 - r 2 2 + (2ζr)2 ar 2 2 sin aωt - tan-1 2ζr 1 - r2 Similarly, equation (5.74) has steady-state solution of the form x(t) = 311 - r 2 ar 2 2 2 + (2ζr) 2 cos aωt - tan-1 2ζr 1 - r2 b (5.76) b (5.77) since the solution given by equations (2.84) and (2.85) is 90° out of phase and the phase angle ϕ does not depend on the phase of the exciting force. The angle ϕ given by equation (2.85) becomes the angle between the lines OE and EG. From Figure 5.33, the angle θ made between the x axis and the line OE is tanθ = sin (ωt - ϕ) y = tan(ωt - ϕ) = x cos(ωt - ϕ) (5.78) or θ = ωt - ϕ (5.79) # Differentiating equation (5.79) with respect to t yields θ = ω. # The velocity θ is the velocity of whirling. Whirling is the angular motion of the deflected shaft rotating about the neutral axis of the shaft. The calculation leading to equation (5.79) and its derivative shows that the whirling velocity is the same as # the speed with which the disk rotates about the shaft (i.e., θ = ω). This is called synchronous whirl. The amplitude of motion of the center of the shaft about its neutral axis is the line r = OE in the end view of Figure 5.33. Note that vector OE = r = x: OE = r = x ni + yjn . The magnitude of this vector is just 0 r(t) 0 = 2x2 + y2 = X 2 sin 2(ωt - ϕ) + cos2(ωt - ϕ) = X (5.80) where X is the magnitude of x(t). Note that X = Y, where Y is the magnitude of y(t) as given in equation (5.76). This calculation indicates that the distance between the shaft and its neutral axis is constant and has magnitude X = 3 11 - r 22 + (2ζr)2 ar 2 2 (5.81) 488 Design for Vibration Suppression Chap. 5 X a Normalized magnitude 5   0.1 4 3 2   0.25   0.707 1   1.0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 r  r Figure 5.34 A plot of the ratio of radius of deflection (OE) to the distance to the center of mass of the disk (a) versus the frequency ratio for four different values of the damping ratio for the disk and shaft system of Figure 5.33. This, of course, is exactly the same form as the magnitude plot of equation (2.84) given in Figure 2.21 for a spring–mass–damper system driven by a rotating out-ofbalance mass. This plot is repeated for the rotational amplitude case of interest in Figure 5.34. Note that a resonance phenomenon occurs near r = 1, as expected. For lightly damped shafts this corresponds to unacceptably high amplitudes of rotation. The special case of r = 1 (i.e., ωr = 1k>m) is called the rotor system’s critical speed. If a rotor system runs at its critical speed, the large deflection will cause a large force to be transmitted to the bearings and eventually lead to failure. From the design point of view, the running speed, mass, and stiffness are examined for a given rotor and redesigned until r 7 3, so that the deflections are limited to the size of the distance to the center of the mass of the disk. However, when the rotor system is started up, it must pass through the region near r = 1. If this startup procedure occurs too slowly, the resonance phenomena could damage the rotor bearings. Hence some damping in the system is desirable to avoid excessive amplitude at resonance. Note from Figure 5.34 that as ζ increases, X>a at resonance becomes substantially smaller. Example 5.7.1 Consider a 55-kg compressor rotor with a shaft stiffness of 1.4 * 107 N>m, with an operation speed of 6000 rpm, and a measured internal damping providing a damping ratio of ζ = 0.05. The rotor is assumed to have a worst-case eccentricity of 1000 µm (a = 0.001 m). Calculate (a) the rotor’s critical speed, (b) the radial amplitude at operating speed, and (c) the whirl amplitude at the system’s critical speed. Sec. 5.7 489 Critical Speeds of Rotating Disks Solution (a) The critical speed of the rotor is just the rotor’s natural frequency, so that 1.4 * 107 N>m ωc = 2k>m = C 55 kg = 504.5 rad>s which corresponds to a rotor speed of 504.5 cycle rad 60 s * * = 4817.6 rpm s min 2π rad (b) The value of r at running speed is just r = ω = 2k>m (2π>60) = (2π>60) 2k>m 6000 = 1.2454 4817.6 or about 1.25. The value of the radial amplitude of whirl at the operating speed is then given by equation (5.81) with this value of r: X = 0 r(t) 0 = 3 11 - r 2 2 + (2ζr)2 ar 2 2 = 0.0027116 m = 3 31 - (1.25)24 + 32(0.05)(1.25)4 2 (0.001)(1.25)2 2 or about 2.7 mm. Here r = 1.25, ζ = 0.05, and a = 0.001. Note that if r = 1.2454 is used, X = 0.0027455 m results. This gives some feeling for the sensitivity of the value of X to knowing exact values of r. Minor speed variation of, say, 10% in the running speed would cause r to vary between 1.12 and 1.37. (c) At critical speed, r = 1 and X becomes X = a 0.001 = = 0.01 m 2ζ 2(0.05) or 1 cm, an order of magnitude larger than the whirl amplitude at running speed. n Example 5.7.2 In designing a rotor system, there are many factors besides the deflection calculation indicated previously that determine the damping, stiffness, mass, and operating speed of the rotor system. Hence the designer concerned about dynamic deflections and critical speeds is often only allowed to change the design a little. Otherwise, an entire redesign must be performed, which may become very costly. With this in mind, again consider the rotor of Example 5.7.1. The clearance specification for the rotor inside the compressor housing limits the whirl amplitude at resonance to be 2 mm. Since the whirl amplitude at operating speed is greater than the allowable clearance, what percent of change in mass is required to redesign this system? What percent change in stiffness would result in the same design? Discuss the feasibility of such a change. 490 Design for Vibration Suppression Chap. 5 Solution The required mass for a 2-mm deflection can be calculated from equation (5.81) by first determining a value of r corresponding to 2 mm. This yields X = 0.002 m = 2(1 - r)2 + 3 2(0.05)r 24 2 (0.001)r 2 or r must satisfy r4 - 2.653r2 + 1.3332 = 0. This is a quadratic equation in r 2, which has solutions r2 = 0.6737, 1.979 or r = 0.8207, 1.406, since the values of frequency ratio must be positive and real. Examination of the plot in Figure 5.34 of the magnitude yields that the value of r of interest is r = 1.406. At running speed, 6000 rpm r = min 60 s # 2π rad rev 2k>m rad>s Solving for the mass m yields = 21.4 * 107 >m 628.12 = 1.406 m = 70.15 kg Since the original design value of the mass of the disk is 55 kg, the mass must be increased by 27.5% to produce a design that has its running speed deflection limited to 2 mm. If the compressor is to be used in an application fixed to ground (such as a building), then adding 15 kg of mass to the disk may be a perfectly reasonable solution, provided that the bearings are capable of the increased force. However, if the compressor is to be used in a vehicle where weight is a consideration, such as an airplane, a 27% increase in mass may not be an acceptable design. In this case equation (5.81) can be used to examine a possible redesign by making a change in stiffness. Equation (5.81) with the appropriate parameter values yields r = 1.406 = 628.12 rad>s 2k>55 Solving for k yields k = 1.0977 * 107 N>m This amounts to about a 27% change in the stiffness. Unfortunately, the stiffness of the shaft cannot be changed very easily. It is determined by geometric and material properties. The material is often determined by temperature and cost considerations as well as toughness. It can be difficult to change the stiffness by 27%. n Note from Example 5.7.2 that the amplitude of whirling is sensitive to changes in mass and changes in stiffness. Also note from Figure 5.34 that the damping value is of little concern when choosing the design for whirl amplitude if chosen far enough from resonance (i.e., r 7 2). Rather, damping is chosen to limit the Chap. 5 491 Problems amplitude near resonance, which should occur only during startup and run down (i.e., X = a>2ζ at resonance). The analysis of critical speeds and rotor dynamics presented here provides a quick introduction to the topic with simplifying assumptions. The topic of rotor dynamics constitutes a separate field of study, and a text on rotor dynamics should be consulted for complete details [see, for example, Ehrich (1992), or Childs (1993)]. PROBLEMS Section 5.1 (Problems 5.1 through 5.5) 5.1. Using the nomograph of Figure 5.1, determine the frequency range of vibration for which a machine oscillation remains at a satisfactory level under rms acceleration of 8000 mm>s2. 5.2. Using the nomograph of Figure 5.1, determine the frequency range of vibration for which a structure’s rms acceleration will not cause wall damage if vibrating with an rms displacement of 2 mm or less. 5.3. What natural frequency must a hand drill have if its vibration must be limited to a minimum rms displacement of 10 μm and rms acceleration of 0.1 m>s2? What rms velocity will the drill have? 5.4. A machine part has a maximum rms acceleration of 5 m>s2 at a frequency of 4 Hz, however its rms amplitude must be less than 1 cm. Will this part satisfy these vibration requirements? 5.5. Using the expression for the amplitude of the displacement, velocity, and acceleration of an undamped single-degree-of-freedom system, calculate the velocity and acceleration amplitude of a system with a maximum displacement of 10 cm and a natural frequency of 10 Hz. If this corresponds to the vibration of the wall of a building under a wind load, is it an acceptable level? Section 5.2 (Problems 5.6 through 5.26) 5.6. A 100-kg machine is supported on an isolator of stiffness 700 * 103 N>m. The machine causes a vertical disturbance force of 350 N at a revolution of 3000 rpm. The damping ratio of the isolator is ζ = 0.2. Calculate (a) the amplitude of motion caused by the unbalanced force, (b) the transmissibility ratio, and (c) the magnitude of the force transmitted to ground through the isolator. 5.7. Plot the T.R. of Problem 5.6 for the cases ζ = 0.001, ζ = 0.025, and ζ = 1.1. 5.8. A simplified model of a washing machine is illustrated in Figure P5.8. A bundle of wet clothes forms a mass of 10 kg (mb) in the machine and causes a rotating unbalance. The rotating mass is 20 kg (including mb) and the diameter of the washer basket (2e) is 50 cm. Assume that the spin cycle rotates at 300 rpm. let k be 1000 N>m and ζ = 0.01. Calculate the force transmitted to the sides of the washing machine. Discuss the assumptions made in your analysis in view of what you might know about washing machines. 492 Design for Vibration Suppression Chap. 5 Top view k/2 c/2 e mb m k/2 Figure P5.8 A simple model of the vibration of a washing machine induced by a rotating imbalance such as commonly caused by an uneven distribution of wet clothes during a rinse cycle. c/2 Frictionless support 5.9. Referring to Problem 5.8, let the spring constant and damping rate become variable. The quantities m, mb, e, and ω are all fixed by the previous design of the washing machine. Design the isolation system (i.e., decide on which value of k and c to use) so that the force transmitted to the side of the washing machine (considered as ground) is less than 100 N. 5.10. A harmonic force of maximum value 25 N and frequency of 180 cycles>min acts on a machine of 25 kg mass. Design a support system for the machine (i.e., choose c, k) so that only 10% of the force applied to the machine is transmitted to the base supporting the machine. 5.11. Consider a machine of mass 70 kg mounted to ground through an isolation system of total stiffness 30,000 N>m, with a measured damping ratio of 0.2. The machine produces a harmonic force of 450 N at 13 rad>s during steady-state operating conditions. Determine (a) the amplitude of motion of the machine, (b) the phase shift of the motion (with respect to a zero phase exciting force), (c) the transmissibility ratio, (d) the maximum dynamic force transmitted to the floor, and (e) the maximum velocity of the machine. 5.12. A small compressor weighs about 70 lb and runs at 900 rpm. The compressor is mounted on four supports made of metal with negligible damping. (a) Design the stiffness of these supports so that only 15% of the harmonic force produced by the compressor is transmitted to the foundation. (b) Design a metal spring that provides the appropriate stiffness using Section 1.5 (refer to Table 1.2 for material properties). 5.13. Typically, in designing an isolation system, one cannot choose any continuous value of k and c but rather works from a parts catalog wherein manufacturers list isolators available and their properties (and costs, details of which are ignored here). Table 5.3 lists several made-up examples of available parts. Using this table, design an isolator for a 500-kg compressor running in steady state at 1500 rev>min. Keep in mind that as a rule of thumb compressors usually require a frequency ratio of r = 3. TABLE 5.3 CAtAlog VAluES oF StIFFnESS AnD DAMPIng PRoPERtIES oF VARIouS oFF-tHE-SHElF ISolAtoRS Part No.a 3 k (10 N>m) c (N # s>m) a R-1 R-2 R-3 R-4 R-5 M-1 M-2 M-3 M-4 M-5 250 2000 500 1800 1000 1500 1800 1000 2500 500 75 110 150 115 250 140 500 160 750 200 The “R” in the part number designates that the isolator is made of rubber, and the “M” designates metal. In general, metal isolators are more expensive than rubber isolators. Chap. 5 493 Problems 5.14. An electric motor of mass 10 kg is mounted on four identical springs as indicated in Figure P5.14. The motor operates at a steady-state speed of 1750 rpm. The radius of gyration (see Example 1.4.6 for a definition) is 100 mm. Assume that the springs are undamped and choose a design (i.e., pick k) such that the transmissibility ratio in the vertical direction is 0.0194. With this value of k, determine the transmissibility ratio for the torsional vibration (i.e., using θ rather than x as the displacement coordinates). 250 mm  mg 2k(r)  mg 2 mg  2k(r) 2 Figure P5.14 A vibration model of an electric motor mount. 5.15. A large industrial exhaust fan is mounted on a steel frame in a factory. The plant manager has decided to mount a storage bin on the same platform. Adding mass to a system can change its dynamics substantially and the plant manager wants to know if this is a safe change to make. The original design of the fan support system is not available. Hence measurements of the floor amplitude (horizontal motion) are made at several different motor speeds in an attempt to measure the system dynamics. No resonance is observed in running the fan from zero to 500 rpm. Deflection measurements are made and it is found that the amplitude is 10 mm at 500 rpm and 4.5 mm at 400 rpm. The mass of the fan is 50 kg, and the plant manager would like to store up to 50 kg on the same platform. The best operating speed for the exhaust fan is between 400 and 500 rpm depending on environmental conditions in the plant. 5.16. A 350-kg rotating machine operates at 800 cycles>min. It is desired to reduce the transmissibility ratio by one-fourth of its current value by adding a rubber vibration isolation pad. How much static deflection must the pad be able to withstand? 5.17. A 68-kg electric motor is mounted on an isolator of mass 1200 kg. The natural frequency of the entire system is 160 cycles>min and has a measured damping ratio of ζ = 1. Determine the amplitude of vibration and the force transmitted to the floor if the out-of-balance force produced by the motor is F(t) = 100 sin (31.4t) in newtons. 5.18. The force exerted by an eccentric (e = 0.22 mm) flywheel of 1000 kg, is 600 cos (52.4t) in newtons. Design a mounting to reduce the amplitude of the force exerted on the floor to 1% of the force generated. Use this choice of damping to ensure that the maximum force transmitted is never greater than twice the generated force. 5.19. A rotating machine weighing 4000 lb has an operating speed of 2000 rpm. It is desired to reduce the amplitude of the transmitted force by 80% using isolation pads. Calculate the stiffness required of the isolation pads to accomplish this design goal. 5.20. The mass of a system may be changed to improve the vibration isolation characteristics. Such isolation systems often occur when mounting heavy compressors on factory floors. This is illustrated in Figure P5.20. In this case the soil provides the stiffness of the isolation system (damping is neglected) and the design problem becomes that of 494 Design for Vibration Suppression Chap. 5 choosing the value of the mass of the concrete block>compressor system. Assume that the stiffness of the soil is about k = 2.0 * 107 N>m and design the size of the concrete block (i.e., choose m) such that the isolation system reduces the transmitted force by 75%. Assume that the density of concrete is ρ = 23,000 N>m3. The surface area of the cement block is 4 m2. The steady-state operating speed of the compressor is 1800 rpm. m  2000-lb Compressor Asphalt F0 sin t Asphalt Floor M Floor mM k of soil Soil Figure P5.20 A model of a floor-mounted compressor illustrating the use of added mass to design a vibration-isolation system. 5.21. The instrument board of an aircraft is mounted on an isolation pad to protect the panel from vibration of the aircraft frame. The dominant vibration in the aircraft is measured to be at 2000 rpm. because of size limitation in the aircraft’s cabin, the isolators are only allowed to deflect 1>8 in. Find the percent of motion transmitted to the instrument panel if it weighs 50 lb. 5.22. Design a base isolation system for an electronic module of mass 5 kg so that only 10% of the displacement of the base is transmitted into displacement of the module at 50 Hz. What will the transmissibility be if the frequency of the base motion changes to 100 Hz? What if it reduces to 25 Hz? 5.23. Redesign the system of Problem 5.22 such that the smallest transmissibility ratio possible is obtained over the range 50 to 75 Hz. 5.24. A 2-kg printed circuit board for a computer is to be isolated from external vibration of frequency 3 rad>s at a maximum amplitude of 1 mm, as illustrated in Figure P5.24. Design an undamped isolator such that the transmitted displacement is 10% of the base motion. Also calculate the range of transmitted force. Printed circuit board x(t) m Isolator k Chassis c y(t) Base Figure P5.24 Isolation system for a printed circuit board. 5.25. Change the design of the isolator of Problem 5.24 by using a damping material with damping value ζ chosen such that the maximum T.R. at resonance is 2. 5.26. Calculate the damping ratio required to limit the displacement transmissibility to 4 at resonance for any damped isolation system. Chap. 5 495 Problems Section 5.3 (Problems 5.27 through 5.36) 5.27. A motor is mounted on a platform that is observed to vibrate excessively at an operating speed of 6000 rpm, producing a 250-N force. Design a vibration absorber (undamped) to add to the platform. Note that in this case, the absorber mass will only be allowed to move 2 mm because of geometric and size constraints. 5.28. Consider an undamped vibration absorber with β = 1 and μ = 0.2. Determine the operating range of frequencies for which 0 Xk>F0 0 … 0.5. 5.29. Consider an internal combustion engine that is modeled as a lumped inertia attached to ground through a spring. Assuming that the system has a measured resonance of 100 rad>s, design an absorber so that the amplitude is 0.01 m for a (measured) force input of 102 N. 5.30. A small rotating machine weighing 50 lb runs at a constant speed of 6000 rpm. The machine was installed in a building, and it was discovered that the system was operating at resonance. Design a retrofit undamped absorber such that the nearest resonance is at least 20% away from the driving frequency. 5.31. A 3000-kg machine tool exhibits a large resonance at 120 Hz. The plant manager attaches an absorber to the machine of 600 kg tuned to 120 Hz. Calculate the range of frequencies at which the amplitude of the machine vibration is less with the absorber fitted than without the absorber. 5.32. A motor-generator set is designed with steady-state operating speed between 2000 and 4000 rpm. Unfortunately, due to an imbalance in the machine, a large violent vibration occurs at around 3000 rpm. An initial absorber design is implemented with a mass of 2 kg tuned to 3000 rpm. This, however, causes the combined system natural frequencies to occur at 2500 and 3500 rpm. Redesign the absorber so that ω1 6 2000 rpm and ω2 7 4000 rpm, rendering the system safe for operation. 5.33. A rotating machine is mounted on the floor of a building. Together, the mass of the machine and the floor is 2000 lb. The machine operates in steady state at 600 rpm and causes the floor of the building to shake. The floor–machine system can be modeled as a spring–mass system similar to the optical table of Figure 5.14. Design an undamped absorber system to correct this problem. Make sure you consider the bandwidth. 5.34. A pipe carrying steam through a section of a factory vibrates violently when the driving pump hits a speed of 300 rpm. (See Figure P5.34.) In an attempt to design an absorber, a trial 9-kg absorber tuned to 300 rpm was attached. by changing the pump speed, it was found that the pipe–absorber system has a resonance at 207 rpm. Redesign the absorber so that the natural frequencies are 40% away from the driving frequency. Ceiling Pipe Pump Absorber Figure P5.34 A schematic of a steam-pipe system with an absorber attached. 5.35. A machine sorts bolts according to their size by moving a screen back and forth using a primary system of 2500 kg with a natural frequency of 400 cycle>min. Design a 496 Design for Vibration Suppression Chap. 5 vibration absorber so that the machine–absorber system has natural frequencies below 160 cycles>min and above 320 rpm. The machine is illustrated in Figure P5.35. Screen Loose bolts to be sorted Parts tray Absorber 600-cycle/min motor System ma kg Figure P5.35 Model of a parts sorting machine. The parts (bolts here) are placed on a screen that shakes. Parts that are small enough fall through the screen into the tray below. The larger ones remain on the screen. 5.36. A dynamic absorber is designed with μ = 1>4 and ωa = ωp. Calculate the frequency range for which the ratio 0 Xk>F0 0 6 1. Section 5.4 (Problems 5.37 through 5.52) 5.37. A machine, largely made of aluminum, is modeled as a simple mass (of 100 kg) attached to ground through a spring of 2000 N>m. The machine is subjected to a 100-N harmonic force at 20 rad>s. Design an undamped tuned absorber system (i.e., calculate ma and ka) so that the machine is stationary at steady state. Aluminum, of course, is not completely undamped and has internal damping that gives rise to a damping ratio of about ζ = 0.001. Similarly, the steel spring for the absorber gives rise to internal damping of about ζa = 0.0015. Calculate how much this spoils the absorber design by determining the magnitude X using equation (5.32). 5.38. Plot the magnitude of the primary system calculated in Problem 5.37 with and without the internal damping. Discuss how the damping affects the bandwidth and performance of the absorber designed without knowledge of internal damping. 5.39. Derive equation (5.35) for the damped absorber from equations (5.34) and (5.32) along with Window 5.4. Also derive the nondimensional form of equation (5.37) from equation (5.35). Note that the definition of ζ given in equation (5.36) is not the same as the ζ values used in Problems 5.37 and 5.38. *5.40. (Project) If you have a three-dimensional graphics routine available, plot equation (5.37) [i.e., plot (X> ∆) versus both r and ζ for 0 6 ζ 6 1 and 0 6 r 6 3, and a fixed μ and β]. Discuss the nature of your results. Does this plot indicate any obvious design choices? How does it compare to the information obtained by the series of plots given in Figures 5.19 to 5.21? (Three-dimensional plots such as these are commonplace.) *5.41. (Project) Repeat Problem 5.40 by plotting 0 X> ∆ 0 versus r and β for a fixed ζ and μ. *5.42. (Project) The full damped vibration absorber equations (5.32) and (5.33) have not historically been used in absorber design because of the complicated nature of the complex arithmetic involved. However, if you have a symbolic manipulation code available to you, calculate an expression for the magnitude X by using the code to calculate the magnitude Chap. 5 497 Problems and phase of equation (5.32). Apply your results to the absorber design indicated in Problem 5.37 by using ma, ka, and ζa as design variables (i.e., design the absorber). 5.43. A machine of mass 200 kg is driven harmonically by a 100-N force at 10 rad>s. The stiffness of the machine is 20,000 N>m. Design a broadband vibration absorber [i.e., equation (5.37)] to limit the machine’s motion as much as possible over the frequency range 8 to 12 rad>s. Note that other physical constraints limit the added absorber mass to be, at most, 50 kg. 5.44. Often, absorber designs are afterthoughts, such as indicated in Example 5.3.1. Add a damper to the absorber design of Figure 5.17 to increase the useful bandwidth of operation of the absorber system in the event the driving frequency drifts beyond the range indicated in Example 5.3.2 (Recall that m = 73.16 kg, k = 2600 N>m, ma = 18.29 kg, ka = 6500 N>m, and 7.4059 6 ω 6 21.0821 rad>s). 5.45. Again consider the absorber design of Example 5.3.1 (m = 73.16 kg, k = 2600 N>m subject to a force of 13 N at 180 cycles>min constrained to a maximum deflection of 0.2 cm). If the absorber spring is made of aluminum and introduces a damping ratio of ζ = 0.001, calculate the effect of this on the deflection of the saw (primary system) with the design given in Example 5.3.1. 5.46. Consider the undamped primary system with a viscous absorber as modeled in Figure 5.22 and the rotational counterpart of Figure 5.23 repeated in Figure P5.46. Calculate the magnification factor 0 Xk>M0 0 for a 400-kg compressor having a natural frequency of 16.2 Hz if driven at resonance, for an absorber system defined by μ = 0.133 and ζ = 0.025. Torsional stiffness, k 1 2 Primary system Viscous absorber Internal disk of inertia J2 and rotational coordinate 2 ca k m x(t) ma xa(t) Casing of rotational inertia J1 Viscous oil with damping coefficient ca 5.47. Recalculate the magnification factor 0 Xk>M0 0 for the compressor of Problem 5.46 if the damping factor is changed to ζ = 0.1. Which absorber design produces the smallest displacement of the primary system ζ = 0.025 or ζ = 0.1? Figure P5.46 A repeat of Figures 5.22 and 5.23. 5.48. Consider a one-degree-of-freedom model of the nose of an aircraft (A-10) as illustrated in Figure P5.48. The nose cracked under fatigue during battle conditions. This problem has been fixed by adding a viscoelastic material to the inside of the skin to act as a damped vibration absorber as illustrated in Figure P5.48. This fixed the problem and the vibration fatigue cracking disappeared in the A-10s after they were retrofitted with viscoelastic damping treatments. While the actual values remain classified, use the following data to calculate the required damping ratio given the following: M = 100 kg, f = 31 Hz, F0 = 100 N, and k = 3.533 * 106 N>m such that the maximum response is less than 0.25 mm. Since mass always needs to be limited in an aircraft, use μ = 0.1 in your design. 498 Design for Vibration Suppression Chap. 5 F0 cos t Damping layer F0 cos t m Gun k ma ca m k Skin Nose cannon of A-10 aircraft Vibration model without damping Vibration model with damping layer treatment Figure P5.48 Simplified model of the A-10 nose cannon vibration problem. The nose cannon of the A-10 aircraft can be modeled as applying a harmonic force of F0 cos ωt to the skin of the aircraft nose. The skin can be modeled as a spring–mass system based on the stiffness model of Figure 1.26 (i.e., k = 3EI >l3). 5.49. Plot an amplification curve (such as Figure 5.24) by using equation (5.39) for ζ = 0.02 after several values of μ (μ = 0.1, 0.25, 0.5, and 1). Can you form any conclusions about the effect of the mass ratio on the response of the primary system? Note that as μ gets large, 0 (Xk>M0) 0 gets very small. What is wrong with using very large μ in an absorber design? 5.50. A Houdaille damper is to be designed for an automobile engine. Choose a value for ζ and μ if the magnification 0 (Xk>M0) 0 is to be limited to 4 at resonance. (One solution is μ = 1, ζ = 0.129.) 5.51. Determine the amplitude of vibration for the various dampers of Problem 5.46 if ζ = 0.1 and F0 = 100 N. 5.52. (Project) Use your knowledge of absorbers and isolation to design a device that will protect a mass from both shock inputs and harmonic inputs. It may help to have a particular device in mind such as the module discussed in Figure 5.6. Section 5.5 (Problems 5.53 through 5.66) 5.53. Design a Houdaille damper for an engine modeled as having an inertia of 1.5 kg # m2 and a natural frequency of 33 Hz. Choose a design such that the maximum dynamic magnification is less than 6: ` Xk ` 6 6 M0 The design consists of choosing J2 and ca, the required optimal damping. 5.54. Recall the optimal vibration absorber of Problem 5.53. This design is based on a steady-state response. Calculate the response of the primary system to an impulse of magnitude M0 applied to the primary inertia J1. How does the maximum amplitude of the transient compare to that in steady state? 5.55. Consider the damped vibration absorber of equation (5.37) with β fixed at β = 1>2 and µ fixed at µ = 0.25. Calculate the value of ζ that minimizes 0 X> ∆ 0 . Plot this function for several values of 0 6 ζ 6 1 to check your design. If you cannot solve this analytically, consider using a three-dimensional plot of 0 X> ∆ 0 versus r and ζ to determine your design. Chap. 5 499 Problems 5.56. For a Houdaille damper with mass ratio µ = 0.25, calculate the optimum damping ratio and the frequency at which the damper is most effective at reducing the amplitude of vibration of the primary system. 5.57. Consider again the system of Problem 5.53. If the damping ratio is changed to ζ = 0.1, what happens to 0 Xk>M0 0 ? 5.58. Derive equation (5.42) from equation (5.35) and derive equation (5.49) for the optimal damping ratio. 5.59. Consider the design suggested in Example 5.5.1 (mass moment of inertia of 1.5 kg m2 >rad, torsional stiffness of 6 * 103 N m>rad, and a running speed of 2000 rpm). Calculate the percent change in the maximum deflection if the damping constant changes 10% from its optimal value. If the optimal damping is fixed but the mass of the absorber changes by 10%, what percent change in 0 Xk>M0 0 max results? Is the optimal absorber design more sensitive to changes in ca or ma? 5.60. Consider the elastic isolation problem described in Figure 5.26 and repeated in Figure P5.60. Derive equations (5.54) and (5.55) from equation (5.53). F0 cost m x1(t) c k1 k2 x2(t) Figure P5.60 A repeat of Figure 5.26 for reference in the following problems. 5.61. Use the derivative calculation for finding maximum and minimum to derive equations (5.57) and (5.58) for the elastic damper system. 5.62. A 1000-kg mass is suspended from ground by a 40,000-N>m spring. A viscoelastic damper is added, as indicated in Figure P5.60. Design the isolator (choose k2 and c) such that when a 70-N sinusoidal force is applied to the mass, no more than 100 N is transmitted to ground. 5.63. Consider the isolation design of Example 5.5.2 (c = 188.56 kg>s and k2 = 200 N>m with r = 5 and γ = 0.5). If the value of the damping coefficient changes 10% from the optimal value (of 188.56 kg>s), what percent change occurs in (T.R.)max? If c remains at its optimal value and k2 changes by 10%, what percent change occurs in (T.R.)max? Is the design of this type of isolation more sensitive to changes in damping or stiffness? 5.64. A 3000-kg machine is mounted on an isolator with an elastically coupled viscous damper such as indicated in Figure P5.60. The machine stiffness (k1) is 2.943 * 106 N>m, γ = 0.5, and c = 56.4 * 103 N # s>m. The machine, a large compressor, develops a harmonic force of 1000 N at 7 Hz. Determine the amplitude of vibration of the machine. 5.65. Again consider the compressor isolation design given in Problem 5.64. If the isolation material is changed so that the damping in the isolator is changed to ζ = 0.15, what is the force transmitted? Next determine the optimal value for the damping ratio and calculate the resulting transmitted force. 500 Design for Vibration Suppression Chap. 5 5.66. Consider the optimal vibration isolation design of Problem 5.65. Calculate the optimal design if the compressor’s steady-state driving frequency changes to 24.7 Hz. If the wrong optimal point is used (i.e., if the optimal damping for the 7-Hz driving frequency is used), what happens to the transmissibility ratio? Section 5.6 (Problems 5.67 through 5.73) 5.67. Compare the resonant amplitude at steady state (assume a driving frequency of 100 Hz) of a piece of nitrile rubber at 50°F versus the value at 75°F. Use the values for η from Table 5.2. 5.68. Using equation (5.67), calculate the new modulus of a 0.05 * 0.01 * 1 m piece of pinned–pinned aluminum covered with a 1-cm-thick piece of nitrile rubber at 75°F driven at 100 Hz. 5.69. Calculate Problem 5.68 again at 50°F. What percent effect does this change in temperature have on the modulus of the layered material? 5.70. Repeat the design of Example 5.6.1 (recall E1 = 7.1 * 1010 N>m2 and h2 = 1) by (a) changing the operating frequency to 1000 Hz, and (b) changing the operating temperature to 50°F. Discuss which of these designs yields the most favorable system. 5.71. Reconsider Example 5.6.2. Make a plot of thickness of the damping treatment versus loss factor. 5.72. Calculate the maximum transmissibility coefficient of the center of the shelf of Example 5.6.1. Make a plot of the maximum transmissibility ratio for this system frequency using Table 5.1 for each temperature. 5.73. The damping ratio associated with steel is about ζ = 0.001. Does it make any difference whether the shelf in Example 5.6.1 is made out of aluminum or steel? What percent improvement in damping ratio at resonance does the rubber layer provide the steel shelf? Section 5.7 (Problems 5.74 through 5.80) 5.74. A 100-kg compressor rotor has a shaft stiffness of 1.4 * 107 N>m. The compressor is designed to operate at a speed of 6000 rpm. The internal damping of the rotor shaft system is measured to be ζ = 0.01. (a) If the rotor has an eccentric radius of 1 cm, what is the rotor system’s critical speed? (b) Calculate the whirl amplitude at critical speed. Compare your results to those of Example 5.7.1. 5.75. Redesign the rotor system of Problem 5.74 such that the whirl amplitude at critical speed is less than 1 cm by changing the mass of the rotor. 5.76. Determine the effect of the rotor system’s damping ratio on the design of the whirl amplitude at critical speed for the system of Example 5.7.1 (r = 1 and α = 0.001 m) by plotting X at r = 1 for ζ between 0 6 ζ 6 1. 5.77. Consider the design of the compressor rotor system of Example 5.7.1 (r = 1 and α = 0.001 m). The amplitude of the whirling motion depends on the parameters a, ζ, m, k, and the driving frequency. Which parameter has the greatest effect on the amplitude? Discuss your results. Chap. 5 501 Problems 5.78. The flywheel of an automobile engine has a mass of about 50 kg and an eccentricity of about 1 cm. The operating speed ranges from 1200 rpm (idle) to 5000 rpm (red line). Choose the remaining parameters so that the whirling amplitude is never more than 1 mm. 5.79. At critical speed the amplitude is determined entirely by the damping ratio and the eccentricity. If a rotor has an eccentricity of 1 cm, what value of damping ratio is required to limit the deflection to 1 cm? 5.80. A rotor system has damping limited by ζ 6 0.05. What is the maximum value of eccentricity allowable in the rotor design if the maximum amplitude at critical speed must be less than 1 cm? MATLAB ENGINEERING VIBRATION TOOLBOX If you have not yet used the Engineering Vibration Toolbox program, return to the end of Chapter 1 or Appendix G for a brief introduction to using MATlAb files. The files contained in folder VTb5 may be used to help solve the preceding problems. The M-files from earlier chapters (VTbx.x.M, etc.) may also be useful. The following Toolbox problems are intended to help you gain some experience with the concepts introduced in this chapter for designing vibrating systems and to build experience with the various formulas. These problems may also be solved using any of the codes introduced in Sections 1.9, 1.10, 2.8, 2.9, 3.8, 4.9, and 4.10. TOOLBOX PROBLEMS TB5.1. Use file VTb5_1 to reproduce the plots of Figure 5.5 by inputting various values of m, c, and k. First fix m and k and vary c. Then fix m and c and vary k. TB5.2. Use file VTb5_2 to verify the solution of Example 5.2.1 for the magnitude of the force transmitted to the electronic module. TB5.3. Use file VTb5_3 to examine what happens to the shaded region in Figure 5.15 as μ is varied. Do this by increasing the absorber mass (μ = ma >m) so that µ varies in increments of 0.1 from 0.1 to 1 for a fixed value of β. TB5.4. Examine the effect of β on Figure 5.16 by using file VTb5_4 to plot Figure 5.16 over again for several different values of β (β = 0.1, 0.5, 1, 1.5). What do you notice? What does changing β correspond to in terms of choosing the values of the absorber design? TB5.5. Consider the amplitude plot of the damped absorber given in Figure 5.19. Use file VTb5_5 to see the effect changing the primary mass m has on the design. Choose ma = 10, ca = 1, k = 1000, ka = 1000. Plot 0 Xk>F0 0 for various values of the primary mass m (i.e., m = 1, 10, 100, 1000) versus r. Can you draw any conclusions? TB5.6. File VTb5_6 plots the three-dimensional mesh used to generate Figure 5.25, which illustrates the optimal damped absorber design. Use this program to note the effects of changing the spring–mass on the values of ζ corresponding to the absolute minimum of the maximum response. 6 Distributed-Parameter Systems So far this book has focused on the vibration of rigid bodies. This chapter introduces the analysis needed to describe the vibration of systems that have flexible components. Flexibility of structural components arises when the mass and stiffness properties are modeled as being distributed throughout the spatial definition of the component rather than at lumped positions, as done in Chapter 4. Examples of such systems are the wings and panels of aircraft such as the Reaper pictured on the left. The vibrations of the wing of a commercial aircraft can usually be seen during takeoff and landing or during turbulence. The blades of the wind turbine in the bottom photo form another example of a distributed-parameter system. increased reliance on wind energy has promoted larger, and hence more flexible, wind turbine blades. many structures, such as wings, blades, and other components, can be modeled by the simple string, beam, and plate models discussed in this chapter. many systems, such as truck chassis, buildings, dance floors, and computer disk drives, can be modeled and analyzed by the methods presented in the chapter. The major concept presented here is that distributed-parameter systems have an infinite number of natural frequencies. The concepts of mode shapes and modal analysis used in Chapter 4 are extended here to treat the vibration of distributedparameter systems. 502 Chap. 6 Distributed-Parameter Systems 503 In previous chapters, all systems considered are modeled as lumped-parameter systems; that is, the motion of each point in the system under consideration is modeled as if the mass were concentrated at that point. Multiple-degree-offreedom systems are considered as arrangements of various lumped masses separated by springs and dampers. In this sense, the parameters of the system are discrete sets of finite numbers. Hence, such systems are also called discrete systems or finite dimensional systems. In this chapter, the flexibility of structures is considered. Here the mass of an object is considered to be distributed throughout the structure as a series of infinitely small elements. When a structure vibrates, each of these infinite number of elements move relative to each other in a continuous fashion. Hence these systems are called infinite-dimensional systems, continuous systems, or distributed-parameter systems. The choice of modeling a given mechanical system as a lumped-parameter system or a distributed-parameter system depends on the purpose at hand as well as the nature of the object. There are only a few distributed-parameter models that have closed-form solutions. However, these solutions provide insight into a large number of problems that cannot be solved in closed form. The time response of a distributed-parameter system is described spatially by a continuous function of the relative position along the system. In contrast, the time response of a lumped-parameter system is described spatially by labeling a discrete number of points throughout the system in the form of a vector. Here the terms lumped parameter and distributed parameter are used rather than discrete and continuous to avoid confusion with discrete-time systems (used in numerical integration and measurement). The specific cases considered here are the vibrations of strings, rods, beams, membranes, and plates. Common examples of such structures are a vibrating guitar string and the swaying motion of a bridge or tall building. In addition, systems having both lumped parameters and distributed parameters are considered. The single-degree-of-freedom systems discussed in Chapters 1 through 3 have only one natural frequency ωn = 1k>m. In Chapter 4, multiple-degree-of-freedom systems introduce the concept of multiple natural frequencies, denoted by ωi, one frequency for each degree of freedom. Design to avoid resonance becomes more difficult with multiple natural frequencies because of the increased chance that a harmonic driving frequency will correspond to one of the natural frequencies, causing resonance. The distributed-parameter systems considered in this chapter have an infinite number of degrees of freedom and hence an infinite number of natural frequencies, again increasing concern about resonance in design. The numerous frequencies of a distributed-parameter system are also denoted by ωn for one-dimensional structures and ωnm for structures defined in a plane. There is thus a slight notational inconsistency that persists in the vibration literature by using ωn for both the natural frequency of a single-degree-of-freedom system and to mean the nth natural frequency of a distributed system. The distinction is ultimately clear from the context of usage. 504 Distributed-Parameter Systems Chap. 6 6.1 VIBRATION OF A STRING OR CABLE String instruments (guitars, violins, etc.) provide an excellent and intuitive example of the vibration of a distributed-parameter object. Strings are also the easiest system to solve and provide a systematic way to approach other distributed-parameter structures, much like the simple spring–mass system formed the basics for analysis for the lumped-parameter system analysis. Consider the string of Figure 6.1 with mass density ρ and cross-sectional area A, fixed at both ends and under a tension denoted by τ. The string moves up and down in the y direction. The motion at any point on the string must be a function of both the time, t, and the position along the string, x. The deflection of the string is thus denoted by (x, t). let f(x, t) be an external force per unit length also distributed along the string, and consider the infinitesimal element (∆x long) of the displaced string indicated in Figure 6.1. The net force acting on the infinitesimal element in the y direction must be equal to the inertial force in the y direction, ρA∆x(02w>0t 2) so that -τ1 sin θ1 + τ2 sin θ2 + f(x, t)∆x = ρA∆x 02w(x, t) (6.1) Note that the acceleration is stated in terms of partial derivatives (02 >0t 2) because is a function of two variables. The expressions in equation (6.1) can be approximated in the case of small deflections so that θ1 and θ2 are small. In this case, τ1 and τ2 can easily be related to the initial tension in the string τ by noting that the horizontal component of the deflected string tension is τ1 cos θ1 at end 1, and τ2 cos θ2 at y 0t 2 0 w(x, t) l x f(x, t) Guitar f(x, t)  0 2 1 l 2 1 w(x, t) x1 x2  x1  x Infinitesimal element of string displaced from rest Figure 6.1 The geometry of a vibrating string with applied force f(x, t) and displacement (x, t). Sec. 6.1 505 Vibration of a String or Cable end 2. In the small-angle approximation, cos θ1 ≃ 1 and cos θ2 ≃ 1, so it is reasonable to set τ1 = τ2 = τ. Also, for small θ1, sin θ1 ≃ tan θ1 = and 0w(x, t) ` 0x x1 (6.2) 0w(x, t) ` 0x x2 (6.3) where (0w>0x) 0 x1 is the slope of the string at point x1 and (0w>0x) 0 x2 is the slope of the string at point x2 = x1 + ∆x. The notation for partial and total derivatives is reviewed in Window 6.1. With these approximations, equation (6.1) now becomes sin θ2 ≃ tan θ2 = aτ 0w(x, t) 0w(x, t) 02w(x, t) ∆x b ` - aτ b ` + f(x, t)∆x = ρA 0x 0x 0t 2 x2 x1 (6.4) The slopes can be evaluated further by recalling the Taylor series expansion for the function τ(0w>0x) around the point x1 from calculus. This yields aτ 0w 0w 0 0w b ` = aτ b ` + ∆x aτ b ` + O(∆x2) 0x x2 0x x1 0x 0x x1 (6.5) Window 6.1 Notation for Vibrations Described by Distributed Mass and Stiffness Derivatives of a function of multiple variables, such as f(x, y), deal with partial derivatives with the following notation: fx(x, y) = f (x + ∆x, y) - f (x, y) 0f(x, y) = lim a b 0x ∆x ∆x S 0 In contrast, the total derivative of a function of a single variable, say T(t), is denoted by $ dT(t) T(t + ∆t) - T(t) = T(t) = lim dt ∆t ∆t S 0 The exact differential is df = 0f 0f dx + dy 0x 0y 506 Distributed-Parameter Systems Chap. 6 where O(∆x2) denotes the rest of the Taylor series, which consists of terms of order ∆x2 and higher. Since ∆x is small, O(∆x2) is even smaller and hence neglected. Substitution of expression (6.5) into equation (6.4) yields 02w(x, t) 0 0w ∆x aτ b ` ∆x + f(x, t)∆x = ρA 0x 0x x1 0t 2 (6.6) Dividing by ∆x and realizing that since ∆x is infinitesimal, the designation of point 1 becomes unnecessary, the equation of motion for the string becomes 0w(x, t) 02w(x, t) 0 aτ b + f(x, t) = ρA 0x 0x 0t 2 (6.7) Since the tension τ is constant, and if the external force is zero, this becomes c2 02w(x, t) 0x2 = 02w(x, t) 0t 2 or 02w(x, t) 0x2 = 1 02w(x, t) c2 0t 2 (6.8) where c = 1τ>ρA depends only on the physical properties of the string (called the wave speed and is not to be confused with the symbol used for the damping coefficient in earlier chapters). Equation (6.8) is the one-dimensional wave equation, also called the string equation, and is subject to two initial conditions in time because of the dependence on the second time derivative. These are written as # w(x, 0) = w0(x) and wt(x, 0) = w0(x), where the subscript t is an alternative nota# tion for the partial derivative 0>0t, and where w0(x) and w0(x) are the initial displacement and velocity distributions of the string, respectively. The second spatial derivative in equation (6.8) implies that two other conditions must be applied to the solution w(x, t) in order to calculate the two constants of integration arising from integrating these spatial derivatives. These conditions come from examining the boundaries of the string. In the configuration of Figure 6.1 the string is fixed at both ends (i.e., at x = 0 and x = l). This means that the deflection w(x, t) must be zero at these points so that w(0, t) = w(l, t) = 0 t 7 0 (6.9) These two conditions at the boundary provide the other two constants of integration resulting from the two spatial derivatives of w(x, t). because these conditions occur at the boundaries, the problem described by equations (6.8) and (6.9) and the initial conditions is called a boundary-value problem. Using the subscript notation for partial differentiation, the various derivatives of the deflection, w(x, t), have the following physical interpretations. The quantity wx(x, t) denotes the slope of the string, while τwxx(x, t) corresponds to the restoring force of the string (i.e., the string’s stiffness or elastic property). The quantity wt(x, t) is the velocity and wtt(x, t) is the acceleration of the string at any point x and time t. Sec. 6.1 507 Vibration of a String or Cable The string vibration problem just described forms a convenient and simple model to study the vibration of distributed-parameter systems. This is analogous to the spring–mass model of Section 1.1, which provided a building block for the study of lumped-parameter systems. To that end, the string with fixed endpoints is used in the next section to develop general techniques of solving for the vibration response of distributed-parameter systems. More about the string equation and its use in wave propagation can be found in introductory physics texts. Note that these developments apply to cables as well as strings. Example 6.1.1 Consider the cable of Figure 6.2, which is pinned at one end and attached to a spring at the other end held in a frictionless slider such that the cable remains in constant tension. Determine the governing equation for the vibration of the system. 0 . w(x, t) l k y Figure 6.2 A cable fixed at one end and attached to a spring at the other end with a frictionless slide. Note that the motion w(x, t) is in the y direction. x Solution The equations of motion for the spring and the cable are the same and are given by equation (6.8). The initial conditions are also unaffected. However, the boundary condition at x = l changes. Writing a force balance in the y direction at point x = l yields a F 0 x = l = τ sin θ + kw(l, t) = 0 y where k is the stiffness of the (lumped) spring. Again, enforcing the small-angle approximation, this becomes τ 0w(x, t) 0x ` x=l = - kw(x, t) 0 x = l The boundary condition at x = 0 remains unchanged. n 508 Distributed-Parameter Systems Chap. 6 6.2 MODES AND NATURAL FREQUENCIES In this section the string equation is solved for the case of fixed–fixed boundary conditions using the technique of separation of variables. This method leads in a natural way to modal analysis and the concepts of mode shapes and natural frequencies for distributed-parameter systems that are used extensively for lumped-parameter systems in Chapter 4. The solution procedures are described in detail in introductory differential equations (see boyce and DiPrima, 2009, for instance) and reviewed here. First, it is assumed that the displacement w(x, t) can be written as the product of two functions, one depending only on x and the other depending only on t (hence separation of variables). Thus w(x, t) = X(x)T(t) Substitution of this separated form into the string equation (6.8) yields $ c 2X″(x)T(t) = X(x)T(t) (6.10) (6.11) where the primes on X″(x) denote total differentiation (twice in this case) with respect to x, and the overdots indicate total differentiation (twice in this case) with respect to the time, t. These derivatives become total derivatives, instead of partial derivatives, because the functions X(x) and T(t) are each a function of a single variable. A simple rearrangement of equation (6.11) yields $ T(t) X″(x) = 2 X(x) c T(t) (6.12) Note that the choice of which side in equation (6.12) to put the constant c2 is arbitrary. Some choose to place it on the left side and some, as done here, on the right side. The final solution remains the same. Since each side of the equation is a function of a different variable, it is argued that each side must be constant. To see this, differentiate with respect to x. This yields d X″ a b = 0 dx X (6.13) X″ = constant = -σ2 X (6.14) which becomes, upon integration, In this case –σ2 is the constant chosen to ensure that the quantity on the right side of equation (6.14) is negative. Actually, all possible choices (negative, positive, and zero) for this constant need to be considered. The other two choices (positive and Sec. 6.2 509 Modes and natural Frequencies zero) lead to physically unacceptable results as discussed in Example 6.2.1 to follow. Equation (6.14) also requires that $ T(t) (6.15) = -σ2 c 2T(t) in order to satisfy equation (6.12). Rearranging equation (6.14) yields the result that the function X(x) must satisfy X″(x) + σ2X(x) = 0 (6.16) Equation (6.16) has the solution (see Example 6.2.1) X(x) = a1 sin σx + a2 cos σx (6.17) where a1 and a2 are constants of integration. To solve for these constants of integration, consider the boundary conditions of equation (6.9) in the separated form implied by equation (6.10). They become X(0)T(t) = 0 and X(l)T(t) = 0 t 7 0 (6.18) Since it is assumed that T(t) cannot be zero for all t [this would yield only the uninteresting solution w(x, t) = 0 for all time], equation (6.18) reduces to X(0) = 0 X(l) = 0 (6.19) Applying these two conditions to the solution, equation (6.17), yields the two simultaneous equations X(0) = a2 = 0 X(l) = a1 sin σl = 0 (6.20) The first of these two expressions eliminates the cosine term in the solution, and the last of these two expressions yields values of σ by requiring that sin σl = 0. This is called the characteristic equation, which has solutions σl = nπ. Since the sine function vanishes when its argument is 0 and any integer multiple of π, there is one solution of the characteristic equation for each value of n = 0, 1, 2, 3, c Hence, there exists an infinite number of values of σ that satisfy the condition σl = nπ. Thus, σ is indexed to be σn and takes on the following values (n = 0 implies a zero solution): σn = nπ l n = 1, 2, 3, c (6.21) The two simultaneous equations resulting from the boundary conditions given by equation (6.20) can also be written in matrix form as J sin σl 0 0 a1 0 RJ R = J R 1 a2 0 510 Distributed-Parameter Systems Chap. 6 Recall from Section 4.1, equation (4.19), that this vector equation has a nonzero solution (for a1 and a2) as long as the coefficient matrix has a zero determinant. Thus this alternative formulation yields the characteristic equation det J sin σl 0 0 R = sin σl = 0 1 This provides a more systematic calculation of the characteristic equation from the statement of the boundary conditions. Since there are an infinite number of values of σn, the solution (6.17) then becomes the infinite number of solutions Xn(x) = an sin a nπ xb l n = 1, 2, c (6.22) Here X is now indexed by n because of its dependence on σn, and the an are stillto-be-determined arbitrary constants, potentially depending on the index n as well. Equation (6.22) forms the spatial solution of the vibrating string problem. The functions Xn(x) satisfy the boundary value problem 02 (Xn) = λnXn 0x2 Xn(0) = Xn(l) = 0 (6.23) where λ n are constants ( λn = σ2n ) and where the function Xn is never identically zero over all values of x. Comparison of this with the definition of the matrix eigenvalue and eigenvector problem of Section 4.2 yields some very strong similarities. Instead of an eigenvector consisting of a column of constants, equation (6.23) defines the eigenfunctions, Xn(x), which are nonzero functions of x satisfying boundary conditions as well as a differential equation. The constants λ n are called eigenvalues just as in the matrix case. The differential operator -02 >0x2 takes the place of the matrix in this eigenproblem. Similar to the eigenvector of Chapter 4, the eigenfunction is only known to within a constant. That is, if Xn(x) is an eigenfunction, so is aXn(x), where a is any constant. In fact, the concept of eigenvector and that of eigenfunction are mathematically identical. As is illustrated in the following, the eigenvalues, as determined by the characteristic equation given in this case by expression (6.20), and the eigenfunctions, described in (6.22), will determine the natural frequencies and mode shapes of the vibrating string. To this end, consider next the temporal equation given by (6.15) with the quantities of (6.21) substituted for σn: $ (6.24) Tn(t) + σ2nc 2Tn(t) = 0 n = 1, 2, c Sec. 6.2 511 Modes and natural Frequencies where T(t) is now indexed because there is one solution for each value of σn. The coefficient of Tn(t) in the temporal equation defines the natural frequency by noting that ωn = cσn and hence ωn = cσn = τ nπ rad>s l A ρA The general form of the solution of (6.24) is given in Window 1.4 as Tn(t) = An sin ωnt + Bn cos ωnt (6.25) where An and Bn are constants of integration. Since both of the functions Xn(x) and Tn(t) are found to be dependent on n, the solution w(x, t) = Xn(x)Tn(t) must also be a function of n, so that wn(x, t) = cn sin a nπ nπc nπ nπc xb sin a tb + d n sin a xb cos a tb l l l l n = 1, 2, c (6.26) where cn and dn are new constants to be determined. Note that an unknown constant an times another unknown constant An is the unknown constant cn (similarly, dn = anBn). Since the string equation is linear, any linear combination of solutions is a solution. Hence the general solution is of the form w(x, t) = a (cn sin σnx sin σnct + d n sin σnx cos σnct) ∞ (6.27) n=1 The set of constants {cn} and {d n} can be determined by applying the initial conditions on (x, t) and the orthogonality of the set of functions sin (nπx>l). The orthogonality of the set of functions sin (nπx>l) states that l nπx mπx sin dx = • 2 sin l l L0 0 l n = m (6.28) n ≠ m which is identical to the orthogonality of mode shape vectors discussed in Section 4.2. This orthogonality condition can be derived by simple trigonometric identities and integration (as suggested in Section 3.3). Consider the initial condition on the displacement applied to equation (6.27): w(x, 0) = w0(x) = a d n sin ∞ n=1 nπx cos(0) l (6.29) 512 Distributed-Parameter Systems Chap. 6 Multiplying both sides of this equality by sin mπx>l and integrating over the length of the string yields L0 l ∞ mπx nπx mπx l dx = a d n sin sin dx = d m a b l l l 2 L0 n=1 l w0(x) sin (6.30) where each term in the summation on the right-hand side of equation (6.29) is zero except for the mth term, by direct application of the orthogonality condition of (6.28). Equation (6.30) must hold for each value of m, so that l dm = 2 mπx w0(x) sin dx m = 1, 2, 3, c l L0 l (6.31) It is customary to replace m by n in the preceding, since the index runs over all positive values [i.e., the index in equation (6.31) is a free index and it does not matter what it is named]. It is most convenient to rename it dn. A similar expression for the constants {cn} is obtained by using the initial velocity condition. Time differentiation of the summation of equation (6.27) yields ∞ nπx # cos (0) w0(x) = wt (x, 0) = a cnσnc sin l n=1 (6.32) Again, multiplying by sin mπx>l, integrating over the length of the string, and applying the orthogonality condition of (6.28) yields l cn = 2 nπx # w0(x) sin dx n = 1, 2, 3, c nπc L0 l (6.33) where the index has been renamed n. Equations (6.31) and (6.33) combined with equation (6.27) form the complete solution for the vibrating string (i.e., they describe the vibration response of the string at any point x and any time t). Example 6.2.1 The solution of a second-order ordinary differential equation with constant coefficients subject to boundary conditions is used throughout this section. This example clarifies the choice of a negative constant for the separation-of-variables procedure and reviews the solution technique for second-order differential equations with constant coefficients (see, e.g., boyce and DiPrima, 2009). Consider again equation (6.14), where this time the separation constant is chosen to be –β, where β is of arbitrary sign. This yields X″ + βX = 0 (6.34) Assume that the solution of (6.34) is of the form X(x) = eλx, where λ is to be determined. Substitution into (6.34) yields (λ2 + β)e λx = 0 (6.35) Sec. 6.2 513 Modes and natural Frequencies Since e λx is never zero, this requires that λ = { 1- β (6.36) Here λ will be purely imaginary or real, depending on the sign of the separation constant β. Thus there are two solutions and the general solution is the sum X(x) = Ae -1-βx + Be +1-βx (6.37) where A and B are constants of integration to be determined by the boundary conditions. Applying the boundary conditions to (6.37) yields X(0) = A + B = 0 X(l) = Ae -1-βl + Be 1-βl = 0 (6.38) This system of equations can be written in the matrix form J 1 e -1-βl 1 A 0 RJ R = J R e 1-βl B 0 which has a nonzero solution for A and B if and only if (recall Section 4.1) the determinant of the matrix coefficient is zero. This yields e +1-βl - e -1-βl = 0 (6.39) 2 which must be satisfied. For negative real values of β, say β = –σ , this becomes [recall the definition sinh u = (eu – e–u)>2] e σl - e -σl = 2 sinh σl = 0 (6.40) which has only the trivial solution and hence β ≠ -σ2. This means that the separation constant σ in the development of equation (6.14) cannot be positive. Thus 1-β must be complex (i.e., β = σ2), so that equation (6.39) becomes e σjl - e -σjl = 0 (6.41) where j = 1- 1. Euler’s formula for the sine function is sin u = (euj – e–uj)>2j, so that (6.41) becomes sin σl = 0 (6.42) which has the solution σ = nπ>l as used in equation (6.20). The only possibility that remains to check is the case β = 0, which yields X″ = 0, or, upon integrating twice, X(x) = a + bx (6.43) Applying the boundary conditions results in X(0) = a = 0 X(l ) = bl = 0 (6.44) which yields only the trivial solution a = b = 0. Hence the choice of separation constant in equation (6.14) as –σ2 is fully justified, and the solution of the spatial equation for the string fixed at both ends is of the form 514 Distributed-Parameter Systems Xn(x) = an sin nπx l Chap. 6 n = 1, 2, 3, c Here the subscript n has been added to X(x) to denote its dependence on the index n and to indicate that more than one solution results. In this case, an infinite number of solutions result, one for each integer n. n Note that the solution of this spatial equation is very much like the solution of the equation of the single-degree-of-freedom oscillator in Section 1.1 and reviewed in Window 6.2. There are, however, two main differences. First, the sign of the coefficient in the single-degree-of-freedom oscillator is determined on physical grounds (i.e., the ratio of stiffness to mass is positive), and second, the constants of integration were both evaluated at the beginning of the interval instead of one at each end. The spatial string equation is a boundary-value problem, while the single-degreeof-freedom oscillator equation is an initial-value problem. Note that the temporal equation for the string, however, is identical to the equation of a single-degree-offreedom oscillator. The temporal equation is also an initial-value problem. Window 6.2 Review of the Solution of a Single-Degree-of-Freedom System $ # The solution to mx + kx = 0, x(0) = x0, x(0) = v0 for m, k 7 0 is x(t) = where ωn = 1k>m. ωnx0 2ω2nx20 + v20 b sin aωnt + tan-1 ωn v0 (1.10) Now that the mathematical solution of the vibrating string is established, consider the physical interpretation of the various terms in equation (6.27). Consider giving the string an initial displacement of the form w0(x) = sin πx l (6.45) # and an initial velocity of w0(x) = 0. With these values of the initial conditions, the computation of the coefficients in (6.27) by using equations (6.31) and (6.33) yields cn = 0 for n = 1, 2, c, dn = 0 for n = 2, 3, c, and d1 = 1. Substitution of these coefficients back into equation (6.27) yields the solution w(x, t) = sin πx πc cos t l l (6.46) Sec. 6.2 515 Modes and natural Frequencies which is the first term of the series. Comparing the second factor of equation (6.46) to cos ωt states that the string is oscillating in time at a frequency of ω1 = ω1 πc π τ = rad>s or f1 = Hz l l A ρA 2π (6.47) in the spatial shape of sin (πx>l). Hence for a fixed time t, the string will be deformed in the shape of a sinusoid. Each point of the string is moving up and down in time (i.e., vibrating at frequency πc>l). borrowing the jargon of Chapter 4, the function X1(x) = sin (πx>l) is called a mode shape, or mode, of the string and the quantity (πc>l) is called a natural frequency of the string. This procedure can be repeated for each value of the index n by choosing the initial displacement to be sin (nπx>l) and the velocity to be zero. This gives rise to an infinite number of mode shapes, sin (nπx>l), and natural frequencies, (nπc>l), and is the reason distributed-parameter systems are called infinite-dimensional systems. Figure 6.3 illustrates the first two mode shapes of the fixed-endpoint string. If the string is excited in the second mode by using the initial conditions w0 = sin (2πx>l), # w0 = 0, and the resulting motion is viewed by a stroboscope flashing with frequency 2πc>l, the curve labeled n = 2 in Figure 6.3 would be visible. An interesting property of the modes of a string are the points where the modes sin (nπx>l) are zero. These points are called nodes. Note from Figure 6.3 that the node of the second mode is at the point l>2. If the string is given an initial displacement equal to the second mode, there will be no motion at the point l>2 for any value of t. This node, as in the case of lumped-parameter systems, is a point on the string that does not move if the structure is excited only in that mode. The modes and natural frequencies for a string and those for a multiple-degreeof-freedom system are very similar. For a multiple-degree-of-freedom system, the mode shape is a vector, the elements of which yield the relative amplitude of vibration of each coordinate of the system. The same is true for an eigenfunction of the string, the difference being that instead of elements of a vector, the eigenfunction gives the modal response magnitude at each value of the position x along the string. Equation (6.27) is the distributed-parameter equivalent of the modal expansion given by equation (4.103) for lumped-parameter systems. Instead of an expansion of the solution vector x(t) in terms of the modal vectors ui indicated in equation (4.103) w(x, t) n1 l/2 0 l n2 x Figure 6.3 A plot of the deflection w(x, t) versus the position x for a fixed time illustrating the first two mode shapes of a vibrating string fixed at both ends. 516 Distributed-Parameter Systems Chap. 6 Window 6.3 A Review of the Modal Expansion Solution of a Multiple-Degree-of-Freedom System $ # # The solution of Mx + Kx = 0, x(0) = x0, x(0) = x0 is x(t) = a ci sin (ωit + ϕi)u i n (4.103) i=1 where ui are the eigenvectors of the matrix M -1>2KM -1>2 and ω2i are the associated eigenvalues of M -1>2KM -1>2. The constants ci and ϕi are determined from the initial conditions by using the orthogonality of ui to obtain ci = u Ti Mx(0) sin ϕi ϕi = tan-1 ωiu Ti Mx(0) # u Ti Mx(0) and reviewed in Window 6.3, the distributed-parameter equivalent consists of an expansion in terms of the mode shape functions sin σnx. Just as the mode shape vectors determine the relative magnitude of the motion of various masses of the lumped system, the mode shape functions determine the magnitude of the motion of the mass distribution of the distributed-parameter system. The procedures outlined in this section constitute the modal analysis of a distributed-parameter system. This basic procedure is repeated in the following sections to determine the vibration response of a variety of different distributed-parameter systems. Modal analysis for damped systems is given in Section 6.7 and for the forced response in Section 6.8. Example 6.2.2 For the sake of obtaining a feeling for the units of a vibrating string, consider a piano wire. A reasonable model of a piano wire is the string fixed at both ends of Figure 6.1. A piano wire is 1.4- m long, has a mass of about 110 g, and has a tension of about τ = 11.1 * 104 N. Its first natural frequency is then ω1 = πc π τ = rad>s l 1.4 m A ρA Since ρA = 110 g per 1.4 m = 0.0786 kg>m, this yields ω1 = π 11.1 * 104 N 1>2 a b = 2666.69 rad>s or f1 = 424 Hz 1.4 m 0.0786 kg>m n Sec. 6.2 517 Modes and natural Frequencies Example 6.2.3 Calculate the mode shapes and natural frequencies of the spring–cable system of Example 6.1.1. The solution of the string equation will be the same as that presented for the string fixed at both ends except for the evaluation of the constants by using the boundary conditions. Solution Referring to equation (6.17) and applying the boundary conditions given in Example 6.1.1 yields X(0) = a2 = 0 (6.48) at one end. At the other end, substitution of w(x, t) = X(x)T(t) into the boundary condition yields τX =(l)T(t) = - kX(l)T(t) (6.49) Upon further substitution of X = a1 sin σx this becomes τσ cos σl = - k sin σl (6.50) so that tan σl = - τσ k (6.51) As in the case of the string fixed at both ends, satisfying the second boundary condition yields an infinite number of values of σ. Hence σ becomes σn, where σn are the solutions of equation (6.51). These values of σn can be visualized (and computed) by plotting tan σl and –τσ >k versus σ on the same plot, as illustrated in Figure 6.4. Thus the eigenfunctions or mode shapes of a cable connected to a spring are an sin σnx, n = 1, 2, 3, c∞ (6.52) where the σn must be calculated numerically for given values of k, τ, and l from equation (6.51). Note that by examining the curves of Figure 6.4, the values of σn (the point where the two curves cross) for large values of n become very close to (2n – 1)π>(2l). n tan l k  800 N/m l n1 n2 f(l)   /k Figure 6.4 A graphical solution of the transcendental equation tan σl = –τσ >k. 518 Distributed-Parameter Systems Chap. 6 The separation-of-variables modal analysis method illustrated in this section is summarized in Window 6.4. Note that the procedure is analogous to that summarized in Window 6.3 for lumped-parameter systems. Window 6.4 Summary of the Separation-of-Variables Solutions Method 1. A solution to a partial differential equation is assumed to be of the form w(x, t) = X(x)T(t) (i.e., separated). 2. The separated form is next substituted into the partial differential equation of motion and separated by trying to position all of the terms containing X(x) on one side of the equality, and all of the terms containing T(t) on the other side. (Note that this cannot always be done.) 3. Once written in separated form, equation (6.12), for example, each side of the equality is set equal to the same constant to obtain two new equations: a spatial equation in the single-valued variable X(x) and a temporal equation in the single-valued variable T(t). This constant is called the separation constant and is denoted by σ2. 4. The boundary conditions are applied to the spatial equation which results in an algebraic eigenvector problem that yields an infinite number of values of the separation constant, σn, and the mode shapes Xn(x). 5. The separation constants obtained in step 4 are substituted into the temporal equation, and the solutions Tn(t) are obtained in terms of two arbitrary constants. 6. The infinite number of solutions wn(x, t) = Xn(x)Tn(t) are then formed and summed to produce the series solution w(x, t) = a Xn(x)Tn(t) ∞ n=1 which contains two sets of undetermined constants. 7. The unknown constants are determined by applying the initial conditions w(x, 0) and wt(x, 0) to the series solution, multiplying by a mode Xm(x), and integrating over x. The orthogonality condition then yields simple equations for the unknown constants in terms of the initial conditions. See equations (6.31) and (6.33), for example. 8. The series solution is written with the constants evaluated from step 7, and the solution is complete. Sec. 6.3 519 Vibration of Rods and bars 6.3 VIBRATION OF RODS AND BARS Next consider the vibration of an elastic bar (or rod) of length l and of varying cross-sectional area in the direction indicated in Figure 6.5. The density of the bar is denoted by ρ (not to be confused with mass per unit length used for the string) and the cross-sectional area by A(x). Using the coordinate system indicated in the figure, the forces on the infinitesimal element summed in the x direction are F + dF - F = ρA(x)dx 02w(x, t) (6.53) 0t 2 where w(x, t) is the deflection of the rod in the x direction, F denotes the force acting on the infinitesimal element to the left, and F + dF denotes the force to the right, as the element is displaced. From introductory strength of materials, the force F is given by F = σsA, where σs is the unit stress in the x direction and has the value Ewx(x, t), where E is the Young’s modulus (or modulus of elasticity), and wx is the unit strain. This yields F = EA(x) 0w(x, t) 0x (6.54) The differential force becomes dF = (0F>0x) dx, from the chain rule for partial derivatives. Substitution of these quantities into equation (6.53) and dividing by dx yields 0w(x, t) 02w(x, t) 0 aEA(x) b = ρA(x) 0x 0x 0t 2 (6.55) y Displaced position x F F  dF l 0 w(x, t) dx Equilibrium x x  dx position Infinitesimal element of bar Figure 6.5 A cantilevered bar in longitudinal vibration along x. 520 Distributed-Parameter Systems Chap. 6 In those cases where A(x) is a constant this becomes 02w(x, t) E 02w(x, t) = a b ρ 0x2 0t 2 (6.56) which has exactly the same form as the string equation of Section 6.2. The quantity c = 1E>ρ defines the velocity of propagation of the displacement (or stress wave) in the bar. Hence the solution technique used in Section 6.2 is exactly applicable here. because the bar can support its own weight, a variety of boundary conditions are possible. Various springs, masses, or dashpots can be attached to one end of the bar or the other, to model a variety of situations. Consider the boundary conditions for the clamped–free configuration (called cantilevered) of Figure 6.5. At the clamped end, x = 0, the displacement must be zero, so that w(0, t) = 0 t 7 0 (6.57) At the free end, x = l, the force in the bar must be zero or EA 0w(x, t) ` = 0 0x x=l (6.58) These are the simplest boundary conditions. Example 6.3.1 Calculate the mode shapes and natural frequencies of the cantilevered (clamped–free) bar of Figure 6.5 for the case of constant cross section. Solution Following the method of separation of variables outlined in Window 6.4, let w(x, t) = X(x)T(t) in equation (6.56) to get X″(x) X(x) = $ T(t) c 2T(t) = - σ2 This last expression yields the following spatial equation: X″(x) + σ2X(x) = 0, and X(0) = 0, AEX =(l) = 0 The spatial equation has a solution of the form X(x) = a sin σx + b cos σx (6.59) Sec. 6.3 521 Vibration of Rods and bars Applying the condition at the fixed end, x = 0, yields the result that b = 0, so that X(x) has the form X(x) = a sin σx At x = l, the boundary condition yields AEX =(l) = 0 = AEσa cos σl (6.60) Since A, E, a, and σ are nonzero, this requires that cos σl = 0, or that σn = 2n - 1 π n = 1, 2, 3, c 2l (6.61) The mode shapes are thus of the form an sin (2n - 1)πx (6.62) 2l Now consider the temporal equation resulting from equation (6.59): $ T(t) + c 2σ2T(t) = 0 1 T(t) = A sin σct + B cos σct This suggests an oscillation with frequency σc, or natural frequencies of the form ωn = (2n - 1)πc 2l = (2n - 1)π 2l E rad>s Aρ n = 1, 2, 3, c (6.63) The solution is given by equation (6.27) with c = 2E>ρ and σ n as given by expression (6.61). n In distributed-parameter systems, the form of the spatial differential equation is often determined by the stiffness term. The differential form of the stiffness term along with the boundary conditions determine whether or not the eigenfunctions, and hence the mode shapes, are orthogonal. The eigenfunction orthogonality condition for distributed-parameter systems is an exact replica of the eigenvector and mode shape orthogonality condition for lumped-parameter systems. Precise orthogonality conditions are beyond the scope of this chapter and can be found in Inman (2006), for instance. Note in Example 6.3.1 that the mode shapes given by expression (6.62) are orthogonal [recall equation (6.28)]. Just as in the lumpedparameter case, the orthogonality of mode shapes for a distributed-parameter system is extremely useful in calculating the vibration response of the system. The following example illustrates the use of modes to calculate the response to a given set of initial conditions. 522 Distributed-Parameter Systems Chap. 6 Example 6.3.2 Calculate the response of the bar of Example 6.3.1 to an initial velocity of 3 cm>s at the free end and a zero initial displacement. Assume that the bar is 5 m long, has a density of ρ = 8 * 103 kg>m3, and has a modulus of E = 20 * 1010 N>m2. Plot the response at x = l and x = l>2. Solution The solution w(x, t) for the longitudinal vibration of a clamped–free bar is given by equation (6.27) as w(x, t) = a (cn sin σnct + d n cos σnct ) sin ∞ (2n - 1)π x n=1 2l where σn is calculated in Example 6.3.1 and given in equation (6.61). To calculate the coefficients cn and dn, note that w(x, 0) = 0 and wt(l, 0) = 0.03 m>s. Thus from equation (6.31) l dn = l 2 nπx 2 nπx w (x) sin dx = (0) sin dx = 0 l L0 0 l l L0 l The coefficients cn are computed from differentiating the series w(x, t) and setting t = 0: wt(x, 0) = 0.03δ(x - l) = a cnσnc sin σnx cos (0) ∞ n=1 where δ is the Dirac delta function. Multiplying by sin σmx and integrating yields 0.03 L0 δ(x - l) sin(σmx)dx = a cnσnc l ∞ n=1 L0 l sin (σnx) sin (σmx)dx because {sin σmx} is an orthogonal set of functions, each term on the right side in the summation is zero except for the one term when n = m: L0 sin a l 2n - 1 2m - 1 l>2, πxb sin a πxbdx = e 2l 2l 0, n = m n ≠ m and the integral on the left side can be evaluated using the filtering property of the Dirac delta function from Section 3.1, 1 f(x)δ(x - a)dx = f(a). Thus the preceding equation becomes 0.03 sin a (2m - 1)π 2 b = cσml c 2 m 523 Vibration of Rods and bars w(l/2,t) 1 3.82  105 m 3.82  105 m w(l/2,t) 0 1 0.00 0.02 0.04 0.06 0.08 1 0 1 t 0.10 0.00 0.02 0.04 (a) 0.08 t 0.10 0.06 0.08 t 0.10 w(l/2,t) 3.82  105 m 1 0 1 0.00 0.06 (b) w(l/2,t) 3.82  105 m Sec. 6.3 0.02 0.04 0.06 0.08 1 0 1 t 0.10 0.00 0.02 0.04 (c) (d) Figure 6.6 Response of the bar of Example 6.3.2 plotted for x = l >2 for (a) one, (b) two, (c) five, and (d) ten terms of the series solution. Note the slight amplitude changes. The numerator on the left-hand side can be written as {1 depending on the value of the sine function. Solving for cm, evaluating the physical constants, and renaming m by n yields the equation cn = 0.06( -1)n + 1 cσnl = 0.6l( -1)n + 1 cπ(2n - 1) = 3.82 * 10-6 ( - 1)n + 1 (2n - 1) m, n = 1, 2, 3, c Thus the total solution becomes ∞ ( -1)n + 1 (2n - 1)πx w(x, t) = 3.82 * 10-6 a sin sin [512.35(2n - 1)πt]m 2l n = 1 (2n - 1) At x = l this becomes ∞ ( -1)n + 1 w(l, t) = 3.82 * 10-6 a sin [512.35(2n - 1)πt]m n = 1 (2n - 1) and at x = l>2 the response is given by w(l>2, t) = 3.82 * 10-5 a N ( -1)n + 1 n=1 (2n - 1)2 sin (512.35)(2n - 1)πt m where N = ∞ . To plot these, the infinite series must be approximated by truncating the series taking finite values of N. In Figure 6.6, the response of w(l>2, t) is plotted for 524 Distributed-Parameter Systems Chap. 6 one, two, five, and ten terms. Note that the only noticeable difference in the plot of the first term in the sum, and that of the first two terms, is the increase in amplitude. After five terms, not much change in the plots occurs. The Toolbox problems at the end of this chapter investigate this further, as can any math software. n The previous example is solved by the method of mode summation introduced in Chapter 4 for lumped-parameter systems. The procedure is the same both here and in Section 4.4. basically, the solution is written as a sum of modes with constants to be evaluated by the initial conditions. Then the orthogonality of the mode shapes is used along with the initial conditions to evaluate these constants. The analysis of a lumped-parameter system and that of a distributed-parameter system are similar in as much as they both require the computation of natural frequencies and mode shapes (eigenvectors in the lumped case, eigenfunctions in the distributed case). However, the time response of the distributed-parameter system is distributed spatially in a continuous manner, whereas that of a lumped-parameter system is spatially discrete. Thus for a lumped system it is useful to plot the time response of a single coordinate xi(t), and for a distributed-parameter system the response plot is at a single point [i.e., w(l>2, t)] as in Figure 6.6. For the lumped case, there is a finite number of degrees of freedom and hence a finite number of responses xi(t)(i = 1, 2, cn) to consider. In the distributed-parameter case, there is an infinite number of responses w(x, t) to consider, as the spatial variable x can take on any value between 0 and l. boundary conditions for various configurations of the bar are listed in Table 6.1. Table 6.2 lists the various frequencies and mode shapes using these boundary conditions. TABLE 6.1 A SuMMARy oF VARIouS bounDARy ConDItIonS FoR tHE longItuDInAl VIbRAtIon oF tHE bAR oF FIguRE 6.5 Fixed at left end: w(x, t) 0 x = 0 = 0 Fixed at right end: w(x, t) 0 x = l = 0 0w(x, t) Free at left end: EA ` = 0 0x x=0 Free at right end: EA 0w(x, t) 0x ` = 0 x=l Attached to a mass of mass m at left end: AE 0w(x, t) Attached to a mass of mass m at right end: AE 0x ` Attached to a spring of stiffness k at left end: AE ` 0t 2 0x ` = -m x=l 0w(x, t) Attached to a spring of stiffness k at right end: AE 02w(x, t) x=0 0w(x, t) 0x = m 0w(x, t) 0x ` x=0 ` 0t 2 = kw(x, t) ` x=l ` x=0 02w(x, t) x=l = - kw(x, t) ` x=0 x=l Sec. 6.4 525 torsional Vibration TABLE 6.2 VARIouS ConFIguRAtIonS oF A unIFoRM bAR oF lEngtH l In longItuDInAl VIbRAtIon IlluStRAtIng tHE nAtuRAl FREquEnCIES AnD MoDE SHAPESa Frequency (rad/s) or characteristic equation Configuration Mode shape x n  nc , n  0, 1, 2, . . . l cos nxb l Free–free x n  (2n  1)c 2l Fixed–free x n  , n  1, 2, . . . sin (2n  1)x 2l nc , n  1, 2, . . . l sin nx l sin nx l sin nx l Fixed–fixed kl x k n cot n   n  Fixed–spring x cot n  EA nc l m n Al m Fixed–mass n  nc l a Note that the last two examples require a numerical procedure (suggested in the Toolbox problems at the end of the chapter) in order to calculate the values of the natural frequencies. Here c = 1E>ρ. b The free mode shape is a constant. 6.4 TORSIONAL VIBRATION The bar of Section 6.3 may also vibrate in the torsional direction as indicated by the circular shaft of Figure 6.7. In this case, the vibration occurs in an angular direction around the center axis of the shaft in a plane perpendicular to the cross section of the shaft or rod. The rotation of the shaft, θ, about the center axis is a function of both the position along the length of the rod, x, and the time, t. Thus θ is a function of two variables denoted θ(x, t). The equation of motion can be determined by considering a moment balance of an infinitesimal element of the rod of length dx illustrated in 526 Distributed-Parameter Systems (x,t) Chap. 6 dx G  x l x  x  dx    dx x (x,t)  d Figure 6.7 A circular shaft illustrating an angular motion, θ(x, t), as the result of a moment acting on a differential element dx of the shaft of density ρ, length l, and given modulus G. The function θ(x, t) denotes the angle of twist. (x,t) dx the figure. Referring to Figure 6.7, the torque at the right face of the element dx (i.e., at position x) is τ (here τ is used as a torque, not as tension as in Section 6.1), while 0τ that at the left end, at position x + dx, is τ + dx. From solid mechanics, the applied 0x torque is related to the torsional deflection by (Shames, 1989) τ = GJ 0θ(x, t) 0x (6.64) where GJ is the torsional stiffness composed of the shear modulus G and the polar moment of area J of the cross section. Note that J could be a function of x as well, but is considered constant here. The total torque acting on dx becomes τ + 0τ 02θ dx - τ = J0 2 dx 0x 0t (6.65) where J0 is the polar moment of inertia of the shaft per unit length and 02θ/0t 2 is the angular acceleration. If the shaft is of uniform circular cross section, J0 becomes simply J0 = ρJ, where ρ is the shaft’s material mass density. Substitution of the expression for the torque given by equation (6.64) into equation (6.65) yields 0 0θ 02θ aGJ b = ρJ 2 0x 0x 0t Simplifying for the case of constant stiffness GJ yields 02θ(x, t) 0t 2 = a G 02θ(x, t) b ρ 0x2 (6.66) for the equation of twisting vibration of a rod. This is, again, identical in mathematical form to the formula of equation (6.8) for the string or cable and the formula of equation (6.56) for the longitudinal vibrations of a rod or bar. Sec. 6.4 527 torsional Vibration For other types of cross sections, the torsional equation of a shaft can still be used to approximate the torsional motion by replacing J in equation (6.64) with a torsional constant γ defined to be the moment required to produce a torsional rotation of 1 rad on a unit length of shaft divided by the shear modulus. Thus a shaft with noncircular cross section can be approximated by the equation 02θ(x, t) 0t 2 = a Gγ 02θ(x, t) b ρJ 0x2 (6.67) Some values of γ are presented in Table 6.3 for several common cross sections. Keep in mind that equation (6.67) is approximate and assumes in particular that the TABLE 6.3 SoME VAluES oF tHE toRSIonAl ConStAnt FoR VARIouS CRoSS-SECtIonAl SHAPES uSED In APPRoxIMAtIng toRSIonAl VIbRAtIon FoR nonCIRCulAR CRoSS SECtIonS Cross section Torsional constant,  R Circular shaft R 4 2 R2 Hollow circular shaft R1  (R 4  R 4 ) 1 2 2 a Square shaft 0.1406a4 a Hollow rectangular B A a b shaft 2AB(a  A)2 (b  B)2 aA  bB  A2  B2 528 Distributed-Parameter Systems Chap. 6 center of mass and center of rotation coincide so that torsional and flexural vibrations do not couple. The solution of equation (6.66) depends on two initial conditions in time [i.e., θ(x, 0) and θt(x, 0)] and two boundary conditions, one at each end of the rod. The possible choices of boundary conditions are similar to those for the string and the bar (i.e., either the deflection is zero if the rod is fixed at a boundary or the torque is zero if the rod is free at a boundary). For example, the clamped–free rod of Figure 6.7 has boundary conditions (deflection at 0) (torque at l) θ(0, t) = 0 (6.68) γGθx(l, t) = 0 (6.69) Table 6.4 lists some common boundary conditions for the torsional vibration of shafts. TABLE 6.4 A SuMMARy oF VARIouS bounDARy ConDItIonS FoR tHE toRSIonAl VIbRAtIon oF tHE SHAFt oF FIguRE 6.7 Fixed at left end: θ(x, t) 0 x = 0 = 0 Fixed at right end: θ(x, t) 0 x = l = 0 Free at left end: γG 0θ(x, t) Free at right end: γG 0x ` 0θ(x, t) 0x ` = 0 x=0 = 0 ` 0t 2 x=0 0θ(x, t) 02θ(x, t) Attached to an inertial mass J1 at right end: γG ` = -J1 ` 0x 0t 2 x=l x=l 0θ(x, t) Attached to a rotational spring of stiffness k at left end: γG ` = kθ(x, t) ` 0x x=0 x=0 x=l Attached to an inertial mass J1 at left end: γG 0θ(x, t) 0x ` x=0 = J1 Attached to a rotational spring of stiffness k at right end: γG 02θ(x, t) 0θ(x, t) 0x ` x=l = -kθ(x, t) ` x=l Example 6.4.1 The vibration of a large grinding machine can be modeled as a shaft or rod with a disk at one end as illustrated in Figure 6.8. The top end of the shaft, at x = 0, is connected to a pulley. The effects of the drive belt and motor are accounted for by including their collective inertia in with the mass moment of inertia of the pulley, denoted by J1. Determine the natural frequencies of the system. Sec. 6.4 529 torsional Vibration Pulley of inertia J1 Motor Drive belt Rod of length l and torsional rigidity GJ x Figure 6.8 A simple model of a large grinding machine. Grinding head, a disk of inertia J2 Solution The frequencies of vibration are determined by equation (6.66) subject to the appropriate boundary conditions. The boundary conditions are determined by examining either the deflections or torques at the boundaries. In this example, the deflection is unspecified at x = 0, but the torque must match that supplied by the pulley or GJ 0θ(0, t) 0x = J1 02θ(0, t) 0t 2 (6.70) Similarly, at x = l a balance of torques yields GJ 0θ(l, t) 0x = -J2 02θ(l, t) 0t 2 (6.71) where the minus sign arises from the right-hand rule. Following the method of Section 6.2, the solution for θ(x, t) is assumed to separate and be of the form θ(x, t) = Θ(x)T(t). Substitution into equation (6.66) and rearranging yields $ Θ″(x) ρ T(t) = a b = - σ2 (6.72) Θ(x) G T(t) Defining (ρ>G) = 1>c 2 and realizing that equation (6.72) splits into a temporal equation and a spatial equation, the spatial equation becomes Θ″(x) + σ2Θ(x) = 0 (6.73) where σ is the separation constant and is related to the natural frequencies of the system by ω = σc = σ G Aρ From equation (6.70) the boundary condition at x = 0 becomes $ GJΘ=(0)T(t) = J1Θ(0)T(t) (6.74) (6.75) 530 Distributed-Parameter Systems Chap. 6 or GJΘ=(0) J1Θ(0) = $ T(t) T(t) = -c 2σ2 (6.76) where the last equality follows from the right-hand side of equation (6.72). Upon further manipulation, equation (6.76) yields Θ=(0) = - σ2J1 Θ(0) ρJ (6.77) Similarly, the boundary conditions at x = l given by equation (6.71) yield Θ=(l) = σ2J2 Θ(l) ρJ (6.78) The general solution of equation (6.73) is Θ(x) = a1 sin σx + a2 cos σx (6.79) Θ=(x) = a1 σ cos σx - a2 σ sin σx (6.80) so that Substitution of these expressions into equation (6.77) for the boundary condition at x = 0 yields a1 = - σJ1 a ρJ 2 (6.81) Application of the boundary conditions at x = l by substitution of equations (6.79), (6.80), and (6.81) into equation (6.78) yields the characteristic equation given by tan σI = ρJl(J1 + J2)(σl) J1J2(σl)2 - ρ2J2l 2 (6.82) This expression is a transcendental equation in the quantity σl, which must be solved numerically, similar to Example 6.2.3. because of the tangent term, equation (6.82) has an infinite number of solutions, which can be denoted σnl, n = 1, 2, 3, c ∞ . The numerical solutions σn determine the natural frequencies of vibrations according to equation (6.74): ω n = σn G Aρ (6.83) Note that the first solution of equation (6.82) corresponding to the first eigenvalue yields ω1 = 0 Sec. 6.4 3 2 2 1.0 tan x and x bx2 a 531 torsional Vibration 5 2 7 2 9 2 x bx2 a 0.5 0.0 0.5 1.0 tan x 0 2 4 6 8 10 12 14 x Figure 6.9 The plot of tan x versus x and x>(bx2 – a) versus x. These points of intersection are determined using the fzero command in MATlAb, as discussed at the end of the problem section. This yields the roots of equation (6.84). $ Since ω1 = σ1 = 0, the time equation (6.72) becomes T1(t) = 0, or T1(t) = a + bt, where a and b are constants determined by the initial conditions. This solution corresponds to the rigid-body mode of the system, which in this case is a constant shaft rotation. The rigid-body mode shape is calculated from equation (6.73) with σ = 0. This yields Θ1″(x) = 0, so that Θ1(x) = a1 + b1x. Applying the boundary condition at x = 0 yields GJ (b1) = 0, so that b1 = 0. The boundary condition at x = l also results in b1 = 0, so that the first mode shape, or first eigenfunction, is simply Θ1(x) = a1, a nonzero constant. This defines the rigid-body mode shape. For very large values of σl, equation (6.82) reduces to tan σl = 0, so that the high frequencies of vibration approach ωn = nπc. Examining the form of the characteristic equation indicates that it can be written as (bx2 - a)tan x = x (6.84) where x = σl, a = ρ Jl>(J1 + J2), and b = J1J2 >[(J1 + J2) ρ Jl]. The solution of this expression for the case J1 = 10 kg # m2 >rad, J2 = 10 kg # m2 >rad, ρ = 7870 kg>m3, J = 5 m4, l = 0.425 m is illustrated by the points of intersection of the two curves in Figure 6.9. These points of intersection are determined by using MATlAb to solve for the roots of equation (6.84). For G = 80 * 109 Pa, this results in f1 = 0 f2 = 3,980 Hz f3 = 7,961 Hz f4 = 11,940 Hz f5 = 15,920 Hz f6 = 19,900 Hz with the higher frequencies approaching nπc in rad>s. n The natural frequencies for various configurations of torsional systems are listed in Table 6.5. Note the similarity to Table 6.2 for longitudinal vibration. The only difference is the physical interpretation of the motion. Table 6.5 can be combined with Table 6.3 to approximate the natural frequencies of torsional vibration for noncircular cross sections as well. A more complete tabulation can be found in blevins (1987). 532 Distributed-Parameter Systems Chap. 6 TABLE 6.5 SAMPlE oF VARIouS ConFIguRAtIonS oF A unIFoRM SHAFt In toRSIonAl VIbRAtIon, oF lEngtH l, IlluStRAtIng tHE nAtuRAl FREquEnCIES AnD MoDE SHAPESa Frequency (rad/s) or characteristic equation Configuration x  n  nc , l Free–free x n = 0, 1, 2, . . . b cos nx l  n  (2n  1)c , 2l Fixed–free x Mode shape n = 1, 2, . . . sin (2n  1)x 2l  n  nc , l n = 1, 2, . . . n cot n   kl G sin nx l Fixed–fixed x  Fixed–spring (rotational) x  n  cot n  n  nx l sin nx l nc l J0 Fixed–inertia sin nc l J0 n l a Here c = 2Gγ>ρJ. The values of γ can be found in Table 6.3. Note that a numerical procedure is required for the last two cases, as illustrated in Example 6.4.1. b The free mode shape is a constant. 6.5 BENDING VIBRATION OF A BEAM This section again considers the vibration of the bar or beam of Figure 6.5. In this case, however, vibration of the beam in the direction perpendicular to its length is considered. Such vibrations are often called transverse vibrations, or flexural vibrations, because they move across the length of the beam. Transverse vibration is easily felt by humans when walking over bridges, for example. Sec. 6.5 533 bending Vibration of a beam Euler–Bournoulli Beam Theory Figure 6.10 illustrates a cantilevered beam with the transverse direction of vibration indicated [i.e., the deflection, w(x, t), is in the y direction]. The beam is of rectangular cross section A(x) with width hy, thickness hz, and length l. Also associated with the beam is a flexural (bending) stiffness EI(x), where E is Young’s elastic modulus for the beam and I(x) is the cross-sectional area moment of inertia about the “z axis.” From mechanics of materials, the beam sustains a bending moment M(x, t), which is related to the beam deflection, or bending deformation, w(x, t), by M(x, t) = EI(x) 02w(x, t) (6.85) 0x2 A model of bending vibration may be derived from examining the force diagram of an infinitesimal element of the beam as indicated in Figure 6.10. Assuming the deformation to be small enough such that the shear deformation is much smaller than w(x,t) y dx f(x,t) hy x hz z f(x, t) M(x, t)  M(x, t) dx x M(x, t) V(x, t)  Q w(x, t) V(x, t) A dx 2 w t2 V(x, t) dx x Undeformed x axis dx x x  dx Figure 6.10 A simple Euler–bernoulli beam of length (l) in transverse vibration and a free-body diagram of a small element of the beam as it is deformed by a distributed force-per-unit length, denoted by f(x, t). A(x) 534 Distributed-Parameter Systems Chap. 6 w(x, t) (i.e., so that the sides of the element dx do not bend), a summation of forces in the y direction yields aV(x, t) + 0V(x, t) 02w(x, t) dxb - V(x, t) + f(x, t)dx = ρA(x) dx 0x 0t 2 (6.86) Here V(x, t) is the shear force at the left end of the element dx, V(x, t) + Vx(x, t) dx is the shear force at the right end of the element dx, f(x, t) is the total external force applied to the element per unit length, and the term on the right side of the equality is the inertial force of the element. The assumption of small shear deformation used in the force balance of equation (6.86) is true if l>hz Ú 10 and l>hy Ú 10 (i.e., for long slender beams or Euler–bernoulli beams). Next, the moments acting on the element dx about the z axis through point Q are summed. This yields c M(x, t) + 0M(x, t) 0V(x, t) dx dx d - M(x, t) + c V(x, t) + dx d dx + [ f(x, t)dx] = 0 0x 0x 2 (6.87) Here the right-hand side of the equation is zero since it is also assumed that the rotary inertia of the element dx is negligible. Simplifying this expression yields c 0M(x, t) 0V(x, t) f(x, t) + V(x, t) d dx + c + d (dx)2 = 0 0x 0x 2 (6.88) Since dx is assumed to be very small, (dx)2 is assumed to be almost zero, so that this moment expression yields (dx is small, but not zero) V(x, t) = - 0M(x, t) 0x (6.89) This states that the shear force is proportional to the spatial change in the bending moment. Substitution of this expression for the shear force into equation (6.86) yields - 02w(x, t) 02 [M(x, t)]dx + f(x, t)dx = ρA(x)dx 0x2 0t 2 (6.90) Further substitution of equation (6.85) into (6.90) and dividing by dx yields ρA(x) 02w(x, t) 0t 2 + 02w(x, t) 02 c EI(x) d = f(x, t) 0x2 0x2 (6.91) If no external force is applied so that f(x, t) = 0 and if EI(x) and A(x) are assumed to be constant, equation (6.91) simplifies so that free vibration is governed by 02w(x, t) 0t 2 + c2 04w(x, t) 0x 4 = 0, c = EI A ρA (6.92) Note that unlike the previous equations, the free vibration equation (6.92) contains four spatial derivatives and hence requires four (instead of two) boundary conditions Sec. 6.5 535 bending Vibration of a beam in calculating a solution. The presence of the two time derivatives again requires that two initial conditions, one for the displacement and one for the velocity, be specified. The boundary conditions required to solve the spatial equation in a separationof-variables solution of equation (6.92) are obtained by examining the deflection w(x, t), the slope of the deflection 0w(x, t)>0x, the bending moment EI02w(x, t)>0x2, and the shear force 0[EI02w(x, t)>0x2]>0x at each end of the beam. A common configuration is clamped–free or cantilevered as illustrated in Figure 6.10. In addition to a boundary being clamped or free, the end of a beam could be resting on a support restrained from bending or deflecting. The situation is called simply supported or pinned. A sliding boundary is one in which displacement is allowed but rotation is not. The shear load at a sliding boundary is zero. If a beam in transverse vibration is free at one end, the deflection and slope at that end are unrestricted, but the bending moment and shear force must vanish: bending moment = EI 02w = 0 0x2 0 02w shear force = c EI 2 d = 0 0x 0x (6.93) If, on the other hand, the end of a beam is clamped (or fixed), the bending moment and shear force are unrestricted, but the deflection and slope must vanish at that end: deflection = w = 0 0w slope = = 0 0x (6.94) At a simply supported or pinned end, the slope and shear force are unrestricted and the deflection and bending moment must vanish: deflection = w = 0 02w bending moment = EI 2 = 0 0x (6.95) At a sliding end, the slope or rotation is zero and no shear force is allowed. On the other hand, the deflection and bending moment are unrestricted. Hence, at a sliding boundary, 0w = 0 0x 0 02w shear force = aEI 2 b = 0 0x 0x slope = (6.96) Other boundary conditions are possible by connecting the ends of a beam to a variety of devices such as lumped masses, springs, and so on. These boundary conditions can be determined by force and moment balances. In addition to satisfying four boundary conditions, the solution of equation (6.92) for free vibration can be calculated only if two initial conditions (in time) are 536 Distributed-Parameter Systems Chap. 6 specified. As in the case of the rod, string, and bar, these initial conditions are specified initial deflection and velocity profiles: # w(x, 0) = w0(x) and wt(x, 0) = w0(x) # assuming that t = 0 is the initial time. Note that if w0 and w0 were both zero, no motion would result. The solution of equation (6.92) subject to four boundary conditions and two initial conditions proceeds following exactly the same steps used in previous sections. A separation-of-variables solution of the form w(x, t) = X(x)T(t) is assumed. This is substituted into the equation of motion, equation (6.92), to yield (after rearrangement) $ T(t) X″″(x) 2 c = = ω2 (6.97) X(x) T(t) where the partial derivatives have been replaced with total derivatives as before $ (note: X″″ = d 4X>dx4, T = d 2T>dt 2). Here the choice of separation constant, ω2, is made, based on experience with the systems of Section 6.4 that the natural frequency comes from the temporal equation $ T(t) + ω2T(t) = 0 (6.98) which is the right side of equation (6.97). This temporal equation has a solution of the form T(t) = A sin ωt + B cos ωt (6.99) where the constants A and B will eventually be determined by the specified initial conditions after being combined with the spatial solution. The spatial equation comes from rearranging equation (6.97), which yields ω 2 X″″(x) - a b X(x) = 0 c (6.100) by defining [recall equation (6.92)] β4 = ρAω2 ω2 EI = aso that ω = β2 rad>sb 2 EI A ρA c (6.101) and assuming a solution to equation (6.100) of the form Aeσx, the general solution of equation (6.100) can be calculated to be of the form (see Problem 6.44) X(x) = a1 sin βx + a2 cos βx + a3 sinh βx + a4 cosh βx (102) Here the value for β and three of the four constants of integration a1, a2, a3, and a4 will be determined from the four boundary conditions. The fourth constant becomes combined with the constants A and B from the temporal equation, which are then determined from the initial conditions. The following example illustrates the solution procedure for a beam fixed at one end and simply supported at the other end. Sec. 6.5 537 bending Vibration of a beam Example 6.5.1 Calculate the natural frequencies and mode shapes for the transverse vibration of a beam of length l that is fixed at one end and pinned at the other end. Solution The boundary conditions in this case are given by equation (6.94) at the fixed end and equation (6.95) at the pinned end. Substitution of the general solution given by equation (6.102) into equation (6.94) at x = 0 yields X(0) = 0 1 a2 + a4 = 0 (a) X =(0) = 0 1 β(a1 + a3) = 0 (b) Similarly, at x = l the boundary conditions result in X(l) = 0 1 a1 sin βl + a2 cos βl + a3 sinh βl + a4 cosh βl = 0 EIX″(l) = 0 1 β2( - a1 sin βl - a2 cos βl + a3 sinh βl + a4 cosh βl) = 0 (c) (d) These four boundary conditions thus yield four equations [(a) through (d)] in the four unknown coefficients a1, a2, a3, and a4. These can be written as the single vector equation 1 β D sin βl - β2 sin βl 1 0 cos βl - β2 cos βl 0 β sinh βl β2 sin βl 1 a1 0 0 a2 0 TD T = D T cosh βl a3 0 β2 cosh βl a4 0 Recall from Chapter 4 that this vector equation can have a nonzero solution for the vector a = [a1 a2 a3 a4]T only if the determinant of the coefficient matrix vanishes (i.e., if the coefficient matrix is singular). Furthermore, recall that since the coefficient matrix is singular, not all of the elements of the vector a can be calculated. Setting the determinant above equal to zero yields the characteristic equation tan βl = tanh βl This equality is satisfied for an infinite number of choices for β, denoted βn. The solution can be visualized by plotting both tan βl and tanh βl versus βl on the same plot. This is similar to the solution technique used in Example 6.4.1 and illustrated in Figure 6.9. The first five solutions are β1l = 3.926602 β2l = 7.068583 β3l = 10.210176 β4l = 13.351768 β5l = 16.49336143 For the rest of the modes (i.e., for values of the index n 7 5), the solutions to the characteristic equation are well approximated by βn l = (4n + 1)π 4 The weighted frequencies determine the system natural frequencies by ωn = β2n EI rad>s, A ρA and fn = β2n EI Hz 2π A ρA 538 Distributed-Parameter Systems Chap. 6 With these values of the weighted frequencies βnl, the individual modes of vibration can be calculated. Solving the preceding matrix equation for the individual coefficients ai yields a1 = –a3, a2 = –a4, and (sinh βnl - sin βnl)a3 + (cosh βnl - cos βnl)a4 = 0 Thus a3 = - cosh βnl - cos βnl a sinh βnl - sin βnl 4 for each n. The fourth coefficient a4 cannot be determined by this set of equations, because the coefficient matrix is singular (otherwise, each ai would be zero). This remaining coefficient becomes the arbitrary magnitude of the eigenfunction. As this constant depends on n, denote it by (a4)n. Substitution of these values of ai in the expression X(x) for the spatial solution yields the result that the eigenfunctions or mode shapes have the form Xn(x) = (a4)n c cosh βnl - cos βnl (sinh βnx - sin βnx) - cosh βnx + cos βnx d , sinh βnl - sin βnl n = 1, 2, 3, c The first three mode shapes are plotted in Figure 6.11 for (a2)n = 1 and n = 1, 2, 3. Mode 3 Mode shape Xn(x) 1 Mode 2 Mode 1 0 1 0 0.2 0.4 0.6 0.8 x/l 1.0 Figure 6.11 A plot of the first three mode shapes of the clamped–pinned beam of Example 6.5.1, arbitrarily normalized to unity. These mode shapes can be shown to be orthogonal, so that L0 l Xn(x)Xm(x)dx = 0 for n ≠ m (see Problem 6.47). As in Example 6.3.2, this orthogonality, along with the initial conditions, can be used to calculate the constants An and Bn in the series solution for the displacement w(x, t) = a (An sin ωnt + Bn cos ωnt)Xn(x) ∞ n=1 n Table 6.6 summarizes a number of different boundary configurations for the slender beam. The slender-beam model given in equation (6.91) is often referred to TABLE 6.6 SAMPlE oF VARIouS bounDARy ConFIguRAtIonS oF A SlEnDER bEAM In tRAnSVERSE VIbRAtIon oF lEngtH l IlluStRAtIng wEIgHtED nAtuRAl FREquEnCIES AnD MoDE SHAPESa Configuration x Free–free x Clamped–free x Clamped–pinned x Clamped–sliding x Clamped–clamped x Weighted frequencies nl and characteristic equation Mode shape 0 (rigid-body mode) 4.73004074 7.85320462 10.9956078 14.1371655 17.2787597 (2n  1) for n  5 2 cos l cosh l  1 cosh nx  cos nx 1.87510407 4.69409113 7.85475744 10.99554073 14.13716839 (2n  1) for n  5 2 cosh nx  cos nx cos l cosh l  1 3.92660231 7.06858275 10.21017612 13.35176878 16.49336143 (4n  1) for n  5 4 tan l  tanh l 2.36502037 5.49780392 8.63937983 11.78097245 14.92256510 (4n  1) for n  5 4 tan l  tanh l  0 4.73004074 7.85320462 10.9956079 14.1371655 17.2787597 (2n  1) for n  5 2 cos l cosh l  1 n sin n (sinh nx  sin nx)b n 0.9825 1.0008 0.9999 1.0000 0.9999 1 for n  5 n (sinh nx  sin nx) 0.7341 1.0185 0.9992 1.0000 1.0000 1 for n  5 cosh nx  cos nx 1.0008 1 for n  1 n (sinh nx  sin nx) cosh nx  cos nx 0.9825 1 for n  1 n (sinh nx  sin nx) cosh nx  cos nx n (sinh nx  sin nx) nx l 0.982502 1.00078 0.999966 1.0000 1.0000 1 for n  5 none sin l  0 Pinned–pinned a Here the weighted natural frequencies βnl are related to the natural frequencies by equation (6.101) or ωn = β2n 1EI>ρA, as used in Example 6.5.1. The values of σi for the mode shapes are computed from the formulas given in Table 6.5. b There are two free–free mode shapes: X0 = constant and X0 = A(x - l>2); the first is translational, the second rotational. 540 Distributed-Parameter Systems Chap. 6 as the Euler–Bernoulli or bernoulli–Euler beam equation. The assumptions used in formulating this model are that the beam be • • • • Uniform along its span, or length, and slender (l 7 10h) Composed of a linear, homogeneous, isotropic elastic material without axial loads Such that plane sections remain plane Such that the plane of symmetry of the beam is also the plane of vibration so that rotation and translation are decoupled • Such that rotary inertia and shear deformation can be neglected The key to solving for the time response of distributed-parameter systems is the orthogonality of the mode shapes. Note from Table 6.7 that the mode shapes are quite complicated in many configurations. This does not mean that orthogonality is necessarily violated, just that evaluating the integrals in the modal analysis procedure becomes more difficult. TABLE 6.7 EquAtIonS FoR tHE MoDE SHAPE CoEFFICIEntS σn FoR uSE In tAblE 6.4a boundary condition Free–free Clamped–free Clamped–pinned Clamped–sliding Clamped–clamped Formula for σn cosh βnl - cos βnl sinh βnl - sin βnl sinh βnl - sin βnl σn = cosh βnl + cos βnl cosh βnl - cos βnl σn = sinh βnl - sin βnl sinh βnl - sin βnl σn = cosh βnl + cos βnl Same as free–free σn = a These coefficients are used in the calculations for the mode shapes, as illustrated in Example 6.5.1. Timoshenko Beam Theory The model of the transverse vibration of the beam presented in equation (6.91) ignores the effects of shear deformation and rotary inertia. beams modeled including the effects of rotary inertia and shear deformation are called Timoshenko beams. These effects are considered next. As mentioned previously, it is safe to ignore the shear deformation as long as the hz and hy illustrated in Figure 6.10 are small compared with the length of the beam. As the beam becomes shorter, the effect of shear deformation becomes evident. This is illustrated in Figure 6.12, which is a repeat of the element dx of Figure 6.10 with shear deformation included. Sec. 6.5 541 bending Vibration of a beam  (Rotation of the cross section) (Rotation of the cross section dw dx due to pure bending) Shear angle M  dM V  dV (Slope due to bending) dw dx Tangent to centerline of the beam O V Sides no longer parallel M Figure 6.12 The effects of shear deformation on an element of a bending beam. Referring to the figure, the line OA is a line through the center of the element dx perpendicular to the face at the right side. The line OB, on the other hand, is the line through the center tangent to the centerline of the beam, while the line OC is the centerline of the beam while at rest. As the beam bends, the shear angle appears as the length l is decreased relative to the beam width. For the case of a long beam, the lines OB and OA coincide, as in Figure 6.10. The presence of significant shear causes the rectangular infinitesimal element of Figure 6.10 to deform into the distorted almost-diamond shape of Figure 6.12. Referring to Figure 6.12, note that the shear angle given by ψ - dw/dx (i.e., the difference between the total angle due to bending, ψ, and the slope of the centerline of the beam, dw/dx), represents the effect of shear deformation. From elastic considerations [see, e.g., Reismann and Pawlik (1974)] the bending moment equation becomes EI dψ(x, t) = M(x, t) dx and the shear force equation becomes κ2AGc ψ(x, t) - dw(x, t) d = V(x, t) dx (6.103) (6.104) where E, I, A, ψ, V, and M are as defined previously, G is the shear modulus, and κ2 is a dimensionless factor that depends on the shape of the cross-sectional area. Note that some texts use κ instead of κ2 for the shear coefficient, also sometimes referred to as the Timoshenko shear coefficient. The constant κ2 is called a shear coefficient and has been tabulated by Cowper (1966). The shear modulus G and the elastic modulus E are related by the Poisson’s ratio , according the relationship E = 2(1 + v)G. The 542 Distributed-Parameter Systems Chap. 6 Poisson’s ratio is a function of material properties. As in the case of equation (6.86), a dynamic force balance yields ρA(x)dx 02w(x, t) 0t 2 = - c V(x, t) + 0V(x, t) dx d + V(x, t) + f(x, t)dx (6.105) 0x If the rotary inertia is included, then the moment balance on dx previously given by equation (6.87) becomes ρI(x)dx 02ψ(x, t) 0t 2 = c M(x, t) + 0M(x, t) dx d - M(x, t) 0x + c V(x, t) + 0V(x, t) dx2 dx d dx + f(x, t) 0x 2 (6.106) Substitution of equations (6.103) and (6.104) into (6.105) and (6.106) yields the two coupled equations and 0ψ 02ψ 0 0w c EI d + κ2AG a - ψb = ρI 2 0x 0x 0x 0t (6.107) 0 0w 02w c κ2AG a - ψb d + f(x, t) = ρA 2 0x 0x 0t (6.108) governing the vibration of a beam, including the effects of rotary inertia and shear deformation. Assuming that the coefficients are all constant and that no external force is applied, ψ(x, t) can be eliminated and the coupled equations can be reduced to one single equation for free vibration of uniform beams. This is EI ρ2I 04w 02w 04w E 04w + ρA 2 - ρI a1 + 2 b 2 2 + 2 = 0 4 0t κ G 0x 0t κ G 0t 4 0x (6.109) Equation (6.109) is now subject to four initial conditions and four boundary conditions. For a clamped end, the boundary conditions become (at x = 0, say) ψ(0, t) = w(0, t) = 0 (6.110) At a simply supported end, they become EI 0ψ(0, t) = w(0, t) = 0 0x (6.111) and at a free end they become κ2AG a 0ψ 0w - ψb = EI = 0 0x 0x (6.112) Sec. 6.5 543 bending Vibration of a beam These equations can be solved by the methods suggested for the beam model given by equation (6.91). Equation (6.91) is called the Euler–Bernoulli beam model or classical beam model and equation (6.109) is called the Timoshenko beam model. Which of these two beam models to use is largely dependent on the beam geometry, which modes are of interest, and how many modes are important. A steel beam 12- m long, 15- cm wide, and 0.6- m deep shows a difference of only 0.4% between the first natural frequency of the Euler–bernoulli model and that of the Timoshenko model. This grows to a 10% difference in the fifth natural frequency. Hence if only the first mode is of interest, the Euler–bernoulli model for this system would be good enough. On the other hand, if the fifth mode is of interest, the Timoshenko model might be worth the extra complexity. For a beam of typical metal with rectangular cross section, the shear deformation is about three times more significant than rotary inertia effects. This is examined in the next example. Example 6.5.2 To obtain a feeling for the differences between the Euler–bernoulli beam model and the more complicated Timoshenko model, calculate the natural frequencies of a pinned– pinned beam using the Timoshenko equation and compare them to the natural frequencies predicted by the beam without shear deformation or rotary inertia as given in Table 6.6. Solution Equation (6.109) cannot easily be solved by separation of variables as suggested in Window 6.4. This results because substitution of w(x, t) = X(x)T(t) into equation (6.3) yields an equation of the form (ai are constants) $ $$ a1X″″T + (a2X + a3X″)T + a4X T = 0 which does not readily separate because of the middle term and the fourth time derivative. This expression can be solved by assuming that the mode shapes of pinned–pinned Euler– bernoulli beam and that of the Timoshenko beam are the same (a reasonable assumption, as one is a more complete account of the other) and that the temporal response is periodic (also reasonable, as the system is undamped). Proceeding with these assumptions, a solution of equation (6.109) is assumed to be of the specific separated form wn(x, t) = sin nπx cos ωnt l where sin (nπx>l) is the assumed nth mode shape of the pinned–pinned configuration and ωn is, at this point, the unknown natural frequency. Substitution of this form into equation (6.109) yields EI a nπ 4 nπx nπ 2 nπ E b sin cos ωnt - ρI a1 + 2 ba b ω2n sin cos ωnt l l l l κG = - ρ2I κ2G ω4n sin nπx nπx cos ωnt + ρAω2n sin cos ωnt l l Each term contains the factor sin (nπx>l)cos (ωnt), which can be factored out to reveal the characteristic equation ω4n ρr 2 κ2G - a1 + n2π2r 2 E n2π2r 2 α2n4π4 2 b = 0 + ω + n l2 l 2 κ2G l4 544 Distributed-Parameter Systems Chap. 6 where r and α are defined by α2 = EI 2 I r = ρA A The characteristic equation for the frequencies ωn is quadratic in ω2n, and hence easily solved. This expression for ωn provides a mechanism for observing the effects of shear deformation and rotary inertia on the natural frequencies of a pinned–pinned beam. Note the following comparisons: 1. Of the two roots for each value of n determined by the frequency equation, the smaller value is associated with bending deformation, and the larger root is associated with shear deformation. 2. The natural frequencies for just the Euler–bernoulli beam are ω2n = α2n4π4 l4 3. The natural frequencies for including just rotary inertia (i.e., no shear, so that terms involving κ are eliminated) are ω2n = α2n4π4 l 4(1 + n2π2r 2 >l 2) 4. The natural frequencies for neglecting the rotary inertia and including the shear deformation are ω2n = α2n4π4 l [1 + (n π r >l 2) E>κG] 4 2 2 2 These expressions can be used to investigate the various effects on the natural frequencies of prismatic beams in the pinned–pinned configuration. These expressions can also be used to gain insight into the effects of rotary inertia and shear deformation for other configurations. by comparing notes 2, 3, and 4, the general effect of shear deformation and rotary inertia is to reduce the value of the natural frequencies. Also note that for high frequencies (large n) the effects of shear deformation and rotary inertia are more pronounced because of the 1 + (nπr>l)2 term in the denominator. This is investigated further in Problem Tb.6.4 at the end of this chapter. n 6.6 VIBRATION OF MEMBRANES AND PLATES The string, rod, and beam models considered in the previous five sections have displacements that are a function of a single direction, x, along the mass. In this sense they are one-dimensional problems. In this section membranes and plates are considered having displacements which are functions of two dimensions (i.e., they are Sec. 6.6 545 Vibration of Membranes and Plates z y w(x,y,t) Figure 6.13 A schematic of a rectangular membrane illustrating the vibration perpendicular to its surface. The boundary conditions are specified around the edge of the membrane (i.e., along the lines x = a, x = 0, y = b, and y = 0). b  x a defined in a plane region in space as illustrated in Figure 6.13). A membrane is essentially a two-dimensional string, and a plate is essentially a two-dimensional beam. The equation for a membrane and plate are not derived here but follow similar arguments used in developing the string and Euler–bernoulli beam equations, respectively. First, consider the vibrational equations for a membrane. A membrane is basically a two-dimensional system that lies in a plane when in equilibrium. A drumhead and a soap film are physical examples of objects that may be modeled as a membrane. The structure itself provides no resistance to bending, so the restoring force is due only to the tension in the membrane. Thus a membrane is similar to a string. The reader is referred to Weaver, Timoshenko, and Young (1990) for the derivation of the membrane equation. let w(x, y, t) represent the displacement in the z direction of a membrane, lying in the xy plane at the point (x, y) and time t. The displacement is assumed to be small, with small slopes, and is perpendicular to the xy plane. let τ be the constant tension per unit length of the membrane and ρ be the mass per unit area of the membrane. Then the equation for free vibration is given by τ∇ 2w(x, y, t) = ρwtt(x, y, t) (6.113) where x and