Scribd Logo
Upload

Welcome to Scribd!

  • Newland An Introduction To Random Vibration

    Uploaded by

    Malena Rufino
    11 views175 pages
    Newland-an-Introduction-to-Random-Vibration

    Original Title

    Newland an Introduction to Random Vibration

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    Newland-an-Introduction-to-Random-Vibration

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    11 views175 pages

    Newland An Introduction To Random Vibration

    Uploaded by

    Malena Rufino
    Newland-an-Introduction-to-Random-Vibration

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 175
    An introduction to Random vibrations, spectral and wavelet analysis Third edition D. E. Newland Professor of Engineering University of Cambridge >> man man Longman Scientific &| Technical author's companion book Mechanical Vibration Analysis and vutation is also published by Longma Longman Scientitic & Technical, wan Group UK Limited, House, Burnt Mill, Hi Essex CM20 26. England and Assctated Companies throughout the world. Copublished in the United States with John Wiley & Sons, Inc., 605 Third Avenue, New York, NY 10 «D.E. Newland 1975, 1984, 1993 All rights reserved: no part of this publication may be reproduced, stored in a retrieval system, oF transmitted in any form or by any means, electronic, mechanical. photocopying, recording, of otherwise Publishers or a licence permutting restricted copying in the United Kingdom issued by the Copyright Licensing rency Ltd, 90 Towenham Court Road, London WIP 91E. First published 1975 Reprinted 19/8, IY8U, 1981, 1985 Second edition 1984 Reprinted 1986, 1987. 1959, 1991 Third edition 1993 British Library Cataloeuing in Public: Acatalogue entry for this title aval the British Library ISBI 0982 21989 6 in Data able from Library of Congress Cataloying-in-Publication Data ISBN 0-470-22153-4 (USA only) Set in 10 on 12pt ‘Limes by 6 Printed in Malaysia by GPS Contents Preface to the first edition Preface to the second edition Preface to the third edition Acknowledgemenis List of symbols 1 Introduction to probability distributions and averages Probability density function Gaussian distribution Calculation of averages Probability distribution function ry Joint probabi ty dis Second-order probability functions Second-urder averages Conditional probability Second-order Gaussian distribution Ensemble averaging 3 Correlation Autocorrelation Cross-correlation 4 Fourier analysis Fourier integral Complex form of the Fourier transform 5 Spectral density Narrow band and broad band processes Spectral density of a derived process Cross-spectral density Note on the units of spectral density Comtents Excitation — response rela Classical approach Frequency response method Impulse response method Relationship between the frequency response and impulse response functions Calculation of response to an arbitrary input ear systems Transmission of random vibration Autocorrelation Spectral density Mean square response Cross-correlation Cross-spectral density Probability distributions Statistics of narrow band processes Crossing anaiysis Distribution of peaks Frequer y of maxima Accuracy of measurements Analogue spectrum analysis Variance of the measurement Analysis of finite length records Confidence limits Digital spectral analysis I: Discrete Fourier transforms Disercte Fourie: Wransforms Fourier transforms of periodic functions Aliasing Calculation of spectral estimates Digital spectral analysis Il: Windows and smoothing Relationship between linear and circular correlation Fourier transform of a train of aperiodic functions Basic lag and spectral windows Smoothing spectral estimates Extending record length by adding zeros Summary Practical considerations The fast Fourier transform Rasic theary Sample calculation 33 54 55 58 60 62 67 6s 69 103 108 113 4 116 118 120 125 126 129 131 137 139 142 146 150 150 152 13 14 15 16 17 Contents Programming flow charts Pracucal value of FFT Alternative algorithms Pseudo random processes Random binary process Pseudo random binary signals Random multi-level process Spectrum of a mulu-level process Generation of random numbers Synthesis of correlated noise sources Application notes Response of a resonant mode to broad band excitation Fatigue and failure due to random vibration Excitation by random surface irregularities Simulation of random environments Frequency response function and coherency measurements Weibull distribution of peaks Multi-dimensional spectral analysis Two-dimensional Fourier series Properties of the two-dimensional DFT Spectral density of a multi-dimensional random process Discrete spectral density and circular correlation functions for a two-dimensional random process Two-dimensional windows ‘Two-dimensional smoothing Artificial generation of a two-dimensional random process Generation of an isotropic surface Cross-spectral density between parallel tracks a random surface oss a Response of continuous linear systems to stationary random excitation Response to excitation applied at a point Response to distributed excitation Normal mode anal; Kinetic energy of a flat plate subjected to uncorrelated random excitation Single degree-of-freedom analogy Discrete wavelet analysis Basic ideas Dilation equations viii Contents Dilation wavelets Properties of the wavelet coefficients Circular wavelet transforms Discrete wavelet transforms Properties of the DWT Mean-square maps Convolution by wavelets Two-dimensional wavelet transforms Harmonic wavelets Concluding comments Appendices Appendix 1 Table of integrals Appendix 2. Computer programs Appendix 3 The Gaussian probability integral Appendix 4 Distribution of y2 Appendix 5 Random numbers Appendix © Distribution of reverse arrangements for a random series Appendix 7 Wavelet programs Problems Answers to problems References Index 303 307 322 326 339 348 353 355 359 370 371 373 379 381 383 487 465 472 Preface to the first edition The importance of random vibrations has increased in recent years. Whereas ten years ago undergraduate courses in engineering and applied science barely mentioned the subject, now a well-educated engineer needs at least some familiarity with the concepts and methods involved Furthermore the applications of random process theory extend far beyond traditional engincering fields. The same methods of analysis which apply to the time-varying pressure in an aircraft's turbulent boundary layer apply also to the daily fluctuations of an economic index. Because of this growing importance, the theory of random vibrations is now being introduced to undergraduate students, and the methods of measurement and analysis are being increasingly used in research laboratories and in industry. This book has two main objectives: first, to introduce the fundamental ideas of random vibrations, and, secondly, to deal in some depth with digital spectral analysis, which involves the measurement and analysis of random vibrations. Chaptets 1 to 9, which take up about half the book, try to meet the first objective and provide the background for an introduction to the subject at undergraduate level. Chapters 10, 11 and 12 are then more specialized and cover digital spectral analysis and the fast Fourier transform. They deal in detail with the application of discrete Fourier transforms to the analysis of sampled data. The emphasis is on the so-called “direct method” of spectrum estimation based on the FFT (fast Fourier transform) and the approximations involved in this method are discussed at length. In the author's experience, many people who find themselves using the FFT for spectral analysis have an incomplete understanding of the nature of the approximations involved, and it is hoped that these chapters will provide the necessary explanation. The last two chapters of the book deal with the properties of pseudo-random processes and a variety of more specialized subjects including the measurement and application of coherence functions and a brief introduction to the analysis of non-stationary processes. Au impuriani feature of and answers; most of these have been specially prepared to illustrate points in the text. They serve the dua! purpose of allowing readers to test their 1c Luva is the sei uf Laicfully selecied prob x Preface understanding while at the same time allowing the author to include alternative developments of the theory and different applications which could not otherwise have been discussed because of lack of space. Although there 18 a good deal of mathematics, this has been kept as simple as a reasonably complete treatment of the subject allows. The knowledge assumed is limited to that normally possessed by final year undergraduates in the applied sciences, and no additional statistical background is required, This slightly restricts the mathcmatical scope of the book. but the text is supported by references to more specialized books and to the original literature where appropri a ' < if they wish. The author will be pleased to hear from readers who spot errors and misprints, or who find points where amplification is desirable or see other ways of improving the book. All such suggestions will be gratefully received and most carefully studied, Sheffield, 1974 D. E. NewLand Disclaimer of warranty Neither the author nor publisher makes any warranty. express or implied, that the calculation methods, proc i are free from error. The application of these methods and procedures is at the user's own risk and the author and publisher disclaim all liability for damages, whether direct. incidental or consequential, arising from such application or from any other use of this book Preface to the second edition For the sccond edition, two additional chapters have been added. Chapter 15 introduces multi-dimensional random processes and discusses multi-dimensional spectral analysis in detail. This is important when a random process depends on more than one independent variable. For example, the height of the sea depends an time and on position The logic of the multi-dimensional discrete Fourier transform is explained and Appendix 2 now contains a two-dimensional DFT program as well as the original one-dimensional FFT program, which it uses. Chapter 16 extends the previous theory to the random vibration of a continuous linear system subjected to distributed random excitation. This introduces a specialized but growing field of random vibration analysis A new section on the Weibull distribution of peaks has been added to Chapter 14 and a number of other small additions made. The computer programs in Appendix 2 are now listed in both Fortran and Basic languages. Readers with suitabie mini-computers will be able to use the latter programs for some of the exercises. Because HP-enhanced Basic has been uscd, some modifications may be necessary for computers using other versions of Basic, but it is hoped that this will not present serious difficulties. As before, the author will be glad to hear from readers and will be most grateful for the notification of errors and for suggestions for improvement. Cambridge, 1984 D. E. NewLand Preface to the third edition Since completing the second edition in 1984, the most important development in signal analysis has been the wavelet transform. This overcomes a fundamental disadvantage of the Fourier transform. When applied to non-stationary processes, the Fourier transform gives frequency data averaged over the record length being analysed. The local position of particular features is lost. In contrast, waveler analysis allows a general function of time to be decomposed into a series of orthogonal basis functions, called wavelets, of different lengths and different positions along the time axis. A particular feature can be located from the positions of the wavelets into which it is decomposed. Results of the wavelet transform can be presented in the form of a contour map in the frequency-time plane. This display, which has been compared to the way that notes are shown on a musical score, allows the changing spectral composition of non-stationary signals to be measured and compared. Therefore it has scemcd desieable to add a chapter on the with three objectives: (i) to include enough of the basic theory of wavelet analysis to establish the background for practical calculations; (ii) to explain how the discrete wavelet transform (DWT) works: (iii) to indicate some of the potential applications of the DWT. In addition it has been necessary to prepare computer programs for the DWT and its inverse, and for a two-dimensional DWT and and also for displaying their results. Considerable thought has been given to the language for these programs and MATLAB** has been chosen as currently the most appropriate. To use the programs in Appendix 7 it will therefore be necessary to have MATLAB software. This is becoming increasingly popular in the teaching world as well as in industry and many readers will have access to it already, Others may not, and information is given at the end of the book about how MATLAB may be obtained. Although the program listings will not be usable by readers without MATLAB, the program logic iy fully described in Chapter 17, and it is possible for readers who wish to recode the listings given in the language of their choice. This edition includes all the material in the second edition (subject to minor editing and corrections) together with the following additions *MATLAB 1s a registered trademark of The MathWorks Ine. Preface to the third edition for Chapter 14 on local spectral density calculations: an extensive new Chapter 17, Discrete wavelet analysis; a new Appendix 7 with computer programs for wavelet analysis; and a set of supporting problems for Chapter 17. The objective of the new material is to put readers into the same position vis-d-vis wavelet analysis as readers of earlier editions have been in respect of spectral analysis. The program listings in Appendix 7 allow readers to make their own wavelet calculations, just as readers can use the programs in Appendix 2 to make their own spectral analysis calculations. To recognize the importance of wavelet analysis in the overall scope of the new book, it has deen decided Lo change the titie from Random Vibrations and Spectral Analysis to Random Vibrations, Spectral and Wavelet Analysis. Although wavelets have been studied widely in the mathematical world, their application to vibration analysis is still new. The DWT provides an important procedure for analysing non-stationary vibration records, and it is hoped that this third edition will help vibration practitioners to understand and use the wavelet transform. As for the earlier editions, the author would be glad to hear from readers with corrections and suggestions for improvement. Cambridge, 1993 D. E. NEWLAND Acknowledgements for the first edition During the early 1960s I had the privilege of working with Professor J. P. Den Hartog and Professor Stephen H. Crandall at the Massachusetts Institute of Technology. My professional interest in vibrations dates from that period and it is a pleasure to acknowledge my debt to both these famous teachers. The textbook Random Vibration in Mechanical Systems by S. H. Crandall and W. D. Mark (Academic Press, 1963) has been a valuable reference during preparation of the first half of the present book. T have been most fortunate to have had the untiring assistance of my father-in-law, Philip Mayne, M.A., formerly scholar of King's College, Cambridge. Beginning with the first handwitten draft, he has read every page, checked every formula and worked out every problem. Serious readers will appreciate the enthusiasm needed to undertake such a task and the resolution needed to pursue it to a conclusion. As a result of this help I have made many couetiions aud alieraiiuns iv the uriginal manuscript and T beticve that many obscurities have been removed. J am more than grateful. Completion of the manuscript coincided with a visit to Sheffield by Professor Philip G. Hill of Queen's University, Kingston, Ontario, who has been kind enough to read and comment helpfully on a major part of the hook. Again 1 have been pleased to incorporate many of the changes suggested. I am also indebted to my colleague Dr H. K. Kohler, who provided information and advice on a wide range of topics, and to Professor R. E. D. Bishop, who reviewed the manuscript for Longman and gave encouragement at a critical stage. My publishers have always been enthusiastic and 1 would like to thank them for the patience and care with which this book has been prepared for publication. Lastly I must mention my secretary, Elaine Ibbotson, who toiled painstakingly with successive drafts, cheerfully accepting my determination to alter every section after, but never before, it had been typed. Only an author who has written a book in his “spare time” knows the tremendous extra demands that this inevitably places on those nearest to him, Without the su not hi DEN. Acknowledgements for the second edition During the preparation of the second edition, I have been glad to have had the help of my doctoral research student David Cebon, formerly of Melbourne University. He has read and commented on the two new chapters and tested the Fortran version of the new computer program. A first draft of the new chapters was used as lecture notes at a short course at Monash University in 1983 and this allowed a number of errors and misprints to be identified. Professor J. D. C. Crisp organized this course and I am very glad to have had that opportunity to try out the new material for the second edition. I am grateful also for discussions at Monash with Professor W. H. Melbourne, asa result of which I decided to adda section on the Weibull distribution of peaks to Chapter 14. Thave been encouraged by the comments I have received from readers of the first edition. As a result of readers’ suggestions, 1 have been able to make a number of corrections and improvements to the text, and Iam most grateful to all those who have heiped in this way. D. The author and publishers are indebted to the following for permission to reproduce copyright material: the McGraw-Hill Book Company for Fig. 12.7(a) which is based on ig. 6.23 of Gold and Rader, Digital Processing of Signals, 1969; the Institute of Electrical and Flectronics Engineers for the computer program in Appendix 2 which is based on that listed on p. 29 of the IEEE Transactions on Lducation, Vol. 12, No. 1, March 1969 (paper by Cooley, Lewis and Welch, “The Fast Fourier Transform and its Applications”): and to the authors and publishers of Fisher and Yates, Statistical Tables for Biological Agricultural and Medical Research (6th edition), Longman, 1974 (previously published by Oliver and Boyd), from which Tables H(i), IV and XXXIII(i) are reproduced here as Appendices 3, 4 and 5 respectively. Acknowledgements for the third edition In preparing the new chapter on wavelets for this edition, I have studied many papers on the mathematics of wavelets. I am indebted to all their authors, but particularly to Professors Ingrid Daubechies and Stéphane Mallat whose pivuccting papers inave deen inmenseiy important in tie subject’s deveiopment For me, a most valuable contribution has been the review paper by Professor Gilbert Strang [115]; this made clear many ideas that T had only gleaned roughly from other publications Many colleagues have helped in various ways, but I would like to mention particularly Dr Bill Fitzgerald who suggested key publications and commented on a first draft of Chapter 17, and Mrs Margaret Margereson who word-processed the text with care and good humour. My thanks are due to both of them. There would not have been time to prepare this new edition without the opportunity afforded by sabbatical leave from the furious pace of university teaching, research and administration in the 1990s. Lam grateful to Cambridge University for that opportunity. DEN, List of symbols The principal symbols in chapters 1 to 14 are defined below with, where appropriate, reference to the chapter or equation in which each symbol first appears, a Go, Gy, a {a,} a) Boy Bas Bay eos Diy eee Glee (i) Amplitude of a periodic function. (i) Constant level y= during the analysis of crossings (Ch. 8). (iii) Peak height for one cycle of a narrow band random process (Ch. 8). (iv) Amplitude of a random binary signal (Ch. 13). (i) Coefficients of Fourier series (4.1). (ii) Coefficients of Fourier series (10.1) (which, 1m (4.1) by a factor 2) (ii) Coefficients of differential equations of motion (6.1). Re-ordered array of terms, r = 1,2,. to the FFT calculation, Fig. 12.4. N, for entry Array of terms, r = 1, 2,..., NV, after completion of one row of butterfly calculations in the FFT alguritiun, Fig, 12.4. Deterministic (non-random) function of time (14.84), (i) Coefficients of Fourier series (4.1). (ii) Coefficients of Fourier series (101) (which differ from (4.1) by a factor 2). (iii) Coefficients of differential equations of motion (6.1) Viscous damping coefficients. List of symbols fi S(O) S(x,y) A) Ay(o), hls. my m, function d(t), Discrete value of the data weighti defined by d, = d{t = ra) (Ch. 11). Data weighting function or “data window” (Ch 11). cosw@t + isinwt. Frequency (Hz = c/s). (i) Filter centre frequency (Hz) (Ch. 9). Gi) Maximum frequency of siguificam spectrai components present in a signal (Hz) (Ch. 10), (iii) Clock frequency of a random binary process (Hz) (Ch. 13). Lower cut-off frequency (Hz) of band-pass filter (Ch. 9). Mathematical function of t, usually representing force. Mathematical function of x and y. Impulse response function for a linear system sub- jected to an impulsive input of unit magnitude ((6.14) and (6.15). Impulse response functions relating the output to a unit impulse at input 1, input 2, ..., respectively. v-t Variable integer. (i) Spring stiffness. (1) Variable integer. Spring stiffnesses. Variable integer. (i) Mean value of a random variable. (ii) Mass. (iii) Variable integer. Mean value of the random variable x, y or of the stationary random process x(t), j{t). (i) Variable integer. (ii) Number of sample values satisfying a certain condition. ng (T) Pa) wT) Px) Bx, 9) pxly) Pan) tose ntaseey 4, u(r) (a) List of symbols xix Number of crossings of the level y = a with posi- tive slope in time T tor a single member function of the random process y(t) (Ch. 8). Probability density function describing the dis- tribution of the random variable a which represents the height of peaks of a narrow band random process (Ch. 8). Probability density function for the distribution of the time T at which a first crossing occurs (Ch. 14) (First order) probability density function describing the distribution of the random variable x (Ch. 1). Second-order probability density function des- cribing the joint distribution of the random variables x and y (Ch. 2). Conditional probability density function for the distribution of x when y is specified (2.10). Probability density function describing the dis- tribution of the random variable 2 (defined in problem 9.3). Variable integer. Variable integer. (i) Time. ii) Variabic integer. Instants of time. Weighted version of R, defined by {R,} as ¥ linear correlation function R(z) (Ch. 11). Continuous aperiodic function of t from: which the values u, can be derived by putting =ur=rd) 0< | will lie within certain limits. The subject of probability is therefore at the heart of random vibration theory and we begin by considering some of the fundamental ideas of probability theory. Probability density function Suppose, first, that we are dealing with a time history which is non-random or deterministic and is, in fact, a sinc wave, Fig, 1.2. In this case we can exactly 4 x(t) = x9 sin wt Fig. 1.2 Waveform for steady state deterministic vibration (1) = Xp sin wt predict the value of x for any given value of t. We can therefore calculate the proportion of time that the waveform spends between any two levels of x. With reference to Fig, 1.3, during one complete cycle, x(t) lies in the band x x) Fig. 1.3 Illustrating calculation of time for which x < x(t)

    You might also like