Mechanical Vibration and Shock Analysis, Volume 3, Random Vibration E-Book

Christian Lalanne

• Herausgeber: John Wiley & Sons
• Kategorie: Fachliteratur
• Sprache: Englisch
• Veröffentlichungsjahr: 2014

Seitenzahl: 441

Beschreibung

The vast majority of vibrations encountered in the real environment are random in nature. Such vibrations are intrinsically complicated and this volume describes the process that enables us to simplify the required analysis, along with the analysis of the signal in the frequency domain. The power spectrum density is also defined, together with the requisite precautions to be taken in its calculations as well as the processes (windowing, overlapping) necessary to obtain improved results. An additional complementary method - the analysis of statistical properties of the time signal - is also described. This enables the distribution law of the maxima of a random Gaussian signal to be determined and simplifies the calculation of fatigue damage by avoiding direct peak counting.

Leseprobe

Table of Contents

Foreword to Series

Introduction

List of Symbols

Chapter 1 Statistical Properties of a Random Process

1.1. Definitions

1.2. Random vibration in real environments

1.3. Random vibration in laboratory tests

1.4. Methods of random vibration analysis

1.5. Distribution of instantaneous values

1.6. Gaussian random process

1.7. Rayleigh distribution

1.8. Ensemble averages: through the process

1.9. Temporal averages: along the process

1.10. Significance of the statistical analysis (ensemble or temporal)

1.11. Stationary and pseudo-stationary signals

1.12. Summary chart of main definitions

1.13. Sliding mean

1.14. Test of stationarity

1.15 Identification of shocks and/or signal problems

1.16. Breakdown of vibratory signal into “events”: choice of signal samples

1.17. Interpretation and taking into account of environment variation

Chapter 2 Random Vibration Properties in the Frequency Domain

2.1. Fourier transform

2.2. Power spectral density

2.3. Amplitude Spectral Density

2.4. Cross-power spectral density

2.5. Power spectral density of a random process

2.6. Cross-power spectral density of two processes

2.7. Relationship between the PSD and correlation function of a process

2.8. Quadspectrum – cospectrum

2.9. Definitions

2.10. Autocorrelation function of white noise

2.11. Autocorrelation function of band-limited white noise

2.12. Peak factor

2.13. Effects of truncation of peaks of acceleration signal on the PSD

2.14. Standardized PSD/density of probability analogy

2.15. Spectral density as a function of time

2.16. Sum of two random processes

2.17. Relationship between the PSD of the excitation and the response of a linear system

2.18. Relationship between the PSD of the excitation and the cross-power spectral density of the response of a linear system

2.19. Coherence function

2.20. Transfer function calculation from random vibration measurements

Chapter 3 Rms Value of Random Vibration

3.1. Rms value of a signal as a function of its PSD

3.2. Relationships between the PSD of acceleration, velocity and displacement

3.3. Graphical representation of the PSD

3.4. Practical calculation of acceleration, velocity and displacement rms values

3.5. Rms value according to the frequency

3.6. Case of periodic signals

3.7. Case of a periodic signal superimposed onto random noise

Chapter 4 Practical Calculation of the Power Spectral Density

4.1. Sampling of signal

4.2. PSD calculation methods

4.3. PSD calculation steps

4.4. FFT

4.5. Particular case of a periodic excitation

4.6. Statistical error

4.7. Statistical error calculation

4.8. Influence of duration and frequency step on the PSD

4.9. Overlapping

4.10. Information to provide with a PSD

4.11. Difference between rms values calculated from a signal according to time and from its PSD

4.12. Calculation of a PSD from a Fourier transform

4.13. Amplitude based on frequency: relationship with the PSD

4.14. Calculation of the PSD for given statistical error

4.15. Choice of filter bandwidth

4.16. Probability that the measured PSD lies between ± one standard deviation

4.17. Statistical error: other quantities

4.18. Peak hold spectrum

4.19. Generation of random signal of given PSD

4.20. Using a window during the creation of a random signal from a PSD

Chapter 5 Statistical Properties of Random Vibration in the Time Domain

5.1. Distribution of instantaneous values

5.2. Properties of derivative process

5.3. Number of threshold crossings per unit time

5.4. Average frequency

5.5. Threshold level crossing curves

5.6. Moments

5.7. Average frequency of PSD defined by straight line segments

5.8. Fourth moment of PSD defined by straight line segments

5.9. Generalization: moment of order n

Chapter 6 Probability Distribution of Maxima of Random Vibration

6.1. Probability density of maxima

6.2. Moments of the maxima probability distribution

6.3. Expected number of maxima per unit time

6.4. Average time interval between two successive maxima

6.5. Average correlation between two successive maxima

6.6. Properties of the irregularity factor

6.7. Error related to the use of Rayleigh’s law instead of a complete probability density function

6.8. Peak distribution function

6.9. Mean number of maxima greater than the given threshold (by unit time)

6.10. Mean number of maxima above given threshold between two times

6.11. Mean time interval between two successive maxima

6.12. Mean number of maxima above given level reached by signal excursion above this threshold

6.13. Time during which the signal is above a given value

6.14. Probability that a maximum is positive or negative

6.15. Probability density of the positive maxima

6.16. Probability that the positive maxima is lower than a given threshold

6.17. Average number of positive maxima per unit of time

6.18. Average amplitude jump between two successive extrema

6.19. Average number of inflection points per unit of time

Chapter 7 Statistics of Extreme Values

7.1. Probability density of maxima greater than a given value

7.2. Return period

7.3. Peak ℓp expected among Np peaks

7.4. Logarithmic rise

7.5. Average maximum of Np peaks

7.6. Variance of maximum

7.7. Mode (most probable maximum value)

7.8. Maximum value exceeded with risk α

7.9. Application to the case of a centered narrowband normal process

7.10. Wideband centered normal process

7.11. Asymptotic laws

7.12. Choice of type of analysis

7.13. Study of the envelope of a narrowband process

Chapter 8 Response of a One-Degree-of-Freedom Linear System to Random Vibration

8.1. Average value of the response of a linear system

8.2. Response of perfect bandpass filter to random vibration

8.3. The PSD of the response of a one-dof linear system

8.4. Rms value of response to white noise

8.5. Rms value of response of a linear one-degree of freedom system subjected to bands of random noise

8.6. Rms value of the absolute acceleration of the response

8.7. Transitory response of a dynamic system under stationary random excitation

8.8. Transitory response of a dynamic system under amplitude modulated white noise excitation

Chapter 9 Characteristics of the Response of a One-Degree-of-Freedom Linear System to Random Vibration

9.1. Moments of response of a one-degree-of-freedom linear system: irregularity factor of response

9.2. Autocorrelation function of response displacement

9.3. Average numbers of maxima and minima per second

9.4. Equivalence between the transfer functions of a bandpass filter and a one-degree-of-freedom linear system

Chapter 10 First Passage at a Given Level of Response of a One-Degree-of-Freedom Linear System to a Random Vibration

10.1. Assumptions

10.2. Definitions

10.3. Statistically independent threshold crossings

10.4. Statistically independent response maxima

10.5. Independent threshold crossings by the envelope of maxima

10.6. Independent envelope peaks

10.7. Markov process assumption

10.8. E.H. Vanmarcke model

Appendix

A1. Laws of probability

A2. 1/nth octave analysis

A3. Conversion of an acoustic spectrum into a PSD

A4. Mathematical functions

A5. Complements to the transfer functions

A6. Calculation of integrals

Bibliography

Index

Summary of Other Volumes in the Series

Summary of Volume 1 Sinusoidal Vibration

Summary of Volume 2 Mechanical Shock

Summary of Volume 4 Fatigue Damage

Summary of Volume 5 Specification Development

First edition published 2002 by Hermes Penton Ltd © Hermes Penton Ltd 2002Second edition published 2009 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc. © ISTE Ltd 2009Third edition published 2014 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc.

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the undermentioned address:

ISTE Ltd27-37 St George’s RoadLondon SW19 4EUUK

www.iste.co.uk

John Wiley & Sons, Inc.111 River StreetHoboken, NJ 07030USA

www.wiley.com

© ISTE Ltd 2014The rights of Christian Lalanne to be identified as the author of this work have been asserted by him in accordance with the Copyright, Designs and Patents Act 1988.

Library of Congress Control Number: 2014933739

British Library Cataloguing-in-Publication DataA CIP record for this book is available from the British LibraryISBN 978-1-84821-643-3 (Set of 5 volumes)ISBN 978-1-84821-646-4 (Volume 3)

Foreword to Series

In the course of their lifetime simple items in everyday use such as mobile telephones, wristwatches, electronic components in cars or more specific items such as satellite equipment or flight systems in aircraft, can be subjected to various conditions of temperature and humidity, and more particularly to mechanical shock and vibrations, which form the subject of this work. They must therefore be designed in such a way that they can withstand the effects of the environmental conditions to which they are exposed without being damaged. Their design must be verified using a prototype or by calculations and/or significant laboratory testing.

Sizing, and later, testing are performed on the basis of specifications taken from national or international standards. The initial standards, drawn up in the 1940s, were blanket specifications, often extremely stringent, consisting of a sinusoidal vibration, the frequency of which was set to the resonance of the equipment. They were essentially designed to demonstrate a certain standard resistance of the equipment, with the implicit hypothesis that if the equipment survived the particular environment it would withstand, undamaged, the vibrations to which it would be subjected in service. Sometimes with a delay due to a certain conservatism, the evolution of these standards followed that of the testing facilities: the possibility of producing swept sine tests, the production of narrowband random vibrations swept over a wide range and finally the generation of wideband random vibrations. At the end of the 1970s, it was felt that there was a basic need to reduce the weight and cost of on-board equipment and to produce specifications closer to the real conditions of use. This evolution was taken into account between 1980 and 1985 concerning American standards (MIL-STD 810), French standards (GAM EG 13) or international standards (NATO), which all recommended the tailoring of tests. Current preference is to talk of the tailoring of the product to its environment in order to assert more clearly that the environment must be taken into account from the very start of the project, rather than to check the behavior of the material aposteriori. These concepts, originating with the military, are currently being increasingly echoed in the civil field.

Tailoring is based on an analysis of the life profile of the equipment, on the measurement of the environmental conditions associated with each condition of use and on the synthesis of all the data into a simple specification, which should be of the same severity as the actual environment.

This approach presupposes a proper understanding of the mechanical systems subjected to dynamic loads and knowledge of the most frequent failure modes.

Generally speaking, a good assessment of the stresses in a system subjected to vibration is possible only on the basis of a finite element model and relatively complex calculations. Such calculations can only be undertaken at a relatively advanced stage of the project once the structure has been sufficiently defined for such a model to be established.

Considerable work on the environment must be performed independently of the equipment concerned either at the very beginning of the project, at a time where there are no drawings available, or at the qualification stage, in order to define the test conditions.

In the absence of a precise and validated model of the structure, the simplest possible mechanical system is frequently used consisting of mass, stiffness and damping (a linear system with one degree of freedom), especially for:

– the comparison of the severity of several shocks (shock response spectrum) or of several vibrations (extreme response and fatigue damage spectra);

– the drafting of specifications: determining a vibration which produces the same effects on the model as the real environment, with the underlying hypothesis that the equivalent value will remain valid on the real, more complex structure;

– the calculations for pre-sizing at the start of the project;

– the establishment of rules for analysis of the vibrations (choice of the number of calculation points of a power spectral density) or for the definition of the tests (choice of the sweep rate of a swept sine test).

This explains the importance given to this simple model in this work of five volumes on “Mechanical Vibration and Shock Analysis”.

Volume 1 of this series is devoted to sinusoidal vibration. After several reminders about the main vibratory environments which can affect materials during their working life and also about the methods used to take them into account, following several fundamental mechanical concepts, the responses (relative and absolute) of a mechanical one-degree-of-freedom system to an arbitrary excitation are considered, and its transfer function in various forms are defined. By placing the properties of sinusoidal vibrations in the contexts of the real environment and of laboratory tests, the transitory and steady state response of a single-degree-of-freedom system with viscous and then with non-linear damping is evolved. The various sinusoidal modes of sweeping with their properties are described, and then, starting from the response of a one-degree-of-freedom system, the consequences of an unsuitable choice of sweep rate are shown and a rule for choice of this rate is deduced from it.

Volume 2 deals with mechanical shock. This volume presents the shock response spectrum (SRS) with its different definitions, its properties and the precautions to be taken in calculating it. The shock shapes most widely used with the usual test facilities are presented with their characteristics, with indications how to establish test specifications of the same severity as the real, measured environment. A demonstration is then given on how these specifications can be made with classic laboratory equipment: shock machines, electrodynamic exciters driven by a time signal or by a response spectrum, indicating the limits, advantages and disadvantages of each solution.

Volume 3 examines the analysis of random vibration which encompasses the vast majority of the vibrations encountered in the real environment. This volume describes the properties of the process, enabling simplification of the analysis, before presenting the analysis of the signal in the frequency domain. The definition of the power spectral density is reviewed, as well as the precautions to be taken in calculating it, together with the processes used to improve results (windowing, overlapping). A complementary third approach consists of analyzing the statistical properties of the time signal. In particular, this study makes it possible to determine the distribution law of the maxima of a random Gaussian signal and to simplify the calculations of fatigue damage by avoiding direct counting of the peaks (Volumes 4 and 5). The relationships that provide the response of a one-degree-of-freedom linear system to a random vibration are established.

Volume 4 is devoted to the calculation of damage fatigue. It presents the hypotheses adopted to describe the behavior of a material subjected to fatigue, the laws of damage accumulation and the methods for counting the peaks of the response (used to establish a histogram when it is impossible to use the probability density of the peaks obtained with a Gaussian signal). The expressions of mean damage and its standard deviation are established. A few cases are then examined using other hypotheses (mean not equal to zero, taking account of the fatigue limit, non-linear accumulation law, etc.). The main laws governing low cycle fatigue and fracture mechanics are also presented.

Volume 5 is dedicated to presenting the method of specification development according to the principle of tailoring. The extreme response and fatigue damage spectra are defined for each type of stress (sinusoidal vibrations, swept sine, shocks, random vibrations, etc.). The process for establishing a specification as from the lifecycle profile of the equipment is then detailed taking into account the uncertainty factor (uncertainties related to the dispersion of the real environment and of the mechanical strength) and the test factor (function of the number of tests performed to demonstrate the resistance of the equipment).

First and foremost, this work is intended for engineers and technicians working in design teams responsible for sizing equipment, for project teams given the task of writing the various sizing and testing specifications (validation, qualification, certification, etc.) and for laboratories in charge of defining the tests and their performance following the choice of the most suitable simulation means.

Introduction

The vibratory environment found in the majority of vehicles essentially consists of random vibrations. Each recording of the same phenomenon results in a signal different from the previous ones. The characterization of a random environment therefore requires an infinite number of measurements to cover all the possibilities. Such vibrations can only be analyzed statistically.

The first stage consists of defining the properties of the processes comprising all the measurements, making it possible to reduce the study to the more realistic measurement of single or several short samples. This means evidencing the stationary character of the process, making it possible to demonstrate that its statistical properties are conserved in time, thus its ergodicity, with each recording representative of the entire process. As a result, only a small sample consisting of one recording has to be analyzed (Chapter 1).

The value of this sample gives an overall idea of the severity of the vibration, but the vibration has a continuous frequency spectrum that must be determined in order to understand its effects on a structure. This frequency analysis is performed using the power spectral density (PSD) (Chapter 2) which is the ideal tool for describing random vibrations. This spectrum, a basic element for many other treatments, has numerous applications, the first being the calculation of the rms (root mean square) value of the vibration in a given frequency band (Chapter 3).

The practical calculation of the PSD, completed on a small signal sample, provides only an estimate of its mean value, with a statistical error that must be evaluated. Chapter 4 shows how this error can be evaluated according to the analysis conditions and how it can be reduced, before providing rules for the determination of the PSD.

The majority of signals measured in the real environment have a Gaussian distribution of instantaneous values. The study of the properties of such a signal is extremely rich in content (Chapter 5). For example, knowledge of the PSD alone gives access, without having to count the peaks, to the distribution of the maxima of a random signal (Chapter 6), and to the law of distribution of the largest peaks, in itself useful information for the pre-sizing of a structure (Chapter 7).

It is also used to determine the response of a system with one degree-of-freedom (Chapters 8 and 9), which is necessary to calculate the fatigue damage caused by the vibration in question (Volume 4).

The study of the first crossing of a given response threshold for a one-degree-of-freedom system can also be useful in estimating the greatest stress value over a given duration. Different methods are presented (Chapter 10).

List of Symbols

The list below gives the most frequent definition of the main symbols used in this book. Some of the symbols can have another meaning which will be defined in the text to avoid any confusion.

a

Threshold value of

ℓ

(t) or maximum of

ℓ

(t)

A

Maximum of A (t)

A(t)

Envelope of a signal

b

Exponent

c

Viscous damping constant

e(t)

Narrow band white noise

E()

Expectation of…

E

1

( )

First definition of error function

E

2

( )

Second definition of error function

Erf

Error function

E( )

Expected function of …

f

Frequency of excitation

f

samp

.

Sampling frequency

f

max

Maximum frequency

f

0

Natural frequency

g

Acceleration due to gravity

G

Particular value of power spectral density

G( )

Power spectral density for 0 ≤ f ≤ ∞

Measured value of G( )

G

ℓ

u

( )

Cross-power spectral density

h

Interval (f/f

0

) or f

2

/f

1

h(t)

Impulse response

H( )

Transfer function

i

k

Stiffness

K

Number of subsamples

ℓ

Value of

ℓ

(t)

Mean value of

ℓ

(t)

Average maximum of N

p

peaks

ℓ

rms

Rms value of

ℓ

(t)

Rms value of

ℓ

(t)

Generalized excitation (displacement)

First derivative of

ℓ

(t)

Second derivative of

ℓ

(t)

L

Given value of

ℓ

(t)

L

rms

Rms value of filtered signal

L(Ω)

Fourier transform of

ℓ

(t)

Fourier transform of

m

Mean

M

Number of points of PSD

M

a

Average number of maxima which exceeds threshold per unit time

M

n

Moment of order n

n

Order of moment or number of degrees of freedom

n

a

Average number of crossings of threshold a per unit time

Average number of crossings of threshold

a

with positive slope per unit time

n

0

Average number of zero-crossings per unit time

Average number of zero-crossings with positive slope per second (average frequency)

Average number of maxima per unit time

N

Number of curves or Number of points of signal orNumbers of dB

N

p

Number of peaks

Average number of crossings of threshold

a

with positive slope for given length of time

Average number of zero-crossings with positive slope for given length of time

Average number of positive maxima for given length of time

p( )

Probability density

p

N

( )

Probability density of largest maximum over given duration

P

Probability

PSD

Power spectral density

q

q

E

Probability that a maximum is positive

Probability that a maximum is negative

q( )

Probability density of maxima of

ℓ

(t)

q(θ)

Reduced response

First derivative of q(θ)

Second derivative of q(θ)

Q

Q factor (quality factor)

Q( )

Distribution function of maxima of

ℓ

(t)

Q(u)

Probability that a maximum is higher than a given threshold

r

Irregularity factor

rms

Root mean square (value)

r(t)

Temporal window

R

Slope in dB/octave or Ratio of the number of minima to the number of maxima

R

E

( )

Auto-correlation function of envelope

R

ℓ

u

Cross-correlation function between

ℓ

(t) and u(t)

R(f)

Fourier transform of r(t)

R(t)

Envelope of maxima of u(t)

First derivative of R(t)

R(

τ

)

Auto-correlation function

s

Standard deviation

S

0

Value of constant PSD

S( )

Power spectral density for –∞ ≤ f ≤ +∞

t

Time

T

Duration of sample of signal or duration of vibration

T

a

Average time between two successive maxima

u

Ratio of threshold a to rms value

ℓ

rms

of

ℓ

(t)

u

0

Initial value of u(t)

Initial value of

Average of highest peaks

u

rms

Rms value of u(t)

Rms value of

Rms value of

u(t)

Generalized response

First derivative of u(t)

Second derivative of u(t)

v

Ratio a/u

rms

v

rms

Rms value of

x

rms

Rms value of x(t)

Absolute acceleration of base of one-degree-of-freedom system

Rms value of

Maximum value of

y

rms

Rms value of y(t)

Rms value of

Rms value of

z

rms

Rms value of z(t)

Rms value of

Rms value of

α

Time-constant of the probability density of the first passage of a threshold orRisk of up-crossing or

β

2 (1 – 2

ξ

2

)

Variable of chi-square with n degrees of freedom

δt

Time step

δ( )

Dirac delta function

Δ

τ

Effective time interval

Δf

Frequency interval between half-power points or frequency step of the PSD

ΔF

Bandwidth of analysis filter

Δ

ℓ

Interval of amplitude of

ℓ

(t)

Δt

Time interval

ε

Statistical error or Euler’s constant (0.57721566490…)

γ

ℓ

u

Coherence function between

ℓ

(t) and u(t)

φ

Phase

Φ(t)

Gauss complementary distribution function

λ( )

Reduced excitation

π

3.14159265…

θ

Reduced time (ω

0

t)

μ

n

Central moment of order n

μ′

n

Reduced central moment of order n

π

3.14159265 …

ρ

Correlation coefficient

τ

Delay

τ

m

Average time between two successive maxima

ω

0

Natural pulsation (2 π f

0

)

Ω

Pulsation of excitation (2 π f)

ξ

Damping factor

ψ

Phase

Chapter 1

Statistical Properties of a Random Process

1.1. Definitions

1.1.1. Random variable

A random variable is a quantity whose instantaneous value cannot be predicted. Knowledge of the values of the variable before time t does not make it possible to deduce the value at the time t from it.

Example: the Brownian movement of a particle.

If a vibration was perfectly random, its analysis would be impossible. The points that define the signal would have an amplitude that varied in a completely unpredictable way. Thankfully, in practice, it is possible to associate with all the points that characterize the signal a probability law which will enable a statistical analysis [AND 11].

The principal characteristic of a random vibration is to simultaneously excite all the frequencies of a structure [TUS 67]. In contrast to sinusoidal functions, random vibrations are made up of a continuous range of frequencies, the amplitude of the signal and its phase varying with respect to time in a random fashion [TIP 77] [TUS 79]. Thus, the random vibrations are also called noise.

Random functions are sometimes defined as a continuous distribution of sinusoids of all frequencies whose amplitudes and phases vary randomly with time [CUR 64], [CUR 88].

1.1.2. Random process

Let us consider, as an example, the acceleration recorded at a given point on the dial of a truck traveling on a good road between two cities A and B. For a journey, the recorded acceleration obeys the definition of a random variable. The vibration characterized by this acceleration is said to be random or stochastic.

Complexity of the analysis

Even in the most simple hypothesis where a vehicle runs at a constant speed on a straight road in the same state, each vibration measure (t) at one point of the vehicle is different from the other. An infinity of measures to completely characterize the trip should be completed