(Foundations of Engineering Mechanics) Prof. DR V. I. Babitsky (Auth.) - Theory of Vibro-Impact Systems and Applications-Springer-Verlag Berlin Heidelberg (1998) PDF

Foundations of Engineering Mechanics
V. I. Babitsky, Theory of Vibro-Impact Systems and

Applications
Springer-Verlag
Berlin Heidelberg GmbH
v. I. Babitsky
Theory of Vibro-Impact
Systems and Applications
Translated by N. Birkett
With 146 Figures
Springer
Series Editors:
Prof. Dr V. I. Babitsky Prof. Dr J. Wittenburg
Loughborough University Universität Karlsruhe (TH)
Department of Mechanical Engineering Institut für Mechanik
SEn 3TU Loughborough, Leicestershire Kaiserstraße 12
UK D-76128 Karlsruhe/Germany
Author:
Prof. Dr V. I. Babitsky
Translator:
Nicolas Birkett
Grangefield Drive 28
LE7 7NB Rothley, Leicester
UK
ISBN 978-3-662-22534-9 ISBN 978-3-540-69635-3 (eBook)

DOI 10.1007/978-3-540-69635-3
Cataloging-in-Publication Data applied for

Die Deutsche Bibliothek - CIP Einheitsaufnahme
Babickij, Vladimir I.:
Theory of vibro-impact systems with applications / V, I. Babitsky.
Trans!. by N. Birkett. -Berlin; Heidelberg; New York; Barcelona; Budapest; Hong Kong; London;
Milan; Paris; Singapore; Tokyo:
Springer, 1998
(Foundations of engineering mechanics)
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broad-
casting, reproduction on microfIlm or in other ways, and storage in data banks. Duplication of this
publication or parts thereof is permitted only under the provisions of the German Copyright Law of
September 9, 1965, in its current version, and permission for use must always be obtained from
Springer-Verlag. Violations are liable for prosecution act under German Copyright Law.
© Springer-Verlag Berlin Heidelberg 1998

Originally published by Springer-Verlag Berlin Heidelberg New York in 1998.
Softcover reprint ofthe hardcover Ist edition 1998
The use of general descriptive names, registered names, trademarks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant pro-
tective laws and regulations and therefore free for general use.
Product liability: The publisher cannot guarantee the accuracy of any information ab out dosage and
application contained in this book. In every individual case the user must check such information by
consulting the relevant literature.
Cover design: de 'blik, Berlin

Typesetting: Camera ready by author
SPIN: 10630548 6213020 - 5 4 3 2 1 0 - Printed on acid-free paper

To Eleonora and Ilya
Preface to the English edition
The drop of rain maketh a hole in the stone, not by vio/ence, but by oft falling.
Bishop Hugh Latimer

(Born 1485 in Leicestershire, was burned in 1555 for heresy at Oxford)
The first edition of this book, published in Russian, generated wide interest in
the problem within Russia but was not widely known outside. The aim of the
author in writing this book on the synergistic effects of multiple impacts was to
ac hieve the following:
• to understand the unusual resonant behaviour ofvibro-impact systems,

• to accommodate vibro-impact systems within the general theory of
nonlinear vibrations for all forms of mechanical model and excitations,
• to compare different mathematical methods for the analysis of vibro-
impact systems and to develop effective approximate methods based on
frequency and structural approaches,
• to apply frequency methods for analysis of vibro-impact processes in
complex mechanical structures interacting with driving and control units,
• to develop an approach for the structural synthesis of vibro-impact
systems,
• to explain the dynamics of some important engineering systems.
In order to solve these problems, the author extended the mechanical

description of vibro-impact systems by applying contact compliances, distributed
flexible mechanical structures and complex excitations.
The theory of vibro-impact systems that was developed, revealed new
nonlinear phenomena. The exploitation of these phenomena gave an impetus for
the promotion of different engineering applications. To achieve this the author
founded the Laboratory at the Blagonravov Institute of Machine Studies of the
Russian Academy of Sciences which carried out research and development on a
wide range of engineering systems that resulted in the design of effective rigs.
These activities were generally described in the article included as Appendix II of
this book, which we recommend to the reader for initial acquaintance with the
problem. Regular Symposiums and Seminars on "Dynamics of Vibro-Impact
Systems" were conducted within the mentioned activity, which stimulated
intensive development ofthis area and mutual applications.
The following lectures and demonstrations in different countries together with
VIII Preface to English edition
interest in this topic from industry, inspired the author to propose the revised
English translation of this book. The development in recent years of powerful
software such as Matlab-Simulink gave new opportunities to simulate vibro-
impact structures using frequency and structural methods. The book can also be
used as a foundation for these new applications. In order to demonstrate these
possibilities, we have added to the book Appendix I, written together with Dr A
Veprik who also helped the author with the description of recently discovered
vibro-impact phenomena in § 14 section 5 and § 16 section 4. The book is further
upgraded with an additional list of new references.
We hope that the book will introduce these general approaches for important
areas of mechanical engineering to English speaking readers and help the
generation of new research and development specialists in the analysis of such
complex mechanical structures. From the book, students and their teachers can
acquire knowledge of the mechanical and engineering aspects of dynamics of
strongly nonlinear systems.
The translator of the book, Mr Birkett, and the staff of Springer-Verlag were
extremely co-operative during all stages in the preparation ofthe book. The author
is very grateful to all the individuals who participated in accomplishing this
project.
Vladimir Babitsky Loughborough, Leicestershire

Preface to the Russian edition
It is difficult to think of a physical effect other than impact that generates more
significant results for so small a resource requirement. It is constantly attracts
interest to study it and its varied applications. The use of impact principles in
household and industrial practices has evolved on its own over the years. For
thousands of years, impact tools have developed in a complex manner, remaining
one of the most productive and effective tools throughout. From the first primitive
hammers, to contemporary forging and riveting machines, from ancient pick-axes
to novel types of hammers and pneumatic drills, from battering rams to perfectly
designed concrete breakers and perforators, these represent the extremes of this
long lasting process.
From the wide circle of engineering problems linked to the study of impact
interactions, it is reasonable to separate the cIass of problems where systematic
collisions occur. Systems in which such a form of interaction occurs are known as
vibro-impact. This type of interaction becomes significant compared to a single
impact only when the collision frequency matches the natural frequency of the
interacting elastic systems. In the state of affairs specified, collisions occurring in
the excited system, which has some energy resource that can be built up from
impact to impact, cause various forms of vibration effects. What has been said
defines the subject of the theory of vibro-impact systems as a self-contained
section within the theory of mechanical vibrations.
The book systematically presents the theory of vibration for mechanical
systems whose motion is accompanied by collisions between elements. Loss of
kinematic links within machines and collisions between the elements of kinematic
pairs and structures cause the establishment of intensive vibro-impact processes,
as the result of wh ich the tuning of elastic systems changes significantly, as do
their natural frequency distributions and the character of transient processes, and a
variety of nonlinear effects can arise. Thus the study of problems in dynamics in
terms of accuracy, stability and vibration activity, requires vibro-impact processes
to be considered in systems with various structural forms, having both distributed
and lumped masses, nonlinearities and being exposed to the effects of a variety of
regular and random forces.
In addition, vibro-impact regimes form the basic working process for a broad
cIass of machines, drives and installations with the widest functional purposes. To
this cIass belong machines for driving and extracting piles, tamping and
cultivating ground, crushing and grinding materials, compacting concrete and
foundry mixtures, vibro-impact test devices, various examples of drill, percussion
devices, riveters, nut-tightening machines, etc. For this cIass of system, the
characteristic periodic motion arises either due to forced vibration or self-
X Preface to Russian edition
sustaining vibration.
The range of constructional variatIons for vibro-impact mechanisms and
systems requires the development of methods not only to solve the problems of
analysis and tuning, but also the development of methods to synthesise dynamic
structures which best fulfil the optimum form of motion.
Nowadays, the basic techniques for the theoretical study of vibro-impact
processes are the exact methods of nonlinear mechanics, based on stitching
solutions describing adjacent intervals of motion, which are separated by the
instants of impact. These methods allowed a detailed study of the complicated
dynamics picture for the motion of aseries of systems and revealed a number of
fundamental properties inherent to them. Results obtained in this way are reflected
to a considerable extent in the monograph by A E and A A Kobrinsky [IISf,
Searching for solutions by exact methods is a very tedious procedure, which
limits the scope of their application. It is particularly difficult to use them as the
system dimensions increase in number, when it becomes necessary to take into
account additional nonlinear factors and the complieated eharaeter of exeitation
from the effects of non-periodic and random forees.
In addition, to solve a large eircle of problems in the theory or vibro-impact
systems, and generally in the theory of non linear vibrations, it is unneeessary to
define motion in a detailed manner. Foremost in this eontext, are problems
eonneeted with the tuning of sueh systems: finding their resonance eonditions, the
number and type of resonanee bands, studying questions of stability and self-
exeitation ete. All these effeets already appear when the basie harmonie
eomponents of vibration are analysed, sinee one of the effeetive methods of
studying nonlinear systems is to separate from the general mathematieal
deseription the simpler relationships which eharacterise most ofthe corresponding
patterns of vibration. These ideas, which come from the general proeedure for the
asymptotic presentation of solutions as power series of small parameters, have
shown themselves to be particularly effective when transformed to spectral
representation and the ideas of equivalent linearisation.
The first reported eonnection was established by N M Krylov and N N
Bogolyubov, who provided the first approximation for the asymptotie method they
developed in the form of equivalent (harmonie) linearisation. Later the harmonie
linearisation method was widely developed and applied in the works of V A
Kotelnikov, L S Goldfarb, W Oppelt, R Kochenburger, J D Loeb, E P Popov, K
Magnus, L A Zadeh, and many other investigators, prineipally being applied to
problems in the theory of automatie contro!. The eontemporary state of this
question and a bibliography appear in aseries of eolleetive monographs [162, 163,
173] published under the general editorship ofE P Popov.
In the 1950s, P Booton [66] and I E Kazakov [112] independently presented an
effeetive form of equivalent linearisation for the study of non linear stoehastie
proeesses, which is known as the statistieal linearisation method. Teehniques for
applying the method were developed in the books by I E Kazakov and B G
Dostupov [113] and A A Pervozvansky [177]. Finally we must mention the
suggestions of M Z Kolovsky and A A Pervozvansky for the original form of
• See also Chapters 7 to 9 in (117) written by R E Brunstein and A E Kobrinsky,

Preface to Russian edition XI
equivalent linearisation by distributionfunction [128, 130] which made it possible

to study multiharmonic processes in nonlinear systems. This method and its
application to problems in the theory of nonlinear vibration protection is described
in the monograph by M Z Kolovsky [111].
Systematic application of the range of methods described for equivalent
linearisation and the solution of problems in the theory of vibro-impact systems
forms the basis of the material in this book. In this way it was possible to consider
the basic forms of vibration motion for vibro-impact systems of arbitrary
dimensions in a unified manner using simple methods and to develop a
sufficiently general theory for tuning a wide range of systems, such as impact
vibration absorbers, pneumatic impact machines, devices for ultrasonic machining
of materials, etc., and to reveal new nonlinear effects.
Sometimes in the analysis of a range of problems in the theory of vibro-impact
systems, information yielded by equivalent linearisation methods is insufficient.
This happens when it is necessary to obtain an idea of the law of motion or the
trajectory of the system elements, for example, when estimating collision
velocities, studying the spectral characteristics of motion, defining existence
domains of dynamic models in configurational space, etc. For all of these,
methods are suggested, based on full approximations of the laws of motion for
vibro-impact systems with the help of so-called impulse-frequency characteristics,
widely studied in the works of E N Rosenvasser [191, 192,]. Following the
traditions of mechanics these characteristics will be known as periodic Green 's
functions in this book
Together with these, procedures for exact methods are presented and applied,
particularly in those cases when they can help the understanding of the nature and
accuracy of approximate methods, but also for the better understanding of the
effects studied. All the methods described and applied in this book do not assume
that the reader has any previous knowledge ofthem.
The book consists of five chapters. The first chapter presents a short summary
of contemporary ideas on the theory of collisions between solids and gives
examples of simple vibration systems showing exact methods of analysis for the
basic nonlinear behaviour of these systems in the presence of collisions. The
character of resonance phenomena in vibro-impact systems and dynamics of the
basic structure of vibro-impact machines are considered.
The second chapter introduces equations describing the process of vibro-
impact interaction in general types of system undergoing arbitrary excitation and
deals with the analysis of frequency characteristics for vibro-impact systems and
the processes they cause. The concept of equivalent linearisation of vibro-impact
systems is explained and harmonic and statistical linearisation coefficients for
basic types of link between colliding elements are calculated.
The third chapter gives a systematic presentation of vibro-impact processes in
systems with one degree of freedom based on approximate methods. Dynamics,
stability, transient processes of periodic force vibration, parametric vibration and
self-sustaining vibration are considered. Systems with random excitation are also
studied. Special attention is given to sub-periodic regimes, and effective
techniques for calculating these are developed. Throughout the explanations,
descriptions of corresponding calculation methods are given and amplitude-
XII Preface to Russian edition
frequency characteristics are revealed and analysed, and a comparison of

approximate and exact solutions is developed. The influence of the characteristics
ofthe source of excitation on vibro-impact processes is estimated.
The fourth chapter develops analytical methods extended to systems with
arbitrary degrees of freedom and is applied to the study of vibro-impact processes
occurring during axial vibration of rods and bending vibration of beams for the
analysis of systems with impact dynamic absorbers.
Finally, the fifth chapter presents problems in the structural synthesis of self-
sustaining vibro-impact systems. Exact solutions are presented as aseries of
problems based on optimal control methods, and approximate methods to find the
structure of controllers for general types of system are given using harmonic
linearisation. These methods are applied to the solution of problems of excitation
and stabilisation of autoresonant regimes in vibro-impact systems.
Trying to reflect the qualitative features of vibro-impact systems as objectively
as possible, we have tried everywhere to develop the results up to the final
analytical expressions, giving clear physical interpretations. In addition,
experimental data is presented to confirm the basic effects.
The book is designed for engineers and researchers immersed in the study and
design of systems involving mechanical vibrations, and also for lecturers and
students interested in non linear vibrations. To study the book a reasonable
knowledge of engineering based general mathematics will be adequate, covering
elements of Fourier and Laplace transform theory and probability theory, together
with a knowledge of basic theory of mechanical vibrations and stability of motion,
as presented in the book by I M Babakov [24 r.
All additional mathematical
information is presented in the text during the explanation with references to the
source where a detailed explanation is given.
The concluding bibliography lists only works either mentioned or used in the
text, together with aseries of related works of a general character.
Those wishing to use approximate methods immediately can begin their study
with Chapter 2. To a considerable extent, Chapter 5 can be read as stand-alone. A
self-contained concentrated explanation is also give for the statistical aspects of
vibro-impact theory, to be found in §8, § 13 and partially in § 14 section 6 and § 17
section 4. Before beginning to study these topics it is better to read through §6
first.
A E Kobrinsky first raised the author's interest in vibro-impact systems. The
original ideas for the approximate methods developed in this book started as a
series of papers completed in conjunction with M Z Kolovsky; whose advice and
constant support was given freely to the author both during the planning of this
book and during the writing process. V 0 Kononenko and Y G Panovko endorsed
the plan ofthe book, and P F Nagaev made kind comments when the book was in
manuscript form. V K Astashev helped greatly and § I section 7, §2, §4, § 15 were
written with his help. The deep interest in the methods developed, shown by my
colleagues in the Institute of Machine Studies such as T S Akinfiev, B A
"Translator's Note. For texts in English covering similar material the reader should consult W T
Thomson's "Theory of vibrations with applications", 4th edition, (Chapman and Hall) or S S Rao's
"Mechanical vibrations", 3'd edition (Addison-Wesley).
Preface to Russian edition XIII
Borovkov, M EHertz. M Y Israilovich, V L Krupenin and A N Tresviatsky, who

were concemed in the development of aseries of problems and in discussion of
the results, was very stimulating for the author. The author sincerely thanks all the
people named.
V Babitsky Moscow
Contents
Preface to English edition ...................................................................................... vii

Preface to Russian edition ...................................................................................... ix
Contents .............................................................................................................. xv
Chapter 1 Vibro-impact systems
§ 1. Collision of so lids ................................................................................... 1

§2. Oscillator with fixed stop ..................................................................... 20
§3. Oscillator with symmetrical stops ......................................................... 37
§4. Principles ofpercussion machine dynamics ......................................... 53
§5. Stitching method ................................................................................... 69
Chapter 2 Equivalent linearisation of vibro-impact systems
§6. Equations ofvibro-impact systems ....................................................... 75

§7. Harmonic linearisation ofvibro-impact systems .................................. 94
§8. Statisticallinearisation ofvibro-impact systems ................................. 108
Chapter 3 Vibro-impact processes in systems with single degree of

freedom
§9. Forced vibration ................................................................................. 125

§ 10. Stability and transients offorced vibration ......................................... 140
§ 11. Parametric vibration ........................................................................... 148
§ 12. Vibration of self-sustaining systems ................................................... 165
§ 13. Random vibration ............................................................................... 185
Chapter 4 Vibro-impact processes in systems with multiple degrees

of freedom
§ 14. Vibro-impact interactions of elastic systems ...................................... 193
§ 15. Vibro-impact interactions ofvisco-elastic bars .................................. 212
§ 16. Vibro-impact regimes in vibrating beams ........................................... 229
§ 17. Theory of impact dynamic absorber ................................................... 243
XVI Contents
Chapter 5 Structural synthesis of self·sustaining vibro·impact

systems
§ 18. Optimal periodic motions ofvibro-impact systems ............................ 251

§ 19 Synthesis of self-sustaining systems ................................................... 260
§20. Synthesis of autoresonant systems ...................................................... 271
Bibliography ........................................................................................................ 279

Additional Bibliography ...................................................................................... 287
Index ............................................................................................................ 292
Appendix I .......................................................................................................... A1-1

Appendix II ...................................................................................................... AII-l
Chapter 1 Vibro·impact systems
Awaiting (he sensation 0/ a short. sharp knock.

From a cheap and chippy chopper on a big black block.
W S Gilbert: Bab Ballads
§ 1. Collision of solids
1. We shall examine some aspects of impact theory for so lids [60, 80, 103,
115], limiting the analysis to those points necessary to understand the processes
occurring in vibro-impact systems, and generate mathematical models for them.
We shall mainly be interested in the force changes and the kinematic
characteristics of the motion of solids and their elements: particles, plane sections,
etc.
Impact is characterised by the appearance of strain in the contact zone due to
the transfer of kinetic energy arising from the relative motion of the contacting
elements into deformation energy, and the spread ofwaves of strain throughout the
reacting bodies from these zones. As a result, the energy spreads through the
volume of the bodies, causing vibration of the elements and gradually dissipating
throughout the material.
Consider initially the longitudinal impact of a short rod (Fig. 1.1, a) of length
10 , moving with velocity v o ' against a long rod of length 1 of identical cross-
section S, fixed at the right hand end and initially at rest. At the moment of impact
between the rods, interaction forces of equal magnitude, F, are developed, acting
internally in the body of the rods. These forces deform a portion of the rod of
length dl for time dt. The value
Y= dll dt (1.1)
defines the velocity at which the disturbance in the rods spreads.
Let us now examine in greater detail how these waves of strain travel through
one ofthe rods. In time dt its deformed element is compressed by an amount
du = f.dl , where f. is the relative strain. As a result, as the wave reaches sections
of the rod, they move with velocity
v=duldt=f.dlldt=f.Y, (1.2)
or, assuming that the deformation is elastic,
v=cry/E, (1.3)
where a = FIS is the stress in the disturbed rod and E is the elastic modulus.
Denoting the mass of the deformed element as dm, we can write an expression,
2 Chapter 1 Vibro-impact systems
using the principle of impulse and momentum, for the vibrating motion of this
element as vdm = vpSdl = Fdt or
v=a l(py) (1.4)
where p is the density of the rod material. Eliminating v from equations (1.3),
(1.4) we obtain
y=JE/p (I.S)
For steel, E:::d .96.101IN l m2;p"'7.8S.103kg l m3, and as a result,
y",S.10 3mls. Using these values in equation (1.2), we can now calculate the
velocity of displacement, v', at which plastic deformation will occur for an
element ofthe material, Assuming that at the elastic limit for steel c' '" 0.001 , we
find v' '" Sm I s .
The working velocity range of tools should not exceed this critical value. By
analogy, this imposes a limit on the collision velocity that can be allowed to occur
between machine parts with clearances or in linkages. However, it must be
mentioned that in systems where vibro-impact occurs, plastic deformation can
occur even at low impact velocities as a
result of an accumulation of deformation
under regular collisions. Some special tools
permit impact velocities of the order of 10
a) mls.
v~-
In the example studied, the motion of
u.a- 1)-0 the material elements occurred in the same
z direction as the motion of the wave, hence
/J -D UDIZ - u- D
J such waves are called longitudinal or
v-D uoll - expansion waves, since they occur as a
/J-D result of the resistance of the material to
5 lilj" j 'll
IJ - tl - uul2 change in volume. In cases when the
-vol2 v-g transverse dimensions of the waveguides
11 111111111111111111111111111111111111
_ uo /J-D are equal to or greater than the length of the
8 longitudinal waves, their velocity increases
-1IIJ1t /J -D
~ because of the complementary material
u- D - uolZ resistance, expanding into the transverse
fD
/J-D direction and becomes equal to
If '"" clil;.'>. .
/J-D IJO,Z --' E[(1-v)/(l + v)(1-2v)pf2, where v is
12 -
Vo12~ v-D Poisson's ratio.
fJ -
vo_ u- D Another kind of waves, in which
f4
Vol?- uoo_ portions of the material move
15 perpendicularly to the direction of the
h)
Fig. 1.1 spreading waves, occurs due to the
resistance of so lids to changes of shape.
These transverse or shear waves have a
velocity of (G/p)I /2 , where G is the shear modulus. The values of E, G and v are
§ 1 Collision of solids 3
related through the expression G = E /[2(1 + v)].

In addition, on the surfaces of solids, waves of a different type can occur,
whose importance was first recognised by Rayleigh. These produce trajectories in
the element that are almost circular, and lie on planes normal to the surface and
coinciding with the direction of the longitudinal waves. The amplitude of the
surface waves decreases exponentially with depth, while their velocity is about
90% ofthat ofthe shear waves [l31]. The types ofwaves mentioned determine the
general picture of the propagation of disturbances in solids, which can be obtained
from the theory of elasticity [201].
2. Let us continue to examine the one-dimensional case (Fig. 1.1 a). We
shall take the x-axis along the central axis of the longer rod, with the origin at the
interaction section at the instant of the start of impact, when 1 = O. We shall write
the absolute displacement of an arbitrary section as a function u(x ,I), and the axial
force occurring in it will be called F(x ,I) .
According to Hooke's law
F(x,t) = ESE = ES au~:,t) (1.6)
The change in axial force over an element dx is as folIows:
F(x + dx, t) - F(x, t) = ES

a
2 u(x t)
; dx (1.7)
ax
This force difference produces an acceleration a2 u(x,t)/ al 2 transferred to the
element ofmass pSdx. From Newton's second law, we can write
(1.8)
Dividing both sides of (1.8) by Sdx and using (1.5), we obtain the differential
equation for the longitudinal vibration of a uniform rod as
a2U(X,/) 2 a2u(x,t)
at2 =y ax 2 . (1.9)
Direct substitution shows that equation (1.9) is satisfied by a general solution

ofthe form
u(x,t) = cp(yt - x) + ljI(yt + x) (1.1 0)
where cp and ljI are arbitrary double differentiable functions. These functions
represent waves of deformation moving to right and to left within the rods at
velocity y.
lmmediately after impact the compression wave in the short rod moves to the
left while in the longer rod it moves to the right away from the contact zone. The
waveform is defined by the character of the force of interaction between the
contacting surfaces ofthe rods. As long as these forces are equal in magnitude they
produce equal waves, hence the wave propagating along the longer rod takes the
form cp( yt - x), while that in the shorter rod is represented by -cp( yt - x). The
first wave involves motion into the elements ofthe rod under investigation
u:(x,t) = <p(yt-x) (0"'; x <I) (l.lI)

while conversely, the second causes deceleration in the corresponding section of
the short rod given by the relationship
u:I(x,t) = vot-<p(yt+x) (-1 0 <x<O) (1.12)
Here the superscript shows the rod number while the suffix shows the time interval
examined.
The relationship between the shape of the generated wave <pe yt - x) and the
impact force F(t) can be found from the following relationship for the longitudinal
force at the zero section ofthe longer rod:
- F(t) = ES au I (x, t)/ax 1,=0 (1.13)
where the negative sign indicates a compressive force. Substituting (1.11) into
(1.13), we obtain
<p'(yt) = F(t)/(ES) (1.14)
where the dash indicates differentiation with respect to the argument. Writing the
argument as a and integrating (1.14) using the value <p(0) = 0, since the section at
x = 0 was static at time t = 0 , we obtain
= _I fF(a \la
<p(a)
ES 0 Y r
For the example we have been examining (Fig. 1.1, a) let us define the
(1.15)
velocity, v, of the displaced section taking part in the wave process. From the law
of conservation of momentum, we can put
vodm = 2vdm , (1.16)
from which
v=v o /2 (1.17)
Comparing (1.17) with (1.3) we find
F = ESv o /(2y) (I.l8)
Substituting (1.18) into (1.5), we obtain after integration for a = yt - x:
Vo x
<p(yt-x)=-(t--). (1.19)
2 y
Using (1.l9), we find from (1.11) and (1.12)
(1.20)
(-l() <x<O) (1.21 )
Suppose we have impact of a rigid body of mass Mo at velocity v() , onto a long
rod. Assuming that after impact the rigid body moves together with the section of
the rod x = 0, we have, according to (1.13)
2 l
_ F(t) = -M a u (X,t)1 = ES a2ul (X,t)1
o at 2 (]x
x-=ü x=O
or, introducing a term covering the relation of the mass of the rod and a striker
f-l = pSI/ Mo and using (1.5), we obtain
a 2 u l (x,t) _ f-lL aul (X,t)] = 0 .

(1.22)
[
at 2 I ax x=o
Substituting (1.11) into (1.22) we obtain the following equation for the shape
ofthe wave after impact
<p"(a) + l::<p'(a) = O.
/
lntegrating this, we find
<p'(a) = C exp( - f-la/I) ,
where C is an arbitrary constant. Making use of the initial condition
aul(O,t)/ato
= v o , using (l.lI) we obtain
<p'(O) = vo/Y =C
Hence,
<p'(a) = V; exp( -Ta). (1.23)
lntegrating (1.23) with the initial condition rp( 0) = 0, we find
<p(a) = ~:[I-exP(-Ta)l
When a = yt - x we finally obtain
(05, x < I) (1.24)
At the ends of the rods the waves are reflected. For the fixed end (x = I)
according to (1.10) we find
u(/,t) = <p(yt -/)+ \jJ(yt + I) == 0,
from which
\jJ(y (+/)= -<p(y t-/). (1.25)
The resuIt is that the reflection at the fixed end causes the appearance of an
additional wave, similar to the incident wave and spreading to meet them with
opposite phase. Using (1.25), equation (l.l0) takes the form
u~ (X,l) = <p(yl - x) - <p(yt + x - 2/) . (1.26)
(The suffix 2 shows the time interval after reflection of the waves.) In the same
way, it is possible to show that reflection from the free end is characterised by
absence ofaxial force: ESau / ax = 0, acting in phase, since for the short rod the
following will apply after reflection:
U~I (x,t) = Vol - <p(Yl + x)- <p(yt - x - 2/ 0 ), (1.27)
Bearing in mind (1.19), from (1.27) we find
u 2Il()
x,t =v ot- vor t+
2 yxJ -2 - -y- J=-y-=const
vor t -X-21 0vol o
Hence, as the initial compression wave reaches the left end of the short rod, it
is retlected as an expansion wave moving towards the right which relieves the rod
sections from their deformed conditions.
The wave behaviour we have been examining can be visualised graphically. Fig
1.1, b shows the spread of waves through the rod at successive moments in time.
Transverse shading shows compressed sections, axial shading - expansion. Areas
without shading show that elements are relatively undeformed.
Once the compression waves have passed through the contact zone (position 3,
Fig. 1.1, b), the striker has been completely relieved of stress and all its energy has
been transferred to wave energy moving along the longer rod. However, the rods
remain in contact until the wave from the fixed end returns as a retlection.
The duration of impact, '(, is taken to represent the time during which the
strained regime in the contact zone is maintained. In this case, '( = 21 0 1y . If the
short rod is removed at the end of impact, then wave (1.20) or (1.24) is retlected as
it reaches the fixed end and according to (1.26) will circulate along the long rod,
causing vibration of its elements with period 411 y (Fig. 1.1, b).
The periodicity that appears in our model fails to ac count for the real
imperfections of elastic conditions, which become apparent when energy is
absorbed by internal damping of the material. This idea of internal damping is
common to a range of mechanisms that occur in the deformation of materials that
are cyclically strained; in particular, internal damping is the cause of hysterisis
loops that occur in the stress1strain curve. Creep and relaxation intluence energy
dissipation and make it impossible to obtain equations for force distribution as
functions of current deformations. Using integral-differential equations to describe
the evolution of material properties makes the study of wave processes more
difficult.
At the same time, numerous tests have shown that under a periodic deformation
the area of hysterisis loops for most materials depend on the amplitude of the
deformation for a wide range of change and not, in practice, on the frequency. For
this reason, in the study of periodic processes, which form the basis of this book,
we will usually introduce an equivalent force proportional to the velocity of
deformation, selecting a suitable factor so that the general dissipative effect
corresponds to the experimental data [176]. As a result, equation (1.9) for the
study of periodic vibration of the rod will be written in the form
a2 u(x,t) _ 2 a2 u(x,t) b aJu(x,t) = 0 (1.28)
at 2 y ax 2 ax 2at '
where b = y2%/(2TCm) is the linearised coefficient ofthe force of internal damping;
0) the frequency; % the absorbtion coefficient found from tests as the ratio of
absorbtion of the energy during the cycle (proportional to the area of the hysterisis
loop) to the basic deformation energy. For steel, %",0.015.
Because of internal damping, the periodic process can only be established
when there is a regular addition of energy into the system. In the example
considered, this depends on periodic motion of the short rod caused by an extemal
source to induce systematic impacts.
Under periodic excitation due to the result of superposing direct and reflected
waves to the system, stationary fields of waves can develop (standing waves),
whose intensity can greatly exceed the amplitude of the generating waves. As an
example, for a contact force F(t) = aj.$in rot we find from (1.15)
<p ()
a =-
aF
f . ooa a Y ooa
sm-da = - - - c o s - ,
Cl F
ES 0 Y wES Y
Hence a direct wave along the long rod is generated, given by
<p(yt-x)=- aj.y cosoolt-!..)=- aFy cosloot- xoo )

wES y wES y
We define the amplitude of the standing waves, forrned as the result of repeated
reflection of the gene rating waves from the fixed and free ends, as a. According to
(1.26), by superposing the direct and reflected waves, to satisty the conditions for
the fixed right hand end
u' (x, t) = acos(ffit - xoo/ y) - acos[ffit + (x - 2!)w/ y]
or, using the well-known trigonometric forrnula
· a+ß . ß-a
cosa - cos ß = 2sm - - sm - - ,
2 2
we obtain the equation ofthe standing wave
u'(x,t)= 2asin~ (I -x~inoolt-~). (1.29)
Introducing phase \jf into the exciting force F(t),

F(t) = aj.sin(oot-\jf), (1.30)
l
and imposing the free end conditions (1.13) onto (1.29) and (1.30), we obtain
00 Scos-sm
2a-E oo!. oot--
OO!) = . (oot-\jf ) ,
aFsm
! Y Y
whence
aFy
a=-----C...-,--,
200EScos(00l! y) .
As a result, at values of
00= ny (2m-I) (m = 1,2,... )
21
the amplitude of the standing waves increases infinitely. The values of 00
correspond to the natural frequencies of vibration of the rod. Stability of the
standing waves during resonance is due to the forces of internal damping.
This standing wave phenomenon that occurs during vibro-impact interaction of
rods will be examined in § 15.
3. So far, we have assumed that during impact the contacting elements start
to move' instantaneously at a velocity deterrnined by the law of conservation of
momentum and move together before the impact is finished. Such an interpretation
for the theory ofaxial impact of rods was developed by Navier, Poisson,
Boussinesque, St Venant and Flamant [103]. This theory gives a convenient
approach to study the spread of disturbances produced by impact.
Coupled with this disturbance, as we have shown above, contact forces occur,
and their idealised representation can lead to the essential distinction in the
explanation ofthe behaviour ofthe strained elements involved in the wave process.
In particular, in the examples considered, the effects we neglected in the contact
zone cause the appearance of discontinuities in deformation and strain in front of
the wave due to a sudden increase in the velocity at t = O. For example, according
to (1.24)
y
These results stimulated research on the development of models of contact
interaction of bodies during impact and the matching of them with the equation of
wave propagation.
Contact interaction is defined as a combination of effects that occur in, and
near, the contact zone and are related in a complicated manner to the spread of the
disturbance through the actual geometry of contact elements in the coIIiding bodies
and their mechanical characteristics. The curvature ofthe contacting surfaces, their
micro-geometry, the presence of strengthening layers, any special processing
caIIed for, etc., aII have important influences. When the interacting surfaces are
smaII in size and complicated in configuration changes of the strained condition in
the neighbourhood of the contact zone occur more rapidly than in the main part of
the coIIiding bodies. These circumstances led Sears to advance the hypothesis of
quasi-static stresses in this region at impact and to describe contact effects by using
a static solution [197]. Sears examined the problem of
axial impact of rods with circular ends, using the
Hertz equations to determine the contact
deformations. Earlier, Hertz himself used the same
technique to study impact of short bodies, in which
the natural periods are considerably shorter than the
contact time, and for this study ignored basic wave
processes (see section 5).
We shaII develop the simple form of the Hertzian
formula, as produced by I. Y. Staerman [202]. Let us
assume that under compression by a force F (Fig. Fig. 1.2
1.2), two elastic bodies deform so that the surface
points which Iie on a common normal to the plane constructed through the point of
initial contact, are displaced along this plane. The relationship between the
increase of force dF and the change ofaxial displacement U o due to deformation of
value duo takes the form
E
dF = c - - 2 8du o ' (1.31)
1- v
where C is a constant of proportionality, e is the length of the perimeter of the

r
compressed surface, and (1- v 2 l is a factor accounting for the conditions of
three-dimensional compression. Assuming that the cross-section of the
contacting bodies can be represented by a second degree curve, and putting
(Fig. 1.2) ul = y2 /(2rl ), u2 = y2 /(2r2 ), it follows that
U o = ul + u2 =(~+~)L.
r r 2
l 2
(1.32)
For a sphere e = 2ny , hence integrating (1.31) using (1.32) we find

F = 2n C~(~ + ~)yJ
3 1- v rl r2
. (1.33)
Substituting in (1.33) for y in terms of U using (1.32), we finally obtain
F = 4fin
3
C~(~ + ~)-1/2 U~/2
1- v rl r2
. (1.34)
The coefficient c is determined experimentally. We note that when

equation (1.34) takes the Hertzian form. Writing
CI = 4fin c~(~+~)-1/2 (1.35)

3 1- v rl r2
we obtain
(1.36)
Hence, using this assumption, the contact interaction of colliding solids can be
represented by introducing an elastic "buffer" zone between them, as represented
by (1.36) - an expression that shows the smoothing effect of the contact force
increases during impact. According to (1.35) the slope of the impact wave front
depends significantly on the curvature of the contacting surfaces and increases as
the surfaces become less curved, reaching the limiting situation examined earlier
for impact of rods with flat ends. These qualitative ideas agree weil with
experimental data [5].
According to I Y Staerman [202] as the bodies approach more closely to the
limiting case of flat surfaces, equation (1.36) can be transformed into the
following:
2m+l
(1.37)
(m = I, 2, ... ), which increases due to increase in the adjoining value of m. Since
I :s; (2m + 1) / 2m :s; 3/2, then as the contact surface becomes flatter, equation
(1.37) becomes more linear. In vibro-impact systems, this often becomes the
situation because of additional wear of the surfaces under repeated impact.
Coefficient c2 has the order of the value of CI.
As a result, the buffer zone mentioned earlier can be considered as a thin
boundary layer that may be neglected when studying wave effects during impact
between bodies. The static elastic behaviour of the layer, represented by a
monotonically increasing function, can be taken as linear over its working range.
As mentioned in [115], the strengthening effect of the surface layer influences its
elastic properties to a great extent. This can cause additional effects, for example
generating pure transverse waves on surfaces (Love's waves), vibration of shell
surfaces on elastic foundations etc. For our purposes, however, it is important that
these effects occur over a faster time scale, both because of the small dimensions
of the contact zone and because of the increase in elastic properties, and the effect
that they have on the spread of excitations can be calculated using a generalised
model of static elastic behaviour of the form
F=tI'(u o ), (1.38)
which we will call the static force characteristic of contact interaction. Because of
the strength of the filtering effects of vibro-impact systems (see §6) the nature of
the vibration process shows little sensitivity to the basic static character of the
contact interaction, so that various approximations, whose accuracy depends on the
system frequency, can be used when creating dynamic models of vibro-impact
systems. Of course, contact phenomena require more careful consideration in the
analysis of strength due to impact [219, 108, 115, 187].
Note that force characteristics can be used to describe contact interaction of
bodies during impact and in cases when plastic deformation begins. T Pöschl
proposed the following relationship [182]:
F=CIU~!2 ±bl(duo/dtr,
where the positive sign corresponds to compression up to the maximum and the
negative sign to unloading. Other relationships derived as approximations of
experimental data, are given in [49, 217]. Obviously, when calculating the plastic
deformation, material behaviour must take account of hysterisis.
Accordingly, a more general form of the force characteristic of contact
interaction can be written as
(1.39)
In contrast to (1.38), equation (1.39) will be called the dynamic characteristic of
conlacl interaction.
4. A convenient way to describe impact taking into account contact and
wave behaviour using integral equations, was suggested by S. P. Timoshenko when
studying the bending impact of a sphere on a beam [208]. Such an approach has a
universal character and can be used to study similar processes in systems of
differing structure. We shall show the application of this method to the example
problem ofaxial collision of rods with rounded ends (Fig. 1.3), one of which is
stationary while the other moves with an initial velocity va' As before, we will
assume that the impact interaction starts at
section x = 0 at t = 0 .
Vo v =O We define the absolute displacement ofthe
+'=A 1[&i:ss~. . .=====t~~
T2 / Tl
end elements of the relative rods, neglecting
any compression on contact, as uIO(t) and
u20 (t) and the value of the contact
Fig.1.3
deformation on impact between the bodies as
ulO(t) + u 20 (i) + 1/,, (1) = V,'" (1.40)

We shall write the displacement of the end elements under the influence of the
contact force F(t) through their reactions to a single impulse
I I
u\O(t) = fF(t)hl(t-t)dr, u 20 (t) = f F (t)h 2 (t-t)dt, (lAI)
where hl(l) and hP) are the impuLse response functions of the rods, occurring as
reactions at the end elements on the Dirac delta-function excitation. Impulse
response functions are non-zero only for 1 > O. Substituting (1041) into (1040) and
taking into account that
(1042)
obtained using the static characteristic of contact interaction (1.38), we obtain
eventually
I
fh{t - t )F(t )dt + X[F(t)] = VI'! . (1043)
Here h(t) = h l (t) + h2 (t). Equation (1.43) defines the law of changing contact
force. Once it has been solved, the wave process in each of the rods can be
investigated separately by solving the boundary problem with the given force F(l)
acting on the ends.
Note that we can obtain integral equations in the form of (1043) for axial
collision in systems of arbitrary configuration by using the proper impulse
response function in (1041). Hence, the framework of the present example gives us
a general method of describing impact processes using nonlinear non-
homogeneous integral equations. The lack 01' standard methods of finding solutions
led to the development of numeri ca I methods to find particular solutions, by
substituting discrete summation in place of integration and the solving the resulting
difference equations iteratively [178] .
The right hand side of equation (1.43) defines the influence of external effects
on the system. Denoting this in the general ca se as &(1) and using (1.38) and
(1.42), we obtain the equation of contact deformation from (1.43) as
I
fh(I-·r)\fI[U o(l)]dl + lIJt) = &(1). ( 1.44)
We will use integral equations like (1.44) later to investigate vibro-impact systems.
We shall apply approximate analytical methods to find
solutions.
5. Examining effects in the contact zone
together with the way the waves spread throughollt the
structure of interacting badies gives an adeqllate
physical picture of impact processes. It reveals the
mechanism of such processes, including the way that
forces change together with the kinematic factors ,
elastic and dissipative effects of impact. To develap these objectives for the theory
of impact requires a study of waves produced in impact of bodies of nonlinear and
elastic-plastic materials, or materials with post-effects and relaxation etc. [187].

Studying such examples is rather time-consuming and can be made more
straightforward only for simple configurations of colliding bodies. For this reason,
there has been a wide spread examination of various methods to simplify the study
of impact [60,115]. For axial impact (direct impact) of short bodies with
reasonably compliant contact zones, for example, small diameter elastic balls (Fig.
1.4), impact acts for a considerably longer time than that required for the repeated
passage of the waves through the interacting bodies. In such cases, it is usual to
neglect the finite time for the spread of the excitation and to apply the Hertzian
theory outlined earlier, assuming that loading follows the static loading law.
Defining the masses of the impacting bodies as MI and M 2 , the axial co-
ordinate of the centre of mass as x land x 2 and neglecting changes in the position of
the centre of mass due to contact deformation, for central impact (the line of action
ofthe principal vector ofthe contact force passes through the centres ofmasses)
(l.45)
where x =x2 - XI; q.t(x) is the static characteristic of contact interaction. The dot
above the function indicates differentiation with respect to time. Subtracting the
second of these equations from the first, we obtain
Mi + q.t{x) =0. (1.46)
Here M = M I M 2 /(M I + M 2 ) represents the reduced mass of the colliding
bodies. Taking into account that at the moment of contact, t = 0, the relative co-
ordinate x = 0 , while velocity is given by i =v o, we obtain the first integral of
equation (1.46) in the form
x2 -v~ = -~fq.t(x)crr. (1.47)

Mo
The maximum deformation x' corresponds to x = 0 , i.e.
2 2 x'
Vo =- fq.t(x)crr. (1.48)
Mo
Defining the impact duration from (1.47) as T, we find
-112
x· 2 x·
T = 2 f [
v~ -- fq.t{x)crr ]
crr. (1.49)
o m 0
Assuming that deformation occurs according to the Hertz equation (1.36), we

find from (1.48)
x' = (2. Mv~ )2/5 (1.50)

4 CI '
and as a consequence, the maximum force

F* = C 2/ 5
I
(2.4 Mv 2 )3/5
0
(1.51)
Integrating (1.49) gives [115]

§ 1 Collision of so1ids 13
'( = 2.9432~=2.9432 (5M J

2/5
V~1/5; (l.52)
Vo 4c 1
for example, for collision of two identical spheres with radii r l = r2 = 0.01 m,
moving with relative velocity Vo = 0.1 m / s , putting c = (n,J8 r l in (1.35), then
from equations (l.50) to (l.52), when v = 0.28, we find x* = 3.3 X 10-6 m ;
F* = 61.4N; '( = 9.8 x 10-5 s.
According to (l.52), the duration of contact depends weakly on the velocity of
impact. For a linear static characteristic, it will of course be constant. Estimating
the time of impact far real short bodies to be 10-3 to 10- 5 s, we find that for
Hertzian theory the characteristic natural period of bodies must be shorter than 10- 5
to 10- 7 s.
6. The relatively brief duration of impact for short bodies allows us to
exclude the impact process from consideration when defining the resuIting
kinematic characteristics of motion and we obtain the required relationship from
basic theory of mechanics for rigid bodies. The impact can be evaluated from its
impulse
t
J = fF(t)dt
o
a function that is assumed instantaneous. All additional slowly changing forces
have a vanishingly small impulse during impact, whose value on the effects of
impact can be ignored. The resulting relationships for pre- and post-impact
velocities of motion of the impacting bodies can be obtained by examining
autonomous systems.
The theory of impact, based on assumptions by Huygens and Newton, has been
called stereomechanics [103]. In stereomechanics, using integral estimates of
the force factors occurring during impact does not allow the study of the strength
aspects of impact, hence the field of application is limited to the study of the
motion of bodies under the action of impulses. Such motion must occur very
slowly by comparison to impact processes. The natural period of such motion
should not exceed 10-2 to 10-3 s, according to previous estimates for the duration of
impact. Under these conditions motion can be separated into pre- and post-impact,
coupled with kinematic conditions of impact. The substance ofthe last approach is
that the co-ordinates of the centres of mass remain unchanged but their velocity
undergoes sudden instantaneous change, as defined by the laws of conservation of
impulse and energy.
We shall examine several typical examples. Under straight central impact of
two bodies moving with velocities Xl and X2 (Fig. 1.4), indicating pre- and post-
impact velocities of each body with minus and plus signs and assuming x1_ > x2 _,
we can write the law of conservation of momentum as
M1x l _ + M 2X2- = M1x l + + M 2X2+ . (1.53)
To determine the unknown post-impact velocities, an additional equation is
required. This can be obtained from the law of conservation of energy, defining the
proportion of kinetic energy remaining in the post-impact motion to the energy
dissipated during impact. Newton proposed a convenient way of showing this
proportion as a coejjicient 0/ restitution for the relative velocities of the colliding
bodies. Calling this R, we find
(1.54)
in which 0 ::; R ::; l.

Stereomechanic theory describes all forms of direct impact between bodies
uniformly in the range between absolutely elastic (R = 1) and completely plastic (R
= 0). A reduction in the coefficient of restitution is related to an increase in the
proportion of energy lost irreversibly during impact. The energy of the waves
remaining in the bodies after the impact has ended come in this latter proportion so
that, even during collision of completely elastic bodies, the value of R can differ
from unity.
In agreement with the basic ideas presented in this section, the true values of
the coefficient of restitution must be defined by the entire combination of
parameters reflecting the conditions of collision. For this reason, it is hard to
calculate true values because it often involves equations that are difficult to solve.
The problem can be considerably simplified by finding the data from experiments.
The experimental conditions must represent as closely as possible those occurring
during the intended collision. Taking into account the relatively weak relationship
between the coefficient of restitution and the velocity of collision can reduce the
volume of experiments [206].
From the system of equations (1.53), (1.54), we find post-impact velocities
given by:
(1.55)
MI +M 2
M 2 (1 + R)x l _ + (M 2 - RM I )X 2 _
x 2+ = (1.56)
The conditions for impact of a body onto asolid stop moving with velocity Vo are
obtained from (1.55), by putting M 2 ~ 00, :\:2_ = v o :
XI + = V o + R(v o - x I _) . (1.57)
Ifthe stop does not move (v o = 0) , then
xI+ = -RX I · (1.58)
Tuming now to the dynamic treatment of the coefficient of restitution [149], we
will make use of the concept of centre of inertia of colliding bodies. In view of the
autonomous nature of such systems, the velocity of their centres of inertia does not
change. For the case considered (Fig. lA), calling this velocity v o , we find
Mlx l _ + M 2 x 2 _ Mlx l+ + M 2 x 2+
Vo = -'----'---=--=- (1.59)
MI +M 2 MI +M 2
Using (1.59), equations (1.55) and (1.56) take the form
XI+=vo+R(vo-xl~)' x2+=vo+R(vo-xz~). (1.60)

Comparing (1.60) and (1.57), we see that direct central impact of two bodies can
be interpreted as collision of each of them with a stop, moving at the velocity of
the centre of mass.
We can distinguish two phases of impact. Firstly - a loading phase, from the
start of contact up to the maximum deformation, when the relative velocity of the
bodies is zero, but their general absolute velocity is v o • and secondly - an
unloading phase, from the moment of maximum deformation until the bodies
separate. Writing the impulse of the contact force in the first and second phases as
J l and J z' using the impulse-momentum equations, we can write for the first body
J I = -MI (XI~ - vo), Jz = -MI (v o - xIJ·
Examining the ratio between them, we obtain, using (1.59)
~= Vo -x l+ = x z+ -x l+ =R (1.6 I)
JI XI~ - Vo XI~ - xz~
Since the value of the impulse acting on each body is equal, then we find an
identical result when considering the second body. Hence, the coefficient of
restitution can be treated as the relation of the normal
'r
impulse of the contact force during the first and
second phases of impact. Such a representation is
known as Poisson's hypothesis.
Hrsj ,
7. We will use this latter treatment to examine '
,I ,
~
oblique impact, retaining x as the normal

t
displacement and defining the tangential I
I
!I
displacement as y. We shall begin with the simplest Fig. I.5
case of impact of a solid point of mass M onto a non-
moving stop (Fig. 1.5).
When considering oblique impact we assume that the change in the normal
component of velocity occurs independently from the tangential one, and follows
the law of direct impact. To determine the tangential component, we use the
hypothesis of dry friction, establishing a relationship between tangential and
normal impulse according to Coulomb's law.
Applying the impulse-momentum equation to the two load phases, we find
J lx = Mx_, (1.62)
(1.63)
Here yo is the sliding velocity at the instant when the first phase ends; the sense of
the remaining quantities is obvious. In agreement with the accepted hypothesis
J ly = eJ lx (1.64)
where e is the coefficient offriction. Substituting (1.62) and (1.63) into (1.64), we
obtain
(1.65)
Since the tangential velocity at the moment of impact can not change direction,
then Yo ~ 0, and from (1.65), we find
( 1.66)
This is the condition for the absence of jamming. When it is not satisfi ed or is
satisfied at the boundary, the tangential displacement stops even during the first
phase of impact and the point returns along the normal to the obstacle. If this does
not occur, then after the end ofthe second phase, we find
J 2x =Mx , ' J 2y =(y+ - Yo)' (1.67)
Attached to equations (1.67) are conditions equivalent to (1.64), and when we take
into account that the normal impulse in the first and second phases becomes
subordinate to equation (1.61), we eventually obtain, taking account of (1.62) and
(1.65),
(1.68)
Putting y + > 0 in (1.68), we find the condition for the absence of jamming
during impact as
y_ /x _ > 8(l+R). (1.69)
We will call oblique impact when the tangential component
does not become zero, sliding impact.
A more complicated picture emerges during constrained
oblique impact, when supplementary links are present on
the colliding bodies. Aseries of typical situations of this
kind is considered below.
First of all let us consider a mass M (Fig. 1.6) sliding
freely on a guiding bar and colliding with an inclined cam
on a fixed anvil.
The relative approach velocity ofthe colliding elements along the normal to the
cam profile will be
v N = -xcos\V -q,rsin \V (1.70)
where x is the linear velocity of the striker along the bar; q, is the angular
velocity of the striker; \V is the angle of the oblique profile, r is the average
radius of the cam. The impulse-momentum equation for the first phase of impact
takes the form
M(x o - x _) = -J I COS\jI
(1.71)
I(q,o -q, _) = -J I sin\jl
where 1 is the moment of inertia of the body; J I is the normal impulse during the
first phase of impact.
At the end of the first phase of impact v N = 0, hence from equation (1.70) we
find at that instant
. Xo
<Po = --cot\V (1.72)
r
From equations (1.71),(1.72) we find the value of the normal impulse of the
first phase of impact as
J = M(x30S\V + q,Jsin \V) (1.73)
I f-l sin 2 \jI + COS 2 \V '
where f-l = Mr 2 / I .
The post-impact velocity of the bodies can be found by writing the impulse-
momentum equation for impact ofthe whole as
(1.74)
and noting that the complete normal impulse is, according to (1.61)
J=J 1 +J 2 =(l+R)J , . (1.75)
Eventually, using (1.73) to (1.75) we obtain
x_ (f-lsin 2 \jJ - Rcos 2 \jJ) - 0.5(1 + R)<pJsin2\jJ
x+ =
11 sin 2 \jJ + cos 2 \jJ
. <pJ(1l cos 2 \jJ - RIl sin 2 \jJ) - 0.5(1 + R)f!:i:Jsin 2\jJ
q> + = r(1l sin 2 \jJ + cos 2 \jJ)
We note that similar relationships can be written for the impact of a free body,
without a guiding bar, when the cams are symmetrical by assuming that contact of
the cams occurs simultaneously.
When calculating constrained impact with
friction, the problem becomes complicated. In Fig.
1.7 a body of mass M, impacting against a stop, is
coupled kinematically to an inclined plane having an
unlimited large mass and moving with velocity v o'
>. The constraint is assumed to be ideal.
Fig.17
Denoting by x and y the normal and tangential
components of the absolute velocity of the impacting body and introducing the
value i for its relative velocity along the oblique plane, we find from Fig. 1.7
x=v o +zcos\jJo (1.76)
y=zsin\jJo (1.77)
where \jJ 0 is the slope of the oblique plane. At the end of the first phase of impact
x = 0, hence from (1.76), we find at that instant
zo=-vo/cos\jJo (1.78)
Comparing (1.78) and (1.77), we see that in the first phase of impact Z, and
consequently y, change sign, hence we can divide the phase into two intervals
having different signs for these velocities. To agree with (1.77) on the interval
boundaries y = i = 0 .
The impulse-momentum equation for the first interval of the first phase takes
the form
- Mz_ = J1'x cos\jJo + 8J,\ sin \jJo (1.79)
where J1'x is the normal impulse during the first interval. From (1.79), we find
1 Mz_
J 1 =------- (1.80)
x cos \jJ 0 + 8 sin \jJ 0
The impulse-momentum equation for the second interval is written as
Mz o =JII~COS\jJo -8JII~sin\jJ0'
From this, and using (1.78) we obtain
JII = _ Mv o
Ix COS \jJ 0(cos \jJ 0 - 8 sin \jJ 0)
Since the normal impulse can not change sign during impact, we obtain the
condition
8< cot\jJ 0. (1.81 )
This condition ensures that j amming does not occur.
The full normal impulse in the first phase of impact is as follows
J lx =J 1\ +Ji~ =-M[ cos \jJ 0 +z_8 sin \jJ 0 + cos \jJ 0(cos \jJ o V
0 - 8 sin \jJ 0)
] (1.82)
The impulse-momentum equation for the second phase of impact gives

M(i+ - zo) = J 2x COS \jJ 0 - 8J 2x sin \jJo
From wh ich, using (1.78), we find
J 2x = M(z+ +vo /c.os\jJo) .
cos \jJ 0 - 8sm \jJ 0
Writing the expressions for the impulses during the first and second phases m
terms of the coefficient of restitution (1.61) we eventually obtain
. . cos \jJ 0 - 8 sin \jJ 0 v ()
z+ = -Ri_ . - (I + R ) - - . (1.83)
cos \jJ 0 + 8sm \jJ 0 cos \jJ 0
In reference [17], it is shown that, for the impact arrangements of Fig. 1.8, a
the expressions for the linear velocity Z of mass M before and after impact are
similar to (1.83), where \jJ 0 is the value of the angle of crank rotation \jJ, at the
moment of impact, measured from the vertical axis counter-cJockwise. Bearing in
mind that in this case Z = \jJ!, where I is the length of the crank, we obtain the
angular velocity as
(1.84 )
c)
a)
Let us examine a type of machine arrangement frequently found in vibro-impact equipment (Fig.
1.8, b). Carriage MI carries tlywheel M2 with an unbalanced mass M3 We shall assume that a dry
friction force acts between the carriage and the guiding surfacc. The impulse-momentum equations,
§ I Collision 01' solids 19
describing an increase in the velocity ox, of the carriage and the angular velocity Ow of the flywheel,
can be written as
J x = ox,(M, + M 2 + M) - oIjiM)/cOSIjI-eIJylsignxl ,
J y = oIjiM)/sinljl (1.8S)
ox,MiCOSIjI- olji(/2 + Mi) = 0

where /2 is the moment of inertia ofthe flywheel; / is the eccentricity ofthe mass M 3 ,
From equation (1,85) we find
~' MicosljI~,
üljl = 2 üX, (186)
/2 + M)/
'l
J x =ox,M,+M 2 +M)
/2 + M 3 / 2(sin 2 \V-0,S9Isin21j1IsignX)]
2 ' (1.87)
/2 + M)I
Noting that in the first phase of impact ox, = -x,_ , J x = J,x' while in the second,
ox, = x" ,J, = J 2x = RJ,x and the change in the sign of velocity x,occurs on the phase boundary,
with the help 01' equation (1,87) we find the value ofthe post-impact velocity ofthe carriage as
(1.88)
where
M, + M 2 + M) 12 + Mi 2 (sin 21j1- 0,Selsin21j11)/(l2 + M)/2)
:fl = R----'---=---"l'-"--..::....,;-'--;,--'--~-+..:....='----=-."..:..j
MI + M 2 + M) 12 + M)P(sin 2 1j1 + 0,S9Isin2\vl)/(/2 + M)/2)
The complete change of velocity as a result of collision is
ox+=x,+-x,_,OW=W+-Iji_,
hence, in agreement with (1,86) and taking account of (I ,88), we obtain
, , MicosljI( ')'
\V+ =1jI-- 2 1+9'1 XI_' (189)
12 +Mi
Note that the normal impulse in (1,87), J x 2: O,As a result, in (1,88) 9'12:0 , The change of sign of
these quantities shows that the mechanism becomes jammed in the first impact phase. Thus, initially,
when /1 = 0, \jI = rr/4 the following condition will guarantee that this situation will not occur:
- 9<1+2(M,+M2 )/M),
In conclusion let us estimate the eftect of collision in a similar system with a moving boundary
mass M p assuming 9 = 0 (Fig. 1,8, cl, In this case, the corresponding equations for the impacting
system (1.8S) to (1,87) remain valid but for the boundary mass M 4 we find
-J x = OX4M4 (1.90)
Noting that in the first phase of impact oXI = Xo - x,_ , oX4 = Xo - x4_' and in the second phase
ox, = xl+ - Xo ' oX4 = x4+ - Xo where Xo =xlO = X40 common velocity of masses MI and M. on the
phase boundary, and lIsing eqllation (1,61), we obtain the following equations for the post-impact
velocities analogous to (I.5S) and (IS6)
X _(M-RM4 )XI_+M4 (l+R)x4_ X _ M(l+R)xI_(M 4 -RM)x4_

,+ - M + M4 ' 4+ - M + M4
Here M=MI+M2+M)(/2+M)/2sin21j1)/(I2+M)/2) is a value, which represents an equivalent

mass of the impacting systems
In agreement with equation (1,86), the post-impact angular velocity ofthe flywheel is defined as
, , M)/cosljI, '
1jI+=IjI-+ 2 (XI+- X '_)'
/2 + M)I
Problems involving constrained impact in systems with dry friction have received special
consideration here becallse such problems are not adequately covered in the literature,
Hence, within areas the application of stereomechanics gives simple kinematic solutions during
the study of coll ision.
§2 Oscillator with fixed stop

1. Using stereomechanics in impact theory makes it possible to find exact
solutions for some dynamic models of vibro-impact systems. Finding such
solutions is based on the stitching procedure using kinematic conditions of impact
to stitch the solutions of differential equations describing motion in the intervals
between collisions. Hence, in parallel with the some form of idealisation,
characteristic of stereomechanics, the development of suitable models, requires the
introduction of additional simplifying assumptions, which allow us to find
solutions during the intervals. As a rule, this leads to the use of linear differential
equations to describe the motion during the intervals.
Thus, exact solutions can be obtained by
significant idealisation of real systems, at the D
I
price of results that do not match quantitatively

the real situation and sometimes also give a
distorted qualitative picture of the true motion. " »Tm7/ffmr/l~
~If
For example, it is characteristic of such idealised d)
simplifications that dissipative effects in systems
frequently leads to unrepresentative motion of the
system.
Exact methods can generally only produce
O~I -
,j
1;
r~l\'
b)
analytical results for models of small dimension.
At the same time, exact methods can allow Fig.2.1
detailed analysis of motion of these models to be
carried out, and these are important not only in themselves as areas of study, but
also from the point of view of obtaining standard results with which various
approximate methods can be compared. This latter point will be dealt with as we
study approximate methods.
Using this approach, throughout this chapter we will examine the results of
exact analysis for several types of dynamic models of vibro-impact systems,
emphasising the qualitative side of the examined effects and studying the
frequency behaviour of systems.
The case of simple restricted movement of the mass of a linear oscillator near a
rigid stop (Fig. 2.1, a) reveals important features of the behaviour of such systems
following collision. Many references in the literature [23, 48, 58, 146, 193] are
devoted to the study of this dynamic model, and we shall use their results
frequently in what folIows.
We shall look initially at the natural vibration of this system, assumed to be a
conservative oscillator, with totally elastic impact (coefficient of restitution R = 1).
In the intervals between impacts, the motion of the mass M, measured from the
state coinciding with the unstrained condition of the spring of stiffness c, satisfies
the differential equation of an oscillator
§2 Oscillator with fixed stop 21
Mx+cx=O (2.1)
At the instant when the mass impacts against the stop, which occurs when co-
ordinate x = ~, its velocity instantaneously changes sign. We note that the
positive value of ~ occurs when there is a gap between stops and mass, and
becomes negative with initial interference.
Elements, which contact during impact form an impact pair. As we examine
one-dimensional conditions, during which relative motion of the contacting
elements can be written as a single co-ordinate, we will recognise single-sided and
double-sided impact pairs depending on the number of contact surfaces for each of
the interacting elements.
Rewriting equation (2.1) in the form
(2.2)
where D = Jcl M is the natural frequency of the linear oscillator, we obtain after
integration
x = CI cosDt + C 2 sinDt, (2.3)
where CI and C2 are arbitrary constants.
Using the boundary conditions t = 0, x = ~; t = 2n / co o ' x = ~ , where CO o is the
natural frequency of the oscillator interacting with the stops, we find the motion of
the system in the intervals between collisions as
x = ~
cos(nD/ co o )
cos(Dt - nE..J .
CO o
(2.4)
Relationship (2.4) represents the vibration of the system only for one per iod
I E [0,2n/co o ] (Fig. 2.1, b) however, this information, naturally, is sufficient to
represent the entire process. For example, expanding this solution as a Fourier
series, using (2.4), we find for the coefficients a XI (i - 0, 1, ... ) ofthe harmonics
(2.5)
and, as a result, the process examined can be represented along the entire time axis
in the form
(2.6)
where m," = a(o /2 .

Figure 2.2 shows phase planes for the process examined for various stop
arrangements. Since the motion of the representative points on the curved
boundaries occurs uniformly with constant angular velocity D, but impact occurs
instantaneously, the length of the curves for the corresponding c10sed cycles is
proportional to the period of vibration of the system T = 2n / CO o with collisions.
From the diagram it follows directly that this period can, in general, depend on the
amplitude of vibration and will not exceed thc pcriod of natural vibration of linear
oscillator, i.e.
Q / wo:S: I. (2.7)
while in the case when the stop and gap are arranged as in Fig. 2.2, a, as the
11>0 z/fl
(/ /
amplitude of vibration Increases, ti'cqllency increases and consequently,

nonlinearity, introducing impact, takes a "hard" character; with interference,
conversely, (Fig. 2.2, c), increasing amplitude causes a redllction in the vibration
frequency, which we will call "soft" nonlinearity; finally . in the boundary case,
L'1 = 0 (Fig. 2.2, b) corresponds to an isochronolls system. although the nonlinear
character ofthe vibration is retained. Wc notc that
0.5<Q / w u :S: 1 when .'I>1i
Q/w o =0.5 when i\ = () (2.8)
Q ! w o <0.5 when /1< ()
We will construct backbone curves for all three types of system, i.e. the
relationship between amplitude and frequency 01' ti-ee vibration. When we refer to
amplitude a, ofprocess X(/), then we willmean thc half~swing value (Fig. 2.1, b) as
often used in non-harmonic vibration studies. E\(pressing the co-ordinate 01' the
maximum deflection dllring process x(1) as\,. we find
G, = +(,'" -x o) (u,,? 0) . (2.9)
According to (2.4)
Xo ==L'1 ! COS(TCQ / CIl,,) (2.10)
Substituting (2.10) into (2.9) we obtain the desired relationship
" , =
\( I)
-211- ~-;-s(;D!~~, . (2 . 11 )
In Fig. 2.3 b,lckbon e cllrves are presented for

systems with variolls arrangements of stop (full
lines). When the 5top is installed with a gap, the
vertical line of the backbone cllrve is added for
isochronolls vibration 01' a linear oscillator until
they contact thc 5top. Later we will see that the type
Fig 23 of nonlincarilV obtaincd is characteristic for vibro-
impact systems with single sided stops. Note that in all the examples examined
(Fig. 2.3), as amplitude is increased, their frequencies approach more cJosely to
double that ofthe linear oscillator.
In accordance with (2.7), series (2.6) converges fairly quickly and it is
interesting to discover to what extent the described nonlinear behaviour of the
backbone curve is shown by the first harmonic (fundamental) of vibration.
According to (2.5) the amplitude ofthe fundamental is given by
211 0
axl = tann- (a xl ~ 0) (2.12)
n(O/coo-coo/n) COo
This relationship is shown dotted in Fig. 2.3. Comparison shows that the first
harmonic of the process quite accurately represents the nonlinear behaviour of the
system. The situation indicated is studied in Chapters 2 to 4 using approximate
methods based on analysis of the fundamental.
2. Let us now consider forced vibration of a linear oscillators with stop,
arising when a periodic force P(t) = P cos( cot + <p) (see Fig. 2.1, a) is applied to
°
the vibrating mass. We shall find the loss of energy in the system using the
coefficient of restitution R, assuming that < R ~ 1.
In the interval between collisions in the system, the following differential
equation is valid
(2.13)
where a p = P / M; <p is the unknown initial phase of excitation. The general
solution of the non-homogeneous linear equation (2.13) will be written in the form
x = CI cosOt + C 2 sin ru + acos(cot + <p), (2.14)
where
ap
a= -....,--'----,- (2.15)
021 1 -s 21
is the amplitude offorced vibration ofthe oscillator without a stop; s= co/O. We
note that, in nonlinear systems, as weil as the periodic motion with frequency co,
sub-periodic vibrations can also occur with aperiod that is a multiple of the
exciting period . By taking the start time at the instant after impact, we write the
boundary conditions for all combinations of periodic motion considered in the
following form:
t=O,x=l1,x=x+; t=2nl!co,x=l1,x=x_, (2.16)
where I is regime multiple; x_,x+ are velocities before and after impact, related by
(l.58)* Using boundary conditions (2.16) to find the arbitrary constants CI and C 2
and phase <p in (2.14) we obtain
• As weil as periodic regimes with a single impact per period of a mass against a stop as examined
here. in vibro-impact systems it is also possible to find, similarly, periodic multi-impact processes,
characterised by several impacts against the stop during the course of a single period. In general,
single-impact regimes show the most intensive (resonant) character. They occur commonIy in working
processes and research is being carried out into their harmfuI effects. For this reason they are given
special attention in this book.
(2.17)
. BD.
sm<p=-x_, (2.18)
Oa
(
cos<p = 1 - O~ x D')~a' (2.19)
where
1(1-
B=-c, I+R tan~,
R) nl
D = --(1 + R)cot-
2 S
1 nl
(2.20)
Eliminating the trigonometric functions from equations (2.28) and (2.19), we

can express the value ofthe velocity x_ by the following
[1 ± ~1 - (1- a 2 / ~2)(I + B 2 ) ]O~
X = (2.21)
- D(1 + B 2 )
Assuming in (2.21) that the expression under the radical 1- (1- a 2 / ~2)(1 + B2 ) ~ 0,
we obtain the condition defining the domain for real values ofthe velocity x_:
I~I /a ::; Jl + 11 B 2 • (2.22)
According to (2.22) vibro-impact regimes can occur not only when the stops
are arranged within the limits of the amplitude of vibration of linear oscillators
(0 < ~ < a) , but also outside this region (I ~ I> a) . In the latter case, to develop
into a vibro-impact regime the system must be supplied with sufficient energy for
the process to start up, i.e. to receive what is called hard excitation. Later, when
examining the motion of systems with double-sided stops, we will consider such
features of vibro-impact systems in greater detai I.
We note that when equation (2.21) and the values (2.17) to (2.19) obtained
from it, are substituted into the motion law (2.14) we obtain double-valued
functions for the system parameters. Hence, multivalued solutions that are
characteristic of nonlinear vibration, also occur in vibro-impact systems. In
addition, not all solutions satisfYing (2.22) represent regimes that can exist.
Searching for periodic solutions, we assumed that the mass received one collision
per period of motion, making use of boundary conditions (2.16). From the
structure of the model, it can be seen that the motion traj ectory must not reach the
stop in the interval between two calculated collisions, and this is not taken into
account in the analysis carried out. By making motion satisfY conditions for the
absence of additional intersections
x(t) ::; ~ , (2.23)
it is possible to isolate a narrower existence domain for the required solution from
the domain ofreal values (2.22).
It is obligatory that the following inequality is satisfied in this case
x_ ~ O. (2.24)
This obvious condition was not apparent either in the equations of motion or the
boundary conditions and must be introduced additionally.
From (2.21), it follows that, if the term under the radical is greater than unity,
then velocity x can take different signs and as a result, equation (2.24) is always
satisfied for one and one only of them. Substituting into (2.21)
1- (1- a 2 I ~2 )(1 + b 2 ) > 1
we find the following condition for a single solution I~I < a. This solution
corresponds to the positive sign in front of the radical in (2.21) which occurs
when MD > 0, or, using the notation of (2.20), in the following frequency bands:
l/ rn < C, < 2l/(2rn - 1) for ~ > 0 ;
2l!(2rn - 1) < C, <l I(rn -1) for ~ <0 (2.25)
(rn = 1, ... , 2, ... ) ,
In the case when ~ID < 0, i.e. adjacent to frequency areas of (2.25), a regime
occurs which corresponds to the negative sign ofthe radical.
Hence, for devices with gaps (interference),which do not exceed the amplitude
of vibration of the linear oscillator, only one vibro-impact regime can exist for
each multiple.
Equation (2.24) is satisfied for both values of velocity (2.21), if I~I > a and as a
result ~ID > 0, i.e. in the frequency range found earlier (2.25). Here, in fact,
equation (2.22) defines the upper boundary for the area of double-valued solutions,
the lower boundary ofwhich gives the ratio.
I~II a = 1. (2.26)
The inequality obtained still does not guarantee the absence of additional
intersections. In general, it is difficult to find obvious analytical equations for the
boundaries of existence domains as defined in equation (2.23). It would be
necessary to find extreme values of deflection for the motion (2.14), using the sum
of separate harmonic functions with different frequencies. Such proofs are usually
obtained by checking the motion directly. This laborious procedure can be
automated using computers.
We shall limit the examination of frequency effects, assuming R = 1, and
neglecting energy dissipation during impact. For convenience in further analysis,
we will change the law ofmotion (2.14) using (2.17) into the form
~ - a cos <p rot - TII
x= cos---+acos(rot+<p). (2.27)
cos( TII I c,) C,
In agreement with equations (2.18) to (2.21), when R=l
sin<p=O, cos<p=±l,
and one of the extreme values (2.27) occurs at the instant t = TII I ro . This value
must satisfy condition (2.23)
X( -
J -
TII = ~ a cos <p + (-1) I a cos <p :<:; ~ (2.28)
ro cos( TII I c,)
Defining the va1ue x_ = ilDA (J ± a / lAI) from (2.21) and substituting this into (2.19)
we find
acos<p=+A1aIAI· (2.29)
Both now and 1ater, the sign that is disposed Iike the
simi1ar sign before the radica1 in (2.21) must be taken.
Transforming inequa1ity (2.28) using relationship
(2.29), we eventually obtain the following conditions for
the absence of the additional intersections
~ < +cot 2 ~ when A > 0 (2.30)

a 2s cos(nl/ s)
~>+coe~ when A <0

a 2s cos( nl I s)
foroddratios(l= 1,3, .. . );
lAI< +1 when A >0 (2.31)

a cos( nl I s)
lAI> +1 when A <0

a cos( nl I s)
for even ratios (l = 2, 4, ... ).
Fig. 2.4 shows examp1es of the existence domains
for regimes with various ratios (l = 1, 2, 3) constructed
from inequalities, (2.22), (2.25), (2.26), (2.30), (2.31)
and constant coefficient of restitution R = 0.5. Here the
areas hatched with positive slope correspond to regimes Fig. 2.4
with positive radicals in (2.21), while the negative
shaded areas correspond to the negative radical. Areas where double-valued
solutions occur are indicated by cJoser hatching.
The vibro-impact system considered occurs due to single-sided limitation ofthe
vibration of the simplest form of linear oscillator, whose behaviour is weH known .
It is now simple to evaluate those effects that cause such limitations on the
frequency range ofvibro-impact regimes. Above all, we can see immediately (Fig.
2.4) that the domains in which vibro-impact regimes can exist mainly in the post-
resonance range of the linear oscillator. (l, > I), whi le for a system with
interference (A < 0) in the pre-resonance frequencies (l, < I) such regimes do not
exist in general.
Adding the curve of Ma to Fig. 2.4, using equation (2.15) for the amplitude-
frequency characteristics of a linear system
A = AQ211_ s21 (2.32)
a ap
we can visualise the frequency range for the existence of different regimes. The
intersection points of this curve with the boundary of the existence domains define
R-D,5 ~>O
a)
Idl!i 6-<0
(/p
bJ
Fig.2.5
the limiting frequencies for the occurrence of vibro-impact processes. Hence, for
example, for systems with a gap (ö- > 0) near resonance S = 1 and a unique vibro-
impact regime will occur. At the same time, at post-resonance frequencies in the
band I < S < 2/ it is possible to cause dual regimes of ratio I.
In Fig. 2.5, a, for a system with an initial gap (ö- > 0) the existence domains
are drawn against amplitude-frequency characteristics for a linear oscillator (2.15).
The hatched areas, show the lower curves (2.15), which exist only for single-
valued regimes, but in the areas included between the lines (2.15) and (2.22) -
pairs can exist. Existence domains for regimes with ratios I > 1 are bounded on the
left by frequency S = I . The diagrams presented allow us to define the frequency
range in which vibro-impact occurs for given values of a gap and to find the value
of the gap for which such a process is possible at a given frequency. Fig. 2.5, a
illustrates graphically the fact mentioned earlier of the existence of vibro-impact
regimes for a device with a gap, as the amplitude of linear oscillation is increased
(L'. > a). The value of such increase defines the boundaries (2.22). Notice that
existence domains for various frequency ratios partially overlap. Obviously, in
these zones development of one or other regime of existence depends initially on

the way the transient process takes place.
It is interesting to note that neither increasing the gap nor reducing the exciting
force will lead to the elimination of vibro-impact systems whatever the ratio, even
in the presence of impact dissipation, which only contracts the frequency range.
(As R increases, the limiting value (2.22) approaches more closely to the vertical
line S = 21 .) We must remember, however, that we have not taken account of
other reasons for dissipation of energy, such as viscous damping, internal damping,
etc. When we consider double-sided stops later, this aspect will be analysed.
Jumping ahead, we will show that viscous losses influence the limits (2.22) of
existence domains near frequencies S = 1 , limiting the value of Li
at which vibro-
impact regimes occur and reducing existence domains for regimes as 1 increases.
For a system with initial interference, (Li< 0) the impact dissipation causes the
elimination of regimes of higher multiples. In Fig. 2.4, such a situation occurs
when 1 = 3, where the lower portion of curve (2.32) does not intersect with the
boundary (2.22) of the area of existence. The limiting value of the related
parameters for which a regime of multiple 1 occurs, is found by substituting
"c: inoot1'lT' ft -,'I

equation (2.32) into condition (2.22), and writing it using (2.20) in the form
ILlI< ~; [1 +
Simplifying this inequality, as S --+ 00 , we obtain
ILiI ::; ~ (I + R) . (2.33)
1(/0. 2 (1- R)
Fig. 2.5, b shows existence domains of vibro-impact regimes for systems with
initial interference. Note that existence domains for regimes with high ratios are
completely enclosed within areas of lower ratio, while regimes of ratio 1 exist in
the frequency range S ~ / . For odd values of I, single valued domains are arranged
below boundary (2.30), but for even values, below boundary (2.15). Double-
valued domains occur above these lines, limited by (2.22). Equation (2.33) is
shown as asymptotic to the upper boundary of the area of existence (2.22) (Fig.
2.5, b). The asymptotes, shown as dotted lines for the lower boundary (2.30) ofthe
area of double values, are found similarly and defined by the relationship
Li
= 4al' /(1(/0.)2 (2.34)
Note that relationship (2.33) can easily be used to choose the initial
interference, guaranteeing the suppression of undesirable vibro-impact processes,
occurring, for example, in kinematic coupled mechanisms when they are
disconnected.
We shall examine regimes for various ratios at greater length and try to clarify
the physical meaning. When the boundary conditions (2.16) were written, the ratio
1 was defined as the ratio of the period of system motion to the period of
excitation. Hence a regime of ratio I is a process of frequency 0)1 which is related
to the excitation frequency 0) by the expression
w, = w/l . (2.35)
Referring to Fig. 2.4 and Fig. 2.5, we note that frequency w, lies within the limits
o ::; w, < 20 when ~ > 0,
20 ::; w, < 00 when ~ < o.
Comparing this with the results obtained in section 1 during the examination of
free vibration, we see that the frequency 0), of the regime for any ratio coincides
with one of the natural frequencies Wo (2.8) of the
vibro-impact system, and hence the corresponding
tuning ofthe system is called resonance.
Fig. 2.6 shows the motion for regimes with
ratios I = 1 and I = 3, excited by an extemal force
t of frequency ~ = 1.75, and ~ = 5.25 . For

Fig.2.6 comparison the dotted lines show motion during
free vibration with frequency ~o = 1.75 . From the
diagram it is obvious that the regimes for various ratio I are not very different from
each other and at the same time they are quite cIose to natural vibration. lt is also
possible to establish this directly from the motion (2.27), rewriting it taking into
account (2.35), in the form
x= ~ - acosep
cos( nO / w, )
no)
cos( O t - - +acos(wt+ep)
w,
(2.36)
and comparing (2.36) to the free vibration (2.4). As the frequency of excitation
w increases, or the regime ratio increases, so the value a ~ 0 and the law of
motion (2.36) approaches more closely to (2.4).
Thus re sonant motion of vibro-impact systems can be treated as free vibration
sustained by external excitation. Such treatment is widely used in nonlinear
mechanics and will be repeatedly made use of in what folIows. The fact that
resonant motion is cIose to free vibration can be used to estimate the amplitude of
forced vibration using the backbone curve (2.11).
Fig. 2.7 shows backbone curves for various ratios (l = 1, 2, 3), constructed
using equation (2.11), using relationship (2.35). In this way we can approximate
the branches of the resonance curves when R = 1 . They are increasing infinite as
C; ~ 21. Dissipation by impact (R"* 1) leads to the limiting value of vibration
amplitude. Searching for the limiting frequency of vibro-impact regimes
corresponding to the boundaries (2.22) of existence domains (Fig. 2.5), we find the
higher points of the resonance curves, the achieved parts of which are shown in
Fig. 2.7 by bold lines.
Later we will carry out detailed analysis of the amplitude-frequency
characteristics of systems, and we will prove that the branches of the resonance
curves, found from exact relationships, are arranged near the lines we have
constructed, while the interpretation of resonance regimes given here, is
convenient and sufficiently accurate, especially when estimating the maximum
amplitude.
3. When the differential equations and boundary conditions were defined,

only basic factors influencing the system were
accounted for, based on the assumption that we 11.
could neglect small perturbations. Yet the
extemal perturbations cause changes to the real
motion from the solution found. For one period
of motion, these small effects cause slight
changes in the starting conditions of motion in
the following interval from those that match
the conditions of periodicity. However, it can 11.
be shown that as time elapses, small
perturbations will accumulate and significantly l -Z
distort the final result of motion of the system,

so that it will differ considerably from the \111
solutions found for undisturbed conditions. z

Such solutions are called unstable. Naturally, FIg.2.7
they are not produced in real systems. Thus the
analysis of stability leading to the separation of
physically possible solutions has a very important practical meaning.
A stable solution is one, which changes negligibly due to small perturbations of
the initial conditions during any time interval. In this case, it is sufficient to find
the tendency for perturbations to cause changes at the instant of a following
collision. Stable motion occurs if, during continuous collisions, the perturbation
either decreases or stays within limits; when unlimited growth of perturbation
occurs, this represents instability. The procedure for such analysis was developed
by R E Brunstein and A. E Kobrinsky [I 17].
Suppose the system realises a disturbing motion, such that at the instant of an
arbitrary v-th collision the mass reaches the stops with a velocity (x _)v , differing
from the undisturbed value x_ by a small quantity oi v while the disturbing phase
<p has increment o<Pv. This is equivalent to an increment oC lv and oC 2v in the
integration constants Cl and C2 for the solution (2.14). Assuming initial start time
at the instant of impact, we can write the equation for the disturbed motion of the
v-th interval as
x+ox" =(C l +oClv)cosQt v +
+(C 2 +oC2JsinQlv + (2.37)
+ acos(wt v + <P + o<p.,)
The (v+ 1)-th collision occurs at the instant t v = 2nl / w + 0 TI' where 0 Tl' is a small
change in the period between collisions following the initial disturbance. Suppose
that in (2.37) t v = t + ot v ' where ot v = 0 at the v-th instant and 01" = oTv at the
instant ofthe (v+I)-th impact, we can expand (2.37) in a Taylor series for small
increments. From the equations obtained we subtract those for undisturbed motion
(2.14) and we find the following equation for the disturbances in the v-th interval
to second-order accuracy:
8x" = 8C I " cosQt + 8C 2" sinQt-

(2.38)
- Q8t v(CI sin Qt - C 2 cosQt) - a( ro8t v + 8q>v )sin( rot + q»
Writing the boundary conditions for the disturbed motion (2.38) at the instant
after the v-th and before the (v+ 1)th impact
t = 0, 8t v = 0, 8x v = 0, 8i" = 8x v+;
2nl (2.39)
t = -,8t" = 8T,,, 8x,. = 0, 8i" = 8x(v+I)_'
ro
and the corresponding increment in the velocity during the (v+ 1)th collision
8x("+I)+ = -R8x(v+I)_ . (2.40)

Applying conditions (2.39), (2.40), to (2.38) we obtain a recurrent relationship
connecting the value of the disturbing initial conditions 8C I, 8C 2 and phase 8q>
to the two adjacent intervals ofthe disturbed motion
8C I" - a8q>" sin q> = 0,
(2.41 )
CI nl
+ (8q>"+1 -8q>,,)~tan~=0,
. 2nl 2nl
8C I"RsmT- 8C 2(v+I) -8C 2"RcosT+
+ 8q>(,,+])[ R~I + a~(l + R)COSq>] - 8q>" R~I = °

(~= ro/Q)
In (2.41) and taking equation (2.17) into account as weil, there is an obvious
relationship between the incremental phase and period 8q>("+1) = 8q>" + ro8T",
wh ich reflects the fact of the accumulation of phase for v cycles
"
8q>" = 8q>o + ro8Tv I
i=1
The relationship obtained represents the linear finite difference equations [99],
the solution of which take the form
(2.42)
where hl' h2, h] are coefficients independent from v; ß is a constant.
Substituting the assumed solution (2.42) into equation (2.41), we obtain the
following system of linear ordinary equations for the relative co-efficient hl' h2, h3 :
h l - h 3 a sin q> = °,
2n/J . 2nl CI nl
h ( ß-cos- -h sm-+h (ß-l)-tan-=O
I ~ 2 ~ 3 ~ ~'
. 2rcl 2rcl
hjRslll- - h 2 (ß + Rcos-) +
I;; I;;
+ hj {ß[ R~j + al;;(1 + R)COS<P] - R~j } = 0

This set of equations has a non-zero solution only when the determinant of the
coefficients is zero, i.e.
1 0 - Gsin <P
2rcl . 2nl CI nl
ß-cos- -Stn- (ß -I)-tan- =0.
I;; I;; I;; I;;
. l2nl
RS ll- - ( ß+ Rcos~ ~ + al;;(1 + R) cos <P ] - ~
2n/) ß[RC RC
I;;
Expanding the determinant and using equation (2.17) to (2.19) we obtain the
chGracteristic equation in the form
aoß2 + ajß + a 2 = 0 (2.43)
where
ao =~; Gj = t.1;;(1 + R)sin 2nl +

co I;;
x_[ 2( 2 2n/) 2
+ 2co (1 + R) l21;; cos ~ -1 - (1- R) cos~ ;
2n/]
Hence solving the stability problem leads to an estimate of the roots of the
characteristic equation (2.43). lfroot ß I,2 is disposed within a unit circle, i.e.,
(2.44)
then, in accordance with (2.42), when v ~ 00, the value of the perturbations
oC lv ~ 0, oC 2v ~ 0, o<Pv ~ 0, and the motion under investigation has
Gsymptotic stability.
To estimate whether roots correspond to condition (2.44) it is not necessary to
solve the characteristic equation. There exists for this aim Schur's criteria [160],
which can be constructed directly trom the coefficients of the equation or trom the
determinant ofthe characteristic equation [14]. For equations ofthe second degree,
these criteria take the following form:
121
a o < 1' 1
Go + G21 0, (2.46)
a j - ao - a 2 < 0 . (2.47)
Substituting in (2.46) the values of the coefficients from (2.43) and making use
of relationship (2.20) during any further conversion we have
ßQ/(DxJ< 1+ B 2 (2.48)
r
Using the value given by equation (2.21) for x_, we obtain
[l± ~1-(1- a 2 / ß2)(1 + B 2 ) < 1. (2.49)
From this it follows immediately that the conditions for stability of the solution
with the positive sign before the radical are satisfied when
IßI /a < ~1 + 11 B 2 , (2.50)
but solutions with the negative sign when
IßI /a < 1. (2.51 )
Note that the stability inequalities found in (2.50), (2.51) coincide with the
conditions for existence obtained earlier, (2.22), (2.26). Hence it appears that these
conditions are always satisfied for those regimes corresponding to domains
containing single valued solutions (2.4).
In domains with double-valued solutions, stability can only exist for a regime
having a value of x_ together with the positive sign before the radical in equation
(2.21 ).
Let us now examine the criterion (2.47). After conversion, similarly to that
earlier, we obtain the following condition for stability:
I 1- C 2
--;=======>-- (2.52)
I ± ~I- (1- a 2 / ß2 )(1- B 2 ) 1+ B 2 '
where
c 2 = (1 + R
)[1 + cos 2(nil <:;)] + 2R sin 2(nil <:;)
2
[<:;(1 + R)cos(nl I <:;)]2

When analysing (2.52) we will assurne R = I for simplicity, taking into account
that this equation depends uniformlyon R. As a result inequality (2.52) takes the
form
1 > 1- I (2.53)
1± a IIßI <:; 2 cos 2(nil <:;) .
Using the positive sign, the stability conditions take the following form:
IßI
I a > <:; 2 cos 2(nil <:;) - 1 (2.54)
Using the negative sign, and taking account of the conditions for existence
IßII a < I (Fig. 2.5), we obtain
IßI I a <1- <:;2 cos 2(nI! <:;). (2.55)
and in practice these all exhibit instability, excepting domains lying in the narrow
zones elose to frequency <:; = 21 .
Fig. 2.8 shows boundaries for existence domains, hatched portions ofwhich are
stability domains, obtained according to (2.54, (2.55). Note that stability domains
in general fall in the frequency range 1 < <:; < 21 when ß > 0 and <:; ~ 21 when
L'l < 0 ,i.e., exactly where we constructed the backbone curves and discovered the
possibility oflarge amplitude vibrations (Fig. 2.7).
4. Let us turn to the construction of amplitude-frequency characteristics for
the system, taking the value given by (2.9) for amplitude, as before. Finding exact
values for the amplitude is laborious and requires a search for the maximum
deflection X o in equation (2.27). A simple solution occurs only when R = I, and the
maximum deflection is at the instant t = nl / 00 . However, the construction of the
laws of motion for various values R '" 1 (see, e.g., Fig 2.6) as discussed in section
2, and comparing the resonant regimes with free vibration shows that, even he re
the maximum deflection occurs close to the middle of period t = nl / 00 . Hence we
will assurne X o "" x(nl / (0) . Finding this value from (2.27) and substituting it into
(2.9) we obtain for the amplitude ax :
- 1-
a x =L'l(
2
1 J- acos<p[(-1)I - cos(nI/
cos( nI/ c,)
1] c,)
. (2.56)
Let us construct the resonance curves for regimes with ratios 1= 1. In this case,
the first term in (2.56) coincides with equation (2.11). According to (2.19), (2.31),
the value cos<p is a double-valued function of the system parameters. Thus
equation (2.56) defines two branches of the resonance curve, and (2.11) the
backbone curve which separates them. The
resonance curves for various stop arrangements
R-O,5
are shown in Fig. 2.9. Here, item 1 indicates the
backbone curves. H
We note that in a system with initial gap 1
(L'l > 0), linear harmonic vibration can also
occur, of which the amplitude a <I L'll. It is easy
to show that resonance curves (2.56) result from -2
points where amplitude frequency characteristics
of linear oscillators intersect the stop level
a =1 L'll. Really, with this condition, equation
(2.21) gives a velocity value x_= 0, in
agreement with (2.19), a cos <p = L'l and as a
result (2.56) gives a x = L'l. It is interesting that
the same value of the amplitude a x of linear
resonance is reached when C, = 1 . In this case, the
11/17
second term (2.56) tends to zero as C, ~ 1. 12
Hence for pre-resonance frequencies near linear 5
resonance, the amplitude-frequency O~~;PP~~9~(
characteristics of a nonlinear system show a
-5
characteristic fall, when amplitude a x is less than
-11
the value of the initial gap L'l. Later, we will find
that this effect leads to stability of systems with Fig.2.8
double-sided symmetrical stops when impact
occurs against one stop only (see §3, section 6).

Let us now find the maximum vibration amplitude. From equation (2.56) it is
obvious that when cos<p = 0 the resonance curve intersects the backbone curve .
According to (2 .19) and (2.21) one of the values of cos<p is zero, when
ß = a / B . Multiplying both sides of the latter equation by ..!..(I-

2
I ), using
cos( 1t/ / ~ )
values (2.15), (2.20) we obtain
~ (1- COS(~l/ S)) = a;, (2.57)
where
• ap (1 + R) Stan[1tl/(2S)]
a = 2
(2.58)
x 20 (l-R) (s2- 1)
Comparing equations (2.56) and 2.57) it can be seen that equation (2.58)
defines a curve in the (r" a) plane, which intersects the backbone curves exactly
at resonance. This is the highest point on the resonance curve. When R = 1,
branches of the resonance curves continue infinitely along the backbone curves
and do not cross them anywhere (dotted lines in Fig.2.9, c) . This follows directly
from (2.56), taking into account that cos 2/ .
From Fig. 2.9 it is obvious that such behaviour exists when curve (2 .58) intersects
the backbone line higher. Substituting into (2.57) the limit when S ~ 00, we
obtain the limiting value for the conditions of existence of vibro-impact regimes
lill<~I+R .
1t / 0 2 1-R
Earlier this equation was obtained from other considerations (see (2.33)).
d>O /J.-O /J.<O

'I
2
a) bl
Fig. 2.9
Finally, systems with no initial gap (il = 0), give an isochronous resonance
curve for all ranges of existence, the maximum of which reaches frequency
S = 2/ . At this value, equation (2.58) gives the following value for the resonance
amplitude:
• al' I (l + R)
a =---,--- (2.59)
x 0 2 (4/ 2 -I) (l-R)
From this it can be seen that amplitude of resonant vibration dies away faster as
the ratio 1 for the regime increases.
Let us look at limiting points ofthe resonance curves (Fig. 2.9, a, c) where the
upper and lower branches merge. Such merging occurs when cos<p in (2.56) takes
unit value, i.e., according to (2.19), (2.21) on the boundaries (2.22) of existence
domains which coincide with the stability boundary (2.50). For these boundaries
the value is
Substituting this in the equation for the resonance curves (2.56), we obtain the
following equation for the stability bound:
a = L'l(I_ I )_ L'l [(_1)1 ___1_ _] (2.60)
x 2 cos( TI 1/ c,) 1 + 11 B 2 cos( TI 1/ c,) .
Here the first term defines the backbone curve with which the stability boundaries
coincide when R = I. In general, the bound (2.60) occurs below the backbone
curve, separating the branches of the resonance curve, from which it follows that
the stable one in the double-valued existence domain is disposed above the
backbone curve.
All that has been discussed here holds true for regimes of any ratio. It folIows,
however, that as 1 increases, the frequency range of regimes moves further away
from the frequency of linear resonance, the second term in (2.56) becomes
negligibly smalI, and branches of the resonance curves become eloser to the
backbone curve. However, higher points on the resonance curves approach the
stability boundaries. Thus, the calculation of the amplitude of forced vibration set
out in section 2 (Fig. 2.7) appears reasonably accurate.
To concIude this section we will check how the full analysis of resonance
effects in systems can be represented by the first harmonic component of the
process. With this aim, we shall present motion on all time axes, expanding (2.27)
as a Fourier series:
~ 2(L'l- a cos<p) TIO
X = mx + L. 2 tan-cosiw{ + acos(w{ + <p), (2.61)
i~l (TIO/w)([I-(iw/O) w
nn.
where mx=(ll-acos<p)-tan-lsaconstantlort
O) .., heprocess. Usmgreatlons
. I ' h'IpS
nn 0)
(2.19) and (2.20) we can modify the expression for m< to the form
(I + R)x M
J G
m x =- (2.62)
Tc Tc c
In (2.62) J = - M(! + R)x_ is the impact impulse acting on mass M; T = 21[/0) is the
period of the process; G = J/T is the average value of the elastic force in the spring
during the period. Hence, in the vibro-impact process examined, displacement of
the middle of vibration occurs due to unsymrnetrical single-sided collisions. In
§3 Oscillator with symmetrical stops 37
practice, similar effects are revealed as the development of static components of

deflection in the operation of various measuring devices such as gyroscopes,
indicators etc., and the effect has been given the name drift. The numerical value
of drift is measured by mx '
Substituting a = 0 into (2.51), we obtain aseries, which is identical to that
given by the analysis of free vibration (2.6). We have seen that for free vibration
the backbone curves constructed for the fundamental coincide with the exact
solution (Fig. 2.3). From (2.61) it follows that the amplitude of the fundamental (i
= 1) of forced vibration is double-valued as a result of the double value of cos<p ,
hence the branches of the resonance curve are grouped close to the backbone
curve, and intersect it when cos<p = 0 .
§3. Oscillator with symmetrical stops

1. In the preceding section, we came across various effects, which occur
during mechanical vibration of simple systems when collisions develop with a
stop. A stop is regarded as a structural element of the system, which impart a
non linear character to the motion. This is characterised by ambiguous behaviour of
the system at the same values of the parameters, and considerably complicates the
"'{lI
0
I
C
l,
{lr;oSf,Jt
n
%Jffd//I7I///;;1~
Fig.3.1 Fig.3.2
vibration spectrum, by generating sub-periodic motion, etc. Of course, it is
interesting to compare the specific nonlinearity introduced by collisions with
similar nonlinearities widely studied in vibration theory. For this reason we shall
turn to the examination of an exact periodic solution, describing the vibration of a
linear oscillator with a symmetrical arrangement of elastic stops (Fig. 3.1) [40],
and retaining the basic definitions of §2.
Fig. 3.2 shows the static characteristics ofthe system studied, <D(x). To some
extent, it limits characteristics to the usual representation of the elastic behaviour
of a material with hardening. This case serves as the basis for studying the
evolution of a typical dynamic model. At the same time, the specific structure of
the system described allows us to overcome the inherent difficulty of exact analysis
ofthe amplitude-frequency characteristics ofvibro-impact systems.
We begin by studying free vibration, assuming the system as conservative.
Assuming zero time at the instant after impact on the right hand stop, we obtain for
equation (2.3) the following boundary conditions:

t=O, x=L\; t=n l wo, x=-L\. (3 .1)
Defining an arbitrary constant, with the help of (3.1) we obtain the motion of the
system in the interval between collisions
x= L\
sin[n/(2(, 0)]
. (1""\
Sill ~d--
2(, 0
n) (3.2)
where (, = wo/O. Differentiating (3 .2) we find an expression for the velocity

change
. OL\ ( n ) (3 .3)
x = - sin[ n 1(2(, 0)] cos Ot - 2(,0
From the condition x(O) = x+ ~ 0 we obtain cot[n / 2(,0};::' 0 and consequently,

(, 0 ;::. 1 . (3.4)
This result also follows immediately from a study
i / fl on ofthe phase plane for the system (Fig. 3.3).
Thus the free vibration frequency of the
Fig.3.3 Fig. 3.4

system cannot be less than the natural frequency of the linear oscillator. Since the
amplitude ofvibration is constant and equal to L\, the backbone curve ofthe system
takes the form shown as asolid line in Fig. 3.4. The vertical portion of the
backbone curve represents linear vibration of the oscillator with no impact against
the stops.
Let us analyse this process as a Fourier series. For the amplitude of the i-th
harmonic a xi we find (i = 1, 3, ... )
__ 2w o 1tl "'O
aX/ - fL\
n 0 sin[n /(2(, 0)]
. (1""\ _~)
Sill ~d
2(,0
. d _ 4L\(,0 cot[ n 1(2(,0)] (3 5)
COSZW ol t - 2 2
n(i (, 0 - 1)
• •
As a result, free vibration can be represented by the following series:

~ 4L\(, 0 cot[n 1(2(, 0)] .
x = L. .2 2 coszwot (i = I, 3, ... ) . (3.6)
i~1 n(z (,0 - 1)
Using (3.4) we can see that the series converges rapidly, hence we can look at the
expression for the amplitude ofthe fundamental
a I = 4L\(,0 cot[n / 2(,0)] (a xl ;::' 0). (3.7)
x n((,~ - l)
§3 OseilJator with symmetrie al stops 39
This relationship is shown dotted in Fig. 3.4. As in the case of a system with one
stop, the backbone curve for the fundamental is very similar to the full
characteristics of the system. In particular, as 1;0 increases, the value ofaxi rapidly
approaches to the constant value 8l7tz.
2. Let us turn to the examination of forced vibration of systems due to a
periodic force Pcosw t taking identical coefficients of restitution for both stops and
describing the continuous energy dissipation of the system by means of a viscous
damper whose action is proportional to the velocity of a reduced mass, through b
(Fig.3.1 ).
In the interval between collisions, the motion ofthe system can be described by
the differential equation
x + bx + OZx = al' cos(mt + <Po)' (3.8)
where <PI! is the unknown initial phase of excitation, b = n / M, a p = P / M , we
find as in the case ofthe single-sided stop, the periodic solution
x = a cosm{ (3.9)
Here
a= al' al' (r=b/(20», (3.10)

J(OZ - wZ)z + bZw z OZ J(1- SZ)2 + 4rzsz
.
sm<po =
2rs
. cos<po
I-s z
=-,=========
~(1- I:/)z + 4r2sZ . ~(1- S2)Z + 4rZs2
This solution corresponds to the motion of the system without impact occurring
naturally when a < Ll .
To find the vibro-impact behaviour, we can write the general solution for
equation (3.8) as
x = exp( -rnt)(C I cosOlt + C z sin Oll) + acos(wt + <p) . (3.11)
Here 0 1 = O~, CI' Cz are arbitrary constants: <p is the phase of the force
and depends on the choice of starting condition. The general solution written in the
form (3.11) assurnes that r < 1. The basic effects of vibro-impact behaviour, which
we shall study for the remainder of this book, become more obvious as the result
of small values of the viscous damping coefficient. This allows us to simplify the
representation by taking 0 I = 0 .
We will study the periodic regimes characterised by alternate impacts of the
mass against the stops. Because of symmetry of the system, the period of motion
consists of the sum of two identical intervals, each of which covers the time
intervals between two consecutive collisions. Consequently, to get a fuH picture of
the motion, it is sufficient to find the law of motion for one of these intervals. For
analysis we will look at the interval beginning after the impact on the right hand
stop (Fig. 3.1). Then the boundary condition of the regimes for various ratios I
(I = 1,3, ... ) takes the form
t = 0, x = L'l, x = -RX_; (3.12)

t=rcl/oo,x=-L'l, x=-x_.
Substituting boundary conditions (3.12) into (3.11) we can express the arbitrary
constants and the phase, in terms of system parameters. As a result the motion in
the regimes examined will be as folIows:
x = exp(-rOt) ( cosOt- exp r3. + cos 3.smnt ) -DR

x_
.
+
sm3 0 (3.13)
+ acos(oot + <j» (t E [0, rcl / 00],)
where, similar to §2
. BD .
sm<j>=-x_, cos<j> = I - - x (D .. ) L'la (3.14)
aO OL'l
. [1±~I-(l-a2 /L'l2)(l+ B 2 )]OL'l
x = ----"-----...,----- (3.15)
- D(l + B 2 )
Here we make use ofthe following non-dimensional terms:
B = (~_ expr3 + cos3 + rsin 3)~
D sin 3 t:;'
(3.16)
D= (1 + R) sin 3 3 = rcl
expr3 + 2cos3 + exp( -r3)' S
3. As long as the amplitude ofthe vibro-impact systems studied is constant,
a x = L'l, the structure of the amplitude-frequency characteristics is defined by
existence domains for regimes in the S, L'l plane. We shall study the boundaries of
the areas of existence. To begin with, we shall isolate the areas of validity for the
value of x_. Assuming that the term under the radical in (3.15) is non-negative, we
obtain similarly to (2.22)
(3.17)
As we found already in §2 areas of validity are wider than existence domains
since the structure of the model is based on the fact that in between collisions the
mass must be situated between the stops, not exceeding the level 1x 1= L'l (i.e., with
no additional intersections) which was not reflected in the previous analysis and
consequently in (3.17). Meanwhile, this limitation is not satisfied with the given
model for all values of the system parameters, hence later we shall study at length
the effect of this additional limit on the results of the analysis, but now we shall
°,
examine some basic dynamic features, resulting from inequality (3.17).
Since in the basic case B"* then, similarly to the situation described in §2,
vibro-impact regimes of systems can exist not only when the inter-stop distance is
less than the amplitude of the linear oscillator (L'l< a) but also when the gap is
expanded to a value higher than the linear vibration of the system (L'l> a) up to
the value identified with the equals sign in (3.17). Naturally, when L'l > a the same
tuning results in harmonie vibration (3.9) ofthe linear system, hence development
into a vibro-impact regime can occur as the result of an additional impulse.
§3 Oscillator with symmetrie al stops 41
Within the domains 01' real values, we shall separate existence domains for this
study, i.e., by combining the system parameters during which the law 01' motion in
these regimes satisfies the conditions for the absence 01' additional intersections
IX(l)l::::tl. (3.18)
In the general case, it is not possible to find the analytic bounds 01' the areas 01'
existence from the general condition (3.18) for the given model, since the law 01'
motion is written as the sum 01' components 01' harmonic functions with non-
multiple frequency. Instead, as was shown in §2, for systems with reasonably
simple structures, analytic methods can expose the approximate boundaries 01' the
existence domains, and we will use these here.
According to (3.12) for the model examined
x_ :?: O. (3.19)
Turning to expression (3.15) we find that when tl < Cl one, and only one 01' the two
values 01' satisfies condition (3.19) x. Hence, for th is model due to the
arrangement 01' stops, within the limits 01' vibration of the linear system only one
regime can exist for each ratio studied. Noting that the term D (see(3.16» depends
on the sign 01' sin 3 we obtain that when
C, :?: f (3.20)
these regimes correspond to the positive sign, and when C, < I on the minus sign.
We will now look at how condition (3.19) influences the areas 01' existence 01'
vibro-impact regimes tor which tl:?: a . According to
(3.15) in this case the expression below the radical is
N-f} $ r - N
less than or equal to unity and consequently both
1-/
values 01' ~y take the sign of the coefficient in the
denominator, in other words the sign of the value of
sin 3. With the requirement that sin 3 :?: 0 , we again
Q
co me to the condition (3.20).
Consequently, in the frequency range (3.20) for
"lI every point in the area 01' real values, satisf)ring the
I
condition tl:?: a two periodic vibro-impact regimes
can exist with ratios ti'om I to I inclusive. Note that
when tl = CI, one 01' the values of X is equal to zero
independently 01' the other system parameters. This
Flg. represents linear vibration 01' the oscillator just
touching the stops but without impact.
From the backbone curves for the system (Fig. 3.4) we can convince ourselves
that inequality (3.20) represents the condition when the excitation frequency
coincides with one 01' the natural frequencies 01' the vibro-impact system, and this
correspondence can be interpreted as resonance. Later, such a treatment will
receive a fuller grounding and will be developed effectively.
Fig. 3.5 shows examples of cross-sections of the parameter space with the
existence domains hatched. The sections are constructed using (3.17) for constant
values of R, r, and using I = 1 and 1= 3. Here the area hatched with positive slope
lines denotes the existence domains with the positive sign in (3.15), with opposite
hatehing for the negative sign. The closer spaeed hatehing when LI, > a shows the
existenee of eoupled regimes.
4. Interest is naturally aroused as to the extent to whieh the qualitative
analysis obtained is preserved when eonstrueting additional boundaries for
existenee domains on the basis of (3.18). We will study these boundaries for the
ease R = 1, r = O.
Based on (3.13) to (3.15), the laws of motion will be written for this ease in the
form
x= LI, ± a
sin[ nl f(2s)]
.
sm (r.
~d--
nl ) +acos (oot+<P )
2s
(t E [0, nl/ 00]) (3.21 )
in which <p =n with the positive sign and <p =0 with the negative sign. Hence, the
movement of the system in the intervals between impacts is represented by the sum
of portions of two harmonie funetions, whieh are anti-symmetrieal values relative
to t = TCl f(2oo) .
Initially, let us look at motion when s< I . In the range of t examined , both
harmonie funetions reaeh extremes. In general, these extremes are shifted in time,
however, we will assurne for eonvenienee in the situation of absence of additional
interseetions, the ease when both extremes oecur simultaneously and with opposite
phase. At this instant
I
lxi = Isin[LI,±a
nl f(2S)] - a .
(3.22)
Earlier, it was shown that in the frequency range examined only regimes with
the negative sign ean exist satisfying the eondition LI, < a . Then, from (3.22) we
find lxi> LI, . Henee, when S < 1 regimes with alternate impacts on the stops do not
oceur.
At the same time, when LI, < a, the stops present an obstacle to the
development of harmonie vibration, whieh leads to the development of eollisions,
meaning that the system develops vibro-impaet regimes of a different type. From
Fig.2.9, a, it can be seen that these can be regimes with exeitation frequeney and
eharacterised by impacts on one stop only.
We shall now study the effeets of equation (3.18) on the frequeney domains
satisfying (3.20). Initially we note that, studying the stability of these regimes as
earried out in the previous seetion, for the given model the requirements of
stability are also satisfied by regimes with the plus sign before the radical in (3.15)
and properly in (3.21), hence, we shall now only study such regimes. This is a
typical feature ofnonlinear resonance, which we will regularly come aeross.
When S < I in the examined time interval t, the first harmonie term of (3.21)
deereases monotonieally (Fig. 3.6) and the second term depends on I. When 1 = 1,
this term grows monotonieally whieh ean mean violation of the eondition x ~ -LI, .
Making the situation worse, let us replaee the harmonie terms J, 2 by the linear
terms 3, 4. Obviously, if the eonditions for the absence of additional interseetions
are satisfied for the linear approximation when t = 0, then they are known to be
satisfied for the harmonie terms for the arbitrary value t eonsidered. Consequently,
§3 Oseill;:tor with symmetrie al stops 43
we can put 11 + a - na /2;::: -11 , whence

l1/a;:::(n-2)/4",,0.29. (3.23)
Hence, even for such artificial narrowing, the existence domains for basic vibro-
impact regimes found (l = 1) cover almost all pre-resonance areas of the parameter
range. Note that when condition (3.23) is not satisfied and the stops are arranged
with relatively sm all gap, the character of motion is preserved; however, the
methods used here are of Iittle use for general estimation, since fluctuating effects
occur according to the value of the basic displacement. For this reason, inequality
(3.23), which guarantees the existence of the regimes examined, will be taken
,-, I-J 1-5
Fig.3.6
conditionally as the boundary of an existence regime.

When I = 3, 5, ... the first harmonic component does not change, while the
second oscillates, and achieves a positive value of amplitude (Fig. 3.6). As a result,
when 11 < a , condition (3.18) is violated and for this reason sub-periodic regimes
do not occur here. When 11;::: a in the case when condition (3.18) is violated, this
happens at the boundary x = 11. We can limit our study to the
segment! E [0, n/co]. As long as condition (3.18) is satisfied when t = 0, it is
sufficient that in the examined segment the second term of the harmonie does not
grow so fast that it extinguishes the first. Such a requirement can be majorised by
two conditions, the modulus of the derivative of the first harmonic term when
t = 0, must not be less than the modulus of the slope of line 3 (this guarantees
condition (3.19) in the portion tE [0, n/co]), but when t = n/(2co) it must not be
less than the derivative of the second term (this guarantees (3.18) when
tE [n/(2co), n/co]). The first condition gives
11 + a Q> 2aco
tan[ nl/(2s)] - n '
or after conversion
(3.24)
We will now examine the second condition at the instant t = n/(2co) . It takes
the form
11 + a Q cos n(l - 1) ;::: aco .
sin[nl/(2s)] 2s
Assuming s » 1 we obtain, putting cos(O) =1

L1r. nl
-2::~sm--l. (3.25)
a 2S
Using (3.20) conditions (3.24), (3.25) can be joined together in the majorising
relationship:
L1 nl
-2::stan--I, (3.26)
a 2S
wh ich we will take as a boundary for the existence of the sub-periodic vibro-
impact regime. Beginning with some value S > I, condition (3.26) is always
satisfied, while the boundaries corresponding to various values of 1 > 1 are
arranged one above the other, higher than the level L1 / a = 1 , and do not intersect
each other..
5. Study of the existence domains on the basis of condition (3.18) allowed
us to refine the boundaries and to detail the structure of the system parameters
extensively, hence, no matter how laborious this procedure such investigation is an
important element of analysis to find accurate solutions. From this springs a
question, in what measure are the results obtained when R = 1, r = 0, pertinent to a
real system that inc1udes dissipative effects?
First of all we notice that if inequality (3.17) defining the dissipative effects of
the system gave us the upper limit for the area of existence for frequency of
excitation Sand amplitude L1, then condition (3.18) or its more particular form
(3.19) lead to the appearance of the lower boundaries (3.20), (3.23), (3.26). This
reflects various physical characteristics ofthe limits specified.
Inequality (3.17) defines the energy requirements of a system to cause
vibration. Hence, any solutions satisfying the differential equation (3.8) and
boundary conditions (3.13) and also with additional intersections satisfy (3.17)
because they bring about the conditions of energy balance. Therefore, it is natural
that increasing dissipation of energy in the
system affects only the upper limits by the
~ H -~ r~Q.f1 swing and rapidity ofvibration. Y I Neimark
J \'
/ showed that for the general system this
l,t H- tJ-.f r-ß·f boundary always coincides with one of the
~" ~ r-- boundaries of stability (corresponding to the
i)-. l'- ~
o
H-I).5 r-tl-Z -
-f-..
J ~
positive unit root when studying the roots of
the characteristic equation inside a unit
z
Fig. 3.7
circ1e) [148]. In future we shall refer to this
as the energy boundary 0/ vibration.
Conversely, condition (3.18) produces the structurallimitations on the defined
vibration trajectory (with additional intersections) and does not, in principle,
prevent an increase in the amplitude and frequency of vibration. The boundaries
(3.20), (3.23), (3.26) obtained on the basis of (3.18) represent existence domains
of vibro-impact regimes, and will be referred to as the structural boundaries 0/
vibration.
To estimate the influence of dissipative factors on the structural boundaries we
§3 Oscillator with symmetrie al stops 45
shall examine in more detail the specific law of motion of the system. Comparing
(3.21) with (3.2) we see that the first harmonic term in (3.21) is similar to the free
vibration of a conservative system with frequency C:;O = C:;/ [, investigated in
section 1, differing only in the value a. As c:; increases from unity, then, according
to (3.10), the value of a rapidly decreases and tends to zero, hence the basic
motion of the system (3.21) becomes more and more like a pure conservative
system, but the law of motion becomes more and more saw-toothed. Analysing
(3.13) it can be shown that there is a similar tendency also when taking ac count of
the dissipation of energy in the system due to the compensating energy loss of the
extern al excitation. This confirms the re sonant character of the vibro-impact
regimes studied in section 3. As a result, accounting for sm all dissipation during
vibratory motion, cannot change the qualitative character of conditions (3.20),
(3.23), (3.26), concerning high frequency motion. The specific result of (3.22)
shows that similar concIusions can be obtained with regard to low frequency
motion. Calculated results gives good confirmation ofthese concIusions.
Hence, dissipative effects of systems, in practice, influence only the energy
boundaries of vibration (3.17). We will study this in more detail. Fig. 3.7 shows
examples of these boundaries constructed for [= 1, for various values of the
coefficients of viscous and impact dissipation of energy. It can be seen that both
mechanisms for energy dissipation give numerically identical configurations to the
energy boundary of vibration, decreasing monotonically as excitation frequency
increases. This result allows us to make an equivalent substitution of one
mechanism of dissipation for the other, a process we will practice widely when
presenting approximate methods.
In addition, it is useful to notice that the influences of viscous and impact
dissipation have several special features. In particular, as c:; / [ increases towards
un ity, according to (3.16), B -+and consequently condition (3.17) gives
OCJ ,
6. / a ~ 1 (3.27)
which is independent of R. This follows from the fact that the regimes on the
boundaries are impact free (see section 3). When we neglect viscous friction
(r = 0). we find the limit of B equal to zero, and from condition (3.17)
notwithstanding (3.27) we obtain 6. / a < OCJ • Consequently, cIose to this boundary
for the frequency range for the existence of vibro-impact regimes, accounting for
viscous damping remains important but impact dissipation can be neglected.
However, study of effects in this area is of little interest, because of the sluggish
impact regimes in these areas.
For the more important high frequency area, tending to the limit when c:; -+ OCJ
in (3.16), we find from (3.17)
6. rc 2[2(1+R)2
lim - ~ 1 + (3.28)
C--.CD a 4(1- R)2
The right hand side is, of course, finite only when R *- land does not depend on r,
i.e., impact dissipation at high frequency appears stronger than viscous dissipation.
For this reason, as we search for the exact solutions of re sonant motion, we will
concentrate mainly on impact dissipation of energy.
Expanding (3.16) as aseries in terms of the small parameter rand limiting the
expansion to linear terms we obtain directly the equation of condition (3.17) in
terms ofbasic system parameters in the following form:
~:S 1+ ( Ssin(rcl / s) )2 (3.29)

a (1- R)[I + cos(rcl / S)] /(1 + R) + E
where
E =r ( ~ . ~
rcl + sm rcl )/( I + cos 2S
rcl ) . (3.30)
From (3.29) it can be seen that as the energy dissipation in the system
increases, so the energy boundary of vibration reduces steadily. However, even for
the limiting value of impact dissipation corresponding to R = 0 this boundary
exceeds the amplitude of non-impact vibration a for all values of the ratio 1 for all
frequency ranges in existing regimes.
Comparing (3.29) with (3.23), (3.26) and using the fact that
tan ;~ = sin? /(1 + cos ?) , (3.31)
we can establish that for domains of any ratio, beginning with some value S > 1 ,
the energy boundary must be above the structural boundary which leads to the
existence of a regime.
In section 4, we have already mentioned
that the lower boundaries for existence
R- N r - P.f
domains of sub-periodic regimes for
various ratios (see (3.26» are disposed
above one another with increase of I, and
do not intersect one another for all 1
frequency ranges (3.20) while in the whole

of this range Ll / a > 1 for any I > I. This p.
implies that it is not feasible to transfer
directly from harmonic vibration to the
sub-periodic regimes, whatever the
frequency ratio, and requires the use of
additional energy for the starting process
(hard excitation). Soft transfer from
harmonic vibration to vibro-impact can
occur only for the basic regime (l = 1) c.
according to (3.23), because their structural
boundaries are very close to zero.
Fig 3.8
However, in agreement with (3.29), as R
increases the existence domains of vibro-
impact regimes for various ratios overlap and consequently, due to the small
changes in the system parameters the regimes can transfer continuously from one
§3 Oseillator with symmetrie al stops 47
to another. In areas of overlap, establishment of one or other regime depends on

the character ofthe preceding transfer process.
Hence, an investigation of the influence of dissipation shows that it
considerably decreases the size of existence domains for vibro-impact regimes, but
it cannot completely eliminate them. This explains why it is so difficult to combat
collisions in machines, which occurs in systems of vibration isolation, bearing
assemblies and various other types of mechanical structures. In such cases, when it
is occurs in actual assemblies, then according to (3 .30), an effective measure to
eliminate collisions can be to increase the gap L'1 beyond the limits ofthe existence
domains .
Fig. 3.8 shows examples of existence domains for regimes on the basis of
equations (3.20), (3.23), (3.26) and (3 .29). It is convenient to show them with the
help ofthe basic amplitude-frequency characteristic corresponding to linear system
I . These areas are shown hatched with positive slope, while those with negative
slope hatching are the existence domains for vibro-impact regimes having the same
frequency as the excitation and characterised by impact on one stop only with
amplitude a , < L'1 (see section 4). The frequency range for existence ofthis area is
plotted from the results obtained in §2. Naturally, depending on the transfer
process such a regime remains on one or other of the stops.
6. Transferring to the s,a x 0 2 / ap plane, and drawing a horizontal straight
line corresponding with the level L'1 of the stops, together with the existence
domains for vibro-impact (Fig. 3.9, a), we produce the amplitude-frequency
characteristics of the system. In Fig. 3.9, a, this is indicated with asolid line
corresponding to the basic regime. On the horizontal portion 1 a symmetrical
vibro-impact regime exists having alternate collisions of the mass against each
R- U$ ,. - O·f
-·==n
I
I
I
Q
a} h)
Fig. 3.9
stop. In the pre-resonance frequency zone (s < 1), branch 2 of the frequency-
amplitude characteristics is constructed from the results of §2, for an actual system
ofvibro-impact regimes with frequency equal to that ofthe excitation and colliding
against one stop. Moreover, the added branches 3, 4, representing the amplitude
characteristics for harmonie vibration of a linear oscillator, lie below the level of
the stops.
From Fig. 3.9, a, it can be seen that as the exciting frequency is gradually
increased (see arrows) from the zero regime of the system, passing through the
frequency zone corresponding to harmonic vibration, then after the first collisions,
the regime changes initially to one with single- sided impact. Subsequently, when
it passes through the resonant frequency of a linear oscillator, a symmetrical vibro-
impact regime is established, with alternate collisions on each stop, which can be
prolonged in high frequency domains until the system reaches the existence
boundary. When the system frequency intersects the boundary values, the vibrator
"breaks-away" from the stops and once again harmonic vibration becomes
established for all subsequent frequency ranges.
A different process occurs as the frequency is decreased from a high value
Here the vibro-impact regime is established at a frequency lower than the
breakaway frequency and then the system develops in the opposite direction to its
amplitude-frequency characteristics. Note that in this case vibro-impact regimes
can exist only within the limits of the frequency range in which the stops obstruct
harmonic vibration of a linear oscillator.
Thus, during run-up to the high frequency zone, basic vibro-impact regimes
occur more widely than during run-down, i.e., as was to be expected, a similar
pulling occurs as for systems of the nonlinear hard type. At the same time, in
contrast to systems with smooth nonlinearity, in vibro-impact systems, linear and
nonlinear vibration is clearly separated. Transfer from regime to regime in a
pulling field can only happen due to special additional excitation.
As weil as the description of a pulling effect in regimes due to jrequency, in
vibro-impact systems it is also possible to obtain pulling effects due to amplitude.
In the system studied this can occur for every excitation frequency which exceeds
the natural frequency ofthe linear oscillator (e.g., C; in Fig. 3.9, a) by means of
the introduction ofstops in the harmonic vibration zone ofthe oscillator (Fig. 3.9,
b) before a vibro-impact regime becomes established and then slowly expanding
the gap. As a result the vibro-impact regime is maintained up to the value of the
amplitude at breakaway, when the stops reach the level defining the boundaries for
the existence domains. After breakaway in the system, the original harmonie
vibration is re-established.
Specifically, the structure of the system leads to the fact that its amplitude-
frequency characteristics (Fig. 3.9, a), coincide with the backbone curves (Fig.
3.4), within the limits of the existence domains of symmetrical regimes .
Moreover, in zones of pulling (~> a), each point of the characteristic exists in
two regimes with different collision velocities, in agreement with the results of
section 3. We have already mentioned that, as is usual for non linear systems in
double-valued areas, the amplitude-frequency characteristics corresponding to
lesser velocity are unstable. In § 14, we will show that this condition is typical for a
large number ofbasic systems.
Let us consider the behaviour of the fundamental for systems in the
symmetrical regimes studied. For simplification, we will ignore energy dissipation
in the system. Expanding (3.21) as a Fourier series, we obtain, similarly to (3.5),
(3.6)
§3 Osciliator with symmetrieal stops 49
-----~ ..._ - - - - - - - ----
=[ S
4 (t. ± a)cot[nl 1(2S)] _ ]
x 2 + a cos rot +
n/(s -1)
(3.32)
~ 4S(t. ± a)cot[nl 1(2S)] iro
+ ~ cos-t (i = 1, 3, ... ).
;=1 nl[(isll)2 -1] I
,."
When s> I the value a tends to zero, as does I I S2, and the series (3.32)
Fig. 3.10
converges as rapidly as (3.6) in practice. Consequently, independent of the
existence of the nonlinearity of the system when the periodie regime oeeurs it
causes a filtering effect on the fundamental Ci = 1) .
The amplitude of the first harmonie when I = I is found from (3.32) after re-
arrangement as
a
xl
= 4t.Scot[nl(2S)]
n(s2 -I)
± a[4Scot[nl(2S)]
n(s2 -1)
1]. (3.33)
Comparing (3.33) with (3.7) we see that the two branehes of the amplitude-
frequeney eharaeteristies for the fundamental are loeated on either side close to the
backbone eurve, and rapidly eonverge to it as S is inereased (see (3.10)). Sinee
existence domains for the fundamental coincide naturally with the calculated
values for the existence domain already determined, then the study of the first
harmonie gives all the required qualitative information relevant to the process
studied. In the next chapters we shall try to obtain this same information using
approximate methods.
Let us find the equation for the gain eoeffieient of the fundamental due to
harmonie exeitation. Denoting this by k, we find from (3 .32) using (3.10)
k =!!.E... = 4t.scot[nl/(2s)] + 1 [4 s cot[nll(2S)] -1] (3.34)
ap nla p (s2 _/ 2) - ro 2 _n 2 nl(s2 _/ 2)
As is usual in nonlinear systems, this coefficient depends on both amplitude
and frequency of vibration. From (3.34) it follows that in the frequeney range of
existenee for the solution s>
I, the gain eoeffieient ean be reasonably well
approximated by its first term. Thus, when studying vibro-impaet processes of the
type discussed, the value of the gain coeffieient due to harmonie exeitation in the
high frequeney range ean be obtained with suffieient aeeuraey by analysing free
vibration of the system. This will be explained in detail in §9, during the
presentation of suitable approximate proeedures.

To examine the behaviour of solutions during transient processes, a study was
made of a simulation of the system under eonsideration [43]. Fig. 3.10 shows
vibration traees from the model as the frequeney of exeitation of the system was
gradually inereased and, superposed on it, the vibration during sub se quent
deereasing frequeney. The pulling in the high frequeney areas ean be seen clearly
as weIl as the single sided eharaeter of eollisions in pre-resonant zones.
CaIculation of the eollision velocity eonfirmed that when co > n, vibration oeeurs
in all regions eorresponding to the positive sign in (3.15).
To study the transition from the harmonie regime to vibro-impaet, the domains
of attraetion were eonstrueted on the phase plane inside the pulling zone. Fig. 3.11
shows such a eonstruetion for point A, of the plane of parameters, (see Fig. 3.8). In
this ease, harmonie vibration and basic vibro-impaet regimes ex ist, the limiting
eycle of eaeh being marked in Fig.3.11 by the numbers 1 and 2. The proeedure for
eonstructing the zones of attraetion was to establish the limiting eycle of a
transition proeess during start-up of the system from various starting eonditions
eorresponding to points on the sealed grid applied to the phase plane. The initial
phase of exeitation during all start-ups was identieal (zero in the given example).
From Fig. 3.11 it ean be seen that, when the starting impulse exeeeds eertain
Flg, :I 11
FIg, 3 12
"l{ll
Q
i"--I
" 'I:
(I
;~
0'
'7/ IJ "b
g 1_. __ -
.J
hg .114
threshold values, it always lead to the establishment of a vibro-impaet regime. Fig.

3.12 shows an example ofthe phase trajeetory for such a transition.
Fig. 3.13 shows examples recorded for a model of various aetual regimes.
§3 Oscillator with symmetrical stops 51
These all correspond to identical rrequency of excitation. Here 1 represents

harmonic vibration, 2 represents the basic vibro-impact regime, 3 represents the
sub-periodic regime with rrequency ratio 3. The transition of systems into vibro-
impact regimes occurs as the result of a starting impulse.
Summarising the analysis carried out, it can be said that the dynamic model
studied is the limiting case of a hardening type of nonlinear system not only in
terms of structure but also in function. For comparison, Fig. 3.14 shows various
static elastic characteristics of the hardening type, and the amplitude-rrequency
characteristics resulting rrom them. In this figure, unstable branches are shown
dotted, i.e., as already mentioned, , the resonance with the highest velocity of the
two possible resonances is the one that becomes established. Similar results were
obtained in the systems studied. The cause of such similarity can be understood if
one takes into account the change rrom elastic characteristics of Fig. 14, b to the
characteristics in Fig. 3.14, c corresponds to the smooth increase in the stiffness of
the edge portions . As the analysis shows, the sloping portions of the backbone
curves gradually lengthen, the steepening portions move away in the direction of
higher rrequency, in the limit tending to infinity, but double-valued branches ofthe
resonance curves approach more c\osely to the backbone line. It is significant that
the appearance of nonlinearity beyond the limits of small effects does not lead to a
qualitative transformation of the system behaviour. Hence, no matter what
expressions are specified, the basic character of nonlinear behaviour for vibro-
impact systems, obtained using exact methods, are similar to "basic" non linear
hardening systems, showing the success possible with approximate methods. This
analogy leads to the conc\usion about the usefulness of applying approximate
methods for the analysis ofvibro-impact systems. In §9, the model studied will be
investigated using a harmonic linearisation technique and the results obtained will
be compared with the results of exact analysis.
7. The effects of pulling of vibro-impact regimes described here were studied in an

experimental device designed for the high frequency system of photoelectric scanning [26]. Fig. 3.15
shows the experimental arrangement. The be am I, fixed at 4. passes through the gap between two
electromagnets 2, which is supplied with altemating current from a generator. On the end of the beam
a tixed ball 3 is installed in the gap between stops 5. The stops were designed as a slot narrowing in
the direction perpendicular to the picture plane. The
beam and the excitation system were assembled on a
single body, arranged so that they could be moved
along the slot through an angle formed by the stops to
change the size of the gap. The vibration of the ball
was registered by a wire strain gauge whose signal
was recorded lIsing an oscillograph.
Fig. 3.16 shows several typical oscillograms
obtained [rom the vibrating system with a natural Fig.3.15
linear freqllency 0./2n", 750Hz. On the
oscillograms the central line shows the signal from the strain gauge, which is proportional to the
displacement of the ball from central symmetry, and the lower curve shows a timebase with a
frequency of 500 Hz.
The two antinodes in Fig. 3.16, a represent the uniform increase of the exciting frequency and the
subsequent uniform decrease. An example of the frequency scale of the excitation is given above the
oscillogram. The frequency curve was recorded with a stationary gap !'1 =5 x 10-4 m . Various lengths
of antinode demonstrate various pulling of the vibro-impact regime as the frequency of excitation was
increased. The beat at the end of the oscillogram occurs after the breakdown of the vibro-impact
regime into a mixture of damped free vibration and forced vibration at the fundamental. At the ends of
the oscillogram, vibration without collision is recorded.
730Hz 760 880 810 780 710Hz
Il)
..
b)
0.45mm 0.3 1.2mm
c)
Fig. :UO
Fig. 3.16 b shows the results near the breakdown point at a frequeney w/(2n) = 880Hz in the
vibro-impact and harmonie regimes, into which the system ehanges as a rcsult of breakdown.
Comparison with the oscillogram in Fig. 3.16 b shows that the breakdown from the vibro-impaet
regime to harmonie represents a fivetold deerease in amplitude.
The oscillogram in Fig. 3.16 c was recorded at a fixed trequency w/(2n) = 880Hz and represents
amplitude pulling whieh happens in the system when the gap between the stops initially starts to
deerease to a value at which vibro-impaet is triggered. This instant is eharacterised by a sudden peak
in the left-hand part of the oscillogram as the gap is further decreased to !'1 = 3 x I 0- 4 m and then
inereased up to !'1 = 12 x 10-4 m . A vibro-impaet regime eontinues to hold in the system. After
breakdown of this regime as it eros ses the boundaries 01' the existenee domains (Fig. 3.7, a) the
original form of harmonie vibration reappears. Thus, as a result of a gradual inerease in the distanee
between the stops, the system is pulled to a regime whose amplitude is several limes greater than the
regime without collisions.
By measuring the frequency of breakdown of vibro-impael regimes for various values of gap it is
possible to construet a boundary for their existenee domains (Fig. 3.17). Its eonfiguration corresponds
to the energy boundary ofvibration (3.29) (see Fig. 3.7)
§4 Principles of percussion machine dynamics 53
§4. Principles of percussion machine dynamics

I. Percussion machines are widely used in various fields of technology,
particularly for processes that inc1ude demolition or deformation of workpiece or
materials. Their principal advantage is the simple production of large forces
arising from the collision of a machine
striker with workpiece either directly onto
the workpiece or onto a tool which
transforms impact energy into a working
process.
The dynamic system for percussion
Fig.4.1 must inc1ude a drive arrangement, the
models of the vibro-impact mechanism
and the working process. The model of
impact interaction, various examples of which were studied in § I, describes the
character of the work process. The character of the drive produces a significant
influence on the dynamics of the system, interactions between the vibrator and its
f C, Cz
Fig.4.2
drive will be studied in § 12 (see sections 3 to 7). Here we will examine the basic
dynamics of a typical two-mass model of a percussion machine (Fig. 4.1) taking
into account the mass of the striker and body with connecting elements isolated
from the striker by an elastic spring. For simplicity we will assurne that the drive
provides periodic excitation of the system independent of its vibration, while the
working process (impact) can be described kinematically using stereomechanic
theory. This allows us to develop the analysis with exact methods and to evaluate
the effect introduced to take account of the mass of the body on the dynamics of
the vibro-impact mechanism by comparison to the more highly idealised model
examined in §2 (Fig. 2.1), also used to model percussion machines.
As an example, Fig. 4.2 shows schematically an actual example of a percussion
machine. Striker I assisted by spring CI connected to body 2, while spring c2 is
attached to the piston rod of a simple crankshaft driven mechanism. The whole
device is pressed against a stop by static force G (the feed). We shall assurne that
the driving device for the crank shaft supplies it with a constant angular velocity
co , while the motion of the piston rod 3 relative to the body is simple harmonic
motion ofthe form x 3 =rcos(ro!+<p), where r is the crank radius. Neglecting
dissipative forces we can write the differential equations ofmotion in the form
Mix i =-CI(X I -X 2)-C 2[X I -(x 2 + x))],

M 2 x 2 =-C I (X 2 -X I )-C 2 [(X 2 +xJ)-xl]+G,
where XI ,X 2 are coordinates ofthe striker and body: MI' M 2' - their masses.
Writing C = CI + c2 ' P = c2 r , we obtain
Mix i = CI (x 2 - XI) + Pcos(wt + <p), (4.1)
(4.2)
Thus, performance analysis of this machine leads to the study of the dynamic
model shown in Fig. 4.1, in which vibration is excited by harmonie forces acting
with opposite phase on both masses. The study of similar models leads to the
modelling of a wide range of percussion machines.
If mass M 2 is sufficiently large, then its vibration can be neglected. Then
according to (4.2) x2 --70 when M 2 -+ 00. It would appear that the model
described is identical to that of an oscillator with a stop studied earlier in §2.
However, this is not so.
The characteristic feature of the model consists of the following. In the static
condition under the influence of force G, mass M 2 can take up some equilibrium
position and creates astrain L\ = -G / C in the system. Let us assurne that under the
influence of an excitation force, the system undergoes vibration with period
T = 2nl / 00 , during which time mass MI collides once with the stop. In this case,
equilibrium must be satisfied between an impulse of force G for the period and the
increment of momentum, i.e.
GT = MI (x I_ - x I+) . (4.3)
From this, using the relationship between the velocities xI+ = - Rx l _, we can
immediately find the value ofthe pre-impact velocity
x
I _ = GT /[M I (I + R)] . (4.4)
From this expression it can be seen that velocity xI _ does not depend on the
force of excitation. Thus, if mass M2 is large, it takes some position of dynamic
equilibrium, differing from the static value mx2 and ofvalue
L\ = -(ma + G / c) (4.5)
similar to the gap (interference) in the system studied in §2. Thus, here the gap
does not have a fixed value, but changes according to the vibration regime. This
predetermines aseries of specific features for the behaviour of systems. Weshall
examine several of these for the case M 2 -+ 00 , using the solutions of §2.
During natural vibration ofan oscillator impacting on a stop, according to (2.4)
the velocity is
(4.6)
Assuming in (4.4) that T = 2n / Wo and R = 1 , after substitution in (4.6) we obtain
the following value for the equivalent gap:
L\ = -n G ~col n~). (4.7)

C Wo \, Wo
§4 Principles ofpercussion machine dynamics 55
Substituting (4.7) in equation (2.11) we obtain the equation of the backbone curve
a
x, = ~~tan(n~J.
2c 00 200
0 0
(4.8)
This is shown as asolid line in Fig. 4.3.

The equivalent gap for forced vibration is found by solving equation (2.21) for
the value of ~ :
xl_D I I 2 2 ·2
~=n±2oo V(2aoo) -(l-R) XI (4.9)
Assuming for the sub-radical quantity in (4.9)

2aoo - (I - R)< ?: 0
and using (2.15) and (4.4) when T = 2nl / 00 we obtain the condition defining the
existence domains for the value of ~, as
G 1 (l + R) 00 2
-<
P -ni (l-R) In 2 _0021.
(4.10)
Obviously, of all the solutions satisfying (4.10) the system can achieve only those
when the gap (interference) ~?: G / c. Making (4.9) comply with this condition
and limiting the value of R to R = I , we obtain
(4.11 )
The analysis shows that for regimes with frequency ratio I = 1 , both values of
(4.9) satisfy condition (4.11) only in the range
,., 00 > n , and consequently
G/P>1I2, (4.12)
but only one of the solutions lies in the remaining
parts of the area (4.10). Regimes with high ratios
always exist in the frequency range
-r _ _ _ _
oo>/n. (4.13)
I
Fig. 4.4 shows existence domains ofthe solution for
Fig.4.3
frequency ratios I = I, 2, corresponding to the
inequalities (4.10), (4.13). Note that existence
domains for regimes ofhigh ratios completely overlap those oflower ratio.
The resonance curves for values of the amplitude of forced vibration, Fig. 4.5,
are constructed using the same method used to construct Fig. 2.7. Here the thin
lines show resonance curves when R = 1, constructed using the backbone curve
equation (4.8) using expression (2.35). The portions of the resonance curves
shown by thick lines arise when energy dissipation occurs during collisions
*'
(R 1) . The limiting frequency for the development of vibro-impact behaviour is
found from the boundaries of the existence domains (Fig. 4.4). Here, we meet an
interesting feature of the systems examined: it appears that neither the increase in
dissipation during impact, nor increase in clamping force can eliminate vibration
with infinite amplitude at the fundamental resonance frequency 00 = n . (Later we
will show that the limit arises because of viscous damping.) The vibrations shown
can also occur at multiple frequencies 00 = In, however, they can be is suppressed.
In agreement with (4.10), sub-periodic resonances occur when

G < _(I_+_R_) _ _1_
(4.14)
P - (1- R) 1t(l2 - 1) .
R-IN C/P-UO
':~/I-l
,-,
f
f 1 ~
-':=:==Fi41:::/f}
Fig.4.4 Fig.4.5
2. Let us now consider the general case of vibration of the system under
investigation (4.1) described by the complete set of equations (4.1), (4.2). We shall
limit this study to the basic periodic regimes when ratio I = 1. Introducing relative
coordinates x = x I - x 2' from (4.1), (4.2) we obtain the equations for relative motion
.. 2 G
x + 0 x = al' cos(rot + <p) - - , (4.15)
M2
where Q =JcM I(M IM2 ); ap = PMI(M IM2 ); M= MI + Mz.
The general solution of (4.15) takes the form
x = CI cosOt + C z sinOt + acos(rot + <p) - G I(M z0 2) , (4.16)
where a = al' 11 0 2 _ ro 2 I.
Substituting (4.16) in the right hand side of equations (4.1), (4.2) we can write
the general solution for a system of non-homogeneous linear equations in the
following form:
XI = M 2 [CI cosnt + C2 sin Ot + acos(rot + <p)] + C3 + C4t + ~t2

M 2M
(4.17)
x 2 =_!:!.J..[C I cosOt+C 2 sinOl+acos(rol+<p)- G]+C 3 +C4l+~t2
M C 2M
Taking the time origin at the instant after impact we can write the boundary
conditions for the desired periodic motion
t=O, XI =0, XI=X I+' X2 =X 20 ' X 2 =X 20 ;
t=21t/ro, XI =0, Xl =X I__ , x 2 =x 20 ' x 2 =x 20 '

where x 20 ' x20 are the co-ordinates and velocity of mass M 2 at the moment of
collision.
Using the boundary conditions to find the constants Ci' C2' C3' C4 and phase
<p in solution (4.17) we obtain
nG n
C =----cot-
I MIQoo S'
(4.18)
C =_ nG
4 Moo'
. (l-R)M:<
sm<p =- (4.19)
2M 2 aoo
where S = oo/Q ,
. 2nG
x =---- (4.20)
I~ (1+R)M l oo
Using (4.20) and the notation of (4.15), equation (4.19) can be modified to the
form
(4.21)
Assuming Isin <p I~ 1 in (4.21), we obtain the condition for existence of the
required periodic solution as
G 1 (1 + R) S2
-<-------
P - n (1- R) 11 - 21 ' s
which coincides with the inequa1ity (4.10) found earlier. Hence, regardless of the
ratio of mass MI to mass M 2' all regimes correspond to points in the existence
domains as shown in Fig. 4.4
Substituting the constants given by (4.18) in the motion (4.17) we obtain
XI = ~ (t~ - : t)+ ~ {a[cos(oot+<p)-cos<p]-

(4.22)
- nG
MIQoosin(n 1s)
[cos(Qt - ~J
s - cos ~]}
s
M 200M
(t n) M [MI
x 2 = -G - - - t + -{a
2 -cos(oot + <p) - cos<p -
M2
]
(4.23)
- nG [MI
-cos ( Qt-- n]} +
nJ +cos- MI-G
-
MIQoosin(nls) M 2 S S M c
As previously, by the term 'amplitude of vibration' we mean the value of half
the sweep. It can be shown that when R = I, functions (4.22), (4.23) take extreme
values at the instants t = 0 and t = nloo . Bearing in mind that masses MI and M 2
re ach their extreme displacements at these instants, we will take the value
a x, = Yz[x;(O)-x;(nlw)] (i=1,2)
as the amplitude oftheir vibration.
As a result, using (4.22), (4.23) we obtain the following equations for the
amplitude-frequency characteristics:
2G G1t
a =1t- - - 2 )2 --tan-+-acos<p,
+ (M 1t M2 (4.24)
XI 4Mw 2 M C 2S 2S M
1t 2 G M]M 2 G 1t 1t MI
a X2 =---+-----tan-+-acos<p. (4.25)
4Mw 2 M 2 C 2S 2S M
Initially, let us examine the special case mentioned earlier when mass M 2 is
considerably greater than mass MI. Converting (4.24), (4.25), to the limit when
M2 ~ Cf) and as a result M 2 / M ~ 1, we find
G 1t
2s 2s
1t
a XI = -;; tan + a cos <p, a y, == 0 . (4.26)
Here the first term is identical to the equation of the backbone curve (4.8). Since
the value of cos<p is a double-valued function of the system parameters, equation
(4.26) defines two branches of the resonance curves separated by the backbone
lines. The amplitude-frequency characteristics are shown in Fig. 4.6, where the
thin lines show the backbone curves. When R = 1, according to (4.21),
1cos<p 1= 1 . Substituting into (4.26), the obvious requirement
a XI ~ 0, (4.27)
the apparent necessary conditions for the absence of additional intersections, we
find that both branches of the resonance curve exist for all frequencies in the range
s>
1, when inequality (4.12) obtained earlier as the
result of other reasons, is satisfied.
When R *- 1 the value of the initial phase <p is
defined by equation (4.21). For this case, the
branches of the resonance curve are enclosed
between the proper branches of the system without
dissipation. From (4.26), it can be seen that when the
value of cos<p = 0 , the resonance curves intersect the
Fig.4.6 backbone curves at points which lie on the boundary
ofthe existence domain.
In general, for arbitrary ratio of the mass of the system, the mass M 2 is
moving. The resonance curves for both masses look similar as in Fig. 4.6. This can
be confirmed by comparing equations (4.24), (4.25) with expression (4.26).
They differ because of the appearance of the term 1t 2G /(4M(j)2) , which
represents the possibility of vibration for all systems as a unit mass M , impacting
against stops. The additional multiplying factors before both ofthe other two terms
represent a combined mass term, representing a degree of coupling between the
masses due to their relative vibration. In this case, inequality (4.27) gives the
conditions for existence of both branches of the resonance curves in the form
G I(MI+M?)
->- - (4.28)
P- 2 M2
94 Principles of percussion machine dynamics 59
a particular case of which appears in condition (4.12). We note that even here the
loss of energy during collision reduces the frequency range of the vibro-impact
regime and does not lead to the elimination of resonance with infinite amplitude at
frequency w = n. In this case at the amplitude of resonance, a considerable
influence is shown by other forms of damping, inherent always in real systems.
The amplitude of vibration of mass M 2 also influences the character of the
system considerably, since it defines the vibration activity of the body of the
machine. The tendency to obtain maximum amplitude ofthe striker naturally leads
to an increase in the vibration of the body, the allowable level of which, especially
for hand machines, is Iimited by rigid standards. Examining the equations for
amplitude (4.24), (4.25), we see that in the neighbourhood of resonance, t;; ---) 1,
the first term is negligibly small by comparison with the two others and the
amp Iitude of vibration of mass M 2 becomes
aX2 '" (MI / M 2 )ax, .
Hence, the level of vibration in machines depends mostlyon the ratio of the
mass ofthe body to that ofthe striker.
3. In §3 we found that viscous damping changes the boundaries of the
existence domains in the vicinity of linear resonance. In the models examined
earlier, the impact regime had a weak character in this frequency range. Hence a
study of these regions was not of particular interest, and special attention was
given to impact dissipation and its influence on vibro-impact resonances.
However, in the two-mass system examined regimes with maximum amplitude
occur particularly in the vicinity of the linear resonance frequency and it can be
assumed that viscous damping shows an important influence here. To ca\culate this
influence we will use an approximate process based on the establishment of the
equations for energy balance during motion, displayed by the first harmonie of
natural vibration. Weshall examine the limiting case as M 2 ---) 00 •
The equations for the amplitude of the fundamental are obtained with the help
of expression (2.12) and equation (4.17), which defines the equivalent gap. We
have
20
(4.29)
In Fig. 4.3, this expression is shown as a dotted line for comparison with the
exact backbone curve (4.8) drawn with asolid line.
Suppose under the action of a harmonic excitation P(t) = P cos( wt + <p) and
the force of viscous friction ni, mass MI achieves harmonic vibration
XI = a xl coswt . During this motion, the work ofthe excitation force for one period
is equal to
fP(t)xldt = na
2n!ru
Ee = x1 P sin <p (4.30)

o
and the work ofthe viscous friction force is
JnX dt = nroa;1 n .
21t/w
Edl = 1 (4.31 )
From the condition of balance between the energy entering the system due to
excitation, Ee and the energy Edl dispersed by the dissipative forces during the
per iod we obtain
sin<p = axlnro/ P.
Since 1sin <p I::; 11, we obtain the following condition of existence:
a XI ::; P /(nro). (4.32)
Hence, viscous damping limits the maximum vibration amplitude. It is
~k"~"'~
, 1 J( , t J( 1 2 S~
a) c)
hg.4.7
convenient to find the limiting amplitude graphically as the points where the
backbone curves (4.29) intersect the energy boundary of vibration, defined exactly
by equation (4.32). The resulting resonance curves take the form shown in Fig. 4.7
a.
By similar methods we can find the conditions for existence when impact
dissipation only is present. The loss of energy during impact takes the form
E
d2
=~(x2_x2)=MI«I_R2)
2 1- 1+ 2 '
or using (4.20)
2n 2G 2Q2 (1- R)
E =---- (4.33)
d2 cro 2 (l + R)
From the conditions of energy balance, (4.31) and (4.33) we obtain
. 2nG 2Q2 (1- R)
sm -- (4.35)
XI - cro 2P (1 + R)
From this it follows that impact dissipation provides the lower limit of the
resonance curve. When ca\culating this form of dissipation only, the resonance
curves takes the form shown in Fig. 4.7 b. The minimum amplitude is defined by
the points where the backbone curve intersects the line given by expression (4.35).
When combining the effects of viscous and impact dissipation, the energy
balances (4.30), (4.31), (4.33) give the following condition for existence:
~[1-
2nco
1- 2
4G nC]<a
cop 2 -
<~[1+
x, - 2nco
1- 2
4G nc]
coP 2 '
(4.36)
where C=2nO-R)/[M 10+R)].

In this case the resonance curves (Fig. 4.7 b) take the peculiar fonn of loops,
the frequency range of which becomes compressed as the dissipative condition of
the system increases.
Using the fact that the tenn under the radical in (4.36) is positive we find the
ratio between the parameters at which it is possible to obtain resonance
G< coM, (I + R) (4.37)
P - 8nn(l- R)
When equality applies in (4.37), the resonance curves degenerate to a point
defining the intersection ofthe backbone curve
with the line a XI = P /(2nco) .
Percussion machines often try to perfonn as ,.l'lQ
resonant ones. It is obvious that regimes that J
confonn to the maximum resonance curves are
suitable for optimum productivity as weil as energy
consumption. However, these conditions require
z
that the relationship between the system parameters
f...::.._~ _ _ _-=-
is fully maintained. Hence, the frequency of ()
p
excitation co', corresponding to the maximum
resonance curve, depends on the value of the
excitation force P, the applied force G and
dissipative factors. Assuming that in the vicinity of co * , the influence of impact
damping is negligible, from (4.29) and (4.32) we find the relationship of the
parameters during tuning of resonance.
G= M 1 .(co·2 _n 2 ). (4.38)
P 2nco
Fig. 4.8 shows the dependence of the resonant frequency co' on the feed force
G, according to equation (4.38). From this, the resonance amplitude changes in
accordance with expression (4.32).
In §20 we will concentrate especially on excitation and stabilisation of
resonance in vibro-impact systems.
4. Up to this point, during the study of a range of vibro-impact systems, it
has been assumed that the excitation is applied in the prescribed periodic fonn. In
percussion machines, such effects are created by a various range of vibration
exciters. Amongst them, and described above, are drives which convert mechanical
rotation to reciprocating motion (Fig. 4.2) and numerous electromagnetic and
electrodynamic exciters, pneumatic and hydraulic vibration exciters.
a) 5
11 I A-"(
*'
-
2
Fig_49
However, there is a wide c\ass of machine typically using vibro-impact

principles of excitation, based on systematic interaction of an oscillating mass
moving relative to aseries of stops [17]. An obvious examples of such motion is
often seen when boys run a stick along a fence.
H
a)
,~~-----H------~*-
/J)
Fig. 4. 10
An example of such a method appears as an impact rotating mechanism

(FigA.9, a) widely used in the construction of screwthread wrenches, screwjacks,
c\amping devices etc. The mechanism consists of a striker 1, surrounding a driving
shaft 2, with a helical v-shaped groove on the shaft surface, 6. The ball 5 is
mounted in the groove of the shaft, and lies in a similar v-shaped groove on the
internal surface of the striker. With the help of aspring 4, the lugs on the striker
mesh with those on the anvil 3. The mechanism operates in the following manner.
Ouring the time when the extemal force on the anvil is less than the value of the
moment created as a result of the preliminary force of compression, the anvil and
the striker move as one body. Ouring braking of the anvil 3, rod 2 continues to
rotate and exerts a screwing effect on the striker 1 and disconnects the lugs, which
after jumping once again, are put into mesh under the influence ofspring 4.
In FigA.10 a, is shown a plane analogue of this mechanism (in Fig. 4.9 and
4.10 similar elements are denoted by identical figures). The driving shaft is
mode lied by a moving carrier which has a sloping plane 2 and the lug of the anvil
by the stop 3 against which rides the striker 1, elastically suspended by aspring of
stiffness c, moves along the sloping plane.
This system represents one of the dynamic models of a double mass system
studied earlier. Its features, including superposition of additional constraints, is
related to the vibrating body itself, which does not have a straight line trajectory
but requires plane motion to be considered. At the same time, the arrangement of
the striker on the carrying body leads to a geometrie summation of its relative
(vibratory) and its translational motion, significantly increasing the impact
efficiency.
Superposing various constraints allows us to put into practice various types of
mechanism producing vibration from the possibilities described. Fig, 4.9 b shows
yet another different construction for a vibro-rotation mechanism. The striker 1,
eccentrically mounted on the driving shaft 2 and connected to it by spring 4,
er) ;~
} .\'
I
h)
a)
Fig.4.11
meshes with the lugs on the anvil 3 through its own teeth. Ouring braking the lugs
ofthe rotating shaft 2, because of its eccentricity, take the striker out of mesh with
the anvil. The striker rotates the spring 4, whieh on release, gives the striker an
additional rotation relative to the eccentric. As a result the striker applies an impact
to the following lug of the anvil and the whole process repeats. Fig. 4.11 b shows
the plane analogy for this mechanism. Here, striker 1 rotates in conjunction with
straight-line motion of the carriage 2 and the spring of stiffness c associated with
it. Later it will be shown that the character of additional constraints between
vibrating and supporting bodies significantly influences the vibration
characteristics.
Note that in this manner we can produce excitation of many different
mechanisms. One of the earliest to be produced practically was the watch

mechanism known as the escapement (Fig. 4.11 a). Ouring vibration of a
pendulum 1, one of the fingers attached to it, 2, 3, becomes clear of the teeth on
the moving wheel 4, which rotates through one step for one swing ofthe pendulum
while the other finger gradually meshes with a tooth on the moving wheel. At the
same time the impulse from the moving wheel is transmitted through the fingers to
the pendulum and in this fashion maintains its vibration. A simplified dynamic
model of a similar mechanism (Fig. 4.11, b) was described in [9] and examined in
detail in [9, 51 to 53]. The moving wheel was modelIed as an endless belt 4, fitted
with teeth. A spring-mounted mass 1 imitates the pendulum, reciprocating between
the branches ofthe belt and supplying the fingers 2, 3 with delicate impacts. In the
interval between impacts, the belt can only move forward through one tooth, which
instantaneously retracts after impact, so that the point of its attachment can freely
move past the finger of the oscillator. Since only the tooth on the lower branch
retracts itself, the tooth on the upper branch moves accordingly and so on, in turn.
Further refinement of the models requires calculation of the geometric
constraints superposed on the vibrating body and provision for unimpeded bending
of the fingers and the corresponding teeth of the rotating wheel after collision,
similar to that shown in the model in Fig. 4.10 b. The models of Fig. 4.10 can be
used also to explain spurious vibro-impact effects which happen when operating a
rotating clutch mechanism, or rotating a ratchet mechanism etc.
We shall study the dynamic model described, (Fig. 4.10) assuming that the
main body 2 moves uniformly with velocity v o ' Such an assumption allows us to
simplity the study considerably, reducing it to the analysis of a system with a
single degree of freedom, whose vibrations are synchronised by a periodic change
with time of its kinematic structure.
Using the stitching method we find existence conditions and the law of motion
for regimes characterised by regular impact of the striker onto each stop. The
motion of the striker relative to the carrying body has a periodic character and
completely defines the regime sought. We shall use z as the relative coordinate of
the point mass MI of the striker, measured from the free position of the elastic
element. For the second model (Fig. 4.10, b) co ordinate z = (\jf - a)L introduces a
condition to identity the situation. Here \jf is the current angular coordinate: a is
the angle characterising the free position of the elastic element; both angles are
measured from the vertical axis. Neglecting the loss of energy in the intervals
between impacts we can write the differential equation for the relative motion in
the form
Z+02 Z =0. (4.39)
For the first model 0 2 = c sin 2 \jfo I MI and for the second 0 2 = C I(M IL2 ). The
boundary conditions for the regimes examined take the form
t = 0,
(4.40)
t=T=Hlv o, z=zo, z=z.
Using the boundary condition when t = 0, the solution of equation (4.39) is as
folIows:
§4 Principles ofpercussion machine dynamics 65
z = Zo cosOt +~sinOt (4.41)

o
After substitution ofthe boundary conditions when t =T we find
i Ot
z =--- cot- (4.42)
o 0 2
z + =-z - (4.43)
The kinematic relationships of impact interactions with constraints between the
striker and the supporting body (constrained impact) have already been obtained in
§ 1, section 7. Using relationship (l.83) connecting the velocity before and after
impact and using equations (4.42) and (4.43) we obtain
(1 + R)(cos \11 0 + 8sin \11 0 )v o cot(Ot / 2)
Zo = - . (4.44)
[(1 - R) cos \11 0 + 8(1 + R) sin \11 0 ]0 cos \11 0
For the first model (Fig. 4.10, a) \110 is a fixed parameter and hence formula
(4.44) defines a unique solution for the problem. For the second model (Fig. 4.10,
b) Zo = (\110 - a)L, where \110 is the unknown angular coordinate at collision
defined by the transcendental equation (4.44). In this case, the solution is not
unique, which corresponds with the muItiplicity of periodic regimes for the system.
For both models, relationship (4.44) gives as a condition for existence of the
solutions found, OT"* 2nn (n = 1, 2, ... ) which has an obvious physical
meaning. For the second model, moreover, the solution \110 must comply with
inequality (1.81) the condition developed to ensure the absence of jamming during
restricted impact.
The relative motion of the striker (4.41), with the help of equations (4.42),
(4.43) can be converted to the form
z = Zo(cosOt + tan Ot sinOt) = Zo COSo(t-!) (4.45)

2 cos(Ot / 2) 2
(t E [0, Tl).
An impression ofthe character ofthe periodic regimes for the second model can be
found easily by substituting 8 = 0 in (4.44).
The result can be written as
1/(\110 -a) = Kcos\llo, (4.46)
where
K = _ OL(1 -R) tan Ot . (4.47)
voO+R) 2
Fig. 4.12 shows examples for graphic solutions of equation (4.46). These solutions
produce multiple values, the elements of which are related to the number of
revolutions of the angular coordinate of the striker at the moment of impact.
Rearranging (4.42) in the form
\11- = -0(\110 -a)tan(OT /2), (4.48)
when tan( 0 T / 2) > 0, coordinate \11 0 - a and velocity \11 take opposite signs at
the moment of impact, when tan(OT /2) < 0, they take the same sign. The first
66 Chapter I Vibro-impact systems
case represents impact during conditions of pre-arranged strain, the second

represents conditions for a gap. This situation was similarly obtained earlier for the
double mass model (Fig. 4.1) where the value of the equivalent gap (interference)
(4.9) changed depending on the vibration regime.
From (4.46) it can be seen that
as the solution of equation (4.46)
moves further away from the
asymptote \110 = a (Fig. 4.12) then
it rapidly approaches the value
±nn/2 (n = 1,3, ... ) . Hence,
practical interest is aroused only by
those few va lues of \11 0 nearest the
Fig.4.12
asymptote which can represent real
processes satistying the conditions for the absence of jamming (1.81).
5. Let us study the stability of solutions due to small perturbations of the
phase coordinates. For simplification we shall limit the presentation to the case
e = 0, which will preserve all the basic features of the qualitative picture of
stability. The study will be carried out simultaneously for both models, using the
finite difference scheme presented in §2 for the study of asymptotic behaviour near
stationary solutions.
In agreement with (4.41) we will write the equation for the disturbed motion
after the v-th impact
Z +öz .
z=(zo+özy)cosQt+ + Y+smQt
Q (4.49)
Z = -(Zo + ÖZy)Qsin nt + (z, + ÖZ y+ )cosQt
where ÖZ y and ÖZy+ are small increases in coordinate and velocity at the instant
after the v-th impact. We will find the increment in the phase coordinate due to
the following (v + l)-th impact. Substituting in (4.49)
t=T+ÖTy , z=zo+öz y+l , z=z+ÖZ(Y+I)+
we obtain after linearisation the equations for disturbances
öZ Y +1 =öz,cosQT+ Ö~y sinQT+öT,z_,

(4.50)
ÖZ(Y+Il_ = -özyQsinQT + öZ vt cosQT + ÖTJ],2 Z0
The increment in the period during the v-th interval öTy will be expressed
through the increment of the phase coordinates (Fig. 4.10). Taking into account
that for the first model, the impact condition takes the form
t = T + öT,,, x = Vol + Zcos \jf 0 = H ,
we find, incrementing this equation
öX y = vOöTy + (ÖZ"I - ÖZJcos \110 = 0,
whence
cos Wo
öTv = ---(Öz v +1 - ÖZ v ) ' (4.51)
Vo
A similar relationship is obtained for the second model by incrementing

x = Vol + L sin W in the vicinity of W = Wo'
Incrementing equation (1.83) through velocity i and angle W in the vicinity
ofthe stationary value when e = 0, we find the relationship between the increment
of velocity before and after impact for the second model
(4.52)
where
D = _(l_+_R_)_v-,:-o2 _si_n_w--,-o.' s:
uzv=LuWv
s:
L cos Wo
Note that as Wo nears the value ±rr.nI2 (n = I, 3, ... ) the condition for small
value of öz v + is violated, hence the stability analysis for these values does not
provide true results. However, as was shown earlier, a periodic regime has no
practical meaning in this case.
For the first model, the relationship between the increment of velocity before
and after impact can be obtained formally from(4.52) when D = O.
Substituting (4.51. (4.52) into (4.50), and substituting for the value i_ from
(4.42) we obtain a system of linear homogeneous equations in finite-difference
form
2öz
- s:
- v-+1u z (I
+R n. n. )
v --+cos~~T--sm~~T
D. S:' R. n.
+uzv_-sm~~T=O,
l-R l-R Q Q
Öz 1+R QT
--Qcot--Öz +R QT. (1 )
--Qcot--QsmQT-DcosQT +
v+1 1- R 2 v 1- R 2
+ Ö2:(V+1)_ + Özv_RcosQT = O.
Seeking their solution in a form similar to (2.42), we find the following
characteristic equation after simple transformation:
2ß2 - ß[ (I + R)2 + (1- R)2 cosQT - ~ (1- R)SinQT] + 2R 2 = 0 .

Eventually, using Schur's criterion (2.45) we find the conditions for stability
(1 + R)2 + (1- R)2 cosQT - D (1- R)sinQT
I----------,-----=.Q=------I < 1. (4.53)
2(1 + R 2 )
Substituting D = 0 in (4.53) we obtain for the first modell cosQT I< 1, i.e., the
condition for stability is satisfied for all the corresponding parameters for which
the periodic regimes exist.
Substituting in (4.53) the value of D from (4.52) we obtain, using (4.46), (4.47)
conditions for stability for the second model in the form
1(1 + R)2 + (1- R)2 eosQT + (1- R)2 (1- eosQT)(\jf 0 - a)tan \jfo 1<
(4.54)
<2(I+R 2)
Equation (4.54) is equivalent to two inequalities
(\jf 0 - a) tan \jf 0 < I, (\jf 0 - a) tan \jf 0 > -d , (4.55)
where
d= 1 [2(l+R 2)+(1+R)2 +eOSQT].

1- eosQT (I - R)2
Thus, depending on the sign of the quantity \jf 0 - a the eonditions for stability
are as folIows:
1 0 tan \jf
---~tan'Vo' --->---- (4.56)
\jfo-a \jfo-a< d
Here the upper value eorresponds to
positive value and the lower to a negative
value of \jfo-a. In Fig. 4.13, stability
domains eonstrueted in agreement with
(4.56) are shown hatehed. The dotted line
shows the graphie solution for equation
(4.46) (Fig. 4.12).
6. The formulas obtained make possible Fig. 413

complete design ca1culations for impact rotary
mechanisms similar to those shown in Fig. 4.9, and
to synthesise their structural elements and to tune the parameters ofvibratory systems. Fig. 4.14 shows
examples of a comparison between relative, 1, and absolute, 2, solutions (full line), and the practical
(dotted line) trajectory for a striker taken from areal machine manufactured according to the design of
Fig. 4.9, a. The calculated trajectory can be constructed either using the parametric equations
x=vol+zcoslVo, y=-zsinlVo (tE[O,H/voD
or with the help of graphical construction (Fig. 4.14) offsetting relative coordinate z along the direction
ofthe inclined plane from point x = val along the abscissa. Since the striker performs complex helical
motion in the real mechanism, for ca1culation it is necessary to take into account that in the model the
mass of the striker MI has a corrected value related to the actual inertia parameters of the striker
through the relationship
1
M I = -ZCOS 2 M·2
~Jo + sm ~Jo,
r
where 1, Mare the moment of inertia and the mass of the striker; r is the radius of the trajectory of the
centre ofthe ball (Fig. 4.9, a). It is not difficult to use the known absolute trajectory, to choose such a
configuration of colliding elements, so that the striker will bend the teeth of the anvil without touching
in the interval between calculated impacts.
An important parameter of the machine is the power of its drive mechanism. By analogy with
(4.3) we will find the average force G developed by the drive. lt is related to the value of the normal
impulse J appearing on the striker during impact, by the expression
G = J / T = JVI) / ff. (4.57)
Using equation (1.82) we find when e = 0
J =(I+R)J I = MI(I+R)[(vo/coS2~Jo)-(Z_/COSlVo)1 (4.58)
Substituting in equation (4.57) the value (4.58) and the value of L from 4.42) we find using
§5 Stitching method 69
(444)
G= 2Ml v 6 (I+R)
(4.59)
Hcos 2 Wo (1- R)'
Fig. 4.14
The nominal power ofthe drive mechanism is found from the formula
N=Gv".
Note that equation (4.59) also gives the feed, i.e., the static force which must be applied to the
body of the machine during operation.
§5. Stitching method

I. To conclude the presentation of the results
01' analysis 01' vibro-impact systems using exact
methods. we will give a short generalisation of the
calculation procedures used for the analysis of
existence conditions and stability of periodic solutions
in systems with an arbitrary finite number of degrees
of freedom, vibrating with several impacts per period.
For an explanation of the procedure, it is convenient to
illustrate it by dealing with the phase space of the
system.
We will look at a vibro-impact system described z,
by a set of differential equations of the n-th degree,
i.e., having nl2 degrees offreedom, which describe the
system behaviour in the interval between collisions.
The system described has n phase coordinates, z,
Fig. 5.1
xI , "', XII as a measure of which we can take the
generalised coordinates and velocities of its point

masses.
The study of the structure of any vibro-impact system reveals the functional interrelations between
its generalised coordinates corresponding to all possible system configurations in which collisions of
its elements occur. We shall call these unilateral constraints impact constraints.
[n the general case, each impact constraint is described by a function, expressed through phase
coordinates, time t, and time shift (phase) t", between extern al periodic force and system motion,
definable from the chosen starting time point and written as:
nk(x\, ... , x n ' t", t) =0 (k = [, 2, ... , m) (5.1 )
[t is necessary to take into account the dependence of impact constraints on time when elements of the
system are lead kinematically to the assigned periodic law. That is why, it is assumed that function
(5. [) is periodic in t and t" with excitation period T.
As we study each time interval 01' the moving system between alternate collisions, it is convenient
to combine the start of the interval with the instant of impacl. As long as impact constraints given by
(5.1) are applied to the system only at this instant (t = 0) then the expression can be contracted:
n'(Xq,t,;) = 0 (q = 1,2, ... ,m) . (5.2)
Expression (5.2) originates in the phase space surfaces (Fig 5.1), called surfaces of discontinuity. The
number of discontinuity surfaces depends on the structure of the system and in the general case is an
arbitrary finite quantity. The appearance of the representative point (end 01' the phase vector) on the
discontinuity surface corresponds with the collision of some 01' the system elements, as the result of
which their velocity instantaneously changes in agreement with the kinematic relationship of
stereomechanical theory and the phase point moves along these surfaces to a new position. Hence, for
each discontinuity surface n k = 0 some functional transformations of the phase coordinate occur
which in the general form can be written as folIows:
X,- =<t>;(x,+ .... ,x; ,1o,) 0= 1,2,... ,n). (5.3)
Here, the negative and positive signs relate to values before and after impact. The time dependency in
(5.3) arises in the same way following kinematic excitation of some elements and for this reason can
be considered as T - periodic. Taking into account that the transformation (5.3) occurs when t = 0 it
can be contracted and written in the form
x,- = <t>; (x,; ,I,;) . (54)
Noting that the transformation can exist on the discontinuity surface only if the following
condition is satisfied
an k aXq an k
I - - .+ - ; " 0
n
(5.5)
'10' aXq at~ at{,
which denotes that at the moment of intersection with a discontinuity surface the points shown possess
non-zero veiocity relative to them.
Surfaces nk =0 divide the phase space ofthe system into the zones C', .. , C' , in the limits 01'
which the point described fulfils continuous motion, corresponding to the movement of the system in
the interval between impacts. This movement can be written as the solution of some systems of
differential equations. Just suppose C' is a zone between surfaces n' = 0 and n'+' = O. The
solution of the system of differential equations can be written provisionally as:
x;'=x:(C; .. ,c~,t;;,t)=x:(C;,t,:,t) (j=I .... ,n). (56)
where C j is a set of integration constants.
Using the accepted meanings, each type of periodical movement of the system with an arbitrary
finite number, m, of collisions for aperiod 01' movement T, can be written as a sequence of values of
index k defining aseries of collision elements. For a study 01' regimes 01' the defined type it is
expedient to arrange these sequences in order ofthe values
1,2, .... m (5.7)
by means of a corresponding relabelling of the discontinuity surfaces (it is possible that various
numbers may be repeated). This sequence characterises also types of non-periodic motions to which
the series of colliding elements correspond (5.7). The interval of motion during which the sequence
(5.7) is realised on ce will be called the cycle ofnon-periodic motion. For convenience in our study the
beginning and end ofthe cycle will be arranged on the surface n1 = 0 .
In phase space (Fig. 5.1), periodic motion of the form (5.7) will correspond with the closed
trajectory shown, intersecting the discontinuity surfaces and the areas between them in accordance
with the given sequence (5.7) and in agreement with express ions (54) and (5.6). With small
perturbations, the system motion, deviating from the periodic regime (shown dotted in Fig. 5.1),
undergoes n' = 0 at point A, different from .4.,. corresponding to a periodic trajectory, and, passing
in sequence (5.7) to the remaining discontinuity surfaces, arrives again at n' = 0 in the new point
A, . In this case also, the trajectory of the phase point in phase space also 1'ollows the same sequence of
transformations of its phase coordinates. Hence, the aggregate motion 01' the system characterised by
the defined series of collisions between its elements is adequately represented by several groups of
sequential transformations. Denoting the group of sequential transformations by the letter Q, the
motion ofform (5.7) can be interpreted as a form of a point mapping for some surfaces (for example,
rr I = 0 ) onto itself after transformation Q. Conditionally this can be written as QA = AI .
Periodic motion of the type studied will be correlated with stationary point Ao by transformation
Q, i.e., QAo = Ao . If, as a result of multiple application of transformation Q to the points, we find
ourselves in the low E-neighbourhood of static point Ao , as the points continuously approach it, then
the point Ao is called asymptotically stahle.
The geometric interpretation of the examined combination of transformation Q in the form of the
mapping of the point of some surfaces on themselves, has been called the method 0/ point mapping.
Such an interpretation allows the use of topological representation and methods to establish many
important effects of the transformations described. The idea for point mapping which arose from the
work of H. Poincare, and G. Birkhoff, has found wide applications in the theory of nonlinear
vibrations, thanks to the researehes of A. A. Andronov and his school [10]. Contemporary methods for
point mapping and a detailed bibliography, are more fully explained in the monograph by Y. I.
Neimark [171]. In particular, a number ofworks on the application ofthis approach in the theory of
vibro-impact systems is presented (see also [57]).
The analytical relationship between the phase coordinate of the points constructed before and after
transformation Q produce expressions in the form of difference equations. This means we may use the
methods and theorems of difference equations for the study of the type of transformation described
[99]. Most important here is the distribution of Lyapunov's method for stability for a system of
difference equations. The contemporary state of this problem and its applied aspects are set out in the
book by P V Bromberg [68]. A E Kobrinsky and A A Kobrinsky iIIuminate problems and provide a
bibliography for using this method in the theory ofvibro-impact systems in the monograph [118].
Both approaches are adequate since they are different mathematical interpretations exist for one
and the same transformation based on the idea of stitching associated with names of L LeComu, A D
Papaleksy and A Sommerfeld [8,141].
We have al ready met examples of the application of difference equations for the analysis of
stability in vibro-impact systems (see §2, section 3, §4, section 4), and in Chapter 5, when
synthesising vibro-impact systems we will make use ofthe method ofpoint mapping.
2. We will find the conditions for existence of a T-periodic regime oftype (5.7) (see Fig. 5.1).
We will denote the passage of time of the phase point of area G·' as r' . We will take n l = 0 as the
original surface for examination . At the moment of leaving it, the point described has coordinates
Ix:(c;ti,.o)} .The next entry ofthe point described on the surface nl = 0 occurs during its motion
from surface n m = 0 when t = tm . The values of the phase coordinates at this point will be as
folIows: IX;" (C;'. tg'. t m )} . Using (5.4) we obtain the conditions for periodicity in the form
X;' (C;' ,tg' ,tm) = <1>: [x~(C) ,/,\,0),/,\] (i = 1,... ,n) . (5.8)
These equations must be supplemented with the conditions connecting the coordinates on the
remaining discontinuity surfaces
X~-I (C~-I ,/;'-1 /-1) = ~ [x~ (C~ ,I,~ ,0),/;']
(5.9)
(i = I, ... ,n;k = 2, ... ,m)
and the conditions attached to the phase coordinates converted according to (5.8), (5.9) on to the
discontinuity surfaces, i.e.,
(5.10)
Since
r~ = 1;'+1 - t;;' when s = 1,... ,m - 1,
(5.11)
m-I
Im =T - Lr'
.~=I
express ions (5.8) to (5.10) provide sufficient conditions for determining the arbitrary constants
I "
C,() ... ,n, = I, ... ,m) an d phase 10k (k = I,... ,m).
= I,k
Let us look at the stability of the periodic regimes found. As the result of small perturbations
applied to the system, the value of the phase coordinates for the point describing a stable solution at
the instant it appears on the discontinuity surface and its departure therefrom, will be different from
the corresponding value for periodic motion by a small increment. These increments can be equated
through the incremental values of C~ ,1,: and 1I . The quest ion of stability is decided by studying
recurrent relationships coupling the variation of these values belonging to various cycles of excitation,
for some discontinuity surfaces, for example, IT" = 0 .
Let us follow the motion of the representative point on surface n l = 0 (Fig. 5.1). On this surface
it appears at point A, after finishing the v-Ih cycle of disturbed motion. At this time its coordinates will
be as folIows: x,[C;, + (8C;'),.,I;;' +(81;;'),,1'" +(81"'),], where 8 means a small variation in the
corresponding value. At the instant of departure of the point from surface nl = 0 its coordinate will
take the form [x}(C; + 8C; ,1/, + 81/,,0),0]. Comparing variations pertaining to various cycles
produced on surface n"' = 0 , those which have no index m will not be indicated by number of the
cycle, since they are under consideration for the one cycle v+ I only.
Coupling the phase coordinates in the system on surface nl = 0 before and after collision with
the help of equation (5.4) we find
x;'[C;' + (8C;' ),. ,I~' +(81;;' ),,,1"' + (Öl"'), 1=
(5.12)
= <P;[X:,(C~ +8C~,I,~ +81,~,0),1,~ +81,~1.
Expanding (5.12) as a Taylor series for small variations and noting that (Öl"')" = ÖI/, -(ö(;')" we
obtain to second degree accuracy
n a "'
I~(öC"') + ~-~ (81"')-I-"'-',I-
(a", a"' J
q 8C 1 +
n ami axn
l
1=1 ac;, ; ,. alg' al'" '" '1=1 ax: H ac; J
(5.\3)
+ fa "' ami - "_"'_'

~-_",_, "ami -_'Il 81 1 = 0 ax J
L malai"
I ~a'al
'1=1 x'l I"
Cl •
It is not difficult to see that equation (5.13) can be obtained by means of formal variation of equation
(5.8). Carrying out this operation on equation (59), (5.10) and noting that according to (5.11)
Öl' = ö15+1 - Ö15, we obtain
na I-I '-I Ol-IJ (0
n amI n ax'
I~ÖC'-I+ ~-~ öt'-I-I-"'-' I-'-' 8C' +
,=1 aC;"1 1 al,;-I al'-I " '1=1 ax:, 1=1 ae: 1
I-I I , 0 I
+ [ ~_ a<t>, _ ~ a<t>, ..!.cL 81' = 0
J (k 2 )
(5.14)
ar' al,; .;Si ox~ 01,;" = ,'. ,m ,
n an' n ox'
I-,-. I-T.-8C~ + -,-. + I - - ,
1=1 oX q )=1 ac; 01" 1=1 aX q
(on l
n on ox')81,; =0
al"
l
-+ (k = I, ... ,m) (5.15)
Note that in equations (5,14), (5,15) when k = m variations 8C;' ,Öl;;' do arise, however, in
contrast to the similar variation in equation (5.13) they belong to the following v+ I-th cycle of
disturbed motion. Thus equations (5.13) to (5.15) cause recurrent relationships between the variations
of the arbitrary constants 8C;' ,81,~' bel on ging to various cycles of the disturbed motion in the form of
a system of difference equations. Taking into account the character of the solutions of these equations
(see §2, section. 3), we can put in equations (5.14), (5.15) ÖC;' =ß(ÖC;'),.,ÖI;;' =ß(ÖI~'),. (ß is a
numerical constant). In such a way, equations (5.13) to (515) converge to a system of linear ordinary
algebraic equations far the variations. Setting the determinant of this system to zero, we obtain the
characteristic determinant, the roots of which represent stability for the motion studied. It is not
difficult to convince oneself that maximum order of the determinant for finding the constant ß is equal
to n, since when using equation (5.15) the phase variations in equations (5.13), (5.14) can be linear
express ions through corresponding variations in the constants of integration.
The stability conditions for the periodic solution considered exist by finding the roots of the
characteristic determinant inside a unit circle 1ßi I< I (i = I, ... ,n) . If any of the roots lie outside the
limits ofthis area ofthe complex plane, then it represents an unstable solution. Estimating the roots of
the characteristic determinant is possible using Schur's criterion, or equivalent (see §2, section
3). Using the substitution ß = (I + s)/(l- s) it is possible to transform the unit circle on the complex
plane for root ß onto the left-hand half-space of the complex plane of root s, and to use the Rouse-
Hurwitz criterion for stability [24,161].
The stitching method appears convenient for studying low order systems, examples of which have
been presented in this chapter, or for systems having a regular structure [116, 166, 168]. In references
[125, 127, 172] this procedure is simplified to find approximate solutions for periodic regimes of
vibro-impact systems, with small nonlinearities.
In the following section we will acquaint the reader with another method to find exact periodic
solutions, which is more convenient for dealing with systems of large dimension since it allows the
wide use of linear analysis ofthe structures composing vibro-impact systems.
Chapter 2. Equivalent linearisation of vibro·impact systems
For this reliefmuch thanks.
Williarn Shakespeare: Harnlet.
§6. Equations 01 vibro-impact systems

I. When we constructed exact solutions we took one mathematical model to
describe motion between impacts and another for the instant of collision. The first
we described by linear differential equations; and the second using finite
express ions from stereomechanical impact theory, used as the boundary conditions
of the process. Such a description hampers the general analysis of vibro-impact
systems, largely because it is impossible to predict the successive instants when
the bodies will collide. Using the procedure of finding a periodic solution, which
makes it possible to postulate regularity of collisions, does, in some measure,
make it possible to overcome these difficulties, but, even in these cases, the
solution constructed does not appear completely adequate due to the additional
possibility of the body trajectory crossing the boundary levels. Separate
descriptions of inter-impact motion and the impact process are even more
problematic when using a more general concept of collisions between solids. For
this reason, we will turn to the construction and analysis of single forms to record
the equation of motion for the colliding elements of a vibro-impact system on all
time axes, describing the fully combined motion occurring. This is possible by
introducing nonlinear expressions reflecting the process of force interaction of
colliding bodies or their elements.
The required express ions are constructed on the basis of force characteristics
of contact interaction \fex) or \f(x,x) (see §1), and in the case of single sided
impact interaction when relative coordinate x =.1 take the form
<1>(x) = \f(x - MT](x -.1), (6.1 )
<1>(x, x) = \f(x - .1,x)T](x -.1) (6.2)
x <0,
T](x) = {~ when x ~ o.
In the case of a symmetrical double sided interaction, the elements of the impact
pairs fixed at the basic gap 2.1 when x = 0, the similar characteristic will be:
<1>(x) = \f(1 x 1-.1)T](1 x 1-.1)signx (6.3)
76 Chapter 2 Equivalent linearisation of vibro-impact systems
(x,x) = '1'(1 x 1-L\,x)ll(1 x I-L\)signx (6.4)

Expressions ofthe type (6.1), (6.3) will be called staticforce characteristics ofthe
impact pair, while type (6.2), (6.4) will be called the dynamic force characteristics
~
W~(Z)' /
(ZI
- /I
o d z
a) b)
Fig.6.1
of the impact pair. The dynamic characteristics are frequently presented in the
form of the sum of elastic components <1>" (x) or ,,(x,x) and dissipative
constituents D (x) or D (x ,x) .
In particular, when describing contact interaction using Hertzian theory (see
(1.35), (1.36» the static characteristics take the form (Fig. 6.1, a)
(x) = CI (x - /),Y/2 11 (x - /),) (6.5)
in the case of single sided interaction and
(x) = CI (I x 1-L\)3/2 11(1 x I-L\)signx (6.6)
for double sided collisions. To approximate the force characteristics of contact
interaction linear expressions take the form (Fig. 6.1, b)
(x) = c2 (x - L\)ll(x - L\), (6.7)
(x) = c2 (I x 1-L\)1l(1 x I-L\)signx . (6.8)
Similar characteristics can also be constructed using the hypotheses of
stereomechanical theory of impact [37]. Let us look at the direct central impact of
two forward moving bodies with masses M I and M 2 • Let us assurne that collision
occurs when the relative coordinate x = L\. According to (1.46) the equation for
relative motion ofthe bodies in this case takes the form
Mx=- (M=M I M 2 /(M I +MJ), (6 .9)
where is the principal vector of the interaction force . An effect equivalent to
collision can be obtained by introducing an "instantaneous" force acting only
in the narrow layer L\ S; x < L\ + E , where E is a small quantity, but, as it
approach es the limit, by decreasing the thickness of the layer to zero. Let us
substitute this force in the sum oftwo components: the elastic component
 Li = <1>1 {ll(x - L\)-ll[x -(L\ +E)]} (6.10)
and the dissipative
 D = <1>2 {ll(X - L\) -ll[x - (L\ + E)]}signi . (6.11)
The constant coefficients I and 2 wi 11 be defined so that as x changes
from L\ to L\ + E ,the value of the relative velocity x changes from x_ to zero,
§6 Equations of vibro-impact systems 77
but during the change of x from Ll + I: to Ll the velocity changes from zero to
-R:ic_, where:ic is the relative velocity ofthe bodies at the start ofcontact. As a
result the work ofthe force during impact will be as folIows:
(<1>1 +<1>2)1: = Mx: /2, (<1>1 -<1>2)1: = M:ic:R 2 /2.
Solving this system of equations for the coefficients <1>1 and <1>2 we find
<1>1 =M(1+R 2 )x:/(41:), <1>2 =M(I-R 2 )x:/(41:). (6.12)
Substituting the value ofthe coefficients <1>1 and <1>2 in (6.10), (6.11) and noting
°
that the usual limit when I: ~ is not valid here, we apply the full transition in
the generalised sense [98, 198]. As a result we find
<1>. = Mx:(1+R 2 ) ö ( -Ll) (6.13)
/, 4 + x ,
(6.14)
Here ö+(x) is the right-hand Dirac function since its singularity is arranged to the
right from the coordinates of the triggering of the Ö - function. The symbol
8+ (x - Ll)signx is understood in the sense
ö+(x-Ll) when x>°
0+ (x - Ll)sigm ={
when -o+(x-Ll) °
x<
Due to symmetrical double-sided interaction, elastic and dissipative constants
take the form
Mx 2 (1 + R 2 )
<1> /, = - 4 Ö(lxl- Ll)Slgnx ,
•
(6.15)
Mx: (1- R 2) Us:(1 X 1_ L\A)' .

'*'
,n
f)
_
- slgnx . (6.16)
4
The symbols used here take the following meanings:
8+(X-Ll) when x>o,
o( 11
x - Ll)signx = {
-o_(x+Ll) when x<o,
.. {O+(X-Ll) when X>O,

8(11
x - Ll)slgnx =
-8_(x+Ll) when X<O,
where 8 _(x) is the left-hand Dirac function. Remember that dimensions of the
Dirac function are [Xl-I. The resulting outcome assumes that the impact process is
limited to the level x = ±Ll in the forward and reverse directions and
consequently the double triggering of the 0- function corresponds to the two
phases of impact. In this way,
,1,+E ~-6.-e
fo+(x-Ll)dx= fO_Cx+Ll)dx=l. (6.17)

Il. -Il.
Thus, when using stereomechanical impact theory, the elastic and dissipative
components of the dynamic characteristics of impact pairs appear to be dependent
on the coordinates and velocity of impact, i.e.

 (x, x) = j,, (x ,x) + D(X,X) . (6 .18)
When R = I, equations (6.14), (6.16) change to zero, and from (6 .13) to (6.15)
we obtain the force characteristic for the case of an totally elastic impact
. Mx
<l>j (x , x ) = -
2
- 8 , (x-ll), (6.19)
. 2
(6.20)
The force characteristics of the impact pair can also be obtained by considering
the direct central impact of two forward moving bodies as the result of the effects
ofa nonlinear elastic force oftype (6.5) or (6 .7) with an unlimited increase in the
coefficients CI or c2 ·
For single-sided and double-sided interactions, the limiting cases conform to
the static characteristic drawn in Fig. 6.2 a, b. Due to the presence of elastic bonds
of stiffness C between the interacting bodies, these characteristics acquire the form
shown in Fig, 6.2 c, d. It is natural that in this case, the extra elasticity can be
taken into account as the effective parallel to the nonlinear characteristic shown in
Fig. 6.2 a, b:
<l>(x) = cx + j'(x). (6.21)
In the approach to the study of vibro-impact systems using the elastic
interaction characteristics shown in Fig. 6,2, energy los ses during impact can be
ca1culated independently by introducing an equivalent dissipative force j) (x) ,
convenient for analysis, found from the conditions for an energy balance. In
addition, it is often expedient to take into account a specific regime of an actual
system, therefore the practical possibilities of such ca1culation will be presented
during the investigation of real systems. For us at the moment it is important that
with the help of different nonlinear expressions of the type studied, all the basic
forms of force interactions for colliding bodies or their contacting elements can be
written.
i+. W ~
o
a)
4:&
-.1
o
b)
.1 z
Fig, 6,2
cd
r:)
.1 J:
d)
2. No matter which of the nonlinear force characteristics, described above,

we substitute into equation (6.9), we will obtain the differential equation of
relative motion of the colliding bodies describing the total actual motion on all
time axes
Mx+(x,x)=O. (6.22)
The effects of the external force on the collision of the bodies and also the
additional elastic and dissipative connections are taken into account by the
introduction into equation (6.22) ofcorresponding additive terms.
Substituting M 2 ~ 00 in (6.9) we obtain M = MI' and equation (6.22) then
represents the impact of a body against a limiter. In the event, if the latter moves
according to the law xo(t) we must substitute x = XI - xo(t) in (6.22), where XI
is the absolute coordinate of the body. Using the impact pair characteristic
<l>{x, x») we can build-up, for example, the equation for oblique sliding impact of
the material point mass M on a fixed obstruction (Fig. 1.5). Based on the
hypothesis of dry friction we take for force G acting in the tangential direction,
G(x,x,y) = -e<l>(x,x)signy, (6.23)
where e is the coefficient of dry friction.
The equation for the normal coordinate X takes the form (6.22). For the
tangential coordinate y using (6.23)
My + e<l>(x,x)signy = O. (6.24)
The methods used to describe impact interactions can naturally be extended
into more general cases for two-dimensional and three-dimensional interactions.
Such cases are not of particular interest here since the purpose of this book is to
deal specifically with a study of vibro-impact systems, hence, later on we will
limit the systems considered to those with one
dimensional impact interactions. By introducing to
the structure of the dynamic model for such
systems, a non linear elastic dissipative element
<l>(x, x) to characterise the one-dimensional
interaction forces in impact pairs, it is possible to
construct the differential equations for vibro-
b
..
impact systems with an arbitrary number of
degrees of freedom by the usual methods. Fig. 6.3
We shall examine several examples. Fig. 6.3
shows the dynamic model of a simple vibrating system with an impact vibration
absorber. Writing the coordinates of displacement of the damper, relative to the
damped mass X z - XI through X we will write the equations of motion for each of
the masses in the form
Mix i = -nix i -CIX I -<l>(x,x) + P(t),
(6.25)
M 2X2 = <l>(x,x).
Writing the differential operator as s = d / dt , we can rewrite the equations of the
system (6.25) in the following form:
(M l s 2 + nls + CI )x I + <l>(x,sx) = P(t)
(6.26)
M ZS 2 X 2 -<l>(x,sx)=O.
Solving equations (6.26) accordingly relative to XI and x 2 and subtracting the
first from the second, after reorganisation we obtain the equation for the relative
coordinate x:
(M IM 2s 4 +M 2 n ls 3 +M 2 c IS2)X+
(6.27)
+ [(MI + M 2 )S2 + nls + CI ](x,SX) = -M 2s 2pet).
In general, for a stationary holonomic mechanical system with n degrees of
freedom in the presence of a single nonlinearity in the form of a one-dimensional
impact pair between the k-th and the m-th masses, the equations can be arranged in
the form
q/!ICs)x\+ ... +9Ms)xk+. "+~m(s)xm +. ·.+0')n(s)xn = !;(t),
+. ··+!?4n(s)x n = P,,(t),
(6.28)
91,I(S)X I+" .+qo"k (s )xk+.. '+~,m(s )xm+.. .+91J1lI(s)xn = P,,(t),

where 9?lp/S) = Mpqs2+npqs+cpq; Mpq,npq,cpq(p=l, ... ,n;q=l, ... ,n) are
coefficients ofpositive definite quadratic form
T=t LMpqxpxq, B=t L I1=t LCpqXpXq

11 11 11
npqxpxq,
p,q=1 p,q=1 p,q=l
characterising the corresponding kinetic energy, dissipative functions and

potential energy of linear parts of the system [24], obtained from the original
system by neglecting the impacting interacting mass (x,sx) == 0; xI' . ",x n are
generalised coordinates of the system; x = Xm - Xk are coordinates of the relative
disposition of the colliding masses; PI (t), ... , P" (t) are generalised external
forces.
Transferring the nonlinear terms in (6.28) to the right-hand side and solving
the system of equations as linear in terms of coordinates x k and x m , we find
x (t) = [?1.,(S) - ~mk(S)][X(t),sx(t)]+

k go(s) !Yl(s)
t ?J?k(S) P(t),
i=1 Q(I(s) ,
(6.29)
xm(t) = [- ?/km(s) + ~.m(S)][x(t)'SX(t)]+ t ::Jj~m.(S) P'(t),
Q(I(s)01(s) i=1 (yJ(s)
where
9I1n(S)
9?l(S) = (6.30)
911i(S) ~1lI(s)
is the characteristic determinant defining the linear part of the system; Y'A'pq (s) are
the algebraic adjunct of the (pq) - th elements. Because of symmetry of the
matrix, the coefficients of the quadratic form y,,?)(/ (s) == Y'I\/P (s). Both here and
later, we shall consider systems with total dissipation in which any motion is
accompanied by energy dissipation. For this reason, we will not consider

asymptotic time-damped processes caused by non-zero initial conditions.
Subtracting the first equation in (6.29) from the second we obtain for
coordinate x, similar to (6.26)
9l!(s)x(t) + 9II(s)<l>[x(t),sx(t)] = ta;(s)p,(t), (6.31)
1=1
where i?It'(s) = ~k (s) + ~m(s) - 29t,(s);J;(s) = ~m(s) - ~k(S). The equations for
9l!(s ),9\'(s ),a; (s) are represented as polynomials of the differential operator s in
which the degree of 9l(s) is 2n, the degree of i?It'(s) is 2n - 2 and the degree of
rt/; (s) does not exceed 2n - 2.
After defining the law of change of x(t) from (6.31), the displacement of the
impacting mass can be found from equation (6.29). For the remaining masses we
obtain similarly
x,(t) = [9~AS) - ~'(S)]<l>[X(t),sx(t)] + f ~,(s) P,(t) (6.32)
~(s) 9l!(s) i=\ 9l!(s)
(/ = 1,2,.. . ,n) .
We notice that in equations (6.29), (6.32) terms of the form
01/X/s)/9Jl(s) = L/X/s) represent dynamie eomplianee operators of the linear
portion of the system connecting the displacement of the p-th mass with the force
applied to the q-th mass
(6.33)
while Lpq(s) == Lqp(s). Later the operator Lpq(s) for the transmission of excitation
from one coordinate to another will be called the transfer dynamic compliance
operator to distinguish it from the loeal dynamic compliance operator obtained
from the transfer one when p == q .
Because (6.33) is linear, the expression for the operator Lpq(s) coincides with
the relationship for the Laplace transforms of the functions transformed by
operator Lpq(s) [79]
<Xl
Ix p(t)exp( -st)dt
o
00
I Pq(t) exp( -st)dt

o
Here, .[ is the Laplace transform operator; s is a complex value. For this reason,
when using the operator description (6.33) later, we will widely use the conditions
of identity between the operator Lpq(s) and the Laplace transform. In particular, it
follows that
I
xp(t) = Ihpq(t-t)Pq{'t)dt, (6.34)

o
where the impulse response hpq(t) is the inverse Laplace integral from the
dynamic compliance operator [78]
82 Chapter 2 Equivalent Iinearisation of vibro-impact systems
I 1 a+joo
hpq(t) = [ {Lpl/(s)} = - . fLpl/(s)exp(st)ds (t> 0). (6.35)
27CJ a- ( Xc
Here, .L,I is the inverse Laplace transform operator; / = -I; a is the abscissa
of convergence chosen to satisfy the condition
ce
flh p,/t)1 exp( -at)dt < OC! •

o
For a system with total dissipation a = 0 . In this case, when s = jm the Laplace
trans form be comes the Fourier transform.
Dynamic compliance operators can be
constructed not only for a system with a finite
number of degrees of freedom, but for also for
continuous systems and for systems consisting of a
I PI~'t) ~
-.JL!J
mixture of structures including elements with
lumped and distributed parameters. In general, these
operators can direct1y form equations like (6.29) for 2 Mt}
.
JI,
elastic systems of sufficiently general form .
Let us examine initially a one-dimensional Fig 6.4
vibro-impact interaction, (Fig. 6.4) between two
elements 1, 2 (particles, plane sections, etc.) of a linear stationary system of
general structure under the action of external forces ~ (t) and P2(t) applied to the
contacting elements. Calling their absolute displacements XI (t) and Xz(t) and
introducing the characteristic of impact pair cl>(x, sx) , where x(t) = Xz (t) - XI (t) ,
we will write the corresponding displacement through dynamic compliance
operators for the elements
XI (t) = L (s ){cl>[ x(t), sx(t)] + PI (t)} + L z, (s){ -cl>[x(t), sx(t)] + Pz (t)},
" (6.36)
Thus, to describe the interaction of impact pairs, it is necessary to know the matrix
of dynamic compliance operators for impact pair A(s)
A(s) = II L II (s) L 2 J S)11 . (6.37)

L,z(s) L22 (S)
as weil as the force characteristics, cl>(x, sx) .This matrix is symmetrical since
L12 (s):; L 21 (s). In the event when the elements of the impact pairs belong to a
disconnected system, LI2 (s) :; 0 and matrix (6.37) takes the diagonal form .
Introducing the nomenclature LII(S)PI(t)+L21(S)P2(t)=XIO(t);
LI2(S)PI(t)+L22(S)P2(t)=X20(t) and rewriting the system of equations (6.36)
using this nomenclature, we obtain
XI (I) = x lO (t) + [LI I (s) - L 21 (s )]cl>[ x(t), sx(t)],
(6.38)
The express ions for xlO(t) and x 20 (t) define the vibrations ofthe elements ofthe
impact pair in the absence of impact interactions between them and completely
characterise the external excitation applied to the impact pairs. In many cases, this
relationship and the elements of the dynamic compliance operator matrix can be
defined from the results of direct measurement on the object under study. For
example, it is easy to apply when investigating vibro-impact interactions of two
independent mechanical systems since in this case, L12 (s) == 0 and the
characteristics specified can be defined separately for each system. Applying some
standard of force excitation in the form of a unit function 1'](/) or an impulse
function 0(/) to the elements of the impacting pair under zero initial conditions
and taking the reactions as g(t) or h(t) we can define the dynamic compliance of
i - th element (i = I, 2) by the formula for the Laplace transform
00 00
Lii(s) = s fg(t)exp(-st)dt, Lii(s) = fh(t)exp(-st)dt.

o 0
Since, when s = jm, the corresponding operator coincides with the dynamic
compliance of the systems, they can be measured directly as the response on the
standard sweep sine test. As a result, we find
Lii (jm) = ILii (jm)1 exp[j<p ii (m)] ,
where IL'i (jm)1 is the gain coefficient of the harmonic signal at frequency
m; <Pii(m) is the value of its phase shift. Currently, special apparatus is available
to measure dynamic compliance [75, 199]. The value of x iO can also be directly
recorded by applying an extemal force ~ (I) .
Similar methods to describe the characteristics can usually be used even in the
presence of linear coupling between the elements of the impact pairs, i.e., when
L]2 (s) i O. In order to exc1ude vibro-impact interaction in this case, it is
necessary, with great care, to normalise the standard signal. To record the value of
x,o (I) we can alternately switch on the sources of excitation with sub se quent
summation ofthe resulting measurements.
Let us return to the examination of the system of equations (6.38). Subtracting
the first equation from the second we obtain the expression for the relative
coordinate x:
2
x(t) + L(s)<D[x(/), sx(t)] = Ißi (s)~ (t) . (6.39)
1=1
Here, L(s) = LII(S)+L22(S)-2LI2(S); BI(s) = L 12 (s)-L 11 (s);

B2(s)=L22(s)-L21(s). Writing the operator equations (6.39) in the form
L( s) = PII( s) / '{J;l( s); B, (s) = rV; (s) / !JZ)( s), we come once again to equations of
type (6.31). However, in the presence of elements with distributed parameters in
the system, the corresponding operator will not be rational but transcendental
functions for the argument s.
Proceeding from the description of the operators as transfer functions to the
time domain similar to (6.34) we obtain from (6.39) nonlinear integral equations
which completely describe the process of the relative change in coordinate x(/),
either in transient or stationary regimes

2 I
I
f
x(t) + f h(t - t)[x( T), x('r)]dt = f hi (t - t)~ (T)dT . (6.40)

i~1 0
Here, h(t) = I I {L(s)}; hi(t) =I I {Bi(s)}. The right-hand side of equation

(6.40) is a known time function, characterising the relative motion of the impact
pairs, in the absence of impact interactions. Calling this g(t) we obtain
f
x(t) + fh(t - t)[x( t), x( t)]dt = g(t) . (6.41)
Equation (6.41) is similar to the integral equation for contact interaction (1.44)
considered in § land is generalised in cases of vibro-impact interaction,
characterised by the systematic change from contacting to non-contacting motion
of the elements of the impact pair. This describes a process starting from zero
initial conditions. Taking into account the extinction of transient motion in the
systems described, integral equations for the stationary process can be written in
the following form, which do not depend on the initial conditions
f
x(t)+ fh(t-T)[X(T),X(t)]dT = g(t). (6.42)
3. The transformation ofthe variables in equation (6.41) can be represented

by the schematic diagram shown in Fig. 6.5. Here, I is a nonlinear element
transforming the dynamic process x(t) studied into the force impulses
F(t) = [x(t),x(t)] corresponding to the characteristics of the impact pair, 2 is a
linear dynamic block describing the elastic vibration X(t)
f
X(!) = fh(t - T)F(t)dT (6.43)
under the effect of the force impulse, the

shaded sector of the comparator 3 denotes
subtraction. A similar structural scheme has
traditionally been used for a wide range of
nonlinear systems: relay operated [212], with
variable structure [83] etc. The specific
features of each cIass of system reflects to a
large extent the character of the transformation
l lt)
produced by the nonlinear elements.
As already mentioned, the characteristic of "lg. 6.5
vibro-impact systems appears more strongly in
the equations when using the hypotheses of stereomechanical theory for impact,.
hence we will consider transformations wh ich, for example, consist of nonlinear
elements with the characteristics given in (6.19). We will assume that collisions
occur at a certain instant of time t v with relative velocity x,,_ (v = I, 2, ... ) . We
will define the impulse J" of the force (6.19) for both phases of impact at the
instant t" :
Jv ='7 MX;- Ö + [X(t)_~]dt=

1,, -0 2
Mx ;_
2
[7ö+ (X~~)dx + ö+ (X~~)dx],
<'> XI
f
<'> +e XII
where x p XII are the average velocities in the land II phases of impact. Since
XI = < /2,x lI = -x v_ /2 in the case considered using (6.15) we find J v = 2Mx v _ •
This result, of course, can be obtained straight away from impulse-momentum
equation. Thus, as a result of transformation of process x(t) by the non linear
element, the outcome will be aseries offorce impulses, e.g.
F(t) = <l> tJx(t), x(t)] = 2M~>v_ ö(t - IJ
where Iv is the instant when x(/) reaches the value X = ~. For a periodic process
with one collision per period T, taking into account the energy loss during impact,
we obtain, similarly
00
F, (I) = M(l + R)x_ Iö(t - vT). (6.44)
For symmetrical vibro-impact process, caused by double-sided impact pairs with

alternate impacts on the stops
Fz (t) = M(1 + Rh_[.tooÖ(1 - vT) - "too ö(t - vT - ~)] = F, (t) - F{I - ~) .(6.45)
Consequently, the nonlinear elements in the schematic structure of vibro-
IA ~(ZJI / F{t)1 A ~
wnuw
F(f) fI
~ o I
Fig. 6.6
impact systems trans form the continuous process into successive impulses,
modulated by the velocity of the input process at the instant when the process
reaches the threshold value. Such nonlinear components of the structural scheme
we will call impacI elements. The impact elements have posilional-impulse effects
which specifically combine the effects ofthe relayed and impulse elements. When
using descriptions of vibro-impact systems based on other impact theories, the
actual form of transformation applied to the impacting elements changes and is
reflected by the form of the generating impulse, but the general character of the
effects of impact elements is preserved. Fig. 6.6 shows examples of
transformations representing impact elements with characteristics (6.7), (6.8).
The force impulse generated by the impacting elements is absorbed by the
linear block 2 (Fig, 6.5) for which, the transfer fun ction, according to the results of
section 2, takes the form of a dynamic compliance operator for the elastic system
L( s) = ~(s) / 0J( s) . We will pause here to consider the characteristics of dynamic
compliance which has considerable importance for the study that folIows.
For the linear part of the system (6.28), L(s)takes the form of a rational
fraction
Oll() h 211-2 h S 211-3 h
L(s) = _,7l_S_ = oS + I + ... + 211-2 (6.46) 0
@(s) d oS211 + d l s 211 - 1 + ... + d 211

This shows that the dynamic compliance L(jm) has a filtering effect, i.e.,
limIL(jm)1 = O. (6.47)
Ol-->""
Remembering that the impulse response he,) for the system with operators of the
form (6.46) can be defined by the formula [97]
(6.48)
where
H = 1 m I
- [d
(S-Sk)"k9l'(S)]
km (m -1)! ds m - I 0J(s)
S=.\'"
Here, I is the number of different roots Sk of the characteristic equation for the
system
@(S)=O; (6.49)
where vk is a multiple for the k - th root (VI + V 2 + ... + VI == 2n) . In the following
we will limit consideration to systems having only simple roots. In this case,
211 9l'(s )
h(t) = I-.-k-exp(skl) (l;::: 0). (6.50)
k"1 9tJ'(Sk)
where :JjI'(Sk) = (do/J/ds)'"'k .
We are interested mainly in the resonant response of vibro-impact systems.
These are displayed if the system is weakly dissipative, i.e .• if the coefficient n pq
of the quadratic form B which represents the dissipative effect of the linear part of
(6.28) is small.
Let us rewrite the equations for the linear part ofthe system (6.28) with normal
coordinates zp"., Z2 • Using [95]
11
S2 Zp + Ibpqsz q +O!zp == Qp(t) (p == l, .... n), (6.51 )
q"1
whence
n
I I
11
Zp == ApqX q , Qp (t) = ApqPq (t) , (6.52)

'1"1 q"1
where 0 p are the natural frequencies of the conservative system; bpq are small
quantities: Apq is the normalised coefficient of the p - th mode of vibration.
According to (6.51) the characteristic equation (6.49) now takes the following
form:
9ZI(s) = =0. (6.53)
Expanding the determinant and neglecting the products of sm all terms we obtain
n
9JI(s) = g(s' + 2~,Ovs + O~) = 0 (~ = bw /(20J). (6.54)

As a result, the roots of the characteristic equation will be complex with the
conjugate pairs
Sv "" -r"Ov ± jOv , (6.55)
On the basis of the assumptions made the solutions for the equations (6.51) can be
written in the following operator form:
Qp(t)
Zp(t)= 2 2 (6.56)
s + 2r/:>.ps + Op
Using (6.52), (6.56) we find the connection between the original coordinates and
the excitation ofthe linear system
n 11 A n
Xp(t)=LAvpzv(t)=L 2 ']' 2LA vk Pk (t)· (6.57)
v~1 1'=1 s + 2rvO"s + 0v k=1
Assuming that Pk (t) == 0 when k :cF q in (6.56), P/t) = 1 , we obtain in agreement
with (6.33) the equation for the dynamic compliance operator
~ AvpA"q
Lpq(s) = L.. 2 2 • (6.58)
1'=1 S + 2r"Ov s + 0v
According to (6.58)
Lpq(s) == Lqp(s) (6.59)
When s = JOO , ignoring small terms of second order we find
L (')
P'l JOO
~
= L..
2
AvpA"q
2 2 2
(r'\2 2· )
2 2 ~'v - 00 - j2rvO vOO . (6.60)
,,=1 (Ov - 00 ) + 4rv Ov OO
Note that when the energy dissipation is neglected, then the imaginary part of
the dynamic compliance reduces to zero. The modulus of dynamic compliance is
found to be
. I= L..
ILp/jOO) ~
I
AvpA vq
. (6.61)
1'=1 ,,(O~ _00 2)2 +4rv20~oo2
Hence, the dynamic compliance operator for the linear system studied with n
degrees of freedom, can be represented in the form of a sum of n resonant filters,
with natural frequencies ofthe conservative system. When 00 = 0v' a peak occurs
in the dynamic compliance due to the appearance of the small quantity 2r"O~ in
the denominator of the v-th term. The width of the peak defines the band-width of
the filter, and is proportional to the value rvO v ' As the number of modes of
vibration v increases the "amplitude" ofthe peak decreases rapidly. Fig 6.7 shows
examples of the expression (6.61) for the case p = q. A detailed analysis of
frequency characteristics of dynamic compliance as described here can be found
in [123].
In a similar way, the equations for the dynamic compliance operators of a
series of systems with distributed parameters can be presented, for example,
straight elastic rods undergoing axial, torsional or transverse vibrations [24].* The
number of vibration modes for this increases infinitely, and the coefficients for
each mode remain functions of the continuous coordinate x, which represents the
position of the section under consideration. Denoting these coefficients as Av (x)
we find when transferring the action to section x = p from the concentrated force
applied at the section x = q,
L () = ~ Av(p)Av(q)
pqs L..2 2' (6.62)
v~1 S + 2rvQ vs + Qv
During the treatment of such systems, the mechanism of energy dissipation is
frequently represented in the form of internal friction in the material expressed in
terms of the absorbtion coefficient % (see § I section 2). In this case,
1"" = %Q)(41tCo) [176] and the equations for the dynamic compliance take the
form
(6.63)
whence
IL (jm)1 = "~I~(Q~
pq
i: A, (p) A" ( q)
-m')' +%'Q~ /(4n')
. (6.64)
As the mode number v increases, the second term under the radieal grows
rapidly, therefore the practical value can only take a few initial modes.
Frequently during the study of the mathematical model of real linear systems,
the equations for dynamic compliance can be
determined directly by searching for a steady
state solution due to the effect of harmonic
forces with unit amplitude. A convenient
procedure will be explained later (see § 15
section 1, §16, section 2).
Fig. 6.7
4. To understand the character of the
transformation of the variables occurring in the description of vibro-impact
systems of the type which can be reduced to the structural scheme of Fig. 6.5, let
us consider a process undergoing a T-periodic sequence of impulses of the form
(6.44) or (6.45), through a linear dynamic block with characteristic L(s),
represented as the sum of dynamic compliance operators of the form (6.58).
Impulses are produced by impacting elements when the system performs periodic
motion with one impact against each stop during aperiod.
* In [241 instead of dynamic compliance, the name harmonie influenee eoefJieient is used.
Functions (6.44), (6.45) define the following Fourier series [191]:

J 00
F, (t) = - 2)xp(mjrot) (/ = -I), (6.65)

T m~-oo
2J
=-
00
F 2 (t) 2)xp(2m + I)jrot (6.66)

T m~ -oc
where J is the value of the impact impulse; ro = 2n / T .
These series are divergent, however, they can be considered as generalised
periodic functions, wh ich allows regular operations to be performed with them,
such as addition and multiplication, differentiation, integration etc. [191, 196].
Applying the principle of superposition as the excitation (6.65) or (6.66) pass
through a linear stationary system with frequency characteristic L(jro) the
resulting periodic process will take the following form :
X\J(t)=:!.... 'IL(mjro)exp(mjro() (6.67)

T m~ -oo
in the non-symmetrical case, and
2J
=-
00
Xv(t) IL[(2m -1)jro]exp(2m -I)jrot (6.68)

T m~-oo
in the symmetrical. For some functions L(jro), series (6.67), (6.68) can converge.
Since only these cases conform to actual process in vibro-impact systems, we can
find the general criterion for convergence ofthe series (6.67), (6.68). Assuming an
integration of quadratic function of Xi.! (t) and using Parseval's equality [90] we
find for the non-symmetrical case
r J2 00 J2 2J 2 00
fx~J(I)dt = - IL(mjro)L(-mjro) = - , IL(0)1 2 +-IIL(mjro)1 2 . (6.69)

o T m ~ -oo 1 T m~'
A similar equality can be found for the series (6.68).
The filtering effect of the linear part of the system, (6.47), is a necessary
condition for the convergence of series (6.69). Since for sufficiently large m, each
term in (6.61) is of the same order as the quantity (mror 2 , series (6.69) also
demonstrates sufficient conditions for convergence [90] . For this it is necessary
that the purely imaginary poles and poles with positive real parts are absent from
the operator L(s). Since the poles of the operator L(s) are roots of the
characteristic equation (6.53), according to (6.65) the conditions indicated are
satisfied for the elastic system considered, because
of total dissipation. The existence of zero poles for
the linear part of the system is also permissible.
This corresponds to free motion as a rigid body.
Thus, the real qualities of a mechanical system
ensure that the harmonic series (6.68), (6.69)
converge to the periodic functions XII (t)
or XV (I) with period T. These functions, which
are the stationary reactions of the linear system to Fig. 6.8
the periodic 0 - function, have been given the
name impulse-frequency characteristics 01 the first and second types [191].

Below, we shall call them periodic Green 's lunctions, following mechanical
traditions. It can be shown that for linear operators of type (6.46), with simple
poles Sk' the periodic Green's functions XI(t) or X2(t) (the absence ofthe index
J implies that the functions are considered when J = 1) can be defined by the
formulas
(t E (0, T»,
(6.70)
and if, among the roots Sk there are purely imaginary ones, then frequency
Cü = 21t 1T must not be equal to or a multiple of them. From now on, for brevity in
such cases, when the specific characteristics XI (t) or X2 (t) have no role we will
omit the numerical index.
The periodic Green's functions are linearly related to the value ofthe impulse
i.e.,
X./ (I) = h(t) , (6.71 )
We note that the degree ofthe polynomial in the denominator which is two orders
greater than that of the numerator in the operator L(s) ensures that the function
X(t) is continuous along the whole time axis. However, the derivative, Xt (t)
takes ajump when t = mT, and X2(t) when 1= mT 12 (m = 0, ± 1, ... ) [191].
Suppose, for example,
(6.72)
i.e., the linear part of the system has represents an oscillator with viscous
damping. Then from formulae (6.67), (6.68) we find
( ) =~ ~ exp(mjCüt)
XI t L. .
°-
T m~-oo bmJCü +
2 2 2'
m co
(6.73)
2 x exp[(2m - l)jcot]
(6.74)
X2(t) =T m"foo b(2m -1)jco + 0 2 - (2m _1)2 co 2
Assuming that 0 2 > b2 14 we find from formula (6.70)

exp( -bt 12) sin 0ll + exp( -bT 12)sin 01 (T - t)
XI ()
I =~~--~----~--~~----~--~--~- (t E [0, T]), (6.75)
01 1+ exp( -bT) - 2 exp( -bT 12)cosO I T
X2(t) =
exp(-btI2) sinOJ+exp(-bTI4)sinO t(t-TI2) (tE[0,T12]), (6.76)
0t 1 + exp( -bT) + 2exp( -bT 14)cosOtT 12
where 01 = ~0 2 - b2 14 is the natural frequency of the oscillator. Fig. 6.8 shows

examples obtained for this relationship constructed when T = 21t S, b = 0,
0= 0.51/ s . Note that periodic Green's functions have the dimensions of [s/kg].
5. With the help of periodic Green's functions, the periodic single impact
solution x(t) of the equation of a vibro-impact system for relative motion of the
elements ofthe impact pair can be written in the form (see Fig. 6.5)
x(t) = Set) - x./ (t) , (6.77)
where Set) is the periodic relative motion of the elements obtained from the
solution of the corresponding linear equations with prescribed excitation when
impact interactions are neglected. Note that in the same form (6.77) we have
previously obtained the final expression for the law of motion when solving the
problem of forced vibration of an oscillator impacting against stops (see (2.14),
(3.l3)). In those cases, S(t) was written in the form
S(t) = acos(cot+<p), (6.78)
where a is a value found from the solution of the linear problem; <p is the
unknown phase of the motion relative to the instant of collision which is
considered as the origin for time measurement. The solution S(t) can be written
in the form (6.78) for systems of arbitrary dimension excited by harmonic external
force.
The circumstances specified allow us to develop the foUowing compact form
for exact periodic solutions. Using (6.71), (6.78), the periodic single impact
solution (6.77) takes the form
x(t) = a cos( cot + <p) - JX(t) , (6.79)
where the unknown phase <p and impulse J are defined by the conditions of
collisions
x(O)= ~, J = (1 + R)Mx-CO), (6.80)
Taking account ofthe coordinate axis direction we must also have J? O.
Applying the conditions (6.80) to (6.79), we find
acos<p=~+JX(O), aSin<p=-J[L(O)+
co
1
(1 + R)M
]. (6.81)
Here, X_CO) is the left value ofthe derivative ofthe periodic Green's functions at
the point of discontinuity. Since the periodic Green's function is usually assigned
in the sector [0, 11 in the non-symmetrical case or [0,T/2] in the symmetrical, then
the value of L (0) is defined in the first case from the equation X- (0) = XI- (T)
and in the second case from the equation L (0) = -X2- (T / 2) .
Solving the system of equations (6.81) in terms of the unknowns we obtain
finally
- ~X(O) ±
(J ? 0), (6.82)
sin<p = -~[L (0) + 1 ], cos<p = Lid JX(O) , (6.83)

ma (l+R)M a
For perodic multi-impact regimes with v consecutive impacts per period against each stop the
solution can be found in the form
l'
x(t) = 'Ö(I) - LJ,X(t - IJ, (6.84)

I:~ J
The unknown value of the impact impulse J, and the instant of colliswn I, can be found from a system
oftranscendental equations obtained from (6.84) using collision conditions like (6.80)
x( I,) = Ll, J, = (I + R) Mx_ (t, ) (Jj 2 0)
The solutions (6.79), (6.84) must be tested in the absence of additional

crossings: x(t):<;; L'1 in the case of single-sided stops and Ix(t)I:<;; L'1 for
symmetrical double-sided stops. In § 14 section 2, a method to study the stability
ofthese solutions will be described.
Calculation schemes which represent exact methods to find periodic solutions,
are to be preferred when dealing with large systems by comparison to the
procedures applied in Chapter I, since the use of periodic Green' s functions in the
form of series (6.67), (6.68) allows us to avoid time-consuming procedures to find
the roots ofthe characteristic equations obtained when writing the general solution
in the form (5.6).
Moreover, this method allows the study of systems with distributed
parameters, and also of systems where the linear elements are described only in
terms of the dynamic compliance characteristics of the contacting elements. An
important detail is also that, by limiting series (6.67), (6.68) to finite terms using
filtering effects of linear parts of the system, a continuous approximation to the
exact solution can be obtained. In particular, by neglecting terms of the series
corresponding to m > 1 we obtain a harmonic approximation, which has an
important meaning for the understanding of the qualitative picture of the process,
as will be shown later.
It is just as important that it is possible to expand the operator L(jm), which
appears in (6.67), (6.68) as aseries in terms of the natural mode of vibration for
the linear part of the system, and to reduce the basic mode, which in practice
influences the regimes considered in the frequency range investigated. This
procedure was introduced into the study of vibro-impact systems in the works of
Zevin [224, 226].
Using periodic Green's functions allows us to convert equation (6.42) to
Hammerstein's type of integral equation when we describe the periodic motion of
a vibro-impact system in terms of any form of force characteristics for an impact
pair (x, x), allowing us to construct effective schemes for continuous
approximation and to estimate the errors of the approximate solutions [191]. For a
steady periodic solution x(t) expanding the periodic function F(t) = [x(t),x(t)]
as a Fourier series, we find
L Fm exp(mjmt) ,
x
F(t) =
m",,-r
where the coefficients of the series are defined by the formula

l'
Fm = ~ fF('r)exp( -mjoYC)dr. (6.85)
To
For the basic structural scheme of Fig. 6.5, we find, using superposition of the
motion on the output from the linear block,
I
00
x(t) = 3(t) - L(mjro) Fm exp(mjrot). (6.86)
Substituting (6.85) into (6.86) and changing the order of integration and
summation we obtain
x(t) = 3(t) - I
1'[T1 m"'f",L(mjro)eXP[mjro(t - t)] F(-r)dr
00 ]
Using (6.67), (6.71) we can see that the term in the square brackets represents the
periodic Green's function x(t -1) for the linear part ofthe system. Using this we
eventually obtain
T
x(t) = 3(t) - fx(t -l)<1>[X(l),X(t)]dl, (6.87)
where 3(t + T) = 3(t) .

According to equation (6.87) the character ofthe periodic motion ofthe vibro-
impact system depends weakly on the actual description of the force interaction
during impact and consequently on the form of the function <1>( x, x). Obviously,
examining the integral on the right-hand side
l' e
fx(t - t)<1>[x(l),x(t)]dl = fx(t -l)<1>[X(l),X(l)]dl,
o 0
where the sm all value E limits the length of the interval during which
<1>[X(l),X(l)] ~ 0 we find using average theorem and assuming continuous
periodic Green's functions *,
e e
f X(t -l)<1>[x( 1), X( l)]dl = X[t - E8(t)] f<1>[X( 1), X( l)]dl =
o 0
(6.88)
=Jx[t - E8(t)] (0< 8(t) < 1).
Here
e
J = f<1>[X(t),x(t)]dt (6.89)
o
is the impulse force interaction during impact.
Noting that E8(t)« T and consequently for continuous periodic Green's
function X[t - E8(t)] '" X(t), we conclude that the force characteristic <1>(x,x) in
practice only influences the motion through its integral (6.89) in small areas of
* A sufficient condition for continuous periodic Green's function corresponds to the conditions for
uniform convergence ofthe Fourier series ofthe form (6.67), (6.68) and consequently can be written in
the form [185] 1L(mjm) 1= O(l/m 1+ A) , where A is an arbitrary positive number.
integration E and consequently it can be approximated by any convenient

expression leading to the identical value for the integral. For this reason, during
the study of motion of vibro-impact systems, it is expedient to use the simplest
method to describe the contact interaction.
From the transformation (6.88) the following important point folIows, that
whatever method is used to describe contact interaction for vibro-impact systems,
when the linear part is represented by continuous periodic Green' s function, then
the single impact solution of (6.87) is elose to the following:
x(t) "" S(t) - Jx(t) . (6.90)
The general scheme for successive approximation for equation (6.87) takes the
form
fX(t - T)cI>[X, (T),

7'
Xi +1 (t) = S(t) - Xi (T)]dr , (6.91)
where the correct choice for the initial approximation Xo (t) has an important
influence on the convergence of the iteration process. Calculation by continuous
approximation does not require us to revert to the integral equation (6.87) but can
in practice be done by introducing the general operator
Xi + 1 (t) = Set) - L(jro)cI> [x, (t),x,(t)] . (6.92)
A routine method to find the initial approximation for the periodic solution of a
nonlinear system is the method of harmonic linearisation. A description of this
method, adapted for systems with vibro-impact interactions and used for the
analysis of the basic qualitative effects of solutions for various models of vibro-
impact systems, will be dealt with at considerable length in the following two
chapters of this book.
In conclusion, we direct the reader's attention to the possibility of directly
constructing transfer processes for vibro-impact systems based on (6.42) using
numerical procedures as described, for example, for integral equations in the
monograph [169].
§ 7. Harmonie linearisation of vibro-impaet systems

1. Let us now turn to a description of the idea and methods of harmonic
linearisation of vibro-impact systems [34, 37]. The basis of the harmonic
linearisation method lies in the principle of harmonie balance.
Let us examine the operator equation (see §6)
x(t)+L(s)cI>[x(t),sx(t)] = S(t) , (7.1)
where S(t) is a given periodic function with period T = 2rc / ro. Assuming that
(7.1) has a periodic solution of identical period T, we can write it in the form of a
Fourier series
00
x(t) = l>m exp(mjrot) , (7.2)

nl=-OO
00
x(t) = jro ~>mmexp(mjrot). (7.3)

m=,,··YJ
Substituting (7.2), (7.3) in the equation for the nonlinearity cI>(x,x), we obtain
§7 Hmmonic linearisation of vibro-impact systems 95
a similar T-periodic function of time F(t) = [x(t),x(t)] , which can be

represented by the following Fourier series:
00
F(t) = L Fm exp(mjrot), (7.4)

m= - oo
where the coefficients Fm are defined by formula (6.85). Finally, the expansion of
the external periodic excitation 3(t) can be written in the form
00
3(t) = L3 mexp(mjrot) (7.5)

m=-cc
Substituting series (7.2) to (7.5) in equation (7.1) and equating the coefficients for
identical exponents we find
X m + L(mjro)Fm(xO'xpx_ p. .. ) = 3 m (m = O,± 1,±2,... ) (7.6)
The solution of the transcendental equation (7.6) as an infinite series is not
gene rally possible, but with several additional assumptions, this system can be
truncated to finite values.
a)
Fig. 7.1
We have seen above (see §7 section 3), that the actual filtering effect of a
linear system occurs only in a limited frequency band, hence, assuming that for
some value m > k
L(mjro) == 0, (7.7)
we find from (7.6) for these values
(I m I> k) (7.8)
The linear system with frequency characteristics (7.7) is known as an ideal
filter . In such cases, when the actuaI system can be ideaIised in the same way, then
using (7.8) the system of equations (7.6) becomes as folIows:
XIII + L(mjro) Fm (X o, X p X _P",'Xk'X _k ) = 3 m (7.9)
(m=O,±I, ... ,±k)
and can be solved by either numericalor analytical procedures.
The simplest such system of finite equations arises when k = 1, i.e. in the case
when the ideal filter does not pass harmonics higher than the fundamental.
The predominance of the fundamental over other harmonics in the vibration
spectrum of the system is not only possible when limiting the bandwidth passed
by the linear system (Fig. 7.1, a), but also when it demonstrates a strong resonance
response (Fig. 7.1, b). Assuming that the vibration frequency coincides with
resonance, we may say in this case that L(jm)>> L(mjm) (Iml>
1), and
consequently idealisation (7.7) remains justifiable.

For the actual system, condition (7.7) is not exactly satisfied, and so we can
only proceed to an approximate search for a periodic solution in the form of
limited Fourier series when either jiltering or resonance effects are present within
the linear parts ofthe system [1]. Obviously, the closer the equation represents the
effects described for the system, the better the degree of approximation. As was
shown in §6 section 3, the actual characteristics of elastic systems ensure the
effects described, hence it is completely justifiable to apply the approximations
described in the study ofvibro-impact systems.
When considering exact solutions to the problem of vibration of oscillators
with stops, we saw that the basic qualitative information of the nonlinear
behaviour of vibro-impact systems is completely reflected by the first harmonic,
thus the search for and analysis of the first harmonic is expedient even in those
cases when the filtering or resonance effects of the system do not appear strongly
in the equations. The quickness and clearness of information about the basic
non linear effects of the system obtained in this way allow us to form the initial
concepts to construct rational numerical procedures and to estimate them
qualitatively.
Suppose S(t) = mg + acos(mt + <p), then So = m B , S tI = ~exp(±j<p),
2
Sm == 0 when 1m I> 1 . Searching for a solution x(t) in the form
x(t)=m x +axcosmt+xll(t) = xh(t) + xll(t) (7.10)
where XII (t) contains higher harmonic components only, we find
x a = mx ' X±l = a x /2. From the system of equations (7.6) we obtain for the first
two components
m x + L(O)m f = m g (m" == Fa)' (7.11 )
a a
_x + L(jm)F1 = -exp(j<p). (7.12)
2 2
Equation (7.11) determines the connection between the periodic components
ofthe solution and the constant component mx :
m x = mg - L(O)m". (7.13)
Using (6.85) we can rewrite equation (7.12) in the form
2T
ax + L(jm)y: fF(l)exp( -jml)dr = aexp(j<p). (7.14)
o
We can represent the integral in equation (7.14) in the following form:
~ fF (l)exp( - jml)dl =
l
To
(7.15)
= ~ f[X(l),X(l)](COsml - jsin ml)dl = aJk + jmß),
T o
where
§7 Harmonie linearisation of vibro-impaet systems 97
k = _2_f<D[x(t),x(t)]COSWtdt,
Tax 0
(7.16)
r
ß = _2_ f<D[x(t),x(t)]sinwtdt,
Twa x 0
As a result, equation (7.14) can be written as:
a x [1 + L(jw)(k + jwß)] = aexp(j<p) (7.17)
This is an exact expression for the parameters of the fundamental. For known
values of k and ß, separating (7.17) into real and imaginary parts, we obtain a
pair of equations to define a x and <p :
a
sin<p = _x [kImL(jw) + wß ReL(jw)], (7.18)
a
a
cos<p = _x [I + k Re L(jw) - wß Im(jw)] . (7.19)
a
From these equations we find
a
ay = , (7.20)
. ~[I + k Re L(jw) - wß Im L(jW)]2 + [k Im L(jw) + wß Re L(jW)]2
k Im L(jw) + wß Re L(jw)
tan<p = . (7.21)
I + k Re L(jw) - wß Im L(jw)
In fact, according to (6.85) and (7.16) the coefficients m,., k and ß in turn
depend on the parameters ofthe solution for x(t), hence express ions (7.13), (7.20),
(7.21) can be used as equations to find these parameters.
Note that the system of defining equations (7.11), (7.17) which are appropriate
in the search for components ofthe solution xh(t) = m x + a x coswt, can be found
direct1y from the original equation (7.1) by the substituting in it for the nonlinear
function <D(x,x), the linearised equation
<D(x,x) = m/o + k(x h - mJ + ßX h (7.22)
The substitution indicated differs from the usual linearisation for a non linear
function in the fact that the coefficients m /., k and ß depend on unknown
parameters in the solution x(t).
Finding the coefficients mF' k and ß is connected with the formation of
several apriori concepts concerning the relative character of the periodic solution
x(t). In such cases, when the operator for the linear part of system L(s) gives rise to
vigorous qualities for the filtering or resonance effect, the value of XII (t) in
equation (7.10) becomes small and from (6.85), (7.16), changing the integration
variable, we obtain
I 2n
mF =- f<D(m x + a x cos\jJ, - axwsin \jJ)d\jJ (\jJ = Wt), (7.23)
2n 0
I 2n
k = -- f<D(m x + a x cos\jJ, - axwsin \jJ)cos\jJd\jJ, (7.24)
na x 0
98 Chapter 2 Equiva1ent linearisation of vibro-impact systems
1 2n
ß =- - - f(m x + a x cos \jJ, - axffisin \jJ)sin \jJd\jJ . (7.25)
naxffi 0
The selection of coefficients for equation (7.22) in the form (7.23) to (7.25) is
called harmonie linearisation [181]. Here, the linearisation coefficients depend on
m x and a x ' Substituting (7.23) to (7.25) into (7.13), (7.20) we obtain a pair of
coupled equations to find m x and a x • After finding these values the phase <p is
found from (7.21).
As an example, we will look at the forced vibration of a linear oscillator with symmetric rigid
stops (Fig. 3.1). Introducing the static characteristics for a double-sided impact pair (x) (loss of
energy during impact is not taken into account here), we will write, bearing in mi nd condition (3.8), the
differential equation ofthe system in the following form:
Mx+ nX + (x) = Pcos(rot+<p), (7.26)
where, because ofthe symmetry ofthe system <1>( -x) = -(x) and, as a result mx '" 0 . Searching for
an approximate solution x(t) for equation (7.26) in the form
(7.27)
we expand the symmetrical periodic function (ax cosrot) as a Fourier series, limiting the expansion
to the first term
(7.28)
where
Zn
k(ax)=_I- f<l>(a x COS\If)cos\lfd\lf (\If = rot) . (7.29)
]tax 0
Substituting (7.28) in (7.26) we will rewrite the latter in operator form
[Ms 2 +ns+c+k(ax)]x = Pcos(rot + <p) . (7.30)
For the solution ofthe linearised equation (7.29) in the form (7.27) we find
P P
a = =-;========= (7.31)
r IM(jco)2 +ry'ro+c+k(a,)1 Jlc+k(arl- MC0 2]2 +n2co2
tan<p = nro ![c + k(a x ) - Mro 2 ] . (7.32)

These equations are similar to (7.20), (7.21). The first,
taking into account (7.29) defines the unknown amplitude
2
of the fundamental ax ' Once that has been found the
solution for <p is obtained from (7.32). 11::
For the system considered (7.26) the amplitude- ,1
frequency characteristics can be constructed without
solving equation (7.31) based on the following reasoning.
The function roo(ax)=~c+k(ax) which fixes the 0
relationship between the natural frequency of the Fig. 7.2
linearised system and the amplitude of vibration,
represents an approximate equation for the backbone curve. We will use it to replace the exact
backbone curve for the nonlinear system considered, which takes the simple form I (Fig, 7.2).
Substituting ro 2 = c + k(a x ) in (7.31), we obtain the equation for the energy boundary ofvibration
a, = P !(nro). (7.33)
According to (7.31), the intersection of curve 2 (7.33 1 with the backbone curve gives the maximum
value of the amplitude characteristic. Adding the branches of the amplitude characteristics for the
linear oscillator 3 to Fig 7.2, we can draw the full amplitude frequency characteristics for the system at
the fundamental (compare to Fig. 3.9).
§7 Hannonic linearisation of vibro-impact systems 99
Hence, this approach gives a simple method to understand the basic qualitative principles of forced
vibration in systems and to find express ions for them that are convenient to calculate. Later, we will
make a detailed study of forced vibrations for the systems considered, using approximate methods and
making comparisons with exact solutions (see §9).
2. The approximation of periodie funetions as a finite sum of Fourier series
ensures minimum mean square deviation [90]. On this basis, harmonie
linearisation methods, where similar approximations to the real solution are
limited to the basie harmonie eomponents, ean be given another interpretation
allowing it to be related to other forms of equivalent linearisation and to obtain
eonvenient formula for the ealculation ofthe !inearisation eoeffieients [124].
Let us substitute the nonlinear funetion <D(x, x) by the linear one
mF + k( x - mx ) + ßx , ehoosing the eoeffieients mF' k, and ß from the eonditions
of minimum of mean square deviation in an infinite time interval. Minimising the
value
1
e(m J• ,k,ß) = lim ~ f{<D[x(t), x(t)] - mF - kx° - ßXO}2 dt ,
7 -4CD T °
where XO = x-mx we find
T
=-lim~f[(x,x)-m.
aBe
mF
T 1-400 F
-kx0-ßxo]dt=O,
°
Be = -lim ~ f[<D(x,x) -m F _kx° -ßxo]xOdt = 0, (7.34)
ak 1-4CD T °
T
ae = -lim~ f[<D(x,x)-m. _kx° -ßxo]xOdt = O.
aß T° 1-4X J
Assuming
(7.35)
we find
lim ~ fxO(t)x°(t)dt = 0, !im ~ fx°(t)dt = 0,

T-->CL T ° T-->x T °
lim ~ fx(t)dt = O.
1'-->00 T°
Using these relationships we find from (7.34)
1
mF = lim ~ f<D(x,x)dt, (7.36)
T -->00 T °
1
k=~!im~ f<D(x,x)xOdt, (7.37)
cr x I-->CD T °
1
ß= ~!im ~ f<l>(x,x)xO dt, (7.38)
cr x 1-->00 T °
where we make use ofthe following notation:
l' 2 l' 2 2
cr; = lim~ f[X°(t)]2dt=~, cr~ = lim~ f[X O(t)]2dt= axO) . (7.39)
1'-->00 T 0 2 1'-->>0 T ° 2
In [124] it is shown that in a solution of the form (7.35) eondition (7.34) is

suffieient to find the minimum funetion e(m F' k, ß) .
Introdueing the mean square value (7.39) for the (7.35) allows the
identifieation of formulae for harmonie and statistical linearisation. To eomplete
this synthesis we introduee distribution function W(u) for the deterministie
proeess x(t) as relative length of time, during the eourse of whieh the eondition
x(t) ~ u [128] is satisfied. Using Heaviside's funetion T](x) we find
W(u) = lim ~ fT][u - x(t)]dt. (7.40)

T -,>YJ T 0
In agreement with (7.40), we obtain for the density function w = dW / dt
w(u) = lim ~ 7f8 [u - x(t)]dt , (7.41 )

7'-'>00 T 0
where 8(x) = dT](x) / dx is the Dirae funetion. By definition

x2
fw(u)du = I, (7.42)
where XI and x 2 are the minimum and maximum values of x(t) .

For the proeess (7.35), for example, we find (Fig. 7.3)
T
1
u = T-(t2 -tl) =-
W()
T
[r (n 1 . u-m x ) +-aresJn--
- ---aresJn--
(j) (j) ax
1
(j)
. u-m x ] =
ax
1 1 . u - mx
=-+-aresJn--
2 n ax
whenee
w(u) = [n~a; -(u - mJ2

Similarly introdueing a joint distribution funetion
r (7.43)
W(u, v) for the proeesses x(t) and x(t) as relative

time durations satisfYing the conditions
x(t) ~ u, x(t) ~ v , i.e.
W(u, v) = lim ~ fT][u - x(t)]T][v - x(t)]dt .(7.44) Flg 7J

T-,>YJ T 0
To satisfY the density funetion w(u, v) we obtain from (7.44)
w(u,v) = 02W(U,V) = lim f 8[u-x(t)]8[v-x(t)]dt, (7.45)

oUOv 1'-'>00 0
whenee
X2 1'2
f fw(u, v)dudv = I, (7.46)
where segment [x p x 2 ], [V 1 ,V 2 ] includes the region of values of funetions

X(t) and i(t).

In particular, taking into account that between the coordinates and velocity in
process (7.35) there exists a deterministic link
V -_ +
_co ~ a x2 _( u _ mx )2 , (7.47)
we obtain using (7.43),
w(u,v)=
1t
Ja; -(u-mJ2
1 o[v±mJa;-(u-mJ 2 ]. (7.48)
The ± sign in the argument of the 0 - function is an abbreviation meaning the

sum of two 0 - functions, one of which takes the positive sign and the other the
negative sign.
Introducing the representation (7.45), equations (7.36) to (7.38) for the
calculation of the linearisation coefficients from the conditions of minimum mean
square deviation, can be transformed to the following form:
mr = lim ~ 7f <l>(X,x)dt =
1'->00 T °
1 l' X2"2
= lim - fdt f f<l>(u, v)O[u - x(t)]o[v - x(t)]dudv =
7 ->00 T
o xI vI
(7.49)
x2 "2 1 T
= f f<l>(u, v)dudv lim - fo[u - x(t)]o[v - x(t)]dt =

r ->00 T
XlVI 0
X2 VI
= f f<l>(U, v)w(u, v)dudv,
l'
k = ~Iim f<l>(x,x)(x - mx)dt =
°
(J1 ->00
X
(7.50)
1 X2"2
= -2 f f(u - mJ<l>(u, v)w(u, v)dudv,
cr x xI PI
ß = J, 7->00
Iim lf<l>(x, x)xo (t)dt =
°
(J -
i:
(7.51 )
1 Xz 1'2
=-2 f fV<l>(U, v)w(u, v)dudv

(J.
x xI PI
similar to the equations for the statisticallinearisation coefficients (see §8). During
the development of equations (7.49) to (7.51) we made use of the following
quality for the 0 - function :
X2
f<l>(x)o(x - a)dx = <l>(a) (XI< a < x 2 ) (7.52)
for any continuous function <l>(x).

The linearisation of the nonlinearities <l>(x,x) using (7.49) to (7.51) on the
distribution w(u, v) of a deterministic process x(t) was called distribution

function linearisation by the authors M Z Kolovsky and A A Pervosvansky, (128,
130]. They showed that this method could be used effectively to study
polyharmonic processes. This is due to the fact that the definition of the
linearisation coefficients is based on the initial moments of distribution and are
only weakly dependent on higher moments of distribution. As a result, distribution
functions can be approximated for a wide class of function which makes it
possible to simplify their application considerably. On this basis, similar simple
methods are proposed to calculate the harmonie linearisation coefficients [124].
3. We will now present the forrnulae for the harmonic linearisation coetftcients for the basic
concept of force characteristics of impact pairs. Methods for calculating the Iinearisation coefficients
for a nonlinear function <!J(x) oftype (6.7), (6.8) by the formula (7.23) to (7.25) are presented in detail
in the monograph [181], hence we will immediately introduce the final equations. For the
characteristics of a single-sided interaction (6.7) we find
/:,-m 1 [/:,-m
- a rx c s.J/',.-m - mx
1- ('/:,- 2]1
1
x
mF = -c2a< - - - n - x+ J . (7.53)
2a x 11 a< ax ax
J2]1 ,
1
1- -
k=C2 -
2
I [ a r. c/:,s
1t
-J /:, +
mxn - - m-
Qx
x
Qx
- - mx
1- (/:,
Qx
(7.54)
where a x ;:: /:, - mx . The latter condition represents the presence ofthe impact interaction.
For the double-sided impact pair (7.8) when a r ;:: /:,+ Im x I we find
- c-2a-
mF - x [
1t
I - (/:,
- --
Qx
my J2 - -+-
I + (/:,
Qx
mx J'] +c,m x +
(7.55)
. /:'-m X - arCSIn-_
- m x ( arCSIn-_ . /:'+m X JJ
0[ .
Qx Qx
A-m . A+m
k = c, - - arcsJn-- +arcsJn--' +X
1t ax Qr
(7.56)
When studying symmetrical processes, ifthe condition x(1 + T /2) = -x(t) is satisfied, the values
m x '" 0, mF '" 0 and forrnula (7.56) is substantially simplified.
The characteristics (6.7) and (6.8) reflect only the effect of elastic interaction during impact. To
calculate the loss of energy in this case arising from the introduction of the coetftcient of restitution R,
the following considerations can be used. Let us choose the Iinearisation coetftcient ß so that the
equivalent dissipation due to the force of viscous friction ßx in (7.22) gives rise to the same energy
loss as there was during impact. The energy loss of mass M for one impact
Ed2 = Mx: (l-R 2 )

2
(M = MI M2
MI +M 2
J. (7.57)
§7 Hannonic Iinearisation of vibro-impact systems 103
The value of the pre-impact velocity x_

is found approximately from the first harmonic of
process x(t). Bearing in mind. for example, the resonance characteristics of the vibro-impact regime of
interest, we can assurne that impact occurs at maximum velocity
(7.58)
Substituting (7.58) into (7.57) we find
Ed2 = Yz Ma;ro 2 (1- R2 ) . (7.59)
We define the work ofthe linear force ßx during harmonic vibration (7.35):
27too 21[00
Ed1 = fßx 2 dl = fßa;ro 2 sin 2 roldl = 1tßroa;. (7.60)
o 0
Equating the energy values (7.59 and (7.60) we find an equation for the linearisation coefficient ß :
ß =Mro(I- R 2 )/21t,., pMro (0:'> r:'> 0.16) . (7.61)
For double-sided collisions, the value ofthe coefficient ß must be double:
ß= MCJ)(I-R 2 )/7t,.,pMCJ) (0:'> r:'> 0.32). (7.62)
Note that in the formulae (7.53) and (7.54) we can define the transfer when C2 ~ co to the
situation for a.rigid stop *. Since a system with such limits satisfies the condition
mx+ax=l::>., (7.63)
as c2 ~ co the value of (I::>. - mx )/ a x ~ IX = I To define the limiting transition we eliminate c2 from
(7.53) and (7.54), dividing the first relationship by the second. Applying the limit when IX ~ 1 and
expanding the indeterminacy, we find
mF/k=ax /2. (7.64)
Using condition (7.13) we can obtain formulae for the linearisation coefficients which depend only on
one unknown parameter a x ofprocess x(t) from equations (7.63), (7.64):
mF =(nz., -1::>.+ax)C1(0), (7.65)
k=_2_(I+ nz., -1::>.). (7.66)

L(O) ax
Hence in the case of rigid stops, (7.53), (7.54) can be replaced by the simple relationships (7.55),
(7.66). As weil as significantly simplifying the formula, the convenience of such a substitution is that it
is possible to eliminate the coefficient C2 , whose actual value is complex to define.
Let us now move to the calculation of the harmonic linearisation coefficients for force
characteristics ofthe type (6.13) to (6.16). Calculation with such functions can conveniently be carried
out using the basic formula (7.48) to (7.50) linearised in terms ofthe distribution function.
From §6, section 3, it follows that when applying force characteristics impact pair containing the
ö - function, when finding the change of impact force with time using transformation
F(t) = <1>[x(t),x(I)], the presence ofthe discontinuous function x(l) at the instant ofimpact formally
causes the value of the coefficient to be doubled during transfer by comparison with the continuous
case because of using the average velocity of the corresponding impact phase. During the
approximation of such force characteristics for regular (linear) functions this information appears lost
in the process of equivalent linearisation and the succeeding calculations for the solution of the
linearised equations. To compensate for this defect when calculating the linearisation coefficients (both
harmonic and statistical) we will double the force functions (6.13) to (6.16).
From (7.49) to (7.51), bearing in mind (6.18), and also taking account ofthe fact that for the force
characteristic under consideration <1>E(U,-V) = <1>E(U,V), <1>[)(u,-v) = <1>[)(u,v) (u '" x, v '" x), we
obtain for the harmonic approximation (7.35) bearing in mind the symmetrical functions w(u,v)
relative to v = °:
*The occurrence ofthis limiting transformation was brought to the author's attention by V K Astashev.
(767)
(7.68)
mx+a x GxÜl
ß = ~2 f fVct>D(U,V)W(u,v)dudv (769)
x mx-ar -axw
Substituting in (7.67) to (7.69) the double value ofthe force characteristics (6.13) to (6.14) and the
density function (7.48), (retaining only the velocity distribution before impact by using the sense of
equations (6.13) to (6.14» we find for single-sided impact pairs, using condition (7.52) and the
terminology (7.39)
mx-a
Naturally, the expressions shown apply when ax ~ L'1 - mx'

The similar calculation carried out for a double-sided impact pair using regulations (6.15) to (6.16)
we find when a x ~I m x I +L'1
mF =
MOl(I+R 2 )[ I
21t
2 2
I 2
"V a r -(li-mx) -"Va< -(L'1+mx )
2] , (7.73)
§7 Hannonic linearisation of vibro-impact systems 105
k = Mm (1;R ) [(IJ.-mx)~a; -(IJ.-m x )2 +(IJ.+mx)~a; -(IJ.+m x )2]

2 2
(7.74)
lta x
ß= 2 Mw(l- R2 ) (
2
2 _ ,2 _
ax u mx·
2)
(7.75)
Mx
It must be mentioned that formula (7.70) to (775) can only be used over a comparatively narrow
area of change in the argument a x / IJ., i.e., the change of input parameters within small limits
corresponds to every possible form of vibro-impact regime. This implies that in the form presented, the
harmonic linearisation method is too crude. Small errors in the parameters of the process x(/) can
significantly influence the value of the instantaneous force when higher harmonics are ignored.
The principal method of improving the accuracy of the values of the harmonie linearisation
coefficients is to by improve the initial information relative to the character of the solution sought. It is
possible for example, to calculate the coefficients using more harmonics [124, 209, 214], or to improve
the accuracy ofthe distribution density, w(u,v) for process x(t) by using limited regions for change in
x(t) , i.e., limiting w(u, v) by the condition w(u, v) = 0 when x> IJ. in the non-symmetrical case and
1x I> IJ. in the symmetrical (see §8, section 4). However, the complexity of the corresponding
calculation procedure is expediently matched with the possibility of direct numerical solution by
successive approximations of the type (6.91), (6.92). It is often possible to refine the coefficients of
linearisation arising from the specific solution of the problem. Thus in proper cases we will continue
with this problem.
4. Let us now look at another modification to method of harmonic
linearisation of vibro-impact systems, which is distinguished by the fact that the
linearisation coefficients are not found from the parameters of the process x(t),
but as parameters of the force F(t) = <l>[x(t),x(t)], which is developed at the
output of the impacting elements (Fig. 6.5) [37]. In the following, to distinguish
these two linearisation methods, we shall refer to them as motion linearisation and
force linearisation.
It is convenient to use force linearisation when the force elastic characteristics
of impact pairs are described by the relationship F = <l>(x) , shown in Fig. 6.2.
Direct linearisation of these functions is not possible since their values are
ambiguous and unlimited during impact, hence we will linearise them by reverse
function x = X (F), which characterises the relationship of relative motion of the
elements of the impact pair to the impact forces occurring between them. For the
relationships shown in Fig. 6.2, b - d, the function x = X(F) can be represented
by the fOllOwing equations (Fig. 7.4):
+ß when F> 0,
x ={ -ß when F< 0, (7.76)
1
x = -[F - (F - Cß)l1(F - Cß)] , (7.77)
C
1
x =-ß+-[(F + cMT](F + cß)-(F -cß)T](F -cß)]. (7.78)
C
In the case ofthe characteristic <l>(x), shown in Fig. 7.2, a, the reverse
fI 11
.11----
o r f
u)
b) c)
Fig.7.4
relationship does not exist, hence we will not use this approach in this case.
To linearise the function x = X(F), we will assume
x",;mx+q(F-m j . ) (q=l/k) (7.79)
The linearisation coefficients m x and q are found from the condition for the
minimum mean square error (see section 3) given by the formula
F,
m x = fX(F)w(F)dF, (7.80)
F,
(7.81)
Here F p F2 are minimum and maximum values of process F(t), cr7, is the mean
square value of the centred process FO (t) = F(t) - mF; w( F) is the density
function of F(t).
U sing the low sensitivity to the input process for nonlinearities of types (7.76)
to (7.78), on the type of output action [214], when they are linearise, a fairly crude
assumption can be made for the relative form of F(t). Thus, limiting our choice
to the simplest, we assume
F(t)",; mjo + a F coscot. (7.82)
Bearing in mind that the density function of F(t) in this case takes the form
w(F) = [nJa; -(F-m j )2r', (7.83)
cr;- = ai, /2, and from (7.80), (7.81) we find the following values of the
coefficients ofharmonic linearisation.
For the characteristic (7.76) (Fig. 7.4, a) when aF 2:1 mF 1
m = 2D. arcsin mF (7.84)
x 1t QF '
(7.85)
For the characteristic (7.77) (Fig. 7.4, b) when ajo 2: cD. - mF
mx =aF cD.-
- -
c
j
+-
2aI'
I [CD.
mF- -
1t
- mF . cD. - mF
---arcstn---+
aF aF 1-[ "":,m,J' II (7.86)
1 -+-
q=- 11
1 [ arcsm---+---
. cD. - mF cD. - mF (7.87)
c 2 1t aF aF
For characteristic (7.78) (Fig. 7.4, c) when aF I

~ cD.+ mF I
m, = ~[
1tC
I _ (CD. + m F)2
aF
_ cD.-mF
1- ( - - -
aF
)2 +
(7.88)
I [ . CD.-mF . cD.+mF
q = - arcsm---+arcsm--.-+
1tC aF a,.
(7.89)
To study the latter case for double-sided impact pairs with e1astic coupling the linearisation
coefficients can be found in another way using the following argument. Let us represent the static
elastic characteristics of this impact pair F = <t>(x) (Fig. 6.2, c) as the sum of two characteristics, one
linear (Fig. 7.5, a) and one nonlinear (Fig. 7.5, b). For this we assume conditionally that when the
maximum force developed by the linear spring is constant and equal to cD. during the occurrence of a
vibro-impact regime, then the maximum force in nonlinear "elasticity" will be aF - cD. . Linearising
these characteristics in relation to this force with the help of the examples described above and using
(7.84), (7.85) we obtain when aF ~ CD.+ ImF I
2D. . mF
mx =-arcsm--- (7.90)
1t aF -cD.
q- [ c+
-
_--r=~1t(~a,=.. -=c~D.
4D.~I-[mF/(a,;-cD.)1
)==]-1 (7.91 )
Comparing the numerical value of the Iinearisation coefficients in the forms (7.90), (7.91) to the
corresponding values (7.88) , (7.89) shows close agreement. This detail can be used by simplii)'ing
formula (7.87) for the case when m" = 0, which will be studied frequently later. The simplification
shown is using the following analogy.
According to (7.87, (7.89) the expression for q when mF '" 0 takes the form
2 . cD. cD.
[
ql=- arcsm-+- 1- -
1tc aF a,.
cD.
a" ~)2l (7.92)
m)
for a double-sided impact pair and
I I I . cD. cD. cD.
1 [ 2ll
q2 =- -+- arcsm-+- 1- - (7.93)
c 2 1t aF aF aF
for a single-sided impact pair. Note that the relationship in the square brackets in (7.92), (7.93) differ
only because for the first case the value of D. from the physical sense is essentially positive, but in the
second case, changes sign to negative when gap changes to interference. At the same time, even in this
case only a positive value of D. can be used, chan ging the sign before the radical when gap changes to
interference. Taking this into account, we can eliminate the general expression in square brackets from
(7.92), (7.93). As a result we find
(7.94)
Substituting in (7.94) for ql' the equivalent value (7 .91), taken when mF ,,0 we eventually find
q2 =~(l±-
2c
___l___-J
1-1t/4+1tQ{./(4ci'l)
(7.95)
Here the positive sign indicates a gap, and the minus sign interference, the quantity '" is equal to the
absolute value of the gap or interference.
F When it is necessary to calculate energy
dissipation during impact when using
harmonie Iinearisation by force the equivalent
-,i eoefficients of viseous friction ß ean be taken
in the form (761), (762)
In conclusion, note that the eoeffieient of
harmonie Iinearisation ean be ealculated in
principle for even more eomplieated contact
a) b)
interactions, for example, in the form of
Fig.7.5 eharaeteristics (6.5), (6.6) although in this ease
integrals are developed whieh it is not possible
to evaluate through elementary funetions. This diffieulty ean be overeome by simplifying the
approximate express ions for ealeulating the Iinearisation eoeffieients [124] or using speeial graphical
methods to eonstruet them [211]. However, taking aceount of the insensitivity of periodie solutions to
the speeifie eharaeter terms of eontaet interactions as defined in §7, such eomplieated proeedures for
finding the first approximation are hardly justified. Characteristies oftype (6.5), (6.6) remain useful in
schemes for successive approximation (6.91) when, for example, the eontaet forees are to be further
refined, however, even in this case it is eonvenient to eonstruet the initial approximation from whatever
simplification possible, resulting, say, from harmonie Iinearisation of the equations with eharacteristies
of the type (6.7), (6.8). After finding the highest approximation, the maximum eontaet forces are
defined using (7.5), (7.6) from one ofthe formula
max 1<!lex) 1= <!l(x max ) or max 1<!lex) 1= <!l(Xmin) (7.96)
§8. Statisticallinearisation of vibro·impact systems

I. The description of the equations of motion for vibro-impact systems in
§6 gives a convenient basis for studying the statistical dynamics of such systems.
Problems in statistical dynamics for vibro-impact systems occur when studying
the vibration of machines and structures with regard to colliding elements of
kinematic pairs or couplings, when designing systems of vibration isolation,
during the study of interference effects in systems for automatic control, or in
similar devices consisting of mechanical elements with clearance, etc. In the
majority ofthe cases mentioned, random vibration is caused by "external" sources
of wide band excitation: jet engines, turbulent media in which components move,
irregularity in the profile of the path of transport devices, noise in electronic
instruments in control circuits, etc. However, even in the case of periodic external
excitation the presence in the structure of a large number of couplings, creates
"internaI" sources of random excitation. These irregular impact interactions,
caused by additional mobility in gaps, leading to dynamic instability of the
structure, or caused by specific impact regimes in conditions of poor
synchronisation, are characterised by the appearance of irregular clatter *.
• The study of vibration accompanying repeated attenuated eollisions, was carried out in the work of R
F Nagaev [165. 167] and MI Feigin [84, 871
§8 Statisticallinearisation of vibro-impact systems 109
The methods of statistical dynamics allow us to study a combination of various

real processes which arise under identical conditions, and to combine them by
common probability descriptions. The basis for this description lies in the
introduction of random funclions, hence we need to deal with some background
knowledge from the theory ofrandom functions, before further investigation [184,
204 ].**
The random process or random function x(t) with deterministic argument I
(time) is a function lying in the range -00 < I < 00, which is defined to have a
random variable X(tl) at each fixed instant of time 11 , The full probability
characteristic of random variable X(tl) is known as the distribution funclion
WJu"II) ' which characterises the probability that the variable X(tl) of the
random variable X(II) satisfies the condition X(tl) < u l . Hence
WI(u"/l) = 9'{X(t l ):-::; u l }, (8.1)
where f//' is the probability occurrence (0:-::; p/, :-::; 1). For any argument t l , the
value W(u\,1 I) characterises the probability ofthe condition ofprocess x(t]) in
each fixed instant of time and is known as its one-dimensional distribution
funclion. In order to establish the connection between the state of process x(t) m
consecutive instants of time, the multi-dimensional distribution function is
introduced as
Wn(u\, I" ... , un, t n) = fJl'{X(tI) :-: ; u" ... ,X(tn) :-: ; uJ . (8.2)
Alongside the distribution function for describing random processes, the n-
dimensional density funclion of a random process is given by
8n W
wn(u"tp ... ,un,tn )= n (8.3)
8u l .. ·8u z
When analysing random processes an important quantity are their moments,
which are functions defined generally as follows
'" '"
m'l'" (t p ... , t n ) = f .. · fU;1 ... u;; wl/ (u p t p ... , un' tn )du l ... dUI/ . (8.4)
Function (8.4) is called the n-dimensional moment of the q-th order

(q = i] + ... + in)' Moment characteristics are symmetrical relative to all their
arguments. Apart from moments of form (8.4), called initial, we also introduce
central moments defined by
00 00
J.l'I'n(I" ... ,tI/)= f .. · f[UI-ml(tI)P ... [Un-ml/(tn)]'n x

(8.5)
X w(u l , t l , ... ul/' tn)du l ·· . dUn
where ml (tl)' ... , ml (t n ) are expected values characterising the statistical average
value.
"Translator's Note. For texts in English covering sirnilar material, the reader can consult A B Clarke,
R L Disney "Probability and randorn processes for engineers," (Wiley) or M O'Flynn "Probabilities,
randorn variables and randorn processes," (Harper & Row).
A typical feature of moments is that as their order increases, the information in

them about the random processes principally reflects the part played by the
occasionallarge amplitude peaks. Thus the practical study of random processes, is
limited, as a rule, to several fundamental moments. The most important is the first
order moment
ce
m$t) = mx(t) = juw\(u,t)du, (8.6)
known as the expected, mean or average value of random process x(t), and the
central second order moment
00
1l11(tpt2)=Rx(tpt2)= j[u 1 -m (t$][u 2 -mx(t2)]w(uptpu2,tJduldu2

X (8.7)
known as the covariance ofthe process. Substituting t\ = t 2 = t in (8.7) we obtain

the mean square value of the random process as
x
cr;(t) = R«t,t) = j[u-m«t)]2 w\(u,t)du. (8.8)
The function cr x (t) is called the standard deviation of random process x( t) .

Using moments we can reduce the density function ofrandom processes.
In applications, the normal or Gaussian distribution plays an important role.
ClarifYing this, the sum of the largest quantity of statistically independent random
processes is always cIose to the normal for the process, provided none of the
components dominates the others. The probability characteristics of anormal
process are completely defined by its expected value and covariance. For example,
the one-dimensional density function takes the form
w\ (u,t) = r;;-I { [u-m x(t)]2}

exp - 2 • (8.9)
v21tcr x (t) 2cr x (t)
Weshall study only stationary random processes. A randorn process is called
stationary in the narrow sense when its distribution function does not change
when the time origin changes
WI/(u\,t\ +t, ... ,ul/,tl/ +t) == WI/(U1,tl, ... ,ul/,tl/) (8.10)
for all n and t . As a result for such processes
WI =WI(U I ); W2 =W2(U p u2,t), (8.11)
Le. a one-dimensional distribution function is independent of time, but a two-
dimensional depends only on the difference t = l2 - ll' Sirnilar features are
possessed by the density function of stationary processes. As a result, from (8.6) to
(8.8) we find for the stationary case
mx(t) = m x = const, R x(tl,t 2 ) = Rx(t2 -tl) = Rx(t),
(8.12)
cr; (t) = cr; = Rx (0) = const.
A random process satisfYing conditions (8.12) belongs to the cIass of process
known as stationary in the wide sense or a strictly stationary randorn process. Of
course, conditions (8.12) is, in general, insufficient to satisfY the identity (8.10),
however, for anormal process the rneanings of stationary in wide and narrow
senses eoineide.
An important feature of stationary proeesses is the possibility of representing
them speetralJy. Applying a Fourier transformation to the eovarianee, we find
00
Sx(ro) = JRJr)exp(- jroT)dt. (8.13)
Funetion Sx(ro) is ealJed the spectral density ofthe stationary random proeess.
The integral of this term
defines a value proportional to the mean power of the proeess in the frequeney
range [ro"roJ. From the expression for the inverse Fourier transform we find
from (8.13)
(8.14)
Relationships (8.13) and (8.14) are generalJy known as the Wiener-Khinehine

relations.
Assuming t = 0 in (8.14), we find
1
0-;
00
R.(O) = =- JS.(ro)dro (8.15)

27t -00
As both are even funetions of their arguments, the eonvolution and speetral
density are eonneeted to those eharaeteristie eonditions so that the larger the time
correlation of random proeess to ensuring the eondition I R x (T) I~ 0 when t > t o '
then the narrower the speetrum oeeurring in the frequeney proeess S.( ro) .
A stationary proeess, whose speetral density is eonstant throughout the
frequeney range from -00 to + 00 is ealJed white noise. Aeeording to (8.15) white
noise has an infinitely large mean square value, henee it ean be eonsidered only as
the idealisation of an aetual wideband random proeess. Taking
S x (ro) = So = const in (8.14) we obtain for white noise
S 00
Rx(T) = _0 Jexp(jroT)dro = SoÖ(T), (8.16)

27t -00
where Ö(T) is the Dirae funetion.

A stationary random proeess proeeeds uniformly with time, Le. its statistical
eharaeteristies do not depend on the chosen initial instant of observation. Hence,
one aetual waveform of a stationary proeess eontinuing for a suffieiently long time
ean include alJ data and eonditions of the random proeess. Stationary random
processes, alJ forms of whieh can be represented in this way, are called ergodie*.
A sufficient condition for a stationary normal proeess to be ergodie takes the form
[204]
* Strictly speaking, in an ergodie process the existence of separate unrepresentative waveforms is also
accepted (producing a set ofzero dimensions).
(8.17)
For ergodic processes

W1 (u 1 ) = lim _1_ 7f l'][u 1 - x(t)]dt (8.18)
2T -T
7 ·->00
where l'](u) is the unit step function; x(t) is the waveform ofthe ergodic process,
mx = lim-I- fX{t)dt, (8.19)

2T -T
T-->w
RJr) = \im _1_ 7fx (t)X(t + l)dt . (8.20)

T-->oc 2T -T
These relationships provide the basis for calculating the statistical characteristics
of stationary ergodic processes by experimentally recording their waveform.
To describe the statistical connection between two stationary processes
x(t) and y(t) we introduce cross-correlationfunctions as
R xy (tp t 2 ) = ([x(t]) - m x J[y(t 2 ) - m y ]),
(8.21)
Ryx{t], t 2 ) = ([y(t]) - m y ][X(t2) - m x ]).
Both here and later, square brackets mean the operation of statistical averaging of
an ensemble of actual waveforms. If function (8.21) depends only on the
difference 1 - t 2 - t 1 , then processes x(t) and y(t) are said to be stationary coupled.
For such processes
(8.22)
Using the cross-correlation function the density function for a normally
distributed random vector [x, y] is defined as
a
2 .
w(u, v) = --Ei!'{X ::;; u, Y ::;; v} =
I
x
auBv 2nJa~a~ - R;y
In the absence of a correlation between processes x(t) and y(t) the value Rxy == 0
and expression (8.23) is substantially simplified
_ I [_(u-m x )2 (v-m y )2]
w( u, v ) - exp 2 2' (8.24)
2na x a y 2a x 2a y
Comparing (8.24) and (8.9) we can see that the joint density function in this case
is equal to the product of the density functions produced for these processes. Of
course this fact follows directly from the nature of the statistical independence of
the occurrences.
Random process x(t) is considered to be differentiable at point t, if for any
number E convergent to zero, there is a limit for a random sequence ofvalues*

[x(t + E) - x(t)]1E .
A random value appearing as the limit of a random sequence is called the
derivative 01 the random process (function) x(t) at point t and will be designated
as x(t). A sufficient condition for differentiability of a stationary random function
is when the second derivative ofthe correlation function exists at 't = 0 [204]:
d 2 RJr)/ de 2 It~O .
The knowledge of the joint probability density and its derivative w(u, v)
allows the calculation of the average value I of the events when the process x(t)
crosses level ~ in unit time. This value is defined by the formula [204]
0()
I = fvw(~, v)dv. (8.25)

o
The inverse value t o = IIj gives the average time between crossings. It can be
shown that for a differentiable stationary process the ordinate of the random
function and its derivative at the same instant are noncorrelated, and for anormal
process and independent random variables. From the normality of process x(t),
comes normality x(t) [184]. For this reason, during calculation from (8.25) the
normallaw is used in form (8.24) where y(t)= x(t).
Under the influence of a stationary differentiable random process x(t) on a
linear system with constant parameters described by the operator L(s), the
mathematical expectation and the spectral density reaction of the system z(t) are
expressed through similar characteristics of excitation in the following form [184]:
mz = L(O)mx' (8.26)
Sz(m) =1 LUm) 12 S,(m). (8.27)
During this, within the limits of the actual frequency range, function LOm) must
not take an infinite value. It is important to note that the specified transformation
does not violate the normality ofthe process.
2. We will now turn to the application of ideas and methods of statistical
linearisation to vibro-impact systems [37, 38]. Suppose that for the system
described by equation (6.39), a stationary random normal excitation P;(t) acts.
Taking into account that during linear stationary transformation the random
process remains stationary and normal, we can write the resulting equations in the
form
x(t) + L(s)[x(t),sx(t)] = S(t), (8.28)
where 'Ö(t) is the given stationary normal random process.
Following the statistical linearisation method [113], we will substitute the
non linear function (x,x) by the linear one
(x,x)=m F +k(x-mx)+ßx, (8.29)
selecting the linearisation coefficients mF' k and ß so as to minimise the
• The limit of a sequence of random variables is a random value, the root-mean-square variation of
which, from the elements ofthe sequence, tends to zero when E-t O.
statistical mean square value of the residual
(e(mF,k,ß»)= f f[cI>(u,v)-m F -k(u-mJ-ßv]2 w (u,v)dudv. (8.30)
Here w(u, v) is the joint density function of processes x(t) and y(t). From the
necessary conditions of extrem um existence
o(e)/om F = o(e)/ ok = o(e)/ aß = 0 we obtain equations to determine the
coefficients of statistical linearisation
mF = f fcI>(u, v)w(u, v)dudv, (8.31 )
k= -2
1
cr x
00
f f(u - m )cI>(u, v)w(u, v)dudv ,

._XJ-x;
'"
x (8.32)
I x X'
ß = -2 f fvcI>(u, v)w(u, v)dudv. (8.33)

Cf Y -Cf.-' -Xl
Carrying out the substitution (8.29) in (8.28) and assuming that

11(1)= ms + 11° (I) where ms = const, 11°(1) is the centred component of the
process: 11 0 (t) = l1(t)-ms,
x(t)+L(s){m F +k[x(t)-m<]+ßsx(I)}=m s +11°(1). (8.34)
Searching for the solution of(8.34) in the form
X(I) = mx +x°(t), (8.35)
where mx = const, XO (t) = x(t) - mx , we find from (8.34) for constant and centred
components
mx + L(O)mj- = ms , (8.36)
x ° (I) + L( S ) [ kx ° (I) + ßsx ° (I)] = 11 0 ( I) . (8.37)
From (8.37) we find
(8.38)
Making use of the formula for the linear transformation of the spectral density
(8.27) we obtain from (8.38):
S w _ Ss(w)
(8.39)
x( ) - 1I + L(jw)(k + ßjw) 12
The mean square value of x(/) can be calculated from (8.15) by using equation
(8.39) as
2 1 1 Ss(w)dw
(8.40)
cr x = 2rc -00 1I + L(jw)(k + ßjw) 12
and noting that according (8.27) S/ül )=ül 2S/ül), the mean square value of velocity
ofprocess x(/) as:
*8 Statisticallinearisation of vibro-impact systems 115
2 1 ."J Ss ( ro )ro 2 dro

er - - (8.41)
y - 2rc 11 + L(jro )(k + ßJro) 12
-00
For integrals of type (8.40), (8.41)

I =_1 g ,,(jro)droI
" 2rc 1 h,,(jro) 12 '
-00
h
were h"s( ) =aos "+a\s
,,-\ . ()
+ ... +a",gll s =cos
2,,-2
+c\s 2,,-4 + ... +c lI _\'
having the general formula, presented, for example in [184].
In what folIows, we shall be interested in general in the specified statistical
characteristics of process x(t), the calculation of which will consist of the nature of
speclral Iheory for stationary random processes. In addition, note that the values
of m x ' er x' er, in equations (8.36), (8.40). (8.41) have been shown to depend on
the linearisation coefficients, which , in their turn, depend on the unknown
statistical characteristics of process x(/).
According to (8.31) to (8.33), in order to find the final relationship it is
necessary, as in the harmonic linearisation method, to express some initial
considerations relative to the character of the process examined, particularly its
density function w(u, v) . In order to achieve that, we shall return to the structural
scheme of the system examined (Fig. 6.5), taking into account the effects the
random excitation S(t) has on it.
During the transformation ofa random process in nonlinear elements <l>(x,sx)
the normality of the process is violated. That is why the input signal F(t) of the
linear filter L(s) can be substantially different from the normal. We shall examine
the transformation of F(t) by a linear filter [177]. Using (6.43) we find, chan ging
the variable 1-1 = l' and extending the upper limit to infinity, bearing in mind
that h(t -1) == 0 when 1> I ,
x
X(t) = Jh(1')F(t-1')d1'. (8.42)
Designating a variable further as 1, we will divide the interval of its variation

as (0,1\), (1\,1 2 ), .•• , (1",1,,+\), ... , transforming the integral into an infinite
sum, using the averaging theorem
X(t)
00
= L,hmF(t -
m=O
<,) = L,Sm .
00
m=O
(8.43)
Here
fh(1)d1,
tm+1
hm = Sm=hmF(t-<,), 1m~<,:Slm+\'
We will choose the sub-intervals in such a way that the random value
F(t - 1:,) and consequently Sm are statistically independent. Obviously, the more
strongly process F(t) IS correlated then the Ion ger the sub interval
Tm = 1 m+\ - 1 m' from wh ich one can say that R F (t) "" 0 when I> Tm' Here
RI. (t) is the correlation function of process F(t) .

As we already pointed out in section 1, the distribution sum of large
statistically independent random processes, and consequently their values, tends to
normality, if the summed components are statistically non dominant. In the case
considered, this condition is satisfied if function her:) changes sufficiently slowly,
i.e. if its value does not change significantly over a 1arge number of sub-intervals.
The maximum rate of damping hCr) according to (6.50) is defined the magnitude
of the greatest value of the real part rkQk of root Sk (see 6.55) of the
characteristic equation ofthe system (6.54). Denoting the inverse value of rkQk ,
as
(8.44)
we can define the condition for the normality of a random process when passing
through a linear filter in the following manner:
Tm / Ta « 1 , (8.45)
i.e. the correlation time for an external excitation Tm must be considerably less
than the characteristic time of damping for transient components of a linear
system. Since the time correlation defines the width of the frequency band of a
random process, and the value Ta according to (6.61) the pass band of the filter,
condition (8A5) can be thought of in the following physical terms: to normalise a
random process transformed by a linear filter, the pass band of the frequency
characteristics ofthe latter IL(jeo)1 must be considerably narrower than the range of
signuificance of the frequency of the spectral density SI'(eo) of the external
excitation F(!). Above all it is important that Tm « T, where T = 2rc/Qk . In the
strongest sense, by normalising a process, it is understood that the signal at the
output from a linear filter is closer to normality than at input.
To satisfy the requirements described above it is possible to assurne
approximately that the incoming non linear element <l>(x,sx) (Fig. 6.5) of a
random excitation x(t), existing as the solution of equation (8.28), has anormal
distribution, which allows us to use coefficients (8.31) to (8.33) with the density
function (8.24) during calculation, taking x(t) = y(t).
As a result the specified coefficients are dependent on m x , (J x' (J y and the
system of finite equations (8.36), (8,AO), (8A1) allows us finally to define these
values.
In such cases, when the non linear characteristics take the form <l>(x), i.e. they
reflect only the effect of elastic interaction, ß == 0 and the formula for calculating
the linearisation coefficients is simplified considerably:
cn
mr = f<l>(u)w(u)du, (8.46)
1 ., "
k =-2 J(u - mJ<l>(u)w(u)du . (8.47)
cr x -x;
Here w(u) is the single-dimensional density function of process x(t), given by

expression (8.9) when m x = const, (J x = const .
It is interesting to note that, as shown in [177] for anormal process x(t),
approximating the nonlinear function <l>(x) by the linear relationship
m F +k(x-m x ) is the best possible in terms ofthe minimum root-mean-square
divergence of all possible linear approximations. It is also shown that calculation
using the method described of the linearised value of the correlation function of
process x(t) and consequently the standard deviation, always show less than their
exact values. There exists the possibility of statistical linearisation leading to the
known overestimation ofthe va lues [113], however calculation ofthe linearisation
coefficients using this method is considerably more difficult.
The practical calculation frequently uses values, existing as the half-sum ofthe
coefficients, obtained from both methods. We will limit the first method to the
calculation of linearisation coefficients, giving an adequate idea of the basic
qualitative features of statistical dynamics of vibro-impact systems.
3. Let us turn to the coefficients for the statistical linearisation for the basic force
characteristics of an impact pair. For nonlinearities of type (6.7), (6.8) the calculation according to
(8.46) to (8.47) and using (8.9) is performed in monograph [113], hence we arrive directly at the final
expressions. For the characteristics of single-sided interaction (7.7) we find
mF =-C20 x { Il- m" [~+;;r( Il- m x )] _ _

I exp[
J2n
(Il- ~j ]}, (8.48)
::x )l
0x 2 0r 20 x
k =C2 [ ++ 9-( Il (8.49)
where .';(u) = .)z; jexp[ - ~ Ja"'. is the probability integral, the value of which is tabulated in
[184]. For double-sided impact pairs (6.8)
mF =c {m -(ll+m
2 X x );';( Il::xJ+(Il-m x );';( Il::xJ+
(8.50)
+ 1z;[exp[ - (Il~:;j )-exp[ (Il;:;j )])'

(8.51)
When it is necessary to take energy dissipation du ring collisions into ac count in cases when using
the statistical characteristic (x) it is possible to make use of reasoning, similar to that which was
used during harmonie Iinearisation.
Let us define the mean power of the energy dissipated during impact as N Oue to the suggested
density function of process x(t) (8.24), the standard deviation of its velocity does not depend on the
state u and is equal to CY y . As a result, the standard deviation of the collision velocity, when u = Il ,
is also equal to CY; As a result. using (7.57), we find during impact onto one stop
1-R 2
N d2 =--Mf(J~· (8.52)
2
Here J is the average number of collisions in unit time against one stop.
The average power dispersed by the force of viscous friction ßx under normal distribution of
process x( I) = y(t) equals
X>
N dl = fßV2W(v)dv=ßO"~ (8.53)
Equating the values (8.52) and (8.53) we find
ß=
l-R 2 Mf· (8.54)
2
The value off is found from (8.25) as the average number of interseetions when y > 0 of process
x(l) at level u = A. Substituting (8.24) in (8.25) we find
f=~expl (A-rn j ].
27lO"x l 20";
(8.55)
Substituting ( 8.55) into (8.54) we finally obtain

(1- R2 )MO"Y [
ß= exp (8.56)
47lO"x
Naturally, for regimes with double-sided impacts the value ofJand ß must be doubled.
From time to time to simplity the ca1culation of the value off, the problem can be simplified as
before on the basis of a spectral analysis of the extemal excitation and natural characteristics of the
system. For example, in (8.54) we can put
f = Wo /(27l) , (8.57)
where Wo is the dominant frequency in the spectrum of the process.
4. When calculating the coefficients of the statistically linearised force characteristics of the
impact pair <l>(x,x) of type (6.13) to (6.16) from equation (8.31) to (8.33) the accuracy of ca1culation
can be increased by refining the law of distribution w(u, v) making use of additional information
reflecting the specific nonlinearity examined. Ouring ca1culation for systems with rigid stops it is
convenient to take into account the limiting region of measurement of X(I) • i.e. to apply to w(u, v)
the condition
w(u,v)"O when u>A (8.58)
for single-sided impacts pairs and
w(u,v) " 0 when 1u I> A (8.59)
for double-sided. Frequently, when disturbing systems with normal random processes it is expedient
to choose as the value of w(u,v) the truncated normallaw [38] (see also §13, section 3)
(u-rnx)2 v2 ]
w(u,v) = C1exp [ - .2 - [l-ll(u-A)] (8.60)
20" r 20";
for the single-sided case and
(u_rn)2
w(u,v)=C 2 exp [ - 2x (8.61 )
20"x
for the double-sided. Here ll(u) is the unit step function; CI, C2 are constants defined from the
normalisation conditions
ß CL
f fW(u, v)dudv = I (8.62)
for the single-sided case and

ß x
f fw(u,V)dudv=1 (8.63)
--~-x
§8 Statisticallincarisation of vibro-impact systems 119
Eventually we find
(8.64)
Substituting (8.61) into (8.63) and following a similar calculation, we find for the double-sided impact
pairs
(865)
Noting that for the characteristics (6.13) to (6.16) CPE(u,-v) =CP(u,v), CPD(U,-V) = CPD(u,v) and
also, according to (8.60), (8.61) w(u,-v) = w(u,v) from the formulae (8.31) to (8.33) we find for the
single-sided impact pair, taking (6.18) into account
6 00
mF = f fCPj,.(U,v)W(u,v)dudu, (8.66)
6 x
k=~ f f(u-mx)CPt;(u,V)W(u,v)dudv, (8.67)
(Jx -00-00
6 x
ß=~ f fVCPE(U,V)W(u,v)dudV (8.68)
O'y -oo-'Xl
We shall substitute into (8.66) to (8.68) the double value of force characteristics (6.13) to (6.14) (see
the note on p103) and also the formula for the law of distribution (8.60). During this, note that, since
the value of the velocity before impact appears in equations (6.13) to (6.14), during calculation of the
coefficients of linearised density distribution then the velocity in (8.60) and (8.61) must be halved.
Using (8.64) we find
mF = f1 Mv 2(l + R 2)o+(u - ,1) exp[ (u _~x)2 4][1-11(U - Mldudu =

-00-00 ~
81ler x erY[ + :7( ,1: ~x J] 2er x 2erY
(8.69)
(~- m x)expl- _(~_-_m-2-r)~l

l
2 2
M0+R) 2~ ~ (870)
= 4..& 1+.rYrI--
J ~ - mr ) er~ ,
\ er,
[ (~-mx)2l
_ M(I-R 2 ) exp --~- er y
(8.71 )
- 2n ~+:;<_(~-mx) er,'
2 er r
For the double-sided impact pair equations (8.3 I) to (8.33) take the following form:
ß x·
m" = f flP,Ju,V)W(u,v)dudv, (8.72)

-ß-OO
'" w
k = ~ f f(u - mx )lPj,(u, v)w(u, v)dudv, (8.73)
a x -d-'X'
'" CD
ß= f fVIPD(U,V)W(u,V)dudv, (8.74)
-ß-X!
where IPE and IP J) are defined by doubling equations (6.15) to (6.16). Substituting for them in (8.72) to
(8.74) and carrying out the caIculation similar to that above, we find using (8.61),(8.65)
2 exp - [ (~-mx)2l -
~~2-- exp
[(~+mr)2]
- - - 2- - ?
M(l+R) 2er x 2er x er;
(8.75)
mF = 4..& (~-mr). (~+mx)
0' - - + . 0 ' - - er;
ax O"x
§8 Statisticallineafisation of vibro-impact systems 121
-------------------------
2 (L1-m x )exp[- (L1-~x)2J+(L1+mx)exp[- (L1+~x)2J 2

M(I+R) 2a x 2a x ay
(8.76)
k= 4.,{i; . (L1-
:Y ----- xJ or(
L1 +m
m +::;. ----- xJ a~ ,
ax (J' x
2 exp[- (L1-~jJ+exp[- (L1+~x)2J

M (1- R ) 2a x 2a x ay
ß=~--~--~~~-+--~--~~ (8.77)
2rr ;;r( L1:~x J+Y7( L1:~x J ax
5. The coefficients of linearisation obtained above depend on the statistical characteristics

mx '0;,0;. for process x(t) appears in the input signal for the nonlinearity examined describing the
impact. Making use of the terminology described earlier (see § 7) the described method to find the
coefficients we will call slalislical motion linearisalion. By analogy with harmonic linearisation we
can construct a method of statistical force linearisation, during which the coefficients of linearisation
are expressed in terms of the statistical characteristics of process F(t), acting as the output of the
impacting elements. As for the harmonic case, such a method allows the use of elastic statistical
characteristics of the impact pair F = <l> (x), given by the form shown in Fig. 7.2, c-d, leading to a
simple calculation relationship.
We shall replace the approximate inverse function x = X(F) (7.76) to (7.78) (Fig. 7.4) by the
linear relationship
x,,"mx+q(F-m,.) (q=I1k) (878)
and defining the linearisation coefficients mx and q by the formulae
mx = [X(F)w(F)dF , (8.79)
q=~ [(F-mF)X(F)W(F)dF (8.80)

aF 00
minimising the mean square value of the error. Here mF ,a7; and w( F) are mathematical expectation,
mean square value and one-dimensional density function ofprocess F(I).
The relationship for w(F) is chosen as the result of the following considerations. The spectral
characteristics ofprocess F(t) during transformation F(t) = <l>[x(t)] are defined corresponding to the
characteristics of the input process x(t) and additional distortions arising from the nonlinearity of the
transformation. The use of inertia-free Iinearisation relationships during statistical linearisation does
not allow the us to catch these distortions. Thus in the approximate solution when applying
normalisation to process x(t) the distribution function of process F(t) also shows normality. In other
words, the approximate solution reflects only the characteristics of the main harmonic components in
the transmission band of the linearised system. However, as mentioned in §7 the weak sensitivity of
the inverse transformation x(t) = X[F(t)] of types (7.76) to (7.78) to the particular form F(t), the
presence in this process of additional components by comparison with the basic components cannot
substantially influence the character of the transformed main harmonics. Hence, even in the case of
statistical Iinearisation it is possible to use quite crude approximations for process F(t). From this
discussion, we can take as an approximation for the normallaw of distribution of F(t):
W(F)=_I_ exp[- (F-~';}J. (8.81)

.,{i;aF 2aF
As a result, for functions oftype (7.76), (7.78) (Fig. 7.4 a, b), we can make use ofthe coefficients
ofstatisticallinearisation calculated from the formulae similar to (8.79) to (8.81) in monograph [ll3],
because they are treated there as typical nonlinearities of automatic control systems: ideal relay
characteristics and linear characteristics with saturation. We will quote the resulting formula using our
designations here.
For function (7.76), we have
(8.82)
il
~11: crF (mJ,l
q= --exp --2-
2crF J
, (8.83)
(8.84)
(8.85)
For function (7.77) from formulae (8.79), (8.80) we calculate directly using (8.81)
mx = l!.[F-(F-Cil)l](F-il)] ~ exp[_(F-~f-}]dF=
-00 C V211: crF 2crF
= __ I_CJ
J2;cr FC -00
Fex p[
2cr7:
il_fexp[_ (F-m F)2]dF
(F-mF)2}r+ __
J2;cr F cll 2cr7: .
(8.86)
The first integral is obtained with the help ofthe substitution (F -mj.")/crF = A. As a result we
find, using the abbreviation (cil- mF ) / cr F '" v ,
r:::-I cll
fFexp
[(F-m r )2}
2· r=
1
r:::- l' ( A2
f(mF+crFA)exp--dA=
J
V21tcrFC_00 2crF v211:c_ oo 2
=-mF-
J2;c -00
f (2J
v
A cr F
f- (2J
A
exp - - dA - - - v Aexp - - dA
2 J2;c -XJ 2
=
_mF[1
- - - - - r:::-
=(Cil-mFJ]
- -+.,. crF exp [(Cil-m
- F )2]
C 2 crF v211:C 2cr}
Noting that the second integral in the right-hand part of (8.86) can be calculated using the
probability integral we find eventually
mx - F {m
_ cr- 1
- F [ -+#"
e crF 2
- -m-
(Cil-
crF
F )] - r:::-
1 exp[(Cil-
V211:
- mF
2
2crF
)2]} +LlA[--:7
1
2
(eil-
- -m,..)]
crF
- . (8.87)
Further
1 1
q=-2 00f(F-mF)-[F-(F-eil)l](F-eil)] r:::- exp [
1 (F-mF)
2
2]dF=
crF -00 e V211: crF 2crI-"
Carrying out in both integrals the change ofvariable (F - mF)/oF= A, we find
q= ~I I'
2
il
f(mF+crFA)Aexp [ - - dA+ r:::-AJ 00
2
fAexp - - dA= [A J
211:cr Fe -00 2 V211: cr F v 2
-------------------------------
=~ "fAexp (-~
2 J dt...+--I- " (-~
f,,2 exp 2J dt...+---L'>--fAexp
"" (-~
2 J dt....
Ji;crFc -oc 2 Ji; -00 2 Ji;crF" 2
lntegrating the second term by parts, we obtain finally
q= -mI- exp [(CL'>-ml.)2]

- I
---Aexp ( -A- +
2
JI"
~21[cr I'c 2cr7' Ji;c 2_ x
= ~[~+:?T( cL'>- mF )] . (8.88)

c 2 cr F
The dissipation of energy in the system equivalent to the loss during collisions, can be calculated
during the given method from formulae oftype (8.54), (8.57). Later we will use the values obtained for
the coefficients of statistical linearisation to examine the statistical dynamics of some characteristic
vibro-impact systems. We will also compare approximate and exact solutions (see § 13).
Chapter 3. Vibroolmpact processes In systems with single
degree of freedom
I owe a lot to my engineering training beeause it taught me to tolerate approximations.

Previously to that I thought ... one should just eoneentrate on exaet equations all the time.
Then 1 got the idea that in the aetual world all our equations are only approximate.
Pau] Dirae.
§9. Forced vibration

1. Using the methods of harmonic linearisation described in §7, we will
continue by analysing forced vibration of a linear oscillator with rigid stops, which
will occasionally be referred to as impact oscillators later on. We will analyse the
basis of using the method of harmonic force and motion linearisation, and
compare the results with the exact solutions obtained in §2 to §3. We will begin by
studying the non-symmetrical system (Fig. 2.1, a), assuming as before, that the
stop is arranged with gap (interference) ß relative to the equilibrium condition of
the system without stop.
A characteristic of unsymmetrical vibration, as mentioned in §2, consists of a
constant component mx in the periodic solution x(t), arising because of the basic
non-symmetry of the structure or extemal force acting on it. In addition, it is
important to emphasise that the constant component of vibration in a nonlinear
system can occur even during symmetrical excitation (see §2). Moreover, when a
constant extemal load mp is applied to a nonlinear system with a symmetrical
structure, a constant displacement (drift) m x can occur during vibration, which is
not proportional to m p •
A suitable case for the study of systems with both factors mentioned above
influencing the appearance of drift, can be found in the following general form of
the differential equation of the system:
(S2 +bs +02)X+E(X,SX) + <l>D(X,SX) = P(t) (9.1)
Here E (x, sx )and D (x, sx) are the elastic and dissipative forces of the
characteristic impact pair, while /, (- x,sx) =1'= E (x,sx) ,
pet) = mp + a p cos( Olt + <p), where in the general case ap = ap ( ol).
Searching for an approximate periodic solution for equation (9.1) in the form
x(t)""mx+axcosOlt, sx(t)""-axOlsinOlt, (9.2)
we can transform the periodic applied extemal excitation PU) using (9.2):
. . ) =a-
a p (cosOltcos<p-smOltsm<p p ( sin <p) x 0 ,
cos<p+s-- (9.3)
ax ol
126 Chapter 3 Systems with single degree of freedom
-----------------------------
where XO = x-mx'
Substituting (9 .3) in (9 .1), we obtain
[ s2 + (b - a p sin <p) s +~~

r'\ 2
- a p cos <p] x ° +
axffi ax (9.4)
+ 02 mx + <1> Ax,sx)+ <1> f)(x, sx) = mp .
Thus, using the character of the assumed solution the initial non-autonomous
system can be converted to an "equivalent" autonomous one. Such a concept
allows us to consider various types of vibration of nonlinear systems such as
forced, parametric or self-sustained vibration in a similar manner, and this
approach will be widely used later.
For the nonlinear functions <1> E (x, sx ),<1> f) (x, sx), using their linearised
expressions (see §7):
<1>r;(x,sx)"'m F +k(x-mJ, <1>1J(x,sx) "'ßsx, (9.5)
where the coefficients of harmonie linearisation ml ., k and ß are found using the
unknown parameters ofthe solution x(t).
+,'
Substituting (9.5) into (9.4) and separating into similar terms, we obtain
[" ++ ß- a'a::~ } + a, :~S~ l]x" +m,. +n'm, ~ m,. k- (9.6)
Separating the constant and centred components in (9.6), we find

(9.7)
[
S2 + (b + ß _ a p sin<p)s + 0 2 + k_al' cos<P]X O = O. (9.8)
axffi ax
Equation (9.7) represents the final relationship, defining the connection
m x = mx(aJ . (9.9)
This expression gives the relationship between the value of the drift to
vibration of the system. In particular, using the method shown, relationship (9.9)
can be found in practice without solving the differential equation for the system.
Expression (9.9) allows us to exclude the value of m, from the equation for
coefficients k and ß, after which equation (9.8) can be studied independently.
The linearised form of equation (9.8) has the following characteristic equation:
@(s) = S2 +(b+ ß _ al' sin<p)s +0 2 +k _ al' cos<p = O. (9.10)

axw a,
The conditions for the existence of periodic solution (9.2) for the Iinearised system
(9.8) is the occurrence of a pair of imaginary roots s = ±jw to its characteristic
equation (9.10). Substituting s = jw into (9.10) and setting the real and imaginary
parts to zero, we obtain a system of equations of the form
sin<p = (b + ß)wa x laI" (9.11 )
cos<p = (0 2 + k - (J) 2 )a x I a I' . (9.12)
These represent a particular case of expressions (7.18), (7.19).
§9 Forced vibration 127
Squaring the equations (9.11) and (9.12) and adding the left and right parts, we
obtain the vibration amplitude in terms of the system parameters and the
linearisation coefficients
a x = a p [(Q2 + k - ( 2)2 + (b + ß)2 w2rl/ 2 . (9.13)
Using the relationship for the linearisation coefficients and equation (9.7),
equation (9.13) is the equation for amplitude ax' which can be found using either
graphically or by caIculation. Hence, from formulae (9.11), (9.12) the value of the
phase <p can be found.
Let us examine equation (9.13). The relationship
0)0(ax)=~Q2+k(ax)' (9.14)
as already discussed in §7, section 1, represents an approximation to the equation
of the backbone curve on the basis of the fundamental for the system. Substituting
0) = 0)0 in (9.13), we obtain the equation for the energy boundary of vibration
a x =ap/[(b+ß)O)], (9.15)
which defines the maximum obtainable amplitude, since any violation of the
equality 0) = 0)0 leads to a reduction in the value ax in (9.13) by comparison with
(9.15). Note that equation (9.15) takes the same physical meaning as that obtained
in § 3 far the exact boundary (3.17), however the structure of equation (9.15) is
considerably simpler. In section 3 we will compare these relationships.
The intersections of curve I for the relationship (9.15) (Fig. 9.1) with the
backbone curve 2 represents a solution to equation (9.13), defining an
approximation for the maximum point of the amplitude-frequency characteristics
of the system, the resonance branches of which are situated along the backbone
curve. The remaining points of the amplitude curve can be found conveniently
using ofthe following graphical example, fully described in [124]. Examining the
natural frequency of the system using 0)0 as argument, we construct the graphical
function 3 (Fig. 9.1)
ax(w o) = ap[(w~ _( 2)2 +(b+ß)2 w2r l/2
for various fixed values 0) with the remaining parameters of the system fixed. In
this figure we draw the backbone curve 2
(9.14), for the identical parameter values.
flz/& The intersection points of both curves
determine the solution required. From Fig.
9.1 it can be seen that due to the character
of the non-isochronous system several
1 periodic solutions can occur for the same
values ofthe constant parameters.
For systems with single-sided impact
DL-.----::f~--7Z--....,J:!:--"-1I-=a pairs acceptable accuracy is obtained using
the Iinearisation coefficients (7.65) to
Fig.9.1 (7.66) and estimating the energy
dissipation during impact from (7.61). In this case the procedure for an analytic
solution can be considerably simplified, since the equation for the constant and
periodic terms ofthe solution can be immediately separated: expression (9.9) takes
the simple form (7.63), while equation (9.8) can be solved independently, since k
is a function of only one unknown parameter ax ' We will now find this solution.
Noting that in the case examined the operator L(s) takes the form (6.72) we
find for formula (7.66)
L(O) = 1/0 2 .
The value m s in (7.66) is expressed in terms of the parameters of the system
considered in the following form:
ms = mp 10 2 • (9.16)
As a result, in this case the linearisation coefficients (7.65), (7.66) take the
form
mF = m p + (a x _ ~)o 2 , (9.17)
k = 20 2 ( 1- ~ - : : 102 J. (9.18)
Substituting (9.18) in (9.l4) we find for
r
the natural frequency (00 of the nonlinear 11;/4
system
(6
2
=3- 2(~-::102). (9.19)
1
Expression (9.19) gives an obvious
representation of the relative configuration
of the backbone curves of the system. By
satisfying the condition
o f Z
m p 10 2 = ~, (9.20) Fig.9.2
i.e. when the static force m p completely takes up the arranged gap or interference
~, the system remains isochronous at the natural frequency
(00 = JjO "" 1.730 (9.21)
and its backbone curve takes the form of a vertical line 1 (Fig. 9.2). When m p == 0
we find from (9.20) that the isochronous state in this case appears when ~ = O. The
fact that the system considered is isochronous when ~ = 0 was described in § 2
when considering exact solutions of the equations for free vibration. In that case
the exact value of the frequency obtained was
(00 = 20. (9.22)
Comparing (9.21) with (9.22) shows the size of the error produced in this case
by using the approximate approach.
The value ~o = ~ - (m p /0 2 ) determines the resulting set of the system.
When ß o > 0 in the system implies a gap; when ß o < 0 implies interference. Fig.
9.2 shows the branches of the backbone curve for systems 2, 3 using formula
(9.19) with fulliines for the cases ~o > 0 and ß o < O. Comparing Fig. 9.2 with the
graphical exact solution (Fig. 2.3) shows the very high quality correlation obtained
by the approximation.
The amplitude-frequency characteristics were drawn for the system according

to formula (9.13) using (9.18), (7.61). Fig. 9.3 shows several typical
configurations for the amplitude-frequency characteristics of systems considered.
The backbone lines and energy boundaries of vibration are shown dashed. For
simplification the latter were drawn for R = 1. Naturally, when Ll o ~ 0 systems
can only develop vibro-impact regimes of motion and the development of such
regimes takes a different character. In the case when GI' > 0 2 Ll o , the system
I1p·Gonst Jo>O flp-G(J2
11: 11;
/J o /J(J
0 1 lJ/f1 (J f (J/Q
a)
b)
-10 .0
Ilz flz
cl)
()
excitation has a soft character, i.e. vibration exists whatever the initial conditions.
However, vibro-impact regimes can exist in systems even when 0 2 Ll o (Fig.
GI'<
9.3, e,f). [n these cases the change to a vibro-impact regime can occur only as the
result of hard excitation (an additional starting impulse). The value of interference
does not influence the character of the backbone curves, so that increasing the
contact force can lead to an increase in the value of the starting impulse needed
but does not eliminate the possibility of establishing vibro-impact regimes. It
follows from Fig. 9.3, e, fthat such regimes can lead to particularly dangerous
oscillation when slowing down (00-70).
With a gap (ß > 0) vibro-impact regimes can exist only when ax > ß o' In the
case ap = const, it is possible for regimes to develop into impact types smoothly as
the frequency of excitation increases and there is a specific pulling typieal for
nonlinearities ofhardening types (Fig. 9.3, a). When al' = E(j)2 a situation can arise
(Fig. 9.3, b) when the change from non-impact regimes to vibro-impact regimes
requires hard excitation, hence the latter can be undetected when testing systems
on vibration devices.
As can be seen from Fig. 9.3, b (see (9.15», an effective method ofcombating
vibro-impact regimes of the types considered is to increase the dissipation of the
system.
2. We shall continue the basic search for non-symmetrical solutions to the
equation of a linear oscillator with rigid stop using the method of harmonie force
linearisation. In this case the elastic effects of a linear spring must be incJuded in
the basic nonlinear elastic characteristics ofthe system (x) , and the energy loss
leads to the introduction of an equivalent dissipative force ßx with coefficients
given by (7.61).
The differential equation wh ich results from applying the method studied can
be written as:
[S2 +(b+ß)s]x+(x)=P(t) (9.23)
or following the method described above for reducing the system using the
character ofthe excitation and solution ofthe type (9.2), to equivalent autonomous
differential equation (see (9.3» ofthe following form:
.[S2 +(b+ ß - ap:~s<p}_ al':~s<P}o +(x)=ml" (9.24)
To linearise equation (9.24), we put

(x) :::: mF + k(x - mx ) • (9.25)
As a result from (9.24), (9.25) we find for the constant terms
m F = ml' (9.26)
Condition (9.26) can be accounted for immediately in the expression for
k = I / q (see §7, section 4). Hence the value of k depends only on aF and the
linearised equation for the periodie component, as in the previous case, can be
solved independently. It will take the following form:
[s 2+ (b + ß- ap Sin<p] s+ k-
axro
ap cos<p] x 0= 0 .
ax
(9.27)
Substituting s=jro in its characteristic equation and setting the real and
imaginary parts to zero, we obtain a system of equations similar to (9.11), (9.12):
sin q> = (b + ß)wa x / a p , (9.28)

cosq>=(k-w 2 )ax/a p . (9.29)
Solving these equations for ax ' we find
a x =ap[(e - ( 2)+(b+ß)2 ro2rI/2 . (9.30)
Multiplying both sides of (9.30) by ~e + W2ß2 and using the relationship

aF = ax ~ e + 00 2ß 2 , we obtain eventually
al- = al'(e +ro 2ß2)1/2[(e _ro 2)2 +(b+ß)2 ro2rI/2 . (9.31 )
Equation (9.31) makes it possible to find ap from which all the remaining
parameters ofthe regime considered can be determined.
The value
(9.32)
represents the natural frequency of the linearised system, which depends on the
force amplitude aF realised by the nonlinear element <l>(x), while
al- = k(aj.)a x · (9.33)
We shall use the linearisation coefficients k and ß in their forrns (7.95), (7.61)
in our analysis. Substituting (7.95) into (9.32) using here c = Q2, we find the
following equation for the natural frequency ofthe system:
('; H2(1± 1- rr/4H~" 1(4Q,JJ'

Expressing aF in terms ofax in (9.34) using (9.33), (7.95), we obtain the
(9.34)
equation for the backbone curve ofthe system
~ =;[±2-(W:/Q)2 - (W :Q)2 (l-~)l O

(9.35)
Here the positive sign before the first term implies a gap, while the negative
sign implies interference. Equation (9.35) is shown by a dashed line in Fig. 9.2. It
can be seen that in this case, the configuration of the backbone curves coincide
closely with the exact results (Fig. 2.3): for an initial gap, (L'l > 0) these are of
hardening type, with interference (L'l < 0) they are of softening type, while when L'l
= 0 the system remains isochronous. In addition, the isochronous frequency is
found with a somewhat larger error and according to (9.35) takes the value
WO =.J2Q"'I.41Q (9.36)
The construction of the amplitude-frequency characteristics in the case
considered according to formula (9.31) requires the values ofcoefficients k and ß.
For this the graphical method described in section 1 can be used (see Fig. 9.1). As
a result we obtain the configuration of amplitude curves similar to those shown in
Fig. 9.3. The unknown phase q> is found from (9.28), (9.29), using the known
value of aF"
Because the vibro-impact regimes studied are resonant, the accuracy of their
resonance curves depends essentially on the accuracy of the relationship roo(a),
R-0-5 IJ-O 1l-D-5 /)-0
~>O \.! \
f f
o ~ __ ____ __
~ ~ ~
o f z t.J/O 1 2 3 (,J/fl
Fig. 9.4
expressed in terms of the linearisation coefficient k. For this relationship can use
the exact backbone curves for systems similar to that which we already used in the
example of §7, section I, or the exact equations for the backbone curves for the
fundamental. It is convenient to make such a substitution for systems with one
degree of freedom, for which the relationship shown is easily found by the
"stitching" method or its analogy_ Nevertheless, for systems with high er degrees of
freedom this refinement is not worthwhile in practice, since the application their
stronger filtering effects automatically increases the accuracy of investigation
using harmonic linearisation.
For the unsymmetrical system considered we will use the exact equations for
their backbone curves (2.11), found from a study of the phase portrait for free
vibration. Applying an external force mp causes the value of L'l to change with
increment (decrement) mp/L'l2, hence in the general case we find
Q
(9.37)
ro
o
= arccos[l- ~axQ2 ]-1
Q L'l-m p
Fig. 9.4 shows examples of the amplitude-frequency characteristics for the
system according to (9.13) using (9.14),
(9.37), (7.61). For comparison the dotted
line shows the exact relationship obtained in
§2 ( Fig. 2.9).
3. Let us now consider the problem
f of forced vibration of a linear oscillator with
symmetrical stops, excited by a force
P(t) = a p (ro) cos( ro( + <p) (Fig. 3.1). Using
o f 2 3 fJo/O the method of harmonic force Iinearisation
in this case gives a considerably simpler
Fig. 9.5 calculation procedure.
The equations for the system as before take the form (9.23) where (x) is now
the characteristic of a symmetrical impact pair (Fig. 6.2, b). As a result we can
make use of the solution of equation (9.23) found in section 2, putting mF == 0 to

account for the symmetry of the system and using the linearisation coefficient k
and ß in their forms (8.91), (8.92). In (7.91) we must have mF = 0, c = 0 2 , and so
we find
(9.38)
Eliminating aI- and k from equations (9.32), (9.33), (9.38), we find the equation
for the backbone curves
~=~-~(~-1)
~ 11: ())~ 11:
(9.39)
According to (9.39) I ::; a/~ ::; 4/11:. Fig. 9.5 shows the backbone curves for the
fundamental of the system, drawn using (9.39) (compare with the exact values in
Fig. 3.4).
Substituting ()) == ())o == Jk in (9.30) we obtain the equation for the energy
boundary of vibration in the form (9.15). When R = I, ß = 0 and equation (9.15)
takes the form
(9.40)
i.e. it coincides with the boundary of the
existence domains for harmonic vibration
\
of a linear oscillator with viscous friction.
When ap = const this is a hyperbola, hence
f
'-.
its intersection with the hard backbone
curve can occur at one point only and the
amplitude-frequency characteristics will
show pulling in domains of high frequency f 2 3 (j/!1
excitation (Fig. 9.6, a). Of course, impact in
a)
the system corresponds to the branches of /
the curves satisfying the condition a/~ > I.
Ire
./ .-1:--._ .-
When ap = E())2, i.e. during excitation, f
.
caused for example by an external
unbalanced mass, equation (9.40) becomes
linear at a relative frequency
f z J fJ/U
a x = E())/ b (9.41) b)
and its intersection with the backbone
curves leads to the configuration of the
amp I itude-frequency characteristics shown
in Fig. 9.6, b, c. Hence depending on the &/A ........,,..,,,,,,-L...
ratio E/b it is possible to have either smooth Q~--~--~----~-
f 2 J ,4D
transformation to regimes with impacts due
c)
to the acceleration of the unbalanced mass
(Fig. 9.6, b), or to proceed to such regimes Fig.9.6
as the result ofhard excitation in the frequency range satisfying the condition (Fig.
9.6, c)
(9.42)
In real systems the upper boundary for the indicated range is limited, because
the finite power ofthe applied excitation at higher frequencies, prevents the law of
excitation considered from being achieved
(see § 12).
R- tJ.J 6/0 - 0'l Let us estimate the influence ofboth types
of damping. Substituting (7.62) in (9.15), we
3
find
2
(9.43)
oL ___""======:::::;::=
f
The concept presented agrees with the
f l J. GJ/n result of analysis ofthe exact solution (see §3,
Fig. 9.7 section 5) on the dominant influence of the
impact damping in the high-frequency range.
Fig. 9.7 shows compares the boundaries of existence domains for the regimes
considered (energy boundaries of vibration), found using exact methods and the
approximations from (9.15), (7.62). In the high-frequency range they virtually
coincide. The discrepancy between these values close to the natural frequency of
the linear oscillator is due to the fact that in this frequency range impacts have a
weak character (see, for example, Fig. 3.10), since the value of the damping
coefficient used (7.62), calculated from the assumption of aresonant phase
regime, is over-estimated.
When al' = 100)2 from (9.43) we find
10
ar = . (9.44)
. (b/O))+p
From this we can obtain the upper limits for the lower boundary of the
frequency band, in which the vibro-impact regime considered exists. Solving
(9.44) for 0), we obtain the existence condition in the form
b b
0) ;:0: "" (9.45)
(elaJ-p (eI~)-p
From this its folIows, for example, that when
~ ~ e/p (9.46)
vibro-impact regimes exist in all post-resonance frequency bands III which the
excitation is of the type considered.
As an example of use of the formula we will study the influence of energy dissipation on the
clearance in a vibration isolation system, during the design of which the dynamic model considered is
widely used. When the isolator has either natural or special stops, motion can arise due to the
application of random jolts which violate the conditions tor isolation, resulting in the establishment of
vibro-impact regimes ofthe type considered ("clatter on the stops").
Suppose that an isolation system with a natural fTequency of 10Hz (Q = 62.8l/s) is affected by
vibration excitation at a frequency of IOD Hz (Q = 628.l/s) and magnitude E = 1O-"m. The damping
coefficient is b = 12.51/s. It is necessary to lind the value of the fTee movement /\., which will ensure
the absence of clatter against the stops.
The formulae of the linear theory of vibration, neglecting the possibility 01' vibro-impact regimes
due to hard excitation of the system gives
/I.>a x = Eüi[(02_ ro 2)2 +b 2ro 2 r1/ 2 '" 1.01·1O-4 m .
From equation (9.25), for example, when R = 0.5, we obtain using (7.62), /I. > ax = 3.24 10 -4 m,
finally from equation (9.42), neglecting the loss of energy during impact, we find /I. > ax = ero/b
= 52 10 -4 m. Hence the gap taken into account must be about 4 times larger than the calculated
amplitude of deformation of the suspension. If we do not take account of the energy losses during
impact, then in the example considered the clearance in the isolator must be increased by a factor not
less than 13. Note that the theoretical study of the phenomenon of deterioration of the isolator and the
emergence in post-resonance domains of excitation frequencies in high-frequency regimes with impact
against stops was made by Y I Iorish [109] and is investigated in detail in the monograph by M Z
Kolovsky [124].
Later, as we study the quantitative features of various types of vibration of

vibro-impaet systems, for the ease of single-sided impact pairs, we will use as a
basis the harmonie motion linearisation, taking the eoeffieients k and ß in the form
(7.66), (7.61), while for double-sided pairs we shall use the method of harmonie
force linearisation with eoeffieients (7.91), (7.62), sinee this will lead to
eonsiderably simpler ealculation of the relationships and to satisfaetory qualitative
answers.
4. Harmonie linearisation of vibro-impaet systems, while it gives a cIear
representation of the eharaeter of the fundamental during periodic vibration, does
not provide fuH information about the form of the periodie motion. Of particular
importanee here is the question ofwhether or not sub-periodie regimes exist, sinee
in these regimes the addition of energy to the system oeeurs at higher frequeneies.
In praetiee, suppose that x(t) represents a partieular periodie motion for a
vibro-impaet system with one degree of freedom, having frequeney I times less
than the exeitation frequeney 00. This ean be expanded in the form of a Fourier
series
xCt) x
= --.2.. +
L>m eos - t - </>m .
00 (mw )
(9.47)
2 m=! I
Let us ealculate the work Ee of the external force pet) = apeos( wt + </» on the
motion eonsidered (9.47). Using the eonditions of orthogonality of the harmonie
funetions we find for the period of vibration
2n' Im
Ee = jP(t)x(t)dt = l1a p x,lsin</>o' (9.48)
where </>0 = </> + </>, . Henee, sub-periodie regimes ean oeeur only when harmonies
are present whieh have the exeitation frequeney (x, t 0) in the speetrum of
motion. Correspondingly, we must identify approximate sub-periodie solutions
from the equation ofmotion, taking aeeount ofthis harmonie.
Effeetive methods to approximate multi-harmonie solutions exists by the
distribution funetion linearisation diseussed in §7 [124]. Other modifieations to the
ealculation sehemes for multi-harmonie linearisation are known [107]. All these
examples ean be applied to the study of sub-periodie motion of vibro-impaet
systems, but, sinee such applieations have no special features eompared to the
traditionally used methods deseribed, we will not eontinue with this here.
The speeifie eharaeter of vibro-impaet systems makes it possible to propose a
136 Chapter 3 Systems with single degree offreedom
method of finding existence conditions for sub-periodic regimes using a much

more complete representation ofthe solution sought. Let us find the work done by
the dissipative force on the motion studied (9.47) for one period of vibration. For
the system with viscous damping b:ic we find
nbw
Zn/! ro . 2 2 2 00
Ed1 = fbxdt=-Ixmm. (9.49)

o I m~l
To estimate the loss of energy during impact using the coefficient of restitution
R, we shall substitute the equation for the pre-impact velocity :ic through the
impulse J, reduced to a single mass in (7.57),
:ic_ = J 1(1 + R) (9.50)
and we find for the period in the case of single-sided collision
E = M:ic= (1-R2) = (1-R)J 2 .. (9.51 )
d2 2 2(1 + R)
For double-sided collisions this value must be doubled.
Using (9.48), (9.49), (9.51) we can write the equation for the balance of work
ofthe dissipative and excitation forces in the periodic motion considered (9.47)
. nbw ~ 2 2 (1 - R)J
nal'x,lslll<po =--L.,xmm + . (9.52)
I m~1 2(1 + R)
Expression (9.52) represents the energy conditions for the existence ofmotion
(9.47). From this it can be seen that sub-periodic regimes occur in the system only
for relatively small value of the coefficients used for defining energy dissipation.
In addition, for the given value of ap the defined levels of excitation can be
reached only with significant amplitude of vibration, since due to the filtering
effects of the linear part of the system, as I increases the value of x, rapidly
decreases. For large amplitudes of vibration larger inertia and elastic forces arise.
The situation here is similar to resonance in a linear system, when vibration with
considerable amplitude occurs in comparison to the relatively sm all value of the
disturbing force, compensating for energy dissipation. For this reason, motion of
the type described is known as sub-periodic resonance of order I or fractional
resonance of order 1/1. Resonance, corresponding to I = 1, is known as
fundamental. Of course, resonant motion which appears at phase <Po in the region
of n12, ensures the best excitation.
It is possible to maintain vibration at high amplitude with a small excitation
force only when the vibration is dose to the natural vibration of the system, hence
the latter is a good representation for the synthesis of resonant motion. Since the
natural frequency Wo of the vibro-impact system considered satisfies the condition
Wo 2 0 (see (9.14)), the frequency range in which sub-periodic resonance can exist
is limited by the following condition:
wll=w o 20, (9.53)
which corresponds, for example, with that found above by the exact method in
expression (3.20). By finding the law of motion for vibration of the system by the
stitching method and presenting this in the form (9.47), we can obtain from (9.52)
the conditions for the existence of sub-periodic resonance of various multiples.
This procedure was developed by M Z Kolovsky and is explained In his

mono graph [124], to which the interested reader is referred.
At this stage, we will describe another representation of vibration during sub-
periodic resonances, based on the application of periodic Green's functions [36].
Reverting to the basic description of single-impact periodic solutions for vibro-
impact systems in the form (6.79), we note that during vibro-impact resonance the
second term on the right side of equation (8.79) is dominant. This occurs because
the first term describes forced vibration of the impact element pairs with
negligible impact interaction, but the resonant frequency of such motion differs
from the frequencies of the vibro-impact resonances. As a result vibration during
vibro-impact resonances of various multiplicities is weH described by the
simplified expression
x(t) "" -JX(t) . (9.54)
As an example we will look at the exact solution (3.32) found in §3 for the
periodic vibro-impact motion of an oscillator arranged between stops, writing this
in the form:
_
x(t) = +acosmt + L. ~
r
~ 4 ß cot[rcl I(Y)] (
~
a)
mm
1± - cos-t. (9.55)
m=1 rc/[(ms /l)2 - 1 ß I
Here a is found from formula (3.10). At sub-periodic resonance frequencies S > I,
the value a becomes negligibly small and the exact solution (9.55) is weIl
approximated by the term under the summation sign. Comparing (9.55) with
(7.79), it can be seen that the term mentioned represents the periodic Green's
functions of the system xl!) = JX(t). Comparing this to the equation for free
vibration (3.6), it can be seen that when VI = So they differ only by terms
proportional to the value alß. The latter is less than one, since according to the
results of analysis for the exact solution (see §3, section 3) vibro-impact resonance
as described only exists when a < ß .. Hence, just as the law of natural vibration,
the periodic Green's functions represent a good approximation to the resonance
solution for the equations ofvibro-impact systems.
Using the representation of periodic Green's functions in the form of series
(6.67), we can write the approximate solution sought for x(t) of period
T (T = 2rcl I m) in accordance with (9.47), (9.54) in the following form:
x(t) = - ~ L(O) - 2: ~IL( m~m )Ico{ ~m t+ argL( m~m)]. (9.56)
As a result the amplitude x m ofthe corresponding harmonics takes the form
x m = 2: IL( m~m )1. (9.57)
The value J is found from the condition of impact, that when t = 0, x = ß.

Substituting in this condition the equation (9.54), we find
J = -ß/X(O). (9.58)
Since J has a non-negative value, in agreement with the coordinate axis direction
assumed in this book, then the sign ofthe value X(O) is opposite to the sign of ß.
Using (9.57), (9.58) the equation for energy balance (9.52) takes the form
Ll bw' m·w I-R

I L _Y- m
XI (0) {1T.l 'I
m=1 (J1
00
apwIL(jw)lsin<po = - - - --3
2 2 + , (9.59)
I 2(1 + R) }
in wh ich <Po = <p- argL(jw). From (9.59) we find two possible va lues for the phase
. apw I L(jw) I
<Po = arcsm { 2 } ,
~IL(-l-)I m + 2(J~R)
L1 bw 300 mjw
-XI
2 l-R
(0) nl 3
(9.60)
. apw I L(jw) I
<Po = 1T. - arcsm _ ._~{bW3 :tI L('!!1.ü}JI 2
m2 + 1- _I!_}.
XI (0) nl 3 m=1 I 2(1 + R)
Since Isin<Pol :::; I,the existence conditions for re sonant regimes can be written as:
Ll bw m·w 1- R
I L(-Y-)
3 2
z--
Xo
apwIL(jw)1 { -3
2
m + }
. (9.61 )
X1(0) m=1 1 1
nl l 2(1 + R)
Similar conditions for the case of double-sided collisions using (6.68) will be as
folIows:
. L1 2bw 3 2m - 1 1- R }
apwIL(jw)1 z--
X2 (0)
{ -3
nl
I 00
m=1 1 I
2
L[(--)jw] 12 (2m -I) +
2(1 + R)
. (9.62)
Using solutions in the form (9.54) makes it possible to obtain the law of
motion in the form (using (9.58))
L1
-X(t). x(t) ~ (9.63)
X(O)
From this all the remaining characteristics of motion can be found. Taking into
consideration, for example, that energy dissipation in the system has been assumed
to be smalI, and consequently, the periodic Green's functions are only slightly
non-symmetrie relative to t = TI 2, the value of the amplitude of vibration ax
during single-sided collisions can be obtained from the approximate relationship
ax ~±[L1-X(~)]~ ~[I- XI~o/{~)l (9.64)
The value ofthe pre-impact velocity x_ is found from (9.50), (9.58):

Ll
x_ = (9.65)
X(O)(l + R)
Obviously when b is small the damping in the equation for X(t) can be neglected,
since away from the frequency of linear resonance
minimal influence.
*- Q) the damping has (wll
The permissibility of using the relationship obtained is proved since the
condition
Gx > G, (9.66)
is satisfied, where G as before is calculated from (3.10). When ~ :;t0 we can use a
simpler and sufficient condition
(9.67)
which can be checked before the construction of approximate solutions. Let us
express the conditions (9.61), (9.62) with the help ofthe parameters ofthe linear
part ofthe vibro-impact system with one degree offreedom. On the basis of(6.72)
we find
---;========. (9.68)
Taking into account in equation (9.68) that b is assumed to be sm all and using
condition (9.53) it is possible to ignore the value (bmwIP). When summing over
index m for all m > I, terms Q2 also can be neglected. Using this simplification,
substituting (9.68) in (9.61) and summing the corresponding numeri ca I series, we
obtain
GpW > __~_ bl

(9.69)
!W 2 _Q2!- xJO) 1(W
where according to (6.75), the value XI(O) takes the form
( nb) . 2nO I
exp - -;- sm -m-
XI (0) = - - = - - - - - - ' - - - - ' - - - - - - - - - - - = -
°1[1 + exp( - 2:b) _2exr( _ ~ )cos 2n~1 ]
[ Q, • ~Q' ~ b: 1 (9.70)
For symmetrical double-sided impact pairs using a similar simplification we

find from (9.62)
GpW > __ ~_J2bl[ 1 +~-ll+ l-R} (9.71)

!W 2 _Q2!- X2(0)11(W (1_Q 2 12Iw 2 Y 8 2(l+R) ,
where according to (6.76), X2(0) is found from the equation
( nb) . nO I
exp - - sm--
X2(0)= 2m m . (9.72)
°1[1+ exp(- -~
2nb) + 2exp(nb) nO
- 2m cos--;-
I ]
In the case, when wll» Q, equation (9.71) can be simplified by neglecting

the value ofQ 2:
Cl [7tbl 0)(1- R)]

a p 2':- X2(0) 4+ 2(1+R) .
(9.73)
From (9.73) it follows that as the excitation frequency increases, the relative
effect of dissipation during impact grows by comparison with viscous damping.
This result agrees with that presented during the analysis of exact solutions (see
§3, section 5).
As an example we will find the limiting value of ~ in an arrangement of stops and the
corresponding impact velocity x_ for a linear oscillator with double-sided collisions for the case of
sub-periodic resonance of the third order (I = 3).Suppose the parameter system and excitation take the
values n = 62.8\ Is; R = 0.5; b = \2.51 Is; ap = 4g = 40 rnli; (() = 236.1 Is.
From equation (9.7\), we calculate
J = -~/X2(0) =0.189kgm/s, (974)
consequently, according to (965), x = 0.\26m/s. In (972), if we neglect the value of viscous
friction (b", 0) we find X2(0) = -24.5.IO'3 s/kg Finally, from (9.74), we find ~ = 4.65.10,3 m. For
comparison we will find the similar value of ~ from the exact expression (3.\7). After a considerably
more difticult calculation using (3.\ 0), (3.16) we obtain a = 0.82.10,3 rn; ~ = 4.4310,3 m. We can see
that condition (9.67) is satisfied, which explains the high accuracy ofthe approximate solution.
In conclusion, we note that the approximate approach described can be used

for other forms of dissipation and excitation force, guaranteeing resonant vibration
of the system with arbitrary degrees of freedom (see § 14, section 4), and, taking
account of (6.90), with the use of any mathematical model of impact.
§ 1O. Stability and transients of forced vibration

1. Let us now look at the stability of the steady-state solutions found in §9
for the equations of forced vibration in relation to small disturbanees. For this it is
convenient to make use of methods which correspond to the approximate character
of the solution sought. The majority of such methods are based on the assumption
that small disturbances do not change the form of the solution, but will only lead
to the slow drift of its parameters. In the case of a harmonie solution this means
that the disturbed motion takes the form of "harmonie" vibration with term (I) , m,
with amplitude a,(I) and phase (p(t) which change slowly with time:
x(t) = m,(I) + a,(I)COS[O)I - (p(t)]. (10.1)
For convenience, the phase from the expression for the disturbing force will in
future be transferred to the solution, i.e. by using a shift ofthe starting time.
By presenting the disturbed motion in the form (10.1) we can conveniently
develop caIculation methods to study the stability of approximate solutions based
on various modifications to the method of slowly changing amplitude [221].
These assumptions also make it possible to study stability directly from the
harmonie linearisation equations for systems. The reduction of systems to the
equivalent autonomous ones makes it possible to formulate the necessary
conditions on the basis of the stability of systems at the vibration boundary.
The following convenient form for the stability conditions can be suggested:
§ 10 Stability and transients offorced vibration 141
[de(ax'ro)! daxlax=a x > o. (10.2)

where e(a x' ro) is the imaginary part of the characteristic comp/ex 9lXjro) of the
linearised system, reduced to the equivalent autonomous one, at the frequency of
the periodic solution considered. The tilde above the corresponding value implies
that when substituted into (10.2) it is considered as a slow variable, changing
around the constant value, corresponding to the steady-state solution. We will
show that condition (10.2) represents the disturbed balance of work of a non-
conservative force for the period of stable motion. For example, for a linearised
system described by equation (9.1) and the expressions (9.5), the work of the
dissipative force (b+ß)x for the period of motion (10.1) with "frozen" slow
parameters is found from the expression
2x/CiJ
Ed = f(b + ß)[X(t)]2 dt = n(b + ß)roa; . (10.3)
The work done by the external force P(t) = apcosrot during this motion will be as
folIows:
2rr./w
Ee = jP(t)X(t)dt=1tapaxsin. (10.4)
The energy condition for stability [22 I], reflects the fact that, during change of the
a
disturbed amplitude x in the dose neighbourhood to the steady-state value a x '
the balance of work due to the dissipative and disturbing force is disrupted in such
a way so as to compensate this disturbance:
d
-da( Ed -Eehx =ax >0.
x
Substituting here (10.3), (10.4), we obtain
~(b+ß- ap_Sin] >0. (10.5)

da x axro_x lI =U x
At the same time, by applying condition (10.2) to the characteristic equation

(9. I 0), we obtain expression (10.5) directly. Thus, condition (10.2) represents a
more convenient form for the energy conditions for stability.
2. In (10.1) three unknown functions appear instead of one, and we can
connect them in turn to two arbitrary conditions. Let us assume that in the
disturbed motion mx(t)andax(t) are related in the same way as in the steady-state
case, and that this applies also to (t)andax(t).
Differentiating (10.5) using the relationship for in terms ofax , we obtain
dß + a::' 2 (sin - dC{J

[ da da x
a cos]l
x > o. (10.6)
x roa x _
Qx=U r
The term (d(j)/ daJax=a x is found by differentiating (10.12) as an implicit function

ofa.:
(10.7)
Substituting the expression for simp from (9.11) into (10.7), we obtain
( cfi{> ) = I [(0 2 + k _ co 2 )+ a ~] (10.8)

Ga x ax=a x (b + ß)coa x x Ga x uxea x '
;J+
Using (10.8), (9.11) and (9.12), condition (10.6) takes the final form:
[C0 2 (b+ ß{b+ ß +ax
;J1,._,
(10.9)
+(n' +k -oo'{n' +k -00' +0, >0
Expression (10.9) coincides with the stability conditions obtained by Van der
Pol's method using a much longer transformation [221]. In addition, the study of
stability using Van der Pol's method gives another condition
[2(b+ß)+a xdß/Ga xk o ax >0, (10.10)
which, however, is satisfied for all forms of dissipative force.
When using the method of harmonie linearisation by force it is possible to
retain the above reasoning, bearing in mind that the linearisation coefficients k and
a
ß are implicitly related to x . Then, using the transformations
dk _ dk Gal- _ dk k dß _ dß Gajo _ dß k
Ga x - Gajo Ga x - Gal- 'Ga x - Ga, Ga x - Ga F
condition (10.9) leads to a convenient form for applying this method:
dk )+CO\b+ß)(b+ß+al- dß )]
[(k-C0 )(k-C0 +aF Gal-
2 2 . >0. (10.11)
GaF _
ll,.-=a!<
Here k and ß are the force linearisation coefficients, dependent on ajo; aF is the
steady-state value of al- , corresponding to the periodic solution studied.
It is shown in [221] that the boundaries of the stability domains satisfying
condition (10.9), appear on the (co, aJ plane as the locus of points, in which the
resonance curves constructed from equation (9.13), have vertical tangents. This
implies that the branches of the amplitude characteristics for nonlinear regimes,
which lie above the backbone curves, occur in stability domains. In double-valued
regions below the backbone curve, the branches correspond to unstable solutions.
We will demonstrate this using as an example vibro-impact systems in which the
coefficient of harmonie linearisation k is described by expression (9.18).
Assuming for simplicity b + ß '= 0 , we obtain from Cl 0.9) two boundary branches
for the stability domains
[0 2 +k(ax)-co2k=a, =0, (10.12)
[0 2 +k(ax )-co 2 +axdk(aJ/Gaxk=a x =0. (10.13)

§ lOStability and transients of forced vibration 143
Expression (10.12) represents the equations of the backbone curves.

Substituting (10.18) in (10.13), we find
ro=.J3n . (10.14)
This expression defines the asymptotes of the backbone curve (see (9.21 )).
Analysing inequality (10.9) we conclude that the stable branches ofthe amplitude-
frequency characteristics are arranged above the backbone line, and the unstable
branches between the backbone curves and their asymptotes, which proves the
point. Fig. 10.1 shows the unstable domains hatched, over the curves of the
amplitude-frequency characteristics for systems with single-sided stops having
either gap (Fig. 10.1, a) or interference (Fig. 10.1, b). Of course, the results
obtained are valid only for nonlinear solutions, since Fig. 10.1 , a shows another
boundary ax = ~, which separates existence domains for linear and nonlinear
solutions. Similar results are obtained using condition (10.11). Note that the
stability conclusions agree weil with the known conditions for nonlinear systems
and with the results obtained when applying exact methods to the analysis of the
vibro-impact systems considered (see §2).
3. We shall now look at transient processes for forced vibration, from some
initial conditions under the action of a harmonic force P(t). The study of such
effects is important in the use of re sonant processes, which we often try to fulfil in
vibro-impact systems. At the same time the dynamics of such processes is
interesting, in terms of the conditions studied, for the passing of resonance in
machines during run-up or run-down, i.e. during growth or decay ofthe excitation
frequency. Note also, that the approach used allows the limits ofapplication ofthe
methods of studying stability of periodic motion described above to be further
refined, and also to examine stability of such systems "at large" with finite values
of disturbance .
az
~ I~l
f 1
0
Q
f Z llJ/fl I 2
a) b)
Fig. 10.1
We shall now turn to an example of a second order symmetrical system,

described by the equation
(S 2 +bs+n 2 )x+<l>(x,sx)=apcos(rot+(j»), (10.15)
where (j) is a variable, found during the solution process and inserted on the right
hand side for convenience of analysis. The solution required for x(t) will be
approximated by a single frequency function with variable amplitude Gx and phase
<p:
x(t) = ax(t)cosffit. (10.16)
By including phase on the right hand side, the study of process x(t) leads strictly to
systems of time measurement displaced relative to the external excitation by a
phase shift <P(t).
In contrast to the transient of process (l 0.1), considered during the study of
stability, we will now discard the assumption that amplitude ax (t) changes
slowly, since resonance can be realised for a small number of periods of the
disturbing force even from zero initial conditions. With this in mind, it is
convenient to approximate the change in ax (t) by the following relationship
(l0.17)
where ~ represents a slowly changing quantity, found from the solution process.
As a result the rate of change of amplitude will depend principally on the absolute
value of the term ~,and not on its rate of change, since even for a small change
in ~ the expression (10.17) can describe a rapid change of amplitude through the
expression
I
ax(t) = Cexp gdt, (10.18)
where Cis defined by the initial conditions.

Using the above equation we will search for an approximate solution to
equation (10.15) ofthe form (10.16), assuming ~ and <p are change slowly.
To find a single-frequency periodic process ofthe type described, occurring in
a nonlinear autonomous system, E P Popov proposed an effective approach [181]
based on the asymptotic method ofKrylov-Bogolyubov
Applying harmonie Iinearisation to equation (10.15) to bring it to the
equivalent autonomous form using the form of solution (10.16), (l0.17), this
method can be extended to the class of system considered. Since according to
(10.16), (10.17)
(10.19)
the expressions for the harmonie linearisation coefficients according to (7.16) will
take the form
k 1
= --;:::;-
1ta x
f"<D (~a x cos \jf, - a~.
2
0
~~ ) d
x sm \jf + a x'" cos \jf cos \jf \jf,
ffi (10.20)
ß= --1- 2]<D(ax cos \jf, - axffi sin \jf + a ~ cos \jf) sin \jfd\jf ,
1ta x 0
x (10.21)
i.e. they are functions of three unknowns a x ,0) and ~ . In cases when the value of
~ is small or non linear or <D(x) is independent of the derivatives of the process,
the usual equations for the coefficients of harmonie Iinearisation can be used: in
the first case in the form (7.24), (7.25), in the second case in the form
§ 10 Stability and transients of forced vibration 145
1 2n
k=- f<l>(a x eos\jf)eos\jfd\jf (10.22)
na x 0
a
substituting ax for x in the final expression.
To eonvert equation (10.15) to the equivalent homogeneous form we will re-
write the right hand side in the form
a p eos(co! + q;) = a p eos co! eosq; - a p sin co! sin q; , (10.23)
expressing eosco! and sinco! in terms of the solution sought for Xe!). From (10.16)
we find
eosco! = x/ax ; (10.24)
from (10.19) using (10.16) we find
.
Sill co! = - -sx- + -~x- . (10.25)
axco axco
Substituting (10.24), (10.25) into (10.23), we find
~
a p eos(co! + <p) =~
a p sin q; ap [ ~ -Sill
sx + -;::;- eos(x,sx) = kx + ßsx
in (10.15) and substituting in (10.26) we obtain the following autonomous system:
[" ++ß -a'a:;~~lHQ' +k - ;Jos~- s;n~)}" 0! (10.27)
Transient proeesses of type (10.16), (10.17) eorrespond to the existenee of pairs of

eomplex roots s = ~ ± jco to the eharaeteristie equation of the linearised system
(10.27). Substituting s = ~ + jco in the latter and setting the real and imaginary
parts to zero, we obtain
(10.28)
~ = ~ ( a p sin q; _ b -
2 axco
ß) . (10.29)
Equations (10.28), (10.29) define how q; depends on axand ~ on a x ' When ~ == 0

we obtain the existenee eonditions for steady-state regimes (a x = a x = eonst),
which coincide, naturally, with (9.11), (9.12).
In the case when expression (10.16), (10.17) represent the disturbed motion of
the system, assuming C = ax in (10.18), we see that as amplitude x varies from a
the steady-state value ax ' stability will be maintained as ~ ehanges sign aeeording
to the inequality
(10.30)
Substituting in (10.30) the expression for ~ from (10.29), we obtain the eondition
for stability in the form (10.5). Henee, eonditions (10.2) and (10.30) are
equivalent.
Expression (10.28) also shows that the calculated expression (depl aaJux:a x
used in section 2 based on (9.12), satisfying the steady-state case, appears correct
only for sufficiently small values of the damping factor b + ß. This also defines
the limits of application for methods developed in sections 1 and 2 to investigate
stability of forced vibration in non linear systems with sm all disturbances. With
larger dissipative losses, the expression (d~1 aaJax:a x must be calculated on the
basis of the complete expressions (10.28), (10.29). We will now find this value,
assuming for simplification that ß == O. In this case, neglecting ~2 in (10.28) as of
second order accuracy and differentiating (10.28), (10.29) as implicit functions of
a x ' in the vicinity ofthe steady-state value ax we find
[b d~ + dk +
aax aax
~~ (a x aad({Jx sin (j5 + cos(j5)] _
ax
=0 ,
ax::::a x
[ - d~ ) = a p (~d({J
[ --:::z- a x -cos<p~ - sm<p
. ~)] .
cRix _ 2a.oo cRi x _
ax=a x . ax=a x
Eliminating the value (d~ 1 aa x hx:a x from these expressions, we obtain using
[0
(9.11), (9.12):
( - d({J) - -
2
+ k - 00 2 + axdk 1aax - b 2 /2] (10.31)
aax u,:a, - a xboo[I+(02 +k_oo 2)/(2oo 2 )] a,-a x '
Comparing equations (10.31) and (10.8) and noting that in the domains
considered, the range ofthe system parameters is given by
(0 2 + k - (02 )/(2(02)« 1
since they are close to resonance, we find that, for small b they differ only to
second order accuracy. Consequently, it is unsuitable to change from the
simplified expression (10.8) to the full expression (10.31), during the study of
qualitative principles, as indicated.
4. The transient process is c1early illustrated by the dependence of ~ on ax '
To find it we can eliminate phase (j5 from the system of equations (10.28),
(10.29). We find
(10.32)
Expression (10.32) represents an algebraic equation of the fourth degree in ~. To
simplify further analysis we will put b + ß= O. In this case equation (10.32) is
transformed to the equation
p +0 2 +k-00 2 -~apla; _4oo2~2 =0
which can be converted to the biquadratic equation
• ~4 +2~2(02 +k _ (02) +(0 2 +k - (02) - a p ja; = O. (10.33)
From equation (10.33) we find eventually
§ 10 Stability and transients of forced vibration 147
~2 =_(02 +k_(2)+(02 +k_(2) 1+ apl'a; _(0 2

+k_( 2 )2 . (10.34)
(0 2 +k_( 2)2
We shall look at the case when the transient process satisfies the resonance
condition
0 2 +k-m 2 =0. (10.35)
Substituting (10.35) into (10.34) and assuming that ap is a sm all value, we can
expand the radical as apower series, limited to the linear terms. As a result we
obtain
2
-2 ap
S =2a;(02+k-m 2 )
whence
~ = a p /(2a x ffi) . (10.36)
Substituting (10.36) into (10.17), we find
Mx / dt = a l , /(2m) (10.37)
and consequently using (10.35)
_ ap ap
ax =-t = t. (10.38)
2m 2~02 +k
In the absence of nonlinearity (k == 0) we obtain the known formula for the
development of resonance of a linear system. Substituting in (10.38) the value k
appropriate to various nonlinearities, we find how the resonance amplitude
changes during run-up and run-down of a non linear system. Since expression
(10.38) is obtained from condition (10.35), it corresponds to the most stressed
transient process.
As an example we shall look at the character of passing of resonance band of frequencies of an

oscillatoL colliding against a stop. We shall make use of the expressions for the linearisation
coefficients k in the form (9.18). When mp =' 0, we find
k = 20 2 (\ - f\./a x ). (\039)
Substituting (10.39) in (10.38), we obtain
-.. Qp
(10.40)
a x = 20)3 _ 2M axt.
From (\0.40) it follows that, when the stops are arranged with initial gap (f\. > 0), with increase in ax
the rate of growth of the vibration amplitude during run-up of the system within the limits of the
resonance zone decreases from the rate oflinear resonance (when ax '" f\. ) to the value
(10.41)
when ax -+ 00. When the stops are arranged with interference (f\. < 0), then during run-down an
increase in the velocity of growth of the system occurs from the infinitely small value for low values of
a x to a value (10.41) (when ax -+ 00 ). Hence, the velocity of growth of resonant vibration during
run-up is in principle higher than during run-down. When f\. = 0 the system swings at resonance with
velocity (10.41), almost halfthat ofthe linear case.
§ 11. Parametrie vibration.

1. Parametric vibration of mechanisms, machines and structures is an
essential factor of their vibration activity, often causing complete break down of
their nonnal operation due to parametrie resonance. A range of such examples is
presented in [91]. As a rule, this study leads to a linear formulation based on the
analysis of Hill or Mathieu equations [203] and the analysis is restricted to finding
the conditions which cause parametric resonance to emerge.
However, an estimate of the level of vibration for such systems can be
obtained only by evaluating aseries of additional nonlinear factors, accompanying
the actual vibration. These include fuller characteristics for elastic and dissipative
features of systems, gaps, stops, etc.
I~
- g(t)
a)
I
JJ
Fig. 11.1
--g(t)
b)
A study of non linear parametric effects makes it necessary to use approximate

methods of nonlinear mechanics. In particular, N N Sogolyubov and N M Krylov
applied asymptotic methods [138], and in the monograph by V V Bolotin [65], the
study of nonlinear parametric effects is combined with the problem of dynamic
stability of structures using Van der Pol's method and by complete expansion of
Fourier series. These methods give identical first approximations, and it is normal
to study nonlinear parametric effects within such a framework. In addition,
another investigation, described by V 0 Kononenko [132] and based on
asymptotic methods, consisted of crossing through the parametric resonance
domains ofnonlinear systems when studying interaction with the exciting vibrator.
The existence of gaps and rigid stops leads to the onset of collisions,
considerably influences the character of parametric vibration. It is quite natural to
extend the methods explained to the process indicated. This section is devoted to
the description of such an approach [42]. Sy using frequency methods and the
ideas of hannonic linearisation, we can obtain a c\ear physical interpretation of
transfonnation, which will lead to solutions, and will give a much shorter route to
the basic expressions for the first approximation.
We shall look at a second degree system with one periodicallY changing
parameter, writing the non linear equation in the form
x + bx + n 2 [1 - h cos( rot + ( x, x) = 0. (1 1. 1)
Here h is the amplitude of modulation, where, in the general case h = h(ro); (x,x) is the non linear elastic
and dissipative characteristic ofthe system. Such a type of system is representative
of a wide class of problems in machine dynamics: vibration of mechanisms
§ 11 Parametrie vibration 149
arranged on oscillating bases [117, 186], vibration of anisotropie shafts in bearings

[67], vibration ofmechanisms with non linear transformation functions [220], etc.
As an example, Fig. 11.1 shows two dynamic
models of mechanisms with vibrating bases,
having double-sided (Fig. 11.1, a) and single-sided
(Fig. 1l.l, b) rigid stops for the motion of an
impacting mass M = l,with a ball-end attached to
the base by a weightless rigid bar. The single-sided
stop can be arranged (Fig. 11.1, b) either with
: (t)
initial gap or interference ~. The zero position of
the system with a single-sided stop coincides with
the equilibrium of its linear part and we find that
situations of gap or interference differ only in the
sign of~.
Let us assume that the base vibrates according
to a harmonie law of the form Fig. 11.2
y(t) = Yo cos(mt + <p). (11.2)
We will show that the equation for small relative vibration ofthe impacting mass
of the mechanism takes the form (11.1). Let us call the angle of rotation of the
mechanism relative to the base, \jf, and let us assume that the projections of the
viscous and elastic forces on the x -axis are proportional to the velocity and x
displacement x of the impacting mass in this direction through the coefficients b
and 0 2 . Applying the inertia force oftranslational motion to the mass as ji(t) , we
can write Newton's second law for the force projected onto the x-axis:
x = -bx - 0 2 - cD(x,x) - jisin<p (11.3)
Here cD(x, x) is the force characteristic for the impacting pair. Substituting (11.2)
into (11.3) and using the expression
h = y om2 /(0 2 L), 11.4)
where L is the distance from the centre of inertia of the mass to the joint (see Fig.
1l.l), we obtain, assuming sin<p "" <p, equation (1l.l).
Let us rewrite equation (11.1) in the following form:
(S2 + bs + 0 2 )X = xh0 2 COS(ffit + <p) - cD(x,sx) . (11.5)
Equation (11.5) corresponds with the schematic diagram for the signal
transformation shown in Fig. 11.2. Here I is the nonlinear block with
characteristic cD(x, sx); 2 is the multiplying block, in which the product of process
x(t) with a parametrie effect, h02cos( mt + <p) is produced; 3 is the linear block with
operator L(s) = (S2 + bs +02)'1; 4 is the summer. From Fig. 11.2, it can be seen that
the solution sought for x(t) exists as the output from the linear block, which
represents an oscillator with viscous damping. Using the obvious filtering effects
of this block, it is natural to search for an approximate solution of equation (11.1)
in harmonie form. Such an approach allows us to make use of the output from
blocks land 2 only for the basic harmonie components and to develop a solution
using harmonie linearisation. A description for such treatment of parametric
vibration was given by A Leongard in [143].
The appearance of the arbitrary harmonie signal a x cos vt at the output of the
linear filter 3 (Fig. 11.2), after its passage through the multiplying block 2, leads to
a pattern combining tones with frequency Cü - v and Cü + v as the result of the
transformation
a x cosvt· hQ,2 cos(mt + <p) = Yz hQ,2 ax {cos[(m - v)t + <p] + cos[(m + v)t + <p]}
which moves further through the linear blocks without altering the frequency. As a
result, the frequency condition for steady-state vibration at the fundamental to
exist takes the form
Cü -v = v (11.6)
and consequently v = Cü/2. Such behaviour is called primary parametrie
resonance. By studying mutual passage of the signal throughout the loop of the
schematic, we can obtain the frequency conditions for secondary parametrie
resonances to exist
v = mCü/2 (m = 2,3, ... ) . (11.7)
To form the corresponding m-th resonance, high er harmonics, occurring in the
vibration spectrum, must pass through the loop m times until the conditions of
frequency balance arise, similar to (11.6). Noting that for each passage through the
linear block, a weakening ofthe amplitude ofthe higher harmonics occurs, so that
as the number of resonances increases so each requires higher levels of
modulation h to maintain them. It is obvious that any form of parametrie
resonance occurs only when v is close to some natural frequency ofthe system.
Let us find the existence conditions in the system for the principal parametric
resonance, occurring at minimum level of modulation. Vibration on output from
the linear part of the system using its filtering effect will be expressed in the form
x(t) m x + a x cos( mt 12) = m x + X O(t)
';:;j (11.8)
Process (11.8) exists at the entry of blocks 1 and 2 (Fig. 11.2). Let us carry out
harmonie linearisation for the characteristics ofthese blocks. For the multiplier we
caIculate directly
F1 (t) = [mx + a x cos(Cüt 12)]hQ,2 cos(Cüt + <p) =
= YzhQ,2 a x cos(Cüt /2 + <p) + hQ,2 mx cos(Cüt + <p) + (11.9)
+ Yz hQ,2 a x cos(3Cüt /2 + <p)
Noting that the second and third harmonics in (11.9) are suppressed due to the
filtering effect of the linear system, the transformation of process x(t) in the
multiplier can be approximately written using the complex coefficient of
amplification kl' carrying out the change of amplitude and phase of the
fundamental harmonic of vibration of system x(t) according to (11.9). Formally
such a coefficient can be obtained in the following way. Using Euler's formula, we
can represent the basic harmonic input and output signals of the multiplier in the
form of the sum of an exponential function
x °(t)=axcos 2Cüt =2
a x [ exp(Cüt)
J2 +exp (Cüt)]
- J2 ' (11.10)
§ 11 Parametric vibration 151
(11.11)
Identifying the input and output signals with the first harmonies in equations
(1l.l 0), (1l.lI) we finally obtain
F (t) hn 2 . hn 2 •
k l = - 01 -- = --exPN = --(eos<p+ jsm<p). (11.12)
x (t) 2 2
Henee, the linearisation eoeffieients of the multiplier depend only on the unknown
phase for the proeess.
Using harmonie linearisation for the funetion <l>(x, xs) we find
<l>(x,sx) ~ mF +(k + ßs)xo . (11.13)
Substituting from the linearised expression for the funetion on the right hand side
in (11.5), we fmd, using (1l.l2), (11.13)
[S2 +(b+ß)s+n 2 +k-kl]xo +n 2 mx +m F =0. (11.14)
Separating the eonstant and eentred eomponents in (11.14), we find
m F +n 2 m X =0 (11.15)
[S2 +(b+ß)s+n 2 +k-kl]xo =0 (11.16)
Expression (11.15) is similar to (9.7). It gives the eonneetion between mx and ax '
Substituting this into (11.16), we obtain the equation to find ax '
From (11.15), it ean be seen that the eonstant eomponents of parametrie
vibration only arise beeause of the non-symmetry of the funetion <l>(x, sx). In the
symmetrieal ease m,. == 0, and we ean find ax from equation (11.16) direetly. Let
us look at how to solve it.
As in the ease of foreed vibration, equation (.11.16) represents some equivalent
autonomous systems, undergoing harmonie vibration similar to the parametric.
The existenee eonditions for periodie regimes of type (11.8) for sueh systems are
represented by the presenee of pairs of imaginary roots s = ±joo/2 for their
eharaeteristie equations
9ll(S)=S2 +(b+ß)s+n 2 +k-k l =0. (11.17)
Substituting s = joo/2 in (11 17) and taking aeeount of (11.12), we obtain the
eharaeteristie equation ofthe systems in the form
!?/)(j (0)
2
= 0.2 + k _ ~ _ hn
4 2
2
eos<p + j[(b + ß) 00
2
_ hn 2 sin <p].
2
(11.18)
From (11.18) it follows that parametrie exeitation eauses "negative damping,"

whieh explains the effeets it eauses. Setting the real and imaginary parts of (11.18)
to zero, we find
sin<p =(b+ß)oo/(hn 2 ) , (11.19)
eos<p = 2(0. 2 + k - 00 2 / 4)/(hn 2 ). (11.20)
Eliminating <p from (11.19), (11.20), we obtain the final expression for ax ' It is
convenient to write it in the following form:
02+k=(~Y ±~~h202-(b+ß)2oo2. (11.21)
To calculate ax from this equation, it is necessary to substitute into it the values of

linearisation coefficients k and ß.
As can be seen from (11.21) it is possible that two different periodic solutions
may exist for certain values ofthe system parameters. The existence conditions for
parametrie vibration according to (11.21) take the form
h::::(b+ß)m/0 2 (11.22)
This defines the modulation for which the periodic regime of motion is sustained.
Of course, this value is proportional to damping in the system.
2. It is possible to use the method described in § 10 to study the stability of
parametrie vibration, based on analysis of the stability of solutions for the
equivalent linearised autonomous system. In the case considered, the imaginary
part of the characteristic complex, according to (11.18), takes the form
00 h0 2
8(a<,OO) = (b+ ß)"2--2-sin<P. (11.23)
Applying condition (10.2) to (11.23) and using the connection between 0. (11.24)
Gax 00 Ga x _
ux=a x
The expression ~ / Gax is found by differentiating (11.20) as an implicit function

0 (11.26)
To use force harmonie linearisation, the corresponding conditions for stability

are obtained by differentiating the equivalent force linearisation coefficients in
(11.26), k (instead of 0 2 + k) and ß, as implicit functions ofax . Since, for force
characteristics of impact pairs
Gar/aax=k>O
we find,
2
[ dß 00 dk (k 00 2 )] >0 (11.27)
Gal- + 4(b + ß) Gar - 4 ZiF~lIl'
where aF is the steady-state value of al-' corresponding to the periodic regime
considered.
In eases when the eoeffieient ß is independent of the proeess parameters or is

zero, the eonditions of stability ean be eonsiderably simplified. For example, from
ß == 0
[::J
(1 1.26) we find when
0' + k - : ' 1Lo, >0 , (11.28)
i.e. for a hard system (dk/dax > 0) the eondition for stability be comes
[22 + k > ((0/2)2, (11.29)
wh ich eorresponds to the positive sign in (11.21); for soft systems (dkl dax < 0) the
inequality reverses to:
(11.30)
and the regime is stable, eorresponding to the negative sign in (11.21). A similar
relationship is obtain from (11.27).
Note that, so far, we have not in praetiee used the speeifie of vibro-impaet
systems, sinee the results obtained for the dynamies and stability of parametric
vibration are suitable for all nonlinear systems, deseribed by equations of the type
(11.1), the nonlinear terms of whieh allow harmonie linearisation. The eonditions
for applying the methods eonsidered coineide with the usual requirements set out
for harmonie Iinearisation (see §7). In partieular it is satisfaetory to use this for
systems with filtering effeets in their linear part.
Seeondary parametrie resonanees ean be studied by a similar seheme to
determine the eorresponding harmonies at the output of the linear part of the
system. The effeets ofhigher harmonies ean be found in the same way.
As an example let us look at the case of cubic nonlinearity <t>(x, sx) = <t>(x) = yx 3 , for which the
dynamies solution is deseribed, using asymptotic methods for the first approximation, in reference [64].
For these functions the harmonie linearisation eoefficients take the form
k = 4"3 ya 2
p ß= 0 (11.31)
J
a; 3~[(~r _n2±+~h2n4_b2co2
= (11.32)
whieh eoincides with the solution speeified. We will evaluate its stability using eondition (11.28).
Noting that dk/dax = I. 5yax ' we obtain a stable solution when y > 0 , corresponding to the positive
sign of (11.32), and to the negative sign when y < O. The conclusions regarding stability have
obvious physieal meanings [64].
3. Let us now look at how parametrie vibration beeomes established. For

this, we ean use the method similar to the study of foreed vibration stability in
§ 1O. With this aim, the value of<p in equation (1l.l) will be replaeed by a slowly
changing value <p and we shall seareh for a solution for the prineipal resonanee in
the form
x(t) "" Cix eos( rut 12) (11.33)
where
(1l.34)
in which S is a slowly changing variable. The existence of a constant component
in x(t) has no specific influence on the calculation scheme to be described later,
hence we shall assume for brevity that the system is symmetrical <l>(-x, sx) = -
<l>(x, sx) and consequently mx == O.

Carrying out the transformation of the signal (1l.33) in the multiplier similar
to (11.9), it can be seen that the same gain coefficient (11.l2) will occur for
harmonie linearisation, but, in contrast to the steady-state process, coefficient k is J
nowa slowly changing function oftime

k J = lShQ2(COSQ) + jsinQ)), (1l.35)
The linearisation coefficients of function <l>(x, sx) can be found from formulae
(10.20), (10.21), while for the nonlinear form <l>(x) they coincide with the usual
harmonie linearisation coefficients, using Ci x instead ofax'
As a result, we can use the characteristic equation in the form (11.17) to study
the establishment of the process, treating the linearisation coefficients wh ich
appear in it in a similar manner.
Transient processes of the linearised system (11 .16) of the type (11.3 3),
(11.34) correspond to the existence of pairs of complex roots s = s± jm / 2 for its
characteristic equation (11.17). Substituting s = S+ jm/ 2 in the latter and setting
the real and imaginary parts to zero, we obtain using (11.35)
p +(b+ß)~+Q2 +k-±hQ2CosQ)-(~r =0, (1l.36)
~'" = 2"1 ( --;-sm<p

hQ 2
- • - b- ß) . (1l.37)
The system of equations (1l.36), (11.37) determines the relationships between

Q) and Cix'~ and Cix ' When ~ == 0 we obtain the existence condition for steady-
state regimes (Ci x = ax = const), which coincide with (11.19), (11.20).
Introducing the stability condition (10.30) into (1l.37), we obtain condition
(11.24). Let us find the error arising in the value of (di.p/ da, hx=a x when using the
abbreviated equation (11.19), using the method described in section 2 to study
stability. We shall find the value of the specified term on the basis of the full
relationships (1l.36), (1l.37) differentiating them as implicit functions of Ci x in
the vicinity of the steady-state value ax ' For simplification, assuming ß = 0 and
neglecting the square of small values in (1l.36) near the specified value of S , we
find
(b -dad~- +dadk- + -21

x x
h,",2
~l
. - di.p
sm<p-
dax
1
_
0
=,
{/xo:o{/x
x
Eliminating the value (d~ / da hy~ay from these expressions, we obtain using
(11.19), (11.20)
(:t~ay ={~bOO[I+2(~~k:~~ol/4)/002 Lx=ax (11.38)
Comparing expressions (11.38) and (11.25) and noting that for parametrie
resonance domains the value
2(0 2 +k-(j//4)(U} «I,
we find that they practically coincide.
To find the relationship of ~ to a x ' which characterises the transient process,
we shall eliminate phase <p from equations (11.36), (11.37). We find
~~2 ~
+(b+ß)~+02 +k- -
(00)2 + h 20 2 00 2 ~
----(2~+b+ß)2.
2 4 4
To simplify this equation, let us assume b + ß == O. As a result we obtain a
biquadratic equation, whose real solution takes the following form:
According to (11.39) the conditions for soft excitation of parametrie

vibration (~ :::: 0) lead to the following relationships:
from which we find
(11.40)
Assuming that when a

x =0 the system is not nonlinear (k == 0), we obtain the
well-known relationships for self-excitation of parametrie vibration for a linear
system (see, for example, [65])
~I-~ ~ 2~ ~~I+~. (11.41)
üf course, this relationship can also be obtained from (11.21) when

a x = 0, b + ß = 0 . Expression (11.41) is obtained in reference [65] as the result of
the first approximation from the equation for critical frequency, wh ich represents
Hill's infinite determinant. Assuming that h has a small value and expanding the
radical as aseries oflinear terms in h, we obtain the formula ofN M Beliayev [54]
h 00
1-- ~- ~ 1+-.
h
(11.42)
4 2.Jc 4
Let us examine the case of a transient process when resonance conditions are
156 Chapter 3 Systems with single degre.: of freedom
supported throughout the entire process

Q2+k-(ro/2)2 =0. (11.43)
Substituting (11.43) into (11.39) and expanding the radical as aseries in terms of
the small parameter h, we obtain, limiting the expansion to linear
_ h 2Q2
terms, S2 = 2 ' whence, using (11.43) we find
16(00/2)
(11.44)
When k == 0 this corresponds to the well-known characteristic index for a linear

system (see, for example, [65]). Substituting the value of k for various
nonlinearities, we find the character of the change of resonant amplitude for start-
up or run-down conditions.
Taking k = Xya; (cubic nonlinearity), then
- ho. 2
~----r=== (1145)
- 4~o.2 + Xya;
i.e. pulling du ring run-down ofthe system with soft nonlinearity (y < 0) occurs as S increases,
characterised by the rate of oscillation (see (10.18)), but run-up of a system with hard non linearity
(y > 0) is accompanied by a decrease in this term.
4. Let us now turn to examine parametric vibro-impact regimes. From the

formulas obtained above, calculation of energy dissipation due to impact is
equivalent to increasing the equivalent coefficient of viscous damping b to a value
ß, thus without impairing the qualitative side of the problem the coefficient of
restitution can be taken as unity (ß == 0) .
To study the case of symmetrical double-sided collisions we will make use of
the harmonic force linearisation. Substituting the expression for the linearisation
r
coefficients in the form (9.38) in (11.21) in place of Q2+k, we obtain
", ~ 4: [(; -0'(1- ~H~h'O' - l,

b'oo' (11.46)
where, because the nonlinearity considered is of hardening type, the negative sign
before the radical corresponds to an unstable solution.
Since impacts correspond to the value a F > Q2 Cl, to find the existence
conditions for periodic vibro-impact regimes we shall use in (11.46)
aF~QCl. (11.47)
As a result we obtain the relationship, defining the frequency range of the
vibration examined as
(~J ~Q2 -lJh 2Q 4 _b oo 2 2 • (11.48)
Because of the structure of the system (Fig. 11.1, a) the amplitude is constant in
the range shown and equals Cl.
§ 11 Parametrie vibration 157
Fig. 11.3 shows the amplitude-frequency

characteristics for symmetrical systems. It can be
" ~(}5 A~O.01 R-'seen that as the excitation frequency increases,
pulling can occur up to the break-down frequency
defined by (11.22). By moving the stops apart when
excitation frequency is constant, it is possible to
obtain amplitude pulling. It is interesting to note
o A f /J 1 that in contrast to forced vibration, amplitude
pulling for parametric vibration can not be limited
Fig. 11.3
by the dissipative factors considered. In actual
systems limitation occurs due to the fact that at
large amplitudes additional nonlinear factors occur and, moreover, the loading of
the vibratory source of energy leads to a decrease in the level of modulation and,
consequently existence conditions (11.22) are infringed.
In the frequency range AB. (Fig. 11.3), when the relationship
0 2 _~Jh204 _b 20)2 «0))2 <0 2+~Jh204 _b20)2 (11.49)

2 2 2 '
:/1
obtained using (11.21) when k == 0, is satisfied,
h"0-5 ll/fI - {}Qf 11-1
the system's state of equilibrium is unstable, and 0.'' ' I
the system develops into a regime with impacts.
This instability domain for a linear parametric , I~
system is disposed between the stops,
h I
consequently, even taking into account the loss
of energy during collisions, vibro-impact regimes
rr/ji
J I
I
·L
will always exist in this frequency range. With AI" 2
the help of hard excitation at the pulling a)
J(fnl
1a,.1"'1
frequency, situated between point Band the
d<O
break-down frequency, the behaviour of systems
can change from stable equilibrium to vibro-
impact.
Let us now look at the case of single-sided
Iimited motion (Fig. 11.1, b), using the
express ions for the linearisation coefficients in 2
the form (9.18). When ~ > 0 the nonlinearity is b)
hard, so regimes corresponding to the positive

sign in (11.21) are stable; when ~ < 0 the
nonlinearity becomes soft and the sign for stable
regimes changes to negative. Substituting the known value of k in (11.21), we
obtain when ß == 0
a
x
=_2~[(~)2
20
-3± h2 _~(~)2l-1
0 220
(11.50)
Fig. 11.4 shows amplitude-frequency characteristics for non-symmetrical systems

according to formula (11.50), corresponding to stops arranged with a gap (Fig.
11.4, a) and with interference (Fig. 11.4, b). Only the branches for stable regimes
are shown.
For systems with a gap, the same zone of unstable equilibrium occurs (11.49)
close to the excitation frequency, at almost double the natural frequency. With
interference the equilibrium state is always stable and the shift to a vibration
regime can happen only with additional excitation in the frequency range found
from the relationship
( 2~ ) 2;::: 3 +"21 h2 -
4b2 (
02 2~
)2 (11.51)
In both cases the vibration of the system can be pulled to infinite amplitude.
When the system is arranged with a gap this corresponds to the smooth increase of
the excitation frequency, while with interference, with decreasing frequency. For
example, the latter can lead to dangerous vibration of systems during run-down
when force-locked components are linked. According to (11.50) infinite
oscillation of systems is observed as the frequency nears the value given by
(2~)2 =3±~ h2 - ~: (2~)2 . (11.52)
Here the pOSItIVe sign corresponds to resonant frequency of a system with

interference and the negative sign to that with a
gap. As the system passes through the re sonant
frequency (11.52) the vibration breaks down and
dies away completely. Note that in the examples
studied damping does not limit the maximum
value of amplitude. However, according to (11.22)
when damping is increased, parametrie effects are
eliminated.
The resonance amplitude is not limited as is
!1ft} -
the increase of interference ~ (see (11.50». The Fig.l1.5
latter does not even affect the re sonant frequency.
Hence increasing the force locking mechanism
(Fig. 11.1, b) leads to a growth in the value needed for the starting impulse, but
does not eliminate the possibility of parametric resonance. The situation shown is
particularly dangerous, since it can not be discovered even during machine testing.
To estimate how the vibrations grow for regimes with pulling while
maintaining resonance conditions (11.43), when we substitute (9.18) in (11.44),
we find
- hO
S = ----;::=== (11.53)
4~3-2~/ax .
This means that for a stable system with gap (~> 0) , as the excitation frequency
increases, index S decreases from the corresponding value for a linear system to
the value
(11.54)
almost twice as small. During run-down of a system with interference (~< 0) an
increase in ~ occurs from a small value to a value given by (11.54). This same
value characterises the rate of parametric resonance when (~= 0). In the latter
case the system can not develop stable steady-state regimes.
5. The schematic diagram of Fig. 1 1.2 represents a wide class of systems
which can be studied in a similar way to the above presentation. To conclude the
description of methods to study periodic motion of non-autonomous systems, we
shall use as an example the problems associated with the simultaneous effects of
extern al and parametric excitation [44]. Such excitation is exhibited by the
simplest form of mechanism with force locking, fixed on the base (Fig. 11.5),
which moves according to the law y(t) = Yo cos( cot + <p) with a non-zero angle \jfo
between the axis ofthe bar and the direction of the vibration [117].
Keeping the same assumptions of mechanism properties and parameter
designation as in section 1, we arrive, taking x = \jfL, at equation (11.3).
Substituting (11.2) in (11.3), we find using (11.4)
(S2 + bs + n 2)x = hLn 2 cos(CO{ + <p )sin \jf - <D(x, sx) . (11.55)
Expanding sin\jf as a Taylor series near the reference angle \jfo' and limiting the
expansion to linear terms we find
sin \jf ", sin \jf 0 + (\jf - \jf 0) cos \jf 0 . (11.56)
Substituting (11.56) into (11.55), we obtain
(S2 + bs + n 2)x = hft cos(cot + <p) + hd COS(ffit + <p) - <D(x, sx) , (11.57)
where j = n 2 cos\jf o ; d = n 2 (Lsin \jf o - ~COS\jfo); ~ = \jfoL. The first term on
the right hand side of equation (11 . 57) represents a parametric excitation, and the
second term extern al excitation. It can be seen that as \jfo increases the external
excitation grows but the parametric excitation decreases. Equation (11.57)
corresponds to the schematic diagram of Fig.
11 .6. It differs from the arrangement of Fig. 11.2
by the addition of extern al excitation
P(t) = hd cos( cot + <p) and the introduction of
coefficientjin the parametric excitation.
An interesting feature of this system is that
when it is instal1ed with pre-arranged interference
(\jf 0 < 0) due to accidental break away of the
mass from the stop, stable equilibrium can be
violated and an intensive vibro-impact regime is
established. In this case the amplitude of the
exciting force can be below the level of force due
to the initial strain ofthe spring.
Let us find the existence conditions for the specified regimes. The joint effects
of parametric and external excitation produce the main dynamic effects on the
frequency of the principle and secondary parametric resonances. This occurs
because of the development of a strong "damping decreasing" parametric effect at
these frequencies causing favourable conditions for the external force to produce
excitation corresponding to sub-periodic resonance of ratio 2 and fundamental
resonance. A method to find an approximate solution with frequency ro/2 was

developed above and recurs in this case in its principle feature, hence we will
continue by studying vibration at frequency ro. In addition, with external and
parametrie vibration interaction, resonance at frequency ro is more vigorous.
An approximate periodic solution of (11.57) will be sought in the form
x(t) '" m x + a x cosrot = m x + XO (t) . (11.58)
The action (1l.58) appears at entry to the multiplier and nonlinear blocks. Let
us establish the harmonie linearisation chara.::teristics of these blocks. In the first
case we find directIy
F j (t) = (my + a x cosrot)hf cos(ro( + <p) =
(1l.59)
= Yz hfa, cos<p + hfm, cos( rot + <p) + Yz hfa, cos(2ro( + <p).
Discarding higher harmonics in (11.59) by taking account of the filtering effects of
the system, we can write the transformation of the harmonie process in the
multiplier using a complex gain factor
FJ(t) '" Yz hfa x cos<p + kJx o(t). (1l.60)
where
hfmx kj .
= --exp J<p. (1l.61)
ax
The nonlinear block will be linearised, as usual, using formula (11.13).
Using (1l.58), the external excitation P(t) is transformed to the form
P(t) = hd(cosro( cos<p - sin rotsin <p) = hd (cos<p + sin<p s)X O (t) . (1l.62)
ax ro
Substituting express ions (1l.60), (11.13), (1l.62) in the right hand side of (11.57)
and separating the constant and centred components we find
rim x +m F - Yzhfa x cos<p = 0, (1l.63)
[
S2 + (b + ß - hd sin<pJs + 0 2 + k- k j - hd]x o (I) = '?O(s)x o(t) = o. (1l.64)
axro ax
The existence conditions for periodic regimes of type (11.58) of the Iinearised
system (1l.64) are the presence of roots s = ±jro in the characteristic equation.
Substituting s = jro and setting the real and imaginary parts to zero we obtain,
using (1l.61)
(b + ß) ro- h( d + m J). 0
sm<p = , (1l.65)
ax
(1l.66)
Eliminating the trigonometrie functions from (11.65), (11.66), we find

a x =hld+mJI[(ro~ _ro 2)2 +ro 2(b+ß)2r J / 2 , (11.67)
where roo = ~02 + k . Comparing (1l.67) with the similar expression (9.13) for
forced vibration, it is easy to establish the additional effects on the amplitude due
to parametric excitation. From (11.67) it follows that vibro-impact regimes exist

for any small values of h, and, as a consequence, when the conditions provide a
"force locking" mechanism
I~I > hL sin \jf 0 • (11.68)
Obviously, such a regime can only result with hard excitation.
Let us re-write (11.63) with the help of(11.66), (11.67) in the following form:
_ r. 2 f(oo~ -oo2)a;
mr - -~4 m +
. x 2(d+mJ)
(11.69)
=_Q2 m + h2f(d+mJ)(oo~-oo2)
x 2[(oo~_oo2)2+oo2(b+ß)2]
Equations (11.67), (11.69) make it possible to define the values mx and ax using
the equations for the linearisation coefficients mF and k. Subsequently, from the
relationship
q> = arctan[oo(b + ß)/(oo~ - (0 2 )], (11.70)
obtained from (11.65), (11.66), the phase ofthe process can be found.
We will use the method described in § 10 to study the stability of the periodic
solutions. Applying condition (10.2) to the characteristic complex 9/Xjoo) for the
linearised system (11.64), we find, using (11.61)
~ [b + ß - h( d + ~ J) sin q>] >0. (11.71)

dax axOO-
Clx=Cl x
We will assume that slowly chan ging functions in x (t),a, (t) and (j5(t) are
connected during the disturbed motion by equations (11.63), (11.66).
Differentiating (11.71), we find
hftix sin (j5 din, + h(d + inJ)(ax cos(j5 d(j5 - sin(j5)

dß da, da x
> O. (11.72)
da,
~2
G<OO
The derivative d(j5 / da x is found by differentiating (11.66) as an implicit function

(j5 ofax . Using (11.67) we find
dß 2 2 dk
00-(00 0 -00 )--oo(b+ß)
d(j5 da, dax
(11.73)
(oo~ -oo2)+oo(b+ß)2
To simplify the relationship we will put ß == O. Substituting expression (11.73) in

(11. 72) we obtain after transformation and using the steady-state values sinq>, cosq>
and h(d + mJ) found from equations (11.65) to (11.67),
When b == 0 condition (11.74) can be separated into two factors
[(wo2-00 2) (2
Wo -00 2+ax--hf--
- dk
Mx
diii
M
x )]
_
>0. (11.75)
x llx = lI x
Comparing (1l.75) with the similar equation (10.11) when b + ß == 0, obtained for
forced vibration, we see that the additional effects of parametric excitation led to
the appearance in the condition of a term proportional to the value of modulation.
In the resulting expressions, describing the conditions for existence and
stability of periodic solutions, linearisation coefficient k exists due to the
dependence of the natural frequency Wo on amplitude ax of the fundamental.
Noting that the motion considered has aresonant character and consequently
appears elose to the natural frequency of the conservative system, for the
relationship wo(a), it is possible to apply the exact equations for the backbone
curve of the fundamental. Expanding the law of motion for a conservative system
(2.6) found by the "stitching" method, as a Fourier series, we find for the
coefficients the two first terms ofthe series
L'lw o ren
m x =--tan- (11.76)
tHl9 ren Wo
Il-g.j P'q- - {/{)7 2L'lw on ren
6/fl - tHlf a =- tan- (11.77)
x 1{(w~ _ (0 2 ) 00
(}UD
Obviously, we could use the general equation for
the harmonic linearisation coefficients obtained in
Q{}J §7 to obtain a solution for the problem considered.
We simply wanted to demonstrate here yet another
method to caIculate the linearisation coefficients,
based on a fuller description of the particular
problem. According to (11.67), (11.70) the
Fig. 11 7 consideration of energy losses during impact does
not provide any qualitative peculiarities, hence we estimate these los ses with the
simplified relationship (7.61). Greater accuracy can be obtained by using in (7.57)
the value of the pre-impact velocity x ,found for free vibration of a conservative
system.
Using express ions (11.76), (11.77) allows us to solve equation (11.67)
independently and to construct amplitude-frequency characteristics for the system.
Fig. 11.7 shows examples of such construction for systems with interference
(L'l < 0). Since a x ~ 0 , then in this case according to (11.77) Wo ~ 2n .
Substituting in (11.77) 00 = Wo ,we find the expression for the energy boundary of
vibration using (11.76)
§ ll Parametric vibration 163
a = h Id + !100 0! tan n:ol (11.78)

x oo(h + ß) n:0 00 0 '
The intersection of the energy boundary / with the backbone curve 2 found from
expression (11.77) gives the highest point of the resonance curve. The remaining
points (curve 3) are conveniently obtained as described in §9, section 1,
constructed graphically as the intersections of the family of curves ax(ooo),
produced from (11.67) using (11.76), (7.61) for various values 00 = const and from
the backbone curves.
To investigate stability according to condition (11.75) we find from (11.76)
diii x = !1dk I dax (tan n:0 _ n:0 /cos 2 n:0) =
dax 2n:oo oO 00 0 00 0 00 0
(11.79)
= !1dk I dax (sin 2n:0 _ 2n:0) < 0
4n:oooOcos 2 (n:O/oo o) 00 0 00 0
:<t
.
since the sign of !1 coincides with the sign of .-fM.,., , ... 1/11-0111
dk I da x ' but the factor in curved brackets is of
negative sign. Thus, in the example considered the 0''1 ///
two final in the second factor of condition (11.75) //
take the same sign. Using an argument similar to 0·2 ~~;';';;;;/

that explained in § 10, it can be shown that in Fig.
11.7 these correspond to stable regimes, Iying on o~--~--~--~·~-
115 f
--
fJ ~#
branch 3 of the curve, above the backbone line for
the system. The dotted line shows unstable Fig. 118
solutions.
From Fig. 11.7 it can be seen that as a result of hard excitation, systems can
shift from an equilibrium condition to regimes of periodic vibration with impact.
Increasing interference also causes an increase in the pushing impulse and does
not eliminate the possibility of a vibro-impact regime. Fig. 11.8 shows the
relationship between the amplitude of steady vibration a x and the value of the
initial angular displacement \jIo' The maximum amplitude occurs for forced
vibration (\jI 0 = n:/2) .
6. To demonstrate some ofthe effects described an experimental arrangement was constructed
as shown in Fig. 119. The beam J, fixed in the support 2, is installed on base 3 of a horizontal
vibration test-bed, and is subjected to vibration in the direction of the arrows shown. At the free end of
the be am a ball 4 is firmly attached, and its motion is Iimited by one or two stops, 5.
The motion of the ball was measured by a strain-gauge 6 fixed on the beam and whose signal was
recorded on an oscillograph; the vibration of the table was recorded by a similar method using an
mductance transducer. The instant when the ball and the stop
touched was recorded by a contact.
With lJIo = 0 an oscillogram was recorded of the
fundamental parametric resonance of a system with a natural
linear frequency 0/21t = 10Hz. Fig. 11.10, a shows the
oscillation of the system (curve J) without stops at a fixed
excitation frequency w/21t = 20Hz. Curve 2 in the oscillogram
Fig.l1.9 corresponds to the vibration ve10city of the table when
i
- - /sec -
~~ ~~N~A~M ~~ ~; I ~l\~~~~M~t· ~\1

., . ,2
.'. .-
vv,.·oNI/Wff'N~~.V'JVWVVV~ 'VV~,~~, I'ij~: Il~Jn~n~ I!
f ' . .. :
I~ '!' ':" fNNI''''''''' 'MW'W/M

1 l ,f •
\it I \ I \1lt.l fin li

l " .
.. . ''', ......... . - . .~ l: .... ~. __
,b.WtNNNHJI WtNfNNINNNNN.'NN'i1NNNffl t 1 I
j A,j I
• I I.
(n) (ii>
Flg. 11.10
Ya '" IO- J m , at a frequency of 50Hz. Noting that L '" 7.1 x IO- J m , we find from (114) that h = 0.56.
Substituting the value of h in (/144) we find when k '" 0, ~ = hD./4 '" 0.88/ S . Direct measurement of
n)
b)
Fig. 11.11
the amplitude in Fig. 11.10, a at intervals of I second gives a value ~ '" 0.831/ S .
Fig. 11.10, b shows a steady-state vibro-impact process for double-sided alternate collisions
Fig. 11.12
against the stop when L'. = 10- 2 m (The instant of impact corresponds to the peaks in curves 3 and 4.)
The process arises due to oscillation of the system from a linear instability domain (Fig. 11.1 I, al, or
with the help of additional push (Fig. 11.11, b)
The two fragments in Fig. 11.12 represent the passing of an instability domain as the frequency of
excitation is increased and subsequently (by time expansion) decreased. The effect of pulling is c1early
shown and the exponential character of the amplitude increase in a linear domain. After breakdown of
the vibro-impact regime, a beat is observed caused by the c10seness of the natural frequency of the
system.
§ 12 Vibration of se1f-sustaining systems 165
Fig. lU3, a shows the process of establishing a vibro-impact regime for a single-sided stop with
,,)
gap ~ = 10- 2 m . Fig. IU3, b shows the shift to a similar regime assisted by an additional push.
Finally, Fig.lU4 shows the effects considered in section 5 of the loss of stability of the system with
pre-arranged interference as the result of an initial push. Here curve 3 represents the position of the ball
without
Fig. 11.l4
interference The case described corresponds to point A ofthe amplitude characteristics in Fig. 11.7.
§ 12. Vibration of self-sustaining systems

I, Let us look at solution methods and the basic principles of self-sustaining
vibration of systems with collisions. A necessary condition for steady-state
vibration in self-sustaining systems is the existence of a source of energy with
which the vibrating system interacts, as the result of which compensating
dissipative losses arises in it and vibration at the established level is maintained.
Considering vibrating systems with stationary stops, we shall deal with various
dynamic models of vibration excitation. The simplest approach is to introduce a
special force, nonlinearly related to the velocity E(X), into the differential
equation of the system. In this case, the equation for a vibro-impact system will
contain not less than two nonlinear terms, as a result of which it is not only
effective to use approximate methods, but frequently the only possible way.
A more general model of a self-sustaining system can be made by using a
description for the source of the excitation force which allows us to find the strain
due to the loading of the source by the vibrating system. This extension of the
dynamic model for the system leads to the introduction of additional equations,
which describe the condition of the energy source, and consequently increasing
the basic degree of the system of differential equations. A feature of the system of
equations described is the existence of nonlinear terms

o z which represent the interaction of vibrating systems with
--+,--_. the excitation source, usuaIly, in the form of a product of
QZ ~A ..
~ ~ functions of generalised coordinates which determine the
- state of the sub-system described.
(+~ 6 ( :+"\
' To study these systems we will examine the effects of
V Y resonance types (autoresonance) in self-sustaining systems
which exist near the natural frequencies of the impacting
Fig. \2 .\
osciIlator, assuming that the latter are significantly
different from the natural frequencies ofthe energy source.
As a result the equations for the source serve principally to define the level of
excitation of the vibrator, which then brings us conveniently to the study of the
process described in the present chapter.
The cIassical example of a mechanism of vibration excitation in self-sustaining
systems is represented by frictional interaction, caused by various types bodies
sliding relative to one another (Fig. 12.1).
Depending on the character of the contact, the
interaction force B(i) takes various functional
forms, however the principal feature is the
existence of a range of velocities over which the
force linking the rubbing surfaces decreases as the
velocity of sliding increases. This occurs, for
example, due to the appearance of additional Fig. \2.2
adhesion with relatively "dry" surfaces (static
friction) or with low sliding velocity between greased surfaces [134].
For vibro-impact systems with one degree of freedom the operator equation for
self-sustained vibration in similar cases can be written as:
(12.1)
where (aB / ax) x~o > 0 . For sliding of greased surfaces this is usually taken as
B(X) = hX - bosigni, (12.2)
where hand bo are given positive values.
Equation (12.1) can be used to describe other mechanisms for self-sustaining
vibration with impacts. For example, in the construction ofmachines with electro-,
pneumatic, and hydro-impact effects to obtain greater effectiveness of the
excitation one tries to produce
B(i) = Usignx, (12.3)
where U is the maximum gain, developed by the energy source. Excitation (12.3)
guarantees the maximum contribution of the energy source to the vibrating system
(see Chapter 5). Similar excitation can be used when modelling electromagnetic
contact breakers (Fig. 12.2), for the construction of vibrating relays, beils, etc.
Finally, for several types of gas-dynamic processes, self-sustaining vibration can
be produced by a force like
B(X) = H(x-ai 3 ) , (12.4)
where H, a are positive quantities.
§ 12 Vibration of self-sustaining systems 167
Fig. 12.3 shows a schematic diagram for the transformation of signals

corresponding to equation (\2.1). Existing in the linear part of the system
r
L(s) = (S2 + bs = n 2 1 a filtering effect
gives the basis for carrying out harmonic
linearisation of all nonlinear elements of the
system. It is preferable to use the identical
zrt) linearisation method for all nonlinearities,
I-....,....--I'---t--~ however it can be expedient to use a
'--_....J
different method, when the impact pair
characteristics are Iinearised by force, but
the vibration exciter characteristics B(x) by
displacement. In this case it is necessary to
add to the resulting expression conditions
connecting the amplitude of the contact
Fig.12J
force and displacement a F = ka x •
As in previous studies, the solution x(t) of equation (\2.1) is searched for
approximately in the form (9.2), and replacing the force characteristics for the
impact pair <l> /, (x, sx) and <l> D (x, sx) with the linearisation expression (9.5). Also
applying the harmonic linearisation function B(sx) using the form
B(sx) '" ASX, (\2.5)
where A is the harmonic Iinearisation coefficient which depends on ax and m in the
general case, we obtain from (\ 2.1)
[S 2 +(b+ß-A)S+n 2 +k]xo +m F +n 2m x =0 (xo =x-m x )' (\2.6)
Separating the constant and centred components in (12.6), we obtain the equations
for self-sustaining vibration in the standardised form studied earlier (see, for
example (9.7), (9.8) or (11.15), (11.16»
m F +n 2 m x =0, (\2.7)
[S2 +(b+ß-A)S+n 2 +k]xo =O. (\2.8)
The relationship (\2.7) defines the connection m/a), which after substitution into
equation (12.8) makes the latter independent.
Existence conditions for steady-state vibration at frequency m appear as a pair
of imaginary roots to the characteristic equation ofthe system
.r?t1(s)=s2+(b+ß-A)s+n 2 +k=0. (12.9)
When s = jm we obtain the characteristic complex system in the form
~(jm)=n2 +k-m 2 + j(b+ß-A). (12.10)
Setting the real and imaginary parts of (12.10) to zero we find
m=Jn 2 +k, (12.11)
A=b+ß· (12.12)
Thus, according to (12.11) the connection between frequency and the
amplitude of steady-state vibration is found from the equations for the backbone
curves of the system, i.e. steady-state vibration in the system considered can be
treated as free vibration supported by excitation forces. Condition (12.12)

represents the equation for the energy boundary of vibration and defines the level
of free vibration at which an energy balance is achieved between the excitation
force and damping. Examining (12.11), (12.12) as implicit expressions ax(m), it
is convenient to obtain a graphical solution for this system, by locating the
intersections between these curves on the (00, a x) plane.
To study stability of self-sustaining vibration found by harmonie linearisation,
we can apply different criteria, based on the behaviour of the system considered
near the steady-state. In particular, it is convenient to make use of condition
(10.2). In the case considered, according to (12.10), the imaginary part of the
characteristic complex takes the form
8(a x ,m)=b+ß-A. (12.13)
Applying condition (10.2) to (12.13), we obtain
( ~-~J
dax dax - U"(=llx
>0
.
(12.14)
Condition (12.14) has a simple physical meaning, implying that as the amplitude
of stable vibration increases then the dissipative forces should perform more work
than the excitation fore es but the reverse is true with decreasing amplitude. This is
shown in many works devoted to the study of self-sustaining vibration.
Together with a study of the stability of self-sustaining vibration, it is also
necessary to examine stability of equilibrium when studying autonomous systems,
since solutions of the type x = mx = const also satisfy equation (12.1). This
problem can be dealt with by considering the linear equations with constant
coefficients, obtained from (12.1) by replacing all nonlinear functions by their
Taylor's series expansions to linear accuracy. For these, stability conditions are
generated using the Routh-Hurwitz criterium or its equivalent.
2. As an example, let us look at vibro-impact regimes of systems as
described by expressions (12.1), (12.4). The harmonie linearisation coefficient A
for function (12.4) takes the following form
A = H(l- Xaa;m 2 ). (12.15)
Substituting (12.15) in (12.12) and assuming ß=pm (see (7.61), (7.62)), we
obtain the following equation for the energy boundaries of vibration
ax = ~
00
_1 (1- b + pm) .
3a H
(12.16)
From this we find the existence condition für steady-state vibration in the form
H > b+pm. (12.17)
With the stops arranged to have a gap, satisfying condition (12.17) leads to
unstable equilibrium for the system and soft excitation of vibration. With
interference, self-sustaining vibration can only be established through an initial
push, which allows it to overcome the initial force ofthe spring Q 2L1 ..
Superposing express ions (12.16) (curve 1 in Fig. 12.4) on the graph of the
backbone curve 2 for the system found from equation (12.11), we find the values
of amplitude and frequency for steady-state vibration. Fig. 12.4, a to c correspond

to different variations of single-sided stops, and Fig. 12.4, d to symmetrical
double-sided stops. From Fig. 12.4, a it can be seen that with a sufficiently large
gap, vibration can occur without touching the stops. The vibration frequency in this
case is w = Q . Substituting this frequency into (12.16), we find, for example, that
the following condition for excitation of vibration with impacts occurs for the case
ofsymmetrically arranged stops (ax = L1) (Fig. 12.4, d)
(12.18)
According to Fig. 12.4, b, when L1 =0 the vibro-impact regime is isochronous.
4>0 d~O
d. Q'c
f
.d
2
0 . ~/Q 0
u)
f via
b)
4<0
I1z dz
0
~
f
c)
I
6110 0
1
d)
61lfl
Fig.12A
When L1 < 0 (Fig. 12.4, c) the system is in equilibrium, and, as can be easily
found, in contrast to other variants of stop arrangement, is stable, so that shift into
a vibration regime in this case can only happen with the help of hard excitation.
Let us now look at the stability of vibration in the system considered.
Substituting (12.15) into (12.14) we find
YzcrHw 2 a x > O. (12.19)
Comparing express ions (12.19) and (12.17) shows that when cr > 0 the regime is
stable for all existence domains.
Let us now look at the case when self-sustaining vibration is excited in the
system described by equation (12.1) with the function B(i) taking the form
(12.3). The harmonie linearisation coefficient for this function is
A=~. (12.20)
Substituting (12.20) into (12.12) we find the energy boundary for vibration
4U
(12.21)
Superposing expression (12.21) on the backbone curves we find the amplitude and
frequency for self-sustaining vibration. For symmetrical double-sided stops,
assuming a x = ~ in (12.21), we obtain the following expression for the vibration
frequency:
0) = ~[
2p
1+ 16pU
nb2~
-1], (12.22)
where the formula is only valid in the natural frequency range of the oscillator
(0);::: 0). Using this we find from (12.22) the existence condition for a vibro-
impact regime
U ;::: y; n~O( b + PO) . (12.23)
Substituting (12.20) in (12.14) we find the conditions for stable self-sustaining
vibration U > O. Thus vibration is stable for all existence domains in the given
system.
3. Let us turn to the study of self-sustaining vibration of vibro-impact
systems with a non-ideal energy source. The work of many authors has been
devoted to the study of interaction of elastic systems with the energy source
causing vibration [61, 62, 101, 133,216,227 and others]. The traditional method
for analytic study of similar problems is the asymptotic methods of nonlinear
mechanics. We shall extend the harmonic linearisation method to the class of
problem shown, tuming our attention specifically to the behaviour of vibro-
impacts systems with "Iimited" excitation [25].
We shall start by looking at a system (Fig. 12.5), which consists of an electric
or hydraulic motor producing a variable torque L, driving a
shaft with moment of inertia I, wh ich excites vibration of an
impacting oscillator having mass M, through a crank of radius r
attached to the shaft, and a connecting-rod mechanism which is
an elastically sprung element of stiffness c. The damping in the
guide and other losses, will be represented by a viscous force
proportional to the absolute velocity ofthe striker by a factor n.
The zero position of the striker corresponds when the spring is
undeformed and the crank arm horizontal; from this position,
the displacement of the striker is limited by the gap
(interference) ~. An example of a similar system are shown in
Fig.4.2.
Fig.12.5
Measuring displacement of the vibrating mass as x(t) and
rotation angle of the shaft as \jI(t) and assuming that the length
of the crank is essentially less than that of the connecting rod,
we can write the differential equation in terms of these basic coordinates for a unit
oscillator mass and shaft moment of inertia, in the form
x + bx + 0 2 X + E (x, x) + D (x, x) = r0 2 sin \jI , (12.24)
(12.25)
where b = nl M; 0 2 = cl M
. The seeond term on the right hand side of (12.25)
represents the load whieh the vibrating system transmits to the motor shaft.
To find the torque L we shall use the equation for the dynamic eharaeteristics
of the motor [215]
(12.26)
Here t = T;,/: TM is the time eonstant of the motor (eleetromagnetie for an
eleetrie motor and hydraulie for a hydraulic motor); Il is a eoeffieient of
proportionality (Il > 0); ffio is the idling velocity.
Assuming that rand L are proportional to the sm all parameter e: we establish
from (12.25) that
\jJ = ffit + eV(t) (12.27)
where ffi = eonst is the unknown average angular veloeity of the erank, while v(t)
is an oseillatory funetion of time. When L is small we may assume a small
variation in the angular veloeity ofthe motor by eomparison to its average value.
Substituting (12.27) in equations (12.24), (12.25) and expanding their right
hand sides as series in terms of the sm all parameter e, and limiting the expansion
to linear terms, we obtain:
x+bx + 02 X +<PE(x,x) +<PD(x,x) = er0 2 sinffit, (12.28)
(12.29)
Henee, (12.28) ean be eonsidered independently. It represents the vibration of an
impaeting oseillator under the influenee of harmonie exeitation as represented by
the right hand side. Thus, we ean use the periodie solution for the problem
indieated, found by using either the approximation deseribed above or by using an
exaet method (see §2, §9). An important feature of sueh a solution in this ease is
that it depends on the unknown parameter ffi, whieh must be found subsequently
from (12.26), (12.29) using this solution.
To obtain the final expression, we will take advantage of the simplified
approximate solution of (9.1), similar to (12.24), found in §9 seetion 1 using
harmonie linearisation, bearing in mind that here a p = er0 2, m p '" 0 , while phase
qJ now represents the time lag ofthe approximate solution
x(t) ~ m x + a x sin( ffit - q» (12.30)

relative to the exeitation. As a result, similar to (9.11) to (9.13) we find
ax = er0 2 [(02 +k-ffi2)+(b+ß)2ü) 2 r 1/ 2 ,
(12.31)
. (b+ß)ffia x
smq> = 2' (12.32)
erO
(0 2 +k-ffi 2 )ffia x
eosq> = 2 ' (12.33)
erO
while for the linearisation eoeffieients m,., k and ß we will use their forms (9.17),
(9.18), (7.61). In this ease the eonneetion between mx and ax is defined by (7.63),
but expressions oftype (9.7) for the eonstant eomponents
m F +Q2 mx =0 (12.34)
coincide with (9.17) in view of (7.63).
Resonant vibration ofthe oscillator, which will be of interest to us later, occurs
for sm all values of the detuning Q2 + k - ü} and sm all coefficients of dissipation
b + ß, since no matter how sm all the excitation, amplitude ax at resonance (see
(12.31)) is not a small value. The value of mx follows the same order.
Let us find the corresponding periodic solution for the system of equations
(12.26), (12.29). Using (12.30) the equations described can be considered as the
equation of a linear non-autononmous system subjected to the effects of a constant
moment and biharmonic excitation at frequencies 00 and 200. Consequently,
substituting (12.30) into (12.29) and transforming the resulting trigonometrie
functions on the right hand side, we find
\V - EL = ErQ2 [- a; sin<p + m, cos oot + a; sin(2oot - <P)]. (12.35)
Searching for a solution to the system of equations (12.26), (12.35) in the form
\jJ = 00 + a: sin( oot - SI) + a~I sin(2oot - S2)' (12.36)
L = m I. - a LI sin(oot - G I ) - a LII sin(2oo{ - G 2 ),
we find using harmonic balance, neglecting small values of a: / co and a2 /00 as
explained above in the description of (12.27) since after multiplying by E they
become second order smaJl quantities:
m l. 1,....2 a sm<p
--r~~
. =
0 (12.37)
x
2
I ErQ2 II ErQ2
aro = ---mx' Gm =- - ax ' (12.38)
00 400
Q2+k_oo 2
SI = 11:, tan S2 = - - - - - - (12.39)
(h + ß)oo
ETWQ2 II ETWQ2
mx,a/ ax ' (12.40)
oo.JT 2 + 00 2 •4oo.JT 2 + 400 2
00 Tsin S2 + 200 cos S2
tanG I = - , tanG 2 = . (12.41)
r TCOSS 2 -2oosmS z
Note that harmonic vibration of the shaft with frequency co occurs only as a result
of the non-symmetry of the law of motion of the oscillator ( m x i= 0), arising due
to single-sided collisions, and proportional to this non-symmetry (drift), while at
the same time harmonic double frequency is formed in accordance with the law
for change of amplitude of vibration of the oscillator ax '
Expression (12.37) establishes the relationship between the average torques
and load on the motor shaft, caused by the vibrating system. This defines the
energy boundary of vibration. Substituting the value of sin<p from (12.32) into
(12.37), we obtain eventually
l1(co o - co) - Yz (h + ß)wa; = O. (12.42)
The components of the torque proportional to band ß, determine the load

produced by energy dissipation in the vibrating system and by the work process
causing collisions. The power control ofthe motor leads to the shift ofidling value
of the angular velocity (00 .
The system of finite equations (12.31), (12.42) is similar to the corresponding
ones (12.11), (12.12) considered in section 1, and enables us to find the amplitude
ax and frequency (0 for self-sustaining vibration of the impact oscillator, from
which all the remaining parameters for motion of the system can the found.
Simi1ar equations were obtained in [132] using the asymptotic method.
The character of possible values for (0 in relation to the arrangement of the
stops and motor power, proportional to the value of (00, is conveniently found
graphically from the solution of equation (12.42) (Fig. 12.6) by substituting
(12.31) into it and using the equations for the harmonic linearisation coefficients
(9.18), (7.61). In Fig. 12.6
m, = Yz(b+ß)(Oa; (12.43)
is the average moment of the force in the motor, caused by vibration of the
oscillator. The intersection of the lines for moments mlro) and mlro) set the
required value of (0. The remaining unknown values can then be found from
formulas (12.31) to (12.33), (12.38) to (12.41). Note that for the given system
Ipo t,-o /Ilt .m, ~<Q
m~ , 111,
Fig. 12.6
when ~ < 0, a vibro-impact regime can occur only when an additional hard
excitation is supplied.
4. Let us look at the stability of the periodic solutions obtained in terms of
small disturbances. Representing the disturbance of motion of the oscillator in the
form
x(t) = iii x (t) + Ci, (t)sin[05(t)t - <p(t)] , (12.44)
we shall as before use the stability condition (10.2), which, when we apply this to
the imaginary parts of the characteristic complex of equation (12.28), gives,
similar to (10.5)
(12.45)
However, an important feature of the case considered is the fact that now
frequency must also be considered as a slowly changing function ffi(t). Hence,

assuming Ci, (t) is an independent function it is necessary to apply three arbitrary
conditions to its connections with the remaining slowly changing functions in
(12.44). Retaining the conditions (12.34), (12.33) for the connections between
in, (t) and Cix (t), <pet) and Ci, (l) we add expression (12.42) as the connection
between ffi(t) and Cix (l) .
Differentiating (12.45) using the express ions above, we find
[-
dß Er0 2 (~ . ~
+ ~2~2 ~ ~ d 0. (12.46)
da, w ax da, da, _
ux=a x
Differentiating (12.33) as an implicit function <p of Ci" we find
( - d<p ) = 1 (r.2 + k -w~2 +dk- a~

~l
aw
- 2 - w~~
a ) (12.47)
da x iix =lI x Er0 2 sin<p da,' dax X iix =lI x
As the result of differentiating (12.42) we find
j..lWOCix[~Cix
[daJ ~ l
aw da
+2(b+ß)])
x
(12.48)
,,co, - 2[" + Y, (b + ßlO; ]' _
ax=u x
To simplify the final relationship, we assume ß == 0, and we find from (12.46)

using (12.31), (12.32), (12.47), (12.48)
ob 2
b2w2_ j..lw wa x [b 2 _2(02+k_w 2)]+
(j..l+liba;) (12.49)
+(0 2 +k_w 2)[02 +k-w 2 +u.(dkldaJiix =lI,.]> 0.
Comparing (12.49) with the stability condition for forced vibration (10.9) we
can see that when energy dissipation in the system is neglected, they coincide;
consequently in the system considered, instability will occur only in the lower
branch of the double-valued domain of the amplitude-frequency characteristics
drawn using (12.31). In the general case the stability picture changes. Substituting
the value of k from (9.18) in (12.49), when m p == 0, we obtain eventually
b2ü:/- j..lw
°bwa x2 [b2_2(302+20~_W2)]+
2
(j..l+liba;) ax
(12.50)
The unstable solution found using condition (12.50) is drawn in Fig. 12.7 as a
dotted line. In addition, notice that when the stops are arranged with initial
interference, satisfying the stability conditions depends strongly on parameter J.l,
wh ich is defined by the "rigidity" of the static characteristic of the motor. As the
value of J.l changes from an infinitely large value to zero four distinctive domains
for the value of J.l are developed similar to [132]. These domains are represented in
the form of sectors in Fig. 12.7, c. In sector I, corresponding to steep
characteristics, the system behaves similarly to the analogous non-autonomous
system, excited by a periodic force with fixed frequency ro. The stability picture
closely resembles this condition, since the lower branch of m,Cro) remains unstable
in this case.
Another picture is observed when mL(ro) lies in sector 3. This case is also
11>0 /JoO
Iltlfl,
a) b) c)
Fig.12.7
shown in Fig. 12.7, c. Here an unstable solution appears on the fragment of the
upper branches of expression m,( ro).
In sector 2 condition (12.50) is satisfied by all steady-state solutions which
produces so me ambiguity. The experiment described in [132] for a similar
situation, showed the absence of stable periodic motion. Finally, in cases when
rn/(ro) lies in sector 4, calculation using condition (12.50) shows that the system is
near the boundaries of stable domains. According to [132], steady-state regimes
were not achieved experimentally in such situations.
Assuming that power control of the motor corresponds to a shift of mL(ro)
parallel to itself (because of a proportional change in roo), we find that, as we
increase the power, the applied moment m,(ro) and correspondingly, the angular
velocity ro will change in the direction shown in Fig. 12.7 by continuous arrows,
but when the power is decreased, the angular velocity changes in the direction
shown by dotted arrows. At the instant of jumps of the working point of
characteristic m L (w) there will be transient processes leading to the
corresponding final change in the value of w. Note that when L\ < 0, increasing
the motor power can sometimes lead to a decrease in the average angular velocity
w, because ofrapid growth in the vibration amplitude ofthe oscillator. As a result,
for any stop arrangement the velocity of the motor changes negligibly as the
amplitude increases during resonance, no matter how the power is controlIed, hut
changes significantly after the oscillator passes maximum amplitude (the
Sommerfeld effect).
Fig. 12.8, in agreement with the previous construction (compare to Fig. 12.7)
shows with full lines the branches of the amplitude-frequency characteristics for
the oscillator corresponding to run-up and slowing down of the motor. Fig. 12.9
shows similar branches for the amplitude-frequency eharaeteristics aceording to
(12.38), (12.40) for the fundamental harmonie component of vibration of the
angular velocity and moment for an electrie motor.
5. Using the methods outlined in § 10, seetion 3, we shall now look at the
.l1z d>V
c:
,
I
(
/
I
(
\
-
'\
".... -------
(J
IJ -O
- (J
0, 0:
11
11
, 1
,
1 I
~
- (J
11<0
- (J
11. fI:
(J
~ (J
Fig. 12.8
process by which steady-state vibration of the system eonsidered (Fig. 12.5) is

established. Rejecting the assumption of slowly changing function ax(t) in the
transfer process, we shall approximate it with expression (10. 17) and will look for
an approximate solution ofequation (12.28) in the form (12.30), (10.17), assuming
~(t),<p(t) and (ö(t) are slowly changing functions . In this situation the
linearisation coefficients (9.17), (9.18) (7.61) retain the same structure as in the
steady-state case.
Taking this into account, and substituting (12.30), (10.17) in the harmonically
linearised equation (12.28), we obtain similar to (10.28), (10.29)
P + (b + ß)~ + n 2 + k - 05 2 - crn~ cosq5 = o. (12.51)

ax
~ = ~(cr~2 sinq5 _ b _ ß). (12.52)

2 ffia x
For constant components expression (12.34) is retained. The angular velocity

05(t) is found from the static characteristic ofthe motor
(12.53)
This assumes a sufficiently slow rate of transient processes in the shaft of the
motor so that the Sommerfeld effect is observed. When ~ == 0, we obtain the
existence conditions for a steady-state regime (Zi x = a x = const) from (12.51),
(12.52), which coincide, of course, with (12.32), (12.33).
The establishment of vibration is defined by the dependence of ~ on Zix . To
find this, let us eliminate phase q5 from equation (12.51), (12.52). We find
~2 +(b+ß)~+n2 +k-05 2 -~(crn2 /Zi x ,)2 -(2~+b+ß)2052 =0. (12.54)
Equation (12.54) represents an algebraic equation of the fourth degree in ~. For

simplicity we shall assume for further analysis that b + ß == 0, and from equation
(12.54) we find eventually
11:' l1t, d-O, i\d<O

" \
.::' ~ \~
N:i!,
,J'\"
\
I ' .......
11
litt '\l ,,J

,~
C' ......- "
11/
11: I1t
Fig. 12.9
Let us find now the acceleration of the motor shaft during the transfer process.
Substituting (12.44) in (12.29) and bearing in mind (12.27) we obtain, averaged
over rapidly oscillating terms,
-dffi = E[ ~(Cl) 0 -
~
Cl) - -
1 rO 2 a~ sm
. ~]
q> . (12.56)
dt 2 x
From equation (12.56) using (12.52) we find when b + ß == 0
dffi
- = E[(
dt
~ Cl) 0 -) - - -
-Cl) ~05a;
E
1 (12.57)
or, substituting (10.17),

dffi= E[(
- ~ Cl)
0 - ) - - - - -] ,
-Cl) 05a x da x (12.58)
dt E dt
i.e. the average torque of the motor load caused by vibration of the system is
proportional to the product of speed of amplitude variation and maximum
vibration velocity in the period T = 211 / 05 .
Let us look at the case when the resonance correlation during the transfer
process
0 2 + k - 05 2 = 0 . (12.59)
is satisfied. Substituting (12.59) into (12.55) and expanding the radical as aseries
in terms of parameter E, we obtain, limiting the expansion to linear terms,
~ = Er0 2 /(2maJ. (12.60)
dax / dt = Er0 2 /(2m)
and consequently
(12.61)
Comparing (12.61) with (10.38), we can see that the transfer process in this
case evolves similarly to forced vibration. Using (12.60), equation (12.57) takes
the form
-dffi = E[ ~(Cl) 0 - -
Cl) - 1 rO 2-
- a ] . (12.62)
dt 2 x
,
Contrasting (12.62) with (12.56) we can see that the torque on the motor is a
1iEJ
maximum in the transfer process considered. Hence,
express ions (12.61), (12.62) correspond to the most
stressed transient process. f
6. Vibration is significantly influenced by the

QZ
characteristics of the excitation source in pneumatic
, Z .
impact systems. Let us look at a vibro-impact system
with single-sided pneumatic action (Fig. 12.10),
Fig. 12. \0
assuming that the connection between the working
chamber of the cylinder at pressure PI and the
atmosphere at press ure P 2 occurs through ports 1 or 2 depending on the motion of
the piston.
Taking the origin for the motion of the piston, x, from its position when the
spring is undeformed, we can write the equation of motion for a unit mass piston
in the form
i + bosigrrX + nZx + <D E(X,X) + <D f) (x,x) = -(p - pz)S . (12.63)
Here S is the cross-sectional area ofthe piston; n is the elasticity ofthe spring; bo
z
is the amplitude ofthe dry friction force; pis the pressure inside the cylinder.
An equation describing the change of pressure in the cylinder in relation to the
vibration of the piston can be found based on the equation for an ideal gas
pV = R o WT , (12.64)
where V is the volume of the working chamber of the cylinder; T is the absolute
temperature of the gas; W is the weight of gas; Ro is the specific gas constant.
For simplification later we shall assurne that expansion and compression ofthe
gas in the cylinder is isothermal. In [13] it is shown that with this assumption the
calculated time for change of pressure in the cylinder is increased by 10 to 20%
compared to reality. Differentiating equation (12.64) with respect to time, using
the assumption stated, we find
pV + pV = RoTW
Noting that V = (1- x)S , where I is the coordinate ofthe right end ofthe cylinder
(Fig. 12.10), we find V = -Sx . Denoting the gas flow rate W= GI - G z where GI'
Gz are the flow rates for entry to and exit from the cylinder we find
. I
p( - x) = -RoT (GI .
- G z) + px . (12.65)
S
Let us find the existence conditions for autoresonant tuning of pneumatic
vibro-impact systems. Such tuning assurnes significant vibration of the piston for
small variation of the pressure in the cylinder relative to the average value. As
result, compression and exhaust occur with large drops in pressure. This allows us
to assurne in both cases that the gas flow is e10se to being supercritical and to
describe the appropriate flow rate by the expressions [106]
GI = nSIP p G z = nSzp, (12.66)
where SI' Sz are the effective areas of the inlet and oudet ports, n"" 0.350/ JT .
Apart from the simplicity of calculating this expression, the assumption suggested
decreases the time for pressure change in the cylinder by comparison with the real
value and also compensates to so me measure for errors arising by accepting the
isothermal character of process.
The control of overlap for the corresponding ports will be characterised using
logic functions TJI(X,X), TJz(x,x) which are assumed zero for a e10sed port and
one for an open port. To accommodate this and bearing in mind (12.66), we can
write equation (12.65) in the following final form:
p(l- x) = n l [SI PI TJI (x, x) - SzPTJz (x, x)] + px (12.67)
where n l = RoTnl S .
Let us look for approximate solutions to the system of equations (12.63),
(12.67) in the form
x(t) "" m x + a x COSOO[, (12.68)
p( () ~ m" - a" cos( 00 ( + <p ), (12.69)

where 00 is the unknown frequency of self-sustaining vibration; <p is the phase lag
between vibration ofthe piston and pressure in the cylinder.
Carrying out harmonic linearisation ofthe dry friction force in (12.63), we find
bosigni: ~ ßIX (12.70)
where the harmonic linearisation coefficient ßI by analogy with (12.20) will be as
folIows:
ß1 = 4bo /(naxoo) (12.71)
Substituting (12.69), (12.70) into equation (12.63), we obtain
x + ßIX + OZ X + E (x, x) + f) (x, x) =
(12.72)
=-(m" - pz)S + apScos(OO( + <p),
i.e. we obtain onee again the equivalent equation for forced vibration, similar to
(9.1).
Carrying out harmonic linearisation of the impact force characteristic in the
form (9.5) we obtain, similar to (9. 11) to (9.13)
a x = apS[(OZ +k _ooz)z +(ßI +ß)zoozr l / z (12.73)
. (ß 1 + ß)ooa x (Oz + k - OOZ )a x
sm<p = ,cos<p = (12.74)
a"S a"S
where, for the present, we shall use the forms (9.17), (9.18), (7.61) for the
harmonic linearisation coefficients m" k and ß.
By analogy with (9. 7), we find from (12.72) for the constant terms
mF +Ozm, = -(mI' - pz)S. (12.75)
This expression also follows from (9.17), since now
m p = -(m" - pz)S (12.76)
and, moreover, condition (7.63) is satisfied. Comparing (12.73) to (\2.75) with
similar solutions (12.31) to (12. 34) for the problem above, we notice that in this
case, not only is the frequency of vibration 00 unknown, but also the value of m p
and an which represent the average value and amplitude of the excitation for an
impact oscillator. These values can be found using the gas-dynamic equation
(12.67). When we substitute the solution (12.68) into it, we obtain a linear non-
homogeneous equation of the first order relative in p with periodic coefficients
and periodic "extemal" excitation.
Under conditions of autoresonance, when the pressure is in phase with the
piston velocity, the term px ofthe equation considered prevents abrupt change of
pressure (the piston enlarges the working volume during inlet and decreases it
during expansion). As the piston vibration intensifies then its smoothing effects
become stronger. Similar influences on pressure are shown during operation by
the term I - x , representing time for the pressure drop to change. As vibration of
the piston increases, the average value of these coefficients increases (similar to
the time constant of the response), because m, < O. As a result, according to
(12.67) it is possible to ensure low vibration of the pressure in the chamber
relative to the average value when the inlet and exhaust ports have small area, no
matter how intensive the vibration of the piston. This shows that the assumptions
taken when introdueing this equation are valid.
The solution is eonveniently earried out in the following manner. Performing
harmonie linearisation of the logie funetions, we find
n IS I11I(x,x) "'m rh +ul(x-mx)+Alx,
(12.77)
n IS 2112(X,X) "'m 112 +u 2(x-m x )+A 2x,
where the linearisation eoeffieients depend on the parameters (12.68). Replaeing
the nonlinear terms in equations (12.62), (12.67) by their linearised
approximations (9.5), (12.76) we obtain, substituting (12.68), (12.69):
-axoo2eosoo(-ßlaXoosinoo(+122(mx +axeosoo()+m F +
(12.78)
+ ka x eosoo( - ßaxoosin oo( = -[m p - a p eos(oo( + <p) - P2 ]S,
-a poo sin( oo( + <p )(1- m x - a x eosoo() = PI (m 111 + ula x eosoo(-
- AI axOO sin oot) - [m p - a p eos( oo( + <p )](m 112 + u 2a x eos 00(- (12.79)
- A2 a xoosin oo( - axoosin 00(.
Differentiating (12.79) and eliminating p and P from (12.79) using the
expressions obtained and (12.78), we reeeive express ions whieh depend only on
the motion parameters ofthe piston. Using harmonie balance we ean find, eorreet
on the basis of the main harmonie eomponent
[m 1l2 (P2 S - 122mx ) - PI Sm 111 + Yz (ß I + ß)A 2 a;oo 2 +
+ Yzu 2a;(122 +k-00 2]
m F = --------------'--=---=----"---'------" (12.80)
m 1l2
2 +k)+U 2(P2 S - m F _122 m J-U IPI S
00 2 (12 _
m 11 _ _ _ _ _ _ _ _ _ _ _ _ ___
(12.81)
=_~2
m ll1 +(I-mx)(ßI +ß) ,

[(1- mJ(122 + k) + m 112 (ßI + ß) + (11,2 + 1)(p2S - m F - 122 mx )-
AIPISI]
o
00" = ------------------..:..::....;:........:..=---. -
(12.82)
I-mx
The system of finite equations (12.81), (12.82) is similar to the expressions
(12.11), (12.12): from (12.81) the eonneetion between amplitude and frequeney of
"natural" vibration of the piston in the system examined (using gas-dynamie
processes) is established, (12.82) defines the energy boundaries of vibration. As a
result the joint solution of equations (12.80) to (12.82), using the equations for the
harmonie linearisation eoeffieients, gives the frequeney of vibration 00, the
amplitude ofthe piston a x and the value of m F • Using these values it is possible
to find the proeess parameters a p , <p and m p (12.69), subsequently from (12.73)
to (12.75).
To eomplete the solution of this problem it is neeessary to settle the strueture
of the eontrol arrangement whieh defines the eyclie reversal of the inlet and
exhaust ports during motion of the piston. Here we ean reverse the problem so
that, given the requirements of the motion parameters, we can look for the
linearisation coefficients of the logic functions and from them synthesise the
control system. Chapter 5 will be devoted to the problems of structural synthesis
of self-sustaining vibro-impact systems. In particular, §20 will consider the
synthesis of auto-re sonant systems based on harmonie Iinearisation methods.
The strategy of port switching has some arbitrariness but does not always lead
to the establishment of autoresonance. Based on the simplified scheme described
in section 2, in which contro) was described by equation (12.3), it is possible, for
example, by analogy to choose the switching sequence for the ports, defining them
by the following logic functions:
Th(X,x) = l1(-x), 112(X,x) = l1(x), (12.83)
where l1(x) is a unit step function, i.e. during motion ofthe piston away from the
stops, only the inlet port is open, and only the outlet port during motion towards
the stop. For greater effectiveness of excitation, it follows of course, that we must
adjust the advance of switching to compensate for the inertia of the gas-dynamic
process, however for simplification we will not dweil on this point.
The harmonic linearisation coefficients for expression (12.77) take the
following form, using (12.83):
m = nlSI m = n lS 2 'AI = 2n lSI 'A 2 = _ 2n lS 2 ,
~I 2' ~2 2' na.oo na.oo (12.84)
== 0, u 2 == 0.
°
UI
7. Substituting u l == 0, u 2 == in (12.81), we obtain
00 2 = - - - - - - - - -
n 2 +k
(12.85)
I+(ßI +ß)(l-m.)lm~2
Noting that autoresonance regimes arise with relatively small values of the
dissipation coefficients, and, in the denominator of (12.85), neglecting values
which are small compared to unity, we obtain again expression (12.11). Hence,
even in the system studied, autoresonance occurs ne ar to the natural frequency of
the impacting oscillator. In practice, when conditions (12.11) are satisfied, then
according to (12.74) the excitation phase remains equal to n12.
From (12.80), when U z == 0 ,we obtain
mF = [ P2 - PI-- m~1 Js r'\2

-~~
ßI + ß 12 a.oo
m. +--1\. Z Z
. (12.86)
m~2 2m~2
Substituting (12.86) into (12.75) using the value of the linearisation coefficients
(12.84), we find
(12.87)
Expression (12.87) is similar in appearance to that considered above when

solving the previous equation (12.42) for the balance of average torques of the
shaft of an electrie motor.
Let us rewrite (12.87) in the form
(m p - PISI I S2)S = 2(ßI + ß)a,OO In. (12.88)
The right hand side of this expression defines the value of the average force in the
pneumatic system during vibration of the impact oscillator. As before, such a force
is proportional to the loss of energy in the vibrating system due to damping and
the performance of the working process, but also to the intensity of the vibration.
The left-hand side of the expression represents the average force generated by the
pneumatic system to compensate the force indicated.
In cases when the vibration load is negligibly smalI, we find from (12.87)
mp = PIS, / S2· (12.89)
The excitation at which condition (12.89) is satisfied will be called idling. Since
m p < p, the port areas must satisfy the condition
S2 > S, . (12.90)
As the load increases, m p also increases, and as a result the average force
developed by a pneumatic motor increases, and according to (12.88) tends in the
limit to the value
(12.91)
which by analogy with the electric motor, will be called stalling. It must be
understood that this value is only hypothetical, since by letting
m p approach to p, , the initial assumption offlow character is infringed, since the
approach of mp to PI indicates insufficient power in the pneumatic actuator to
develop autoresonance. According to (12.91), a radical method to increase the
intensity of actuation is to increase the inlet pressure p, . Limited opportunity can
also be achieved by decreasing the ratio of area of the ports SI / S2 . According to
(12.88) for large power reserves in the pneumatic system:
P,S,/S2 »2(ß, +ß)axO)/n,
the vibration load can be neglected, using (12.89) to calculate the average
pressure. In this case one talks of ideal sources of excitation.
For an actual excitation source with limited power, substituting the resonance
value found from (12.73) when satisfying condition (12. I I), for a x in (12.87), the
connection between m p and a p for autoresonance conditions is established
mp=p,SI/S2+2ap/n. (12.92)
In addition to the assumption already made, the flow must be supercritical
mp -a p >P2' mp+ap<p" (12.93)
and we obtain from (12.92) the following limits:
ap <2.74(p,SI/ S2 -P2)' a p <0.6IPI(I-S,/Sz} (12.94)
Given the required value of ap , it is possible to select suitable values of the
pressures p" P2 and the port areas SI' S2 from these inequalities.
Let us examine relationship (12.82), which defines the energy boundaries of
vibration. Assuming, using (12.85), 0)2 '" Q2 + k , we find
m~2 (ß, + ß) + (1."2 + 1)(pzS - mF - Q2 mx ) - A,p,S = o.
Substituting here the value of m F from (12.86) for the case of an ideal exciter
(neglecting sm all terms proportional to the dissipation coefficient)

mT\2 (ßI + ß) + (1"2 + I)SPI S I / S2 - AIPI S = ° (12.95)
Expressing the harmonic linearisation coefficients appearing in (12.95) using
(12.71), (7.61), (12.84), we obtain after transformation
2(2PI S - bOS 2/ S I)
a x
(S
= nE._2 eo 2 + ~-eo S) (p=0.16(l-R 2)). (12.96)
2 SI n l S2
Comparing (12.96) with the similar relationship (12.21) obtained for a
suddenly increasing excitation (12.3) in the direction of the velocity of the
impacting oscillator, we can establish that gasdynamic processes are basically
similar to additional viscous damping, the parameters of wh ich are reflected in the
second term ofthe denominator.
According to (12.96), increasing the dry friction force bo leads to a decrease in
the amplitude of vibration of the piston, from wh ich there must be a critical value
of friction, when vibration ceases (a x = 0) . It is given by
bo = 2PI SSI / S2' (12.97)
Decreasing the coefficient of restitution R also leads to a decrease in aX> however,
for all values of R, regimes do exist.
Superposing the curves for the vibration energy boundaries onto the backbone
curves for the system, we obtain at the interseetion points amplitude a x and
frequency eo similar to that found in Fig. 12.4. Let us find the equation for the
backbone curves. Substituting the values for the linearisation coefficients (9.18)
using (12.76), (12.89) into the approximate relationship (12.11), we find
2[0 2L1 + (PISI / S2 - P2 )S]
ax = 30 2 _ eo 2 ' (12.98)
where as before, gap (interference) L1 is measured from the piston position when
the spring is undeformed. Setting the numerator in (12.98) to zero, we find
L1 = (P2 - PISI / S2 )S02 (12.99)
To satisfy this condition, the vibration frequency eo is independent of amplitude
a x and must equal .J30. This corresponds to the set of parameters for which the
piston, under the effects of average force due to air pressure and elasticity of the
spring, just touches the stops with neither gap or interference. As L1 is increased
from this value the nonlinearity of the system remains of a soft type; as L1 is
decreased, the system becomes harder.
Finding the values ofa x and eo from the system of finite equations (12.96),
(12.98), we can find from these the value al" assuming 0 2 + k - eo 2 = ° in
(12.73). Using (12.71), (7.61) we obtain
a p = axeo [°.160 - R 2 )eo

S
+~]
naxeo
. (12.100)
To verify the accuracy of the assumption regarding the ideal excitation source,
used to construct the energy boundaries for vibration, it follows that the value of
§13 Random vibration 185
ml' must be close to (12.89) and this must be satisfied with the help of (12.92).
When using the simplified expressions (12.96), (12.98), it is necessary, in the
same way, to check the results of caIculation of small order of the second term in
the denominator of(12.85). Above all, inequality (12.94) must always be satisfied.
Finally, note that the frequency of autoresonance, similar to (12.85), can be
obtained from other logic structures for the switching arrangement. Really,
assuming in (12.81) that
U2(P2S - mjo - Q2 mx ) - U1PIS;: 0
and substituting (12.86), we find, for example, neglecting values proportional to
the dissipation coefficient (the ideal excitation source):
ul/m~1 =u2/m~2· (12.101)
Linearisation coefficients in form (12.84) represent only one ofthe possible ways
of achieving condition (12.10 I).
Reference [3] presents a numerical method which can be used to find solutions
for the general set of equations which describe the dynamics of pneumatically
driven strikers.
§ 13. Random vibration

I. Let us turn to consider statistical dynamics for impact oscillators,
comparing statistical motion and force linearisation methods as examples and
comparing the results with exact solutions [37, 38]. The study will be carried out
for a system with double-sided stops (Fig. 3.1), assuming now that the applied
external excitation pet) is a stationary normal random process. The dynamic model
described has numerous technical applications, so it is worthwhile to study its
statistical dynamics from this point ofview also.
Assuming that P(t)=mp+po(t), where mp=const;p°(t) is a centred
random process, we shall write the equations for the system for an oscillator of
unit mass for the statistical motion linearisation method in the form
(S2 +bs +Q2)X +cDE(x,sx) + cDf)(x,sx) = m p + pO(t). (13.1)
Here the impact interaction force characteristics cD E(X,SX) and cD f) (x, sx) are
found from expressions similar to (6.15), (6.16).
Equation (13.1) corresponds to the structural scheme shown in Fig. 13.1. From
this it can be seen that the additive component x(t) , acting on the input to the
non linear block, appears as the output signal from the linear filter, of high factor
when Q» b/2. In agreement with the presentation of §8, section 2, this allows
us to consider the situation described as approximately normal and consequently,
to assurne normality for the complete process at entry to the nonlinear block. As a
result, we can describe its statistical linearisation according to the formulae
obtained in §8.
Assuming the solution of equation (13.1) in the form
x(t) = mx +X°(t) , (13.2)
where mx = const; xO(t) = x(t) - mx is the centred random process, we find the
linearised force characteristics for impact interaction
<D F;CX,S) "'" m F + k(x - mJ = m F + kx° , (13.3)
<DD(X,SX) "'" ßsx. (13.4)
Here m F' k and ß are expressed in terms of unknown statistical characteristics of
the solution with the help of (8.75) to (8.77), where y(t) = x(t).
Fig. 13.1
Substituting (13.3), (13.4) into (13. 1) and separating the constant and centred
components, we find
(13.5)
(13.6)
From (13.6) using formula (8.27) for the transformation by the linear system of
a stationary random process, we find the spectral density ofthe Iinearised system
Sx(w) = Sp(w) 2 ' (13.7)
102 + k - w + (b + ß)jWI
2
where S( w) represents the spectral density of the process indicated by the index.
Using (13.7) we find from (8.15)
a; = _1_} Sp(w)dw 2 '
(l3.8)
2n -00 102 + k - w + (b + ß)jWI
2
cr~ = _1_} w 2 Sp(w)dw 2 •

(13.9)
2n -00 02 + k - w2 + (b + ß)jWI
1
For simplification of the later relationships we will assurne that Po(t) has
constant spectral density Sp( w) = So = const, i.e. white noise appears. Applying
this and using the values ofintegrals oftype (13.8), (13.9) [184], we find
So
a2 = (13.10)
x 2(b+ß)(02 +k)'
2 So
cr Y =2(b+ß) (13.11)
The relationships (13.5), (13.10), (13.11) together with the values of the
linearisation coefficients, complete the system of equations to find the unknowns
Let us look at some particular cases of solution for the equations shown.
Assuming mp == 0, then mx == 0 and consequently m,. == O. Using this with (8.76),
(8.77) we find when M= I
k = ~(1 + R 2 ) exp[ _~2 /(2cr; )] cr~
(13.12)
4..}2;cr x ;:0(~/crx) cr;'
ß= 1 - R 2 exp[ _~2 /(2cr; )] cr y ,
(13.13)
21t ::0(~/crx) cr x
where 9{u) is the probability integral.
Dividing (13.11) by (13.10) we find
cr~ / cr; = 0 2 + k . (13.14)
Substituting (13.14) into (13.12), (13.13), we obtain
k 2[ 4..}2;cr/:0(~/crx> l-I (13.15)

=0 ~(l + R 2 ) exp[ _~2 /(2cr;)]
ß = 1- R 2 exp[ _~2 /(2cr;)] ~02 + k =
21t gT(~/crx)
-1/2 (13.16)
=(l-R2)OeXP[-~2/(2cr;)][I_ ~ exP[-~2/(2cr;)]l .
21t ST(~/crJ 4..}2;cr x ST(~/crJ
Substituting (13.15), (13.16) into (13.10), we obtain an equation to find crx . When
R = I (ß = 0) it can be simplified significantly and takes the form
2 So [ ~ eXP[-~2/(2cr;)]l
cr x = 2b0 2 1- 4..}2;cr x ::0(~/crx). (13.17)
When ~ ~ OCJ from (13.17) we find, noting that ::0 (OCJ ) = ~ :

(13.18)
which coincides with the the expression for the mean square value of the linear
oscillator, excited by white noise [77). As ~ increases, as anticipated, the intluence
of the stops remains of less significance.
Let us find the average frequency 0)" of x(t),i.e. the mathematical expectation
of the frequency when the process will intersect the level x = 0 with x > o. In
agreement with formula (8.25), we find
00
0)0 = 21t Jvw(O, v)dv , (13.19)
where w(u, v) is the joint probability density of x(t) and y(t) = x(t). Using
(8.61), (8.65) we find from (13.19)
0)
o
=21tC ooJvexp(-~ldv=
2 0 2cr~
cr = ~
2crx::0(~/crx) ST(~/crX>
y (13.20)
According to (13.15) when ß ~ OCJ the value of k ~ 0 and consequently the

average frequency tends to the natural frequency of the linear system ro ~ n .
Let us find the mathematical expectation for the impact frequency as double
the number of crossings when x(t) intersects level ß with x(t) > O. Using (8.25),
we find using (8.61), (8.65)
f = 2 fvw(ß, v)dv = 2C 2 exp

00 ( ß2
---2
2] dv =
Joofvexp [ -~
° 2cr x ° 2cr y
ß2 '\
(--j (13.21 )
exp
_--'---_cr_~-,-----1n 2 + k .
2n9T(: )
2. Let us now look at the solution ofthe problem offorced random vibration
of an impacting oscillator using the statistical force linearisation method. In this
case, by analogy with (8.28) the equation for the system can be written in the
following form:
(13.22)
where ß is found from (8.54); poet) is a stationary centred normal random process
as before with constant spectral density.
Let us linearise equation (13.22), assuming
Cl>(x) "" mF + k(x - mx ) = m,. + kx° . (13.23)
Here k = lIq is found from (8.85) and depends on the basic moment ofthe random
process when F(t) = Cl>[x(t)]. Substituting (13.23) into (13.22) we obtain the
condition
m,. =m p .
It can be used directly in the formula for k, which as a result depends on motion
only through the value cr,..
From (8.15), (8.27), for the system studied, we find similar to (13.8)
cr; = _1_ } Sodro 2 •
2n ~OO Ik - ro 2 + (b + ß)jrol
Multiplying both sides ofthis expression by e, and noting that
ecr; "" cr;, , (13.24)
we find
cr~ = _1_} eSodro 2
kS o
(13.25)
2n ~oolk - ro + (b + ß)jrol
2 2(b + ß)
For numerical analysis of the quantity ß, as we already found in §8, we can

assign approximate prior values according to (8.54), (8.57). Then using the
equation for k, equation (13.25) contains a single unknown quantity crF, which
defines its solution. Knowing crf', it is possible to find all the remaining
parameters. In general, coefficient ß can be found using the more complete

expression (8.56), however the caIculation process in this case is significantly
more difficult.
=
Let us now study in detail the case when m p 0, and consequently, m F 0, =
while the coefficient of restitution R = 1 (ß = 0). In this case, after substituting
(8.85), equation (13.25) takes the form
2 02So
aJ = (13.26)
. 4b;:7"(02 Ll I a /.)
When Ll ~ OCJ we find for the linear case
a~ = QSo 1(2b)
CaIculating a,drom equation (13.26), we can find a; from this using (13.24)
2 So;:7"(02 Ll l aj;)
ax =2 • (13.27)
bO
We will find the average frequency of F(t) as the mathematical expectation of
the frequency intersecting F(t) at F = 0 when F(t) > O. Obviously this coincides
with the average frequency 00 0 ofx(t), since <D(O) = O. By analogy with (13.19) we
find
f
00
00 0 = 27t Fw(O, F)dF , (13.28)
where w(F, F) is the joint probability density of F(t) and F(t) . Assuming the
simplification
w(F , F·) -- 1 exp[- (F _m2 .)2

j -~l
2 ' (13.29)
27ta F o p 2a F 2a p
we find from (13.28)
ai'
00 0-
- - (13.30)
al-
For a p we find, similarly to (13.25)
o}=_I_} e00
2
Sodoo 2 Soe (13.31)
27t -",Ik - 00 2 + (b + ß)jOOI 2(b + ß)
Substituting the values (13.31), (13.25) in (13.30), we find using (8.85)
00 0 = Jk = 0[2;:7"(02Lllo j.W I/2 . (13.32)
The average number of collisions in unit time is found as twice the
mathematical expectation of the frequency at exit from F(t) at level 0 2 Ll when
F(t) > O. Similarly to (13.21) using (13.29), (8.85) we find
f=2JpW(Q2~,F)dP=~exp[-(Q2~)2lGF =
o 7t 2G F GI'
(13.33)
= Jk exp[- (Q2 ~)2 ] = ~ exp[ _(Q2 ~)2 /(2G~) .
7t 2G F J27t l~(Q2~/GF)
3. To estimate the accuracy of the approximate solutions found we use the
fact that when R = 1 an exact solution for the problem considered can be
constructed for the case when excitation pet) exists as centred white noise
(P(t) = pO(t)). As before, assuming the system is symmetrical (Fig. 3.1) and
noting that in the interval between collisions the vibration of the oscillator is
defined by the equation
(13.34)
it can be considered as a two-dimensional Markov process [204], in terms of
vector phase coordinates {x(t), y(t) = x(t)} , since under conditions of 0-
correlation of the process, the excitation of the vector components in any time
instant to{x(to)' y(to)} completely defines the statistical state when t;::: t o .
The common probability density w( u, v) of x(t) and its derivative with respect
to time y(t) for the stationary solution of equation (13.34) satisfies the Fokker-
Planck equation [204]
a a I a2w
2
2
au (vw) - 00 [(bv + Q u)w] = So 00 2 (13.35)
This partial differential equation is ofthe parabolic type. To find an unambiguous
solution it is necessary to assign boundary conditions for w(u, v), which is non-
zero in regions where the arguments take the values -~ S; u S; ~, - 00 < v < 00 .
Taking into account the law which represents rebound of the mass from the stops
(R. = I), we arrive at the following boundary conditions:
w(~,v)==w(~,-v), w(-~,v)==w(-~,-v). (13.36)
When v ~ ±oo the density function must reduce to zero:
w(u,-oo) == 0, w(u, (0) == o. (13.37)
The known solution of(13.35) is satisfied with conditions (13.36), (13.37), [204]
w(u,v) = Cexp [ -S;

2b (Q2 U2
-2-+2 v 2
Jl ' (13.38)
where Cis a constant found from the condition ofnormalisation

d 00
fdu fw(u, v)dv = I. (13.39)

-d -00
Substituting (13.38) in the left hand side of(13.39) we find after integration
C fexp(- bQ2 u 2 )dU Jexp(- bv 2 )dV = 27tSo ry(.J2b~Ql.

-d So -00 So bQ ..[S;
Consequently,
where according to (13.18) 0';0 represents the standard deviation of x(l) in the
absence of stops. Hence corresponding to (13.38), (13.40) the distribution of
velocity for any coordinate is normal, the distribution for any coordinateis limited
to the Gaussian curve bounded by the values u = ±L1 on the graph of density for
the coordinate.
Let us find the mean square value of x(t):
0'; = C !;fu 2 exp ( - bQ

-- 2U
-2)du f exp(bv2
- -)dv =
00
_!; So -00 So
(13.41)
For the mean square value ofvelocity we find

b 2
Q
O'~ = C !;fexp ( ___ 2)
u_ du oefv 2 exp (bv2
- - )du = 0. 2 0';0' (13.42)
_!; So -00 So
Let us find the average frequency Wo of x(t). Using (13.38), (13.40), we find
from (13.19)
(13.43)
Finally, by analogy with (13.21) let us find the mathematical expectancy for the
coIlision frequency
f = 2Cexp (- - /),2) (bv2 )

2 - f vexp - - du =
00
20' rO -00 So
(13.44)
Let us compare the results of approximate and exact solutions. Fig. 13.2 shows
express ions for the standard deviation of x(t)
for the systems considered dependent on the
similar value with the absence of stops. The
IN curves correspond to the formulae found by
the exact method (13.41) and the methods of
statistical motion (13.17) and force (13.27)
IJ (,0 ~Z8/iJ
Iinearisation. The gradual "saturation" of
standard deviation is cIearly seen for the
Fig. 13.2
exiting process x(l) as the intensity of the
entry excitation pet) increases. Comparison
shows that the statistical force linearisation method produces higher accuracy here
than motion linearisation.
The curves for the expressions 0)0(L', / 0' xO) and J(L', / 0' xO) (Fig. 13.3) obtained
by different methods almost coincide. It can be seen that as L1 increases, the
OJ,/fl fjf)
8 8
\ \
"
4
"
~~
D 2 A/ISJ4 D f 2 1J/lSz '
Fig. 13.3
average frequency and frequency of collisions is reduced. It is interesting to note

that when t:./ cr xO > I the influence of increasing gap, as a method of combating
impact against the stops, is considerably reduced. This must be taken into account
in the design ofvibration isolation systems.
Chapter 4 Vibro-impact processes In systems with multiple
degrees of freedom
It don 't mean a thing if it ain 't got that swing.
Duke Ellington.
§ 14. Vibro-impact interactions of elastic systems

I. We shall now look at vibro-impact interactions in systems with multiple
degrees of freedom, implying either lumped or continuous models. As we showed
in §6, using dynamic compliance operators makes it possible to develop uniform
equations of motion in both cases wh ich differ only in the actual form of the
dynamic compliance, The description of analysis
methods for problems of existence and stability
of periodic motion in the previous chapter can be
extended to the class of system described here. A
solution for such problems can often be
successfully obtained in the general form with
the final formula [39].
We shall examine two elastic linear systems
1 and 2 (Fig. 14.1) the contact elements ofwhich
(assuming particles, plane sections) arranged at
distance L'l apart, undergo one-dimensional
motion with collisions und er the effect of an
exciting force ~II (I), applied to element n of
system I. We shall assume that both systems are
Fig 14.1
loaded by constant forces
Gi' (i = 1, 2; r = 1, 2, ... ) , acting on elements r.
Later, to agree with the way vibration energy spreads, system 1 will sometimes be
called aclive and system 2 passive.
The displacement of element x of the i-th system will be described by the
function U/X(l), measured from the undeformed position of the element. We shall
represents the dynamic compliance operator of the i-th system as L,p(x,s) which
connects the movement of element x due to an arbitrary force applied to element p
(p = r, n). In this respect, for a system with distributed parameters simplies alat.
If the contacting element has index zero, we can represent the force properties of
the impact interaction in the form
F = (v, sv) (14,1)
194 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
where v(t) = u lO (t) - u 20 (t)

As a result the operator equations for vibration of the system elements can be
written in the form
ulx(t) = L Llr(x,O)G lr + L III (x,s)PIII (t) - LIQ(x,s)cD(V,sv) , (14.2)
(14.3)
Here the terms under the summation sign describe the static deformation of the
°
elements.
Subtracting (14.3) from (14.2) we obtain when x = the equation of relative
motion ofthe colliding elements, which we can write in the following form:
v(t) = L III (0, S )PIII (t) + L(sHG - cD[ v(t), sv(t)]} , (14.4)
where
L(s) = LIQ(O,s) + L 20 (0,s) (14.5)
is the total dynamic compliance ofthe contacting elements;
G = L[LI!' (O,O)G lr - L 2r (0,0)G 2r ]L 1(0) (14.6)
is the reduced force oftheir relative static deformations.

Equation (14.4) is similar to (6.39), and, consequently, operator L(s) produces
the filtering effect described in §6. The equivalent linearisation method can
effectively be used here for the solution.
We shall assume ~I/(t) = ~II cosmt. We shall search for an approximate
periodic solution ofequation (14.4) in the form
v(t) "" m,. + va (t), VO (I) = a , cos( mt - <PI) . (14.7)
Applying harmonie linearisation to (14.1)
<l>(v,sv) "" m,. (m" A,) + k(m" ,al' )v() + ß(m,., al' )sv o (14.8)
and substituting (14.7),(14.8) into (14.4), we obtain for the constant and periodic
components
mp
L(O) = G - m,(m,.,a p), (14.9)
1 ] 0 LIII(O,s)
[ L(s) +k(m",a")+sß(m,,,a p) v (t)= L(s) PI"cosm(. (14.10)
The right hand side of the operator equation (14.10) contains the reduced force of
extemal excitation
LII/(O,s)? )
11/ cosm( = a p cos(m[ + \jf , (14.11)
L(s)
where a" = ILII/ (0, jm)r l (jm)lpllI ; \jf = arg L III (0, jm) - arg L(jm) .
We shall rewrite equation (14.10) in the form
W(s,mp,a,,)v()(t) = PIII cosmt . (14.12)
Here we introduce the definition
§ 14 Vibro-impact interactions of elastic systems 195
----
_I+[k(m",a")+sß(m,,,a,,)]L(s)
W ( s,m p ,a 11 ) - • (14.13)
LI,,(O,s)
We will transform equation (14.12) to the equivalent homogeneous form
taking aeeount of the form of solution sought (14.7), using the expression
PI"
ll"eoswt=-eos
D [( wt-<p" ) +<p,. ] =PI"
- ( eos<p,,+s--
sin<p" v o( t.
) J
a" a" w
Substituting this into (14.12), we obtain
[W( s,m",a" )--;;: sin<p" J] v o()

PI" ( eos<p" +s-w- t =!YJ ( s)v o()
t =0. (14.14)
The existenee eonditions for a periodie solution (14.7) of equation (14.14)

oeeur when the roots s = ±jw of its eharaeteristie equation (JI,(s) = 0 exist. Putting
s jw into the latter and separating real and imaginary parts we find
=
sin<p" = ay(w,m",a,,)/ PI,,' (14.15)

eos<p" = a"U(w,m,,,a,')/ PI". (14.16)
HereU(w,m",a,,) = ReW(jw,m",a");V(w,m,,,a,,) = Im W(jw,m,,,a,,) .
Eliminating the trigonometrie funetions from (14.15), (14.16), we obtain
~" PI" (14.17)
a" = JU 2 (w,m",a")+V 2 (w,m,,,a,,) IW(jw,m,,,a,.)I'
The system of finite equations (14.9), (14.15) to (14.17) define the unknown
parameters for the relative motion (14.7) of the eolliding elements. For systems
with one degree of freedom these are similar to (9.7), (9.11) to (9.13). Their
solutions are eonveniently found by using the following sequenee. Finding from
(14.9) the term m,,(a,') , and substituting this into (14.17) we obtain an
independent equation to find a". Further, we ean find the phase <p" using (14.15),
(14.16).
After finding the parameters for the relative motion v(t) and using (14.2),
(14.3) we ean find the motion of the remaining elements of the system. The
approximate periodie solution for equations (14.2), (14.3) will be taken in the form
"" mix + Ui~ (t),
U ,X (t) Ui~ (t) = a ,x eos( wt - <p IX) • (14.18)
Substituting (14.18), (14.8) into (14.2) and (14.3) and separating eonstant and
eentred eomponents, we obtain, using (14.10)
mix = ILlr(x,0)G 1r - LIO(x,O)m,.(m,,,a,,) ,
(14.19)
(14.20)
where
_I+[k(m,,,a,,)+sß(m,,,a,,)]L(s)
W1x ( s,ml"a ll ) - • (14.21)
A
Here
A = L 111 (x, s){1 + k(m,,, a,,) + sß(m" , a,,)]L(s)} -

- [k(m", a,') + sß(m", a" )]L IO (x, s )L III (0, s), (14.22)
W2x ( s,m l1 ,a 11 ) -_ 1+[k(m",a")+sß(m,,,a,,)]L(s)

[k(m", a,') + sß(m,,, a,J]L 20 (x, 5 )L III (0, s)
From (14.20) we find the amplitude and phase ofthe periodic components
aix = PlnIW,x(Jw,m,,,a,JI- 1 , (14.23)
sin </),x = aiXV,x (w, m,,, a p ) / PIII ' (14.24)
cos<p,x = a'XUiX(w,m,,,a,,)/ PIII · (14.25)
Here
U,x(w,m",a,,) = ReW,x(Jw,m,,,a,');
V,x(w,m",a,,) = ImW,x(Jw,m,,,a,,).
2. To study the stability ofthe periodic solutions against small perturbations
we can use the approach described in §10. We shall assume that in the vicinity of
the stable solution (14.7) relative vibration ofthe colliding elements takes the form
v(t) = m" (t) + G" (t) cos[ W{ - <pp (I)], (14.26)
where m,,(t),G,,(t) and <p,,(t) are slowly changing functions oftime.
From equation (14.12) we find the work ofthe dissipative and excitation forces
for one period of the steady motion (14.7). The loss of energy E d is calculated
from the equation
f{W(s,m",a,,)v (t)}sv °(t)dt = na,,V(s,m,,,a,,).

i
Ed = 0 2 (14.27)
For the work ofthe excitation forces Ee , we find

i
Ee = fPIIICOSW{.svo(t)dl=nPllla,sin<p" (14.28)
°
A specific feature of expression (14.27) in the case of a system with several
degrees of freedom is that it can take either a positive or a negative value. The
latter situation can arise in the case when the observed coordinates, in terms of
which the expression has been written (for example VO (I) ), does not coincide with
the coordinate of the applied force (here unCt)). As a result, when so me forms of
vibration occur, an odd number of nodes can be situated between the coordinates
and consequently the phase shift between the force and the observed coordinates
<p" is greater than n.
If we require that, when the amplitude a" diverges from the steady value a",
the balance of energy is distributed in such a way as to compensate this
divergence, we find the energy conditions of stability
~ 14 Vibro-impact interactions of e1astic systems 197
> 0, if sin <j)" > 0,

(14.29)
< 0, if sin<j)" < O.
It is easy to see that the first inequality in (14.29) can be obtained directly from the
characteristic equation of the linearised system (see (14.14)) using condition
(10.2). Hence the formal application of condition (10.2) to systems with several
degrees of freedom requires that the excitation force is reduced first to the
coordinate observed and a study of the phase shift between this coordinate and the
reduced force. In the given case, for example, during the search for a solution of
type (14.7) it is sufficient to set the right hand side of equation (14.10) equal to
al,coswt, while the resulting force Pll/(t) , using (14.11) can be considered as
PlI/ cos( wt -1jJ) .
An important detail, however, is that, as will be shown below, both cases of
(14.29) lead to the same final expression to find stability, hence this can be taken
as the basic case for the phase relationship, for example, in equations of the form
(14.12). Really, differentiating the first inequality (14.29) and assuming, using an
arbitrary choice of function defined by (14.26), that in the disturbed motion
in,,(t) and Ci,. are related by conditions (14.9) but (ji" and Ci,. by conditions
(14.16), we find using (14.15)
dV a , sin(ji" a l ,
( --+ l -2
- d(jl"
--;:::-cos<j),,----;:::-
J =
da" a" a" da" __ "l'-U p
(14.30)
=( dV +~-ud(jl"J >0.
l da" Ci" da" ,._(1,,-(/(,
Differentiating (14.16) as an implicit function (ji, of Ci,. , we find using (14.15)
(14.31)
Substituting this equation into (14.30), we obtain eventually
l J v(v + Ci" ~Jl

[Ur U + Ci" dU +
da" da,. .
>0. ( 14.32)
"I'=a(t
When differentiating the second inequality (14.29) we have the opposite sign
In (14.30), however, noting that now V < 0, after substitution (14.31) and
multiplication of both sides of the inequality by the value Ci,y < 0 we obtain
(14.32) again.
Later in this chapter we will use condition (14.32) to determine stability, which
will allow us to use the general approach for phase reference for all vibrating
elements of the system, for example, relative to the applied external excitation at
any point.
In the absence of dissipation in the system V (w, m (" a r ) =' 0 and inequality
(14.32) gives the following boundaries for the stability domains:
U(w,m,,,a,') == 0,
[ u(w,m,,, Ci,,) + Ci" ~U(w,m,,,

aa"
Ci,,)]
__
== O.
(14.33)
(f{,-(lI'
In the particular case of a linear oscillator impacting against the stops,

U == 0 2 + k - (02), V == (b + ß)(O and condition (14.32) coincides with equation
(10.9) obtained ear1ier.
The description of the approximate approach, of course, does not completely
exhaust the study of the question of stability of the periodic solution for systems
with arbitrary degrees of freedom. In the context of applying approximate methods
to find periodic solutions, astriet study of this quest ion is often impossible. At the
same time it must be borne in mind that, when using approximate methods, as a
rule periodic motion is assumed apriori, mostly for some physical reason. Thus
the problem consists only of separating stable and unstable approximate solutions,
which can be done using the procedure set out.
Note that a similar approach can be used to study stability of the exact solutions (682), (6.83)
obtained §6 section 5. Let us assume that the disturbed motion takes the form
x(t) = acos[wl + <p(t)]- J(t)X(t) ,
where <P(t), J(t) are slowly changing functions oftime.

Using the first expression (6.83) as the redllced condition tor the balance of work between the
dissipative and exciting forces, we shall formlilate the energy condition tor stability in the torm
d { J [.
-=-X_(O)+ I +SllHpr J .J
>0. (1434)
dJ wa (l+R)M J]=J
lntrodllcing two arbitrary slow functions in place of one, we will connect them by an expression
similar to the second eqllation in (683). Differentiating (14.34) using this connection. we obtain. taking
account of (6.83)
J> -t.X(O) 0 .
,
X-(O)+-I [.X_(O)+-----
(02
I
(I+R)M
J-
Comparing (6.82) with the equation obtained, we can establish that the stability condition is
satisfied only by regimes which have the positive sign bdore the radical.
3. Assuming that the impact interaction between the contacting elements

(Fig. 14.1) is purely elastic (ß =' 0), we will use the expressions for the harmonie
linearisation coefficients in the form (7.65), (7.66) to find the final solution,
bearing in mind, that here we assume ms == GL(O) . Introducing the designation ~o
for the reduced value of the gap (interference) taking account of the static loading
of the system
~o == ~ - GL(O), (14.35)
we obtain the equations for the linearisation coefficients in the form
m F == (a" - ~o)rl (0), (14.36)
§ j 4 Vibro-impact interactions of elastic systems 199
k = 2(1- L1 0 / a" )L I (0), (14,37)

by making use ofthe fact also that, according to (7,63), m" + a" = L1 ,
We will estimate the behaviour of the system during vibro-impact interactions
through the dynamic compliance of its elements at the fundamental. This is
adequate, since higher harmonics are localised near the impact pairs and
consequently have little influence on the vibration of the elements outside the
contact zone (see § 15, section 5),
For brevity we will restrict the calculation to the dynamic compliance of the
colliding elements, From (14.23) using (14.21), (14.22) we find
1
a lO
-----,,---""------;- - 1 + k(m",a,,)L 2o (0,jm)1 (14.3 8)
PIIlILIIl(O,jm)l- 1+ k(m,,,a,,)L(jm) ,
a 20 Ik(I11",a,,)L 20 (0,jm)1 (14.39)

PI"ILI,,(O,jm)l- l+k(ml',a,,)L(jm) .
For convenience, we take here the relative value of dynamic compliance. From
(14.37) it follows that when a" » L1 0 the value k ~2/ L(O). Substituting this value
of k in (14.38), (14.39), we obtain for the limiting resonance case, the final
equations:
a lo IL(O) + 2L 2o (0,jm)1 ( 14.40)
PI"IL I" (0, jm)1 - L(O) + 2L(jm) ,
a 20 _I
2L 20 (0,jm) I (14.41)
PIIlILIIl(O,jm)l- L(O) + 2L(jm) ,
In cases when vibro-impact resonance occurs near to the natural frequency of
the impacting linear system, the static compliance in (14.40), (14.41) need not be
taken into account compared to the dynamic value of these quantities, and both
formulas lead to the identical result
I G IO
PI" LI,,(ü,jm) I= IL L20 (ü,jm)
IO (O,jm)+2L(jm)
I (i = I, 2). (14.42)
From (14.42) it follows that for colliding elements with identical dynamic
compliance the latter due to interaction is reduced by half.
To study the basic case we will initially solve equation (14.17) neglecting the
dissipation in the system for simplification. In this case the dynamic compliance
considered remains real (see §6, section 3). Substituting (14.13), (14.37) into
(14.17) and writing using the earlier assumption
LI,,(O,jm) = fll,,(m), L(jm) = fl(m) = fllo(m) + ~l20(m),
(14.43)
fl( m) / fl( 0) = fl 0 (m),
we find after transformation
± PI"fllll(m) + 2flo(m)L1 o
(14.44)
a" = 1+ 2fl o(m) ,
where
1 + 2~l 0 (w)( 1 - ~ () / a)
U = , V = 0. (14.45)
IlIII (w)
Since a" is positive, according to (14.44) when satisfying the condition
PI 11 III 111 (w)1 > 211l()(w)~()1 (14.46)

a unique solution occurs , defined by the express ions
± PIIIIlIII (w) > ° when I + 2~l () (w) > 0,
± P11I1l 11, (w) < °

when I + 2~lo(W) < 0.
(14.47)
When the expression (14.46) has opposite sign

P11I III 111 (w)1 < 21~l()(W)~ol ( 14.48)
solutions exist in pairs and only when the value of the parameters satisfies the
condition
i.e.
llo(wM o > ° when I + 2~lu(W) > 0,
< ° when
(14.49)
~lo(W)~o 1 + 2~l()(W) < 0.
Let us look at stability of the solutions obtained. Substituting (14.45) into
(14.32) we obtain stability, conditions using (14.44) which coincide with the
existence conditions (14.47). Hence, in domains where solutions which satisfy the
conditions of uniqueness exist, the existing solutions are always stable. In such
cases when solutions exist in pairs, one and only olle of them is always stable.
This is defined by condition (14.47). Using (14.47), (14.49), (14.44) shows that
this solution always corresponds to large amplitude vibration.
Substituting (14.37), (14.43), (14.44) into (14.38), (14.39) and after re-
arrangement we finally obtain
a lO
1+J
1l(0)
[± ~llU(W)~o +~l7()(W)l
PIII~llll({J) -
(14.50)
2~l20(W) +
~l(O)
[I ~()
PIII~llll (w)
l
(14.51)
1+2~l()(w)
Note, that because of nonlinearity of the system examined, the expressIOns

obtained for dynamic compliance depend on the excitation.
Let us now look at the case (Fig. 14.2) when passive system 2 is not fixed and
is c1amped to active system J by a constant force. Then L 2() (0,0) = CIJ and
consequently 1/ L(O) == O. As a result from (14.35) to (14.37) we find
m} = G, (14.52)
k = 2G / G" (14.53)
Neglecting dissipative effects in the system and assuming that the passive
system is sem i-infinite, we find
LIII(O,jO) = 11111(0), Llo(O,jO) = 1110(0),
(14.54)
Here we take account that wave dissipation of energy occurs in the sem i-infinite
system.
Substituting (14.13), (14.53), (14.54) into (14.17), we obtain
(14.55)
for which
2G
1+-1110(0)
U= G" v= 2G~l20(0). (14.56)
11111 (0) a,,11111 (0)
Solving (14.55) for a", we find
2GIl20(0))]2
a,,=±PIlIllil'(O) 1- [ - - - -2Glllo(Ul). (14.57)
PIIIIlIII(O)
According to (14.57) vibro-impact regimes exist when
Os 2Gf..l2o(0) s PIII!f..lIIl(O)!' ( 14.58)
whence, if
PIII !11111 (O)! > 12G~I1~o(0) + 11;0(0)1, (14.59)
then a unique regime exists, wh ich, when

11111 (0) > 0 corresponds to the positive sign before
the radical, but when 11111 (0) < 0 , to the negative
sign. If condition (14.59) is not satisfied, then
regimes exist in pairs and only when 11 10 (0) < 0 .
From its physical appearance, inequality
(14.59) is similar to that obtained when solving the
previous problems for condition (14.46). Both of
them determine levels of excitation at which vibro-
impact regimes exist for all values of the system
parameters.
Let us verify the stability of the solution. Fig.14.2
Substituting (14.56), (14.57) into (14.32), we
obtain after transformation the condition for
stability in the form
(14.60)
Hence, a unique solution which satisfies the existence conditions is always stable,
but when pairs of solutions exist, stability of one and only one of them
corresponds to inequality (14.60). According to (14 .57) and in the given case, it
corresponds to the largest amplitude or vibration.
Let us now find the dynamic compliance of the colliding elements. From
(14.38), (14.39) and using equations (14.53). (14.54), (14.57) we obtain after
transformation the formula
(14.61 )
(14.62)
In domains defined by the conditions (14 .58) where vibro-impact regimes exist,
the value of(14.61) changes within the following limits:
----c:-a:..;:..lo--,- E
Plnl!llnCW)1 1L
=+= 2G!lIO(w). 1 +
Pln!lln '
[2G~IIO(W) l-
PI 11 !l 111
(14.63)
Vibration of system 2 (Fig. 14.2) in a vibro-impact regime appears to be

independent of excitation in the active system and its dynamic qualities according
to (14.62), and is determined only by the dynamic compliance of the passive
system and the value of static clamping. This situation can be used to regulate the
transfer of vibrational energy across the joint.
In the following paragraph we will turn to the analysis of the general
expressions found and applying them to the problem of vibro-impact processes
occurring during collision of rods. Note that the solutions developed can also be
used when the active system is excited by the application of several sources of
identical frequencies, applied to elements n and having amplitude P III . In this case
we must use ~n!lln(w) everywhere in place of LPl ll f-l ln (W) ,
4. To consider fundamental and sub-periodic resonances of vibro-impact

systems of general form, the same approach can
be used as was taken in §9 section 4 for a system
elf
with one degree of freedom. Basically, such a
method to approximate solutions using periodic
Green's functions will be realistic only when the
contacting elements have mass, hence at the
points shown in Fig. 14.3, we will assume that
the colliding elements consist of rigid bodies
with masses MI and M2 and coefficient of
restitution R. In addition for generality we
assume that the active system I has several
synchronised sources of excitation with
frequency w applied, acting on elements n (n =
1,2, ... ) with forces P ln (l) .
Taking into account that in the system
considered, vibro-impact interaction with period
§ 14 Vibro-impact interactions of e1astic systems 203
T can be characterised now as a periodic sequence of 8-impulses (see (6.44)), then

the equations ofvibration ofthe elements by analogy with (14.2), (14.3) take the
following form:
ulx(t) = LLli(x,O)G lr + LL,,,(x,s)P,,,(t)-
x (14.64)
-L,o(x,s)J L8(t-vT),
LW - vT).
CL
u 2x (t) = LL 2r (x,0)G 2r + L 20 (x,s)J (14.65)
Subtracting the second equation from the first, and using (14.5), (14.6), when x = 0
we obtain
LW - vT).
x
v 2x (t) = LLI,,(O,S)Pli,(t) + L(O)G - L(s)J (14.66)
"
The last term on the right hand side of equation (14.66) represents the periodic
Green's functions of the first order X\J(t) (see §6 section 4) for the systems.
Writing
(14.67)
"
where Set) as before represents the solution with frequency (J) of the
corresponding linear problem, obtained by neglecting impact interactions.
Measuring the displacement from the static equilibrium condition of the system
using the function
v()(t) = v(t) - L(O)G , (14.68)
we obtain once again the general solution in the form (6.77)
v()(t) = S(t) - JXI (t).
Time measurement begins at the instant of impact.
The approximate solution for aresonant vibro-impact regime can be written in
the form
v()(t) ",,-JXI(t). (14.69)
Finding J from the impact conditions v()(O) "" -JXI (0) = v(O) - L(O)G =
= L'. - L(O)G = L'.o' we find J = -L'.o / XI (0) and, consequently,
L'.
vo(t) ",,-O-XI(t). (14.70)
XI (0)
Existence conditions for this solution are found from an energy balance for one
period of motion T = 2rrl/(J) (l = I, 2, ... ). The energy loss during impact we find
similarly to (9.51)
E d2 = (1- R)J2 = (1- R~L'.~
(14.71)
2(1 + R)M 2(1 + R)X; (O)M
From (14.66) taking account of (14.68), (14.70) we find the work done by
dissipative force, acting on linear sub-systems J, 2 (Fig. 14.3):
Edl = J{rl(s)vo(t)}vo(t)dt. (14.72)
Having in mind the appearance of the functions in the integrand, and taking
account ofthe expressions for periodic Green's functions which appear in it in the
form ofFourier series (see §6 section4)
vo(t)=~
XI (O)T
I
m=-x
L(mjOJJexp(mJOJtl,
I I)
(14.73)
vo(t) = t-,.oJOJ
XI (O)TI m=-x
I mL(mJOJ lexp(mJOJ tJ
I) I I
(14.74)
we find from (14.72) after integration
E,II = ~2~~ ImlmL I (mJOJIIL(mJOJ 1/

2
(14.75)
nl X~ (0) m-.I I) I)
The work done by the excitation force is found from (14.66) with the help of
the expression
i
Ee = J{rl(s)S(t)}vll(t)dt. (14.76)
The value of this integral depends on the phase of the periodic process S(t) in
relation to the instant of impact. Suppose, tor example, Set) = a cos( OJt + <p) , then
r l (s)S(t) = al' cos(wt + \jJ), (14.77)
where ap=IL-I(jOJ)la;\jJ=<p-argL(jOJ). Substituting (14.74), (14.77) into
(14.76), we obtain after integration
Ee apOJt-,.o
=-~L(jOJ)sm<po
I I. (<pI) =\jJ-argL(jOJ)). (14.78)
As a result the equation of energy balance

E"I + E d2 = Ee
H f 2(l"-;~M
in the case considered takes the following form:
Xf;Ol 1~', ~ mL ,( ~ro m;ro + ) -

(14.79)
t-,.
-u"OJ--O-IL(jw)lsin<po' =
XI (0)
From this two possible values for the unknown phase for the process
<Po (0 s <Po sn) can be found:
(14.80)
be written as:
Ql'ooIL(joo)1 ~ ~L~
XI (0) 1TCl mcl I I
2
Im Im Cl (m)ooJIL(m)ooJI + 1- R ).
2(1 + R)M
(14.81)
In particular, for systems with one degree of freedom, with an operator for the
linear portion from (6.72) given by Im L I (m)ooll) = bmooll, we obtain the
expression obtained earlier for this case (9.61).
For symmetrical double-sided interaction of the elements of the impacting pair
the equation for relative motion ofthese elements takes the form
vo(t) = ILIII(O,s)P11/(t)-
1/
(14.82)
-L(s)J [ l'~xÖ(t-VT)-l'~XÖ
' L X (
t-vT-"2
T)] .
Writing the general form of solution using periodic Green's functions of
second order X2 (I) :
vo(t) = S(t) - JX2(f), (14.83)
we arrive at the following approximate solution:
Ll o
vo(t) "" - - X , (I). (14.84)
X 2 (0) -
The existence conditions for this solution, found from the equation for energy
balance, now take the following form
Q" oo IL(joo)l:2: _~{2oo22 I(2m -I) Im L [(2m -1»)00] x

I
Xz (0) TCI m=1 1
XIL[(2m-I»)ooJI 2 + I-R } (14.85)

I 2(1 + R)M
The fitness ofthe approximation used as before can be conveniently calculated
by using one ofthe expressions oftype (10.66), (10.67).
An important simplification to the approximate solutions (14.70), (14.84) can
be obtained by expanding the operator L(mjoo / I) which appears in the equation,
as aseries oftype (6.61), (6.64) for natural modes ofvibration ofthe linear part of
the system with the subsequent use of the basic vibration modes. Suppose, for
example, during vibro-impact interaction of identical systems with a finite number
of degrees of freedom, only the v - th form of vibration appears. Then In

agreement with (14.5), (6.61)
L( m~OJ ) = 2L IO ( 0, ~OJ
[Q;- (~oo rr
)= r=====2=A="o=====
+ ~,Q;r_2OJ_2
(14.86)
where Qy is the frequency of the v - th mode; r y is its reduced damping

coefficient; A yo is the normalisation coefficient of the structure (see §6 section 3).
The frequency response (14.86) is similar to (9.68), hence to find the existence
conditions for resonant regimes in this case we may use express ions of types
(9.69), (9.71).
Series in m in the equations for the periodic Green' s tunction can be turned
into finite expressions (6.75), (6.76) or trimmed for further simplification of the
solution by taking the filtering effect of the response (14.86) into account. When
m = 1 we obtain the harmonic approximation for resonant vibration of the vibro-
impact system.
Thus the approximate method described allows us to study vibro-impact
interactions by examining the equivalent linear problem for the effect of a periodic
sequence of impulses and consequently to use the wide opportunities of linear
analysis of interacting systems. Note that the form of approximations for the
periodic solutions (14.70), (14.84) does not depend on the character ofthe periodic
excitation in the system. The excitation, according to (14.76) intluences only the
existence conditions of for solutions. This fact retlects the characteristic
peculiarity of resonant motion, that its form is determined to considerable extent
by the natural characteristics of the system. A convenient detail exists also that,
particularly for nonlinear systems, a non-unique solution does not affect the form
of the law of motion, but occurs only when finding the value of phase from the
existence conditions for the regime. It is clear that tor systems with hard
nonlinearity, i.e. with a gap ("" > 0), the phase (jlo, corresponding to a stable
solution, appears in the range 0::; <Po ::; 1[ / 2 .When ""0 < 0, the corresponding
phase range is 1[ /2 ::; (jlo ::; 1[. Using (14.80) we conclude that stable solutions
correspond with one and only one of the two phase values which satisfy the
existence conditions.
5. The procedure described can be generalised when searching for resonant
regimes in systems with many one-dimensional impact pairs. lntroducing vector
x(l) as the coordinate of the relative disposition of the elements in pairs, we can
write an integral equation for it, similar to (6.87):
r
x(t) = S(t) - JX(t -1)<l>[x(1),x:('r)]d1
()
Here <l>(x, x) is the vector for the force characteristics of the impacting pairs;
S(t) is the vector of relative vibration of the elements of the impacting pairs when
the impact interaction is neglected; X(t) is the matrix of periodic Green's
functions for the linear part ofthe system
----
X(t)
1 '
=- I A(mjw)exp(mjwI).
T 111:.:-1
A(jw) is the matrix of dynamic compliances.

In systems ofthe type examined it is possible to discover nonlinear (impacting)
modes of vibration, characterised by the fact that impact occurs simultaneously for
all pairs. The number of impact pairs sharing vibration in various modes can vary.
Vibro-impact regimes of the resonant type can occur when such vibration is
maintained in the system by external excitatlOn.
For the regimes described we find similarly to (14.69)
x(t) ~ -X(I)];
where ] is the vector of impact impulses. From the impact conditions
x(O) =,1\, where,l\ is the vector of gap (interference) arrangements, we find
eventually
x(t) ~ X(I)X 1(0),1\. (14.87)
The existence conditions for regimes of type (14.87) and the unknown phase of
excitation are found from an energy balance 01' the dissipative and excitation
forces in these solutions
r r (1_ R )J 2
f{A-1(s)S(t)}.X(t)dt= f{A1(s)x(I)}'X(l)dl+ I I I
o 0 I 2(1+R I )M ,
where i is the index ofthe impact pair under consideration; R , is the coefficient of
restitution for the i - th pair; J , is their impact impulse; M, is the reduced mass.
The dot indicates a sc al ar product
Fig. 14.4 demonstrates the
experiment that confirmed the
aEAOS
existence of the vibro-impact
modes described [246]. A
rubber cord with three fixed
plastic beads was stretched
between the pin of a force ~IClOwAt. .. J
transducer and the pushrod of
Fig. 14.4
an electrodynamic shaker
producing trans verse excitation
ofthe cord. The beads had unilateral rigid limitation 01' motion and it was possible
to fix the instant of contact by means üf conlacl pick-ups. The pictures of
nonlinear modes (standing waves) wen: laken by using a stroboscope
synchronised with the voltage applied to the exciter.
Fig. 14.5, a represents the cooperative mode described when the set of beads
impacted their Iimiters simultaneously. Fig. 14.5, h confirms this from the records
of the transducer signals. Here J, 2 and 3 are the signals of the contact pick-ups
from the limiters, 4 is the force transducer signal, 5 is the pushrod vibration signal.
The set of beads involved in the cooperative mode of vibration behaved like an
impact oscillator with single degree of ti·eedom. These results demonstrate
frequency and amplitude pulling, transition t1'Oln vibro-impact mode to linear
mode and back as a result of hard excitation (additional impulse). The frequency
band of the cooperative modes is located in the right hand side close to the lowest
natural frequency of linear vibration.
Fig. 14 .5, a
.. ._-----
----_ .. - ..
Fig. 14.5. b
In other conditions, the system described demonstrated a localised nonlinear

mode of vibration when only one of the beads vibrated with impact against a
limiter and others were almost stationary (see Fig. 14.6). The frequency band of
these modes was located in the right hand side close to the third natural frequency
of the linear system. The impacting particle also demonstrated frequency and
amplitude pulling and transitions from vibro-impact to linear mode and back as
the result of hard excitation
6. Vibro-impact regimes with large amplitude can exist not only under
------------
*14 Vibro-irnpact interartions of elastic systems 209
conditions of harmonic excitation of the system but also during random action on
the system from an exciting force . In the majority of cases such vibration is not
stable and after aseries of collisions collapses to a regime without impacts.
Engineering analysis of similar motion is concerned in principle with finding the
conditions under which impact interaction with the limiters will be reduced to a
minimum. Using the description of resonant motion of vibro-impact systems in the
form (\4.70), (14.84) from section 4, allows us to use the method described by M
Z Kolovsky in reference [126] to find
such estimates.
Suppose that in the system
described in section 4 (Fig. 14.3), the
applied external excitation
P(t) = Cl (s)S(f) is represented by a
stationary normal ergodic random
process with zero average value. We
shall ass urne that as a result of random
thrust in the system a vibro-impact
regime develops with frequency (().
For a sufficiently small level of ~ --= .-;--==------::-::..- -._-::-:-.
excitation (compared to the level of
the inertia and elastic forces) such a
regime can only folIowaresonant law
of motion and, consequently, the
corresponding vibration of the relative
coordinate vo(t) of the impacting Fig_ 14.6
elements (Fig. 14.3) can be
approximated by relationships of type
(14.70), (14 .84). Let us find the conditions to maintain such a regime with the help
of random external excitation P(t).
Representing the power of pet) for motion V o(t) by Ne' we find
I Ji,
Ne =- f P(t)v(t)dt ,
To 0
where To is the observation time. We find the expected value and mean square
value of Ne:
I Ji, ) I Ji,
(N e )= ( - fP(f)Vo(t)dt = - fVo(t)(P(t))dt=O,
To 0 To 0
(N;) = (~lJ P(t)v o(f)d/Jp(t')V o(t')df') =

To 0 0
1 Ji,lj,
= -2 f fVo(t)vo(t')RI'(t,t')dtdt' .
To 00
Here, angle brackets mean statistical averaging; RI'(t,t') is the cross-correlation

function of pet). Taking into account the periodicity of v()(t) and the stationarity
P(t), we find
1 li,li,
(Nn=cr~, =-2 f fvo(t)Vo(t+T)RI'(T)dtdr. (14.88)
To 00
The value of Ne follows the normal distribution law w(r]):
W(l1) = ~ I
V 21[cr N I!
exp[11
--2-
2cr N
2
('
1
.
From this the probability of breakdown of the vibro-impact regime will be as

folIows:
N,,(ül)
YJ{N e < Nd (oo)} = f W (l1)dll , (14.89)
where Nd(oo) is the power ofthe dissipative force:

Nd (00) = (E dl + Ed2 )/T , (14.90)
where T= 21[/00; Edh EdZ are found from formulae oftype (14.75), (14.71).
In those cases where pet) does not contain a periodic component then
lim R p (r) = 0 and, consequently, as the observation time To increases, the value
t->oo
of cr~, decreases. As a result, with the probability prescribed, we can find time To
during which the expression
(14.91)
becomes established.
Suppose that P(t) is white noise with correlation function R,,(T) = So8(t).
From (14.88) we find
1 li,li, S
cr~, f fvo(t)V(t+T)S08(t)dtdT=~(V;(t)).
=-2
~ 00 ~
Assuming, for example, o/'{N" < Nd(oo)} = 0.98, we find using the normal
distribution law [184]
( 14.92)
Thus with 98% probability
(14.93)
With the help of Parseval's equality, using (14.74) for the case of a single-sided
impact interaction
(v()'2) =
2ß~oo2 er 2
2 2
. 2
Im
IL(mjoo)1 . (14.94)
T XI (0) m~1
Substituting (14.90), (14.94) into (14.93) we obtain using (14.71), (14.75)
§14 Vibro-impact interactions of elastic systems 211
To >
8S ooo 2 I m2 IL(mjoo)1
m:::::1
2
2 • (14.95)
!1.20- [00
-2- -Im
2
x (1- R) ]
ImL-I(mjoo)IL(mjool 2 +--'----=-----
XI (0) TC m=1 2(1 + R)M
For systems with double-sided symmetrieal stops we find as the result of a similar
ealculation
8Som2 ~::C2m -1)2IL(2m - l)jm 2 l

To > m=1 2 .(14.96)
~~
- 2 - - [2m
-L.,(2m
2
~ 1
-1)lmL_I (2m -1)jmL(2m (I-R)]
-I)jm12 + .--'------'-----
X2 (0) 1t m=1 2(1 + R)M
Let us assurne now that P(t) is a harmonie disturbanee with random amplitude
and phase
P(t) = CI sin vt + C 2 eos vt .
Here CI and C2 have identieal mean square values, equal to (Yz P02 (t)) , where
Po(t) is the envelope of the random proeess [204]. Assuming that the distribution
ofphase in P(t) is symmetrieal we find [204]
Rp(T) = (Yzpneosvt. (14.97)
Substituting (14.97) into (14.88) we ealeulate using (14.74) for the ease of
single-sided collisions
(Yz pnoo 4!1.~ ir xc
a~, = 2.2 2 2 fImIL(mjoo)lsin(moot)dtx

TC I T XI (0) om=1
(14.98)
il" XJ
x fImIL(mjCD)lsin[moo(t + T)]eosvrdT (i = I, 2, ... ).

()m::::;1
When v*- 00 the value a~, ~ 0 as i inereases and, eonsequently, a stable regime
at frequeney 00 is not possible. When v = CD, from (14.98) we find
2 2
( 1/ p2 )oo !1.
a 2 = /2 0 0 IL '00 I?-
(14.99)
N" 4TC2X~(0) ( j ) .
As a result eondition (14.92) using (14.90), (14.71), (14.75) takes the form
2(Yz Po2 ) I/ 2 ooIL(joo)1 <
2 } (14.100)
<_~ { ~:tmlmL-I(mjoo)IL(mjoof+ l-R .
XI (0) TC m=1 2(1 + R)M
n
For regimes with symmetrieal double-sided eollisions we find similarly
2(Yz p l 2
/ mIL(jCD)1 <
< -~o ~ 2
- - {2m
--L.,(2m-I)lmL -I [(2m-l)jm]L[(2m-l)jm]
1 12 + 1- R } .
X2(0) 1t 11/=1 2(1+R)M
(14.101)
212 Chapter 4 Vibro-impact processes in systems with many degrees of freedorn
Conditions (14.100), (14.101) with probability ?/J==0.98 guarantee the

absence in the system of stable vibro-impact processes.
Sufficient conditions for the satisfactory use of approximation in statistical
cases can be framed, by analogy with (9.67), in the form
I~ol > 0'1'0' (14.102)
where 0'1'0 is the standard deviation of the process calculated when impact
disturbance is neglected.
§ 15. Vibro-impact interaction of visco-elastic bars

1. Using the general approach of § 14 for the analysis of vibro-impact
interactions in elastic systems, we shall now study the behaviour of straight bars
under the effect of periodic lengthwise impacts [19, 20].
We shall begin by examining a simple system (Fig.
15.1). In a bar of linear visco-elastic material, length land
o plane cross-sectional area S, longitudinal waves are excited
,..,."'1'"1'=f"
by a force PI/(f) = a p coswt, applied at the section with
coordinate x == n, measured from the fixed end of the bar.
During vibration the bottom end of the bar, with
coordinate x == I, interacts with a limiting stop placed at
distance ~ relative to the position of section x = 1 in the
undeformed bar. We shall also write the vibrational
displacement of arbitrary section x as a function uJt).
By introducing a force characteristic of impact
interaction <l>(u/,su/) and dynamic compliance operators
Fig. 15.1 LI/(x,s) , coupling displacement uxCt) to the force acting at
section x == n, we can write the operator equation for the
unknown function ux(t) in the form
ux(t) = LI/(x,s)P',(t) - L/(x,s)<l>(u/,su/). (15.1)
When x == I, equation (15.1) becomes similar to (14.4), hence we can use the
solution of the latter.
Suppose, initially, that an exciting force ~(t) == a p coswl is applied to the
lower end of the bar which couples with gap (interference) ~ to an elastic stop of
stiffness c2 • Under these conditions, the force characteristic of impact interaction
takes the form (6.7). For the present situation, to agree with (14.13) we have
W(s,m"a,) = L;I(l,s)+k(m"a,L (15.2)
where k(rn" a,) is defined by equation (7.54). Equation (15.2) has a c1ear physical
implication: the dynamic stiffness due to relative motion between the lower end of
the bar and the stop is equal to the sum ofthe dynamic stiffnesses ofthe elements.
Hence, to complete the solution, it is necessary to find an equation for the
operator qx. jm), i.e. to establish the reaction at the section with coordinate x on
§ 15 Vibro-impact interaction of visco-elastic bars 213
the application of a unit harmonie force (a" = I) to the lower end of the bar. We
shall solve a suitable linear boundary problem. Using (1.28), we can write the
differential equation for the bar in the form
S2 U - ( / +bs)8 2u/8x 2 = 0, (15.3)
where u = u(x,t), s = 8/ 8t. The boundary conditions for the fixed and free ends
are as folIows:
u(O,t) = 0, ES ( 1+ -b2 S ][8U(X,t)]

--- = exp{jwt)
. (15.4)
y 8x x='
where E is Young's Modulus.

Assuming solutions ofthe form
u(x,t) = ux(t) = L,(x,jw)exp(jwt) (15.5)
we obtain, substituting (15.5) into (15.3),15.4):
d 2L (x,jw)
(JW)2 L,(x,jw) - (y2 + jwb) '2 =0 (15.6)
dx
L,(O,jw) = 0, ES(I + jw y~ { dL,;jW) 1=, = 1. ( 15.7)
The solution of the linear differential equation (15.6) of second order with
constant coefficients satisfied by boundary conditions (15.7) takes the form
L ( .) _ 'Ay 2 exp('Ax)-exp(-'Ax)
, X,JW - 2 ' (15.8)
ESw exp('A/) + exp( -Al)
where
(15.9)
Since b = y21</( 1tw) , then, taking into account the smaller value of the damping
coefficient 1<, the solution obtained can be simplified, by expansion as aseries in I<
and neglecting terms of higher order. From (15.9), with first degree accuracy we
find
'A = jw + I<W . (15.10)
Y 41ty
Substituting (15.10) into (15.8) and restricting the expansion of the exponent, for
small values of I< , i.e. linear terms only, we obtain at x c-. I
. ) __
L, (I ,JW -
1_( tan r_·J~
I< t;+ Yz sin2t;]
2' (15.11)
ESt; 41t cos t;
where t; = wl / y. In §6 section 3, we obtained a different form for the dynamic
compliance operator for a system with distributed parameters in the form of a
series for the natural vibration modes of the system. Thus equation (15.11) is
equivalent to equation (6.63) by substituting p = q = I into the latter.
Clearly, calculation using equation (15.11) is more convenient, since it does
not require aseries summation. Moreover, under certain conditions, it is not
necessary to calculate the natural frequencies of the system, whieh is required

when using operators of the form (7.63). At the same time the idea of a closed
presentation for the dynamic compliance operator is often rather unwieldy and its
decomposition into aseries for natural vibration, using truncated series including
principal terms, can substantially help to simplify the expressions. For a more
detailed discussion of the questions mentioned, the reader is reeommended to see
the book by I M Babakov [24].
Substituting (15.11) into (15.2), we find when s = jw
. ES[ .X(,(,+hsin2(,] (15.12)

W(jw,m"a,)=- (,cot(,+ko(m"a,)+ J - . 2 '
! 4n sm (,
where k o(m" a,) = k(m" a,)1 I ES . Hence,
ES
U(w,m"a,) = -[(,eot(, + k o(m"u,)1
I
(15.13)
V(w) = ESx(, (, + h~in2(,.
4nl sin - (,
By analogy with (15.17), we find
l[
a, =X (,cot(,+ko(m"u,) ] 2 (X(,(,+hSin2(,]2l-Y;
+ -
411
. ,
sm- (,
' (15.14)
where X = apll(ES) is the displacement of the lower end of the bar due to the
action of static force apo
We can write the coefficients of the harmonic linearisation for the static
characteristics ofthe stops (8.53), (8.54) in the following form:
h=-a,c{~-;(aarcsina+JI-a2)1, (15.15)
k o =co[±-;(arCsina+aJI-a 2 )]. (15.16)
Here, h = m,.l/(ES), Co = c 2 ! I(ES), a = (D. - m,) I u" when -1 :s; a :s; I since only
under these conditions that the rod and stop will interact.
To define the basic parameters of motion of the lower end we must add to
express ions (15.14) - (15.16) the express ions for the eonstant components, similar
to (15.9)
(15.17)
Using (15.11), (15.15) this can be transformed to the form
a = D.{ a + co[~ -;(aarCSina + ~)Jr (15.18)
2. Using equations (15.14), (15.16), (15.18), we can construct the

amplitude-frequency response for the lower section of the bar. According to
(\ 5 .14), the expression
§15 Vibro-impact interaction of visco-elastic bars 215
Socots o +ko(m"a,)=O, (15.19)

where So = 00 0 1/ c defines all the natural frequencies 00" for vibration of the
elements of the system examined . Because the system is nonlinear these
frequencies depend on the vibration amplitude al" Thus, using (15.16) and (15.18),
.1-0
1}fIl<O
11
11
,\--
I1 \
I I \
\
\
\
-1 (J
b) 17ig 15.2
equation (15 .19) determines the configuration for the backbone curves of the
system. To find them, it is convenient to use the graphical method shown in Fig.
15.2. Relationships (15.18), (15.16), (15.19) are shown in Fig. 15.2, a - c with
coordinates shown on the axes. In addition, by choosing various values of a" as
the ordinate in Fig. 15.2 a, and following the construction shown by the arrows,
we can determine the corresponding value a on the abscissa and from that the
value of ko in Fig. 15.2 band So in Fig. 15.2 c. The resulting backbone curve
data for aiso) is drawn in Fig. 15.2 d.
For each value of al there are many natural frequencies , So' corresponding to
the different modes of vibration (in Fig. 15.2d the backbone curves for the lower
two modes are shown). For a system with gap (~> 0) the backbone curve for
each mode has a hard character while for a system with interference (~< 0) the
curve is of soft type. In addition, linear vibrations can also occur with both cases.
For (~ > 0) vibration with amplitude a, < ~ occurs; for (~< 0) vibration occurs
without loss of contact between the lower end ofthe bar and the elastic stop. In the
latter case, to obtain movement without impact the amplitude ofvibration must not
exceed the relative deformation of the contacting elements. Taking into account
that for an interference ~ the static force of interaction equals -~ES / I , while the
stiffness of the relative deformation occurring is equal to the sum of the stiffness
ofthe components, we obtain the conditions for non-impact motion in the form
lf/Z r. Z" (
'I b
_/~d&.
~ I
/FII Jf 5lf/Z 2Tr (
FIg 15.3
-MS/l Ö
a/ < =---.
ES / I + c 2 1+ Co
For (Ö = 0), the system has a spectrum of constant natural frequencies, shifted
towards high er values from the corresponding natural frequencies of the bar (Fig.
15.2 c).
The degree to which the nonlinear effects manifest themselves depends
directly on the increase in rigidity ofthe stops. In the limit as Co -7 00 the equation
for ko by comparison with (15.37), takes the form
k o =2(l-ö/a/). (15.20)
Substituting this value into (15.19) we obtain the equations for the backbone
curves
(15.21)
Imposing the requirement that a/ > 0 in (15.21) then in the case where there is a
gap the equation becomes
So cot So + 2 ;?: 0 , (15.22)
The sign ofinequality (15.22) changes to the opposite value for interference. From
this it follows that the natural frequency of the system lies in the following bands:
(2m - I)n12:::; So< SOm when 6. > 0,
SOm < So < nm when 6. < 0,
So = SOm when 6. =0 (m = 1,2,. .. ),
where ~Om is the solution of equation (15 .22). The backbone curves for the limited
system are drawn for comparison in Fig. 15.2 das dotted lines.
Putting ~ = ~o in (15 .14) and taking account of (15.19), we obtain
(15.23)
This relationship defines the energy boundary of vibration. The intersection of

curve I (Fig. 15.3), given by equation (15.23), and the backbone line 2, from
(15.14), gives upper values of the amplitude-frequency response. From (15.23) it
follows that with increasing mode number and consequently SO' the amplitude of
resonance of the relevant forms dies away rapidly. In order to achieve the
condition max al > 6. with the initial requirement for 6. > 0 we find from (15.23)
that vibration with impact against the stops only occurs for the condition
S < J4n X/(6.'X) . (15.24)
The relationship (15.24) has a clear physical meaning.
The remaining points on the amplitude curve that correspond to the nonlinear
4.0 J>O f regime are found from the simultaneous solution
of the final equations (15. 14), (15. I 5) and (15.18).
To achieve this the following method for
calculation is recommended. The equation for k o
(15.15) is substituted into (15.18), and the values
subsequently obtained for various values
S = const are plotted as curve I (Fig. 15.4); on
I -_ _-':---'-_ _ ...J (I;the relationships (15.18) already used to construct
-1 D 1 the backbone line (see Fig. 15 .2 a). The
Flg. ISA intersection points of the curves shown define
corresponding values of the amplitude response
a l and parameter a. Using these values in the expression a = (6.- ml ) 1a l the
value of m l can be found. Fig. 15.3 shows examples of amplitude curves for three
basic positions ofthe stop. Only the frequencies ofthe two lowest resonance zones
are shown. The branches of the amplitude curves which correspond to the linear
regime ofvibration (al< 6. in Fig. 15.3 a and a l < -6./(1 + co) in Fig. 15.3 b) are
found from the same relationships by substituting in them the value
a = I when 6. > 0 and a = -1 when ~ < o. Fig. 15.3, also shows dotted the
unrealised branches of resonance curves for the linear system, constructed using
(15 .14)for k o =0 when 6.>0 (Fig. 15.3 a)and k o =c o for ~<o (Fig. 15.3,b).
The first case represents vibration of the bar without stops and the second case
represents permanent connection ofthe lower end ofthe bar with the stop.
Fig. 15.3 shows that as the result of vibro-impact interaction of the lower end
of the bar with the stop a non linear distortion occurs for aII resonance curves,
similar to that observed for systems with one degree of freedom.
For the Iimiting case as Co ~ 00 when we substitute (J 5.20) in (J 5. 14) and put
% == 0, we obtain a simple relationship for the resonance curves of nonlinear
systems
n±X
a, = , ( 15.25)
scoU~ + 2
the existence domains of which, according to (15.25), are defined by the
I-J..l+tt-t:::>"", p IZ-d I
x- Iut
z<lz-d1
....
Fig.15.5
relationships a, > 0 when J'!. > 0 and a, > 0 when J'!. < 0 . Comparison of
equations (J 5.25) and (J 5.21) shows that as the value of excitation X increases the
amplitude-frequency response, then envelopes the backbone lines (Fig. 15.5) and
when X> 2J'!. change from double to single values. It is easy to see that this
condition represents a particular case similar to the general expression (14.46).
In agreement with (15 .25) for X> 2J'!. the vibro-impact regime exists for aII
frequency bands; for values of X< 2J'!. the regime exists only for frequencies
satisfYing the relationships
Scot S + 2 > 0 when J'!. > 0
(15.26)
Scot S + 2 < 0 when J'!. < 0
These conditions can also be obtained from the basic relationship (J 4.49). In
accordance with the results of § 14 section 3 stability exists for solutions with
greater amplitude, i.e. corresponding to the positive sign in (15.25) for J'!. > 0 and
negative sign for J'!. < O. In Fig. 15.5 unstable conditions are shown dotted.
3. Let us now examine the vibration of a system in which interference
occurs as the result of a compressive force (Fig. 15.6 a). This consists of a bar,
rigidly attached at its upper end to an infinitely large mass whose vibration can be
ignored. The bar is excited by a harmonic axial force . The complete system is
compressed against the stop by a static force G.
The system shown represents a device for ultrasonic material processing [74,
178]. Fig. 15.6 b shows such a device for ultrasonic erosion. The main device of
the machine is an acoustic head comprising a body I, a piezostrictive or
magnetostristive vibrator 2, with a stepped concentrator 3, and tool 4, attached.
§15 Vibro-impact interaction of visco-e1astic bars 219
"1
The piezostrictive or magnetostristive device receives impulses from a powerful

high frequency generator (at a frequency of 20kHz). The tool is compressed to
workpiece item 5 with a static force G. In the erosion zone 6, an abrasive
suspension is placed. Processing takes place as the result of impact of the tool on
the abrasive particles which settle on the surface ofthe workpiece.
The specific feature of the dynamic model described (Fig. 15.6 a) consists of
the displacement of an upper mass wh ich when vibrating, finds some dynamic
equilibrium position. The origin of the displacement function U, (I) is taken at a
prescribed section of the undeformed rod while coordinates of section x are
measured from the static position of the top surface of the bar under the action of
force G. In the system examined mF == G. Using this, the relative deformation of
the bottom section of the bar and the stop of rigidity c2 under the action of force
G , takes the form
Ei
ES
+!2 = h(1 +~)
c2 CO
(h GI) .
=
ES
( 15.27)
Now we can make use of equations for the harmonie linearisation coefficients in
the form (15.15), (15.16) using interference values ~ from (15.27). From (15.15)
we obtain
h
(15.28)
a =
l
c{;(aarCsina+~)-~]'
Assuming as previously that the harmonie excitation is applied to the lower
end ofthe bar we obtain again relationship (15.14). Hence the problem is reduced
to the solution of a system of finite equations (15.14), 15.15) and (15.28). This
solution is carried out using a technique identical to that explained in section 2.
We shall confine ourselves here to the examination of the limiting case
consisting of a rigid stop (c 2 ~ 00). Taking (15.53) into account, we find the
following equation for resonance curves from (15.14) when % == 0 :
-2h± X
al = , (15.29)
scots
and the existence domains are defined rrom the condition a l > 0 .
For X == 0 this becomes the equation for the backbone curve defined in
rrequency bands
I-i g.. 15.7
Yzn(2m -I) < So < nm (m = I, 2, ... ) (15.30)

and having fixed asymptotes coinciding with the natural frequencies of the bar
So =n(2m-l)/2 (see (15.11)).
According to (15.29) when X< 2h solutions for the parameters exist in pairs
in the natural rrequency bands of the system (15.30). In accordance with the
results of § 14 section 3, stable regimes correspond to large amplitude values, i.e.
having the negative sign in (15.29) .When X> 2h, the conditions for a unique
solution are satisfied, and these are stable for any excitation rrequency .
Fig. 15.7 shows the amplitude-frequency response of the system, constructed
for the two lower resonance zones with various values of X and h. Here the
chain-dotted line shows the backbone curves, the thin line shows the energy
boundaries for vibration (15.23), the thicker line shows branches of the amplitude
curve corresponding to stable solutions, while the dotted line shows unstable
values . It can be seen that as the compression force
increases, so does the frequency of maximum vibration
amplitude. When X< 2h vibration can occur causing loss
of contact at the stops either because of pulling at low
rrequencies or because of hard excitation.
Note that a problem close to the one examined at this
point has been solved in reference [154] using a f"ull Fourier
series. The results presented agree with results rrom this
reference. The nonlinear phenomena described were
Fig. 15.8 observed experimentally in [21].
4 Let us now look at the longitudinal impact
interactions between two straight bars with co-linear axes
(Fig. 15.8). The two contacting bar ends, whose coordinates are taken as the
§ 15 Vlbro-impact interaction of visco-elastic bars 221
origin, are arranged with a gap (interference) ~ between the two and which come
into contact under the action of a periodic force ~1I(t) == a" cosoot , applied at the
section with coordinates x == n in bar I. For generality, we will assume that both
bars are loaded by static axial forces G, Ci == I, 2) , applied at the origin of each.
As before, assuming that the displacement of sections is u,x(t) from their
undeformed position and introducing a characteristic force impact interaction
<D(v,sv) where v(t) == u lO (t) - u20 (t) , we obtain the following operator equations:
u 1x (t) == L IO (x,O)G I + L III (x, s)PIII (t) - L IO (x, s)<D(v,sv) , (15.3 I)
u 2x (t) == L 20 (x,0)G 2 + L 20 (x,s)<D(s,sv) (15.32)
where L,ix, s) is the dynamic compliance operator of the i-th bar acting at an
arbitrary section with coordinates n on the section with coordinate x.
Subtracting (15.32) from (15.31) when x = 0, we obtain the equation for the
relative motion ofthe contact section
v(t) == L III (0, s)PIII (t) + L(s){G - <D[ v(t), sv(t(]} (15.33)
where
L(s) == L IO (O,s)+ L 20 (0,s);
G == [L IO (O,O)G I - L 20 (0,0)G 2 ]L1 (0).
Equations (15.31) to (15.33) are similar to the
U
' ~
general equations (14.2) to (14.4) examined in §14,
hence we may use the solutions found in § 14 for
the problem of interaction between the bars. I
f f T I
To simplify the resuIts, we will assume that the
~ ~
!TI2 r
excitation force is applied to the section at the PIQ (f )
P,,(fJ
origin of bar I, i.e. it takes the form
PIO (t) == a p cos oot. Then, to obtain solutions
relative to the section at the origin, we can solve
the problem using only a dynamic compliance
I L
factor L,o (0, ioo) for the contacting section. uJ 11)
Suppose, as before, that the interacting bars are fig.15.9

made from a linear visco-elastic material and are of
finite length either buiIt-in at their ends (Fig. 15.9 a) or compressed by an axial
force G (Fig. 15. 9b) (in the system shown in Fig. 15.9 b, bar 1 is attached to a
large mass whose vibration can be ignored), then we may write the equations for
the dynamic compliance in the form (15.11) as:
L;o(O,ioo)==-'-'-(tan(-i%' S,+Yz:in2(). (15.34)

E,S,s, cos S,
4n
The interaction, as in the previous case, will be considered to be perfectly elastic,
corresponding to an unlimited rise in the contact stiffness Co --+ 00 .
We can rewrite (15.17), using (15.13) and (15.5) in the following form:
I LIO(O,ioo) I (15.35)
a" ==a p l+k(a,,)[LIO(0,ioo)+L20(0,ioo)]'
We shall eonsider the interaetion of identieal bars

L10(0,jm) = L 20(0,jm), (15.36)
Substituting (15.34) and (15.36) into (15.35) we obtain the equation for amplitude
a", similarto (15.14):
a" = x.1[seots + 2k o(a,,)Y + (xs

4n
S+ ~ ~in 2s J2
sm S
j
-rz
2
(15.37)
( ko =~J.
ES
The term k(a,') for the ease of fixed bars (Fig. 15.9 a) to agree with (15.37)
and (15.36), takes the form:
k(a)= 1 ( 1 -~o- ) =ES ~o-)

-( 1- (15.38)
" L,o(O,O) a,. I a,, '
In the ease ofeompressed bars (Fig. 15.9 b), equation (15.53) remains valid.
From (15.37) the baekbone eurves of the amplitude response for relative
motion ofthe system eonsidered beeome
seots+2k o(a,.) =0. (15.39)
Substituting in turn from (15.38) and (15.53) into (15.39) we obtain the
relationships
2~ o
a,. = - --"--- (15.40)
seots + 2
4h
a" =- seots
(h GI
= ES .
J (15.41)
From (15.37) and taking into aeeount (15.39) we ean find the equation for the
::/z
hg. 15.IO
energy boundary of vibration, similar to (15.23)

• 4nxsin 2 S (15.42)
al) = ,
xs( s + Yz sin 2s)
Aeeording to (15.37), interseetions on the (S,a,.) plane between the baekbone
eurves and the energy boundary of vibration define the maximum points of the
§ 15 Vibro-impact interaction of visco-e1astic bars 223
resonance response a". The remaining points can be found using a simple
graphical construction, similar to that shown in section 2. When % == 0 ,
substituting in turn from (15.38) and (15.53) into (15.37), we obtain the explicit
expression for the amplitude-frequency response
a" = 2L1 0 ± X , (15.43)
I;, cot I;, + 2
-4h± X
alJ = (15.44)
I;,cotl;,
Examples of resonance curves for the relative motion of contacting sections for
fixed bars are shown in Fig. 15.10. In addition, the branches 1 and 2
corresponding to linear vibration without impact are also shown. Fig. 15.10 a and
b represent systems with pre-arranged gap or interference. Similar results for
compressed bars are shown in Fig. 15.11. Examples for soft and hard excitation of
the vibro-impact regime are shown in Fig. 15.11 a and b respectively. 80th here
and later in this section, the curves are drawn at the lowest mode of vibration of
the bar, with the dotted \ines showing branches for unstable solutions.
Let us now examine the absolute vibration of the contacting sections.
Substituting (15.34) into (15.38) and (15.39) and re-arranging we obtain
I;,cotl;, + k(a,,) + jB I (15.45)
a lO = XII;, cot 1;,[1;, cot I;, + 2k(a,,)] + 2jB[ I;, cot I;, + k(a,J] ,
u)
Fig. IS.II
xk(a,,)
(15.46)
a 20 = 11;, cot 1;,[1;, cot I;, + 2k(a,,)] + 2jB[ I;, cot I;, + k(a,J]1
Where, for convenience, we make the substitution
B = %1;,(1;, + Yzsin 21;,)
(15.47)
41t sin 2 I;, .
Neglecting energy dissipation in the system (% == 0), we can obtain the
amplitude equations in an explicit form, which has a cIear physical interpretation.
For the case of fixed bars (Fig. 15.9 a), using (15.38) and (15.43) we find from
(15.45) and (15.46) when B == 0
aIQ -
_1~O~cot~ ± X(~cot~ + 1)1 , (15.48)
~ cot ~(~ cot ~ + 2)
± X- ~ o~cot~ I
a 20 = I~cot~(~cot~ + 2)
( 15.49)
where existence domains for the solution are found using (15.43) and the
conditions a" > ~o for ~o > 0 and al' > 0 for t.. o < 0 .
When X == 0 from (15.48) and (15.49) we obtain the equation for the backbone
curves for absolute vibration wh ich has half the value of the amplitudes compared
to the similar expression (15.43) for relative vibration
~
aiO = (i=1,2). (15.50)
2
~cot~ +
At the same time, according to (15.48) and (15.49), in the presence of
excitation together with vibration corresponding to the backbone curves (15.50),
which we will call clapping, the system resonates in absolute motion at the natural
frequencies ofthe bars, satisfying the condition cot~ = 0 , i.e. at frequencies
Fig. 15. 12
~ = Yz n(2m + I) (m =: 0, L . ..) . (15 .51)

Under these conditions, the displacement of the contacting surfaces has an
"accompanying" character and hence does not reveal itself during the relative
motion considered. Note that the vibration described for bars arranged with a gap
will only be maintained as the result of systematic interaction and consequently,
can not be explained within the framework of linear behaviour. For this reason
they will be ca lied grazing.
The maximum vibration amplitude occurs at the intersection of the backbone
curves (15 .50), (15 .5\) with the energy boundary for absolute vibration, defined
from(15.45) and (15.46). Substituting into these express ions the values from
(15.39) and bearing in mind that aeeording to (15.42), (15.47)

B=X/a;', (15.52)
we obtain using (15.38)
(a;' _~)2 +X 2
(15.53)
4(a;' - ~)2 + X2 '
• 4' a; -~ (15.54)
a 20 = ai' I • J ?
V4(al' - ~t + X-
In these relationships a;' is found from equation (J 5.42). From (J 5.53) and
(J 5.54) it follows for a suffieiently large value of a;' - ~ eompared to X,
eorresponding to low internal damping, that
(15.55)
Fig. 15.12 shows examples of the amplitude-frequeney response for absolute
fllg
.1I<,l/4
010
o>X/4
~
1
n/l '3n/4
11)
11
k .)
tt/Z 31T/4 fT
I!/Z
b)
JIf/ 4 ns
-~
nil 344
t1 )
1C (
Fig. 15.13
displaeements of the eontaeting seetions of fixed bars. Fig. 15.12 a, b eorresponds

to bars installed with a gap while Fig. 12 C, d to those with interferenee. In Fig.
15.12 a, apart from branehes relative to the vibro-impaet regimes, eurves 1 and 2
are also shown representing linear vibration of the aetive bar within the limits of
the gap without eollisions. In Fig. 15.12 c, d, eurve I shows linear vibration
without loss of eontaet.
For bars with a gap, as the exeitation frequeney inereases at the free end of the
aetive bar, then the free end of the aetive bar inereases its amplitude until
interaetion with the passive bar oeeurs. 80th bars oseillate in phase, i.e. with
identieal direetion of displaeement of similar seetions and reaeh maximum
amplitude at a frequeney L; = TI /2, equal to their natural linear frequeneies.
Onee anti-resonanee has passed, similar seetions move in c1apping eharaeter,
i.e. they move in opposite directions. Under these conditions, impact has the
strongest character. As the frequency increases, the amplitude of such motion
increases in accordance with the resonance curves for relative vibration (see Fig.
15.10 a). When vibration breaks down, as the maximum is passed, the active bar
vibrates Iinearly within the limits of the gap, but the vibration of the passive bar
decays. A similar picture occurs, in principle, for all frequency bands
corresponding to higher modes of vibration; however, with practical va lues of
interna I damping the resonance phenomena at these frequencies is weak.
As the frequency reduces from post-resonance regions (Fig. 15.12 a, b)
resonance in c1apping motion does not happen. The active bar oscillates to the
origination of contact, with the result of a decrease in its amplitude after contact.
Furthermore, both contact sections pass anti-resonance and consequently resonate
with grazing at frequency 1:; = 1t / 2 . Thus, as the frequency of excitation increases,
two resonance peaks occur, but with decreasing frequency, only one.
A contrasting picture emerges for bar devices with pre-arranged interference
(Fig. 15.12 c, d). In this case, as the frequency increases, the bars do not separate
and resonate at their own natural frequencies. As the frequency decreases from the
post-resonance domain, the bars separate as a result of vibration and they move
into a c1apping regime until they break apart as they cross the maximum amplitude
response, corresponding to the resonance curves of relative vibration (see Fig.
15.10b).
We shall now analyse absolute vibration of contacting sections of compressed
bars (Fig. 15.9 b). From (15.45) and (15.46) when ')( = 0 and, consequently
B == 0, taking into account (15.53) and bearing in mind that a,. in this case is
defined by (15.44), we obtain
-2h±X
a 10 = I:;cotl:; , (15.56)
-2h
a 20 = 1:; cot I:; . (15.57)
When X == 0 we have the equation for the backbone curves for absolute
vibration
-2h
a =-- (15.58)
,0 I:; cot 1:; ,
which also has twice the minimum amplitude compared to the similar expression
(15.44) for relative vibration. For maximum amplitude a;o by using (15.53),
(15.39) and 15.52) we find from (15.45) and (15.46)
, ,4h 2 + X2
a lO = a,. 2 2' (15.59)
16h +X
, , 2h
a20 = a,. I ' (15.60)
V16h2+X2
where a;' is defined by (15.42). When X« 2h we obtain the corresponding
amplitude values similar to (15.55).
Fig. 15.13 shows examples of the amplitude-frequency response for absolute

motion of the contacting sections of compressed bars. The additional curve 1
corresponds to linear vibration without separation.
Fig. 15.13 a, b shows the case of smooth excitation of vibro-impact regimes,
occurring with relatively small compressive forces (compare with Fig. 15.11 a).
As the compressive force increases above the critical value h = X/ 4 (see Fig.
15.11 b) the configuration ofthe amplitude-frequency response changes form (Fig.
15.13 C, ci) and becomes similar to the response of the
system with fixed bars, arranged with initial interference
(Fig. 15.12 c, ci), i.e. as the excitation frequency
increases linear vibration occurs, but as the frequency
decreases through resonance, intense vibro-impact
regimes can be established. Such regimes can also occur
when the system is excited with an especially large push.
A characteristic feature in the case of compressed
z bars is that according to (15.57) vibration ofthe passive
t bar does not depend on the excitation amplitude of the
U
Fig.15.14
active bar. This situation, as mentioned in §14, is typical
of the interactions examined for systems without
damping.
5. Let us now look at the vibration of systems
(Fig. 15.14) in which bar 1 length 11 impacts with a sem i-infinite bar 2 under the
action ofa harmonic force PIO(t) = a p coswt.
Interference is caused by a static force G. The system described is a special
case of the general problem examined in § 14 section 3, hence we will use the
results obtained there. To obtain the final expression we must define the dynamic
compliance of a sem i-infinite bar.
Using equation (15 .8), we obtain for a finite bar, measuring coordinates from
the free end
(15.61)
where
A _ jw
2 - ~Y; +b2 jw
Letting 12 ~ 00 in (15.61), we obtain for a sem i-infinite bar
.
L20 (x,jw) = - Y;A 2 2 exp( -A 2 X). (15.62)
E 2 S2 W
Linearising this equation in terms of %2' which occurs in A2 , we find
L 20 (x,jw) = _ _
Y2_(j
E2 S 2W 4n
+2)ex [-(j +2)
p
4n Y2
wx]. (15.63)
From this it follows that the transmission of dynamic effect to the semi-infinite bar
weakens exponentially as distanee x from the souree of exeitation inereases. The

essential feature of this is that the exponent is proportional to the frequeney of
exeitation. Henee, the high frequeneies generated in the eontaet zone due to
impact remain loealised to that zone . This onee again emphasises the signifieanee
of the lower frequeney harmonie eomponents of the vibro-impaet proeess. They
provide the basis for the transmission of vibro-impaet intluence in structural
elements.
When x = 0 we obtain from (15.63) for the eontacting seetion
. ) = - -Y-
L 20 (0 ,jro 2 -(.
j +%-
2 = - -- J \ (. j +%2
- J
, (15.64)
E 2S 2ro 471: N 2ro 471:
where N 2 = S2 ~ E 2 P2 is the specijic impedance ofthe bar.
When %;: 0 the dynamic complianee L 20 (0 , jro) becomes imaginary, and this
demonstrates the effeet of wave dissipation of the vibration energy through the
semi-infinite bar.
To determine the amplitude of relative vibration a " of the eontaeting seetions
for the system eonsidered using (15.35) and taking into account the value of the
dynamic compliance (15.34) with i = 1 and using (\5.64), we obtain, neglecting
-1
viscosity ofthe bar materials (%1 = % 2 ;: 0) and bearing in mind (15.53):
a, ~ I;:~I;[ ± ;h 1- (2h:~tl;)' (15 .65)
Here, h = G/ I I(EIS I ); X = apl l I(EIS I ); I;; = roll I YI; w = N 2 I NI is the ratio of the
impedances of the impacting bars. This formula
represents a special case of the general expression
(15 .57). Aceording to (15 .65) vibro-impaet regimes
exist when the following eondition is satisfied
Xl h > 2!eoU;;!1 w . (15.66)
Fig. 15.15 shows examples of the amplitude-
frequency response for relative vibration of the
contacting sections for the case X = \, h = I, W = 1 . rr/2
For absolute vibration of these sections we
Fig. 15. 15
obtain from the general conditions (15.6\) and
(15.62)
aJO = _X_ 1+ (2h)2 ± 4h 1_ (2heot l;; )2 , ( 15.67)

!I;;cotl;;! X X XW
a20 =2hl(l;;w). (15.68)
These expressions are shown in Fig. 15.15 for the above parameter values. The
line I in the figure shows amplitude response for vibration without loss of contact.
Henee, in the examples considered, the vibro-impact regime can oeeur only as the
result of hard excitation.
Let us find the conditions for smooth opening of the joint. This occurs if the
§ 16 Vibro-impact regimes of vibrating beams 229
amplitude of the variable force at the joint when

moving without loss of contact exceeds
compressive force G, i.e.
E2S2(8a2xI3xt=o >G, (15.69)
where a 2x is the vibration amplitude ofthe sections
(see (15.18)).
Function u~x (t) describing vibration without
-h (
f
loss of contact, is found from (15.20) when
ß == 0, k ~ 00. Bearing in mind that PI" = al" we
find
Fig. \5.\(, 0 L 20 (x, s)
u?x(t) = \ al' COSOlt .(15.70)
- 1+ L 20 (0,s)L;o(0,s)
For a dynamic compliance LIO(O,jOl) , found from (15.34) when i = I,
L20 (x,jOl) and L 20 (0,jOl) are given by (15.63) and (15.64), with %1 = %2 == 0, we
obtain from (15.70)
(15.71)
Using (15.71), condition (15.69) takes the form
*> 1+( c:sr

Using (15.66) and (15.72) it is possible to estimate the safety of force-Iocked
(15.72)
rods acted on by dynamic loading.
§ 16. Vibro·impact regimes of vibrating beams

1. Weshall use the approach described
1/
above to study bending vibration of a
cantilever beam (Fig. 16.1) of length I ~--I/---..j
produced from a linear visco-elastic material

' 0
and with a point mass M attached to the free
end, arranged symmetrically between stops
with gap ~ [45]. The coordinate x of a section
is measured from the fixed end. The displacement of a section at position x will be
denoted by ux(t), measured from the equilibrium position ofthe beam. The stops
will be considered as elastic with stiffness c 2 , later we shall limit the transfer to
solutions when C2 ~ 00 . Using (6.8) the static characteristics of the impact pairs
take the form
(16.1 )
The beam undergoes bending vibration under the effect of a trans verse force
P,,(/) = a p cosffil applied to the section with coordinate x = n. As a result mass M
interacts with the stops.
Use ofthe concept of adynamie compliance operator for the be am Ln(x,s),
establishes the connection between the trans verse force in the section with
coordinate n and displacement Ur (t), writing the following operator equation for
Ur (I) :
(16.2)
The final term on the right hand side of equation (16.2) represents the
influence of inertia of the point mass on bending vibration.
The approximate periodic solution of equation (16.2) will be sought in the
form
ur(t) = ar COS(ffit - <Pr)' (16.3)
Applying harmonie linearisation to function (16.1), we find
<D(u,) "" k(a, )u, , (16.4)
where according to (7.56) the linearisation coefficient takes the form
k(a,) = ~[",,,o,~-
1t a, a,
!') -(ll )2].
a,
(16.5)
Substituting (16.4) into (16.2) we find when x = I

{I + L,(l,s)[k(a, + Ms 2]}u,(/) = L,,(l,s)P'I(t). (16.6)
Introducing the nomenclature
W(s,a,) = {I + L, (I, s)[ k(a,) + Ms 2]} L: 1(I, s) , (16.7)
we obtain once more an equation oftype (14.12):
W(s,a,)u,(t) = al' COSffit . (16.8)
Bringing this to the equivalent autonomous form using (16.3), we find
[w(s,a,) - ~ (cos<p, + s SI:<P, ) },(t) = (J'l(s)u,(t) = O. (16.9)
Substituting into the characteristic equation (J'l(s) = 0 the root s = jffi, we obtain,
setting the real and imaginary parts to zero:
sin <PI = aY(ffi, a,)/ a" , (16.10)
cos<p, = a,U(ffi, a,)/ a". (16.11)
Here, as before, U = ReW, V = ImW. From (16.10), (16.11) we obtain the final
equation to find a l:
a, = a p [U 2(ffi,a,)+ V 2(ffi,a,)}-Jl2 == a/,IW(jffi,a,)11 (16.12)

After finding ai from equations (16.10), (16.11) we can find the phase <Pi' The
periodic motion parameters for the remaining sections of the beam are found from
(16.2) using (16.3) to (16.5).
To examine stability of the solutions relative to small perturbations it is

possible to use condition (14.32), substituting in it a" = a,.
2. To simplify the final expressions let us assume that the excitation force is
applied at the free end of the be am (x = l), where the mass M is attached. Now to
apply the formula of seetion I we must put LJx, s) = L, (x, s) , and consequently,
to solve the problem it is sufficient to find the dynamic compliance L, (x, jOO) .
With this in mind we will turn to the differential equation for bending vibration of
a homogeneous beam of constant cross-section, appreciating the dissipation due to
the internal damping force with coefficient 'X [176]:
PS2U+(I+~sjE/f~ =0. (16.13)

2TC 00 3x
Here u = u(x, t); P is the mass of a unit length of the beam; E is the modulus of
elasticity; I is the second moment of area of the beam section. Applying a unit
harmonie force to the free end of the beam, we obtain the following boundary
conditions for the system under consideration:
=0 3 2 u(l,/) = 0
u(0 ,I ) , 2 '
3x
(16.14)
3u(0,/)
- --- -0 , - EI(1 +--
S'X j3 J U(l,/) -exp
_ (joot.
')
3x 2rroo 3x 3
Searching for a stable solution for equation (16.13) in the form
u(x,/) = ux(t) = L,(x,joo)exp(joot), (16.15)
we find, substituting (16.15) into (l6.l3), (16.14) when s = joo
d 4 L,(x,jOO) . 2 p .
dx 4 +(joo) ('X jL,(X,jOO)=O, (16.16)
EI 1+ j -
2TC
2
L(O' )=0 d L,(l,jOO)=0
, ,jOO , dx 2 '
(16.17)
dL,(O,jOO) = 0 d J L,(l,jOO) = _ _ _ _ __
dx 'dx3 EI (1 + j'X/(2rr))
Searching for a solution of the ordinary linear differential equation (16.16) of
the fourth order with constant coefficients in the form
LtCx,jOO) = Cexp(qx), (16.18)
where C is an arbitrary constant, we arrive at the characteristic equation of the
fourth degree, the roots of which to first order accuracy in terms of 'X take the
form
q, = ),,[1- j'X/(8rr)], qJ = ),,[-1 + j'X/(8rr)],
(16.19)
q2 = )"['X/(8rr) + j], q4 = -)"['X/(8rr) + j].
Here ).,4 = poo2/(EI). As a result, using (16.18), (16.19) the general solution of
equation (16.16) can be written as:
4
L,(x,jm) = IC,(jm)exp(q,x). (16.20)
Applying the boundary conditions (16.17) to (16.20) we obtain a non-

homogeneous set oflinear equations ofthe fourth order for the coefficients Ci :
Cl +C 2 +C 1 +C4 =0,
Clql +C2q2 +C3q3 +C 4 q4 =0,
C l q12 exp(q/) + C 2qi exp(q)) + C 3 Q; exp(q31) +
+ C Q; exp(Q41) = 0,
4
(16.21)
C 1Q; exp(q1 /) + C2q~ exp(q2 1) + C 3Q{ exp(qJI) +
+C Q3 exp(q I) = _ I
4 4 4 EI() + jx/(2n)
Solving the system of equations (16.21), we find with the help of (16.20)
equations for the operators L, (x, jm), which we have not introduced because of
the awkwardness ofthe final expressions.
To construct the amplitude-frequency response for the point mass from
equation (16.12) it is satisfactory to use the equation for the local dynamic
compliance of the free end of the beam L, (I, jm). Linearising the expression for
L, (I, jm) in terms of the small parameter x using (16.19), we find when x = 1
L,(l,jm) = IJ(coshssins-sinhscosi:;)_
EIs 3 (1 + cosh s cos s)
xl J [~(COSh 2s - cos2s) + (1 + cosh s cos s) x (16.22)
x (cosh ssin s - sinh S coss) + 2ssinh ssin s]

-j--------~--~~~--~--~--~--~~
8nEls J (1 + cosh s cos s) 2
where C;; = AI. Substituting (16.22) into (16.7) and noting that now n = I, we
finally obtain
.
W( Jm, a, ) = EIJ { S J (1 + cosh S cos S) + k 0 ( a, ) _ M 0':>r 4} +
I coshssins-sinhScosS
[%CCOSh2 s - cos2S) + (1 + cosh SCOSS) x (16.23)
. EIxsJ x (cosh ssin S - sinh SCOSS) + 2Ssinh ssin s]

+ J 8ne (coshssins-sinhscoss)2
Here ko(a,) = k(a, )/ J I(E/); Mo = M /(pl) .
From now on, to eliminate awkward calculation it is convenient to use the
frequency functions tabulated in [5] during the solution, which will use the same
definitions with a zero index added:
Yz
U0 (c,) = (cosh C, - cos c,), Bo(c,) = cosh C, sin C, - sinh S cos c"
(16.24)
From (16.12) using (16.23), (16.24) we obtain the following equation for a,
1
a, =XJ[C/Eo(t;,) +k o(a,)+M o(,4]+
Bo«(,)
(16.25)
where X = al'13 /(EJ) .

To find the basic motion parameters of the impacting mass we must add the
equation for ko to equation (16.25), which, using (16.5) is represented in the form
(16.26)
where
Cl = L'1 / a, (0 < Cl ::; 1) . (16.27)
3. Let us construct the amplitude-frequency response for the point mass.
According to (16.25), the expression
(,~Eo«(,o)
Bo«(,o)
+k (a )-M r
0' 0':>0
4 =0 (r o
':>
=~PEOO/~ll' (16.28)
l -,
81 fli
r I
f 1l 1
11
0
Jr4 kD I
1
1
1
1
1
1
I
_ ._ ......L-
I
1
Fig. 16.2
where 00 0 is the frequency of free vibration, defines the configuration of the

backbone curves for the system. It is convenient to use the graphical method
described in § 15, section 2 to show them. Expressions (16 .27), (16 .26), (16.28) are
drawn respectively in Fig. 16.2, a to c (curves 1 and 2 in Fig. 16.2, b represent

different stop stiffness c2 , where the larger numbers corresponds to greater
stiffness). Choosing a point with value a; in Fig. 16.2 a and making the
construction as shown by the chain dotted line, we find in Fig. 16.2, d the
corresponding point of the backbone curve. Here, for each value a ; there is a
countable set of natural frequencies So' Connecting the points obtained gives a
fam ily ofbackbone curves. Fig. 16.2, d shows curves for the three lower modes of
vibration. The portions ofthe backbone curve added correspond to linear vibration
ofthe mass without striking the stops. We will refer to the modes later on by their
roman numerals.
Because of the presence of the interacting mass with the elastic stops the
frequency of vibration for each mode depends on the amplitude, and this
relationship, naturally, has an hard character. As can be seen in Fig. 16.2, d, the
degree to which the nonlinear effects develop is weakened as the number ofmodes
increases. It can also be seen that the nonlinearity of the system increases as the
11l
z
f
fi g. 16. 3
stiffness of the stops increases. The asymptotes of the backbone curve, defined by
the sequence of points in expression (16.26) when a = 0, converge, when
c 2 ~ 00 , to the asymptotes of expression (16.28). These asymptotes are shown
dotted in Fig. 16.2 c, d. They are situated at the anti-resonance frequency of the
linear system. Hence in the limit when c 2 ~ 00 for rigid stops we obtain the
backbone lines of form 3 (Fig. 16.2, d) with the length of the horizontal lines
increasing as the number of modes increases and vertical lines situated at the anti-
resonance frequency ofthe linear system . Of course, in the investigation of actual
vibration only the horizontal portions of the backbone lines described are of
interest.
Putting in (16.25) S = So ' we obtain the energy boundary of vibration, using
(16.28)
8nx B~ (So )
(16.29)
a, = ')(. s~ISoUo (2So)1 + Eo(so )Bo(s o) + soSo(so) .
According to (16.25), the intersection of the energy boundary of vibration 1
(Fig. 16.3) with the backbone line 2 gives the maximum points of the amplitude-
frequency response for the system. It can be shown, that zero points on the energy
boundary of vibration correspond to anti-resonance of the linear system arranged
between stops. It follows from Fig. 16.3 that the maximum of curve 1 rapidly dies
away as the number ofthe modes increases, hence practical values only occur for
some initial modes of vibration.
The remaining points on the amplitude curves, corresponding to non linear
regimes of vibration, are found by jointly solving equations (16.25) to (16.27). For
this, the expression (16.26) for k o(a) is substituted into (16.25), where to
simplify calculation it is convenient to substitute x == 0 in (16.25), since as we
move away from the peak values the intluence of internal damping on the
amplitude curves remains insignificant. Subsequently the expression obtain for
a l = (a'(J is drawn on the (a,a l ) plane (Fig. 16.4) for various values of
1; = const. In Fig. 16.4 this is shown only for the basic modes of vibration for
frequencies 1;1 < 1;2 < 1;3' The intersection of these curves with a l (a) (16.27)
(curve 1 in Fig. 16.4), which was already used to construct the backbone curves,
also produces points for the required values of the expression al (1;) , drawn in Fig.
16.3 for a l > L'1. When al < L'1 a portion of the amplitude-frequency response for
linear systems appears. Their continuation to the level al = Ö is shown dotted in
Fig.16.3.
From Fig. 16.3 it follows that in the examples studied, as distinct from the
linear case, it is possible to obtain larger amplitudes in the second mode of
vibration compared to the first, while they exceed the amplitude of linear vibration
substantially in the second mode. It can also be seen that in the third mode only
linear regimes of vibration can occur without hitting the stops. Note that the
principal qualitative results can also be obtained on the base of the backbone
curves and the energy boundary of vibration.
Let us examine the stability of the solutions found. Assuming for simplicity
that x == 0 and, consequently V == 0 , we obtain the following expression for the
boundaries of the existence domains from (14.33) using (16.23), (16.24), when
S3E o(S) +k (a )-M 1"'4 =0, ( 16.30)

Ro(S) 0' 0'0
s3 Eo(S)+k(a)_Mr 4+ a / dk =0. (16.31)

Ro(S) 0' 0'0 EI da,
Here, we take into account that dU / da, = dk / da,. Expression (16.32) coincides
with the equation ofthe backbone curves (16.28), shown by line 2 in Fig. 16.3. To
draw (16.31) we will calcu1ate the derivative which appear in it. According to
(16.5) we find
dk
( 16.32)
da, rra;
Using (16 .32) expression (16.31) can be written in a form similar to (16 .30):
S3 Eo(s) + k (a) _ M 1"'4 = 0 (16.33)
Ro(S) z 0'0'
where k z (a) is found from the expression
kz(a) = 2c 2 13 (arccosa + aJI- a Z ). (16.34)

rr.EI
As a result, using the graphical construction in Fig. 16.2, we can obtain the
second boundary branch of the existence domain. The corresponding express ions
(16.27), (16.33) are already drawn. It remains to plot expression (16.34) in Fig.
16.2, band to carry out the construction, similar to that which was used when
searching for the backbone curves. The limiting branches found in this way,
described by expression (16.31) are represented by curve 3 in Fig. 16.3. As a
result the regions of instability arranged between curves 2 and 3 enclose vibro-
impact regimes, where the lower branches correspond to the double-valued
domains ofthe resonance curves. The upper branches represent stable solutions.
Let us look at the amplitude-frequency response, obtained in the limit as the
stops become rigid. As was shown above, when c z ~ OCJ, the backbone lines
e . . . __
Fig.16.5
change into line 3 (Fig. 16.2, d). During this limited transfer, pairs of nonlinear
branches of the amplitude-frequency response (Fig. 16.3) join the horizontal
portion of the backbone curves. The resulting characteristics of the system with
rigid stops are shown in Fig. 16.5, where the numbering of the curves is similar to
that ofFig. 16.3.
In Fig. 16.5, the amplitude curve configuration for the first and second modes
of vibration have a similar form and are different from the response of the third
mode. From Fig. 16.5 it follows that in the system examined, pulling can occur in
vibro-impact regimes in terms of frequency as weIl as amplitude, similar to that
shown for systems with one degree of freedom (see §3). However, in the present
system similar effects can arise in several frequency bands, corresponding to
different modes of vibration of the beam. Frequency pulling occurs for fixed stops
with slowly changing excitation frequency as shown by the arrows, for example,
with the second mode of vibration. It can be seen that, depending on the direction
in which the frequency changes, the existence band for regimes also changes. In
the example shown in Fig. 16.5, pulling in the second mode is of greater
significance than in the first. By fixing the excitation frequency, for example at
point d, it is possible to vary the amplitude of the vibro-impact regime by slowly
changing the distance between the stops in range dd J• At point d J the vibration
breaks down.
In the third mode of vibration as weil as linear resonance, an intensive vibro-
impact regime can arise (portion ee J), caused by hard excitation and whose
amplitude exceeds the amplitude of linear resonance.
Hence, the presence of various natural modes of vibration for a point mass on
elastic supports increases the difficulty ofpreventing cIatter against the stops. This
occurs particularly with ben ding shafts where the destabilising effect of internal
damping, occurring at angular velocities greater than the critical, lead to collisions
with the bearings.
4. [n [296, 298, 333, 334] the term distributed impact element was
introduced as applied to the study of vibro-impact processes in the tube bundles of
heat exchangers and similar extended objects. In such systems a significant
variable length of continuous vibrating element (thin-wall tube) may be involved
in vibro-impact interaction either with the neighbouring elements or a rigid wall.
The object of study was an axially stretched beam located in parallel with a rigid
wall that limited the transverse deflections of the beam.
In order to reveal the re sonant behaviour of the system, preliminary
experiments were carried out. Figure 16.6
represents schematically an experimental
rig [333]. A rubber cord was stretched
between the pin of a force transducer and
the pushrod of an electrodynamic shaker
producing trans verse excitation of the
Fig 16,6
cord. A rigid wall Iimited the trans verse
movement of the cord. The pictures of the standing waves were taken by using a
stroboscope synchronised with the voltage applied to the shaker.
As the frequency of excitation exceeded the frequency of the first resonance of
the linear system (the cord without motion limiter) the nonlinear mode of cord
vibration took a trapezoidal form with simultaneous impact ofthe entire plane part
against the limiter. Fig. 16.7 demonstrates this slapping mode at different instants:
impact (trapezoidal shape), static equilibrium (straight line) and backward
'. " ~}
~.
-----...---.-
-------- ~
Fig.16.7
maximum deflection (triangle shape). The intermediate shapes are always

trapezoidal ones with variable length of sides formed by the cord.
The dynamic behaviour of the cord in slapping mode is reminiscent of an
impact oscillator with single degree of freedom. The mode demonstrates
frequency and amplitude pulling, transition from linear to nonlinear mode and
back with the help of hard excitation. However, the cord has many non linear
slapping modes disposed in the right hand side in the vicinity of each natural
frequency of the linear system . The non linear modes have the same quantity of
nodes as a proper linear mode. The frequency bands of non linear modes become
narrower with increase ofmode order.
As it was shown in [333], the method of periodic Green's functions can be

applied effectively for regular analysis of slapping modes. This is based on the
assumption that the resonant motions are close to free vibration in the
corresponding conservative system. The conditions of existence of such modes
can be found with the help of an energy balance.
As an example ofthe approach mentioned, let us consider a continuous system
as shown in Fig. 16.8, a . The
straight uniform beam of length I
is stretched parallel to the rigid
limiter with axial force N. The
following definitions will be used
for description of the system: p
is the density of the beam
material, E, Gare elastic and
shear modulus, S is cross
sectional area, I is area moment
X, :c;
of inertia, E is a small parameter.
Fig. 16.8 The beam is exposed to the
extern al force with density of
distribution EP(X, t) . It vibrates in the paper plane and can collide with a straight
obstacle displaced parallel to the beam with a clearance ~ relative to the position
of static equilibrium of the beam.
Following Timoshenko's model ofbeam bending [207], the following notation
will be used for a description of the beam vibration: u(x,t) is the shape of the
beam transverse deflection at instant t; y(x, t) is the shape of the beam cross
section rotation relative to the neutralline at instant t.
The differential equations of beam oscillations with provision for collisions
with the rigid obstacle may be written in a following form:
8 2u 8 2u 8y'
pS-2 -(aSG+ N)-2 +aSG-+(U)=EP(X,t)
8t 8x 8x
(16.35)
8 2y
K--aSG (8U) 2
Ip -8 =
y-- -- y 0
8x 2 8x S 8t 2
with boundary conditions: u(O,t)= u(l,t) = 8u(0,t) = 8u(l,t) = O. Here K = EI is

8x 8x
the bending stiffness, ais a form factor, (u) is the impact force density.
Considering T-periodic motions with one impact over the one period of
excitation (slapping modes), the impact force density may be presented in the time
domain as folIows:
(u) = J(x)8 7[t - q>(x)] (16.36)
where J(x);::: 0 and q>(x) are the distribution of impact impulse density and the
impact phase respectively.
In order to study the beam vibration in a first slapping mode (as described in
the experiment) the following initial conditions will be applied (see Fig 16.8,b):
r 0. x. x E[O.X,J
"
U-<X,O) = U+(X,O) = U(X,O) =
1-x+t.\+-(l-x).
t.\
0.. x E[xl.I-xll
0.
X E[/-X, ,!]
(16.37)
t
X, Xl
ou_ (x,O) = o, XE [x ,,/- XI]. iJu,,(X,O) = -R OU _(x,O) , (16.38)

ot lO,X~[XIJ-XI] ('/ cl
where R,., 1- E is the coefficient of restitution, plus or minus in the subscript

denotes pre-impact or post-impact values.
As a result:
J(x)= (1 + R)PSVo =)=const. XE[XI,I-X , ]
( 16.39)
q>(x) = const = 0, X ~ [x I' ! - X I ]
For the description of the slapping mode, let lIS consider first the conservative
case, assuming E = O. Using the T-periodic Green's function of the distributed
conservative system, the following integral presentation may be produced for the
T-periodic free vibration with single elastic impact over the excitation period:
-f f X(x,z,t -s)<D[u(z,s)]dsdz,
! 7
u(x,t) = / E[O,!], X E[O,!] ( 16.40)

o0
where X(x,z,t) is a T-periodic Green's fllnction representing a steady state
response of the section x on periodic impulses applied to the section z . The
periodic Green 's function can be expressed by using the transfer dynamic
compliance ofthe linear system L(x, z. jüJ) as folIows:
2 '
X(x. z, t) = - I L(x, z, kjüJ )cos(küJt)
TH
(16.41)
In a number of cases, such as the thin-wall tllbes of heat exchangers, we can

neglect the rotary inertia of the cross sections and ignore. therefore. the last term
in the second equation of (16.35). Under such assumption. the expression for the
dynamic compliance can be presented in the form ofthe infinite series:
TII17X . TImz
Sll1- Sll1- -
_. _ 2 ! !
L(x,,,,jüJ)- --I (2 2)
L
( 16.42)
pSl ",=1 üJ - D",
n r
with the natural frequencies defined in accordance with the following expression
(T aKSG( nm 'NK( 7) +aNSG]
ps[ K( 71' + aNSG

Substituting (16.42) into the (16.41) and using the finite expression for the
Fourier series on the interval of periodicity [191] we find
1 00
sin 1tmx sin 1tmz cos n
I 1 m
(t _~)
2
X(x,z,tj=-L ' XE[O,/j, tE[O,Tj. (16.43)
pSI mol n sin n"t
m 111 2
After substituting (16.36), (16.39) and (16.43) into (16.40) and integrating, we
have
21
, l)m+1 . 1t(2m -1)/ 0 1t(2m -I)x
n2m-I 2
(t -~)
u(x,t j = - - z:
1tpSm=ln
ce (-
(2m-l)
sm
21
sm
•
I
cos---'--...:...
. n T ,(16.44)
2m _ I
sm 2m-I "2
XE [O,Tj, tE (0, T)
In the expression (16.44) the unknown values J and 10 = 2/-x l are functions
of period T of nonlinear vibration. For the first nonlinear impact mode (as shown
in Fig. 16.8, b), from the condition of impact of the middle point of the beam with
coordinate x = / / 2 , which must be involved in the impact, we have
(//2) 2) ~ 1 . n(2m -1)/0 0.?m_I T
,x = - - L
A
U sm cot---- = L\
npS m=1 0. 2m _1(2m - I) 2/ 2
or
) = npSil . (16.45)
2~ 1 . n(2m - 1)/0 0. 2m _1T
L --,---"sm cot---
m=1 0. 2m - 1(2m - 1) 21 2
From the simplest geometrical structure (see Fig. 16.8, b), it follows that
10 = 1- 2x I = 1- ~~O,O) . Considering further the case of small vibration
tan---
8x
( tan 8u(O,O) "" 8u(O,O

8u 8u
)J ' we find finally an equation for defining 1 0 :
/ = lilpS
(16.46)
o J'~ (_I)m+1 .
0. zm _I T· n(2m -1)/0
L---sm cot---
m=1 0. Zm - 1 21 2
By substituting (16.45) into (16.46) we have the following transcendental
equation for the non-dimensional variable 71, = 10 / I :
~ 1 . n(2m -1)71, n0. _
L --..,----" sm cot - -2m-1
71, = 1_3..m=10. 2m_I(2m-l) 2 C;0. 1
(16.47)
n ~(-lrl. n(2m-I)A n0. 2m .1
L---sm cot---
m=1 2m - 1 2 C;0. 1
where S= ~ is non-dimensional frequency of vibration.

In l
For the numerical example, round steel tube of total length 1= 2m with extern al
diameter D= O.OI6m and internal diameter d= O.013m was considered. The
remaining parameters of the model were as folIows: N=2.10 4 N, E=2.1O '1 Nlm 2,
G=O.75.10 11 N/m 2 , p=8.10 3 kg/m 3 , t.=O.Olm, a "" 1. After calculations, we have
pS""lkglm, K=360N/m 2•
The function A(C;;) was calculated numerically from (16.47). By substituting
A(C;;) into (16.45) and taking into consideration that A{C;;)E [0,1] and J(C;;)~ 0, the
frequency band of the first slapping mode was evaluated numerically as
C;; E[I, 1.16]. Figure 16.9 shows the full impact impulse J = l/o as function of S.
As a result of the study of resonant behaviour, a force of excitation distributed
in accordance with the first linear mode p{x,t):= qsin nx cos(wt+<p) was
/
considered, where q is intensity, <p is the phase of excitation relative to impact.
The resonant slapping mode u:= (x,l) was taken in the form (16.44) with values
of A(C;;) and J(C;;) being evaluated numerically according to (16.45), (16.47). Then
the generative solution (16.44) was substituted into the first equation from (16.35),
both parts of the equality obtained were multiplied by the function ou(x,t)
ot
evaluated analytically from (16.44). The equality prepared in such a manner was
integrated along the length of the beam and over the period of vibration
T := 2n / w . As a result the equation of energy balance between excitation and
dissipation was estimated:
ff ci>(u) ou(x,ot t) dxdt [ff) () p{x.t) au~x, I) dxdt .

00
:=
ot
E (16.48)
Because a considerable part of the beam is involved in simultaneous collision, the

energy loss can be calculated as the impact of a rigid body with coefficient of
restitution R:
ffci>(u) ou(x, t) dxdt:= (J/o
00 ot
! 2/[){11-+ RR)pS "" J\ 4pS
_E_.
After substituting in the right side of equation (16.48) the excitation and
integrating, the equation of energy balance was finally yielded
2 qw/ sm . . nlo
<psm-
21 Jl o
--n-r-(n-'~:-_-w72"")CO- 4 (16.49)
The condition for the existence of the process is Isin <pI :;:; 1 . This condition can
be used to define the energy boundary of the slapping mode:
§ 17 Theory of the impact vibration absorber 243
q ~ 7tJ(I- 1;/
8qsin~
fl . (16.50)
2
This defines the minimum excitation to support the resonant mode studied.
The procedure described allows the quantitative analysis of the experimentally
observed slapping modes. Equation (16.45) describes the phenomena of frequency
and amplitude pulling, equation (16.44) presents the trapezoidal shape of the
standing wave (slapping mode), and the equality in (16.50) defines the backward
jump frequency and through it the frequency band ofnonlinear resonance.
§ 17. Theory of the impact vibration absorber

I. Many references have been devoted to the theoretical and experimental
investigation of dynamic systems undergoing impact vibration absorbtion [7, 11,
47,82,85,86,119,105, 145,157 to 159,205,226 et al]. Theoretical analysis
often uses exact methods and is limited to the examination of a double mass
dynamic model, showed in Fig. 6.3, by taking the external excitation in the form
of a harmonic function.
Using approximate methods described in this book enabled us to obtain several
general results, characterising the effect of impact absorbers on linear systems of
arbitrary structures, subjected to different forms of excitation [35]. We shall turn
to consider these results.
Let us look at the dynamic model shown in Fig. 17.1. Here 1 is an arbitrary
linear system, to which a force pet) is
applied at point A. To reduce the vibration
of point A an impact vibration absorber 2
of mass M 2 is attached to it. Various
arrangements of absorber are possible. We
shall look at three variations: "free
absorber" (Fig. 17.1, b), and "absorber
with natural frequency" arranged either
with a gap or with interference (Fig. 17.1,
c), and with double-sided stops (Fig. 17.1,
d). The model described is a special case
hg.!7! of impact interaction of two linear
dynamic systems, described in § 14. Here
system 1 is active while the passive system takes the form of a free mass (Fig.
17.1, b) or of a linear oscillator (Fig. 17.1, c, d). In §14 we used harmonic
linearisation by displacement, so here to demonstrate the use of the method of
equivalent force linearisation, we will concentrate our principal attention on the
analysis of results of ca1culation and to explaining the basic character of impact
vibration absorbtion.
The three variations for attaching the damper correspond to the ideas presented
in §6 (Fig. 6.2, b to d) for the static elastic characteristics <l>(x) relative to the
displacement, x, of the mass of the absorber and the damped point J (Fig. 17.1).
Then the equation for the system with an absorber considered here can be written
in the following form:
Xl (t) = LI (s){P(t) + <l>[x(t)] + ßsx} , (17.1)
(17.2)
where XI is the absolute displacement of point A, measured from the free position
of the system; ß is the coefficient of equivalent linearisation for the dissipative
losses during impact. We can use here the values (7.61), (7.62) found in §7 when
studying harmonie extern al excitation, since when we introduced the formula, it
was stated that for subsequent cases it would only be necessary to adjust the
resonance of the "partial frequency" of the impacting element wh ich, as will we
show later, is a basic condition for the effectiveness of impact vibration
absorbtion. For stationary random excitation we use (8.54) for the value of ß, or
double for a symmetrical impact pair.
Let us linearise function <l>(x) using the formula
<l>(x)",m F +k(x-mJ=m F +kx°. (17.3)
Substituting (17.3) into (17.2) and noting that the left hand side of (17 .2) contains
no constant component, we find
mF=O. (17.4)
Furthermore, from (17.1), (17.2) using (17.3), (17.4) we obtain a system of
linear equations relative to the generalised coordinates XI (t) and x(t). Suppose
that P(t) = mp + pO(t) (mI' = const) , we can separate this system into constant
and centred components
O()= (M 2s 2 +ßs+k)LI(s) pO()
XI t 2 2 t, (17.5)
[1 + M 2 s LI (s)](ßs + k)+ M 2 s
X°(t) = 2M 2s 2L I(s) ? p0(t) ,

(17.6)
[1+M 2 s L I (s)](ßs+k)+M 2 s-
mXt = LI (O)m J,. (17.7)
Using (17.6), we find the equations for the centred components of process
F(t) = <l>[x(t)] :
(17.8)
2. Let us suppose P(t) = ml' + al' cos( W( + <p). Then for various methods of
attaching the absorber we can use corresponding formula for the harmonic
linearisation coefficients obtained in §7, section 4, bearing in mind that k = llq
and substituting condition (17.4) in them directly.
At the same time, according to (17.8) we find for amplitude a F of process
F°(t)
M 2ro2 k(a F )IL I(jro)la p

a - (17.9)
} -I[l-LI(jro)M2ro2][k(aj.)+ßjro]-M2ro21·
Hence, in the basic case, equation (17.9) together with one of the equations for k
allows us to find the value of a F , with the help of which we can then find all the
remaining parameters for the process in the system examined. When using the
formula for k in the forms (7.85), (7.91), (7.95) the equations for aI- can be solved
directly.
After finding a F , the amplitude of vibration of point A in the fundamental is
found from (17.5)
Ik(a}) M + ßjro -
)Ia p 2ro211LI (jro
a =----'.-------':..-----'---~
x, 1[l-LI(jro)M 2ro 2][k(a j .)+ßjro]- M 2ro2 1
or, using (17.9)
Ik(a F ) + ßjro - M 2ro21
ax = , ar · (17.10)
, M 2ro-k(a j.)
Supposing that the absorber parameters have been chosen so that
k(a F ) = M 2(02, (17.11)
then for the corresponding frequency ro we find from (17.10)
a x , =ßa x /(M 2 ro) (a x =aj./k). (17.12)
Hence, similar to the linear dynamic vibration absorber [207], the best tuning
of the impact vibration absorber corresponds to the equality of the partial
frequency of the absorber
roo =~k(aF)/M2
and to the frequency of excitation ro. The "residual vibration" ofthe first harmonie
is proportional to the loss of energy in the absorber. However, with impact
absorbtion the equivalent elasticity k depends not only on the system parameters,
but also on the parameters of its own regime, i.e. feedback exists. Thus it is
possible to use impact damping to damp vibration with changing frequency.
Substituting the expression for ß from (7.61) or (7.62) into (17.12) and bearing
in mind that during impact vibration damping M ::::; M 2' since M 2 «MI (MI is
the mass ofthe protecting system J reduced to point A), we obtain
(17.13)
where p = 0.32(1- R 2 ) for a double-sided impact pair and p = 0.16(1 - R 2 ) for a
single-sided pair. Hence complete damping of the fundamental is possible only
when the coefficient of restitution R = 1. Taking into account that, according to
(17.9) a F =a p and substituting this value into (17.1), we can obtain the
expression for a p (ro), with which complete damping of the fundamental can be
obtained in all frequency ranges, which satisfy the existence conditions of the
vibro-impact regimes. For example, for the free absorber, using (7.85), (17.4) we
find
a p =4M 2 L1m 2 /1C, (17.14)

i.e. it can effectively be applied to damp vibration in regimes with slowly
changing frequency due to acceleration or braking of unbalanced masses, since the
excitation amplitude here changes according to the law a l , = I:m 2 . Substituting
this expression into (17.14) we obtain the condition for tuning the damper in the
given case
M 2 L1 = ~ 1C1: "" 0.781: . (17.15)
According to (17.15) the mass and gap equally influence the tuning of the free
damper. This situation is confirmed by experiments [7].
In engineering practice, the value of the exciting force P(t) is not, as a rule,
known in advance, but only the vibrational motion of point A (Fig. 17.1, a)
without damping. Assuming this is given by acos(m( + \11), we find
a
ap = ILI(jm)l'
In this case the formula for the best arrangement of the free absorber, using
(17.14) takes the form
1C a
M 2 L1=-1 (17.16)
4 LI (jm) m-l ' 1
Let us continue with the study of dynamic systems with free damper (Fig. 17.1,
b). Using (7.85) we find
a< =qa r =4L1/1C. (17.17)
Substituting (17.17) in (17.13), we obtain
a x , =4pL1/1C.
Hence, for the worst ca se when R = 0 we find a" "" 0.4L1 .

To find the existence conditions of solutions and to construct the amplitude-
frequency response for systems with free damper, substitute (17.17) and (7.62)
into (17.9). We find
[1- M 2m2 LI (jm)](}PM 2m2 +

1
~aF)
4L1
- M ,u/
-
1= M 2m2 LI
1 (jm)l~ap
4L1
.(17.18)
Assuming L I (jm)=IlI(oo)+ }vl(oo) and solving equation (17.18) for a F , we

find
where
B = 7t{[I- M 2oo\1 1(00)]2 +
+M iW4v~ (W)}{ C:~2 aJ, - p 2M~W4 }~~ (w) + v~ (w)] +

+ 2M2W2p[~II(W) + pvl(w)] - p2}_ Mim 4yf(m).
In reality, of course, only positive values of a l have a real sense. Analysing
*17 Theory of the impact vibration absorber 247
(17.19) it is easy to show that when

M 2 C02 1l,(CO) < 1, B>O, (17.20)
rc 2 2
1--- 2 a~[Il~(co) +v~(co)] +p {I + Mico4[1l~(CO) +v~(co}]}-
1M
- 2M 2C02p[1l1 (co) + pv, (co)] > 0
both values of al- are positive, but when
rc 2 2 2
2 a~[Il~(co) + V , (CO)] + p {l + Mico4[1l~(CO)+ v~(co}]}-
1---
1M (17.21)
- 2M zco zp[ll, (co) + pv, (co)] < 0
only one positive value of al- occurs. Ifneither ofthese conditions is not satisfied,
then the solution sought does not generally exist. We can assume that in this case
vibration at frequency co can not occur in the system with two impacts per period.
Suppose v, (co) ;: 0, i.e. energy dissipation in the damped system is negligible.
Then in agreement with (17.19)
B = [1 - M 2C02 1l1 (co )]2 {~a~ll;

1M
(co) - p2 [1- M 2C02 1l1 (co )]2}
and, consequently,
al-= 4 L'lM 2co

2
2
rc[l- Mzco Ill(co)
1 rc 2 2ll l2(co)-p[I-M
1± --,ap
1611-
2 2CO 2IlJco)] 2) (17.22)
Conditions (17.20), (17.21) take the corresponding form

M 2C021l1 (co) < 1,
pll- M 2COZ Il I (co)1 < 4: apllll(co)1 < Jl +pz[l_ M 2C02 1l1(CO)]2,

(17.23)
~apllli (co)1 > Jl + p2 [1- M 2C02 1l1 (CO)]2 . (17.24)

4L'l
For the amplitude a x , ' the following expression is obtained:
4L'l
a = x
x, I zco 2 Il,(co)I
rcl-M
z
rc 2 2 z
r
2 2 2 )
--2 aplll(co)-P [1-M 2 co Il,(co)] + (17.25)
1M
+p'[I- M,oo'~,(oo)]'
The condition for maximum damping of vibration now appear as:
2
M2C021l1 (co) ± ~a;'Il~(co) _pZ[I_ M2COzll,(CO)]z = O. (17.26)
1M
If condition (17.23) is satisfied, then according to (17.26) the maximum

damping can only be achieved in one of the possible regimes. In the case when
(17.24) is satisfied damping is achieved if /-11(00) < 0, since in this case only the
solution corresponding to the positive sign in (17.25) is valid.
Thus the final solution to the behaviour of the system considered can be
established only after studying the solution stability. However, before turning to
look at stability, we shall, as an example, look at the simpler problem of impact
vibration damping of a suspended mass MI as an example (Fig. 6.3). In this case
when n,=0,Lj(s)=(Mjs2+c,r'. Then IlI(eü)=(c l -M l ül 2 )1, and from
(17.25) we obtain
4~
a = x
x, nil - (1 + /-1)E, 21 (17.27)
x~'{/-1-S-2-±-~rX=2=_=p2=[=1_=(=I=+=/-1=)s=2]~}-2-+-p2-[-I--(-I-+-~l-)s-2-f,
where /-1=M 2 /M I ; x=4~aP/(nc,); S2 =M I 00 2 /C I .
According to (17.24), (17.23) when

X > ~r(1-_-S-2-)2-+-p-2-[1---(-I-+-/-1-)S-2-]
2
only one solution is possible, and when

/-1S 2 /(1-s2)<1, pI1-(1+/-1)s21<x<~(1-s2)2+p2[1-(1+/-1)s2]2
two solutions are possible. Fig. 17.2 shows curves for the amplitude-frequency
response for the damping mass M, when p = 0,
/-1 = 0.1, X = 0.2. The positive sign in (17.27)
corresponds to the full line, the negative sign
to the dotted line. The latter can not exist in
i domains where 11 - S 21 < X. Later we will
,/" ! ...
/1 _- show that the solution corresponding to the
dotted curves is unstable.
~ i I ~ r-.... --- 3. We shall now turn to look at stability
P (/-5 'P ,.$ 2'P ~z
~ of approximate periodic solutions for the basic
Fig.17.2
model of the system studied (Fig. 17.1, a),
using for this the results of § 14, section 2. In
agreement with condition (14.32) and bearing
in mind that now a" = ax , to solve the stability problem we need to find the
reduced dynamic stiffness W, connecting the external force P(t) with the
"deformation" xCt) of the linearised impacting element. According to (17.6), this
characteristic takes the form
W( . .) _ [1- M 2 00 2 LI (joo)] [ßjoo + k(a F )] - M 2 002
loo,a F - 2 . (17.28)
M 2 00 L,(joo)
When using harmonic linearisation by force the frequency response (17.28)
depends on a F hence to use condition (14.32) it follows that
dU(ro,G/.) = dU(ro,GF ) aal' = dU(ro,GF ) k

(17.29)
aax aaF aax aal' '
dV(ro,GF ) = dV(ro,G/.) aal' = dV(ro, GI') k (17.30)
aax aaF aax aal-
( U = Re W, V = Im W ). Substituting (17.29), (17.30) into (14.32) and noting that
Gxk = GI- , we obtain corresponding conditions for stability of the solution, using
the harmonic linearisation by force method
[u(u + GI' dU)+

aal-
v(v + GI- dV)]
aal- _ aF='''F
> O. (17.31)
When energy dissipation in the system is neglected, V = 0 and condition (17.31) is

transformed to the form
(17.32)
In (17.28), putting ß = 0, LI (jro) = !! J ro) , we find

U= [1-M 2 ro 2 !!I(ro)]k-M 2 ro 2
(17.33)
M 2ro 2!!1 (ro)
As a result, for example, for systems with free damper (k = na I- j( 4/l) ), when
we substitute (17.33) into (17.32), we obtain the following necessary conditions
for stability:
(17.34)
The first ofthese conditions includes the second.

From (17.22) when p = 0 we find the value of a/. wh ich corresponds to the
regime studied:
(17.35)
Substituting (17.35) into the first ofthe conditions (17.34), we obtain eventually
± ~apl!!l (ro)1 > O. (17.36)

4,1
Hence, regimes which correspond to the negative sign in the solution of the
dynamics problem are unstable.
In the application to the example studied (Fig. 17.2) this means that the dotted
curves correspond to unstable regimes and, consequently, vibration damping is
obtained only when passing through the natural frequency of the damped system.
In pre-resonant domains it is possible to get strong oscillation of the system at a
frequency ro = ~ CI j(M I + M 2) • These results agree with known results, obtained
by using exact methods [84, 85, 86, 157 to 159] and with experimental data [7].
4. Let us now look at the effects of random excitation on systems with an
impact vibration absorber, i.e. we will assurne that P(t) = mp + pO (t) where
pO (t) is a stationary normal random process with a given spectral density S p (00) .
Then, for different forms of absorber attachment, we can use the corresponding
formula for the statistical linearisation coefficients found in §8 section 4, bearing
in mind that k = l/q and directly substituting in them the condition (17.4).
From (17.8), we obtain also
cr;. =_1 1 M;oo4k!L j(joo)!2 Sp(oo)doo 2 (17.37)

21t -00 1[1- M 2002 Lj(joo)](ßjoo + k) - M 20021
The relationship (17.37), together with one of the equations for k allows us to find
the value of a F' which we can use to find the remaining characteristics of the
process in the system studied. In particular, for cr;1 we find from (17.5)
a;1 =_I_ll(k+ßjOO-M2OO2)Lj(joofSp(OO)dOO2
(17.38)
21t -00 1[1- M 2oo2 Lj(joo)](ßjoo + k) - M 2oo21
As an example, we shall look at the effects of white noise
( S p (00) = So = const ) on a single mass damped system, supplied impact
absorbtion (Fig. 7.3). In this case, L j (s) = (M S 2 + n, S + c, r' . As a result, from
j
(17.38), assuming for simplification R = 1 (ß = 0) and using the tabulated values

for similar integrals [184] we find
2 _ I "'f (k - M 2 (02 )2 Sodoo So
cr - (17.39)
XI -21t-001(Cj-M,002+n,jOO)(k-M2oo2)-M2oo2k2r 2n,c,'
i.e. the mean square value of x j (t) does not depend on k and coincides with
expressions for the mean square value of linear single mass systems without
absorbtion (see (13.18)). Hence, in the case considered the mean square value of
the damped system is not changed with any "elastic" attachment ofthe absorber.
The reason for this apparent paradoxical effect can be explained by the nature
of dynamic damping. During harmonic excitation the absorber, thanks to its
natural non-isochronous nature, automatically adjusts its own partial frequency to
the frequency of excitation. When there is practically no such dominant frequency,
for example when excited by white noise, then the influence of the absorber can
'*
only dissipate energy when ß 0 . Such a method for damping random vibrations
would appear to be ineffective, since significant losses in the absorber will be
possible only by maintaining intensive vibration of the damping mass.
Chapter 5. Structural synthesis of self·sustaining vibro·impact
systems
The seientist deseribes what is; the engineer ereates what never was.
Theodore von Karman
§ 18. Optimal periodic motions of vibro-impact systems

1. In § 12 we have already analysed systems in which the action of a force
on a striker appears as a function of its phase
coordinates only. Similar methods of realising periodic
motion have been widely used in percussion machines
wh ich use electrical, pneumatic or hydraulic methods of
exciting vibration. Such machines include pneumatic
chiseis and perforators, hydraulic or pneumatic rotary
impact drills, and various types of test machines or test
rigs. Applying self-sustaining vibration gives stability
to the process compared to the application of forced
vibration and often leads to a more simple form of
design.
The general schematic for self-sustaining vibro-
impact systems can be represented by an impacting Fig. 18.1
mass J (Fig. 18.1) with motion x(t), to which is applied
a varying force of excitation u(t) = B[x(t),x(t)] trom an unregulated source of
energy 3, using a controller 2 having characteristic B(x,x) as the result ofwhich
the mass undergoes systematic impacts against an anvil 4.
The form of the controller characteristic B(x,x) defines the dynamic
behaviour of the system to a considerable degree, making it possible to achieve
any type of motion which satisfies the pre-defined technological and energy
requirements. Hence, this remaining chapter is devoted to the problem of synthesis
of B( x, x) , to produce the required type of periodic system motion under optimal
conditions.
In practice, such a problem is not very different trom the known problem of
the analytic design ofcontrollers [144]. However, using the traditional approach in
this case, based on the design of a stabilising controller to compensate the
mismatch between the ideal motion required and the real observed motion, does
not appear to be applicable for vibro-impact systems because of the great
complexity of such control systems.
252 Chapter 5 Structural synthesis of self-sustaining vibro-impact systems
The specific features of vibro-impact systems provide the opportunity to

suggest another approach for synthesis [28, 30 to 32, 29].We seek characteristics
for the controller wh ich will enable it to satisfy two basic ftmctions
simultaneously: to achieve the required optimal working motion for the system
and to compensate extemal disturbances. Moreover, the required reaction of the
system can be pre-programmed also in the structure of the controller within
defmed limits, as parameters of the system drift from the nominal. These
combined effects can be obtained within an autonomous system, which ensures
simple and safe engineering arrangements.
The specific feature of vibro-impact systems mentioned is that they should
undergo periodic motion, in such a way that the working process (impact) occurs
during short time intervals, which are considerably shorter than the period of
vibration. This allows us to control the periodic regime of the striker using relay
type devices, applying the defined level of force excitation in the given phase
plane domains for the striker. In this case, synthesising the form of the controller
characteristic B(x, x) reduces to finding the values of the corresponding control
effort Vi and to finding the switching curves IIi(x,x) = o.
Later, we will show that for a wide class of functionals which represent the
quality of vibro-impact systems, optimal control, u(t) = B[x(t),x(t)), prescribed
fulfils the required type of periodic motion for the system x(t), is a piecewise-
constant function of time with finite number of switches per period
(discontinuities). Thus, the search for time control function u(t) and the
corresponding system motion x(t) allows us to find aseries of points for the
switching curve on the phase plane, through which the phase trajectory of the
optimal limit cycle for the system passes, and the correct control levels V" which
support the portion of the trajectory between the switching points (Fig. 18.2). The
points indicated have coordinate {x(t;),x(t, )} where t i is the instant at wh ich
u(t) is switched, and completely defines the
optimal periodic regime in combination with the
given control levels. By adding points defined in
this way up to the switching curve, we can
guarantee stability of the regime and supply
additional useful qualities to the system wh ich
characterise its reaction to a variation of the
parameters. Q
The search for optimal control function u(t) for fig.18.2
vibro-impacting systems has specific features [27],
since control, which is strong and non linear, is sought for the periodic motion of
the system. By taking the system as linear in the interval between collisions, we
can effectively use the mathematical theory of moments [32, 188] to find the
required control function, based on the work of M G Krein [2], and applied to
optimal control by N N Krasovsky [137] and R Kulikowsky [139]
§l8 Optimal periodic motions of vibro-impact systems 253
2. Let us formulate the mathematical problem of synthesis described above.

In the interval between impacts the equations of the system (Fig. 18.1), reduced to
unit mass ofthe striker can be written in the following form:
x=R(x,x)-g, (18.1)
where R(x,x) is the control characteristic sought; g is the acceleration due to
gravity. Impact, originating when x = 0 is described by Newton's law
x+ = -Ri_ RE [0,1]. (18.2)
Taking into account the !imitations of the actuator on the value of the forces
developed, we naturally assume that
IR(x,x)l:s; U = const. (18.3)
As already mentioned, the form of motion for a striker is limited in practice to
a periodic function with period T. In addition, during the course of one period a
finite number of impacts q of required periodicity and intensity can occur,
depending on the technological requirements of the system. Such a situation
arises, for example, in actual impact processes where each impact during a cycle
represents a separate operation.
AIthough the problem of producing periodic motion by itself is complex due to
the character of such nonlinear systems with ambiguous behaviour, it is
convenient to solve the problem of synthesis on the basis of some optimised
criteria.
In cases when the working cycle ofthe proposed system is strictly defined, i.e.
given the number of impacts per period, their intensity and time distribution,
naturally one tries to obtain such a process using a minimal intensity drive, and
choosing as the minimised value of functional I the value
1= U . (18.4)
For the majority of vibro-impact machines, working process is single impact
(q = 1) and the regime parameters are not strictly prescribed. In this case we can
pose the situation of a maximum intensity impact process with limited effort from
the excitation as given by (18.3). The natural intensity of the impacting process is
represented by an impacting velocity x_ and impact frequency. As a rule, one of
the properties is defined from strength or technological consideration, hence the
problem of optimisation can be formulated as a maximisation
I=x (18.5)
with the given values U and T, or minimising
1= T (18.6)
with the given va lues U and x_ .
Note that with achanging coefficient of restitution R, a more objective
characteristic of intensity of the impact process will be the value of the impact
impulse J, transmitted to the anvil by the striker, or to the corresponding value of
the energy E. Using the linear relationship between impulse and impact velocity
J = (1 + R)x_ , this is preferable in calculations, since later we will also look at the
problem of maximisation
I=J (18.7)
given the values of U and T. The value JfT is a measure of the force and can be
used to demonstrate the average efficiency of the machine.
Thus, the problem of synthesis can be put in the following form. To find the
characteristic B( x, x) wh ich satisfies (19.3), producing q impact processes with
period T in systems described by (18.1), (18.2), and optimising the quality
functional L defined by one olthe expressions (18.4), (18.5), (18.6) or (18.7).
3. Denoting B[x(t), x(t)] = u(t), we can rewrite (18.1) in the form
x
= u(t) - g (18.8)
and we will look at the preliminary problem of finding the optimal control
function u(t), realising the periodic motion ofthe system given in section 2.
We will limit the analysis to single-impact motion (q = 1), considered, for
generality as a number I sub-periodic regime, i.e. a regime with one impact per I
periods of excitation. In this case equation (18.8) acts during the interval
t E (O,IT) , where the instant after impact is taken as the starting point for time
measurement. It is shown in [225] that for the functionals examined, single impact
motion exists as the optimal among all periodic vibro-impact processes. The
boundary conditions at the instant of collision with the stationary stop using the
conditions ofperiodicity will be as folIows:
x(O) = x(lT) = 0, x(O) = x(lT). (18.9)
Rewriting equation (18.8) in the Cauchy form [142]
I
x(t) = x+ - gt + fu('t)dt,
o
(18.10)
t2 I
x(t) = x(O) + tX+ _L+ f(t-'t)u('t)dt.

2 0
Applying condition (18.9) to (18.10) we obtain, using (18.2)

/1' l'
(1 + R)x_ + glT = fu('t)dt = I fu('t)d't, (18.11)
[2T2 /7'
RITx_ +~ = f(lT -'t)u('t)d't. (18.12)
o
Changing the interval of integration in equation (18.12) to the period of control
't E (0, T) . We find
/7' I_I (;+1)7'
fuT - 't)u('t)dt =L f(lT - 't)u('t - iT)dr,

o ;=0 ;7'
due to the periodicity of the control function u('t) = u('t-iT) (i =0,1, ... ).
Introducing a new variable t = 't - iT , we obtain
§ 18 Optimal periodic motions of vibro-impact systems 255
H O+I)r I_I r
I JetT - t)u(t - iT)dt = I SUT - iT - t)u(t)dt =
;::::0 iT j::::Q 0
r r
SI (lT - iT - t)u(t)dt SI (iT - t)u(t)dt =
I_I I
= =
0;=0 01=0
= I[ IU ; 1) T -lt }(t)dt.
As a result conditions (18.11), (18.12) can be rewritten in the following form:

7'
(1 + R)x_ + glT = Slu(t)dt,
o
(18.13)
RITx_ + g(lT)2
- - = rSI - T -t (t)dt. (/+1 }
2 0 2
Let us discuss from first principles the problem of minimisation of functional
(18.4), assuming x_ and T have given values. On the left hand side of (18.13) we
have a known constant aJ• Regarding the functions
hl (t) = I, h2 (l) = {I; 1 T - t) and u(t) on the right hand side of equation (18.13)
and defined in the sector t E [0, T] as the vectors of some normalised space, we
note that the right hand side of expression (18.13) represents a scalar product of
the unknown vector u(t) with the given vector
I h;(t) (i = 1,2) .
Choosing as a norm ("length") of the vector in
fß* space for the function u(t) with the value
U = IE[O,T]
maxlu(t)I, (18.14)
we obtain that finding the optimal control function
u(t) from expression (18.13) can be interpreted as
the problem of finding vector U O(t) E .'1:1 *, having
Q~----------~~
~ "lt) the given projection a , on vectors h; (t) and
Fig. IX3 possessing the minimum norm (Fig. 18.3). From Fig.
18.3 it can be seen that vector U O(t) is unique and is found among a number of
vectors u(t) the ends of which are arranged along the line of intersection of the
planes orthogonal to the given system of vectors h; (I), which must be linearly
independent. In [137] it is shown that in addition, the space .'13* of function u(t)
must be "conjugated" to the .CJ3 space of function h, (t) so that choosing the space
,UjJ* in the form (18.14) gives the norm ,CJ3 in the form
(18.15)
i.e. the functions h(t) must be limited by the norm (18.15), and this is always
ensured in problems examined later.
4. The problem described above can be considered as part of a more general
mathematieal problem. Let a set of m express ions exist
r
U; = fh;(t)u(t)dt (i = 1,2, ... ,m), (18.16)
where u, are known real numbers; h(t) are linear independent functions, defined in
the sector tE [0. T], is limited in the sense of (18.15). It is required to find a
function u(t), satisfying (18.16) with minimum modulus (18.14).
The problem formulated belongs to that known as a moment problem, which
has recently received wide application in the theory of optimal contro!. The
complete deduction and proof of the existence of solutions to this problem will
emerge in the framework of the present exposition since it is based on the
application of functional analysis, with which the reader is not assumed to be
familiar. For this reason only a simple explanation is given here, leading to a
solution ofthis partieular problem.
Suppose n; are arbitrary real numbers. Multiplying eaeh of the expressions in
(18.16) by n; and adding these products, we obtain
m Tm
"Lu,n, = f"Ln,h;Ct)u(t)dt. (18.17)
i=l 0 1=1
Since the numbers n, are assumed arbitrary, then they can be chosen, in
partieular, so that
m
"Lu;n, =1. (18.18)
i=1
Then from (18.17) we find

T m
1 = S"Ln;h,(t)u(t)dt. (18.19)
o ,~I
Applying to the right hand side of (18.19) an obvious inequality, we obtain
I = f~n;h;(t)U(t)dt:S:; fl~n;h;(t)U(t)ldt:S:;
(18.20)
:s:; ma)(lu(t)IISltn;h;(t)ldt.
tE[O.1 J 0 ,~I
From which we find
(18.21)
Inequality (18.21) must be satisfied for any number n" satisfying (18.18) and
especially for sueh n~ , for which the right hand side of (18.21) takes a maximum
value
maxlu(t)1
tE[O,T]
~ [min f!i n;h; (t)!dt]-I
11; 0 ;=1
(18.22)
From (18.22) it follows that the minimum value of the maximum lu(t)1 found
among all possible functions u(t), satisfies the condition
V o = min ma"lu(t)1 = [f!in;Oh;(t)!dt]-1 (18.23)

11 IE[O,l] 0 ,=1
Now it is necessary to find the function u(t) in such a form that the series of
two inequalities (18,20) can be converted to an equality, since this is a necessary
condition to satisfy expression (18.16). It is not difficult to prove that this takes
place when and only when
m
u(t) = Vsign L n;h; (t), (18.24)
1=1
where V is a constant. Really, the first ofthe inequalities (18.20) can be converted
m
to an equation, when and only when the function Ln;h;(t) and u(t) take
;==1
identical signs at every instant of time, and secondly when and only when u(t)
m
takes the same sign as Ln;h;(t) , and also, a constant value
lu(t)1 = V for all tE [0, T] . Obviously, if V was not a constant then it can not be
factorised out of the integral and the second inequal ity can not be converted to an
equation.
Thus to satisfy (18.16), function u(t) must take the form (18.24) but since we
want the value V = maxlu(t)1 to be aminimum, i.e. equal to V o , then it must be
1
found from (18.23), where the number n,o is calculated from the solution to the
problem: to find
(18.25)
under condition (18.18). Consequently, on a basis of (18.23), (18.24) the solution

ofthe resulting problem takes the form
m
sign Ln~ h;(t)
u(t) = ;=1 (18.26)
II~n~h;(t+t
5. We shall apply the results of section 4 to find the optimal control
function u(t) for vibro-impact systems. From (18.26), using (18.13) we find
(18.27)
where the numbers n~ and ng are solutions of the expression (18.25): to find
. ~ n [I(l
mm 1 I 1
I) - I I ] +n Idl=-
- -+T 2 ( 18.28)
1/1,1/2 0 2 Va
with the condition
n 1[ RITx_ + g (1:)2 ] + n 2[(l + R)X. + glT] =I . (18.29)
In the case under consideration it is not necessary to follow the standard

procedure, since the solution can be obtained more simply using only the
information from solution of the moment problem that, according to (18.27), the
optimal control function is discontinuous with one switching in the interval
1 E (0, T) , since the 'sign' is applied to a linear function. Thus,
+ Va when I E [0"1]'
u(t)= { (18.30)
-Va when IE[/I'T]
(the opposite combination of signs is excluded for the obvious mechanical
reasons).
Substituting (18.30) into (18.13) we obtain after integrating the system of two
finite equations for finding V o and 11 :
(l + R)x_ + glT = (2/ 1 - T)lU o'
(lT)2 [/+1
RITx. +g--= T2 ]
- ( 2 / 1 -T)T-/ 12 +-lU
(18.31)
a.
- 2 2 2
Solving these, we find
V = (1 + R)x_ + glT
(18.32)
a (2/ 1 - T ) / '
11 =a±~a2 -aT+ YzT 2 , (18.33)

where
12 T(1 + R)x_ + IT(l- R) + g/3T 2
a= --'----'-----'----'---'='---
2[(l + R)x_ + glT]
The sought solution is only satisfied when the values I E (0, T). If both roots 1
I1 E (0, T) , then it is necessary to choose that corresponding to the smaller value

of Vo in (18.32).
The equations (18.31) allow us to find the optimal control function U(/), giving
the maximum for (18.5) for the given va lues of Vand T. To eliminate from x
(18.31) we find a quadratic equation in terms of t 1:
t; - ( 1+1--
l-R) Tt
I+R
l
l-R
+1--(l+y)T
I+R
2 =0,
where y = g / U 0 .
Solving this, we find
t
I
= ![1
2
+ 1 1- R ± 1_ 2y/ l-R +/ 2(I-R)2].
I+R I+R I+R
(18.34)
Limiting the value of y trom the inequality y < 1, we find trom (18.34) that for
a positive sign t l > T , hence the conditions for the problem are satisfied only if
the solution (18.34) agrees with the negative sign. Using this we find trom (18.31)
Ix-I=
.
[
y+
1- R 2 1- R
1-2yl--+1 ( - -)
I+R I+R
2 1
1 - R ITU 0
-1-- - - .
I+R I+R
(18.35)
From (18.35) it can be seen that the maximum lx_I occurs for the maximum
possible value of Ua, i.e. when Ua = U.
For the problem of maximising the impact impulse, we find in agreement with
(18.35)
max J
11
= (1 + R) maxlx _I =
11
l-R 2 l-R l-R (18.36)

= [
y+ 1-2y--+1 ( - - ) 2 - 1 - - ] ITU.
I+R I+R I+R
In the same way we can find expressions for the problem of optimisation of high
speed action, by minimising condition (18.6).
Note that when solving the problem of
optimisation the condition x(t) ~ 0 was not
used, leading, in general, to the intringement of
the fact that a unique optimal solution exists.
However, depending on the form of the solution
obtained, it is easy to show that they always
satisfy the required conditions.
t
Fig. 1804 shows examples of optimal
excitation u(t) and the corresponding motion of
the striker when 1=2, R = 0.5, Y = DA, U = 1 .
Expression (18.36) is drawn in Fig. 18.5 for
various values of R, y, I. It can be seen that
increasing the value of 1 can lead to an increase
in the energy of impact. t
Fig. 18.4
Note that for the case when I == 1 reference [188] gives a solution for the
problem considered by optimising condition (18.4); in [16] the optimal control
I/V! r{}* I/ur R-O
~-(}8
2
/
~
(}f
f
0
o {}5~___~______~___~
1 Z S t f 2 SI
Fig. 18.5
function was found for the maximum speed of systems when I == 1, R == 0 using
special procedures; in [110] the problem is solved for a striker with elastic
suspension.
The solutions obtained above for the problem of optimal control will be
applied later to the synthesis of regulators, however they are also of separate
interest, since they can be used in etalon during the design, adjustment and
evaluation ofmachines with impact effects. Thus, for example, references [16, 16]
searched for the best parameters to construct hammer drills with impact effects
using similar calculation principles. Here, the optimallaws of motion were used in
etalon, and in reference [140] the optimal law was used in a similar manner to
synthesise the parameters for various mechanisms with impact effects. Finally,
note that frequently the optimal control u(t) can be used in the form of periodic
excitation to produce forced vibration of vibro-impact systems. In this case, the
question of stability of the optimal motion still remains. The problem of stability
of vibro-impact systems excited by arbitrary periodic forces has been studied by
the stitching method in references [30, 69].
§ 19. Synthesis of self·sustaining systems

1. The above solution for optimising striker excitation makes it possible to
synthesise self-sustaining systems to give the
optimal motion for the striker. This can be
achieved by eliminating time from the optimised
functions found for x(t), x(t) and u(t) (Fig. 19.1).
The problem of synthesis now becomes a
question of finding a function B(x,x), which
coincides in time with the optimised control
function (18.30) B[x(t), x(t)] = u(t) as the
system moves periodically. This function will be
determined completely, given the switching curve
II(x,x) = 0, which divides the phase plane ofthe
system into two domains, corresponding to the
different control levels ±U . It is obvious that the
switching curve must pass through the point with Fig. 19.1
§ 19 Synthesls of self-sustaining systems 261
coordinate {X(tI),X(tI)} (Fig. 19.1), which we shall call

the nominal point, since at this point control switching
occurs in periodic regimes of systems whose parameters
are nominal. o;t-- - - ' o - -- ---4-
Let us look at the equations for the phase coordinates
of the nominal point, obtained from the solution to the
problem of finding the control function u(t), optimising
the system for various criteria (see § 18).
a) maximising J or x_, given the values of U and
T:
x(t ) =I [x(t 1) _ 1 - Y!J..-]UT (19.1 )
1 1 UT 2 T '
.
X(/I)= - - I--+y1 [R ( 2(I 1 ] UT,
) +(1-y)- (19.2)
I+R T T
2- ~2(1 + R 2)-2y(l- R 2)
I = T. (19.3)
1 2(1 + R)
b)' minimising T, given the values of J and U:
(1- R)J2
x(t ) - -'-----'-- (19.4)
1 - 4(1 + R)U '
x(t)= II+R2-y(J-R2 )~ (19.5)

1 V 2 1+R'
I1
=J 1 ~1+R2-y(1-R2)_R • (19.6)
U (1 + R)(1- y) 2
c) minimising T, given the values of x_ and U:
l-y 2)<x_)2
X(tl) = ( Rt l + -2- tl -U' (19.7)
X(t l ) = [R + (J - y)/1 ]Ix-I, (19.8)
tl = _1_[ V/I
(1- y)
+ R2
- y(l-
2
R R]tl .
2
) -
U
(19.9)
Note that the nominal point completely defines the optimal limit cycle of the
synthesised system . This is typical of the synthesis of systems to produce optimal
periodic motion, in contrast to the usual problem of synthesis in the theory of
optimal control, meant to produce optimal transient processes [55, 88, 179]. When
the problem is restricted to the optimisation of periodic motion, an element of
arbitrariness is introduced to the shape of the switching curves, so it is possible to
expand nominal switching points to the switching curve in such a way that, apart
from the necessary singularity and stability conditions of the nominal limit cycle,
the system would ensure the required reaction to drift the parameters from their
nominal values, for example, stabilise some parameters of the working regime:
period T, impact velocity x_ or impulse J as the coefficient of restitution R

changes.
2. Later we will make a detailed study of the additional potential of the
synthesis method indicated, but for the moment we will study the switching curve
1 (Fig. 19.2)which is simpler to achieve, and has the form
n(x,X)={x-x' =0 when X>O, (19.10)

x = 0 when x < 0,
where x· = X(tl) (Fig.19.1).
As a result the optimised force nmction (control characteristic) B(x,x) will be
as follows (Fig. 19.2):
+ U when x > 0, 0 ~ x < x· ,
B(x,x)= { -U when X>O, x~x', (19.11)
x< 0, x> 0,
i.e. as the striker move upwards to the switching point x· it takes the positive
value, and is negative for all other conditions.
Let us investigate the state of the vibro-impact system, defined by expressions
(18.1), (18.2), during the application of the force characteristic (19.11). We shall
take x' as an arbitrary value.
We will rewrite (18.1) in the form
dx/ dt = x, dX/ dt = B(x,x) - g (19.12)
and we will look at the phase trajectory of system 2 (Fig. 19.2). The straight line
x = 0 represents the position of the stop. The domain of positive values for the
force function lies above the abscissa between the lines x = 0, x = x'. The
remainder ofthe phase plane corresponds to negative va lues ofthe force function.
At the instant t = 0, following the v-th collision, the coordinates of the system
will be (0, xv+)' B( x, x) = U . Dividing the second equation in (19.12) by the first"
we obtain
dX/dx=U(1-y)/x.
Integrating this equation using the initial conditions, we find
x = J(x v +)2 + 2U(l- y)x , (19.13)
i.e. the system velocity increases continuously and when it reaches the switching
point becomes equal to
X· =J(Xv+)2 +2U(1-y)x' . (19.14)
Similarly for the phase trajectory ofthe system after switching we obtain
x = ±J(xv +)2 +2U[2x' -(1+y)x], (19.15)
where the upper half-space corresponds to the positive sign, and the lower to the
x
negative sign. Assuming = 0 in (19.15), we find the maximum displacement of
the striker in the v-th cycle:
§ 19 Synthesis of self-sustaining systems 263
x = (X,,+)2 +4Ux' (19.16)

rn.x 2U(1+y)
As the representative point reaches the line x = 0 the next (v + 1)-th collision
occurs. In this situation, in agreement with (19.15)
X(V+I)_ = _~R2(xv_)2 +4Ux' . (19.17)
Expression (19.17) defined the point mapping of the line x = 0 onto itself.
When the regime is periodic, we find from (19.17)
X(V+I)_ =x,,_ =x_ =-2J'U-x-'/-(I---R-2- ). (19.18)
Hence, the system has a unique limiting cycle which must be the optimal.
Let us study its stability. From (19.17) we obtain
· I=
laX(v+I)_ R zx v_ <1
ai" v- ~R2(xv_)Z +4Ux'
for any starting conditions and parameters of the system. Consequently, the
periodic regime ofthe synthesised self-sustaining system is optimal and stable.
The connection between the period T of self-sustaining vibration and the
switching coordinate x' can be obtained conveniently using (18.10). We find
X· = x(t l ) = -RX_t l -
gt2 I,
T
(z
+ J(tl - T)UdT = -tIRX_ + U (1- y). (19.19)
o
t
Substituting he re the values for x_ and (I from (18.34), (18.35), we obtain after
transformation
T=_1
1- y
z[ 2(~-Y)+
1- R
0+R(~-y)lW.
VI=R 1+ R Vu (19.20)
Substituting the optimal value ofthe impact velocity from (19.18) into (19.16) we
find a formula to estimate the length ofthe system
2x'
xrn.x = (l-y)(1-R 2)' (19.21)
The value of the optimal impact impulse for the system in agreement with (19.18)
will be as folIows:
J =2 Ux •(1 + R) . (19.22)
I-R
Formulae (19.20) to (19.22) show that changing the switching coordinate x·
and force amplitude U is a convenient method to regulate the regime parameters in
a synthesised vibro-impact system. Note that by virtue ofthe unique limit cycle of
the synthesised system, the control method shown allows us to obtain optimal
regimes for all bands of control.
To estimate the quality of transient processes in systems, using (19.17) we can
obtain a general expression to find the sequence of impact velocities for systems
during the period when it settles into a steady-state regime. Summing the resulting
geometrie progression, we find
(v = 1,2, ... )). (19.23)
where xo_ is the impact velocity at the start ofthe transient process.
Let us look at a numerical example of the transient process for a system with
the following parameter values: R = 0.5, T = Is; Y = 0, U = 10m/ S2 . From (19.1)
to (19.3) we find the switching point x' = 0.45m; from (19.18) the impact
velocity in the steady-state regime x = -4.9m / s. Supposing xo_ = 0, then from
(19.23) we find
x l _ =-4.2m/s, x 2_ =-4.75m/s, x J _ =-4.87m/s
Hence, in practice, the system becomes stable after the third impact.
3. In the system synthesised in section 2 the force function B(x,x) has two
switching points x = 0 and x = /. Accordingly, one switching must occur at the
instant of impact. We will look at the influence of delay or advance in switching
relative to the instant of impact on the dynamics.
Consider first the case of delayed switching. The phase plane for such a system
is drawn in Fig. 19.3. The first switching occurs at some value x = x~ . As a result
the force function B(x,x) takes the form
+ U when x > 0, x ~ :::; x < x' ,
B(x, x) =
- U when x > 0, 0:::; x < x ~ ,
x;::: x',
x< 0, x> O.
Let us look at the motion of the representative point
after the instant of the v-th collision (Fig. 19.3). Until the
instant of the first switching B( x, x) = -U , the equation
ofthe phase trajectory takes the form (see section 2)
x= ~(xv+)2 - 2U(l + y)x .
According to this law the point will move until such
time as it reaches the switching line x = x~ or its
Fig 1\1.1
projection reaches abscissa x = O. The first case will occur
if
(19.24)
the second when the inequality has the opposite sign. Hence, inequality (19.24)
defines the value of initial velocity at which self-sustained vibration can possibly
occur, since in the second case vibration will be damped out (Fig. 19.3). As a
result, the limiting cycle in this case can be achieved with hard excitation.
When condition (19.24) is satisfied after intersecting the line = x x;
the
equation ofthe phase trajectory takes the form
~X2 =U(l-y)x+C. (19.25)
Finding constant C from the condition X = ~(xv+)2 - 2U(l + y)x~ when x = x',
we obtain the law of motion for the representative point before the second
switching
(19.26)
Similarly we find the law of motion for the chosen point after the second
switching
(19.27)
°
where the positive sign corresponds to motion in the upper half-plane, and the
°
negative sign to the lower half-plane. Assuming x = in (19.27), we obtain the
equation for the point mapping ofthe line x = onto itself
i Xc V+ 11 = _~r-R-2-(x-·v---)2-+-4U-(x-·---x-~-). (19.28)
+u -u The unique fixed point ofthis mapping
z X = -2 U(x' - x~) (19.29)

1- R 2
determines the stable limit cycle, since
axev+11-1 R2xv_
hg.19.4
l-ax,,_- -- ~R2(xv_)2 +4U(x' -x~)
< 1.(19.30)
However, this limit cycle will not be stable for all

initial conditions and parameters, since the initial velocity must satisfy condition
(19.24) for it to happen. The latter condition is only a necessary condition for the
limit cycle to exist. In reality, if velocity x+ corresponding to periodic motion
does not satisfy inequality (19.24) then it will be unstable. Hence, from (19.24),
(19.29) we can obtain the following condition for stability of the limit cycle:
• 2R 2 /(1-R 2) •
XI < 2 2 X . (19.31)
1+ Y+ 2R /(1- R )
Let us now look at a system where switching is advanced (Fig. 19.4). Its force
function is defined by the express ions
+ U when °: ;
x > 0, x < x· ,
B(x,x) =
x< 0, 0::; x < x~, (19.32)
-U when x> 0, x;::-: x',
x< 0, x;::-: x;.
The law ofmotion ofthe chosen point after impact is described by (19.13). After
switching, (19.15) will apply, according to which the point will move until it
reaches the line x = X; . Thus the motion will be defined by the formula
x = -~(xv+)2 + 2U(1- y)x + 4U(x' - x~) . (19.33)
If x· > x~ ,then as it moves according to the (19.33), the chosen point will
always reach the line x = 0. The corresponding point mapping for this coincides
with (19.28) and on the basis of (19.30) we find that the limit cyc\e in this case is
stable for all parameters values and for any initial conditions. From (19.29) for the
case considered we find
J=2 l+RU(x·_x·).
1- R I
4. Let us turn now to explore systems with more general forms of switching
curve Il(x, x) = O. In section 1 we already have shown that widening the c\ass of
the system considered allows us to the synthesise the characteristics for regulators,
which add adaptive properties to systems, for example, the invariance of certain of
the regime parameters as external conditions change. The simplest method to
construct such a family of switching curves is as folIows: the coordinates of
nominal points are dealt with as equations for the switching curve in terms of
parameter changes, for example R, while invariant parameters T or x remain
fixed.
Fig. 19.5 shows examples of locus for
switching points, constructed using formulae i'lur
(19 .1) to (19.3) for various values of R with
Tconstant. Using portions ofthese curves as
working parts of the switching curve
appropriate to the control of the system over (}Ij
the working zone, it is possible to secure

invariance of the period T as R changes over tJ.l f--+--j
the interval defined. 80th here and later we
shall assume that change of R occurs slowly 0 (}IJ2 Mit (}(JQ M3 z·/ur Z
by comparison with the time taken for Fig. 19.5
transition of the system process.
Similarly, we can construct curves for switching which ensure the invariance
of J or x_ .
The lengths of the portions obtained are limited by the conditions for the
absence of additional intersections with this portion and of the limit cyc\e for the
system. Switching curves which satisfy such conditions will be called permissible.
The chosen working portion of the switching curve should supplement the
curves which ensure the transition of the system onto this part of the switching
curve from the required domains of initial conditions. In the simplest case these
curves are produced, as shown, using a portion of the line x = const. In the more
general case such portions can be drawn using the required quality of the transfer
process.
Let us look at a system with an arbitrary switching curve Il(x, x) = O. The
force function in this case takes the form
. {Usi gnIl(X,x) when x> 0,
B(x,x) = (19.34)
-U when x< O.
After an arbitrary v-th collision the representative point will move in
agreement with (19.13) until it intersects the switching curve. The coordinates of
the switching point X:, x: are defined by the system of equations
(19.35)
After reversal, in agreement with (19.15), the equation for the phase trajectory
takes the form
(19.36)
Here the positive sign corresponds with motion in the upper half-plane, and the
negative sign with the lower. Substituting x = 0 into (19.36), we find the
expression for the point mapping
X(V+IJ- = -~(R.Xv_)2 + 4Ux: . (19.37)
Assuming X(V+IJ- = x v _ = x_, x: = x', x: = x', we obtain a system of three
equations to find the stationary point, corresponding to the periodic regime:
(X')2 _R 2(X_)2 =2U(1-y)x',
(19.38)
Eliminating (X_)2 from (19.38) we come to a system of equations which define

the switching point
(X')2 = 2(~+
1- R
l-Y)UX'2 '
(19.39)
II(x·,x*)=O. (19.40)
In the general case, the system of equations (19.39), (19.40) can not be solved
in an explicit form. As a convenient calculation procedure we can take the
parabolic approximation of the switching curve in the neighbourhood of the
nominal points x~,x~ .
Rearranging (19.40) in the form
x' = al(x' _X~)2 +a 2 (x' -x~)+x~. (19.41)
where ai' a2 are numerical coefficients, and using conditions (19.39) which can
be written in the form (X~)2 = ax~, where
2R2 )
a = 2( 1_ R 2 + 1- Y U ,
we obtain from (19.41) a quadratic equation in x':

(1- aal )(X')2 + a(2alx~ - a 2 )x' - a[a l (X~)2 - a2x~ + x~] = o. (19.42)
Using equation (19.42) it is possible to find the coordinates ofthe switching curve
(19.41) at which switching of control excitation occurs as the system parameters
moves away from the nominal. This leads to the selection of the numerical
parameters aJ, a2 to ensure that the required properties arise in systems with
slowly changing parameters. Note that the values of T and J in the case of an
arbitrary switching curve are defined as before by (19.20), (19.22), however the
value of x* is now found from the set of equations (19.39), (19.40).
5. Let us look at the stability of systems in the region of the nominal limit
switching cycle for an arbitrary switching curve. Suppose x~, x~ are the reversal
points corresponding to this cycle; Xo is the pre-impact velocity. We shall
introduce differences 8x v = x" - x, 8x" = Xv - X (v = 1,2, ... ) which characterise
the variation of the phase trajectory from the nominal cycle. Linearising (19.13)
near the nominal cycle, we obtain
8x" _-.I_[R 2 x 0 8x v +U(I-y)8x..l. (19.43)
X o_
After linearising the switching curve, we find for the switching points in the
disturbed motion
(19.44)
an '.'
where k = - [ a;(xo,x o )
]-1 an '.'
ax (xo,x o )' From (19.43), (19.44) we find for the
instant of switching
.'
8x v =-
R 2 x- .
8xl" (19.45)
x o_ + kU(1- y)
For the increment in velocity before the following (v + I)-th impact, linearising
(19.37), we obtain
8x(l'+l) =-.I_(R 2 x o_8x v _ +2U8x:). (19.46)
X o_
From this using (19.44), (19.45) we find the difference equations for the disturbed
motion
(19.47)
where
K = x o_ -kU(1+Y).
(19.48)
x o_ + kU(l- y)
According to (19.47), the stability condition takes the form
IR KI < I.
2 (19.49)
Note that, in addition to condition (19.49), the
value of k must also satisfy another condition by
intersecting the switching curve and the limiting
cyc1e at the nominal switching point (Fig. 19.6).
The latter takes the form
~> dX(x')= U(l-y) when ~>O
k ~ .' k'
x (19.50)
I~Ik > U(lx'+ y) k

when ~ < O.
When y = 0 both conditions (19.50) can be :< [,ig. 19.6
combined into one:
§ 19 Synthesis cf se1f-sustaining systems 269
1*1 > ~ . (19.51)

The table below shows the limiting values of the term T / k , calculated from
(19.50) and corresponding to the problem of maximising the impact impulse for
given values of T and U .
R 0 0.2 0.4 0.5 0.6 0.7 0.8 0.9
..
U(l-y)T
x
3.34
2
3.14
1.61
2.78
1.22
2.63
1.09
2.5
0.93
2.32
0.86
2.22
0.8
2.12
0.72
-'L= 0
..
U(I +y)T 6 4.83 3.66 3.27 2.79 2.58 2.4 2.16 y= 0.5
---
x
Numerical calculation shows that for the majority of values of the parameters
condition (19.50) is stronger than (19.49). For example, for problem a) (see
section 1) when y = 0, R = 0.2 we obtain Ik / TI<
0.415 using condition (19.49),
while at the same time from the table we find Ik / TI < 0.32 .
It follows from (19.47), (19.48), that the corresponding choice of value for k
can also influence the speed ofthe transient process.
6. Let us now look at the problem of synthesising working portions of the
switching curve for small changes in the coefficient of restitution relative to the
nominal value Ro ' Suppose that as the result of changing the coefficient of
restitution by an amount öR a transient process occurs in the system. Let us
linearise the motion of the system ne ar the nominal limit cycle using the change in
R. From (19.13) we obtain
ö.:<_ =_-.I_[Rgx o_öx v_ +U(l-y)öxv]-Roxo_öR (19.52)
x o_
(x o_ is the nominal impact velocity). At the instant of switching, expression
(19.44) is satisfied. Using this, we find from (19.52)
\:: .• = _ Rgx o_ \::. _ ROCXO_)2 \::R
ux v ux v u . (19.53)
x o_ + kU(l'-y) - x o_ + kU(l- y)
Linearising (19.37), we find the increment in velocity before impact
ÖX(V+l)_ =-.I_(Rgx o_öx v_ +2uö<)+Rox o_öR. (19.54)
x o_
Substituting into (19.54) the expression for öx: using (19.44), (19.53), we obtain
the difference equation for öx_ :
ÖX(V+l)_ = Rg Köx v_ + Roxo_KöR. (19.55)
From equation (19.47) this changes in the presence of a non-homogeneous term. If
the solution of the homogeneous equation (19.47) is stable, Le. satisfies condition
(19.49), then as v ~ 00 , the transient component of the solution of (19.55) tends
to zero and only non-homogeneous components remain, which we find by putting

ÖX(I'+I)_ = öX v _ = öx_ in (19.55):
(19.56)
Expression (19.56) defines the increment in the impact velocity for stable regimes
due to a change in the coefficient of restitution. Let us find the corresponding
change in the remaining parameters ofmotion. From (19.53) using (19.56) we find
öx' = - R (R; + I-R;K

(X O_)2 o
xo_ +kU(I-y)
K JöR.
For öJ we obtain directly

öJ = -(1 + Ro + öR)(Xo_ + öx_) - (1 + Ro)x o_ =
= _ (1 + KR o )x o_ öR. (19.57)
1- KR;
To find öT we will express Tthrough and x_ x'.
Taking into account that in the
interval between impacts and switchings the velocity of the system changes
uniformly, we obtain
x' +RX x _ -x' 2x'+xj(l+y)R-I+y]
T=--- (19.58)
U(l-y) U(l+y) U(l_y2)
Linearising (19.58) in terms of öR at the point Ro , we find
öT= {28x' +[(1+y)Ro -l+y]öx} + xo_öR
(19.59)
U(1- y)2 U(1 - y)
(19.56), (19.57) and (19.59) allow
us to calculate the sensitivity of
~.T/17/~R ~T/TtfR
the process parameters to changes r----r----.---~r---~
in the coefficient of restitution.
With this aim we can introduce //-9
sensitivity characteristics 0'51---::=t::::::::;::=-r~r-t---"i
8T / öR, 8x_ / 8R, öJ / öR, which
depend on the gradient of the 0·3
switching function k. Choosing
this value in the proper mann er, it 03 k
is possible to influence the way in -1 f
which the parameters of the
-2
periodic regime of the system
change due to varying the -J
coefficient of restitution from its
nominal value. For this the dijTtltfR
specified regimes wi 11
. ll y cause t he process to
automatlca Fig.19. 7
function, which gives the adaptive
condition required. Fig. 19.7 shows curves for the sensitivity coefficients when
§20 Synthesis of autoresonant systems 271
y = 0, R = 0.2. It was shown above that for these values of the parameters the
slope of the switching curve for problem a) is limited to the value Ik / TI ~ 0.32 . In
the whole ofthis range stability ofthe system is guaranteed for sm all disturbances.
Reference [29] describes methods for the synthesis for the development of
self-sustained vibration in systems causing periodic multi-impact regimes with
given cycle of operation and intensity of collision.
§20. Synthesis of autoresonant systems

1. Wehave shown earlier that an effective scheme to excite the striker is to
create a driving excitation force on the striker which is a function of its phase
coordinate, i.e. using the principle of feedback. The best way to obtain similar
excitation of striker I (Fig. 20.1) or of the vibro-impact system 2, including elastic
Converter
suspension and drive mechanism or wave guide, is to measure and analyse the
motion parameters of the striker and the basic loads (for example, the static feed
force G, with which an impact machine compresses the work piece), to transform
and amplify the signal and to generate a driving excitation force to be produced by
the actuating arrangement ofthe drive, based on these signals.
In actual self-sustaining systems, the elements of the control system specified
may differ, both in structure and construction. At the same time, as we saw when
analysing a pneumatic impacter (see § 12 sections 6, 7), the internal dynamics of
the system drive elements (such as nonlinearity, hysterisis, the relationship
between the output power of the excitation source and the regime in which the
system is operating, etc) have a significant influence on system operation. A
combination of the internal nonlinear dynamic in vibro-impact systems, with the
chan ging conditions during the operation of the working process (typified by
changes in the coefficient of restitution R), show that universal control methods,
based, for example, on the application of search systems or other adaptive
principles, are both expensive and ineffective in such situations.
The best arrangement for systems described would appear to be a self-
sustained system tuned to autoresonance, stabilised by nonlinear corrections of
perturbation [18] (rigid adaptation [89]). With such an approach, the basic
emphasis in the construction and adjustment of systems is not to widen the
functional potential of the control system, and so increase its complexity by

introducing a different analytic structure, but to make a thorough study of systems
in order to match all the basic interacting factors. This makes serious demands on
the caIculation methods to achieve these goals in terms of their universality,
flexibility and clarity. The latter quality appears significant, since partially
changing, or adding structure to the caIculation model by introducing correcting
arrangements can often lead directly to the synthesis process from the results of
caIculations.
Similar to the analysis of vibro-impact systems, the complicated calculation
model can give considerable, and often insuperable, difficulties when exact
synthesis methods, like those above, are used. The most essential difficulties are
related to the increased dimensions of systems and to the estimation of the
influence of other nonlinear factors. However, by moving to approximate methods
based on equivalent linearisation, the problems of synthesising solutions with the
difficuIties described can be overcome. The use of equivalent linearisation for
nonlinearities makes it possible to reduce the problem of synthesising the required
structure to the problem of optimising its linearisation coefficients in terms of the
system parameters. Equivalent Iinearisation is a routine method to reduce a
variational problem of optimisation in functional space to a problem of finding
extrem um for functionals in the parameter space. From the values found for the
optimallinearisation coefficients the required structure can be produced.
Note that regenerating the functions from their equivalent Iinearisation
coefficients is ambiguous since at this stage of synthesis, additional requirements
and limits on characteristics of the synthesised structure can also be taken into
account.
The use of harmonic linearisation methods to control vibration can lead to the
regulation of amplitude, frequency and phase of the fundamental. In vibro-impact
systems of the autoresonant type, the frequency of self-sustaining vibration is
defined by the natural characteristics of the vibro-impact system (see § 12), hence
the problem, as shown, consists of obtaining the maximum intensity of impacts as
the tuning of the system changes due to slowly changing the feed G, which
influences the system frequency.
A solution to the problem is guaranteed by introducing feedback (Fig. 20.2),
Fig. 20.2
---
amplifying and transforming the signal ur(t) proportional to the resulting motion
ofpoint x ofthe striker 1 (the observed point) by function B[ur(t)] in block 5 and
attenuated through the delay component 4 at point n (the control point) of the
vibrating system 2. As a harmonic signal with frequency ro passes through the
delay component, a phase lag q> appears between the input and output signals
whose value is q> = rot o , so that, by choosing the delay time 10 it is possible to
guarantee resonant phase at frequency roto = 37t/2 (since
cos(rot - 37t/2) = cos(rot + 7t/2» on the given autoresonant frequency ro,
corresponding to the frequency of resonance with maximum amplitude for the
vibro-impact system. By changing the value of feed G applied at point r (the point
of disturbance), drift of the natural properties of the vibro-impact system takes
place, hence, in order to maintain autoresonance under these conditions it is
necessary to control the value of delay 10 in relation to G to maintain the resonant
value of phase q> = -37t/2. Calculating the relationship of ro to G makes it
possible to apply such control into practice using a rigid correction coupling using
the transformer 3. By adjusting the transformer 5 using correction component 6,
by changing G, it is possible to obtain the required value of autoresonant vibration
amplitude in relation to feed G. Hence by arranging the scheme described
correctly, autoresonant conditions will be maintained in the system with the
prescribed level of vibration whatever the value of G.
2. Let us find the structure of elements 3, 5, 6 for the feedback loop,
assuming system 2 is linear and describing the working process executed by the
striker by means of a nonlinear reaction cD(ur,su r ) representing the force
characteristics of the impact interaction between the striker and the work-piece.
We will consider the value ofthe maximum force, generated by amplifier 5 to be
limited to
(20.1)
Using an operator for the transfer dynamic compliance for system 2 in the
form L,,(x,s) , connecting the displacement of point x with the force applied at
point n, we can write the equation for vibration of the striker in the following
form:
Ur (!) = L" (x,s)P" (t) - Lr(x, s)cD[uAt), sU r (I)] + Lr(x,O)G , (20.2)
where, in agreement with the feedback error
P" (t) = B[u r (t - (0 )] = B[ur(t)]exp( -sto ) • (20.3)
Let us search for an approximate periodic solution to equation (20.2) in the
form
Ur(t) mr + u~(t),
I'l:: u~(t) = a r cosrot . (20.4)
Applying harmonic linearisation to the characteristic B(u r ) , we obtain
B(u r ) = B(u~) I'l:: Z(ar)u~, (20.5)
where Z(aJ is the linearisation coefficient. Here we assume that the feedback
filters only the variable components of signal (t) . Ux
Taking into account that because of the properties of Fourier series, the
limiting level U o of the periodic excitation P,,(t) can not have a first harmonie
amplitude greater than 4Uo In we find from (20.5) using (20.4)
(20.6)
Linearising the force characteristics of the impact interaction similarly to
(14.8), we obtain, substituting (20.4), (20.5) into (20.2) and separating the constant
and periodic components:
m x = -Lx(x,O)mF(m x' a) + Lr(x,O)G (20.7)
{I + Lx(x,s)[k(mx,a x ) +sß(mx,a x )]-
(20.8)
- LII(x,s)Z(ax)exp( -sto) = 9JJ(s)u~(t) = 0,
where 9!l(s) is the characteristic equation for the linearised system depending on
m x and ax as weil as the other parameters. Substituting s = jw into this and
setting the real and imaginary parts to zero we find the existence conditions for
self-sustained vibration:
Z(ax)sinwt o = -V(w, m x' aJ, (20.9)
Z(ax)coswt o = U(w, m x ' a x )' (20.10)
Here U(w,mx,a x ) = ReW(jw,mx,aJ; V(w,mx,aJ = ImW(jw,m x aJ;
W(jm,mx,aJ = [I + k + jmß)Lx(x,jm)]C I (x,jw) ..
Eliminating the trigonometrie functions from (20.9), (20.10) we obtain
Z(a x ) = ~U\w, m x' a x ) + V 2 (w,mx' a x ) = IW(jw,m x' aJI. (20.11)
We will rewrite this expression in the form
a x = aJ(ax)[U\w,mx,a x )+ V 2 (w,m x,aJr Il2 (20.12).
Comparing expression (20.12) with (20.6) shows that for all frequencies the
maximum amplitude ofthe striker vibration occurs when the following equation is
satisfied
whence
Z(aJ = 4U o l(naJ. (20.13)
As a result we obtain the value of the harmonie linearisation coefficients for
the nonlinear amplifier 5 (Fig. 20.2), which ensure maximum excitation of the
striker. lt is not difficult to construct the amplifier
4(u~}
UfJl--- - - structural characteristic from it. By analogy with
(12.3), (\2.20) we observe that expression (20.13)
can be obtained, for example, by using a
characteristic of saturation for the amplifier
----4-"" B(u x ) in the form (Fig. 20.3):
Fig. ZO.] B(ux ) = Uosignux' (20.14)
By analogy with (14.17) we can establish that
-----------=-
when an extemal periodic excitation Pn (t) = p" cos( rot + q» is applied at point n
instead of excitation through feedback, the vibration amplitude of point x takes the
form
a x = p,,[U 2(ro,m x,aJ + V2(ro,mx,a.)r'/2. (20.15)
Comparing equation (20.12) and (20.15) we can conclude that, taking into account
(20.13), when U o = np', /4, the amplitude and frequency of self-sustained
vibration of point x, obtained for various delay values to, correspond to points on
the resonance curves of the striker.
Let us find the delay for wh ich the largest amplitude of self-sustaining
vibration occurs (autoresonance). The value V(ro,mx,aJ is proportional to the
dissipative coefficients ofthe vibro-impact system. Assuming that these values are
sufficiently smalI, it is natural to assume that, in agreement with (20.12), the
maximum amplitude of vibration a; occurs at the frequency ro' , wh ich satisfies
the condition
U(ro' ,mx,a;) = O. (20.16)
Using (20.13), (20.16) we find from (20.12)
• 4U o
ax = * .' (20.17)
nV(ro ,mx,a.)
The system of equations (20.16), (20.17) is similar to expressions (12.11) (12.12)
and establishes the expression for the backbone curves of elastic systems and the
energy boundary ofvibration. Taking into account (20.7), the relationship between
m x and a x ' this system of equations allows us to define the amplitude and
frequency of autoresonance as was done in §12. When condition (20.16) is
satisfied in agreement with (20.9), (20.10), the delay is found from the expression
3n 2ni
=-,+-,
to (i=0,1, ... ). (20.18)
2ro ro
The system of equations (20.7), (20.16), (20.17) allows us to define the
relationship between the autoresonant frequency ro' and force G by eliminating
m x and a x ' Ifwe define this relationship as ro'(G) and substitute it into (20.18),
we obtain the characteristic which links the controlling delay 3 (Fig. 20.2) to the
feed G. This characteristic depends on the system parameters only and thus the
control of delay can be obtained using a rigid nonlinear coupling.
According to (20.17) the amplitude of autoresonance can be easily controlled
by varying the saturation value Uo in amplifier 5, so that, by eliminating the
values of mx and ro' from equations (20.7), (20.16), (20.17) we can obtain the
relationship a x = ax(G,U o)' Given the required value of amplitude a x which
depends on feed G, we can find the corresponding value of Uo and consequently
the characteristic Uo(G) linking the amplitude controller 6 (Fig. 20.2). In
particular when it is necessary to stabilise the autoresonant amplitude for all
values of G the corresponding characteristic U 0 (G) is found from the implicit
expression ax(G,U o) = const. Since the characteristic Uo(G) is defined in terms

of system parameters only, it also can be obtained using a rigid correction
coupling.
3. Let us now look at the stability of autoresonance in synthesised systems,
using an approach similar to that described in § 14 section 2. We will assume that
in the neighbourhood of the stable periodic solution (20.4) the vibration of the
striker takes the form
(20.19)
where the tilde means a slowly changing time function. We shall consider
equation (20.9) as a condition for the energy balance between dissipative and
excitation forces for the steady-state motion considered, forming the stability
conditions for this motion due to the small deviations of amplitude x from the a
steady-state value ax in the form
~[V(ro,mx,ax)+Z(ax>sinrotola
da x
=a
x
>0 when sinoot o <0,
(20.20)
Noting that as distinct from the case studied in § 14, in this system both situations
represented by (20.20) are possible when the observation and control points
coincide. However, both conditions here lead to identical final forms of the
stability condition.
Introducing three unknown functions instead of one in (20.19), we shall
connect them through two conditions, assuming that during disturbed
motionmx(t) and ax(t) are connected by equation (20.7) but ro(t) and ax(t) by
equation (20.10). Eliminating sin oot o from (20.20) using (20.10), we find
~[V(ro,mx,ax)
da
- ~Z2(ax) - U 2(ro,m x,aJ b=a
' x
>0
x
when sinoot o < 0,

(20.21)
~[V(05,mx,ax> -~Z2(ax) -U 2«ö,m x,ax Hi
~ x
=a
x
<0
when sinoot o > o.

Differentiating (20.21) and bearing in mind (20.11) we obtain the following
conditions for stability for both cases:
~ ~ ~) dU(05,m x,ax ) V(~ ~ ~) dV(ro,mx,a x )
[U( oo,mx,a x + oo,mx,a x -
dax da x
(20.22)
_ Z(ax) dZ(ax )] . > o.
da x - ax::.a;r
•
Let us apply this to investigate the stability of autoresonance. According to

(20.16), (20.11) we find in this case from (20.22)
Condition (20.23) generalised expression (12.14) obtained in § 12. In accordance

with (20.18), (20.9) in the conditions for autoresonance V(ro,mx,a.) >0, while
this value is linear in terms of coefficient ß. As a result, for all forms of the
harmonie linearisation coefficients
[ dV(ffi,iii x,G.)] ~ O.
dax - '
UxO;:;"x
Since according to (20.13)
[ d~(Gx)]
da x - •
<0,
U X= Ul'
the stability condition (20.23) is always satisfied for synthesised systems.

4. The method described was used in
references [21 , 22] to synthesise a system for the
automatie adjustment of an ultrasonic machine
undergoing an autoresonant regime under various
feed forces. The dynamic model of the vibration
system for an ultrasonic machine was considered
in § 15 section 3 where the excitation P" (I) was
produced by an external generator (Fig. 15.6).
Let us now synthesise an autoresonant
sehe me for excitation of the object described
(Fig. 20.4). In this case the excitation is applied
in the form of delayed feedback at some section
of the bar. Suppose, for example, that a
displacement pickup measures the vibration of ~
Fig. 20.4
the bottom section of the bar (x = I). After
amplification and transformation of this signal in block 5 it passes through the
delay block 4 at the section with coordinate x = n . Maintaining the system
description of § 15, section 3 and assuming for simplicity that n = 1 we can write
the equation ofvibration for the bottom section ofthe bar in the form
u, (I) = L,(l, s){B[u, (I)] exp( -SI) - [u, (I)] + G} . (20.24)
Here (u,) is the static force characteristic ofthe stop.
Equation (20.24) is a particular case of equation (20.2), hence choosing
B(u,) and 10 in the forms (20.14), (20.18), we find the required structure for the
excitation feedback.
Let us define characteristic 10 (G) as the correction coupling by perturbation of
the feed (Fig. 20.4). We shall take advantage ofthe fact that the equation for the
operator W(jro ,m"a,) coincides in this case with (15 .12) (see §15 section 3).
Assuming that the stop is rigid (15.20) we find from (15.29) the following
equation for the backbone curves (20.16) when X == 0 :
* 2h
a, = - r* r* (20.25)
~ cot~
The equations for the energy boundary of vibration (20.17) eoincide here with
(15.23) when X = 4Vol l(nES):
* 4nxsin (2
a = -:--7-'---'------::-- (20.26)
I 'X.s* «( + Yz sin 2() .
Eliminating a; from (20.25),(20.26), we find
G 8sin2(
(20.27)
Vo x(2( + sin 2() .
In (20.27) expressing t; in terms of lo using (20.18), we obtain the relationship
sought which defines the characteristic lo(G) ofthe delay block regulator. When
i = 0 it takes the form
8 . 3n SIn-
G
Vo xe; +sin 3;)

(20.28)
--
The curve of the characteristic described is drawn in Fig. 20.5 using (20.28). The
T range of delay change corresponds to
r-..... autoresonance of the system in the first mode
"-.. of vibration. Changing the delay outside the
band considered ean, in prineiple, induce
~ other modes of vibration, however internal
2·g damping in the material of the bar limits such
o 1-0 1·5
possibilities in practice to only a few 10wer
Fig. 20.5 modes. The monotonie behaviour of the
expression obtained for (0 (G) makes it easy to produee in praetiee.
Bibliography
There were giants in the earth in those days.
Genesis, 6: 4
I. Aizerman M A, Smirnova N M, On the application of methods of small parameters in the

investigation of periodic regimes in automatic control systems not containing small parameters.
Coll. "Tribute tor A A Andronov", 1955. (In Russian).
2. Akhiezer N I, Krein M G, On some questions in the theory of moments. Gonti, Kharkov, 1938.
(In Russian).
3. Akynfiev T S, Babitsky V I, Autoresonant regimes in pneumatically driven systems. Machine
Studies, 1977, No. I. (In Russian).
4. Alabuzhev P M, Tsivinsky Yu P, Amplitude-frequency characteristics of vibration for a two-
mass system with impact of mass against a stop. Symposium "Mechanisms and machines with
impact and rotary impact effects", Novosibirsk, 1963. (In Russian).
5. Aleksandrov E V, Sokolinsky V B, Applied theory and design of impacting systems. Nauka,
Moscow, 1969. (In Russian).
6. Ananiev I V Manual for the computation of natural vibration of elastic systems. Gostekhizdat,
1946. (In Russian).
7. Ananiev I V, Timofeev P G, Vibration of elastic systems in aircraft construction and their
damping. Mashinostroenie, 1965. (In Russian).
8. Andronov A A, L I, Mandelstam and the theory of non linear vibration. Collected works, 1956.
(In Russian).
9. Andronov A A, Neimark Yu I. On the motion of an idealised model of a clock with two degrees
offreedom. Models ofpre-Galilean clocks, 1946, vol. LI, No. I. (In Russian).
10. Andronov A A, Vitt A A, Khaikyn S E, Theory of oscillations, 1987. Dover Publications Inc
(Paperback version of translation pub. Pergamon Press, 1966). (Translation from Russian,
Fizmatgiz, 1959)
11. Arnold RN, Response of an impact vibration absorber to forced vibration. Proc IX Cong. Appl
Mechs, Brussels, 1956.
12. Arsenin, V Y, Mathematical physics. Basic equations and special functions. Nauka, Moscow,
1966. (In Russian)
13. Artobolevsky I I, Hertz E V, Kobrinsky A E, Raevsky N P, On the dynamics of pneumatic
installations. Proceedings of seminars on theory of machines and mechanisms, 1953, No. 51. (In
Russian).
14. Arzhanikh I S On novel inequalities in stability. Automation and telemechanics, 1961, No. 4. (In
Russian).
15. Ashavsky A M, Sheinbaum V S, On one isoperimetrical problem in the optimisation ofthe speed
ofresponse for a single mass vibro-impact system. Machine Studies, 1970, No. 3. (In Russian).
16. Ashavsky A M, Vladov M V, On the synthesis of optimal vibro-impact machines. Dynamics of
machines, Mashinostroenie, 1969. (In Russian).
17. Astashev V K, Babitsky V I, Dolnik E S, About one method of oscillation excitation, Mechanies
of Solids, 1972, No. I. (In Russian).
18. Astashev V K, Babitsky V I, Hertz M E, On the synthesis of autoresonant systems. Vibro-
Engineering, Kaunas, 1973, No. 3 (20). (In Russian).
19. Astashev V K, Babitsky V I, Resonant oscillations of a visco-elastic rod with limiter. Mechanics
of Solids, 1972, No. 4. (In Russian).
20. Astashev V K, Babitsky V I, Vibro-impact interactions of visco-elastic rods. Machine Studies,
1974, No. 5. (In Russian).
21. Astashev V K, Hertz M E, Excitation and stabilisation of resonant vibration of ultrasonic rod
systems. Acoustic Journal, 1976, No. 2. (In Russian).
280 Bibliography
22. Astashev V K, Hertz M E, Se\f-sustaining vibration of a visco-elastic rod with limiter under
delayed force. Machine Studies, 1973, No. 5. (ln Russian).
23. Astashev V K, On the dynamics ofan oscillator impacting against a stop. Machine Studies, 1971,
No. 2. (In Russian).
24. Babakov I M, Theory ofvibrations. Nauka, Moscow, 1968. (In Russian).
25. Babitsky V I, Borovkov B A, Vibro-impact processes in systems with Iimited excitation.
Mechanics ofSolids, 1977, No. I. (In Russian).
26. Babitsky V I, Dzhokhadze Sh R, Kobrinsky A E, Komdorf S F, Polunov Yu L, Pulling effects in
vibro-impact systems. Machine Studies, 1968, No. 6. (In Russian).
27. Babitsky V I, Izrailovich M Y, On the optimal motion ofvibro-impact systems. Machine Studies,
28. Babitsky V I, Izrailovich M Y, On the synthesis of self-sustaining systems. Machine Studies,
29. Babitsky V I, Izrailovich M Y, Optimal multiple motions ofvibro-impact systems. Mechanics of
Solids, 1969, No. 3. (In Russian).
30. Babitsky V I, Izrailovich M Y, Some aspects of the optimisation of vibro-impact systems.
Machine Studies, 1969, No. 3. (In Russian).
31. Babitsky V I, Izrailovich M Y, Some problems in the synthesis of optimal periodic vibro-impact
systems. Mechanics ofMachines, 1972, Nos. 33-34. (In Russian).
32. Babitsky V I, Izrailovich M Y, Synthesis of se\f-sustaining vibro-impact systems with prescribed
cycles. Mechanics ofSolids, 1970, No. 5. (In Russian).
33. Babitsky V I, Kobrinsky A E, Periodic motion of two-mass vibrating systems in a container.
Theory ofMachine and Mechanisms, 1964, No. 104. (In Russian).
34. Babitsky V I, Kolovsky M Z, Investigations of linear vibrating systems with limiters by exact and
approximate methods. Machine Studies, 1967, No.4. (In Russian).
35. Babitsky V I, Kolovsky M Z, On the dynamics of systems with impact damper. Machine Studies,
36. Babitsky V I, Kolovsky M Z, On the investigation of resonance in vibro-impact systems.
Mechanics ofSolids, 1976, No. 4. (In Russian).
37. Babitsky V I, Kolovsky M Z, On the theory of vibro-impact systems. Machine Studies, 1970,
No. I. (In Russian).
38. Babitsky V I, Kolovsky M Z, Random oscillations of linear systems with Iimiters. Mechanics of
Solids, 1967, No. 3. (In Russian).
39. Babitsky V I, On the analysis of elastic systems with vibro-impact interaction. Mechanics of
SoIids, 1975, No. I. (In Russian).
40. Babitsky V I, On the existence of high-frequency oscillations with large amplitude in linear
systems with Iimiters. Machine Studies, 1966, No. I. (In Russian).
41. Babitsky V 1, Oscillations of an inertially suspended pendulum in a vibrating container.
Dynamics ofmachines Mashinostroenie, M., 1966. (ln Russian).
42. Babitsky V I, Parametric oscillations of vibro-impact systems. Machine Studies, 1971, No. I. (In
Russian).
43. Babitsky V I, Suvorov G A, Investigation of vibro-impact systems using the universal analogue
computer Analak-IIO, Analogues and analogue-numerical techniques, Mashinstroenie, 1968. (In
Russian).
44. Babitsky V I, Tresviatsky A N, About one case of stability loss in apreloaded mechanism,
Mechanics ofsolids, 1975, NO.6. (In Russian).
45. Babitsky V I, Tresviatsky A N, Forced oscillations of a visco-elastic beam with fixed particle
between limiters, Machine Studies, 1976, No. 2. (In Russian).
46. Babitsky V I, Vibro-impact motion of a pendulum with inertial suspension in a vibrating
container. Analysis and synthesis ofautomatic machines, Nauka, Moscow 1964. (In Russian).
47. Ballo I, Contribution to the investigation of dynamical properties of nonlinear systems with
acceleration damper, "Zagadnienia drgan nieliniowych", Warszawa, 1963. (In Polish).
48. Barkan D D, Shekhter 0 Y, On the theory offorced oscillation of a vibrator with limiter. J. Tech.
Phys., 1955, v. 25, No. 13. (In Russian).
49. Batuev G S, Golubkov Yu V, Efremov A K, Fedosov A A, Engineering methods for the study of
impact processes, Mashinostroenie, 1969 (In Russian).
50. Bautin N N, Behaviour of dynamic systems near boundary of stability domains. Gostekhizdat,
1949. (In Russian).
51. Bautin N N, Dynamic theory of clock operation without stopping the driving wheel
(instantaneous impulse). Engineering, 1955, vol. XXI. (In Russian).
Bibliography 281
52. Bautin N N, On the MandeJstam problem in the theory of clocks. DAN SSSR, 1949, v. LXV, No.
3. (In Russian).
53. Bautin N N, On the Motion of an ideal model ofa clock with two degrees offreedom. Modelling
ofGalileo-Huygens clocks. DAN SSSR, 1948, v. LXI, No. I (In Russian).
54. Beliayev N M, Stability of a prismatic rod subject to variable longitudinal forces. Collection of
papers on engineering construction and structural mechanics. Pul. Leningrad, 1924. (In Russian).
55. Bellman P, Dynamic programming. Princeton University Press, 1957.
56. Beserkersky V A, Popov E P, Theory of automatie control. Nauka, Moscow, 1966 (In Russian).
57. Bespalova L V, Neimark Yu I, Feigin M I, Dynamies ofsystems with impact interactions and the
theory of nonlinear vibration. Mechanies of solids, 1966, No. I. (In Russian).
58. Bespalova L V, On the theory ofvibro-impact mechanism. OTN, 1957, NoS (In Russian).
59. Bespechniy V M, Viktorov E D, Propagation ofvibration in beams having a gap in the assembly.
Mechanics ofsolids, 1970, No. 4. (In Russian).
60. Biderman V L, Theory ofimpact Mashgiz, 1952. (In Russian).
61. Blekhman I I, Dzhanelidze G Yu, Dynamics ofthe Buass-Sard regulator. OTN, 1955, No. 10. (In
Russian).
62. Blekhman I I, SeJf-synchronisation in some vibration machines. Engineering, 1953, v. XVI. (In
Russian).
63. Blekhman I I, Synchronisation of dynamic systems. Nauka, Moscow, 1971. (In Russian).
64. Bogolyubov N N, Mitropolsky Yu A, Asymptotic methods in the theory of nonlinear oscillation.
Gordon and Breach, NY. 1961. (Tranlation from Russian, Fizmatgiz, 1958).
65. Bolotin V V, Dynamic stability of elastic systems. Holden Day, San Francisco, 1964.(Translation
from Russian, Goztekhizdat, 1956)
66. Booton R C, The analysis of non linear control systems with random inputs. Proc. Symp.
Nonlinear Circuit Analysis, 1953, v. 2.
67. Boyarinov V S, Neimark Yu I, On vibrations of shafts in ball bearings, Mechanics, 1965, No. 3.
(ln Russian).
68. Bromberg P V, Matrix methods in the theory of relay and impulse control systems. Nauka,
Moscow, 1967. (In Russian).
69. Brunshtein R E, Kobrinsky A E, Periodic motion of systems consisting of a ball in a cavity. OTN,
1959, No. I. (In Russian).
70. Brunstein R E, Dynamics and stability of oscillating systems occurring due to the effects of
arbitrary periodic forces. Mechanics of machines, 1969, Nos. 25-26. (In Russian).
71. Brunstein R E, Kobrinsky A E, On the stability of periodic motion of vibro-impact systems.
OTN, 1960, No. 5. (In Russian).
72. Butenin N V, Elements of theory of nonlinear oscillation. Blundell Publishing, NY, 1965.
(Translation from Russian, Sudpromgiz, 1962).
73. CempeJ Cz, The equations of motion of a system with impacts. Engineering reports, Warsaw,
1969, No.17. (In Polish).
74. Clarc N J, An ultrasonic machine tool. Trans IRE, 1954, No. 10.
75. Colloquium on mechanical impedance methods for mechanical vibrations. Ed . by R Plunkett.
ASME, New York, 1958.
76. Crandall S H (Ed.), Random vibration, v. 2. MIT press, Cambridge, Mass., 1963.
77. Crandall S H, Mark W D, Random vibration in mechanical systems. Academic Press, 1963.
78. Deutsch G, Anleitung zum praktischen gebrauch der Laplace-Transformation. R. Oldenburg,
Munieh, 1956 ..
79. Ditkyn V A, Prudnikov A P, Integral transformations and operational calculus. Nauka, Moscow,
1974. (In Russian).
80. Dobrogursky S 0, On the problem of stresses and forces during impact In book: "Problems in
the design and construction ofmachine details", Proc. AN SSSR, 1942. (In Russian).
81. Dubowsky S, Freudenstein F, Dynamic analysis of mechanical systems with clearances, parts I,
11. 1. of eng. for industry, 1971,No. I.
82. Egle D M, An investigation ofan impact vibration absorber. J ofEng. For Industry, 1967, No. 4.
83. Emelianov S V, Systems for automatic control with variable structure. Nauka, Moscow 1967. (ln
Russian).
84. Feigin M I, Dynamic systems with impact interactions, with the existence of sliding motion.
Mechanics ofmachines, 1972, Nos. 33-34. (In Russian).
85. Feigin M I, On forced vibration of two masses with a gap. Mechanics and mechanical
engineering, 1960, NoS (In Russian).
282 Bibliography
86. Feigin M I, On the theory of nonlinear damping. Proc. Vuzov, Radio physics, 1959, No.4. (In
Russian).
87. Feigin M I, Sliding regimes in dynamic systems with impact interactions. Applied maths. and
mechanics, 1967, No. 3. (In Russian).
88. Feldbaum A A, Basic theory of optimal automatic systems. Nauka, Moscow, 1966. (In Russian).
89. Feldbaum A A, Problems in self-adjusting (adaptive) systems. Proc. all-union conf. on self-
adjustment, Nauka, Moscow, 1965. (In Russian).
90. Fikhtengolts G M, Course of differential and integral calculus. Nauka, Moscow, 1969, vols. I-IIl.
(In Russian).
91. Frolov K V, Some problems in parametric vibration of machines. Col!. "Vibration and stability of
instruments, machines and elements of control systems", Nauka, Moscow, 1968. (In Russian).
92. Fu C C, Dynamic stability of an impact system connected with rock drilling. J. of App!.
Mechanics, 1969, No. 4.
93. Fu C C, Paul B, Stability ofmotion ofimpact tools.lnt J. ofSolids and Structures, 1968, v. 4.
94. Fufaev N A, Theory of electromagnetic contact breakers. Coll. "Memoirs for A A Andronov",
1955. (In Russian).
95. Gantmakher F R, Lectures on analytical mechanics. Fizmatgiz, 1961. (In Russian).
96. Garber E D, Rozenvasser E N, On the study of periodic regimes in systems based on the filter
hypothesis. Automation and telemechanics, 1965, No. 2. (In Russian).
97. Gardner M F, Bums J L, Transients in linear systems, vo!.1. John Wiley, NY, 1942.
98. Gelfand I M, Shilov G E, Generalised functions and their applications. Fizmatgiz, 1958. (In
Russian).
99. Gelfond A 0, Calculation using finite differences. Fizmatgiz, 1959. (ln Russian).
100. Glatenok 1 V, On the estimation of domains of occurrence of real periodic solutions found by the
harmonic balance method. Automation and telemechanics, 1957, No. 12. (In Russian).
101. Glukharev K K, Frolov K V, Interaction of oscillating systems with two energy sources.
Mechanics ofsolids, 1971, No. 4. (ln Russian).
102. Goldfarb L S, On some nonlinearities in control systems. Automation and telemechanics, 1947,
103. Goldsmith W, Impact. The theory and physical behaviour of colliding solids. London
104. Gorokhov V A, On periodic motion of systems with impact interactions. Radio physics, 1966,
105. Grubin C, On the theory ofthe acceleration damper, 1. of App!. Mechanics, 1956, No. 3.
106. Hertz E V, Kreinin G V, Theory and design of pneumatically powered systems. Mashgiz, 1960.
(In Russian).
107. Hsu J C, Meyer A U, Modem control principles and applications. McGraw-Hill, New York,
1968.
108. lonov V N, Ogiba10v P M, Stresses in bodies due to impulse loading. "Vysshaia Shkola", 1975.
(In Russian).
109. lorish Yu I,The protection ofaircraft equipment from vibration. Oborongiz, 1949. (In Russian).
110. Israi10vich M Y, The optimallaw ofmotion for single mass oscillating systems. Machine studies,
1969, No. I. (In Russian).
111. Katkovnik V Y, Vibro-impact excitation of high frequency elastic oscillation. Mechanics of
solids, 1968, No. 5. (In Russian).
112. Kazakov I E, Approximate methods of statistical study of nonlinear systems. Zhukovsky
Aviation Academy, 1954, No. 394. (In Russian).
113. Kazakov I E, Dostupov B G, Statistical dynamics of nonlinear automatic systems .. Fizmatgiz,
1962. (In Russian).
114. Kharkevich A A, Self-sustaining vibration. Gostekhizdat, 1953. (In Russian).
115. Kilchevsky N A, Theory of collisions for solids. "Naukova dumka", Kiev, 1969. (In Russian).
116. Kobrinsky A A, Dynamics and stability of one-dimensional vibro-impact systems. Mechanics of
solids, 1970, No. 5. (In Russian).
117. Kobrinsky A E, Dynamics of mechanisms with elastic connections and impact systems. lIiffe
Books Ltd., London, 1969. (Translated from Russian, Nauka, Moscow, 1964.)
118. Kobrinsky A E, Kobrinsky A A, Vibro-impact systems. Nauka, Moscow, 1973. (In Russian).
119. Kobrinsky A E, On the theory ofimpact vibration absorbers. OTN, 1957 No. 5. (In Russian).
120. Kobrinsky A E, Oscillation oftwo mass systems, moving with periodic collisions. OTN, 1956,
121. Kobrinsky A E, Some problems in the theory of vibro-impact systems. In "Collection of
problems in the theory ofmachines and mechanisms", Nauka, Moscow, 1965. (In Russian).
Bibliography 283
122. Kochenburger R J, Nonlinearity in control systems. Control Engineering, 1956, No. 3.

123. Kolovsky M Z, Automatie control ofvibration protection systems. Nauka, Moscow, 1976, (In
Russian).
124. Kolovsky M Z, Nonlinear theory of vibration protection systems. Nauka, Moscow, 1966. (In
Russian).
125. Kolovsky M Z, On the application of the small parameter method to the calculation of
discontinuous periodic solutions. Int. symp. on non linear vibration .. AN USSR, Kiev, 1963, v. I.
(In Russian).
126. Kolovsky M Z, On the design ofnonlinear vibration isolators with one degree offreedom. OTN,
Mechanics and machine engineering, 1960, No. 4. (In Russian).
127. Kolovsky M Z, On the existence conditions for periodic regimes in systems of differential
equations with discontinuities, consisting of a sm all parameter. App!. maths. and mechanies,
128. Kolovsky M Z, Pervozvansky A A, On linearisation using the distribution funetion in problems
ofnonlinear oscillation. OTN, Mechanies and maehine engineering, 1962, No. 5. (In Russian).
129. Kolovsky M Z, Pervozvansky A A, On the stability of solutions found using the statistical
Iinearisation method. OTN, Meehanics and machine engineering, 1963, No. I.
130. Kolovsky M Z, Pervozvansky A A, The applieation ofthe distribution function method for poly-
harmonie solution of nonlinear equations. Proe. of Vuzov, Mashinostroenie, 1963, No. 4. (In
Russian).
131. Kolsky G, Stress waves in solids. Clarendon Press, Oxford, 1953.
132. Kononenko V 0, Oscillating systems with restricted exeitation. I1iffe, London, 1969. (Translation
from Russian, Nauka, Moscow, 1964)
133. Kononenko V 0, Problems in the theory of dynamie interaction of machines and sources of
energy. Mechanies ofsolids, 1975, No. 5. (In Russian).
134. Kosterin Yu I, Meehanieal seJf-exeited vibration due to dry frietion, AN USSR, 1960. (In
Russian).
135. Kotelnikov V A, Longitudinal stability of aireraft with autopilots. Trudy LII, M. 1941, No 2. (In
Russian).
136. Kozhevnikov S N, Dynamies ofmaehines with elastic eomponents. AN USSR, Kiev, 1961. (In
Russian).
137. Krasovsky N N, Theory ofmotion eontro!." Nauka, Moseow", Moseow, 1968. (In Russian).
138. Krylov N M, Bogolyubov N N, Introduction to nonlinear meehanics. Prineeton U P,
I 947.(Translation from Russian, Kiev, AN USSR, 1937)
139. Kulikovsky P, Optimal and adaptive processes in automatie eontrol systems. Nauka, Moseow,
1967. (In Russian).
140. Lavendel E E, Synthesis ofoptimal vibration machines. "Zinatne", Riga, 1970. (In Russian)
141. Leeomu L F A, Regularisation du mouvement dans les machines. Cauthier-Villas et fils, Paris,
1898.
142. Lefshetz S, Differential equations: Geometrie theory. Wiley, NY, 1957.
143. Leongard A, Generalised methods of harmonie balance in the analysis of parametrie vibration.
Proe. II congress IFAK, Nauka, Moscow, 1965, No. I.
144. Letov A M, Analytic design of controllers, I - IV Automation and telemechanies, 1960, No. 4 -
No. 6; 1961, No. 4. (In Russian).
145. Lieber P, Jensen D, An acceleration damper: development, design and some applications. Trans
ASME, 1945, v. 67.
146. Lifshits P S, On the problem of forced vibration of a systems impacting against a stop. Teeh.
Phys., 1952, No. 6. (In Russian).
147. Litovchenko Ts G, Yakovenko Yu P, Analytical and struetural deseription of mechanical
transmission systems for automatie controllers with stops and c1earanees. Automation and
telemechanies, 1961, No. 8. (In Russian).
148. Loeb J D, De la mechanique Iineaire a la meehanique non-Iineaire. Ann. Telecommuns., 1950,
No. 2.
149. Loitsiansky L G, Lurie A I, Theoretical mechanies v. 11. Gostekhizdat. 1955. (In Russian).
150. Lukomsky S I, Investigation of working regimes of percussion hammers.
VNIISTROIDORMASH, 1959, No. 24. (In Russian).
151. Lurie A I, Operator calculus and its application in problems of mechanies. Gostekhizdat, 1951.
(In Russian).
152. Lurie A I, Some nonlinear problems in the theory of automatie contro!. Gostekhizdat, 1951 (In
Russian).
284 Bibliography
153. Lyusternik L A, Sobolev V I, Elements 01' functlOnal analysis. Nauka, Moscow, 1965. (In
Russian).
154. Maezawa S, Steady, forced vibration of unsymmetrical piecewise linear system. Bulletin of
JSME, 1961, No. 4.
155. Magnus K, Uber ein Verfahren zur Untersuchung nichtlinearer Schwingungs und Regelungs
Systeme. VDI-Forschungshelft, 1955, No. 45 I
156. Mandelstam L I. Complete works. AN SSSR. 1955, v. IV. (In Russian)
157. Masri S F, Analytical and experimental studies 01' multiple-unit impact dampers . .1. of the
Acoustical Soc. oF Amer., 1969. No. 5.
158. Masri S F, Stability boundaries ofthe impact damper..1. Appl. Meehanics, 1968, No. 2.
159. Masri, S F, Caughey T K, On the stability of the impact damper. .1. Appl. Mechanics, 1966, No.
3.
160. Meiman N I, Some problems on the distribution of zero in polynomials. UMN, 1949, No. 6. (In
Russian).
161. Merkin D R, Introduetion to the theory of stability of motion. Nauka, Moseow, 1971. (In
Russian).
162. Methods for harmonic Iinearisation in the design 01' nonlinear systems for automatie eontrol. Ed.
by Yu I Topeheev, "Mashinostroenie", M.. 1970. (In Russian)
163. Methods for the synthesis of nonlinear systems in automatie control. Ed. By S M Fedorov,
"Mashinostroenie", M., 1970. (In Russian)
164. Nagaev R F, Dynamics of a vibro-impact crusher with a pair of self-synchronising vibrators.
Meehanics and mechanical engineering, 1963, No. I (In Russian).
165. Nagaev R F, General problems of quasi-plastic impacl. Mechanies of solids, 1971, No. 3. (In
Russian).
166. Nagaev R F, Khodzhaev K Sh, Oscillation 01' a mcehanieal system with a periodic structure.
"FAN", Tashkent, 1973. (In Russian).
167. Nagaev R F, On the analytical description oflJuasi-plastic impacl. Mechanics ofsolids, 1970, No.
4. (In Russian)
168. Nagaev R F, Regular impulse motion 111 one-dimensional systems. Applied maths. and
mechanies, 1967, No 2.
169. Naumov B N, Theory of non linear automatie systems. Nauka. Moscow, 1972 (In Russian).
170. Neimark Yu I, Point mapping methods in the theory of nonlinear vibration, ch I-lll. Radio
physics, 1958, No.l, NO.2 .. No. 5 - No. 6. (In Russian)
171. Neimark Yu I, Point mapping methods in the theory 01' nonlinear vibration., Nauka, Moscow,
1972. (In Russian).
172. Neimark Yu I, Shilnikov L p, On the application 01' the small parameter method to systems of
differential equations with discontinuous right hand sides. Mechanics and mechanical
engineering, 1959, No. 6. (In Russian).
173. Nonlinear devices in automatic eontrol systems. Ed. by Yu I Topeheev, "Mashinostroenie", M.,
1970. (In Russian).
174. Oppelt W, Method for the frequency loeus tor controllers with triction. Z. VDl, 1948,90
175. Paget A L, Vibration in steam turbine buckelS and damping by impael. Engineering, 1937, v.
143:305-7.
176. Panovko Y G, Internal damping during oseillalion 01' clastie systems. Fizmatgiz, 1960. (In
Russian).
177. Pervozvansky A A. Random processes in nonlll1car aUlomatie systems. Fizmalgiz, 1962. (In
Russian).
178. Ponomarev S D, Biderman V L, Likharev ]( S, Makushll1 V M. Malinin N N. Feodosiev V I,
Strength in mechanieal engineering, v. 111. MashglZ. 1962. (In Russian)
179. Pontriagin L S, Boltiansky V G, Gamkrelidze R V. Mishenko E F, Mathematieal theory of
optimal proeesses. Wiley Interseience, NY, I 962.(Translation trom Russian).
180. Popov E P, On the approximate study ofselt~excited and foreed vibration. DAN SSSR, 1954, v.
XCV, NoS (In Russian).
181. Popov E P, Paltov I P, Approximate methods in thc study 01' nonlinear automatie systems.
Fizmatgiz, 1960. (In Russian).
182. Poschi T, Der Stoss. Handbuch der Physik. Hd. VI. Spnnger. 1926.
183. Pugachev V S, Introduction to theory ofprobabilily. Nauka, Moscow. 1968. (In Russian).
184. Pugachev V S, Theory of randolll functions allli thm application to problems in automatic
controL Fizmatgiz, 1960. (In Russian)
185. Ragulskene V L Vibro-illlpactsystems. Mintis, Vilnius. 1974. (In Russian).
Bibliography 285
IS6. Ragulskis K M, Meehanisms on vibration basis. Proc. A," Lit. SSR, Kaunas, 1963. (In Russian).
187. Rakhmatulin Kh A, Demianov Yu A, Stabilitv undcr intense short time loading. Fizmatgiz, 1961.
(In Russian).
188. Raskyn Kh I, I) Applieation o1'the moment mdhod to so me problems in vibration transportation.
2) On estimation 01' deviations 01' real laws trom the ideal. Col!.. "Problems in dynamies and
stability", "Zinatne", Riga, 1965, No. 16. (In Russian).
189. Riabov Yu A, On determining tields o1'applieation o1'the small parameter method in problems 01'
nonlinear oseillation. Proe. int. symp. on nonlinear oseillation, Kiev. 1963, v. 1. (In Russian).
190. Rosenberg L D, Kazantsev V F, Makarov L 0, Akhimovieh L F. Ultrasonie eutting. Proe. AN
USSR, 1962. (In Russian).
191. Rozenvasser E N, Oseillation 01' nonlinear systems. Nauka. Moscow, 1969. (In Russian).
192. Rozenvasser E N, Periodie non-stationary control systems. Nauka. Moscow. 1973. (In Russian).
193. Rusakov I G. Kharkevieh A A. Forced vibration 01' a system colliding against a stop. 1. 1'ech.
Phys .. 1942. Nos. 11 -12. (In Russian).
194. Sadek M M, The behaviou o1'the impact damper. Proc. Ins!. Mech Eng., 1965-1966, v. 180.
195. Samolenko AM, Averaging methods in systems with impulses. In book "Mathematieal Physies",
"Naukova dumka", Kiev, 1971, No. 9. (In Russian).
196. Schwartz L, Methodes mathematiques pour les scienees physillues. Paris, 1961 (In Freneh).
197. Sears J E, On the longitudinal impact of metal rods with rounded ends. Trans. Camb. Phil. Soe.,
1912. v. 21(2).
198. Shilov G E, Mathematieal analysis. Seeond special course. Nauka. Moseow, 1965. (In Russian).
199. Shoek and vibration handbook. Ed. By C M Hanis ami CE Crelk McGraw-Hili Co .. New York,
1975.
200. Sikarskye D 1." Paul B, Periodic motions 01' a two-body system sllbjected to repititive impact. J.
ofEng, tor Ind., 1969, No. 4.
201. Sneddon I N, Beny D S, Classical theory 01' elasticity. Handbuch der Physik, v.6, Ed. S Flugge.
Springer Verlag, Berlin. 1958.
202. Staerman I Y, Contaet problems in the theory 01' e1astieity. Gostekhizdat, 1949. (In Russian).
203. Stoker J, Nonlinear vibration in mechanieal ami eleetrieal systems. Interseienee Pub. NY, 1950.
204. Sveshnikov A A. Applied methods in the theory ofrandom nmctions. Nauka, Moseow, 196K (In
Russian).
205. Sysoev V I. Pendulum vibration absorber installation 01' tower type. Col!. "Study 01' dynamie
installations". Gosstroiizdat. 1957. (In Rllssian).
206. Takeda. On the eodlieient 01' restitution. Mekhanika. 19)(,. No. 12S2. (ln Russian).
207. Timoshenko S P, Vibration probkms in ~ngin~ering. D Van Nostrand Co. Ltd., 1955.
208. Timoshenko S P, Zur Frage nach der Wirkung ~ines Stosses auf einen Balken. Zeits. Math Phys.,
Bd. 62, 1913.
209. Topeheev Yu I, Analysis 01' s!ability and quality 01' nonlinear systems 01' automatie eontrol by the
logarithmic frequency eharaeteristie method. In book Theory 01' automatie eontro!. Ed. by. V V
Solodovnikov, "Mashinostroenie", M., 1969, v. J, part 1. (In Russian).
210. Tsuda K, Studies 01' oseillating systems. transmiUing energy in the presenee 01' angular gaps.
Col!. translations, "Mekhanika", M., 196 I, No. I, (ln Russian).
211. Tsypkyn Ya Z, On the eonnection betwe~n the equivaknt amplitication tactor 01' a nonlinear
element and its eharaeteristics. Automation ami tdemechanics. 1956. No.4. (In Russian).
212. Tsypkyn Ya Z, Relay automatie systems. Nauka, Moscow. 1974. (In Russian).
213. Tsypkyn Ya Z, Sampling systems and their application. Pergamon. Oxiord, 1962. (Translation
trom Russian).
214. Vavilov A A, Periodie mehods in the study 01' nonlinear systems. Energia, Leningrad,. 1970. (In
Russian).
215. Veits V L, Dynamies 01' maehine units. Mashinostroenie, Leningrad. 1969. (In Russian).
216. Veits V L, Koehura A E, The etreet oflimited ~"citation in machines with internal eombustion
engines. eh. 1,11. Vibrotekhnika. 1973, No. 2 (19). (In Russian).
217. Veluswami M A, Crossley F R E, Horvay G, Multipl~ impacts of a ball b~tw~en two plates, p II,
Mathematieal Modelling. 1. 01' Eng. For Industry. 1975, No. J.
218. Ventsel ES, Theory o1'probability. Nauka, Moscowl969.
219. Voloshenko-Klimovitsky Yu Y, Dynamie limits o1'yield. Nauka, Moscow 1965. (In Russian).
220. Vul1'son I I, Dynamie design 01' eyelie meehanisms. Mashinostroenie, Leningrad, 1976. (In
Russian).
221, Vul1'son I I, Kolovsky M Z, Nonlinear probl~ms in machine dynamies. Mashinostroenie.
Leningrad, 1968. (In Russian).
286 Bibliography
222. Zadeh L A, On the identification problem. Trans IRE, 1956, No. 4.

223. Zaretsky L B, Synchronisation of centrifugal excitation in systems with discontinuous
characteristics. Mechanics of solids, 1968, No. 1. (In Russian).
224. Zevin A A, On the stability of periodic motion due to the collision of masses. Machine studies,
225. Zevin A A, Optimal regimes in vibro-impact systems having one degree of freedom. Mechanics
of solids, 1976, No. 2. (In Russian)
226. Zevin A A, Periodic regimes in systems with impact vibration absorbers. Mechanics ofmachines,
1969, No. 17-18. (In Russian).
227. Zhivkov V S, The influence of electromagnetic inertia of motors on the stability of oscillation of
mechanical systems with centrifugal excitation. Machine Studies. 1971, No.4. (In Russian).
Additional bibliography
This list compiles the publications that arrived after the first edition of the book
was published. We have tried here to present both development ofthe ideas ofthe
book and achievements in other directions. Preference was given to publications
where analysis is supplemented with physical experiments or engineering
applications.
228. Astashev V K, About the influence of high frequency vibration on the processes of plastic
deformation. Machine Studies, 1983, No. 2. (in Russian).
229. Astashev V K, Effect of ultrasonic vibration of a single-point tool on the process of cutting. 1. of
Machinery Manufacture and Reliability, 1992, No. 3. (in Russian).
230. Astashev V K, Babitsky V 1. Dynamics ofresonant machines. In book "Dynamics and control of
machines". Ed. Kreinin G V, Machinostroenie, Moscow, 1988, (in Russian).
231. Astashev V K, Babitsky V I, Efficiency of resonant tuning in ultrasonic machines. Machine
Studies, 1981, No. 6. (in Russian).
232. Astashev V K, Babitsky V I, Methods of efficiency improvement of ultrasonic machines. Stanki i
Instrument (Machine Tools and TooIs) 1982, No. 3. (in Russian).
233. Astashev V K, Babitsky V I, Methods of resonant machine tuning. Machine Studies, 1982, No. I.
(in Russian).
234. Astashev V K, Babitsky V I, Ultrasonic cutting as a nonlinear (vibro-impact) process.
Ultrasonics, 1998, v. 36, No. 1-5.
235. Astashev V K, Babitsky V I, Hertz M E, Adaptive control system of ultrasonic machine.
Instruments & Control systems, 1977, No. 2. (in Russian).
236. Astashev V K, Krupenin V L, Tresviatsky AN, Experimental investigation ofsynchronisation of
impacts in continuous systems with variable quantity of impact pairs. 1. of Machinery
Manufacture and Reliability, 1996, No. 2. (in Russian).
237. Babitsky V I, Autoresonant mechatronic systems. Mechatronics, 1995, v. 5 No. 5.
238. Babitsky V I, Hand-held percussion machine as discrete non-linear converter. J. of Sound and
Vibration, 1998, v. 214, No. I.
239. Babitsky V I, Some applications of autoresonant strongly nonlinear vibratory systems. Machine
Vibration, 1992, v. 1, NO.2.
240. Babitsky V I, Borovkov B A, Stabilisation of resonance in vibro-impact machines. Machine
Studies, 1979, No. I. (in Russian).
241. Babitsky V I, Kovaleva A S, Optimal control of non symmetric vibro-impact systems. Mechanics
of Solids, 1988, No. I. (in Russian).
242. Babitsky V I, Kovaleva A S, Krupenin V L, Investigation of quasi-conservative vibro-impact
systems by averaging method. Mechanics of Solids, 1982, No. I. (in Russian).
243. Babitsky V I, Krupenin V L, Oscillations in strongly non linear systems. Nauka, Moscow, 1985.
(in Russian).
244. Babitsky V I, Krupenin V L, To the theory of parametric resonances in vibro-impact systems.
Mechanics ofSolids, 1978, No. 4. (in Russian).
245. Babitsky V I, Vedenova E G. Goihman LV, Manevich L I, Oscillating Iocalisation in non linear
beam system with periodic structure. Proc. USSR Academy ofSciences, 1985, v. 281, No. 2. (in
Russian).
288 Additional bibliography
246. Babitsky V I, Veprik A M, Krupenin V L, Vibro-impact phenomena due to Iimited oscillations of

one-dimensional elasto-connected particles. Proc. USSR Academy of Sciences, 1988, v. 300,
No. 3. (in Russian).
247. Babitsky V I, Veprik A M, Universal bumpered vibration isolator for severe environment. J. of
Sound and Vibration, 1988 (in print).
248. Babitsky V I, Vozniuk P D, Perminov S M, Excitation of wide band vibration using auto-
stochasticity. Machine Studies, 1986, No. 4. (in Russian).
249. Bansevicius R, Ragulskis K, Vibromotors. Mokslas, Vilnius, 1981. (in Russian).
250. Bapat C N, The general motion of an inclined impact damper with friction. J of Sound and
Vibration, 1995, v. 184, No. 3.
251. Bapat C N, Sancar S, PoppieweIl N, Repeated impacts on a sinusoidally vibrating table
reappraised. J of Sound and Vibration, 1986, v. 108, No. 1.
252. Bapat C N, PoppieweIl N, Several similar vibro-impact systems. J. of Sound and Vibration, 1987,
v. 113, No. 1.
253. Bapat C N, PoppieweIl N, MacLachlan K, Stable periodic motion of an impact pair. 1. of Sound
and Vibration, 1983, v. 87, No. 1.
254. Bapat C N, Sancar S, Single unit impact damper in free and force vibration. J of Sound and
Vibration, 1985, v. 99, No. 1.
255. Bishop S R, Impact oscillators. Phil. Trans. ofthe Royal Society, London, 1994.347 A.
256. Brach R M. Mechanical impact dynamics. Rigid body collisions. John Wiley & Sons. New York.
1991.
257. Brogliato B, Nonsmooth impact mechanics. Models, Dynamics and Control. Lectures notes in
control and information sciences 220, Springer-Verlag, London, 1996.
258. Budd C, Dux F, Chattering and related behaviour in impact oscillator. Phil. Trans. Royal Society,
London, 1994, A347.
259. Budd C, Dux F, Cliffe A. The effect of frequency and c1earance variations on single-degree-of-
freedom impact oscillator. J. of Sound and Vibration, 1995, v.184, No. 3.
260. Cabannes H. Cordes virantes avec obstacles. Acustica, 1984, v. 55.
261. Cabannes H, Haraux A, Almost periodic motion of astring against tixed obstacle. Nonlinear
Analysis: Theory, Methods and Applications, 1983, No. 7.
262. Chatterjee S, Mallik A K, Ghosh A, Impact dampers for controlling selt~excited oscillation. 1. of
Sound and Vibration, 1996, v. 193, No. 5.
263. Chu F, Zhang Z, Bifurcation and chaos in a rub-impact Jeffcott rotor system. J of Sound and
Vibration, 1998, v. 210, No. 1.
264. Comparin R J, Singh R, Nonlinear frequency response characteristics of an impact pair. J of
265. Dimentberg M, Statistical dynamics of nonlinear and time-varying systems. Research Studies
Press, Taunton, 1988 (Translation from Russian, Nauka, Moscow, 1980)
266. Feigin M I, Forced oscillation in systems with discolltinuous nonlinearities. Nauka, Moscow,
1994 (in Russian).
267. Foale S, Bishop S, Bifurcations in impact oscillator. Nonlillear Dynamies, 1994. v. 6.
268. Haug E J, Wu S C, Yang S M, Dynamics alld mechanical systems with Coulomb friction,
stiction, impact and constraint addition-deletion I. Mechanisms and Machine Theory, 1986, v. 21,
No. 5.
269. Heiman M S, Sherman P J, Bajaj A K, On the dYllamics and stability of an inclined impact pair.
J. ofSound and Vibration, 1987, v. 114, No. 3.
270. Hertz M E, Makarov N S, Tresviatsky A M, Experimental investigation of autoresonant system
with eleetro-dynamic exeitation. Maehine Studies, 1989, No. 3. (in Russian).
271. Hogan S J, On the motion of a rigid block, tethered at one corner, under harmonic forcing. Proc.
Royal Society, 1992, London A 439.
272. Hogan S J, Heteroclinic bifureations in damped rigid block oscillations. Proc. Royal Society,
1992, London A 439.
273. Hogan S J, Siender rigid block motion. 1. ofEngineering Mechanics, 1994, 120.
274. Hogan S J, On the dynamics of rigid block motion under harmonie forcing. Proc. Royal Soc,
1989, London A 425.
275. Holmes P J, The dynamics ofrepeated impacts with sinusoidally vibrating table. J of Sound and
Vibration, 1982, v. 84, No. 2.
276. Hongler M 0, Streit L, On the origin of c'mos in gearbox models. Physica, 29D, 1988.
Additional bibliography 289
277. Ivanov A P, Dynamics of systems with mechanic collisions. Int. Programme of Education,
Moscow, 1997. (in Russian).
278. Ivanov A P, Bifurcations in impact systems. Chaos, Solitons and Fractals, v. 7, No. 10.
279. Ivanov A P, Bifurcations associated with grazing collisions. In Nonlinear dynamics: New
theoretical and applied results. Academic Verlag, Berlin, 1995.
280. Ivanov A P, Impact oscillations: linear theory stability and bifurcations. 1. of Sound and
Vibration, 1994, v. 178, No. 3.
281. Jean M, Moreau J J, Dynamics in the presence of unilateral contacts and dry friction; a numerical
approach. In "Unilateral problems in structural analysis 2", Ed. G. Dei Piero and F. Maceri, 1987,
CISM courses and lectures 304, Springer-Verlag.
282. Kahraman A, Singh R, Nonlinear dynamics of a spur gear train. 1. of Sound and Vibration, 1990,
v. 142, No. 1.
283. Kahraman A, Singh R, Nonlinear dynamics ofa gear rotor system with multiple cJearances. 1. of
284. Kahraman A, Singh R, Dynamics of an oscillator with both clearance and continuous
nonlinearities. 1992, v. 153, No. 1.
285. Karagiannis K, Pfeiffer F, Theoretical and experimental investigations of gear rattling. Nonlinear
Dynamics, 1991, v. 2.
286. Karyeaclis M P, Caughey T K, Stability of a semi-active impact damper: part 1- Global
behaviour. ASME J. of Applied Mechanics 1989, v. 56.
287. Karyeaclis M P, Caughey T K, Stability of a semi-active impact damper: part 2- Periodic
solutions. ASME 1. of Applied Mechanics 1989, v. 56.
288. Keller J B, Impact with friction. ASME J. of Applied Mechanics, 1986, v.53.
289. Keller J B, Impact with an impulsive frictional moment. ASME 1. of Applied Mechanics, 1987,
v.54.
290. Khulief Y A, Shabana A A, A continuous force model for the impact analysis of flexible
multibody systems. Mechanisms and Machine Theory, 1987, v. 22.
291. Kobrinsky A A, Kobrinsky A E, Two-dimensional vibro-impact systems. Nauka, Moscow, 1981.
(in Russian).
292. Kovaleva A S, Control ofvibratory and vibro-impact systems. Nauka, Moscow, (in Russian).
293. Kovaleva A S, Optimal control of periodic motions in vibro-impact systems. Mechanics of
Solids, 1982, No. 6. (in Russian).
294. Kovaleva A S, Optimisation of periodic motions of some vibro-impact systems. Mechanics of
Solids, 1983, No. 3. (in Russian).
295. Kovaleva A S, Investigation of vibro-impact resonances in systems with random parametric
excitation. Machine Studies, 1983, No. 2. (in Russian).
296. Krupenin V L, About oscillation of vibro-impact systems with distributed impact elements.
Machine Studies, 1985, No. 4. (in Russian)
297. Krupenin V L, To the calculation ofresonant oscillation ofthread interacting with point limiter. J
ofMachinery Manufacture and Reliability, 1992, No. 2. (in Russian).
298. Krupenin V L, To the theory of vibro-impact systems with distributed impact elements.
Mechanics ofSolids, 1986, No. 1. (in Russian).
299. Krupenin V L, Transformation of modes in astring interacting with two lengthy limiters. Proc.
USSR Academy ofSciences, 1990, v. 313, No. 6. (in Russian).
300. Kumabe D, Vibratory cutting. Dzikke Sjuppan, Tokio, 1979. (in Japanese).
301. Kurilo R, Ragulskene V, Two-dimensional vibration drives. Mokslas, Vilnius, 1986. (in
Russian).
302. Lancarani H M, Nikravesh P E, Continuous contact force models for impact analysis in
multibody systems, Nonlinear Dynamics, 1994, No. 5.
303. Landa P S, Stratonovich R L, Calculation of stationary distribution function for one of the
slmplest strange attractors. Proc. USSR Academy of Sciences, 1982, v. 267, No. 4. (in Russian).
304. Landa P S, Stratonovich R L, Probability characteristics of stochastic oscillations for controlled
pendulum. Mechanics ofSolids, 1984, No. 4. (in Russian).
305. Luo G.-W, Xie J.-H, Hopf bifurcation of a two-degree-of freedom vibro-impact system. J. of
306. Manevish L I, Michlin J V, Pilipchuk V N, Method of normal forms for essentially nonlinear
systems, Nauka, Moscow, 1989. (in Russian).
307. Moore D B, Shaw S W, The experimental response of an impacting pendulum system., Int. 1.
Non-linear Mechanics, 1990, v. 25.
290 Additional bibliography
308. Moreau J J, Standard inelastic shocks and the dynamics of unilateral constraints. CISM courses
and lectures, 1985, No 288, Springer-Verlag.
309. Moreau J J, Some numerical methods in multibody dynamics: applications to granular materials.
European J. ofMechanics AlSolids, 1994, v.13, No. 4.
310. Moreau J J, Panagiotopoulos P D, (Eds) Nonsmooth mechanics and applications. CI SM courses
and lectures, 1988, No 302, Springer-Verlag.
311. Muszunska A, Rotor-to-stationary element rub-related vibration phenomena in rotating
machinery - literature survey. The Shock and Vibration Digest, v. 21, No. 9.
312. Nagaev R F, Mechanical processes with repeated attenuated impacts. Nauka, Moscow, 1985. (in
Russian).
313. Natsiavas S, Stability and bifurcation analysis for oscillators with motion limiting constraints. J.
ofSound and Vibration, 1990, v. 141, No. I.
314. Natsiavas S, Dynamics ofmultiple-degree-oftreedom oscillators with colliding components. J. of
315. Neimark J 1, Landa P S, Stochastic and chaotic oscillations. Kluwer, 1992. (Translation from
Russian, Nauka, Moscow. 1987).
316. Nordmark A B, Non-periodic motions caused by grazing incidence in an impact oscillator. J. of
317. Paidoussis M P, Li G X, Moon F C, Chaotic oscillations of the autonomous system of a
constrained pipe conveying fluid. J. of Sound and Vibration, 1985, v. 135, No. I.
318. Panovko J G, Introduction in the theory of mechanical impact. Nauka, Moscow, 1977. (in
Russian).
319. Peterka F, Introduction to oscillation of mechanical systems with impacts. Academia, Prague,
1981. (in Czech).
320. Peterka F, Vacik J, Transition to chaotic motion in mechanical systems with impacts. 1. of Sound
and Vibration, 1992, v. 154, No. I.
32\. Pfeiffer F, Glocker Ch, Multibody dynamics with unilateral contacts. Wiley-Interscience, New
York, 1996.
322. PoppieweIl N, Bapat C N, MacLachlan K, Stable periodic vibro-impacts of an oscillator. J. of
Sound and Vibration, 1983, v. 87, No.\.
323. Sharif-Bakthiar M, Shaw S W, The dynamic response of a centrifugal pendulum vibration
absorber with motion Iimiting stops. J. ofSound and Vibration, 1988. v. 126. No. 2.
324. Shaw S W, Holmes P J, A periodically forced impact oscillator with large dissipation. 1. of Appl.
Mechanics, 1983, v.90, No. I.
325. Shorr BF, Melnikova G V Analysis of structures by direct mathematical simulation.
Mashinostroenie, Moscow, 1988 (in Russian)
326. Sinopoli A, Unilaterality and dry triction: A geometric formation for two-dimensional rigid body
dynamics. Nonlinear Dynanics, 1997. v. 12, No. 4.
327. Stoianovici D, Hurmuzlu Y, A critical study of the applicability of rigid body collision theory.
ASME J. of Applied Mechanics, 1996, v. 63, No. 2.
328. Stronge W J, Rigid body collision with friction. Proc. Royal Society, London, 1990, A 43 I.
329. Stronge W J, Friction in collisions: resolution ofa paradox. 1. ofAppl. Phys., 1991, v. 69, No. 2.
330. Thompson J M T, Ghaffari R, Chaos after period-doubling bifurcations in the response of an
impact oscillator. Phys. Lett., 1982, 91 A(l).
331. Thompson J M T, Ghaffari R, Chaotic dynamics of an impact oscillator. Phys. Rev. A, 1983,
27(3).
332. Veden ova E G, Manevich L I, Periodical and localised waves in vibro-impact systems with
regular structures. Machine Studies, 1984, No. 4. (in Russian).
333. Veprik A M, Krupenin V L, About resonant oscillations of a system with a distributed impact
element. Machine Studies, 1988, No. 6. (in Russian).
334. Veprik A M, Krupenin V L, On the solution of the system containing distributed impact
elements. Machine Studies, 1985, No. 6. (in Russian).
335. Watanabe T, Forced vibrations of continuous systems with non linear boundary conditions.
ASME J. ofMechanical Design, 1978, v. 100.
336. Wehage R A, Haug E J, Dynamic analysis of mechanical systems with intermittent motion.
ASME J. ofMechanical Design, 1982, v. 104.
337. Whiston G S, Global dynamics of vibro-impacting linear oscillators. J. of Sound and Vibration,
1987, v. 118, No. 3.
Additional bibliography 291
338. Whiston G S, Singularities in vibro-impact dynamics. 1. of Sound and Vibration, 1992, v. 152,
No. 3.
339. Wood L A, Byme K P, Analysis of a random repeated impact process. 1. of Sound and
Vibration, 1981, v.78, No. 3.
340. Wu S C, Yang S M, Haug E J, Dynamics and mechanical systems with Coulomb friction,
stiction, impact and constraint addition-deletion 11, Planar Systems. Mechanisms and Machine
Theory, 1986, v. 21, No. 5.
341. Wu S C, Yang S M, Haug E J, Dynamics and mechanical systems with Coulomb friction,
stiction, impact and constraint addition-deletion Ill, Spatial Systems. Mechanisms and Machine
Theory, 1986, v. 21, No. 5.
342. Zhong Z H, Mackerle J, Contact-impact problems: A review with bibliography. Applied
Mechanics Reviews, 1994, v. 47, No. 2.
343. Zhuravlev V F, Methods of analysis of the vibro-impact systems with the help of special
functions. Mechanics ofSolids, 1976, No. 2. (in Russian).
344. Zhuravlev V F, Motion equations of mechanical systems with unilateral constraints. Appl.
Mechanics and Mathematics, 1978, No. 5. (in Russian).
345. Zhuravlev V F, Klimov D M, Applied methods in oscillation theory. Nauka, Moscow, 1988. (in
Russian).
Subject Index
Absorber, impact, 243 Coefficient of restitution, 14, 15, 18,

Absorber, vibration, 79, 243, 245, 20, 23, 26, 102, 136, 156, 202,
250 207, 240, 242, 245, 253, 262,
Amplitude pulling, 52, 157, 207, 269 et seq.
238,243 Collision of spheres, 11
Autonomous system, 13, 144, 145, Collision velocity, 2, 117
151,168,252 Compliance, dynamic, 81 et seq.,
Autoresonance, 166, 271, 273, 275 92,193,194 et seq., 227 et seq.,
et seq. 240,273
Autoresonant frequency, 273, 275 Constraints, impact, 69, 70
Autoresonant system, 271 Contact zone, I, 3, 6, 8, 10, 11, 12,
Axial impact, 8, 12 199,228
Backbone curve, 22, 29, 34, 35, 37, Control characteristic, 253, 262
41, 48, 51, 58, 60, 128, 130 et Control function, 252 et seq.
seq., 142, 163, 167, 170,215 et Control point, 273, 276
seq., 233, 235, 236, 275, 278 Convergence, 82,89,93
Beam slapping modes, 237 Damping, 39, 45, 55, 59, 60, 61, 90,
Beam, vibrating, 229 116, 134 et seq., 146, 149, 156,
Bearing,47 168, 170,206,213,227,245 et
Boundaries, stability, 36 seq.
Characteristic, linear static, 13 Damping, internal, 6, 7, 28, 225,
Characteristic, amplitude-frequency, 226,231,235,237,278
26,27,34,37,40,47,48,49,51, Decay, 143
58,98, 127, 129, 131, 132. 143, Determinant, 32, 73,80,87,155
157,162,174,176 Difference equation, 11, 31, 71, 268,
Characteristic equation, 32, 44, 67, 269
86, 87, 89, 126, 130, 141, 145, Dirac function, 77, 100, 111
151, 154, 166, 167, 195, 197, Discontinuity surface, 70, 71, 72
230, 231, 274 Dissipation, impact, 28, 45, 46, 59,
Characteristic of contact interaction, 60
10 Dissipation, total, 80, 82, 89
Characteristics, dynamic, 76, 171 Dissipation, viscous, 45
Characteristics, force, 10, 75, 76, 78, Dissipative effects, 11, 20, 44, 45,
82, 92, 102 et seq., 117 et seq., 201
149,152,167,206,273 Distributed impact element, 237
Characteristics, kinematic, 1, 13 Drift, 37, 125, 140, 172, 252, 261,
Clapping, 224 273
Subject Index 293
Dry friction, 15, 18, 19, 79 Holonomic systems, 80

Elastic modulus, 1 Hysterisis, 6, 10,271
Energy balance, 44,59,60,78, 137, Hysterisis loops, 6
168,203 et seq., 239, 242, 276 Impact impulse, 36, 89, 92, 207,
Energy boundary of vibration, 44, 239,242,253,259,263,269
45,46,52,60,98, 127, 133, 162, Impact oscillator, 125, 173,207,238
168,172,217,222,275,278 Impact pair, 21, 75 et seq.,91, 98,
Energy conditions for stability, 141, 102, 125, 127, 132, 135, 139,
152, 168, 196 et seq., 276 152, 167, 199, 206, 207, 229,
Equivalent linearisation, 99, 103, 244
194,244,272 Impact theory, 1
Escapement, watch, 64 Impact velocity, 140, 253, 262, 263,
Excitation frequency, 28, 41, 42, 45, 264,269,270
48, 135, 140, 143, 157, 163,220, Impact, constrained, 16 et seq.
225,227,237 Impact, oblique, 15, 16
Excitation, external, 29, 45, 83, 108, Impact, sliding, 16, 79
116, 118, 125, 144, 159, 160, Impulse-momentum equation, 2, 15
194, 197,207,240 et seq., 85
Excitation, hard, 24, 46, 129 et seq., Impulse response function, 11
157, 161, 169, 173, 207, 220, Inertia, 14, 16, 19, 68, 136, 149,
228,237,264 170,209,230,239
Excitation, soft, 155, 168 Integral equations, 10, 83, 94
Existence domains, 25 et seq., 33, Integral-differential equations, 6
36, 40 et seq., 52, 55, 57, 59, Interaction, contact, 8 et seq., 757t6,
133, 143, 169, 170, 218, 220, 84,94,108
224,236 Interaction, symmetrical double
Feedback, 271 et seq. sided,75
Fi1tering effect, 86, 96 Interference, 21, 25, 28 et seq., 35,
First harmonic, 23, 36, 42, 43, 45, 54,66
49,59,96,103,151,245,274 Jamming, 16, 18,65,66
Forced vibration, 23,29, 36, 37, 39, Kinetic energy, 1, 14,80
52, 55, 91, 98, 125, 132, 137, Laplace trans form, 81 et seq.
140, 143, 136, 151, 157, 160, Linear oscillator, 20 et seq., 34, 37,
174,251,260 40,47,98, 125, 130, 132 et seq.,
Fourier series, 21, 36, 38, 48, 89, 92 140, 198, 243
et seq., 128, 132, 1327, 162, 168, Linearisation, 66, 75, 94, 97 et seq.,
204,220,241,274 125 et seq., 142, 147, 151 et seq.,
Free vibration, 22, 29, 34, 37, 38, 167, 171, 176, 230, 243, 272,
45,49,52,233,239 274
Functions, generalised periodic, 89 Linearisation by distribution
Gain coefficient, 49, 83, 154 function, 99 et seq., 102
GeneraJised coordinates, 69, 80, Linearisation coefficients, 98 et seq.,
166,244 113, 121, 127, 128, 151 et seq.,
Grazing, 224 161, 171, 176, 198, 272
Harmonic vibration, 34, 40, 42, 46 Linearisation, force, 105, 121
et seq., 59,103,133,151,172 Linearisation, harmonic, 51, 94, 98
Hertz equations, 8 et seq., 115, 117, 121, 125, 132,
294 Subject Index
140 et seq., 167 et seq., 194, 198, Pulling, 48,51 et seq., 130, 133,156
214,219,230,243,248, 272 ~ et seq., 164, 220, 237
274,277 Random vibration, 108, 250
Linearisation, motion, 105, 121 Regime, periodic, 39,43,46, 49, 51,
Linearisation, statistical, 100, 113, 65, 67, 70, 151, 152, 160,253,
117,121,123,250 261,267,270
Longitudinal vibration, 3 Regime, symmetrieal, 48
Lower boundaries, 44, 46 Resonance curve, 29, 34 et seq., 51,
Moment, central, 109 55, 58, 60, 61, 132, 142, 163,
Moment problem, 256, 258 217 et seq., 226, 236, 275
Momentum, conservation of, 4, 7, Resonance, isochronous, 35
13 Resonance, nonlinear, 42, 243
Motion, stable, 141 Resonance, parametrie, 148 et seq.
Natural frequency, 7,21,38,48,90, Resonance, sub-periodic, 136,202
98,127,128,131,136,150,158, Resonant response, 86
162, 170, 199, 208, 217, 220, Response, 11, 81, 83, 86, 95, 206,
238,243,249 214 et seq., 240, 246, 248
Nominal point, 261, 266 Rigid body, 4
Nonlinear mechanics, 29, 148, 170 Screwthread wrench, 62
Nonlinearity, hard, 206 Self-sustaining system, 165, 166,
Nonlinearity, soft, 156 260,263,271
Operator, dynamic compliance, 81 Self-sustaining vibration, 165 et
et seq., 193,212 et seq., 221, 230 seq., 173, 251, 263, 272, 275
Optimal control function, 252 et seq. Single degree of freedom, 64, 125,
Optimal periodic motion, 251 207,238
Optimisation, 253, 259, 261, 272 Sommerfeld effect, 176
Oscillator with symmetrical stops, Stability, 30, 32 et seq., 42, 44, 66 et
132 seq., 92, 140 et seq.,161 et seq.,
Oscillator with viscous damping, 90 193, 196 et seq., 218, 231, 235,
Parametric vibration, 148 et seq., 248, 251, 260, 261, 263, 268,
160 271,276
Pendulum, 64 Stability, asymptotic, 32, 71
Percussion machine, 53, 54, 61, 251 Stable solution, 30, 153, 196, 206,
Periodic Green's functions, 90 et 220,231,236
seq., 137 et seq., 202 et seq., 239 Stationary process, 84, 110, 111 et
et seq. seq.
Perturbations, smalI, 30, 66, 70, 72, Stereomechanics, 13, 14, 53
196,231 Stitching method, 64, 73, 136, 260
Phase boundary, 19 Striker, 5, 16, 53, 62, 68
Phase plane, 21, 38, 50, 252, 260, Structural boundaries of vibration,
262,264 44 et seq.
Phase point, 70 Switching curve, 252, 261, 266 et
Phase trajectory, 50, 252, 262, 264, seq.
267 Switching point, 252, 261 et seq.
Point mapping, 71, 263, 265, 267 Taylor series, 30, 72, 159
Point mass, 64, 69, 79, 229, 237 Transfer function, 83, 85
Poisson's ratio, 2 Transformation, 51, 67, 70, 71, 84,
Subject Index 295
85,88,94, 103, 111 et seq., 121, Vibro-impact regime, 24 et seq., 35,

133, 142, 148 et seq., 160, 167, 40 et seq., 59,103,105,107,129
199,201,263,277 et seq., 156 et seq., 168, 169,
Transient processes, 50, 143, 175, 173,201 et seq., 209, 218, 223,
261,263 225,227,228,236,245
Transition, 50,51, 77,103,207,238, Visco-elastic bar, 212
266 Viscous friction, 45, 59, 102, 108,
Unstable solution, 30, 142, 156, 163, 118,133,140
175,223 Waves, longitudinal, 2, 212
Vibration, structural boundaries of, Wave reflection, 5
44 Waves, shear, 2
Vibro-impact interactions in elastic Waves, standing, 7, 207, 237
systems, 212 Waves, surface, 3
Appendix I Simulation of impact on viscoelastic Iimiter
I. Numerous models based on the application of mechanical elements have

been developed to explain the viscoelasticity. They use successive and parallel
combinations of weightless springs and dashpots to retlect the elastic and viscous
behaviour of material.
Figure A 1.1 demonstrates the model ofaxial collision of a free rigid body of
mass Mwith the simplest viscoelastic limiter. The limiter is modeled schematically
as a parallel working linear spring with elasticity coefficient c and dashpot with
viscous coefficient b. Such a model describes accurately the strain and
deformation occuring in a viscoelastic sampIe under dynamic loading. This is
known as a Kelvin-Voigt model of viscoelastic material. The approaching velocity
of a free body with mass M is denoted as v .
This model of impact has an exact analytical solution and can be conveniently
used, therefore, as a test model for the development of PC-oriented simulation
procedures, such as Matlab-Simulink.
v
•
o M
•
o x
Fig. Al I
Dynamic response of the limiter F = cx + bx governs the development of

collision. The differential equation of body motion during the impact is
Mx + cx + bx = 0 t E [0, t], (A l.l)
with initial conditions: x{O) = 0, x{O) = v , where t is the duration of impact.
The condition of the body rebound in instant t = t is the change of the sign of
the limiter response or ofthe body acceleration:
F{t)=O or x(t)=O. (Al.2)
The equation of impact (A 1.1) has a known solution:
AI-2 Appendix I
x(t) = A exp( -ht)sin Qt, tE [0,,], (Al.3)

where:
h=~ A=~n2=n2_h2 n2=~.
2M' 0.' 0 ' 0 M
The corresponding derivatives with respect to time have been obtained as folIows:
x(t) = Aexp(- ht XncosQt - hsin nt)
x{t) = -Aexp{- ht)[(n 2 - h 2)sinQt + 2hncosnt] (A1.4)
X-{t) = -Aexp(- ht )[(3hn 2 - hJ )sin nt + (3h 2n - 0. 3 )cosnt]
The instant of impact completion can be found by solving the equation
x{,)= -Aexp{- h, )[(0. 2 - h2)sin 0., + 2hncosn,]= 0
from which
I -2hn
, = - arctan 2 2 '
0. 0. -h
or, using the notation for fracture of critical damping (loss factor) ~=~
0. 0
, =-
1
~ arctan
2~R
2' (AI.5)
0. 0 " 1_ ~ 2 1- 2~
The peak value of acceleration may be either at the instant collision starts (due
to the immediate step-wise response F = vb or at the instant when X-{t) = 0 . The
value ofthe initial acceleration may be calculated as folIows: Mx{O) = vb and
x,,()
0 = -vb = 2v~no . (A1.6)
M
The maximum value of acceleration can be achieved in the moment when
X'{t = 'I) = 0 if 0<, I <, , therefore using results of (A IA) for the third-order
derivative we have
(3hn z -hJ)sinn'l +(3hzn-nJ)cosn'l =0 and 'I =-arctan

I
0.
(3h
- n-n
3hn -h
Z
2 3
J
)
or
1 ~~2_1~
'I =- ~ arctan ( 2 J (Al.7)
no"I-~2 ~3-4~
By substituting the value 'I from (A 1. 7) into the expression for the acceleration
from (AIA) we obtain after rearrangement the following maximum value of
acceleration
x(t 1 )=
=
vQo [~
r:--::2 arctan
o( - 4~ z ) exp - ,,1 - ~ z
(4~2 (-I~l cos arctan (4~2 (-1~1
0 - 4~ - J
>\
[ 2\
~ 3 - 4~ J
(A1.8)
Finally, combining (A 1.6) with (A 1.8), we obtain for the maximum of acceleration
Appendix I AI-3
where
r(~) = {y(~) if y(~) ~ 2~ (AU 0)
2~ if y(~) < 2~
with
y(~)= ( I
~3-4~
2 t [~
~ arctan (4~2 CI~l
XP -
VI-~2
:) cos[ arctan (4~2 (- I~l
~3-4~
-) .
~3-4~2
The value of ~ = 0.5 is the exact solution for the equation y(~) = 2~ . That means,
that for the viscoelastic limiter with ~ > 0.5 the peak of acceleration will be at the
initial instant of collision.
The peak deflection will be when i (I = ' 2 ) = 0 (, 2 < ,), therefore
with the help of expression for the first-order derivative from (AI.4) we have:
1 0
OCOSO'2 - hsinOt 2 = 0 and '2 = -arctan-
o h
or
arctan R
~ (AI.1l)
t2=--~"=~-
0 0 1- ~2
By substituting the corresponding values of the time instant t 2 from (A 1.11)
into the expression for the deflection (A 1.3) and after rearrangement the following
peak value was obtained:
x(t 2 ) = r:--::2
v
00VI-~2
(~
exp - r:--::2 arctan
VI-~2
R] ( R]
~
sin arctan
~
or
(A1.12)
where
A(~)= p e xp(- 1_~2

R"CIan
1_~2
RJsin(arctan RJ.(Al.13)
~ ~
Figure Al.2 depicts the functions (A 1.1 0) and (Al. 13):

By using the expression for the velocity from (Al.4) and for the instant of
rebound (AI.5), we can calculate the velocity ofrebound:
• These results are somewhat similar to those obtained for the maximum deflection and transmitted
force for the single degree of freedom system with impulse excitation in the book "Shock and
vibration concepts in engineering design." by C E Crede. Prentice-Hall, NY, 1965.
AI-4 Appendix I
2.S
T,A
2 - T
-A
1.5
0.5
~
o L---------------------------------~--
o 0.1 0.2 0.3 0.4 0.5 0.$ 0.1 0.8 0.9
Fig. AU
x(r):::vex p[- h-arctan(-

I- e 2~F?llsin[arctan[-
I- 2~
2~F?ll
I- 2~
By definition, the coefficient of restitution will then be:
x(r) exp -
R::: ---:::
v
[~
vI -
~
~2
arctan [ - 2~gll
I- -
sin rarctan (2~gll
L l I-
2~
-? (ALI4)
2~
2
&...
...
"§ CL7
.~
~Q.I
50.4
~ o.)
!! cu
u
0.,
oL-------------------------------
o 0., 0.'o.s 0.] 0..' 0.' 0.1
L f,c'()l
Fig. AU
It appears that at viscoelastic impact the coefficient of restitution is a ftmction of

the bumper loss factor ~ only.
2. For experimental validation of the model of viscoelastic impact, the
experimental rig shown in Fig. A 1.4 was assembled. A plastic pad (E-A-R
Isodamp C-I 000 plastic, 6 mm thick) was mounted in a vice fixture with the plain
surface facing a pendulum. The weighted steel rod (010mm) with a rounded tip,
was fixed in a bifilar suspension at a point directly above the sampie vice. It was
Appendix I AI-5
swing up and then held in pre-start position. The height defined the approaching
speed ofthe pendulum at impact. When released, the pendulum impacted the pad.
Fig. A1.4
The process of collision was registered by a miniature accelerometer (type
4393, Bruel & Kjaer) mounted on the plane edge of the rod, charge conditioner
(type 501 OA, Kistler, not shown) and signal analyzer (model SD-390, Scientific
Atlanta, not shown). A typical example ofthe time history of acceleration obtained
is shown in Fig. Al.5.
7
I
,""'"'-
.F
\
\
;.) \
" \.
\
1
"" "-
~
Fig. Al.S
The characteristic instants of impact, such as the start instant, instant of peak
acceleration and impact duration may be easily measured by using the cursor ofthe
signal analyser. In this manner the impact duration T and time to peak
acceleration TI were measured as 3.83ms and 1.54ms respectively. Expressions
(AI.5) and (AI.7) were then solved as a set of transcendental equations with
unknown 00 and sand known values T and TI' Dividing (A 1.5) by (A 1. 7) we
can eliminate parameter 00 :

A 1-6 Appendix I
2~R--
arctan -~~-----
, 1- 2~2
(A 1.15)
'I 0~2-1~
arctan - ~(3--=-4~ 2T-
Solving firstly (A 1.15) for ~ and then (A 1.5) for ° 0 , we obtained

~ = 0.14, 00 = 755 rad / s .
The expression for acceleration from (A 1.4) was used further with the newly
detennined parameters 00 and ~ . The time history of acceleration was generated
numerically. The initially unknown impact velocity v was identified by
minimizing the integral mean square difference between measured and curve fitted
shape of impact. For this particular case it was found that v = 1.45 m/s _ Figure
Al.6 shows superimposed the almost coincident time histories of acceleration at
impact obtained from the experiment ("TEST") and that obtained theoretically
("CURVE").
100
00 "0
~. !!O
E "0
~ W
"t: 50
40
j/J
10
/0
°O~--~--~----~--~--~f~
' --~--~7 -WMM
nm~. ms
3. The differential equation of the motion (Al.I) was valid for the finite
interval of time covering the impact process. Expanding the previously obtained
results for the infinite time interval the following differential equation may be
written for unit mass:
x+<D(x,x)=o, tE[o,oo] (AI.16)
with initial conditions: x(O) = 0, x(O) = v , where the force characteristic of impact
interaction <D(x, x) is as folIows:
o~x + 2~00x if x > ° and <D(x,x) > 0,

<D(x,x) = l Oifx>O and <D(x,x):o:O,
if x < ° °
In accordance with (A 1.17), the impact interaction begins when the impacting
body crosses the zero coordinate level in the positive direction. The impact is
Appendix I AI-7
.---_~ D 0 1>-- - ---,
Fig. Al.?
completed when the value of the limiter response crosses its zero level in the
"negative" direction .
Figure AI. 7 depicts the schematics of the Simulink model representing
equation (A 1.16).
100
hg. ALS
The result of simulation with I; = 0.14,0 0 = 755rad / sand initial conditions

x(O) = 0, v = 1.45m / s is presented in Fig. A 1.8. It practically coincides with the
experimentally obtained shape of impact.
Appendix 11 Some Applications of Autoresonant Strongly
Nonlinear Vibratory Systems *
The use of resonance phenomena to develop useful mechanical work is discussed.

Principles of strongly nonlinear resonant behaviour are presented, with particular
reference to achievement of stable operating modes and their practical
implementation in machines.
Starting from vibro-impact systems, the concept of autoresonance is developed
and its implementation in a laser scanner for surfacing treatment and
autoresonant ultrasonic heads for surface impact treatments and cutting are
described. Development and operation of a modular robot manipulator utilizing
electromechanical self excited shift and rotate modules is reported.
Finally same future directions 0/ research and development are discussed. The
potential for increased electromechanical integration within complex machines
and recent research on svnchronized vibration excitation in weighted strings are
briejly considered.
Keywords: Applications; Autoresonance; Homeostat; Laser surfacing;
Mechatronics; Nonlinear vibration; Oscillation; Percussion machines; Q-factor;
Resonance; Robotic manipulator; Surface treatment equipment; Ultrasonics;
Vibro- impact.
1. Introduction
In striving towards speed and efficiency machines sometimes suffer
mysterious catastrophes; most of these are the result of hardly discernible
oscillatory processes. However, penetration into the amazing and complex world
of oscillations also opens up new horizons for engineering. Elaboration of the
technology ofthe future requires of scientists and engineers intensive work on the
understanding and use of economic forms of motion, of wh ich resonance proves to
be one of the most efficient.
Resonance makes possible excitation and maintenance of high-energy
oscillations through the use of low-power sources. It is familiar to anyone who has
set a heavy swing in motion with a large amplitude by giving it light pushes.
Anything that oscillates may be easily swung ifthe force is applied in unison with
the natural oscillations. A rapid increase of the swinging amplitude is the most
familiar property of resonance. It is extremely attractive to realize such highly
economical motions with large amplitudes, speeds and rates of acceleration and
thereby to achieve considerable force. For example, a motor car, which has stuck
fast, may be freed by rocking to and fro. This explains the wide use of resonant
• Frorn "Machine Vibration", (1992) I: 11 0-119

AII-2 Appendix II
oscillations in various fields of engineering.

In traditional resonant systems, whose general arrangement is commonly
represented by a load rocking on aspring under the effect of the harmonic force
(Fig. I a), all processes occur in accordance with harmonic laws (Fig. I b)
characterized by smooth alteration of the acceleration and deceleration with
defined maximum and minimum of all parameters. The maximum values of the
oscillatory process parameters depend on the frequency of the extemal periodic
action in accordance with the curve known as the amplitude-frequency
characteristic shown in Fig. 1c. The peak of this curve at point A represents
resonance. The extemal periodic action coincides with the natural frequency of
oscillation. The system is swung by any convenient mechanism and it readily
absorbs the energy supplied from the source of excitation. This mode of operation
is also optimal for the excitation source. Extremely low forces are required from
the source since the system imposes virtually no load upon it, and the source
functions without strain. The ratio of the natural frequency to the factor
characterizing attenuation of the oscillations is termed 'the Q-factor of the
oscillatory system.' The larger the Q-factor, the narrower and higher is the peak of
the amplitude-frequency characteristic. For comparison, the dashed line in Fig. 1c
shows the characteristic curve of a system with low Q-factor.
.,
'~
E
~ A.
A
~~\J~ V time. 1
a) "
o 2Wo freQuency. w b)
c)
Fig I Traditional hannolllc rcsonant y tcm
2. Vibro-impact Systems.
Is it possible to intensify the resonant oscillations and thus develop forces
within the oscillation cycle which considerably exceed the conventional maximum
values of harmonic curves? This may be attained through inclusion of shock into
the oscillatory motion since shock is the most efficient method developed by
nature for conversion of mechanical motion into force. Now we approach a new
important type of mechanical oscillatory systems - the vibro-impact system. The
most simple model of such a system is illustrated in Fig. 2a. Now the oscillations
of the load are accompanied by periodic shocks against a fixed limiter. The
smooth flow of the process is disturbed by abrupt jumps of the speed at the
Appendix II AII-3
11\ !! II
O "----~2..---''---'''----''-----,.
tim• . t
cl Fi~. 2 V,bro-impuc, o".;ill"lol'\ s}"I~m

bl
instants of collision, with production of considerable forces of interaction (Fig.

2b). Though the time of the shock is short (in solid bodies it ranges from 10- 3 to
10- 5 s depending on the dimensions and configuration), the resultant forces are
tremendous and capable of execution of considerable work. This work may be
efficiently used in two ways: on one hand, demolition or treatment of some
material (the workpiece replaces the limiter) and on the other hand, rapid revers al
(change of direction of motion) of a heavy load, exemplified by the percussion
tools which have been permanent companions of humanity throughout its history
and remain some of the most efficient and effective work tools. From the first
primitive hammers to modern forging and riveting hammers, from ancient
pickaxes to up-to-date drilling installations, from wall-demolition rams to
perfectly designed concrete breakers and perforators - such are the extremes of
this long-established process. Both useful properties of the vibro-impact process
are used to the maximum in all these machines. Large forces developed during the
shock improve malleability of the hardest materials and simultaneously ensure
instantaneous reversal of the striker, thereby contributing to a high rate of
recurrence of the process. To ensure the maximum economy of machine operation
throughout the entire cycle, the oscillation exciter (drive) should be synchronized
to the bouncing load which will again be accelerated to the required shock speed
in unison with its natural motion, i.e. the load should again swing in resonance but
now in resonance with the frequency ofthe vibro-impact system.
What are the characteristics of vibro-impact resonance? For a detailed analysis
of the resonance, it has been necessary to develop special methods of nonlinear
mechanics, perform time-consuming calculations and make numerous
experiments. Research conducted at the A .A. Blagonravov Machine Studies
Institute of the Academy of Sciences of the USSR have revealed a number of
amazing properties peculiar to vibro-impact resonance. We shall describe only
some of these properties, commencing with analysis of the most simple model
illustrated in Fig. 2a.
Bear in mind that when the limiter is installed with a certain clearance L\
relative to the spring, two kinds of oscillation are possible: without contact of the
AII-4 Appendix II
load with the limiter (motion within the limits of the clearance) and with impacts
against the limiter. If such oscillations are excited by a low-power source of
periodic force, e.g. by an electromagnet or other pulser, steady oscillations within
the clearance limits have conventional properties of traditional oscillatory systems
such as are shown in Fig. ta. Hence, the amplitude-frequency characteristic curve
(Fig. 2c) of a vibro-impact system in non-shock-motion (branches land 3)
coincides with the segments ofthe resonant curve shown in Fig. 1c.
Steady modes of motion with impacts against the limiter are no longer
harmonic (Fig. 2b) therefore half of the full swing is commonly used for
estimation of the oscillation amplitude . Just this magnitude is shown in Fig. 2c
(branch 2). As is evident from Fig. 2c, when the excitation frequency is smoothly
increased (arrows pointing to the right), the load swings to the limiter and starts
striking against it, with a slight initial decrease of the full swing and with
subsequent increase ofthe swing to the maximum values at point A corresponding
to frequency O)A close to double the frequency of free oscillation of the load
suspended from the spring. Ifan attempt is made to further increase the oscillation
frequency the vibro-impact nature of motion of the load is discontinued and
harmonic oscillations within the clearance limits are set up after a short transient
process. If the excitation frequency is subsequently decreased (arrows pointing to
the left), this will not result in the load swinging to the previous values. The vibro-
impact motion will commence by a jump at point Band resonance at point A will
not be displayed.
- .:l r--::::,.-.,.,."..,TnJ/
oH+++if+1t+it
d 1---"C>..l.L---'-I.J..UJJ.
x
a)
time. t
Fig. 3. Frequency pulling 01' oscillalions
Thus, the system behaves in different ways depending on the direction of
h·_
.
held frequency change. Fig. 3
JL Ixcitltion
hard
blako illustrates variation with time of
U the vibro-impact system
osci1lations first as the excitation
of+-:f+.r4+-+--*--I!---I!-- . _.. - - frequency is slowly increased
_. 11;"••-1 and then as it is decreased. Such
behaviour is termed 'pulling of
01 bl oscillations' and is peculiar to
nonlinear dynamic systems
Fig. 4. llard excitation 01' llscillations
whose restoring force acting on
Appendix II AII-5
the load is not proportional to load displacement. In the given case, the non-
proportionality is explained by installation ofthe motion limiter.
Let us now consider some other remarkable peculiarities of the vibro-impact
system. It is evident from Fig. 2c, that in the frequency band corresponding to
points Band A (double-valued zone of the amplitude-frequency characteristic
curve) the system may realize two absolutely different types of motion for the
same parameters and conditions of excitation: energetic oscillations with strong
impacts (vibro-impact resonance) and weak harmonic oscillations near the
equilibrium position, without contact with the limiter. Realization of one or the
other behaviour depends, as we have shown, on the pre-history of the excitation
process.
However, there exists another method of transition from one state to the other.
Let us fix excitation
frequency W at the
intermediate point Wo (see
Fig. 2c) of this frequency
band at realization of the
harmonic oscillations and
give a strong accelerating ~ ~--------------~~~~~----
impulse to the load in the
direction of motion to
Fig. 5. Amplitude pulling or oscillations
ensure collision of the load
with the limiter (Fig. 4a).
The vibro-impact mode of motion will be set up after a short transient process. As
distinguished from smooth ('soft') commencement of the motion mode, such a
method is termed ' hard' excitation of the oscillations. Having discontinued the
shock oscillations ofthe system for an instant (e.g. by application of a deceleration
impulse), we may return the system to the shockless mode of oscillation (Fig. 4b).
The system apparently 'memorizes' the last action.
It is interesting 'to contro\' such vibro-impact mode of oscillation during
displacement of the limiter (Fig. 5). Let us assume that the system experiences
harmonic oscillations. When the limiter is displaced into the swinging zone, the
process will be changed, and vibro-impact oscillations will originate. Ifthe limiter
is now slowly moved in the opposite
direction (giving increased c1earance),
the vibro-impact mode of oscillation will
be preserved, with an increase of
oscillation amplitude. The load appears
to follow the limiter, as in basketball
dribbling when the player starts hitting
the ball near the floor and then rises it
higher and higher. Only at considerable
amplitude of swing is the vibro-impact
mode of oscillations disrupted and the
system returns to harmonic oscillation.
Fig. 6. Modell"r lypical vibro-impacl machine. As distinguished from the 'frequency
pulling' described earlier, this
AII-6 Appendix II
phenomenon is termed 'amplitude pulling.'

There are major problems associated with the transition of the vibro-impact
system into the resonant mode. Point A (see Fig. 2c) corresponding to the most
efficient processes lies on the boundary of existence of this mode. Accidental
tluctuation will transfer the system into quite another type of oscillation from
which return to the initial process is hardly practicable.
Realization of vibro-impact resonance in the system shown in Fig. 6 is even
more difficult. This system approximately simulates a wide range of vibro-impact
machines, in particular, hand-operated hammers, concrete breakers, tampers,
riveting machines and many other appliances.
Figure 7 illustrates variation of the half-swing (amplitude) of the striker
oscillations versus frequency at various feeds. With increase of the feed, the
resonant frequency of the vibro-impact mode of oscillation becomes greater and
the form of the amplitude-frequency characteristic curve is changed. At high rates
of feed ensuring effective treatment (since the impact force impulse is
proportional to the feed force),
the resonant mode of
..
..,,;
oscillation again approaches
the boundary of existence of
a the vibro-impact process
~
.
E (point A). Now the zone of
oscillation is located to the
right of the doubled frequency
of the natural oscillations of
o ... the striker. Any slight decrease
Fig.7. Effect of feed force on amplitude-frequency characteristlcs. of the excitation frequency
will result in complete
discontinuation of the
oscillations (see the arrows in Fig. 7).
To start the system under these conditions, It IS necessary to considerably
increase the frequency, make transition to the vibro-impact mode of oscillation
with the use of hard excitation (additional push) and, smoothly decreasing the
frequency, establish resonance. This method is hardly practicable and frequently
cannot be realized at all. Even automatic frequency contral systems commonly
used in radio engineering for tracking bell-shaped resonance (as in Fig. Ic) prove
to be inefficient. This is because exploratory steps which change the excitation
frequency to both sides of the effective value are continuously made in these
systems and the direction of trimming is determined by the results. When the
mode exists on the boundary of its realization, such a search is not permissible
since it may lead to disruption and complete discontinuation of the process.
Therefore, the most economic dynamic mode of operation of the system turns out
to be the most inconvenient for realization.
Since reliability of operation is the main parameter of a machine, an alternative
approach to development of the engineering was selected. Doing their best to
avoid instability and having no clear theoretical concepts, the development
engineers started empirically de-tuning the machines, withdrawing them in fact
from highly efficient vibro-impact resonance to the high-stability processes
Appendix II AII-7
associated with relatively low rates of feed and low Q-factors of oscillatory
systems. As a result, efficiency is decreased, effectiveness is lost, and energy
requirement is too great.
3. Autoresonance
The investigators of vibro-impact systems could not resign themselves to this
situation . It was necessary to find some other method of excitation of resonant
vibro-impact modes ensuring that these modes of oscillations become most stable
and, if practicable, the only state of the system at all changes of its parameters and
loads. And such a way was found. The researchers directed their attention to self-
excited oscillation processes originating under the effect of forces depending on
the motion ofthe system.
Let us remember how the swing is set in motion . As the swing amplitude is
increased, the time between stops in the upper position becomes longer, though
the speed of passage through the
lower position continuously increases.
Therefore, to accelerate the swing, it
is necessary to change the frequency
o.t of the pushes, each time trying to push
the swing in the direction of the speed
of motion. Such feedback between the
motion of the system and the force
causmg this motion is termed
' positive feedback.' In our case, speed
positive feedback is meant. In practice
such feedback is realized, for instance,
by the use of a speed transducer (Fig.
- - - - - -+ ----'..- n-od-Uc.-'-ou-,.,.-, Sa) whose amplified output signal is
fed to the system which is being
swung through an electromagnet or
other source of power. As a result, the
folg R PosItive speed feedhaek hv trunsdueer equilibrium of such a system becomes
unstable: any infinitesimal deviation
gives rise to a push in the direction of
the acquired speed. The system starts self-swinging in unison with its natural
oscillations. Autoresonance takes place; this has been defined by Academician A
A Andronov as ' ... the resonance under the effect of the forces depending on the
motion of the system.' Speed positive feedback is the main feature of self-excited
oscillation systems and autoresonance is their most characteristic behaviour.
For limitation of self-excited oscillations, it is necessary to reduce the
swinging force when the system approaches the required level of oscillation. To
this end, special technical means are used, for instance, 'roll-off of the amplifier
characteristic curve in the course of increase of the current fed to the
electromagnet (Fig. Sb).
Man successfully uses self-excited oscillation systems. They efficiently
convert source energy into the oscillatory mode of the driven member. The most
common implementation is anormal timepiece - from pendulum clock to modern
AII-8 Appendix II
crystal-operated watch. It is no mere chance that the saying goes 'Reliable as a

watch.' In all spheres of modern radio engineering self-excited oscillations are the
foundation of diverse oscillators. Whistles, beils and many types of musical
instrument are actually examples of self-excited oscillation systems. Air-operated
and electric hammers are self-excited oscillation vibro-impact systems.
The progress of modern electronics and control means provides for the use of self-
excited oscillation processes in machine engineering as methods of conversion of
source energy into the production process. Just such a direction has been taken in
the Vibration Engineering Systems Laboratory of the Machine Studies Institute
whose scientists have studied nonlinear dynamic processes of machines for many
years.
4. Vibro-impact Scanners
The auto re sonant vibro-impact scanners that are necessary primarily for laser
engineering represent one of the first developments of our Laboratory. Physical
fields do not deflect a laser beam, therefore movable mirrors are used for laser
beam displacement in space. Fig. 9a illustrates laser treatment of a surface using a
vibro-impact scanning system. Laser beam I is directed into scanner mirror 2
which reflects it onto surface 3 wh ich is being displaced at speed V. The mirror is
spring-loaded relative to casing 4 (Fig. 9b). Tilting oscillations of the mirror
through angle <p are generated by the use of an electro-magnet 5 and are limited by
...... - .
.
--
'Ci'~'
--
-.11-';
~ frtqu.ncy. 61 __ I
Fig. 9. Vibro-impact laser surface treatment scanner
rigid stops 6.
The mechanical model of such a system is shown in Fig. 9c. As distinguished
from the system shown in Fig. 2a, now the motion of the load is limited from both
sides. The amplitude-frequency characteristic curve of such a system is illustrated
in Fig. 9d. The frequency 'pulling' effect described earlier becomes more
pronounced in such a model. Having commenced striking against the stops, the
mirror preserves the vibro-impact mode of motion with constant amplitude, equal
to the amount of c1earance to the limiters, up to point A, i.e. at frequencies which
materially exceed the natural frequency (wo) of the oscillatory system arranged
between the limiters (see the arrows in Fig. 9c). When this mode is realized, the
Appendix II AII-9
motion of the mirror in time approximates the saw-tooth motion desirable for
uniformity of laser treatment. Thus, high frequency, linearity and large scanning
amplitudes are attained in the 'pulling' mode. The re sonant nature of the process
leads to a decrease in energy consumption and reduction of the dimensions of the
feeding magnets.
The problem is reduced to ensuring stability ofthe mode, since in this case the
best characteristics are also obtained at point A (Fig. 9d) on the boundary of
existence of the vibro-impact mode. To this end, a positive feedback is provided
(Fig. 9b): the signal from the transducer measuring mirror oscillations is converted
(7), amplified (8) and fed into the electromagnet windings. The circuit is now
completed and the system starts functioning in the autoresonant vibro-impact
mode with characteristics corresponding to those at point A. This is the only stable
state ofthe system.
The developed theory and methods of design of auto-resonant vibro-impact
scanning systems have made it possible to preserve saw-tooth motion accurate to
1% or better for mirror oscillation amplitudes of 3 0 and scanning frequencies of up
to 400 Hz. The diameter of the water-cooled mirror equals 80 mm . Vibro-impact
scanners with such unique characteristics are paten ted in all industrially developed
countries. Figure ge, demonstrates scanner for gas laser, Fig. 9f for solid state
laser.
Stable scanning devices for light beams provide also for making high-accuracy
measurements exercising reception and transmission of information signals and
image analysis. The information vibro-impact scanners designed in the Laboratory
c)
Fig. 10. Surface treatment llsing an llltrasonic converter
for these purposes feature high stability, sm all overall dimensions and mass,
making it possible to build these scanners into complex electronic control
instruments.
5. Autoresonant Ultrasonic Heads

As already stated in the beginning, the impact method oftransmission of action
from driven members ofthe vibration machine to the medium being treated makes
it possible to increase considerably the efficiency of technological processes
An-lO Appendix 11
whose realization calls for application of great force. When this method is used, a
high Q-factor of the system and hence its economy are preserved. These
advantages may be realized in processes intensified by high-frequency vibration.
Fig. 10 a is a diagram of an ultrasonic converter used for superposition of high-
frequency force impulses on the process of treatment of article I with tool 2. The
main component of the converter is the acoustic head consisting of casing 3,
magnetic or piezostriction vibrator 4, and stepped waveguide (concentrator) 5
carrying tool 2 on its end. The waveguide is attached to the end-face of the
vibrator. The acoustic head functions as folIows. When a high-frequency electric
impulse is fed to the input of vibrator 4, it starts making longitudinal oscillations
because of the magnetic or piezostriction effect. These oscillations excite standing
waves in waveguide 5. The amplitude of the standing waves at the lower end
carrying the tool exceeds the amplitude of the waves in the joint between the
waveguide and vibrator. As a result, the oscillations are intensified. Having
pressed the head to the blank with feed force P, treatment of the article occurs by
systematic separation of the tool from the surface and subsequent collision caused
by the waves.
Let us remember the analysis of the mechanical model (see Fig. 6) and try to
understand what happens in the course of ultrasonic treatment. In modem
ultrasonic converters, the vibrators are supplied with power from powerful sources
of high frequency (oscillators). The frequency is selected directly behind the
boundary of the audibility threshold (somewhat above 20 kHz) to avoid
unpleasant sound. The waveguide-concentrator is tuned to the same frequency. As
a result, the tool starts oscillating like a load suspended from aspring. The peak-
to-peak amplitude of the oscillations is low (several micrometres) since at such
high frequencies the attenuation of oscillations in the material is extremely large.
Nevertheless the researchers try to ensure a maximum Q-factor for the system by
selection of material properties - otherwise the system will not oscillate: the solid
metal ofthe waveguide must be made to 'breathe.'
Pressing the head to the article gives rise to the vibro-impact mode. The
natural properties of the system are modified as shown in Fig. 7 and the selected
frequency is found aside to the most favourable modes corresponding to vibro-
impact resonance. Again progress of the technology occurred by way of ensuring
stability 'at any cost.' The power of ultrasonic installations kept growing, special
cooling systems were designed, and the overall dimensions of the oscillators were
increased. As a result, in modem ultrasonic technological installations, power over
I kW is required for development of feed forces of the order of 100 N. In this
case, efficiency equals ~3%. Therefore, ultrasonic technology is used only when it
cannot be replaced by any other means, namely for treatment of extra-hard and
brittle materials (diamonds, corundum, glass, ceramics, semi-conductor materials).
Even here autoresonance provides for improvement of efficiency of ultrasonic
equipment. Multi-year researches conducted by staff members of our Laboratory
have resulted in development of the principles of design of autoresonant ultrasonic
systems and to production of pilot installations. The principle of operation of such
installations is rather simple. Instead of a master oscillator, a feedback transducer
and special amplifier feeding the vibrator (Fig. lOb) are introduced into the
system. Since we are dealing with ultrasound, even a microphone may be used as
Appendix 11 All-li
a trans duc er. As a result, positive feedback is provided, the system loses stability,
and starts generating ultrasonic oscillations at the frequencies of the vibro-impact
system. The conditions of treatment, feeding forces, and properties of the article
material are changed. The frequency of the autoresonant vibro-impact mode is
changed as weil. The amplitude-frequency characteristic curve of an autoresonant
system is the envelope of the peaks of the resonant curves (Fig. I Oc)
corresponding to various feeds. The process then becomes fully automated; it is
possible to change the feed in accordance with any program - the system will
automatically select the optimal mode for each feed. The energy consumption of
the installation is changed as weil. The efficiency may be increased to the
theoretical optimum of 50%. The feeding forces have sharply increased, though
the power required for this increase is negligible.
The principal features of an autoresonant ultrasonic machine consist of the
dynamic connection and coordination of the oscillatory processes in the machine
with the technological processes of treatment. The main efforts of the scientists
have been directed towards
ensuring such coordination. As a
result, the machine has acquired
a new property: ability to quickly
and independently adapt itself to
the treatment process. Now it has
become practicable to exclude
complex analyzing processes and
their support facilities from the
control system. As a result, a
new type of machine has been
produced - homeostat which, like
a living being, responds in an
optimal way to any change of the
Fig. 11 . Autoresonant ultrasonic cutting lathe.
functioning conditions, while the
type of behaviour
autoresonance, is not changed. This considerably extends the field of application
of high-power ultrasound.
If oscillations from the ultrasonic head shown in Fig. 11 are brought to the
treatment zone when cutting metals, the high cutting load will be replaced by a
series of micro-collisions. This will obviate the need for transmission of
considerable forces via the machine drive that imposes additional loads on the
transmission, supports and casing. Machines intended for plastic deformation of
materials may be similarly relieved of loading. Research to boost the power of
technological processes by use of vibration and ultrasonic actions is conducted in
several countries. Autoresonance will bring about a technological era of intensive
ultrasound.
6. Resonant Robots
AII-12 Appendix 11
Is it possible to employ the developed principles of coordination and relief of

resonant systems in new types of machines? To answer this question, our
Laboratory began work in a new direction associated with the production of
resonant robots - manipulation systems designed for displacement of loads
between assigned positions with stops made for loading and offloading or for
Fig. 12. Manipulation Auto-Resonance System.
performance of technological operations. To this end, the complex motion of a

load in space was divided into simple operations: translational and rotational
motions. Special translational motion modules (Fig. l2a) and rotational motion
modules (Fig. 12b) have been designed. The robot is assembled from such
modules in accordance with the 'outlet-base' principle in which the moving
member of one module serves for attachment of the next module. The moving link
of the last module mounts a gripper. Depending on the modules selected, the
manipulator operates in the desired coordinate system. Fig. 12c, is a diagram of
the MARS (Manipulation Auto-Resonance System) robot incorporating two
translational modules and one rotational module.
Each module (Fig. 12) is an electromechanical self-excited oscillation system
consisting of movable carriage I attached to the base by the use of spring 2.
Latches 3 mounted on the ends of the carriage path preclude motion of the
carriage in the opposite direction on ce it has stopped. The carriage is connected
with reversible electric motor 4. Gripper 5 is installed on the end of one of the
modules.
Appendix II AII-13
The module operates in the following way. Bring the carriage into the
operating state by fixing one of the latches. To this end, select the automatie
swinging mode by successive reversing of the electric motor. When a work stroke
is to be made, the latch is automatically withdrawn and the carriage moves to the
other extreme position under the force of the spring and electric motor, where it is
again fixed by a latch. The next work stroke takes place after release of the latch
and revers al of the electric motor. The program cycle of manipulator operation is
realized by a controller that establishes the sequence of module operation. The
autoresonant control adapts the system to change of parameters and external
conditions.
The MARS robot (Fig. 12d) is four times as light and three times as movable
as its industrial analogues. When the robot manipulates loads of up to 0.5 kg, the
power consumption in all coordinates does not exceed 20W which is an order of
magnitude less than the power consumption of fellow-robots. This consumption of
power has been attained by virtue of autoresonance. All displacements represent
parts of the self-excited oscillation processes interrupted by latches when the
system is stopped in positions corresponding to the maximum deformation of the
springs. The task of the electric motors consists only in replenishment of the
energy reserves in the oscillatory systems. If the system is swung in unison, in
resonance, the energy consumption is low. Other useful properties have been
attained: soft start-up and smooth stopping which improves accuracy of
positioning and reliability of operation. Mechanisms operating on this principle
have record-breaking specific parameters such as consumption of materials,
consumption of energy, as weil as maximum speeds and rates of acceleration.
7. Mechatronics and Machine Engineering

Vibration engineering represents a boundless field of application of
autoresonance. Control of the displacement of materials and parts, mixing,
separation, grinding and many other processes may be intensified by
autoresonance. The burgeoning
ofnew technical possibilities has
~
_ ~ _ _ provided the incentive for studies
in the field of nonlinear process
dynamics, development of
effective calculation methods,
Flg. 11 Synchrolllscu vihro-impactcxcital1on 111 a 'lring design of control facilities and
manufacture of new materials.
Such interaction is characteristic of technical progress. A peculiar feature of the
present per iod is the ever more close juxtaposition of mechanical and electric
processes in the executive and control functions. This calls for complete
'cooperation ' of the above-mentioned sub-systems, their transformation into a
single dynamic system. This in turn involves in-depth coordination of dynamic
and algorithmic functions of the multi-link system necessary to fully utilise the
merits of each subsystem. A new term 'mechatronics' has been coined for
designation of this permanent unity. Developments in this direction have just
commenced. There is opportunity for use of new materials, employment of
superconductivity properties and creation of new methods of solution of
AII-14 Appendix II
differential equations.
A proliferation of functional qualities is attained when the control means
utilise to the maximum extent the executive opportunities of the mechanical
devices. This places quite new requirements on such established branches of
science as mechanics. Each phase opens up new horizons for development of
technology, provides the outline of new generation machines wh ich are integrated
dynamic complexes of mechanical, electric and information-calculation processes.
A peculiar sort of resonance is displayed here: progress in one sphere gives a push
to development in other spheres. While the design bureaux and plant shops are
busy with development of next generation machines, the scientific laboratories
proceed with studies of new
§ dynamic processes and
g-~I methods for their mathematical

analysis, conduct experiments,
and discover hitherto unknown
Fig. 14. Isolated vlbro-nnpact excitation in astring
phenomena. Such intensive
work is conducted in our
Laboratory.
We have recently observed experimentally new vibro-impact effects
previously predicted by the theorists. The possibility of synchronization of the
process of vibration excitation of astring mounting hard particles and located in
the vicinity of a fixed limiter has been discovered. In this case, all the particles, as
if coming to an agreement, simultaneously strike against the limiter (Fig. 13). The
'committee' of particles behaves as a single load suspended from the spring
(string). The system is 'pulI' in frequency and amplitude and the oscillations are
excited in a hard manner (similar to the simplest system shown in Fig. 2a). Such
synchronization of impacts in structural elements or in materials may lead to
intensive weakening of these systems and their rapid destruction. Quite opposite
behaviour of the particles has been observed under other conditions. Each particle
may fall into resonant vibro-impact interaction with the limiter (Fig. 14) resulting
in the same nonlinear effects whereas the remaining particles are practically
immobile. The future will show what new machines and technologies will be
designed on the basis ofthese effects.
Springer
and the
environment
At Springer we firmly believe that an
international science publisher has a
special obligation to the environment,
and our corporate policies consistently
reflect this conviction.
We also expect our business partners -
paper mills, printers, packaging
manufacturers, etc. - to commit
themselves to using materials and
production processes that do not harm
the environment. The paper in this
book is made from low- or no-chlorine
pulp and is acid free, in conformance
with international standards for paper
permanency.
t Springer

(Foundations of Engineering Mechanics) Prof. DR V. I. Babitsky (Auth.) - Theory of Vibro-Impact Systems and Applications-Springer-Verlag Berlin Heidelberg (1998) PDF

Uploaded by

Copyright:

Available Formats

(Foundations of Engineering Mechanics) Prof. DR V. I. Babitsky (Auth.) - Theory of Vibro-Impact Systems and Applications-Springer-Verlag Berlin Heidelberg (1998) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Foundations of Engineering Mechanics) Prof. DR V. I. Babitsky (Auth.) - Theory of Vibro-Impact Systems and Applications-Springer-Verlag Berlin Heidelberg (1998) PDF

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What are some examples of vibro-impact effects discussed in the book?

What are some examples of vibro-impact effects discussed in the book?

Foundations of Engineering Mechanics

V. I. Babitsky, Theory of Vibro-Impact Systems and

With 146 Figures

ISBN 978-3-662-22534-9 ISBN 978-3-540-69635-3 (eBook)

Cataloging-in-Publication Data applied for

© Springer-Verlag Berlin Heidelberg 1998

Cover design: de 'blik, Berlin

SPIN: 10630548 6213020 - 5 4 3 2 1 0 - Printed on acid-free paper

Bishop Hugh Latimer

• to understand the unusual resonant behaviour ofvibro-impact systems,

In order to solve these problems, the author extended the mechanical

Vladimir Babitsky Loughborough, Leicestershire

• See also Chapters 7 to 9 in (117) written by R E Brunstein and A E Kobrinsky,

equivalent linearisation by distributionfunction [128, 130] which made it possible

frequency characteristics are revealed and analysed, and a comparison of

Borovkov, M EHertz. M Y Israilovich, V L Krupenin and A N Tresviatsky, who

Preface to English edition ...................................................................................... vii

Chapter 1 Vibro-impact systems

§ 1. Collision of so lids ................................................................................... 1

Chapter 2 Equivalent linearisation of vibro-impact systems

§6. Equations ofvibro-impact systems ....................................................... 75

Chapter 3 Vibro-impact processes in systems with single degree of

§9. Forced vibration ................................................................................. 125

Chapter 4 Vibro-impact processes in systems with multiple degrees

Chapter 5 Structural synthesis of self·sustaining vibro·impact

§ 18. Optimal periodic motions ofvibro-impact systems ............................ 251

Bibliography ........................................................................................................ 279

Index ............................................................................................................ 292

Appendix I .......................................................................................................... A1-1

Awaiting (he sensation 0/ a short. sharp knock.

W S Gilbert: Bab Ballads

related through the expression G = E /[2(1 + v)].

F(x,t) = ESE = ES au~:,t) (1.6)

The change in axial force over an element dx is as folIows:

F(x + dx, t) - F(x, t) = ES

Direct substitution shows that equation (1.9) is satisfied by a general solution

u:(x,t) = <p(yt-x) (0"'; x <I) (l.lI)

(-l() <x<O) (1.21 )

a 2 u l (x,t) _ f-lL aul (X,t)] = 0 .

<p'(a) = V; exp( -Ta). (1.23)

lntegrating (1.23) with the initial condition rp( 0) = 0, we find

(05, x < I) (1.24)

<p(yt-x)=- aj.y cosoolt-!..)=- aFy cosloot- xoo )

u'(x,t)= 2asin~ (I -x~inoolt-~). (1.29)

Introducing phase \jf into the exciting force F(t),

where C is a constant of proportionality, e is the length of the perimeter of the

For a sphere e = 2ny , hence integrating (1.31) using (1.32) we find

Substituting in (1.33) for y in terms of U using (1.32), we finally obtain

The coefficient c is determined experimentally. We note that when

CI = 4fin c~(~+~)-1/2 (1.35)

ulO(t) + u 20 (i) + 1/,, (1) = V,'" (1.40)

u\O(t) = fF(t)hl(t-t)dr, u 20 (t) = f F (t)h 2 (t-t)dt, (lAI)

fh{t - t )F(t )dt + X[F(t)] = VI'! . (1043)

fh(I-·r)\fI[U o(l)]dl + lIJt) = &(1). ( 1.44)

elastic-plastic materials, or materials with post-effects and relaxation etc. [187].