(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) Prof. DR V. I. Babitsky (Auth.) - Theory of Vibro-Impact Systems and Applications-Springer-Verlag Berlin Heidelberg (1998) PDF
(Foundations of Engineering Mechanics) Prof. DR V. I. Babitsky (Auth.) - Theory of Vibro-Impact Systems and Applications-Springer-Verlag Berlin Heidelberg (1998) PDF
Theory of Vibro-Impact
Systems and Applications
Translated by N. Birkett
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
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.
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.
The drop of rain maketh a hole in the stone, not by vio/ence, but by oft falling.
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:
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.
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
"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
V Babitsky Moscow
Contents
§ 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
(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)
= _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)
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
§ 1 Collision of solids 5
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
<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) = ~:[I-exP(-Ta)l
When a = yt - x we finally obtain
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
6 Chapter 1 Vibro-impact systems
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
§ 1 Collision of solids 7
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
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
8 Chapter 1 Vibro-impact systems
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
§ 1 Collision of solids 9
U o = ul + u2 =(~+~)L.
r r 2
l 2
(1.32)
F = 4fin
3
C~(~ + ~)-1/2 U~/2
1- v rl r2
. (1.34)
(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
10 Chapter 1 Vibro-impact systems
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
§ 1 Collision of solids 11
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
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
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
12 Chapter 1 Vibro-impact systems
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
14 Chapter 1 Vibro-impact systems
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)
(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
§ 1 Collision of solids 15
'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 ,
~
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
16 Chapter 1 Vibro-impact systems
( 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 '
§ 1 Collision of solids 17
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
18 Chapter 1 Vibro-impact systems
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)
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)
'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)
Hence, within areas the application of stereomechanics gives simple kinematic solutions during
the study of coll ision.
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)
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
(/ /
" , =
\( I)
-211- ~-;-s(;D!~~, . (2 . 11 )
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.
24 Chapter 1 Vibro-impact systems
(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)
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,)
26 Chapter 1 Vibro-impact systems
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
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
28 Chapter 1 Vibro-impact systems
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
§2 Oscillator with fixed stop 29
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
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.
30 Chapter 1 Vibro-impact systems
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
(2.41 )
CI nl
+ (8q>"+1 -8q>,,)~tan~=0,
. 2nl 2nl
8C I"RsmT- 8C 2(v+I) -8C 2"RcosT+
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 ~ ~'
32 Chapter 1 Vibro-impact systems
. 2rcl 2rcl
hjRslll- - h 2 (ß + Rcos-) +
I;; I;;
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
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< I .
(2.45)
j
G
The first of these gives the condition R < I, which is always satisfied. The
second is equivalent to the following two relationships
G j + Go + G2 > 0, (2.46)
a j - ao - a 2 < 0 . (2.47)
§2 Oscillator with fixed stop 33
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
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
§2 Oscillator with fixed stop 35
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<p = ±l , but also follows from equation
(2 .57).
For systems with interference (ß < 0) the resonance curve takes the form of a
loop enciosing the backbone line and extending to high frequency areas S > 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)).
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
36 Chapter 1 Vibro-impact systems
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
hence the branches of the resonance curve are grouped close to the backbone
curve, and intersect it when cos<p = 0 .
"'{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
38 Chapter 1 Vibro-impact systems
x= L\
sin[n/(2(, 0)]
. (1""\
Sill ~d--
2(, 0
n) (3.2)
. OL\ ( n ) (3 .3)
x = - sin[ n 1(2(, 0)] cos Ot - 2(,0
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
.
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
40 Chapter 1 Vibro-impact systems
°,
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
42 Chapter 1 Vibro-impact systems
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
Fig.3.6
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
44 Chapter 1 Vibro-impact systems
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.
46 Chapter 1 Vibro-impact systems
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:
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
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
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
48 Chapter 1 Vibro-impact systems
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
50 Chapter 1 Vibro-impact systems
Flg, :I 11
FIg, 3 12
"l{ll
Q
i"--I
" 'I:
(I
;~
0'
'7/ IJ "b
g 1_. __ -
.J
hg .114
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.
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
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
54 Chapter 1 Vibro-impact systems
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)
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.
56 Chapter 1 Vibro-impact systems
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:
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
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
58 Chapter 1 Vibro-impact systems
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
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<p = 2 (4.34)
cro a x, P (1 + R)
Demanding that 1sin <p I::; 1 we obtain the following condition for existence:
2nG 2Q2 (1- R)
a > -- (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:
§4 Principles of percussion machine dynamics 61
~[1-
2nco
1- 2
4G nC]<a
cop 2 -
<~[1+
x, - 2nco
1- 2
4G nc]
coP 2 '
(4.36)
a) 5
11 I A-"(
*'
-
2
Fig_49
H
a)
,~~-----H------~*-
/J)
Fig. 4. 10
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
64 Chapter 1 Vibro-impact systems
cos Wo
öTv = ---(Öz v +1 - ÖZ v ) ' (4.51)
Vo
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:
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
68 Chapter 1 Vibro-impact systems
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
(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.
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)
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
§5 Stitching method 71
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) = <I>~ [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
72 Chapter 1 Vibro-impact systems
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
(5.\3)
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)
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
§5 Stitching method 73
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
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
~
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 <I>,,(x,x) and dissipative
constituents <I> D (x) or <I> 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)
<I>(x) = CI (x - /),Y/2 11 (x - /),) (6.5)
in the case of single sided interaction and
<I>(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)
<I>(x) = c2 (x - L\)ll(x - L\), (6.7)
<I>(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=-<I> (M=M I M 2 /(M I +MJ), (6 .9)
where <I> is the principal vector of the interaction force . An effect equivalent to
collision can be obtained by introducing an "instantaneous" force <I> 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
<I> Li = <1>1 {ll(x - L\)-ll[x -(L\ +E)]} (6.10)
and the dissipative
<I> D = <1>2 {ll(X - L\) -ll[x - (L\ + E)]}signi . (6.11)
The constant coefficients <I> I and <I> 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)
Thus, when using stereomechanical impact theory, the elastic and dissipative
components of the dynamic characteristics of impact pairs appear to be dependent
78 Chapter 2 Equivalent linearisation of vibro-impact systems
(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 + <I> 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 <I> 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)
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:
80 Chapter 2 Equivalent linearisation of vibro-impact systems
(M IM 2s 4 +M 2 n ls 3 +M 2 c IS2)X+
(6.27)
+ [(MI + M 2 )S2 + nls + CI ]<I>(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)
npqxpxq,
p,q=1 p,q=1 p,q=l
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 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
Thus, to describe the interaction of impact pairs, it is necessary to know the matrix
of dynamic compliance operators for impact pair A(s)
The express ions for xlO(t) and x 20 (t) define the vibrations ofthe elements ofthe
§6 Equations of vibro-impact systems 83
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
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
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
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
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
86 Chapter 2 Equivalent linearisation of vibro-impact systems
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
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
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
§6 Equations of vibro-impact systems 87
form:
Expanding the determinant and neglecting the products of sm all terms we obtain
n
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
88 Chapter 2 Equivalent linearisation of vibro-impact systems
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.
§6 Equations of vibro-impact systems 89
(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
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)
92 Chapter 2 Equivalent linearisation of vibro-impact systems
For perodic multi-impact regimes with v consecutive impacts per period against each stop the
solution can be found in the form
l'
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)
L Fm exp(mjmt) ,
x
F(t) =
m",,-r
§6 Equations of vibro-impact systems 93
I
00
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 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.
94 Chapter 2 Equivalent linearisation of vibro-impact systems
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].
00
Substituting (7.2), (7.3) in the equation for the nonlinearity cI>(x,x), we obtain
§7 Hmmonic linearisation of vibro-impact systems 95
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
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
96 Chapter 2 Equivalent linearisation of vibro-impact systems
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<I>(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 <I>(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 + <I>(x) = Pcos(rot+<p), (7.26)
where, because ofthe symmetry ofthe system <1>( -x) = -<I>(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 <I>(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
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[<I>(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
fw(u)du = I, (7.42)
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
whenee
X2 1'2
w(u,v)=
1t
Ja; -(u-mJ2
1 o[v±mJa;-(u-mJ 2 ]. (7.48)
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
X2 VI
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
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
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
*The occurrence ofthis limiting transformation was brought to the author's attention by V K Astashev.
104 Chapter 2 Equivalent linearisation of vibro-impact systems
(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
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
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
106 Chapter 2 Equivalent linearisation of vibro-impact systems
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)
mx =aF cD.-
- -
c
j
+-
2aI'
I [CD.
mF- -
1t
- mF . cD. - mF
---arcstn---+
aF aF 1-[ "":,m,J' II (7.86)
§7 Harmonie linearisation of vibro-impaet systems 107
1 -+-
q=- 11
1 [ arcsm---+---
. cD. - mF cD. - mF (7.87)
c 2 1t aF aF
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
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)
108 Chapter 2 Equivalent linearisation of vibro-impact systems
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
• 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
"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).
110 Chapter 2 Equivalent linearisation of vibro-impact systems
known as the expected, mean or average value of random process x(t), and the
central second order moment
00
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
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)
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
* Strictly speaking, in an ergodie process the existence of separate unrepresentative waveforms is also
accepted (producing a set ofzero dimensions).
112 Chapter 2 Equivalent linearisation of vibro-impact systems
(8.17)
where l'](u) is the unit step function; x(t) is the waveform ofthe ergodic process,
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
§8 Statisticallinearisation of vibro-impact systems 113
• 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.
114 Chapter 2 Equivalent linearisation of vibro-impact systems
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
k= -2
1
cr x
00
x (8.32)
I x X'
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
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)
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
116 Chapter 2 Equivalent linearisation of vibro-impact systems
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;
§8 Statisticallinearisation of vibro-impact systems 117
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
::x )l
0x 2 0r 20 x
where .';(u) = .)z; jexp[ - ~ Ja"'. is the probability integral, the value of which is tabulated in
mF =c {m -(ll+m
2 X x );';( Il::xJ+(Il-m x );';( Il::xJ+
(8.50)
When it is necessary to take energy dissipation du ring collisions into ac count in cases when using
the statistical characteristic <I>(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
118 Chapter 2 Equivalent linearisation of vibro-impact systems
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)
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)
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=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
(8.69)
120 Chapter 2 Equivalent linearisation of vibro-impact systems
[ (~-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·
'" 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
-------------------------
mx = [X(F)w(F)dF , (8.79)
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):
(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-"
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
§8 Statisticallinearisation of vibro-impact systems 123
-------------------------------
=~ "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
Pau] Dirae.
where XO = x-mx'
Substituting (9 .3) in (9 .1), we obtain
+,'
Substituting (9.5) into (9.4) and separating into similar terms, we obtain
[
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:
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
128 Chapter 3 Systems with single degree of freedom
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)
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.
§9 Forced vibration 129
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
130 Chapter 3 Systems with single degree of freedom
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 <I>(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+<I>(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:
[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):
§9 Forced vibration 131
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
132 Chapter 3 Systems with single degree of freedom
~>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 <I>(x) is now
the characteristic of a symmetrical impact pair (Fig. 6.2, b). As a result we can
§9 Forced vibration 133
(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.
134 Chapter 3 Systems with single degree of freedom
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
§9 Forced vibration 135
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].
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
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
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
( nb) . 2nO I
exp - -;- sm -m-
XI (0) = - - = - - - - - - ' - - - - ' - - - - - - - - - - - = -
°1[1 + exp( - 2:b) _2exr( _ ~ )cos 2n~1 ]
[ Q, • ~Q' ~ b: 1 (9.70)
( nb) . nO I
exp - - sm--
X2(0)= 2m m . (9.72)
°1[1+ exp(- -~
2nb) + 2exp(nb) nO
- 2m cos--;-
I ]
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.
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<i>. (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
(10.7)
Substituting the expression for simp from (9.11) into (10.7), we obtain
;J+
Using (10.8), (9.11) and (9.12), condition (10.6) takes the final form:
;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)
az
~ I~l
f 1
0
Q
f Z llJ/fl I 2
a) b)
Fig. 10.1
<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
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<p -
~.~) <p x. (10.26)
axco ax co
Carrying out harmonie linearisation ofthe nonlinear funetion <l>(x,sx) = kx + ßsx
in (10.15) and substituting in (10.26) we obtain the following autonomous system:
(10.28)
~ = ~ ( a p sin q; _ b -
2 axco
ß) . (10.29)
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
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.
148 Chapter 3 Systems with single degree offreedom
I~
- g(t)
a)
I
JJ
Fig. 11.1
--g(t)
b)
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)
Eliminating <p from (11.19), (11.20), we obtain the final expression for ax ' It is
convenient to write it in the following form:
Applying condition (10.2) to (11.23) and using the connection between <p andax '
we obtain
dß h0 2 d<p )
( ------cos<p >0. (11.24)
Gax 00 Ga x _
ux=a x
(: t~ax = - OO(b
2
+ ß) ( :x t""X . (11.25)
Substituting the va lues (11.20) and (11.25) into (11.24), we obtain finally
[::J
(1 1.26) we find when
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
Substituting (11.31) into (11.21), we find
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].
(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) = -
x
Eliminating the value (d~ / da hy~ay from these expressions, we obtain using
(11.19), (11.20)
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:
(11.40)
(11.44)
- 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.
r
coefficients in the form (9.38) in (11.21) in place of Q2+k, we obtain
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
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
:/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)
a
x
=_2~[(~)2
20
-3± h2 _~(~)2l-1
0 220
(11.50)
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
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
160 Chapter 3 Systems with single degree of freedom
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)
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
§ 11 Parametric vibration 161
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)
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
[(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
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) //
i
- - /sec -
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.
( ~-~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
§ 12 Vibration of self-sustaining systems 169
(12.18)
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
170 Chapter 3 Systems with single degree of freedom
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 + <I> E (x, x) + <I> D (x, x) = r0 2 sin \jI , (12.24)
(12.25)
§ 12 Vibration of self-sustaining systems 171
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
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
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)
§ 12 Vibration of se1f-sustaining systems 173
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
174 Chapter 3 Systems with single degree of freedom
[-
dß Er0 2 (~ . ~
+ ~2~2 ~ ~ d<p ~~ aw ~)l
wsm<p + - a x sm<p --wax cos<p
. > 0. (12.46)
da, w ax da, da, _
ux=a x
j..lWOCix[~Cix
[daJ ~ l
aw da
+2(b+ß)])
x
(12.48)
,,co, - 2[" + Y, (b + ßlO; ]' _
ax=u x
b2ü:/- j..lw
°bwa x2 [b2_2(302+20~_W2)]+
2
(j..l+liba;) ax
(12.50)
§ 12 Vibration of self-sustaining systems 175
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
176 Chapter 3 Systems with single degree of freedom
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
.::' ~ \~
N:i!,
,J'\"
\
I ' .......
11
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,
178 Chapter 3 Systems with single degree of freedom
-dffi = E[ ~(Cl) 0 -
~
Cl) - -
1 rO 2 a~ sm
. ~]
q> . (12.56)
dt 2 x
dffi
- = E[(
dt
~ Cl) 0 -) - - -
-Cl) ~05a;
E
1 (12.57)
(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
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)
180 Chapter 3 Systems with single degree of freedom
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
§ 12 Vibration of self-sustaining systems 181
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
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
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
(12.87)
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
184 Chapter 3 Systems with single degree offreedom
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
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.
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
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
§13 Random vibration 187
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
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)
188 Chapter 3 Systems with single degree offreedom
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)
f
00
where w(F, F) is the joint probability density of F(t) and F(t) . Assuming the
simplification
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
190 Chapter 3 Systems with single degree of freedom
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]
Substituting (13.38) in the left hand side of(13.39) we find after integration
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):
_!; So -00 So
(13.41)
(13.43)
Finally, by analogy with (13.21) let us find the mathematical expectancy for the
coIlision frequency
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
192 Chapter 3 Systems with single degree of freedom
OJ,/fl fjf)
8 8
\ \
"
4
"
~~
D 2 A/ISJ4 D f 2 1J/lSz '
Fig. 13.3
Duke Ellington.
(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)
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
(14.19)
(14.20)
where
_I+[k(m,,,a,,)+sß(m,,,a,,)]L(s)
W1x ( s,ml"a ll ) - • (14.21)
A
Here
196 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
Ed = 0 2 (14.27)
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
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,,-(/(,
(14.31)
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
198 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
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,
(f{,-(lI'
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) ,
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.
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
200 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
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
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)
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
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)
x (14.64)
-L,o(x,s)J L8(t-vT),
LW - vT).
CL
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):
204 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
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)
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)
H f 2(l"-;~M
in the case considered takes the following form:
(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
L( m~OJ ) = 2L IO ( 0, ~OJ
[Q;- (~oo rr
)= r=====2=A="o=====
+ ~,Q;r_2OJ_2
(14.86)
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
§ 14 Vibro-impact interactions of elastic systems 207
----
X(t)
1 '
=- I A(mjw)exp(mjwI).
T 111:.:-1
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
208 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
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
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
1 Ji,lj,
= -2 f fVo(t)vo(t')RI'(t,t')dtdt' .
To 00
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
.
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
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
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
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:
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
214 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
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)
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)
.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
216 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
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
§ 15 Vibro-impact interaction of visco-elastic bars 217
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)
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
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 :
220 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
-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
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)
I LIO(O,ioo) I (15.35)
a" ==a p l+k(a,,)[LIO(0,ioo)+L20(0,ioo)]'
222 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
( 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:
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
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
224 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
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
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
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).
§ 15 Vibro-impact interaction of visco-elastic bars 227
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
228 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
. ) = - -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):
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)
(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)
(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 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:
(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:
232 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
4
L,(x,jm) = IC,(jm)exp(q,x). (16.20)
+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)
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)
(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
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-
§ 16 Vibro-impact regimes of vibrating beams 235
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
236 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
boundaries of the existence domains from (14.33) using (16.23), (16.24), when
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
§ 16 Vibro-impact regimes of vibrating beams 237
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
238 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
'. " ~}
~.
-----...---.-
-------- ~
Fig.16.7
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
-f f X(x,z,t -s)<D[u(z,s)]dsdz,
! 7
( 16.42)
pSl ",=1 üJ - D",
n r
with the natural frequencies defined in accordance with the following expression
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
/ = 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
242 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
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:
ot
E (16.48)
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.
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
(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)
§ 17 Theory of the impact vibration absorber 245
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
246 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
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.
where
B = 7t{[I- M 2oo\1 1(00)]2 +
and, consequently,
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
248 Chapter 4 Vibro-impact processes in systems with many degrees of freedom
,/" ! ...
/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
§ 17 Theory of the impact vibration absorber 249
(17.32)
(17.34)
(17.35)
Substituting (17.35) into the first ofthe conditions (17.34), we obtain eventually
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
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
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.
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
[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'
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.
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)
256 Chapter 5 Structural synthesis of self-sustaining vibro-impact systems
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
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
(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
§ 18 Optimal periodic motions of vibro-impact systems 257
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
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)
(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
+ 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 ) / '
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
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
260 Chapter 5 Structural synthesis of self-sustaining vibro-impact systems
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].
.
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 '
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)
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:
262 Chapter 5 Structural synthesis of self-sustaining vibro-impact systems
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
· 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
264 Chapter 5 Structural synthesis of self-sustaining vibro-impact systems
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)
§ 19 Synthesis of self-sustaining systems 265
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
axev+11-1 R2xv_
hg.19.4
l-ax,,_- -- ~R2(xv_)2 +4U(x' -x~)
< 1.(19.30)
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
(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)
(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 ,
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)
..
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
270 Chapter 5 Structural synthesis of self-sustaining vibro-impact systems
(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
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.
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
272 Chapter 5 Structural synthesis of self-sustaining vibro-impact systems
Fig. 20.2
§20 Synthesis of autoresonant systems 273
---
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)
274 Chapter 5 Structural synthesis of self-sustaining vibro-impact systems
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
§20 Synthesis of autoresonant systems 275
-----------=-
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
276 Chapter 5 Structural synthesis of self-sustaining vibro-impact systems
~[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
[ dV(ffi,iii x,G.)] ~ O.
dax - '
UxO;:;"x
[ d~(Gx)]
da x - •
<0,
U X= Ul'
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
--
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
Genesis, 6: 4
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,
1968, No. 3. (In Russian).
28. Babitsky V I, Izrailovich M Y, On the synthesis of self-sustaining systems. Machine Studies,
1971, No. 3. (In Russian).
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,
1970, No. 2. (In Russian).
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,
No. 5. (In Russian).
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,
No. 2. (In Russian).
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,
No. 5. (In Russian).
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
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
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
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
Sound and Vibration, 1990, v. 144, No. 3.
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
Sound and Vibration, 1998, v. 213, No. 3.
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
Sound and Vibration, 1993, v. 165, No. 3.
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
Sound and Vibration, 1991, v. 145, No. 2.
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
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
v
•
o M
•
o x
Fig. Al I
, =-
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
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
• 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) 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
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
2~R--
arctan -~~-----
, 1- 2~2
(A 1.15)
'I 0~2-1~
arctan - ~(3--=-4~ 2T-
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:
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
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
.,
'~
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
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
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
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
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
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)
for these purposes feature high stability, sm all overall dimensions and mass,
making it possible to build these scanners into complex electronic control
instruments.
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
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.
~
_ ~ _ _ 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
t Springer