You are on page 1of 15

International Journal of Non-Linear Mechanics 38 (2003) 389403

Analytical approximations for stickslip vibration amplitudes


Jon Juel Thomsen
a,
, Alexander Fidlin
b
a
Department of Mechanical Engineering, Solid Mechanics, Technical University of Denmark Building 404,
DK-2800 Lyngby, Denmark
b
Tennesseealee 35, D-76149 Karlsruhe, Germany
Abstract
The classical mass-on-moving-belt model for describing friction-induced vibrations is considered, with a fric-
tion law describing friction forces that rst decreases and then increases smoothly with relative interface speed.
Approximate analytical expressions are derived for the conditions, the amplitudes, and the base frequencies of
friction-induced stickslip and pure-slip oscillations. For stickslip oscillations, this is accomplished by using per-
turbation analysis for the nite time interval of the stick phase, which is linked to the subsequent slip phase through
conditions of continuity and periodicity. The results are illustrated and tested by time-series, phase plots and am-
plitude response diagrams, which compare very favorably with results obtained by numerical simulation of the
equation of motion, as long as the dierence in static and kinetic friction is not too large. ? 2002 Elsevier Science
Ltd. All rights reserved.
Keywords: Friction; Stickslip; Self-excited vibrations
1. Introduction
We consider the classical mass-on-moving-belt model for describing friction-induced vibrations. Tak-
ing the model for grantedwe do not discuss its validity for mimicking real systemsthe purpose is
to contribute further understanding on what it predicts, i.e. to provide expressions and graphs illustrating
the change of vibration character and amplitude with excitation speed. For a common friction law, with
a kinetic coecient of friction that rst decreases and then increases smoothly with sliding speed, we
provide simple approximate expressions for the system state during each cycle, and for the stationary
vibration amplitudes and base frequencies. This supplements the many works in this area who deals
mainly with setting up criterions for the onset of such vibrations, considers only non-sticking motions,
assumes piecewise linear friction laws, or rely solely on numerical simulation.
Assuming small but nite dierences in static and kinetic friction, a perturbation approach is used to
rst set up expressions describing the occurrence, the stability, and amplitude characteristics of pure-slip
motions, also showing that such motions are typically stable only for a narrow range of sliding speeds.

Corresponding author. Fax: +45-4593-1475.


E-mail addresses: jjt@mek.dtu.dk (J. Juel Thomsen), dlin@luk.de (A. Fidlin).
0020-7462/03/$ - see front matter ? 2002 Elsevier Science Ltd. All rights reserved.
PII: S0020- 7462( 01)00073- 7
390 J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403
Below a certain excitation speed, stickslip motions occur, for which we also provide simple analytical
expressions. This is accomplished by using perturbation analysis for the stick phase, which is linked
to the subsequent slip phase by requirements of continuity and periodicity. The results are illustrated
and tested by showing time-series, phase plots and amplitude response diagrams, which compare very
favorably to results obtained by numerical simulation of the equation of motion.
Stickslip vibrations show up in many kinds of engineering systems and everyday life, e.g. as sounds
form a bowed violin, squeaking chalks and shoes, creaking doors, squealing tramways, chattering machine
tools, and grating brakesmost of which are attributed to surfaces sliding with friction. Furthermore,
many mechanical interfaces are characterized by a form of dry friction where the forcevelocity curve
has negative slope at low velocities. Initially, friction decreases as the contacting object starts to move,
whereas at higher velocities the friction force increases again; in particular this characterizes surfaces
with boundary lubrication. The initial negative slope corresponds to negative damping, and may thus
cause oscillations that grow in amplitude, until a balance of dissipated and induced energy is attained,
as pointed out already by Lord Rayleigh ([1], Vol. I, p. 212). Typically there are two phases of such
oscillations: a stick phase with no slippage between parts and friction forces limited by static friction,
followed by a slip phase with a somewhat lower friction force; though, on some conditions there is no
stick phase.
Numerous works has been devoted to the study of friction-induced oscillations. For ease of setup
and interpretation an idealized physical system consisting of a mass sliding on a moving belt has been
considered very often, as it will be in this present study. Panovko and Gubanova [2], for such a system
with a friction characteristic having minimum coecient at velocity t
m
, show that self-excited oscillations
occur only when the belt velocity is lower than t
m
. Tondl [3], Nayfeh and Mook [4], and Mitropolskii
and Nguyen [5] describe self-excited oscillations of the mass-on-belt system, presenting approximate
expressions for the vibration amplitudes for the case, where there is no sticking between mass and belt.
Popp [6] presents models and numerical and experimental results for four systems that are similar to
the mass-on-moving-belt. Ibrahim [7,8] and McMillan [9] presents and discusses the basic mechanics
of friction and friction models, and provides reviews on relevant literature. A very readable historical
review on dry friction and stickslip phenomena is given by Feeny et al. [10], and a large survey on
friction literature until 1992 by Armstrong-H elouvry et al. [11].
There are, however, very few works providing expressions for stationary stickslip amplitudes.
Armstrong-H elouvry [12] performed a perturbation analysis for a system with Stribeck friction and
frictional lag, predicting the onset of stickslip for a robot arm. Gao et al. [13] derive analytical expres-
sions for the change in position during the stick phase for systems that can be described by a linearized
friction law, and a static friction coecient that increases with time. Elmer [14] discusses stickslip
and pure-slip oscillations of the mass-on-belt system with no damping and dierent kinds of friction
functions, provides analytical expressions for the transfer between stickslip and pure-slip oscillations,
and sketch typical local and global bifurcation scenarios. Thomsen [15] sets up approximate expressions
for stickslip oscillations of a friction slider, which are accurate for very small dierences in static and
kinetic friction.
Much research seems concentrated in determining the onset of stick-vibrations in order to avoid these
totally. However, stickslip vibrations might be acceptable in applications, provided their amplitudes are
suciently small. Therefore, simple expressions providing immediate insight into the inuence of param-
eters on vibration amplitudes are believed to be useful. Since the equations of motion are non-linear and
discontinuous in the highest derivatives, there are no straightforward standard procedures for calculating
stationary amplitudes. However, indeed, it is possible in specic cases, as demonstrated in this paper.
Also, it is sometimes possible to use perturbation methods for problems involving non-smooth functions
and non-small terms, provided they are non-small only in a small time interval, as has been recently
exemplied in Fidlin and Thomsen [16] and proven more generally in Fidlin [17]. Alternatively, the
J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403 391
system can be considered as having a variable number of degrees of freedom (decreasing by one during
times where a part sticks), as demonstrated in Fidlin [18] and Pechenev and Fidlin [19].
Section 2 of the paper denes the specic problem to be solved, and Section 3 provides solutions
in the form of analytical expressions based on perturbation analysis. Section 4 then illustrates typical
solutions, including amplitude diagrams showing how the stickslip oscillation amplitude varies with
excitation speed, and also compare to results obtained by using numerical simulation of the equation of
motion.
2. The problem
Fig. 1(a) shows the physical example system: a mass M on a belt that moves at constant speed J
b
, to
be termed the excitation speed below. The mass is a rigid body, at time

t positioned at X(

t) in a xed
frame of coordinates. It is subjected to a normal static load F, linear spring-loading KX, damping force
CdX}d

t, and a friction force Fj(J


r
) where j is the friction force as a function of the relative velocity,
J
r
=dX}d

t J
b
, and the static friction force is Fj
s
.
The equation of motion is, in non-dimensional form:
x + 2[ x +x +j( x t
b
) =0 for x =t
b
(slip), (1)
x =0, x + 2[t
b
6j
s
for x =t
b
(stick), (2)
where all parameters are positive, and variables and parameters are non-dimensionalized by
t =c
0

t, c
2
0
=
K
M
, x =
X
L
, L=
F
K
, t
b
=
J
b
c
0
L
, 2[ =
C

KM
, (3)
where x =dx}dt is the non-dimensional velocity of the mass at non-dimensional time t. Here lengths
have been normalized by the characteristic length L and time by the linear natural frequency c
0
of free
oscillations of the mass when there is no damping and friction.
For the friction function j(t
r
) we assume, as in e.g. Panovko and Gubanova [2] and Ibrahim [8],
j(t
r
) =j
s
sgn(t
r
)
1
t
r
+
3
t
3
r
, (4)
where t
r
= x t
b
is the (non-dimensional) relative velocity between mass and belt, and

1

3
2
(j
s
j
m
)}t
m
,
3

1
2
(j
s
j
m
)}t
3
m
, (5)
Fig. 1. (a) The example system: a mass at position X on a belt that moves at constant speed J
b
. (b) Friction function as given
by (4) with j
s
=0.4, j
m
=0.25, and t
m
=0.5.
392 J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403
where j
s
is the coecient of static friction, and t
m
is the velocity corresponding to the minimum
coecient j
m
of kinetic friction, j
m
6j
s
, and
1
,
2
0. Thus, the friction function satises |j(0)| 6j
s
,
j(t
m
) =j
m
, j

(t
m
) =0, j(t
r
) =j(t
r
), and j

(t
r
) =j

(t
r
), where j

=dj}dt
r
. Fig. 1(b) depicts this
function for typical parameter values (t
m
, j
m
, j
s
) to be used in this study. As appears |j| 6j
s
when the
mass is at rest on the moving belt (t
r
=0, stick phase), whereas when the mass starts sliding the friction
forces initially decrease with increasing velocity (t
r
=0, slip phase). This particular form of the friction
law is not overly restricted; it resembles characteristic features of friction models in common use, e.g.
see [11,20]. The so-called Stribeck curve, describing the friction-velocity relationship for systems with
boundary lubrication, e.g. [12], also share the essential features of (4), though they dier in detail.
The problem to be solved below is to determine stable periodic solutions of (1) with (4). Closed form
solutions are unavailable, due to the discontinuity and non-linearity of the friction function j. However,
for the important case of relatively small dierence between static and kinetic friction coecient, we
can employ perturbation analysis to set up approximate analytical expressions, and check the validity of
results by using numerical simulation.
3. Solutions
After having determined the dierent types of motion that can occur, we solve for each of these, and
then summarize the results in Section 3.4.
3.1. Types of motion
According to (1) and (4), the mass has a static equilibrium at x = x,
x =j(t
b
) =j
s

1
t
b
+
3
t
3
b
(6)
since then x = x =0. To study motions near this equilibrium we shift the origin by dening
u(t) =x(t) x (7)
by which (1) and (2) transform into
u +u +ch( u) =0 (slip), (8)
u =0, u +c(
1
t
b
+
3
t
3
b
) + 2c[t
b
60 for u =t
b
(stick), (9)
where
h( u) 2[ u +j( u t
b
) j(t
b
)
=2[ u +j
s
(1 + sgn( u t
b
)) + (
1
+ 3
3
t
2
b
) u 3
3
t
b
u
2
+
3
u
3
. (10)
Here c1 has been introduced merely as a book-keeping parameter, to indicate that the damping and
the dierence in static and kinetic friction coecient are assumed small; in the nal results c will be set
to unity.
The equilibrium u = u =0 of (8) corresponds to a state of steady sliding, with the mass being at rest
and the belt sliding at constant velocity t
b
below it. This static equilibrium can be stable or unstable.
If it is unstable, then stable periodic motion takes over; this is the only possibility, since generally the
steady state must be either static equilibrium, periodic motion, or chaotic motionand chaotic solutions
cannot occur for a single second-order autonomous ordinary dierential equation (e.g. [21]).
Two dierent kinds of periodic solutions to (8) are considered below: pure-slip oscillations, where
u(t) t
b
at all times, i.e. the mass never catches up with the beltand stickslip oscillations where
u(t) 6t
b
; i.e. the mass occasionally sticks to the belt.
J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403 393
It is important to note that during stationary oscillations the velocity of the mass will never exceed
that of the belt, i.e. u(t) 6t
b
for t t
0
where t
0
is a nite. This is so because the energy storing spring
cannot accelerate the mass to a velocity exceeding the maximum velocity during the previous oscillation
period, and the energy-providing belt cannot accelerate the mass to a velocity beyond its own. It is of
course possible to start the system from a state with u t
b
, however, viscous damping and dry friction
will then drain energy until a stationary state is achieved with u 6t
b
.
3.2. Pure-slip oscillations
With pure slip u t
b
for all t, so that the discontinuity of the friction function is never met or crossed.
For this case the function h in (10) can be written
ch( u) =h
1
u +h
2
u
2
+h
3
u
3
for u t
b
, (11)
where
h
1
=2[
1
+ 3
3
t
2
b
, h
2
=3
3
t
b
, h
3
=
3
. (12)
The calculation of non-linear oscillation amplitudes follows Thomsen [15]; here we just give the main
results to be used subsequently. Using standard averaging for solving (8), one nds that
u =Asin , u =Acos , (t) t +0(t), (13)
where A(t) and 0(t) are solutions of the averaged equations

A=
1
2
cA(h
1
+
3
4
h
3
A
2
), A

0 =0 (14)
for which there are two equilibriums: A trivial solution A(t) =0, corresponding to the static equilibrium
u(t) =0, and a nontrivial solution given by
A(t) =A
1

_

4
3
h
1
}h
3
, 0(t) =0
1
=constant for t (15)
corresponding to periodic solutions u(t) =A
1
sin(t +0
1
).
As for the stability of solutions, one nds that u =0 is unstable when h
1
0; by (12) and (5) this
condition becomes
t
b
t
b1
t
m

1
4[t
m
3(j
s
j
m
)
. (16)
As appears, when there is no viscous damping ([ =0) the static equilibrium is unstable for excitation
speeds lower than t
m
. Viscous damping stabilizes the equilibrium, and at suciently large damping,
[
3
4
(j
s
j
m
), the static equilibrium is always stable.
Periodic motions exist and are stable when h
1
0 and h
3
0. Since j
s
j
m
the latter requirement
is automatically fullled. The amplitude A
1
of the stable periodic motion is found by inserting into (15)
and using (12) and (5) to give
A
1
=2t
m

1 (t
b
}t
m
)
2

4[t
m
3(j
s
j
m
)
=2
_
t
2
b1
t
2
b
, t
b0
t
b
t
b1
, (19)
where t
b1
is the speed below which pure-slip oscillations rst occur, as given by (16). This expression
for A
1
assumes pure slip, so the increase in amplitude for decreasing t
b
will cease when the mass starts
sticking to the belt, i.e. when max( u) =t
b
. With max( u) =A
1
(by (13) and (15)), it is found that sticking
rst occurs when A
1
=t
b
. Inserting this into (19) and solving for t
b
we nd that stickslip oscillations
394 J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403
occurs when t
b
t
b0
where
t
b0
=
_
4
5
t
b1
. (20)
Hence, the range of belt velocities where pure-slip oscillations occur is rather small, its width (t
b1
t
b0
)
being only 1
_
4}5 10% of t
b1
. It forms a transition zone to a wider range of belt velocities where
stickslip motions occur.
When sticking just starts, the amplitude of oscillations is given by inserting t
b
=t
b0
in (19), and then
use (16) and (20) to nd
A
1
|
t
b
=t
b0
=A
1, max
=t
b0
=
_
4
5
t
m

1
4[t
m
3(j
s
j
m
)
. (21)
Hence, for vanishing damping [ the maximum amplitude grows linearly with the velocity t
m
of minimum
kinetic friction. As appears from (13), the non-dimensional displacement amplitude equals the velocity
amplitude; thus the velocity amplitude A
v1
=A
1
.
3.3. Stickslip oscillations
We here assume that t
b
t
b0
, cf. (20), so that during part of an oscillation period the mass sticks
to the belt. This case cannot be analyzed using the above averaging procedure for pure-slip oscillations,
since the switch from slip to stick is accompanied by a discontinuous change in acceleration. However, it
is possible to analyze the stick and the slip phases of the motion separately, and link the results together
to obtain an approximate expression for one full oscillation period.
3.3.1. Slip phase
During stick the mass moves with the belt, u =t
b
. This continues until the force from the restoring
spring and the damper has increased to the maximum static friction force, i.e. until the strict inequality
in (2) or (9) is no longer satised. We consider this the initial condition at time t =0, where the stick
phase ends and the slip phase begins (cf. Fig. 2), i.e.
u(0) =2c[t
b
+c(
1
t
b

3
t
3
b
), u(0) =t
b
, (slip starts). (22)
Motions during the subsequent slip phase are then governed by (8). This equation is non-linear, so
approximate methods are in need. However, during slip it is continuous in its highest derivatives, since
u t
b
during slip. Further, since the solution is only needed for the nite time interval of the slip phase
we can use a straightforward perturbation approach. Letting
u(t) =u
0
(t) +cu
1
(t), t [0; t
s
], c1, (23)
we substitute into (8) and (22), balance terms of like powers of c, insert (10), and obtain two new
initial value problems for the determination of u
0
and u
1
:
u
0
+u
0
=0, u
0
(0) =0, u
0
(0) =t
b
, (24)
u
1
+u
1
=h( u
0
), u
1
(0) =2[t
b
+
1
t
b

3
t
3
b
, u
1
(0) =0. (25)
The solution to (24) for the zero-order approximation u
0
is
u
0
=t
b
sin (t). (26)
Inserting this and (10) into (25), the equation for the rst-order correction u
1
becomes
u
1
+u
1
=
3
2
c
3
c
1
cos (t) +
3
2
c
3
cos (2t)
1
4
c
3
cos (3t), (27)
J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403 395
Fig. 2. Displacement u(t), velocity u(t), and acceleration u(t) during one cycle of stickslip oscillation. Denitions of displacement
and velocity amplitudes A
0
and A
t0
, switch time t
s
, and the times t
m
, t
m+
, t
m
of maximum absolute displacements and velocities.
where
c
1
2[t
b

1
t
b
+
15
4
c
3
=2[t
b

3
2
(j
s
j
m
)
t
b
t
m
_
1
5
4
_
t
b
t
m
_
2
_
,
c
3

3
t
3
b
=
1
2
(j
s
j
m
)
_
t
b
t
m
_
3
. (28)
The solution of this linear equation satisfying the initial conditions in (25) is
u
1
=
3
2
c
3

1
2
c
1
t sin (t) + (
55
32
c
3
c
1
) cos (t)
1
2
c
3
cos (2t) +
1
32
c
3
cos (3t), (29)
where the secular term t sin t is fully acceptable, since it remains bounded in the nite time of slipping.
Hence, by (23), (26), (29), and (5), motions during the slip phase are approximately given by
u(t) =t
b
sin (t) +c[
3
2
c
3

1
2
c
1
t sin (t) + (
55
32
c
3
c
1
) cos (t)
1
2
c
3
cos (2t) +
1
32
c
3
cos (3t)]
+O(c
2
), t [0, t
s
] (slip phase), (30)
where O(c
2
) denote small terms. The corresponding velocities u and accelerations u are obtained simply
by dierentiation.
We still need to determine t
s
, the time where slip stops and stick starts. This occurs after the mass
has slipped back on the belt, and then has been accelerated forward by the spring and the friction, until
396 J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403
the velocity of the mass again equals the excitation speed. Hence, we determine t
s
as the rst solution
to u(t
s
) =t
b
for t
s
0. Using (30) directly to compute u, a transcendental equation would have to be
solved numerically to determine t
s
. The most consistent way to estimate its solution seems to be the
asymptotic analysis used above. So the zero-order approximation according to (26) is t
b
cos t
s
=t
b
, with
solution
t
0
s
=2. (31)
The perturbed Eq. (30) should then determine the rst-order approximation for t
s
. However (31) is an
extreme point of the cosine function, and so the accuracy of the rst approximation will be O(

c).
Hence, for the typical values of friction dierences used in this study, i.e. (j
s
j
m
)}t
m
=0.3, the error
in t
s
will be about 55%. This problem especially concerns t
s
, whereas similar approximations for the
vibration amplitudes are fairly well (see below). If a better approximation for t
s
is in need, then one could
either consider the higher-level approximations for the solutions of (8), or use the following approach:
The simplest way is to notice, that the solution of (30) must be in the interval t
s
[; 2], and then use
a Taylor-expansion of (30) near some point within this interval. Taking e.g. the center of the interval,
one obtains
u(t
s
) = u
_
3
2
_
+ u
_
3
2
__
t
s

3
2
_
+O
_
t
s

3
2
_
2
. (32)
Letting u(t
s
) =t
b
and solving for t
s
yields an approximate solution t

s
for t
s
:
t

s
=
3
2
+
t
b
u(3}2)
u(3}2)
+O
_
t
s

3
2
_
2
(33)
or, using (30) to compute u, and u and omitting higher-order terms
t

s
=
3
2
+
t
b
+ (1}2)c
1
(13}8)c
3
t
b
(3}4)c
1
2c
3
. (34)
This approximation is very accurate for typical cases, e.g. for the parameters given in the caption for
Fig. 3 the error in t
s
as compared with numerical solutions is less than 0.3%, and is even lower for higher
dierences in friction. However, for lower levels of friction dierence, the true value of t
s
approaches
2, and expression (34) becomes more inaccurate. For such cases a more accurate approximation for t
s
can be obtained by Taylor expanding to second-order near t =2, which gives the approximation t

s
for
t
s
:
t

s
=2

2c
1
t
b
+c
1
+ 27}32c
3
+O(t
s
2)
3
(35)
for which the error is less than 1.4% for (j
s
j
m
)}t
m
0.3, smaller for the lower friction dierences,
with other parameters as above.
It does not seem possible to set up a single expression for t
s
, based on Taylor-expansion near any
particular t, that will provide good approximations for a wide range of system parameters. In essence,
the approximation t

s
is very accurate, except for situations where the stick phase is very short, for which
instead the approximation t

s
is very accurate. What is needed is therefore to merge these solutions
in a suitable manner, e.g. as in the following heuristically based expression, that favors t

s
unless the
parameters corresponds to low levels of friction dierence:
t
s
=t

s
+ e
(t
b
t
b0
)}
(t

s
t

s
),
4[t
m
3(j
s
j
m
)
, (36)
J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403 397
where is identical to the constant under the radical in (16) ( [0; 1[ for stable stickslip oscillations).
This expression for t
s
gives approximations that, compared to numerical solutions, are in error by (typi-
cally much) less than 1% for all t
b
t
b0
for [0.11; 1.00] (corresponding to (j
s
j
m
)}t
m
[0.07; 0.61]
when ([, t
m
, j
m
, j
s
) are as for Fig. 3). It should be recalled that better numerical solutions for t
s
can
always be obtained simply by solving the algebraic equation u(t
s
) =t
b
numerically, with u given by
(30).
3.3.2. Stick phase
Stick starts at t =t
s
, and then the mass just follows the belt, i.e.
u =t
b
, u =0,
u(t) =u(t
s
) +t
b
(t t
s
), t ]t
s
; 1[, (stick), (37)
where u(t
s
) is known by having applied (30) for the just completed phase of slip. The time t =1 where
stick ends is determined by applying the periodicity condition to (37), i.e. u(1) =u(0), so that
1 =t
s
+
u(0) u(t
s
)
t
b
, (38)
where u(0) and u(t
s
) is determined by (30).
3.3.3. Stickslip vibration amplitude
Considering velocity amplitudes during one stickslip cycle, we note that during stick the velocity is
constant, u =t
b
, while during slip the velocity changes continuously with a maximum absolute value at
time t =t
m
dened by the solution to u(t
m
) =0. Seeking an approximate solution, we let
t
m
=t
m0
+ct
m1
, c1. (39)
Inserting this into (23) and Taylor-expanding for small c, one nds
u(t
m
) = u
0
(t
m0
) +c(t
m1
u
0
(t
m0
) + u
1
(t
m0
)) +O(c
2
). (40)
Balancing terms of like orders of magnitude it is found that u(t
m
) =0 is approximately satised when
u
0
(t
m0
) =0, u
1
(t
m0
) =t
m1
u
0
(t
m0
). (41)
By (26) the rst equation is satised by t
m0
=. Inserting this and ((26), (29)) for (u
0
, u
1
) into the
second equation, one can solve for t
m1
and nally insert into (39) to nd (letting c =1):
t
m
=
c
1
+ (73}32)c
3
t
b
. (42)
The velocity at this time is, by (23),(39), (26), (29), t
m0
=, and Taylor expanding for c1:
u(t
m
) = u
0
(t
m0
) +c[t
m1
u
0
(t
m0
) + u
1
(t
m0
)] +O(c
2
)
=t
b
+c(}2)c
1
+O(c
2
). (43)
As a measure indicating the magnitude of oscillations, which are asymmetric with respect to u =0, we
dene the velocity amplitude A
t0
of stickslip oscillations as half the peak-to-peak velocity, i.e.
A
t0
=
1
2
(t
b
u(t
m
)), (44)
which on account of (43) and (28) becomes, ignoring terms of order c
2
and higher
A
t0
=
_
1

2
[
_
t
b
+
3
8
(j
s
j
m
)
t
b
t
m
_
1
5
4
_
t
b
t
m
_
2
_
, t
b
t
b0
. (45)
398 J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403
The corresponding displacement amplitude A
0
can be determined similarly, by calculating approximations
to the times t
m+
and t
m
where u =0 (cf. Fig. 2); this yields
t
m
=
3
2

1
2
c
1
+
52
32
c
3
, u(t
m
) =t
b
+ 2c
3
+
3
4
c
1
,
t
m+
=

2
+
1
2
c
1

52
32
c
3
, u(t
m+
) =t
b
+ 2c
3


4
c
1
. (46)
Dening the displacement amplitude A
1
as half the peak-to-peak displacement, it becomes
A
0
=
1
2
(u(t
m+
) u(t
m
)) =t
b


2
c
1
=(1 [)t
b
+
3
4
(j
s
j
m
)
t
b
t
m
_
1
5
4
_
t
b
t
m
_
2
_
, t
b
t
b0
. (47)
One can show that this function has a maximum value at t
b
=t

b
,
t

b
=
_

_
t
b
0
for (j
s
j
m
)}t
m
6
3
8
(1 ( 3)[),
t
m

4
15
_
1 +
(1 [)t
m
3}4(j
s
j
m
)
_
otherwise
(48)
where t
b0
is given by (20). Thus, if the dierence in static and kinetic friction is not too large, the
strongest stickslip oscillations occur when the excitation speed reaches the value separating stickslip
oscillations from pure-slip oscillations, t
b
=t
b0
. For larger friction dierences the strongest oscillations
occur at an excitation speed in the range t
b
]0; t
b0
[, as given by the second expression in (48). Since
the analysis assumes friction dierences that are small (but nite), then the rst case applies, so that we
conclude that the strongest oscillations occur at t
b
=t
b0
, hence
A
0
|
t
b
=t
b0
=A
0, max
=t
b0
=
_
4
5
t
m

1
4[t
m
3(j
s
j
m
)
=A
1, max
. (49)
The last equality expresses that the predicted amplitude A
1
of pure-slip oscillations equals the predicted
amplitude A
0
of stickslip oscillations at the value of excitation speed separating these dierent kinds
of motion; this exact continuity is neither obvious nor required, since the two expressions were derived
using approximate methods.
It can be noted from (47) that at small values of the excitation speed, the stickslip oscillation
amplitude grows approximately linear with this speed
A
0

_
1 [ +
3
4
_
j
s
j
m
t
m
__
t
b
for t
b
1. (50)
Finally, since the velocity of the mass must change continuously with time, the maximum and minimum
displacements of the mass must occur during the slip phase; they cannot occur during stick, because
displacements here increase linearly with time until the time of slip. Hence, the amplitude A
1
, determining
displacements during the slip phase, also determines the oscillation amplitude of the complete stickslip
oscillation.
3.3.4. Stickslip base frequency
Since at t =1 one cycle of slip and stick is completed, the base angular frequency of stickslip
oscillations is
c
ss
=
2
1
=
2t
b
t
b
t
s
+u(0) u(t
s
)
. (51)
J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403 399
This frequency is generally somewhat lower than the linear natural frequency of the system and the
pure-slip oscillation frequency, c
ss
6c
ps
1.
3.4. Summary of results
For the system (1)(2) with (4), or equivalently (8)(9) with (10), the type of stationary motion
depends on the excitation speed t
b
as follows, where t
b0
and t
b1
are given by (20) and (16), and all
results holds approximately for small but nite levels of vibration amplitudes, viscous damping [, and
relative friction dierence (j
s
j
m
)}t
m
.
For excitation speeds t
b
t
b0
: Stickslip oscillations, with stationary displacement and velocity am-
plitude A
0
and A
t0
, as given by (47) and (44), and base frequency c
ss
given by (51) (slightly less than
the linear natural frequency of the system).
For excitation speeds t
b
[t
b0
; t
b1
]: Pure-slip oscillations, with stationary displacement amplitude A
1
(equal to the velocity amplitude A
t1
) as given by (19), and base frequency c
ps
=1 (equal to the linear
natural frequency of the system). For some parameters it may happen that t
b0
=t
b1
, so that pure-slip
oscillations cannot occur at all. This occurs when the radical in (16) becomes negative, i.e. if viscous
damping is suciently large, or the friction forces or dierence between static and kinetic friction is
suciently small.
For excitation speeds t
b
t
b1
: Static equilibrium at position x(t) = x, cf. (6). This corresponds to a
state of steady sliding of the mass.
4. Illustration and testing of results
4.1. Oscillation cycles at dierent excitation speeds
Fig. 3 depicts time series and phase plane data for one cycle of stationary stickslip oscillations. Curves
in solid line show analytical predictions for displacements and velocities, u(t) and u(t), as obtained by
(30) for the slip phase and (37) for the stick phase; Curves in dashed line represent results of numerical
simulation of (1) with (4) and (7), for comparison. Each column of gures shows result for one value of
excitation speed t
b
, as given in the gure legend along with other parameters used. With these parameters
(16) and (20) gives t
b0
=0.3944 and t
b1
=0.4410; hence stickslip oscillations are predicted to occur for
belt velocities t
b
t
b0
=0.3944, whereas above that value there should be pure-slip oscillations (until
t
b
=t
b1
=0.4410), and then no oscillations at all.
For Fig. 3(a) the excitation speed is very low, and consequently the oscillation amplitudes are small
and the stick phase occupies a signicant part of the full cycle (sticking appears as horizontal lines in
the velocity plots and the phase plots). For Fig. 3(b) the excitation speed is in the midrange of velocities
yielding stable oscillations. The oscillation magnitude is signicantly increased as compared to Fig. 3(a),
but the sticking phase is relatively shorter. This increase in oscillation magnitude and decrease in relative
stick time continues until t
b
=t
b0
, where there is no stick and all and pure-slip oscillations occur; Fig.
3(c) shows this border case.
As appears, the agreement of the analytical predictions with numerical simulation is generally good,
considering that substantial approximations have been introduced to solve the highly non-linear problem
analytically. Figs. 3(a)(c) seem not to reveal any systematic error; e.g. for Fig. 3(a) the main deviation
is in the minimal displacement, which is somewhat underestimated, for Fig. 3(b) this same displacement
is slightly overestimated, whereas for Fig. 3(c) the main deviation is in the minimal velocity.
Of course, the agreement to numerical simulations ceases as the assumptions behind the analysis fail
to be fullled, i.e. as the level of friction dierence increases in magnitude; an example of that will
400 J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403
Fig. 3. One period of stickslip displacements u(t) (top row), velocities u(t) (middle), and phase plane orbits u(u) (bottom row)
for three values of excitation speed t
b
(a, b, c). () Analytical prediction (Eqs. (30) and (37)); (- - - - - -) numerical simulation
of (1) with (4) and (7). Parameters: [ =0.05, t
m
=0.5, j
m
=0.25, j
s
=0.4, and (a) t
b
=0.05, (b) t
b
=0.25, (c) t
b
=0.3944.
appear below. Also, as the friction dierence is decreased the approximations become more accurate;
for example, if Fig. 3 is re-computed for (j
s
j
m
)}t
m
=0.15 instead of 0.3, then the results from using
analytical approximations is almost indistinguishable from numerical simulation (only Figs. 3(a) and (b)
would apply in that case).
4.2. Displacement amplitude as a function of excitation speed
Fig. 4(a) shows the predicted variation of displacement amplitude with excitation speed for typical
parameters. Here A
0
indicates the amplitude during stickslip oscillations for t
b
t
b0
according to (47),
and A
1
is the amplitude for pure-slip oscillations for t
b
[t
b0
; t
b1
] according to (19). As appears, when
the excitation speed is increased from zero, stickslip oscillations occur with increasing amplitude until,
at t
b
=t
b0
pure-slip oscillations take over. These prevail only in a narrow range of belt velocities, until
t
b1
, above which oscillations cease and steady slip becomes the stable type of motion. The agreement
J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403 401
Fig. 4. (a) Amplitude A and (b) base frequency c of stable periodic motions as a function of excitation speed t
b
. (-) Analytical
prediction (Eq. (19), (47), (51)); (- - - - - -) analytical prediction from Thomsen [15]; (O) Numerical simulation of (1) with
(4). Parameters as for Fig. 3.
of the analytical predictions with numerical simulation (encircled points) is seen to be very good; this
will hold true the more closely the actual parameters satises the assumptions underlying the analytical
expressions, i.e. as long as the dierence in static and kinetic friction, the amount of viscous damping,
and the amplitudes are not too large. It also appears that the predictions for the stickslip amplitudes A
0
are much better than the low-order approximation (indicated in dashed line) given in [15].
Fig. 4(b) depicts the predicted variation of base frequency c with excitation speed, showing a slight
drop in frequency for the lower velocities. Seemingly, the agreement with numerical simulation is quite
good; however, since the change in frequency is so small, the relative error in the prediction is somewhat
larger than for the displacement and amplitude predictions. This is a consequence of the diculty in
predicting, with high accuracy, the end of the slip phase of a stickslip cycle, based on an approximate
expression for the displacements during that cycle.
4.3. The signicance of friction dierence
Fig. 5 depicts how the quality of the analytical predictions changes as the assumptions behind the
analysis fails to hold true. The gure shows the variation of displacement amplitudes with excitation
speedjust as Fig. 4(a), but for dierent levels of friction dierence (j
s
j
m
)}t
m
and compare these
to results of numerical simulation.
It is evident from the gure that the quality of the analytical predictions deteriorates, from good to
poor, as (j
s
j
m
)}t
m
is raised from a value well within the assumed smallness (0.075, circle symbols)
to a value closer to unity (0.6, diamond symbols). The latter quite high level of friction dierence is an
example where the second equation in (48) applies, e.g. the amplitude has a maximum at an excitation
speed lower than t
b0
, actually at t
b
=t

b
0.33, according to (48). Higher-order approximations will be
in need to improve on the accuracy for such large levels of friction dierence.
The gure also illustrates how, quite expectedly, the magnitude of oscillation amplitudes and the range
of oscillation-producing velocities, both increases with the level of friction dierence. For the smaller
values of friction dierence, the relation between stickslip amplitudes and excitation speed is very close
to being linear, as predicted by (50).
402 J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403
Fig. 5. As Fig. 4(a), but for varying levels of relative friction dierence (j
s
j
m
)}t
m
. (-) Analytical prediction (Eq. (19),
(47)); (, , , , ) numerical simulation of (1) with (4) and, respectively (j
s
j
m
)}t
m
=(0.075, 0.099, 0.15, 0.3, 0.6). Other
parameters: [ =0.05, t
m
=0.5, j
s
=0.4.
5. Conclusions
The analytical expressions for stationary friction-induced oscillations produce predictions that agree
favorably with numerical simulations, as long as the assumption of relatively small dierences in static
and kinetic friction is fullled. This holds for stickslip oscillations, that occur for relatively small
excitation speeds t
b
t
b0
, and for pure-slip oscillations that occur for a narrow range of slightly higher
speeds t
b
[t
b0
; t
b1
]. Beyond this range, according to the predictions and numerical simulations, the
system exhibits steady sliding with no oscillations.
The expressions provide insight into the inuence of parameters on vibration characteristics such as
amplitude and base frequency; features that can be measured in experiments and compared to assess the
validity of the underlying friction model.
Although results are given in terms of a specic friction model, the approach used is not tied to this.
It can readily be applied to set up and test expressions for other friction models having the properties of
being (1) discontinuous only at zero-valued relative velocity, and (2) suciently smooth (for non-zero
relative velocities) to enable Taylor-expansion to an order high enough to include motion-restricting
non-linearities (usually three).
References
[1] J.W.S. Rayleigh, The Theory of Sound, 2nd Edition, 1945 re-issue. Dover Publications, New York, 1877.
[2] Y.G. Panovko, I.I. Gubanova, Stability and Oscillations of Elastic Systems; Paradoxes, Fallacies and New Concepts,
Consultants Bureau, New York, 1965.
[3] A. Tondl, Quenching of Self-Excited Vibrations, Elsevier, Amsterdam, 1991.
[4] A.H. Nayfeh, D.T. Mook, Nonlinear Oscillations, Wiley, New York, 1979.
[5] Y.A. Mitropolskii, V.D. Nguyen, Applied Asymptotic Methods in Nonlinear Oscillations, Kluwer, Dordrecht, 1997.
[6] K. Popp, Some model problems showing stickslip motion and chaos, Friction-Induced Vibration, Chatter, Squeal, and Chaos,
ASME DE-Vol. 49, ASME Design Engineering Divison, New York, 1992, pp. 112.
[7] R.A. Ibrahim, Friction-induced vibration, chatter, squeal, and chaos: part Imechanics of friction, Friction-Induced Vibration,
Chatter, Squeal, and Chaos, ASME DE-Vol. 49, ASME Design Engineering Divison, New York, 1992, pp. 107121.
[8] R.A. Ibrahim, Friction-induced vibration, chatter, squeal, and chaos: part IIdynamics and modeling, Friction-Induced
Vibration, Chatter, Squeal, and Chaos, ASME DE-Vol. 49, ASME Design Engineering Divison, New York, 1992,
pp. 123138.
[9] J. McMillan, A non-linear friction model for self-excited vibrations, J. Sound Vibration 205 (3) (1997) 323335.
[10] B. Feeny, Ard eshir Guran, N. Hinrichs, K. Popp, A historical review on dry friction and stickslip phenomena, ASME Appl.
Mech. Rev. 51 (5) (1998) 321341.
J. Juel Thomsen, A. Fidlin / International Journal of Non-Linear Mechanics 38 (2003) 389403 403
[11] B. Armstrong-H elouvry, P. Dupont, C. Canudas de Wit, A survey of models, analysis tools and compensation methods for
the control of machines with friction, Automatica 30 (7) (1994) 10831138.
[12] B. Armstrong-H elouvry, A perturbation analysis of stickslip, Friction-Induced Vibration, Chatter, Squeal, and Chaos, ASME
DE-Vol. 49, ASME Design Engineering Divison, New York, 1992, pp. 4148.
[13] C. Gao, D. Kuhlmann-Wilsdorf, D.D. Makel, The dynamic analysis of stickslip motion, Wear 173 (1994) 112.
[14] F.-J. Elmer, Nonlinear dynamics of dry friction, J. Phys. AMath. General 30 (1997) 60576063.
[15] J.J. Thomsen, Using fast vibrations to quench friction-induced oscillations, J. Sound Vibration 228 (5) (1999) 10791102.
[16] A. Fidlin, J.J. Thomsen, Predicting vibration-induced displacement for a resonant friction slider, European J. Mech. A}Solids
20 (2001) 155166.
[17] A. Fidlin, On the asymptotic analysis of discontinuous systems, DCAMM Report 646, Technical University of Denmark,
2000.
[18] A.Y. Fidlin, On averaging in systems with a variable number of degrees of freedom, J. Appl. Math. Mech. 55 (4) (1991)
507510.
[19] A.V. Pechenev, A.Y. Fidlin, Hierachy of resonant motions excited in a vibroimpact system with contact zones by an inertial
source of limited power, Izv. AN SSSR. Mechanika Tverdogo Tela 27 (4) (1992) 4653.
[20] N. Hinrichs, M. Oestreich, K. Popp, On the modelling of friction oscillators, J. Sound Vibration 216 (3) (1998) 435459.
[21] J.J. Thomsen, Vibrations and Stability, Order and Chaos, McGraw-Hill, London, 1997.

You might also like