Ouyang 1999
Ouyang 1999
SAE TECHNICAL
PAPER SERIES 1999-01-3409
M. P. Cartmell
University of Glasgow
400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
The appearance of this ISSN code at the bottom of this page indicates SAE’s consent that copies of the
paper may be made for personal or internal use of specific clients. This consent is given on the condition,
however, that the copier pay a $7.00 per article copy fee through the Copyright Clearance Center, Inc.
Operations Center, 222 Rosewood Drive, Danvers, MA 01923 for copying beyond that permitted by Sec-
tions 107 or 108 of the U.S. Copyright Law. This consent does not extend to other kinds of copying such as
copying for general distribution, for advertising or promotional purposes, for creating new collective works,
or for resale.
SAE routinely stocks printed papers for a period of three years following date of publication. Direct your
orders to SAE Customer Sales and Satisfaction Department.
Quantity reprint rates can be obtained from the Customer Sales and Satisfaction Department.
To request permission to reprint a technical paper or permission to use copyrighted SAE publications in
other works, contact the SAE Publications Group.
No part of this publication may be reproduced in any form, in an electronic retrieval system or otherwise, without the prior written
permission of the publisher.
ISSN 0148-7191
Copyright 1999 Society of Automotive Engineers, Inc.
Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely
responsible for the content of the paper. A process is available by which discussions will be printed with the paper if it is published in
SAE Transactions. For permission to publish this paper in full or in part, contact the SAE Publications Group.
Persons wishing to submit papers to be considered for presentation or publication through SAE should send the manuscript or a 300
word abstract of a proposed manuscript to: Secretary, Engineering Meetings Board, SAE.
Printed in USA
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
1999-01-3409
M. P. Cartmell
University of Glasgow
1
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
of thousand of degrees-of-freedom. However, because the trailing edge. Effectively, the contact elements are of
we want to include the relative rotation between the disc orthotropic material. If the Young’s modulus of some con-
and the pads, a simple finite element model, which cap- tact elements is taken to be zero, then these elements
tures the basic feature of the real disc brake system, represent a very narrow gap, for example between the
such as the pads, the caliper with fingers, piston and car- pad and the piston-side caliper. Therefore, it is advanta-
rier, the cage and their mounting, should be sufficient for geous to put a layer of contact elements at the interface
a preliminary study. We model the disc of regular geome- between any two separate parts. If it is judged that part of
try as an annular, thin, flat plate. As a result, a neat ana- the interface is actually a gap, the Young’s modulus of the
lytic description of the dynamics of the disc can be corresponding contact elements are assigned zero. The
derived. The equation of motion for the whole disc brake simple finite element model of the pads, caliper and
system is established by coupling the dynamics of the mounting is shown in Figure 1. It contains about four hun-
disc and the dynamics of the pads, caliper and mounting. dred 3-D brick elements and spring elements. The con-
Pads, caliper and cage are separate parts. There are tact elements are brick elements with one size being vary
contact elements modeling frictional interfaces between small (between the pads and the caliper) or dimension-
the pads and the disc and between the pads and the cal- less spring elements (between the pad side faces and
iper. In the present paper, all materials are assumed to the cage). The schematic of the whole disc brake system
be linear and elastic. The effect of wear on the pad sur- is given in Figure 2.
face could be simplistically modeled as a thin layer of
When the finite element model is constructed, the equa-
contact elements. Temperature variation and its effect are
tion of motion for the pads, caliper and mounting, which
not considered here but can be included in subsequent
are stationary, can be established as
work. The state space method is employed to solve the
eigenvalues of the whole dynamic system after the trun- Mx&& + Cx& + Kx = f
cation of the modes of both the disc and the pads, caliper (1)
and mounting. These eigenvalues are shown to be influ- The nodal force vector and nodal displacement vector are
enced by the system parameters such as the damping divided into the following two parts
and the stiffness of the pads and the operating conditions
such as the rotating speed of the disc. The real parts of f T = [ f pT f oT ] x T = [ x pT xoT ]
the eigenvalues, indicating whether the system is stable (2)
or not, are presented to show the effects of the system where
parameters of the disc brake system. Squeal is supposed
to be the consequence of the unstable parametric reso- x pT = [ u1 , v1 , w1 ,......, u j , v j , w j ,......, u2 j , v2 j , w2 j ]
nances induced by the rotating dry friction between the (3)
disc and the pads. The first part in equation (2) corresponds to the nodes at
the disc/pad interfaces while the second is for all the
FINITE ELEMENT MODEL OF THE PADS, other nodes in the stationary components. We want to
CALIPER AND MOUNTING derive the nodal force vector at the pad/disc interface in
terms of the nodal displacement vector at the same loca-
A disc brake system consists of the following major com- tion, as well as the modal information of the pads, caliper
ponents: the disc, the two pads with back plates, the cali- and mounting. Since the disc brake is an autonomous
per, the cage and their mounting to the hub. As the disc system without excitation and the mechanism for self-
rotates past the other components which are stationary, excited vibration is the rotating disc rubbing past the pads
the disc and the other components may be treated differ- through dry friction,
ently.
f pT = [ 0, µ 1 ( p1 − n1 ), p1 − n1 ,......, 0, )
In a disc brake system, the pads sit loosely within the cal-
iper and bounded by the cage. The pads, caliper and µ j ( p j − n j , p j − n j , p j − n j ,......, 0,
cage are separate parts which come into contact when µ 2 j ( p2 j − n2 j ), p2 j − n2 j ] f oT = 0
the brake is applied. The finite element model has to (4)
reflect this feature. As such, a very thin layer of contact The friction forces at the disc/pad interface is assumed to
elements are built between the pads and the caliper and be proportional to the normal forces by a phenomenolog-
between the pads and the disc. The contact elements ical friction coefficient µi at the ith node at the disc/pads
have very high stiffness across their thickness (in the z interface, which can be a function of the relative speed
direction) and very low stiffness in the other two direc- (rotating speed Ω in this instance), the normal force (or
tions. This means that the two surfaces in contact pressure), or its position, should reliable data be avail-
through the contact elements can have relative displace- able. For simplicity, we take µ to be a constant in this
ments in the plane of the contact surfaces while the dis- paper. When data of the dependency of µ on relative
placements normal to the surfaces are nearly equal. speed, or pressure or temperature are available, they can
Similarly, there are contact elements between the pad be accommodated into the model in this paper. Such a
side surfaces and the cage. However, those contact ele- simple friction law allows better modeling on other
ments at leading edge have lower stiffness than that of aspects of the disc brake system. Other uncertainties
2
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
due to the frictional contact are taken care of by a layer of where Xwp are formed by those rows of Xp which corre-
contact elements in a simplistic way. spond to the w components.
The natural frequencies and modes can be computed to Now we derive the expression of the nodal forces in z
any degree of accuracy in theory. As far as squeal is con- direction at the disc/pads interface, p, in terms of the
cerned, the frequencies of the pads, caliper and mount- nodal displacements at the same positions, xw, as
ing around the range of squeal frequencies are thought to
contribute to the dynamics of the disc. Hence, not all of d2 d
p(t ) = {X wp ( I + 2diag[ β cω 2 ] +
them would be needed. The exact number of modes dt 2
dt
needed to be retained can be determined through a few diag[ω 2 ]) −1 X µT }−1 x w (t ) + n
trial runs of the final computer program. (8)
To facilitate subsequent mathematical development, These forces fully represent the dynamics of the station-
equation (1) is decoupled using the orthogonality of the ary components of the system. It should be stressed at
modal vector matrix X and expressed as this point that modal truncation has not been performed
in equation (8) and so far there has not been any further
q&& + 2diag[ β c ω 2 ]q& + diag[ω 2 ]q = X T f mathematical approximation besides finite element dis-
= X pT f p = X µT ( p − n) cretization. The force vector n due to the static pressure
(5) appears in equations (5) to (8) because the vibration con-
cerned occurs as being superimposed on an initial stress
when a Kelvin-type damping is assumed.
problem caused by the braking pressure so that the brak-
The matrix X µT is derived by rearrangement of the terms ing-pressure induced nodal force vector n should be
in X p fp to give X µ ( p − n) . Thus,
T T
deducted from the overall nodal force vector p. The net
normal force due to the static pressure across the disc is
zero since the disc is in equilibrium under the static brak-
ing pressure. The dynamics of the pads, caliper and
mounting is manifested in X, diag[βcω2] and diag[ω2],
which are obtained by running a commercial symbolic
software package. The inverse matrix operations in equa-
tion (8) are calculated as pseudo-inverses by the same
software.
3
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
lytical effort, which should be obvious in the following MODELING OF SLIDING CONTACT BETWEEN
development. THE PADS AND THE DISC
As squeal is much more likely at low speed, we will omit
the gyroscopic and centrifugal effects. The equation of When the brake is applied, the pads are pressed into slid-
free transverse motion of an annular thin plate in a cylin- ing contact with the rotating disc. The dry friction
drical co-ordinate fixed to the disc, is between the disc and the pads creates a thin layer of
special material and heats up the disc and the pads. The
∂2 w physics of the thin layer of special material is hard to cap-
ρh + D∇ 4 w = 0 ture and model. As we are interested in the vibration of
∂t 2 (9) the whole brake system, which is a measure of global
where behavior of the whole system, we feel justified to use a
simple friction law and contact model, instead of a com-
∂2 1 ∂ 1 ∂2 2 plicated friction law. Additionally, as the temperature at
∇4 = ( + + 2 ) squealing during each testing is kept constant, we can
∂r 2
r ∂r r ∂θ 2 (10) omit the effects of temperature variation in this prelimi-
The boundary conditions in this case are nary investigation on squeal. Of course, different con-
stant temperatures are allowed in different runs.
∂w
w = 0, =0 (at r = a ) As we use the finite element method for the pads, the
∂r normal forces and frictional forces at the contact interface
Mr = 0 , Vr = 0 (at r = b) are concentrated forces. The simplest contact elements
(11)
are dimensionless and massless springs linking the mid
where plane of the disc to the opposing nodes on the pad sur-
face, as illustrated in Figure 4. These kind of elements
∂2 w 1 ∂w 1 ∂ 2 w
M r = − D[ + ν ( + )] are used by Ripin (1995) in his finite element model of a
∂r 2 r ∂r r 2 ∂θ 2 car disc brake. The equation of equilibrium for the spring
∂ ∂ element is
Vr = −[ D (∇ 2 w) + (1 − ν ) ×
∂r r∂θ
~−x ) = p−n
diag[ k c ]( w
( 1 ∂ w − 12 ∂w )]
2 w
(16)
r ∂r∂θ r ∂θ (12) where the spring constant kc is a constant for all the con-
tact elements at the disc/pads interface, though it would
The equivalent inner and outer radii of the reduced thin
better to be a function of local normal force, temperature
plate are to be determined by means of modal testing of
and so on.
the brake disc bolted to the wheel hub, and subsequent
adjustment of the parameters of the model. Substituting equation (7) into (16) leads to
The free vibration solution of equation (9) can be repre-
d2
sented through a modal expansion as ( p − n) = {diag[1 / k c ] + X wp ( I +
dt 2
∞ ∞ d
w( r , θ , t ) = ∑ ∑ ψ kl ( r ,θ ) q kl ( t ) 2diag[ β c ω 2 ] ~
+ diag[ω 2 ]) −1 X µT } −1 w
k = 0 l = −∞ dt
(13) (17)
where As w~ is the vector of the deflection of the rotating disc,
then
1
ψ kl (r ,θ ) = Rkl (r ) exp( ilθ ) ~ T = {..., w(r , θ − Ω t , t ), ...} =
ρhb 2 w i i
(14)
∞ ∞ Rkl (ri )
The modal functions satisfy the ortho-normality condi- {..., ∑ ∑ exp[il (θ i − Ωt )]q kl (t ), ...}
k = 0 l = −∞
ρbh 2
tions, (18)
b If a complete contact between the pads and the disc is
∫ ρhψ klψ mn rdrdθ = δ kmδ ln , assumed, kc → ∞, then equations (16) and (17) reduce to
a
b
∫ Dψ kl ∇ ψ mn rdrdθ = ω mn δ kmδ ln .
4 2
~, d2
a
(15) xw = w ( p − n) = { X wp ( I +
dt 2
where the overbar denotes complex conjugation. d ~
2diag[ β c ω 2 ] + diag[ω 2 ]) −1 X µT } −1 w
dt (19)
4
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
DYNAMICS OF THE DISC UNDER BRAKING on each face of the pads in contact with the disc. This is a
simple model for a real brake arrangement. However, it
As the pads are modeled with finite elements, the disc is includes all basic features, with fingers, piston, carrier
in contact with the pads through a number of nodes at the and caliper cage.
disc/pad interfaces and is subjected to the nodal forces at In each separate analysis, one parameter is chosen to
the interface. Thus, the equation of motion of the disc vary while all the other parameters are kept constant.
under brake application is Therefore, the influence of this specific parameter is
determined. More specifically, the eigenvalues of the
∂2 w ∂w
ρh + D *∇ 4 + D∇ 4 w = whole system are computed. The positive real parts of
∂t 2
∂t the eigenvalues denotes unstable modes whereas the
1 imaginary parts represent the natural frequencies of the
− d T (r , θ , t )( p(t ) − n)
r (20) brake system. After all the parameters of interest are
studied in this way, a general picture of the instability
where behavior of the whole brake system can be conceived.
Unless otherwise stated, fifteen disc modes and five hun-
d T = [δ (r − r1 )δ(θ − θ 1 − Ω t ) ,......, dred modes of the stationary components of the system
δ(r − rj )δ (θ − θ j − Ω t ) ,......, are used. The use of a deliberately high number of
δ(r − r2 j )δ (θ − θ 2 j − Ω t )] modes of the pads, caliper and mounting is to ensure
(21) high accuracy in determining the eigenvalues. When a
and rs and θs (s = 1,2,…) specify the positions of j nodes reduction process is employed, this number can be
at either side of the disc on the disc/pad interfaces. expected to be reduced to a few tens of modes. This
reduction is essential when a large, realistic finite ele-
Using equation (15), equation (20) can be re-written in ment model for the stationary components of the brake
terms of the modal co-ordinates of the disc as system is built and used in future analysis.
5
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
gators that above certain speed disc brakes tend to stiff pads is the 12th pair rather than the 14th in the case
become less noisy. Numerical results show that smaller of normal pads and this pair of modes become unstable
friction coefficients lead to reduced regions of instability. at a lower rotating speed of the disc. Thus, very stiff pads
Therefore, when a speed-dependent friction coefficient is have adverse effect on noise behavior.
used, which decreases with increasing relative speed
On the other hand, soft pads do not help either. The real
within a certain speed range, it is expected that the
parts of eigenvalues for soft pads (Young’s modulus
present model can predict this observation.
reduced by 20%) are given in Table 5 (βc = 5 × 10-8,
Secondly, we want to see how damping of the pads βd = 0, µ = 0.4).
affects the instability. The numerical results are illustrated
Finally, it can be observed that the instability behavior is
for damping of the pads in Table 2 (Ω = 10rad/s, βd = 0,
very much speed dependent, even at low speeds (see
µ = 0.4). As can be seen from Table 2, most modes would
the 14th pair of modes in Table 5).
be unstable under rotating friction when damping is
absent. However, a very small damping would bring It should be stressed that all the above findings are
many of those modes to stability. Another way of stating based on several important assumptions made in this
this fact is that the system stability is highly dependent on paper. Once reliable data on friction characteristics are
the damping present in the pads. It can also be seen that available, a better frictional law can be use, which may
some modes are more sensitive to damping than others. also take care of wearing between the disc and the pads.
Note that the frequencies (the imaginary parts of the sys- The assumption that the disc is in complete contact with
tem eigenvalues) of the whole system are not listed here the pads can be removed. However, this will lead to a
as variations of neither the rotating speed nor the damp- computationally expensive time-stepping scheme. Ther-
ing of the pads affects them. mal effect may be included in the model similarly. It is
easy to accommodate nonhomogeneous or nonisotropic
Thirdly, we examine the stiffness of the stationary compo-
material properties into the stationary components. Gyro-
nents of the brake system. To do this, we introduce α as a
scopic and centrifugal effects are negligible. This
stiffness ratio such that αΕ is used as Young’s modulus of
assumption should be removed for high speed rotors.
the stationary components in the following computations.
However, it is considered to be appropriate for the car
The results are shown in Table 3 (βc = 5 × 10-8, βd = 0,
disc brake squeal problem.
µ = 0.4, Ω = 10rad/s). This is a very interesting system
parameter in that some unstable modes disappear and
some stable modes become unstable as the parameter
CONCLUSIONS
value increases.
Based on our assumptions (that the disc is a thin, elastic
Table 3 shows that the influence of the stiffness change plate in complete contact with the pads, that the pads
on the stability of the system is more dramatic. As α and the disc are rotating relative to each other at constant
increases from 0.25 to 2.25 (the frequencies of the sta- speed, gyroscopic and centrifugal effects are negligible
tionary components alone from 0.5 to 1.5), the real parts and the friction characteristics are pressure- and temper-
of the 1st pair, the 8th pair and 14th pair of eigenvalues ature-independent) and the numerical results obtained,
all experience fluctuation, though the 1st pair of modes we can draw the following conclusions:
are quite insensitive to the stiffness ratio and are always
1. The 1st pair of modes related to the lowest frequency
close to the boundary of the instability regions. The fluc-
of the disc brake system is always unstable or close
tuation of real parts of the eigenvalues means that the
to the instability boundary when an instability is
degree of instability depends on the range of α values
present.
involved. For example, in the range of α =0.25 to α =0.64,
the instability associated with the 14th pair of modes 2. The increase of the rotating speed of the disc make
increases while it decreases in the adjacent range of some modes more stable and some modes more
α =0.64 to α =1. There is an opposite trend for the 8th unstable. The stability of the system is speed-depen-
pair of modes in the above two ranges of α values. In dent.
terms of instability, one has to examine individual modes 3. The damping of the pads is always stabilizing. Which
one by one. modes become unstable and which unstable mode
appears first highly depend on the damping level and
If a stiff pad material (say, the Young’s modulus is
the stiffness of the stationary components of the
increased by 40%) is used, the stability of the system
brake system.
improves. But if a very stiff pad material (say, the Young’s
modulus is doubled) is used, the stability of the system 4. Neither the damping of the disc nor the rotating
becomes worse. Such a case is presented in Table 4 speed of the disc affects the frequencies of the brake
(βc = 5 × 10-8, βd = 0, µ = 0.4).Table 4 indicates that some system to a significant degree.
modes (the 1st and the 5th pairs) are insensitive to the 5. The stiffness of the stationary components has a
speed change while other modes are sensitive. The 1st subtle influence on the dynamic instability of the sys-
pair of modes are always around the boundary of the tem. However, a general trend does not exist. Evalua-
instability region. Compared with Table 1, it can be seen tion of system instability affected by the stiffness
that the first unstable pair of modes in this case of very must be carried out on an individual basis.
6
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
7
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
30. Shen, I. Y. (1993) ‘Response of a stationary damped r radial co-ordinate in cylindrical co-ordinate
circular plate under a rotating slider bearing system’, system.
Trans. ASME, J. Vibr. Acoust., Vol.115, pp.65-69 t time.
31. Yang, S. and Gibson, R. F. (1997) ‘Brake vibration u,v,w displacements in cylindrical co-ordinate system.
and noise: review, comments and proposals’, Int. J.
wi displacement in z direction of i th node on the
Mater. & Product Tech., Vol.12, Nos.4-6, pp.496-513
disc/pad interfaces.
32. Yuan, Y. (1995) ‘A study of the effects of negative fric- x nodal displacement vector for the pads, caliper
tion-speed slope on brake squeal’, 1995 Design
and mounting.
Engineering Technical Conferences Proceedings,
Boston, Vol.3, Part A, pp.1153-1162 xo nodal displacement vector of x less xp.
xp nodal displacement vector of nodes on the pad
CONTACT surfaces in contact with the disc.
xw displacement vector formed by the elements of
Please contact Dr. H Ouyang or Professor J E Motters- w in xp.
head for further information regarding this paper. Their z axial co-ordinate of the cylindrical co-ordinate
email addresses are: [email protected] and J.E.Mot- system.
[email protected]. D flexural rigidity of the disc.
D* Kelvin-type damping coefficient of the disc.
NONMENCLATURE E Young’s modulus.
Rld combination of Bessel functions to represent
a,b inner and outer radii of the thin, flat plate as a the mode shape of the disc in the radial
substitute for the real disc. direction.
ƒ nodal force vector for pads, caliper and X modal vector matrix for the pads, caliper and
mounting. mounting.
ƒo nodal force vector of ƒ less ƒp. Xp part of X corresponding to xp.
ƒp nodal force vector for nodes on the pad surfaces α stiffness ratio for the pads, caliper and mounting
in contact with the disc. (stationary components).
h thickness of the disc. βc Kelvin-type damping coefficients for the pads,
i −1 . caliper and mounting.
j number of nodes on the either pad face which is βd Kelvin-type damping coefficients for the disc.
in contact with the disc. δ (•) Dirac delta function.
kc spring constant of the contact elements δld Kronecker delta.
between the disc and the pads. θ circumferential co-ordinate of cylindrical
k, l number of nodal circles and number of nodal co-ordinate system.
diameters in a mode of the disc. µ kinetic dry fiction coefficient between the pads
m, n number of nodal circles and number of nodal and the disc.
diameters in a mode of the disc. ν Poisson’s ratio of the disc material.
n nodal vector of initial normal loads on the disc ρ mass-density of the disc.
from the pads due to braking pressure. Ψld mode shape function for the transverse
p total concentrated force in z direction on the vibration of the disc corresponding to qld.
disc at the pad/disc interfaces. ω natural frequencies of the pads, caliper and
p force vector consisting of all p’s at the pad/disc mounting.
interfaces. ωld natural frequency corresponding to qld.
qld(qmn) modal co-ordinate for k(m) nodal circles and Ω constant rotating speed of the disc in radians
l (n) nodal diameters for the disc. per second.
q modal co-ordinate vector corresponding to x.
8
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
βc × 108 0 1 3 5
1st Pair 0.0006 0.0004 -0.0001 -0.0006
2nd Pair 0.0045 0.0025 -0.0015 -0.0056
3rd Pair -0.0005 -0.0110 -0.0320 -0.0530
4th Pair 0.017 -0.010 -0.065 -0.119
5th Pair -0.0013 -0.0720 -0.210 -0.353
6th Pair 0.0385 -0.105 -0.390 -0.677
11th Pair 0.132 -4.97 -15.19 -25.40
12th Pair 4.59 -0.97 -12.16 -23.33
13th Pair -4.96 -10.89 -22.76 -34.64
14th Pair 19.31 16.35 10.44 4.53
Table 4. Real parts of some representative eigenvalues for very stiff pads
Ω (rad/s) 0 5 10 15 20 25 30
1st Pair -0.0015 -0.0012 -0.0009 -0.0007 -0.0004 -0.0002 0.0001
5th Pair -0.36 -0.36 -0.36 -0.36 -0.36 -0.35 -0.35
12th Pair -5.97 0.56 1.17 1.29 1.32 1.33 1.33
14th Pair -4.74 -3.25 -1.16 1.67 3.58 5.61 6.13
9
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
Figure 1. Finite element model of the stationary components of a car disc brake
10
Downloaded from SAE International by Univ of California Berkeley, Tuesday, July 31, 2018
11