Linear friction damper consisting of cylindrical friction

block and inclined lever (Investigation of fundamental

property)
H. Yamaguchi1, H. Yoshida2
1
National Defense Academy, Department of Mechanical Systems Engineering,
Hashirimizu 1-10-20, Yokosuka, Kanagawa, 239-8686 Japan
e-mail: hideya@nda.ac.jp
2

National Defense Academy, Department of Mechanical Systems Engineering,

Hashirimizu 1-10-20, Yokosuka, Kanagawa, 239-8686 Japan

Abstract
For passive isolation systems, a friction damper is often used because it is simple, inexpensive and
reliable. However, the ordinary friction damper of constant friction force has performance limitations.
This is because the isolation characteristic declines when the friction force is large, while the resonance
peak becomes large when the friction force is small. Furthermore, the displacement remains apart from
the equilibrium position after the disturbance disappears. In order to overcome the above drawbacks, a
new type of friction damper is proposed. In the new friction damper, the friction lever mechanism is
devised to make the magnitude of the friction force proportional to the relative displacement. This is
achieved by use of an inclined lever, which contacts the cylindrical block by means of a rotational
spring. When the cylindrical block moves, the angle of lever inclination varies, causing the normal and
friction forces at the contact surface to vary proportional to the displacement of the cylindrical block.

Introduction

Performance of precision machinery such as the semiconductor inspection equipment is diminished by

the vibration transmitted from the outside. Further, any instrument that generates vibration such as the
pressing machine has a negative influence on the performance of other machinery. Therefore, vigorous
studies of the vibration isolation system have been conducted by many researchers. Among various
types of the vibration isolation system, the system consisting of a spring and a friction damper has the
advantage that substantial vibration damping is achieved using simple mechanism with low maintenance
requirements.
In most of the previous work on the friction damper, it has been assumed that the magnitude of the
friction force is constant. However, it is known that the conventional damper of constant friction force
has performance limitations; A) a resonant peak becomes infinite when the friction force is small, while
the frequency range where the vibration is not isolated will expand when the friction force is large [1,
2], and B) the displacement remains apart from the equilibrium position after the vibration ceases.
Previous works have included various approaches to the problem. Beucke et al. [3] described a friction
damping model based on experimentation where the friction coefficient is linearly dependent on the slip
amplitude. Their model is the combination of viscous damping, constant Coulomb friction and linear
Coulomb friction. Makris et al. [4, 5] analyzed the transient and steady state responses of the constant
and linear Coulomb friction oscillators addressing the analysis of multiple stops of the motion in detail.
Tadjbakhsh et al. [6] investigated the energy absorbing base isolation system that provides friction
resistance in proportion to the slip between the building and its foundation. Two sets of friction plates
are attached between the footing wall and the building. When slip occurs, the plates are tightened

1317

1318

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

together by a pair of cables arranged in series with a spring and dashpot.

Authors [7] have also proposed a vibration isolation system with a new linear friction damping
mechanism where the magnitude of the friction force increases proportionally to the relative
displacement. In this mechanism, the friction lever that rotates around the pivot on the isolation stage is
connected to the restoring spring on the base. When the relative displacement occurs between the stage
and the base, the lever is pressed to the friction plate on the base by the restoring spring. On this
occasion, the friction force between the end of the lever and the friction plate is proportional to the
relative displacement because the magnitude of the normal force at the sliding interface varies
proportionally to the relative displacement. By numerical simulation and experimentation, effectiveness
of the proposed mechanism was verified. However, optimal and simultaneous tuning of the friction
force and the restoring force was complex.
Recently, authors proposed [8] the simple mechanism of a linear friction damper in which the friction
force can be set up independent of the restoring spring. This is achieved by contacting the inclined
lever with the rotational spring to the cylindrical block. When the cylindrical block moves and the
angle of inclination of the lever varies, the normal and friction forces at the contact surface vary
depending on the displacement of the cylindrical block. The experimental and the numerical simulation
results showed that the expected performance was obtained. However, depending on the experimental
conditions, vibration due to "Sprag-slip" occurred in some cases. Through additional experiments, the
main contributing factor of the vibration was found to be the flexural deformation and the large
moment of inertia of the lever [9].
This paper shows the results of improved numerical simulations and experiments which take the above
mentioned factor into consideration.

Mechanism and working principle

The model of the linear friction damper proposed in this paper is shown in Figure 1. The inclined lever
that can rotate around the pivot OL is contacting with the cylindrical block of radius R at Point C.
When the cylindrical block moves to the right and the lever is forced to rotate clockwise, the moment
M caused by the rotational spring at the pivot OL increases. Therefore, the friction force F increases,
since the normal force P at the contact point C increases.
It is noted that the rotational spring is set up so that the normal force P is set to zero when the center
of the cylindrical block, OB, is located at x= 0 i.e. the origin O. In this case, we define the distance
between the origin O and the pivot OL as the fundamental length l; and the inclination angle of the
lever as the fundamental angle of inclination q0. Furthermore, distances from the pivot OL to the
contacting surface of the lever (frictional interface) and the x-axis are denoted by the frictional interface
offset e and the axial offset h, respectively. When the center of the cylindrical block OB is located in

Lever

s
C

q' e
h

OB

O
x

M
OL

Pivot

Cylindrical
Rotational
Block
Spring

Figure 1: Linear friction damper model.

DAMPING

1319

7
x

y
0

11
10

1:Table
2:Base
3:Linear Guide
4:Restoring Spring
5,6:Spring Support
7:Cylindrical Block
8:Lever
9:Pivot
10:Spring
11:Spring Support

(a) Top view.

z
0

x 57 8

6 4

11 3

(b) Side view.

Figure 2: Example of vibration isolation system.
the position x, the angle of lever is denoted by q ' and the distance from OL to the contact point C is
denoted by s.
In order to clarify the working principle of this friction damper, the example of the damper being
applied in the vibration isolation equipment is shown in Figure 2. This vibration isolation system will
be researched in the near future. Figure 2(b) is the side view of the vibration isolation equipment, and
Figure 2(a) is the top view when the table 1 is removed. The table 1 is connected to the base 2 via the
linear guide 3 and the restoring spring 4, and the table moves in the x direction. The cylindrical block 7
is attached under the table 1, and the lever 8 is placed so as to contact with the cylindrical block. This
lever, which can rotate around the pivot 9 fixed to the base, is pressed on the cylindrical block by the
spring 10 that works like the rotational spring. In addition, the cylindrical block is held with two levers
in order to cancel the force in the y direction. It is noticed that a pair of the cylindrical block and the
lever in the right-hand side generates the friction force, only when the table moves in the positive x
region. Therefore, a second pair of the cylindrical block and the lever is installed in a mirror image
position at the left side. This pair works separately when the table moves in the negative x region.

3
3.1

Formulation of the model

Basic equation

The rotation angle of the lever q is defined as

q = q - q0
Then, the moment M by the rotational spring of the spring constant k is expressed as
M = kq
The equation of motion of the lever is given by
d 2q
J 2 = sP + eF - M
dt

(1)
(2)
(3)

1320

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

where J is the moment of inertia of the lever about the pivot OL. It is noticed that M=0 when the
cylindrical block is located at x=0. If Coulomb's friction with the friction coefficient m is assumed, the
friction force F is expressed as
F = sgn( x& ) m P
(4)
where sgn( x& ) represents a function having the value +1 if its argument x& is positive and the value -1
if x& is negative. Substituting equations (2) and (4) into equation (3), we have
d 2q
+ kq
(5)
dt 2
P=
s + sgn( x& ) m e
It is noted from Figure 1 that q increases and s decreases when the position of the cylindrical block x
increases. Accordingly, it is deduced from equation (5) that P increases together with x. In addition, the
2nd term of the denominator of equation (5) indicates that P changes with moving directions (sign of
velocity x& ), even if x is in the same position. This characteristic is similar to the self-servo effect of
the automobile drum brake [10].
J

The force that behaves, in practice, as the friction damper is the x direction component of the resultant
of the forces P and F, which is given by
Q = QF + Q P
(6)
where
QF = F cos q = sgn( x& ) m P cos q

(7)

QP = P sin q

Since Q includes the restoring force QP due to the rotational spring in addition to the essential friction
force QF, we call it quasi-friction force henceforth.

3.2

Fundamental performance

In view of the geometric relationship shown in Figure 3, we have

( R - e) cosq - h
s=
sin q
R-e

h
q = sin -1 2
- tan -1
2
l-x
h + (l - x )

(8)
(9)

Substitution of equation (9) into equation (8) indicates that s is a function of x. From equations (1), (5)
and (6), it is also deduced that the forces P and Q are functions of x. In fact, as shown by the
numerical simulations and experiments in the later chapter, forces P and Q vary roughly proportional to
the displacement x. Therefore, the fundamental performance of the proposed friction damper is

q'

A
R

B
OB D

O
x

C
q'

e
h

l-x

Figure 3: Geometric relationship.

OL

DAMPING

1321

estimated by the rate of change. Now, the representative slope CQ is defined by the differentiation of
equation (6) with respect to x at x= 0 when the inertia of the lever is ignored:
CQ =

dQ
dx

=
x =0

dQ
dq '
d q ' q '=q0 dx

(10)
x =0

where

dQ
dP
= {sgn( x& ) m cos q0 + sin q 0 }
dq q =q0
dq q =q0
sin 2 q 0
dq
=h - ( R - e) cosq 0
dx x=0
and

k sin q0
dP
=
dq q =q0 ( R - e ) cos q0 - h + sgn( x& ) m e sin q0

(11)

(12)

Experimental apparatus

The schematic and picture of the experimental setup are shown in Figures 4 and 5. The cylindrical
block 1 is mounted on the linear actuator table 12, and is arbitrarily moved in the x direction. The lever
frame 2 is consisted of the parts 2a and 2b and it can rotate around the pivot 6. The friction plate 3 is
connected to the lever frame 2 via spacers 4 and the force sensors 5 (there are four couples in total)
that measure normal force P, and the friction plate 3 is in contact with the cylindrical block 1. In order
to tune the friction coefficient m between the friction plate 3 and the cylindrical block 1, which are
made of aluminum, the abrasion-resistance cork sheet is pasted by double-sided adhesive tape on the
face of the friction plate 3. The linear guide 7 is attached in order that the normal force P is transmitted
to the lever frame 2 through only the force sensor 4, while the friction force F is transmitted directly.
The leaf springs 9 of two sets of the levers are combined together with the spring joint 11, and they
perform the role of the rotational spring. In order to measure the resultant force that act to the pivot

54

2a 8 7
10 9 2b 11

15 B

1
6

1:Cylindrical Block
2 (2a, 2b) :Lever Frame
3:Friction Plate, 4:Spacer
5:Force Sensor, 6:Pivot
7:Linear Guide, 8:Leaf Spring
9:Leaf Spring
10:Spring Holder, 11:Spring Joint
12:Linear Actuator Table
13:Pivot Support Frame
14:Linear Guide, 15:Force Sensor

(a) Top view.

4

2a

10 9 11

2b
12
(b) Side view (section A-A).

13

15

14
(c) Side view (section B-B).

Figure 4: Schematic of experimental apparatus.

1322

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

9 6

11

15

7
2a
1
3

Figure 5: Picture of experimental apparatus.

item
R [mm]
e [mm]
l [mm]
h [mm]
q 0 [deg]
k [Nmm/rad]
J [kgmm2]

model A
56.0
45.6
185.5
0
3.2
7.40104
6.45103

model B
21.1
0
193.9
0
6.2
6.41104
6.86103

Table 1: System parameters.

Figure 6: Friction coefficient m.

support frame 13 in x direction, the force sensor 15 is attached. This force corresponds to the
quasi-friction force Q.
Table 1 shows system parameters of the models used in subsequent experiments. It is noticed that the
frictional interface offset e is large in the case of model A, and zero in the case of model B.

5
5.1

Numerical simulation and experimental results

Friction coefficient

Figure 6 shows the experimental result of the friction coefficient m between the abrasion-resistance cork
sheet and the aluminum. Since the friction coefficient varies depending on the relative velocity v, the

DAMPING

1323

(a) Cylindrical block displacement x.

(a) Normal force P.

(b) Normal force P.

(b) Quasi-friction force Q.

Figure 8: Variation of forces versus displacement x
(model A).

(c) Quasi-friction force Q.

Figure 7: Time history results of model A.

empirical formulae are derived as:

in the velocity region 0v<200
m1 = -9.235 10 -6 v 2 + 4.422 10-3 v + 0.3491
in the velocity region 200v<750
m2 = -7.086 10-7 v 2 + 1.058 10-3 v + 0.6809
in the velocity region 750v
m 3 = 1.076

(13)
(14)
(15)

These empirical formulae are used in subsequent numerical simulations.

5.2

Variation of forces and evaluation of energy dissipation

Figure 7 shows the experimental and numerical simulation results of the time history of the normal
force P and the quasi-friction force Q for the case of the model A, in which the cylindrical block is
moved back and forth in the displacement range, x=0-65 mm, with constant velocity of about v=112
mm/s. The normal force P and the quasi-friction force Q versus the displacement x are shown in Figure
8, which are modified from the results of Figure 7. The arrowed line in Figure 8 expresses the direction
of change of x. The forces P and Q increase gradually when the displacement x increases. At a moment
that x starts to decrease, P increases abruptly and Q inverts its sign, then P and Q asymptotically
approach zero together with decreasing x. It is noticed that magnitudes of P and Q in the case of v<0

1324

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

(a) Normal force P.

(b) Quasi-friction force Q.

Figure 9: Variation of forces versus displacement x
(model B).
are greater than ones in the case of v>0, even if x is in the same position. This characteristic is similar
to the self-servo effect of the automobile drum brake [10] as mentioned in section 3.1. Moreover,
vibration occurs due to abrupt change of Q immediately after x turns around. Through additional
experiments, it is verified that magnitude of the vibration becomes small when the velocity v is small or
the displacement x is harmonic.
Furthermore, the dotted straight line shown in Figure 8(b) indicates the linear equation with
representative slope CQ defined by equation (10), which is given by
Q L = CQ x
(16)
In the range of small displacement x, it is considered that the relationship between the quasi-friction
force Q and the displacement x can be approximated by this straight line.
Figure 9 shows variations of the normal force P and the quasi-friction force Q versus the displacement
x for the case of model B when the cylindrical block is moved in the same manner as the case of
model A. In this model, abrupt increase of P at a moment that x turns around does not occur, and
magnitudes of P are the same in both cases v>0 and v<0. This is because the frictional interface offset
e is zero and the self-servo effect does not occur. In addition, it is noticed that magnitude of Q in the
case of v<0 is smaller than that in the case of v>0. This feature is contrary to the case of model A
shown in Figure 8(b).
In addition, the area inside the hysteresis loop of the quasi-friction force Q shown in Figure 8(b) or
9(b) is equal to the energy dissipated by this force during back and forth motion [11]. Thus, the
performance of the damper is attributed to this area. It is noted that the area becomes so large that Q is
large in the case v>0 and Q is small in the case v<0. Accordingly, the performance of the damper is
evaluated by the representative slope CQ. In the following section, the characteristic of CQ is
investigated in detail.

DAMPING

5.3

1325

Evaluation of the damper performance by the non-dimensional

representative slope

In order to evaluate in a general way the performance of the friction damper proposed in this paper,
non-dimensional parameters are introduced:
x
e
h
l
x= , e= , h= , l =
(17)
R
R
R
R
P
Q
P= , Q=
(18)
PR
PR
where
M0
, M0 = k
(19)
l
It is noted that M0 is the moment to rotate the lever by one radian, and PR is the force applied at the
position of distance l from the pivot OL to balance the moment. Taking into account equations (17)-(19)
and rewriting equations (10)-(12), the non-dimensional representative slope C Q is given as follows:
PR =

CQ =

dQ
dx

q ' =q 0

dq '
= dQ
d q ' q '=q 0 d x

(20)

x =0

where
dQ
dP
= {sgn( x& ) m cos q 0 + sin q 0 }
d q q =q
d q q =q
0

dq
dx

x =0

sin 2 q 0
=h - (1 - e ) cos q 0

(21)

and
l sin q 0
dP
=
d q q =q 0 (1 - e ) cos q 0 - h + sgn( x& ) m e sin q 0

In addition, since q'=q0 at x=0, equation (9) is rewritten as

1 - e - h cos q 0
l =
sin q 0

(22)

(23)

From equations (20)-(23), it is apparent that C Q varies depending on the parameters e , h , q0 and m.
Table 2 shows non-dimensional parameters of the models A and B. Henceforth, effects of the
parameters on C Q are shown. Figure 10 shows the variation of C Q of the models A and B when the
and
indicate the case of the mean value
friction coefficient m is varied. In the figure, symbols
m=0.73 in the experiments of the preceding paragraph. In a positive velocity case, C Q is always positive
and increases in proportion to m. In the case of a negative velocity, C Q decreases in proportion to m. It
is also noticed in a negative velocity case that C Q is positive in the small range of m, and it needs to
enlarge m above to some extent in order to make C Q <0. Figure 11 shows the variation of C Q versus q0
with various e for the case of the friction coefficient m=0.73. Symbols
and
show the case

item
e

model A
0.8143
0

model B
0
0

Table 2: Non-dimensional parameters.

1326

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

(a) Positive velocity, x& >0.

(a) Positive velocity, x& >0.

(b) Negative velocity, x& <0.

(b) Negative velocity, x& <0.

Figure 10: Variation of C Q versus m.

Figure 11: Variation of C Q versus q 0,

m=0.73.

corresponding to models A and B, respectively. In a positive velocity case, C Q is always positive and
increases in proportion to q0 and e . On the other hand, in the case of the negative velocity, C Q is
negative and decreases together with q0 in the small range of q0, but it turns to increase at the vicinity
of q0=25 degree. In addition, the curves of e >0.2 have broken off. This is because the cylindrical block
contacts to pivot OL due to a geometric relationship, and q0 cannot be increased any more. Furthermore,
in the case of a negative velocity, C Q is zero at q0=36.1 degree irrespective of the value of e , and C Q
increases together with q0 in the range more than 36.1 degree. Inserting the condition C Q=0 into
equations (20) and (21), we have
sgn( x& ) m cos q0 + sin q 0 = 0, x& < 0
(24)
Rewriting q0 to q0B to denote that equation (24) has been satisfied, equation is modified as
q0 B = tan -1 m
(25)
The value q0B=36.1 degree corresponds to the case m=0.73 in Equation (25). It is noted that the 1st and
the 2nd terms of equation (24) correspond to QF and QP in equation (7), respectively. Thus, because the
normal force component is larger than the essential friction force component in the range of q0>q0B, the
performance as a damper may be impaired.

Conclusion

A new type of linear friction damper consisting of the cylindrical block and the inclined lever to make
the magnitude of the friction force proportional to the relative displacement has been proposed, and the
characteristics are investigated by numerical simulations and experiments. The following conclusions
are derived:

DAMPING

1327

(1) It is verified that the friction force varies proportional to the displacement of the cylindrical block.
(2) Rate of variation of the friction force versus the displacement of the cylindrical block becomes
larger together with the frictional interface offset e. This is caused by the self-servo effect.
(3) Rate of variation of the friction force also becomes larger together with the fundamental angle of
inclination q0 of less than a specific value q0B. However, the performance as a damper is impaired
when q0 is set excessively large, because the normal force applied by the rotational spring becomes
larger than the essential friction force.
(4) The problem of vibration occurring immediately after the cylindrical block turns around, needs to
be examined in detail in applying the proposed damper to the vibration isolation system.

Acknowledgements
This work was supported by Japan Society for the Promotion of Science, Grant-in-Aid for Scientific
Research (C), No. 21560255.

References
[1] J.P. Den Hartog, Forced Vibrations with Combined Coulomb and Viscous Friction, Transaction of
the ASME Journal of Applied Mechanics, Vol. 53, No. 9 (1931), pp. 107-115.
[2] A. Schlesinger, Vibration Isolation in the Presence of Coulomb Friction, Journal of Sound and
Vibration, Vol. 63, No. 2, Academic Press (1979), pp. 213-224.
[3] K.E. Beucke, J.M. Kelly, Equivalent Linearizations for Practical Hysteretic Systems, International
Journal of Non-Linear Mechanics, Vol. 23, No. 4, Pergamon Press (1985), pp. 211-238.
[4] N. Makris, M.C. Constantinou, Analysis of Motion Resisted by Friction. I. Constant Coulomb and
Linear/Coulomb Friction, Mechanics of Structure and Machines, Vol. 19, No. 4, Marcel Dekker
(1991), pp. 477-500.
[5] N. Makris, M.C. Constantinou, Analysis of Motion Resisted by Friction. II. Velocity-Dependent
Friction, Mechanics of Structures and Machines, Vol. 19, No. 4, Marcel Dekker (1991), pp.
501-526.
[6] I. Tadjbakhsh, B.C. Lin, Displacement-Proportional Friction (DPF) in Base Isolation, Earthquake
Engineering and Structural Dynamics, Vol. 15, John Wiley & Sons (1987), pp. 799-813.
[7] H. Yamaguchi, Y. Satoh, H. Ookubo, Vibration Isolation and Reduction by a Linear Friction
Damper, in Proceedings of the 7th International Conference on Motion and Vibration Control, St.
Louis, USA, 2004 August 8-11, CD-ROM 32.
[8] H. Yamaguchi, H. Yoshida, Linear Friction Damper Consisted of Cylindrical Friction Block and
Inclined Lever, in Proceedings of the Dynamics and Design Conference 2009 (in Japanese),
Sapporo, Japan, 2009 August 3-8, CD-ROM 460.
[9] H. Yamaguchi, H. Yoshida, Linear Friction Damper Consisted of Cylindrical Friction Block and
Inclined Lever (Investigation of Sprag-Slip), in Proceedings of the 16th Kanto Branch Regular
Meeting of the Japan Society of Mechanical Engineers (in Japanese), Tokyo, Japan, 2010 March
10-11, pp. 31-32.
[10] T. Ozaki, Automotive Engineering (in Japanese), Morikita Press (1979), p. 136.
[11] R.S.H. Richardson, H. Nole, Energy Dissipation in Rotary Structural Joints, Journal of Sound and
Vibration, Vol. 54, No. 4, Academic Press (1977), pp. 577-588.

1328

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010