Original Article

Proc IMechE Part I:

J Systems and Control Engineering
1–9
Effect of friction model on simulation Ó IMechE 2014
Reprints and permissions:
of hydraulic actuator sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0959651814539476
pii.sagepub.com

XB Tran1, WH Khaing2, H Endo2 and H Yanada2

Abstract
This article examines the effect of friction model on the simulation accuracy of hydraulic cylinders by using three friction
models: a steady-state friction model, the LuGre model, and the new modified LuGre model. Hydraulic cylinder’s beha-
viors are measured under different sinusoidal inputs to the servo valve. Simulations are conducted under the same con-
ditions as the experiments. The comparisons of simulated results with measured ones show that the new modified
LuGre model can predict accurately the hydraulic cylinder’s behaviors while the LuGre model may cause high-frequency
oscillations in velocity, friction force, and pressures, which are not observed in experiments.

Keywords
Friction, hydraulic cylinder, friction model, simulation, new modified LuGre model

Date received: 7 February 2014; accepted: 19 May 2014

Introduction Several dynamic friction models have been proposed

so far,4–9 and among them, the LuGre model5 is most
Hydraulic actuation systems are widely used in many widely utilized. The LuGre model, however, cannot
applications ranging from construction and mining to simulate well the dynamic friction behaviors of a
robotics, aerospace, and underwater manipulators hydraulic cylinder in the sliding regime as shown in
because of its high force/torque and power density. Yanada and Sekikawa.9 Yanada and Sekikawa9 have
However, the dynamic characteristics of the hydraulic made a modification to the LuGre model by incorpor-
system are relatively complicated due to its high nonli- ating lubricant film dynamics into the model, and it
nearities. If the motion of the hydraulic system can be has been shown that the proposed model, called the
accurately predicted at its design stage by simulation, modified LuGre model, can simulate the dynamic
the design of the system including the selection of the behaviors of friction observed in hydraulic cylinders
components and the design of the controller will be with a relatively good accuracy.10
able to be made appropriately, and the design process Tran et al.11 have shown that the modified LuGre
may be shortened. model is valid only in the negative resistance regime
One of the nonlinearities of the hydraulic system is and cannot simulate the hysteretic behaviors observed
friction. Friction may cause control errors, limit cycles, in hydraulic cylinders in the fluid lubrication regime. In
and poor performance of the system. It is, therefore, addition, they have revised the modified LuGre model
necessary to find an accurate mathematical model of
by replacing the usual fluid friction term with a first-
friction to predict accurately the motions of the hydrau-
order lead dynamics and have shown the usefulness of
lic system.
Mathematical models to describe the steady-state
1
friction characteristics have been proposed1–3 and are Department of Fluid Power & Automation Engineering, School of
Transportation Engineering, Hanoi University of Science and Technology,
widely used in mechanical systems including a hydrau-
Hanoi, Vietnam
lic system. Those models are very useful when the 2
Department of Mechanical Engineering, Toyohashi University of
motions of a hydraulic system are predicted or analyzed Technology, Toyohashi, Japan
under steady-state operating conditions. However, the
steady-state friction models are not enough or useless Corresponding author:
XB Tran, Department of Fluid Power & Automation Engineering, School
to predict or control the motion of a hydraulic system of Transportation Engineering, Hanoi University of Science and
when the system operates under oscillating velocity Technology, C6-205, No. 1, Dai Co Viet, Hai Ba Trung, Hanoi, Vietnam.
conditions. Email: [email protected]

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016

2 Proc IMechE Part I: J Systems and Control Engineering

the new modified LuGre model in the entire sliding

regime.
Although the usefulness of the LuGre model and the
new modified LuGre model have been verified, the
validity of those models in predicting the motion of a
hydraulic system has not been investigated.
In this article, the effect of three friction models, that
is, a steady-state friction model (static + Coulomb +
viscous friction), the LuGre model, and the new modi-
fied LuGre model on the simulation accuracy of two
hydraulic cylinders is examined. Hydraulic cylinder’s
behaviors such as piston velocity and pressures are mea-
sured under various operating conditions of sinusoidal
input to an electrohydraulic servo valve. Hydraulic
cylinder’s behaviors are simulated using MATLAB/
Figure 1. Steady-state friction model.
Simulink by incorporating one of the three friction
models with identified parameters into the entire system
model. The simulated behaviors are compared with
measured ones, and how the simulated behaviors are
affected by friction model and which model is the best
are discussed.
The rest of this article is organized as follows.
Section ‘‘Friction models’’ describes in brief the steady-
state friction model, the LuGre model, and the new
modified LuGre model. Section ‘‘Electrohydraulic
servo system and its mathematical model’’ describes the
electrohydraulic servo system and its mathematical
model. The experimental and simulation results are
presented and discussed in section ‘‘Results and discus-
Figure 2. Bristle model.
sion.’’ Finally, the main conclusions are drawn in sec-
tion ‘‘Conclusion.’’
LuGre model
Friction models De Wit et al.5 have proposed the LuGre model that
In this section, short descriptions of a steady-state fric- combined the stiction behavior with an arbitrary
tion model, the LuGre model, and the new modified steady-state friction characteristic. The LuGre model is
LuGre model are presented. based on the bristle model shown in Figure 2.
Contacting asperities on the surfaces are modeled as
rigid bristles on one surface and elastic ones on another
Steady-state friction model surface. The LuGre model is given by
The steady-state friction models that are the combina- dz s0 z
tion of Coulomb friction, viscous friction, and static =v  v ð2Þ
dt gs (v)
friction have been proposed and are summarized in the
literature1,2 and are most commonly used in engineer- dz
Fr = s0 z + s1 + s2 v ð3Þ
ing fields. The friction force is given by a function of dt
velocity as follows where z is the mean deflection of the elastic bristles, s0
n is the stiffness of the elastic bristles, s1 is the micro-
Fr = Fc + (Fs  Fc )e(v=vs ) + s2 v ð1Þ viscous friction coefficient, and gs(v) is a Stribeck func-
where Fr is the friction force, Fc is the Coulomb friction tion given by
and is independent of the magnitude of the velocity, Fs n
gs (v) = Fc + (Fs  Fc )e(v=vs ) ð4Þ
is the static friction force that is observed immediately
before there is a slide of the contacting surfaces, vs is For steady state, friction force is given by equation
the Stribeck velocity and is related to the velocity range (1).
of the negative resistance regime, n is the exponent that
affects the slope of the Stribeck curve, s2 is the viscous
friction coefficient, and v is the velocity between the New modified LuGre model
two surfaces in contact. The characteristics of the Tran et al.11 have extended the modified LuGre model9
steady-state friction model are shown in Figure 1. for simulating the dynamic behaviors of friction of

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016

Tran et al. 3

Figure 3. Schema of experimental apparatus: (a) side view and (b) front view.

n
hydraulic cylinders in the fluid lubrication regime by Frss = Fc + ½(1  hss )Fs  Fc e(v=vs ) + s2 v ð12Þ
replacing the usual fluid friction term with a first-order
lead dynamics. The model is called the new modified The static parameters, Fs, Fc, vs, vb, n, and s2, of the
LuGre model and is described by three models were identified from the measured steady-
state friction characteristics using the least-squares
dz s0 z method, and the dynamic parameters, s0, s1, th, and
=v  v ð5Þ
dt gs (v, h) T, were identified from the measured dynamic friction
  characteristics by the methods proposed in Tran et al.11
dz dv
Fr = s0 z + s1 + s2 v + T ð6Þ
dt dt
where T is the time constant for fluid friction dynamics; Electrohydraulic servo system and its
gs(v, h) is a Stribeck function that expresses the mathematical model
Coulomb friction and the Stribeck effect and is In this section, a test setup of the electrohydraulic servo
obtained by incorporating a dimensionless lubricant system for measuring the hydraulic cylinder’s behaviors
film thickness, h, into the Stribeck function gs(v) of the is first described and then the mathematical model of
LuGre model in equation (4) as follows the system is derived.
n
gs (v, h) = Fc + ½(1  h)Fs  Fc e(v=vs ) ð7Þ
The lubricant film dynamics can be given by
Electrohydraulic servo system
Figure 3 shows the test setup used in this investigation.
dh 1 In this test setup, a single-rod hydraulic cylinder (2) was
= (hss  h) ð8Þ
dt th fixed vertically on a frame (1) made of U-shape bars
8
< t hp (v 6¼ 0, h 4 hss ) and its piston was connected to the load mass (5) made
t h = t hn (v 6¼ 0, h . hss ) ð9Þ of steel circular plates through a rectangular steel plate
:
t h0 (v = 0) (4). The motion of the hydraulic piston was controlled
( by a servo valve (9). Two pressure sensors (8) with an
Kf jvj2=3 (jvj 4 jvb j) accuracy of 0.5% R.O. were used to measure the pres-
hss = ð10Þ
Kf jvb j2=3 (v . vb ) sures, P1 and P2, in the cylinder chambers; the piston
  velocity, v, was measured using a tacho-generator (6)
Fc
Kf = 1  jvb j2=3 ð11Þ with a ripple of less than 2% by converting linear
Fs
motion of the piston to rotational motion through a
where hss is the dimensionless steady-state lubricant film ball screw (3) and a belt (7). The ball screw and the
thickness parameter; Kf is the proportional constant for tacho-generator were mounted on the frame (1) as
lubricant film thickness; vb is the velocity within which shown in Figure 3(b). Signals from the sensors were
the lubricant film thickness is varied; and t hp, t hn, and read into a computer (10) through a 12-bit analogue-to-
th0 are the time constants for acceleration, deceleration, digital (A/D) converter and a signal from the computer
and dwell periods, respectively. For steady state, fric- was supplied to the servo valve through a 12-bit digital-
tion force is given by to-analogue (D/A) converter.

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016

4 Proc IMechE Part I: J Systems and Control Engineering

Table 1. Specifications of hydraulic cylinders used.

Type Bore diameter Rod diameter Stroke Packing material

1 0.032 m 0.018 m 0.2 m Nitrile rubber

2 Hydrogenated nitrile rubber

Measured data, that is, velocity, v, and pressures, P1 V1 _

Q1  A1 x_ = P1 ð15Þ
and P2, were recorded at the interval of 0.5 ms (2 kHz). bh
An acausal low-pass filter with a bandwidth of 150 Hz V2 _
was used to reduce the measurement noise. The accel- A2 x_  Q2 = P2 ð16Þ
bh
eration, a, of the piston was calculated by an approxi-
mate differentiation of the measured piston velocity. where bh is the effective bulk modulus of the fluid, and
The noise in the calculated acceleration signal was fil- V1 and V2 are the fluid volumes in the two cylinder
tered by an acausal low-pass filter with a bandwidth of chambers and are given by
32 Hz.
V1 = V10 + A1 x ð17Þ
The friction force, Fr, is obtained from the equation
of motion of the hydraulic piston using the measured V2 = V20 + A2 (L  x) ð18Þ
values of the pressures in the cylinder chambers, the
where L stands for the stroke length of the cylinder and
acceleration of the piston, and the weight of the load
x is the displacement of the mass. V10 and V20 are the
mass as follows
dead volumes in the two cylinder chambers,
Fr = P1 A1  P2 A2  ma  mg ð13Þ respectively.
The volumetric flow rates, Q1 and Q2, into or out of
where m is the load mass, A1 and A2 are the piston the chambers 1 and 2 of the cylinder can be written in
areas, and g is the acceleration of gravity. It is noted in the following forms
equation (13) that the friction force, Fr, exactly stands For jxv j 4 U
for the sum of the friction force generated in the qffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
hydraulic cylinder and that generated in the ball screw Q1 = cd r2w(U + xv ) P2s + jxxvv j P2s  P1
mechanism. However, the majority of the friction force qffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ð19Þ
results from the hydraulic cylinder and, therefore, in cd r2w(U  xv ) P2s + jxxvv j P1  P2s
this article, Fr is treated as the friction force in the qffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
hydraulic cylinder. Q2 = cd r2w(U + xv ) P2s + jxxvv j P2  P2s
qffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ð20Þ
Two hydraulic cylinders described in Table 1 with
cd r2w(U  xv ) P2s + jxxvv j P2s  P2
the same structure and dimensions but different pack-
ing materials were used for this investigation. The
For jxv j . U
experiments were conducted at the oil temperature in
the oil tank of 30 °C6 2 °C and at the supply pressure sffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
2 Ps xv Ps
of Ps = 5 MPa under open-loop condition. Sinusoidal Q 1 = cd w(U + xv ) +  P1 ð21Þ
inputs were supplied to the servo valve to move the r 2 jxv j 2
hydraulic piston sinusoidally. Every experiment was sffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
2 Ps xv Ps
conducted three times to ensure the repeatability. Q 2 = cd w(U + xv ) + P2  ð22Þ
r 2 jxv j 2

Mathematical model where Ps is the supply pressure to the servo valve, w is

the circumferential width of the rectangular port cut into
The relationship between the displacement of the valve the valve sleeve, U is the valve underlap, cd is the dis-
spool, xv, and the servo current (control signal), u, can charge coefficient of the valve orifice, and r is the density
be approximated by a second-order model as follows of the hydraulic fluid. The analysis of hydraulic servo
kv v2v systems of this type is well described in the literature.12
xv = u ð14Þ The motion of the hydraulic cylinder can be
s2 + 2zv vv s + v2v
described as follows
where kv is the valve spool position gain, vv is the valve
dv
natural angular frequency, zv is the damping ratio of m = P1 A1  P2 A2  Fr  mg ð23Þ
dt
the servo valve, and s is the Laplace variable. The rated
servo current is 30 mA. The system parameters used in the simulation are
The relationship between the volumetric flow rates given in Table 2. The values of the effective bulk
Q1 and Q2 and the pressures P1 and P2 in both cham- modulus and the discharge coefficient were identified
bers of the cylinder are described in the following forms by comparing measured velocity and pressures with

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016

Tran et al. 5

Table 2. System parameters. simulated ones. The value of the underlap was esti-
mated by the pressure gain characteristic of the servo
Parameters Value valve. In this article, simulation was done using
kv (m/A) 0.0227 MATLAB/Simulink. The sinusoidal current supplied
vv (rad/s) 440 to the servo valve was used as the input to the entire
zv 0.75 system model.
bh (Pa) 1 3 108
V10 (m3) 5 3 1026
V20 (m3) 5 3 1026
U (m) 4 3 1025
Results and discussion
cd 0.32 In this section, experimental results of the two hydrau-
w (m) 0.0228
r (kg/m3) 862
lic cylinders and simulation results using the three fric-
tion models are presented and discussed.

Experiment
Figure 4 shows the steady-state friction characteristics
of the two hydraulic cylinders measured at the load
mass of m = 18 kg. Positive velocity corresponds to
the extending stroke of the piston and negative one to
the retracting stroke. It can be seen from Figure 4
that the friction forces of both Type 1 and 2 are larger
in the extending stroke than in the retracting one. It
seems that such an asymmetrical nature of the friction
results from the asymmetrical structures of the piston
and rod seals as well as the hydraulic piston used. It
can also be seen from Figure 4 that the steady-state
friction characteristics are different between the two
Figure 4. Steady-state friction characteristics measured for
hydraulic cylinders. The friction force of Type 2 cylin-
Type 1 and Type 2 at load mass m = 18 kg.
der is larger than that of Type 1 cylinder and Type 2
cylinder does not enter the fluid lubrication regime
within the experimental conditions. These differences
may be caused by the difference in the viscoelastic
nature of the packing material between the two types
of cylinder.11
Figure 5 shows an example of the measured dynamic
characteristics of Type 1 cylinder under the sinusoidal
current input of the amplitude juj = 4:5 mA and the fre-
quency f = 1 Hz at the load mass of 118 kg. The piston
velocity is varied between 20.14 and + 0.14 m/s as
shown in Figure 5(a). Figure 5(b) shows the friction
force variation and shows that the maximum friction
force (break-away force) can be seen instantaneously
after starting from the rest. After the friction force has
attained its maximum, it decreases continuously in the
following period of velocity variation in the first half
cycle. After that, the sign of the friction force is reversed
almost at the same time as the velocity reversal. After
the first cycle of the velocity variation, almost the same
friction behavior is repeated. Regarding the pressure
variation in Figure 5(c), it can be noticed that the pres-
sure P1 is always greater than the pressure P2 because
of the large load mass.
Based on the measured steady-state and dynamic
friction characteristics and using the identification
methods proposed in Tran et al.,11 the parameters of
Figure 5. Dynamic characteristics measured at juj = 4:5 mA, the three models were identified and are shown for
f = 1 Hz, and m = 118 kg of Type 1 cylinder: (a) velocity, (b) Type 1 cylinder at 118 kg and Type 2 cylinder at 18 kg
friction force, and (c) pressure. in Table 3.

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016

6 Proc IMechE Part I: J Systems and Control Engineering

Table 3. Values of the parameters of three friction models for

Type 1 and Type 2.

Parameters Type 1 (118 kg) Type 2 (18 kg)

v.0 v\0 v.0 v\0

Fs (N) 830 2500 1300 21100

Fc (N) 210 280 400 2350
vs (m/s) 0.0125 20.01 0.08 20.06
vb (m/s) 0.7 20.9 0.2 21.5
n 0.5 0.5 1 0.6
s2 (N s/m) 330 350 80 80
T (s) 0.33 0.07 0.62 0.8
s0 (N/m) 5 3 106 5 3 105
s1 (N s/m) 0.1 0.1
t hp (s) 0.25 0.12
t hn (s) 1.5 0.5
t h0 (s) 40 45

Figure 7. Comparison between measured and simulated

results using the LuGre model at juj = 2 mA, f = 0.5 Hz, and
m = 118 kg for Type 1 cylinder: (a) velocity, (b) friction force, (c)
pressure p1, and (d) pressure p2.

new modified LuGre model at the conditions of

juj = 2 mA, f = 0.5 Hz, and m = 118 kg for Type 1
cylinder. As can be seen from Figure 6, the simulated
results of the velocity and the friction force are in
good overall agreement with the measured ones. The
pressures are not in so good agreement between the
simulation and experiment, especially in the negative
velocity range, but the overall tendency is similar
between both.
Figure 7 shows the comparisons between the
dynamic characteristics measured and the ones simu-
Figure 6. Comparison between measured and simulated lated by the LuGre model at juj = 2 mA, f = 0.5 Hz,
results using the new modified LuGre model at juj = 2 mA, and m = 118 kg for Type 1 cylinder. It is shown in
f = 0.5 Hz, and m = 118 kg for Type 1 cylinder: (a) velocity, (b) Figure 7(b) that the friction force cannot be simulated
friction force, (c) pressure p1, and (d) pressure p2.
precisely by the LuGre model as has already been
shown in Yanada and Sekikawa.9 In addition, oscilla-
Simulation tions are observed in the velocity. Such velocity oscilla-
Figure 6 shows the comparisons between the dynamic tions cause the oscillations of friction force as well as
characteristics measured and the ones simulated by the pressures.

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016

Tran et al. 7

Figure 9. Comparison between measured and simulated

Figure 8. Comparison between measured and simulated results using the new modified LuGre model at juj = 4 mA, f = 1
results using the steady-state friction model at juj = 2 mA, f = 0.5 Hz, and m = 18 kg for Type 2 cylinder: (a) velocity, (b) friction
Hz, and m = 118 kg for Type 1 cylinder: (a) velocity, (b) friction force, (c) pressure p1, and (d) pressure p2.
force, (c) pressure p1, and (d) pressure p2.

Figure 11 that the steady-state friction model is not

Figure 8 shows the comparisons between the appropriate for the simulation of the dynamic behaviors.
dynamic characteristics measured and the ones simu-
lated by the steady-state friction model at juj = 2 mA,
f = 0.5 Hz, and m = 118 kg for Type 1 cylinder. For Conclusion
this model, the variation in the friction force at velocity
reversal becomes larger than that of the LuGre model, In this article, both experiments and simulations are
and because of this, much larger oscillations are seen in conducted to investigate the effect of friction model on
the velocity, friction force, and pressures. the simulation accuracy of hydraulic actuators. The
Figures 9–11 show the comparisons between the results show that the new modified LuGre model can
dynamic characteristics measured and the ones simu- predict the hydraulic actuator’s behaviors with a good
lated by the new modified LuGre model, the LuGre accuracy and that the LuGre model can give relatively
model, and the steady-state model, respectively, for good results except for some oscillatory behaviors. It is
Type 2 cylinder at juj = 4 mA, f = 1 Hz, and m = 18 confirmed that the steady-state model is not appropri-
kg. As shown in Figure 9, oscillatory behaviors appear ate. The new modified LuGre model includes more
also in the new modified LuGre model at start, but the parameters than the LuGre model and it is not so easy
overall behaviors are well simulated by the model. to identify them. A certain simplification of the model
The behaviors of the velocity simulated by the is the subject for a future study. In addition, the appli-
LuGre model (Figure 10) are almost the same as cations of the new modified LuGre model to the simu-
those shown in Figure 9, but there are some differences lations of pneumatic actuators are also the subject for a
in the friction force and pressures. It is confirmed from future study.

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016

8 Proc IMechE Part I: J Systems and Control Engineering

Figure 11. Comparison between measured and simulated

results using the steady-state friction model at juj = 4 mA, f = 1
Hz, and m = 18 kg for Type 2 cylinder: (a) velocity, (b) friction
Figure 10. Comparison between measured and simulated force, (c) pressure p1, and (d) pressure p2.
results using the LuGre model at juj = 4 mA, f = 1 Hz, and m = 18
kg for Type 2 cylinder: (a) velocity, (b) friction force, (c)
pressure p1, and (d) pressure p2.
3. Hibi A and Ichikawa T. Mathematical model of the tor-
que characteristics for hydraulic motors. Bull JSME
1977; 20(143): 616–621.
Declaration of conflicting interests 4. Haessig DA and Friedland B. On the modeling of fric-
The authors declare that there is no conflict of interest. tion and simulation. J Dyn Syst Meas Contr 1991;
113(3): 354–362.
5. De Wit CC, Olsson H, Åström KJ, et al. A new model
Funding for control of systems with friction. IEEE Trans Automat
Contr 1995; 40(3): 419–425.
This research received no specific grant from any fund- 6. Dupont PE and Dunlop EP. Friction modeling and PD
ing agency in the public, commercial, or not-for-profit compensation at very low velocities. J Dyn Syst Meas
sectors. Contr 1995; 117(1): 8–14.
7. Swevers J, Al-Bencer F, Ganseman CG, et al. An inte-
References grated friction model structure with improved presliding
behavior for accurate friction compensation. IEEE Trans
1. Armstrong HB. Control of machines with friction. Boston, Automat Contr 2000; 45(4): 675–686.
MA: Springer, 1991. 8. Dupont P, Hayward V, Armstrong B, et al. Single state
2. Armstrong HB, Dupont P and Canudas DWC. A survey elastoplastic friction models. IEEE Trans Automat Contr
of models, analysis tools and compensation methods for 2002; 47(5): 787–792.
the control of machines with friction. Automatica 1994; 9. Yanada H and Sekikawa Y. Modeling of dynamic beha-
30(7): 1083–1138. viors of friction. Mechatronics 2008; 18(7): 330–339.

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016

Tran et al. 9

10. Yanada H, Takahashi K and Matsui A. Identification of Ps supply pressure

dynamic parameters of modified LuGre model and appli- Qi flow rate (i = 1, 2)
cation to hydraulic actuator. Trans Jpn Fluid Power Syst T time constant for fluid friction dynamics
Soc 2009; 40(4): 57–64. u control input (servo current)
11. Tran XB, Hafizah N and Yanada H. Modeling of juj amplitude of u
dynamic friction behaviors of hydraulic cylinders. Mecha- U valve underlap
tronics 2012; 22(1): 65–75. v velocity
12. Merrit HE. Hydraulic control systems. New York: Wiley, vb upper (for v . 0) or lower (for v \ 0)
1967.
limit of velocity in film thickness variation
vs Stribeck velocity
Appendix 1 Vi total fluid volumes in cylinder chambers
Notation (i = 1, 2)
Vi0 dead volumes in cylinder chambers
a piston acceleration (i = 1, 2)
Ai piston area (i = 1, 2) w circumferential width of rectangular port
cd discharge coefficient cut into valve sleeve
f frequency of sinusoidal input x piston position
Fc Coulomb friction force xv valve spool position
Fr friction force z mean deflection of bristles
Frss steady-state friction force bh effective bulk modulus of fluid
Fs maximum static friction force zv damping ratio of servo valve
g acceleration of gravity r oil density
gs Stribeck function s0 stiffness of bristles
h dimensionless unsteady-state lubricant s1 micro-viscous friction coefficient for
film thickness bristles
hss dimensionless steady-state lubricant film s2 viscous friction coefficient
thickness th time constant for lubricant film dynamics
kv valve spool position gain thp time constant for acceleration period
Kf proportional constant for lubricant film thn time constant for deceleration period
thickness th0 time constant for dwell period
L cylinder stroke length vv natural angular frequency of servo valve
m load mass
n exponent for Stribeck curve
Pi pressure (i = 1, 2)

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016