Mechatronics 21 (2011) 1156–1162

Contents lists available at SciVerse ScienceDirect

Mechatronics
journal homepage: www.elsevier.com/locate/mechatronics

Novel clamping force control for electric parking brake systems

Young Ok Lee a, Minseok Jang a, Wongoo Lee a, Choong Woo Lee a, Chung Choo Chung b,⇑,
Young Seop Son a,c
a
Department of Electrical Engineering, Hanyang University, Seoul 133-791, Republic of Korea
b
Division of Electrical and Biomedical Engineering, Hanyang University, Seoul 133-791, Republic of Korea
c
Central R&D Center, MANDO Corporation, Giheung-gu, Gyeonggi-Do 446-901, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, we present a novel control method with clamping force estimation for an electric parking
Received 5 October 2010 brake system. This simple control structure can be implemented at low cost as it does not require a
Accepted 5 July 2011 clamping force sensor. The characteristic curve is conventionally used to estimate the clamping force
Available online 10 September 2011
through the angular displacement of the DC motor; however, this can result in error because of the
dependence of the curve on the brake clearance between the brake pads and brake disk at release. We
Keywords: solve this problem by approximating the initial contact point using the angular velocity of the motor.
Electric-parking brake
We then propose a novel on–off control method to avoid excessive clamping forces by predicting the
Characteristic curve
Clamping force estimation
additional angular displacement after power-off caused by the inertia effect. Finally, we experimentally
Inertia effect validate our proposed control method.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction tracking observer provides accurate measurement of both absolute

Brake-by-wire systems have manufacturing advantages and are
therefore installed in many vehicles [1]. An electric parking brake
(EPB) system is a type of electromechanical brake-by-wire system
that replaces the conventional lever parking system by generating
a clamping force for parking using electric motor torque [2]. At the
push of a button, a driver can easily apply or release the parking
brake; this enables elderly or disabled persons to easily apply a full
braking load on highly inclined surfaces [3]. The EPB system oper-
ates quickly and over a wide force range through the use of electri-
cal components. It supports dynamic braking, anti-lock braking,
key-out application, hill hold, and drive-away release, in addition
to its conventional parking brake function [1,4,5].
The general control architecture used for the brake-by-wire sys-
tem is a force control loop with a force sensor that is used to measure
the clamping force [6–8]. Recently, due to cost issues and implemen-
tation difficulties, research has been conducted to determine if the
force sensor can be eliminated from the system and if the clamping
force can be estimated based on the sensory outputs of an alterna-
tive system. These outputs include the angular displacement of
the motor [9–12] or the combination of motor voltage, current,
and speed measurements [13]. The motor position can be obtained
from a resolver or an encoder. A resolver integrated with an angle-
⇑ Corresponding author. Tel.: +82 2 2220 1724; fax: +82 2 2291 5307.
E-mail addresses: yo.lee929@gmail.com (Y.O. Lee), coowolf0312@naver.com
(M. Jang), hillfolk219@gmail.com (W. Lee), chungwoo.lee@gmail.com (C.W. Lee),
cchung@hanyang.ac.kr (C.C. Chung), ysson@mando.com (Y.S. Son).
position and speed. Hoseinnezhad developed a robust estimator
module that is stable in applications with high speed and accelera-
tion [14]. A real-time adaptive estimator was investigated for the
calibration of resolver parameters dependent on age and tempera-
ture [15]. Using these estimators, the resolver has been used to
determine the clamp forces in brake-by-wire systems [8,13,14]. In
contrast, force estimation based on the output of the incremental en-
coder has been studied only with a high resolution encoder, which
provides over 8000 counts per revolution [8,12]. In this paper, we
show that a low resolution encoder can be used for force estimation.
In a previous study, a nonlinear control method was applied to the
EPB system and it showed improved robustness and tracking perfor-
mance compared to those of linear and on–off controllers [16]. How-
ever, this method requires an additional electric circuit to drive a DC
motor. In this paper, we use an on–off control method for the force
control logic to supply the maximum voltage until the clamping
force reaches the desired final force. Using this method, the DC mo-
tor continues to rotate after the power is cut-off due to its momen-
tum, resulting in an excessive clamping force. For functions that
need fast or repeated apply-release operations such as anti-lock
brake systems (ABS) and drive-away release, the excessive clamping
force may cause a longer release time. Thus, the excessive force
caused by the inertia effect should be compensated for. By compen-
sating for the inertia effect through the novel on–off control method,
our control logic can be implemented using a simple control unit
without a PWM driver.
This paper focuses on a low cost and simple mechanism for both
force estimation and control of an EPB system. Our proposed

0957-4158/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechatronics.2011.07.006
 Y.O. Lee et al. / Mechatronics 21 (2011) 1156–1162 1157

Fig. 1. Clamping force control architecture for an EPB system.

Motor
Controller Driver
Control
command Gear
Angular Signal Box
displacement
Encoder
Gear
Force logging Current Motor Box
Screw
Joint Brake disk
system Sensor
Brake pad
Load Force
cell sensor
Parking cable
Monitoring
clearance
PC

Fig. 2. Structure of the EPB system.

alternative control architecture for an EPB system is shown in
Fig. 1. The omission of a force sensor leads to the problem that
the initial contact point (where the clamping force starts increas-
ing) between the brake pads and the brake disk cannot be sensed.
Measuring the angular displacement is not sufficient to complete
clamping force estimation without knowledge of the brake pads’
clearance. In [9], both the current and angular displacements of
the motor were used to detect the initial contact point. We propose
an initial contact point detection method using only the angular
velocity of the DC motor. We demonstrate through theoretical
analysis that contact occurs when the angular velocity of the DC
motor nears its maximum value. The clamping force, f(heff), can
then be estimated as a function of the effective angular displace-
ment, heff, from the initial contact point, h(tmax). The characteristic
curve that represents the relationship between force and angular
displacement has been used to estimate the clamping force for
electric mechanical brake (EMB) systems [9,10]. We approximate
the characteristic curve as a second-order polynomial. Finally, to
address the limitation of our low-cost on–off control, we propose
a novel on–off control method that considers the inertia effect
[17]. We derive the additional angular values, h1, of the motor
after power-off for each desired final force. The target force, ftar,
is re-calculated independently from the desired final clamping
force, ffin, based on these additional angular values. The clamping
force error can be minimized by deactivating the power supply
to the motor in consideration of these additional angular values.
We experimentally validate the proposed novel on–off control
method; our experimental results illustrate that the proposed
method meets control specifications and results in a smaller
clamping force error than that of the simple on–off control method.
This paper is organized as follows: in Section 2, we introduce
the structure and specifications of the EPB system, as well as the
detection method used to determine the initial contact point and
the estimation method used to calculate the clamping force. In Sec-
tion 3, we analyze the inertia effect to decrease the error between
the target clamping force and the desired final clamping force. The
experiments performed to validate the proposed control method
are described in Section 4. In Section 5, we present our conclusions.
2. Clamping force estimation
2.1. Structure and specifications of the EPB system
In this paper, we study the stretch type EPB system depicted in
Fig. 2. This system includes a controller, parking cables, a DC mo-
tor, a gear box, an assembly of a screw and nuts, a motor driver,
a force sensor, a magnetic encoder, and a current sensor. The park-
ing cables of the EPB system are connected to the two sets of brake
pads. There are two operating modes: brake-applying and brake-
releasing. In brake-applying mode, the clamping force is increased
by pulling the parking cables using the DC motor until the target
force is achieved [16]. In brake-releasing mode, the clamping force
Table 1
Specifications of the EPB system.
Target force 80–100 kgf
Settling time Less than 1 s
Permitted error bound ±10% of target force
1158 Y.O. Lee et al. / Mechatronics 21 (2011) 1156–1162

100
5
90 Release for long time Angular velocity
4.5
Release for mid time Motor current
80 4
Release for short time

magnitude [not to scale]

Clamping force
70 3.5
force [kgf]

60 3

50 2.5
2
40
1.5
30
1
20
0.5
10
0
The initial contact time
0 −0.5
0 20 40 60 80 100 120 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
angular displacement [rev] time [s]

Fig. 3. Characteristic curves for different clearances. Fig. 4. Clamping force, angular velocity, and current in brake-applying mode.

is decreased by releasing the parking cables. To measure the angu- is the back emf, Tm is the torque developed by the motor, and Km
lar displacement of the motor with high resolution, a current sen- is the motor torque constant. The torque relationship in the EPB sys-
sor can be used by counting the number of current ripples tem is given by
[18,19,24]. In our study, a low cost magnetic encoder that has
one pulse per revolution is used for clamping force control, and Tm ¼ Ta þ Ti þ Tf ; ð4Þ
current ripples are only used to confirm the angular displacement where Ta is the applied torque of the EPB system, Ti is the inertia
measured from the magnetic encoder. In the experimental tests, torque, and Tf is the friction torque. Thus, Ta is proportional to the
we use a hall effect force sensor to measure the absolute clamping clamping force, f(heff), which is defined by
force and then compare it with the estimated force. The hall effect
force sensor is calibrated using a load cell at one of the parking T a ¼ ct f ðheff Þ; ð5Þ
cables. Table 1 lists the specifications of the EPB system [20]. where h is the angular displacement of the DC motor and
heff(t): = h(t)  h(tmax). The angular velocity reaches its maximum
2.2. Detection of the initial contact point value at time tmax. The constant ct is determined by the screw gain
and gear ratio. Details of the modeling can be found in [2]. The
The clearance between the brake pads and brake disk varies clamping force, f(heff), is a function of the effective angular displace-
depending on the level of pad wear and the point at which the ment, heff, from the initial contact point as derived in Section 2.3. In
braking force is released. Fig. 3 shows the relationship between addition, Ti is proportional to the angular acceleration, x _ ðtÞ, as de-
the angular displacement and clamping force. There is a significant fined by (6):
change in the characteristic curve depending on the degree of ini-
T i ¼ Jt x
_ ðtÞ; ð6Þ
tial clearance between the brake pads and the brake disk. Thus,
clamping force control estimation using angular displacement re- where Jt is the total inertia constant. In (4), Tf is the sum of the vis-
sults in final force error. The initial contact point between the pads cous friction, Tv, and Coulomb friction, Tc, such that
and the disk must be known to provide high quality clamping force
Tf ¼ Tc þ Tv : ð7Þ
estimation and control using motor displacement. Schwarz et al.
[9] used the motor displacement and the motor current to detect Coulomb friction dominates the viscous friction in the EPB system
the initial contact point by evaluating the stiffness quotient. Unlike so Tc  Tv. Thus (7) can be approximated as Tf ffi Tc [9]. We can then
the method in [9], we propose a detection method that uses only rewrite Tc as a function of f(heff) with some load dependency, l, and
the angular velocity of the motor, which can be obtained with an a constant offset, a [11,23]:
encoder. To derive this algorithm, we need to study the relation-
T f ffi T c ¼ lf ðheff Þ þ a: ð8Þ
ship between the angular velocity of the motor and the initial con-
tact time. Fig. 4 shows the motor current, angular velocity, and Eq. (4) can be rewritten as
clamping force in brake-applying mode at full power. It is clear _ ðtÞ þ lf ðheff Þ þ a:
T m ¼ ct f ðheff Þ þ J t x ð9Þ
from this figure that the three signals are correlated [21]. When
the clamping force starts increasing, the angular velocity begins L can be disregarded in (1) because the response of the armature
decreasing, but the motor current begins increasing. Thus, we circuit is much faster than the mechanical response. The applied in-
can derive a relationship between the angular velocity and the put voltage, Va(t), can then be rewritten by substituting (2), (3), and
clamping force. The electrical dynamics of the DC motor can be (9) into (1) to obtain
represented by Eqs. (1)–(3) [22]:
dim ðtÞ RT m
V a ðtÞ ¼ Rim ðtÞ þ L þ v emf ðtÞ ffi þ K b xðtÞ
dim ðtÞ dt Km
V a ðtÞ ¼ Rim ðtÞ þ L þ v emf ðtÞ; ð1Þ  
dt _ ðtÞ þ lf ðheff Þ þ a
R ct f ðheff Þ þ J t x
v emf ðtÞ ¼ K b xðtÞ; ð2Þ ¼ þ K b xðtÞ:
Km
T m ¼ K m im ðtÞ; ð3Þ
The above equation can be rearranged for x(t):
where R is the resistance, L is the inductance, Kb is the back emf con-
stant of the motor, Va(t) is the applied input voltage, im(t) is the V a ðtÞ Rðct þ lÞf ðheff Þ RJ t x
_ ðtÞ þ Ra
xðtÞ ¼   : ð10Þ
motor current, x(t) is the angular velocity of the DC motor, vemf(t) Kb KbKm KbKm
 Y.O. Lee et al. / Mechatronics 21 (2011) 1156–1162 1159

Let ta be the time of the initial contact between the brake pads and 3.5
the brake disk. In (10), we assume that Va(t) is a constant with mag-
nitude Va for t > 0 s until the power is deactivated (i.e., t = tf) because 3
in this paper, the controller uses an on–off control for low-cost
implementation. Eq. (10) can then be rewritten as 2.5
8
< K m V a K bRJK m xðtÞRa ; 0 6 t 6 ta

Δθ [rev]
t 2
x_ ðtÞ ¼ ð11Þ
: K m V a K b K m xðtÞRaRðct þlÞf ðheff Þ ; ta < t < tf :
RJt
1.5
If t 6 ta, (12) is a solution of the differential equation for x(t):
  1
ðK m V a  RaÞ 1  eðK b K m =RJt Þt
xðtÞ ¼ : ð12Þ
KbKm 0.5

In (12), if Va is large enough such that KmVa  Ra > 0, x(t) will in-
0
crease at least until the initial contact time, as depicted in Fig. 5. 0 0.05 0.1 0.15 0.2 0.25
If ta < t < tf, (11) is no longer a linear differential equation. Because ta [s]
the system becomes a closed chain right after the contact happens,
the closed-form solution of the differential equation for x(t) is not Fig. 6. Comparison of Dh = h(tmax)  h(ta) at different ta.
possible. Instead, we got Fig. 6 through numerical simulation using
Dymola/Modelica with the parameters used in [20]. It is possible to
detect the time tmax at which the angular velocity of the motor is Table 2
The worst force error due to Dh at each final force.
maximized using current ripples [19] because of their high resolu-
tion. However, it is not possible to detect the time ta at which the Final force kgf Force error kgf
parking cable force begins to increase without using a force sensor. 80 2.7
Thus, in this paper, we use h(tmax) instead of h(ta) to estimate the 90 3.8
clamping force. It is essential to investigate the influence of this po- 100 4.2

sition difference on the force control, Dh = h(tmax)  h(ta). Dh is plot-

ted as a function of the initial contact time ta. Here, ta = 0 s is the
case when the initial clearance is zero. For ta > 0.2 s, Dh is nearly
zero. In other words, we see that ta ffi tmax for a large clearance. Fur-
thermore, the worst case of Dh is less than 3.3 rev. The possible
clamping force error due to the maximum Dh for each target force
is listed in Table 2. These errors are acceptable in the operating tar-
get force range based on the specifications in Table 1.
the method requires less computational effort than do model-
based estimation methods or a state observer. Unlike in Fig. 3,
the f(heff) values shown in Fig. 7 agree within a tolerable bound.
Based on the experimental results determined using both the abso-
lute force sensor and high-resolution current ripples, we obtain an
approximated characteristic function f(n):

2.3. The clamping force vs. the effective angular displacement f ðnÞ ¼ a1 n2 þ a2 n þ a3 ; n ¼ modðh; 2pÞ; ð13Þ

where n is the rotational number of the motor measured by an en-

We showed that the parking cables begin applying the clamping
coder after the maximum velocity point. To obtain the ai values of
force near the time of maximum angular velocity of the motor. In
(13), we use the weighed least square estimation method (14) [25]:
this section, we validate the claim that the clamping force can be
estimated from the angular displacement of the motor after the ^i ¼ ðUT W UÞ1 UT WY;
a 8i ¼ 1; 2; 3; ð14Þ
time of maximum angular velocity. The blue-solid lines in Fig. 7
T
^i are coefficients, U is the input matrix, W is a diagonal ma-
show the relationship between the clamping force and effective where a
angular displacement obtained experimentally. The absolute trix with the weights in the diagonal, and Y is the output matrix. We
clamping force is measured by the force sensor, and the angular use higher weighting factors for larger clamping forces. In Fig. 7, the
displacement is determined using the current ripples. This method
estimates the angular displacement and angular velocity of the
140
motor by detecting periodic oscillations of the armature current
caused by rotor slots [18,19,24]. The method has higher resolution 1st
120 2nd
than inexpensive digital position sensor like an encoder. Moreover,
3rd
100 4th
5th
force [kgf]

80 estimated force

60

40

20

0
0 20 40 60 80
angular displacement [rev]

Fig. 7. Comparison of the approximated function, f(heff), and the absolute clamping
Fig. 5. Comparison between h(ta) and h(tmax). force measured by the force sensor.
1160 Y.O. Lee et al. / Mechatronics 21 (2011) 1156–1162

6 70
additional angular displacement [rev]

75

final clamping force [kgf]

5.8 80

85

5.6 90

95

5.4 100

105 (a)
5.2 110
74 76 78 80 82 84 86 88 90 92 94 96 98 100
target force [kgf]

Fig. 8. The additional angular displacement and the final clamping force under each
target force.

clamping forces measured by the force sensor and the estimated

clamping force using (13) are plotted. The estimated clamping
forces are represented by the red dashed lines. The maximum esti-
mation error is less than 3.5 kgf. Some factors that influence the
force-angular displacement characteristic curve have been re- (b)
searched [8–10]. For any external factors that influence the stiff-
nesses of the brake pads, caliper, and parking cables, the
characteristic curve stored in the product memory may not be valid,
because of the level of softness of the brake pads from OEM or var-
iation in brake pad stiffness based on temperature. To overcome
this problem, the characteristic curve has to be adapted through
parameter estimation [8–10,12] or real-time calibration [26].

3. Analysis of the inertia effect

First, we tested a simple on–off control for the EPB system in

which the power supply is deactivated when the clamping force
(c)
reaches its target. Such a simple control allows the motor to rotate Fig. 9. (a) Experimental environment of the EPB system, (b) zoomed-in view of the
for a longer amount of time due to its momentum, even without EPB module box, (c) zoomed-in view of the parking cable.
power. Although this simple on–off control meets the specifica-
tions, a large error was produced. The EPB system supports addi-
tional functions such as dynamic braking, anti-lock braking, key- at the desired clamping force. The linearized spring constant kl is
out application, hill hold, and drive-away release. Excessive forces obtained from the clamping force function (13) as follows:
may not cause damage when the EPB operates on a flat plane for a 
@f ðheff Þ
parking function. For functions that need fast or repeated applica- kl ðhd Þ ¼ ¼ 2a1 hd þ a2 ; ð16Þ
@h h¼hd
tion-release operations such as ABS and drive-away release, how-
ever, the excessive clamping force affects the operation time. The where hd is the angular displacement at the desired clamping force.
error can be reduced if one can predict the effect of inertia on mo- Assuming that hðt 
f Þ ¼ 0, rearrangement of the result of the Laplace
tor rotation. In this section, we propose a novel prediction method transform of (15) yields
for analyzing the inertia effect, and validate it experimentally [17].   
When the power is off, the input voltage, Va(), becomes zero. a 1
HðsÞ ¼ xðt f Þ  ; ð17Þ
Eq. (10) can then be rearranged as follows: Jt s s2 þ ðK b K m =RJ t Þs þ ðct þ lÞkl =J t
K b K m xðtÞ ¼ Rðct þ lÞf ðheff Þ  RJ t x
_ ðtÞ  Ra: ð15Þ where tf is the time at which the input signal becomes zero. From
the final value theorem, the steady-state values of h are given by
In (13), we obtained an approximated characteristic function, f(heff),
as a 2nd-order polynomial. The Taylor series of f(heff) is a
 hð1Þ ¼  ; 8a < 0: ð18Þ
 ðct þ lÞkl
d f 
2
df  ðh  hd Þ2
f ðheff Þ ¼ f ðhd Þ þ  ðh  hd Þ þ 2  þ :
dh h¼hd dh  2 The additional angular displacement of the motor is h(1) after the
h¼hd
power is deactivated due to the inertia effect. Thus, we can calculate
The high-order terms are written in the form of O(kh  hdk2). We the final clamping force using (13) and (18). Fig. 8 plots the addi-
use the linearized spring model for the additional angular tional angular displacement and the final clamping forces at each
displacement at the desired clamping force. Ignoring the higher-or- target force when the power is deactivated. The left y axis is the
der terms, we obtain additional angular displacement of the DC motor and the right y axis
is the final clamping force. We constructed a look-up table (LUT) that
f ðheff Þ ¼ kl heff
maps each target force to the final clamping force. This LUT is shown
 Y.O. Lee et al. / Mechatronics 21 (2011) 1156–1162 1161

(a) (b) (c)

90 100
meas. force meas. force meas. force
80 est. force est. force 100 est. force

70 80
80
60

force [kgf]

force [kgf]
force [kgf]

60
50 60

40
40
40
30

20
20 20
10

0 0 0
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
time [s] time [s] time [s]

Fig. 10. Experimental results obtained using the simple on–off control: (a) 80 kgf, (b) 90 kgf, (c) 100 kgf.

in the Re-calculation of target force block in Fig. 1. If a high level con-

troller determines the final clamping force, ffin, then the target force, Simple on-off control
Novel on-off control
ftar, is calculated using the LUT, and the power is shut-off when the
estimated force, f(heff), is equal to or greater than ftar.

4. Experimental results
Error [kgf]

4.1. Experimental environment

In order to control the clamping force of the EPB system, we use

a stretch type EPB system, as shown in Fig. 9a. The system consists
of (1) a module box including a DC motor, a gear box, a hall-effect
force sensor, and a magnetic encoder, (2) a parking cable, (3) a 12 V
battery, (4) a motor driver including a current sensor, (5) a control-
ler board with a digital signal processor (TMS320F2812), (6) a com-
puter, and (7) an oscilloscope. Fig. 9b is a zoomed-in image of the
module box (1) in Fig. 9a. Fig. 9c shows a zoomed-in image of a
parking cable. The end of the wire is clamped by a load cell, which
we use to measure the absolute tension of the parking cable. The
output of the load cell is connected to the force indicator.
4.2. Control method
Three control methods were evaluated in [16]: on–off, linear
proportional (P), and nonlinear P. The nonlinear P controller was
the most robust and had the most uniform tracking performance.
However, the nonlinear P controller needs a PWM driver. For low-
cost control, we use a simple on–off control, which applies the max-
imum input signal until the estimated clamping force reaches the
target force for braking force application. The input signal becomes
zero when the estimated force reaches or exceeds the target force:

umax ; e > 0
uðtÞ ¼ ; ð19Þ
0; e60
where u(t) is the input voltage for the motor, umax is the maximum
voltage of the battery, and e is the error between the target force,
ftar, and the estimated clamping force, f(heff), i.e., e = ftar  f(heff).
4.3. Results and discussions
We controlled the EPB system using the clamping force function
(13) with the simple on–off controller (19). In order to validate the
Target force [kgf]
Fig. 11. Clamping force errors for simple and novel on–off control methods.
proposed method for different clearances, we performed experi-
ments for three different target clamping forces under different
clearances; the results are shown in Fig. 10. To compare the esti-
mated forces with the measured forces, which are obtained from
the force sensor embedded in the module of the EPB system, the
measured forces are plotted as solid lines in Fig. 10. There is good
agreement in the force estimation. The initial points where the esti-
mated force follows the output of the force sensor are different be-
cause the experiment was performed under different clearances,
and the maximum error between the estimated forces and the mea-
sured forces was about 3.5 kgf, which is within the error bounds. We
observed that the estimated force was much less than the measured
force after the power was deactivated because the current ripple was
not reliably detected after that point. The EPB system had a maxi-
mum brake-applying time of less than 0.8 s. Thus, it satisfies the
application time specifications. The average and the deviation of
the errors are plotted in Fig. 11. We repeated the application and re-
lease 30 times for five different target clamping forces using the two
on–off control methods (the simple on–off and the novel on–off).
The negative error in Fig. 11 indicates that over-clamping force
was applied. All experiments satisfies the specifications for both
methods. The simple on–off controller results in over clamping at
every target force. The maximum amplitude of error is 5 kgf and
the minimum is 2 kgf. This controller discontinues the input voltage
1162 Y.O. Lee et al. / Mechatronics 21 (2011) 1156–1162
immediately after the estimated force reaches the target force and
does not have any compensation algorithm that considers the iner-
tia effect. The novel on–off control method recalculates that the tar-
get force should be less than the final clamping force based on the
LUT. The maximum amplitude error of the novel on–off controller
is 2 kgf, which is an improved averaged error compared to that of
the simple on–off control method. With the simple on–off controller,
we obtained errors of 5 kgf and 4.2 kgf at 80 kgf and 90 kgf,
respectively. Applying the novel on–off control reduces the averaged
errors to 1 kgf at 80 kgf and 0.2 kgf at 90 kgf, respectively. At a
target force of 100 kgf, both controllers have the same magnitude
of averaged error with the largest error deviation among the five tar-
gets. The differences in the deviation of the error are due to mechan-
ical vibrations of the test bench [27]. We observed that the
misalignment between the gear and screw could cause vibrations.
These vibrations appear in the current and force sensors signals as
low frequency signals, as shown in Fig. 4. As the braking force in-
creased, this vibration and the error deviation also increased. How-
ever, no significant vibration was observed in the Manufacturing
Verification Test (MVT) EPB systems. Therefore, use of our novel
on–off controller in manufactured EPB systems could provide more
regular and decreased error deviation compared to that in the test
rig. Low-cost control of the EPB system is therefore possible by using
our novel on–off controller.
5. Conclusions
We described novel clamping force estimation and control
methods for automotive EPB systems that can be realized using a
simple and low cost control structure without a clamping force
sensor. These methods exploit the relationship between the clamp-
ing force and the effective angular displacement of the motor from
the initial contact point between the brake pads and disk. Angular
velocity measurements were used to approximately detect the ini-
tial contact point. The clamping force was estimated using the
clamping force function of the angular displacement after the max-
imum angular velocity. We then applied a novel on–off control
method that reduces the inertia effects of the motor by predicting
the additional angular displacement due to momentum using sim-
ple control logic. We validated the proposed methods experimen-
tally; all results indicated that the methods satisfied the
specifications of the EPB system. Our proposed method can facili-
tate low-cost EPB system implementation.
References
[1] Bretz EA. By-wire cars turn the corner. IEEE Spectrum 2001;38(4):68–73.
[2] Lee YO, Lee CW, Chung CC, Son Y, Yoon P, Hwang I. Stability analysis of electric
parking brake (EPB) systems with a nonlinear proportional controller. In: Proc
of 17th IFAC world congress, Seoul, Korea; 2008. p. 14247–53.
[3] Wang N. Electric brake actuator. 2002. US patent app. 10/216,078.
[4] Jung HM, Park CH, Jeon JW, Yeo H. Development of electric parking brake
system. In: Proceedings of the KSAE autumn conference, Deoksan, Korea; 2007.
p. 962–7.
[5] Hyun D, Kim IS, Seo H. Study on a dynamic braking algorithm for an electronic
parking brake. In: Proceedings of the KSAE autumn conference, Deoksan,
Korea; 2007. p. 950–5.
[6] Godlewsky G, Maur T, Keller M, Leiter R, Kinder R, Poertzgen G. Electrically
actuatable vehicle brake and method for controlling an electrically actuatable
vehicle brake. 2005. US patent app. 11/126,902.
[7] Ringlstetter M. Parking brake system equipped with a sensor; 2005. US patent
app. 11/569,299.
[8] Hoseinnezhad R, Saric S, Bab-Hadiashar A. Estimation of clamp force in brake-
by-wire systems: a step-by-step identification approach. SAE transactions
journal of passenger cars: mechanical systems 2006. SAE paper 2006-01-1154.
p. 1088–97.
[9] Schwarz R, Isermann R, Bohm J, Nell J, Rieth P. Clamping force estimation for a
brake-by-wire actuator. SAE technical paper 1999. No. 1999-01-0482.
[10] Saric S, Bab-Hadiashar A, Walt J. Estimating clamp force for brake-by-wire
systems: thermal considerations. Mechatronics 2009;19(6):886–95.
[11] Line C, Manzie C, Good M. Control of an electromechanical brake for
automotive brake-by-wire systems with an adapted motion control
architecture. SAE Trans 2004;113(6):1047–56.
[12] Saric S, Bab-Hadiashar A, Hoseinnezhad R. Clamp-force estimation for a brake-
by-wire system: a sensor-fusion approach. IEEE Trans Vehicular Technol
2008;57(2):778–86.
[13] Hoseinnezhad R, Bab-Hadiashar A. Recent patents on measurement and
estimation in brake-by-wire technology. Bentham Science Publishers
2009;2(1):54–64.
[14] Hoseinnezhad R. Position sensing in brake-by-wire callipers using resolvers.
Vehicular Technol, IEEE Trans 2006;55(3):924–32.
[15] Hoseinnezhad R, Bab-Hadiashar A, Harding P. Calibration of resolver sensors in
electromechanical braking systems: a modified recursive weighted least-
squares approach. IEEE Trans Indus Electron 2007;54(2):1052–60.
[16] Lee YO, Lee CW, Chung HB, Chung CC, Son Y, Yoon P, et al. A nonlinear
proportional controller for electric parking brake (EPB) systems. In:
Proceedings of the 14th Asia Pacific automotive engineering conference,
Hollywood, CA, USA; 2007. No. 2007-01-3657.
[17] Son YS, Chung CC, Lee CW, Lee YO, Jang M, Lee W. Force control method of
electronic parking brake system; 2009a. Korea patent app. 10-2009-0054018.
[18] Consoli A, Bottiglieri A, Letor R, Ruggeri R, Testa A, De Caro S. Sensorless
position control of DC actuators for automotive applications. In: IEEE industry
applications conference, vol. 2. 2004. p. 1217–24.
[19] Lee CW, Chung HB, Lee YO, Chung CC, Son YS, Yoon P. Fault detection method
for electric parking brake (EPB) systems with sensorless estimation using
current ripples. Int J Automotive Technol 2010;11(3):387–94.
[20] Yang I, Son Y, Park I, Noh K, Yoon P. Development and performance evaluation
of electronic parking brake. In: Proceedings of the KSAE spring conference,
Korea; 2006. p. 953–8.
[21] Son Y.S, Chung C.C, Lee C.W, Lee Y.O, Jang M, Lee W. Force control method of
electronic parking brake system; 2009b. Korea patent app. 10-2009-0026441.
[22] Nise NS. Control systems engineering. John Wiley & Sons; 2004.
[23] Olsson H, Åström KJ, Canudas De Wit C, Gäfvert M, Lischinsky P. Friction
models and friction compensation. Europ J Control 1998;4:176–95.
[24] Son YS, Chung CC, Lee CW, Lee YO, Jang M, Lee W. Force control method of
electronic parking brake system; 2009c. Korea patent app. 10-2009-0026439.
[25] Astrom KJ, Wittenmark B. Adaptive control. Boston, MA, USA: Addison-Wesley
Longman Publishing Co., Inc.; 1994.
[26] Hoseinnezhad R, Bab-Hadiashar A, Rocco T. Real-time clamp force
measurement in electromechanical brake calipers. IEEE Trans Vehicular
Technol 2008;57(2):770–7.
[27] Lee W, Lee YO, Jang M, Lee CW, Chung CC, Chung HB. Fault tolerant control of
sensor fault of EPB system. Trans Korean Soc Automotive Eng
2010;18(4):8–17.