You are on page 1of 5

2012

International Symposium on Power Electronics,


Electrical Drives, Automation and Motion

PID Feedback Control Method


for Linear Resonant Actuator
Using an Estimated External Load
from the Back-EMF as a Target Voltage
Y. Asai*, K. Hirata* and T. Ota**
* Dept. of Adaptive Machine Systems, Graduate School of Eng., Osaka University, Suita, Osaka 565-0871, (Japan)
** Panasonic Corporation, Kadoma, Osaka 571-8686, (Japan)

Abstract—This paper proposes a feedback control This actuator mainly consists of two movers, a common
method for a Linear Resonant Actuator (LRA), in which an stator and resonant springs that support the air-gap
external load estimated from two signals of the back-EMF is (0.36mm). Two parallelly arranged movers are composed
used as a target voltage in PID control. The effectiveness of of two opposite pole magnets fixed on the back yoke. The
this method was verified through the FEM analysis in which common stator is composed of E-type laminated yoke
the magnetic field equation is coupled with electric circuit
with an exciting coil of 68 turns at its midleg. The
equation, control method, and motion equation. As a result,
it was found to be possible to maintain arbitrary amplitude resonant spring is composed of main-springs and link-
responded to the estimated load. springs which link two movers.

Index Terms--Linear resonant actuator, External load


estimation, PWM control, PID control, Finite element
method, Dynamic analysis. Resonant spring

I. INTRODUCTION Main-spring Back yoke


Recently, linear resonant actuators (LRA) have been Link-spring Magnet
used in a wide range of applications [1]-[3] because they
have a lot of advantages; high efficiency, simple structure,
easy control, and so on. On the other hand, they have a
problem that the amplitude severely decreases in response Stator yoke
to an external load. To control this, we proposed a PWM
feedback control method, in which the back-EMF of the Coil
coil is detected to determine the current duty [4] [5].
Successfully, we proposed the more effective feedback
Fig. 1. Basic structure of the LRA
control using PID control in PWM feedback control to
keep the amplitude constant against an external load [6].
Resonant spring
However, it was found to be difficult to maintain arbitrary
Main-spring
amplitude.
Link-spring
In this paper, we propose an external load estimation
method for a LRA using the back-EMF. The
effectiveness of this method was verified through the
FEM analysis [7]-[11] in which the magnetic field
equation is coupled with electric circuit equation, control
method, and motion equation. As a result, it was found to Stator yoke
be possible to maintain arbitrary amplitude responded to Back yoke
the estimated load. 16
Coil
Magnet
II. BASIC STRUCTURE AND CONTROL
A. Basic Structure and Operating Principle
10
The basic structure of the LRA in this study is shown
in Fig. 1 and the schematic diagram is shown in Fig. 2. 19
Fig. 2. Schematic diagram

978-1-4673-1301-8/12/$31.00 ©2012 IEEE


1547

Authorized licensed use limited to: UNIVERSITE DE TUNIS EL MANAR. Downloaded on February 07,2023 at 09:27:34 UTC from IEEE Xplore. Restrictions apply.
Fig. 3 shows the operating principle. The magnetic Amplitude
Input Voltage
Back EMF
poles of the stator are magnetized when the coil is ton
excited, and the thrust occurs between the stator and two
movers. Two movers move to reverse direction by Current
exciting the coil. The operation frequency of the LRA is
substantially determined by the spring constant and the

V1
mass of the mover. This actuator is operated under PWM toff
feedback control, and excitation time (duty) is decided by
detecting the back-EMF while the coil is not excited.
a b c d e
B. Unipolar PWM Feedback Control
Fig. 4. Unipolar PWM feedback control
We proposed the unipolar PWM feedback control in
the previous paper to simplify the control circuit
Battery + Battery + Battery +
configuration and extend the voltage detecting interval.
Fig. 4 shows the unipolar PWM feedback control of
ON OFF OFF
this actuator. First of all, in interval (a), the coil detects
the maximum value of back EMF V1 after the back EMF
is reduced to zero. In interval (b), the coil has a delayed
time (30s). In interval (c), the coil is exited under PWM
control according to the duty determined by PID control
using the back-EMF V1. The circuit diagram of this ON ON OFF
operating mode is shown in Fig. 5(a). In interval (d),
current circulates through a diode as shown in Fig. 5(b).
Finally, in interval (e), the circuit is opened as shown in
Battery - Battery - Battery -
Fig. 5(c). It is highly effective to drive the actuator under
the advanced PWM control, in which the input voltage of (a) Excitation mode (b) Circulation mode (c) Open mode
Fig. 5. Control circuit
multi-square wave is applied, and the actuator can be
operated in low electric power because the current does
not decrease soon even if the input voltage is turned off. III. ANALYSIS METHOD
In interval (c), the coil is exited according to the duty A. Magnetic Field Analysis
determined by PID control using the back-EMF V1. The The equations of the magnetic field and the electric
duty is defined by equations (1) and (2). circuit are coupled using the FEM, which are given by the
de(t ) magnetic vector potential A and the exciting current I0 as
Duty (%) = Kpe(t ) + KI ³ e(t ) dt +KD (1)
follows:
dt
e(t ) = VS − V1 (2) rot(ν rot A) = J 0 + ν 0 rot M (3)

Where KP is the proportional gain, KI is the integral dΨ


gain, KD is the differential gain, e(t) is the deviation, and E = V0 − RI 0 − =0 (4)
dt
VS is the target voltage.
nc
J0 = I 0 ns (5)
䂦䋺Pole by current Sc
䂔䋺Pole by magnet
Where ν is the reluctivity, J0 is the exciting current
moving direction density, ν0 is the reluctivity of the vacuum, M is the
magnetization of permanent magnet, V0 is the applied
voltage, R is the resistance, Ψ is the interlinkage flux of
exciting coil, nc is the number of turns, Sc is the cross-
N S sectional area of the coil and ns is the unit vector along
S N S with the direction of exciting current.
B. Coupled Analysis with Motion Equation
The motion equations are given as follow:
current
dx
M d 2x + C
2
+ Kx ± F = Fx (6)
dt dt
Fig. 3. Operating principle Where M is the mass of mover, x is the displacement of
mover, Fx is thrust, K is spring constant, F is load and C
is the viscous damping coefficient.
The thrust of mover is calculated using the Maxwell
stress tensor method, and is substituted in the equation

1548

Authorized licensed use limited to: UNIVERSITE DE TUNIS EL MANAR. Downloaded on February 07,2023 at 09:27:34 UTC from IEEE Xplore. Restrictions apply.
(6). The position of mover is calculated by the time step.
PWM feedback control is taken into consideration in this
analysis. Fig. 6 shows the flowchart for this coupled
analysis.

IV. DYNAMIC ANALYSIS UNDER UNIPOLAR PWM Back yoke


FEEDBACK CONTROL Stator yoke Magnet
A. Analyzed Model and Condition
Coil
Fig. 7 shows the FEM model. The number of elements
is about 19,000, and unknown variables are about 9,500. y
Table 1 shows the analysis conditions. Total resistance of
a coil and FET is given by resistance(on) when control x
circuit is in excitation mode, and by resistance(off) when Fig. 7. FEM model
control circuit is in circulation mode. The number of
steps is 12,000, time division is 5s, and total CPU time Table 1.Analysis condition
is about 2 hours. Input voltage (V) 3.6
Resistance(on) (m) 350
Coil
B. Dynamic Characteristic Resistance(off) (m) 290
Number of turns (turn) 68
Fig. 8 shows the computed results when it is operated Magnetization of magnets (T) 1.42
with an external load under PID feedback control. The Mass of mover (g) 6.95
target voltages VS = 0.99, 1.48, 1.97, 2.26[V] are decided Spring constant (N/mm) 13.36
by the induced voltage when each amplitude of 1.0 mm, Viscous damping coefficient (N·s/m) 0.14
1.5 mm, 2.0 mm, and 2.4 mm is maintained. Diode voltage (V) 0.5
The following gains are decided by the analysis. KP = 1,
KI = 0.05, KD = 0.5. 3 Target
As can be seen, the amplitude is kept constant by PID Vs=0.99
2.5
control. With the load of more than 1N at VS=2.26[V], the Vs=1.48
Amplitudepk-pk[mm]

amplitude decreases because its duty reaches nearly 2 Vs=1.97


100%. Moreover, it is found that the maintained Vs=2.26
amplitude well responses when the target voltage is 1.5
Amplitude
changed. The amplitude can be kept constant against an
1 Vs=0.99
external load in all target voltages.
Vs=1.48
From these results, the effectiveness of the feedback 0.5
Vs=1.97
control using PID control is clarified.
0 Vs=2.26
0 0.2 0.4 0.6 0.8 1 1.2
Load[N]

Fig. 8. Relation between target voltage and amplitude

V. PROPOSAL OF LOAD ESTIMATION PWM FEEDBACK


CONTROL
A. Load Estimation Method
It is described in the previous chapter that the constant
amplitude could be kept against an external load by
unipolar PWM feedback control. However, it was found
to be difficult to maintain arbitrary amplitude. It is
necessary to estimate the external load in order to
maintain arbitrary amplitude according to the external
load.
Under unipolar PWM feedback control, it is difficult
to estimate the external load only by V1. Therefore, the
load estimation method is proposed using two signals of
back-EMF V1 and V2. V2 is the back-EMF when a
constant time (250s) passes after it is increased to zero.
Fig. 6. Flowchart for analysis In order to clarify the relation of two detection voltage,
the damped oscillation was analyzed when an initial
position and the external load were given. Fig. 9 shows
the relationship between V1 and V2 when the external load
was varied. From a result, it is found that the external

1549

Authorized licensed use limited to: UNIVERSITE DE TUNIS EL MANAR. Downloaded on February 07,2023 at 09:27:34 UTC from IEEE Xplore. Restrictions apply.
load is decided by V1 and V2. 0.7
The external load is estimated by equations (7) and (8) 0.6
derived motion equation.
0.5
External
K a {V1 exp(−ξπ ) − KvV2 }
F'= k (7) 0.4 load

V2[V]
2{exp(−ξπ ) + 1} 0.0N
0.3
0.4N
c
ξ= (8) 0.2 0.8N
2 mk 0.1 1.2N
Where F’ is the estimated load, Ka and Kv are 0
constants, ξ is damping ratio. 0 0.5 1 1.5 2 2.5
V1[V]
B. Amplitude Control Method
Fig. 9. Relation of V1, V2. and the external load
Under unipolar PWM feedback control, the amplitude
is able to keep constant against an external load. And
Amplitude Input Voltage Back EMF
when the target voltage is changed, the amplitude well ton
Current
responses. Therefore it is possible to maintain the
arbitrary amplitude by changing the target voltage
responded to the estimated load as follow.

V2
VS = f ( F ' ) (9)

V1
toff
C. Load Estimation PWM Feedback Control
Fig. 10 shows load estimation PWM feedback control
a b c d e f
of this actuator. First of all, in interval (a), the coil detects
Fig. 10. Load estimation PWM feedback control
the maximum value of back-EMF V1 after the back-EMF
is reduced to zero. In interval (b), the coil detects the
3
back-EMF V2 when a constant time (250s) passes after it Target
is increased to zero. In interval (c), the coil has a delayed 2.5
A=0
time (30s). In interval (d), the coil is exited under PWM A=0.5
Amplitudepk-pk[mm]

control according to the duty determined by PID control 2 A=1.0


and equations (7) and (9) using the back-EMF V1 and V2. A=1.5
1.5
The circuit diagram of this operating mode is shown in Amplitude
Fig. 5(a). In interval (e), current circulates through a 1 A=0
diode as shown in Fig. 5(b). Finally, in interval (f), the
A=0.5
circuit is opened as shown in Fig. 5(c). 0.5
A=1.0
0 A=1.5
VI. DYNAMIC ANALYSIS UNDER LOAD ESTIMATION 0 0.2 0.4 0.6 0.8 1 1.2
UNIPOLAR PWM FEEDBACK CONTROL Load[N]
Fig. 11 shows the analyzed results when it is operated (a) VS=A*F’+1
with the external load under load estimation PWM
3
Feedback control. In this analysis, target voltage Vs is
given by equations (10) or (11). 2.5 Target
B=1
Amplitudepk-pk[mm]

VS = A * F '+1 (10) 2 B=2

VS = B * ( F '−0.6) + 2 2
(11) 1.5
B=3

Amplitude
Where A and B are constants. 1
From these results, it is possible to maintain the B=1
0.5 B=2
arbitrary amplitude by changing the target voltage
responded to the estimated load. When the load is larger 0 B=3
than 1.0N at A= 1.5, the amplitude decreases because its 0 0.2 0.4 0.6 0.8 1 1.2
duty becomes nearly 100%. Load[N]
Fig. 12 and Fig. 13 shows transient property of (b) VS=B*(F’-0.6)2+1
amplitude and estimated loads at A =0.5. As can be seen, Fig. 11. Load characteristic
it is found that the external load can be estimated
correctly.
It was shown that arbitrary amplitude can be almost
maintained according to the estimated load which was
difficult under the sensorless control.

1550

Authorized licensed use limited to: UNIVERSITE DE TUNIS EL MANAR. Downloaded on February 07,2023 at 09:27:34 UTC from IEEE Xplore. Restrictions apply.
1 VII. CONCLUSION
This paper presented the load estimation method for
Amplitude[mm]

0.5
our LRA using the back-EMF. The effectiveness of this
0 method was verified through the FEM analysis, in which
the magnetic field equation was coupled with electric
-0.5 circuit equation, control method, and motion equation. As
a result, it was found that it was possible to maintain
-1
0 0.02 0.04 0.06 0.08
arbitrary amplitude responded to the estimated load.
Time[sec]
REFERENCES
(a) 0.0N
[1] Tadashi Yamaguchi, Yoshihiro Kawase, Koichi Sato,
1 Satoshi Suzuki, Katsuhiro Hirata, Tomohiro Ota and Yuya
Hasegawa, “Trajectory Analysis of 2-D Magnetic Resonant
Amplitude[mm]

0.5
Actuator” , IEEE Trans. Magn., vol.45, No.3, pp.1732-
1735, Mar. 2009.
0
[2] T. Yamaguchi, Y. Kawase, S. Suzuki, K. Hirata, T. Ota and
-0.5 Y. Hasegawa “Dynamic Analysis of Linear Resonant
Actuator Driven by DC Motor Taking into Account
-1 Contact Resistance between Brush and Commutator”,
0 0.02 0.04 0.06 0.08 IEEE Trans. Magn., vol.44, No.6, pp.1510-1513, Jun. 2008.
Time[sec] [3] K. Hirata, T. Yamamoto, T. Yamaguchi, Y. Kawase and Y.
Hasegawa, ”Dynamic Analysis Method of Two-
(b) 0.4N Dimensional Linear Oscillatory Actuator Employing Finite
1 Element Method”, IEEE Trans. Magn., vol.43, No.4,
pp.1441-1444, 2007.
Amplitude[mm]

0.5 [4] K. Matsui, K.Hirata, and T.Ota, “Dynamic Analysis of


Linear Resonant Actuator under PWM Control Employing
0 the 3-D Finite Element Method”, Proceedings of the 13th
Biennial IEEE CEFC (Conference on Electromagnetic
-0.5 Field Computation), OC1-1, p.192, 2008.
[5] Y. Asai, K. Hirata, and T. Ota, “Dynamic Analysis Method
-1
of Linear Resonant Actuator with Multi-Movers
0 0.02 0.04 0.06 0.08
Employing 3-D Finite Element Method”, IEEE Trans.
Time[sec]
Magn., vol.46, no.8, pp.2971-2974, Aug. 2010.
(c) 0.8N [6] Y. Asai, K. Hirata, and T. Ota, “3-D Finite Element
1
Analysis of Linear Resonance Actuator under PID
Control”, Proceedings of the 14th Biennial IEEE CEFC
(Conference on Electromagnetic Field Computation), 32P8,
Amplitude[mm]

0.5
2010.
0 [7] Satoshi Suzuki, Yoshihiro Kawase, Tadashi Yamaguchi,
Katsuhiro Hirata and Yu Okaue, “Dynamic Response
-0.5 Analysis of Shear-Type Compact MR Brake”, IEEE Trans.
Magn., vol.45, No.3, pp.1-4, Mar. 2009.
-1 [8] T. Ota, Y. Mitsutake, Y. Hasegawa, K. Hirata and T.
0 0.02 0.04 0.06 0.08 Tanaka, ”Dynamic Analysis of Electromagnetic Impact
Time[sec] Drive Mechanism Using Eddy Current”, IEEE Trans.
Magn., VOL.43, No.4, pp.1421-1424, 2007.
(d) 1.2N [9] T. Yamaguchi, Y. Kawase, H. Kodama, K. Hirata, T. Ota
Fig. 12. Transient property of amplitude
and Y. Hasegawa, “Eddy Current Damping Analysis of
Laser Marker Using 3-D Finite Element Method”, IEEE
1.6 Trans. Magn., vol.42, No.4, pp.1011-1014, Apr. 2006.
1.4 [10] T. Yamaguchi, Y. Kawase, A. Nakase, K. Hirata and T. Ota,
1.2 External “Multi-motion Analysis of Opening and Closing Sensor for
Estimated load[N]

1 load Windows Using 3-D Finite Element Method”, IEEE Trans.


0.8
0 Magn., vol.42, No.4, pp.1015-1018, Apr. 2006.
0.6
0.4 [11] T. Ota, K. Hirata, T. Yamaguchi, Y. Kawase, K.
0.4
0.8 Watanabeand A. Nakase, “Dynamic Response Analysis of
0.2
Opening and Closing Sensor for Windows”, IEEE Trans.
0 1.2
Magn., vol.41, No.5, pp.1604-1607, May. 2005.
-0.2
-0.4
0 0.02 0.04 0.06 0.08
Time[sec]

Fig. 13. Estimated loads

1551

Authorized licensed use limited to: UNIVERSITE DE TUNIS EL MANAR. Downloaded on February 07,2023 at 09:27:34 UTC from IEEE Xplore. Restrictions apply.

You might also like