Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Proceedings of the 2017 IEEE/SICE International Symposium

on System Integration, Taipei, Taiwan, December 11-14, 2017 :H&

The extended Bouc-Wen model with structural element and its


application to the compensation of rate-dependent hysteresis of a
piezoelectric actuator
Ken’ichi Tatebatake∗ and Fumitake Fujii∗

Abstract— This paper proposes the extended Bouc-Wen which suits well to the model-based hysteresis compensator
model to capture the hysteretic behavior of the piezoelectric design is the Bouc-Wen model[7].
actuator. The proposed model adds additional terms to the One has to increase the number of parameters in the model
basic Bouc-Wen model to explain the behaviors commonly
observed with the thin piezoelectric actuators which include to increase its accuracy if the effort is put on the hysteresis
the rate-dependence, asymmetric loop formation and the phase operator based mathematical models. The Bouc-Wen model
lag. Remaining part of the paper is devoted to the synthesis on the other hand requires small number of parameters to
of the hysteresis compensator using the proposed model. It capture the hysteretic behavior. The basic Bouc-Wen model
has been shown through the experimental results that the requires only 4 parameters. The authors recently proposed
identified model provides good modeling accuracy and high
control performance as well, as compared to the results obtained the enhanced Bouc-Wen model[8] to capture the aforemen-
with the classical Bouc-Wen model. tioned eccentric behavior of the real world piezoelectric
actuator which includes 20 parameters, whereas our previous
I. INTRODUCTION work[9] shows that we need to determine 5050 parameters
when we try to establish the Preisach model which exhibits
Recent improvements on the manufacturing technology almost the same precision as the enhanced Bouc-Wen model.
enable the production of small and highly integrated devices Various controllers have been developed to compensate
both for consumer and industry markets. This invokes the the hysteretic behavior of the piezoelectric actuators using
demands for the development of small actuators. Piezoelec- the Bouc-Wen model. Rakotondrabe[6] recently proposed the
tric actuators attract great attentions of the researchers and direct inverse multiplication method for the compensation of
practitioners in the field because of the advantages in their hysteresis of the piezoelectric actuator. The structure of the
size, fast response, high force generation and fine positioning Bouc-Wen model plays a key role on the derivation of the
resolution. Applications can be found in the literature which algorithm. Its practical advantage lies in the extreme easiness
include the atomic force microscope nanopositioner[1], micro of the implementation and the low computation cost. Yang
robot[2], and the positioning control of the head of the hard et al.[10] synthesized a 2 d.o.f. control based on the Bouc-
disk drives[3]. Wen model to improve the trajectory tracking accuracy of
However, it is known that the piezoelectric actuator ex- the MFC actuator. The parameters of the Bouc-wen model
hibits hysteresis between its input and the output. Positioning for the actuator were identified by the genetic algorithm.
accuracy of the piezoelectric actuators will be severely Zhu and Rui[11] formulated a generalized Bouc-Wen model
deteriorated if no adequate hysteresis compensation is given. for the hysteresis characteristics of the piezoelectric actuator
It is also known that the characteristics of the hysteresis may by combining the modified Bouc-Wen model with a 2nd
differ when the rate/frequency of the driving signal varies. order system element which is responsible for the structural
This phenomenon is often referred to as the rate-dependent dynamics of the actuator. They also constructed a 2 d.o.f.
hysteresis[4]. Great efforts have been devoted to the modeling control system whose controllers are determined from the
and compensation of the rate-dependent hysteresis of the identified models. The compensators developed in the latter
piezoelectric actuators. two are the extended versions of the compensator proposed
Many mathematical models have been developed to cap- by Rakotondrabe[6].
ture the hysteretic behavior of the piezoelectric actuator. Two However, as the basic Bouc-Wen model forms a clockwise
major lines exist in the development of the phenomenological hysteresis loop which is symmetric with respect to the origin
models of hysteresis. One is to use the hysteretic operators of the input/output plane, and as it cannot exhibit rate-
with associated distribution functions to capture the behavior dependent behavior, straightforward application of the Bouc-
of the hysteresis. Examples include the Preisach model[12], Wen model to the real world piezoelectric actuators only pro-
the Prandtl-Ishlinskii model[5], the Play model and the Stop duces limited accuracy. Modification of the model structure
model[13]. The other is to utilize the differential equation as and/or the augmentation of the model should therefore be
the key component of the model. Prime example of the latter considered when one tries to synthesize a high performance
∗ K. Tatebatake and F. Fujii are with the Graduate School of
motion controller of the piezoelectric actuator based on the
Science and Technology for Innovation, Yamaguchi University, Japan Bouc-Wen model.
w033vd@yamaguchi-u.ac.jp Along with the lines, this paper proposes the new extended

978-1-5386-2263-6/17/$31.00 ©2017 IEEE



Bouc-Wen model which provides enhanced modeling capa-
bility. The proposed extended Bouc-Wen model combines the
modified Bouc-Wen model with the 3rd order linear model
to capture the behavior of the bimorph type piezoelectric
actuator which exhibits rate-dependent hysteresis. The pro-
posed extended Bouc-Wen model can form the asymmetric
hysteresis loop and exhibit the rate-dependent behavior as
well. We also propose the hysteresis compensator design for
the piezoelectric actuator which is the augmented/improved
version of the recent excellent work by Rakotondrabe[6].
This paper has six sections. Section 2 describes the basic Fig. 1 An example of the response of the basic Bouc-Wen
Bouc-Wen model and introduces the proposed extended model to a sinusoidal input. The Bouc-Wen model exhibits a
Bouc-Wen model. Section 3 briefly states the procedure of clockwise loop whose center is the origin of the input/output
the on-line parameter identification. Details of the hysteresis plane.
compensator design are disclosed in section 4. Section 5
shows the result of the experimental verification of the
performance of the proposed control system. Summaries
together with some future work implications will be given
in section 6 to conclude the paper.

II. M ODELING THE BEHAVIOR OF THE PIEZOELECTRIC


ACTUATOR

A. Bouc-Wen model
The basic version of the Bouc-Wen model[14] is a form
of the phenomenological mathematical models of hysteresis Fig. 2 An example of the observed hysteresis loop of the
which uses a differential equation. The basic Bouc-Wen piezoelectric actuator used in this study. It is an input/output
model is characterized by trajectory for a 5Hz sinusoidal wave input of 1V amplitude.
Notice that the hysteresis loop is not centered at the origin
ḣ = Au̇ − γ |h|n u̇ − β |u̇||h|n−1 h, (1) and the loop circulates in the counter-clockwise direction.

where u is the input and h is the model output. The


parameters A, γ , β and n are related to the geometrical char- where A0 is the additionally employed model parameter of
acteristics of the hysteresis loop. To be specific, A governs the modified Bouc-Wen model. Introduction of the additional
the amplitude, γ and β determine the shape, and n defines term A0 |u̇| to the model enables us to modify the sensitivity
the smoothness, respectively. Fig. 1 shows an example of of the model response to u̇ depending on its sign sgn(u̇), as
a hysteresis loop depicted numerically with the Bouc-Wen described by the equation
model (1). Following the result disclosed by Ismail et al.[16], 
n is set to be unity throughout this paper. The Bouc-Wen (A + A0 )u̇ u̇ > 0
Au̇ + A0 |u̇| = (4)
model in this case is simplified to (A − A0 )u̇ u̇ < 0.

ḣ = Au̇ − γ |h|u̇ − β |u̇|h. (2) Fig. 3 shows an example of the hysteresis loop produced
by the modified Bouc-Wen model (3). It can be seen from
the figure that the model (3) exhibits a hysteresis loop
whose center is off the origin of the input/output plane. The
B. Modification of the Bouc-Wen model modified Bouc-Wen model is further altered here to form
Fig. 2 depicts the relationship between the input voltage a counterclockwise hysteresis loop. Fig. 4 shows the idea.
and the output displacement of the piezoelectric actuator This additional modification of the model results in the set
used in this study. It can be seen from the figure that the of equations represented by
center of the hysteresis loop is not located at the origin of
ḣ = Au̇ − γ |ḣ|u̇ − β |u̇|ḣ + A0 |u̇|
the input/output plane and the loop is counterclockwise. The
Bouc-Wen model on the other hand forms a clockwise loop yt = A1 u − h, (5)
which is symmetric about the origin, as evidenced from Fig.
1. This observation reveals that modification of the structure where another additional parameter A1 defines the slope of
of the Bouc-Wen model is necessary to capture the hysteresis the linear response and yt is the output of the model. This
loop whose center is off the origin of the input/output plane. model (5) is the extended version of the classical Bouc-
The modified model is given by Wen model fitted to the piezoelectric actuator. The proposed
model (5) is discretized here to facilitate the digital com-
ḣ = Au̇ − γ |ḣ|u̇ − β |u̇|ḣ + A0 |u̇|, (3) puter implementation of the hysteresis compensator based


Fig. 6 Input/output trajectory of the piezoelectric actuator
when a 2Hz sinusoidal wave input of 1V amplitude is given
Fig. 3 An example of the hysteresis loop of the modified to the piezo driver.
Bouc-Wen model. The modified Bouc-Wen model can form
asymmetric hysteresis loop whose center is off the origin of
the input/output plane.

Fig. 7 Input/output trajectory of the piezoelectric actuator


Fig. 4 The 2nd modification of the Bouc-Wen model. It
when a 30Hz sinusoidal wave input of 1V amplitude is fed
includes the negation of h and the linear augmentation based
to the piezo driver. Phase lag can be clearly observed in its
on the input signal.
response.

response can often be observed in the real world piezoelectric


actuator. Some fundamental thought in physics raises a
hypothesis that the model should equip a component which
is able to capture the structural behavior of the thin bimorph
type piezoelectric actuator. We accordingly introduce the IIR
filter into the model to express the phase lag of the actuator.
The final form of the proposed extended Bouc-Wen model
results in the set of equations described by
Fig. 5 An example of the hysteresis loop of the extended
Bouc-Wen model (6). It is a counterclockwise hysteresis h[k] = h[k − 1] + A(u[k] − u[k − 1]) − A0 |u[k] − u[k − 1]|
loop. −γ |h[k − 1]|(u[k] − u[k − 1]) − β |u[k] − u[k − 1]|h[k − 1]
y[k] = y[k − 1] + a1 y[k − 2] + a2 y[k − 3] + A1 u[k] − h[k], (7)
on this model. The result is expressed by the following two
where a1 ,a2 ,A1 ,A,γ ,β and A0 are the parameters which
equations
determine the behavior of this extended Bouc-Wen model
h[k] = h[k − 1] + A(u[k] − u[k − 1]) − A0 |u[k] − u[k − 1]| and y[k] represents the displacement output of the bimorph
type piezoelectric actuator. An example of the hysteresis loop
−γ |h[k − 1]|(u[k] − u[k − 1]) − β |u[k] − u[k − 1]|h[k − 1] generated with the extended Bouc-Wen model is given in Fig.
8. It can be seen from the figure that the proposed extended
yt [k] = A1 u[k] − h[k], (6)
Bouc-Wen model shows a hysteresis loop with phase lag.
where k(k = 1, 2, · · · ) denotes the sampling number. Fig. 5
III. C OMPENSATION OF HYSTERESIS USING THE
shows an example of the hysteresis loop obtained with (6).
PROPOSED EXTENDED B OUC -W EN MODEL
C. Model for a bimorph type piezoelectric actuator A. Parameter identification of the extended Bouc-Wen model
Fig. 6 shows the input voltage excitation of 2Hz sinusoid The extended Bouc-Wen model has been shown to have
and the output displacement of the piezoelectric actuator the potential to express the characteristic behaviors which
used in this study, whereas Fig. 7 shows the same input and are observed with the piezoelectric actuators exhibiting the
output plots where the input frequency is altered to 30Hz. rate-dependent hysteresis. However, it is crucially important
Comparison of Fig.7 with Fig.6 clearly reveals that the phase to identify the parameters of the model to capture the
lag is observed in the 30Hz response. The phase lag in the behavior of the actuator with satisfactory precision. We use


Fig. 8 An example of the hysteretic responses of the
proposed extended Bouc-Wen model (7) for sinusoidal inputs Fig. 10 Schematic diagram of the experimental apparatus.
of 1Hz and 25Hz. It can be seen that the model can form a The PC sends command signal to the piezo driver. The driven
counterclockwise hysteresis loop with phase lag when the amplifies the command signal and drives the actuator. The
input signal frequency is high. It can also be evidenced displacement of the actuator is measured with the capacitance
from the figure that the proposed extended Bouc-Wen model type displacement sensor M-2218.
exhibits rate-dependence.
TABLE I Initial parameters of the proposed extended Bouc-
wen model.

The initial parameters


a1 0
a2 0
A1 0.01
A 5.0× 10−3
A0 0
γ 1.0× 10−3
β 1.0× 10−3

where ŷ[k] is a signal characterized by

Fig. 9 The schematic diagram of feedforward control. ŷ[k] = yd [k] − y[k − 1],
The hysteresis element of the identified model is used to
with yd [k] being the output measurement of the actuator at
counteract the hysteresis element of the piezoelectric actuator
the sampling instance k in the identification experiment.
during compensation.
B. Hysteresis Compensation based on the extended Bouc-
Wen model
the recursive least squares (RLS) algorithm to determine the
The feedforward controller based on the extended Bouc-
parameters of the model. It is easy to see that the extended
Wen model is proposed here in order to achieve the linear
Bouc-Wen model (7) can be written in the linear regression
drive of the piezoelectric actuator. This controller owes its
form
theoretical basis to the direct inverse multiplication proposed
y[k] = y[k − 1] + θ T [k]X
X [k], (8)
by Rakotondrabe[6].
where Suppose that the parameters of the extended Bouc-Wen
 T model are appropriately identified to capture the behavior of
θ= a1 , a2 , A1 , A, A0 , β, γ , (9) the actuator accurately. Then the problem is to determine the
is the vector of parameters and control input ν [k] so that y[k] in (7) would be identical to
the desired trajectory d[k]. If we replace y[k] of (7) with the
 k
desired trajectory d[k], a simple algebraic calculation derives
X = y[k − 2], y[k − 1], u[k], ∑ (u[i] − u[i − 1]),
i=1 the control law ν [k] which is described by
k k h[k] = h[k − 1] + A(ν [k] − ν [k − 1]) − A0 |ν [k] − ν [k − 1]|
∑ |u[i] − u[i − 1]|, ∑ |u[i] − u[i − 1]|h[i − 1],
i=1 i=1 −γ |h[k − 1]|(ν [k] − ν [k − 1]) − β |ν [k] − ν [k − 1]|h[k − 1]
k T
d[k] − d[k − 1] − a1 d[k − 2] − a2 d[k − 3] + h[k − 1]
∑ |h[i − 1]|(u[i] − u[i − 1]) (10) ν [k] =
A1
.
i=1
(12)
is the regression vector. The model error e is then defined The desired trajectory d[k] is determined by
by
e[k] = ŷ[k] − θ T [k]X
X [k], (11) d[k] = α u[k], (13)


Fig. 11 RMS errors between the actuator response and the Fig. 14 RMS errors between the actuator response and the
model output calculated for 1 to 30Hz sinusoidal inputs. desired displacement calculated for 1 to 30Hz sinusoidal
Dotted line shows the RMS value of the ambient sensor inputs. Dotted line shows the RMS value of the ambient
noise. sensor noise.

Fig. 12 The phase difference between the model output(18) Fig. 15 The phase difference between the actuator response
and the actuator response calculated for 1 to 30Hz. and the desired displacement calculated for 1 to 30Hz, when
the feedforward compensator is synthesized with the basic
Bouc-Wen model. Intrinsic phase lag of the actuator cannot
be compensated in almost all frequencies tested.

Fig. 13 The phase difference between the proposed extend


Bouc-Wen model output and the actuator response calculated
for 1 to 30Hz. Phase lag characteristics are successfully
captured at most of the input frequencies. Fig. 16 The phase difference between the actuator response
and the desired displacement calculated for 1 to 30Hz,
when the feedforward compensator is synthesized with the
where α is a frequency dependent parameter to specify the proposed extended Bouc-Wen model. Phase lags have been
desired trajectory. As the amplitude of the response varies as eliminated successfully with only a few exceptional frequen-
a function of input sequence frequency, it is natural to use cies.
k=1 yd [k]u[k]
∑m
α= (14)
∑mk=1 u[k]
2
in this study is a bimorph piezoelectric actuator (FDK,PZBA-
for equation (13) where u[k] is the input sequence used in 00030). The experimental calculations are done in real time
the identification experiment and yd [k] is the corresponding by the PC in which a real-time OS ART-Linux is installed.
displacement output of the actuator. Fig. 9 shows a schematic The PC has a CPU (AMD, phenom X4 955 Processor)
diagram of the proposed feedforward controller. The pro- driven at 3.2GHz and a 16bit A/D and D/A conversion board
posed extended Bouc-Wen model identified is used directly (Interface, LPC-361116). The piezo driver (NF, As-904-150
to cancel the hysteresis characteristics of the piezoelectric B) amplifies the command voltage signal within the range
actuator. of bipolar 75V to drive the piezoelectric actuator. A non-
contacting displacement sensor (MESS-TEK, M-2218) and
IV. E XPERIMENTAL VALIDATION
the corresponding sensor probe (MESS-TEK, TRA1020-1K-
A. Experimental device V3-1057) is used to measure the displacement of the actuator.
Fig. 10 shows a schematic diagram of the experimental The resolution of the non-contacting displacement sensor is
equipment used in this study. The piezoelectric actuator used 10nm.


Fig. 17 Actuator response and desired displacement for Fig. 20 Actuator response and desired displacement for
10Hz sinusoidal input 10Hz sinusoidal desired trajectory.

Fig. 18 Actuator response and desired displacement for Fig. 21 Actuator response and desired displacement for
23Hz sinusoidal input 23Hz sinusoidal desired trajectory.

Fig. 19 Actuator response and desired displacement for Fig. 22 Actuator response and desired displacement for
30Hz sinusoidal input 30Hz sinusoidal desired trajectory.

B. Identification experiment and output of the piezoelectric actuator for every frequency
A sinusoidal wave is fed to the piezo driver as the from 1 to 30Hz is calculated to verify tracking accuracy.
described mathematically by Tracking error et [k] in the control experiment is

u[k] = sin(2π f Δt(k − 1)), (15) et [k] = d[k] − ycon [k], (16)

where f is the input frequency, Δt is the sampling interval where ycon [k] is the measurement of the actuator displace-
which is 1ms in this experiment. The experimental output of ment at time k. We also conducted the same control exper-
the piezoelectric actuator is measured for 100 seconds after iment with the model of the piezoelectric actuator used in
the output converges to the steady state oscillation. The pa- Rakotondrabe[6] in order to verify the advantage of the model
rameters of the extended Bouc-Wen model are identified by proposed in this study. The model used by Rakotondrabe is
the RLS algorithm using that data for each input frequency. described by
The input frequency is altered from 1Hz to 30Hz with the
y(t) = d pU(t) − h(t) y(t0 ) = y0
increment of 1Hz. Table 1 shows the initial values of the
parameters to be updated for the extended Bouc-Wen model. ḣ = AU̇ − B|U̇|h − ΓU̇|h| h(t0 ) = h0 , (17)
The root mean squared error (RMSE) between the actuator
displacement and the model output is calculated to verify the where y is the displacement output and U is the control
modeling accuracy. input. d p ,A,B and Γ are the parameters of the model. (17) is
discretized accordingly which finally results in the following
C. Control experiment two equations
We conducted the compensation experiments for each h[k] = h[k − 1] − Γ|h[k − 1]|(U[k] −U[k − 1])
frequency using the identified parameters. Each attempt lasts
for 100 seconds. The RMSE between the desired trajectory +A(U[k] −U[k − 1]) − β |U[k] −U[k − 1]|h[k − 1]


y[k] = d pU[k] − h[k]. (18)
thesized by combining the modified Bouc-Wen model and
Both the identification and the controller synthesis were the 3rd order linear model. We have also proposed the
carried out in the same manner as explained with our model. hysteresis compensator design for the piezoelectric ac-
The only difference is the use of the model (18). tuator which is the extended version of the work by
Fig. 11 shows the RMSEs on the modeling error sequence Rakotondrabe[6]. The extended Bouc-Wen model outper-
for each input frequency and the ambient sensor noise. forms the model (18) at most input frequencies. Our future
Fig. 12 shows the difference of the phase between the work is on the treatment of the 3rd harmonic of the actuator
model(18) and the actuator response, whereas Fig. 13 plots response for perfect linear drive of the actuator.
the difference between the proposed extended Bouc-Wen
model and the actuator response. It can be inferred from R EFERENCES
the figures that the extended Bouc-Wen model has the better [1] Y. Yong, B. Bhikkaji and S. O. R. Moheimani, ʠ Design, modeling,
modeling accuracy than the model (18) as the proposed and FPAA-based control of a high-speed atomic force microscope
nanopositioner ʡ, IEEE/ASME Transactions on Mechatronics, vol. 18,
model exhibits small RMS modeling errors and the phase no. 3, 2013, pp. 1060-1071.
lag of the model response is negligible at most of the tested [2] M. Karpelson, G. Wei and R. J. Wood, ʠ Driving high voltage
frequencies. piezoelectric actuators in microrobotic applications ʡ, Sensors and
Actuators A: Physical, vol. 176, 2012, pp. 78-89.
As for the control performance, Fig. 14 shows the tracking [3] M. A. Rahman, A. A. Mamun and K. Yao, ʠ Rate dependent direct
error RMSEs between the desired trajectory and the out- inverse hysteresis compensation of piezoelectric micro-actuator used
put displacement of the piezoelectric actuator for 1-30Hz in dual-stage hard disk drive head positioning system ʡ, Review of
Scientific Instruments, vol. 86, no. 8, 2015, pp. 085002
frequencies. The figure also shows level of the the ambi- [4] X. Zhang and Y. Tan,ʠ A hybrid model for rate-dependent hysteresis
ent sensor noise. It is evident that the proposed controller in piezoelectric actuators ʡ, Sensors and Actuators A: Physical, vol.
outperforms the counterpart. Figs. 15 and 16 both show 157, 2010, pp. 54-60.
[5] Z. Li, O. Aljanaideh, S. Rakheja and C. Su, ʠ Compensation of
the phase difference between the actuator output the the play operator-based Prandtl-Ishlinskii hysteresis model using a stop
reference: the only difference is the model. Fig.15 shows operator with application to piezoelectric actuators ʡ, International
the results with the basic model (18) whereas Fig.16 shows Journal of Advanced Mechatronic Systems, vol. 4, No.1 (2012), pp.
25-31.
the results with the proposed extended Bouc-Wen model. [6] M. Rakotondrabe, Bouc-Wen modeling and inverse multiplicative
The controller based on the proposed model basically com- structure to compensate hysteresis nonlinearity in piezoelectric ac-
pensates the intrinsic phase lag of the actuator with fine tuators, IEEE Transactions on Automation Science and Engineering,
Vol.8, No.2(2011), pp.428-431.
precision but performance degradation can be observed at [7] Y. Wen, Method for random vibration of hyseteretic system, ASCE J
some frequencies. Eng Mech Div, 102(2), pp. 249-263.
In order to gain insights on the cause of performance [8] K. Tatebatake, Y. Ueda, D. Liu and and F. Fujii, Modeling of hysteresis
characteristics of the piezoelectric actuator using the bouc-wen model,
degradation observed at some frequencies, we drew the in- Transactions of the JSME, Vol.83, No.852(2017), pp. 1-14.
put/output plots using both the identification and the control [9] Y. Ueda, F. Fujii and D. Liu, “Identification of hysteresis using
experiment results at 10, 23 and 30Hz frequencies in Figs. adaptive Preisach model,” Transactions of the JSME, Vol. 81, No.
830, 2015
17, 18 and 19 for open loop responses and in Figs. 20, [10] Y. Yang, Y. Wei, J. Lou, L. Fu, G. Tian and M. Wu, Hysteresis model-
21 and 22 for controlled outputs. All plots in these six ing and precision trajectory control for a new MFC micromanipulator,
figures are with the extended Bouc-Wen model. It can be Sensor and Actuator A:Physical, Vol. 247, 2016, pp. 37-52.
[11] W. Zhu and X. T. Rui, Hysteresis modeling and displacement control
seen from these figures that the identification accuracy and of piezoelectric actuators with the frequency-dependent behavior using
compensation accuracy at the input frequency of 23Hz are a generalized Bouc-Wen model, Precision Engineering, Vol.43 (2016),
worse than the other two input frequencies. It is shown in pp.299-307.
[12] S. Xiao and Y. Li, ʠ Modeling and high dynamic compensating the
Fig. 21 that the actuator response for 23Hz input has a unique rate-dependent hysteresis of piezoelectric actuators via novel modified
shape. Recently Li, et al.[17] has reported that the basic inverse Preisach model ʡ, IEEE Transaction on Control Systems
input frequency and its 3rd order harmonic would affect the Technology, vol. 21, no. 5, 2013, pp. 1549-1557.
[13] T. Matsuo, Y. Osaka and M. Shimasaki, ʠ Eddy-current analysis
behavior of the piezoelectric actuator. It might be feasible using vector hysteresis models with play and stop hysterons ʡ, IEEE
to presume that the unique shape of the actuator response is Transactions on magnetics, vol. 36, no. 4, pp. 1172-1177, 2000.
dominated by the 3rd harmonic at the input frequency around [14] J. song and A. D. Kiureghian, Generalized Bouc-Wen model for
Highly asymmetric Hysteresis, Journal of engineering mechanics,
23Hz. The deviation of phase in Fig.13 and Fig. 16 might Vol.132, No.6(2006), pp610-618.
also be affected greatly by the 3rd harmonic at the input [15] M.Zhou and J. Wang, Research on Hysteresis of Piezoceramic Ac-
frequency around 23Hz. tuator Based on the Duhem model, The Scientific World Journal,
Vol.2013(2013).
[16] Ismail, M., Ikhouane, F. and Rodellar, J., The Hysteresis Bouc-Wen
V. C ONCLUSION Model, a Survey, Archives of Computational Methods in Engneering,
Vol.16, Issue2(2009), pp.161-188.
This paper presents the new extended Bouc-Wen model [17] C. X. Li, G. Y. Gu, L. M. Zhu and C. Y. Su, Odd-harmonic repetitive
which forms the asymmetrical hysteresis loop and exhibits control for high-speed raster scanning of piezo-actuated nanoposi-
the rate-dependent behavior as well to capture the hysteresis tioning stages with hysteresis nonlinearity, Sensor and Actuators A,
Vol.244(2016), pp.95-105.
of the real-world bimorph type piezoelectric actuator used in [18] Adachi, S.,Basis of system identification, Tokyo Denki University
this study. The proposed extended Bouc-Wen model is syn- Press(2009) (in Japanese).



You might also like