Chinese

Journal of
Aeronautics
Chinese Journal of Aeronautics 23(2010) 240-245 www.elsevier.com/locate/cja

Compound Control for Hydraulic Flight Motion Simulator

Wang Benyonga,*, Dong Yanliangb, Zhao Kedingb
a
School of Mechanical Engineering, Heilongjiang Institute of Science and Technology, Harbin 150027, China
b
School of Mechatronics Engineering, Harbin Institute of Technology, Harbin 150001, China

Received 20 February 2009; accepted 30 April 2009

Abstract
The design of a compound control is presented for the servo system of hydraulic flight motion simulator, which suffers from
highly nonlinear dynamics, large parameter time-variation and severe load coupling among channels. The compound control is
composed of a robust feedback controller and a feedforward compensator. The design aim is to achieve high tracking perform-
ance even in the presence of considerable uncertainty, external disturbance and load coupling among channels. Toward this aim
the feedback controller for rejecting perturbation and disturbance is designed by using μ synthesis optimization technique and
the feedforward compensator for compensating time lag of dynamic system is established based on the basic idea of zero phase
error tracking. To validate the proposed control strategy, simulations and experiments are implemented, and show that the result-
ing system is highly robust against model perturbation and possesses excellent capability of suppressing the load coupling and
improving the tracking performance.

Keywords: hydraulic flight motion simulator; system identification; compound control; μ synthesis; feedforward compensation

1. Introduction1 Although HFMS has many distinct advantages, the

Hydraulic flight motion simulator (HFMS) is a
three-degree-of-freedom complex high-accuracy me-
chanism as shown in Fig.1. HFMS has been widely
used in flight motion simulation of various aircraft
because it possesses greater power to weight ratio,
higher stiffness and less motion error than its electrical
rival[1].
1, 8, 11, 15—Servo motor; 2, 7, 10, 14—Angular sensor;
3, 6, 9, 13—Electro-hydraulic servo valve; 4—Middle gimbal;
5—Inner gimbal; 12—Outer gimbal; 16—Base
Fig.1 Schematic structure of a HFMS.
*Corresponding author. Tel.: +86-451-88036252.
E-mail address: wangbenyong2000@163.com
Foundation item: National “211 Project” Foundation of China
1000-9361/$ - see front matter © 2010 Elsevier Ltd. All rights reserved.
doi: 10.1016/S1000-9361(09)60211-9
characteristics of highly nonlinear dynamics, large par-
ameter time-variation and severe load coupling distur-
bance among channels[2] make various controller de-
sign techniques, such as proportion integral derivative
(PID), lead-lag compensation, adaptive control[3], slid-
ing model control[4] and others, not be suitable for such
high-accuracy plant. The μ synthesis optimization tec-
hnique[5] can excellently solve this problem due to its
extremely effective robustness and enables the three-
channel controllers of HFMS to be designed separately.
Considering the characteristics of HFMS and the re-
quirement for high tracking performance even in the
presence of considerable uncertainty, external distur-
bance and load coupling among channels, a compound
control for HFMS electro-hydraulic servo system is
presented in this article. The compound control con-
sists of a robust feedback controller and a feedforward
compensator. The feedback controller for attenuating
perturbation and disturbance is designed by using μ
synthesis optimization technique. The feedforward
compensator for compensating time lag of dynamic
system is established based on the basic idea of zero
phase error tracking. The simulation and experiment
results show that the proposed compound control for
HFMS servo system is highly robust against large per-
turbation and extremely effective for suppressing the
No.2 Wang Benyong et al. / Chinese Journal of Aeronautics 23(2010) 240-245 · 241 ·

severe load-coupling disturbance among channels and

improving system trajectory tracking performance.

2. Modeling

The single channel system model for HFMS can be

written as[6]
Kv
G = G0 + Wm Δg = + Wm Δg (1)
§s 2
2ξ h ·
s¨ 2 + s + 1¸
© ωh ωh ¹
where G is the real system model, G0 the nominal Fig.3 Output signals of inner gimbal single channel system.
model, Ȧh the hydraulic undamped natural frequency,
ȟh the hydraulic damping ratio, Kv the open-loop gain,
s the Laplace operator, Wm Δg the additive error of sys-
tem model, Wm the weighting function reflecting the
amount of uncertainty in the model and Δg denotes the
normalized uncertainty satisfying ||Δg||’<1.
The modeling is accomplished by using system
identification[7]. To collect the sufficiently informative
data for model identification, the persistently exciting
input is chosen to be a pseudo-random binary signal
(PRBS) which has the clock time of 0.009 s, the period
length of 2 046 and shift between −0.5 V and 0.5 V. A
part of the PRBS is shown in Fig.2. The identification Fig.4 Examination of inner gimbal system model.
procedure is illustrated with the HFMS inner gimbal
servo system. 2.2. Estimation of model uncertainty

Here the additive error Wm Δg, namely, model un-

Fig.2 A part of PRBS for system identification.
2.1. Parameter identification of nominal model
certainty, is modeled by model error modeling tech-
nique based on prediction error algorithm[8]. The dy-
namic model of Wm Δg can be expressed by
B C
A( y − G0u ) = u+ v (2)
F D
where y is the system output, u the system input, v the
noise, y-G0 u the residual and A, B, C, D, F are the
polynomials with the different orders na, nb, nc, nd, nf.
Start with the design of the model error model with
a low order model. Finally, for the parameter set na=1,
nb=8, nc=5, nd =5, nf =9, we receive an unfalsified er-
ror model with tight bounds, nearly over all frequen-
cies, as shown in Fig.5.

As the first step, the above PRBS is input to the in-

ner gimbal single channel system of HFMS, and the
output data are recorded and preprocessed as shown in
Fig.3. Then the preprocessed data are split into two
halves, the first to be used for estimation, and the sec-
ond one for validation. Based on the estimation data
the best nominal model can be computed with predic-
tion error identification algorithm. The obtained iden-
tification parameters for inner gimbal servo system are
Ȧh = 220 rad/s, ȟh = 0.27 and Kv = 30.3 (°)/s/V. In order
to validate the identification results, the simulation
output and experiment output are compared based on
the validation data, and shown in Fig.4. The fit is
75.14%.
· 242 · Wang Benyong et al. / Chinese Journal of Aeronautics 23(2010) 240-245 No.2

output magnitude of controller K, is selected to be 0.1,

i.e. Ws=0.1, because the saturation value of the control
input voltage to the amplifier is ±10 V. The weighting
function Wp is used to reflect the relative importance of
various frequency ranges for which performance is
desired, and determines the tracking performance
boundary of control system. According to the per-
formance requirement for the HFMS servo system and
the principle for selection of performance weighting
function[9], choose Wp to be
40( s /170 + 1)
Wp = (4)
s
In addition, in Fig.6, e is the tracking error, e1 and e2
are the evaluated errors, u is the control signal, și the
Fig.5 Model error modeling for inner gimbal servo system. command signal, șm the angular position of motor
shaft and d, caused by coupling torque, friction torque
The over-bounding weight Wm can be expressed by and others, the disturbance at the plant input.
20(0.2s + 1)
Wm = (3)
27 s + 1
The identification procedures for middle gimbal and
outer gimbal are similar to that for the inner gimbal
just introduced above, and are omitted here.

3. Robust Compound Control

3.1. Design of robust feedback controller

Fig.6 Closed-loop interconnection structure of inner gim-
In order to attenuate the effect of model uncertainty bal single channel system.
and external disturbance on the trajectory tracking
performance of HFMS servo system, here we intro- Transform the interconnection structure shown in
duce robust μ synthesis optimization technique for Fig.6 into the basic framework of μ synthesis for the
feedback controller design. Compared with the classi- inner gimbal single channel system, namely M-ǻ struc-
cal design approaches, the μ synthesis technique pro- ture, shown in Fig.7, where Δp is the fictitious uncer-
vides a flexible systematical framework for feedback tainty performance block, the dashed box P represents
controller design and can make the controlled system the transfer function of four-input and four-output
simultaneously achieve robust stability and robust per- open-loop interconnection, and the dashed box M repre-
formance. sents the transfer function from [w d și u]T to [z
Before designing robust feedback controller via μ e1 e2 e]T and can be expressed as the lower linear
synthesis, describe the performance requirement, pa- fractional transformation about P and K, i.e. M =
rametric uncertainty, high-frequency un-modeled un- Fl (P,K).
certainty and others with weighting functions, and then
construct the closed-loop interconnection system with
the weighting functions and nominal model. An inter-
connection diagram for the HFMS inner gimbal
closed-loop system, which includes the feedback
structure of the plant and controller, and the elements
associated with the uncertainty models and perform-
ance objectives, is shown in Fig.6. The dashed box
represents the true plant, with associated transfer func-
tion G. Inside the box are the nominal model of the
inner gimbal system dynamics, G0, and two elements,
Wm and Δg, which parameterize the uncertainty in the
model. The transfer functions G0 and Wm have been
obtained with the aforementioned identification pro-
cedure. The transfer function Δg is assumed to be sta-
ble and unknown, except for the norm condition Fig.7 M-ǻ structure of μ synthesis for inner gimbal single
||Δg||’<1. The weighting function Ws, used to limit the channel system.
No.2 Wang Benyong et al. / Chinese Journal of Aeronautics 23(2010) 240-245
The design objective of the robust feedback control
for HFMS servo system is to design a stabilizing con-
troller K such that for all stable perturbations Δg, with
||Δg||’<1, the perturbed closed-loop system remains
stable, and the perturbed weighted performance trans-
fer function has H’ norm not more than 1 for all such
perturbations. According to main loop theorem and
robust performance theorem[10], this robust design ob-
jective exactly fits in with the structured singular value
μ framework, and can be solved with D-K iteration[11]
program in MATLAB μ analysis and synthesis tool-
box[12]. After two D-K iterations, a controller of order
7 is found with the maximum of μ less than 1. The μ
plot is shown in Fig.8. The maximum upper bound of
μ is 0.985, which guarantees the stability and per-
formance robustness of the closed-loop system with
such a control.
Fig.8 μ curve of inner gimbal single channel system.
From the point of implementation in engineering, a
controller order reduction is definitely needed. Here is
used a balanced realization model reduction meth-
od[13-15], which results in an iterative procedure. Start
with reducing the controller order to the first order and
then carry out the μ test. If the μ test is satisfied, then
stop, otherwise reduce the controller to the second
order and begin another run. The process goes on until
the μ test is satisfied. By such an order reduction pro-
cedure, a controller of third order is obtained as fol-
lows:
3.502 6( s + 17.36)( s 2 + 124.1s + 48 570)
K=
s ( s 2 + 249.4 s + 71 250)
In order to examine the above order reduced D-K
iteration robust controller, simulations have been car-
ried out. Fig.9 shows the dynamic responses to the
same step disturbance at the plant input.
In Fig.9, the curve 1 and curve 2 denote the time
responses of the inner gimbal system with the order
reduced D-K iteration robust controller at Δg =0 and
Δg = 0.9, respectively, and the curves 3 and 4 those
with the well-tuned PID controller at the same Δg as
the above. From the simulation results, it is clear that,
in the presence of model perturbation, the proposed
robust controller can keep the responses of the con-
· 243 ·
trolled plant almost unchanged, and suppress the ex-
ternal disturbance more effectively than the well-tuned
PID controller.
Fig.9 Step responses of disturbance attenuation of inner
gimbal single channel system.
3.2. Design of feedforward compensator
The block diagram of compound control is shown
in Fig.10, where Cff is the feedforward compensator
and Δa the additive uncertainty which satisfies |Δa|=
|Wm Δg| < |Wm|. The error transfer function ĭe from și
to e is
1 − GCff
Φe = (6)
1 + GK
It is clear that when Cff = 1/G, the above error trans-
fer function ĭe is 0, which means that with this in-
verse-feedforward, exact output tracking can be ob-
tained. In practice, the exact plant G may not be
known due to modeling errors. Therefore, we replace
the true plant G with nominal model G0 to design the
feedforward compensator, namely, Cff = 1/G0. Substi-
tuting G = G0+ǻa and Cff = 1/G0 into Eq.(6), we obtain
Δa ( jω )G0−1 ( jω ) Δ ( jω )
Φe ( jω ) = = a Φe0 ( jω ) (7)
1 + G ( jω ) K ( jω ) G0 ( jω )
where Φe0 represents the transfer function from și to e
with the feedforward compensator inactive, i.e., Cff = 0.
From Eq.(7), it can be seen that |ĭe(jȦ)| < |ĭe0(jȦ)|
is obtained when |ǻa(jȦ)| < |G0(jȦ)| or |Wm(jȦ)| <
|G0(jȦ)| holds. This means that only if the size of the
(5)
plant uncertainty is less than the size of the nominal
model, the feedforward compensator based on the in-
verse nominal model can correct tracking error. Gen-
erally, the ratio |ǻa(jȦ)|/|G0(jȦ)| is more than 1 at the
high frequency. We can shut off the effect of high fre-
quency uncertainty on the tracking performance by
introducing non-causal zero phase low-pass filter.
Fig.10 Block diagram of compound control.
· 244 · Wang Benyong et al. / Chinese Journal of Aeronautics 23(2010) 240-245 No.2

The inverse-feedforward Cff = 1/G0, in fact, is the
design method based on the zero-pole cancellation and
phase compensation. As the realization problem of
digital control is concerned, the system model might
have uncancelable zeros which cause unacceptable
feedforward compensation. Therefore, we applies the
basic idea of zero phase error tracking, which can
guarantee the reliability of digital control[16-17], to this
inverse-feedforward compensator Cff. Let z transfer fu-
nction of nominal plant be
z − d B ( z −1 )
G0 ( z −1 ) =
A( z −1 )
where z−d denotes the d-step pure delay, and B(z−1) and
A(z−1) are the polynomials about z−1. Decompose B(z−1)
into two parts as
B( z −1 ) = B a ( z −1 ) B u ( z −1 )
where B (z ) contains cancelable zeros of B(z−1) and
a −1
The experimental results of the inner gimbal single
channel are shown in Fig.12, where the curve 1 de-
notes the desired trajectory, the curve 2 and curve 3
denote the motion trajectories controlled by the pro-
posed approach and well-tuned PID controller, respec-
tively, and the curves 4 and 5 the desired error band. It
is seen that the proposed approach possesses better
trajectory tracking performance.
The results from the experiments on HFMS load
coupling suppression are shown in Fig.13, where the
curves represent the influences of load coupling on the
(8) inner and middle gimbals when HFMS is controlled by
the proposed approach and the well-tuned PID con-
troller, respectively. From them, it is clear that the
proposed approach exhibits stronger capability of at-
tenuating the load coupling among channels.
(9)

Bu(z−1) uncancelable zeros of B(z−1). According to

Ref.[16], we can obtain the inverse-feedforward
compensator based on zero phase error tracking as
follows: z d A( z −1 ) B u ( z )
Cff = a −1 u (10)
B ( z )( B (1)) 2
where (Bu(1))2 is a scaling factor.
The final feedforward compensator is the cascade of
the inverse-feedforward compensator based on zero
phase error tracking and non-causal zero phase
low-pass filter based on |Wm(jȦ)| < |G0(jȦ)|.
The procedure to design the compound control for
HFMS inner gimbal servo system has been presented.
The control design procedures for the other two gim-
bals are similar and omitted here.

4. Experimental

This section is devoted to the experiments on single

channel trajectory tracking and load coupling attenua-
tion for evaluating the proposed control strategy.
Fig.11 shows the photo of the flight motion simulator
with the self-developed motor in the laboratory. The
servo valve in the outer gimbal is of MG35 type with a
170 L/min rated flow, the servo valve in the middle
gimbal is of YFW06A066AK type with a 66 L/min
rated flow, and the servo valve in the inner gimbal is of Fig.12 Experimental results of inner gimbal single channel.
FF102 type with a 30 L/min rated flow. The experi-
ments run at the oil source pressure of 12 MPa and the
oil temperature of 25-35 °C.

Fig.11 Photo of HFMS test prototype.

No.2 Wang Benyong et al. / Chinese Journal of Aeronautics 23(2010) 240-245
Fig.13 Experimental results on HFMS load coupling sup-
pression.
5. Conclusions
In order to achieve high control performance, in this
article we first accurately establish the nominal model
and model uncertainty bound for HFMS servo system
by system identification, and then propose a robust
compound control which is composed of a robust
feedback control with μ synthesis and an in-
verse-feedforward compensation. The results from the
simulations and experiments have evidenced that the
proposed control strategy not only exhibits better ca-
pability of trajectory tracking and attenuating load
coupling among channels, but also appears insensitive
to external disturbance and strongly robust in the pres-
ence of a large model perturbation. Therefore, it stands
to reason that the proposed control method completely
fits for HFMS hydraulic servo system.
· 245 ·
References
[1] Liu C. Research on low-speed capability of hydraulic
three-axis simulator and its control strategy. PhD thesis,
Harbin Institute of Technology, 2003. [in Chinese]
[2] Wang P. Theory analysis and experiment research of
3-axis angle vibration hydraulic turntable. MS thesis,
Harbin Institute of Technology, 2006. [in Chinese]
[3] Xie Y, Mahinda V, Tseng K J, et al. Modeling and ro-
bust adaptive control of a 3-axis motion simulator.
Proceedings of IEEE Industry Applications Conference.
2001; 1: 553-560.
[4] Liu J, Sun F. Nominal model-based sliding mode con-
trol with backstepping for 3-axis flight table. Chinese
Journal of Aeronautics 2006; 19(1): 65-71.
[5] Mei S, Shen T, Liu K. Modern robust control theory
and application. 1st ed. Beijing: Tsinghua University
Press, 2003; 108-115. [in Chinese]
[6] Wang B. Research on three-axis hydraulic simulator
control system based on H’ control theory. PhD thesis,
Harbin Institute of Technology, 2008. [in Chinese]
[7] Ljung L. System identification theory for the user. 1st
ed. Beijing: Tsinghua University Press, 2002; 408-457.
[8] Esmaeilsabzali H, Montazeri A, Poshtan J, et al. Ro-
bust identification of lightly damped flexible beam us-
ing set-membership and model error modeling tech-
niques. Proceeding of the 2006 IEEE international
conference on control applications. 2006; 2945-2949.
[9] Franchek M A. Slecting the performance weights for
the μ and H’ synthesis methods for SISO regulating
systems. ASME Journal of Dynamic Systems Meas-
urement and Control 1996; 118(3): 126-131.
[10] Packard A. The complex structured singular value.
Automatica 1993; 29(1): 71-109.
[11] Stein G, Doyle J C. Beyond singular values and loop
shapes. AIAA Journal of Guidance and Control 1991;
14(1): 5-16.
[12] Balas G J, Doyle J C, Glover K, et al. μ-analysis and
synthesis toolbox. 1st ed. Natick: The MathWorks Inc,
2001; 601-639.
[13] Glover K. All optimal Hankel-norm approximations of
linear multivariable systems and their L’ error bounds.
International Journal of Control 1984; 39: 1115-1193.
[14] Moore B C. Principal components analysis in linear
systems: Controllability, observability and model re-
duction. IEEE Transactions on Automatic Control 1981;
26(1): 17-31.
[15] Zhou K, Doyle J C, Glover K. Robust and optimal
control. New Jersey: Prentice Hall, 1996; 155-172.
[16] Tomizuka M. Zero phase error tracking algorithm for
digital control. ASME Journal of Dynamic Systems
Measurement and Control 1987; 109(1): 65-68.
[17] Tsao T C, Tomizuka M. Adaptive zero phase error
tracking algorithm for digital control. ASME Journal of
Dynamic Systems Measurement and Control 1987;
109(4): 349-354.
Biography:
Wang Benyong Born in 1968, he received Ph.D. degree
from Harbin Institute of Technology in 2008. Now, he is a
teacher in School of Mechanical Engineering, Heilongjiang
Institute of Science and Technology. His main research in-
terest is hydraulic servo control.
E-mail: wangbenyong2000@163.com