Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/257725388

GPC Controller Design for an Intelligent Pneumatic Actuator

Article in Procedia Engineering · December 2012


DOI: 10.1016/j.proeng.2012.07.226

CITATIONS READS

16 671

5 authors, including:

Ahmad Athif Mohd Faudzi Koichi Suzumori


Universiti Teknologi Malaysia Tokyo Institute of Technology
200 PUBLICATIONS 1,790 CITATIONS 759 PUBLICATIONS 6,333 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Ahmad Athif Mohd Faudzi on 10 October 2014.

The user has requested enhancement of the downloaded file.


Available online at www.sciencedirect.com

Procedia Engineering 41 (2012) 657 – 663

GPC Controller design for an Intelligent Pneumatic Actuator


Ahmad 'Athif Mohd Faudzi a,*, Nu’man Din Mustafaa, Khairuddin bin Osmanb, M. Asyraf
Azmana, Koichi Suzumori c
a
Department of Mechatronics and Robotics Engineering, Faculty of Electrical Engineering, Universiti Teknologi Malaysia, 81310 Skudai, Malaysia
b
Department of Industrial Electronics, Faculty of Electrical and Electronics, Universiti Teknikal Malaysia Melaka, Hang Tuah J aya, 76100 Durian
Tunggal, Melaka, Malaysia
c
Graduate School of Natural Science and Technology, Okayama University, Okayama, Japan

Abstract

This paper proposed a Generalized Predictive Controller (GPC) to the Intelligent Pneumatic Actuator (IPA) in order to analyze the control
approach and the performance of the actuator. First, an estimation model of the IPA is obtained by using a Reaction Curve M ethod and
the controller is designed based on the model obtained. By using the exist ing function proposed by the previous researcher, the
implementation and tuning process of the GPC become easier. The parameter needed for the function can be obtained from the model and
from simple calculation as presented in this paper. The performance of the designed GPC controller is tested to the actuator position
control and the results shows good accuracy and fast response. Lastly, Proportional-Integral-Derivative controller (PID) is used to validate
the result.

© 2012 The Authors. Published by Elsevier Ltd. Select ion and/or peer-review under responsibility of the Centre o f
Humanoid Robots and Bio-Sensor (HuRoBs), Faculty of Mechanical Engineering, Universiti Teknologi MARA.

Keywords: Intelligent Pneumatic Actuator (IPA), GPC controller, PID controller.

1. Introduction

Pneumatic actuator is a device that converts energy in the form of co mpressed air into motion. Lately, pneumat ic
actuators have been largely used in numerous control applications in the industries. This is main ly because these actuators
have advantages in their high power-to-weight ratio, relat ively low cost, easy to maintain, lighter, an d have simple structure
compare to other actuator that available in the market [1]. Despite of all the advantages compared to electrical actuator used
in robots and mach ines, it is d ifficult to control. This is due to the nonlinear factor involved such the nonlinearity o f the
valve, compliance variation and generating force.
GPC method was proposed by Clarke et al. [2] and it is widely used nowadays in various application in the industry.
This is mainly due to the ability of the controller to control a p lant with various parameters, dead time and higher model
order [3]. It is also recorded that GPC was successfully imp lemented to plants that have non-minimu m phase, unstable open
loop, or model that are over parameterized or under parameterized by the estimation scheme [4]. Lotfi Chikh et al. in [5]
used Generalized Predict ive force control fo r electropneumatic cylinder. Fro m the experiment al result, it shows that GPC
was a good control approach and have high performance in term of capacity of tracking long duration static forces of high
amp litudes. GPC controller also have been recorded for its quick response and accurate tracking based on research done by
[6]. The capability of the GPC to control the pressure with system that used estimate parameter also have been recorded in
[7].
In this paper, the aim is to design a controller for the Intelligent Pneumatic Actuator by using GPC. MATLAB –
Simu link is used as the platform to develop the controller. It is hoped that GPC could eliminate the overshoot for open loop
response and yield a good accuracy for pressure control. PID is used as reference because of its simp licity, easy to use and

* Corresponding author. Tel.: +607-5535291; fax: +607-5566272


E-mail address: athif@fke.utm.my

1877-7058 © 2012 Published by Elsevier Ltd.


doi:10.1016/j.proeng.2012.07.226
658 Ahmad ‘Athif Mohd Faudzi et al. / Procedia Engineering 41 (2012) 657 – 663

do not require the plant model to perform the controller [8, 9]. However, PID parameter is not adaptive and requires some
optimal control method to increase its precision [10]. In this research, the parameter for PID is obtained by using MATLAB
toolbox provided in the Simu lin k. The result for the GPC controller and PID are presented and discussed in the result and
discussion section. The flow of the paper starts with introduction, Intelligent Pneumatic Actuator as plant, model
estimation, GPC controller design, result and discussion and conclusion section.

2. Intelligent Pneumatic Actuator as Plant

Intelligent Pneumatic Actuator (IPA) as shown in Fig. 1 was developed by A.A M. Faudzi et al. for research purposes
[11, 12]. The application proposed was a seating apparatus called Pneu matic Actuator Seat ing System (PASS) where the
system can imitate o ffice chair, stool and other chair shapes with stiffness and damping characteristics [13]. The pneumat ic
actuator consists of five elements which are Programmable System on Ch ip (PSoC) as the controller, laser strip code with
0.169mm accuracy for the position feedback, optical encoder (A EDR – 8300) fo r read ing the laser strip code, pressure
sensor (KOGANEI: PSU-EM-S) for reading the pressure inside the chamber and valves (KOGA NEI: EB10ES1 -PS-6W ) fo r
controlling the injection of air into the chamber. The body part of the actuator is a linear double acting cylinder (KOGA NEI
– HA Twinport Cy linders) with two air inlets and one exhaust outlet. The control algorith m is programmed direct ly into the
microcontroller (PSo C) and it will control the output position based on the input fro m the sensors. The actuator has 200mm
stroke and force up to 100N.

Laser strip code

PSoC Pressure
microcontroller sensor
Optical
encoder Valve

Fig. 1: Intelligent Pneumatic Actuator

3. Model Estimation

Most equipment in the industries is very high order to model and difficult to control. The reason is that most industrial
processes consists of many dynamic elements, usually first order, therefore the full model is of an order equal to the nu mber
of elements [14]. According to [15] it is possible to estimate higher order model processes with first order process combined
with dead t ime. Therefore, the p lant model is estimated by using Reaction Curve Method. Based on [14], the discrete
transfer function for first-order type has the form

bz 1  d
G( z 1 ) z (1)
1  az 1

where

T
 Wd (1)
a e W
d b K (1  a)
T
Ahmad ‘Athif Mohd Faudzi et al. / Procedia Engineering 41 (2012) 657 – 663 659

W 1.5(t1  t2 ) (1)

1 (2)
Wd 1.5(t1  t2 )
3

T is sampling time, t 1 is when the response reach 28.3% of the final value and t 2 is when the response reach 63.2% of the
final value. Thus the model obtained is,

0.0491z 1 1 (3)
G( z 1 ) z
1  0.8261z 1

4. GPC Controller Design

GPC algorith m is based on Controlled Auto-Regressive Integrated Moving Average (CARIMA) model and can be
described after linearizat ion and considering operation around a particular set point of a SISO plant can be described as
below:

e(t ) (4)
A( z 1 ) y(t ) z  d B( z 1 )u (t  1)  C ( z 1 )
'

where, u (t ) and y (t ) are the control input and output sequence of the plant, ' 1  z 1 , e(t ) is zero mean wh ite noise and d
is dead time of the system. Meanwhile A, B and C are polynomial in the backward shift operator z 1 as following:

A( z 1 ) 1  a1z 1  a2 z 2  ...  ana z  na (1)

B( z 1 ) b0  b1z 1  b2 z 2  ...  bnb z  nb (1)

C ( z 1 ) 1  c1z 1  a2 z 2  ...  cnc z  nc (5)

For simplicity, C ( z 1 ) is assumed to be 1, thus the following equation is obtained:

y(t  1) (1  a) y(t )  ay(t  1)  b'(t  d )  H (t  1) (6)

GPC algorithm consists of applying control sequence in order to minimize a multistage cost function as in Equation (7) ,[2].

¦ G ( j )[ yˆ (t  j | t )  w(t  j )]2  ¦ j 1 O ( j )['u(t  j  1)]2


N2 Nu
J ( N1, N2 , N3 ) (7)
j N1

where u is the control input, N u is the control horizon, w is the reference value, ŷ is the plant predict ion on data up to t ime
t, N 1 is the minimu m costing horizon with N1 d  1 , N 2 is the maximu m costing horizon with N 2 d  N and λ is the
control weighting [16].

By using Simu link, the value for yˆ (t  d  j  1 | t ) and y(t  d  j  2 | t ) can be obtained easily, thus making the best
expected value for yˆ (t  d  j ) according to [14] is given by:

yˆ (t  d  j | t ) (1  a) y(t  d  j  1 | t )  ayˆ (t  d  j  2)  b'u(t  j  1) (8)

Equation (8) then is applied recursively j 1,2..., i and thus,

yˆ (t  d  i | t ) Gi ( z 1 ) yˆ (t  d | t )  Di ( z 1 )'u(t  i  1) (9)
660 Ahmad ‘Athif Mohd Faudzi et al. / Procedia Engineering 41 (2012) 657 – 663

where Gi z 1 is of degree 1 and Di ( z 1 ) is of degree i  1 . Minimizing J ( N1, N2 , Nu ) with respect to


'u(t ), 'u(t  1)  'u(t  N  1) leads to
Mu Py  Rw (10)

where

u [u(t ) 'u(t  1)  'u(t  N  1)]T (1)

y [ yˆ (t  d | t ) yˆ (t  d  1 | t )]T (1)

w [w(t  d  1) w(t  d  2)  w(t  d  N )]T (1)

M and R are matrices of dimension N x N, P o f d imension N x 2. When the future set point are unknown, w(t  d  i) is
equal to current references, r (t ). Thus, the control increment is as Equation (11) , the equation for I y1 , I y 2 and I r1 according
to the control scheme in Fig. 2 can be refer to [16].

'u(t ) I y1 yˆ (t  d | t )  I y 2 yˆ (t  d  1 | t )  I r1r (t ) (11)

Fig. 2: GPC Control Scheme

Fig. 3: GPC block diagram in Simulink


Ahmad ‘Athif Mohd Faudzi et al. / Procedia Engineering 41 (2012) 657 – 663 661

5. Result and Discussion

Choosing N1 5 , N 2 6 , N u 6 , O 0.95 and by using the plant model obtained in Equation (2), then the GPC
algorith m is applied in the MATLAB for simulat ion by using Simu lin k as shown in Fig. 3. The simu lation is constructed
based on the control scheme show in Fig. 2. The experimental results for the responses are shown in Fig. 4 and 5 below.

100
100
Position(mm)

GPC
99.5 Input
50
9.085 9.09 9.095

0
0 5 10 15 20 25 30
500
Control Input(%)

Control Input
400

300
0 5 10 15 20 25 30
Time(s)

Fig. 4: Experimental result with step input

101

100
Position(mm)

100
GPC
Input
0
99
9 9.5 10

-100
0 5 10 15 20 25 30

1000
Control Input(%)

Control Input
0

-1000
0 5 10 15 20 25 30
Time(s)

Fig. 5: Experimental result with square wave input

The result show a good performance with time response, TR =1.04s, settling time, TS =2s and dead time=0.1s for the for
step input. The error fo r the step input as show in Fig. 4 is 0.15 and error for the square wave input as shown in Fig. 5 is
0.16. A lthough this controller used first order model as plant, the result showed that the controller manages to follo w the
input given with accuracy up to 99.8%. In order to validate the controller, PID is used as reference and the comparison by
using square wave input is shown Fig. 6.
662 Ahmad ‘Athif Mohd Faudzi et al. / Procedia Engineering 41 (2012) 657 – 663

101

100
Position(mm)
100
GPC
0 PID
99
9 9.5 10 Input
-100
0 5 10 15 20 25 30
Time(s)
Fig. 6: Comparison between PID and GPC

T able 1: Comparison between PID controller and GPC controller analysis results by using step input

Analysis PID controller GPC controller

Percent Overshoot (%OS) 0% 0%


Dead T ime (TU) 0.1s 0.1s
Peak T ime (TP) = Settling Time (TS ) 2.5s 2.0s
Rise T ime (TR) 1.2s 1.0s
Percent Steady State error (%ess) 0.001% 0.15%

In table 1, the results for both controllers show no percentage of overshoot (%OS) and this is mainly because the plant
model used is first order. The results also show that both controllers have same dead time, Tv as the plant model used is
similar. The GPC controller show better performance co mpared to PID with faster settling time, TS and rise time, TR . The
percentage steady error of GPC controller is larger co mpared to PID controller but it is acceptable because still in the range
of 2% of the final value. In Fig. 4 and Fig. 5 the GPC controller shows good response for step and square wave input. In
order to imp lement other type of input such as sinus and random, the controller variable, N 1 , N 2 , N u and O need to be
tuned.
PID controller is a linear controller and does not taking into account the plant or model parameter. Thus the performance
of PID controllers in nonlinear systems such as pneumatic actuator may vary and sometimes unstable. In the PID loop, the
control signal is calculated fro m the erro r by cancelling out the current erro r directly by using gain (Proportional), the
amount of time the error has continued uncorrected (Integral), and anticipate the future error fro m the rate of change of the
error over time (Derivative). The error so metimes cannot be eliminated and as the result the error will grow bigger and the
system will become unstable. Usually PID controller will give a fast and good performance, however in this case the plant
used is pneumatic actuator and sometimes the nonlinearities and the constraint (i.e. the frict ion, the natural weight and the
limitat ion of the valve) involved will make the plant unstable. GPC on the other hand, used past control and output signal to
estimate the future control. In the GPC algorithm the plant parameter involved can be obtained fro m the estimated model as
discussed in previous section. GPC algorithm has the capability to con trol plant with unstable open loop and constraint.

6. Conclusion

In this paper the controller design using GPC and its performance have been recorded and analysed. The PID controller
is used to validate the GPC controller performance. Both controllers have been successfully simulated in the MATLAB by
using same model in Equation (3). The result is analysed and compared based on its percentage of overshoot (%OS), dead
time (T ), peak time (T ), settling time (T ), rise time (T ) and percent steady state error (%ess ). A ll criteria except the percent
U P S R

steady state error show that the GPC controller has the capability to control the plant smoothly and with faster response
compared to PID. In the future this research will be used as a comparison in the validation process for another controller
and will be further improved in order to obtained better response.

Acknowledgements

The authors would like to thank Universit i Teknologi Malaysia (UTM) by UTM -NAS Grant No R.J130000.7723.4P008,
Ministry of Higher Education (MOHE) Malaysia and Okayama University for their support.
Ahmad ‘Athif Mohd Faudzi et al. / Procedia Engineering 41 (2012) 657 – 663 663

References

[1] A. Gentile, N. I. Giannoccaro and G. Reina, 2002. "Experimental tests on position control of a pneumatic actuator using on/off solenoid valves," in
Industrial T echnology - IEEE ICIT '02. 2002 IEEE International Conference on, 2002, vol.1, p. 555 -559.
[2] D.W. Clarke, C.M., and P.S. T uffs., 1987. Generalized predictive control, Part I: Basic algorithm and Part II: Extensions and interpretations.,
Automatica 23, p. 132 - 160.
[3] Dajun, D., and L. Lixiong., 2007. “A Novel Implicit Generalized Predictive Control Algorithm” Control and Automation -ICCA 2007. IEEE
International Conference on. 2007,vol.1,p. 2716-2770.
[4] Zeybek, Z., and S. Çetinkaya., 2006. Generalized Delta Rule (GDR) algorithm with generalized predictive control (GPC) for optimum temperature
tracking of batch polymerization. Chemical Engineering Science 61(20), p. 6691-6700.
[5] Chikh, L.P., Philippe; Pierrot, Francois; Michelin, Micaël.,2010. “ A Generalized Predictive Force Controller for electropneumatic cylinders,” 8 th IFAC
Symposium on Nonlinear Control System (2010) .
[6] Qiang, S., and F. Liu., 2006. “The Direct Approach to Unified GPC Based on ARMAX/CARIMA/CARMA Model and Application for Pneum atic
Actuator Control,” Innovative Computing, Information and Control 2006, ICICIC '06. First International Conference on 2006, Vol.1,p.336-339.
[7] Chaewieang, P., K. Sirisantisamrit, and T. Thepmanee., 2008. “ Pressure control of pneumatic-pressure-load system using generalized predictive
controller,” Mechatronics and Automation - ICMA 2008. IEEE International Conference on. 2008, p.788-791
[8] Kiam Heong, A., G. Chong, and L. Yun., 2005. PID control system analysis, design, and technology. Control System s Technology, IEEE
T ransactionson 13(4), p. 559-576.
[9] Silva, G.J., A. Datta., and S.P. Bhattacharyya., 2002. New results on the synthesis of PID controllers, Automatic Control, IEEE Transactions on 47(2),
p. 241-252.
[10 Jun, W., and H. Jian., 2009. “Fuzzy PID control of a wearable rehabilitation robotic hand driven by pneumatic muscles,” Micro-NanoMechatronics
and Human Science-MHS 2009. International Symposium on. 2009,p.408-413.
[11] A.A.M.Faudzi., K. Suzumori, and S. Wakimoto., 2008. “ Distributed Physical Human Machine Interaction Using Intelligent Pneumatic Cylinders,”
Micro-Nano Mechatronics and Human Science- MHS 2008. International Symposium on. 2008,p.249-254.
[12] A.A.M. Faudzi, K. Suzumori., and S. Wakimoto,, January 2009. Development of an Intelligent Pneumatic Cylinder for Distribut ed Physical Human-
Machine Interaction Advanced Robotics, vol. 23, p. 203-225.
[13] Ahmad 'Athif Mohd Faudzi, 2010. "Development of Intelligent Pneumatic Actuators and T heir Applications to Physical Human -Mechine Interaction
System," Ph.D. thesis, T he Graduate School of Natural Science and Technology, Okayama Univ ersity, Japan.
[14] Bordons, C., and E.F. Camacho.,1998. “ A Generalized Predictive Controller for a Wide Class of Industrial P rocesses,” Control Systems Technology,
IEEE T ransactions on 6(3),p. 372-387.
[15] P.B Deshpande., and R.H.A.,1981. Elements of Computer Process Control, Charlotte,USA.
[16]Comacho, E.F., and Bordon, C., 1998. A generalized predictive controller for a wide class of industrial processes. IEEE Transactions on Robotics and
Automation, 1998. 6(3),p.372-387.

View publication stats

You might also like