Sunteți pe pagina 1din 6

J Control Theory Appl 2012 10 (3) 297302 DOI 10.

1007/s11768-012-1044-4

Design of internal model control based fractional order PID controller


T. VINOPRABA 1 , N. SIVAKUMARAN 1 , S. NARAYANAN 1 , T. K. RADHAKRISHNAN 2
1.Department of Instrumentation and Control Engineering, National Institute of Technology, Tiruchirappalli 620015, India; 2.Department of Chemical Engineering, National Institute of Technology, Tiruchirappalli 620015, India.

Abstract: This article presents a design of the internal model control (IMC) based single degree of freedom (SDF) fractional order (FO) PID controller with a desired bandwidth specication for a class of fractional order system (FOS). The drawbacks of the SDF FO-IMC are eliminated with the help of the two-degree of freedom (TDF) FO PID controller. The robust stability and robust performance of the designed controller are analyzed using an example. Keywords: Fractional order PID controller; Fractional order systems; Model based control; Robust control

1 Introduction
In recent years, the fractional order (FO) PID controller has gained much attention in the control community because it is more robust than the integer order (IO) controller due to the fractional powers of the integral and derivative s terms. Several tuning rules are available for FO PID controller for the IO process [16]. The optimization techniques such as genetic algorithm and the particle swarm optimization techniques are also used to obtain the FO PID parameters [78]. Real world processes are likely to be fractional, though the fractionality may be less [9]. Systems such as voltage-current relation of a semi-innite lossy RC line, the diffusion of heat into a semi-innite solid [9], heating furnace [10] and gas turbine [11] are of fractional order. For FOS, the FO controller parameters are tuned using gain margin and phase margin specications [1213]. Advanced FO controllers such as CRONE controller [14], fractional order lead compensator [15], self tuning regulator [16], model reference adaptive control [17], adaptive high gain controller [18], sliding mode controller [19], iterative learning control [20] have been implemented to improve the performance and robustness in the closed loop control systems. It has been proved that, the stabilizing set of the fractional order PID controllers is wider than the integer order PID controller [2122]. If the stabilizing set of the controller parameters is wider, then the controller is more robust i.e. it can accommodate more model uncertainties compared to integer order controller. In the last decade, the internal model control (IMC) based PID controller design has gained widespread acceptance in the control community because the controller can be easily designed by taking inverse of the model with a single tuning parameter namely the IMC lter time constant. The optimal value of lter time constant is determined by a trade-off between speed of response (small value of time constant) and robustness (large value of time constant). An IMC based PID controller has been designed for the desired bandwidth
Received 7 March 2011; revised 1 July 2011.

specication [23]. Increase in the bandwidth provides less attenuation in reference signal and faster response. An IMC based FO PID controller has been designed for the IO rst order plus time delay process [24] and a class of FOS [25]. In this article, an attempt is made to design IMC FO PID controller with bandwidth specication. In SDF FO-IMC, both tracking performance and disturbance (output) rejection performance cannot be achieved simultaneously which has been mathematically proved in Appendix B. Hence, TDF FO-IMC is designed which ensures good tracking performance and satisfactory disturbance rejection performance. The designed TDF FO-IMC is more robust than the IO-IMC because of fractional powers of integral and derivative s terms. The paper is organized as follows. In Section 2, the problem is formulated. The design of SDF FO-IMC and TDF FO-IMC are discussed in Sections 3 and 4. Simulation results are presented in Section 5. Section 6 describes the robustness analysis of the designed controller. Conclusions are drawn in Section 7.

2 Problem formulation
The basic structure of the IMC is shown in Fig. 1 where Gp is the process, Gm is the model of the process, Gi is the controller, r is the set point, d is the disturbance and y is the output of the process. Fig. 1 can be transformed into standard feedback structure as shown in Fig. 2. Gc is given by Gc (s) = Gi (s) . 1 Gm (s)Gi (s) (1)

The design of the IMC based IO-PID controller for the desired bandwidth specication has been reported by Morari and Zariou [26] and the same has been briey explained in Appendix A. In this paper, an attempt is made to design IMC based FO-PID controller with desired bandwidth specication.

c South China University of Technology and Academy of Mathematics and Systems Science, CAS and Springer-Verlag Berlin Heidelberg 2012

298

T.Vinopraba et al. / J Control Theory Appl 2012 10 (3) 297302

TDF FO-IMC

Fig. 1 Internal model control structure.

In a SDF FO-IMC, the lter time constants obtained for disturbance rejection performance and the tracking performance, with same bandwidth specication, are not same because of fractional power in the s term (proof is shown in Appendix B). To eliminate this drawback, a TDF FO-IMC is designed. 4.1 Design of TDF FO- IMC The IMC structure based on TDF FO controller is shown in Fig. 3.

Fig. 2 Standard feedback controller structure.

3 SDF FO-IMC
The FOS is assumed to be of the form K , (2) Gp FO (s) = as + bs + c where and are real coefcients. Let the fractional model of the process be K . (3) Gm FO (s) = as + bs + c Based on pole-zero cancellation, the controller Gi (s) is assumed to be as + bs + c Gi (s) = , (4) K(1 DF FO s + 1) where is of fractional order ranges from 0 << 2 and 1 DF FO is the lter time constant of the FOC. Substituting equations (3) and (4) in equation (1), Gc (s) is found to be as + bs + c Gc (s) = . (5) K1 DF FO s The general form of FO-PID controller is given by 1 GFOPID (s) = Kc (1 + + Td s ). (6) Ti s Comparing equations (5) and (6), the controller parameters are obtained as follows b Kc = , (7) K1 DF FO b Ti = , (8) c a Td = , (9) b = , (10) = . (11) From Fig. 2, the closed loop transfer function with respect to disturbance rejection is obtained and is given by y(s) 1 1 DF FO s = = . (12) d(s) 1 + Gc (s)Gp (s) 1 DF FO s + 1 The closed loop transfer function with respect to set point tracking is obtained and is given by y(s) Gc (s)Gp (s) 1 = = . (13) r(s) 1 + Gc (s)Gp (s) 1 DF FO s + 1 The lter time constant for the disturbance rejection and tracking performance is determined from the bandwidth specication.

Fig. 3 Two degree of freedom controller.

Here Gc1 (s) is designed based on disturbance rejection and Gc2 (s) is designed to shape the servo response. Choose Gc1 (s) to be as + bs + c , (14) Gc1 (s) = K1 DF FO s where the lter time constant 1 DF FO is obtained from equation (12). Choose Gc2 (s) to be 1 + 1 DF FO s , (15) Gc2 (s) = 1 + 2 DF FO s where the lter time constant 2 DF FO is obtained from equation (13) . 4.2 Robust stability analysis The control system design is usually based on the approximate mathematical model of the system to be controlled. In reality, the system may behave differently than the model indicates, or the system parameters may vary with respect to operating conditions. The closed loop system with the designed controller is robustly stable if the condition |T (j)||G (j)| < 1, (16) is satised [27] where Gp (s)Gc1 (s) T (s) = , (17) 1 + Gp (s)Gc1 (s) and Gp1 (s) Gp (s) G (s) = , (18) Gp (s) where Gp (s) is the model of the process at an operating condition for which the controller is designed and is given by K0 Gp (s) = . (19) a0 s0 + b0 s0 + c0 Hence, by equation (5), a0 s0 + b0 s0 + c0 . (20) Gc1 (s) = K0 1 DF FO s Gp1 (s) is the model of the process at a different operating condition and is given by K1 . (21) Gp1 (s) = a1 s1 + b1 s1 + c1 As the prelter Gc2 (s) is not in feedback path, it has no effect on the robust stability [26]. Substituting (19), (20) and

T.Vinopraba et al. / J Control Theory Appl 2012 10 (3) 297302

299

(21) in equations (17) and (18), 1 T (s) = , (22) 1 + 1 DF FO s G (s) = (K1 a0 s0 + K1 b0 s0 + K1 c0 K0 a1 s1 K0 b1 s1 K0 c1 )/(K0 a1 s1 + K0 b1 s1 + K0 c1 ). (23) 4.3 Robust performance analysis The closed loop system meets the robust performance specication [26] if and only if |T (j)||G (j)| + |S(j)||w(j)| 1, , (24) where S(s) is the sensitivity function, which is given by S(s) = e 1 = . rGc2 (s) d 1 + Gc1 (s)Gp (s) (25)

Using equations (7)(11), the FO-IMC for the plant described by the equation (28) is obtained and is given by Gc (s) = 0.007 7.3846 (1 + 0.8421 + 0.0534s0.8386 ). (30) 1 DF FO s

The lter time constant obtained for FO-IMC with respect to regulatory performance and tracking performance are found and 1 DF FO is 0.1834 and 2 DF FO is 0.1128 respectively. The closed loop frequency response of the system with IO-IMC and FO-IMC are shown in Fig. 5. From Fig. 5, the bandwidth of the closed loop system with different controllers are found to be 10.

By substituting equations (19) and (20) in equation (25), 1 DF FO s . (26) 1 + 1 DF FO s If the control system design is to be carried out to accommodate a specic type of input such as step setpoint change, then the weighting function will be an integrator [26]. S(s) =

5 Simulation results
For simulation, consider the FOS plant given by [11] 103.9705 . (27) Gp (s) = 1.6807 + 0.1356s0.8421 + 1 0.0073s The integer order model of the FOS [28] is given by 68944.5109 . (28) Gp (s) = 2 s + 73.6206s + 663.7429 The frequency response of the fractional order plant and integer order plant is shown in Fig. 4.

Fig. 5 Closed-loop frequency response of the system with different controllers.

The closed loop corner frequency of the process with different controllers is shown in Table 1. The lter time constant tuned for conventional IO-IMC using disturbance rejection meets the desired specication of 10 rad/s used for set point tracking. In case of SDF FO-IMC, the lter time constant 1 DF FO tuned with respect to regulatory performance provides poor tracking performance as shown in Table 1. Hence, TDF FO-IMC is attempted in which the controller tuned for disturbance rejection meets the desired bandwidth specication for set point tracking. Table 1 Closed-loop corner frequency of the system different controllers.
Controller IO-IMC SDF FO-IMC TDF FO-IMC Disturbance rejection/ Set point tracking/ (rads1 ) (rads1 ) 10 10 10 10 5.5872 10

Robustness issues

Fig. 4 Frequency response of the plant.

The robustness of the designed controller is analysed with the performance of the closed loop system at different operating regions. The model of the plant at second operating condition as given by Nataraj et al. [11] is Gp (s) = 0.0130s1.6062 110.9238 + 0.1818s0.7089 + 1 (31)

From Fig. 4, the corner frequency of the plant is found to be 10 rad/s. The desired specication for the controller design is open loop bandwidth is equal to the closed loop bandwidth. The lter time constant is chosen as 0.1 s for the conventional IO controller. The IO-IMC is obtained using equations (a6)(a8) given in Appendix A and is given by 0.001 9.0157 Gc (s) = (1 + + 0.0135s) (29) IO s

and the corresponding IO model of the plant is found to be Gp (s) = 110.9238 . 0.0017s2 + 0.11695s + 1 (32)

The frequency responses of the FO and IO plant for different operating conditions are shown in Fig. 6.

300

T.Vinopraba et al. / J Control Theory Appl 2012 10 (3) 297302

Fig. 6 Frequency response of the plant for different operating condition.

The closed loop frequency responses of the plant with different controllers for two different operating conditions are obtained and shown in Figs. 79.

Fig. 7

Closed-loop frequency response of the IO plant with IO-IMC.

Fig. 8 Closed-loop frequency response of FO plant with SDF FO-IMC.

6.1 Robust stability analysis For robustness analysis, assume Gp (s) to be the model of the plant at the rst operating condition and Gp1 (s) to be the model of the plant at second operating condition. 6.1.1 IO controller For IO controller, Gp (s) is given by equation (28) and Gp1 (s) is given by equation (32). From equation (a4), s2 + 73.6206s + 663.7429 Gc (s) = , (33) 68944.5109(0.1)s Substituting Gp (s) and Gc (s) in equation (a12), 1 T (s) = , (34) 0.1s + 1 Substituting Gp (s) and Gp1 (s) in equation (a13), 6.2818s2 + 57.7374s + 4680.3737 G (s) = , (35) 117.2056s2 +8063.0605s+68944.5109 The maximum value of |T (j)||G (j)| is found to be 0.3951 and hence the closed loop system is robustly stable. 6.1.2 FO controller For FO controller, Gp (s) is given by equation (27) and Gp1 (s) is given by equation (31). Using equation (20), Gc1 (s) is found to be 0.0073s1.6807 + 0.1356s0.8421 + 1 Gc1 (s) = . (36) 103.9705(0.1834)s0.8421 Hence, equation (22) and (23) becomes, 1 T (s) = (37) 0.1834s0.8421 + 1 and G (s) = (0.8097s1.6807 1.3516s1.6062 +15.0412s0.8421 18.9018s0.7089 + 6.9533)/(1.3516s1.6062 +18.9018s0.7089 +103.9705). (38) The maximum value of |T (j)||G (j)| is found to be 0.35 and hence the closed loop system is robustly stable. 6.2 Robust performance analysis 6.2.1 IO controller Substituting equations (28) and (33) in equation (a20), 0.1s . (39) S(s) = 0.1s + 1 The maximum value of |T (j)||G (j)| is found to be 0.4335. 6.2.2 FO controller Substituting equations (27) and (36) in equation (25), 0.1834s0.8421 . (40) S(s) = 0.1834s0.8421 + 1 The maximum value of |T (j)||G (j)| is found to be 0.3905.

Conclusions

Fig. 9 Closed-loop frequency response of FO plant with TDOF FO-IMC.

An IMC based TDF fractional order controller is reported for controlling a class of FO systems. When the power of the integral s term is decreased, the robustness of the system increases. However, in an SDF FO-IMC, the lter time constant tuned for disturbance rejection does not meet the requirement for the set point tracking. Hence, TDF FOIMC has been designed, which eliminates the shortcomings

T.Vinopraba et al. / J Control Theory Appl 2012 10 (3) 297302

301

of the SDF FO-IMC. The designed TDF FO-IMC is also shown to be robustly stable. References
[1] J. Cervera, A. Banos, C. A. Monje, et al. Tuning of fractional PID using QFT. Proceedings of 32nd IEEE Conference on Industrial Electronics, Paris: IEEE, 2006: 5402 5407. [2] Y. Chen, T. Bhaskaran, D. Xue. Practical tuning rule development for fractional order proportional and derivative controllers. Journal of Computational and Nonlinear Dynamics, 2008: 021403.1 021403.8. [3] H. Li, Y. Chen. A Fractional order proportional and derivative (FOPD) controller tuning algorithm. Proceedings of 47th IEEE Conference on Control and Decision, Mexico: IEEE, 2008: 4059 4063. [4] G. Maione, P. Lino. New tuning rules for fractional PI controllers. Nonlinear Dynamics, 2007, 49(1/2): 251 257. [5] C. A. Monje, B. M. Vinagre, V. Feliu, et al. Tuning and auto tuning of fractional order controllers for industry applications. Control Engineering Practice, 2008, 16(7): 798 812. [6] D. Valerio, J. S. Da Costa. Tuning of fractional PID controllers with Ziegler-Nichols-type rules. Signal Processing, 2006, 86(10): 2771 2784. [7] C. Yi, J. Liang, C. Gang. Optimization of fractional order PID controllers based on genetic algorithms. Proceedings of 4th IEEE Conference on Machine Learning and Cybernetics, Guangzhou: IEEE, 2005: 5686 5689. [8] C. Yi, C. Gang. Design of fractional order controllers based on particle swarm optimization. Proceedings of 1st IEEE Conference on Industrial Electronics and Applications, Singapore: IEEE, 2006: 5402 5407. [9] D. Xue, Y. Chen. A comparative introduction of four fractional order controllers. Proceedings of 4th World Congress on Intelligent Control and Automation, Shanghai, 2002: 3228 3235. [10] I. Podlubny, L. Dorcak, I. Kostial. On fractional derivatives, fractional-order dynamic systems and PI D controllers. Proceedings of the 36th IEEE Conference on Decision and Control Conference, California: IEEE, 1997: 4985 4990. [11] P. S. V. Nataraj, R. Kalla, M. M. Deshpande. Computation of spectral sets for uncertain linear fractional order systems using interval constraint propagation. Proceedings of the 17th World Congress international Federation of Automatic Control, South Korea, 2008. [12] Y. Luo, Y. Chen. Fractional order proportional derivative controller for a class of fractional order systems. Automatica, 2009, 45(10): 2446 2450. [13] C. Zhao, D. Xue, Y. Chen. A fractional order PID tuning algorithm for a class of fractional order plants. Proceedings of IEEE Conference on Mechatronics and Automation, Canada: IEEE, 2005: 216 221. [14] A. Oustaloup, M. Bansanol. First generation Crone control. Proceedings of IEEE Conference on Systems, Man and Cybernetics, France: IEEE, 1993: 130 135. [15] L. A. Monje, V. Feliu. The fractional order Lead Compensator. Proceedings of 2nd IEEE Conference on Computational Cybernetics, Vienna, 2004: 347 352. [16] D. Maiti, M. Chakraborty, A. Acharya, et al. Design of fractional order self tuning regulator using optimization algorithm. Proceedings of 2nd IEEE Conference on Computer and Information Technology, Singapore: IEEE, 2008: 470 475. [17] S. Ladaci, A. Charef. Model reference adaptive control with fractional derivative. Proceedings of International Conference on Tele-communication systems, Medical electronics and Automation, Algeria, 2003. [18] S. Ladaci, J. J. Loiseau, A. Charef. Fractional order adaptive high gain controllers for a class of linear systems. Communications in Nonlinear Science and Numerical Simulation, 2008, 13(4): 707 714.

[19] H. Delvari, R. Ghaderi, A. Ranjbar, et al. Fuzzy fractional order sliding mode controller for non linear systems. Communications in Nonlinear Science and Numerical Simulation, 2010, 15(4): 963 978. [20] Y. Chen, K. L. Moore. On D type iterative learning control. Proceedings of 40th IEEE Conference on Decision and Control, Orlando: IEEE, 2001: 4451 4456. [21] S. E. Hamamci, M. Koksal. Calculation of all stabilizing fractionalorder PD controllers for integrating time delay systems. Computers and Mathematics with Applications, 2010, 59(5): 1621 1629. [22] T. Vino praba, N. Sivakumaran, N. Selvaganesan, et al. Stabilization using fractional order PID controllers for rst order time delay system. Proceedings of International Conference on Advances in Computing Control and Telecommunication Technologies, Kerala: IEEE, 2009: 725 728. [23] D. E. Rivera, M. Morari, S. Skogestad. Internal model control for PID controller design. Industrial Engineering Chemical Process, 1986, 25: 252 265. [24] D. Li, W. Fan, Q. Jin, et al. An IMC PI D controller design for fractional calculus system. Proceedings of 29th Chinese Control Conference, Beijing: IEEE, 2010: 3581 3585. [25] T. Vinopraba, N. Sivakumaran, S. Narayanan. IMC based fractional order PID controller. Proceedings of IEEE Conference on Industrial Technology, Auburn: IEEE, 2011: 71 76. [26] M. Morrari, E. Zariou. Robust Process Control, Englewood Cliffs, New Jersy: Prentice Hall, 1989. [27] G. C. Goodwin, S. F. Graebe, M. E. Salgad. Control system design, Valparaiso: Prentice Hall, 2000. [28] S. Narayanan, N. Sivakumaran, T. Vinopraba, et al. IMC Based PI D controller design for gas turbine plant. Proceedings of International Conference on Modeling and Simulation, Kerala, 2009: 386 390.

Appendix A
Consider the process Gp (s) to be K , (a1) Gp (s) = 2 as + bs + c where K is the gain of the process, a, b and c are the real coefcients and the model Gm (s) is same as that of that of the process Gp (s). K Gm (s) = 2 . (a2) as + bs + c Based on pole-zero cancellation; assume Gi (s) to be as2 + bs + c Gi (s) = , (a3) K(IO s + 1) where IO is the lter time constant. The lter time constant has to be tuned according to the desired specication. Substituting Gi (s) and Gm (s) in equation (1), then as2 + bs + c . (a4) Gc (s) = KIO s The PID controller transfer function is given by 1 Gc (s) = Kc (1 + + Td s). (a5) Ti s Based on [26], equating (a4) with (a5), the controller parameters are obtained to be b Kc = , (a6) KIO b (a7) Ti = , c a (a8) Td = . b The closed loop transfer function with respect to disturbance rejection is given by y(s) IO s = . (a9) d(s) IO s + 1

302

T.Vinopraba et al. / J Control Theory Appl 2012 10 (3) 297302 1 |1 s |, = (a23) 1 s + 1 1 1 (a24) 1 s + 1 = 2 . (Because 1 is tuned for the set point tracking) y(s) 1 (a25) d(s) = 2 |1 s |. The desired specication cannot be achieved for the SDF FO-IMC due to fractional power of the s term.
T. VINOPRABA received the B.E. degree in Electronics and Instrumentation Engineering from Annamalai University, Chidambaram, in 2006, M.E. degree in Process Control and Instrumentation Engineering from Annamalai University in 2008, and currently she is pursuing her Ph.D. in National Institute of Technology, Trichy. Her research interests include fractional order controller, controller design for chemical and biomedical systems. E-mail: vinopraba@gmail.com. N. SIVAKUMARAN received the B.E degree in Electronics and Instrumentation Engineering from Bharathidasan University, Trichy, in 1999, M.E. degree in Process Control and Instrumentation Engineering from Annamalai University in 2002 and Ph.D. from National Institute of Technology, Trichy, in 2007. His research interests include system identication, intelligent systems, controller design for chemical and biomedical systems. He is the recipient of the AICTE-National Doctoral Fellowship for the year 20042006. He has research grant from DST, MHRD, CDAC, Government of India. He has teaching experience of 8 years. E-mail: nsk@nitt.edu. S. NARAYANAN received the B.E. degree in Electronics and Instrumentation Engineering from Annamalai University, Chidambaram, and M.E. degree in Instrumentation Engineering from MIT, Anna University, Chennai. He completed his Ph.D. from MIT, Anna University, Chennai. He is currently working as Assistant Professor in the Department of Instrumentation and Control Engineering, National Institute of Technology, Trichy. His research interests include PID control design and control engineering. Email: narayanan@nitt.edu. T. K. RADHAKRISHNAN received his B.E. degree in Chemical Engineering from Annamalai University and M.Tech. degree from Indian Institute of Technology, Chennai, India. After his graduate program he joined Indian Institute of Chemical Technology, Hyderabad in the Simulation, Optimization and Control Group. He later moved to Regional Engineering College, Tiruchirappalli and received Ph.D. He is currently working in National Institute of Technology, Tiruchirappalli as Professor and Head of Chemical Engineering Department. His interests include process control, process modeling and soft computing. E-mail: radha@nitt.edu.

The corner frequency is determined from the bandwidth specication. The closed loop transfer function with respect to set point tracking is given by y(s) 1 = . (a10) r(s) IO s + 1 In the above equation, the lter time constant is tuned according to the bandwidth specication. It is found that for the conventional integer order IMC, the lter time constants tuned for disturbance rejection and for tracking performance are same. Robust stability analysis: For the closed loop system to be robustly stable, the following condition must be satised: |T (j)||G (j)| < 1, , (a11) where Gp (s)Gc (s) T (s) = (a12) 1 + Gp (s)Gc (s) and Gp1 (s) Gp (s) G (s) = . (a13) Gp (s) Gp (s) is the model of the process at an operating condition for which the controller is designed. K0 Gp (s) = . (a14) a0 s2 + b0 s + c0 Hence Gc (s) is given by a0 s2 + b0 s + c0 , (a15) Gc (s) = K0 IO s Gp1 (s) is the model of the process at a different operating condition. K1 Gp1 (s) = . (a16) a1 s2 + b1 s + c1 Substituting (a14), (a15) and (a16) in equations (a12) and (a13) 1 , (a17) T (s) = 1 + IO s G (s) = ((K1 a0 K0 a1 )s2 + (K1 b0 K0 b1 )s +(K1 c0 K0 c1 )/(K0 a1 s2+K0 b1 s+K0 c1 ). (a18) Robust performance analysis: The closed loop system will meet the robust performance specication if and only if |T (j)||G (j)| + |S(j)||w(j)| 1, , (a19) where S(s) is the sensitivity function, which is given by 1 S(s) = . (a20) 1 + Gc (s)Gp (s) By substituting (a14) and (a15) in equation (a20), IO s S(s) = . (a21) 1 + IO s

Appendix B
From the simulation results, the lter time constant tuned for the set point tracking is 1 = 0.1128. While applying the same lter time constant to disturbance rejection, the transfer function becomes y(s) 1 s = (a22) d(s) 1 s + 1

S-ar putea să vă placă și