Research Journal of Applied Sciences, Engineering and Technology 4(23): 5230-5235, 2012 ISSN: 2040-7467 © Maxwell Scientific Organization, 2012 Submitted: May 01, 2012 Accepted: May 22, 2012 Published: December 01, 2012 Model Following Control for Nonlinear System with Time Delay Shujing Wu and Dazhong Wang Shanghai University of Engineering Science, Shanghai, China Abstract: In this study, designing method of Model Following Control System (MFCS) for nonlinear system with time delay and disturbances is discussed. The design of the control system is constructed. The features of this design method are: Bounded property of the inner states for the control system is given and the utility of this control de because there is no necessary to make transformation of this system; It is confirmed on basis of a num sign is guaranteed; Both the physical structure of the system and the physical system variables properties can be preserved because there is no necessary to make transformation of this system; It is confirmed on basis of a numerical example that the output signal of the control system asymptotically follows the reference model signal in the case of the existence of disturbances. Keywords: Model following control system, nonlinear system, time delay INTRODUCTION Metallurgical processing systems, transmission systems, environmental systems, power systems, chemical processing systems and communication systems are all examples of time delay systems (Nazario and Jaime, 2002; Abdellah and Ahmed, 2011; Shi et al., 2001). Also, it has been shown that the existence of time delay usually becomes the source of instability and deteriorates the performance of systems. In recent years, such systems have attracted recurring interest of research community. Much of the research study has been focused on stability analysis and the stabilization of time delay systems using the so-called Lyapunov-Krasovskii functional and the Linear Matrix Inequality (LMI) approach (Zhang and Xie, 2007). The time delay resulted in unpredictable outside disturbances to the system and therefore, it is the dynamic element that is hardest to control in the control system. The neutral system has been studied as on robust stability of neutral systems with time-varying discrete delay and norm-bounded uncertainty (Han, 2004). The State Predictive Model Following Control System for the Linear Time delays is discussed (Wang and Okubo, 2009). It was proposed by (Wu et al., 2008; Wu et al., 2011; Wang and Okubo, 2008) for a family of plants with separable linear and nonlinear part. In previous studies, a method of linear MFCS for time delays system was proposed by (Akiyama et al., 1998). In this study, the method of linear MFCS will be extended to nonlinear system with time delay. The design of the control system is performed using an easy algebraic algorithm of matrices whose elements are polynomials of the operator. The bounded property of internal states for the control system is given and the utility of this control design is guaranteed. The paper studied model following design on the difference differential equation configured by the control system, established input signal that could automatically compensate time delay of the system and there by effectively removed influence caused by outside disturbances on the control system. We set the nonlinear parts f(:(t)) of the controlled object as ||f(:(t))||#"+ $||:(t)||( and show the bounded property of inner states by separating the nonlinear part into 0 # ( < 1. In this case, the effectiveness of this method has verified by a numerical example and simulations. METHODOLOGY The expression of problems: The controlled object is described in (1), (2) and (3): x&(t ) = + k k i=0 i=0 ∑ Ai x(t − ih) + ∑ Ai' x&(t − ih) (1) k ∑ Bi u(t ) + B f f ( µ (t )) + d (t ) i=0 y( t ) = k ∑ C x( t − ih) + d i=0 µ (t ) = i 0 (t ) k ∑ C fi x(t − ih) i=0 where, x(t), Rn, d(t), Rn, u(t), Rl , y(t), Rl, d0(t) , Rl, :(t), Rlf, f(:(t)) , Rlf Corresponding Author: Shujing Wu, Shanghai University of Engineering Science, Shanghai, China 5230 (2) (3) Res. J. Appl. Sci. Eng. Technol., 4(23): 5230-5235, 2012 The available states are output y(t) and measurement output :(t). d(t) is state uncertainty or bounded disturbance which is assumed to be repeatable, d0(t)is output uncertainty or bounded disturbance signal which is assumed to be repeatable. t is the time. The nonlinear function f(:(t)) is available and satisfies the following constraint: ||f (:(t)) || # " + $|| :(t)||( (4) C( ρ ) = (5) ym (t) = Cm xm (t) (6) C(D)[pE (D)!A(D)]!1 B(D) = N(D, p)/D (D, p) C(D)[pE (D)!A(D)]!1 Bf = Nf (D, p)/D (D, p) Here, D (D, p) = | pE(D)! A(D)|, Dm(p) = | pI!Am|. Then the representations of input-output equation are described as followings: (7) Design of nonlinear model following control system: Let: D = (D0, D1, D2, ..., Dk)T, Di = e!pih, Di x(t) = x(t!ih) From this, system (1), (2), (3) are given in system (8)(10): E ( ρ ) x&(t ) = A( ρ ) x(t ) + B( ρ )u(t ) + B f f ( µ (t )) + d (t ) (8) y(t) = C(D) x(t) + d0 (t) (9) :(t) = Cf(D) x(t) B( ρ ) = k ∑ Ai ρi i=0 (11) Dm (p) ym(t) = Nm(p) rm(t) (12) w(t) = C (D) adj [pE (D) !A(D)]d(t) + D(F, p)d0(t) (13) Nf (D, p) = C (D) adj [pE (D)!A(D)] B(D) (14) and A( ρ ) = D (D, p) y (t) = N (D, p) u(t) + Nf (D, p) f(:(t)) + w(t) Define N(D, p), Nf (D, p), Nm (D), Nr (D) as: P = d/dt E = (ρ) = I + i=0 Then the transfer function of controlled system and model are as: converges to zero asymptotically. where, k ∑ C f ρi Cm(D) [pI !Am]!1 B(D) = Nm(p) / Dm(p) In this study, we propose a design of model following control system for nonlinear system with disturbances. We can proof that all the internal states are bounded and output error: e (t) = y(t) ! ym(t) i=0 C f (ρ ) = where, " $ 0 $, $ 0, ||.|| is Euclidean norm. The reference model is given (5) and (6): x&m (t ) = Am xm (t ) + Bmrm (t ) k ∑ Ci ρi k ∑ A'ρi i=0 Nf (D,p) = C (D) adj [pE (D)!A(D)]Bf (15) Nm (p) = Cm adj[pI !Am]Bm (16) N r ( ρ ) = N r ( ρ ) + N$ r (17) ~ ~ ~ where, Mri N (D, p) < 0i, Mri N (D, p) < 0fi, , Mri N m (p) < 0mi, (i = 0, 1, ..., l) is assumed to be regular. The disturbances d(t) and d0(t) are satisfy the following conditions: Dd (p) d(t) = 0, Dd (p) d0(t) = 0 (18) (10) where, Dd (p) is a scalar characteristic polynomial of disturbances, 2Dd (p) = nd and Dd (p) is monic polynomial, thus Dd (p)w(t) = 0. Choose a stable polynomial T(p) which satisfy the following condition: Degree of T(p) is * $ nd + 2n ! nm!1!0i; The coefficient of maximum degree term of T(p) is the same as D(p), R(D, p) and S(D, p) can be obtained: k ∑ Bi ρi T(p)Dm(p) = Dd(p)D(D, p)R(D, p) + S(D, p) i=0 5231 (19) Res. J. Appl. Sci. Eng. Technol., 4(23): 5230-5235, 2012 −1 H1( ρ )[ pI − F1] G1 = N$ r− 1Q − 1( p) Degree of each term as the following: { (26) } ⋅ Dd ( p) R( ρ , p) N ( ρ , p) − Q( p) N r ( ρ ) MT (p) = *, MDm(p) = nm, MDd (p) = nd, MD(D, p) = n, MR(D, p) = * + nm ! nd ! n, MS(D, p) # nd + n!1 J2 ( ρ ) + H2 ( ρ )[ pI − F2 ] G2 −1 (27) = N$ r− 1Q − 1( p)S ( ρ , p) From (7), (11), (12) and (19), we can get easily: T ( p) Dm ( p)e(t ) = Dd ( p) D( ρ , p) R( ρ , p) y(t ) + S ( ρ , p) y(t ) − T ( p) N m ( p)rm (t ) J3 ( ρ ) + H3( ρ )[ pI − F3 ] G3 −1 (28) = N$ r−1Q − 1( p) Dd ( p) R( ρ , p) N f ( ρ , p) (20) J4 ( ρ ) + H4 ( ρ )[ pI − F4 ] G4 −1 We can rewrite e(t) as: { [ ]}{[ D ( p)R( ρ, p) N ( ρ, p) e( t ) = 1 / T ( p) Dm ( p) ] d − Q( p) N r ( ρ ) u( t ) + Dd ( p) R( ρ, p) N f ( ρ , p) f ( µ( t ) ) + Q( p) N r ( ρ) u( t ) + S ( ρ, p) y( t ) The followings must be satisfied: (21) } − T ( p) N m ( p)rm ( t ) and ( Q( p) = diag p ~ ∂ri Q( p) < δ + nm − n + ηi δ + nm − n + ηi ) ~ + Q( p), So, u(t) can be obtained by letting the right-hand side of (21) be equal to zero: } (22) u( t ) = − E 0 ( ρ ) u( t ) − H1 ( ρ )ξ1 ( t ) (31) ξ&3 (t ) = F3ξ3 (t ) + G3 f ( µ (t )) (32) ξ&4 (t ) = F4ξ4 (t ) + G4rm (t ) (33) ⎡ E (ρ ) ⎢ 0 d ⎢⎢ 0 dt ⎢ ⎢ 0 ⎢ 0 ⎣ 0 0 0 0⎤ ⎡ x(t ) ⎤ ⎥ ⎥⎢ | 0 0 0⎥ ⎢ ξ1(t ) ⎥ ⎢ ⎥ 0 | 0 0⎥ ⎢ ξ2 (t )⎥ ⎥ 0 0 | 0⎥ ⎢ ξ3 (t ) ⎥ ⎢ ⎥ 0 0 0 0⎥⎦ ⎢ u(t ) ⎥ ⎣ ⎦ 0 F1 0 0 0 0 0 0 F2 0 0 F3 H1( ρ ) − H2 ( ρ ) − H3 ( ρ ) (24) ⎡ ⎢ ⎢ +⎢ ⎢ ⎢ ⎢− ⎣ Here, E0 ( ρ ) = N$ r− 1 N r ( ρ ) ξ&2 (t ) = F2ξ2 (t ) + G2 y(t ) ⎡ A( ρ ) ⎢ 0 ⎢ = ⎢ G2C( ρ ) ⎢ 0 ⎢ ⎢− J ρ C ρ − ⎣ 2( ) ( ) (23) − E 3 ( ρ ) f ( µ( t ) ) − H 3 ( ρ)ξ3 ( t ) + u( t ) um (t ) = J4 ( ρ )rm (t ) + H4 ( ρ )ξ4 (t ) (30) where, |pI!Fi| = |Q(p)|, (i = 1, 2, 3, 4). If the internal states of control system are bounded, the design of MFCS for nonlinear system can be realized. where, i = 1, 2, ..., l and N$ r ≠ 0 , nm!0mi $ n ! 0i, $ 0fi . Control input u(t) can be described using internal states as followings: + E 2 ( ρ ) y( t ) − H 2 ( ρ )ξ2 ( t ) ξ&1(t ) = F1ξ1(t ) + G1u(t ) Bounded property of internal states: System inputs are reference input signal rm(t) and disturbances d(t), d0(t) which are all assumed to be bounded. The bounded can be easily proved if there is no nonlinear par f(:(t)). But if f(:(t)) exists, the bounded has relation with it. First, the overall system can represented by state space in (34): { ) ) u( t ) = − N r−1 N r ( ρ) u( t ) − N r−1Q −1 ( ρ) Dd ( ρ ) R( ρ, p) N ( ρ, p) ) − Q( p) N r ( ρ) u( t ) − N r−1Q − 1 ( p) Dd ( p) R( ρ, p) ) −1 −1 ⋅ N f ( ρ , p ) f ( µ ( t ) ) − N r Q ( p) S ( ρ , p ) y ( t ) ) + N r−1Q −1 ( p)T ( p) N m ( p)rm ( t ) (29) = N$ r− 1Q − 1( ρ )T ( ρ ) N m ( ρ ) (25) 5232 ⎤ ⎥ ⎥ ⎥ f ( µ (t )) + ⎥ G3 ⎥ J3 ( ρ )⎥⎦ Bf 0 0 ⎡ ⎤ d (t ) ⎢ ⎥ 0 ⎢ ⎥ ⎢ ⎥ G2d0 (t ) ⎢ ⎥ 0 ⎢ ⎥ ⎢ u (t ) − J ( ρ )d (t )⎥ 2 0 ⎦ ⎣ m ⎤ ⎡ x (t ) ⎤ ⎥ ⎥⎢ ⎥ ⎢ ξ1(t ) ⎥ ⎢ ⎥ ξ ( t )⎥ 0 ⎥⎢ 2 ⎥ 0 ⎥ ⎢ ξ3(t ) ⎥ ⎥ ⎢ − I − E0 − H1( ρ )⎥⎦ ⎢ u(t ) ⎥ ⎦ ⎣ B( ρ ) G1 (34) Res. J. Appl. Sci. Eng. Technol., 4(23): 5230-5235, 2012 where, zT (t) = [x(t), >1(t), >2(t), >3, u(t)]T. Necessary part about bounded property is considered, state space can be simplified as: E ( ρ ) z&( t ) = As ( ρ) z( t ) + Bs ( ρ) f ( µ( t ) ) + d s ( t ) (35) y(t) = C(D) x(t) + d0(t) (36) :(t) = Cs (D) z(t) (37) Cs (D) = [Cf (D), 0 ,0, 0, 0] (38) ⎡ Bs1( ρ ) ⎤ ⎥ XBs ( ρ ) = ⎢ ⎢⎣ Bs2 ( ρ )⎥⎦ ⎡ d s1(t ) ⎤ Xd s (t ) = ⎢ ⎥ ⎢⎣ d s2 (t )⎥⎦ So, we have: ⎡ I 0⎤ ⎡ z&1(t ) ⎤ ⎡ As1( ρ ) 0⎤ ⎡ z (t )⎤ ⎥ ⎥⎢ ⎢ 0 0⎥ ⎢ & ⎥ = ⎢ I ⎥⎦ ⎢⎣ z (t )⎥⎦ ⎣ ⎦ ⎢⎣ z2 (t )⎥⎦ ⎢⎣ 0 ⎡ Bs1( ρ )⎤ ⎡ d s1(t ) ⎤ ⎥ f ( µ (t )) ⎢ +⎢ ⎥ ⎢⎣ Bs1( ρ )⎥⎦ ⎢⎣ d s2 (t )⎥⎦ We prove that As(D) is stable. As(D) and its characteristic polynomial is calculated as the followings: (42) The state-space realization of (42) becomes: pE ( ρ ) − As ( ρ ) = N$ r −1 T ( p) Dm ( p) Q( p) Vs ( ρ, p) (39) l l 2 where, Vs(D, p) is the zeros polynomial of C(D)[pE(D)!A(D)]!1B(D) = W(D, p)!1 U(D, p)|, that is Vs(D, p) = |U(D, p)| / | Nr(D)|. As |Q(p)|, Vs(D, p), |T (p)|, |Dm (p)|, are all stable polynomials. Therefore, As(D) is a stable system matrix. In order to obtain desired conclusion, it is sufficient to prove that z(t) is bounded. The nonlinear function is available and satisfies constraint. We set the nonlinear parts f(:(t))) of the controlled object as || f(:(t))|| # " + $ || :(t) ||( and show the bounded property of inner states by separating the nonlinear part into " $ 0, $ $ 0, 0 # ( <1. To use regular transformation z(t) = Y z (t) and X, from (35), we can be rewritten as followings: XE ( ρ )Yz& ( t ) = XAs ( ρ )Yz ( t ) + XBs ( ρ ) f ( µ( t ) ) + Xd s ( t ) z&1 ( t ) = As1 ( ρ ) z1 ( t ) + Bs1 ( ρ ) f ( µ( t ) ) + d s1 ( t ) (43) z2 (t ) = − Bs2 ( ρ ) f ( µ (t )) − d s2 (t ) (44) From (42), we can get easily: X pE ( ρ ) − As ( ρ ) Y = ξ pE − As1( ρ ) Here, > is stable polynomial. Therefore, As(D) is a stable system matrix, we can know As1(D) is also a stable system matrix. Consider the following Lyapunov function: V (t , ρ ) = 1 T z1 (t ) Ps1( ρ ) z1(t ) 2 1 V& (t , ρ ) = − z1T (t )Qs1z1(t ) Ps1( ρ )d s1(t ) 2 (40) + z1T (t ) Ps1( ρ ) Bs1( ρ ) f ( µ (t )) Ez& ( t ) = XAs ( ρ)Yz ( t ) + XBs ( ρ) f ( µ( t ) ) + Xd s ( t ) where ⎡ I 0⎤ E= ⎢ ⎥, ⎣ 0 0⎦ ⎡ AS1 0⎤ XAS ( ρ )Y = ⎢ , I ⎥⎦ ⎣ 0 ⎡ z1(t ) ⎤ z (t ) = ⎢ ⎥, ⎢⎣ z2 (t )⎥⎦ (41) (45) (46) (47) There exists a positive definite matrix Ps1(D) and a semi-positive definite matrix Qs1satisfying (Lyapunov equation): As1T(D)Ps1(D) + Ps1(D) As1(D) = !Qs1 (48) So, we can rewrite (47) as: V& (t , ρ ) ≤ − k1V (t , ρ ) + k2 (49) where, k1 and k2 are exists stable polynomials. We can be written V(t, D) as following: V (t , ρ ) ≤ 5233 k1 + V (0, ρ ) k2 (50) Res. J. Appl. Sci. Eng. Technol., 4(23): 5230-5235, 2012 We conclude that z1 ( t ) is bounded. But with z2 ( t ) is bounded (from (44)), that :(t) and z(t) is bounded. The above result is summarized as Theorem 1. 2 y(t) ym(t) Theorem 1: For the nonlinear system of (1) and (2), using the method in this study to design model following control system, z(t), Rm, :(t) , Rlf , f(:(t)) , Rlf , A(D) , Rn×n. d(t) and d0(t) which are all assumed to be bounded, A(D) is a stable system matrix: d(t) 1 d 0(t) 0 e(t) E ( ρ ) z&(t ) = A( ρ ) z(t ) + B( ρ ) f ( µ (t )) + d (t ) -0.6 0 10 20 30 time (sec) :(t) = C (D)z(t) + d0(t) Fig. 1: Responses of the MFCS with time delay all internal states are bounded if: C C C CONCLUSION ||f (:(t))|| # " + $ ||:(t)||( (" $ 0, $ $ 0, 0 # (<1); As (D), R n×n is bounded; d(t), d0(t) is a stable system matrix. SIMULATION RESULTS In this section, an example is presented to demonstrate the effectiveness and flexibility of the proposed design approach. The system with time delays is given by: ⎡ 1 ρ1 0⎤ ⎢ ⎥ x& ( t ) = ⎢ 0 0 1⎥ x( t ) + ⎢⎣ − 15 − 8 3⎥⎦ ⎡ 0 ⎤ ⎢ ⎥ + ⎢ 0 ⎥ u( t ) + ⎢⎣1 + ρ1 ⎥⎦ ⎡1 0 ⎢ ⎢ 0 1 + ρ2 ⎢⎣ 0 0 ⎡ 2⎤ ⎢ ⎥ ⎢ 1⎥ f ( µ ( t ) ) + ⎢⎣ 2⎥⎦ 2⎤ ⎥ 0⎥ x& ( t ) 0⎥⎦ ACKNOWLEDGMENT This study was financially supported by the Innovation Program of Shanghai Municipal Education Commission (12YZ148), the Project-sponsored by SRF for ROCS, SEM(1568) and the Scientific Research Foundation of SUES (A-0501-12-01; A-0501-10-023). The authors would like to thank the editor and the reviewers for their constructive comments and suggestions which improved the quality of the study. ⎡ 0 ⎤ ⎢ ⎥ ⎢ d ( t )⎥ ⎢⎣ d ( t ) ⎥⎦ y(t) = [2+D3 1 0]x(t) +d0(t) :(t) = [1 + D4 5 0]x(t) f ( µ ( t ) ) = 3µ ( t ) We have presented model following control for nonlinear system with time delay. The illustrative example and the simulation results show the benefits of this proposed design methods. Topics for future include: the discrete control system for the time-delays and the predictive control of the nonlinear system with time-delay will be discussed. REFERENCES 1 5 + 2µ( t ) 1 6 Reference model is given in: 1⎤ ⎡ 0 ⎡ 0⎤ x&m (t ) = ⎢ ⎥ xm (t ) + ⎢ 1⎥ rm (t ) ⎣ − 6 − 5⎦ ⎣ ⎦ ym (t) = [2 1] xm (t) The paper discussed study of a MFCS (Model Following Control System) with time delay (Fig. 1), calculated with practical example, the result showed that both output signal y(t) of the system and output signal ym (t) of the reference model achieved following effect perfectly and error of output converged to zero gradually, therefore, validity of the design could be effectively proven. Abdellah, B. and E.H. Ahmed, 2011. Delay-dependent stabilization conditions of controlled positive t-s fuzzy systems with time varying delay. Int. J. Innov. Comput. Inform. Con., 7(4): 1533-1547. Akiyama, T., H. Hattori and S. Okubo, 1998. Design of the model following control system with time delays. Trans. IEE Japan, 118-c (4): 497-502. Han, Q.L., 2004. On robust stability of neutral systems with time-varying discrete delay and norm-bounded uncertainty. Automatica, 40(6): 1087-1092. Nazario, D.R. and A.M. Jaime, 2002. Neural networks to model dynamic systems with time delays. IE Trans., 34(3): 313-327. Shi, Z., J. Yang, J.J. Foshee, W.B. Hartman, S. Tang, et al., 2001. Photonics for time delay in communication systems. Opt. Eng., 40(7): 1238-1243. 5234 Res. J. Appl. Sci. Eng. Technol., 4(23): 5230-5235, 2012 Wang, D., S. Wu and S. Okubo, 2009. The state predictive model following control system for the linear time delays. Int. J. Automation Comput., 6(2): 186-191. Wang, D. and S. Okubo, 2008. A design of model following control system for linear neutral system with time delays. Trans. Ins. Electr. Eng. Japan, 128(11): 1657-1663. Wu, S., S. Okubo and D. Wang, 2008. Design of model following control system for nonlinear descriptor system in discrete time. Kybernetika, 44(4): 546-556. Wu, S., D. Wang and S. Okubo, 2011. Control for Nonlinear Chemical System. Key Eng. Mat., 467(469): 1450-1455. Zhang, H.S. and L.H. Xie, 2007. Control and Estimation of Systems with Input/Output Delays. SpringerVerlag, Berlin. 5235