Vikram SAINI, Lillie DEWAN

Regularized estimation of Hammerstein systems using a decomposition-based iterative instrumental variable method

This paper presents a two-step instrumental variable (IV) method to obtain the regularized and consistent parameter estimates of the Hammerstein ARMAX model based on the bilinear parameterized form. The two-step identi cation method consists of estimating the bilinear parameters in the rst step, followed by parameter reduction in the second step. An iterative identi cation method is proposed, based on the idea of separating the bilinear form in the two separable forms with partial parameters and solving the decomposed model forms iteratively. The IV-based estimation is integrated into the formulated decomposed structure by introducing the instruments constructed from the estimated auxiliary model outputs. It is shown that in a stochastic environment the proposed IV method produces consistent estimates in the presence of correlated noise disturbances. The validity of the proposed algorithm is veri ed with the help of extensive simulations using a Monte Carlo study.

PDF

___

[1] Narendra KS, Gallman PG. An iterative method for the identi cation of nonlinear systems using Hammerstein model. IEEE T Automat Contr 1966; 11: 546-550.
[2] Thathachar MAL, Ramaswamy S. Identi cation of a class of nonlinear systems. Int J Control 1973; 18: 741-752.
[3] Fruzzetti KP, Palazoglu A, McDonald KA. Nonlinear model predictive control using Hammerstein models. J Process Contr 1977; 7: 31-41.
[4] Janczak A. Identi cation of Nonlinear Systems Using Neural Networks and Polynomial Models: A Block-Oriented Approach. Berlin, Germany: Springer-Verlag, 2005.
[5] Chang FHI, Luus R. A noniterative method for identi cation using Hammerstein models. IEEE T Automat Contr 1971; 16: 464-468.
[6] Bai EW. An optimal two-stage identi cation algorithm for Hammerstein-Wiener nonlinear systems Automatica 1998; 34: 333-338.
[7] Voros J. Iterative algorithm for parameter identi cation of Hammerstein systems with two-segment nonlinearities. IEEE T Automat Contr 1999; 44: 2145-2149.
[8] Bai EW, Chan K.S. Identi cation of an additive nonlinear system and its applications in generalized Hammerstein models. Automatica 2008; 44: 430-436.
[9] Ljung L. System Identi cation: Theory for the User, 2nd ed. Upper Saddle River, NJ, USA: Prentice-Hall, 1999.
[10] Hsia TC. System Identi cation, Least-squares Methods. Lexington, MA, USA: Lexington Books, 1977.
[11] Ding F, Chen T. Identi cation of Hammerstein nonlinear ARMAX systems. Automatica 2005; 41: 1479-1489.
[12] Chaudhary NI, Raja MAZ. Identi cation of Hammerstein nonlinear ARMAX systems using nonlinear adaptive algorithms. Nonlinear Dynam 2015; 79: 1385-1397.
[13] Chen H, Ding F. Hierarchical least squares identi cation for Hammerstein nonlinear controlled autoregressive systems. Circ Syst Signal Pr 2015; 34: 61-75.
[14] Ding F, Shi Y, Chen T. Auxiliary model-based least-squares identi cation methods for Hammerstein output-error systems. Syst Control Lett 2007; 56: 373-380.
[15] Wang DQ, Liu HB, Ding F. Highly efficient identi cation methods for dual-rate Hammerstein systems. IEEE T Contr Syst T 2015; 23: 1952-1960.
[16] Wang Y, Ding F. Filtering-based iterative identi cation for multivariable systems. IET Control Theory A 2016; 10: 894-902.
[17] Saini V, Dewan L. Instrument variable method based on nonlinear transformed instruments for Hammerstein system identi cation. J Vib Control published online ahead of print; January 2017; DOI: 10.1177/1077546317694770.
[18] Laurain V, Garnier H, Gilson M, Young P. Re ned instrumental variable method for identi cation of Hammerstein continuous-time Box-Jenkins models. In: 47th IEEE Conference on Decision and Control; 9{11 December 2008; Cancun, Mexico. pp. 1386-1391.
[19] Svensson A, Schon T.B, Solin A, Sarkka S. Nonlinear state space model identi cation using a regularized basis func- tion expansion. In: 6th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing; 13{16 December 2015; Cancun, Mexico. pp. 481-484.
[20] Hoerl AE, Kennard RW. Ridge regression: biased estimation for nonorthogonal problems. Technometrics 1970; 12: 55-67.
[21] Soderstromt T, Stoica P. Comparison of some instrumental variable methods| consistency and accuracy aspects. Automatica 1981; 17: 101-115.
[22] Stoica P, Nehorai A. Non-iterative optimal min-max instrumental variable method for system identi cation. Int J Control 1988; 47: 1759-1769.
[23] Young PC. An instrumental variable method for real time identi cation of a noisy process. Automatica 1970; 6: 271-287.
[24] Stoica P, Soderstrom T. Optimal instrumental variable estimation and approximate implementations. IEEE T Automat Contr 1983; 28: 757-771.
[25] Grimmett G, Stirzaker D. Probability and Random Processes. Oxford, UK: Oxford University Press, 2001.
[26] Billings SA, Wei HL. A new class of wavelet networks for nonlinear system identi cation. IEEE T Neural Networ 2005; 16: 862-874.
[27] Truong NV, Wang L, Young PC. Nonlinear system modeling based on non- parametric identi cation and linear wavelet estimation of SDP models. Int J Control 2007; 80: 774-788.