A neural identifier for linear dynamic systems

Bu çalışmada, lineer dinamik sistemlerin kimliklendirilmesinde ihtiyaç duyulan system derecesine ve system parametrelerine ihtiyaç duyulmadan yapay sinir ağları (YSA) kullanılarak nasıl kimliklendirme yapılabileceği üzerine bir öneri sunulmuştur. Maksimum 3. dereceye kadar farklı derecelerde bir çok sistemin kimliklendirme sonuçları verilmiştir. Bu çalışmada aynı zamanda Lattice ARMA metodu sonuçları YSA sonuçları ile karşılaştırılmıştır. Bu çalışmadan elde edilen sonuçlara göre, sunulan YSA kimliklendiricinin daha iyi ve kabul edilebilir sonuçlar verdiği, uygulamasının kolay olduğu ve Lattice ARMA'da ve diğer metodlarda sistem kimliklendirilmede ihtiyaç duyulan sistem derecesine bağlı olmaksızın sistem kimliklendirilebileceği anlaşılmıştır.

Lineer dinamik sistemler için bir nörol kimliklendirici

This paper describes the use of multilayered perceptrons for linear dynamic system identification without knowing not only the parameters but also the orders of the systems. Identification results of various linear systems up to the 3-rd orders are given. This study also presents the results of the lattice ARMA method for comparison. The results demonstrate that the neural identifier is versatile, and relatively simple to implement, provides much better results, and does not require the order of the systems to be known in comparison to the lattice ARMA method or the other methods available in the literature.

___

1. Narendra K. S., Parthasarathy K.: Neural networks in dynamical systems. SPIE, Intelligent Control and Adaptive Systems, 1196,230-241, 1989.
2. Chu S. R., Shoureshi R, Tenorio M: Neural networks for system identification. In: Proc. of American Control Conference, Pittsburg, PA, 916-921, June 1989.
3. Narendra K. S., Parthasarathy K.: Identification and control of dynamical systems using neural networks. IEEE Transactions on Neural Networks, 1(1), 4-27, 1990.
4. Pham D. T., Liu, X.: State-space identification of dynamic systems using neural networks, Engineering Applications of AI, 3, 198-203, 1990.
5. Hakim N. Z., Kaufman J. J., Cerf G., Meadows H. E.: A discrete-time neural network model for systems identification. In: Proc. of the INNC90, (IEEE-INNS) San Diego, Califonia, 3, 593-598, June 1990.
6. Bulsari A. B., Saxen H.: A feedforward artificial neural network for system identification of a chemical process. J. Systems Engineering, 1(1), 1991.
7. Pham D. T., Liu, X.: Dynamic system modelling using partically recurrent neural networks. J. of System Engineering, 2(4), 213- 223, 1992.
8. Sagiroglu S.: Identification of linear dynamic systems using multi-layered perceptrons. In: Proc. of the Int. Symposium Computer and Information Science XI, Antalya, Turkey, 2, 517-526, Nov. 6-8, 1996.
9. Elmas C., Sagiroglu S., Çolak I., Bal G.: Modelling of a nonlinear switched reluctance drive based on artificial neural networks. In: Proc. of the IEE PEVD'94, 5 Int. Conf. on Power Electronics and Variable Speed-Drives, London, UK, 7-14, Oct. 26-28 1994.
10. Pham D. T., Sagiroglu S.: Three methods of training multi-layer perceptrons to model a robot sensor. Robotica, 13, 531-538, 1995.
11. Pham D. T., Sagiroglu S.: Processing signal from an inertial sensor using neural networks. Int. J. Machine Tools Manufacture, 36 (11), 1291-1306, 1996.
12. Pham D. T., Sagiroglu S.: Synergistic neural models of a robot sensor for part location. IMechE, 210, 69-76, 1996.
13. Alippi, C. and Piuri, V.: Experimental neural networks for prediction and identification, IEEE Transactions on Instrumentation and Measurement, 45(2), 670-676, April 1996.
14. Bhama, S. and Singh, H.: Single layer neural networks for linear system identification using gradient descent technique. IEEE Transactions on Neural Networks, 4(5), 884-888, 1993.
15. Galvan, J.B. and Ilzarbe, M.J.P: Two neural networks for solving the linear system identification problem, International Conference on Neural Networks ,IEEE World Congress on Computational Intelligence, 5, 3226-3231.
16. Perry F.R.: Parametric modelling of linear and nonlinear systems, Ph.D. Dissertation. Naval Postgraduate School, Monterey, CA, June 1980.
17. Perry F.R., Parker S. R.: Transitions formulas for zero pole modelling, IEEE Trans. ASSP, 32, 178-180, 1984.
18. Rumelhart D. E., McClelland J. L.: Parallel Distributed Processing, 1, MIT Press, Cambridge, MA, 1986.