UDK 62-506.1
ALGORITHMS OF WIENER SYSTEM IDENTIFICATION
N. V. Koplyarova
Siberian Federal University 79, Svobodnyi Аv., Krasnoyarsk, 660041, Russian Federation Е-mail: koplyarovanv@mail. ru
This article is devoted to the problem of identification of nonlinear dynamic systems Wiener class under conditions of incomplete information. It is now widely known parametric identification methods such systems, which often have a discrete-continuous. Normally, when a parametric formulation of the problem identification class of equations describing the dynamic process is known up to a vector of parameters. The next stage is the estimation of parameters from observations of “input-output” process variables. It is important to note that the choice of the parametric model structure of the object is extremely important. Model inaccuracies arising from some errors at the stage of its parametric selection or definition, can not be rectified in the evaluation parameters. This means that the model is in one sense or another, quite rough. In this paper, the linear element model Wiener is not known with an accuracy of parameters, which corresponds to the non-parametric uncertainty. As a non-linear element model is adopted one or other parametric structure. Specifically, we consider the case when the parameter block is represented as a quad and a saturation level. Thus, the problem of identification of objects of a class of stochastic Wiener is seen in partial non-parametric uncertainty. At the first stage model of linear dynamic block is built. To construct a non-parametric model of the last input object must submit the Heaviside function, in this case, the output of the object to within a factor is its transition function. Reconstruction of the weight function is carried out by the observations of transitional methods of nonparametric statistics. To estimate the parameters of non-linear element is necessary to conduct appropriate experiments. We should pay particular attention to the fact that in the identification of nonlinear dynamic system Wiener class, subject only to the control input and output variables. This situation is typical not only in the manufacture of spacecraft, but many of their units and components. Furthermore, these models are useful in creating computer systems of technical diagnostics with vibration testing of spacecraft (SC) on the channel: Vibrate - sensor mounted on the spacecraft. A numerical study of the proposed algorithms, consider the model of the class of Wiener in different conditions (at different levels of noise in the measurement channels, a different sample size and types of input actions). The results of computer studies show efficiency of the proposed algorithms.
identification, Wiener model, nonparametric statistics, nonlinear dynamical system, a priori information.
References
  1. Sung S. W., Lee J. Modeling and control of Wiener-type processes. Chemical Engineering Science, 2004, no. 59, Р. 1515–1521.
  2. Dorf R. C., Bishop R. H. Sovremennye sistemy upravleniya. Per. s angl. B. I. Kopylova. [Modern Control Systems. Per. from English. B. I. Kopylova]. Мoscow, Laboratoriya bazovykh znaniy Publ., 2002, 832 р.
  3. Keesman K. J. System identification. An introduction. London: Springer, 2011. 351 p.
  4. Cypkin Ja. Z. Osnovy informatsionnoi teorii identifikatsii [Fundamentals of the informational theory of identification]. Moscow, Nauka Publ., 1984, 320 p.
  5. Pupkov K. A., Egupov N. D. Metody klassicheskoi i sovremennoi teorii avtomaticheskogo upravleniya. T1: Matematicheskie modeli, dinamicheskie kharakteristiki i analiz sistem upravleniya [Methods of classical and modern control theory. T1: Mathematical models, dynamic characteristics and analysis of control systems]. Moscow, MGTU name. N. E. Baumana Publ., 2004, 656 p.
  6. Eykhoff P. Osnovy identifikatsii sistem upravleniya [Identity-based control systems]. Moscow, Mir Publ, 1975, 681 p.
  7. Medvedev A. V. [Nonparametric algorithms of nonlinear dynamical system identification]. Stokhasticheskie sistemy upravleniya [Stochastic control system]. Novosibirsk, Nauka Publ., 1979, Р. 15–22 (In Russ.).
  8. Medvedev A. V. [Nonparametric systems theory. Modeling]. Vestnic SibGAU, 2010, no. 4 (30), з. 4–9 (In Russ.).
  9. Nadaraya E. A. Neparametricheskoe otsenivanie plotnosti veroyatnostei i krivoi regressii [Nonparametric estimation of probability density function and regression curve]. Tbilissi, Tbilissi Univ. Publ, 1983, 194 p.
  10. Medvedev A. V. Neparametricheskie sistemi adaptatsii [Nonparametric adaptation systems]. Novosibirsk, Nauka Publ., 1983б 173 p.
  11. Medvedev A. V. [Nonparametric systems theory. Control]. Vestnic SibGAU, 2013, no. 2(48), p. 57–63 (In Russ.).
  12. Popkov U. S. Identifikatsiya i optimizatsiya nelineinykh stokhasticheskikh sistem [Identification and optimization of nonlinear stochastic systems]. Moscow, Energya Publ., 1976, 440 p.
  13. Chaika S. N. [To dynamical systems of Hammerstein class identification with partially parameterized model structure]. Dinamika sistem: upravlenie i optimizatsiya [System dynamics: Control and optimization]. Gorki, Gorki university Publ, 1989, P. 24–36 (In Russ.).
  14. Koplyarova N. V., Sergeeva N. A. [Nonparametric algorithms for system identification of Wiener and Hammerstein class]. Sistemy upravleniya i informatsionnie tekhnologii, 2013, no. № 2.1 (52) (In Russ.).
  15. Koplyarova N. V., Sergeeva N. A. [About the nonparametric algorithms of nonlinear dynamical processes identification]. Vestnic SibGAU, 2012, no. 5(51),  p. 39–44 (In Russ.).

Koplyarova Nadezhda Vladimirovna – assistant, Information systems department, Institute of Information and Space Technologies, Siberian Federal University. Е-mail: koplyarovanv@mail.ru