Constellation, le dépôt institutionnel de l'Université du Québec à Chicoutimi

Résumé

Nonlinear systems identification from experimental data without any prior knowledge of the system parameters is a challenge in control and process diagnostic. It determines mathematical model pa-rameters that are able to reproduce the dynamic behavior of a system. This paper combines two fun-damental research areas: MIMO state space system identification and nonlinear control system. This combination produces a technique that leads to robust stabilization of a nonlinear Takagi-Sugeno fuzzy system (T-S). Design/methodology/approach The first part of this paper describes the identification based on the Numerical algorithm for Subspace State Space System IDentification (N4SID). The second part, from the identified models of first part, explains how we use the interpolation of Linear Time Invariants (LTI) models to build a nonlinear multiple model system, T-S model. For demonstration purposes, conditions on stability and stabiliza-tion of discrete time, Takagi-Sugeno (T-S) model were discussed. Findings Stability analysis based on the quadratic Lyapunov function to simplify implementation was ex-plained in this paper. The LMIs (Linear Matrix Inequalities) technique obtained from the linearization of the BMIs (Bilinear Matrix Inequalities) was computed. The suggested N4SID2 algorithm had the smallest error value compared to other algorithms for all estimated system matrices. Originality The stabilization of the closed-loop discrete time T-S system, using the improved PDC control law (Parallel Distributed Compensation), was discussed to reconstruct the state from nonlinear Luen-berger observers.