Recursive Identification Approach of Multivariable Nonlinear Dynamic Systems based on Evolving Fuzzy Hammerstein Models

Abstract

In this paper a novel approach for evolving fuzzy identification of nonlinear Multi-Input Multi-Output (MIMO) dynamic systems, is proposed. The adopted methodology is based on state space multivariable Hammerstein models, in which, the static nonlinearity block is approximated by an evolving Takagi-Sugeno fuzzy inference system and the linear dynamic block is represented by a state space linear model obtained through a Recursive Eigensystem Realization Algorithm (RERA). Computational and experimental results from the online identification of a multivariable dynamic system with a complex combined nonlinearity and an Evaporator Process with threeinput and three-output, which are benchmarks widely cited in the literature, illustrate the efficiency and applicability of the proposed approach.