Metodologia nebulosa evolutiva baseada em modelo de Hammerstein no espaço de estados para identificação de sistemas dinâmicos não lineares multivariáveis

Abstract

In this paper, a novel evolving fuzzy methodology based on state-space Hammerstein model is presented for identification of multivariable nonlinear dynamic systems. In this approach, the static nonlinear block is parameterized through an evolving fuzzy system. This, in turn, is composed of an evolving fuzzy clustering algorithm based on the similarity of the experimental data provided by the dynamic system for estimation of the antecedent, and by a recursive fuzzy eigensystem realization algorithm (RF-ERA) for estimation of the state space linear submodels of the consequent. The linear dynamic block is parameterized via recursive eigensystem realization algorithm (R-ERA). The computational and experimental results to evaluate the proposed methodology consist of: identification of a thermal process; identification of dynamic systems widely used as benchmark in the literature and identification of an evaporator process. The results are evaluated both from point of view of the input and output relationship and from online estimation of the static nonlinearity of the dynamic system.