Adaptive control based on fuzzy process model with estimation of premise variables

Abstract

An adaptive control method based on Takagi-Sugeno fuzzy process model is proposed. It is applicable in cases when the variables in the premises of fuzzy rules, which determine the operating regime of the system, are not measurable. The process dynamics in different operating regimes is described by local linear models, which are combined using fuzzy rules. The premise variables of the fuzzy rules are estimated by minimizing a performance index of the local linear models. The proposed strategy uses the prior knowledge of the process in form of local process models identified offline and stored in the controller's database to simplify the estimation procedure. Thereby, the recursive least-squares identification algorithm used in classic self-tuning control is substituted by much simpler least-squares estimation of a small number of parameters. This makes the proposed method appropriate for implementation on simple platforms, providing, in the same time, the adaptation to changes in operating conditions. The proposed method is experimentally tested on a laboratory liquid level rig. The performance of the proposed control algorithm is compared to the performance of a PI controller.