An approach for the state estimation of Takagi-Sugeno models and application to sensor fault diagnosis

In this paper, a new method to design an observer for nonlinear systems described by Takagi-Sugeno (TS) model, with unmeasurable premise variables, is proposed. Most of existing work on TS models consider models with measurable decision variables. As a consequence, these works cannot be applied when the decision variables are not available to measurement. The idea of the proposed approach is to rewrite the TS model with unmeasurable premise variable into an uncertain TS model by introducing the estimated state in the model. The convergence of the state estimation error is studied using the Lyapunov theory and the stability conditions are given in terms of Linear Matrix Inequalities (LMIs). Finally, an academic example is given to illustrate the proposed approach, with an application to sensor fault detection and isolation using an observer bank.