Fault detection for Takagi–Sugeno fuzzy systems using multiple observers and ellipsoidal analysis

Abstract

Abstract This paper proposes a novel fault detection method for Takagi–Sugeno fuzzy systems via multiple observer design and ellipsoidal analysis. By considering only one fault distribution matrix at a time, the original system is divided into a limited number of fault modes. For each fault mode, an observer with finite‐frequency fault sensitivity performance and full‐frequency unknown inputs robustness performance is designed to generate the residual signal. Based on the designed observer, the ellipsoidal analysis is used to evaluate a residual and to give a fault detection index for each mode. In addition, the fault detection result is provided by combining “or” logical operation of the multiple indices. Numerical simulations and an AC servo motor experimental platform are used to verify the viability and validity of the proposed method.