Data-driven fault detection for closed-loop T-S fuzzy systems with unknown system dynamics and its application to aero-engines

This paper considers the fault detection problem of closed-loop Takagi-Sugeno (T -S) fuzzy systems with unknown system dynamics. However, the unknown dynamics make the modelbased detection methods being infeasible. To tackle this problem, a detection scheme is designed directly by using input/state data, and disturbance attenuation as well as fault sensitivity performance are then introduced within data-driven framework such that a multiobjective optimization problem is formulated to compute the parameters of detector. Moreover, to remove the existing limitation that sensor faults must occur, corresponding design conditions of detector are developed by considering the characteristics of fault frequency. In particular, a linearization approach via searching the minimum singular value of unknown matrix is further developed to handle the nonconvex problem caused by introducing the fault sensitivity performance. Finally, an aero-engine system is used to show the effectiveness and advantages of the developed method.