Fault Detection Approaches for Fuzzy Large-Scale Systems With Unknown Membership Functions

This paper investigates the decentralized fault detection (FD) problem within a type of nonlinear large-scale systems under the parameter uncertainties constraints. First of all, a nonlinear system is treated as the T-S fuzzy large-scale model with unknown membership functions. Then, a switching method is employed in the FD filter integration. Combining the local measurements of each subsystem and the lower and upper bounds information collected from the unknown membership functions, a new decentralized FD filter is built. A cyclic-small-gain condition is introduced to guarantee that the resulted augmented FD system is asymptotically stable with a satisfying H_∞ performance. The comparison results show that the proposed switching-type decentralized FD filter can achieve a better FD performance than linear filters. Finally, the validity and superiority of the proposed method are verified with two examples.

I. Introduction

Recently, more and more researchers are paying their attention to the fault detection (FD) problem of control systems because of the increasing requirement for system security and reliability among engineering systems [1]. Among various FD methods, FD scheme has been widely discussed. The core idea behind an FD scheme is to define a residual evaluation function, and then compare the residual evaluation value with a predefined threshold, such as [2]–[9]. And, more remarkable, many large-scale practical systems appear with the developments of science, for example, power generation systems and distributed systems. The fault diagnosis of these large-scale systems is becoming more obvious and important.