On-line Active Fault Detection Based on Set-membership Ellipsoid and Moving Window

Abstract

On-line active fault detection (AFD) and its optimization problems are proposed based on the set-membership ellipsoid technique in order to solve the problem of on-line fault detection. The design of auxiliary input signal should satisfy two conditions: the signal amplitude is small enough without obvious impact on the system, and it simultaneously separates the system output in the normal and fault operation. Here we describe the output set as an ellipsoid under the framework of set-membership. The system model of moving window is established based on the parity space, and the equivalent optimization design of auxiliary input signal is solved based on this model. The proposed method can significantly reduce the complexity of the optimization calculation and conveniently obtain the auxiliary input signal on-line. The system fault is detected more intuitively by comparing the degree of separation between the output ellipsoid of the actual system and that of the identification normal (or fault) model. The simulation results on a general example verify the effectiveness of the proposed method.