Active Fault Diagnosis for LPV Systems: An Index-Based Approach

This article investigates the active fault diagnosis (AFD) problem for linear parameter-varying (LPV) systems. The system uncertainties, including system disturbances, noises, and initial conditions, are characterized by uncertain Gaussian zonotopes (UGZs). An index-based AFD scheme is proposed, which can avoid the requirement of complete separation of output sets of different system modes and reduce conservatism to a certain extent. A new index is proposed to characterize the similarity between two zonotopes. According to this index, the auxiliary input is designed by solving a quadratic programming problem. A novel fault diagnosis strategy is proposed to determine which mode the system is actually operating in. A numerical example is presented to demonstrate the effectiveness of the proposed approach.