Abstract:

In this paper, a decentralized fault detection (FD) system design approach is proposed for discrete-time large-scale interconnected systems. Such an FD system consists of two parts for each subsystem: a residual generator and a residual evaluator. The residual generator producing weighted output estimation errors is designed to match a proper reference residual model, such that it is robust against system disturbances and sensitive to system faults. The solution to a (sub)-optimal residual generator is given by solving a convex optimization problem with the help of linear matrix inequalities (LMIs). Then norm based evaluation functions are selected for each subsystem and the corresponding thresholds are presented for the residual evaluator design. The computation of thresholds are formulated also as an optimization problem, which can be solved by using LMIs. Finally, a numerical example is given to illustrate the results.