Whole course fault diagnosis based on fuzzy dynamical model

A robust method for whole course fault detection and diagnosis in control systems is proposed. The diagnosis scheme is based on the fuzzy dynamical model that is an aggregation of a set of local linear models representing the local dynamic behavior of the whole system. Then a responding observer-based fault diagnosis scheme is developed to detect and isolate sensors´ faults. In order to quantitatively evaluate the robustness of the generated residual, we present a performance index based on the analysis of residual. Genetic algorithm is then used to optimize the performance index and find the optimal matrices for the generation of robust residual. The diagnosis strategy is verified through simulation studies on temperature sensors fault diagnosis in the superheated steam control system of a power plant.