Application of fault detection based on adaptive threshold in the DAMDDICS benchmark problem

A robust fault detection approach based on self-adaptive threshold was presented to detect the fault of nonlinear systems with uncertain modeling errors, noise and disturbance. With this method, model was constructed by using multi-layer perception neural network, and the adaptive threshold interval was estimated by using outer bounding ellipsoid algorithm. Then, the online fault detection strategy was designed by using weighted moving average remnant and adaptive threshold interval. Finally, the proposed approach was applied in detection of the faults proposed in an industrial actuator used as an Fault Detection and Isolation(FDI)benchmark in the DAMADICS. Simulation results demonstrated the effectiveness of this approach.