Research and realization of sensor fault-tolerance of fusion diagnosis system using evidence redistribution

Abnormal sensor signals may cause missed diagnosis in the fusion fault diagnosis system based on evidence theory. To diagnose the fault in time with abnormal sensor signals, fault-tolerance is required. An evidence redistribution method is proposed for the sensor fault-tolerance of fusion diagnosis system. The evidences directly calculated from sensor characteristic parameters are updated by an evidence redistribution function. Fusion experiments are done on the hardware experimental platform for vibration isolation. The diagnosed fault is vibration divergence. The fusion diagnosis results on the situations of misconnection, disconnection, multichannel abnormity and normal sensor signals are analysed and compared. Experimental results show that the evidence redistribution method can solve the sensor fault-tolerance problem effectively.