Evidence Theory-Based Approach to Sensors Multiple Fault Diagnosis

This paper describes a new method for sensors multiple fault diagnosis and isolation. The information fusion method is based on expanded evidence theory, which offers a new combination rule under different but compatible frames of discernment. By this method, the maximum of available knowledge supported by each source of information is exploited and the uncertainty of the effective state between the potential states of a sensor is decreased. In addition of efficient fault detection and isolation results, the modularized RBF neural network is adopted to get basic probability assignment function of sensor state, which overcomes the disadvantage of being unusable after input parameters changed. Simulation tests demonstrate that the diagnosis strategy works effectively in multisensor fault diagnosis