Fault Diagnosis of Sensor Network Using Information Fusion Defined on Different Reference Sets

This paper describes a novel scheme for fault diagnosis of sensor network based on different frame of discernment information fusion using evidence theory. The information of multisensor redundant or complementary in space or time fusion by the RBF neural network is adopted. The RBF neural network is used as modularization overcome the disadvantage of unusable for input parameters changed. A new combination rule under different but compatible frame of discernment is presented. By the combination operation, 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. This combination guarantees the fault isolability from a practical point of view and is suitable for multiple faults occurring at the same time. Simulation tests demonstrate that the diagnosis strategy works effectively in fault diagnosis of sensor network.