Fault diagnosis method combining multi-relation indexes with D-S evidence theory

Abstract

The fault diagnosis method based on grey relation analysis needs choosing reference pattern vectors which have a strongly ability of classify and identifying fault, otherwise the veracity and reliability of fault diagnosis can be greatly reduced. On basis of traditional grey relation analysis, multi-samples were adopted as reference signals and the relation indexes between multi-sample reference signals and the signal to be diagnosed are calculated by grey relation analysis method and normalized as the mass or basic probability assignment function which are fused to realize fault diagnosis in term of D-S evidence theory. The method provided in this paper is applied to the fault diagnosis of some reducer case operating state. The simulation result is shown that the reliability of fault diagnosis can be improved by fusion and the uncertainty of fault diagnosis depending on single reference pattern vector can be eliminated too.