EMU Fault Prognosis Based on Improved HSMM

Abstract

Due to the limitations of fault diagnosis and forecast using traditional Hidden semi Markov model (HSMM), a model based on fissile PSO is proposed to optimize HSMM parameters. It can avoid local optimum effectively and expand the scope of the search without any impact on training rate. Finally, a more accurate HSMM model verified by experiments is established. The healthy state classifier and lifetime prediction model are established. It shows that the algorithm has high prediction accuracy and can be applied to the fault prediction of complex equipment.