RUL prediction for railway vehicle bearings based on fault diagnosis

In recent years remaining useful life prediction of metro bearing is paid much more attention. Different faults have different prediction models for bearings. In this paper, the remaining useful life prediction based on fault diagnosis is proposed. With the aid of the cyclic stability characteristic of the bearing signal, a spectral correlation density combination method is proposed and used to fault diagnosis and identify the fault scale. In order to ensure the accuracy of analysis and mixed fault, the variable predictive model is used to detect the fault type. As for the remaining useful life prediction, a semi supervised co-training based approach is used to solve the problem of lacking fault data. The effectiveness and better prediction accuracy of this method are demonstrated by a case study.