Intelligent fault diagnosis of roller bearings with multivariable ensemble-based incremental support vector machine

Since roller bearings are the key components in rotating machinery, detecting incipient failure occurring in bearings is an essential attempt to assure machinery operational safety. With a view to design a well intelligent system that can effectively correlate multiple monitored variables with corresponding defect types, a novel intelligent fault diagnosis method with multivariable ensemble-based incremental support vector machine (MEISVM) is proposed, which is testified on a benchmark of roller bearing experiment in comparison with other methods. Moreover, the proposed method is applied in the intelligent fault diagnosis of locomotive roller bearings, which proves the capability of detecting multiple faults including complex compound faults and different severe degrees with the same fault. Both experimental and engineering test results illustrate that the proposed method is effective in intelligent fault diagnosis of roller bearings from vibration signals.