Classification of fault location and the degree of performance degradation of a rolling bearing based on an improved hyper-sphere-structured multi-class support vector machine

Effective classification of a rolling bearing fault location and especially its degree of performance degradation provides an important basis for appropriate fault judgment and processing. Two methods are introduced to extract features of the rolling bearing vibration signal—one combining empirical mode decomposition (EMD) with the autoregressive model, whose model parameters and variances of the remnant can be obtained using the Yule–Walker or Ulrych–Clayton method, and the other combining EMD with singular value decomposition. Feature vector matrices obtained are then regarded as the input of the improved hyper-sphere-structured multi-class support vector machine (HSSMC-SVM) for classification. Thereby, multi-status intelligent diagnosis of normal rolling bearings and faulty rolling bearings at different locations and the degrees of performance degradation of the faulty rolling bearings can be achieved simultaneously. Experimental results show that EMD combined with singular value decomposition and the improved HSSMC-SVM intelligent method requires less time and has a higher recognition rate.

► A diagnosis method is introduced for fault location and the degree of the rolling bearing.
► Two methods that EMD combined with AR model and SVD were used for extracting feature.
► Improved HSSMC–SVM method is applied to multi-fault condition classification.
► Method that EMD combined with SVD and the improved HSSMC–SVM is more effective.