Fault pattern recognition of rolling bearing based on singularity value decomposition and support vector machine

A novel fault diagnosis approach for rolling bearings based on singularity value decomposition(SVD) and support vector machine(SVM) was proposed. The key to the fault bearings diagnosis is condition feature extracting and fault feature classifying. Multidimensional correlated variables were converted into low dimensional independent main-eigenvector by means of singularity value decomposition. The pattern recognition and the nonlinear regression were achieved by the method of support vector machine. In the light of the feature of bearings vibration signals, main-eigenvector was obtained using singularity value decomposition, fault diagnosis of rolling bearing was recognized correspondingly using support vector machine multiple fault classifier. The experimental results show that the combination of main-eigenvector and support vector machine distinguish the normal and fault condition finely, and it also has good recognition ability to unknown fault samples. Comparing with the traditional artificial neural networks, the approach is more efficient, robust and precise.