Application of Multi-scale Principal Component Analysis and SVM to the Motor Fault Diagnosis

Multi-scale principal component analysis (MSPCA) and support vector machine (SVM) are the modern methods, which have much application in classifications. A novel application of them in the motor fault diagnosis is proposed. The multi-scales PCA models are constructed by T2 and Q statistics. As the signal features, T2 and Q statistics are fed to train SVM to diagnose fault. The accuracy of monitoring and fault diagnosis is improved and the experiments illustrate the efficiency of the proposed approach.