Study of Rolling Bearing SVM Pattern Recognition Based on Correlation Dimension of IMF

A method of pattern recognition based on correlation of intrinsic mode function (IMF) and Support Vector Machine (SVM) was proposed. Firstly, the rolling bearing vibration signal was decomposed into a finite series of IMFS by EMD. Secondly, useful IMFS which contained main fault information were chosen through correlation coefficient threshold filtering method. Finally, the correlation dimensions of the main IMFS were computed and served as input characteristic parameters of SVM classifiers to classify normal state, outer and inner fault of the rolling bearing. The method has been applied on pattern recognition of the NO. 6205 rolling bearing. The results show that the proposed approach can identify the working state and fault pattern for the bearing system accurately and effectively and provide a reliable way for the fault diagnosis of mechanical device in the electrical power system.