Rotating Machinery Fault Diagnosis Based on EMD-Approximate Entropy and LS-SVM

Abstract

The precision of fault feature extraction as well as the efficiency of classification and identification is the key point of improving the accuracy and speed of fault diagnosis. Accordingly, a new fault diagnosis method based on EMD (empirical mode decomposition)-approximate entropy and LS-SVM (least square support vector machine) was proposed. The localization characteristics of EMD and approximate entropy were utilized to quantify the fault feature, and then the LS-SVM was combined to identify fault types. Firstly, the vibration signal of fault was decomposed into some IMFs (intrinsic mode functions) by EMD, and then the approximate entropies of the first four IMFs were taken as the eigenvectors. Finally, the eigenvectors were input into LS-SVM classifier for fault type recognition. Simulation results show that this proposed method not only can effectively extract fault features, but also compared with BP (back propagation) networks, the LS-SVM network has smaller samples, shorter training time and higher recognition rate.