Method of fault feature extraction based on EMD sample entropy and LLTSA

A fault feature extraction method based on the empirical mode decomposition (EMD), sample entropy and manifold learning was presented to account for a range of issues of the vibration signal, e.g. Nonlinearities, non-stationary and weak fault features hard to extract. The proposed method combined the EMD, sample entropy and manifold learning techniques. Firstly, on the basis of the property of adaptive multi-resolution for the EMD technique, the sample entropy of the IMF (intrinsic mode function) signal reconstructed by using the EMD was calculated, and the state features of the rolling bearing were preliminarily extracted. Secondly, the extraction performance of the state features was further implemented by using the manifold learning technique. Finally, the SVM(support vector machine) was employed to classify and to evaluate the feature extraction method. Moreover, the proposed method was applied to the experiment of the rolling bearing fault diagnosis. The experimental results show that the proposed fault feature extraction method has more robust clustering performance than the fault diagnosis method based on the sample entropy of wavelet packets. Furthermore, a relatively high precision, namely, 100% of classification result for the SVM, can be obtained. The proposed method not only decreases the complexity of the feature data, but also enhances the classification performance of fault diagnosis and pattern recognition, thus bringing about certain superiority.