An Improved EMD with Second Generation Wavelet and Feature Extraction for Fault Diagnosis of Rotating Machinery

Fault feature extraction is a challenge for fault diagnosis of rotating machinery. The vibration signals measured from rotating machinery are usually non- stationary and nonlinear. Especially, the useful fault characteristics are too weak to be identified at the early stage. In order to solve the problem, a novel method called improved empirical mode decomposition (EMD) with second generation wavelet for fault diagnosis of rotating machinery is proposed. According to the local characteristics of vibration signal and selecting the proper criterion of minimizing the squared error, an optimal predicting operator is constructed for a transforming sample, so that the second generation wavelet basis function is able to fit the local characteristics of the vibration signal. Using the self- adaptive second generation wavelet as the pre-filter to improve EMD decomposition results, EMD is further improved to increase the accuracy and effectiveness of the decomposition results. The proposed method is applied to analyze the rub-impact rotor experimental setup, and the results show that the proposed method is accurate and efficient, and is expected to be applied in engineering practice effectively.