Fault Diagnosis of Roller Bearings Using Wavelet Transform and EEMD-Mahalanobis Distance

The acoustic emission signal of mechanical faults is usually mixed with various kinds of interference and noise. In this article, a method of fault diagnosis of roller bearings was proposed using wavelet transform and EEMD-mahala- nobis distance. First of all, the original acoustic emission signals were disposed by wavelet-denoising and decomposed into several stationary intrinsic mode functions (IMF) by EEMD. Then, the false IMFs of EEMD were eliminated by mahalano- bis distance method so that the IMF components which could reflect the characteristics of bearing faults could be extracted. Finally, the Fourier spectrum of the transient Teager energy was used to recognize the characteristic frequencies of the bear- ing faults. Comparison of simulation signal with the measurement emission signal of the bearing with outer race faults show that the method can effectively remove the noise in the fault mixed signals, and identify the location of the bearing fault accu- rately.