Intelligent bearing faults diagnosis in non-stationary conditions based on angular resampling and support vector machine

Over the past decades, intelligent bearing diagnostic methods have become a research hotspot. These methods require the construction of a Feature Vector (FV) generally composed of indicators calculated from time sampled vibration signals. However, in non-stationary conditions, these signals require the application of complex methods whose calculation time is really important. For this reason, the use of angular resampling techniques is recommended because they make it possible to get rid of speed fluctuations and to employ simple processing methods. In this work, we propose to use angularly resampled acceleration signals for intelligent bearing fault diagnosis under non-stationary conditions. It is the question of comparing three types of FV: classic angular indicators on angular signals x(theta), original order spectrum indicators (peak amplitude), or the combination of the two previous families of indicators. Then, the selection phase is performed by the Minimum Redundancy Maximum Relevance (MRMR) algorithm to select the most relevant features. Finally, the classification is carried out by a cubic support vector machine (SVM) for the detection and identification stages of various bearings fault conditions. The effectiveness of the proposed method achieves a perfect classification rate of 100%.