Fault Detection of Rolling Element Bearings using Advanced Signal Processing Technique

In rotating machinery, rolling element bearings are one of the most critical components and a large majority of system failures arise from faulty bearings. Hence, there is an increasing demand to find an effective and reliable condition monitoring technique. In this paper, a procedure for detecting various types of bearing faults using thermal imaging is presented and assessed. Five different fault cases are tested: No Fault (NF), Line Fault (LF), Small Circle Fault (SCF), Double Line Fault (DLF), and Large Circle Fault (LCF). Experiments were conducted on the BENTLY NEVADA RK4 Rotor Kit. A novel signal processing algorithm is proposed to detect the faults which involves utilizing the Discrete Wavelet Transform (DWT). The HAAR wavelet is used as the mother wavelet and a decomposition level of 7 is used. The inverse of HAAR is applied on the decomposed signal and the envelop spectrum is plotted. A classifier is then created to identify the fault that utilizes the Support Vector Machine (SVM) classifier to extract the features and a confusion matrix is developed to detect the prediction accuracy.