Development of Expert System for Fault Diagnosis and Prognosis of Rolling Element Bearing Using Vibration Signature Analysis

Abstract

The growth of any country depends on its industrial and production performance. Rolling element bearing is the heart component of the industrial and machinery world. In most of the cases, the rotating machines are failed due to the failure of rolling element bearing, resulting in catastrophic and costly systems downtime. In the last decades, a lot of researches have been done on bearing faults investigations at incipient stages. newlineIn present work, an expert system has been developed for fault diagnosis of rolling element bearing using vibration signature analysis and advanced soft computing techniques. The methodology incorporated the experimental work as well as soft computing techniques like wavelet transform, energy entropy approaches, machine learning, and adaptive Neuro-Fuzzy inference system. newlineIn the proposed methodology, vibration signals in the time domain have been captured from an experimental test rig of healthy and different faulty bearings via accelerometer using FFT analyzer. The continuous wavelet transform (CWT) is employed to enhance signal quality and characteristics. The four real-valued and four complex-valued base wavelets are considered. Out of these, the optimal mother wavelet is selected on the basis of wavelet selection criterions to extract statistical features from wavelet coefficients of vibration signals. To choose an appropriate mother wavelet for statistical feature extraction, three wavelet selection criterions such as Maximum energy to Shannon entropy ratio (MEER), Maximum Information (MI), and Maximum relative Wavelet Energy (MRWE) are utilized and compared. On the basis of MEER and MI, Morlet wavelet and on behalf of MRWE, complex Morlet wavelet has been finalized. The statistical features of these wavelets are used as an input to machine newlineii newlinelearning techniques to classify the bearing faults. These machine learning techniques are Artificial neural network (ANN), Support vector machine (SVM), Decision Tree (DT), and k-Nearest Neighbors (kNN).