FAULT DIAGNOSTICS AND PROGNOSTICS OF HIGH SPEED BEARINGS

Abstract

Bearings are very important components of rotating machines and bearing defects, if not detected in time, may cause malfunctioning and catastrophic failures of machinery. Hence online bearing fault diagnostics have drawn great attention of researchers. Many techniques have been developed for the same. But more and more efforts are taken to achieve greater accuracies in this held of bearing fault diagnostics. Along with fault diagnostics, condition based maintenance has played vital role in development of need of bearing fault prognostics. Though it has drawn attention in last decade or so, much progress has been noted in the field of bearing fault prognostics. The main aim of bearing fault prognostics have been the estimation of remaining useful life of bearing. The current study is focused on localised defects of bearings. Frequency based and time-frequency based techniques have been adopted for fault detection. The vibration signals for various defected and healthy bearings have been captured at varying shaft speed and varying radial load. Various features have been extracted from these signals. The extracted features have been used for fault diagnostics and prognostics. The different data mining techniques, based on neural networks have been studied and compared for classification of bearings based on faults induced on them and based on time at which signals have been captured. Different classihiers along with filters have been compared for detecting best suitable classifier for fault diagnostics and prognostics