Bearing Health Monitoring and Fault Diagnosis Using Intelligent Methods

Abstract

Rolling element bearings form a very common and imperative component in almost all types of newlinerotating machines. There are many causes due to which theses bearings get damaged which newlineinclude mainly wear and tear, aging, environmental effects, incorrect mounting, improper bearing newlinelubrication, fatigue etc. The defective bearing often results in reduced efficiency or even severe newlinedamage to the machine under consideration. Therefore, bearing health monitoring and fault newlinediagnosis have received great attention in last many years, which can be conducted based on newlineinformation carriers such as acoustic emission, stress waveform, oil analysis, temperature, newlinevibration etc. The commonly used technique for fault detection is vibration monitoring and newlineanalysis, which offers very important information about anomalies formed in the internal newlinestructure of the bearings. In this work, an experimental setup is prepared to capture the vibration newlineof faulty sample bearings and two new methods are developed for bearing health monitoring and newlinefault diagnosis. newlineThe first method expounds a novel system which includes generation of unique patterns called newlinesignatures of various bearing faults using continuous wavelet transform (CWT) and recognition newlineof these signatures using the neural network. newlineIn the second method, Ensemble Empirical Mode Decomposition (EEMD) is used for extracting newlinethe features in the form of Intrinsic Mode Functions (IMF) values. These IMF values are further newlineused with ANN and knowledge based expert system for recognition and classification of the newlinebearing faults