FAULT DIAGNOSIS OF HIGH SPEED ROLLING ELEMENT BEARING USING ACOUSTIC EMISSION > TECH NI QUES

Abstract

Acoustic emission (AE) is the phenomenon of transient elastic wave generation in materials under stress. When the material is subjected to stress at a certain level, a rapid release of strain energy takes place in the form of elastic waves which can be detected by transducers placed on it. Fault diagnosis of high speed rolling element bearing can be done by using acoustic emission technique. We have aim to work on experimental setup for getting different AE results, it will help to find out the capability of AE technique for detection of bearing abnormalities. The most commonly measured AE parameters are peak amplitude, r.m.s, energy, overall acoustic, power content, ring down counts and distribution of events by peak amplitude and ringdown counts and statistical parameters such as kurtosis, skewness, crest factor, impulse factor and shape factor. The Experiments are carried out for healthy as well as defective (inner race, outer race, ball, combined defect) bearings at various low speed and high speed. The statistical parameters as well as AE parameters (r.m.s., ringdown counts, distribution of events AE energy) on healthy as well as defective bearings has been evaluated. Application of advance signal processing techniques like Hubert Huang Transform for incipient fault diagnosis has been focused. Modified HI-IT based on spectral kurtosis has been developed and discussed. An experimental study conducted for fault diagnosis of ball bearings using various machine learning techniques such as ANN, SVM and KNN etc. These are found to be good in classification accuracy. Soft computing tool can be helpful in Prediction of bearing class. To optimize feature selection in data mining an asymmetric proximity function based KNN (APF-KNN) techniques has been developed and discussed. In this work, feature selection and optimization algorithm has been proposed to enhance the classification accuracy with asymmetry proximity function. The overall classification rate achieved is 96.667%.