Research On Fault Diagnosis Of Ball Bearings Based On Vibration Signal

Ball bearings are considered as critical mechanical components, and statistics show that 30% of defaults in rotating machinery are caused by rolling bearings.Fault diagnosis of ball bearing based on vibration signal is studied in this paper:Wavelet analysis is applied for feature extraction from vibration. As a multi-resolution time-frequency analysis method, wavelet analysis is a powerful mathematical tool for non-stationary signal processing. In this paper wavelet transform is applied to analyze vibration, and the envelope spectrum of wavelet coefficients is made use of to evaluate default characteristic frequency. Consequently, double the default frequency is obtained; however the result is not so satisfactory.Wavelet packet analysis is applied for feature extraction of vibration. In order to extract fault features of ball bearings more precisely, wavelet packet decomposition is used to refine envelope spectrum analysis, and then accurate failure frequency is obtained successfully. After analysis and comparison of extracted statistics index in time domain from vibration frequency band energy is chosen as characteristic parameter for fault diagnosis.Support Vector Machines Classification Methods is used to recognize fault type. Support Vector Machines based on statistical learning theory is introduced to overcome the lack of samples.On the basis of binary classification,the multi-category classifier is built to classify the characteristic vectors and identify failure modes.The experimental simulation result proves the correctness of classification and identification of failure modes ,which indicate that Wavelet Packet Analysis and Support Vector Machines is an excellently effective method for Fault diagnosis of ball bearings.