The Rolling Bearing Fault Diagnosis Based On Wavelet Packet Analysis

The rolling bearing is a common and prone to damage mechanical component.Its working condition has influence on normal operation and product safety in entiredevice group directly. In the process of bearing fault diagnosis, especially the initialfault, the fault feature information which is very weak and often submerged in thestrong noise signal. Signal-noise ratio is very low, the bearing fault is difficult toeffectively extract and diagnose, if not promptly treated will lead to seriousconsequences. Every year for bearing fault caused economic losses more than severalbillions. So effectively detect the rolling bearing fault at the beginning to preventaccidents is very important.This paper is based on the vibration signal testing, the research of rollingbearings entropy and fault feature based on basic theory of wavelet packet, combinedwith EMD method, support vector machines.This paper does the following aspects ofcontents:Firstly, this paper studies the basic theory of wavelet packet、the developmenthistory and give the discriminated method of judge the usual faults, and expounds thetheory of wavelet packet to the impact and significance of machine science. The faulttype of rolling bearings, vibration mechanism and frequency characteristics areanalyzed.Secondly, the theory of wavelet packet theory and EMD theory was studied bythe simulation signal. Signal-noise separation was achieved by wavelet packet; thede-noising effect of the wavelet packet method was obvious. Extracting Componentswith cross-correlation、kurtosis criterion by EMD decomposed de-noised signalavoids blindness of IMF components selection. Meanwhile the traditional waveletpacket has been improved; Not only to maintain the signal frequency resolution, butalso to prevent sub-band signals generate spurious frequency causing seriousfrequency aliasing effect. And using EMD decomposes sub-band signal which in faultfrequency range, can accurately find the bearing fault frequencies. Then, Jiangsu Qian Peng Company produced rolling bearings which in gearbox as the experimentalresearch object, to detect and validate the above method, the results are verysatisfactory.Thirdly, wavelet packet and EEMD detect bearing faults vibration signal. EMDand EEMD were compared to understand EEMD method can effectively reducemodal aliasing degree which in EMD method, but the decomposition rate slower thanEMD.Lastly, Wavelet Packet and Support Vector Machine combined for AmericaSpectra Quest Company produced rolling bearings intelligent fault type classification.The final experiment results show that wavelet packet and EMD can extract andexpress the fault information of the bearing outer ring. Wavelet packet and supportvector machine can effective and accurate distinguish the type of bearing faults, so itcan be seen that this method effectively solves the fault problem of rolling bearing.