The Research Of Rolling Bearing Fault Diagnosis Based On HHT Transform

One of the most important, critical and difficult problem is feature signal of faultinformation extraction.The traditionally time-frequency analysis methods such as the short-time Fouriertransform, Winger-Ville distribution, spectrum, wavelet transform and wavelet packetanalysis, as well as the limitations in the application of these methods in the non-stationarysignals such as rolling bearing fault signal have been described in this paper. On this basis,Hilbert Huang transform (HHT) has been introduced. HHT is a kind of adaptivetime-frequency analysis method, and it can be adaptively time-frequency decomposedbased on local time-frequency characteristics of the signal. It overcomes the traditionalmethods with meaningless harmonic component to represent the non-stationary signals,and can get high time-frequency resolution. It is very suitable for analysis ofnon-stationary signal. Based on the study of the application of HHT in the rolling bearingfault diagnosis, the main work including the following four parts has been finished in thispaper:The rolling bearing operation characteristics and failure mechanism and faultcharacteristics have been discussed, besides, aiming at nonstationary and nonlinearcharacteristics of rolling bearings. The traditional diagnostic methods’ limitations havebeen summarized in rolling bearing fault diagnosisApplications of Hilbert Huang transform on Roller Bearing Fault Diagnosis arestudied. Aming at the problems about Hilbert-Huang transform, currently, this paper putsforward an improved discrete cosine transform (DCT) solution to solve the modal mixing,and the energy threshold and sensitive IMF select is combined to identify true and falseIMF and sensitive IMF.The method based on improved EMD in rolling bearing fault diagnosis, and themethod based on improved DCT and EMD in the rolling bearing fault diagnosis and themethod based on improved DCT-EMD in rolling bearing fault diagnosis have been givenbased on the improved Hilbert-Huang transform, and the feasibility and accuracy of thesemethods have been proven.