Study On Feature Extraction And Analysis Technology For A Gearbox Fault Vibration Signal

Gearbox is the important part of the transmission system, and its performance hasgreat impacted on the normal operation of the mechanical equipment. Affected by thetransmission characteristics of gearbox, the vibration exists during the process ofnormal operation. The components of vibration signal will be enhanced when gearboxfaults occur, and the signal feature will be different from that in conventional vibration,so it is practicable to confirm real-time performance of the gearbox effectively byanalyzing the vibration signal of gearbox.Planned record and maintenance is helpful not only to prolong the service life ofthe gearbox, but also to reduce the loss caused by machine halt and accident.Gear and bearing were focused in this paper, and the techniques of fault featureextraction and analysis were investigated. The efforts are as follows:(1) Vibration mechanism of the failure gearbox is analyzed. Firstly, severalcommon typical faults are listed, and cause of fault is analyzed in this paper. Byestablishing and solving the coupled meshing vibration model of the gear transmission,two of the most important factors resulting in the vibration are meshing rigidity andtransmission error. And then vibration mechanism of the bearing is analyzed.(2) Time domain analysis and frequency domain analysis techniques for vibrationsignal are presented. The main techniques include the time-domain statistical indices,power spectrum analysis, cepstrum analysis, refining spectrum analysis and theenvelope analysis.(3) Analysis of vibration signal on time-frequency feature is suggested. Typicalfaults including exception on tooth surface, broken teeth, abrasion on bear outer ringand rolling abrasion had been located previously and the signal was analyzed andprocessed in time-frequency analysis technique. On one hand, in time domain thetime-domain feature of signal was found in comparison of statistical parameterschanges on oscillogram between fault signal and normal signal. On the other hand, infrequency domain the frequency-domain feature of signal was found in comparison offrequency component between fault signal and normal signal. Finally compared with the results in various time-frequency methods to the fault signals, the conclusion isproposed as followed: whether the fault exists or not it can be detected throughstatistical parameters analysis but it is impossible to be located. It’s undesirable toconfirm fault signal in power spectrum, zoom spectrum and cepstrum. Envelopespectrum analysis can effectively display the fault signal character of the gears andbearings, and can show the failure parts.(4) Experiment example analysis is shown. Firstly the structure of gear box isintroduced and the fault frequency parameters were listed. Secondly, the signal of timeand frequency domain is analyzed for normal gearbox. At last, an unknown gearboxfaults is diagnosed with time-domain statistical parameters analysis method. Thenenvelope spectrum analysis for features extraction is applied to locate the failure parts.The diagnosis result is consistent with the empirical analysis results.