Research On The Fault Diagnosis Methods For Rolling Bearings Based On The Order Multi-scale Chirplet Path Pursuit

Rolling bearing is a common part of mechanical equipment s. The running stateof rolling bearings affects the whole machine’s capability. Therefore, conditionmonitoring and fault diagnosis for the rolling bearing have significant meaning. Thevibration signal contains the fault information, extracting the fault feature from thevibration signals is the key of fault diagnosis.In general, the fault vibration signal of mechanical equipment is usuallynon-stationary. The fault vibration signal of the equipments under varying rotatespeed usually contains more operation and fault information. Yet, the vibration signalswhich measured at equal-time-interval are the multi-component non-stationary signalsand have the low signal-to-noise ratio, thus it is very difficult to extract the faultfeatures from the vibration signals with the existing signal pro cessing techniques.This thesis, supported by National Natural Science Foundation of China(Project’sSerial Number:50875078), proposes a method based on the order Multi-scale ChirpletPath Pursuit(MCPP) to overcome the issues existing in the present signa l processingmethods, and applies this method to the rolling bearing fault diagnosis under varyingrotate speed. In the proposed method, we estimate fault characteristic frequency curveadaptively and acquire angular domain stationary signal according to the order ratioanalysis. Then, the multi-scale morphology demodulating and cyclostationarydemodulating method are used respectively to diagnose rolling bearing faults undervariable rotating speed.The main researches and the acquired innovative achieveme nts in the thesis areas follows：(1) Aiming at the problem that traditional analysis methods have lowtime-frequency concentration and poor anti-noise, the thesis introduces themulti-scale chirplet path pursuit method into the rolling bearing fault diagnosis.Through the comparative analysis of the simulation signal, we prove that the methodhas good time-frequency concentration and strong anti-noise, so it is especiallysuitable for the analysis of non-stationary signals.(2) Aiming at the problem that the fault characteristic frequency of rollingbearing with rotating speed are difficult to extract,a method for the fault diagnosis ofroller bearings with rotating speed fluctuation based on the order Multi-scalemorphology demodulation is proposed in this thesis. In the proposed method, the chirplet path pursuit algorithm is used to obtain the bearing fault characteristicfrequency. According to the bearing fault characteristic frequency, the time domainvibration signal of the roller bearing is resampled at constant angle increments. Analgorithm based on the information of local peaks of the signal is used to figure outthe structuring elements for multiscale morphology analysis. The structuring elementsare used to carry out morphology operations on the resampled signal. Spectrumanalysis on the average value of the operation results is carried out and then the ordermulti-scale morphology demodulation is accomplished. The chirplet path pursuitalgorithm can extract the bearing fault characteristic frequ ency effectively even if thesignal-to-noise ratio of the vibration signal is very low. By averaging the results ofeach scale morphology analysis, the multi-scale morphology demodulation canrestrain the noise effectively. Therefore the present approach has good anti-noiseability, and is suitable for analyzing the actual vibration signal of a bearing withrotating speed fluctuation. Simulation and practical application examples confirmedthe validity and the superiority of the proposed method.(3) A method for the fault diagnosis of roller bearings with rotating speedfluctuation based on the order cyclostationary demodulation is proposed in this thesis.In this proposed method, the chirplet path pursuit algorithm is used to obtain thebearing fault characteristic frequency. According to the bearing fault characteristicfrequency, the envelope of the original signal is resampled at constant angleincrements to obtain the angular domain cyclostationary signal. The autocorrelationfunction of angular domain signal is calculated and the slice of autocorrelationfunction on the characteristic cycle order is computed. The slice demodulationspectrum is then acquired and is used to diagnose the fault of roller bearings. Thechirplet path pursuit algorithm can extract the bearing fault characteristic frequencyprecisely and effectively even if the signal-to-noise ratio of the vibration signal isvery low, moreover, the cyclostationary demodulating approach can ext ract the cyclefault feature from noise signal.Therefore, the present approach has good anti-noiseability, and is suitable for analyzing the actual vibration signal of a roller bearing withrotating speed fluctuation. Simulation and practical application examples show thevalidity and the superiority of the proposed method.This thesis introduces the order multi-scale chirplet path pursuit algorithm intorolling bearing fault diagnosis under variable rotating speed. By combining the ordermulti-scale chirplet path pursuit algorithm with multi-scale morphology demodulatingand cyclostationary demodulating method, two methods for the fault diagnosis of rolling bearings under the variable rotational speed are proposed. Simulation andpractical application examples show that the proposed methods can be used todiagnose the faults of rolling bearings under the variable rotational speed effectively.