A New Method For Fault Diagnosis Of Rolling Bearing Based On Wavelet Noise Reduction And Fractal Technique

Abnormal vibration and failure of rolling bearing will affect the normal operation of equipment and cause great economic loss.In the process of rolling bearing fault diagnosis,it is necessary to reduce the noise interference,retain and strengthen the fault signal characteristics,and effectively extract the fault characteristics,in order to complete the fault diagnosis.Considering that wavelet threshold function has better noise reduction function for bearing vibration signal and multifractal detrend fluctuation analysis has better signal feature extraction function,this thesis combines these two data processing methods to effectively identify various faults of rolling bearing.The structure,damage type and fault type of rolling bearing are studied in this thesis.Aiming at the drive end bearing of a certain type of bearing system and fan end bearings,inner ring and outer ring fault vibration signal is studied,by calculating and comparing the fractal dimension of bearing vibration signal and vibration signal variance,found that when the fractal dimension of between 1.5 and 2,fluctuates variance increases with the amount of data,these phenomena show that in the show at an early stage of fault feature is not obvious,so the conventional spectrum analysis methods are hard to identify the fault.Thesis studies the basic theory and method of wavelet de-noising,aimed at the hard threshold function discrete wavelet coefficients and wavelet coefficient is a constant bias in soft threshold function,this thesis proposes a noise reduction method based on improved wavelet thresholding function,improves the threshold function continuity on the given threshold and asymptotic behavior.The influence of wavelet basis and threshold selection principle on noise reduction effect is analyzed and the optimal selection of wavelet basis and threshold is carried out by means of fault vibration signal simulation.By comparing the noise reduction effects of the three threshold functions,it is verified that the improved threshold function proposed in this thesis has a good noise reduction effect.In this thesis,the change rule of fractal box dimension corresponding to the simulated fault vibration signal before and after noise reduction is studied,and it is pointed out that fractal box dimension has higher recognition ability for bearing inner ring fault and roller fault.The method of fractal correlation dimension for signal feature recognition is studied.The fractal correlation dimension and the error function of the fractal correlation dimension are calculated for the fault signals of the outer ring of the rolling bearing.This thesis studies the application of multifractal detrending wave analysis(MFDFA)method in bearing fault diagnosis,proposes an improved calculation method for four kinds of eigenvector spectral parameters,and uses the stability and sensitivity of eigenvector spectral parameters to select the optimal eigenvectors ao and amin,and determines that a0 and amio can identify bearing faults.Combining the improved threshold function noise reduction method,fractal correlation dimension and multifractal detrend fluctuation analysis method,the analysis flow and parameter setting method of the research method in this thesis are given.Under the condition of the same frequency,the same motor load and the same motor speed,the normal bearing signal and the bearing signal of two different fault types are diagnosed.The method proposed in this thesis has a high identification rate of bearing inner ring and outer ring faults,and can also determine the severity of bearing inner ring and outer ring faults.The rolling bearing fault diagnosis method studied in this thesis is helpful to extract fault features from bearing fault vibration signals,improve fault diagnosis accuracy,and provide a new idea for rolling bearing fault diagnosis.