Research On Fault Feature Extraction Method Of Rolling Bearing Under Different Noise Interferences

As a part of transmission system,rotating machinery undertakes important engineering tasks in industrial production,and rolling bearings are widely used in rotating machinery.In a variety of complex conditions,rolling bearing is prone to failure,which makes the whole transmission system unable to run smoothly and safely.Therefore,it is very necessary to diagnose the fault of rolling bearing,find the fault and deal with it in time.Because the fault of rolling bearing is accompanied by vibration and noise,so extracting the fault feature of rolling bearing is the most common method of fault diagnosis.In signal processing,it is often necessary to eliminate the interference of various noises in order to complete the feature extraction effectively.This paper focuses on the method of rolling bearing fault feature extraction under different noise interference.And it introduces the limitations of classical methods to remove background noise.On this basis,This paper makes an in-depth study on the relevant theories of feature extraction,such as index,optimization algorithm,sparse model and iterative shrinkage algorithm,and completes the feature extraction under the interference of impulse noise,modulation and harmonic and white noise.The specific research contents and achievements are as follows:First of all,aiming at the interference of impulse noise,a rolling bearing feature extraction method based on adaptive harmonic kurtosis and improved bat algorithm is proposed.According to the threshold setting,the adaptive harmonic kurtosis index is constructed,which can calculate the frequency-domain narrow-band kurtosis according to the potential fault type of the signal;then,the bat optimization algorithm is applied and improved to search the optimal resonance frequency band randomly,speedily and finely;through the above two improvements,this method can complete the feature extraction of single fault or composite fault under the interference of impulse noise.Then,the rolling bearing fault feature extraction method under the influence of modulation,harmonic and white noise is studied.Firstly,Fourier dictionary and OMP algorithm are used to remove the modulation and harmonic;secondly,for the residual white noise,an approximateL₁-L₀ sparse model is constructed,and an iterative dictionary learning algorithm based on nonconvex sparse regularization and improved parallel coordinate descent is proposed,which improves the speed of dictionary learning and effectively enhances the impact characteristics of the signal.Finally,by applying the two methods to the fault feature extraction of rolling bearing in 5T-85 gearbox,the fault features are successfully extracted,which further proves the effectiveness of the proposed method and its application potential in engineering.To sum up,considering the influence of different noises,this paper proposes two effective feature extraction methods,and shows the advantages of the two methods through comparative experiments and engineering verification.