Research On Fault Feature Extraction Method Of Transmission Rolling Bearings Based On Adaptive MCKD And Blind Source Separation

Aiming at the shortcomings of existing diagnostic methods for early fault diagnosis of rolling bearings in strong noise environment,a new method based on the combination of spectral clustering algorithm,EWT,L-kurtosis,locust optimization algorithm(GOA)and MCKD denoising is proposed in this paper.The related arrangement of content of this paper are as follows(1)In view of the shortcoming that traditional MCKD algorithm relies on prior knowledge and reasonable selection of relevant parameters,the four important parameters that have the largest influence on the denoising performance of MCKD algorithm are studied,and the influences of each parameter on the calculation time and the denoising performance are analyzed respectively.The parameters of the appropriate number of cycles,the optimal value of the shift number and the optimization range of the length of the filter are determined.An grasshopper optimization algorithm is proposed to optimize the deconvolution cycle which lays a foundation for the proposal of adaptive MCKD method based on grasshopper optimization algorithm.(2)MCKD algorithm is combined with the Grasshopper optimization algorithm(GOA)of swarm intelligence optimization algorithm in this paper,and the Grasshopper optimization algorithm(GOA)is used to realize the adaptive selection of important parameters T and L of MCKD.A new index,the average characteristic amplitude ratio(AFAR),was proposed as the fitness function of GOA,and a method to determine the search range of the deconvolution cycle was proposed,and a new adaptive MCKD bearing fault diagnosis method was summarized.Finally,it is proved that the method can effectively extract the fault characteristic periodic pulse signal components from the mixed signals through the diagnosis of the simulated bearing fault signals.(3)The empirical wavelet transform method is improved by combining the spectral clustering algorithm with the empirical wavelet transform(EWT).The spectral clustering algorithm is used to segment the Fourier spectrum of the signal,which improves the disadvantage that the selection of maximum value often falls into the local optimum in the spectrum segmentation process of the traditional empirical wavelet transform,which leads to the aliasing of the signal components.The SEWT method with better spectral segmentation effect is proposed,which can de-noise the collected mixed signals well and improve the feature extraction effect of the adaptive MCKD method.(4)By combining the SEWT noise elimination method,the adaptive MCKD(GOA-MCKD)method,wtih the new statistic of L kurtosis,a new bearing fault diagnosis method suitable for strong noise background is proposed.The validity of the method is proved by the diagnosis of the simulated bearing fault signal.Compared with MCKD-IEWT and enhanced clustering segmentation,the new method proposed in this paper is proved to have better performance in noise reduction and fault feature extraction.Finally,the practicability and accuracy of this method are proved by the fault diagnosis of the fault bearing of the bearing fault simulation test bed and the fault bearing of the AMT test bed.