| Bearing is one of the important parts of construction machinery.It works for a long time under harsh conditions,and its health status has a major impact on the reliability and stability of the mechanical system.Therefore,it is very important to establish a bearing condition monitoring system for the safe operation of the equipment.In this paper,the bearing fault diagnosis is taken as the research content,and the main work is as follows:(1)In order to eliminate the noise component in the original signal,a signal denoising method based on parameter optimization of variational mode decomposition(VMD)combined with mathematical morphological filtering(MMF)and wavelet threshold denoising(WTD)is proposed.The intrinsic mode component(IMF)is divided into signal modes and mixed modes according to the correlation waveform index(Cwi).The WTD method is used to remove the high frequency noise in the signal mode,and the MMF is used to extract the low frequency component in the mixed mode.Finally,the denoised signal is reconstructed.The denoising effect of the denoising method was tested and verified by using the bearing inner ring analog signal and the bearing inner ring fault signal of Case Western Reserve University(CWRU)respectively.(2)In order to extract denoised signal features effectively,an improved refined composite hierarchical fuzzy entropy(IRCHFE)method is proposed for signal feature extraction.Compared with multi-scale fuzzy entropy(MFE)and hierarchical fuzzy entropy(HFE),the hierarchical decomposition method of IRCHFE method can extract the information of high and low frequency components at the same time,and the hierarchical decomposition operator of IRCHFE method can process the data of any length.Secondly,the improved refined composite method can not only improve the stability of fuzzy entropy and reduce the possibility of undefined fuzzy entropy,but also improve the signal feature expression ability by integrating sequence amplitude information into entropy.Pso-optimized least squares support vector Machine(LSSVM)was used as fault classifier for fault classification.Finally,bearing data from CWRU and Paderborn University(UPB)were used to test and verify the fault diagnosis effectiveness of IRCHFE method,and compared with other methods.(3)In order to improve the feature extraction ability of fuzzy entropy,generalized compound multiscale symbolic fuzzy entropy(GCWMSFE)is proposed for feature extraction.Compared with the multi-scale symbolic fuzzy entropy(MSFE)method,the GCWMSFE method has the following innovations: the fusion of cosine distance and Chebyshev distance can more accurately measure the similarity between vectors;the composite analysis of the sequence at the same scale,the average of the subsequence The weighted average similarity composed of similarity and standard deviation is used to calculate the SFE value at this scale;the standard deviation of each scale sequence is calculated as a weight to form a weighted MSFE;the second-order moment replaces the average operation of the original coarse-grained process.The 5 features with lower scores are selected by Laplacian fraction method to form feature vectors,which are divided into training set and test set.The fault diagnosis effect of GCWMSFE method is tested and verified by using CWRU and UPB bearing data,and compared with other methods. |
