Trivariate Empirical Mode Decomposition Via Convex Optimization And Its Application Research In Bearing Fault Diagnosis

For the rotating machinery,the normal operation of rolling bearings determines the operation status of the whole equipment.Therefore,it is of great significance to research the fault diagnosis technology of rolling bearings.Practically,the characteristics of strong noise interference and non-stationary of vibration signals have become the difficult problems in fault diagnosis of rolling bearings.Based on it,the rolling bearing with frequent faults in mechanical equipment is taken as the research object,and proposed a sparse representation denoising method based on convex optimization.On this account,a Improved Trivariate Empirical Mode Decomposition(ITEMD)method is proposed.Finally,a new bearing fault classification method is proposed by combining ITEMD with Multi-Scale Permutation Entropy(MPE).This paper mainly completes aspects of rolling bearing noise reduction,feature extraction and fault classification.The main work of this paper includes:(1)Aiming at the problem of noise interference in the process of fault feature extraction,the noise suppression of rolling bearing vibration signal is realized by sparse representation.On the basis of L1 norm,a parameterized non-convex penalty function is introduced,which keeps the sparse regularization of the objective function and avoids underestimating the sparse approximate solution.It further suppresses the noise and improves the effect of denoising by sparse representation.(2)Considering the relationship between the directions of trivariate vibration signals of rolling bearings,a Trivariate Empirical Mode Decomposition method based on convex optimization is proposed.Firstly,the low-rank matrix approximation via convex optimization is firstly conducted to achieve the denoising.Then,sample direction represented by principal components obtained by SVD algorithms,and the non-uniform sample scheme is proposed to satisfy the requirement of non-uniform distribution of high-dimensional spatial data.Finally,the local mean is estimated by Multivariate Empirical Mode Decomposition(MEMD),and decomposed the IMF components in all directions.(3)Combining ITEMD with Multi-Scale Permutation Entropy and K-means clustering,a new bearing fault classification method based on ITEMD and MPE is proposed.Firstly,a Trivariate Empirical Mode Decomposition method based on convex optimization is used to decompose and reconstruct the measured trivariate signals,so as to suppress noise and extract main features.Then,the complexity information of the three-direction signal is quantified by MPE,and the fault classification of rolling bearings is realized by inputting K-means clustering into the feature sample set.