Research On Bearing Fault Diagnosis Based On Improved Empirical Mode Decomposition And Depth Feature Extraction

As one of the important components of rotating machinery,the running state of rolling bearing directly affects the operation of the whole machinery and even personal safety.In this paper,a rolling bearing fault diagnosis method based on ICEEMDAN Hilbert marginal spectrum DSELMCAE and IAO-SVM model is proposed from the perspectives of bearing fault feature extraction and fault diagnosis model.Firstly,to solve the problem that it is difficult to extract the fault features due to the nonlinear and non-stationary characteristics of bearing vibration signals,ICEEMDAN algorithm is used to decompose the original bearing fault data and extract the internal feature information of bearing vibration signals.Through the simulation experiment of the empirical mode decomposition algorithm,the overall empirical mode decomposition algorithm and ICEEMDAN algorithm for processing simple multi harmonic signals,it is explained in detail that ICEEMDAN has more advantages in solving mode aliasing and over decomposition,and it is confirmed that ICEEMDAN algorithm is more suitable for decomposition of bearing vibration dynamic signals.ICEEMDAN is used to decompose the bearing signal,and the method of Hilbert marginal spectrum is used to fuse IMF components to form a marginal spectrum signal to achieve the preliminary characteristics of bearing fault features.Then the marginal spectrum signal is input to DSELMCAE for deep level feature extraction,and finally the depth feature extraction method of ICEEMDAN Hilbert marginal spectrum DSELMCAE is formed.Secondly,the influence of different kernel function selection on support vector machine is compared and analyzed,and the SVM model of Gaussian radial basis function kernel function is determined.To solve the problem that it is difficult to select the penalty factor C and the kernel function parameter g in the SVM model of Gaussian radial basis function,a Skyhawk optimization algorithm based on Circle chaotic mapping,reverse learning of dynamic probability switching perturbation and Cauchy Gaussian mutation strategy is introduced.The convergence speed and optimization accuracy of the improved Skyhawk optimization algorithm are proved through the simulation experiment of 12 benchmark test functions and Wilcoxon rank sum test.IAO is used to automatically optimize the two key parameters C and g,and IAO-SVM multi classification model is established.Finally,this paper combines the depth feature extraction method of ICEEMDAN Hilbert marginal spectrum DSELMCAE with IAO-SVM multi classification model.Through the diagnosis results of BP neural network,KELM,SVM,GWO-SVM,AOSVM,IAO-SVM models under preliminary feature extraction of ICEMDAN Hilbert marginal spectrum and GWO-SVM,AO-SVM,IAO-SVM models under depth feature extraction of ICEMDAN Hilbert marginal spectrum DSELMCAE,the scientificity,rationality and high accuracy of ICEMDAN Hilbert marginal spectrum DSELMCAE and IAO-SVM models are verified.The model comprehensively considers the advantages of the algorithm itself,which not only solves the problem of difficult extraction of bearing fault features,but also achieves 99.67% accuracy of fault diagnosis.