Reasearch On Rolling Bearing Fault Feature Extraction And Optimization Based On Variational Mode Decomposition

Rolling bearings are the core components of the rotating machinery system,which directly determines whether the entire rotating machinery system can operate normally.It is widely used in petrochemical,electric power,machinery,aerospace and other fields.It conducts fault diagnosis research on rolling bearings and promptly discovers rolling bearing failure can effectively avoid huge accidents in production,and has important academic significance and engineering application value.This paper studies the feature extraction and optimization of rolling bearing faults under noise background.1.Aiming at the problem that the Variational Mode Decomposition can’t effectively extract the fault feature by using a single penalty factor to decompose the signals of multiple vibration sources,a method for extracting the fault feature of the rolling bearing base on the Parameter Optimized Variational Mode Decomposition is proposed.The penalty factor corresponding to each mode in the VMD decomposition is optimized by the whale algorithm,and the correlation coefficient and kurtosis are integrated as the optimization target function of whale algorithm to realize the adaptive selection of the best penalty factor corresponding to each mode,then the mode is selected for Hilbert envelope to extract the fault feature according to the kurtosis maximum criterion.The effectiveness of the proposed method is applied to the experimental data to successfully extract the fault characteristic frequency,thereby identifying the type of fault.2.Aiming at the problem that the weak features of rolling bearing faults are difficult to extract under strong background noise,a method for extracting weak fault features of rolling bearings based on AVMD and MOMEDAis proposed.The AVMD method is used to optimize parameter selection,define the spectrum overlap degree,and construct an adaptive penalty factor based on the spectrum bandwidth and center frequency.Eliminate interference frequencies with MOMEDA.Simulation and experimental analysis show that the proposed method can obviously highlight the weak fault characteristics.3.Aiming at the problem that the linear weighting in infinite feature selection cannot reflect the information between multiple classifications,a method of rolling bearing feature optimization based on EInf-FS is proposed.Based on the signal AVMD decomposition,the time domain,frequency domain,entropy and other features of each modal component are extracted,and the multi-dimensional morphological high-dimensional feature set is constructed.The infinite feature selection process is optimized through Fisher discrimination,and the sensitive features are selected through experiments.The data verifies the effectiveness of the proposed method.4.The content of the research was verified by the rolling bearing experiment of the laboratory rolling bearing test bench,and the application effect of the two feature extraction methods in engineering was verified.The feature extraction method based on POVMD can extract bearing fault features,and the feature extraction methods based on AVMD and MOMEDA effectively extract the weak fault features of bearings.The feature optimization method based on EInf-FS effectively optimizes the selection of sensitive features and realizes the identification of bearing failure types.