Research On Feature Extraction And Diagnosis Of Fault Rolling Bearing Based On Vibration Signal

Rolling bearings,which serve as core critical components in mechanical equipment,play the role of undertaking and transmitting the loads in the process of equipment operation and are widely used in metallutgical,chemical,aerospace and other important fields.Once the rolling bearings break down with no investigation in time,it will result in production stagnation and casualities.Studying found that the fault bearing vibration signal shows the obvious non-linear,non-stationary characteristics.If it’s possible to monitor the running state of the rolling bearing online and make reasonable judgments on the available information,then it will play great significance to ensure the stability and smmoth operation of the system and prevent the occurrence of safety accidents.The dissertation takes the rolling bearing as the research object,mainly studies the bearing feature extraction and diagnosis method based on bearing vibration signal,the concreate research contents are as follows:① Aiming at the problem that vibration characteristics information is easily submerged in strong vibration noise,an adaptive maximum correlated kurtosis deconvolution method is proposed to realize the filtering of vibration signals of rolling bearings.Taking advantage of the MCKD’s ability of enhancing periodic impulsive signals in noisy signals,the related kurtosis is regarded as the evaluation standard.At the same time,envelope entropy of signal after MCKD served as a fitness function and the enhanced leading particle swarm optimization algorithm is adoption to optimize the filter length and period of MCKD algorithm,and enhance the rolling bearing fault vibration signal by adaptive maximum correlation kurtosis deconvolution;② Aiming at the problem of difficulty of feature extraction of vibration signals of fault rolling bearing,a fault feature extraction method based on improved variational mode decomposition and singular value decomposition is proposed.Taking the Pearson’s correlation coefficient between any IMFs as an index to determine the reasonable number of IMF in the VMD,and achieve the improved variational mode decomposition of the bearing vibration signals.The singular value decomposition is used to calculate the modal matrix of the AVMD,and the eigenvectors of the rolling bearing are characterized to improve the performance of its degree of recognition;③ Aiming at the problem of fault identification of rolling bearing,the genetic algorithm is used to optimize the parameters of support vector machine to realize the fault diagnosis and pattern recognition of rolling bearing.The Gaussian function with strong learning ability and generalization performance is used as the kernel function,and the samples in the low-dimensional space are mapped into the high-dimensional feature space.The genetic algorithm is used to calculate the penalty coefficient of support vector machine and Gaussian kernel function parameters to determine the optimal parameters to achieve fault diagnosis of rolling bearings adaptively with support vector machine.