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

The working state of rolling bearing is very important for the reliable operation of large equipment or mechanical system in the production process.The research on fault diagnosis technology of rolling bearing has always been the focus of the current manufacturing field.Taking rolling bearing as the research object,this paper studies rolling bearing fault diagnosis based on vibration signal analysis from three aspects of signal preprocessing,fault feature extraction and fault type recognition,and proposes a fault diagnosis method based on wavelet threshold denoise,variational mode decomposition and machine learning algorithms.The main contents of this paper are as follows:(1)A denoising method of vibration signal based on wavelet threshold is proposed.The working environment of rolling bearing is often accompanied by strong noise interference,which leads to the vibration signal collected showing strong nonlinear and non-stationary characteristics.To solve this problem,based on the in-depth study of the influence factors of wavelet threshold denoising effect,the wavelet threshold denoising method is used to preprocess the original vibration signal.The simulation results show that the method is effective for bearing signal denoising.(2)This paper proposes an adaptive fault feature extraction method for rolling bearing based on variational mode decomposition(VMD).The algorithm principle of variational mode decomposition is deeply studied,and two problems existing in VMD method are improved: 1)Aiming at the end effect problem of VMD method,the image extension method is used to improve the VMD method,so that the end effect can be effectively suppressed;2)Aiming at the uncertainty caused by the artificial selection of parameter K in VMD method,particle swarm optimization is proposed to make the K value be determined adaptively according to the characteristics of the signal.Simulation results show the effectiveness of the algorithm.(3)Three different machine learning algorithms are used to diagnose the bearing faults.Support vector machine,random forest and extreme learning machine are used for fault classification and diagnosis.Especially for the problem that the selection of penalty parameters and kernel parameters in SVM will affect the classification accuracy of the diagnosis model,particle swarm optimization algorithm,genetic algorithm and grid search algorithm are used to optimize the parameters respectively.Based on the rolling bearing fault data published by Case Western Reserve University,this paper compares and analyzes the fault diagnosis effect of three different machine learning algorithms from two aspects of fault diagnosis accuracy and diagnosis time.The experimental results show that,in the rolling bearing fault diagnosis,the random forest has higher accuracy,faster operation speed and better applicability among the three machine learning algorithms.