Fault Diagnosis And Life Prediction Of LS-2500 Stamping Machine Motor Bearings

In large stamping equipment,the motor bearing is one of the core components,and its running state is closely related to the working performance of the equipment.It is very important to master the healthy state of the motor bearing and prevent the heavy losses caused by the damage of the motor bearing.Through fault diagnosis and life prediction and other health maintenance methods to understand the operation status of motor bearing,provide the basis for the replacement of motor bearing of stamping equipment and the formulation of maintenance plan.In order to remove the interference component of the motor bearing under actual working conditions,effectively extract the fault characteristics.In this paper,singular value decomposition(SVD)is used to denoise the collected vibration signal,and the dimension of the initial matrix is determined by calculating the contribution rate of singular value.Then the butterfly optimization algorithm is used to obtain the best parameter combination in the variational mode decomposition(VMD),and the denoised signal is decomposed by VMD.Finally,according to the kurtosis value selection principle,the effective modal components in the K modal components(IMF)obtained by VMD decomposition are selected for envelope spectrum analysis,and then the fault type is determined.Through simulation analysis and experimental verification,this method can effectively remove the interference of background noise and irrelevant information in the signal,and can accurately identify the fault features in the signal.In order to solve the problems of difficult feature extraction and low prediction accuracy of bearing vibration signal of stamping motor,a bearing residual life prediction model based on CNN-GRU-Attention network is proposed in this paper.Firstly,the vibration signal is selected as the input,and the SVD signal is denoised to make the fault features obvious.Then,the convolution and pooling characteristics of convolution neural network(CNN)are used to extract the deep features from the denoised signal.Finally,combining the advantages of attention mechanism which can screen and highlight key features with the advantages of gated recurrent neural network(GRU)in time series prediction,a quantitative health index is constructed according to the life percentage,and the failure threshold is set.Polynomial curve fitting is used to fit the degradation trend curve to the failure time to complete the residual life prediction.Through the experimental analysis and model comparison,it is verified that the CNN-GRU-Attention prediction model can effectively extract the deep features in the vibration signal,and has the advantages of good fitting effect and high prediction accuracy.The prediction model proposed in this paper is applied to the actual vibration signal of the bearing,which can accurately predict the remaining service life of the bearing.