Research On Intelligent Monitoring Of Tool Wear State And Remaining Life Prediction In Discrete Workshop

With the development of modern industry technology,the production model of the manufacturing industry has gradually changed from large quantities and few varieties to small quantities and many types.At present in the manufacturing process still based on reducing material manufacturing,cutting tool as a key part in the production process,directly affects the quality of parts,production process and production cost.To improve production quality and protect equipment and personal safety,it is of great value to monitor the tool state.In this article,taking milling cutters as the research object,a deep learning based method is proposed to realize the wear state monitoring and remaining life prediction of milling cutters,and the effectiveness of the above method is verified by experiments.The main research contents are as follows:(1)The methods,research state and related technologies of tool wear state monitoring and remaining life prediction were introduced.After analyzing the demand of tool wear state monitoring and remaining life prediction,the development trend of tool wear state monitoring and remaining life prediction and the overall framework of the article were summarized.The tool wear mechanism was analyzed and the torque and vibration signals were selected as monitoring signals.The test platform of milling cutter machining state monitoring was built to record the relevant data in the process from new cutter to blunt grinding,providing a data basis for the subsequent research.(2)Wavelet threshold denoising was used to denoise the acquired monitoring signals containing invalid or interference signals.To maintain the synchronicity of monitoring signals,the torque signal was sampled and the number of samples was expanded by slicing.In addition,to solve the problem of low classification accuracy caused by unbalanced samples,SMOTE oversampling method was used to balance the number of samples in different wear stages,and then extract relevant time-domain,frequency domain and time-frequency domain features.Finally,principal component analysis was used to reduce the dimension of the original features of tool wear state monitoring and remaining life prediction,and the features of low correlation with tool wear were removed.(3)The intelligent monitoring technology of milling cutter wear state was studied,and the P-1DCNN-ELM model was established.In view of the disadvantages of traditional machine learning based on manual feature extraction,a method of one-dimensional convolutional neural network for automatic feature extraction was proposed.Due to the large number of training parameters caused by the traditional method of setting the number of neurons in the neural network model,an inverted pyramid design method was proposed,which reduced the training parameters by 59.03%.Aiming at the problem of insufficient feature classification and slow recognition efficiency of the Softmax classifier,an extreme learning machine was used to replace the Softmax classifier to realize intelligent monitoring of the wear state of milling cutters.(4)The remaining life prediction technology of milling cutter was studied,and the regression model of 1DCNN-Bi GRUs-CP was established.To improve the prediction accuracy under finite samples,a prediction method based on one-dimensional convolutional neural network and bidirectional gated recurrent unit was proposed.In order to improve the convergence speed of the model,the full connection layer behind the bidirectional gated recurrent unit was improved.Combined with the advantage that the scoring function can punish the error to different degrees,a harmonic mean loss function HM-MSE-Score based on MSEScore was constructed by introducing the scoring function on the basis of the mean square error function to realize the prediction of the remaining life of milling cutter.