Research On Milling Cutter Wear Monitoring Method Based On Deep Learning

Timely replacement of cutting tools is of great significance to the improvement of product processing quality and the reduction of production cost.In order to improve tool utilization,reduce the number of tool stops and ensure the quality of workpiece surface machining,it is necessary to accurately and reliably monitor the degree of tool wear in order to provide a reliable decision-making means for machine tool changes.Therefore,based on the deep learning method,this paper conducts in-depth research on the identification of the wear state and the monitoring of the wear quantity of milling cutters,which provides ideas for the intelligent monitoring of milling cutter wear.The main research contents are summarized below:(1)The research status of tool wear monitoring is introduced from different aspects,the shortcomings of artificial feature extraction technology and traditional machine learning models for tool wear monitoring and the advantages of deep learning methods applied to this field are compared and analyzed,and some shortcomings in the current research are summarized.Then,combined with the tool wear law and data set characteristics,the force signal,vibration signal and acoustic emission signal are selected as the monitoring signals of the tool wear monitoring model.Finally,the method of directly using the deep learning model to extract the features of the data and output the final results is determined,which greatly simplifies the monitoring process and is more in line with the development direction of intelligent monitoring.(2)Aiming at the problem that the monitoring signal contains a large amount of noise affecting the model state recognition effect,an MCS module is designed to be introduced into 1D-Res Net for milling cutter wear state recognition,and the designed MCS module improves the feature extraction ability and noise immunity ability of 1DRes Net on different scales of signal data through the combined action of multi-scale convolution algorithm and adaptive soft thresholding algorithm,and proves the effectiveness and superiority of the model structure through the performance of the cross-validation set;Aiming at the problem that the model has a low recall rate in the acute wear stage due to the imbalance of the number of samples in the training dataset,a cost-sensitive loss function is used to further improve the model effect,and the effectiveness and stability of the method are comprehensively verified by a number of evaluation indicators obtained on the cross-validation set and the overall dataset.Finally,the recognition effect of the model under different sensor combination schemes is analyzed experimentally,and the recognition performance and practical significance of the cost-sensitive improved 1D-Res Net model are expounded.(3)Aiming at the problem of monitoring the wear amount of milling cutters,a CNN-Bi LSTM dual-channel shrinkage model is proposed,which can comprehensively and deeply extract local detail features and sequence features of multi-channel signal data through the two feature extraction paths of CNN and Bi LSTM,and filter a large amount of information in the fusion feature map by introducing a shrinkage module to remove redundant information.Compared with the performance of other deep learning models on the cross-validation set,it is verified that the model has better monitoring accuracy and generalization performance.The stability and application significance of the CNN-Bi LSTM dual-channel shrinkage model were comprehensively evaluated by the monitoring effect of the experimental model under the overall dataset and different sensor combination schemes.(4)Taking the milling cutter wear state identification model and wear monitoring model proposed in this paper as the core,the milling cutter wear monitoring system is preliminarily designed by using the Py Qt5 module of Pycharm software,which realizes the complete tool wear monitoring and identification process,and provides a way for the training and monitoring of milling cutter wear monitoring tasks under different working conditions,while meeting the basic practical application needs and having certain practical application value.