Research On The Monitoring And Prediction Method Of Cutting Tool Wear Based On Digital Twin

As the "tooth" of the machine tool,the cutting state of the tool has an important impact on the stability of the whole machining process,the improvement of machining efficiency and quality.Therefore,the timely,accurate and reliable evaluation of tool state has become an important research topic.At present,the research methods for tool wear condition monitoring are mainly divided into finite element model based,sensing data based and model and data fusion based tool wear condition monitoring techniques.However,the wear mechanism model used for tool wear monitoring based on the finite element model has poor adaptability and is difficult to truly and accurately reflect all the wear mechanisms of the tool.The tool wear monitoring method based on sensor data does not consider the physical characteristics,and the result is excessively dependent on the data feature extraction and quality the number of samples,which affects the accuracy of tool wear monitoring results.In the method based on model and data fusion,the digital twin technology has the characteristics of high integration of information and physics and synchronization of virtual and real,which provides an effective means for monitoring and evaluation of tool wear status based on model and data fusion in a complex and changeable processing environment.(1)Based on the concept of digital twin,a digital twin model of the tool is constructed.The constructed tool digital twin model is composed of three parts: twin mechanism model,twin data model and machinability fusion evaluation model.Based on the object-oriented principle,starting from the four dimensions of geometry,physics,rules and behavior,a twin mechanism model is established.The twin data model is established by using convolutional neural network.By determining the number of convolutional layers in the convolutional neural network,the number of pooling layers,and selecting the appropriate convolution kernel size,activation function,dropout and other model parameters,the accuracy of the twin data model is improved.? A machinability fusion evaluation model is constructed using the particle filter algorithm,and the twin mechanism model is updated and revised according to the tool wear results monitored by the twin data model to realize the twin data model and the twin mechanism model fusion.Using the machinability fusion evaluation model,the tool wear monitoring and prediction program is studied.By comparing the fusion results of wear monitoring and prediction with the test results,the validity and accuracy of the machinability fusion evaluation model are demonstrated.(3)A tool wear test platform during turning was built,and two different workpiece materials,AISI-1045 and AISI-4340,were selected for cutting tests.Use the twin data model to mine and analyze the collected sensor signals,and based on the machinability fusion evaluation algorithm,use the results of data analysis to drive the update and correction of the twin mechanism model,and obtain the tool wear status monitoring and prediction based on the digital twin result.The tool wear monitoring and prediction results based on digital twins were compared and analyzed with the tool wear test values respectively,and the accuracy of the tool wear monitoring and prediction program based on digital twins proposed in this paper was verified.