Research On On-line Monitoring Method Of Milling Cutter Wear State Based On Digital Twin

Made in China 2025 promotes the continuous transformation of manufacturing industry from automation to intelligence.As the core equipment of manufacturing industry,CNC machine tools are the core components of machine tools,and their wear status has a decisive impact on machining efficiency and quality.Therefore,how to realize the intelligent monitoring of tool wear status has become an important research topic of intelligent manufacturing.As a key technology to drive the transformation of manufacturing towards intelligence,the digital twin is an effective method to achieve the interactive integration of physical space and information space.It is of great research significance to deeply integrate tool wear status monitoring with digital twin,sensor technology and information processing technology.This paper takes CNC milling as the research object,consider the real-time wear state of the tool,and carry out the research of real-time monitoring of the wear state of milling tools based on digital twin.The main research contents are as follows:A CNC milling digital twin model is constructed.Based on the five-dimensional model architecture of digital twin,the digital twin architecture of CNC milling is proposed;the highfidelity virtual model of physical entities is constructed by using 3ds Max,Unity 3D and other software;the classification and collection methods of data are studied to construct the twin data model;the process of tool wear status monitoring is studied based on the twin data.The tool wear state monitoring model under a single working condition is designed.The tool wear signal is pre-processed to improve the quality of the input data.Based on onedimensional convolutional neural network and bidirectional long short-term memory network,multi-scale convolution and channel attention mechanism are fused for tool wear status monitoring.The multiscale convolution module enhances the ability of the one-dimensional convolutional neural network to extract features at different scales,the channel attention mechanism makes the model focus more on effective features,and the bidirectional long shortterm memory network extracts temporal information in the signal,thus,the integrated model can effectively fuse multi-dimensional feature data and improve the accuracy of tool wear status monitoring.A tool wear condition monitoring model under variable working conditions is designed.To address the problem that the deep learning model constructed under a single working condition has low monitoring accuracy under variable working conditions,a cross-domain monitoring model of tool wear state incorporating maximum mean difference is designed.After completing the extraction of tool wear degradation features under different working conditions,a multi-layer multi-core maximum mean difference alignment feature distribution is used to reduce the distribution differences between features and achieve end-to-end knowledge migration.Various migration tasks are designed to verify the effectiveness of the model.A CNC milling digital twin system is developed.A CNC milling digital twin system is designed in general,and the system design goals,development and operation environment are determined.The system database design is designed to realize the storage of twin data.From the perspective of system application functions,relevant functional modules are designed and the system integration development is completed.