A Digital Twin‐Enabled Hybrid Deep Learning Approach for Tool Wear Monitoring in CNC Milling Based on Multi‐Sensor Fusion

ABSTRACT Aiming at the problems of low monitoring accuracy of tool condition and insufficient intelligence of health management in high‐end CNC machine tools, this paper takes the CNC machining centre (YSV‐855) milling titanium alloy (Ti‐6Al‐4 V) as the research object and proposes a tool life cycle health management method based on digital twin. By constructing a four‐layer digital twin architecture including the physical perception layer, virtual modelling layer, data interconnection layer and intelligent service layer, the bidirectional real‐time interaction between the physical machine tool and the virtual model is realized. Through the fusion of multi‐source sensor (vibration, acoustic emission and cutting force) data, a hybrid prediction model based on multi‐scale convolutional neural network and bidirectional long short‐term memory network is developed, combined with cross‐scale dual attention mechanism, which effectively improves the accuracy of tool wear state recognition. The experimental results show that the proposed method can achieve a wear state recognition accuracy of 99.72% in titanium alloy milling, with a root mean square error and mean absolute error of the remaining life of 0.9537 and 0.7860, respectively. Compared with the traditional method, the prediction accuracy is improved by 2.34% to 7.18%, the standard deviation of the prediction error is reduced by 40.66% to 61.17%, and the average absolute deviation from the true value is reduced by 40.56% to 62.74%. At the same time, the digital twin system developed based on Unity3D realizes the visualization monitoring and intelligent early warning function of tool condition, providing an effective solution for tool health management in the intelligent manufacturing environment.