Digital twin with dynamic mechanistic simulation core for milling tool wear prediction

Abstract

To enhance capability of tool wear prediction, constructing a digital twin (DT) integrated simulation and sensor data is a novel approach to balancing prediction accuracy and real-time performance. However, current DTs mainly remain at the data level and few realize dynamic simulation, which affects prediction efficiency. To address this issue, a DT with dynamic mechanistic simulation core (DMSC-DT) is proposed in this paper for tool wear prediction. In this method, firstly, a core of mechanism models is built based on cutting and wear force. Then, an adaptive extended Kalman filtering is designed, which can automatically calibrate the model coefficients with minimal cutting force signals. Next, the impact of real-time vibration signals on spindle disturbances and chip thickness is analyzed, allowing the data to dynamically adjust the simulation. Finally, the proposed DMSC-DT is deployed based on a series of milling experiments. The results show that the DMSC-DT can achieve deep integration between sensor data and simulation process. In terms of tool wear prediction, the average simulation time per cycle is 3.69 s and the accuracy improved by an average of 11.7 %. The method achieves accurate predictions under varying cutting parameters, demonstrating strong robustness and practical potential.