Life prediction of NC machine tools based on DT technology and LSTM technology

In order to solve the problems of physical model simplification error, insufficient fusion of multi-source monitoring data, and low accuracy of life prediction in traditional CNC machine tool life prediction methods, this study proposes a CNC machine tool remaining useful life (RUL) prediction method that combines digital twin (DT) technology with long short-term memory (LSTM) network. This study constructed a multi physics domain mapping model for CNC machine tools based on DT technology. A multimodal data preprocessing module was introduced into the DT model to extract key degradation features of the machine tool, and an improved LSTM network was developed. By inputting the high-dimensional degraded features output by the DT model into the LSTM network, accurate RUL prediction of CNC machine tools has been achieved. The results show that the proposed model performs well in all core indicators: during the accelerated degradation stage, the prediction accuracy is 96.1%, the average absolute error is only 8.9 h, and the maximum deviation is only 15 h, while maintaining a 100% physical constraint compliance rate and an effective prediction speed of 22 ms. In addition, as the proportion of the system increases, the indicators of the model rapidly improve; When the system proportion reaches 40%, the accuracy exceeds 40%, the recall rate approaches 42%, and the F0.5 score significantly improves. These findings indicate that the proposed method can effectively reduce equipment downtime losses, improve production efficiency, and provide a new technological approach for predictive maintenance of CNC machine tools.