Scholarlink

<p>The precision degradation of machine tools is primarily caused by thermal effects generated during operation. Existing thermal error modeling methods rarely apply the physical mechanism of error formation in real-time error compensation. To address these limitations, a digital twin system for thermal error compensation of numerical control machine tools is proposed, built upon a multi-algorithm collaborative (M-AC) modeling framework. The system integrates a physical mechanism model (PMM) based on finite element analysis (FEA) and a data-driven model (DDM) utilizing a long short-term memory-convolutional neural network-attention architecture with a life-long learning approach (LL-LCA). This M-AC framework, paired with a dynamic reduced-order technique and heat-flow correction strategy, enables real-time thermal error prediction and compensation. Cutting experiments validate the system’s effectiveness under varying conditions. For simple stepped workpieces, the average errors in the Y and Z directions were reduced by 85.09% and 76.23% respectively. For complex boss workpieces, the average errors in the Y and Z directions were reduced by 53.62% and 70.91%. These results demonstrate the effectiveness and robustness of the proposed system in improving machining accuracy.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A digital twin system for thermal error compensation of numerical control machine tools based on multi-algorithm collaborative modeling

  • 期刊论文

  • 发表日期：

    2026年1月25日

查看全文
  • Qi Liu,
  • Zhong Chen,
  • Qisen Wu,
  • Hao Wang,
  • Xianmin Zhang

摘要

The precision degradation of machine tools is primarily caused by thermal effects generated during operation. Existing thermal error modeling methods rarely apply the physical mechanism of error formation in real-time error compensation. To address these limitations, a digital twin system for thermal error compensation of numerical control machine tools is proposed, built upon a multi-algorithm collaborative (M-AC) modeling framework. The system integrates a physical mechanism model (PMM) based on finite element analysis (FEA) and a data-driven model (DDM) utilizing a long short-term memory-convolutional neural network-attention architecture with a life-long learning approach (LL-LCA). This M-AC framework, paired with a dynamic reduced-order technique and heat-flow correction strategy, enables real-time thermal error prediction and compensation. Cutting experiments validate the system’s effectiveness under varying conditions. For simple stepped workpieces, the average errors in the Y and Z directions were reduced by 85.09% and 76.23% respectively. For complex boss workpieces, the average errors in the Y and Z directions were reduced by 53.62% and 70.91%. These results demonstrate the effectiveness and robustness of the proposed system in improving machining accuracy.