Research on thermal error compensation modeling for the machine tool integrated drive system based on energy consumption big data and an optimized bidirectional network

• Published in: Precision engineering Vol. 94; pp. 91 - 112
• Main Authors: Chen, Guo-hua, Zhou, Bo, Li, Tao, Mao, Jie, Li, Bo, Fu, Zhen-xin
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.06.2025
• Subjects: Energy consumption data; Improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN); Integrated drive system; Kernel principal component analysis (KPCA); Scalable crested porcupine optimization (SSCPO); Thermal error compensation modeling; Integrated drive system; Scalable crested porcupine optimization (SSCPO); Kernel principal component analysis (KPCA); Improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN); Energy consumption data; Thermal error compensation modeling
• ISSN: 0141-6359
• DOI: 10.1016/j.precisioneng.2025.02.024

In the era of burgeoning intelligent manufacturing, the thermal errors of the integrated drive system in CNC machine tools manifest intricate dynamic traits, including non-linearity, time-variance, and strong coupling. These thermal errors are intricately associated with multiple factors, such as heat source distribution and energy consumption. Traditional thermal error compensation modeling techniques often fail to account for the influence of multiple thermal factors, primarily relying on temperature data obtained from a limited number of thermally sensitive points. To address this gap, the present research introduces a novel bidirectional spatiotemporal network model (IKSM). This model integrates the Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN), Kernel Principal Component Analysis (KPCA), and Strengthened Scalable Crested Porcupine Optimization (SSCPO). At the onset of the research, experiments were carried out using the S5H Intelligent Precision Machining Center provided by Jiangxi Jiashite Company. Temperature, current, power, and thermal error data of the motorized spindle and the linear motor for driving feed under various working conditions were collected. The ICEEMDAN-KPCA approach was subsequently utilized to reduce data dimensionality, thereby facilitating the efficient extraction of essential features. Similarly, the SSCPO algorithm was applied to optimize the parameters of the network model. Through a series of ablation experiments and comparative analyses, the IKSM demonstrated exceptional performance across varying rotational speeds and feed rates. For instance, at a motorized spindle speed of 10,000 rpm, the Root Mean Square Error (RMSE) decreased by 62.05 % relative to the basic BIGRU model, while the coefficient of determination (R2) increased by 40.23 %. Furthermore, the SHAP method was employed to conduct a comprehensive analysis of the key influencing factors, yielding effective strategies and innovative approaches for enhancing the accuracy of CNC machine tools. •Advanced Algorithm Optimization:The SSCPO algorithm was utilized to optimize the parameters of the MABIGTCN network. It overcomes the limitations of traditional optimization algorithms in thermal - error model training.•Deep Model Fusion:The MABIGTCN network integrates BITCN, BIGRU, and multi-head attention for enhanced feature extraction and interaction.•Enhanced Model Interpretability:The SHAP method was used to conduct an in - depth analysis of the key factors influencing thermal errors, such as P3, P7, and power at 10,000 rpm.•Software ModuleDevelopment and Application:Developed thermal error modeling software and a compensation module for the Huazhong CNC system.