Thermal Error Prediction of Multiple Machine Tools Using Temperature Sensor Network and Spatio-Temporal Graph Transfer Learning

Thermal error is a predominant factor influencing the machining accuracy of CNC machine tools. Recently, deep learning-based data-driven models have demonstrated remarkable success in predicting thermal errors. However, significant variations in data distribution under different operating conditions and across various machine tools impede the direct application of a model trained on one machine tool to another. To address this challenge, we propose a novel spatio-temporal graph transfer learning method named GCN-LSTM-MDD for predicting thermal errors across different machine tools. Specifically, we employ a graph convolutional network (GCN) to extract spatial features from multiple temperature sensors; these extracted features are subsequently fed into a long short-term memory (LSTM) network to capture temporal dependencies. To further enhance the model’s cross-domain adaptability, we integrate a margin disparity discrepancy (MDD) transfer learning approach aimed at mitigating discrepancies in thermal error distributions among different machine tools. Finally, datasets collected from three distinct operational conditions of two identical machine tools were utilized to construct 18 transfer tasks for validation purposes. The experimental results indicate that the proposed method achieves superior predictive performance across all 18 transfer tasks, surpassing other comparative methods.