Thermal error prediction for heavy-duty CNC machine tool based on dynamic graph convolutional recurrent network

Thermal error is one of the main factors affecting the machining accuracy of machine tools, and heavy-duty CNC machine tools are particularly affected by it. The thermal error prediction based on temperature field monitoring is one of the most common approaches for predicting thermal errors, and only the temporal dependencies of temperature field being considered. However, temperature field has dynamic spatial-temporal dependencies, and it is not sufficient to consider the temporal features only. To address these challenges, a model based on dynamic graph convolution recurrent network with attention mechanism (DGCRN-A) is proposed in this paper. The model was validated using a heavy-duty CNC machine tool ZK5540A. Experiments proved that the model outperforms the comparative models under varying working conditions.