Spindle thermal error prediction modeling using vision-based thermal measurement with vision transformer

Modeling and compensation of thermal errors is one important way to enhance the machining accuracy of the machine tool. This paper proposes a thermal error prediction model using the thermal measurement with the infrared thermal imager and Vision Transformer (ViT). Firstly, thermal measurement of spindle is proposed with the vision of the infrared thermal imager. Data enhancement and expansion are realized for data quantity and quality. Secondly, a ViT-based thermal error model is introduced by analyzing the encoder and the multi-headed attention mechanism by inputting the thermal images. Self-attention mechanism can mine the key information in thermal images. Thirdly, the optimization selection of the number of encoder block stacks and the head number of the multi-head attention mechanism affecting the structure of ViT model are proposed. Finally, different validations are proposed to testify the effectiveness and the robustness of the ViT-based model by comparing with existing image prediction models.