MA-CNN based spindle thermal error modeling using the depth feature analysis with thermal error mechanism

The dynamic changes in the temperature fields make it unavoidable for the spindle to undergo thermal deformation. Thermal error modelling is helpful in enhancing the accuracy of machine tools. This paper proposes a spindle thermal error modelling method based on the mayfly algorithm (MA) for complex working conditions. First, different working conditions are set to obtain thermal deformation and temperature data. A thermal error model is established using MA. MA is used to optimize the number and size of convolutional kernels. Next, the depth features of the model are compared with the exponential thermal error model in logarithmic space. The model is compared with five other models. The results indicate that the MA -CNN model has a higher accuracy. Finally, the compensation is conducted in three directions. The results show that the average thermal error in the Z direction is reduced by 63.75 %, which demonstrates the good performance of the model.