摘要
In this study, a thermal balance control method by local heating on the spindle head structure of a vertical milling machine was proposed to improve the squareness error of the Y-Z plane. Thermal deformation and temperature distribution of the spindle head structure was analyzed to determine the key positions on the structure for prediction and control of the spindle pose. A long short-term memory (LSTM) neural network with time series characteristics was used to build the in-situ spindle tilt error prediction model. The particle swarm optimization algorithm was integrated with a neural network model to dynamically determine and find temperature control targets in local positions on the spindle head. Finally, the temperatures at selected local locations of spindle head was adjusted by elective heating to suppress thermal bending and improve the spindle tilt on the Y-Z plane. The proposed method was verified using a five-axis vertical milling machine. The experimental results showed that the proposed method could effectively maintain the spindle tilt in the Y direction at [10, −10] μrad after long-time operation of the spindle at both 7200 and 12,000 rpm.