Modeling Study On Prediction Of Thermal Error In High-Speed Electric Spindle

As one of the core functional components of high-speed machining systems in the manufacturing industry,the electric spindle has a direct impact on the machining accuracy of parts.During the operation of the electric spindle,the heat generated inside has caused temperature rise,which leads to axial deformation of the spindle and directly affects the machining accuracy and operational status of the machine tool.Therefore,based on the precision requirements for high-speed electric spindle operation,it is necessary to analyze and predict the thermal properties,temperature rise,and axial thermal deformation of high-speed electric spindles in detail.This paper takes the electric spindle 170SD24Q15 as the experimental object and systematically investigates the inherent causes of axial thermal error and temperature rise at various thermal sensitive points.Furthermore,a neural network algorithm model for predicting axial thermal error and temperature rise of the electric spindle is proposed.The main research contents of this paper include:(1)Finite element thermal modeling of the electric spindle.By applying thermodynamic transfer principles and considering the heat transfer mechanisms of the electric spindle,the heat generation rates of the stator,rotor,front bearing,rear bearing,heat transfer coefficients between the rotor end and surrounding air,heat transfer coefficients of the stator and cooling system,convective heat transfer coefficients between the stator-rotor and compressed air in the gap,as well as the heat transfer coefficient between the electric spindle casing and air were calculated as boundary conditions.Then,ANSYS Workbench was used to load the boundary conditions and relevant material properties into the finite element model for thermodynamic simulation.A thermal experiment of the electric spindle was designed,and temperature rise data at the stator,rear bearing housing,and front bearing outer ring were collected.The simulation results were compared with the experimental results,showing that the simulation results were slightly higher than the experimental results but followed a similar trend.(2)Three methods,namely Pearson correlation coefficient,system clustering,and fuzzy clustering,were studied for the selection of thermal sensitive points on the electric spindle.Accurate selection of thermal sensitive points is a prerequisite for accurate prediction modeling of thermal errors.Appropriate thermal sensitive points help obtain thermal error models with higher prediction accuracy and generalization ability.Using the three methods mentioned above,the six thermal sensitive points were classified into two categories.Then,by calculating the correlation coefficient between the temperature rise value of each thermal sensitive point and the axial thermal deformation,representative points for each category were determined.(3)Modeling of axial thermal errors and temperature rise prediction for the electric spindle.Based on the optimization of the number of hidden layer nodes,a closed-loop Improved Grey Wolf Optimizer-Long Short-Term Memory(IGWO-LSTM)neural network model was developed for predicting axial thermal errors and internal temperature rise of the electric spindle.The model predicted the temperature rise and axial thermal errors of the stator,spindle,front bearing outer ring,shaft end,rear bearing housing,and rear end cover of the electric spindle.The model was evaluated by comparing four metrics,namely maximum absolute error,mean absolute error(MAE),mean absolute error percentage(MAEP),and root mean square error(RMSE),with a single Long Short-Term Memory(LSTM)neural network.The results showed that the IGWO-LSTM closed-loop neural network outperformed the single LSTM neural network in all metrics,indicating that the proposed neural network can better predict the changes in axial thermal errors and internal temperature rise of the electric spindle.