Research on thermal error modeling of CNC milling head for five-axis machine tools based on GA-ACO optimized BP neural network

Abstract

The predictive performance of the thermal error model determines real-time compensation effect of the Computer Numerical Control (CNC) milling head, and the identification of temperature-sensitive points directly affects the robustness of the modeling. Therefore, a genetic algorithm and ant colony algorithm optimization based BP neural network (GA-ACO-BPNN) model is proposed to improve the generalization and robustness of the thermal error prediction model. This research Takes the 5AS milling head as the research object, the thermal characteristics are analyzed to determine its temperature field distribution. The K-means + + algorithm and gray correlation analysis are used to reduce 11 temperature measurement points to 4, which improves the input quality of the model. Combined with GA and ACO to search for optimal parameters of the BP network, it overcomes the problems of slow convergence speed and quickly falls into local extremes. Compared with the radial basis function neural network (RBFNN) model and ACO-BPNN model, the average value of Root Mean Square Error (RMSE) of the Z-directional thermal errors predicted by the GA-ACO-BPNN model is reduced by 45.3% and 58.3%; and the average value of Mean Absolute Error (MAE) is decreased by 65.2% and 53.8%, respectively.