关键词: 表面残余应力, 铣削, 径向基函数, 神经网络预测, 分子动力学, 仿真

Abstract: To accurately predict the residual stress according to the pre-selected cutting parameters before machining,so as to improve the performance of the workpiece by adjusting the cutting parameters in advance,the research was carried out from the perspectives of macro prediction and micro analysis.At the macro level,to improve the prediction accuracy of residual stress,the Radial Basis Function (RBF) neural network model was constructed after the expansion coefficient SPREAD optimized with 45# steel as the research object,and then the prediction results was compared to the experimental value and the finite element simulation to prove the prediction after the SPREAD optimization had high accuracy and superiority.At the micro level,a molecular dynamics model was established to simulate the milling process under the same conditions.The simulation results of molecular dynamics were compared with the experimental results,and the result showed that there had existed a potential negative correlation between micro and macro of milling residual stress,which provided a certain feasibility to improve the residual stress by adjust the microstructure of the workpiece with process treatment.

Key words: surface residual stress, milling, radial basis function, neural network prediction, molecular dynamics, simulation