According to the energy consumption and workpiece surface roughness information of gear milling machine,a predictive optimization method combining multivariate nonlinear fitting and particle swarm optimization algorithm was proposed to provide the optimal process parameters for gear milling.Based on the dynamic structure of NC gear milling machine,the energy consumption model was established,and the concept of cutting specific energy of gear milling was put forward.Orthogonal and full factorial experiments were carried out to monitor the power and surface roughness data of multi-operating gear milling machine.The prediction model of machine tool specific energy and workpiece roughness was established by multivariate nonlinear fitting function.The fitting objective function group was substituted into particle swarm optimization algorithm to optimize the process parameters.The experimental results show that the fit goodness of the prediction model based on multivariate nonlinear fitting is more than 0.99,and the eight solution sets obtained by particle swarm optimization algorithm take into account the objective needs of machine tool energy saving and efficiency improvement.