基于改进神经网络和 Fluent 的气液固技术的内表面处理

At present,the manufacturing methods of complex multi-bore workpieces in aviation and aerospace often have raised burrs on the inner bore surface of the workpieces,which are not easy to remove. In view of this situation,a gas-liquid-solid multiphase flow technology is proposed to treat the inner surface of complex workpieces. Based on fluent numerical analysis,different gas pressure,water velocity and abrasive particle concentration parameters under multiphase flow are simulated. The parameters of abrasive particle velocity,relative pressure and abrasive particle volume fraction at the inner surface of the near-wall area which affect the machining are obtained. The parameters are fitted by using BP neural network with MATLAB software,and the optimal input value of gas-liquid-solid three-phase flow satisfying the constraint mathematical model is solved by using the BP prediction model and PSO (Particle Swarm Optimization) in the dominant solution set. Based on the optimal input value,an experimental platform is built and L(₉ 3³)orthogonal experiment is designed to machine the inner surface of the workpiece. Finally,the surface accuracy of the inner hole before and after machining is measured by a white light interferometer,which verifies the consistency between the simulation optimal parameters and the experimental optimal parameters. The inner surface of the workpiece is machined by using the gas-liquid-solid three-phase flow with the optimal parameters. The measurement shows that the inner surface accuracy is increased by 75%,which meets the application requirements of aerospace workpieces.