A cutting force prediction model for ultrasonic elliptical vibration cutting of tungsten alloys based on NSGA-II-improved BP neural network

Abstract

Ultrasonic elliptical vibration cutting (UEVC) can effectively solve the problems of tool severe wear and poor surface quality in ultra-precision machining of tungsten alloys. This paper provided a theoretical quantitative analysis of the time-varying states of cutting speed, cutting acceleration, and cutting angle in UEVC. Afterward, a finite element model of UEVC for tungsten alloys was established, and the finite element cutting model was modified using cutting experiments. The influence of vibration frequency and two-phase amplitude on cutting force was studied, and the sensitivity of vibration parameters on cutting force was determined using variance analysis. Finally, by further optimizing the BP neural network model using the NSGA-II algorithm, a NSGA-II/BP cutting force prediction model was established. Compared with the BP model, the prediction accuracy was significantly improved, with a minimum prediction error of 2.61%. This provides a theoretical basis for the optimization of process parameters for UEVC of tungsten alloys, and is of great significance for improving the ultra-precision machining level of tungsten alloys.