Study of cutting forces using FE, ANOVA, and BPNN in elliptical vibration cutting of titanium alloy Ti-6Al-4V

Abstract

In this study, the finite element (FE) analysis, analysis of variance (ANOVA), and the back propagation neural network (BPNN) were employed to evaluate and predict the cutting forces of titanium alloy Ti-6Al-4V depending on vibration frequency, tangential amplitude, and thrust amplitude, as well as cutting speed during elliptical vibration cutting (EVC). A series of EVC simulations were conducted based on the verified FE model to evaluate the impacts of different EVC parameters on cutting forces. The results show that the tangential force decreases with increasing vibration frequency, tangential amplitude, and thrust amplitude, but with decreasing cutting speed. The positive and negative thrust forces decrease with increasing frequency and tangential amplitude, but with decreasing thrust amplitude and cutting speed. In addition, ANOVA results clearly indicated that the tangential amplitude is the dominant parameter affecting the cutting forces, and the percent contributions to cutting forces are 69.56%, 66.03%, and 62.83%, respectively. Further, the BPNN models with three different activation functions and different architectures are utilized to predict the cutting forces, and the best performance, in terms of agreement with the target outputs, can be achieved by the network using logarithmic sigmoid activation function and architecture with 15 neurons in one hidden layer. The correlation coefficients for training and testing the selected network are 0.99993 and 0.99916, and the mean square errors are 0.1963 and 2.6070, respectively, and these reveal that BPNN is fairly successful in predicting the cutting forces.