Prediction of Friction Coefficients During Scratch Based on an Integrated Finite Element and Artificial Neural Network Method

Abstract

Abstract An integrated finite element and artificial neural network method is used to analyze the impact of scratch process parameters on some variables related to elastoplastic deformation of titanium alloy. The elastoplastic constitutive parameters applied for scratch simulations are obtained from the nanoindentation experiments and finite element analysis. The validity of the finite element model of scratch is confirmed by comparing the friction forces from simulations to those from experiments. The input parameters of the artificial neural network are three scratch process parameters: tip normal force, tip radius, and shear friction coefficient. The outputs are four variables related to material deformation measured during scratch: scratch depth, elastic recovery height, plowing height, and plowing friction coefficient. The network is trained with pairs of input and output datasets generated by scratch simulations. The prediction results of the neural network are in agreement with the finite element results. The model provides assistance for the prediction and analysis of complex relationships between scratch process parameters and variables related to material deformation, and between the plowing friction coefficient and the relevant parameters. The results show the independence of scratch depth and the shear friction coefficient, and the positive relationships between the shear friction coefficient and plowing friction coefficient.