GPU-accelerated meshfree simulations for parameter identification of a friction model in metal machining

Abstract

In the analysis of metal machining processes using meshless methods, friction is usually modeled (if at all) by Coulomb’s law with a prescribed constant coefficient. Experimental observations, however, show that the coefficient μ of friction in such processes is not constant but generally a decreasing function of temperature. In this study, an in-process tribometer experiment is initially conducted on a Ti6Al4V workpiece to acknowledge that μ is in fact temperature-dependent. Subsequently, an enhanced Coulomb law is proposed whose coefficient μ(T) is a decreasing function of temperature. The unknown parameters of μ(T) are determined by a force optimization of iterative simulations carried out on several configurations. To this end, the present article takes 5 different cutting geometries from the literature considering 3 alternative sets of Johnson-Cook parameters for the Ti6Al4V constitutive model. This combination leads to 15 case studies in total. To tackle the very expensive cost of computation associated with this massive load of simulations, a GPU-accelerated meshless implementation is employed. Results of the present investigation demonstrate that: (1) friction modeling at the tool-chip interface has a remarkable influence on the numerical simulations of machining; (2) reliability of the friction parameters is substantially interrelated with the choice and reliability of the constitutive model parameters. As a result of this work, the error of force prediction in meshfree cutting simulations can be significantly reduced by adopting an enhanced friction model.