Inverse Identification of Johnson-Cook Parameters Using an Invertible Neural Network

The Johnson-Cook (J-C) model is a widely used constitutive model for describing the deformation and failure behavior of metals under various strains, strain rates and temperatures. However, deviations often exist between simulated and experimental results, such as cutting forces, meaning that reliable cutting simulations require accurate material model parameters. The inverse identification of J-C parameters has therefore been an active research topic. Nevertheless, the inverse problem is inherently ambiguous, as different combinations of J-C parameters can produce similar cutting forces in simulations. This study aims to predict possible J-C parameter sets based on cutting and thrust forces obtained from a series of orthogonal cutting simulations based on AISI 1045 steel. An Invertible Neural Network (INN) is employed to inversely generate J-C parameter combinations, and real orthogonal cutting experiments are conducted to validate the reliability of the generated parameters. The results demonstrate that the INN can generate reliable and physically consistent J-C parameters.