A modeling method for deformation resistance in the forging process integrating physical priors and data-driven approaches

Abstract The dimensional accuracy of precision forgings impacts their mechanical properties and cost, making it a critical research focus in the industry. This accuracy depends on precise control of the forging process, which in turn relies on accurate modeling of the entire forging sequence—particularly the modeling of deformation resistance. To address the strong nonlinearity and complexity of forging processes, this study embeds a data-driven model into a physical framework to establish a real-time online model. The validity of this model was verified using finite element simulation results. Simulations indicate that the relative prediction error of the new model remains largely within ±1%, enabling it to dynamically predict the highly nonlinear forging force in real time, while maintaining strong robustness even at points of abrupt model changes.