Time-dependent surrogate modeling for metal sheet deformation using random forests for a dieless sheet metal forming process

Abstract Accurate prediction of metal sheet deformation during forming processes is essential for optimizing manufacturing performance and minimizing trial-and-error experimentation. However, traditional finite element simulations, such as those performed in FEM software, are computationally intensive and unsuitable for real-time evaluation or iterative design. This study begins by validating the feasibility of using Random Forest Regression as a surrogate modeling method to replace traditional FEM simulations in simplified metal forming scenarios. It then extends this approach to a new machine configuration consisting of two independently rotating cones (acting as dies) and a crease wheel (acting as a punch) demonstrating the model’s potential for more complex, real-world applications. This research highlights the feasibility of integrating surrogate models into intelligent forming systems, significantly reducing simulation time while maintaining accuracy. The proposed framework lays the groundwork for real-time predictive control and digital twin development and robot implementation in advanced sheet metal forming applications.