• 목록

Structural mechanics simulation for process design in the steel industry relies primarily on the finite element method. The finite element method is traditional and accurate, but the higher the desired resolution, the more time and computational resources are required. The computational resource issue is even more significant in the rolling process, which involves continuous rolling mills. This study proposes an artificial neural network-based surrogate model to predict the deformation of H-shaped steel in the intermediate rolling process as an alternative to the finite element method. The finite element simulation data was collected by fractional factorial design, and features reflecting the relationship between geometry and deformation were added to increase the explanatory power. The trained artificial neural network model successfully predicts the deformation of steel under the untrained process condition, showing a performance of R2 = 0.993. The proposed method reduces the prediction time from 6 hours to 5 seconds.