Learning impact of CAD geometry change on finite element analysis results

ABSTRACT: This study examines how CAD geometry variations affect finite element (FE) crash simulations for automotive front rail assembly and motivate the use of combined impact measures that better reflect the physical response. Based on these insights, we outline a machine learning formulation that links geometric modifications to their simulation effects. The study centers on geometric representation, employing UVbased graph encodings to capture local shape changes and provide the basis for advancing and validating the full prediction pipeline.