Machine Learning and Finite Element Simulation Based Shape Optimization of a Rubber Bumper

Rubber components play a critical role in the automotive industry, necessitating significant engineering effort in their design. This study introduces an efficient methodology for optimizing the geometry of axisymmetric rubber bumpers for practical applications. The proposed framework integrates the Finite Element Method (FEM) with a Support Vector Regression (SVR) model. A finite element program based on a so called three-field functional written by the authors was used to determine the rubber bumper's mechanical characteristics efficiently. The optimization objective was defined as minimizing the deviation between the prescribed and computed load-displacement curves, which is the work difference. SVR facilitated the creation of a regression model for the work difference, enabling precise optimization of design variables. The procedure's effectiveness is demonstrated through a numerical example involving three design variables. The results confirm that the proposed methodology delivers accurate and efficient optimization, making it suitable for application in industrial design processes.