Prediction model of surface deflection of rectangular drawing products using finite element analysis and machine learning

The predicted amount of surface deflection that affects the appearance of the vehicle can be quantified as the maximum variation in the curvature. Prediction of the surface deflection according to the material properties is difficult owing to its nonlinearity. In this study, prediction model of surface deflection was studied using artificial neural network technique through supervised learning. Finite element analysis was used to quantify the microfine curvature, and data groups were generated through a variation in the material property values. The effect of various techniques and tools was studied for the construction of optimized artificial neural networks. Comparing the predictions using artificial neural networks with those of statistical regression, we found that the mean error was significantly lower when using such networks.