Gray Box Approach for Prediction of Air Bending

Abstract

Analytical predictions and black box approaches are well-known strategies for the prediction of forming processes. Each of them has its own advantages and weaknesses. Analytical predictions lack flexibility, but usually have a strong theoretical foundation. Black box approach means models based purely on empirical data, these models provides tailored solutions, but do not have any valid physical background. In this paper, we propose to merge these two concepts into a single procedure: a gray box approach. The gray box approach combines the best features from both prediction strategies: on the one hand, it provides a theoretical ground for the considered manufacturing process; on the other hand, it can be tailored for a given task. Prediction of the springback for the air bending process was taken as an example to test the concept. In the experimental investigation stage, construction steel S235JR sheets of various thicknesses have been used, which resulted in 174 bending tests. The obtained results have been used both to train and to test the gray box model. Validation tests revealed that the gray box approach is a suitable option for springback prediction, as it provides more accurate and consistent results than the analytical and black box approaches.