Accurate optimization of blank design in stretch flange based on a forward–inverse prediction scheme

This paper presents a forward–inverse prediction scheme that combines explicit dynamic finite element method (FEM), true strain method (TSM), and adaptive-network-based fuzzy inference system (ANFIS) to determine the anisotropic optimum blank in stretch flange process. It is not easy to guess the blank shape in stretch flange process using traditional trials. The forward–inverse prediction scheme can efficiently determine the optimum blank and improve the shrink phenomenon of the flange target. In the present study, the initial shape of the nearby optimum blank could be predicted quickly by the forward scheme of TSM at first. Then, several blanks with parallel curves to this shape are created and analyzed by explicit dynamic FEM to establish the training database for the inverse scheme of ANFIS. From the hybrid-learning algorithm, ANFIS can be efficiently trained for the optimum blank shape, and the prediction knowledge rule database can be accomplished. Using the trained ANFIS, the exact optimum blank shape of stretch flange can be inversely predicted more correctly and efficiently. Moreover, the stretch flange experiments are performed using non-optimum and optimum blanks. It also shows that a good agreement is achieved from comparison between simulated and experimental results, such as punch load, punch stroke and deform shape. From this investigation, the forward–inverse scheme is proved to be able to supply a useful optimum preform in the metal forming category.