Artificial Neural Network for Predicting Blank Size in Stretch Flange Process

This research utilizes an Artificial Neural Network (ANN) to predict the initial blank size in the stretch flange process. The part material is JIS SPCC steel sheet with a thickness of 1 mm. Finite Element Method (FEM) software, AUTOFORM, was used to simulate and determine the initial blank profile, as well as to define the parameter ranges. The ANN inputs included the part concave radius (R) of 39–45 mm, part width (W) of 40–70 mm, part flange radius (r) of 5–8 mm, and part flange height (H) of 10–30 mm. The ANN outputs were the dimensions of the initial blank size. MATLAB was used to train and test the ANN with 196 training sets and 30 testing sets. The predicted initial blank size from the selected ANN architecture closely matched the FEM simulation results, with a maximum error of 8.74%.