Prediction of Effective Material Property of Tailor Welded Blanks for Air Bending Process

Abstract

A method based on artificial neural network (ANN) for predicting the effective material property is put forward in this paper. The finite element model of tensile test specimen is modeled in LS-DYNA code, which has transverse weld at the middle of the specimen and conforms to the ASTM specification. A statistical error analysis model is used to include the random phenomenon in the result of tensile test finite element model and verify the accuracy of finite element model (FEM) simulation. In order to study the effect of the processing parameter with design of experiment is followed, the simulation trail is conducted in all the levels of parameters. It is assumed that Hollomon’s law is followed by tailor welded blanks. The results obtained from fitting the post-process data of FEM by least square method are used to train and develop ANN model, the prediction average error of ANN model is acceptable compared with simulation trail.