Prediction of welding-induced angular distortions in thin butt-welded plates using artificial neural networks

Abstract

Angular distortions are one of the most common types of distortions frequently observed in thin butt-welded plates. Generally, welding-induced angular distortions lead to additional costs for rework. Therefore, estimation of angular distortions prior to welding can be beneficial from several standpoints. In this study, welding-induced angular distortions in single-pass butt-welded 304 stainless steel plates were predicted using artificial neural networks. In order to predict angular distortions a multilayer feedforward back propagation neural network was created using MATLAB. The input data of the neural network were obtained from a series of finite element simulations for a wide range of plate dimensions. The three-dimensional finite element simulations have been performed using uncoupled thermo-elasto-plastic analysis. In order to validate the numerical results a series of experiments have been performed and temperature histories and angular distortions have been measured for two different cases. The results of this study revealed that the neural network created in this study can accurately predict welding angular distortions produced in single-pass butt welded plates.