Deformation prediction and quality evaluation of the gas metal arc welding butt weld

The control of distortion and the overall quality are the main targets in the design and manufacturing of sound welds. In this paper a numerical approach is presented in order to support the choice of process parameters that can minimize thermal deformation and evaluate weld quality for gas metal arc welding (GMAW) operating with the short-circuiting transfer mode. This numerical approach is based on the integration of artificial intelligence (AI) techniques and the finite element method (FEM). The information to train the AI and to validate the FEM came from experimental trials. Firstly, a number of simple artificial neural networks were trained and validated. They linked the process parameters to the geometry of the molten zone of the welds. In this way it was possible to calculate the geometries throughout the range of the process parameters. Thereafter, the finite element model provided useful information about the residual stress and the shrinkage distortion of the welds. Finally, a concise evaluation of joint quality was possible using a fuzzy C-means clustering algorithm. The 'minimum is the best' rule was applied during the training phase. The numerical model for GMAW was constructed and validated for butt welds of thin plates made of mild steel.