Control of GMA Butt Joint Welding Based on Neural Networks

This paper presents results from an experimentally based research on Gas Metal Arc Welding (GMAW), controlled by the artificial neural network (ANN) technology. A system has been developed for modeling and online adjustment of welding parameters, appropriate to guarantee a high degree of quality in the challenging field of butt joint welding with full penetration under stochastically changing boundary conditions, e.g. major gap width variations.

GMAW experiments performed on mild-steel plates (3 mm of thickness), show that high quality welds with uniform back-bead geometry are achievable for gap width variations from 0.5 mm to 2.3 mm - scanned 10 mm in front of the electrode location. In this research, the mapping from joint geometry and reference weld quality to significant welding parameters has been based on a static multi-layer feed-forward network.

The Levenberg-Marquardt algorithm, for non-linear least square error minimization, has been used with the back-propagation algorithm for training the network, while a Bayesian regularization technique has been successfully applied for minimizing the risk of inexpedient over-training.