Weld pool images provide abundant information about weld penetration while the weld penetration has a close relationship with the back weld width. The key to achieving weld penetration control is to establish a model describing the relationship between the front-side geometrical parameters of the weld pool and the back weld width with sufficient accuracy. By using a three-layer neural network, a new method to establish such a model has been proposed in this paper. The testing results show that the model has sufficient accuracy and can meet the requirements of weld penetration control. The model is used in the back weld width control test and a satisfactory result is obtained.
KovacevicR.ZhangY. M.LiL.Monitoring of weld joint penetration based on weld pool geometrical appearance. Weld. J., 1996, 75(10), 317s–329s.
2.
ChenS. B.LouY. J.ZhaoD. B.Intelligent methodology for sensing, modeling and control of pulsed GTAW. Part I: Bead-on-plate welding. Weld. J., 2000, 79(7), 151s–163s.
3.
HuangS.LiD.SongY.Modeling and controlling of welding process by neural network. Chin. J. Mech. Engng, 1994, 30(3), 24–30.
4.
GaoJ.WuC.Vision-based measuring of weld pool geometry in TIG welding. Acta Metallurgica Sinica, 2000, 36(12), 1284–1288.
5.
DataEngine Overview and User Manual [Z], 1997 (MIT GmbH, Aachen, Germany).
