Prediction of laser butt joint welding parameters using back propagation and learning vector quantization networks

Laser welding parameters include not only the laser power, focused spot siz e, welding speed, focused position, etc., but also the welding gap and the alignment of the laser beam with the center of the welding gap, these latte r two parameters being critical for a butt joint. These parameters are cont rollable in the actual operation of laser welding, but are interconnected a nd extremely non-linear, such problems limit the industrial applicability o f the laser welding for butt joints. The neural network technique is a usef ul tool for predicting the operation parameters of a non-linear model. Back propagation (BP) and learning vector quantization (LVQ) networks are prese nted in this paper to predict the laser welding parameters for butt joints. The input parameters of the network include workpiece thickness and weldin g gap, whilst the output parameters include optimal focused position, accep table welding parameters of laser power and welding speed, and welding qual ity, including weld width, undercut and distortion for the associated power and speed used. The results of this research show a comprehensive and usab le prediction of the laser welding parameters for butt joints using BP and LVQ networks. As a result, the industrial applicability of laser welding fo r butt joints can be expanded widely. (C) 2000 Elsevier Science S.A. All ri ghts reserved.