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

摘要

Laser welding parameters include notonly the laser power, focused spot size, welding speed,focused position, etc., but also the welding gap and thealignment of the laser beam with the center of thewelding gap, these latter two parameters being criticalfor a butt joint. These parameters are controllable inthe actual operation of laser welding, but areinterconnected and extremely non-linear, suchproblems limit the industrial applicability of the laserwelding for butt joints. The neural network techniqueis a useful tool for predicting the operation parametersof a non4inear model. Back propagation (BP) andlearning vector quantization (LVQ) networks arepresented in this paper to predict the laser weldingparameters for butt joints. The input parameters of thenetwork include workpiece thickness and welding gap,whilst the output parameters include optimal focusedposition, acceptable welding parameters of laser powerand welding speed, and welding quality, including weldwidth, undercut and distortion for the associated powerand speed used. The results of this research show acomprehensive and usable prediction of the laserwelding parameters for butt joints using BP and LVQnetworks. As a result, the industrial applicability oflaser welding for butt joints can be expanded widely.