Process parameters optimization of laser beam welded joints by neural network

Laser beam welding of C-Mn steel plates with Ni powder filler metal has been performed. Metallography samples of the welded cross-section have been observed by scanning electron microscopy (SEM) and submitted to energy dispersive spectroscopy to obtain Ni concentration profiles. On the basis of the experimental results, neural networks have been carried out. These networks were first validated and then utilized to foresee Ni concentration along the welded thickness. The objective of obtaining the best Ni penetration and minimizing powder loss was reached optimizing, by numerical simulation, process parameters, such as powder rate and joint geometry.