A Model for Evaluation of Laser Welding Efficiency and Quality Using an Artificial Neural Network and Fuzzy Logic

For any welding process, efficiency and quality strongly depend on the energy input, which is the energy introduced per unit length of weld from a travelling heat source. The focused laser beam is one of the highest power density sources available to the welding industry today, which makes it possible to weld with very low energy input with respect to most of other welding processes. In this paper a number of stainless steel butt joints were produced by laser irradiation. The welding efficiencies were calculated as the melted volume-energy input ratio. Moreover, the weld crown and depth were measured in order to evaluate the joint quality. The collected data were interpolated and correlated to the process parameters using an artificial neural network. They were then clustered using a fuzzy C-means algorithm. During the training stage of the neural network algorithm, the design of experiment (DOE) technique was used for the selection of the optimized network parameters. In practice, using some artificial intelligence, a model was built to choose the most suitable laser welding process for producing high efficiency and good quality, and is now available for supporting design and research.