Performance of ANN in predicting the depth to width ratio and tensile strength of UNS S32750 laser weld joints

Abstract The effectiveness of the pulsed mode laser weld joint is characterized by the supply of optical energy to the interface. The laser welding process parameters, such as laser power, pulse duration, travel speed and focus, determine the efficacy of the weld, which is defined by full penetration and free of pores and defects. However, expressing the relationship between various process parameters and mechanical strength is intricate due to the prevalence of non-linear relationship. The development of computational approaches aids in optimizing the laser welding parameters and reducing trial and error. Hence, the artificial neural network (ANN) model is developed, in a python environment, to predict the optimum process parameter values for a desirable depth to width ratio and maximum tensile strength of UNS S32750 laser butt weld joints. The developed model was assessed by experiments, not utilized for training. The ANN model predicts the depth to width ratio and tensile strength of the weld joints with an accuracy of 90% and less than 10% divergence from the experimental result. Furthermore, for the process, parametric conditions are: (power: 550 W, focus: –1 mm, pulse duration: 13 Hz and travel speed: 136 mm/min) to attain maximum tensile strength.