Optimization of lap-joint laser welding parameters using high-fidelity simulations and machine learning mode

In lap joint laser welding, a common practice is to conduct trial-and-error experiments using various parameter settings to determine processing conditions that enhance the quality of the weld. However, these experiments are both time-consuming and expensive. Therefore, in this study, we propose a more systematic approach for determining the optimal laser power and scanning speed in the lap joint of SS316 by using highly accurate simulations and artificial neural network models. The processing maps were obtained for three criteria: the melt pool depth, melt pool width, and cooling rate, respectively, which were screened using appropriate quality criteria to determine the laser power and scanning speed that could simultaneously minimize porosity, the size of the heat affected zone, and residual stress. The validity of the simulation model was confirmed by comparing the simulation results of the melt pool geometry with the experimental data. The mean deviations of the experimental and simulated results for melt pool depth and width were found to be only 5.34% and 10%, respectively. As a result, the joint welds produced using the optimal processing parameters exhibited minimal porosity, which was reduced from 1.22% in a non-penetration zone to 0.21% in an optimized zone. Additionally, these welds achieved an ultimate shear strength of up to 545.77 MPa, which is approximately 32% higher than that of the original base metal. Therefore, the effectiveness of the proposed framework for determining the optimal processing conditions for joint laser welding of SS316 has been confirmed.