An integrated numerical and experimental investigation of direct laser welding between BOROFLOAT® 33 glass and AA1050 aluminum is presented to address the persistent challenges arising from their substantial thermal and mechanical property mismatches. A three-dimensional COMSOL Multiphysics model was developed to simulate transient heat transfer and thermally induced stress under a wobbling nanosecond laser, incorporating a volumetric heat source for glass and surface heat sources for aluminum. To predict crack initiation and propagation in the brittle glass, a fully coupled thermo-mechanical phase-field model was implemented, enabling the continuous evolution of damage driven by thermal gradients and residual stresses. Simulation results reveal that high laser power and low scanning speed significantly elevate damage-driving energy and promote cracking, whereas moderate energy input confines stress localization. Experimental welding trials validated these predictions, showing close correspondence between simulated damage fields and observed molten morphology, pore formation, and crack behavior. To further optimize process parameters, artificial neural networks (ANNs) were trained on phase-field damage outputs to construct a high-resolution processing map. The resulting map identifies an optimal low-damage window at ∼30–50 W with a scanning speed of 3–5 mm/s, yielding crack-free seams and a maximum shear strength of 8.5 MPa—surpassing many reported glass-metal welding benchmarks. This combined computational–experimental framework provides a quantitative strategy for optimizing glass–metal laser welding without interlayers, offering a robust foundation for future optoelectronic, microfluidic, and packaging applications.
Wang et al. (Sun,) studied this question.