Interpretable physics-guided multi-output learning for optimizing circular-path cold metal transfer wire arc additive manufacturing of ER5356 aluminium alloy

Wire Arc Additive Manufacturing (WAAM) has emerged as a promising technology for fabricating large-scale aluminium components; however, reliable process optimization remains challenging owing to the complex interactions among deposition parameters, thermal history, and mechanical performance. This study proposes an interpretable physics-guided machine learning framework for predictive modelling and multi-objective optimization of circular-path Cold Metal Transfer (CMT)-WAAM of ER5356 aluminium alloy. A Taguchi L27 experimental design with three centre-point replications was employed to investigate the effects of wire feed speed, travel speed, interpass temperature, and shielding gas flow rate on ultimate tensile strength, elongation, flexural strength, and hardness. Physics-guided descriptors representing deposition intensity, thermal accumulation, cooling severity, and shielding adequacy were integrated into a multi-output stacking regressor to embed manufacturing knowledge and provide a physically structured representation of the process variables. The optimized process window predicted an ultimate tensile strength of 295.20 MPa, 40.10% elongation, 397.80 MPa flexural strength, and 71.80 HV hardness. A single confirmatory experiment produced deviations below 5%, with composite desirability decreasing from 0.9989 to 0.9565. Fractographic analysis revealed ductile failure through microvoid coalescence and ligament stretching, providing additional microstructural support for the behaviour observed at the optimized condition. By integrating physics-guided feature engineering with interpretable multi-output learning, the proposed framework provides a data-efficient and data-efficient decision-support approach within the investigated WAAM process window tool for intelligent WAAM process optimization, offering a practical pathway toward AI-enabled, digitally integrated additive manufacturing.