Artificial Intelligence-Assisted Heat Input Design for Melt Pool Depth Consistency in Wire Arc Additive Manufacturing

Thermal accumulation is a characteristic feature of additive manufacturing (AM), which can lead to layer-dependent thermal histories, microstructural heterogeneity, and property variations along the build direction. To improve the consistency of melt pool depth in wire arc AM (WAAM), an artificial intelligence (AI)-assisted heat input design was proposed and evaluated through numerical simulation and experimental feasibility tests. A moving heat source model was first established to calculate the melt pool depth, and the simulated substrate penetration showed a mean absolute percentage error of 5.65% compared with the experimental measurements. The layer thickness and layer number in the numerical simulations were used as the inputs of the neural network, while the corresponding current was selected as the output. The correlation coefficient between the predicted and target current values was 0.99944. Compared with conventional WAAM under fixed heat input, where the melt pool depth varied from 2.72 to 6.67 mm, the proposed AI-assisted heat input design reduced the melt pool depth range to 1.92–2.72 mm under simulated WAAM conditions. In addition, WAAM deposition experiments using the AI-predicted layer-wise current schedule confirmed the practical implementability and macro-forming feasibility of the proposed strategy.