| Abstract Scope |
Conventional superalloys designed for casting, forging, and wrought processing are often incompatible with the steep thermal gradients, rapid solidification, and residual stresses of metal additive manufacturing, leading to cracking, segregation, and microstructural instability. Here, we demonstrate a self-driving alloy design system integrating active learning, autonomous laser-directed energy deposition, and computer-vision-based quality assessment to close the loop between composition selection, fabrication, printability evaluation, and model-guided decision-making. The system was applied to a crack-prone Co-based 706 superalloy and given compositional freedom to introduce Ni, Cr, Zr, and Ni–W additions. Candidate alloys were autonomously deposited, assessed for cracking, and used to guide subsequent experiments. Rather than optimizing processing conditions alone, the framework searched for chemistries intrinsically tolerant to additive manufacturing thermal cycles. Iterative learning converged toward crack-free deposition, demonstrating a pathway for autonomously converting difficult-to-print legacy superalloys into additive-manufacturing-compatible materials. |