LAUSR

Abstract Grain boundary cracking in non-weldable superalloy fabricated with selective electron beam melting is affected by the interaction of multiple factors including mechanical and compositional effects. In this study, we construct process maps in a wide range high-dimensional parameter space for the non-weldable superalloy Alloy713ELC through employing a machine learning approach, and we could fabricate many cracked and crack-free samples under the optimized conditions by excluding the extrinsic effect of process defects on cracking. Comparing between cracked and crack-free samples reveals that the samples with fine columnar grains can be cracked while those with coarse columnar grains can be crack-free, and that the cracking propensity in the optimized samples within a process window with scan speed ≤ 800 mm/s can be ranked by using a quasi-total plastic strain index (QTPSI), which is calculated via thermo-mechanical analysis. The total plastic strain level is a critical cracking factor and a larger scan speed tends to elevate the total plastic strain level, exhibiting a larger deviation beyond the QTPSI. Besides, the non-weldability in Alloy713ELC significantly attributes to its thermal expansion effect, which correlates to the large Al content. This thermal expansion effect combined with the liquation effect and the strain-age cracking effect reveals the intrinsic cause of non-weldability in Alloy713ELC.