Machine-learning assisted design of as-cast NiCoFeCrAlTi multi-principal element alloys with tensile yield strength over 1.35 GPa

Abstract

As-cast alloys have the advantage of short forming processes, but there is currently a lack of research on systematic design alloys with better mechanical properties. Herein, combining a machine-learning with random forest model algorithm, a high-throughput alloy design framework under multidimensional constraints was used to discover new NiCoFeCrAlTi multi-principal element alloys (MPEAs) for superior tensile properties. The as-cast dual-phase Ni28Fe32Cr25Al10Ti5 alloy with 1386 MPa of tensile yield strength and 1.8% uniform elongation was designed, which is much higher than the best value in the original training dataset. This apparent high strength can be attributed to the phase interfacial strengthening, in which the soft face-centered cubic (FCC) phase precipitated extensively aside the grain boundaries of hard body-centered cubic (BCC) matrix. The BCC matrix provides high strength and FCC precipitates play role in ductility. Machine learning is expected to be utilized for designing as-cast MPEAs with superior mechanical properties.