Experimentally Validated and Empirically Compared Machine Learning Approach for Predicting Yield Strength of Additively Manufactured Multi-Principal Element Alloys from Co–Cr–Fe–Mn–Ni System

Traditionally, yield strength prediction relies on detailed and resource-intensive microstructural characterization combined with empirical equations. However, quantifying microstructural feature length scales for novel processes like additive manufacturing, which involves inhomogeneous hierarchical features, poses a challenge. The lack of accurate material constants for broader composition ranges further limits empirical predictions. This study proposes an alternative machine learning (ML) approach for predicting the yield strength of additively manufactured (AM) multi-principal element alloys (MPEAs) from the Co–Cr–Fe–Mn–Ni system by correlating composition, printing parameters, and testing conditions. The top-performing ML model achieved an R² of 0.84, comparable to the best microstructure-based empirical strengthening model. The ambiguities and inconsistencies associated with empirical methods in the literature were critically evaluated for the Co_33.3Cr_33.3Ni_33.3 alloy. The validity of the ML approach was further confirmed by printing and testing two compositions (one novel and one from the dataset). This data-driven approach directly relates yield strength to initial printing parameters, highlighting their significance and individual effects, such as scan velocity's direct impact and laser power's inverse impact on yield strength. This demonstrates ML’s potential to guide AM processes, reducing the need for iterative experiments and enabling rapid exploration of compositional and printing spaces to achieve desired properties.