A phase prediction strategy of high entropy alloys manufactured by selective laser melting based on feature selection and machine learning

The excellent properties of high-entropy alloys (HEAs) and the high cooling rate characteristics of the selective laser melting (SLM) technique make them possess great application prospects. In order to make full use of the reported experimental data of SLM manufactured (SLM-ed) HEAs and to enhance the development efficiency of SLM-ed HEAs, this study carries out the phase prediction work of SLM-ed HEAs by machine learning techniques. We collected 172 SLM-ed HEAs samples (including 88 face-centered cubic (FCC) alloys, 31 body-centered cubic (BCC) alloys, and 53 dual-phase (FCC+BCC) alloys) and developed a machine learning (ML) HEAs phase structure prediction framework using volumetric energy density (VED), a process parameter of SLM, as one of the features, and ultimately achieved a prediction accuracy of 85%, established the linkage between the variation of VED and the phase structure of the SLM-ed HEAs. The reliability of the model was verified by experimentally preparing several Al x Cr y Cu z Fe u Ni w system HEAs. Larger atom volume ( V A ) and elemental density values differences ( δ D ) were found to favour the formation of the SLM-ed HEAs BCC phase by interpretable analysis, while smaller V A and δ D favoured the formation of the FCC phase. In addition to this, the sensitivity matrix and three-dimensional cubic section plots reveal that Al has a great influence on the phase formation of the Al x Cr y Cu z Fe u Ni w system of HEAs, with most of the alloys forming the FCC phase when the Al content is less than 10% (at%), and the majority of the alloys forming the BCC phase when the Al content is greater than 20%.