Integrating machine learning and metallurgical processing for the design of eutectic high-entropy alloys with optimized nano-mechanical performance

High-entropy alloys (HEAs) are promising for structural applications due to their superior mechanical and corrosion properties. Six machine learning (ML) algorithms were evaluated for predicting phase formation, with the random forest (RF) model achieving the highest accuracy (87%) and ROC–AUC score (0.954). Using the optimized ML framework, Fe(25-x)Co25Ni25Cr20V5Tax (x=2.5–10 at.%) HEAs were designed and fabricated. Metallurgical characterization confirmed a single-phase solid solution with increased lattice distortion at higher Ta contents. Tensile strength improved from 498 MPa to 1110 MPa, while nanoindentation revealed hardness from 2.655–5.083 GPa and modulus from 212.71–286.84 GPa. Scratch testing indicated an initial rise in the coefficient of friction during creep, followed by transient and steady-state behavior, consistent with 3D wear surface reconstructions. The close agreement between ML predictions and experimental results demonstrates the effectiveness of data-driven alloy design for tailoring microstructure and optimizing mechanical performance in advanced HEAs.