Synchronized molecular dynamics-machine learning framework for tailoring mechanical properties of FeNiCr medium entropy alloys

FeNiCr medium entropy alloys (MEAs) have emerged as a prominent candidate for extreme environments due to their balanced mechanical robustness, corrosion resistance, and thermal stability. While the conventional experimental approaches are hindered by prohibitive costs and protracted iteration times, the data scarcity further limits the existing machine learning (ML) frameworks for multi-component alloy optimization. Here, a synergistic computational paradigm integrating molecular dynamics (MD) simulation with a back-propagation (BP) neural network is established to systematically decode the composition-property interplay in FeNiCr MEAs. The key findings indicate that Fe primarily enhances strength by amplifying dislocation interactions, but reduces stiffness. Ni exhibits a non-monotonic strengthening effect, peaking at 20-30% to maximize both strength and stiffness via twin boundary refinement. When beyond 20%, Cr degrades the mechanical performance by impeding dislocation glide while promoting brittle phase nucleation. Atomic simulations further correlate Fe/Ni ratio with deformation mode transitions: Fe facilitates extended stacking fault formation, and high concentrations of Ni are beneficial for twin boundary proliferation. This dual-scale framework demonstrates precision in balancing alloy performance metrics, suggesting transformative potential for rapid development of next-generation structural alloys tailored for aerospace propulsion and advanced nuclear systems.