Machine learning-driven design of BCC phase FeCrVTiMoxSiy high-entropy alloy coatings with high hardness to enhance wear resistance

Abstract

High-entropy alloys (HEAs) have been widely considered as promising materials to protect the Ferritic/Martensitic (F/M) steels against the extreme environments in the lead-cooled fast reactors (LFR). Due to the wide diversity of elemental compositions and ratios, the rational design of HEAs with high wear resistance remains a huge challenge. In this work, we employed machine learning (ML) methods to guide the design of HEAs with high wear resistance as the protective coating for the F/M steels. The ML-based models were constructed to predict the phase structure and hardness of HEAs. The constructed SVM and XGBoost models exhibited the best performance in predicting the phase classification and the Vickers hardness of HEAs, respectively. Valence electron concentration (VEC) and ΔH_mix are identified as the most important factors affecting both the phase structures and Vickers hardness of HEAs. With these models, the FeCrVTiMo_xSi_y HEAs were predicted to exhibit a BCC phase and increasing hardness with the decreased ratio of Mo and Si elements. The following experimental results showed that FeCrVTiMo_0.5Si_1.5 exhibited optimal wear resistance with Vickers hardness, Young's modulus, H/E, H³/E², and wear rate of 732.65 HV, 289.6 GPa, 0.0353, 0.0127 GPa, and 8.65 × 10⁻⁷ mm³/(N·m), respectively. Density functional theory (DFT) calculations revealed that decreasing the ratios of Mo and Si elements in FeCrVTiMo_xSi_y HEAs increases lattice distortion and increases the proportion of covalent bonds to enhance solid-solution strengthening, improving wear resistance. This work presents a paradigm shift in quantifying the relationship between elemental compositions and the properties of HEAs.