Machine learning assisted design of Fe-Ni-Cr-Al based multi-principal elements alloys with ultra-high microhardness and unexpected wear resistance

In this work, machine learning (ML) technique was used to discovery new multi-principal elements alloys (MPEAs) with desirable properties. Generalized Regression Neural Network (GRNN) showed high accuracy to construct the composition-microhardness model and was used for microhardness prediction and composition optimization. Based on ML results, Fe0.6Ni0.7CrAl MPEAs were designed and prepared. The proposed GRNN model aligns well with experimental data, Fe0.6Ni0.7CrAl MPEAs exhibit ultra-high microhardness ( ∼ 700 HV) and unexpected wear resistance in the as-cast bulk state. Nano-indentation test showed that an average nano-hardness of 8.69 GPa and an average elastic modulus of 242.8 GPa were derived for the Fe0.6Ni0.7CrAl MPEAs. The hardening increment originates from the synergistic effect of disordered BCC phase and B2 phase. Moreover, Fe0.6Ni0.7CrAl MPEAs showed good wear resistances with friction coefficient of ∼ 0.41, as well as wear rate of 4 . 5 × 1 0 − 6 mm 3 / ( Nm ) . Abrasive wear and oxidative wear occurred during repetitive dry sliding, abrasive wear is the main wear mechanism. MD simulation explains the atomic structure evolution during the nanoscratch process, further revealing the wear mechanism at the atomic scale. This high-performance alloy shows potential for engineering applications.