Prediction of alloy composition and microhardness by random forest in maraging stainless steels based on a cluster formula

Fe–Ni–Cr-based super-high-strength maraging stainless steels were generally realized by multiple-element alloying under a given heat treatment processing. A series of alloy compositions were designed with a uniform cluster formula of [Ni₁₆Fe₁₉₂](Cr₃₂(Ni_{16–x–y–z–m–n}Mo_xTi_yNb_zAl_mV_n)) (at.%) that was developed out of a unique alloy design tool, a cluster-plus-glue-atom model. Alloy rods with a diameter of 6 mm were prepared by copper-mold suction-cast processing under the argon atmosphere. These alloy samples were solid-solutioned at 1273 K for 1 h, followed by water-quenching, and then aged at 783 K for 3 h. The effect of the valence electron concentration, characterized with the number of valence electrons per unit cluster (VE/uc) formula of 16 atoms, on microhardness of these designed maraging stainless steels at both solid-solutioned and aged states was investigated. The relationship between alloy compositions and microhardness in maraging stainless steels was firstly established by the random forest (RF, a kind of machine learning methods) based on the experimental results. It was found that not only the microhardness of any given composition alloy within the frame of cluster formula, but also the alloy composition with a maximum microhardness for any given VE/uc, could be predicted in good agreement with the guidance of the relationship by RF. The contributions of minor-alloying elements to the microhardness of the aged alloys were also discussed.