Page 20 - Read Online

Page 14 of 21 Zhou et al. J Mater Inf 2022;2:18 https://dx.doi.org/10.20517/jmi.2022.27

[115]
Figure 13. Diagram of ML-based alloy design system for HEAs with desired hardness .

[114-116]
In addition to phase prediction, ML is widely used to develop new alloys with the desired properties .
[114]
Wen et al. proposed a property-orientated materials design strategy to design HEAs with high hardness .
ML was used to build a map between hardness and descriptors, such as chemical compositions and
physicochemical properties of elements. The utility functions were employed to guide the search for
Al-Co-Cr-Cu-Fe-Ni HEAs with high hardness from nearly two million possible compositions. This method
developed several HEAs with hardness 10% higher than the best value in the original training dataset via
[114]
only seven experiments . Yang et al. further proposed a new approach called ML-based alloy design
[115]
system (MADS) for developing HEAs with the desired hardness . Figure 13 summarizes the overall
framework of MADS, including database establishment, model construction, composition optimization and
experimental validation. Through a four-step feature selection method, five key features affecting the
hardness of HEAs were screened out, including the average deviation of the column, the average deviation
of the specific volume, the average deviation of the atomic weight, the valance electron concentration (VEC)
and the mean melting point. Among these features, VEC is the most significant variable, which positively
impacts the hardness of HEAs when it is less than 7.5. Through this prediction system, one optimized
sample (Co Cr Fe Ni V , at. %) exhibits superior hardness, which is 24.8% higher than the highest
18 7 35 5 35
hardness in the original dataset.

In contrast, complex metallurgical behavior occurs within the micro-molten pool during L-PBF, which is
greatly affected by the multiple processing parameters, such as laser power, scanning speed, hatch distance,
layer thickness and scanning strategy. It is challenging to obtain suitable processing parameters through
experimental methods due to the numerous combinations of these parameters. Defects, such as porosity,
[117]
incomplete fusion holes and cracks, are inevitable if the processing parameters are unsuitable . It is likely
that there are mature AM parameter packets provided by commercial 3D printing companies for
conventional alloys, such as steel, Al alloys, Ti alloys and Ni-based superalloys. However, it is still quite
difficult to obtain optimal AM parameters for newly designed alloys, such as HEAs, from experimental