AI assisted design of high thermal conductivity Al-Fe-Ni-Mg alloys

Aluminum has a great thermal conductivity ranging from 200 to 240 W/(m·K) but impurities such as Fe and Si will degrade its conductivity significantly, leading to wrought aluminum usually having a thermal conductivity less than 150 W/(m·K) unsuitable for electronic industrial applications. To create a new wrought aluminum alloy with high thermal conductivity, experimental trail-and-errors usually have to be performed. In this study, artificial intelligence code has been used to design high thermal conductivity aluminum alloys. It has been found the new hypoeutectic Al-1.5Fe alloy can transform into eutectic network around primary α-Al and Al₉FeNi after 1.2 % Ni addition. The coarse network can be broken down and transform into a divorced eutectic after a further addition of 0.4 % Mg. We have achieved a thermal conductivity of 202.450 ± 4.052 W/(m·K) for the Al−1.5Fe alloy. From machine learning, mechanical properties can be predicted. Finally, to determine the contribution of free electrons in the heat transfer process, electrical conductivities are measured and a machine learning microstructure-electrical conductivity prediction model was developed and verified by key experiments, achieving a residual error from 0.26 % to 2.04 %.