Abstract: Synchronously improving the ultimate tensile strength (UTS) and elongation (EL) of Al-Cu-Mg alloys at both room and high temperatures (dual-synchronous property improvement) is an important approach to meet the lightweight development requirements of aviation high-end equipment. However, it is very challenging because of the complicated composition of aluminum alloys and distinct strengthening phases at different temperatures. Here we propose a machine learning strategy that enables dual-synchronous improvement of UTS and EL at both room and high temperatures. First, feature selection is used to identify the key alloy factors affecting the UTS and EL of the alloy. Then, machine learning prediction models are developed with key alloy factors and temperature as inputs, while UTS or EL are outputs, respectively. Finally, a utility function of Bayesian optimization is constructed for simultaneously considering four target properties, including UTS and EL at room temperature (RT) and 200 °C, and dual-synchronous improvement is realized based on a genetic algorithm. The designed new alloy has a composition of Al-4.3Cu-1.7Mg-0.5Mn-0.5Zn-0.2Zr-0.2Cr-0.1Ti (wt.%). At RT, the measured UTS, yield strength (YS), and EL of the new alloy C1 samples are 577±4 MPa, 411±4 MPa, and 18.0 %±0.5 %, respectively. At 200 °C, the UTS, YS, and EL of the new alloy C1 samples are 458±7 MPa, 364±8 MPa, and 19.3 %±0.9 %, respectively. Compared to the currently leading AA2524 alloy, the UTS of the new alloy at RT and 200 °C is increased by 19 % and 24 %, the YS is increased by 7 % and 25 %, and the EL is improved by >1/4 and 1/5, respectively. In the new alloy C1, nanoscale S phases with a volume fraction of 4.3 % and the multiple twinned T phases with a volume fraction of 3.1 % result in excellent strength and elongation at both room and high temperatures.

Key words: Al-Cu-Mg alloy, T phase, High-temperature strength, Machine learning, Feature screening