Magnesium (Mg) has a strong application potential as a lightweight metal. Yet, its absolute strength still needs improvement. In this work, we demonstrate that machine learning can be utilized to guide the development of high-strength Mg cast alloys. In the design framework, the composition and heat treatment condition are iteratively optimized by a surrogate model that is also evolving. After two iterations, a new alloy with the composition of Mg-10.0Al-2.0Sn-2.0Zn-0.1Ca-0.1Mn (at. pct) was identified. After aging at 200 °C for 96 hours, this alloy shows a Vickers hardness value of 110.5 Hv, which surpasses the highest value (102.5 Hv) in the initial dataset from literature. Finally, microstructure of the optimized alloy was characterized to understand the origin of its high hardness. This work demonstrates the potential of data-driven approaches for material development.

镁（Mg）作为轻质金属具有很强的应用潜力。但是，其绝对强度仍需要改进。在这项工作中，我们证明了机器学习可以用来指导高强度镁铸造合金的发展。在设计框架中，通过不断发展的替代模型迭代地优化了成分和热处理条件。经过两次迭代，鉴定出一种新的合金，其成分为Mg-10.0Al-2.0Sn-2.0Zn-0.1Ca-0.1Mn（at。pct）。在200°C下老化96小时后，该合金的维氏硬度值为110.5 Hv，超过了文献中初始数据集中的最高值（102.5 Hv）。最后，对经过优化的合金的微观结构进行了表征，以了解其高硬度的起源。