Abstract Optimizing two conflicting properties such as mechanical strength and toughness or dielectric constant and breakdown strength of a material has always been a challenge. Here we propose a machine learning approach to dramatically enhancing the combined ultimate tensile strength (UTS) and electric conductivity (EC) of alloys by identifying a set of key features through correlation screening, recursive elimination and exhaustive screening of existing datasets. We demonstrate that the key features responsible for solid solution strengthened conductive Copper alloys are absolute electronegativity, core electron distance, and atomic radius, based on which, we discovered a series of new alloying elements that can significantly improve the combined UTS and EC. The predictions are then validated by experimentally fabricating four new Cu-In alloys which could potentially replace the more expensive Cu-Ag alloys currently used in railway wiring. We show that the same set of key features can be generally applicable to designing a broad range of conductive alloys.

中文翻译：

---

通过机器学习筛选显着增强合金的极限抗拉强度和电导率的组合

摘要 优化材料的机械强度和韧性或介电常数和击穿强度等两种相互冲突的特性一直是一个挑战。在这里，我们提出了一种机器学习方法，通过相关筛选、递归消除和对现有数据集的详尽筛选来识别一组关键特征，从而显着提高合金的综合极限拉伸强度 (UTS) 和电导率 (EC)。我们证明了固溶强化导电铜合金的关键特征是绝对电负性、核心电子距离和原子半径，在此基础上，我们发现了一系列可以显着提高 UTS 和 EC 组合的新合金元素。然后通过实验制造四种新的 Cu-In 合金来验证这些预测，这些合金可能会取代目前用于铁路布线的更昂贵的 Cu-Ag 合金。我们表明，同一组关键特征可普遍适用于设计广泛的导电合金。