机器学习辅助高性能银合金电接触材料的快速发现

Machine Learning Aided Rapid Discovery of High Performance Silver Alloy Electrical Contact Materials

为了快速发现高性能银合金电接触材料,从文献中收集了32组铸造法制备的银合金电接触材料的成分和性能数据,采用特征量筛选方法识别出影响合金性能的关键合金因子,采用支持向量机算法建立了合金导电率和硬度预测模型,实现了合金成分的快速设计。选取预测性能优异的Ag-19.53Cu-1.36Ni、Ag-10.20Cu-0.20Ni-0.05Ce和Ag-11.43Cu-0.66Ni-0.05Ce (质量分数,%) 3种成分设计方案进行工业生产条件的实验验证,性能预测结果与实验结果误差均小于10%,3种合金导电率均≥79%IACS,Vickers硬度均≥87 HV,综合性能均优于已有铸造法制备的银合金电接触材料。上述研究结果表明,本工作建立的机器学习成分设计方法可靠性好,有助于提高合金成分设计效率,快速发现综合性能优异的银合金电接触材料。

Thirty-two groups of data of composition and performance of silver alloy electrical contact materials prepared via casting were collected from the literature to quickly find high-performance silver alloy electrical contact materials.The key alloy factors affecting the alloy properties were identified using the feature selection method.The prediction model of alloy electrical conductivity and hardness was established using a support vector machine(SVM) algorithm,which achieved the rapid design of alloy composition.Three composition designs of Ag-19.53Cu-1.36Ni,Ag-10.20Cu-0.20Ni-0.05Ce,and Ag-11.43Cu-0.66Ni-0.05Ce(mass fraction,%) with excellent predictive performance were selected for experimental validation under industrial production conditions.The error between the performance prediction and experimental results is less than 10%,the electrical conductivity of the three alloys designed is greater than 79%IACS,and the Vickers hardness is greater than 87 HV.Both the electrical conductivity and hardness are better than those of previous silver alloy electrical contact materials prepared via casting.The above results show that the machine learning composition design method established in this study has good reliability,helps improve the efficiency of alloy composition design,and quickly finds silver alloy electrical contact materials with excellent comprehensive properties.