基于机器学习元素特征量分析的析出强化铜合金的理性设计

Rational Design of Precipitation-Strengthened Copper Alloys via Machine Learning Analysis of Elemental Feature Quantities

  • 摘要: 高端制造用析出强化型铜合金的力学和导电性能相互制约,综合性能提升一直是一个重大挑战。本文采用机器学习方法进行元素特征量筛选,挖掘影响合金性能的关键物理化学特征,实现多元复杂合金的高性能设计。结合相关性筛选、递归消除和穷举法筛选,筛选得到影响时效析出强化型铜合金硬度的5个关键合金因子和影响导电率的5个关键合金因子,以关键合金因子为输入,分别构建了误差小于6 %的硬度预测模型和误差小于5 %的导电率预测模型。应用预测模型,设计了新型合金Cu-2.92Ni-0.92Co-0.74Si。参照Cu-Ni-Co-Si系合金的工业化生产流程和条件进行实验验证,新合金的抗拉强度和导电率分别达到868 MPa和45.6 %IACS,实现了相互制约的合金力电性能的同步提升。

     

    Abstract: The inherently interdependent mechanical and electrical properties of precipitation-strengthening copper alloys for advanced manufacturing, make the simultaneous enhancement of these properties a significant challenge. This study using machine learning techniques to perform elemental feature selection, uncovers the key physicochemical characteristics that govern alloy performance, thereby enabling the high-performance design of multicomponent and complex alloys. Integrating correlation screening, recursive elimination, and exhaustive search methods, the study systematically identifies five key alloy factors influencing hardness and five key alloy factors affecting electrical conductivity (EC) in precipitation-strengthened copper alloys. The identified five key alloy factors affecting strength primarily influence hardness by affecting the outcomes of solid solution strengthening and precipitation strengthening. In contrast, the key alloy factors influencing EC regulate free electron density, adjust the mean free path of electron migration, and affect electron scattering, thereby influencing the EC of copper alloys. Utilizing these key alloy factors as inputs, support vector regression (SVR) models for predicting hardness and EC are developed. Through grid search optimization, the hardness prediction model achieved an error of less than 6%, while the EC model achieved an error of less than 5%. With the candidate compositions are reduced to a limited number of elements in consideration of sustainable development and large-scale industrial production costs, and limit on the content of the expensive element cobalt to below 1 wt.% imposed, the candidate compositions input into the prediction models, the alloys with optimal overall performance are selected. As a result, a novel alloy Cu-2.92Ni-0.92Co-0.74Si is designed. Experimental validation carried out under the industrial production processes for Cu-Ni-Co-Si alloys demonstrated that the newly developed alloy achieved an ultimate tensile strength of 868 MPa and an EC of 45.6% IACS, representing a synergistic enhancement of the typically competing mechanical and electrical properties observed in conventional alloys. Characterization results indicate that the alloy has an average grain size of 10.4 μm. The primary reason for the excellent mechanical and electrical properties of the newly developed alloy is the presence of a high density of fine, uniformly dispersed precipitates, which provide a precipitation strengthening effect while depleting solute atoms in the matrix, thereby enhancing EC. The average diameter of the precipitates in the alloy is 9.84 nm. Additionally, work hardening and grain boundary strengthening are also crucial factors contributing to the improvement of the alloy's mechanical properties. This study by machine learning techniques through predictive modeling and experimental validation to optimize the composition of precipitation-strengthened copper alloys, achieves simultaneous improvements in mechanical strength and EC.

     

/

返回文章
返回