Rapid design of high-end copper alloy processes combining orthogonal experiments, machine learning, and Pareto analysis
Peiwen Yun, Huadong Fu, Hongtao Zhang, Jingtai Sun, Menghe Zhao, Jianxin Xie
Abstract
To overcome the challenge of using data-driven methods for alloy design in extremely limited sample data, this study proposes a rapid alloy design strategy that integrates orthogonal experiments, machine learning, and Pareto analysis . The Cu-0.22Cr-0.23Sn-0.25Zn-0.025Si alloys (C2ZS2) developed in our previous research work is used as a research case for process optimization. A small dataset through 25 groups of orthogonal experiments was established and the support vector machine regression algorithm was used to construct a machine learning model for the relationship of aging parameters and properties. Furthermore, improved Pareto analysis was used to optimize aging parameters and accelerate alloy design efficiency. C2ZS2 alloys exhibited excellent comprehensive properties after optimized aging parameters (primary aging at 490 °C for 9 h and secondary aging at 420 °C for 4 h), achieving the combination of high strength and high electrical conductivity , with a tensile strength of (600 ± 2) MPa and an electrical conductivity of (75.1 ± 0.4) %IACS. Remarkably, it reached the dual upper limits of mechanical and electrical properties of comparable commercial EFTEC-64T-C alloys for high-end lead frame manufacturing, with tensile strength of 490–588 MPa and conductivity of 71 %IACS-75 %IACS.