A dual-drive design strategy integrating knowledge and data for ternary precipitate-strengthened copper alloys

Abstract

Precipitate-strengthened copper alloys are widely used in lead frames and high-speed railway wires due to the enhanced strength and electrical conductivity conferred by nano-precipitates. However, the exploration of novel copper alloys only by data faces the dilemma of insufficient samples. Here we proposed a dual-drive design strategy that integrates knowledge and data for ternary precipitate-strengthened copper alloys. The knowledge from phase diagrams (PD) and open-source databases can be regarded as a dataset for machine learning (ML) models, and alloy candidates predicted by the model can be screened with maximum information using PD. The average compound formation energy, a key factor in the precipitation ability of alloying elements from the matrix, was screened out by the PD/ML dual-drive model. By exploring the composition space of 15 ternary alloys formed by the combination of 6 alloying elements and copper, we developed a Cu-0.44Ti-0.26Si alloy with tensile strength, electrical conductivity, and elongation of 741 MPa, 35.1 %IACS, and 14.0 %. This strategy can improve efficiency and accuracy in the design of precipitation-strengthened alloys with only ML or PD.