Design of a novel Cu-Cr-X alloy with high strength and high electrical conductivity based on mechanical learning

Abstract

Micro-alloying and thermo-mechanical treatments are crucial to the further development of high-strength and high-conductivity Cu-Cr-X alloys. In this study, a high accuracy ultimate tensile strength and electrical conductivity prediction model was obtained by training the BP neural networks with different compositions and processing experimental data. The Cu-Cr-Zr-Mg-Ti alloy with superior properties was optimally designed by genetic algorithm from the massive solutions, which the experimental tensile strength and conductivity reached 668 MPa and 71.5 %IACS, respectively. The atom probe tomography results show that Zr, Mg, and Ti simultaneously segregated in the Cr-rich phase after aging at 440 °C for 480 h, which significantly improves the stability of the precipitated Cr phase and inhibited the Cr-rich phase transition from fcc to bcc structure. With the increase of aging temperature, the bcc structure Cr phase gradually replaces the fine fcc-Cr phase and exhibited a higher coarsening rate. It was found that the addition of Zr tends to nucleation at the interface of the K-S orientation relationship (OR) bcc-Cr phases. The Cu5Zr phase maintains the (1¯13)Cu5Zr//(01¯1)Cr and [011]Cu5Zr//[111]Cr ORs with the bcc structure Cr precipitates. The evolution of microstructure and properties exhibited a narrow aging process region of 80 % cold-rolled Cu-Cr-Zr-Mg-Ti alloy, while the consistent results from prediction model provide a favorable guidance for process designs. The outstanding performance enhancement and rapid prediction of process scopes leads us to recognise that the great potential of multi-objective machine-learning-aid design in the complex composition and process problems.

Keywords

• Machine learning
• Cu-Cr-X alloy
• Thermo-mechanical treatment
• Tensile strength
• Electrical conductivity