Design of Cu–Ni–Si alloy with ultra-high strength and high conductivity based on machine learning model

The low efficiency of multi-element optimization with the coupling of mechanical and electrical properties has been one of the major challenges in traditional trial-and-error methods of Cu alloys. Here, machine learning techniques are employed to design Cu alloys with significantly enhanced overall performance. First, based on the nine key alloy factors collected-including atomic radius difference, electronegativity difference, valence electron concentration, shear modulus difference, shear modulus variance, second ionization energy variance, electron ratio, the sixth power of the work function, and mixed entropy correlation analysis and exhaustive screening are conducted to precisely identify the core factors influencing strength and conductivity. Next, an artificial neural network was used to determine the optimal combination of composition, process, and key factors, and a gradient boosting regression algorithm was employed to construct the optimal prediction models for tensile strength and electrical conductivity, with RMSE values of 60.78 MPa and 5.51% IACS, respectively. After multiple rounds of prediction and experimental validation, designed Cu-1.86Ni-1.1Co-0.62Si-0.12Mg-1.5Ag alloy was achieving a tensile strength of 886.45 MPa and an electrical conductivity of 45.34 %IACS, Compared with Cu–Ni–Si and Cu–Ni–Si–Co–Mg alloys, the synergistic improvement of conductivity and strength is achieved.