Machine Learning Assisted Design Approach for Developing γ′-Strengthened Co-Ni-Base Superalloys

As a new class of promising high-temperature materials, Co–Al–W-base alloys have been developed by alloying additions to improve the microstructure stability and other properties. However, the optimization of Co–Al–W-base alloys becomes more complicated with increasing variety and content of alloying elements. In this study, an accelerated approach to design γ′-strengthened Co–Ni-base superalloys with well-balanced properties was developed, by integrating the diffusion-multiple approach and machine-learning tools. A large amount of experimental data was obtained using the diffusion-multiple approach and fed into machine learning tools to establish the relationship between alloy compositions and important thermodynamic and microstructural parameters such as the phase constituent, the γ′ phase fraction (Fγ′) and the γ′ solvus temperature (Tγ′). The established machine-learning models were then employed to predict the characteristic parameters of multicomponent Co-Ni-base superalloys containing up to nine elements (Co, Ni, Al, W, Ta, Ti, Cr, Mo, Nb), even though most of the collected compositions from experiments were quinary to septenary alloys. Using the predicted results from the models and the computational thermodynamics tools, a multicomponent Co–Ni-base superalloy aimed at the application as single crystal blades was designed and characterized to test the reliability and robustness of the novel design approach.