Machine learning and density functional theory assisted insights into the mechanical and oxidation properties of nickel-based superalloys

includes bibliographical references and index

PhD; 2023; Materials Research Centre

Due to global warming and increasing fuel costs, there is a constant thrust toward increasing fuel efficiency and reducing the emissions of gas-turbine engines, which are made out of superalloys. New superalloy materials, which possess higher temperature capabilities and better oxidation resistance, are desperately needed to address these issues. Using machine learning (ML) and density functional theory (DFT), this work attempts to provide an alternate route for gaining insights that could help develop new superalloys. We develop machine learning-based approaches to predict the mechanical properties of nickel-based superalloys, including yield strength, ultimate tensile strength, creep rupture life, and fatigue life. The ML models developed are highly accurate, and the analysis of the ML models reveals significant trends to optimize the mechanical properties. The ML models also successfully predict physical phenomena, such as the yield point anomaly, in excellent agreement with the physical models. In addition, ML modeling is performed to study the oxidation of nickel-based superalloys. Mass gain due to oxidation and parabolic oxidation rates are predicted using a combined supervised and unsupervised learning approach. The oxidation rates are optimized (minimized) as a function of composition and operational parameters via the genetic algorithm. The approach yielded a set of compositions with improved oxidation resistance while maintaining excellent mechanical properties. Furthermore, we establish structure-property linkages by introducing a new methodology to estimate the Vickers hardness by employing compositions and microstructures. Featurization through image processing and statistics lead to the development of highly accurate ML models for Vickers hardness. We identify essential elements and microstructural features through feature engineering to enhance the hardness. The creep and fatigue properties depend heavily on the interfacial characteristics of the superalloys. Further, we study the interface between nickel-based superalloys' γ and γ' phases. The effects of vacancy formation and doping with different elements are studied at the interface. The solid solution strengtheners are stable at both sides of the interface and increase strength, while corrosion-resisting elements are unstable at the interface and drastically decrease the interface strength. We also study the diffusion across the interface and find that rhenium has the highest barrier for diffusion and inhibits the diffusion of surrounding atoms. Finally, using DFT, we also design high entropy alloys having properties similar to nickel-based superalloys. The problem of the site occupancy of iron in the γ' phase is settled conclusively. The planar fault energies of pristine and Fe-doped γ′ phases are calculated using DFT, and the results are compared to the SEM micrographs to determine the dominant deformation mechanism at room temperature and high temperature. The results presented in this work improve the understanding of the underlying mechanisms of strengthening and could accelerate the development of novel superalloys with enhanced mechanical and corrosion properties.