Crystal Plasticity Simulations and Machine Learning‑Enhanced Modelling of γ/γ′ Sub‑Grain Structure in Ni‑Based Superalloys
Please login to view abstract download link
Nickel-based superalloys play a key role in critical aerospace and energy-sector components because of their excellent mechanical properties, creep resistance, and oxidation stability at elevated temperatures. For most superalloys, their mechanical performance is strongly influenced by the interaction between the gamma (γ) matrix and gamma‑prime (γ′) precipitates at the sub‑grain scale, which governs key deformation mechanisms such as dislocation activity, anti-phase boundary (APB) shearing, Orowan looping, and interfacial strengthening. Although modelling the γ/γ′ interaction and morphology is crucial for resolving these mechanisms, it is computationally prohibitive to do so across an entire polycrystalline aggregate at this scale. This highlights the need to adopt a multiscale framework or develop surrogate models to bridge the gap between mesoscale fidelity and macroscale applicability. In this work, the γ/γ′ microstructure and the associated precipitate‑level deformation mechanisms are explicitly modelled using a crystal‑plasticity finite‑element method (CPFEM). The model captures key physical phenomena such as APB‑mediated γ′ shearing, matrix–precipitate dislocation interactions, and interfacial strengthening mechanisms that govern the sub‑grain‑scale response of γ′‑strengthened superalloys. The numerical predictions are compared and validated against targeted experimental data to ensure physical fidelity at the mesoscale. The insights obtained from these fully γ/γ′‑resolved simulations provide a foundation for future multiscale developments, such as informing grain‑scale constitutive models or generating high‑quality datasets suitable for machine-learning‑based surrogate modelling. Such extensions would enable efficient simulations at the polycrystalline scale while preserving essential γ/γ′‑scale physics, ultimately leading to more accurate and computationally feasible polycrystal‑level simulations.
