Multi-scale modeling of Zircaloy cladding: From mechanistic crystal plasticity to finite element surrogates with scenario-based validation

In light water nuclear reactors, the reliability of fuel/clad assemblies largely depends on the degree to which mechanical interactions between the fuel and cladding materials can be quantified. This work introduces a multi-scale modeling framework for elasto-viscoplasticity and creep in Zircaloy-4 cladding that incorporates out-of-sample validation and explicitly accounts for microstructural evolution, including dislocation density, to improve rod-scale fuel–cladding interaction predictions. Specifically, this work focuses on deriving a surrogate model from correlations in a polycrystal creep simulation database, and integrating the surrogate into a finite element solver for simulations at the component length-scale. The polycrystal simulations employ a mechanistic model that describes the viscoplastic strain rate in terms of contributions from dislocation motion and vacancy-mediated diffusion processes, and includes a dislocation density evolution model. A synthetic database is meticulously constructed by systematically varying relevant input variables in polycrystal creep simulations that influence the mechanical response outputs. A response surface surrogate is trained on the database to capture these critical input–output correlations. The surrogate is then integrated into a finite element modeling framework and validated against industry-standard, experimentally established accident-scenario test cases at the component scale, ensuring robust cross-validation representative of real operational conditions.