Homogenization-Based Softening Model for Fragmented Nuclear Fuel

Crack formation plays an important role in the behavior of oxide nuclear fuel. A conventional approach to account for crack-induced stress relief is the use of isotropic softening models, which explicitly reduce the elastic constants of the fuel. However, these models often fail to capture the strength recovery that occurs when pellet fragments are compressed together upon contact with the cladding. The development of high-fidelity 3D nuclear fuel performance codes now provides the potential to explicitly model the cracked fuel and achieve a more accurate assessment of its stress state. In this study, a more mechanistic softening model has been developed using the open-source high-fidelity fuel performance code OFFBEAT. The softening factors for the bulk and shear moduli were derived based on the hydrostatic and deviatoric energy states of an idealized elastic nuclear fuel fragment subjected to static thermo-mechanical loading conditions. The model is capable of capturing strength recovery under compression and is formulated as an empirical correlation of the linear heat rate, the average fuel surface temperature, and the average pressure applied at fuel outer radius. It accounts for the global effect of radial cracks on the fuel stress state without resolving local cracking behavior, offering high computational efficiency and suitability for axisymmetric simulations. Its simplicity makes it a promising candidate for integration into traditional 1D engineering fuel performance codes to improve stress assessment under pellet-cladding mechanical interaction. The application of the new model is demonstrated in OFFBEAT through axisymmetric 1D and discrete-pellet 2D simulations, including a hypothetical irradiation case and the K11-5 experimental ramp test. The results reveal the effect of strength recovery on the cladding displacements and stresses, as well as on the ridge formation under PCMI, showing improvements over conventional isotropic cracking models.