Abstract: This comprehensive study investigates the complex phenomenon of pore migration in mixed oxide (MOX) nuclear fuels during reactor startup, a critical process that significantly impacts fuel performance and operational safety. The research focuses on addressing the numerical challenges associated with simulating these temperature gradient-driven microstructural changes, particularly the formation of central voids that affect thermal conductivity and mechanical integrity. Traditional simulation approaches using Galerkin finite element methods often fail to accurately capture these convection-dominated processes due to inherent numerical oscillations, leading to unreliable predictions of fuel behavior under operational conditions. To overcome these limitations, an advanced multiphysics modeling framework using the MOOSE simulation platform was developed, which uniquely integrates fully coupled thermal-stress-porosity interactions to provide a more comprehensive understanding of pore migration dynamics. The study systematically evaluates and compares four state-of-the-art numerical stabilization techniques: full upwind convection conservation (FUC), streamline upwind Petrov-Galerkin (SUPG), discontinuous Galerkin (DG), and upwind finite volume method (FVM). Each method was rigorously tested under various conditions to assess its effectiveness in handling the challenging convection-diffusion equation governing pore transport while maintaining computational efficiency. The framework’s validity was confirmed through extensive benchmarking against Idaho National Laboratory’s established Bison fuel performance code, demonstrating excellent agreement in predicting key phenomena such as radial porosity redistribution and the temporal evolution of central void formation. Detailed analysis reveals that SUPG emerges as the most robust approach, successfully suppressing numerical oscillations while preserving solution accuracy and consistency, even when using relatively coarse mesh resolutions. This method proves particularly effective in maintaining stability across different operating conditions and geometric configurations. In contrast, while FUC shows promise in oscillation reduction, it introduces undesirable numerical diffusion that compromised solution fidelity. DG methods, despite their theoretical advantages, demonstrates limited effectiveness for these specific convection-dominated problems, and FVM implementations face practical challenges related to variable mapping within the MOOSE framework. The research was further extended to multidimensional simulations, where SUPG maintaines its superior performance characteristics, offering an optimal balance between computational efficiency and solution accuracy. These findings have significant implications for industrial applications, particularly in the development of next-generation fuel performance analysis tools. The study provides concrete recommendations for implementing these numerical techniques in practical engineering contexts, with SUPG identified as the most suitable approach for large-scale, high-fidelity simulations of nuclear fuel behavior. The results contribute substantially to the ongoing efforts to enhance nuclear fuel safety, reliability, and performance prediction capabilities, while also establishing a foundation for future research in advanced multiphysics modeling of nuclear materials.