A novel approach for full-core mesh generation to enable high-fidelity thermal-hydraulic simulation of nuclear reactor engineering

Thermal-hydraulic analysis is crucial in reactor engineering. High-fidelity simulations, utilizing advanced computing techniques and supercomputing resources, are highly regarded. High-quality fluid mesh models are essential for complex reactors’ high-fidelity simulations. Using existing tools for model construction has limitations in quality control, performance, user dependency, file generation, and visualization. Estimating time and memory consumption for full-core meshing is also not possible. A R-IMG approach is designed, it effortlessly creates mesh models for intricate flow field, demonstrating exceptional modeling performance, robustness, scalability, and reduced user dependency, while its flexible file manner effectively addresses challenges in generating and visualizing large-scale mesh files. Extensive testing validates R-IMG’s effectiveness and reliability in meshing the reactor’s flow field. It efficiently generates high-quality meshes for the complex flow field in the entire fuel region of CEFR, completing the process within 7 h and 10GB of memory. The resulting model has around 14 billion cells and an average quality of 0.7. R-IMG achieves a maximum parallel scale of 3200 processes for file generation, with approximately 90% parallel efficiency. These results demonstrate that R-IMG outperforms existing tools in core meshing and shows significant potential for full-core meshing. Successful visualization of models and benchmark tests provide evidence for models’ correctness. •A novel approach for meshing the entire reactor core is designed for high-fidelity thermal-hydraulic simulations.•A geometric segmentation method is designed for the complex flow field between reactor core assemblies.•A set of high-quality modeling techniques is developed, including mesh refinement, optimization, and stitching.•A hierarchical numbering scheme for components, geometric input formats, and large-scale parallel meshing strategies are designed.•R-IMG constructs models with over 10 billion mesh cells in 7 h and 10 GB, passing benchmark tests.