Uncertainty Analysis Method of Physical-thermal Coupling Based on preCICE

In nuclear reactor systems, precise core modeling and uncertainty assessment are of critical importance for enhancing reactor safety, optimizing design margins, and advancing the development of next-generation reactor technologies. Existing multi-physics coupling platforms exhibit notable limitations in flexibility, scalability, and interoperability, particularly when addressing weakly coupled problems where computational efficiency and extensibility are often compromised. Therefore, this study aims to develop a modular, high-fidelity, and highly scalable multi-physics coupling platform with integrated uncertainty quantification (UQ) capabilities, thereby supporting comprehensive simulations and safety assessments under both steady-state and transient-state reactor conditions. To achieve this objective, the neutron diffusion code KOMODO, the subchannel thermal-hydraulic code SUBCHANFLOW, and the uncertainty quantification tool DAKOTA were integrated into a unified multi-physics coupling and UQ platform based on the open-source preCICE coupling library. To enable effective data exchange, secondary development was conducted on KOMODO and SUBCHANFLOW. New variables and modules were implemented to support data interaction through preCICE. The computational grid in KOMODO was restructured, reflective layer components were filtered, and discrepancies between heterogeneous mesh nodes in the physics solvers were corrected. The steady-state and transient-state calculation modules were decoupled to allow multi-round iterative preconditioning, and the computational architecture was adapted to align with preCICE control modes. OpenMPI was used to optimize communication channels, allowing dynamic switching between dual-configuration file modes. The platform was validated against the NEACRP-L-335 benchmark by simulating multiple rod-ejection transient scenarios under both hot full power and hot zero power conditions. For uncertainty propagation, a quasi-Monte Carlo (QMC) sampling method based on Sobol sequences was employed, combined with orthogonal least-interpolation generalized polynomial chaos expansion (OLI-gPCE) to model and propagate input parameter uncertainties. Output quantities, including peak power factor, final power factor, fuel Doppler temperature, and coolant outlet temperature, exhibit near-normal distributions with narrow 95% confidence intervals, indicating high accuracy and robustness. Sensitivity analysis identifies total core power, coolant inlet temperature, inlet flow rate, and cladding thickness as the dominant contributors to output variability. Furthermore, Sobol variance decomposition reveals significant higher-order nonlinear interaction effects among these key parameters. In conclusion, the proposed platform demonstrates a robust modular architecture, compatibility with both parallel and serial execution modes, strong numerical stability, and support for heterogeneous grid configurations. It shows excellent applicability for nuclear reactor safety analysis and advanced reactor design evaluation. The study confirms the feasibility and effectiveness of weakly coupled multi-physics simulation and uncertainty analysis based on the preCICE framework in the field of nuclear engineering and holds practical value for enhancing the capabilities of domestic simulation platforms.