DEVELOPMENT OF A DATA-DRIVEN TURBULENCE MODEL FOR THERMAL STRATIFICATION ANALYSIS IN REACTOR SYSTEM

Zijing LIU,Pengcheng ZHAO,Badea Aurelian Florin,Tao YU,Xu CHENG
DOI: https://doi.org/10.1299/jsmeicone.2023.30.1125
2023-01-01
The Proceedings of the International Conference on Nuclear Engineering (ICONE)
Abstract:Thermal stratification in large pools or enclosures is a very important phenomenon that is critical to nuclear reactor safety, including the cold- and hot-pool mixing in pool-type reactors, reactor cavity cooling system behavior in high-temperature gas-cooled reactors, passive containment cooling in advanced light-water reactors, and thermal stratification in boiling-water reactor suppression pools. It is very important to accurately predict pool temperature and density distributions for both design optimization and safety analyses of these reactor systems. Currently, effective thermal stratification modeling is a multi-scale method, that couples 1-D system-level code and 3-D CFD code, and can provide detailed thermal stratification information while providing system-level information in other domains. However, coupling 2 codes operating at different spatial and temporal scales remains a challenging issue, and high-resolution CFD simulation is computationally intensive. This study developed a data-driven coarse-mesh turbulence model for thermal stratification to improve the efficiency and accuracy of reactor safety analysis. The 3-D fluid conservation equations are modeled with a data-driven turbulence model based on local flow features on a much coarser scale so that the corresponding mesh setup in the CFD code can be consistent with the 1-D system-level code. A machine learning framework has been introduced for the acceleration of RANS by training an deep neural network with a large amount of fine-mesh CFD-generated data to predict steady-state turbulent eddy viscosity. We choose OpenFOAM as the simulation framework, TensorFlow as machine learning framework, and use TensorFlow C API to realize the coupling of OpenFOAM/TensorFlow and the invocation of the data-driven turbulence model. The coupling is introduced by way of a surrogate model prediction task for the turbulence eddy viscosity for the SUPERCAVNA experimental facility problem. This demonstrates the feasibility of the developed data-driven turbulence model.
What problem does this paper attempt to address?