CFD Prediction of Natural Convection During External Reactor Vessel Cooling

Eulerian-Eulerian computational fluid dynamic models are used in the prediction of multiphase gas-liquid flows in nuclear reactor thermal hydraulics and in many other chemical and process engineering applications. The modelling approach, based on the concept of interpenetrating continua, allows the calculation of complex and large-scale industrial flows with a relatively limited computational load. However, interfacial transfer processes need to be entirely modelled through numerous closure relations. A large number of different optimized closure sets are available, each often showing remarkable accuracy, but generally only over a few experimental data sets. This specificity makes it difficult to compare the overall accuracy of the models and obstructs the development of more general and robust approaches. In this paper, the bubbly flow models developed at the University of Leeds and the Helmholtz-Zentrum Dresden - Rossendorf are benchmarked against relevant experiments. These two research groups follow a similar modelling approach, aimed at identifying a single universal set of widely applicable closures. The models, implemented respectively in Star-CCM+ and CFX, are applied to a large selection of bubbly flows in different geometries. The main focus is on the momentum transfer, mainly responsible for the lateral bubble distribution in any flow, and on turbulence closures. Therefore, monodispersed bubbly flows that can be effectively characterized with a single average bubble diameter are selected. Overall, the models are found to be generally reliable and robust, and additional developments towards further improved accuracy, increased generality and the definition of a common unified set of model closures are identified. In future, additional benchmark exercises of this kind will be performed, and potentially the definition of proven sets of reference experiments will be recommended.