CFD Simulation of Single- and Two-Phase Natural Convection in the Context of External Reactor Vessel Cooling

In recent decades, and with renewed importance after the events in Fukushima, the nuclear community has focused on the opportunity to rely on passive safety and, after plant shutdown, to ensure that reactor cooling can be safely guaranteed by natural processes, at least for a sufficient time before any active power intervention is needed. Buoyancy-driven flows and natural convection provide an efficient and potentially highly reliable and inexpensive heat transfer mechanism but, at the same time, these are rather complex flows because of the strong two-way coupling between the velocity and thermal fields, and the interaction between buoyancy and turbulence. In view of this, numerical tools able to predict these flows with accuracy and confidence are necessary to support the informed design and the safety assessment of present and future nuclear power plants. This paper is focused on research ongoing at the University of Leeds on the development of computational fluid dynamic tools to predict buoyancy-driven flows, of particular relevance to external reactor vessel cooling (ERVC). In ERVC, the aim is to retain the melted corium inside the reactor vessel, which is cooled from the outside by natural convection in the flooded reactor cavity where boiling is expected to occur on the outer vessel wall. The work starts by considering single-phase flow in a square buoyant cavity that is used to assess and improve the accuracy of available Reynolds-averaged Navier-Stokes (RANS) models, some of which, through the turbulence models embedded within them, are known to have shortcomings in predicting natural convection. In the same geometry, the superior accuracy of the large eddy simulation technique, the results from which may underpin further development of RANS approaches, is demonstrated. A two-fluid Eulerian-Eulerian model including boiling at the wall, which will be required to predict the whole ERVC phenomena, is also preliminary tested in the final part of the paper. Overall, encouraging results are found and weaknesses of the available modelling techniques and areas for future development are identified.