Simulation of Buoyancy-Driven Flow for Various Power Levels at the NSTF

Abstract

The Reactor Cavity Cooling System (RCCS) is an important passive safety system that is being incorporated in a number of high temperature reactor design concepts. The Natural convection Shutdown heat removal Test Facility (NSTF), located at Argonne National Laboratory, is an experiment with the objective of investigating the flow and thermal behavior of a particular air-cooled RCCS design. It consists of 12 ducts surrounded by a cavity with a heated wall, through which air flows via natural convection before exiting through two chimneys. The NSTF is a ½-scale facility, and is well instrumented in order to provide data for code validation, including Computational Fluid Dynamics (CFD)-grade data in a number of locations. Instrumentation includes fiber-optic Distributed Temperature Sensors (DTS) throughout one of the riser ducts and in the upper plenum. In conjunction with the experimental tests, CFD simulations were performed to support the design and optimization of these natural convection systems. The CFD simulations were performed using the “as-tested” geometry of the NSTF. All CFD simulations were steady-state. Both a full natural convection model and a smaller forced primary flow model were tested. The influence of boundary conditions, notably at the cavity walls, was tested. Initial simulations assumed adiabatic walls but these were later adapted to simulate heat losses, aided by thermal images taken of the exterior NSTF surfaces during testing. Simulations were run for tests at two different power levels. A number of turbulence models were compared to test their influence. Simulation results were compared with experimental data. Convergence was generally good for both models. It was found that the natural convection model was indeed beneficial for correctly estimating local temperatures in a number of areas, particularly near the top of the riser ducts and from DTS measurements along the flow path. Flow in the heated cavity was complex. In general, the experimental trends were predicted well by CFD, although magnitudes could be improved in some areas. The turbulence models tested had a relatively small effect on the shape of the temperature profile in the ducts and on heated surface temperatures. Results from the simulations have been of direct use in improving test procedures and choosing locations for more accurate instrumentation. In future work, full natural convection simulations of more tests will be performed. After this has been completed, best practices can be established for accurately simulating these general types of natural convection systems across a wide range of operating conditions.