Comparative Analysis of Star-CCM, Star-CD, and CFX Codes for Free-Surface Flow Simulation of the Windowless XT-ADS Spallation Target Water Experiment

Accelerator driven systems for transmutation of nuclear waste require a coupling of an accelerator to a sub-critical reactor core by means of a spallation target. A windowless target is currently developed in the Integrated Project EUROTRANS of the European Community. In the target, HLM (heavy liquid metal) flows downwards through a concentric feeder surrounding the beam tube and forming a conical free surface which is subjected to a beam current. This results in a very high heat generation. The thermal-hydraulics of the target must ensure that the heat deposited into the HLM can be safely removed. Simulations using a volume of fluid method accounting for mass transfer across the free surface can predict the shape of the conical surface, flow detachment, and transient behavior. In the codes used in this study the mass transfer is accomplished via cavitation models, so that evaporation/condensation occurs at a prescribed vapor pressure. Previous simulations have lead to the design rules for the target. In particular, the velocity in feeder nozzle is limited to 2.5 m/s. Forced detachment by enlarging the guide tube stabilizes the free surface. Cavitation in the feeder nozzle is suppressed by fins which induce the required pressure loss. Recirculation flow in the thermally highly loaded zones is minimized by a small injection angle at the feeder nozzle outflow. The simulation results lead to the design of a water experiment, which precedes the test of a HLM target. In the paper the proposed target geometry of the water experiment is presented. Code-to-code comparisons show qualitative agreement between 3 different CFD codes. The considered codes, Star-CD, Star-CCM, and CFX, yield a stable free surface configuration with small recirculation, necessary for safe heat removal. However, differences are observable with respect to the sharpness of the free surface representation and also the behavior near the axis. Also strong differences in the required computational time exist between different codes. Various structured computational meshes are studied. In some simulations a fine mesh is employed within the feeder only and a substantially coarser grid is used in the free surface region and yet yields results comparable to an all-over fine-mesh simulation.