Benchmark exercise TH-27 on natural convection with steam injection and condensation inside the extended THAI facility

Double-blind, blind and open simulations were performed by thirteen participants using lumped parameter (LP) models (ASTEC, COCOSYS, MELCOR) as well as CFD codes, namely FLUENT, CFX, GASFLOW and GOTHIC. The challenge of the double-blind benchmark was to generate a model and simulate a long lasting transient without previous model calibration based on knowledge gained from earlier tests being performed in the facility. The experimental measurements allow quantifying transport mechanisms and flow conditions between the two vessels as well as inhomogeneities of the gas mixtures of the vessels. Overall, the TH-27 benchmark demonstrated the high prediction quality of LP and CFD codes provided that dedicated nodalization guidelines are considered. On the other hand the benchmark revealed noticeable user influence for LP codes, especially due to different nodalization schemes and partial lack of special treatment regarding plumes or upward directed jets.