Analysis of different containment models for IRIS small break LOCA, using GOTHIC and RELAP5 codes

Advanced nuclear water reactors rely on containment behaviour in realization of some of their passive safety functions. Steam condensation on containment walls, where non-condensable gas effects are significant, is an important feature of the new passive containment concepts, like the AP600/1000 ones. In this work the international reactor innovative and secure (IRIS) was taken as reference, and the relevant condensation phenomena involved within its containment were investigated with different computational tools. In particular, IRIS containment response to a small break LOCA (SBLOCA) was calculated with GOTHIC and RELAP5 codes. A simplified model of IRIS containment drywell was implemented with RELAP5 according to a sliced approach, based on the two-pipe-with-junction concept, while it was addressed with GOTHIC using several modelling options, regarding both heat transfer correlations and volume and thermal structure nodalization. The influence on containment behaviour prediction was investigated in terms of drywell temperature and pressure response, heat transfer coefficient (HTC) and steam volume fraction distribution, and internal recirculating mass flow rate. The objective of the paper is to preliminarily compare the capability of the two codes in modelling of the same postulated accident, thus to check the results obtained with RELAP5, when applied in a situation not covered by its validation matrix (comprising SBLOCA and to some extent LBLOCA transients, but not explicitly the modelling of large dry containment volumes). The option to include or not droplets in fluid mass flow discharged to the containment was the most influencing parameter for GOTHIC simulations. Despite some drawbacks, due, e.g. to a marked overestimation of internal natural recirculation, RELAP5 confirmed its capability to satisfactorily model the basic processes in IRIS containment following SBLOCA. (C) 2010 Elsevier B.V. All rights reserved.