Blind benchmark predictions of the NACOK air ingress tests using the CFD code FLUENT

Abstract

The JAERI and NACOK experiments examine the combined effects of natural convection during an air ingress event: diffusion, onset of natural circulation, graphite oxidation and multicomponent chemical reactions. MIT has been benchmarking JAERI tests using the FLUENT code for approximately three years [1]. This work demonstrated that the three fundamental physical phenomena of diffusion, natural circulation and then chemical reactions can be effectively modeled using computational fluid dynamics. The latest series of tests conducted at the NACOK facility were two graphite corrosion experiments: The first test consisted of an open chimney configuration heated to 650C with a pebble bed zones of graphite pebbles and graphite reflectors. The second test is a similar test with a cold leg adjacent to the hot channel with an open return duct below the hot channel. Natural circulation, diffusion and graphite corrosion were studied for both tests. Using and adapting previous computational methods, the FLUENT code is used to blind benchmark these experiments. The objective is to assess the adequacy of the modeling method used in this blind bench-marking analysis by comparing these blind test predictions to the actual data and then modify the model to improve predictive capability. Ultimately, the objective is to develop a benchmarked analysis capability that can be used for real reactors calculations, and to improve the understanding of the physical phenomena taking place during an air ingress event. This thesis presents the modeling process of these experiments, the blind model results and the comparison of the blind computed data with experimental data. Sensitivity studies provide a good understanding of the different phenomena occurring during an air ingress event. The blind benchmarking demonstrates the ability of FLUENT to model satisfactorily in full scale the NACOK air ingress experiment. The blind models are then improved to successfully model air ingress events. An important finding of this work is that there is great variability in graphite corrosion data and that good qualification of specific graphite used is vital to predicting the effects of an air ingress event.