The potential danger of hydrogen was first identified after the Three Mile Island accident in 1989, where a large quantity of hydrogen was released into the containment and started to combust. Since then, numerous studies focusing on mitigating and reducing the potential risk of a hydrogen explosion have been conducted. The recent hydrogen explosions that occurred during the Fukushima Daiichi accident in March 2011 showed that the control and mitigation of risk of a hydrogen explosion are still key safety issues for nuclear power plants [1].

During a loss of coolant accident (LOCA), hydrogen gas may accumulate within the containment of a nuclear power plant. The hydrogen can be generated from (i) a metal-water reaction involving the zirconium fuel cladding and the reactor coolant, (ii) the radiolytic decomposition of water, which also produces oxygen, and (iii) the corrosion of the construction materials [2]. Hydrogen is then induced into the reactor coolant system, and gradually, the entire containment. Assuming the internal conditions of the containment, such as the quality of the steam and air present, a flammable gas mixture may combust, generating a chemical and thermal load with a potential threat to the integrity of the containment [3]-[5].

A catalytic reaction is widely used owing to its lower threshold temperature required for a spontaneous reaction compared to that of a non-catalyzed reaction. Passive autocatalytic recombiners (PARs) are currently implemented in many modern pressurized water reactors (PWRs) as an engineered safety feature for mitigating risk in the event of a core melt-down accompanied by significant releases of hydrogen gas into the reactor containment [6][7]. The catalyst materials are made of platinum and/or palladium, and recombine the hydrogen and oxygen gases into a water vapor upon contact with the surface of the catalyst. Hence, the heat produced during the recombination process creates a strong buoyancy effect, which increases the influx of surrounding gases into the inlet of the PAR [8].

Catalysts are generally developed in the shape of a plate or pellet. For example, PAR manufacturers such as AREVA and AECL utilize a plate-type catalyst, whereas NUKEM developed a specialized cartridge containing pellet-type catalysts. KNT developed a distinctive PAR model with enhanced hydrogen- removal capabilities. The new catalyst model adopts a larger surface area and the characteristics required to enhance the buoyancy-induced convective flow.

Korea Nuclear Technology (KNT) developed a PAR model with enhanced hydrogen removal capabilities. This new model adopts the shape of a honeycomb to create a greater catalyst surface area and an enhancement of the buoyancy-induced convective flow [9]. The KNT PAR is a stainless housing equipped with catalysts inside the lower part of the box. The design of the nuclear containment may cause some of the hydrogen to become trapped in the containment and unable to be reduced using mitigation equipment. Therefore, the location where the PAR is installed will indirectly affect its performance. Hence, a residual amount of hydrogen in a nuclear power plant will accumulate and become a potential risk for a future hydrogen explosion. To investigate and compare the differences in hydrogen reduction, this study proposes the PARs be installed at different locations within the nuclear containment.

This study involves the use of Computational Fluid Dynamics (CFD), which is a widely used computer-based tool for analysis and the design process. By utilizing the advances in computing power and graphics, the creation and analysis of a certain model is much less labor intensive and cheaper than applying experimental methods.

ANSYS CFX solves the unsteady Navier-Stokes equation in its conservation form. The instantaneous equation of mass (continuity) in a stationary frame is expressed as the following equation:

In addition, the instantaneous equation for momentum is expressed as follows:

These instantaneous equations are averaged for turbulent flows leading to additional terms that need to be solved. Whereas the Navier-Stokes equations describe both laminar and turbulent flows without additional terms, realistic flows involve length of scales much smaller than the smallest finite volume mesh. A direct numerical simulation of these flows requires significantly more computing power than what is available now or will be available in the near future.

Therefore, a significant amount of research has been conducted on predicting the effects of turbulence using turbulence models. Such models account for the effects of turbulence without the use of a very fine mesh or a direct numerical simulation.

These turbulence models modify the transport equations by adding in the averaged and fluctuating components. The transport equations have been modified into the following two equations.

The mass equation has not been modified, but the momentum equation contains extra terms, i.e., the Reynolds stresses, ρu⊗u¯, and the Reynolds fluxes, ρu⊗¯. These Reynolds stresses were previously modeled through additional equations to obtain closure. Obtaining closure implies applying equations to obtain closure, and that there is a sufficient number of equations to solve all of the unknowns including the Reynolds stresses and Reynolds fluxes.

Various turbulence models provide various ways to obtain closure. The model utilized in this investigation is the Shear Stress Transport (SST) model. The advantage of using this model is that it combines the advantages of other turbulence models (k-ε, Wilcox k-ω, and BSL k-ω).

The characteristic of the Wilcox model is the strong sensitivity to free-stream conditions. Therefore, a blending of the k- ω model near the surface and the k-ε model in the outer region was achieved by Menter, which resulted in the formulation of the BSL k-ω turbulence model. This model consists of a transformation of the k-ε into a k-ω formulation, subsequently adding in the resulting equations. The Wilcox model is multiplied by a blending function, F1, and transformed k-ε using another function, 1-F1. F1 is a function of the wall distance (a value of 1 near the surface and zero outside the boundary layer). The standard k-ε model is used outside and at the edge of the boundary layer.

However, whereas the BCL k-ω model combines the advantages of both the k-ε and Wilcox k-ω turbulence models, it fails to properly predict the onset and amount of flow separation from a smooth surface. The k-ε and Wilcox k-ω turbulence models do not account for the transport of turbulent shear stress resulting in an over-prediction of the eddy viscosity. A limiter on the formulation can be used to obtain the proper results. Such limiters are given in the following equation:

Where Vt=υtρ

F2 is a blending function that restricts the limiter to the wall boundary, and S is an invariant measure of the strain rate.

The blending functions are given through the following two equations:

where y is the distance to the nearest wall, and v is the kinematic viscosity. In addition,

The hydrogen recombination rate was concluded to be proportional to the distance to the hydrogen induction location. A PAR installed at a bottom location (nearer the hydrogen induction source) has a better hydrogen recombination rate compared to a PAR installed at a higher location, whereas a PAR installed at the center of the containment does not show a significant difference in the hydrogen recombination rate compared to a PAR installed at the side of the containment.