A nuclear waste processing facility under development at the Savannah River Site is designed to remove radioactive Cesium 137 from solution by adding tetraphenylborate (TPB) in order to produce cesium tetraphenylborate, which precipitates out of the solution. The precipitate slurry is filtered and sent to another plant where it is vitrified. The remaining liquid has relatively low radioactivity so it can be processed using less expensive methods.
The TPB added to the tank is subject to radiolytic, thermal, and chemical degradation. One of the products of this degradation is benzene. In order to eliminate the slightest possibility of a fire in the tank, both oxygen and fuel control are exercised. Oxygen control is exercised via a nitrogen purge. The tank is operated at a vacuum to avoid radioactive discharge. At the same time, it must be demonstrated that the vapor space benzene concentrators do not reach the concentration required to support combustion at any location.
The overall interior dimensions of the tank are 85 feet in diameter and 33 feet high. The liquid depth at the end of a precipitation cycle is approximately 4 feet, so there is 29 feet of vapor space. The tank contains a center column 6.75 feet in diameter at the base. The top of the column consists of a conical section with horizontal and vertical sides 12 feet in length. The tank contains a 12 inch diameter outlet located flush with the top and oriented 225 degrees measured counterclockwise from north. Ventilation flow is supplied by a nitrogen purge inlet located in the C3 riser, 36 ft. radial and 340 degrees counterclockwise from north. The 1-inch diameter nitrogen purge inlet pipe orientation is 15 degrees downward below the top surface, tangential to the tank radius and pointing in a counterclockwise direction. The tank contains cooling coil loops that consist of a large number of 2 inch pipes spaced approximately 3 ft. apart and oriented in a vertical direction.
The tank was modeled with the FLUENT CFD code from Fluent Incorporated, Lebanon, New Hampshire. Westinghouse engineers selected this software package because it dramatically reduces the amount of time required to model complicated geometrics such as this tank. FLUENT is a state-of-the-art computer program that uses pressure-based finite volume methods to solve fluid flow and energy equations. FLUENT utilizes an unstructured grid to provide truly automatic mesh generation. This means that once the surface is defined, the grid is generated with little or no user intervention. FLUENT also has a solution-adaptive grid capability, which allows the user to refine the grid, after intermediate solutions, in regions with large gradients in order to achieve a higher level of accuracy. Fluent is the world’s largest CFD software provider with over 1000 users.
The unstructured meshing approach made it quite simple to model the tank despite its complex geometry and large variations in length scale – an 85 foot tank diameter vs. a 1 inch diameter inlet nozzle. The unstructured approach facilitates the construction of finite volumes in the computational domain because it provides an easy transition from a subdomain where a fine mesh is required to a subdomain where a coarser mesh is used. The tank cooling coils were not explicitly modeled, however, frictional losses to the flow field due to their presence were included as a momentum sink. The automatically generated mesh consisted of approximately 17,000 finite volumes. This mesh was later refined to accommodate changes in the location of the benzene source and to support sensitivity studies.
CFD image showing contours of C6H6 Mole Fraction at 2 minutes.
Nitrogen is introduced into the tank from upstream ambient vaporizers. The nitrogen flow rate was specified as a pressure inlet value that provided the correct mass flow rate. The ventilation outlet from the tank was treated as an atmospheric pressure boundary condition. This treatment results in a small positive pressure in the tank. The actual tank has a small negative pressure to prevent leakage, however, the small positive pressure has no effect on the calculated dynamics of the flow field in the tank. Benzene release into the vapor space was modeled as a volumetric source from a single cell layer in a fluid zone centered about the pump location at the bottom of the tank.
Once the main purposes of the analysis was to demonstrate that the model was capable of reproducing the type of large vapor space benzene concentration gradients that were observed in one particular pump run. In this case, pump operations were resumed after a quiescent period of 61 days. During earlier pump operations, release rates on the order of 50 to 100 grams per minute had been observed. It was recognized that the long quiescent period could lead to larger release rates, so only a single pump was employed, and it was started at 600 rpm. The pump was shut down after 14 minutes when multiple high benzene concentration readings were detected. Measurements were taken from two sample ports located 18 inches above the liquid surface that were set to provide measurements at five minute intervals. The average benzene release rate based on the duration and total bulk vapor space concentration increase was calculated to be 1.6 kg per minute.
Four analysis cases were formulated to simulate the pump run with large benzene concentrations. All four cases used benzene release rates to match the calculated release rate of 1.6 kilograms per minute for 14 minutes. An additional 16 minute period with no benzene release was included for a total duration of 30 minutes. The four cases were developed to evaluate specific modeling methods. Case 1 contained a minimum number of computational volumes with no adaption and a fairly large source volume. The mesh was refined around the surface of the liquid in case 2 to show the effects of source layer thickness on benzene concentration. The purpose of case 3 was to illustrate the impact of several changes in the benzene release parameters that could affect benzene distribution. The radius of the source zone was increased from 30 ft. to 40 ft. and the source was ramped up linearly for the first minute of the analysis and ramped down over the last two minutes of the release period. Case 4 was designed to show the effect of further reductions in source layer thickness and grid refinement in the bulk vapor space. The layer of cells directly above the liquid was split and additional mesh refinement was carried out in the bulk vapor space.
CFD image showing contours of C6H6 Mole Fraction at 14 minutes.
The calculated concentration gradients for all cases were in general agreement with the experimental results. In particular, both analytical and experimental results showed the presence of large gradients near the surface of the liquid and in the vapor sampling location 18 inches above the liquid surface. These results demonstrated that CFD analysis is capable of reproducing the general behavior observed in the tank with large benzene release rates. The minor differences between the calculated and measured benzene concentrations at the measurement locations were judged to be primarily due to the simplifying release characterization assumptions made in constructing the model.
Further analysis of the vapor space gradients showed that mixing was occurring at a sufficient speed to avoid dangerous benzene concentrations under the current assumptions for the maximum benzene release rate. This application demonstrates that CFD can accurately model the vapor space mixing characteristics while providing considerably more data than can be obtained from physical testing at a far lower cost and leadtime.