CFD SIMULATIONS OF GAS DISPERSION IN VENTILATED ROOMS T. Gélain, C. Prévost Institut de Radioprotection et de Sûreté Nucléaire (IRSN), Saclay, France Abstract In order to better understand the risks due to dispersion of gases of various densities that can be released accidentally in an industrial plant, especially a nuclear facility, IRSN has been conducting for several years an experimental and numerical research program on gas dispersion in ventilated rooms. Numerical simulations are performed using a commercial CFD code (ANSYS CFX). The program objective is to validate the computer code on gas tracing experiments based on the injection of sulphur hexafluoride SF 6 (heavy gas simulating UF 6 ) or helium (simulating the light gas H 2 ) in two ventilated rooms of the IRSN ( m to 1600 m ), for various ventilation and injection configurations. In an initial study of heavy gas (SF 6 ) dispersion, Ricciardi et al. (2008) performed a first validation of the CFD code ANSYS CFX but some experimental cases were not correctly reproduced. In a second stage, after a study of sensitivity to various parameters, but especially to the discretization scheme, we have retained a data set with the optimal numerical parameters, allowing to greatly improve the simulation of dispersion of heavy gas (SF 6 ) and to well reproduce the experimental results of dispersion of light gas (helium), obtained by Prévost et al. (2007). The simulation results of tests carried out with helium and SF 6, and their comparison with experimental results, are presented in this paper. Keywords: Simulations, CFD, dispersion, helium, SF 6, ventilated room, hydrogen, LEL. 1 Introduction In the nuclear industry, the fuel cycle is made up of a succession of steps in the course of which the fuel undergoes various transformations, before or after its use in the reactor. Although closely monitored, these various steps are susceptible to an incident, even an accident, in the course of which gases can be released and dispersed in the rooms of the nuclear facilities. Depending on the step involved (upstream or downstream of the cycle, reactor), these gases can be of high density compared with air, as it is the case for example for UF 6 (uranium hexafluoride, present during the conversion and uranium enrichment phase), or of low density compared with air, as it is the case for hydrogen. The latter can be produced in the event of a severe accident in a reactor building (Fukushima accident, TMI, etc.), or released in the event of a rupture in pipes carrying hydrogen, used to cool the turbines in the machine room. In the context of safety evaluation of nuclear facilities, this type of incident has been especially studied and monitored, notably the density effects that can lead to a local accumulation of gas and, in the case of hydrogen, to a risk of explosion. This type of risk is obviously not specific of nuclear field and may also occur in the chemical industry (manufacture of sulphuric acid). An initial study on dispersion of a heavy gas in a ventilated room was conducted by Ricciardi et al. (2008). This study, both experimental and computational, shows that even in low quantities, injection of a gas of high density compared with air, in this case SF 6 (tracer Page 1/6
used to simulate a heavy gas), can lead to stratification phenomena and thus creates areas of high concentration. The experimental results acquired in this initial study allowed the ANSYS CFX version code to be satisfactorily validated, although certain phenomena could not be reproduced. In particular, when gas injection into the ventilated room stopped, the decrease in SF 6 concentration was linked, in the simulations, to the air exchange rate in the room, while in the experiments, an additional effect due to the gas falling in the room was observed. Following this study, and in the framework of an experimental and numerical study related to the inverse problem of emission of a light gas (Prévost et al. (2007)), we performed a study of sensitivity highlighting especially the need of a 2 nd order discretization scheme called "High Resolution Scheme" (HRS) in ANSYS CFX. This study of sensitivity is presented in details in Gélain and Prévost (submitted in 2012). In this paper, we first present the program set up in order to study the dispersion of a heavy or light gas in a ventilated room, and the test grids which were then simulated with the ANSYS CFX code version 12 (2009). The experimental study was carried out so as to obtain a multidimensional cartography of the concentration of a gaseous tracer injected into the room. Some numerical results of dispersion of heavy and light gas are then presented and compared to experimental results. 2 Experimental program This experimental study was performed in two rooms of different volumes, the m CEPIA room and the 1600 m room 4, shown in Fig. 1 and Fig. 2. These two rooms also differ in their ventilation system. CEPIA (Fig. 1) has four supply openings located on one side of the room and four exhaust openings on the opposite side, thus allowing the various ventilation configurations to be adjusted. For the tests studied in this article, the ventilation configuration used consists of two blowers in a high position in the room and two exhaust vents on the opposite side in a low position in the room. Figure 1: CEPIA By contrast, the ventilation of room 4 consists of a central duct in the upper part, from which five blower ducts branch off, each with three outlets in the upper part, middle part and lower part, while two exhaust openings are located on the lower part of one of the sides of the room. Figure 2: room 4 Page 2/6
Table 1: test grid for CEPIA (He and The tests performed in these two rooms SF 6 ) consist in injecting a tracer gas (SF 6 or helium) vertically into the lower part of the room for a given time, under stable airflow conditions. During the test, the time evolution of the concentration of tracer gas is monitored at various points in the room to study its dispersion (Fig. 1 and 2). Various parameters have been studied, related either to ventilation (air exchange rate R), or to injection of the tracer (injection velocity U inj, diameter of the injection nozzle d inj, duration of injection t inj ). The test grids are presented in Table 1 for CEPIA and Table 2 for the room 4. In Table 2, we only present the grid for helium; the Test 1 Test 2 Test Test 4 Test Test 6 1 120 120 1.6 6. 22.6 1.6 6. 28. test grid for SF 6 in the room 4 are detailed in Ricciardi et al. (2008). Test R (h -1 ) tinj (s) dinj (mm) Uinj (m.s -1 ) Table 2 : test grid for room 4 (He) Test R (h -1 ) dinj(mm) Uinj(m.s -1 ) Test 1 10 2.6 Test 2 4 2.6 Test 10 2.6 1 Test 4 4 2.6 In Tables 1 and 2, two air exchange rates R (1 h -1 and h -1 ) are used and four injection duct diameters d inj ( mm and mm for CEPIA - 4 mm and 10 mm for room 4) are studied, while the injection velocities U inj vary between 6 m.s -1 and 0 m.s -1 for CEPIA and 2.6 m.s -1 and 2.6 m.s -1 for room 4. The injection times t inj are either 120 s or s for CEPIA while for the room 4, injection continues until the equilibrium of the concentration (around 6 000 s). Numerical simulations.1 Basic equations The numerical simulations of the flows and dispersion of the tracer gas in the ventilated rooms studied are performed with the CFD code ANSYS CFX (version 12). The flows are simulated by solving the non-stationary Navier-Stokes equations with turbulence using a standard RANS (Reynolds Averaged Navier Stokes) formulation. The equations solved in the calculations performed with CFX are based on some assumptions: the fluid considered is a mixture of air and tracer gas (helium or SF 6 ), taken to be mixed at the molecular level (multi-species formulation). The flow is turbulent, isothermal (2 C) and weakly compressible. The gaseous mixture is considered to be an ideal gas. Two gases, air and tracer gas, are considered, but only the tracer gas transport is simulated, the mass fraction of air being calculated as the difference. Before the injection, the flow is taken to be established, and for this an initial stationary calculation is performed, followed by a second transient calculation in two phases: injection of tracer for a given time, then decrease of the tracer concentration after injection ends. Page /6
.2 Geometry and meshes The geometries of CEPIA ( m ) and room 4 (1600 m ) are shown in Fig.. They have been produced using the ANSYS Design Modeler software. (a) CEPIA (b) room 4 Figure : geometries of the rooms studied (a) CEPIA (b) room 4 Figure 4: meshes for the geometries studied The meshes were produced using ANSYS Meshing and are shown in Fig. 4. These meshes are non-structured and composed of tetrahedral cells.. Computational parameters To carry out the calculations, it is necessary to specify the numerical methods used by the solver to discretize the advection terms as well as the convergence parameters for the calculations. For this application, two discretization schemes are available in CFX, 1 st -order (Upwind Differencing Scheme) or hybrid (High Resolution Scheme), the first being more diffusive in areas of high gradients than the second, which adjusts the order of the discretization scheme depending on the local gradients. These different schemes are detailed in the technical documentation for CFX-12 (2009). Finally, the HRS scheme was chosen for all the calculations after the performing of sensitivity study (Gélain and Prévost, submitted). We also specify the convergence parameters for the calculation, namely the number of iterations per time step, chosen to be for the study, and the convergence criterion, which is chosen to be 10-6 for the RMS (Root Mean Square) Max residuals. Finally, the transient calculation has been carried out with a time step of 1 s. A sensitivity study allowed to validate the meshes presented on the Fig. 4, and to choose the k- model of turbulence and the HRS discretization scheme for all the calculations. Page 4/6
4 Results and discussions 4.1 CEPIA room ( m ) Different tests were simulated with the ANSYS CFX code in CEPIA, to assess the influence of some parameters, and were compared with the experimental results for each of the proposed tests; only some examples are presented in this paper. The results are reported in the form of average values per level (high, median, low and floor), calculated from the values measured at the different sampling points. For the case of SF 6 injection, the Test 6 (Fig. ) is considered, because its simulation with CFX has been greatly improved with the HRS scheme in comparison with that presented in Ricciardi et al. Test 6 HRS scheme Figure : variations in SF 6 concentrations in the CEPIA room at different levels (Test 6) Figure 6: variations in helium concentrations in the CEPIA room at different levels (Test 1) (2008). Indeed, in the latter, the phenomena of abrupt decrease in concentration at the low and median levels (blue and red curves) revealed experimentally were not correctly simulated by the CFX version.7.1 code using a first order scheme. For the case of helium injection, the Test 1 (Fig. 6) is presented and shows a good agreement between experimental and numerical results, which confirms the choices considered in the data set. 4.2 Room 4 (1600 m ) As for CEPIA, different tests were simulated with the ANSYS CFX code in the room 4, and compared with the experimental results for all the tests performed. The simulations of SF 6 dispersion in the room 4 are not presented, because no improvement was made by the new data set including the HRS scheme compared to the simulations of Ricciardi et al. (2008). Fig. 7 presents the results for the Test 1 in the room 4 and shows a good agreement between the simulations results and the experimental ones. Page /6
Figure : variations of helium concentrations in room 4 at different levels (Test 1) Conclusion This paper, which follows the study conducted by Ricciardi et al. (2008) on dispersion of a heavy gas (SF 6 ) in a ventilated room, presents simulation results of dispersion of heavy (SF 6 ) and light gas (helium) in ventilated rooms, obtained with the ANSYS CFX version 12 code. These simulations were carried out on the basis of tests performed in two rooms of different volumes, the CEPIA room ( m ) and the room 4 (1600 m ). The data set of the code was improved since the previous study, by the use of a 2 nd order numerical scheme (HRS), retained in the framework of a study of sensitivity. This choice allows to much better reproduce the evolution of SF 6 concentration in CEPIA after the end of the gas injection. Moreover, the numerical results of dispersion of light gas show also a good agreement with the experimental results, validating the ANSYS CFX code and the data set implemented for this application. Taking into account of the heterogeneity and the high levels of the gas concentration observed in some cases, and in the context of safety analyses related to hydrogen risk, a study is currently underway from tests allowing maximum local concentrations of at least 4 %, which represents the Lower Explosive Limit (LEL) of hydrogen, to be attained. 6 References ANSYS CFX-Solver Theory Guide (2009), Release 12.0. Gélain, T., Prévost, C. (2012) Experimental and numerical study of light gas dispersion in a ventilated room, submitted to Computers and Fluids. Prévost, C., Bouilloux, L., Gélain, T., Norvez, O. and Ricciardi, L. (2007) Study of gas dispersion in ventilated rooms. Roomvent, Helsinki. Ricciardi, L., Prévost, C., Bouilloux, L. and Sestier-Carlin, R. (2008) Experimental and numerical study of heavy gas dispersion in a ventilated room, Journal of Hazardous Materials 12, 49 0. Page 6/6