ANALYSIS OF A LARGE SCALE LIQUID HYDROGEN SPILL EXPERIMENT
USING THE MULTI-PHASE HYDRODYNAMICS ANALYSIS CODE
K. Chitose* Y. Ogawa** T. Morii*
*Mitsubishi Heavy Industries, Ltd.
3-3-1, Minatomirai, Nishi-ku, Yokohama-shi , 220-84 Japan
**Advanced Reactor Technology Co.,Ltd.
15-1, Tomihisa-cho, Shinjuku-ku Tokyo, 162 Japan
ABSTRACT
It is planned to use hydrogen extensively as a source of clean energy in the next century. When using hydrogen it is necessary to consider safety because of its potential hazards such as explosions. The data obtained from laboratory-scale experiments have been sufficient to confirm the safety of small-scale uses of hydrogen in the past. However, computer simulation of the behavior of leaked hydrogen will play an important role in confirming the safety of extensive uses of hydrogen in future because of the large quantities involved and the presense of hydrogen in a phase other than the gas phase.
In suppport of the JAPAN MITI/NEDO World Energy Network(WE-NET)project, MHI has been tasked by Institute of the Applied Energy (IAE) to evaluate the accident scenarios of the leakage of liquid hydrogen. The formation and dispersion of flammable vapor clouds resulting from massive liquid hydrogen (LH2.) spills has not yet been simulated because the multi-phase and multi-component nature of the vapor cloud presents one of the most difficult problems for computer simulated fluid dynamics. Since we have developed the multi-phase hydrodynamics analysis code (CHAMPAGNE)[1], we applied the code to simulate the formation and dispersion of@hydrogen vapor clouds[2].It was found through these preliminary calculations that the mass flux evapolated from the leaked liquid hydrogen exerted a serious influence upon the spreading of the hydrogen vapor clouds.
In the present paper, therefore, we have developed the calculation model of the evaporation rate and incorporated the model into the CHAMPAGNE code in order to calculate the behavior of the hydrogen during the experiment in which 5100-liter quasi-instantaneous LH2. spills were performed at the NASA White Sands Test Station. Though we can use the experimental data for evaporation of liquid hydrogen obtained in only small scale under the stagnant condition, we have improved the evaporation model by comparing calculated results with the data.
1. INTRODUCTION
It is planned to use hydrogen extensively as a source of clean energy in the next century. When using hydrogen it is necessary to consider safety because of its potential hazards such as explosions. The data obtained from laboratory-scale experiments have been sufficient to confirm the safety of small-scale uses of hydrogen in the past. However, computer simulation of the behavior of leaked hydrogen will play an important role in confirming the safety of extensive uses of hydrogen in future because of the large quantities involved and the presense of hydrogen in a phase other than the gas phase.
The formation and dispersion of flammable vapor clouds resulting from massive liquid hydrogen (LH2.) spills has not yet been simulated because the multi-phase and multi-component nature of the vapor cloud presents one of the most difficult problems for computer simulated fluid dynamics. Since we have developed the multi-phase hydrodynamics analysis code (CHAMPAGNE), we applied the code to simulate the formation and dispersion of hydrogen vapor clouds. In the present paper, we have calculated the behavior of the hydrogen during the experiment in which 5100-liter quasi-instantaneous LH2. spills were performed at the NASA White Sands Test Station.
2. Experiment[2]
The most important issues to be resolved in the safety assessment of accidental massive LH2. spills are the position, the size and the hydrogen concentration of the vapor clouds formed after the spills. The density of the saturated hydrogen vapor at its boiling point is about 5% higher than the density of air and it is therefore slightly negatively bouyant when it is first vaporized. As the hydrogen is warmed by mixing with ambient air its density decreases and the increasing bouyancy of the gas mixture causes it to rise at an accelerating rate and leads to dispersion of the vapor cloud. Experimental data are necessary to understand the phenomenology of the behavior of the vapor clouds and validate the calculated results.
A LH2. spill facility was constructed at the NASA White Sands Test Station, and a comprehensive collection of specialized instrumentation was deployed on nine towers throughout the spill zone. The vapor cloud was monitored during its formation, expansion and turbulence. The visible cloud was tracked by means of photographs and motion pictures from several cameras. The spill site meteorological conditions were closely monitored and documented.
3.Calculations
3.1 The CHAMPAGNE code[1]
The CHAMPAGNE code has been developed to analyze the thermo-hydraulic behavior of a multi-phase and multi-component fluid which requires for its characterisation more than one set of velocities, temperatures, masses per unit volume, etc., at each location in the calculation domain. Calculations of multi-phase flow often show physical and numerical instability. The upwind-differencing and the fully implicit techniques used in the CHAMPAGNE code gives us convergent solutions more easily than the other techniques.
3.2 Phenomenological model
(1) Number of conservation equations
The number of conservation equations to be solved in the CHAMPAGNE code are shown as follows:
-mass(3) : hydrogen (liquid phase) hydrogen (gas phase) nitrogen (gas phase)
-momentum(2) : hydrogen (liquid phase) gas (hydrogen, nitrogen)
-energy(2) : hydrogen (liquid phase) gas (hydrogen, nitrogen)
(2) Interaction between phases
The CHAMPAGNE code has been developed so that the part of the code which calculates the transfer rates of the mass, momentum and energy generated by the phase transition is independent from the rest of the code (Figure 2). Therefore, it is possible to simulate the various patterns of the multiphase flow, by only modifing the corresponding part of the code. We have modified the code to consider the following interactions.
-mass : evaporation of liquid hydrogen
-momentum : friction between liquid droplets and gas
-energy : heat transfer betweeen liquid hydrogen and gas
(3) Evaporation rates of liquid hydrogen onto the ground
The evaporation rate is one of the dominant factor of the behavior of leaked liquid@hydrogen, so we have developed the calculation model of it. We refer the data measured in laboratory test[3]. The laboratory-scale experiments on the vaporizaion of liquid hydrogen and oxygen were performed and useful data on the evaporation rates were accumulated there.
The test system is illustrated in Figure 2. It has a sufficient thick ground layer in the glass vessel to maintain the temperature at the vessel bottom at it*s initial value. To mesure evaporation rates, the entire vessel was hung from a spring, and a load cell to convert weight into voltage was placed under the vessel. By dividing the total weight into upper and lower parts, small weight changes could be measured. The selected ground layers were sand and limestone concrete. About 1 to 2 seconds elapsed from when the liquid contacted the top surface of the layer to when the pouring process was complete.
When liquid hydrogen evaporates on the concrete surface, the steady-state evaporation rate is propotional to t-1/2(t:time). Since the temperature of the vapor layer above the liquid is exactly at the boiling point of the liquid, the heat flux required for liquid vaporization must be transferred from the concrete surface to the liquid. So the test data could be analysed without loss of accuracy by calculating the thermal conduction within the concrete layer.
Assuming that the temperature of the top surface of the concrete remains constant at the boiling point, the one-dimensional unsteady-state thermal conduction equation of the concrete layer can be solved analytically. The calculated variations of heat flux with time shows good agreemnet with the experimental data except the intial period. Therefore, the evaporation rate can be predicted by calculating the heat conduction within the concrete layer assuming that the heat transfer coefficient between the liquid and the concrete surface is infinite.
When liquid hydrogen evaporates on the dry sand layer, the water or air contained within interparticle cavities in the layer freezes and that the frozen layer acts as a barrier to prevent the liquid from soaking into the layer, since both water and air solidify at the boiling point at the boiling point of liquid hydrogen. The evaporation rate of liquid hydrogen above the sand layer also can be predicted by simple calculation. So if we change the material propertiy, the liquid hydrogen on the sand layer can be treated same to the evaporation on the concrete.
So we conclude to improve the evaporation model to give the heat flux ( function of the time) directly to the liquid hydrogen phase as‚boundary condition.
3.3 Analytical conditions
Figure 3 shows the grid arrangement used in the present calculation. Initial conditions are as follows:
-mass : nitrogen 99%, hydrogen gas 0.999%, liquid hydrogen 0.001%
-momentum : gas velocity of 2.2 m/s in the x-direction, liquid velocity of 0m/s
-energy : gas temperature 15degree (288.15K), liquid temperature 20K
Boundary conditions are as follows:
Nitrogen flows in from the left side of the analytical domain shown in Figure 3 with a velocity of 2.2 m/s and a temperature of 15degree (288.15K). Liquid hydrogen f lows in from the lowest mesh on the left side with the following velocity v and a temperature of 20K.
v=5.7/35/Ģy m/s
The condition on the right side is the pressure boundary. The upper and lower sides are free slip boundaries.
3.4 Results
At first, we assume the evaporation rate as a function of hydrogen droplet size and fix the droplet diameter to 3cm. The calculated distributions of hydrogen concentration as shown in Figure 4 will be compared with the experimental data as shown in Figure 5 which includes the concentration distributions evaluated from the observed temperatures. The calculated results of lower flammability limit of hydrogen (4% by volume) agree with the experimental results. Calculations using the CHAMPAGNE code with improved evaporation model are now in progress.
4. CONCLUSIONS
The formation and dispersion of flammable vapor clouds resulting from massive liquid hydrogen (LH2.) spills has not yet been simulated because the multi-phase and multi-component nature of the vapor cloud presents one of the most difficult problems for computer simulated fluid dynamics. Since we have developed the multi-phase hydrodynamics analysis code (CHAMPAGNE), we applied the code to simulate the formation and dispersion of@hydrogen vapor clouds. It was found through these preliminary calculations that the mass flux evapolated from the leaked liquid hydrogen exerted a serious influence upon the spreading of the hydrogen vapor clouds.
When we used the evaporation model as a function of hydrogen droplet diameter, the results strongly depend the parameter. In the present paper, therefore, we have developed the calculation model of the evaporation rate and incorporated the model into the CHAMPAGNE code in order to calculate the behavior of the hydrogen during the experiment. For the validation of evaporation rates, the experimental data are indispensable, however there are only a few experiments. So we will propose the experiments for evaporation of liquid hydrogen and improve the evaporation model by comparing calculated results with the experimental data.
5. References
[1] T.Morii and Y.Ogawa, "Development of a fully-implicit Fluid Dynamics Code for Multiphase and Multicomponent Flow", Proc., Sixth International Topical Meeting on Nuclear Reactor Thermal Hydraulics, Vol.1, Grenoble, France, October 5-8, 1993.
[2] J.E.Chirivella and R.D.Witcofski, "Experimental Results from Fast 1500-Gallon LH2. Spills", Cryogenic Properties, Processes, and Applications AIChE Symposium Series No.251 Vol.82 1986.
[3] K.Takeno, T.Ichinose, Y.Hyodo and H.Nakamura, "Evaporation Rates of Liquid Hydrogen and Liquid Oxygen Spilled onto the Ground", J. Loss Prev. Process Ind.,Vol.7, Nov. 5 1994.
|
