AIAA-00-0858 Microgravity Geyser and Flow Field Prediction

R.J Thornton and J.I. Hochstein The University of Memphis Memphis, TN

AIAA-00-0858

MICROGRAVITY GEYSER AND FLOW FIELD PREDICTION

R.J. Thornton*, and J.I. Hochstein, The University of Memphis Memphis, Tennessee

ABSTRACT Modeling and prediction of flow fields and geyser formation in microgravity cryogenic propellant tanks was investigated. A computational simulation was used to reproduce the test matrix of experimental results performed by other investigators, as well as to model the flows in a larger tank. An underprediction of geyser height by the CFD code led to a sensitivity study to determine if variations in surface tension coefficient, contact angle, or jet pipe turbulence significantly effect the simulations, it was determined that computational geyser height is not sensitive to slight variations in any of these items. An existing empirical correlation based on dimensionless parameters was re-examined in an effort to improve the accuracy of geyser prediction. This resulted in the proposal for a re-formulation of two dimensionless parameters used in the correlation; the non-dimensional geyser height and the Bond number. It was concluded that the new non-dimensional geyser height shows little promise. Although further data will be required to make a definite judgement, the reformulation of the Bond number provided correlations that are more accurate and appear to be more general than the previously established correlation. NOMENCLATURE a Bo D F G R We = Acceleration = Bond Number = Diameter = Flow Characterization Parameter = Non-Dimensional Geyser Height = Radius = Weber Number = Density = Surface Tension Coefficient

o RF t TH tj

= Jet Pipe Outlet = Re-Fit = Tank = Thornton-Hochstein = Tank/Jet INTRODUCTION

Often spacecraft missions require the vehicle to store rocket propellants for future use. In the case of cryogenic liquid propellants, tank self-pressurization can be a significant problem. Despite insulation, incident solar radiation heats the cryogenic fluid causing liquid to vaporize and raise the pressure inside the tank. If this self-pressurization were allowed to continue unchecked, the tank would rupture. A stronger tank could help, but this would require an undesirable increase in vehicle mass. Another solution to the tank pressurization problem is to vent the tank. Tank venting is undesirable because it wastes valuable propellant. In addition, due to the lack of gravity to positively orient the propellant in a predictable manner, it would be impossible to locate a vent where it could be certain that only vapor would be vented. In fact, even in partially filled tanks, it is possible that the entire tank surface could be wetted due to the influence of surface tension in a microgravity environment. An alternative solution to the tank self-pressurization problem is a Thermodynamic Vent System, or TVS1, 2. The TVS would extract a small portion of the bulk liquid from the tank and pass it through a JouleThomson valve resulting in a reduction in temperature as well as pressure. Once cooled, the fluid would be routed through a heat exchanger, which is used to cool a separate flow of propellant extracted from the bulk liquid. If the fluid leaving the Joule-Thomson valve is a two-phase mixture, it will continue changing phase in the heat exchanger until it is completely vaporized and will eventually be sacrificially vented overboard. The other stream of cooled liquid could then be pumped from the heat exchanger back into the tank and injected through an axial jet pipe located at the fore end of the tank. The jet flow provides several benefits including mixing of the bulk liquid, which helps to reduce temperature gradients and in turn helps prevent

Subscripts AYD = Aydelott j = Jet at the Liquid-Vapor Interface

*

Research Assistant, Mechanical Engineering, Member AIAA Professor/Chair, Mechanical Engineering, Member AIAA

Copyright 2000 The American Institute of Aeronautics and Astronautics Inc. All rights reserved.

1 American Institute of Aeronautics and Astronautics

evaporation of the propellant. Introduction of the cooled liquid also reduces the temperature of the bulk fluid. In addition, if the jet has a moderate amount of momentum it can cause the formation of a geyser at the liquid/vapor interface. The increased surface area of the free surface due to the formation of a geyser would help promote condensation, thus reducing the pressure even further. At higher levels of jet momentum, the geyser will strike the opposite end of the tank and either form a separate pool or roll down the tank walls re-mixing with the bulk fluid. As the fluid comes around the tank walls, there will also be a cooling effect on the wall. However, the addition of excessive kinetic energy to the bulk fluid would eventually result in undesirable heat generation through viscous dissipation. A morphology has been defined based on four distinct flow patterns that occur when an axial jet is injected into the bulk liquid region of a propellant tank in a microgravity environment1. These flow patterns are defined as follows (Figs. 1 and 2): I. Dissipation of the jet in the bulk liquid region. II. Geyser Formation. III. Collection of jet liquid in the aft end (opposite of the jet inlet) of the tank. IV. Liquid circulation over the aft end of the tank and down the tank walls re-mixing with the bulk liquid.

Fig. 2. Flow Patterns III and IV. COMPUTATIONAL MODEL, ECLIPSE The complexity and expense of microgravity experimentation severely limits the amount of data that can be, and has been, collected on microgravity geyser formation. Therefore, an attractive alternative is the use of CFD, computational fluid dynamics, to model the geyser flows. The use of a CFD code can significantly reduce the cost of investigating the microgravity geyser phenomena, as well as allow for easy manipulation of parameters such as tank size, fluid properties, jet momentum levels, and gravitational acceleration levels. The ECLIPSE code was used for the computational studies in this work. ECLIPSE was chosen for its previously demonstrated ability to model the microgravity geyser flows of interest here3, 4. ECLIPSE has evolved from the RIPPLE5 code developed at Los Alamos National Laboratory. RIPPLE is a descendent of the NASA VOF2D code, and was developed to model transient, two-dimensional, laminar, incompressible fluid flows with free surfaces of general topology.5 The flow field is discretized into finite volumes to form a regular non-uniform mesh. RIPPLE models free surfaces with volume of fluid (VOF) data on the mesh, and a continuum surface force (CSF) model is used to model surface tension. Staggered grid differential equation approximations result in a system of algebraic equations that are solved by a two step projection method employing an incomplete Cholesky conjugate gradient (ICCG) solution technique for the pressure Poisson equation (PPE)5. One of the features added to RIPPLE to produce ECLIPSE is the implementation of the two-equation Jones-LaunderPope k turbulence model.

Fig.1. Flow Patterns I and II. The ability to predict the flow pattern for any given combination of tank geometry, tank fill level, gravitational acceleration, and jet momentum is crucial for optimization and implementation of a TVS. A study of geysers and their formation is therefore a necessary task in the development process of the TVS concept.

2 American Institute of Aeronautics and Astronautics

Computational Model Fidelity and Sensitivity It has been demonstrated that ECLIPSE and its predecessors do a credible job of predicting geyser flow pattern as defined by the four-pattern morphology previously presented1. Given the dearth of low-gravity geyser data, test cases in these previous studies included configurations with gravity levels as high as 1g. The focus of the present research was on low-gravity environments with completely turbulent jets, which reduced the pool of available experimental data to only 14 test cases in which a geyser was formed, (Flow Pattern II). Fig. 3 shows the measured geyser height, displayed in dimensionless form, as a function of the Weber number and Bond number as defined by Aydelott2.

Also displayed in Fig. 3 are the geyser heights predicted by ECLIPSE for these cases. Although the correct flow pattern is indeed predicted, the computational predictions consistently underpredict geyser height. This underprediction averages 44% with a standard deviation of only 9.4% about this mean. Although this level of agreement should be useful for predicting trends, and has been useful for predicting flow patterns, it prompted an investigation into the sensitivity of the computational simulation to user specified parameters; surface tension coefficient, contact angle, and inlet turbulence quantities. Aydelotts experiments were conducted in plexiglas tanks that were cleaned with the utmost care prior to filling with ethanol; even a small amount of contaminant can significantly alter the surface tension coefficient, . The literature does not indicate that any of the physical properties of the fluid were measured and it appears that all calculations related to the experiments have been performed using standard published values for these properties. To study the sensitivity of the simulation to the value of , simulations of a case with a measured geyser height of 6.5 cm and a computationally predicted height of 3.66 cm were rerun with specified to be 90% and 80% of

Bo =

aR j

acceleration force surface tension force

(1)

We =

Vo Ro D j
=

inertia force surface tension force

(2)

G=

hg Rt

geyser height tank radius

(3)

Experimental Non-Dimensional Geyser Height Computational Non-Dimensional Geyser Height

Fig. 3. Aydelotts experimental non-dimensional geyser heights1 and computational simulations of these experiments as a function of Weber and Bond number.

3 American Institute of Aeronautics and Astronautics

the published value for ethanol. Reducing sigma by 20% resulted in an 8% (0.31 cm) increase in computationally predicted geyser height. Although this is a small improvement, it does not fully account for the difference between predicted and measured heights. Further, although the lower value improves the predictions, and the lowered value might be appropriate if there was a slight contamination of the ethanol, there is no evidence in the literature that there was contamination and there is no rationale for selecting a specific value for sigma other than the standard published value. In fact, because the value of is so sensitive to even slight contamination of the fluid, it is somewhat comforting to know that the predicted geyser height is not particularly sensitive to the value of this parameter. The second parameter to be examined in the sensitivity study is the contact angle between the liquid and the tank wall, . Actual propellant/aluminum/vapor combinations and the experiment ethanol/plexiglas/air combinations exhibit very small contact angles. To avoid divide-by-zero difficulties in the simulations, the contact angle in the standard simulations has been specified to be 2 degrees. A sequence of simulations was run with increasing values of for the same test case that was used for the -sensitivity study. Increasing all the way to 30 degrees resulted in an increase in predicted geyser height of only 0.3%. It is therefore concluded that geyser height is relatively insensitive to contact angle. The final group of parameters examined as part of the sensitivity study are the turbulence kinetic energy, , and the turbulence energy dissipation rate, , at the exit of the jet-pipe. Standard simulations are run using values computed using published correlations for fully developed flow in a pipe. To examine the sensitivity of geyser height to this fairly crude assumption, three additional simulations were performed. The first used values for these parameters were simply half of the values computed from the correlations. The second and third used values of were selected to produce a turbulent viscosity at the pipe outlet equal to 10% and 1% (respectively) of the fully developed value. The increase in geyser height was less than 8% for any of these cases. Although this is an improvement in the predictive accuracy of the computational simulation, there is no clear rationale for using these reduced values. Again, as concluded for the -sensitivity study, it is comforting to show that geyser height is not particularly sensitive to the specified value of or and all standard simulations for this study were performed using values computed using the standard correlations for full developed pipe flow.

What conclusions can be reached about the fidelity of the computational model? In the past, the computational model has been asked to simply predict which of the four flow patterns will be produced by a given jet/tank/filling combination and it still does so with good reliability. For the present research, the more challenging task of accurately predicting geyser height for flow pattern II configurations was selected as the fidelity criterion. The computational simulation consistently underpredicts geyser height with a standard deviation about the mean error of less than 10%. Although the source of this error has not been identified, it seems to be consistent. A sensitivity study to examine the influence of , , inlet, and inlet on simulated geyser height was conducted because it was felt that these parameters were the most likely to have different values between the simulation and the experiment. Although the simulations predictive accuracy could be improved by changing , inlet, and inlet in a reasonable manner, a sound justification for making these changes could not be established. It is reassuring to note that although the predictions improved, the predicted geyser height is not particularly sensitive to these parameters that may not be known priori with great precision in a spacecraft design environment. Based on all of this information, it was concluded that ECLIPSE is still a very useful tool for predicting jet-induced flow patterns in a propellant tank, that it consistently underpredicts geyser height, and that it should be useful for exploring the relationships between the various parameters that influence geyser production in a low-gravity environment.

DIMENSIONLESS MODELING

Small scale testing was conducted during the late 1970s and 1980s to investigate the geyser phenomena. Through drop tower testing performed in the NASA LeRC Zero Gravity Facility, Aydelott2 determined that the inertia, acceleration, and surface tension forces are the primary factors affecting geyser formation in a microgravity environment. From these forces, he deduced that the main parameters in the dimensional analysis should be the jet velocity, liquid density and surface tension, acceleration environment, geyser height, and selected characteristic lengths. Using the Buckingham Theorem, Aydelott established three dimensionless groups pertinent to predicting geysers and geyser height. These dimensionless groups are the jet-Bond number Bo, the Weber number We, and a nondimensional geyser height G. Aydelott formed his dimensionless groups as presented in Equations (1), (2), and (3) above.

4 American Institute of Aeronautics and Astronautics

Aydelott concluded from his data that a linear relation existed between the non-dimensional geyser height and the Weber number. He also assumed that the nondimensional geyser height should be inversely proportional to the Bond number. Through a leastsquares deviation curve fit of turbulent experimental data the following geyser height prediction correlation was developed by Aydelott:
F= 0.5 + 1.6 We 1 + 0.6 Bo

(4)

where F is the flow characterization parameter and is expected to equal the dimensionless geyser height, G. To formulate a reliable correlation applicable to propellant tank microgravity geyser prediction in general, it is necessary to not only recognize the proper dimensionless groups, but also identify the proper length scales in these parameters. As seen in Equations (1), (2), and (3) Aydelott uses three different length scales, the tank radius Rt, jet pipe radius Ro, and the jet radius at its point of impact with the liquid-vapor interface Rj. The shape and curvature of the liquidvapor interface once a geyser begins to form becomes quite complex (Fig. 4). Therefore, the determination of an appropriate length scale must take into account whether the formation of a geyser is dependent on local effects (i.e. in the vicinity of the actual geyser), global effects (i.e. tank scale effects), or a combination of the two.

Non-Dimensional Geyser Height, G Aydelotts experimental correlation was formulated with Weber and Bond numbers dependent only on the jet properties making them local parameters. If the local effect only model is accepted, and Aydelotts original formulations of Bo and We are assumed to be appropriate, it seems reasonable that the nondimensional geyser height should also be scaled with the jet radius instead of the tank radius. It should be noted that all of Aydelotts experiments were conducted in configurations with the same tank radius to jet radius ratio. Therefore, it is possible that Aydelotts correlation may not perform well when applied to tanks with a tank to jet radius ratio that is different from the experiment configuration; while a reformulation of the non-dimensional geyser height to include the jet radius may provide improved accuracy. The proposed new non-dimensional geyser height uses the jet diameter at the liquid-vapor interface determined by approximate expressions2.
GTH = hg Dj

(5)

Several analyses have been made to determine the validity of this new non-dimensional geyser height, GTH, utilizing Aydelotts experimental data as well as computational data. Bond Number, Bo The Bond number is the ratio of acceleration force to surface tension force, and can have both local and global effects on geyser formation. In a no geyser situation, the tank Bond number will dictate the shape of the free surface, and at least partly so in a geyser situation. Three different formations of the Bond number were considered for the geyser problem. The jet-Bond number Bo or Boj (Equation (1) used by Aydelott), the tank Bond number Bot (using the tank diameter as the characteristic length scale, Equation (7)), and a combination tank/jet-Bond number Botj (using the tank and jet diameters, Equation (8)).
Bot = aDt
2

a Dt D j

(7)

Fig. 4. Typical free surface shape of a microgravity geyser.

Botj =

c h

(8)

5 American Institute of Aeronautics and Astronautics

Although in certain circumstances, the local or global surface tension effects may dominate the physics of geyser formation, most cases are likely to involve both. Incorporating the tank or tank/jet-Bond number may be a feasible way to account for both local and global surface tension effects. An investigation was conducted to determine if either Bot or Botj could be used to produce a correlation with better predictive ability than the existing correlation based on Boj. Weber Number, We The Weber number is the ratio of the inertia force to surface tension force. In the geyser problem, the Weber number primarily describes the interaction between the momentum of the jet impacting on the free surface and the restoring force due to surface tension. Therefore, the jet-Weber number (Equation (2)) used by Aydelott still appears to be the appropriate formulation, and has therefore been employed throughout this study.
NON-DIMENSIONAL GEYSER HEIGHT STUDY

but replaces the tank radius with the jet impact diameter to form the new dimensionless geyser height, GTH. As can be seen from Table 1, the refit of the original correlation provided a slight improvement in the correlations accuracy for the Fig. 3 data, while the use of the newly proposed non-dimensional geyser height resulted in a decrease in accuracy. Fig. 5 presents the measured non-dimensional geyser heights (G) as a function of these three flow characterization parameters (F). It also shows three straight lines representing a linear fit to each of the data sets using these three formulations. Table 1. Experimental data correlation results.
Corelation Aydelott Expression RMS Error 0.203

FAYD = FAYD / RF =
FTH =

0.5 + 16 . We 1 + 0.6 Bo 0.34 + 145 . We 1 + 0.63 Bo

Aydelott Refit New Formulation G TH

0.190

3.43 + 5.69 We 1 + 198 . Bo

0.388

A combination of experimental and computational data was used to evaluate the predictive utility of the proposed dimensionless geyser height. The experimentally measured and computationally predicted geyser data presented in Fig. 3 was supplemented with the results of several additional simulations. Twelve simulations were run using the same tank as the experiment but with parameter values specified to provide a better spread of data over the We-Bo space of Fig. 3 than was available from the experiment. Another set of cases were defined to almost duplicate the experiment conditions except for a tank with a diameter equal to 150% of the diameter of the experiment tank. It should be remembered that the computational simulations consistently underpredict geyser height. Therefore, in the following evaluations, computational and experimental data are segregated as appropriate. All of the correlations developed in this study were produced using a least-squares regression method. The root mean square (RMS) error between a correlations predicted geyser height and that measured in the experiment, or predicted directly by the simulations, was selected as the measure of quality for this study. Experiment Data - Experiment Correlations The first correlation presented in Table 1 is the original correlation, (Equation (4)), developed by Aydelott for all of his experiments including cases not part of the present study. The second correlation is a refit of the original formulation using only the data presented in Fig. 3. The third correlation uses the same formulation

7 New G Correlation 6 Original G Refit Correlation Original Correlation Original G Refit Correlation 5 New G Correlation Original Correlation 4

Non-Dimensional Geyser Height G

0 0 1 2 3 4 5 6 7

Flow Characterization Parameter F

Fig. 5. Experimental non-dimensional geyser height as a function of the three formulations of Table 1 (lines are linear fit to each correlations data set). Computational Data - Experiment Correlations The experimental tests were all simulated with ECLIPSE, as were similar cases with the 50% enlarged tank. The geyser heights of these simulations where predicted using the same three correlations of Table 1 developed strictly from experiment data. Although the quality of the predictions was significantly lower than Table 1 because the simulations consistently

6 American Institute of Aeronautics and Astronautics

underpredict geyser height, the relative performance of the correlations was not different from that of Table 1.
Non-Dimensional Geyser Height G

5 New G Correlation 4 Original Correlation Original Correlation New G Correlation 3

Computational Data - Computational Correlations Computational Data for the Experiment Tank Geyser heights predicted by the computational simulations for the original experiment conditions were used to refit the original correlation formulation, and the new non-dimensional geyser height formulation, to produce the results presented in Table 2. The experiment tank has a 0.4 cm diameter jet pipe, hemispherical heads, a 10 cm diameter, and a total length of 20 cm. The new formulation does not outperform the original formulation. Table 2. Correlations from original tank computational data only.
Correlation Aydelott New Formulation G TH Expression RMS Error 0.155

0 0 1 2 3 4 5

Flow Characterization Parameter F

Fig. 6. Computational non-dimensional geyser height for both tanks as a function of the two formulations of Table 4 (lines are linear fit to each correlations data set). Table 4. Correlations from computational data of both tanks.
Correlation Aydelott Expression RMS Error 0.160

FAYD

0.014 + 0.73We = 1 + 0.71 Bo 0.48 + 176 . We 1 + 116 . Bo

FTH =

0.172

FTH

0.66 We = 1 + 0.83 Bo 0.43 + 185 . We 1 + 122 . Bo

Computational Data for the Enlarged Tank The conditions of the experiment cases were reproduced in a tank with a diameter equal to 150% of the diameter of the experiment tank. This enlarged tank has a 15 cm diameter and a total length of 25 cm. The resulting data was again used to fit the two correlations and once again the new formulation did not outperform the original one. Table 3. Correlations from enlarged tank computational data only.
Correlation Aydelott Expression RMS Error 0.121

New Formulation G TH

FTH =

0.179

Original Tank Correlations Applied to Enlarged Tank In an actual design situation, the correlation from a test tank will be applied to a different sized tank. To see how the two versions of the correlations perform when applied to tanks of differing size, the original tank computational correlations (Table 2) were used to predict the geyser height in the enlarged tank. Table 5 outlines these results. Table 5. Results of the original tank correlations applied to the enlarged tank data.
Correlation Aydelott Expression RMS Error 0.200

FAYD =

0.028 + 0.66 We 1 + 114 . Bo

New Formulation G TH

FTH =

0.29 + 182 . We 1 + 119 . Bo

0.158

FAYD =

0.014 + 0.73We 1 + 0.71 Bo

Correlations Based on Data from Both Tanks Applied to Both Tanks The computational data from both tanks was used to fit correlations for both non-dimensional geyser heights. Table 4 outlines these results. Fig. 6 is a plot of the nondimensional geyser height and flow characterization parameter for both tanks.

New Formulation G TH

FTH =

0.48 + 176 . We 1 + 116 . Bo

0.219

7 American Institute of Aeronautics and Astronautics

Enlarged Tank Correlation Applied to Original Tank The enlarged tank computational correlations (Table 3) were used to predict geyser height in the original tank. This is the only case where the newly proposed nondimensional geyser height matched the accuracy of the original non-dimensional geyser height correlation. Table 6. Results of the enlarged tank correlations applied to the original tank data.
Correlation Aydelott New Formulation G TH Expression RMS Error 0.223

indicated in Fig. 7 where it can be seen that the linear fit to both the Bot and the Botj correlation data sets have slopes near unity and would pass nearly through the origin. Table 7. Experimental data correlations for various Bond number formulations.
Corelation Tank-Bond Number Tank/JetBond Number Jet Bond Number Expression RMS Error 0.234

FBot =
FBotj =

0.57 + 160 . We 1 + 0.016 Bot

0.60 + 186 . We 1 + 0.079 Botj

FAYD =

0.028 + 0.66 We 1 + 114 . Bo 0.29 + 182 . We 1 + 119 . Bo

0.161

FTH =

0.221

FBo j =

0.5 + 16 . We 1 + 0.6 Bo j

0.250

G Study Summary As can be seen from Tables 1-6, the desired improvement using the jet diameter normalized nondimensional geyser height was not realized. In all but one case (the enlarged tank correlations applied to the original tank), the original correlation was more accurate in predicting geyser height. It is possible that as more data for a wider variety of configurations becomes available it may appropriate to reinvestigate the GTH formulation for the non-dimensional geyser height parameter. However, given the currently available data, the present study leads to the conclusion that the original formulation is superior and should be used.
BOND NUMBER RE-FORMULATION

2.5

Non-Dimensional Geyser Height G

1.5 Jet-Bo Correlation (Aydelott) Tank-Bo Correlation 1 Tank/Jet-Bo Correlatioin Jet-Bo Correlation 0.5 Tank/Jet-Bo Correlation Tank-Bo Correlation 0 0 0.5 1 1.5 2 2.5 3 3.5

Flow Characterization Parameter F

To determine if the tank-Bond number, Bot , or the tank/jet-Bond number, Botj , can provide improved geyser prediction, correlations were refitted in a manor similar to the non-dimensional geyser height study. These refitted correlations were based-on the original experimental data as well as the computational predictions for both the original and the enlarged tank. Experimental Data - Experiment Correlations Again, the experiment data was used as a starting point for the correlation study. Table 7 presents a summary of the performance of the Bot correlation, the Botj correlation, and the original Aydelott correlation (Boj), in predicting experimentally measured geyser height. The Bot correlation shows a slight improvement over the original correlation whereas the Botj correlation shows a whopping 35% improvement in predictive accuracy. Ideally, a correlation would result in a direct prediction of G, (i.e. F = G). This is precisely the result

Fig. 7. Experimental non-dimensional geyser height as a function of the three Bond number correlation versions of Table 7 (lines are linear fit to each correlations data set). Computational Data - Computational Correlations Again, to study the effects of different tank configurations the computational data from the original and enlarged tanks were used in various combinations to study the performance of the proposed correlations. Original Tank Data Table 8 presents the predictive performance of the three formulations using coefficients determined from simulated geyser heights for the original tank when the correlations are applied to precisely that data set. Again the Bot correlation slightly outperformed the original correlation and the Botj correlation was the best performer resulting in a 19% improvement over the Boj correlation.

8 American Institute of Aeronautics and Astronautics

Table 8. Original tank computational data correlations for various Bond number formulations.
Corelation Tank-Bond Number Tank/JetBond Number Jet Bond Number Expression RMS Error 0.131

Table 10. Both tank computational data correlations for various Bond number formulations.
Corelation Tank-Bond Number Tank/JetBond Number Jet Bond Number Expression RMS Error 0.108

FBot

0.25 + 0.95We = 1 + 0.023 Bot

0.20 + 0.96 We 1 + 0.081 Botj

FBot

014 . + 0.78 We = 1 + 0.015 Bot

0.084 + 0.77 We 1 + 0.062 Botj

FBotj =

0.125

FBotj =

0.116

FBo j =

0.014 + 0.73We 1 + 0.71 Bo j

0.155

FBo j =

0.66We 1 + 0.83 Bo j

0.160

Non-Dimensional Geyser Height G

Enlarged Tank Data Table 9 presents the predictive performance of the three formulations using coefficients determined from simulated geyser heights for the enlarged tank when the correlations are applied to precisely that data set. In this instance, both the Bot and the Botj correlations significantly outperformed the Boj correlation by 49% and 29% respectively. Table 9. Enlarged tank computational data correlations for various Bond number formulations.
Corelation Tank-Bond Number Tank/JetBond Number Jet Bond Number Expression RMS Error

2 Jet-Bo Correlation Tank-Bo Correlation Tank/Jet-Bo Correlation 1.5 Jet-Bo Correlation Tank-Bo Correlation Tank/Jet-Bo Correlation

0.5

FBot

014 . + 0.76 We = 1 + 0.013 Bot

0.082 + 0.72 We 1 + 0.062 Botj

0.062

0.5

1.5

Flow Characterization Parameter F

FBotj =

0.086

FBo j =

0.028 + 0.66We 1 + 114 . Bo j

0.121

Fig. 8. Computational non-dimensional geyser height for both tanks as a function of the three Bond number correlation versions of Table 10 (lines are linear fit to each correlations data set). Original Tank Correlations Applied to Enlarged Tank Again, the correlations developed from the original tank computational data were applied to the enlarged tank. This was done to give some indication of the applicability of the correlations developed from a test tank to a different size tank. Table 11 shows the results, this time with the Bot correlation yielding an RMS error reduction of 38% over the Boj correlation.

Correlations Based on Data from Both Tanks Applied to Both Tanks The computational data from both tanks was used to fit all three correlations. Table 10 and Fig. 8 present the performance of these correlations. All three correlations seem to provide the desired direct prediction of geyser height, (G = F). However, an error reduction of approximately 30% can be noted for either the Bot or the Botj correlations compared to the Boj correlation.

9 American Institute of Aeronautics and Astronautics

Table 11. Original tank computational data correlations applied to the enlarged tank.
Corelation Tank-Bond Number Tank/JetBond Number Jet Bond Number Expression RMS Error 0.123

SUMMARY AND CONCLUSIONS

FBot =
FBotj =

0.25 + 0.95We 1 + 0.023 Bot

0.20 + 0.96 We 1 + 0.081 Botj

0.148

FBo j =

0.014 + 0.73We 1 + 0.71 Bo j

0.200

Enlarged Tank Correlations Applied to Original Tank Likewise, the enlarged tank correlations were applied to the original tank with similar results, (Table 12). Table 12. Enlarged tank computational data correlations applied to the original tank.
Corelation Tank-Bond Number Tank/JetBond Number Jet Bond Number Expression RMS Error 0.145

Design of future spacecraft propellant management systems employing a Thermodynamic Vent System (TVS) to suppress tank self-pressurization will require prediction of geyser formation in the propellant pool. Although an experiment-based correlation exists for predicting TVS jet-induced flow patterns in a propellant tank, the data on which this correlation has been developed is very limited due to the high cost of experimentation in a reduced-gravity environment. Although very challenging, computational simulation provides a very attractive alternative to experimentation for gaining insight into the geyser formation process and for developing improved correlations for design. The existing correlation assumes that the geyser height can be predicted if the jet-Bond number and the jetWeber number are known. To begin the present study, the limited data set of drop-tower experiments was simulated using the ECLIPSE code. Comparison between experiment and simulation show a consistent underprediction of geyser height within a relatively narrow error band. A sensitivity study was undertaken to determine if the underprediction was due to an unexpectedly strong dependence of simulated geyser height on user-specified parameters that may not be precisely known for the experiments and also may not be well know for real applications. This study showed that geyser height is nearly independent of contact angle. Although it depends on the value of surface tension coefficient and the jet inlet turbulence, geyser height is not unexpectedly sensitive to these parameters. To more fully populate the Bo-We space on which dimensionless geyser height depends, twelve additional computation-only data points were acquired, (using ECLIPSE), to evaluate the predictive capability of proposed correlations. All of the available experiment data is for configurations with the same ratio of tank diameter to jet diameter. Therefore, another set of computation-only data points was acquired, (using ECLIPSE), for a larger tank/jet ratio to better evaluate the range of applicability of the proposed correlations. A re-evaluation of dimensionless parameters and correlations previously used to predict geysers led to an investigation into reformulation of the non-dimensional geyser height and the Bond number. Experimental and computational data were employed to reformulate geyser height prediction correlations. The quality of a correlations prediction was judged by the RMS error between predicted geyser height and either the measured or computationally simulated geyser height.

FBot

014 . + 0.76 We = 1 + 0.013 Bot

0.082 + 0.72 We 1 + 0.062 Botj

FBotj =

0.152

FBo j =

0.028 + 0.66We 1 + 114 . Bo j

0.223

Bond Number Study Summary As can be seen from Tables 7-12, the Bot and Botj correlations consistently outperform the Boj correlations in prediction of geyser height. This holds true for both the experimental and the computational data sets. Figure 7 shows the improved predictive capability with the best performance produced by the Botj correlation.
F = 0.6 + 1.86 We 1 + 0.079 Botj

(9)

Based on the data and analyses presented, it is recommended that this new correlation be used for the design of future spacecraft propellant management systems.

10 American Institute of Aeronautics and Astronautics

Although a new formulation of the non-dimensional geyser height was proposed after a re-examination of the physics of geyser formation, correlations base on this new parameter showed consistently worse performance in predicting geyser height than the existing correlation. In only one instance did the results of the new formulation match the original correlations predictive capability. It is therefore concluded that the existing non-dimensional geyser height results in superior predictive capability and should be used for future work. In contrast with the failure of the dimensionless geyser height reformulation, correlations based on a new formulation of the Bond number show great promise. Correlations were developed based on three different formulations of the Bond number. The existing correlation used the jet-Bond number, Boj , in which the characteristic length is the diameter of the jet at impact with the free surface. The first alternate formulation uses the tank diameter as the characteristic length to produce the tank-Bond number, Bot. The characteristic length of the second alternate formulation uses a combination of the two, (

the design of future experiments and, where required, for future spacecraft.
REFERENCES

1.

Aydelott, J.C. Axial Jet Mixing of Ethanol In Cylindrical Containers During Weightlessness. NASA TP-1487, 1979. Aydelott, J.C. Modeling of Space Vehicle Propellant Mixing. NASA TP-2107, 1983. Wendl, J.C., Hochstein, J.I., and Sasmal, G.P. Modeling of Jet-Induced Geyser Formation in a Reduced Gravity Environment. AIAA-91-0801, 1991. Wendl, M.C. "Jet Induced Mixing of Propellant in Partially Filled Tanks in a Reduced Gravity Environment." Magisterial Thesis. Washington University, St. Louis, Missouri, 1990. Kothe, D.B., Mjolsness, R.C., and Torrey, M.D. RIPPLE: A Computer Program for Incompressible Flows with Free Surfaces. LA12007-MS, 1991

2.

3.

4.

Dt D j ), to produce a

5.

tank/jet-Bond number, Botj . Correlations based on these alternate formulations consistently outperform the correlations based on the original parameter, Boj . Perhaps most striking is the 35% reduction in predictive error when the Botj correlation is compared to the original correlation for the experiment data on which the original correlation was based. It is hoped that the correlations based on small-scale experiments will be useful for predicting the performance of a TVS in an actual spacecraft tank. To test the sensitivity of the correlations predictive accuracy to changes in tank size, correlations with coefficients computed for the experiment tank were used to predict geyser formation in a larger diameter tank, and vice-versa. In both cases, the new Bond number formulations significantly outperform the existing formulation. Future work should focus on increasing the quantity of reliable data on geyser formation in a reduced gravity environment. Experimental data will be expensive and time consuming to obtain. If the quality of the computational simulation can be further improved, it may provide a less expensive path to acquiring the necessary information. The new correlations developed during the present study show clearly improved performance for the available data. These correlations, and the ECLIPSE code, are currently the best available tools for predicting geyser formation in a reduced gravity environment. As such, they should be used for

11 American Institute of Aeronautics and Astronautics