Upload
hangoc
View
220
Download
0
Embed Size (px)
Citation preview
American Institute of Aeronautics and Astronautics
1
Enhanced Numerical Modeling in Simulation of a Generic
Propellant Tank Slosh Baffle
Sunil Chintalapati1, Joel M. Faure
2, Ran Zhou
3, Michael Vergalla
4, Daniel R. Kirk
5, and Héctor Gutiérrez
6
Florida Institute of Technology, Melbourne, Florida, 32901
The slosh dynamics in cryogenic fuel tanks under microgravity is a pressing problem that
severely affects the reliability of spacecraft launching. An accurate prediction of the slosh is
critical for successful mission planning and may influence vehicle control and positioning
during rendezvous, docking, and reorientation maneuvers. This paper defines a novel
method to assess this problem by coupling capabilities of ANSYS FLUENT and
MathWorks-MATLAB. While FLUENT solves the fluid dynamics aspect of problem,
MATLAB solves for dynamics part of motion profile and update CFD with new inputs every
new time step. Experiments performed on the Florida Tech 2-DOF motion table provided a
basis for validating the new method implemented in CFD simulation. Present work focuses
on enhancing the capabilities of CFD tool by reducing coding complexities due to inclusion
of 6-DOF dynamic mesh. The idea is to incorporate resulting forces and other multiple
inputs into FLUENT through MATLAB rather than a user defined function. MATLAB can
independently compute the force feedback due to fluid forces and tank motion and
incorporate it into simulation in the next incremental time-step.
Nomenclature
acv = Average cell volume, m3
Bo = Bond number, ratio of body to surface tension forces, gR2/
bh = Height of baffle relative to bottom of Tank, m
bt = Baffle thickness, m
bw = Baffle width, m
CFD = Computational Fluid Dynamics
Δt = time step
DOF = Degrees of Freedom
FLEVO = Facility for Liquid Experimentation and Verification in Orbit
g/g0 = Gravitational acceleration ratio relative to surface of Earth gravity 9.81 m/s²
NIST = National Institute of Standards and Technology
Re = Reynolds number, ratio of inertial to viscous forces, VD/
t = time, sec
Td = Tank diameter, m
Th = Tank height, m
tr = Computational run time
UDF = User designed function
VOF = Volume of Fluid
= Angular Velocity, rad/s
θ = Angular Orientation, rad
1 Undergraduate Research Assistant, Mechanical and Aerospace Engineering, 150 West University Blvd, Student Member AIAA. 2 Graduate Research Assistant, Mechanical and Aerospace Engineering, 150 West University Blvd, Senior Member AIAA. 3 Graduate Research Assistant, Mechanical and Aerospace Engineering, 150 West University Blvd, Student Member AIAA. 4 Graduate Research Assistant, Mechanical and Aerospace Engineering, 150 West University Blvd, Student Member AIAA. 5 Associate Professor, Mechanical and Aerospace Engineering, 150 West University Blvd, Senior Member AIAA. 6 Associate Professor, Mechanical and Aerospace Engineering, 150 West University Blvd.
46th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit25 - 28 July 2010, Nashville, TN
AIAA 2010-6976
Copyright © 2010 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
2
I. Introduction and Background
losh is a pressing problem for spacecraft stability and control. The motion of the vehicle may excite propellant
remaining inside a tank, and reaction forces and moments caused by slosh can degrade the pointing accuracy of
the system 18, 19
. For example, in preparation for orbital insertion of the payload, the upper-stage of a rocket
undergoes a series of maneuvers, which may lead to large amplitude sloshing motion of the propellants. Liquid
propellant reaching the relief and orbital control vents may result in a significant increase in expelled mass which
may cause mission failure due to loss of the proper orbital attitude. As another example delicate docking maneuvers
between spacecraft and space stations may also be impacted by liquid slosh motion. Although baffles add to the
weight of the tank, they play a vital role in mitigating undesired slosh induced motion. Slosh baffles located within
the Space Shuttle external tank are shown in Figure 1, and Figure 2 shows an example of a baffle used by Armadillo
Aerospace to damp oscillations observed during test flights on a vehicle under development 2.
Figure 1. Space Shuttle External Tank Slosh Baffle 9
Figure 2. Slosh baffle used by
Armadillo Aerospace 2
Depending on the type of disturbance and tank shape, the liquid propellant can experience different types of
motion including simple planer, non-planer, rotational, symmetric, asymmetric, quasi-periodic and chaotic. In low-
gravity, surface tension and capillary action may dominate even in large booster size tanks and the liquid may be
oriented randomly within the tank depending upon the wetting characteristics of the tank wall 11
.
Numerical solution to slosh problem has been emerging in recent times, owing to the major advances in
computational capabilities. CFD models to make slosh predictions during the high acceleration ascent phases of a
rocket have been used8, although very little work has been done in cases of very-low accelerations when the vehicle
is in space 16, 19
. Another approach for predicting slosh motion is to use scaled model testing, such as that done at
Southwest Research Institute, but thus far the results are largely qualitative and there has not yet been direct data
comparison with detailed CFD models. Other studies have focused on analyzing available flight data to identify
conditions leading to mission failure. The FLEVO project, under the direction of the National Aerospace Laboratory
(The Netherlands) has been the most substantial effort devoted to fill the gap between numerical simulations 14, 16
,
and the development of an experimental framework to measure and characterize slosh under microgravity 19
.
Earlier published paper identifies the physics and criteria for evaluation of slosh baffle performance in regular
and microgravity 4. The paper used simple dam break profile to induce slosh and evaluate the baffle performance.
Advances in CFD have resulted in the ability of FLUENT to compute sloshing modes by a 6-DOF motion profile
with inclusion of a UDF. This was performed by slosh dynamics team at Florida Tech in development of a 6-DOF
dynamic mesh numerical model that is able to replicate slosh dynamics in regular gravity 17
. Advances have also
resulted in emergence of numerical solver like Flow-3D, oriented specifically for solving sloshing problems5.
Recent enhancements are multi-physics solver, capable of solving slosh coupled with force feedback from both bulk
fluid movements transferred to the vehicle and from the vehicle motion due to the engines. One such example would
be FLUENT solving for fluid dynamics aspect of the problem and ANSYS Mechanical solving for dynamics of
motion profile updated at every time-step based on both motion of the vehicle and sloshing forces from cryogens
inside the tank.
This paper utilizes computational fluid dynamics (CFD) to assess the performance of a generic propellant tank
baffle in low gravity condition. This is done by simulating a possible motion profile with arbitrary acceleration and
angular velocities, which would induce violent slosh inside the tank. This paper defines a new proof of concept, a
link between MATLAB and FLUENT1. This link is pertinent to enhance the modeling capabilities of CFD by
including complex motion profile and additional inputs including heat transfer, phase change, force feedback, etc.
S
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
3
II. Experimental Setup
Florida Tech’s slosh platform has been in development for past couple of years. Since its inception, the design
has been constantly upgraded to keep discrepancies from creeping into experimentation. The goal of the slosh team
at Florida Tech is to reduce the redundancies that cannot be modeled in a CFD simulation, to have one to one
comparison between experiment and CFD. Figure 3 and Figure 4 shows the overview of the capabilities of slosh
platform. The platform is currently capable of 2-DOF, one being linear translation in x-axis (Figure 3) and other
being free rotation in y-axis (Figure 4). This paper focuses on validation of the linear translation motion along x-
axis, depicted on the left hand side of the picture.
Figure 3. Free Linear Translation Figure 4. Free Rotation Coupling
Data acquisition plays a pivotal role in slosh
experimentation because the assumptions that form
the input for CFD are derived from the experimental
results. The acceleration experienced by the tank and
captured by the accelerometers is used as the input
conditions for the CFD model. The other important
aspect of this work is image-capturing technique,
which plays an important part in comparing results
between experiment and CFD. The whole process of
initializing the experiment, data acquisition and
synchronized image capturing is controlled via
LabView and the data gathered is post processed in
MATLAB. Efforts are still in progress concerning
explicitly compare image data from experiments and
CFD. Preliminary results of which will be discussed
in the next section. Figure 5 shows the overall setup
for the motion table, data acquisition and the
experiment17
.
III. Computational Setup
Analysis for current paper used ANSYS FLUENT, version 12.11 for solving fluid dynamics aspect of the
problem. The 3D geometry was created and meshed using Gambit (version 2.4), preprocessor for FLUENT.
MathWorks-MATLAB15
implemented motion profile, some boundary conditions and post processing the results.
Sloshing in CFD is modeled using the VOF model. Transient Volume of the Fluid (VOF) method utilized is well
suited to multiphase flows involving two immiscible fluids. The method relies on the basis that the fluid state is
described in each cell by one value; the method introduces a function F whose value F=1 correspond to a cell full of
fluid, and F=0 to an empty cell, and a geo-reconstruct scheme is used to track the fluid-air interface 3. This section is
Figure 5. Motion Table Setup: Tank in Lighting
Control box, Work Station, Overview, LabView
instrumentation, and New Encoder Position Sensor
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
4
divided into three major components between grid sensitivity studies, computational setup for ANSYS FLUENT,
and computational setup for MATLAB
A. Grid Sensitivity Studies
Grid sensitivity check is one of the most important verifications for any computational study. The importance of
computational results is not only dependent on assumptions made in the model but also depend on variation of
results with different mesh densities. The ideal grid balance is to capture accurate physics while keeping the number
of cells to a minimum; thereby, reducing computational run-time. A series of simulations performed on a generic
domain with exact initial conditions but varying node spacing yielded the criteria for grid spacing for this current
paper. 1 below displays the parameters involved in conducting the grid sensitivity studies.
Table 1. Grid Sensitivity Study Matrix
GRID SENSITIVITY STUDY MATRIX
Grid # Interval Size (m) Total Cells tr (mins) acv (m3) Mesh Type
1 0.05 9000 75 1.31E-04 structured
2 0.025 52800 160 2.23E-05 structured
3 0.1 1575 45 7.48E-04 structured
4 0.01 825000 1800 1.43E-06 structured
5 0.0325 24150 90 4.88E-05 structured
6 0.0125 422400 900 2.79E-06 structured
The Td is 1.0 m and the Th is 1.5 m, and the
overall tank volume is 1.1775 m3. Coarse and fine
versions of initial node spacing provided results of
grid dependency of solution. Dam break model in
earth’s gravity was preferred because it would induce
vigorous slosh in the simulation. A simulation where
the column of fluid is patched and at time equals zero
is allowed to slosh freely in domain (see Figure 6).
Realizable k-epsilon (two equation turbulence) model
is used due to the violent slosh induced in this
simulation (Re is determined by Td and velocity of
the fluid interface). Properties of water at 297.15K
are taken from NIST13
and earth’s gravity is
9.81m/s2. The wall contact angle is 55 degrees
4.
Capillary effects are small due the gravity being a
dominant factor and thus Bo would be large in
magnitude.
Two methods were determined to assess the impact of grid on the solution. One is setting up pressure and phase
monitors at predefined locations in tank and other being comparing fluid interface at certain time with the help of
MATLAB. Three monitors were set up in domain before the start of simulation. Location of three monitors were set
on the interface of the column of water on and shown in Figure 6. The monitors were to ascertain the static pressure
and interface of the fluid.
Figure 7 shows the results of pressure monitors at location 1 for the different mesh densities considered. The
overall trend follows a similar pattern based on the same initial conditions and location of the monitor. The
difference in results due to mesh density is visible by overlap of data lines (picture on right in Figure 7). The results
from this plots show that data lines from Grid4 and Grid6 overlap each other and hence the results are grid
independent.
Figure 6. Location of monitor in grid study domain
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
5
Pressure vs. time at location 1 Close up of drop of pressure
Figure 7. Plot of pressure monitor at location 1
Figure 8 shows the plot of phase monitor with flow time. This monitor is a measure to track the fluid interface as
the flow passes through the monitor. Fluid interface has no units. At zero, the cell is completely air and at one, it is
completely water. At value 0.5 in phase panel in FLUENT determines the fluid interface. The plot on the left shows
the interface tracking by the monitor while the plot on the right shows the peak of the fluid interface. It is
determined from the plot that Grid4 and Grid 6 are still the best mesh densities to capture fluid interface.
Fluid interface vs. time at location 1 Close up of a spike in plot
Figure 8. Plot of fluid interface monitor at location 1
The third step followed in grid sensitivity check is looking at the overall fluid interface in the domain, since fluid
interface is a primary form of comparing experiments and CFD. A planar slice oriented towards the movement of
initial fluid interface is compared numerically with other mesh densities. Figure 9 shows the resulting difference due
to different mesh densities, performed in MATLAB. Series of picture on top are from grid 6 and series of picture in
the middle column are from grid 5. The last column is the difference between the two grids. The region of complete
black has no difference, but region of green color is the region where the actual difference between the two grid
exists. This technique was used for all the mesh densities considered; Figure 9 shows a typical comparison of fluid
interface between two different grids.
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
6
Grid 6
Grid 5
Difference
0.00 sec 0.50 sec 1.00 sec 1.50 sec 2.00 sec 2.50 sec 3.00 sec t
Figure 9. Numerical difference between Grid6 and Grid5
B. CFD Setup
All tank geometries created and meshed in Gambit, version 2.4. A structured O-grid scheme used in all tank
constructed with node spacing determined from grid sensitivity studies. The mesh was then imported into ANSYS
FLUENT. Pressure based, double precision, transient VOF model used in CFD. Second order discretization was
preferred for momentum and geo-construct scheme for fluid interface. Fluid is patched accordingly in domain with
respect to the problem being analyzed. Table 2 gives the dimensions and statistics for all domains used in this paper.
The required cell volume and node spacing criteria deviates from the grid resolution study for experimental tank due
to the need to validate the CFD method used in this paper. In earth’s gravity, sloshing cause wide spread fluid
splatter, droplets and highly chaotic fluid interface; thus, needs a lot of grid resolution to capture fluid splatter and
random droplets. The grid resolution for microgravity case is less dense due to the fact that fluid tends to be more
viscous, fluid splatter is minimal and thus the droplets in model not taken into consideration. The concern is more
for the bulk fluid capture which all three-grid resolution are adequate to capture and save computational run time at
the same time. Note that generic propellant tank has two baffles and thus two dimensions are noted in the third
column of Table 2.
Table 2. Domain dimensions and statistics
Grid Td (m) Th (m) bh (m) bt (m) bw (m) # cells acv (m3) Type
Grid-study-domain 1 1.5 na na na 422400 2.79E-06 structured
Experimental Tank 0.2413 0.2413 0.138 0.011176 0.01905 411110 2.68E-08 structured
Microgravity Tank 5 4 0.9 0.1 0.3
913176 6.18E-05 structured 1.9 0.1 0.3
Figure 10 shows the initial conditions for the three domains given in Table 2. The picture on left is the domain
for the grid sensitivity study; the picture in middle is the domain of the experimental tank with an initial water patch;
the picture on the right is a generic propellant tank based on a Delta IV Heavy upper stage LH2 tank with an initial
cryogen patch.
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
7
Domain for grid sensitivity
study with initial water patch
Domain of experimental tank
with initial water patch
Generic propellant tank with
initial cryogen patch
Figure 10. Domain overview
C. MATLAB Setup
The MATLAB-FLUENT link allows users to call FLUENT as a function, change the necessary input conditions
in FLUENT, and output results that can be used for various solutions in MATLAB. The current uses for this link
are: high fidelity rocket simulations, slosh dynamics, and automation routines for determining the aerodynamic
characteristics of external flows with the use of neural networks to optimize convergence. The raw experimental
acceleration data is imported in MATLAB, then filtered and stored as an array. A loop iterates time from 1e-3
seconds (Δt) to 10 seconds of experiment time in increments of 1e-3 seconds. The stored acceleration array is
linearly interpolated as a function of time and inputted into the MATLAB-FLUENT link for every time-step. Once
convergence criteria has been met in FLUENT all relevant data is saved, MATLAB then moves onto the next time
step and repeats the process until the loop ends.
IV. Experimental and Computational Results
The time for experimental and CFD simulation was set to 10 seconds. A motion profile for experimental tank
was set and the experimental acceleration data output from all accelerometers used as tank motion profile input for
CFD simulation. Figure 12 shows the motion profile for the CFD simulation recorded by accelerometers from the
experimental tank.
Figure 13 shows liquid surface contours of simulation results compared with liquid surface motion pictures
captured by the cameras. A black surface enclosure is set around the experiment table to disregard any background
disturbance to fluid interface in experiment. A bright light used for illumination of fluid from the bottom of cylinder,
helps in identifying the thickness of gradient of slosh waves. 1e-03 was the Δt used in simulation resulting in 10,000
time-steps. The tr was approximately 136 hours on a six core i7 (3.6 GHz processors) and the two additional cores
were used for MATLAB and post processing data. Saved case and data files from CFD are post processed with
automated journal in FLUENT 12.1.
Automation procedure for getting pictures alone resulted in 6-7 hours of computational time on a single 3.6 GHz
core machine. The image results from FLUENT lacks the sharp crisp quality compared to other professional post
processing software’s (Tecplot, CFD-Post), which is regrettable; but they serve the purpose here. Special note to the
kink in the middle of the tank (in Figure 13) that was unavoidable during rendering process in FLUENT. The
pictures from CFD were extrapolated using an automated journal file and hence the all the CFD pictures are aligned
to positive x-axis or slightly towards x-axis. In comparison to experiment pictures which have a slight angle noted
by the cross-wires on the top surface of tank.
Figure 11 shows the sample of the quantitative analysis done on the sample of results from experiment and CFD.
Three metric are defined to assess the difference between experiment and CFD. They are peak height: which is the
height of the peak of the wave at certain time. Axis height: which is the height of the fluid on the center axis of tank
and Wetted wall: which is height of the fluid on the corner of the tank. Sample results on the top (1.9 second) are
from one of worst comparable sample. The experiment slice has two peaks while the CFD slice has only one bulk
peak. Thus, the difference for the missing second bulk peak comes to about 19%, but on comparison to the peak
below, the difference comes to 4%. Note this is one such sample where the difference between CFD and experiment
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
8
was at its maximum. Comparison with a good case at 3.9 seconds, the difference between peak height is around 4%,
while axis height is around 3% and the wetted wall comes to around 3% too.
TYPE EXP(in) CFD(in) Difference (%)
Axis height 8.79 7.13 18.92
7.36 7.13 3.23
Peak height 8.71 7.69 11.69
7.39 7.69 -4.09
Wetted wall 1.48 1.43 3.64
Evaluation of results from CFD and Experiment at 1.9 sec
TYPE EXP(in) CFD(in) Difference(%)
Axis height 6.91 6.72 2.75
Peak height 8.11 7.79 3.95
Wetted wall 1.00 0.97 2.63
Evaluation of results from CFD and Experiment at 3.9 sec
Figure 11. Quantitative analysis of results from experiment and CFD
As seen in Figure 13, the series of pictures show the results obtained from the experiment and CFD. The bulk
fluid behavior and fluid interface in general agree very well with each other. There are some slight discontinuities in
some slices of simulation; for example, the height of slosh wave obtained in CFD at 1.6 sec is slightly lower than
experimental slice at the same time. The slice at 1.85 second shows the overall bulk fluid motion capture after the
wave hits the baffle. Experiment at this point of time shows a response wave of bulk fluid peak with a partial second
peak on top of the bulk fluid with relatively less fluid. In CFD, we see that the response wave is compromised of
bulk peak with relatively less splatter. Experimental slices 1.3, 2.1, 2.2, 2.9, 6.4 and 6.9 seconds shown in the above
figure agree well with CFD, especially 3.9 where CFD is able to capture the little distinctive hump on the slosh
wave in the experiment.
The assumptions that go into CFD are precise than assumption and discrepancies that might creep into
experiments. The goal of this paper is not to accurately match experiment with CFD but rather create a foundation
for simple modeling enhancements to provide overall accurate answers to the slosh problem.
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
9
Figure 12. Input motion profile for CFD simulation
A: t=1.3 sec B: t=1.6 sec
C: t=1.9 sec D: t=2.1 sec
E: t=2.5 sec F: t=2.9 sec
G: t=3.6 sec H: t=3.9 sec
I: t=6.4 sec J: t=8.8 sec
Figure 13. Bulk fluid and interface comparison between experiment and CFD
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
10
V. Slosh Baffle Performance in Microgravity
Results from the experiment and CFD are in good agreement with each other for bulk motion and fluid interface.
This prompted the need to expand on the capabilities of enhancing the capabilities of modeling slosh baffle
performance in microgravity conditions using the UCAT-FLUENT link. One such simulation was identified to
provide a sample for the capabilities of this new method.
A. Simulation #1
The first simulation is a motion profile in microgravity with tank spinning and moving from a location to
another. The domain tank used in this simulation is based on Delta IV heavy LH2 tank (5 m diameter). The picture
on extreme right in Figure 10 shows the domain used in microgravity simulation. A moving reference frame model
was used on FLUENT side to accomplish this task. The principle behind the model is given by excerpt below, The principal reason for employing a moving reference frame is to render a problem, which is unsteady in the stationary
(inertial) frame steady with respect to the moving frame. For a steadily rotating frame (i.e., the rotational speed is
constant), it is possible to transform the equations of fluid motion to the rotating frame such that steady-state solutions are
possible1.
Simulation time was combination of two
assumptions. The motion profile implemented by
MATLAB for the first 30 seconds of simulation,
followed by a period of no motion input where the
fluid is allowed to re-orient itself in microgravity
condition. The computational run-time was massive
owing to the nature of the problem.
The angular velocity vector inside FLUENT is in
the inertial frame. In order to get slosh inside the tank
it is necessary to induce a jerking motion. The current
model accomplishes this by placing the tank on a
rotating frame and having variable angular
acceleration. The initial position of the tank is along
the -Y-axis in the inertial frame, the tank rotates along
the Z-axis in a counter-clockwise direction (for the
first angular velocity component). The tank’s
rotational motion along the inertial frame begins to
decelerate after it has reached the positive X-axis it
decelerates to zero angular velocity in one second. The
angular velocity is split up into two components time
varying angular velocity along the ―Z-axis‖ in the
inertial frame and a constant angular velocity along the
―y’-axis‖ in the body frame, both as seen in Figure 14
FLUENT requires the angular velocity inputs in the following form: magnitude, and unit vector in the inertial
frame. The second angular velocity component needs to transform to the inertial vector as seen in Eq. 110
.
1
12
12
1
1
2
1
211
11
1
21 cos
sin
0
cos
sin
0
0
0
0
100
0cossin
0sincos
0
0
(1)
Qualitative results of this simulation are shown below in Figure 15 with flow time in simulation. Orientations of the
pictures are different and are not on same axis. Fluid used is liquid hydrogen and gaseous hydrogen forms the other
phase. Laminar flow model is used, based on the Reynolds number with diameter of tank as length. Hydrogen
properties taken from NIST considered are at 18 K and 21 K12
. Gravity is 0.00981m/s2
in negative Y direction and
the wall contact angle is 2 degrees4. Surface tension is 0.0019N/m
7. The effect of Bond number on the outcome of
the slosh is more dominant in a microgravity environment. The tr for the current simulation at 27 seconds is 15 days
on a six-core machine. Present results shows the motion profile input re-orients the fluid towards the upper region of
Figure 14. Inertial and body coordinate frames
for Simulation #1
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
11
tank, thus prompting the need to have a delicate maneuvering of tank. A slow and stable motion profile would not
impart significant momentum to the interior fluid. The latter half of simulation showing the liquid re-orientation is
unavailable at this moment owing to the immense amount computational time needed for current simulation;
therefore, the results from latter half will be presented at the conference. The implementation of complex motion
profile through MATLAB was successful, demonstrating the link between ANSYS FLUENT and MATLAB and
enhancement in numerical modeling.
10 sec 15 sec
20 sec 25 sec 30 sec ( end of motion profile)
Contours of Liquid Hydrogen
Figure 15. Motion profile (including angular and rotational components) in microgravity environment
VI. Conclusion
This paper utilizes the capabilities of ANSYS FLUENT and MATLAB to implement motion profiles into tanks
and to assess the performance of slosh baffle in microgravity. Importance of grid resolution study is emphasized;
methods to find the critical balance between grid independent results and lower cell count are charted out.
Validations of enhancement provided by the link are successful against ground slosh experiments. The bulk fluid
behavior and fluid interface results from the validation case agreed well with the experiment case. Although the
author regrets the rendering capabilities of ANSYS FLUENT, it still does not negate the fact the results provided by
the simulation are outstanding. Performance of a generic propellant tank with slosh baffles and the ability of the
UCAT-FLUENT link are put to test in a typical maneuver applicable for an orbital insertion. Input magnitude of
acceleration and velocities in simulation are exaggerated, to induce immense sloshing mode. Implementation of
motion profile through the link is successful.
The motivation for this work is not only to prove the link between MATLAB and ANSYS FLUENT, but also to
open number of enhancements this link could provide. A typical UDF has its own limitation; namely, only able to
triggering one specific enhancements in a simulation. UDF also needs an understanding of C++ code tailored to
ANSYS standards and lack parallelization in most instances.
Future work will take into account the possibility of adding a neural network component that could potentially
stabilize and control the simulation in real time. This could also lead to huge saving in computational run time by
modifying time-step size based on Courant–Friedrichs–Lewy condition (Courant number).
Acknowledgments
The authors wish to thank Dr. Paul Schallhorn and Laurie Walls of NASA John F. Kennedy Space Center
Launch Services Program for supporting this research effort.
The authors wish to thank Mr. Hao Jiang (Graduate research assistant) for checking the dynamics calculation for
microgravity case. The authors also wish to thank the exceptional undergraduate team (Amber Wright, Brittany
Essink, Namrata Dhingreja, Nicholas Pennington, and Richard Schul) for their assistance during the experiment on
2-DOF motion table.
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976
American Institute of Aeronautics and Astronautics
12
References
1. ANSYS FLUENT 12.1 Documentation. (2010). Retrieved July 15, 2010, from ANSYS FLUENT:
http://www.fluentusers.com/fluent/doc/doc_f.htm
2. Armadillo Aerospace. (2001-2010). Retrieved July 15, 2010, from Armadillo Aerospace:
http://www.armadilloaerospace.com/n.x/Armadillo/Home/News?news_id=344
3. best-practices-vof.pdf. (2006). Retrieved July 15, 2010, from ANSYS FLUENT:
http://www.fluentusers.com/best_prac_archives/best-practices-vof.pdf
4. Chintalapati, S., & Kirk, D. R. (2008). Parametric Study of a Propellant Tank Slosh Baffle . 44th AIAA/ASME/SAE/ASEE
Joint Propulsion Conference and Exhibit. Hartford, CT: AIAA.
5. Computational Fluid Dynamics Software | FLOW -3D. (200-2010). Retrieved July 15, 2010, from http://www.flow3d.com/
6. Dodge, F. T. et al.: "The new dynamic behavior of liquids in moving containers", Technical Report, Southwest Research
Institute, San Antonio, TX, 2000.
7. Engel, A. (Jan, 1962). An Instrumentation System for Temperature Profiling of Large Liquid Hydrogen Dewars. Nuclear
Science, IRE Transactions on , 9 (1), 243-246.
8. Faure, J., Vergalla, M., Zhou, R., Chintalapati, S., Gutierrez, H., & Kirk, D. (2010). Experimental Platform for the Study of
Liquid Slosh Dynamics Using Sounding Rockets. International Review of Aerospace Engineering , 59-66.
9. File:Sts et cutaway. (2007, February 17). Retrieved July 15, 2010, from Wikipedia:
http://en.wikipedia.org/wiki/File:Sts_et_cutaway.jpg
10. Ginsberg, Jerry H., Engineering Dynamics, Cambridge University Press, Cambridge University Press, 2008, pp. 99
11. Ibrahim, R., A., Liquid Sloshing Dynamics: Theory and Applications. Cambridge University Press, © 2005.
12. Isobaric Properties for Hydrogen. (2008). Retrieved July 15, 2010, from NIST Chemistry WebBook:
http://webbook.nist.gov/cgi/fluid.cgi?Action=Load&ID=C1333740&Type=IsoBar&Digits=5&P=1&THigh=25&TLow=16
&TInc=1&RefState=DEF&TUnit=K&PUnit=atm&DUnit=kg%2Fm3&HUnit=kJ%2Fkg&WUnit=m%2Fs&VisUnit=Pa*s&
STUnit=N%2Fm
13. Isobaric Properties for Water. (2008). Retrieved July 15, 2010, from NIST Chemistry WebBook:
http://webbook.nist.gov/cgi/fluid.cgi?Action=Load&ID=C7732185&Type=IsoBar&Digits=5&P=1&THigh=300&TLow=28
5&TInc=1&RefState=DEF&TUnit=K&PUnit=atm&DUnit=kg%2Fm3&HUnit=kJ%2Fkg&WUnit=m%2Fs&VisUnit=uPa*
s&STUnit=N%2Fm
14. Luppes, R., Helder, J. A. and Veldman, A.E.P.: "The Numerical Simulation Of Liquid Sloshing In Microgravity", European
Conference on Computational Fluid Dynamics, Delft, The Netherlands, 2006.
15. MATLAB 7.10.0.499 (R2010a), Simulink 7.5 (R2010a), The MathWorks Inc., 3 Apple Hill Drive, Natick, MA 01760-2098,
2010.
16. Veldman, A.E.P.: "The Simulation of Violent Free-Surface Dynamics At Sea And In Space", European Conference on
Computational Fluid Dynamics, Delft, The Netherlands, 2006.
17. Vergalla, M., Zhou, R., Gutierrez, H., & Kirk, D. (2009). Experimental and Numerical Framework for Characterization of
Slosh Dynamics. International Review of Aerospace Engineering , 52-31.
18. Vreeburg, J. P. B.: "Spacecraft Maneuvers and Slosh Control, IEEE Control Systems Magazine, June 2005.
19. Vreeburg, Jan P.B. and Chato, David, "Models for Liquid Impact Onboard Sloshsat FLEVO", NASA Technical Report
NASA/TM—2000-210475, November 2000.
Dow
nloa
ded
by F
LO
RID
A I
NST
ITU
TE
OF
TE
CH
NO
LO
GY
on
July
27,
201
4 | h
ttp://
arc.
aiaa
.org
| D
OI:
10.
2514
/6.2
010-
6976