two-pendulum model of propellant slosh in europa …...t f aws 06)&Ã two-pendulum model of...
TRANSCRIPT
TFAWSMSFC ∙ 2017
Two-Pendulum Model of
Propellant Slosh in Europa
Clipper PMD TankWanyi Ng & David Benson,
NASA GSFC 597
Presented By
Wanyi Ng
Thermal & Fluids Analysis Workshop
TFAWS 2017
August 21-25, 2017
NASA Marshall Space Flight Center
Huntsville, AL
TFAWS Interdisciplinary Paper Session
https://ntrs.nasa.gov/search.jsp?R=20170007452 2020-03-31T05:03:21+00:00Z
Outline
• Objective
• Background
• Results and literature verification
– Mass
– Frequency
– Damping ratio
– Hinge location
• Conclusions
2
Objective
Model propellant slosh for Europa Clipper
using two pendulums such that controls
engineers can predict slosh behavior
during the mission.
3
BACKGROUND
4
Motivation
• Importance of predicting propellant slosh
– Sloshing changes CM (center of mass) of spacecraft and exerts
forces and torques on spacecraft
– Avoid natural frequencies of structures
– Size ACS (Attitude Control Systems) thrusters to counteract forces
and torques
• Can model sloshing fluid as two pendulums with specific
parameters (mass, length, damping)
5
Background
• Europa Clipper tanks
– Bipropellant system
– Cylindrical with domed top and bottom
– 8-vane PMD (propellant management device)
• CFD (computational fluid dynamics) data
used as “real” slosh behavior
– Have data for two propellants at three fill
fractions each
– Initial condition of 15 degree free surface
offset, released and allowed to settle
– CFD requires long computing time -> Need a
computationally simple model
6
Notional tank
and PMD
CFD Simulation
Background
• Pendulum model
– Model fluid movement as two pendulums
attached to central axis of the tank
– For each CFD data set, find parameters:
mass, frequency, damping ratio, attachment
height
7
𝑚𝐿 ሶ𝜃2
−𝑚𝐿 ሷ𝜃
𝑚𝑎
Forces exerted on
tank by fluid
𝐶𝑀 𝑡 = 𝑚𝐿𝑠𝑖𝑛𝜃 𝑡
= 𝑚𝐿𝑠𝑖𝑛 𝜃0𝑒−𝜉𝜔𝑡
𝜉𝜔
𝜔 1 − 𝜉2sin 𝜔 1 − 𝜉2 𝑡 + cos 𝜔 1 − 𝜉2 𝑡
Existing Literature
• SP-106 (1966), SwRI (2000): Analytical equations and empirical correlations for damping and frequency
– Includes bare cylindrical (no PMD), sector, and annular tanks
• Cassini slosh paper (1994): Two pendulum model
– Slosh around PMD was modeled as combination of sector and annular slosh modes
– Two separate pendulums to model two slosh modes
– Static mass component at bottom that experiences little movement
8
Cassini paper illustration of
double pendulum model
Annular tank
mode (top view)
Sector tank mode
(top view)
Tank
Wall
PMD
METHODS OVERVIEW
9
Generate CFD Data
10
• Propellants: NTO and MMH
• Fill fractions: 25%, 50%, 85%
• Data: CM, Force, Moment (all 3 axes)
Find Initial Guesses
• Curve fitting by finding
parameters in
pendulum equation that
most closely match
CFD
• Trying to resolve CFD
into two pendulums
• Peak-to-peak values ->
• Initial guesses for
damping and frequency
of each pendulum
• Note much higher
damping before first
peak
11
Find Parameters to Fit CM Data
• Matlab’s fsolve(x) ->
• Mass, damping, and frequency parameters to fit CMx CFD data
• Refine and iterate
12
Compare Sum of Pendulums to CFD Data
13
• Sum of two pendulums generates model for propellant slosh
• Should match both CM and Force data
Mean Error in Force
• Metric to quantify accuracy of fit: mean absolute difference between CFD force and pendulum model force
1
𝑛
1
𝑛
𝑎𝑏𝑠 𝐶𝐹𝐷 − 𝑝𝑒𝑛𝑑𝑢𝑙𝑢𝑚
• Select methods that minimize this14
RESULTS AND LITERATURE
COMPARISON
15
Basis for results
• Coordinate system – origin at top
of tank
• Parameters prioritized fitting the
behavior after the first peak
• Two pendulum model is an
approximation only
– PMD does not create a perfectly
sector nor annular tank and is only a
fraction of tank height
– Parameters not constant over time
– Model does not scale well with high
fluid displacements
16
z
xy into page
x
Approximate
shape of PMD
vanes
Mass Participation Fraction
17
0.0480.052
0.145
0.03 0.0290.018
00.020.040.060.08
0.10.120.140.16
0 0.25 0.5 0.75 1
Mas
s fr
acti
on
Fill fraction
Mass participation fraction vs. fill fraction
NTOPendulum 1
MMHPendulum 1
NTOPendulum 2
MMHPendulum 2
• Pendulum mass as a fraction of total fluid mass
• Monotonic trends
• Mass fractions are identical between NTO and MMH
• Piecewise linear fit– First two fill fractions – fluid partially submerges PMD, sloshing occurs between
vanes
– Last fill fraction – fluid completely submerges PMD, different slosh behavior
Frequency
18
0.71190.6575
0.36
0.1831
0.296 0.3322
00.10.20.30.40.50.60.70.8
0 0.25 0.5 0.75 1
Freq
ue
ncy
(ra
d/s
)
Fill fraction
Frequencies vs. Fill Fraction
NTO Sector
MMH Sector
NTO Annular
MMH Annular
• Function of pendulum’s length and acceleration
• Monotonic trends
• Frequencies are identical between NTO and MMH
• Frequencies for the two pendulums converge as fill fraction increases– Sector and annular slosh modes become less distinct as PMD becomes
fully submerged
Frequency - Literature Comparison 1
19
• Left: Cassini paper referenced SP-106 for an analytical equation for slosh frequency in a bare tank (cylindrical tank with no PMD) and compared it to the frequencies of their two pendulums
• Right: Similar trends to Cassini found in Europa pendulum model frequencies
• Sector and annular slosh modes converge towards bare tank frequency as PMD becomes more submerged (fully submerged at 85% fill fraction for Europa tank)
Cassini Paper Frequencies vs. Fill Fraction
(Bare Tank)
(Annular Tank)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 0.25 0.5 0.75 1
Freq
uen
cy (
rad
/s)
Fill fraction
Frequencies vs. Fill Fraction, Comparing to SP-106 Bare Tank
NTO Sector
MMH Sector
SP-106 AnalyticalBare Tank
NTO Annular tank
MMH AnnularTank
Frequency – Literature Comparison 2
• SP-106 references tables (Bauer, 1963) for an analytical equations for sector and annular slosh frequency
• Function of acceleration, geometry, and fluid height• Pendulum frequencies are close to analytical equation frequencies
• Differences between analytical and pendulum fits due to:– PMD is not exactly a sector/annular tank
– Half-dome bottom approximated as flat bottom – at 25% fill fraction, sloshing fluid is almost entirely in the dome
– PMD doesn’t include entire height of tank – at 85% fill fraction, PMD is completely submerged20
0
0.2
0.4
0.6
0.8
1
0 0.25 0.5 0.75 1
Freq
uen
cy (
rad
/s)
Fill fraction
Frequencies vs. Fill Fraction,Comparing to Analytical Sector and Annular Tanks
NTO Sector
MMH Sector
NTO Annular tank
MMH Annular Tank
SP-106 Analytical Sector Tank
SP-106 Analytical Annular Tank
Damping Ratio
• Monotonic trends
• Slightly higher damping ratio for higher dynamic
viscosity (MMH)
21
00.05
0.10.15
0.20.25
0.30.35
0.4
0 0.25 0.5 0.75 1
Dam
pin
g R
atio
Fill fraction
Damping Ratio vs. Fill Fraction
NTOPendulum 1
MMHPendulum 1
NTOPendulum 2
MMHPendulum 2
• Mikishev and Dorozhkin found correlation for damping in a bare tank
• Function of geometry, acceleration, viscosity, and fluid height
• Scales by correction coefficient for domed bottom
• Pendulum damping within order of magnitude of analytical prediction
• Pendulum damping less sensitive to viscosity than analytical prediction – viscous vs. drag forces
22
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 0.25 0.5 0.75 1
Freq
uen
cy (
rad
/s)
Fill fraction
Damping Ratio vs. Fill FractionCFD Fits and Analytical
NTO Pendulum 1
MMH Pendulum 1
NTO Pendulum 2
MMH Pendulum 2
NTO SwRI TheoreticalBare TankMMH SwRI TheoreticalBare Tank
Damping Ratio – Comparison 1
Length and Hinge Location
• Origin is top of tank
• Pendulum bobs stay within fluid
• Monotonic values for pendulum heights
• NTO and MMH heights are close but not identical23
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.25 0.5 0.75 1
Hin
ge H
eigh
t (m
)
Fill fraction
Hinge height vs. fill fraction
NTO Pendulum 1
MMH Pendulum 1
NTO Pendulum 2
MMH Pendulum 2
NTO Static Mass
MMH Static Mass
Length and Hinge Location
24
NTO 25% fill
MMH 25% fill
NTO 50% fill
MMH 50% fill
NTO 85% fill
MMH 85% fill
Approximate
tank wall
Pendulum at
15 degree
offset
PLOTS COMPARING
PENDULUM MODELS
AND CFD DATA25
NTO 25% Fill Fraction
26
NTO 25% Fill Fraction
27
Zoom:
NTO 25% Fill Fraction
28
Zoom:
NTO 25% Fill Fraction
29
NTO 50% Fill Fraction
30
NTO 50% Fill Fraction
31
Zoom:
NTO 50% Fill Fraction
32
Zoom:
NTO 50% Fill Fraction
33
NTO 85% Fill Fraction
34
NTO 85% Fill Fraction
35
NTO 85% Fill Fraction
36
NTO 85% Fill Fraction
37
Summary of Parameters
NTO (nitrogen tetroxide) MMH (monomethyl hydrazine)
25% fill 50% fill 85% fill 25% fill 50% fill 85% fill
Mass fraction1 0.048 0.052 0.145 0.048 0.052 0.145
Mass fraction 2 0.03 0.029 0.018 0.03 0.029 0.018
Mass 1 (kg) 20.09 44.49 210.87 12.12 26.69 126.53
Mass 2 (kg) 12.56 24.81 26.18 7.58 14.89 15.71
Frequency 1 (rad/s) 0.1831 0.296 0.3322 0.1831 0.296 0.3322
Frequency 2 (rad/s) 0.7119 0.6575 0.36 0.7119 0.6575 0.36
Damping Ratio 1 0.34 0.105 0.035 0.35 0.11 0.037
Damping Ratio 2 0.015 0.022 0.035 0.02 0.025 0.037
Hinge Height 1 (m) 0.9 -0.4 -0.5 0.9 -0.5 -0.5
Hinge Height 2 (m) -1.0 -0.7 -0.3 -0.9 -0.7 -0.2
Static Mass Height
(m) -1.12 -0.99 -0.79 -1.14 -0.99 -0.8
Mean Force Error
from t=0 0.0716 0.075 0.1055 0.0398 0.0447 0.0679
Mean Force Error
from First Peak 0.0241 0.018 0.0775 0.0118 0.0119 0.0518
38
CONCLUSIONS
39
Accuracy of Fit
40
• Two-pendulum model can accurately capture either before or
after first peak
• High confidence on frequencies except 85% fill pendulum 2
• Moderate confidence on mass, damping, and hinge location
– Sometimes several sets of parameters could have provided good matching to
CFD
– Selected parameters that made physical sense
• Model parameters may reflect inaccuracies in CFD
• Pendulum model does not scale well for high fluid disturbance
angles
• Damping is actually a function of time and distance traversed by
moving fluid
– Pendulum model assumes damping is constant over time
Observations to Note
• Small initial fluid displacements: Changes have little impact on long-term CFD results
• Large initial displacements: behavior differs drastically
41
Observations to Note
42
• Changing density (NTO vs MMH) only slightly changes damping, has little impact on CFD results
Areas for Further Investigation
• Find literature to support mass fraction parameters
• Potentially to capture first peak – add third pendulum
with damping ratio of one
• Validate with more CFD data:
– At intermediate fill fractions
– At different initial fluid offset angles - 5 degree offset is more
conservative than 15, will be used for deliverable in May
• Validate with experiments
43
44
Thank You
Sources
• Abramson, N.H.: The Dynamic Behavior of Liquids in
Moving Containers. NASA SP-106, 1966
• Bauer, H.F.: Tables and Graphs of Zeros of Cross Product
Bessel Functions. MTP-AERO-63-50 NASA-MSFC, June
1963
• Dodge, F.T.: The New “Dynamic Behavior of Liquids in
Moving Containers”. Southwest Research Institute, 2000
• Enright, P.J. and Wong, E.C.: Propellant Slosh Models for
the Cassini Spacecraft. AIAA-94-3730-CP, 1994
45