Upload
others
View
7
Download
0
Embed Size (px)
Citation preview
MASTER'S THESIS
Implementation of a Computational FluidDynamics Code for Propellant Sloshing
Analysis
Tiago Rebelo2013
Master of Science (120 credits)Space Engineering - Space Master
Luleå University of TechnologyDepartment of Computer Science, Electrical and Space Engineering
CRANFIELD UNIVERSITY
TIAGO ALEXANDRE RAMOS REBELO
IMPLEMENTATION OF A COMPUTATIONAL FLUID DYNAMICS
CODE FOR PROPELLANT SLOSHING ANALYSIS
SCHOOL OF ENGINEERING
MSc in Astronautics and Space Engineering
(SpaceMaster)
MSc Thesis
Academic Year: 2012 - 2013
CRANFIELD UNIVERSITY
SCHOOL OF ENGINEERING
MSc in Astronautics and Space Engineering
(SpaceMaster)
MSc Thesis
Academic Year 2012 - 2013
TIAGO ALEXANDRE RAMOS REBELO
Implementation of a Computational Fluid Dynamics Code for
Propellant Sloshing Analysis
Supervisors: Ph.D. Jennifer Kingston
M.Sc. Manuel Hahn
August 2013
This thesis is submitted in partial fulfilment (45%) of the requirements for the degree of
Master of Science in Astronautics and Space Engineering
© Cranfield University 2013. All rights reserved. No part of this publication may be reproduced without the
written permission of the copyright owner.
Implementation of a Computational FluidDynamics Code for Propellant Sloshing Analysis
MSc Thesis
Tiago Rebelo
Supported by:
Supervisors:
M.Sc. Manuel Hahn - EADS Astrium Satellites
Ph.D. Jennifer Kingston - Cranfield University
Ph.D. Johnny Ejemalm - Lulea University of Technology
August 2013
i
This M.Sc. thesis is dedicated to those whose work, sweat and tears
allowed me to reach this point...
...to my beloved Parents
...to my inspiring Grandparents
iii
“Se todo em cada coisa.
Poe quanto es
No mınimo que fazes.”
- Fernando Pessoa
“Be everything in each thing.
Put all of yourself
Into the slightest thing you do.”
- Fernando Pessoa
Abstract
Liquid propellant sloshing inside spacecraft tanks is of crucial importance to the
dynamics of the space vehicle. The interaction of the disturbance forces and torques,
caused by the moving fuel, with the solid body and the control system, might lead
to an increase in the AOCS actuators commands, which can degrade the vehicle’s
pointing performances and, in critical cases, generate unstable attitude and orbit
control. Thus, it is of major importance to accurately predict the behaviour of liquid
propellants sloshing inside spacecraft tanks.
This M.Sc. thesis is focused on this topic, being its major objective the implementa-
tion of a CFD software in an existing EADS Astrium simulation environment. The
integrated simulation environment is used to assess the influence of liquid propellant
sloshing for specific satellite missions.
From a defined set of requirements an open source CFD software based on FEM is
chosen - Elmer. The software is integrated and the final simulation environment is
evaluated for sloshing purposes using three different sloshing test cases.
The first two test cases deal with rectangular and cylindrical laterally excited tanks
where comparators are available - the results of the tests are validated against nu-
merical and experimental results.
The final test case is defined to reduce the gap between the simple test cases per-
formed to validate the software and the real sloshing problems faced in space vehicles.
A typical liquid propellant tank is selected and real mission conditions are simulated.
The liquid sloshing inside the laterally excited tank is deeply studied, being fully
characterized. The simulation environment is validated for the implemented liquid
sloshing problems.
vi
Acknowledgements
To start with, I would like to express my deepest gratitude to my supervisor at EADS
Astrium, Manuel Hahn. I am heartily thankful for the given opportunity, the guidance, the
encouragement and the constant support. This gratitude is extended to the AOCS/GNC
& Flight Dynamics department of Astrium Satellites, Friedrichshafen, Germany.
Special gratitude goes to my supervisor at Cranfield University, Jennifer Kingston. Her
support and help during the development of this work, but also during my stay in Cranfield,
are not forgotten.
At the Lulea University of Technology my gratitude goes to Victoria Barabash, for her
support in the many different challenges experienced during these 2 years. Also, for his
supervision during the development of this thesis, my gratitude to Johnny Ejemalm.
To Prof. Wolfgang A. Wall from the Institute for Computational Mechanics of the Technical
University of Munich for allowing me to develop my work at his institute, my gratitude.
Special acknowledgement goes to ESA’s directorate of Human Spaceflight and Operations,
for providing a real liquid propellant sloshing problem that brought challenge and value to
this thesis.
For his support and very useful inputs in all matters related with Elmer, my gratitude goes
to D.Sc. Peter Raback from the CSC - IT Center for Science, Finland.
My gratitude to all the entities that financially supported my M.Sc. studies, namely: ESA
Human Spaceflight and Operations directorate, through a study Scholarship; Erasmus
and Erasmus Mundus grants from Lulea University of Technology and the SpaceMaster
consortium; and last but not least the very important support of EADS Astrium during
my internships.
A special thanks goes to Anna Guerman, for giving me the opportunity to learn from her.
Without her I would never have found the beauties of space nor integrated this Master’s
programme.
For those who joined me in this incredible SpaceMaster journey, my deepest gratitude - it
would not have been the same without them.
Without any disregard to all the amazing people I met during these years abroad, my
special gratitude goes to Mauro Aja Prado, Ishan Basyal and Dries Agten, for their true
friendship.
To my family, for their unconditional love and support throughout my life, my deepest love
and gratitude. Special thanks to my parents, Joao and Maria, for providing the conditions
that allowed me to develop and aim higher; and to my sister Mara, for her support and
belief at all moments.
Finally, I want to thank Rita for her love throughout our common life. She gave me the
courage and support to take this programme to its end. Without her I would never have
made it, my unconditional love and gratitude goes to her.
viii
Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviii
List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Sloshing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Lateral Sloshing . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Introduction to Damping . . . . . . . . . . . . . . . . . . . . . 11
2.1.3 Introduction to Non-linear Effects in Slosh . . . . . . . . . . . 12
2.1.4 Introduction to Micro-gravity Effects - Surface Tension . . . . 13
2.1.5 Other Types of Sloshing . . . . . . . . . . . . . . . . . . . . . 14
2.2 Computational Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Fluid Governing Equations . . . . . . . . . . . . . . . . . . . . 15
2.2.2 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.3 Meshing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.4 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.5 Numerical Analysis . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2.6 Solution Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
ix
Contents
3 Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1 Functional Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2 System Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4 CFD Software Selection . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.1 Selection Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.1.1 Available Codes . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.1.2 Satisfactory Codes . . . . . . . . . . . . . . . . . . . . . . . . 31
4.1.3 Top 3 Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.1.4 Final Selection . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2.1 Available Codes . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2.2 Satisfactory Codes . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.3 Top 3 Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.4 Final Selection . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5 Elmer - Open Source Finite Element Software . . . . . . . . . . . . 38
5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2 Models / Solvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3 Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.3.1 Graphical User Interface . . . . . . . . . . . . . . . . . . . . . 44
5.3.2 Command Line . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.4 Pre- and Post- Processing . . . . . . . . . . . . . . . . . . . . . . . . 47
5.4.1 Pre-Processing . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.4.2 Post-Processing . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6 Simulation Environment Setup . . . . . . . . . . . . . . . . . . . . . . 49
6.1 Simulation Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.2 Pre-Processing Methods . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.3 Post-Processing Methods . . . . . . . . . . . . . . . . . . . . . . . . . 51
7 Test case 1: Rectangular Tank . . . . . . . . . . . . . . . . . . . . . . 53
7.1 Test A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.1.1 Test Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.1.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.1.3 Results & Evaluation . . . . . . . . . . . . . . . . . . . . . . . 55
7.2 Test B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7.2.1 Test Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7.2.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 61
7.2.3 Results & Evaluation . . . . . . . . . . . . . . . . . . . . . . . 62
x
Contents
7.3 Test C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.3.1 Test Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.3.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 70
7.3.3 Results & Evaluation . . . . . . . . . . . . . . . . . . . . . . . 73
8 Test case 2: Cylindrical Tank . . . . . . . . . . . . . . . . . . . . . . . 78
8.1 Test Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
8.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
8.3 Results & Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
9 Test case 3: ESA Tank . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
9.1 Test A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
9.1.1 Test Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 85
9.1.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 86
9.1.3 Results & Evaluation . . . . . . . . . . . . . . . . . . . . . . . 88
9.2 Test B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
9.2.1 Test Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 94
9.2.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 95
9.2.3 Results & Evaluation . . . . . . . . . . . . . . . . . . . . . . . 95
9.3 Test C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
9.3.1 Test Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 98
9.3.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 99
9.3.3 Results & Evaluation . . . . . . . . . . . . . . . . . . . . . . . 99
10 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
11 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
A Test case 1 - Results: Test A . . . . . . . . . . . . . . . . . . . . . . . 115
B Test case 1 - Results: Test C . . . . . . . . . . . . . . . . . . . . . . . 122
C Test case 2 - Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
D Test case 3 - Results: Test A . . . . . . . . . . . . . . . . . . . . . . . 135
E Test case 3 - Results: Test B . . . . . . . . . . . . . . . . . . . . . . . 157
F Test case 3 - Results: Test C . . . . . . . . . . . . . . . . . . . . . . . 160
xi
List of Figures
2.1 Slosh wave shapes - first 2 antisymmetric x-modes for a rectangular
tank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Slosh wave shapes - first 2 symmetric x-modes for a rectangular tank 8
2.3 Computational solution procedure process . . . . . . . . . . . . . . . 20
5.1 ElmerGUI main window . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.2 ElmerPost main window & graphics window . . . . . . . . . . . . . . 48
6.1 Software installation diagram . . . . . . . . . . . . . . . . . . . . . . 50
6.2 Simulation flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.3 Complete software installation diagram . . . . . . . . . . . . . . . . . 52
7.1 Rectangular tank - test A a): pressure at t = 0s . . . . . . . . . . . . 56
7.2 Rectangular sloshing tank - test A a): free surface shape evolution . . 57
7.3 Rectangular tank - test A a): CoG plots . . . . . . . . . . . . . . . . 58
7.4 Rectangular tank - test A a): sloshing amplitude plot . . . . . . . . . 59
7.5 Rectangular tank - test B a): pressure at t = 0s . . . . . . . . . . . . 63
7.6 Rectangular tank - test B b): pressure at t = 0s . . . . . . . . . . . . 63
7.7 Rectangular tank - test B a): CoG x-coord. Vs time . . . . . . . . . . 64
7.8 Rectangular tank - test B a): maximum wave amplitude (t = 3.48s) . 64
7.9 Rectangular tank - test B b): CoG plots . . . . . . . . . . . . . . . . 65
7.10 Rectangular tank - test B b): CoG x-coordinate Vs time . . . . . . . 65
7.11 Rectangular tank - test B b): sloshing amplitude plot . . . . . . . . . 66
7.12 Rectangular tank - test B b): PSD plot 1 . . . . . . . . . . . . . . . . 67
7.13 Rectangular tank - test B b): PSD plot 2 . . . . . . . . . . . . . . . . 67
7.14 Rectangular tank - test C: pressure at t = 0s . . . . . . . . . . . . . . 73
7.15 Rect. tank - test C - h = 0.050m longer dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
7.16 Rect. tank - test C - h = 0.050m shorter dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
7.17 Rect. tank - test C - h = 0.050m longer dir.: CoG x-coord. Vs time . 75
7.18 Rect. tank - test C - h = 0.050m shorter dir.: CoG x-coord. Vs time 75
7.19 Rect. tank - test C - h = 0.050m longer dir.: max. wave amplitude . 75
xii
List of Figures
7.20 Rect. tank - test C - h = 0.050m shorter dir.: max. wave amplitude . 76
8.1 Cylindrical tank - test: pressure at t = 0s . . . . . . . . . . . . . . . . 81
8.2 Cylindrical tank test - h = 0.050m: free surface shape at t = 0s . . . 81
8.3 Cylindrical tank test - h = 0.050m: CoG x-coordinate Vs time . . . . 82
8.4 Cylindrical tank test - h = 0.050m: max. wave amplitude . . . . . . . 82
9.1 ESA tank test A - MON-3: pressure at t = 0s . . . . . . . . . . . . . 89
9.2 ESA tank test A - MMH: pressure at t = 0s . . . . . . . . . . . . . . 89
9.3 ESA tank test A - MON-3 (50% fill ratio & fext = 0.70 Hz): CoG plots 90
9.4 ESA tank test A - MON-3 (50% fill ratio & fext = 0.70 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
9.5 ESA tank test A - MON-3 (50% fill ratio & fext = 0.70 Hz): PSD
plot 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
9.6 ESA tank test A - MON-3 (50% fill ratio & fext = 0.70 Hz): PSD
plot 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
9.7 ESA tank test A - MON-3 (50% fill ratio & fext = 1.50 Hz): PSD
plot 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
9.8 ESA tank test A - MON-3 (50% fill ratio & fext = 1.50 Hz): PSD
plot 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
9.9 ESA tank test B - MON-3 or MMH: free surface shape at t = 0s . . . 96
9.10 ESA tank test B - MON-3: CoG x-coordinate Vs time . . . . . . . . 96
9.11 ESA tank test B - MON-3: max. wave amplitude . . . . . . . . . . . 97
9.12 ESA tank test B - MON-3: sloshing amplitude plot . . . . . . . . . . 97
9.13 ESA tank test C - MON-3 (60s simulation): CoG x-coord. Vs time . 100
9.14 ESA tank test C - MON-3 (60s simulation): wave amplitude at t = 30s100
9.15 ESA tank test C - MON-3 (60s simulation): wave amplitude at t = 45.5s101
9.16 ESA tank test C - MON-3 (60s simulation): sloshing amplitude plot . 101
9.17 ESA tank test C - MON-3 (60s simulation): CoG z-coord. Vs time . 102
9.18 ESA tank test C - MON-3 (20s simulation): wave amplitude at t = 0.54s103
9.19 ESA tank test C - MON-3 (20s simulation): wave amplitude at t =
19.60s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
9.20 ESA tank test C - MON-3 (20s simulation): CoG x-coord. Vs time . 104
9.21 ESA tank test C - MON-3 (20s simulation): CoG z-coord. Vs time . 105
A.1 Rectangular tank - test A b): pressure at t = 0s . . . . . . . . . . . . 116
A.2 Rectangular sloshing tank - test A b): free surface shape evolution . . 116
A.3 Rectangular tank - test A b): CoG plots . . . . . . . . . . . . . . . . 117
A.4 Rectangular tank - test A b): sloshing amplitude plot . . . . . . . . . 117
A.5 Rectangular tank - test A c): pressure at t = 0s . . . . . . . . . . . . 118
xiii
List of Figures
A.6 Rectangular sloshing tank - test A c): free surface shape evolution . . 118
A.7 Rectangular tank - test A c): CoG plots . . . . . . . . . . . . . . . . 119
A.8 Rectangular tank - test A c): sloshing amplitude plot . . . . . . . . . 119
A.9 Rectangular tank - test A d): pressure at t = 0s . . . . . . . . . . . . 120
A.10 Rectangular sloshing tank - test A d): free surface shape evolution . . 120
A.11 Rectangular tank - test A d): CoG plots . . . . . . . . . . . . . . . . 121
A.12 Rectangular tank - test A d): sloshing amplitude plot . . . . . . . . . 121
B.1 Rect. tank - test C - h = 0.100m longer dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
B.2 Rect. tank - test C - h = 0.100m longer dir.: CoG x-coord. Vs time . 123
B.3 Rect. tank - test C - h = 0.100m longer dir.: max. wave amplitude . 123
B.4 Rect. tank - test C - h = 0.150m longer dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
B.5 Rect. tank - test C - h = 0.150m longer dir.: CoG x-coord. Vs time . 124
B.6 Rect. tank - test C - h = 0.150m longer dir.: max. wave amplitude . 124
B.7 Rect. tank - test C - h = 0.200m longer dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
B.8 Rect. tank - test C - h = 0.200m longer dir.: CoG x-coord. Vs time . 125
B.9 Rect. tank - test C - h = 0.200m longer dir.: max. wave amplitude . 125
B.10 Rect. tank - test C - h = 0.250m longer dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
B.11 Rect. tank - test C - h = 0.250m longer dir.: CoG x-coord. Vs time . 126
B.12 Rect. tank - test C - h = 0.250m longer dir.: max. wave amplitude . 126
B.13 Rect. tank - test C - h = 0.100m shorter dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
B.14 Rect. tank - test C - h = 0.100m shorter dir.: CoG x-coord. Vs time 127
B.15 Rect. tank - test C - h = 0.100m shorter dir.: max. wave amplitude . 127
B.16 Rect. tank - test C - h = 0.150m shorter dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
B.17 Rect. tank - test C - h = 0.150m shorter dir.: CoG x-coord. Vs time 128
B.18 Rect. tank - test C - h = 0.150m shorter dir.: max. wave amplitude . 128
B.19 Rect. tank - test C - h = 0.200m shorter dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
B.20 Rect. tank - test C - h = 0.200m shorter dir.: CoG x-coord. Vs time 129
B.21 Rect. tank - test C - h = 0.200m shorter dir.: max. wave amplitude . 129
B.22 Rect. tank - test C - h = 0.250m shorter dir.: free surface shape at
t = 0s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
B.23 Rect. tank - test C - h = 0.250m shorter dir.: CoG x-coord. Vs time 130
xiv
List of Figures
B.24 Rect. tank - test C - h = 0.250m shorter dir.: max. wave amplitude . 130
C.1 Cylindrical tank test - h = 0.100m: free surface shape at t = 0s . . . 132
C.2 Cylindrical tank test - h = 0.100m: CoG x-coordinate Vs time . . . . 132
C.3 Cylindrical tank test - h = 0.100m: max. wave amplitude . . . . . . . 132
C.4 Cylindrical tank test - h = 0.150m: free surface shape at t = 0s . . . 133
C.5 Cylindrical tank test - h = 0.150m: CoG x-coordinate Vs time . . . . 133
C.6 Cylindrical tank test - h = 0.150m: max. wave amplitude . . . . . . . 133
C.7 Cylindrical tank test - h = 0.200m: free surface shape at t = 0s . . . 134
C.8 Cylindrical tank test - h = 0.200m: CoG x-coordinate Vs time . . . . 134
C.9 Cylindrical tank test - h = 0.200m: max. wave amplitude . . . . . . . 134
D.1 ESA tank test A - MON-3 (25% fill ratio & fext = 0.70 Hz): CoG plots136
D.2 ESA tank test A - MON-3 (25% fill ratio & fext = 1.50 Hz): CoG plots136
D.3 ESA tank test A - MON-3 (25% fill ratio & fext = 0.70 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
D.4 ESA tank test A - MON-3 (25% fill ratio & fext = 1.50 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
D.5 ESA tank test A - MON-3 (25% fill ratio & fext = 0.70 Hz): PSD
plot 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
D.6 ESA tank test A - MON-3 (25% fill ratio & fext = 0.70 Hz): PSD
plot 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
D.7 ESA tank test A - MON-3 (25% fill ratio & fext = 1.50 Hz): PSD
plot 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
D.8 ESA tank test A - MON-3 (25% fill ratio & fext = 1.50 Hz): PSD
plot 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
D.9 ESA tank test A - MON-3 (50% fill ratio & fext = 1.50 Hz): CoG plots140
D.10 ESA tank test A - MON-3 (50% fill ratio & fext = 1.50 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
D.11 ESA tank test A - MON-3 (75% fill ratio & fext = 0.70 Hz): CoG plots141
D.12 ESA tank test A - MON-3 (75% fill ratio & fext = 1.50 Hz): CoG plots141
D.13 ESA tank test A - MON-3 (75% fill ratio & fext = 0.70 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
D.14 ESA tank test A - MON-3 (75% fill ratio & fext = 1.50 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
D.15 ESA tank test A - MON-3 (75% fill ratio & fext = 0.70 Hz): PSD
plot 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
D.16 ESA tank test A - MON-3 (75% fill ratio & fext = 0.70 Hz): PSD
plot 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
xv
List of Figures
D.17 ESA tank test A - MON-3 (75% fill ratio & fext = 1.50 Hz): PSD
plot 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
D.18 ESA tank test A - MON-3 (75% fill ratio & fext = 1.50 Hz): PSD
plot 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
D.19 ESA tank test A - MMH (25% fill ratio & fext = 0.70 Hz): CoG plots 145
D.20 ESA tank test A - MMH (25% fill ratio & fext = 1.50 Hz): CoG plots 145
D.21 ESA tank test A- MMH (25% fill ratio & fext = 0.70 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
D.22 ESA tank test A - MMH (25% fill ratio & fext = 1.50 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
D.23 ESA tank test A - MMH (25% fill ratio & fext = 0.70 Hz): PSD plot 1147
D.24 ESA tank test A - MMH (25% fill ratio & fext = 0.70 Hz): PSD plot 2147
D.25 ESA tank test A - MMH (25% fill ratio & fext = 1.50 Hz): PSD plot 1148
D.26 ESA tank test A - MMH (25% fill ratio & fext = 1.50 Hz): PSD plot 2148
D.27 ESA tank test A - MMH (50% fill ratio & fext = 0.70 Hz): CoG plots 149
D.28 ESA tank test A - MMH (50% fill ratio & fext = 1.50 Hz): CoG plots 149
D.29 ESA tank test A- MMH (50% fill ratio & fext = 0.70 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
D.30 ESA tank test A - MMH (50% fill ratio & fext = 1.50 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
D.31 ESA tank test A - MMH (50% fill ratio & fext = 0.70 Hz): PSD plot 1151
D.32 ESA tank test A - MMH (50% fill ratio & fext = 0.70 Hz): PSD plot 2151
D.33 ESA tank test A - MMH (50% fill ratio & fext = 1.50 Hz): PSD plot 1152
D.34 ESA tank test A - MMH (50% fill ratio & fext = 1.50 Hz): PSD plot 2152
D.35 ESA tank test A - MMH (75% fill ratio & fext = 0.70 Hz): CoG plots 153
D.36 ESA tank test A - MMH (75% fill ratio & fext = 1.50 Hz): CoG plots 153
D.37 ESA tank test A- MMH (75% fill ratio & fext = 0.70 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
D.38 ESA tank test A - MMH (75% fill ratio & fext = 1.50 Hz): sloshing
amplitude plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
D.39 ESA tank test A - MMH (75% fill ratio & fext = 0.70 Hz): PSD plot 1155
D.40 ESA tank test A - MMH (75% fill ratio & fext = 0.70 Hz): PSD plot 2155
D.41 ESA tank test A - MMH (75% fill ratio & fext = 1.50 Hz): PSD plot 1156
D.42 ESA tank test A - MMH (75% fill ratio & fext = 1.50 Hz): PSD plot 2156
E.1 ESA tank test B - MMH (50% fill ratio): CoG x-coordinate Vs time . 158
E.2 ESA tank test B - MMH (50% fill ratio): max. wave amplitude . . . 158
E.3 ESA tank test B - MMH (50% fill ratio): sloshing amplitude plot . . 159
F.1 ESA tank test C - MON-3: PSD plot . . . . . . . . . . . . . . . . . . 161
xvi
List of Figures
F.2 ESA tank test C - MON-3 (20s simulation): sloshing amplitude plot . 161
F.3 ESA tank test C - MMH: PSD plot . . . . . . . . . . . . . . . . . . . 162
F.4 ESA tank test C - MMH (60s simulation): CoG x-coord. Vs time . . 162
F.5 ESA tank test C - MMH (60s simulation): wave amplitude at t = 30s 163
F.6 ESA tank test C - MMH (60s simulation): wave amplitude at t = 45.46s163
F.7 ESA tank test C - MMH (60s simulation): CoG z-coord. Vs time . . 164
F.8 ESA tank test C - MMH (20s simulation): wave amplitude at t = 0.52s164
F.9 ESA tank test C - MMH (20s simulation): wave amplitude at t = 19.58s165
F.10 ESA tank test C - MMH (20s simulation): CoG x-coord. Vs time . . 165
F.11 ESA tank test C - MMH (20s simulation): CoG z-coord. Vs time . . 166
F.12 ESA tank test C - MMH (60s simulation): sloshing amplitude plot . . 166
F.13 ESA tank test C - MMH (20s simulation): sloshing amplitude plot . . 166
xvii
List of Tables
4.1 Second phase: satisfactory codes - codes and characteristics. . . . . . 34
4.2 Third phase - codes and evaluated characteristics. . . . . . . . . . . . 35
7.1 Results - rectangular tank: test A. . . . . . . . . . . . . . . . . . . . 59
7.2 Comparison of results - rectangular tank: test A. . . . . . . . . . . . 60
7.3 Results - rectangular tank: test B. . . . . . . . . . . . . . . . . . . . . 68
7.4 Comparison of results - rectangular tank: test B. . . . . . . . . . . . 69
7.5 Results - rectangular tank: test C. . . . . . . . . . . . . . . . . . . . . 76
7.6 Comparison of results - rectangular tank: test C - 1 . . . . . . . . . . 76
7.7 Comparison of results - rectangular tank: test C - 2 . . . . . . . . . . 77
8.1 Results - cylindrical tank test. . . . . . . . . . . . . . . . . . . . . . . 83
8.2 Comparison of results - cylindrical tank test. . . . . . . . . . . . . . . 83
9.1 Results - ESA tank: test B. . . . . . . . . . . . . . . . . . . . . . . . 97
xviii
List of Abbreviations
AMG Algebraic Multigrid
AOCS Attitude and Orbit Control System
BDF Backward Differences Formulae
BEM Boundary Element Method
BiCGStab Biconjugate Gradient Stabilized
CFD Computational Fluid Dynamics
CG Conjugate Gradient
CGS Conjugate Gradient Squared
CoG Center of Gravity
ESA European Space Agency
FDM Finite Difference Method
FEM Finite Element Method
FVM Finite Volume Method
GCR Generalized Conjugate Residual
GMG Geometric Multigrid
GMRES Generalized Minimal Residual
GPL General Public License
GUI Graphical User Interface
HSO Human Spaceflight and Operations
ILU Incomplete LU
xix
List of Abbreviations
LAPACK Linear Algebra Package
LGPL Lesser General Public License
NTP Normal Temperature and Pressure
OS Operating System
PDE Partial Differential Equation
PSD Power Spectral Density
Sif Solver Input file
SUPG Streamline-Upwind Petrov-Galerkin
TFQMR Transpose-Free Quasi-Minimal Residual
UMFPACK Unsymmetric Multifrontal Sparse LU Factorization Package
xx
Chapter 1
Introduction
“Ever since there have been people, there have been explorers,
looking in places where others had not been before. Not everyone
does it, but we are part of a species where some members of the
species do, to the benefit of us all.”
- Neil deGrasse Tyson
Since the beginning of times humans have looked into the sky and wondered
at its beauties. For centuries we dreamt about leaving the Earth and going
further, beyond the sky, to achieve space flight. Due to the perseverance
and effort of some, this dream became a reality when the first artificial
satellite - Sputnik I - was launched in 1957 - at that point, a new era
began, the space age just started...
In less than 60 years of space exploration we landed humans on the Moon;
generated conditions to have humans orbiting the Earth on a permanent
basis; alighted several spacecraft in close planetary bodies (Venus, Mars
and Jupiter); studied extraterrestrial bodies; launched thousands of satell-
ites with numerous purposes to orbit the Earth; and are now about to break
another important barrier by flying a spacecraft into outer space - all these
important advances not only contributed to the scientific and technological
development of our society, but also made life on Earth much easier.
To achieve these breakthroughs, many questions had to be addressed and
many studies to be performed. For years, thousands of minds around the
globe worked and are still working to increase the capabilities of modern
space systems.
1
1. Introduction
The complexity level, now reached, allows us to deeply address some ques-
tions which have long been made and yet not fully answered, some of these
are the focus of this work:
• How are the dynamics of a space vehicle affected by the behaviour of
the liquid propellants inside its tanks?
• How accurately can the behaviour of the liquid propellants and their
interaction with the spacecraft be predicted?
Thanks to recent advances in science, engineering and technology, it is
now possible to develop deeper and further studies on this important topic
- liquid propellant sloshing inside spacecraft tanks. Even though, this has
been identified long ago as being of significant and sometimes even critical
influence on the dynamics of a spacecraft, it has not yet been completely
studied, mainly due to the difficulty that is predicting the liquids behaviour
inside the tanks.
When not carefully accounted for, the interaction of the disturbance forces
and torques caused by the moving fuel with the solid body and the control
system through the feedback loop can lead to an increase in the Atti-
tude and Orbit Control System (AOCS) actuators commands, which can
degrade the satellite pointing performances and in some critical cases ge-
nerate unstable attitude and orbit control.
This means that it is of major importance to accurately predict the be-
haviour of the liquid propellants inside the spacecraft tanks. This M.Sc.
thesis is focused on this topic, being its ultimate objective the “Imple-
mentation of a Computational Fluid Dynamics (CFD) Code for Propellant
Sloshing Analysis”.
1.1 Aim
The aim of this project is to implement in the existing EADS Astrium
simulation environment a CFD Code that shall be used to assess the in-
fluence of liquid propellant sloshing in specific satellite missions. Selected
propellant sloshing examples, defined by EADS Astrium and the European
Space Agency’s (ESA) directorate of Human Spaceflight and Operations
(HSO) (which also supported this project through a scholarship), shall be
incorporated, analysed and finally evaluated using the newly implemented
CFD code.
2
1. Introduction
The results of this work are planned to be used in the future by EADS
Astrium. The implemented CFD code is intended to become the favourite
sloshing analysis tool for the AOCS/GNC & Flight Dynamics Department
of Astrium Satellites (Friedrichshafen, Germany) where this work is being
developed under the supervision of M.Sc. Manuel Hahn.
1.2 Objectives
The following milestones were defined for this thesis project. Together they
define the general objectives that shall be accomplished during the project.
• Perform an extensive literature research to select the most suitable
CFD code based on defined requirements;
• Implement the selected CFD code in the existing Astrium’s pre- and
post- processing environment;
• Implement, analyse and evaluate selected propellant sloshing exam-
ples;
• Validate the simulation environment for the selected examples.
1.3 Outline
• Chapter 1 introduces the topic of this M.Sc. thesis, presents its aim
and the general objectives expected to be achieved during its devel-
opment.
• Chapter 2 presents a detailed but not exhaustive literature review
about the sloshing and CFD topics.
• Chapter 3 briefly introduces the requirements for the project.
• Chapter 4 presents the deep state-of-the-art investigation developed
to choose the most suitable CFD software for the purposes of this
project.
• Chapter 5 gives a brief introduction to the chosen CFD software -
Elmer.
3
1. Introduction
• Chapter 6 presents the setup of the complete simulation environment.
• Chapter 7 introduces, defines, explains the implementation, presents
the results and evaluates the tests performed for the rectangular tank
test case.
• Chapter 8 follows the same path of Chapter 7 and introduces the
cylindrical tank test case.
• Chapter 9 similarly to chapters 7 and 8 presents the cylindrical tank
with hemispherical domes (by ESA) test case.
• Chapter 10 presents the final conclusions drawn from this work.
• Chapter 11 attempts to explore future research lines and define work
that could be further developed.
4
Chapter 2
Literature Review
“The learning and knowledge that we have, is, at the most, but
little compared with that of which we are ignorant.”
- Plato
“He who receives ideas from me, receives instruction himself
without lessening mine; as he who lights his taper at mine
receives light without darkening me.”
- Thomas Jefferson
Before getting immersed in the project implementation, it is crucial to have
a general understanding of the relevant topics addressed in this Master’s
thesis. This chapter presents an explanation of the basic concepts related
with liquid slosh and CFD. It shall help the reader to get into the topic
without deeply getting into the overwhelming complexity of the concepts.
The subsequent sections provide an introduction to the liquid sloshing con-
cept - section 2.1, followed by a general overview of the theory behind CFD
- section 2.2.
2.1 Sloshing
As mentioned in chapter 1, liquid propellant sloshing in spacecraft tanks
can be of critical influence to the dynamics of the system, as well as to the
AOCS. The sloshing forces and torques imposed by the liquid motion in
the tank, together with the resulting shifts in the liquid’s center of gravity
(CoG), need to be carefully analysed.
5
2. Literature Review
In the following subsections a general introduction to the sloshing phe-
nomenon is presented. A simple analytical overview of lateral sloshing in
geometrically simple tanks containing ideal liquids in linear regime (small
wave amplitudes) is fully described. More complex sloshing problems -
which include also damping, non-linearity and micro-gravity effects, as well
as the derivation of the equations can be found in the literature, see [1]
and [2]. Even if not deeply described, the concepts of damping, non-linear
effects and micro-gravity effects are still briefly introduced later. However,
before getting deeper in the topic it is important to define the concept of
liquid sloshing inside rigid containers:
- “Any motion of a free liquid surface caused by any disturbance to a rigid
container partially filled with liquid.” [2]
2.1.1 Lateral Sloshing
Lateral sloshing is the simplest way of liquid sloshing inside containers. It
is defined as the formation of a standing wave on the surface of a liquid
when a tank partially filled is laterally excited. Under simplified conditions
the behaviour of the liquid can be defined by a set of equations, which
incorporate a set of liquid sloshing parameters.
The natural frequencies of the liquid, the velocity potential and the forces
and torques generated by its motion can be analytically obtained for sim-
ple tank geometries subjected to small external excitations under accel-
erated environments. Using as a basis the classical potential flow theory,
which involves treating the fluid as incompressible and inviscid, and solving
Laplace’s equation that satisfies the boundary conditions, these parameters
can be found.
In 1952 Graham and Rodriguez [3] introduced for the first time the 3- -
dimensional free surface natural frequencies of a liquid sloshing inside a
rectangular container. Later, in 1955 Housner [4] derived the analytical
solution for the first antisymmetric sloshing frequency of liquids sloshing
inside rectangular and cylindrical tanks. In 1966 Abramson [5] (republished
in 2000 by Dodge [1]) completely derived these parameters for several types
of tanks, having even introduced damping, non-linearity and micro-gravity
effects in its derivations.
6
2. Literature Review
In the following subsections, important sloshing parameters of ideal liquids
sloshing inside simple rigid containers are presented. The fluid is always
considered incompressible and the lateral excitation is considered much
smaller than the vertical acceleration acting on the tank, and therefore
negligible. For more details on the derivation of the equations please refer
to the above mentioned references.
Before presenting the above mentioned liquid sloshing parameters and re-
spective equations for both, rectangular and cylindrical tanks, let us first
introduce the concept of sloshing modes.
2.1.1.1 Sloshing Modes
The definition of sloshing modes results from the multiple configurations
to which the liquid’s surface may evolve when it sloshes inside a container.
Normally, this modes are defined by:
• n - the antisymmetric mode number;
• m - the symmetric mode number.
Therefore, in general, there are two main types of lateral sloshing: the
antisymmetric and the symmetric. Their modes are defined by the number
of wave peaks formed at the liquid’s surface, for example:
• for n = 1 there is a positive peak at one wall and a negative one at
the other;
• for m = 1 a positive peak occurs in the middle of the tank and two
negative ones appear in the walls.
The number of wave peaks increases together with the mode numbers n
and m, respectively for the antisymmetric and symmetric sloshing types. [1]
Antisymmetric Modes
The antisymmetric sloshing modes represent the most severe cases of liquid
slosh that can develop in spacecraft tanks. The slosh wave shapes for the
first two x-modes of a rectangular tank are presented in figure 2.1.
7
2. Literature Review
Figure 2.1: Slosh wave shapes - first 2 antisymmetric x-modes for a rectangulartank.
It is possible to visualize in figure 2.1 that the CoG shifts when the liquid
moves to provoke forces and torques that act on the tank shell. One im-
portant evidence is that the higher the sloshing mode number, the higher
the corresponding natural frequency and the smaller the CoG shift. Thus,
the smaller the generated disturbances and less significant the importance
of the higher sloshing modes.
Symmetric Modes
The symmetric sloshing modes are of less significance regarding the pro-
pellant sloshing disturbing effects in spacecraft tanks. Figure 2.2 presents
the slosh wave shapes for the first two symmetric x-modes in a rectangular
tank.
Figure 2.2: Slosh wave shapes - first 2 symmetric x-modes for a rectangular tank.
8
2. Literature Review
As it can be seen in figure 2.2, there is no lateral CoG shift in the liquid.
This makes the lateral forces and torques acting on the tank shell non-
-existent. As a side note, the frequencies of the symmetric modes are always
higher than those of the corresponding antisymmetric modes.
2.1.1.2 Rectangular Tank
Starting from the classic potential flow theory together with some necessary
assumptions and defined boundary conditions for this specific problem, the
natural frequencies of liquid’s sloshing inside rectangular containers can be
analytically derived. These depend on the height of the liquid inside the
tank, the tank shape and the vertical acceleration.
Antisymmetric modes only - 2D Natural Frequencies [5]
ω2n = π(2n− 1)
(ga
)tanh
[π(2n− 1)
(h
a
)](2.1)
Where:
ω: is the natural frequency
n: is the mode number
g: is the gravitational acceleration
a: is the width of the tank (in the x-direction)
h: is the height of the liquid inside the tank
Symmetric modes only - 2D Natural Frequencies [5]
ω2m = 2mπ
(ga
)tanh
[2mπ
(h
a
)](2.2)
Where:
m: is the mode number
9
2. Literature Review
Equations (2.1) and (2.2) present the natural frequencies for the anti-
symmetric and symmetric sloshing modes when the translational oscillation
of the tank occurs along the x-direction. If this oscillation occurs along the
y-direction, the equations are the same but, the width a is replaced by the
breadth b.
3D Free Surface Natural Frequencies [3]
ω2mn = gKmn tanh (Kmnh) (2.3)
Where:
Kmn = π√
(2m)2
a2 + (2n)2
b2
Equation (2.3) gives the natural frequencies of the modes which vary in
both x and y directions. The resulting wave shapes for this mode are a
combination of the 2D x- and y- mode shapes.
First Antisymmetric Sloshing Frequency - Liquid Water [4]
It is important to note that this equation was developed for inviscid fluids.
However, because in its development Housner applied simpler methods in
the resolution of the Partial Differential Equations (PDEs), the results
that can be obtained using this equation are slightly different from those
obtained by Abramson.
Recent publications [6], have used this formula to obtain the first antisym-
metric natural sloshing frequency of liquid water - according to them, the
results are slightly more accurate than those given by Abramson.
Therefore, for liquid water sloshing inside a rectangular tank vertically
accelerated and laterally excited, the first sloshing frequency is given by
equation (2.4):
ω21 = 2
√5
2
( gL
)tanh
(2
√5
2
(h
L
))(2.4)
Where:
L: is the length of the rectangular tank along the direction of excita-
tion
10
2. Literature Review
2.1.1.3 Cylindrical Tank
The equations previously presented for the rectangular tank case can also
be found in the literature for the cylindrical tank geometry [1]. However,
due to their increased complexity, they are not introduced here. Only the
first sloshing natural frequency for liquid water is presented below.
First Antisymmetric Sloshing Frequency - Liquid Water [4]
For liquid water inside a cylindrical tank vertically accelerated and laterally
excited, the first sloshing frequency is given by equation (2.5):
ω21 = 2
√27
8
( gD
)tanh
(2
√27
8
(h
D
))(2.5)
Where:
D: is the diameter of the cylindrical tank
2.1.2 Introduction to Damping
The analytic equations previously presented (subsection 2.1.1) for laterally
excited simple tanks do not consider the viscosity of the fluids. Meaning
that, damping effects are neglected. Thus, it is being considered that
the oscillation of the sloshing wave will continue over time even when the
external excitation is stopped. This is not representative of the real world,
where such thing does not happen.
For a viscous (non-ideal) fluid, damping will exist. Thus, once the external
excitation is stopped, the sloshing wave decreases in amplitude and will
also eventually stop.
Damping shall then, optimally, be considered, when performing sloshing
analysis. Nevertheless, because its consideration considerably increases the
complexity of the analytical equations previously exposed, CFD tools are
normally used to accurately replicate the effects of damping in sloshing.
As a side note, the main parameters that mostly influence damping are:
the viscosity of the sloshing liquid, the fill level, the shape and the tank
11
2. Literature Review
shell. Meaning that, to increase damping and reduce sloshing, not only
the properties of the liquid matter but also the tank geometry and its
properties. [1]
2.1.3 Introduction to Non-linear Effects in Slosh
It was already stated that the lateral sloshing examples analytically derived
and presented in subsection 2.1.1 include several simplifications, which
made the problem possible to be analytically solved. Among these are
the non-linear effects, which are also not considered. However, non-linear
effects always exist in liquid sloshing and therefore a brief introduction to
these shall be given.
For small sloshing amplitude waves, thus small external perturbations, non-
-linear effects are normally neglected - their effect is almost non-existent.
However, for large wave amplitudes and different forms of sloshing (such
as rotary sloshing), non-linear effects are crucial and dominate the sloshing
response. Thus, they shall also be modelled in CFD in order to accurately
predict the real behaviour of the liquid sloshing. This topic consists itself
in a complex field of studies, but a brief introduction can be found in [1, 2].
To ease the understanding, a simple manner of explaining the importance of
the non-linear effects in a flow is by means of the non-dimensional Reynolds
number (Re) - equation (2.6).
This very useful number gives a measure of the ratio of the inertial forces
by the viscous forces. Consequently, it quantifies the relative importance
of these types of forces for given flow conditions.
Re =ρvL
µ(2.6)
Where:
v: is the mean velocity of the object relative to the fluid
ρ: is the density of the fluid
L: is the characteristic length
µ: is the kinematic viscosity of the fluid
12
2. Literature Review
It is known that an increase in the Reynolds number leads to an increase in
the predominance of the non-linear effects in the fluid. Hence, looking at
equation (2.6) it is possible to see that the Reynolds number increases when
the mean velocity of the fluid increases or the viscosity decreases (keeping
the density constant). So, to avoid non-linear sloshing effects, the velocity
of the fluid shall be kept relatively small and the viscosity relatively high.
2.1.4 Introduction to Micro-gravity Effects - Surface Ten-
sion
Even though one of the simplifications assumed in the lateral sloshing ex-
amples of subsection 2.1.1 was that the liquid inside the spacecraft tank
was under accelerated conditions, the hypothesis of the liquid motion tak-
ing place in a micro-gravity environment exists and is very common in
space missions. Thus, it shall be briefly introduced.
In a micro-gravity environment, where the body forces become so small,
other small forces take place and are dominant in the behaviour of the
fluid. The most important of these are the surface tension forces at the
free surface of the liquid.
Surface tension (or capillary forces) is a complete field of studies by itself.
Thus, it would be totally out of the scope for this project to deeply intro-
duce this topic. Therefore, in this thesis development, it was considered
that the surface tension effects were not accountable for the defined test
cases. Thus, it was only important to understand what affects the predom-
inance of surface tension effects and at which point they become dominant
in the behaviour of the fluid. More details about this topic can also be
found in [1, 2].
To measure the predominance of the surface tension effects the Bond (or
Eotvos) number - equation (2.7), is used. This dimensionless number is
normally used as the most common comparison ratio for gravity and surface
tension forces.
Bo =ρaL2
γ(2.7)
Where:
ρ: is the density of the fluid
13
2. Literature Review
a: is the acceleration associated with the body force
L: is the characteristic length
γ: is the surface tension of the interface
Looking at equation (2.7) it is possible to understand that a high Bond
number represents a system relatively unaffected by the surface tension eff-
ects. Oppositely, a low number (typically less than one) indicates that the
surface tension effects dominate the fluid behaviour. Low Bond numbers
normally occur for fluids under micro-gravity conditions.
2.1.5 Other Types of Sloshing
Besides the lateral sloshing type here presented for 2 geometrically simple
tanks (subsection 2.1.1), there are several other types of sloshing. The most
interesting one is the rotary sloshing type which introduces predominant
non-linear effects in the fluid dynamics.
The study of this or other types of sloshing is out of the scope of this work.
Nevertheless, more information can be found in the literature, see [1] or [2].
2.2 Computational Fluid Dynamics
CFD is considered by many, the new, most interesting, branch of fluid dy-
namics. It acts together with the classical branches of pure experiment and
pure theory which are then supported and complemented by the cost-
-effective CFD tools. The role of CFD in engineering predictions has be-
come so strong that it has taken a permanent place in all the aspects of
fluid dynamics, from basic research to engineering design. [7]
CFD integrates disciplines such as fluid mechanics, mathematics and com-
puter science. The dynamics of a fluid can be characterized by mathemat-
ical equations (often called governing equations), which can then be solved
using numerical methods in sophisticated digital computers, by means of
computer programs or software packages. [8]
Since CFD became a trusted tool, the way in which engineering analy-
ses is performed has totally changed. The use of CFD substantially re-
duces lead times and costs in designs and productions compared to the
14
2. Literature Review
use of experimental approaches. Besides, it also offers the possibility to
solve complicated flow problems which could never be solved by analytical
means.
Nevertheless, one shall not blindly trust CFD results - they are only as
valid as the physical models incorporated in the governing equations and
boundary conditions. Hence, they are subject to various error sources
which can severely influence the accuracy of the results.
Numerical results shall always be thoroughly examined before believed to
be correct. Wonderful bright color pictures may provide a sense of reality
which might lead to mistaken conclusions. Therefore, at least, a basic
understanding of the theory behind CFD is needed to critically judge all
the results before trusting them.
2.2.1 Fluid Governing Equations
A fluid can be described by means of a set of mathematical equations
which represent its physical behaviour - the fluid governing equations. The
fundamental principles on which they are based are:
• Mass conservation - gives the continuity equation;
• Momentum conservation - gives the momentum equations (also known
as Navier-Stokes equations);
• Energy Conservation - gives the energy equation.
Various flow physics are governed by these fundamental principles which
might need to be applied together with some other modelling equations,
such as the turbulence ones.
The governing equations for a fluid flow general case in which the fluid is
considered non-turbulent, unsteady, 3-Dimensional, viscous, incompress-
ible and isothermal are presented here for reference - further in the devel-
opment of this work this is how the fluid is treated. More details, as well
as the derivation of these equations can be found in the literature, see for
example [7, 8, 9, 10] or [11] for a deeper explanation.
15
2. Literature Review
• Continuity equation:
∇ · (ρ~V ) = 0 (2.8)
• Momentum equations (or Navier-Stokes equations):
∇ · (ρu~V ) = −∂p∂x
+∂τxx∂x
+∂τyx∂y
+∂τzx∂z
+ ρfx (2.9)
∇ · (ρv~V ) = −∂p∂y
+∂τxy∂x
+∂τyy∂y
+∂τzy∂z
+ ρfy (2.10)
∇ · (ρw~V ) = −∂p∂z
+∂τxz∂x
+∂τyz∂y
+∂τzz∂z
+ ρfz (2.11)
Where:
~V = u~i+ v~j + w~k: is the vector velocity field
u = u(x, y, z, t): is the velocity component in the x-direction at time
t (unsteady flow)
v = v(x, y, z, t): is the velocity component in the y-direction at time t
(unsteady flow)
w = w(x, y, z, t): is the velocity component in the z-direction at time
t (unsteady flow)
ρ: is the density
τxx = λ∇ · ~V + 2µ∂u∂x
: is the the shear stress xx component
τyy = λ∇ · ~V + 2µ∂v∂y
: is the the shear stress yy component
τzz = λ∇ · ~V + 2µ∂w∂z
: is the the shear stress zz component
τxy = τyx = µ(
∂v∂x
+ ∂u∂y
): is the the shear stress xy or yx component
τxz = τzx = µ(∂u∂z
+ ∂w∂x
): is the the shear stress xz or zx component
τyz = τzy = µ(
∂w∂y
+ ∂v∂z
): is the the shear stress yz or zy component
µ: is the molecular viscosity
λ: is the bulk viscosity
16
2. Literature Review
p: is the pressure
fx, fy, fz: are the body forces per unit of mass acting on the fluid
element in x, y and z directions respectively and ρ~f · ~V = ρ(ufx +
vfy + wfz)
The above system of equations contains 4 equations in terms of 4 unknown
flow-field variables: the velocity field and the pressure.
2.2.2 Boundary Conditions
The mathematical equations previously presented govern the flow of a fluid.
Nevertheless, boundary conditions (and sometimes initial conditions) are
required to dictate the particular solutions to be obtained from the govern-
ing equations. There are 2 types of boundary conditions that, normally,
are defined for a fluid. Without getting into deep details (see [8] for more
information) these are:
• Dirichlet boundary conditions - give the boundary conditions re-
lated with velocity and pressure. The velocity components and/or the
pressure are defined at the boundary.
• Neuman boundary conditions - give the boundary conditions re-
lated with accelerations and forces. The acceleration components
and/or the force are defined at the boundary.
2.2.3 Meshing
A mesh is a discrete representation of the geometry of a problem which is
intended to be solved using CFD. The mesh designates the cells or elements
on which the flow is to be solved. Below an introduction to the different
types of mesh grids, as well as to the different mesh elements is presented.
2.2.3.1 Mesh Grids
In a simple manner, there are two types of meshes which are character-
ized by the connectivity of their points. Structured meshes (or grids) have
a regular connectivity, which means that each point of the mesh has the
17
2. Literature Review
same number of neighbours. The unstructured ones instead, have an irreg-
ular connectivity where each point can have a different number of neigh-
bours. [10]
Below, an introduction to these two types of grids is provided, the most
common types of mesh elements − tetrahedral and hexahedral, are also
discussed.
Structured Grid
In this type of grid, the points of an elemental cell can be easily addressed
by two indices (i, j) in two dimensions or three indices (i, j, k) in three
dimensions. The connectivity is very simple seeing that cells adjacent to
a given element face are identified by the indices and the cell edges form
continuous mesh lines that begin and end on opposite elemental faces. In
2D, the central cell is connected by four neighbouring cells. In 3D, the
central cell is connected by six neighbouring cells. This type of grids have
the advantage of allowing easy data management and connectivity, which
occurs in a regular fashion, making programming easy. [8]
Nevertheless, the disadvantage of adopting such a mesh, particularly for
more complex geometries is the increase in grid non-orthogonality or skew-
ness that can cause non-physical solutions due to the transformation of
the governing equations. The transformed equations that accommodate
the non-orthogonality act as the link between the structured coordinate
system and the body-fitted coordinate system containing additional terms
and thereby augmenting the cost of the numerical calculations and the dif-
ficulties in programming. Consequently, the use of such a mesh may also
affect the accuracy and efficiency of the numerical algorithm that is being
applied.
Unstructured Grid
Unstructured grids are currently the prevalent and widespread grid type in
many CFD applications. In these grids the cells are allowed to be assembled
freely within the computational domain. The connectivity information for
each face thus requires appropriate storage in some form of table. The most
typical shape of an unstructured element is a triangle in two dimensions
or a tetrahedron in three dimensions. However, any other elemental shape
including quadrilateral or hexahedral elements can also be used. [8]
18
2. Literature Review
2.2.3.2 Mesh Elements
Within the various mesh elements supported by the different grids, the
hexahedral and the tetrahedral elements are by far the most used ones.
Even though, for most of the geometries the output results shall not differ
even if using different element types, there are in fact advantages and
disadvantages of using one or another type of elements.
Grids made of tetrahedral elements are known for having the capability to
easily discretize any complex geometry in a fast and simple way, almost
with no user intervention. The use of hexahedral elements instead may
require some more effort to mesh complex geometries. However, grids
made of hexahedral elements do have a significant advantage as they have
the capability to preserve the accuracy in the wall normal direction even
for highly stretched viscous grids. Also, grids composed by these elements
have a reduced number of elements, edges and faces when compared to a
grid made of tetrahedral elements. [9]
2.2.4 Numerical Methods
Analytical solutions obtained from the flow governing equations can only
be found for simple and geometrically well defined problems. Obviously,
this is not the case for most of the real problems. For these problems, the
solution of the governing equations, in a general sense PDEs, can only be
approximated using numerical methods. The most used numerical meth-
ods in CFD are the Finite Element Method (FEM) and the Finite Volume
Method (FVM). A brief overview of these methods is presented here. How-
ever, no extensive nor detailed explanations are given as these can be easily
found in many standard books, see [12, 13].
The non inclusion of the Finite Difference Method (FDM) in this study is
mainly due to its many constrains which make them rarely used for CFD
applications. The most important of its constraints deals with the difficulty
to handle complex geometries − a high degree of mesh regularity is needed.
Due to this and other constraints, only a very small number of engineering
codes rely on this method. [8]
An overview of the process of the computational solution procedure using
FEM or FVM is shown in figure 2.3.
19
2. Literature Review
Figure 2.3: Computational solution procedure process.
2.2.4.1 Finite Element Method
The FEM is one of the available techniques to solve PDEs. The following
fundamental characteristics can be defined for this method (from [7]):
• The continuum field or domain is subdivided into cells, called elements,
which form a grid. The elements can have tetrahedral or hexahedral
forms and can be rectilinear or curved. The grid itself does not nec-
essarily need to be structured. Using unstructured grids and curved
cells, complex geometries can be handled with ease.
• The solution of the discrete problem is assumed a priori to have a
prescribed form. The solution has to belong to a function space, which
is built by varying function values in a given way, for instance linearly
or quadratically, between values in nodal points. The nodal points or
nodes, are typically points of the elements such as vertices, mid-side
points, mid-element points, etc. Due to this, the representation of the
20
2. Literature Review
solution is strongly linked with the geometric representation of the
domain.
• The solution of the PDEs itself is not what FEM looks for, instead a
solution of an integral form of the PDE is what is looked into.
• The discrete equations are constructed from contributions on the ele-
ment level which afterwards are assembled.
2.2.4.2 Finite Volume Method
The FVM is the most used technique in CFD. The following fundamental
characteristics can be defined for this method (from [7]):
• The integral form of the equations are discretized in this method in-
stead of the differential form.
• The flow field or domain is subdivided into a set of non-overlapping
cells that cover the whole domain.
• The conservation laws are applied to determine the flow variables in
some discrete points of the cells, the nodes. These nodes are typical
locations of the cells, such as cell-centers, cell-vertices or mid-sides.
There is considerable freedom in the choice of the cells and nodes.
Cells can be triangular, quadrilateral, etc. They can form structured
or unstructured grids.
• There is geometric flexibility in the choice of the grid and also in
defining the discrete flow variables.
2.2.4.3 Comparison
Both the FEM and FVM have the same geometrical flexibility. Neverthe-
less, the link between the representation of the solution and the geometric
representation of the domain is not as strong in FVM as in FEM. The
FVM tries to combine the best of the FEM, i.e. the geometric flexibility,
with the best of the FDM, i.e. the flexibility in defining the discrete flow
field. [7]
21
2. Literature Review
Some important advantages of FEM over FVM are the ease to obtain high
order accuracy results and also the ease to implement boundary condi-
tions. Nevertheless, the FVM is best suited for flow problems in primitive
variables, where the viscous terms are absent (Euler equations) or are not
dominant (high Reynolds number in Navier-Stokes equations). Further,
curved cell boundaries as used in FEM is difficult to implement in FVM.
FVM is mostly only second-order accurate.
Although, FVM seems to have disadvantages when compared to FEM,
specially in terms of accuracy, historically FVM have been mostly used
in CFD instead of FEM. This is so because FEM is originated from the
field of structural mechanics in which the partial differential formulation
of a problem can be replaced by an equivalent variational formulation -
the minimization of an energy integral over the domain. However, in fluid
dynamics, in general, a variational formulation is less obvious to formulate.
This makes it less obvious to formulate FEM for fluid dynamic purposes.
Hence, most breakthroughs in CFD have first been made in the context of
the FDM or FVM techniques and it has always taken some considerable
time (often more than a decade) to incorporate the same idea into the
FEM. However, in the other hand, once a suitable FEM formulation has
been found, the FEM is almost exclusively used. Obviously, this is due
to the mentioned advantages in the treatment of complex geometries and
obtaining high order accuracy.
Currently, for the simplest problems such as potential flows, both compress-
ible and incompressible Navier-Stokes flows at low Reynolds numbers, the
FEM is already fully grown. Although, new evolutions, specially for Navier-
Stokes problems, are still to come. Complex problems, like compressible
flows governed by Euler or Navier-Stokes equations or incompressible vis-
cous flows at high Reynolds numbers, still form an area of active research.
2.2.5 Numerical Analysis
In this section the methods to obtain the solution of a system of equations
are briefly introduced and explained. Direct, Iterative, Preconditioned and
Multigrid methods are concisely defined inside this chapter. If more details
are needed, please refer to [14].
22
2. Literature Review
Direct Methods
The so-called direct methods compute the solution of a problem in a finite
number of steps. The precise results of a system of equations would be ob-
tained using these methods, if they were performed with infinite arithmetic
precision. As this is not possible, in practice, finite precision is used and
the result is an approximation of the true solution, assuming that stability
exists.
Iterative Methods
Iterative methods, oppositely to direct methods, do not have an expected
number of steps to terminate. The resolution of the system is made by
starting with an initial guess, which then using iterative methods form
successive approximations that will eventually converge to the (acceptable)
exact solution (in the limit). A criteria to define whether a solution is
accurate enough needs to be specified. In these methods, even the use of
infinite arithmetic precision would not allow to reach an exact solution in
a finite number of steps.
Preconditioned Methods
A preconditioning is a procedure in which a transformation called the pre-
conditioner is applied for a given problem. This changes the problem to a
form that is more suitable to obtain numerical solutions. Normally, pre-
conditioning is related with the reduction of a condition number of the
problem. Iterative methods are usually used to solve preconditioned pro-
blems. This is so because the rate of convergence for most of the iterative
solvers increases as the condition number of a matrix decreases as a result
of the preconditioning.
Multigrid Methods
In numerical analysis multigrid methods are considered a specialized group
of algorithms that are used to solve differential equations using a hierarchy
of discretizations. The use of this methods accelerate the convergence of a
basic iterative method by means of a global correction from time to time.
23
2. Literature Review
This is accomplished by solving a coarse problem. This principle is similar
to the interpolation between coarser and finer grids.
Multigrid methods are among the fastest solution techniques known. In
contrast to other methods, these methods are general in a way that they
can treat arbitrary regions and arbitrary conditions. They do not depend
on the separability of the equations or other special properties. A direct
application of these can be used for more complicated, non-symmetric and
non-linear systems of equations, such as the Navier-Stokes equations.
2.2.6 Solution Analysis
The analysis of a computational solution (or approximate solution) repre-
sents one of the most important steps to be performed when using CFD.
As it is known, the system of algebraic equations is solved using numerical
methods which provide an approximate solution of the governing equa-
tions. Because the solution is approximate and not exact, as it would be
if analytical means were used, there are certain properties that one should
care when evaluating the results of a CFD simulation. Some of these prop-
erties are briefly introduced here − more details can be found in literature,
see [8].
Consistency
The consistency of a method concerns the discretization of the PDEs where
the approximation performed should diminish or become exact if the finite
quantities, such as the time step and the mesh spacing, tend to zero. Thus,
for a numerical method to be consistent, the truncation error − the differ-
ence between the result of the discretized equation and the exact one −must become zero when the time step and the mesh spacing tend to zero.
Stability
The existence of stability for a method is related with the growth or decay
of the errors introduced at any stage of the computation. A numerical
solution method is considered to be stable if it does not magnify the errors
that appear in the course of the numerical solution process.
24
2. Literature Review
For temporal problems the existence of stability guarantees that the method
yields a bounded solution whenever the exact solution is bounded. For it-
erative methods existing stability ensures the existence of a solution which
does not diverge.
Convergence
If a numerical method satisfies the consistency and stability properties,
generally the numerical method procedure is also convergent. Convergence
of a numerical process happens when the solution of the system of algebraic
equations approaches the true solution of the PDEs having the same initial
and boundary conditions as the refined grid system.
Accuracy
Because a converged solution does not necessarily mean an accurate one,
some possible sources of solution errors resulting from the numerical cal-
culations of the algebraic equations need to be considered. These require
attentive analysis and if they are to be minimized, systematic steps to per-
form numerical analysis, such as grid independence verification and vali-
dation of numerical methods, are necessary. Possible sources of solution
errors are: discretization errors, round-off errors, iteration or convergence
errors, physical-modelling errors and human errors.
Efficiency
The efficiency of a method is not exactly a property used to evaluate the
results of a CFD study. Nevertheless, if increased efficiency is required
to perform a certain study, this can for example be achieved by means of
parallel computing (e.g. using Message Passing Interface - MPI). Normally,
an increase in efficiency represents a less computationally exigent method
which will thus achieve a solution in less time.
25
2. Literature Review
2.3 Summary
The literature review presented in this chapter (chapter 2) allows a deeper
understanding of both the liquid sloshing concept and CFD. The following
conclusions can be summarized:
• Liquid propellant sloshing in spacecraft tanks can be of critical in-
fluence to the dynamics of the system, as well as to the AOCS. The
generated sloshing forces and torques imposed by the liquid motion
shall then be carefully analysed - section 2.1
• Lateral sloshing is the most common way of liquid sloshing. Its para-
meters can be analytical evaluated using a simplified analytical a-
pproach - subsection 2.1.1
• For non-ideal fluids damping exists and has an important role in the
liquid behaviour. The amplitude of the excited surface wave will de-
crease once the external excitation decreases - subsection 2.1.2
• For small external perturbations non-linear effects can be neglected.
However, for large perturbations or other sloshing types, such as rotary
sloshing, these effects dominate the sloshing response and shall be
accounted for - subsection 2.1.3
• In micro-gravity environments, where body forces are small, surface
tension forces are dominant at the free surface of the liquid. These
forces shall be carefully modelled for problems occurring in micro-
-gravity environments - subsection 2.1.4
• CFD is a very interesting branch of fluid dynamics, having an impor-
tant role in engineering predictions. It integrates disciplines of fluid
mechanics, mathematics and computer science - section 2.2
• CFD results are only as valid as the physical models incorporated
in the governing equations and boundary conditions. Hence, they
are subject to various error sources, which can severely influence the
accuracy of the results - section 2.2
• A set of mathematical equations - governing equations, can be used to
describe a fluid flow, representing its physical behaviour. The govern-
ing equations are based on the mass, the momentum and the energy
conservation principles - subsection 2.2.1
26
2. Literature Review
• Boundary conditions are required to dictate the particular solutions
to be obtained from the governing equations. There are two main
types of boundary conditions, the Dirichlet and the Neuman ones -
subsection 2.2.2
• A mesh represents a discretized geometry, designating the elements on
which the flow is to be solved. Hexahedral elements are preferred as
they allow increased accuracy for more complex problems and have a
reduced number of elements when compared to tetrahedral elements
- subsection 2.2.3
• The fluid flow governing equations can only be solved (for the vast
majority of the problems) using numerical methods, which find an
approximated solution to the problem. From the available numerical
methods, the FEM guarantees an ease to obtain high order accuracy
results and to handle complex geometries. However, it is more difficult
to formulate for fluid dynamics purposes and hence FVM have been
mostly used in CFD - subsection 2.2.4
• There are several solution methods used to solve a system of equa-
tions. These are normally included in the following categories: direct,
iterative, preconditioned or multigrid methods - subsection 2.2.5
• Analysing the solution of a computational method is a very important
step to be performed when using CFD. There are certain properties
which help identifying the performance of the method and the validity
of its solution - subsection 2.2.6
From the conclusions drawn in this literature review, the requirements for
this project were determined, the CFD software selection process was ruled
and the test cases to be validated with the software were defined.
In the following chapters the requirements are presented, followed by the
selection process that led to the choice of the CFD software, the software
characterization and later the implementation, evaluation and validation
of the defined test cases.
27
Chapter 3
Requirements
“Requirements are the What. Design is the How.”
- a System Engineer’s saying
“A factor present in every successful project and absent in every
unsuccessful project is sufficient attention to requirements.”
- Suzanne & James Robertson
In this chapter the requirements for the implementation of this project
are presented. These requirements were defined based on EADS Astrium’s
objectives for this M.Sc. thesis project, but also on some important con-
clusions taken from the literature review presented in chapter 2.
The requirements are divided in functional and system requirements. The
functional requirements include two types of requirements, the ones which
are intended to be accomplished during this project and the ones which
shall be considered for the CFD software choice, but do not necessarily
need to be fulfilled during the project.
3.1 Functional Requirements
• Basic success
1. The CFD software shall be able to perform sloshing analysis for
problems in which the tanks are subjected to translational accele-
rations;
28
3. Requirements
2. The CFD software shall support arbitrary tank geometries;
3. The CFD software shall be able to model real physical problems;
4. The CFD software shall consider viscosity effects in its models;
5. The CFD software shall include linear sloshing effects in its mod-
els (accurate for laminar Reynolds regimes);
6. The results of the CFD software shall be accurate in 1-g condi-
tions;
7. The results of the CFD software shall be validated for the follow-
ing test cases:
(a) Rectangular laterally excited tank;
(b) Cylindrical laterally excited tank;
(c) Cylindrical w/ hemispherical domes laterally excited tank.
• Considered for future developments
1. The CFD software shall be able to perform sloshing analysis for
problems in which the tanks are subjected to rotational accele-
rations;
2. The CFD software shall include non-linear sloshing effects (caused
by high external excitations) in its models;
3. The results of the CFD software shall be accurate for all gravity
conditions, including micro-gravity (surface tension effects dom-
inant in the liquid’s behaviour).
3.2 System Requirements
• Basic success
1. The CFD software shall be based on available open source soft-
ware;
2. The CFD software shall be based on FEM or FVM;
3. The CFD software shall be integrated with EADS Astrium’s pre-
and post- processing environments;
4. The CFD software shall interface with MATLAB R©;
5. The resulting simulation environment shall be used in any oper-
ating system (OS).
29
Chapter 4
CFD Software Selection
“The world’s an exciting place when you know CFD.”
- John Shadid
“Computational fluid dynamics has certainly come of age in
industrial applications and academia research. In the beginning
this popular field of study was primarily limited to high
technology engineering areas of aeronautics and astronautics, but
now it is a widely adopted methodology for solving complex
problems in many modern engineering fields.”
- Tu, J., Yeoh, G.H. & Liu, C.
This chapter presents the extended state-of-the-art study that was made
to find the most suitable CFD software for the purposes of this project. In
section 4.1 the selection process is introduced and later in section 4.2 the
results of this selection are presented.
4.1 Selection Process
The extended selection process to find the most suitable CFD software
was divided in several phases. In a first step, starting from the system
requirements (defined in 3.2), a list of available software which satisfied
them was comprised. In a second phase, important characteristics of the
software were defined and a selection of the solvers, based on these, was
made. The third phase of this selection comprised the choice of the top 3
solvers based on a narrower set of characteristics. Finally, in a last step,
30
4. CFD Software Selection
the final selection was made based on a compromise solution that gave
guarantees of success.
The following subsections deeply define each step that was taken in the
selection process. The characteristics of the software that were looked
into, as well as the chosen approach are presented.
4.1.1 Available Codes
To successfully pass this phase the software had to satisfy the following
criteria, extrapolated from the project’s requirements:
• Be an open source software;
• Be a CFD solver.
4.1.2 Satisfactory Codes
The characteristics which were defined as crucial to be satisfied for the
solvers to pass this phase were:
• High level of maturity;
• Use of known programming languages;
• Licensed as GNU GPL or LGPL;
• Functionalities in-line with the defined requirements;
• Existing interfaces for pre- and post- processing;
• Sufficient support documentation.
4.1.3 Top 3 Codes
The narrower code characteristics that were studied in this third phase
were:
• FEM or FVM based;
31
4. CFD Software Selection
• Mesh elements type (support for at least hexahedral and/or tetrahe-
dral);
• Parallelization supported;
• Surface tension modelled;
• Already used for liquid sloshing purposes.
4.1.4 Final Selection
From the top 3 codes a final selection was performed. At this point, the
codes were analysed individually and the main advantages and disadvan-
tages of each one were defined. A compromise solution was then made.
4.2 Results
The results of the selection process previously introduced (section 4.1) are
presented in the following subsections.
4.2.1 Available Codes
The following list of software passed the first phase:
• ADFC [15]
• arb [16]
• CFD2D [17]
• CFD2k [18]
• cfdpack [19]
• Channelflow [20]
• CLAWPACK [21]
• Code Saturne [22]
• COOLFluiD [23]
• DUNS [24]
• Dolfyn [25]
• Edge [26]
• Elmer [27]
• Featflow [28]
• FEniCS Project [29]
• freeFEM++ [30]
• Fluidity [31]
• HiFlow3 [32]
32
4. CFD Software Selection
• Gerris Flow Solver [33]
• hit3d [34]
• iNavier [35]
• ISAAC [36]
• Kicksey-Winsey [37]
• MFIX [38]
• NaSt2D-2.0 [39]
• NEK5000 [40]
• NUWTUN [41]
• OpenFlower [42]
• OpenFOAM [43]
• OpenLB [44]
• OpenFVM [45]
• PartenovCFD [46]
• PETSc-FEM [47]
• SLFCFD [48]
• SSIIM [49]
• SU2 [50]
• Tochnog [51]
• TYCHO [52]
• Typhon solver [53]
4.2.2 Satisfactory Codes
Table 4.1 presents the CFD solvers that satisfied the conditions established
for the second phase of the selection process. The former EADS Astrium
tool used to analyse sloshing - SLOSHC [54], as well as the most used com-
mercial software for this purpose - FLOW-3D [55], are also characterised.
It is important to note that a set of functionalities was defined and the
solver could only pass this phase if these were satisfied. Besides any other
functionalities that the code might have, at least these had to exist:
• Compressible flows model;
• Incompressible flows model;
• Multiphase flows model (at least two-phase flows);
• Turbulent flows model;
• Fluid structure interaction capabilities (even if by coupling with an-
other software).
33
4. CFD Software Selection
Tab
le4.
1:Sec
ond
phas
e:sa
tisf
acto
ryco
des
-co
des
and
char
acte
rist
ics.
Maturity
Lic
ense
Program
min
gLanguages
Mesh
input
form
at
Post-
Processin
gSupport
Docum
entatio
n
Code
Satu
rne
Develo
pm
ent
start
ed:
1997
Main
develo
per:
ED
FO
pen
sourc
ere
lease
d:
2007
GP
LFort
ran
C Pyth
on
scri
pts
I-deas
univ
ers
al
Gm
shP
ara
Vie
w
Inst
allati
on
Guid
eT
heory
Guid
eP
racti
cal
Use
r’s
Guid
eT
uto
rials
Cours
em
ate
rials
Elm
er
Develo
pm
ent
start
ed:
1995
Main
develo
per:
CSC
-IT
Cente
rfo
rScie
nce
Op
en
sourc
ere
lease
d:
2005
GP
LFort
ran
(Solv
er)
C/C
++
(GU
I)
AN
SY
SG
msh
Abaqus
I-deas
univ
ers
al
Para
vie
wG
nuplo
tO
cta
ve
Matl
ab
Models
,Solv
ers
,P
ara
mete
rs,
Gri
d,
GU
Iand
MA
TC
Manuals
Tuto
rials
Cours
em
ate
rials
FE
niC
SD
evelo
pm
ent
start
ed:
2003
Main
develo
per:
Sim
ula
Rese
arc
hL
ab
ora
tory
Op
en
sourc
ere
lease
d:
2003
LG
PL
C+
+P
yth
on
UF
L
Gm
shA
baqus
Exodus
Para
Vie
wB
ook
Tuto
rials
Cours
em
ate
rials
Flu
idit
yD
evelo
pm
ent
start
ed:
1990s
Main
develo
per:
Imp
eri
al
College
London
Op
en
sourc
ere
lease
d:
No
info
LG
PL
C/C
++
Pyth
on
Fort
ran
Gm
shE
xodus
Para
Vie
w
Manual
Use
r’s
Guid
eD
evelo
per’
sG
uid
eC
ours
em
ate
rials
Op
enF
OA
MD
evelo
pm
ent
start
ed:
1980s
Main
develo
per:
Op
enF
OA
Mfo
undati
on
Op
en
sourc
ere
lease
d:
2004
GP
LC
++
AN
SY
SF
luent
Gm
shI-
deas
univ
ers
al
Para
Vie
wF
luent
Tecplo
t3D
Use
rG
uid
eC
ours
em
ate
rials
Tuto
rials
SU
2D
evelo
pm
ent
start
ed:
2012
Main
develo
per:
Sta
nfo
rdU
niv
ers
ity
Op
en
sourc
ere
lease
d:
2012
GP
LC
++
Pyth
on
scri
pts
CG
NS
(data
standard
)P
ara
Vie
wT
ecplo
t
Use
r’s
Guid
eD
evelo
per’
sG
uid
eT
uto
rials
Cours
em
ate
rials
Ast
rium
SL
OSH
C
Develo
pm
ent
start
ed:
1970s
Main
develo
per:
Ast
rium
Op
en
sourc
ere
lease
d:
N/A
Ast
rium
’sin
-house
use
only
Fort
ran
None
None
(Outp
ut
as
text
file
)
Use
rM
anual
and
Desc
ripti
on
of
Alg
ori
thm
s
FL
OW
-3D
Develo
pm
ent
start
ed:
1980
Main
develo
per:
Flo
wScie
nce,
Inc.
Op
en
sourc
ere
lease
d:
N/A
Com
merc
ial
Copyri
ght
N/A
No
info
EnSig
ht
Tecplo
t
Book
Use
rM
anual
Use
r’s
technic
al
supp
ort
34
4. CFD Software Selection
In terms of functionalities FLOW-3D states to feature all the above. As
for the former SLOSHC tool, it can only compute the parameters of an
equivalent mechanical model for lateral sloshing effects in axially symmetric
containers accelerated in the direction of the symmetry axis (being this only
valid for non-rotating satellites and ideal fluids).
4.2.3 Top 3 Codes
The various CFD software selected in the second phase were once again
evaluated. Table 4.2 presents the evaluation performed for all the selected
codes.
Table 4.2: Third phase - codes and evaluated characteristics.NumericalMethods
Mesh ElementsType
SupportedParallelization
Surface TensionModelled
Used for SloshingPurposes
Code Saturne FVMHexahedralTetrahedral
Yes No infoNone found
Elmer FEMHexahedralTetrahedral
Yes YesSimilar purposes
FEniCS FEM Tetrahedral Yes No infoNone found
Fluidity FVMHexahedralTetrahedral
Yes No infoNone found
OpenFOAM FVMHexahedralTetrahedral
Yes YesYes
SU2 FVMHexahedralTetrahedral
Yes No infoNone found
AstriumSLOSHC
FEMRectangularTriangular
N/A NoYes
FLOW-3D FEMFVM
HexahedralTetrahedral
Yes YesYes
As it is possible to observe in table 4.2, the choice of the top 3 codes is
not clear as they all have pros and cons. Nevertheless, it was decided
that the FEM based solvers were preferred due to the reasons already
stated in chapter 2.2.4. Thus, from the available FEM based methods,
Elmer and FEniCS were considered the most complete ones. Even though
OpenFOAM is a FVM based solver, it has already been used for liquid
sloshing analysis purposes and thus it was also considered for the top 3
codes.
4.2.4 Final Selection
In the last step of the selection process, the top 3 codes were individually
evaluated and a set of advantages and disadvantages was defined for each
one.
35
4. CFD Software Selection
Elmer:
• Advantages:
- FEM based;
- Supports hexahedral mesh elements;
- Surface tension modelled;
- Very accurate for low Reynolds;
- On going development;
- Used for similar purposes;
- Support from developers.
• Disadvantages:
- Programming language: Fortran;
- Not so good for high Reynolds.
FEniCS:
• Advantages:
- FEM based;
- Programming language:
C/C++, Python, UFL;
- Support from developers.
• Disadvantages:
- Does not support hexahedral mesh
elements;
- Surface tension not modelled;
- Never used for similar purposes.
OpenFOAM:
• Advantages:
- Supports hexahedral mesh elements;
- Surface tension modelled;
- Programming language: C++;
- Used for liquid sloshing analysis;
- Support from developers.
• Disadvantages:
- FVM based;
- Not so good for low Reynolds.
The above mentioned advantages and disadvantages for each one of the
codes do not necessarily had the same weight in the final choice. Therefore,
the selection had to be based on a compromise solution that would allow
the choice of the most complete and suitable software for the purposes of
the project.
36
4. CFD Software Selection
Based on the extensive literature review presented in chapter 2, it was
defined that a software based on FEM had to be used. The advantages are
clear, it allows accurater results for the tests planned to be implemented
(low Reynolds), as well as it has a bigger margin for future developments.
It was also defined that the software should support both types of mesh
elements − tetrahedral and hexahedral elements.
Concluding, Elmer was considered the most suitable software to be used for
the purposes of this project, the tests to be implemented and the expected
future developments. Elmer is based on FEM, supports tetrahedral and
hexahedral mesh element types, was already used for similar purposes, shall
give accurate results for the regime in which the fluid was intended to be
treated and it is also supported by a motivated and helpful development
team.
37
Chapter 5
Elmer - Open Source Finite
Element Software
“When it comes to software, I much prefer free software, because
I have very seldom seen a program that has worked well enough
for my needs, and having sources available can be a life-saver.”
- Linus Torvalds
Elmer - Open Source Finite Element Software for Multiphysical Problems
was the chosen software to be used in the implementation of this project.
This chapter presents a brief introduction of Elmer, its functionalities,
models, solvers, interfaces, pre- and post- processing tools.
5.1 Overview
Elmer is an open source finite element software package for multiphysical
problems. Basically, it is a software package used to solve partial differential
equations. Elmer can deal with a great number of different equations, which
can be coupled in a generic manner, making Elmer a versatile tool. Being
an open source software, Elmer gives the user the possibility to modify the
existing solution procedures and thus develop new solvers for equations of
its own interest. [56]
Elmer’s development started in 1995 as part of a national CFD technology
program funded by the Finnish Funding Agency for Technology and Inno-
vation. The initial 5 years project included several partners being CSC -
38
5. Elmer - Open Source Finite Element Software
IT Center for Science [57] the main developer and the one that after this
initial period kept the project under development. In 2005, the software
package was finally released as an open source. Since then, the user com-
munity widened and the number of international users grew together with
the software.
Being a multiphysical problems software package, Elmer contains solvers
for a variety of mathematical models. The following list summarizes Elmer’s
capabilities and integrated physical models for some of the main specialized
fields (from [56]):
• Heat transfer: models for conduction, radiation and phase change;
• Fluid flow: the Navier-Stokes, Stokes and Reynolds equations, k-η
model;
• Species transport: generic convection-diffusion equation;
• Elasticity: general elasticity equations, dimensionally reduced models
for plates and shells;
• Acoustics: the Helmholtz equation, linearized Navier-Stokes equations
in the frequency domain and large amplitude wave motion of an ideal
gas;
• Electromagnetism: electrostatics, magnetostatics, induction;
• Microfluidics: slip conditions, the Poisson-Boltzmann equation;
• Levelset method: Eulerian free boundary problems;
• Quantum Mechanics: density functional theory (Kohn-Sham).
In terms of numerical methods used for approximation and linear systems
solutions, Elmer includes a great number of possibilities. The list below
summarizes the most important ones (from [56]):
• All basic element shapes in 1D, 2D and 3D with the Lagrange shape
functions of degree k ≤ 2;
• Higher degree approximation using p-elements;
• Time integration schemes for the first and second order equations;
39
5. Elmer - Open Source Finite Element Software
• Solution methods for eigenvalue problems;
• Direct linear system solvers;
• Iterative Krylov subspace solvers for linear systems;
• Multigrid solvers for some basic equations;
• ILU preconditioning of linear systems;
• Parallelization of iterative methods;
• The discontinuous Galerkin method;
• Stabilized finite element formulations, including the methods of resi-
dual free bubbles and Streamline-Upwind Petrov-Galerkin (SUPG);
• Adaptivity, particularly in 2D;
• Boundary Element Method (BEM) solvers (without multipole accel-
eration).
As most of the CFD software packages, Elmer is composed of three main
parts: the pre-processor, the solver and the post-processor. These are
separate executables that can also be used independently. Thus, the main
executables included in Elmer’s package are (more details [56]):
• ElmerGUI - graphical user interface (GUI) for Elmer;
• ElmerGrid - provides functionalities for the generation of simple
meshes and conversion of accepted file formats to the native format;
• ElmerSolver - main part of Elmer, the solver;
• ElmerPost - simple GUI post-processor.
Each one of the main parts of Elmer has its own characteristics, which could
be thoroughly explored. Nevertheless, the idea is only to provide the reader
with a general understanding of the software. Thus, the following sections
introduce Elmer’s Models, Solvers, Interfaces, Pre- and Post- processing
main characteristics. Hopefully, by the end of this chapter the reader will
have a general idea about Elmer.
40
5. Elmer - Open Source Finite Element Software
5.2 Models / Solvers
Elmer is a multiphysical problems CFD software package solver composed
of several different modules or solvers. Being a very complete software,
a description of all its solvers is totally out of the scope for this M.Sc.
thesis. Thus, only the important solvers for the work to be performed are
briefly introduced below. More information about other solvers included
in Elmer, as well as more details on those introduced here, can be found
in its manuals [58].
• FlowSolve - solves the Navier-Stokes equations (already introduced
in chapter 2.2.1);
• MeshSolve - moves the current mesh nodes so that the mesh remains
intact when a boundary is moved. It updates the mesh after each time
step;
• FreeSurfaceSolver - allows the specification of a boundary as a free
surface, which can then be solved in combination with the Navier-
-Stokes equations (FlowSolve) and the mesh update solver (Mesh-
Solve).
Elmer includes several solution methods to solve linear and non-linear sys-
tems. These are briefly introduced here together with the time discretiza-
tion strategies included in Elmer. For completeness and extra information,
please refer to [59].
Methods for linear systems
For linear systems, the solution methods included in Elmer fall into two
large categories: direct methods and iterative methods (already briefly
introduced in chapter 2.2.5).
Direct methods
Elmer offers two possibilities to use direct methods:
• The Linear Algebra Package (LAPACK) collection of subroutines [60];
41
5. Elmer - Open Source Finite Element Software
• The Unsymmetric Multifrontal Sparse LU Factorization Package (UMF-
PACK) set of routines [61].
Iterative methods
The iterative methods available in Elmer can be divided in two main cat-
egories, the preconditioned Krylov subspace methods and the multilevel
methods.
Preconditioned Krylov methods
Elmer’s solver includes the following set of Krylov subspace methods:
• Conjugate Gradient (CG);
• Conjugate Gradient Squared (CGS);
• Biconjugate Gradient Stabilized (BiCGStab);
• BiCGStab(l);
• Transpose-Free Quasi-Minimal Residual (TFQMR);
• Generalized Minimal Residual (GMRES);
• Generalized Conjugate Residual (GCR).
A deeper explanation on the integration of these methods in Elmer can
be found in [59]. For a detailed explanation on some of these methods
see [62, 63].
In terms of preconditioning strategies (already introduced in chapter 2.2.5),
Elmer includes some basic strategies (see [59] for details):
• ILU(N) preconditioners;
• ILUT preconditioners;
• Preconditioning by multilevel methods;
• Block pPreconditioning.
42
5. Elmer - Open Source Finite Element Software
Multilevel methods
Even though multilevel methods (introduced in chapter 2.2.5) can be ap-
plied to define preconditioners for the Krylov subspace methods, they are
iteration methods on their own. In Elmer’s solver, two different multilevel
method approaches are available (for details see [59]):
• Geometric Multigrid (GMG);
• Algebraic Multigrid (AMG).
Methods for non-linear systems
The non-linearity of a system might be intrinsically related with the charac-
teristics of the equations to be solved, but it can also result from non-linear
material parameters that depend on the solution. Thus, in Elmer, the ap-
proach used in the linearization of non-linear systems changes from one
solver to another. More details on these strategies can be found in [58, 59].
As an example, for the Navier-Stokes solver there are two different methods
included in Elmer:
• Picard linearization;
• Newton linearization.
Time discretization strategies
The integration of time dependent systems can be performed in Elmer
using one of the following methods (see [59] for more details):
• Crank-Nicolson method;
• Backward Differences Formulae (BDF) of several orders.
43
5. Elmer - Open Source Finite Element Software
5.3 Interfaces
As most open source CFD codes, Elmer can be used directly from the
command line by calling the solver executable - ElmerSolver. The pre- and
post- processing executables, respectively, ElmerGrid and ElmerPost, can
also be called from the command line. Nevertheless, oppositely to most
open source CFD software, Elmer also includes a modern programmable
graphical user interface - ElmerGUI.
In this section, a brief description of Elmer’s GUI capabilities is presented.
The basic concepts about how to run Elmer from the command line are
also introduced.
5.3.1 Graphical User Interface
ElmerGUI is a very complete program, capable of performing almost all
the tasks that can be performed when running Elmer from the command
line. The GUI is capable of importing finite element mesh files in several
formats, generate finite element partitionings for various geometry input
files, setup systems of PDEs to be solved and export model data and results
for ElmerSolver and ElmerPost to solve and post-process, respectively. [64]
Figure 5.1: ElmerGUI main window.
44
5. Elmer - Open Source Finite Element Software
Figure 5.1 shows Elmer’s GUI main window. Several main menus exist in
this window, each one having its purposes. Below an introduction to these
menus (see [64] for more details):
• File - allows the user to load a saved project or to start a new one by
loading a mesh file. The GUI’s definitions and the save buttons are
also located in this menu;
• Mesh - allocates the mesh configuration buttons;
• Model - menu that allows the user to stipulate the model defini-
tions. The Setup, Equation, Material, Body force, Initial condition
and Boundary condition sub-menus are located here. Defining the pa-
rameters located inside each one of these sub-menus defines the model
to be simulated;
• View - allows the user to set view preferences;
• Sif - allows the generation of the Solver Input file (Sif) based on the
Model defined properties. The user can also manually edit the Sif;
• Run - used to start the solver or the post-processor (ElmerPost);
• Help - help menu.
5.3.2 Command Line
Elmer can also be run directly from the command line. To do this, a test
folder needs to be created and at least the following files need to exist:
• Mesh file;
• Solver input file.
The mesh can be generated using ElmerGrid or it can also be converted
from one of Elmer’s mesh accepted input files to the native Elmer mesh file
using ElmerGrid. More details about ElmerGrid are provided in section 5.4
together with Elmer’s post-processing - ElmerPost.
The .Sif file provides the solver with a precise description of the problem.
It contains user-prepared input data that specifies the location of the mesh
45
5. Elmer - Open Source Finite Element Software
files, controls the selection of the physical models and defines the material
parameters, the boundary conditions, the initial conditions, the solver’s
stopping tolerances, etc. The file is organized into different sections which
can form the following general structure:
• Header
• Simulation
• Constants
• Body
• Material
• Body Force
• Equation
• Solver
• Boundary Condition
• Initial Condition
All the required model parameters are defined in the correspondent section
of the Sif. For details on each section, please refer to [59]. To note that,
as stated before, a mesh file representing the geometry of the problem is
also required to completely define the problem, this shall be called inside
the Sif header section.
To start ElmerSolver from the command line, the executable as to be called
together with the .Sif file:
$ ElmerSolver Test.sif
Issuing this command runs the solver and saves the results in the chosen
directory. The results can then be visualized and evaluated using Elmer’s
post-processing.
46
5. Elmer - Open Source Finite Element Software
5.4 Pre- and Post- Processing
As stated before, Elmer includes its own pre- and post- processing tools −ElmerGrid and ElmerPost. In this section a brief introduction to these is
presented.
5.4.1 Pre-Processing
ElmerGrid is one of the executables included in Elmer’s software package.
It is responsible for the pre-processing, consisting of a simple mesh gener-
ator and mesh manipulation utility. It has the capability to read meshes
generated by other programs and manipulate and convert them to a for-
mat accepted by ElmerSolver. The following mesh formats are accepted
by ElmerGrid:
• .grd : ElmerGrid file format;
• .mesh.* : Elmer input format;
• .ep : Elmer output format;
• .ansys : Ansys input format
• .inp : Abaqus input format by Ideas;
• .fil : Abaqus output format;
• .FDNEUT : Gambit (Fidap) neutral file;
• .unv : Universal mesh file format;
• .mphtxt : Comsol Multiphysics mesh format;
• .dat : Fieldview format;
• .node,.ele: Triangle 2D mesh format;
• .mesh : Medit mesh format;
• .msh : GID mesh format;
• .msh : Gmsh mesh format;
• .ep.i : Partitioned ElmerPost format.
47
5. Elmer - Open Source Finite Element Software
These formats can be converted using ElmerGrid to Elmer’s native mesh
format: .mesh.* .
For more details about ElmerGrid, its capabilities and how to use it, please
refer to its manual [65].
5.4.2 Post-Processing
ElmerPost is Elmer’s included post-processing executable. It has the ca-
pability to read the results output by Elmer in the .ep format. Using
this tool, the results can be visualized and evaluated. Figure 5.2 presents
ElmerPost’s main window and graphics window.
Figure 5.2: ElmerPost main window & graphics window.
Even though this simple post-processing tool is enough to evaluate the
simple test cases, other more advanced post-processing tools can also be
used. Therefore, it is important to note that the results of Elmer’s solver
can be saved in several formats. Currently, the supported output formats
include GiD, Gmsh, VTK legacy, XML coded VTK file bearing the suffix
VTU and Open DX. [58]
48
Chapter 6
Simulation Environment Setup
“A simulation is a concrete abstraction of the relevant features of
some real world problem.”
- Unknown
This chapter presents the complete simulation environment including a
description of the simulation flow and the used pre- and post- processing
tools.
As defined in the system requirements for this project (chapter 3.2), the
CFD software had to be integrated with EADS Astrium’s pre- and post-
processing tools, had to interface with MATLAB R© and the resulting sim-
ulation environment had to be used in any operating system.
In the subsequent sections, the final simulation environment which satisfies
the defined requirements is presented. The final software package includes
the CFD solver - Elmer, as the main part, but it also integrates specific
pre- and post- processing tools. First the simulation flow is introduced and
later the pre- and post- processing methods are described.
6.1 Simulation Flow
To satisfy this project’s system requirements (chapter 3.2), a virtual ma-
chine player - VMware R© PlayerTM
, compatible with several operating sys-
tems, was used. FedoraTM
18 OS was built in the virtual machine and the
CFD software - Elmer - was installed in this Linux based OS, figure 6.1.
49
6. Simulation Environment Setup
Figure 6.1: Software installation diagram.
As mentioned in chapter 5, Elmer includes its own pre- and post- pro-
cessing tools: a mesh converter and generator - ElmerGrid, and a simple
post-processing tool - ElmerPost. Nevertheless, some other pre- and post-
processing tools were also to be integrated together with Elmer. Thus, the
diagram of figure 6.2 presents the final, complete, software simulation flow.
Figure 6.2: Simulation flow.
50
6. Simulation Environment Setup
As seen in figure 6.2, the pre-processing part of the simulation flow is re-
sponsible for the generation or conversion of a mesh file to Elmer’s native
mesh file format. In the other extrema, the post-processing part is re-
sponsible for presenting the results and developing further analysis. In
the following sections, the used pre- and post- processing methods are de-
scribed.
6.2 Pre-Processing Methods
As seen in chapter 5.4, ElmerGrid has the capability to generate meshes for
simple geometries. Thus, if a problem is simple enough, its geometry can
be meshed using ElmerGrid. However, if the geometry is rather complex,
more complete mesh generators can be used.
During this project’s implementation, Gmsh [66] and CubitTM
(exporting
the mesh as .unv) were used to generate some of the meshes. Nevertheless,
any other mesh generator may be used, as long as it can export in one
of the formats accepted by ElmerGrid (chapter 5.4) - which can act as a
converter to Elmer’s mesh native format.
6.3 Post-Processing Methods
In terms of post-processing, as mentioned in chapter 5.4, Elmer includes
its own post-processing tool - ElmerPost. However, this is a very simple
tool that did not completely serve the purposes of this project. Therefore,
ParaView [67] was found to be the most suitable tool to visualize and
analyse the results of the simulations.
Using ParaView, the results obtained with ElmerSolver can be directly
visualized and analysed. If needed, a specific sloshing post-processing pro-
vided by EADS Astrium and based on MATLAB R© can also be used. This
specific post-processing has the capability to receive inputs from ParaView,
process them and have as output the following sloshing parameters:
• Liquid sloshing natural frequencies;
• Sloshing modes recognition;
51
6. Simulation Environment Setup
• Sloshing wave amplitudes;
• Mass of liquid participating in the sloshing movement;
• Damping ratio.
As this post-processing is based on MATLAB R©, it is located outside the
virtual machine for convenience. Therefore, figure 6.3 presents the final
simulation environment installation diagram.
Figure 6.3: Complete software installation diagram.
52
Chapter 7
Test case 1: Rectangular Tank
“Testing is a process of gathering information by making
observations and comparing them to expectations.”
- Dale Emery
In this chapter, the tests developed for the rectangular laterally excited
tank test case are introduced. As mentioned before, liquid sloshing inside
a laterally excited rectangular tank is one of the simplest cases of sloshing
that may occur. The simple geometry of the tank allows several relatively
accurate analytical solutions - chapter 2.1.1. Having as a basis the nume-
rical and experimental results available in literature [6, 68, 69, 70, 71] for
sloshing tests performed with rectangular tanks, three different tests were
defined to be implemented, their main objectives were:
• Test A: Evaluate the first antisymmetric sloshing frequency of a liquid
sloshing inside a 2D tank;
• Test B: Recognize the natural sloshing frequencies and modes of a
liquid sloshing inside a 2D tank;
• Test C: Obtain the first antisymmetric sloshing frequency of liquid
water sloshing inside a 3D tank.
53
7. Test case 1: Rectangular Tank
These tests are presented in the following sections. Each test is prima-
rily defined, then its implementation is explained and later the results are
presented, evaluated and validated.
7.1 Test A
7.1.1 Test Definition
The evaluation of the first natural antisymmetric sloshing frequency was
addressed in this test - similarly performed by [68, 69]. For that, a two-
-phase flow in a two-dimensional sloshing tank was considered. The tank
was subjected to a vertical acceleration g = −1 m/s2 and geometrically it
was defined with width a = 1 m and height H = 1.5 m.
The properties of the defined fluids [68, 69] were:
• Fluid 1:
– µ− = 1.0 Pa · s
– ρ− = 1000.0 kg/m3
• Fluid 2:
– µ+ = 0.01 Pa · s
– ρ+ = 1.0 kg/m3
A no-slip boundary condition was prescribed to the bottom of the tank
and slip boundary conditions were defined along the walls. Initially, the
velocity field was assumed to be zero.
The interface separating the two fluids was considered a free surface and it
was initially given as y = 0.26 + 0.1 sin(πx), where x and y have its origin
at the center of the tank. The simulation was performed for t = 20 s.
54
7. Test case 1: Rectangular Tank
7.1.2 Implementation
To implement test A, two different approaches were used. In one approach,
the input mesh was defined as a rectangular tank in which the free surface
shape of the liquid was given by an initial condition - tests a) and c);
oppositely, in the other approach, the input mesh was already generated
with the intended skewed shape - tests b) and d). These two approaches
were implemented using both 2-dimensional meshes (tests a) and b)) and
3-dimensional meshes with single elements in the third direction (tests c)
and d)).
For tests b), c) and d), the origin of the coordinate system was defined to
be located in the center of the tank at a distance equal to 0.75 m from the
bottom. As for test a), its origin was defined to be located at the bottom
center of the tank.
The meshes for these tanks were generated with 40 × 40 elements for test
a), 32 × 32 elements for test b), 26 × 1 × 32 elements for test c) and 31
× 1 × 47 elements for test d).
For all the four tests, the time step sizes were defined to be 0.02 s and the
number of steps 1000. The total simulation time was, as intended, 20 s.
7.1.3 Results & Evaluation
The results obtained for tests A a), b), c) and d) are presented and dis-
cussed in this subsection. Due to their similarity, the results are presented
only for test A a). Nevertheless, the obtained results for tests A b), c) and
d) can be found in appendix A.
To give an impression of the computational effort required by the CFD
software to run these tests, it is important to state the architecture in
which the tests were run, as well as the time that was required by each
simulation. Therefore, the simulations were run in a Samsung Series 9
NP900X4C-A03PT notebook in which a virtual machine was mounted -
see chapter 6 for details. This system is composed by an Intel R© CoreTM
i7-3517U CPU processor (4M Cache, up to 3.00 GHz) and 8.0 GB DDR3
RAM memory. However, the virtual machine was running only on 6.0 GB
of RAM. Thus, for reference, the fastest test performed during this project
- test A a) required approximately 7 minutes to be simulated.
55
7. Test case 1: Rectangular Tank
The first step on the evaluation of the obtained results was to verify the
correct implementation of the tests even before they had started, this is, at
t = 0s - when the fluids were still in rest. Figure 7.1 shows both the fluid
pressure distribution and the free surface position obtained for test A a). A
maximum pressure of about P = 1010 Pa was expected: P = ρgh, where
ρ = 1000kg/m3, g = 1m/s2 and h = 1.01m (non-disturbed fluid). For
the four tests, the obtained values approximately match the expected ones
(even if they are slightly different between tests). As intended, the initial
free surface position is the same for all the tests, even though different
implementations were used.
Figure 7.1: Rectangular tank - test A a): pressure at t = 0s.
The evolution of the free surface position is shown in figure 7.2. When
comparing the results of the different tests, it is possible to conclude that
these slightly differ from test to test. Nevertheless, the results are very
similar for the test pairs that had the same free surface initial position
implementation - tests a) & c) and b) & d). This validates the fact that
a 2D mesh implementation guarantees the same results as the use of a 3D
mesh with single elements in the third direction.
56
7. Test case 1: Rectangular Tank
(a) t = 0.6s (b) t = 1.2s
(c) t = 1.8s
(d) t = 2.4s (e) t = 3.0s
Figure 7.2: Rectangular sloshing tank - test A a): free surface shape evolution.
57
7. Test case 1: Rectangular Tank
The variation of the liquid’s CoG position is presented for test a) in fi-
gure 7.3.
Figure 7.3: Rectangular tank - test A a): CoG plots.
In the presented plot it is possible to visualize that, as the liquid sloshes, the
CoG position in the x-direction describes an harmonic behaviour, moving
together with the liquid slosh wave amplitude. It is possible to conclude
that in those tests where a 2D mesh was used (tests a) and c)) there is
no variation in the third direction, as this is non-existent. In tests b) and
d), where 3D meshes were used, the variation in the third dimension is,
as expected, zero - the visualized disturbances in test d) are only due to
computational errors and thus negligible.
In the vertical direction, a decrease in the CoG position is seen. This
decrease represents a loss of liquid’s mass that occurs due to the fact that
in FEM mass conservation is not guaranteed. For long simulations, the
accumulation of this error can lead to a significant loss of mass that might
not be acceptable. [72]
Therefore, this mass loss shall be kept within reasonable limits. In these
particular sloshing problems, the mass of liquid participating in the sloshing
behaviour was not immediately affected by the general loss of total mass in
the liquid. Hence, a reasonable maximum limit was defined to be located
58
7. Test case 1: Rectangular Tank
below the 5 to 10 % mass loss, depending on the external perturbation and
the liquid height inside the tank.
In the above presented tests, the percentage of mass change was:
• a) ≈ 1, 2 %
• b) ≈ 12 %
• c) ≈ 2 %
• d) ≈ 12 %
The sloshing wave amplitude plot presented in figure 7.4 for test a) shows
the evolution of the sloshing wave amplitude measured in the most top left
point of the tank. From the results obtained in the different tests, it is
possible to conclude that approximately the same behaviour is reproduced
for all. Nevertheless, once again, the tests that had the same free surface
initial position implementation show a much identical behaviour.
Figure 7.4: Rectangular tank - test A a): sloshing amplitude plot.
Table 7.1 presents the resulting first natural antisymmetric sloshing fre-
quencies obtained for tests A a), b), c) and d).
Table 7.1: Results - rectangular tank: test A.
TestFirst natural antisymmetricsloshing frequency [Hz]
a) 0.279b) 0.283c) 0.279d) 0.283
59
7. Test case 1: Rectangular Tank
As seen in table 7.1, the first natural antisymmetric sloshing frequencies
for the different tests are very similar. Nevertheless, due to the constrains
mentioned before, tests a) and c) are validated and assumed more accurate
than tests b) and d). Hence, for test A, the resulting first antisymmetric
sloshing natural frequency is f = 0.279 Hz.
A comparison of the obtained result with the analytical solution of Abram-
son [5] and the results of similar numerical tests available in literature [68,
69] is presented in table 7.2.
Table 7.2: Comparison of results - rectangular tank: test A.
First natural antisymmetricsloshing frequency [Hz]
Abramson [5] 0.282Fries [69] 0.279
Rasthofer et al. [68] 0.274Elmer 0.279
From the results presented in table 7.2 it is possible to verify that the
obtained result is in good agreement with the results available in literature
for similar tests.
7.2 Test B
7.2.1 Test Definition
The evaluation and recognition of the liquid’s natural sloshing frequencies
and modes was the objective of this test - similarly performed by [70, 71].
A two-phase flow in a two-dimensional tank was considered. The width of
the container was a = 1.0 m and the depth of liquid inside the tank was
also h = 1.0 m.
The tank was subjected to a vertical acceleration g = −9.81 m/s2 and
laterally excited. The liquid was to be considered non-viscous with density
ρ = 1000.0 kg/m3.
60
7. Test case 1: Rectangular Tank
Slip boundary conditions were defined along the tank walls and a no-slip
boundary condition was assigned to the bottom. Initially, the velocity field
was assumed to be zero. The surface of the liquid was considered a free
surface initially in rest.
7.2.2 Implementation
Test B was implemented using two different techniques: one more experi-
mentally oriented - a) and another more convenient for numerical simula-
tions - b). These different approaches are introduced below:
a) Experimental Approach
In sloshing experiments, the general technique used to measure the liquid’s
first antisymmetric natural frequency is to oscillate the tank at low ampli-
tude and record the frequency at which the undistorted wave shape reaches
the maximum amplitude without rotation (curl = 0). When the external
frequency matches the first antisymmetric sloshing natural frequency, the
amplitude of the wave reaches its maximum.
Therefore, this same approach was used to find the liquid’s first antisym-
metric natural frequency. A 2-dimensional mesh was used and after some
initial tests, the optimal amplitude of the lateral harmonic acceleration was
found to be 0.6 m/s2 for a time step equal to 0.02 s and a simulation time
of 5 s (250 time steps). The 2D mesh was defined to have 40 × 40 elements
and the origin of the coordinate system was defined to be the center of the
liquid.
Knowing that the expected analytical sloshing frequency was 0.88 Hz
(Abramson [5]), the external frequency was varied between fext = 0.84
Hz and fext = 0.90 Hz.
Because Elmer is not capable of solving the Euler’s equations for inviscid
fluids, the liquid fluid was considered to be liquid water (µwater = 1.0 ·10−3
Pa·s) and the fluid at the free surface was defined to be air (µair = 1.0·10−5
Pa · s and ρair = 1.2 kg/m3), both at Normal Temperature and Pressure
(NTP) conditions. Thus, the first antisymmetric sloshing frequency was
expected to be smaller than that of an ideal fluid.
61
7. Test case 1: Rectangular Tank
It is important to note that with this approach only the first natural an-
tisymmetric frequency of the liquid sloshing could be found. To recognize
the different sloshing modes and respective natural frequencies, a different
approach had to be used.
b) Numerical Approach
In this approach, a random harmonic acceleration was used as the external
excitation, A0 · sin(2πfext · t). The frequency and amplitude of the signal
were chosen carefully to ensure convergence and diminished errors during
the simulation. After some initial tests the amplitude was defined to be
A0 = 0.5 m/s2 and the external frequency fext = 1.20 Hz for a simulation
that ran for 10.5 s using a time step equal to 0.01 s (1050 steps). For con-
venience, a 3-dimensional mesh with single elements in the third dimension
was used. The 3D mesh was defined to have 15 × 1 × 15 elements and the
origin of the coordinate system was defined at the center of the liquid.
The increased number of steps allowed the relatively accurate use of the
Power Spectral Density (PSD) post-processing included in Astrium’s slo-
shing tools. Using the PSD, it was possible to recognize the different slosh-
ing modes and respective natural frequencies.
As in approach a), the liquid fluid was also treated as liquid water and the
fluid at the free surface considered air. Therefore, slightly different natural
sloshing frequencies were, once again, expected.
7.2.3 Results & Evaluation
The results obtained for test B using the two different approaches are
presented and discussed in this subsection.
Similarly to what was done for test A, the validation of the liquid’s pres-
sure distribution is the first point to be checked to evaluate a correct test
implementation. In figures 7.5 and 7.6, the pressure distributions are pre-
sented together with the free surface positions of the liquid still in rest. As
expected, a maximum pressure of about P = 1000 ·9.81 ·1 = 9810 Pa, was
obtained for both cases.
62
7. Test case 1: Rectangular Tank
Figure 7.5: Rectangular tank - test B a): pressure at t = 0s.
Figure 7.6: Rectangular tank - test B b): pressure at t = 0s.
Seeing that for test a) an experimental approach was used, the results
obtained for each external frequency tested would have to be individually
post-processed to find the sloshing wave amplitude (using Astrium’s tool).
This procedure would have been very slow and inefficient. Hence, knowing
that the variation of the position of the liquid’s CoG in the x-direction
is proportional to the sloshing wave amplitude, the post-processing and
visualization of the results could be performed using ParaView. This tool
allows the simultaneous visualization of the results for different tests and
thus different external excitation frequencies can be evaluated at the same
time in a much more efficient process.
Thus, the maximum displacement of the CoG position of the liquid in the
x-direction is what is looked into. In relative terms, the external frequency
that corresponds to the maximum wave amplitude without rotation also
corresponds to the highest CoG displacement. Thus, to find the maxi-
mum wave amplitude without rotation happening, the following criteria
was used:
• Visual evaluation of the liquid behaviour to find the maximum sloshing
wave amplitude without rotation;
• Verification of the occurrence of a sudden increase in mass loss after
rotation occurs.
63
7. Test case 1: Rectangular Tank
Figure 7.7 presents the CoG x-displacement for the three excitation fre-
quencies that gave the highest sloshing wave amplitudes.
(a) t = 0s to t = 5s (250 time steps) (b) Zoom: t = 3.48s
Figure 7.7: Rectangular tank - test B a): CoG x-coord. Vs time for differentexcitation frequencies.
In plot a), the entire simulation time is presented and the point considered
to be correspondent to the maximum sloshing wave amplitude without
rotation is marked with a vertical green bar crossing the time of occurrence.
In plot b), a zoom of this point is presented. As it is possible to see,
fext = 0.86 Hz gives the highest wave amplitude. Thus, the first natural
antisymmetric sloshing frequency for this test is found to be f = 0.86 Hz.
The position of the free surface at the time for which the maximum wave
amplitude is reached is shown in figure 7.8.
Figure 7.8: Rectangular tank - test B a): maximum wave amplitude (t = 3.48s).
For test b), where a more suitable numerical procedure was used, the po-
sition of the CoG of the liquid is presented in figure 7.9.
64
7. Test case 1: Rectangular Tank
Figure 7.9: Rectangular tank - test B b): CoG plots.
It is possible to visualize in the plots that the CoG changes in the y-di-
rection are only due to computational errors and thus negligible. In the
vertical direction (z-direction), there is some mass loss, but very small and
also negligible for the simulation time. In the x-direction it is possible to
see that there is a repeatable pattern of the CoG oscillation. Figure 7.10
shows this pattern in higher detail.
Figure 7.10: Rectangular tank - test B b): CoG x-coordinate Vs time for fext = 1.20Hz.
65
7. Test case 1: Rectangular Tank
It is possible to see that after each cycle, the sloshing movement stalls for
a small time and again the liquid starts sloshing. It is believed that this
happens due to a superposition of waves that lead to the sloshing stall and
restart.
The previously discussed plot of figure 7.7 a) presents only part of the
cycle that is repeated (as seen in figure 7.10). Nevertheless, because in the
experimental approach the objective was to reach the maximum sloshing
wave amplitude in the smallest simulation time, ensuring convergence, the
entire cycle and the repeated pattern were not of significant importance.
The plot presented in figure 7.11 intends to show the above mentioned
proportionality between the CoG displacement in the x-direction and the
sloshing wave amplitude.
Figure 7.11: Rectangular tank - test B b): sloshing amplitude plot.
As it is possible to see, the same behaviour occurs in both plots (figures 7.10
and 7.11) - the waves are inverted by 180 degrees.
In figure 7.12, the PSD plot that presents the antisymmetric modes sloshing
frequencies is shown. Figure 7.13 presents the same PSD plot but for the
symmetric modes. To find the antisymmetric modes sloshing frequencies
the PSD was measured in the top most left point of the tank, as to find
the symmetric modes it was measured in the top center point.
66
7. Test case 1: Rectangular Tank
Figure 7.12: Rectangular tank - test B b): PSD plot (measured at the top leftpoint of the tank).
Figure 7.13: Rectangular tank - test B b): PSD plot (measured at the top centerpoint of the tank).
The following frequencies are highlighted in the PSD plot of figure 7.12:
• f = 0.857 Hz - 1st antisymmetric mode sloshing frequency;
• f = 1.238 Hz - 1st symmetric mode sloshing frequency - possible to
visualize because the sloshing wave is not completely symmetric and
thus the incidence of this frequency is also noticeable in the top most
left point of the tank;
• f = 1.524 Hz - 2nd antisymmetric mode sloshing frequency.
67
7. Test case 1: Rectangular Tank
For the PSD plot of figure 7.13 the following frequencies are highlighted:
• f = 0.857 Hz - 1st antisymmetric mode sloshing frequency - less
stronger than when measured in the top most left point, but still
visible;
• f = 1.238 Hz - 1st symmetric mode sloshing frequency;
• f = 1.809 Hz - 2nd symmetric mode sloshing frequency;
• Other visible frequencies - correspond to the cycle repetition patterns.
The uncertainty of the obtained frequencies is approximately ±0.095 Hz
- increased resolution could be achieved by increasing the number of time
steps (N) or the time step (∆t) itself - frequency resolution ∆f = 1N ·∆t
.
As a summary, the obtained results for the sloshing frequencies are pre-
sented in table 7.3.
Table 7.3: Results - rectangular tank: test B.
Mode Sloshing natural frequencies [Hz]
Test a) Test b)
n=1 0.86 0.857m=1 - 1.238n=2 - 1.524m=2 - 1.809
As it was expected, the first natural antisymmetric sloshing frequencies are
approximately the same for both tests, even though the one obtained with
approach a) is more accurate. Nevertheless, because finding the different
sloshing frequencies and recognizing the different sloshing modes were the
objective of this test, test b) is considered for evaluation purposes.
Table 7.4 presents a comparison of the obtained results, with the calculated
analytical solutions from Abramson [5] and the numerical results of similar
tests available in literature [70, 71].
68
7. Test case 1: Rectangular Tank
Table 7.4: Comparison of results - rectangular tank: test B.
Mode Slosh frequencies [Hz]
Abramson [5] N. C. Pal et al. [70] P. Pal et al. [71] Elmer
n=1 0.881 0.88 0.883 0.857 / 0.86m=1 1.249 1.26 1.253 1.238n=2 1.530 1.54 1.545 1.524m=2 1.767 1.80 1.786 1.809
The results presented above (table 7.4) attest a good agreement of the
obtained results with those available in literature. This is so, despite the
fact that the presented results are for tests in which idealized fluids were
used instead of liquid water. Thus, the small noticeable differences in the
results were already expected.
7.3 Test C
7.3.1 Test Definition
The objective of this test was to obtain the first antisymmetric sloshing
frequency along the longer and shorter directions of a three-dimensional
rectangular tank filled with water - similarly performed by [6].
The width of the container was a = 0.270 m, the breadth b = 0.135 m and
the height H = 0.300 m. The tank was subjected to a vertical acceleration
g = −9.81 m/s2 and laterally excited.
A two-phase flow with air and water fluids was considered. The material
properties of these fluids were:
• Water:
– µwater = 1.0 · 10−3 Pa · s
– ρwater = 1000.0 kg/m3
• Air:
– µair = 1.0 · 10−5 Pa · s
– ρair = 1.2 kg/m3
69
7. Test case 1: Rectangular Tank
Slip boundary conditions were defined along the tank walls and a no-slip
boundary condition was assigned for the bottom. Initially, the velocity
field was assumed to be zero. The surface of the liquid was also considered
a free surface initially in rest.
Different water depths were considered:
• h1 = 0.050 m
• h2 = 0.100 m
• h3 = 0.150 m
• h4 = 0.200 m
• h5 = 0.250 m
7.3.2 Implementation
In this test the experimental approach introduced in test B was used -
approach a). Only the first natural antisymmetric sloshing frequency was
to be found for the different water depths. Therefore, this approach was
more experimentally oriented and less time consuming.
Once again, some initial tests were performed and the optimal external
amplitudes were found. Knowing the expected sloshing natural frequencies
from the analytical solutions of Housner [4], the frequency ranges were
chosen. For each test, the following amplitude, external frequency range
and number of mesh elements were used:
a) Longer Direction
i) h = 0.050 m
• A0 = 0.2 m/s2;
• fext = 1.20 Hz to fext = 1.28 Hz;
• Mesh elements: 20 × 10 × 7.
70
7. Test case 1: Rectangular Tank
ii) h = 0.100 m
• A0 = 0.2 m/s2;
• fext = 1.50 Hz to fext = 1.58 Hz;
• Mesh elements: 20 × 10 × 14.
iii) h = 0.150 m
• A0 = 0.3 m/s2;
• fext = 1.60 Hz to fext = 1.68 Hz;
• Mesh elements: 20 × 10 × 21.
iv) h = 0.200 m
• A0 = 0.3 m/s2;
• fext = 1.64 Hz to fext = 1.72 Hz;
• Mesh elements: 20 × 10 × 28.
v) h = 0.250 m
• A0 = 0.3 m/s2;
• fext = 1.64 Hz to fext = 1.72 Hz;
• Mesh elements: 20 × 10 × 35 .
The simulations ran for 5 s using a time step equal to 0.02 s (250 steps).
b) Shorter Direction
i) h = 0.050 m
• A0 = 0.6 m/s2;
• fext = 2.14 Hz to fext = 2.22 Hz;
• Mesh elements: 10 × 20 × 7.
71
7. Test case 1: Rectangular Tank
ii) h = 0.100 m
• A0 = 0.6 m/s2;
• fext = 2.32 Hz to fext = 2.40 Hz;
• Mesh elements: 10 × 20 × 14.
iii) h = 0.150 m
• A0 = 0.7 m/s2;
• fext = 2.34 Hz to fext = 2.42 Hz;
• Mesh elements: 10 × 20 × 21.
iv) h = 0.200 m
• A0 = 0.7 m/s2;
• fext = 2.34 Hz to fext = 2.42 Hz;
• Mesh elements: 10 × 20 × 28.
v) h = 0.250 m
• A0 = 0.7 m/s2;
• fext = 2.34 Hz to fext = 2.42 Hz;
• Mesh elements: 10 × 20 × 35.
The simulations ran for 2.5 s using a time step equal to 0.02 s (125 steps).
It is important to note that, for all the different liquid heights inside the
tank, the origin of the coordinate system was defined to be the center of
the liquid.
72
7. Test case 1: Rectangular Tank
7.3.3 Results & Evaluation
The results obtained for test C are presented and discussed in this subsec-
tion. Due to their similarity, the results are presented here only for the
cases in which the water depth inside the tank was h = 0.050 m. Never-
theless, the obtained results for the other considered water depths can be
found in appendix B.
Following the same procedure adopted in tests A and B, the first validation
step of test C was to check the liquid’s pressure distribution inside the
containers.
(a) h = 0.050m (b) h = 0.100m
(c) h = 0.150m
(d) h = 0.200m (e) h = 0.250m
Figure 7.14: Rectangular tank - test C: pressure at t = 0s for different waterdepths.
73
7. Test case 1: Rectangular Tank
As it is possible to visualize in figure 7.14, the maximum pressure values
match the analytical expected values for which one of the different water
depths inside the tank.
As a reference, the free surface position for each one of the tests and water
depths inside the tank at t = 0s was evaluated. For the test in which
h = 0.050m and the external excitation occurs along the longer direction
of the tank, this is shown in figure 7.15.
Figure 7.15: Rectangular tank - test C - h = 0.050m longer direction: free surfaceshape at t = 0s.
For the same water depth, but with the excitation occurring along the
shorter direction, the results are presented in figure 7.16.
Figure 7.16: Rectangular tank - test C - h = 0.050m shorter direction: free surfaceshape at t = 0s.
Similarly to test B a), an experimental approach was also used to find the
first natural antisymmetric sloshing frequency. Therefore, the same results
as in test B a) were obtained for each one of the tests and water depths.
Hence, the explanations given in test B a) to describe these results are also
valid here.
In the plots of figures 7.17 (longer direction test) and 7.18 (shorter di-
rection test), the position of the CoG in the x-direction is presented and
the maximum wave sloshing amplitude marked and zoomed to identify the
external frequency that matches the first natural antisymmetric sloshing
frequency.
74
7. Test case 1: Rectangular Tank
(a) t = 0s to t = 5s (250 time steps) (b) Zoom: t = 2.82s
Figure 7.17: Rectangular tank - test C - h = 0.050m longer direction: CoG x-coord.Vs time for different excitation frequencies.
(a) t = 0s to t = 2.5s (125 time steps) (b) Zoom: t = 1.60s
Figure 7.18: Rectangular tank - test C - h = 0.050m shorter direction: CoGx-coord. Vs time for different excitation frequencies.
Figures 7.19 (longer direction test) and 7.20 (shorter direction test) present
the correspondent free surface positions at the time in which the maximum
sloshing wave amplitude occurred.
Figure 7.19: Rectangular tank - test C - h = 0.050m longer direction: maximumwave amplitude (t = 2.82s).
75
7. Test case 1: Rectangular Tank
Figure 7.20: Rectangular tank - test C - h = 0.050m shorter direction: maximumwave amplitude (t = 1.60s).
A summary of the obtained first natural antisymmetric sloshing frequencies
is presented for all tests in table 7.5.
Table 7.5: Results - rectangular tank: test C.
Depth of water [m] First natural antisymmetric sloshing frequencies [Hz]
along longer direction along shorter direction
0.050 1.23 2.150.100 1.53 2.330.150 1.62 2.350.200 1.66 2.350.250 1.68 2.35
A comparison of these results with the analytical solutions of Housner [4]
and the numerical and experimental data available in literature [6] for
similar tests is presented in tables 7.6 for the longer direction tests and 7.7
for the shorter direction ones.
Table 7.6: Comparison of results - rectangular tank: test C - along longer direction.
Depth ofwater [m]
First natural antisymmetric sloshing frequencies [Hz]
Housner [4]Jaiswal et al. [6]- Experimental
Jaiswal et al. [6]- ANSYS R© Elmer
0.050 1.238 1.26 1.23 1.230.100 1.549 1.50 1.53 1.530.150 1.656 1.60 1.62 1.620.200 1.690 1.64 1.66 1.660.250 1.701 1.70 1.68 1.68
76
7. Test case 1: Rectangular Tank
Table 7.7: Comparison of results - rectangular tank: test C - along shorter direction.
Depth ofwater [m]
First natural antisymmetric sloshing frequencies [Hz]
Housner [4]Jaiswal et al. [6]- Experimental
Jaiswal et al. [6]- ANSYS R© Elmer
0.050 2.191 2.17 2.15 2.150.100 2.390 2.33 2.33 2.330.150 2.410 2.38 2.35 2.350.200 2.412 2.40 2.35 2.350.250 2.412 2.40 2.35 2.35
As it is possible to see in the above presented tables (7.6 and 7.7), the
results obtained with Elmer for test C of the rectangular tank test case are
in good agreement with similar test results available in literature.
77
Chapter 8
Test case 2: Cylindrical Tank
“A test is an experiment designed to reveal information, or
answer a specific question, about the software or system.”
- Elisabeth Hendrickson
In this chapter, the tests developed for the cylindrical laterally excited
tank test case are introduced. Having a simple geometry, the cylindrical
tank similarly to the rectangular one also allows accurate analytical solu-
tions for simple sloshing problems - chapter 2.1.1. Based on numerical and
experimental results available in literature [6], a test was defined.
This test is presented in the following sections. Similarly to chapter 7, the
test is primarily defined, then its implementation is explained and later the
results are presented, evaluated and validated.
8.1 Test Definition
Similarly to what was done in test C of the rectangular tank test case,
the objective of this test was to obtain the first antisymmetric sloshing
frequency of a cylindrical tank filled with water - similarly performed by [6].
The tank had diameter D = 0.170 m and height H = 0.230 m. It was
subjected to a vertical acceleration g = −9.81 m/s2 and laterally excited.
A two-phase flow with air and water fluids was considered. The material
properties of these fluids were:
78
8. Test case 2: Cylindrical Tank
• Water:
– µwater = 1.0 · 10−3 Pa · s
– ρwater = 1000.0 kg/m3
• Air:
– µair = 1.0 · 10−5 Pa · s
– ρair = 1.2 kg/m3
Slip boundary conditions were defined along the tank walls and a no-slip
boundary condition was assigned to the bottom. Initially, the velocity field
was assumed to be zero and the surface of the liquid was considered a free
surface initially in rest.
The first antisymmetric sloshing frequency was obtained for different water
depths:
• h1 = 0.050 m
• h2 = 0.100 m
• h3 = 0.150 m
• h4 = 0.200 m
8.2 Implementation
Similarly to test C performed for the rectangular tank, an experimental ap-
proach was used to find the first natural antisymmetric sloshing frequency
for different water depths inside the cylindrical tank. To find the optimal
external amplitude some initial tests were performed.
Knowing the expected sloshing frequency results from the analytical solu-
tions of Housner [4], the frequency ranges were chosen. For each test, the
following amplitude, external frequency range and number of elements per
mesh were used:
i) h = 0.050 m
• A0 = 0.6 m/s2;
• fext = 1.96 Hz to fext = 2.02 Hz;
• Mesh elements: 6144.
79
8. Test case 2: Cylindrical Tank
ii) h = 0.100 m
• A0 = 0.6 m/s2;
• fext = 2.16 Hz to fext = 2.30 Hz;
• Mesh elements: 12288.
iii) h = 0.150 m
• A0 = 0.6 m/s2;
• fext = 2.20 Hz to fext = 2.32 Hz;
• Mesh elements: 18432.
iv) h = 0.200 m
• A0 = 0.6 m/s2;
• fext = 2.20 Hz to fext = 2.32 Hz;
• Mesh elements: 24576
The simulations ran for 2 s using a time step equal to 0.02 s (100 steps).
To provide reference, it is important to note that the coordinate system
was defined to have origin at the bottom center of the tank.
8.3 Results & Evaluation
The results obtained for this test are presented and discussed in this sub-
section. Due to their similarity, the results are presented only for the
considered smaller water depth inside the tank (h = 0.050 m). Neverthe-
less, the obtained results for the other tested water depths can be found in
appendix C.
The procedure adopted in tests B a) and C of the rectangular tank test
case (chapter 7) was also adopted to evaluate the cylindrical tank tests.
Thus, the first step was to check the liquid’s pressure distribution inside
the containers at time t = 0s. In figure 8.1, the pressure distribution for
each one of the different water depths inside the tank is presented.
80
8. Test case 2: Cylindrical Tank
(a) h = 0.050m (b) h = 0.100m
(c) h = 0.150m (d) h = 0.200m
Figure 8.1: Cylindrical tank - test: pressure at t = 0s for different water depths.
As it is possible to visualize, the maximum pressure value clearly matches
the analytical expected ones.
The free surface position for each one of the different water depths inside
the tank still in rest (t = 0s) was studied - in figure 8.2 the results are
shown for the case in which the water depth was h = 0.050m.
Figure 8.2: Cylindrical tank test - h = 0.050m: free surface shape at t = 0s.
81
8. Test case 2: Cylindrical Tank
The same resulting plots presented for tests B a) and C of the rectangular
tank test case (chapter 7) are presented for this test case. Similarly, these
were used to find the first natural antisymmetric sloshing frequency for
which one of the tests.
Figure 8.3 presents the position of the CoG in the x-direction. The max-
imum wave sloshing amplitude is marked and zoomed to identify the cor-
respondent external frequency, which is then equal to the first natural
antisymmetric sloshing frequency.
(a) t = 0s to t = 2s (100 time steps) (b) Zoom: t = 1.48s
Figure 8.3: Cylindrical tank test - h = 0.050m: CoG x-coordinate Vs time.
The correspondent free surface position at the time in which the maximum
sloshing wave amplitude occurred is shown in figure 8.4.
Figure 8.4: Cylindrical tank test - h = 0.050m: maximum wave amplitude (t =1.48s).
A summary of the obtained first natural antisymmetric sloshing frequencies
for the different water depths that were tested is shown in table 8.1.
82
8. Test case 2: Cylindrical Tank
Table 8.1: Results - cylindrical tank test.
Depth of water [m]First natural antisymmetricsloshing frequency [Hz]
0.050 1.990.100 2.180.150 2.220.200 2.22
The above presented results (table 8.1) are compared with the analytical
solutions of Housner [4] and the numerical and experimental data available
in literature [6] for similar tests −table 8.2.
Table 8.2: Comparison of results - cylindrical tank test.
Depth ofwater [m]
First natural antisymmetric sloshing frequencies [Hz]
Housner [4]Jaiswal et al. [6]- Experimental
Jaiswal et al. [6]- ANSYS R© Elmer
0.050 2.064 2.07 1.99 1.990.100 2.287 2.30 2.18 2.180.150 2.314 2.33 2.22 2.220.200 2.317 2.33 2.22 2.22
From table 8.2 it is possible to attest a good agreement of the results
obtained in this test with the results available in the literature.
83
Chapter 9
Test case 3: ESA Tank -
Cylindrical Tank w/
Hemispherical Domes
“Knowing a great deal is not the same as being smart;
intelligence is not information alone but also judgement, the
manner in which information is collected and used.”
- Carl Sagan
In this chapter, the tests developed for a cylindrical tank with hemispherical
domes (from now on called ESA tank) are introduced. This test case was
proposed by ESA’s HSO directorate and its main goal was to reduce the gap
between the previously presented test cases and the real sloshing problems
faced in space vehicles.
In test cases 1 and 2, the software was already validated for simple sloshing
problems where comparators were available. In this test case, a more com-
plex problem in which the liquid propellant tank has a geometry commonly
used in space vehicles is presented.
This type of tank do not have directly accessible accurate analytical so-
lutions for the liquid sloshing parameters. Nevertheless, rough results for
some of the parameters can be obtained using approximated tank geome-
tries.
Because no numerical or experimental data was available for comparison
and validation of the test results, the obtained results had to be critically
84
9. Test case 3: ESA Tank
judged and evaluated before being validated.
To accomplish the goals defined for this test case three different tests were
defined, their objectives were:
• Test A: Recognize the natural sloshing modes frequencies of com-
monly used liquid propellants sloshing inside the tank;
• Test B: Accurately determine the first antisymmetric sloshing fre-
quency of commonly used liquid propellants sloshing inside the tank;
• Test C: Determine the liquid’s damping ratio, as well as the mass
participating in the sloshing movement after the abrupt removal of a
0.1− g lateral acceleration.
In the following sections, the tests that were developed are presented. Each
test is primarily defined, then its implementation is explained and later the
results are presented and evaluated.
9.1 Test A
9.1.1 Test Definition
The objective of this test was to perform a sloshing analysis that would
allow the recognition of the natural sloshing modes frequencies. The test
was performed for different liquid propellants sloshing inside the tank.
The tank was subjected to a vertical acceleration g = −9.81 m/s2 and
laterally excited. A two-phase flow in a three-dimensional sloshing tank
was considered.
Geometrically, the tank was defined to have a main cylindrical part and an
hemispherical dome at each end. The radius of the tank was r = 0.569 m
and the height of the cylindrical part H = 1.206 m.
The fluid at the free surface of the liquid propellant was defined to be
pressurized Helium (P = 20 bar , T = 298 K) with properties:
• µ = 1.9786 · 10−5 Pa · s
• ρ = 3.2312 kg/m3
85
9. Test case 3: ESA Tank
Two different propellant liquids were studied, their properties were:
• MON-3:
– µ = 3.967441 · 10−4 Pa · s
– ρ = 1433.401 kg/m3
• MMH:
– µ = 7.78024 · 10−4 Pa · s
– ρ = 870.372 kg/m3
Different propellant fill ratios were evaluated:
• 25 %
• 50 %
• 75 %
The boundary conditions along the tank walls were defined to be slip
boundary conditions. For the bottom hemisphere, a no-slip boundary con-
dition was prescribed. Initially, the velocity field was assumed to be zero
and the top surface of the liquid propellant was assumed a free surface,
initially in rest.
9.1.2 Implementation
To implement this test, the numerical approach introduced in test B b) of
the rectangular tank test case was used - see chapter 7.2.2 for more details.
An approximated analytical approach was used to have a rough idea of
the expected natural sloshing frequencies. To start with, the first anti-
symmetric natural sloshing frequency of a cylindrical tank (Housner [4] -
chapter 2.1.1) with the same radius r = 0.569 m and variable liquid height
h - dependent on the fill ratio, was used:
• 25 % fill ratio - h = 0.681 m: f = 0.885 Hz
• 50 % fill ratio - h = 1.172 m: f = 0.895 Hz
86
9. Test case 3: ESA Tank
• 75 % fill ratio - h = 1.663 m: f = 0.896 Hz
To have an estimate of the different antisymmetric and symmetric sloshing
frequencies, the first antisymmetric natural frequency found for the cylin-
drical tank was matched with that of a rectangular tank with the same
height of liquid - Abramson [5] (chapter 2.1.1). The corresponding dimen-
sion of the rectangular tank along the excitation direction was found to
be a = 0.972 m. With this dimension as a reference, the different slosh-
ing modes frequencies were then calculated and are believed to be good
estimations for the cylindrical tank with hemispherical domes:
• 25 % fill ratio - h = 0.681 m:
– n=1 - f = 0.885 Hz
– n=2 - f = 1.552 Hz
– m=1 - f = 1.267 Hz
– m=2 - f = 1.792 Hz
• 50 % fill ratio - h = 1.172 m:
– n=1 - f = 0.896 Hz
– n=2 - f = 1.552 Hz
– m=1 - f = 1.267 Hz
– m=2 - f = 1.792 Hz
• 75 % fill ratio - h = 1.663 m:
– n=1 - f = 0.896 Hz
– n=2 - f = 1.552 Hz
– m=1 - f = 1.267 Hz
– m=2 - f = 1.792 Hz
From these rough estimations it was possible to conclude that for the de-
fined fill ratios the sloshing natural frequencies would be approximately the
same.
87
9. Test case 3: ESA Tank
Having as a basis test B b) of the rectangular tank test case, a simulation
with 20 s and 1000 time steps was defined. The external excitation was
defined to start only after 1 s.
After some initial tests, the optimal external acceleration amplitude was
found to be A0 = 0.35 m/s2 for a 25 % fill ratio and A0 = 0.7 m/s2 for
the other fill ratios. Two different external frequencies were tested for each
propellant and fill ratio: fext = 0.70 Hz and fext = 1.50 Hz.
The geometries of the liquids inside the tanks were meshed using unstruc-
tured tetrahedral elements. Even though hexahedral mesh elements are
normally preferred (chapter 2.2.3), for this test tetrahedral elements were
chosen - the meshing of the tank was much easily achieved using these ele-
ments. Moreover, given the relative simplicity of the geometries, the final
results are believed to have a similar accuracy as if hexahedral elements
were used.
After validating the mesh for the 75 % fill ratio tank (the number of ele-
ments was doubled and the same test was performed, having been obtained
similar results), the number of tetrahedral elements per mesh, for each fill
ratio, was settled as:
• 25 % fill ratio - 11816 elements
• 50 % fill ratio - 14528 elements
• 75 % fill ratio - 12600 elements
For reference, the coordinate system was defined to have its origin at the
center of the tank in the plane intersecting the bottom hemispherical dome
and the cylindrical part of the tank.
9.1.3 Results & Evaluation
The results obtained for test A are presented, discussed and evaluated in
this subsection.
As it was done for the other test cases, the first step to validate the correct
implementation of the defined tests is to check the pressure distribution
and the free surface shape at t = 0s.
Figure 9.1 shows the pressure distribution for the MON-3 liquid propellant
tests. Figure 9.2 presents the same but for the MMH propellant case.
88
9. Test case 3: ESA Tank
(a) 25% (b) 50%
(c) 75%
Figure 9.1: ESA tank - test A - MON-3: pressure at t = 0s for the different fillratios.
(a) 25% (b) 50%
(c) 75%
Figure 9.2: ESA tank - test A - MMH: pressure at t = 0s for the different fill ratios.
89
9. Test case 3: ESA Tank
As it is possible to see in figures 9.1 and 9.2, the maximum pressure value
clearly matches the analytical expected values (P = ρ · g · h) for each
propellant and liquid height.
Validated the correct implementation of the test at time t = 0s, the next
step was to evaluate, for each liquid propellant and fill ratio, the results that
would allow the identification of the liquids’ natural sloshing frequencies.
To do this, the CoG variation, the sloshing amplitude measured at the top
most left point of the tank and the PSD plots for both the antisymmetric
and the symmetric modes were evaluated and are presented below.
The obtained results are very similar for the different propellants and fill
ratios. Therefore, in this section, only the most important results of the
MON-3 propellant with a 50 % fill ratio test are shown. The results ob-
tained for all the other developed tests can be found in appendix D.
The variation of the CoG of the liquid is shown in figure 9.3 for the case
in which the external frequency was fext = 0.70 Hz.
Figure 9.3: ESA tank test A - MON-3 (50% fill ratio & fext = 0.70 Hz): CoGplots.
90
9. Test case 3: ESA Tank
A general evidence common to all the developed tests is the fact that in
the vertical direction (z-direction) there is significant decrease of the CoG
position. This is due to the already described mass loss problem, which
in this case is evident. However, because the mass of liquid participating
in the sloshing movement is not immediately affected, it was assumed that
this effect had no major impact in the obtained results. In the y-direction,
the noticeable variations are very small, as expected. In the x-direction,
the CoG position has the expected behaviour, moving together with the
liquid sloshing.
In figure 9.4 the sloshing wave amplitude measured at the top left most
point of the tank (along the excitation direction) is shown.
Figure 9.4: ESA tank test A - MON-3 (50% fill ratio & fext = 0.70 Hz): sloshingamplitude plot.
Figures 9.5 and 9.6 present the PSD plots for the antisymmetric and the
symmetric modes sloshing frequencies for the case in which fext = 0.70 Hz.
Figures 9.7 and 9.8 present the same plots but for fext = 1.50 Hz.
It is important to note that to recognize the antisymmetric modes sloshing
frequencies the PSD was measured in the top most left point of the tank,
and to find the symmetric modes it was measured in the top center point.
91
9. Test case 3: ESA Tank
Figure 9.5: ESA tank test A - MON-3 (50% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top left point of the tank).
Figure 9.6: ESA tank test A - MON-3 (50% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top center point of the tank).
92
9. Test case 3: ESA Tank
Figure 9.7: ESA tank test A - MON-3 (50% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top left point of the tank).
Figure 9.8: ESA tank test A - MON-3 (50% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top center point of the tank).
93
9. Test case 3: ESA Tank
A deep evaluation of the presented PSD plots allowed to draw some conclu-
sions about the first two antisymmetric and symmetric sloshing frequencies.
The following frequencies are highlighted from the plots:
• f = 0.70 Hz - Excitation frequency;
• f = 1.50 Hz - Excitation frequency;
• f = 0.90 Hz - 1st antisymmetric mode sloshing frequency;
• f = 1.40 Hz - 1st symmetric mode sloshing frequency;
• f = 1.60 Hz - 2nd antisymmetric mode sloshing frequency;
• f = 1.95 Hz - 2nd symmetric mode sloshing frequency.
The obtained frequencies have an uncertainty of ±0.05 Hz.
It is important to note that both the 1st antisymmetric and symmetric
sloshing frequencies were obtained with relative confidence, as their power
in the PSD is clearly visible. As for the 2nd antisymmetric and symmetric
sloshing frequencies these were obtained by looking for the points in the
PSD that were in the vicinities of our initial rough analytical estimations.
Therefore, the confidence in these results is slightly diminished.
Regarding the other frequencies visible in the plots, these are not relevant
for this study. They are believed to result from a superposition of waves
originated in the initial time steps of the simulation and then propagated
over time on the top of the general sloshing motion. These waves are
thought to be introduced artificially due to imperfections in the numerical
simulations. Their almost negligible PSD attests their insignificance.
For the different fill ratios, as well as for the other liquid propellant (MMH),
the sloshing frequencies were found to be approximately the same as those
presented and discussed here.
9.2 Test B
9.2.1 Test Definition
This test was defined to accurately determine the first antisymmetric natu-
ral sloshing frequency for the different liquid propellants sloshing inside the
94
9. Test case 3: ESA Tank
tank. Another objective of this test was to identify the maximum sloshing
wave amplitude for a lateral excitation of amplitude 0.1−g.
The tank was subjected to the same set of conditions defined for test A -
subsection 9.1.1.
9.2.2 Implementation
In this test, the same experimental approach used in the rectangular tank
test case - tests 2 a) and 3, as well as in the cylindrical tank test case, was
used.
Using as a basis the initial rough estimations of the first antisymmetric
sloshing frequency, the external excitation frequency was varied between
fext = 0.82 Hz and fext = 0.92 Hz and the amplitude of the external
acceleration was defined to be A0 = 0.981 m/s2.
The same meshes previously defined for test A were also used here. The
tests were defined to run for 4 s with a time step equal to 0.02 s (200 time
steps).
After experimentally finding the first antisymmetric sloshing frequency, the
maximum sloshing amplitude measured at the left most point of the tank
was also evaluated. It is important to note that the used approach is only
valid as long as there is no wave rotation.
Because the results were expected to be approximately the same for the
three different fill ratios, this test was only performed for the case in which
the propellant liquid occupies 50 % of the tank.
9.2.3 Results & Evaluation
The results achieved in test B are presented in this subsection for the MON-
-3 liquid propellant. The results obtained for the MMH liquid propellant
case can be found in appendix E.
The free surface position of the liquid inside the tank still in rest (t = 0s)
is presented together with a study of the CoG position in the x-direction.
From these, it is possible to identify the first antisymmetric sloshing fre-
quency, which corresponds to the maximum sloshing wave without rotation
- see chapter 7.2.3 for details.
95
9. Test case 3: ESA Tank
Figure 9.9 presents the free surface position of the liquid propellants, MON-
-3 or MMH at time t = 0s (as expected, in rest, the free surface of the
liquids was the same).
Figure 9.9: ESA tank test B - MON-3 or MMH: free surface shape at t = 0s.
In figure 9.10, the position of the CoG in the x-direction is shown for
the MON-3 liquid propellant. The maximum wave sloshing amplitude is
also marked and zoomed to allow the identification of the corresponding
external frequency, which is then equal to the first antisymmetric sloshing
natural frequency.
(a) t = 0s to t = 4s (200 time steps) (b) Zoom: t = 3.46s
Figure 9.10: ESA tank test B - MON-3: CoG x-coordinate Vs time.
The free surface position at the time in which the maximum sloshing wave
amplitude occurred is presented in figure 9.11 for the MON-3 liquid pro-
pellant.
96
9. Test case 3: ESA Tank
Figure 9.11: ESA tank test B - MON-3: maximum wave amplitude (t = 3.46s).
In table 9.1, a summary of the obtained first antisymmetric natural sloshing
frequencies is presented.
Table 9.1: Results - ESA tank: test B.
Liquid propellantFirst antisymmetricnatural sloshing frequency [Hz]
MON-3 0.86MMH 0.86
Figure 9.12 presents the sloshing wave amplitude measured at the left most
point of the tank.
Figure 9.12: ESA tank test B - MON-3: sloshing amplitude plot.
97
9. Test case 3: ESA Tank
The maximum wave amplitude for the case presented in figure 9.12 was
approximately 1.521 m, which means that from the initial rest position
of the free surface of the liquid at 0.603 m, the amplitude achieved by
the sloshing wave was 0.918 m. This measurement was performed at t =
3.5s - corresponding to the last wave peak before the rotation of the wave
happened.
For the case in which the MMH liquid propellant was used, the value
obtained was very similar, about 0.915 m. The slightest smaller result
was already expected seeing that the MMH liquid propellant has higher
viscosity than the MON-3.
It is important to note that for this specific test the amplitude of the wave
was only important to measure the sloshing frequency. Hence, the sloshing
wave amplitude values previously presented shall only serve to give an idea
of the expected sloshing wave amplitudes for the established conditions.
In the specific case of the tanks being used, at the referred time t = 3.5s,
the sloshing wave would have already hit the top hemisphere of the tank
and therefore wave rotation and turbulence would have already happened.
As explained in the implementation section, it was assumed that the re-
sults obtained for the tank containing 50% of propellant could be extended
to the tanks containing 25% and 75%. Nevertheless, if certainty in the
results needs to be ensured, the same procedure adopted here could also
be implemented for these two cases.
9.3 Test C
9.3.1 Test Definition
Test C had two main objectives, being the first one obtaining the liquid
propellants’ damping ratio and the second one identifying the mass of pro-
pellant participating in the sloshing phenomenon.
The same set of conditions to which tests A and B had been subjected
were also defined for this test - subsection 9.1.1.
98
9. Test case 3: ESA Tank
9.3.2 Implementation
To implement test C an external perturbation was laterally applied to the
fluid and abruptly removed after some time. The damping ratio of the
liquid was measured for the initial period at which the lateral acceleration
was still acting, as well as for the period from which the lateral perturbation
was removed.
Due to the assumptions considered in the models, the damping ratio is only
dependent on the viscosity of the liquid, the fill ratio and the tank shape.
Two different tests were implemented: one in which a constant 0.1 − g
lateral excitation was applied to the liquid and abruptly removed when
a quasi steady-state was reached, and another in which the same lateral
excitation was removed immediately after the maximum sloshing wave am-
plitude was reached.
For the first test, the simulation was run for 60 s, using a time step equal
to 0.02 s (3000 time steps). In the second test, the total time was reduced
to 20 s and the same time step equal to 0.02 s was used (1000 time steps).
The mass of liquid participating in the sloshing phenomenon was only
measured for the period after the removal of the lateral perturbation.
Following the same logic adopted in test B, this test was only performed
for a propellant fill ratio equal to 50 % .
9.3.3 Results & Evaluation
The results obtained in test C are presented in this subsection for the MON-
-3 liquid propellant. For the MMH liquid propellant case, the resulting
plots can be found in appendix F.
Similarly to what was done in chapter 7 for the fastest test performed
in this project, it is also important to note how long did it take to run
the longest test performed. Therefore, for the 60 s test, approximately 47
hours were required to complete the simulation using the same computer
described in chapter 7.1.3.
The free surface of the liquid propellants still in rest was already presented
in test B - figure 9.9, and shall serve as a reference also in this case.
99
9. Test case 3: ESA Tank
To start with, figure 9.13 presents the CoG x-coordinate development for
the case in which the simulation was run for 60s.
Figure 9.13: ESA tank test C - MON-3 (60s simulation): CoG x-coord. Vs time.
As it is possible to visualize in the figure, it takes about 30 seconds for the
liquid propellant to stabilize after the initial impact of the 0.1 − g lateral
perturbation. Figure 9.14 shows the free surface shape at this point in
time.
Figure 9.14: ESA tank test C - MON-3 (60s simulation): wave amplitude att = 30s.
100
9. Test case 3: ESA Tank
After the abrupt removal of the lateral acceleration, it is visible in fi-
gure 9.13 that it takes about 15s (750 time steps) for the liquid surface
to stabilize around zero. As expected, the damping has an exponential
behaviour going, in the limit, to zero - please refer to [1] for more details.
Figure 9.15 presents the quasi-stabilized free surface at t = 45.5s.
Figure 9.15: ESA tank test C - MON-3 (60s simulation): wave amplitude att = 45.5s.
The variation of the x-coordinate of the CoG at this point (t = 45.5s) has
an amplitude of approximately 1.5 mm and the sloshing wave amplitude
is almost unnoticeable. Figure 9.16 presents the sloshing wave amplitude
measured at the top most right point of the tank.
Figure 9.16: ESA tank test C - MON-3 (60s simulation): sloshing amplitude plot.
101
9. Test case 3: ESA Tank
As it is possible to see in the figure the behaviour is not that which was
expected. Nevertheless, the significant decrease in the amplitude of the
sloshing wave is easily explained by the mass loss phenomenon (already
addressed in chapter 7.1.3), which in this case is clearly visible.
Figure 9.17 presents the variation of the z-coordinate of the CoG. In this
figure it is possible to visualize that the liquid’s mass loss in this test is
significant, as mentioned before.
Figure 9.17: ESA tank test C - MON-3 (60s simulation): CoG z-coord. Vs time.
Besides the already mentioned mass loss phenomenon it was also noticed
that for some tests some other numerical issues would also arise. For ex-
ample, in some tests it was noticed that noise could appear in the vicinities
of zero. This noise is believed to be due to the time integration scheme
that was being used to run the tests using Elmer.
Having addressed all these different numerical constrains and carefully val-
idated the obtained results, the damping ratio was calculated. For the
initial 30s period in which the lateral acceleration was still acting, this was
found to be ζ = 0.023 for the MON-3 propellant and ζ = 0.025 for the
MMH one.
For the period after the abrupt removal of the lateral acceleration these
values were found to be higher, being about ζ = 0.0225 for the two liquid
propellants.
From [1] analytical results were obtained for the damping ratio of these
two liquid propellants. It was found that for the MON-3 propellant -
ζ = 0.00036 and for the MMH - ζ = 0.00064. Meaning that the values that
102
9. Test case 3: ESA Tank
were numerically obtained are most probably overestimated.
This overestimation of the results can be explained by the mass loss phe-
nomenon. The damping ratio is calculated from the maximum wave peaks
of the sloshing wave amplitude, from which an exponential fit is performed
- see [1] for details. As seen in figure 9.16 the sloshing wave amplitude
is being decreased over time because of the liquid’s mass loss. Thus, the
fit being performed to these maximum sloshing wave peaks gives in fact a
damping ratio slightly higher than that which would have been obtained
if mass loss would not be present. In the second part of test the mass loss
phenomenon is even more dominant and therefore for evaluation purposes
the results obtained in the first 30s of the test are considered.
The PSD plots obtained for the 15s period after the abrupt removal of the
lateral acceleration attest the conclusions reached in test B, regarding the
first antisymmetric sloshing frequencies of the two liquids, and therefore
they are also included in appendix F.
For the second tested setup - simulation time equals to t = 20s and lateral
acceleration released after the first sloshing wave peak - the results were
found to be very similar to those obtained in the previously presented test.
Figure 9.18 shows the shape of the surface of the liquid at its maximum
wave amplitude - this corresponds to the point at which the lateral acce-
leration was abruptly removed.
Figure 9.18: ESA tank test C - MON-3 (20s simulation): wave amplitude att = 0.54s.
103
9. Test case 3: ESA Tank
The same results are presented in figure 9.19 for the time t = 19.6s. At
this point in time, the CoG x-coordinate was already in the vicinities of
zero.
Figure 9.19: ESA tank test C - MON-3 (20s simulation): wave amplitude att = 19.60s.
Figures 9.20 and 9.21 present the variation of the CoG x-coordinate and
z-coordinate, respectively.
Figure 9.20: ESA tank test C - MON-3 (20s simulation): CoG x-coord. Vs time.
The damping ratio found in this test was approximately ζ = 0.027 for both
the propellants.
104
9. Test case 3: ESA Tank
Figure 9.21: ESA tank test C - MON-3 (20s simulation): CoG z-coord. Vs time.
However, the results from the previous test are believed to be more ac-
curate. This is so because, as mentioned before, some waves might be
artificially introduced in the sloshing dynamics at the begin of the simu-
lation. Therefore, in a shorter simulation, these are more dominant and
might influence the damping results, specially when no time is given for
stabilization of the lateral acceleration.
The mass participating in the liquid sloshing was roughly estimated using
ParaView and looking at the mesh elements that were moving. The pro-
vided result shall be taken as a very rough estimate and is only intended
to give an approximated idea of the percentage of mass participating in
the sloshing movement. Thus, it was found that about 30 % to 40 % of
the liquid mass inside the tank participated in the sloshing movement af-
ter the abrupt removal of the lateral acceleration. From [1] an analytical
estimation was calculated and found to be about 26 % - this value is in
good agreement with the obtained results.
105
Chapter 10
Conclusions
“The logic of validation allows us to move between the two limits
of dogmatism and scepticism.”
- Paul Ricoeur
The effects of propellant liquids sloshing inside spacecraft tanks have been
defined long ago as being of critical influence to the dynamics of space
vehicles. The interaction of the disturbance forces and torques generated
by the moving fuel with the solid body and the control system can lead
to an increase in the AOCS actuators commands, which can degrade the
vehicle’s pointing performances and, in critical cases, generate unstable
attitude and orbit control.
During years, simplifying analytical models were the only possible way
to predict the dynamics of liquids sloshing inside rigid containers. How-
ever, the necessity of ensuring the correct functioning of any vehicle sent
to space, together with the development of science and technology led to
the development of new tools that can be used to predict the vehicles be-
haviour. Thus, for the cases in which analytical approximate solutions are
not reasonable, two possibilities exist to accurately predict the behaviour
of liquids sloshing inside rigid tanks: an experimental approach - which
brings increased expense and complexity to any project; and a numerical
approach using CFD techniques - which allows early testing and facilitates
the achievement of accurate solutions for complex problems.
This project was focused on the implementation of a CFD code in an
existing EADS Astrium simulation environment. This code was used to
assess the influence of liquid propellant sloshing for specific missions.
106
10. Conclusions
Starting with a defined set of functional and system requirements, an ex-
tended selection process was performed to choose the most suitable CFD
software that would suit the objectives of this project. From a list of
more than 35 available open source CFD software based on FEM or FVM,
Elmer - Open Source Finite Element Software for Multiphysical Problems
was selected.
The CFD solver was integrated with the available pre- and post- processing
environments and the resulting simulation environment allowed a deep and
complete testing of liquid propellants sloshing inside rigid containers.
The geometry of the tanks can be meshed using available meshing software,
such as Gmsh and CubitTM
. For simple geometries, simpler meshing tools,
such as Elmer’s included tool - ElmerGrid, can also be used.
The sloshing problems can be fully defined using Elmer’s simple solver
input file (.Sif ), which is then input to ElmerSolver - the most important
part of Elmer’s package.
The results obtained from the solver can be exported and visualized with
ParaView, which allows a direct visual analysis of the sloshing behaviour.
If a deeper analysis is required, a specific sloshing post-processing based
on MATLAB R© can be used to obtain some of the most important slosh-
ing parameters, such as the liquid sloshing modes natural frequencies, the
sloshing wave amplitudes, the liquid’s damping and the mass of liquid par-
ticipating in the sloshing movement.
To validate the CFD software for sloshing purposes, specific test cases
were defined. The first and the second test cases dealt, respectively, with
rectangular and cylindrical laterally excited tanks, and the final test case -
defined by ESA’s HSO directorate - dealt with a laterally excited cylindrical
tank with hemispherical domes.
For test case 1, three different tests were defined, implemented and va-
lidated against available numerical and experimental data. For all these
tests the obtained results were proven to be in good agreement with the
available comparators.
In test case 2, the first natural antisymmetric sloshing frequency of a li-
quid sloshing inside a laterally excited cylindrical tank was obtained. The
achieved results were also compared and validated against data available
in the literature for similar tests.
107
10. Conclusions
In test case 3, a typical liquid propellant tank was defined and real mis-
sion conditions were simulated. The cylindrical tank with hemispherical
domes was subjected to a 1− g vertical acceleration and laterally excited.
Two different liquid propellants were evaluated - MON-3 and MMH - for
three different liquid fill ratios (25 %, 50 % and 75 %). The liquid’s natu-
ral sloshing frequencies were obtained for the different propellants and fill
ratios. The damping ratio and the mass of fluid participating in the slosh-
ing movement were also estimated. The obtained results were evaluated
and validated against simplified analytical results - a good agreement was
found.
As a final conclusion, the validation of the above mentioned sloshing test
cases together with the fulfilment of the defined requirements allowed the
validation of the final simulation environment for the sloshing problems
that were addressed - chapter 3.1. Hence, it was possible to conclude that
the objectives stipulated for this project were successfully accomplished.
108
Chapter 11
Future Work
“Learn from yesterday, live for today, hope for tomorrow. The
important thing is to not stop questioning.”
- Albert Einstein
The research presented in this thesis addressed and gave answers to several
open questions. Nevertheless, it also raised and left behind many other
important issues that still lack answers. Thus, several lines of investigation
arise from this work and shall be pursued in the future.
Firstly, the several issues related with the numerical problems experienced
during the development of this work shall be addressed. Issues such as mass
loss, numerical noise and unrealistic behaviours appearing at the first time
steps of the simulation (believed to be due to the used time integration
schemes) need to be carefully addressed to increase the robustness of the
software and the validity of the results of sloshing problems.
Rising directly from the future functional requirements defined in chapter 3,
several tests need to be performed to evaluate the limits of the solver for
problems related with liquid sloshing inside spacecraft tanks. More com-
plex tank geometries with internal elements and different configurations
shall be tested; different types of sloshing, such as rotary sloshing, shall
be addressed; sloshing due to high external excitations needs to be tested,
evaluated and validated; tests in micro-gravity environments, where sur-
face tension effects are dominant, need to be performed; and finally, fluid
structure interaction approaches shall be considered to simulate different
types of sloshing problems (e.g. non-isothermal fluids), as well as allow the
simulation of liquids sloshing inside non-rigid tanks.
109
References
[1] F. T. Dodge et al., The new “dynamic behavior of liquids in moving containers”.
Southwest Research Inst., 2000.
[2] R. Ibrahim, Liquid Sloshing Dynamics: Theory and Applications. Cambridge
University Press, 2005.
[3] E. W. Graham and A. M. Rodriguez, “The characteristics of fuel motion which
affect airplane dynamics,” Applied Mechanics, vol. 19, pp. 381–388, 1952.
[4] G. W. Housner, “Dynamic pressures on accelerated fluid containers,” Division
of Engineering, California Institute of Technology, Pasadena, California, USA,
1955.
[5] H. Abramson, The dynamic behavior of liquids in moving containers: with
applications to space vehicle technology, ser. NASA SP. Southwest Research
Institute, Scientific and Technical Information Division, National Aeronautics
and Space Administration, 1966.
[6] O. Jaiswal, S. Kulkarni, and P. Pathak, “A study on sloshing frequencies of
fluid-tank system,” in Proceedings of the 14th World Conference on Earthquake
Engineering, 2008, pp. 12–17.
[7] J. Wendt, J. John D. Anderson, and V. K. I. for Fluid Dynamics, Computational
Fluid Dynamics: An Introduction, ser. A von Karman Institute book. Springer,
2009.
[8] J. Tu, G. Yeoh, and C. Liu, Computational Fluid Dynamics: A Practical Ap-
proach. Elsevier Science, 2007.
[9] J. Blazek, Computational Fluid Dynamics: Principles and Applications. Else-
vier Science, 2005.
[10] A. Sayma, Computational Fluid Dynamics. Bookboon, 2009.
110
References
[11] F. White, Fluid Mechanics, ser. McGraw-Hill Series in Mechanical Engineering.
McGraw-Hill Higher Education, 2008.
[12] D. Logan, A First Course in the Finite Element Method. Cengage Learning,
2011.
[13] V. Voller, Basic Control Volume Finite Element Methods for Fluids and Solids,
ser. IISc research monographs series. World Scientific Publishing Company,
Incorporated, 2009.
[14] R. Burden and J. Faires, Numerical Analysis. Thomson Brooks/Cole Cengage
Learning, 2011.
[15] ADFC webpage (2013/03/12). Available: http://adfc.sourceforge.net/.
[16] Arb finite volume solver webpage (2013/03/12). Available: http://people.eng.
unimelb.edu.au/daltonh/downloads/arb/.
[17] CFD2D webpage (2013/03/12). Available: http://sourceforge.net/projects/
cfd2d/.
[18] CFD2k webpage (2013/03/12). Available: http://www.cfd2k.eu/.
[19] Computational Fluid Dynamics Packages webpage (2013/03/12). Available:
http://cfdpack.net/.
[20] Channelflow webpage (2013/03/12). Available: http://channelflow.org/.
[21] Clawpack webpage (2013/03/12). Available: http://depts.washington.edu/
clawpack/.
[22] “Code saturne: a finite volume code for the computation of turbulent
incompressible flows - industrial applications,” International Journal on Finite
Volumes, vol. 1, 2004. [Online]. Available: http://code-saturne.org/
[23] COOLFluiD webpage (2013/03/12). Available: http://coolfluidsrv.vki.ac.be/
trac/coolfluid.
[24] DUNS webpage (2013/03/12). Available: http://duns.sourceforge.net/.
[25] Dolfyn webpage (2013/03/12). Available: http://www.dolfyn.net/.
[26] Edge webpage (2013/03/12). Available: http://www.foi.se/en/
Customer--Partners/Projects/Edge1/Edge/.
[27] Elmer webpage (2013/03/12). Available: http://www.csc.fi/english/pages/
elmer.
111
References
[28] Featflow webpage (2013/03/12). Available: http://www.featflow.de/.
[29] A. Logg, K. A. Mardal, and G. N. Wells, Eds., Automated Solution of
Differential Equations by the Finite Element Method, ser. Lecture Notes in
Computational Science and Engineering. Springer, 2012, vol. 84. [Online].
Available: http://fenicsproject.org/
[30] FreeFEM webpage (2013/03/12). Available: http://www.freefem.org/.
[31] Fluidity webpage (2013/03/12). Available: http://amcg.ese.ic.ac.uk/index.
php?title=Fluidity.
[32] Hiflow3 webpage (2013/03/12). Available: http://www.hiflow3.org/.
[33] Gerris Flow Solver webpage (2013/03/12). Available: http://gfs.sourceforge.
net/wiki/index.php/Main Page.
[34] Hit3d webpage (2013/03/12). Available: https://code.google.com/p/hit3d/.
[35] INavier webpage (2013/03/12). Available: http://inavier.sourceforge.net/.
[36] ISAAC webpage (2013/03/12). Available: http://isaac-cfd.sourceforge.net/.
[37] Kicksey-Winsey webpage (2013/03/12). Available: http://justpmf.com/
romain/kicksey winsey/.
[38] MFIX webpage (2013/03/12). Available: https://mfix.netl.doe.gov/.
[39] NaSt2D-2.0 webpage (2013/03/12). Available: http://home.arcor.de/drklaus.
bauerfeind/nast/eNaSt2DA.html.
[40] Nek5000 webpage (2013/03/12). Available: http://nek5000.mcs.anl.gov/index.
php/Main Page.
[41] NuWTun webpage (2013/03/12). Available: http://nuwtun.cfdlab.net/.
[42] OpenFlower webpage (2013/03/12). Available: http://openflower.sourceforge.
net/.
[43] OpenFOAM webpage (2013/03/12). Available: http://www.openfoam.org/.
[44] OpenLB webpage (2013/03/12). Available: http://optilb.org/openlb/.
[45] OpenFVM webpage (2013/03/12). Available: http://openfvm.sourceforge.
net/.
[46] Partenov CFD webpage (2013/03/12). Available: http://www.partenovcfd.
com/.
112
References
[47] PETSc-FEM webpage (2013/03/12). Available: http://www.cimec.org.ar/
twiki/bin/view/Cimec/PETScFEM.
[48] SLFCFD webpage (2013/03/12). Available: http://slfcfd.sourceforge.net/.
[49] SSIIM webpage (2013/03/12). Available: http://folk.ntnu.no/nilsol/cfd/.
[50] F. Palacios et al., “Stanford university unstructured (su2): An open-source in-
tegrated computational environment for multi-physics simulation and design,”
in AIAA Paper 2013-0287, 1st AIAA Aerospace Sciences Meeting and Exhibit,
Grapevine, Texas, USA, January 7th - 10th, 2013.
[51] TOCHNOG webpage (2013/03/12). Available: http://tochnog.sourceforge.
net/.
[52] TYCHO webpage (2013/03/12). Available: http://tycho-cfd.at/.
[53] TYPHON webpage (2013/03/12). Available: http://typhon.sourceforge.net/
spip/.
[54] L. Metzger et al., “User manual and description of algorithms for dfh-3 sloshing
model software (program sloshc),” EADS Astrium SLOSCHC, 1990.
[55] FLOW-3D webpage (2013/03/12). Available: http://www.flow3d.com/.
[56] P. Raback and M. Malinen, Overview of Elmer. CSC - IT Center for Science,
February 1, 2013.
[57] CSC - IT Center for Science webpage (2013/05/14). Available: http://www.
csc.fi/english.
[58] P. Raback et al., Elmer Models Manual. CSC - IT Center for Science, February
1, 2013.
[59] J. Ruokolainen et al., ElmerSolver Manual. CSC - IT Center for Science,
February 1, 2013.
[60] E. Anderson et al., LAPACK Users’ Guide, 3rd ed. Philadelphia, PA: Society
for Industrial and Applied Mathematics, 1999.
[61] UMFPACK webpage (2013/05/14). Available: http://www.cise.ufl.edu/
research/sparse/umfpack/.
[62] B. Szabo and I. Babuska, Finite Element Analysis. John Wiley & Sons Ltd,
1991.
113
References
[63] P. Solin et al., Higher-Order Finite Element Methods. Chapman & Hall /
CRC, 2004.
[64] M. Lyly, ElmerGUI Manual v. 0.4. CSC - IT Center for Science, February 1,
2013.
[65] ElmerGrid Manual. CSC - IT Center for Science, February 1, 2013.
[66] C. Geuzaine and J.-F. Remacle, “Gmsh: A 3-d finite element mesh genera-
tor with built-in pre-and post-processing facilities,” International Journal for
Numerical Methods in Engineering, vol. 79, no. 11, pp. 1309–1331, 2009.
[67] ParaView webpage (2013/03/12). Available: http://www.paraview.org/.
[68] U. Rasthofer et al., “An extended residual-based variational multiscale method
for two-phase flow including surface tension,” Computer Methods in Applied
Mechanics and Engineering, vol. 200, no. 2122, pp. 1866 – 1876, 2011.
[69] T. Fries, “The intrinsic xfem for two-phase flows,” Int. J. Numer. Methods
Fluids, vol. 60, pp. 437 – 471, 2009.
[70] N. Pal, S. Bhattacharyya, and P. Sinha, “Non-linear coupled slosh dynamics of
liquid-filled laminated composite containers: a two dimensional finite element
approach,” Journal of Sound Vibration, vol. 261, pp. 729–749, Apr. 2003.
[71] P. Pal and S. Bhattacharyya, “Sloshing in partially filled liquid containers - nu-
merical and experimental study for 2-d problems,” Journal of Sound Vibration,
vol. 329, pp. 4466–4485, Oct. 2010.
[72] S. Bunya, S. Yoshimura, and J. J. Westerink, “Improvements in mass con-
servation using alternative boundary implementations for a quasi-bubble finite
element shallow water model,” International Journal for Numerical Methods in
Fluids, vol. 51, pp. 1277–1296, Aug. 2006.
114
Appendix A
Test case 1 - Results: Test A
115
b) 2D Skewed Tank
Figure A.1: Rectangular tank - test A b): pressure at t = 0s.
(a) t = 0.6s (b) t = 1.2s
(c) t = 1.8s
(d) t = 2.4s (e) t = 3.0s
Figure A.2: Rectangular sloshing tank - test A b): free surface shape evolution.
116
Figure A.3: Rectangular tank - test A b): CoG plots.
Figure A.4: Rectangular tank - test A b): sloshing amplitude plot.
117
c) Simple 3D Tank
Figure A.5: Rectangular tank - test A c): pressure at t = 0s.
(a) t = 0.6s (b) t = 1.2s
(c) t = 1.8s
(d) t = 2.4s (e) t = 3.0s
Figure A.6: Rectangular sloshing tank - test A c): free surface shape evolution.
118
Figure A.7: Rectangular tank - test A c): CoG plots.
Figure A.8: Rectangular tank - test A c): sloshing amplitude plot.
119
d) Simple 3D Skewed Tank
Figure A.9: Rectangular tank - test A d): pressure at t = 0s.
(a) t = 0.6s (b) t = 1.2s
(c) t = 1.8s
(d) t = 2.4s (e) t = 3.0s
Figure A.10: Rectangular sloshing tank - test A d): free surface shape evolution.
120
Figure A.11: Rectangular tank - test A d): CoG plots.
Figure A.12: Rectangular tank - test A d): sloshing amplitude plot.
121
Appendix B
Test case 1 - Results: Test C
122
h = 0.100 m
Figure B.1: Rectangular tank - test C - h = 0.100m longer direction: free surfaceshape at t = 0s.
(a) t = 0s to t = 5s (250 time steps) (b) Zoom: t = 3.56s
Figure B.2: Rectangular tank - test C - h = 0.100m longer direction: CoG x-coord.Vs time for different excitation frequencies.
Figure B.3: Rectangular tank - test C - h = 0.100m longer direction: maximumwave amplitude (t = 3.56s).
123
h = 0.150 m
Figure B.4: Rectangular tank - test C - h = 0.150m longer direction: free surfaceshape at t = 0s.
(a) t = 0s to t = 5s (250 time steps) (b) Zoom: t = 3.06s
Figure B.5: Rectangular tank - test C - h = 0.150m longer direction: CoG x-coord.Vs time for different excitation frequencies.
Figure B.6: Rectangular tank - test C - h = 0.150m longer direction: maximumwave amplitude (t = 3.06s).
124
h = 0.200 m
Figure B.7: Rectangular tank - test C - h = 0.200m longer direction: free surfaceshape at t = 0s.
(a) t = 0s to t = 5s (250 time steps) (b) Zoom: t = 3.30s
Figure B.8: Rectangular tank - test C - h = 0.200m longer direction: CoG x-coord.Vs time for different excitation frequencies.
Figure B.9: Rectangular tank - test C - h = 0.200m longer direction: maximumwave amplitude (t = 3.30s).
125
h = 0.250 m
Figure B.10: Rectangular tank - test C - h = 0.250m longer direction: free surfaceshape at t = 0s.
(a) t = 0s to t = 5s (250 time steps) (b) Zoom: t = 2.66s
Figure B.11: Rectangular tank - test C - h = 0.250m longer direction: CoG x-coord.Vs time for different excitation frequencies.
Figure B.12: Rectangular tank - test C - h = 0.250m longer direction: maximumwave amplitude (t = 2.66s).
126
h = 0.100 m
Figure B.13: Rectangular tank - test C - h = 0.100m shorter direction: free surfaceshape at t = 0s.
(a) t = 0s to t = 2.5s (125 time steps) (b) Zoom: t = 1.48s
Figure B.14: Rectangular tank - test C - h = 0.100m shorter direction: CoGx-coord. Vs time for different excitation frequencies.
Figure B.15: Rectangular tank - test C - h = 0.100m shorter direction: maximumwave amplitude (t = 1.48s).
127
h = 0.150 m
Figure B.16: Rectangular tank - test C - h = 0.150m shorter direction: free surfaceshape at t = 0s.
(a) t = 0s to t = 2.5s 125 time steps) (b) Zoom: t = 1.26s
Figure B.17: Rectangular tank - test C - h = 0.150m shorter direction: CoGx-coord. Vs time for different excitation frequencies.
Figure B.18: Rectangular tank - test C - h = 0.150m shorter direction: maximumwave amplitude (t = 1.26s).
128
h = 0.200 m
Figure B.19: Rectangular tank - test C - h = 0.200m shorter direction: free surfaceshape at t = 0s.
(a) t = 0s to t = 2.5s (125 time steps) (b) Zoom: t = 1.26s
Figure B.20: Rectangular tank - test C - h = 0.200m shorter direction: CoGx-coord. Vs time for different excitation frequencies.
Figure B.21: Rectangular tank - test C - h = 0.200m shorter direction: maximumwave amplitude (t = 1.26s).
129
h = 0.250 m
Figure B.22: Rectangular tank - test C - h = 0.250m shorter direction: free surfaceshape at t = 0s.
(a) t = 0s to t = 2.5s (125 time steps) (b) Zoom: t = 1.26s
Figure B.23: Rectangular tank - test C - h = 0.250m shorter direction: CoGx-coord. Vs time for different excitation frequencies.
Figure B.24: Rectangular tank - test C - h = 0.250m shorter direction: maximumwave amplitude (t = 1.26s).
130
Appendix C
Test case 2 - Results
131
h = 0.100 m
Figure C.1: Cylindrical tank test - h = 0.100m: free surface shape at t = 0s.
(a) t = 0s to t = 2s (100 time steps) (b) Zoom: t = 1.58s
Figure C.2: Cylindrical tank test - h = 0.100m: CoG x-coordinate Vs time.
Figure C.3: Cylindrical tank test - h = 0.100m: maximum wave amplitude (t =1.58s).
132
h = 0.150 m
Figure C.4: Cylindrical tank test - h = 0.150m: free surface shape at t = 0s.
(a) t = 0s to t = 2s (100 time steps) (b) Zoom: t = 1.78s
Figure C.5: Cylindrical tank test - h = 0.150m: CoG x-coordinate Vs time.
Figure C.6: Cylindrical tank test - h = 0.150m: maximum wave amplitude (t =1.78s).
133
h = 0.200 m
Figure C.7: Cylindrical tank test - h = 0.200m: free surface shape at t = 0s.
(a) t = 0s to t = 2s (100 time steps) (b) Zoom: t = 1.78s
Figure C.8: Cylindrical tank test - h = 0.200m: CoG x-coordinate Vs time.
Figure C.9: Cylindrical tank test - h = 0.200m: maximum wave amplitude (t =1.78s).
134
Appendix D
Test case 3 - Results: Test A
135
MON-3 - 25 % fill ratio
Figure D.1: ESA tank test A - MON-3 (25% fill ratio & fext = 0.70 Hz): CoGplots.
Figure D.2: ESA tank test A - MON-3 (25% fill ratio & fext = 1.50 Hz): CoGplots.
136
Figure D.3: ESA tank test A - MON-3 (25% fill ratio & fext = 0.70 Hz): sloshingamplitude plot.
Figure D.4: ESA tank test A - MON-3 (25% fill ratio & fext = 1.50 Hz): sloshingamplitude plot.
137
Figure D.5: ESA tank test A - MON-3 (25% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top left point of the tank).
Figure D.6: ESA tank test A - MON-3 (25% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top center point of the tank).
138
Figure D.7: ESA tank test A - MON-3 (25% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top left point of the tank).
Figure D.8: ESA tank test A - MON-3 (25% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top center point of the tank).
139
MON-3 - 50 % fill ratio
Figure D.9: ESA tank test A - MON-3 (50% fill ratio & fext = 1.50 Hz): CoGplots.
Figure D.10: ESA tank test A - MON-3 (50% fill ratio & fext = 1.50 Hz): sloshingamplitude plot.
140
MON-3 - 75 % fill ratio
Figure D.11: ESA tank test A - MON-3 (75% fill ratio & fext = 0.70 Hz): CoGplots.
Figure D.12: ESA tank test A - MON-3 (75% fill ratio & fext = 1.50 Hz): CoGplots.
141
Figure D.13: ESA tank test A - MON-3 (75% fill ratio & fext = 0.70 Hz): sloshingamplitude plot.
Figure D.14: ESA tank test A - MON-3 (75% fill ratio & fext = 1.50 Hz): sloshingamplitude plot.
142
Figure D.15: ESA tank test A - MON-3 (75% fill ratio & fext = 0.70 Hz): PSDplot (measured at the top left point of the tank).
Figure D.16: ESA tank test A - MON-3 (75% fill ratio & fext = 0.70 Hz): PSDplot (measured at the top center point of the tank).
143
Figure D.17: ESA tank test A - MON-3 (75% fill ratio & fext = 1.50 Hz): PSDplot (measured at the top left point of the tank).
Figure D.18: ESA tank test A - MON-3 (75% fill ratio & fext = 1.50 Hz): PSDplot (measured at the top center point of the tank).
144
MMH - 25 % fill ratio
Figure D.19: ESA tank test A - MMH (25% fill ratio & fext = 0.70 Hz): CoGplots.
Figure D.20: ESA tank test A - MMH (25% fill ratio & fext = 1.50 Hz): CoGplots.
145
Figure D.21: ESA tank test A - MMH (25% fill ratio & fext = 0.70 Hz): sloshingamplitude plot.
Figure D.22: ESA tank test A - MMH (25% fill ratio & fext = 1.50 Hz): sloshingamplitude plot.
146
Figure D.23: ESA tank test A - MMH (25% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top left point of the tank).
Figure D.24: ESA tank test A - MMH (25% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top center point of the tank).
147
Figure D.25: ESA tank test A - MMH (25% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top left point of the tank).
Figure D.26: ESA tank test A - MMH (25% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top center point of the tank).
148
MMH - 50 % fill ratio
Figure D.27: ESA tank test A - MMH (50% fill ratio & fext = 0.70 Hz): CoGplots.
Figure D.28: ESA tank test A - MMH (50% fill ratio & fext = 1.50 Hz): CoGplots.
149
Figure D.29: ESA tank test A - MMH (50% fill ratio & fext = 0.70 Hz): sloshingamplitude plot.
Figure D.30: ESA tank test A - MMH (50% fill ratio & fext = 1.50 Hz): sloshingamplitude plot.
150
Figure D.31: ESA tank test A - MMH (50% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top left point of the tank).
Figure D.32: ESA tank test A - MMH (50% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top center point of the tank).
151
Figure D.33: ESA tank test A - MMH (50% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top left point of the tank).
Figure D.34: ESA tank test A - MMH (50% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top center point of the tank).
152
MMH - 75 % fill ratio
Figure D.35: ESA tank test A - MMH (75% fill ratio & fext = 0.70 Hz): CoGplots.
Figure D.36: ESA tank test A - MMH (75% fill ratio & fext = 1.50 Hz): CoGplots.
153
Figure D.37: ESA tank test A - MMH (75% fill ratio & fext = 0.70 Hz): sloshingamplitude plot.
Figure D.38: ESA tank test A - MMH (75% fill ratio & fext = 1.50 Hz): sloshingamplitude plot.
154
Figure D.39: ESA tank test A - MMH (75% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top left point of the tank).
Figure D.40: ESA tank test A - MMH (75% fill ratio & fext = 0.70 Hz): PSD plot(measured at the top center point of the tank).
155
Figure D.41: ESA tank test A - MMH (75% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top left point of the tank).
Figure D.42: ESA tank test A - MMH (75% fill ratio & fext = 1.50 Hz): PSD plot(measured at the top center point of the tank).
156
Appendix E
Test case 3 - Results: Test B
157
MMH - 50 % fill ratio
(a) t = 0s to t = 4s (200 time steps) (b) Zoom: t = 3.46s
Figure E.1: ESA tank test B - MMH (50% fill ratio): CoG x-coordinate Vs time.
Figure E.2: ESA tank test B - MMH (50% fill ratio): maximum wave amplitude(t = 3.46s).
158
Figure E.3: ESA tank test B - MMH (50% fill ratio): sloshing amplitude plot.
159
Appendix F
Test case 3 - Results: Test C
160
MON-3 - 50 % fill ratio
Figure F.1: ESA tank test C - MON-3: PSD plot (measured at the top right pointof the tank) for the 15 s period after the abrupt removal of a stabilized 0.1−g lateralperturbation.
Figure F.2: ESA tank test C - MON-3 (20s simulation): sloshing amplitude plot.
161
MMH - 50 % fill ratio
Figure F.3: ESA tank test C - MMH: PSD plot (measured at the top right point ofthe tank) for the 15 s period after the abrupt removal of a stabilized 0.1− g lateralperturbation.
Figure F.4: ESA tank test C - MON-3 (60s simulation): CoG x-coord. Vs time.
162
Figure F.5: ESA tank test C - MMH (60s simulation): wave amplitude at t = 30s.
Figure F.6: ESA tank test C - MMH (60s simulation): wave amplitude at t =45.46s.
163
Figure F.7: ESA tank test C - MMH (60s simulation): CoG z-coord. Vs time.
Figure F.8: ESA tank test C - MMH (20s simulation): wave amplitude at t = 0.52s.
164
Figure F.9: ESA tank test C - MMH (20s simulation): wave amplitude at t =19.58s.
Figure F.10: ESA tank test C - MMH (20s simulation): CoG x-coord. Vs time.
165
Figure F.11: ESA tank test C - MMH (20s simulation): CoG z-coord. Vs time.
Figure F.12: ESA tank test C - MMH (60s simulation): sloshing amplitude plot.
Figure F.13: ESA tank test C - MMH (20s simulation): sloshing amplitude plot.
166