RAFAEL DE ANDRADE WATAI

A TIME-DOMAIN BOUNDARY ELEMENTS METHOD FOR THE

SEAKEEPING ANALYSIS OF OFFSHORE SYSTEMS

SÃO PAULO
2015
 RAFAEL DE ANDRADE WATAI

A TIME-DOMAIN BOUNDARY ELEMENTS METHOD FOR THE

SEAKEEPING ANALYSIS OF OFFSHORE SYSTEMS

Thesis submitted to the Escola Politécnica da

Universidade de São Paulo for the partial fulfil-
ment of the requirements for the degree of
Doctor of Science

SÃO PAULO
2015
 RAFAEL DE ANDRADE WATAI

A TIME-DOMAIN BOUNDARY ELEMENTS METHOD FOR THE

SEAKEEPING ANALYSIS OF OFFSHORE SYSTEMS

Thesis submitted to the Escola Politécnica da

Universidade de São Paulo for the partial fulfil-
ment of the requirements for the degree of
Doctor of Science

Research Area:
Naval Architecture and Ocean Engineering

Supervisor: Prof. Alexandre Nicolaos Simos

SÃO PAULO
2015
Este exemplar foi revisado e corrigido em relação à versão original, sob
responsabilidade única do autor e com a anuência de seu orientador.

São Paulo, 02 de Fevereiro de 2015.

Assinatura do autor

Assinatura do orientador

Catalogação-na-publicação

Watai, Rafael de Andrade

A time-domain boundary elements method for the
seakeeping analysis of offshore systems / R.A. Watai. – versão
corr. – São Paulo, 2015.
259 p.

Tese (Doutorado) – Escola Politécnica da Universidade de

São Paulo. Departamento de Engenharia Naval e Oceânica.

1.Comportamento no mar 2.Método de elementos de

contorno no domı́nio do tempo 3.Sistemas multi-corpos
I.Universidade de São Paulo. Escola Politécnica. Departa-
mento de Engenharia Naval e Oceânica II.t.
I dedicate this thesis to my parents Ariovaldo
and Vania and my brother Felipe.
To the woman of my life Isadora, for her love,
inspiration and patience.
 ACKNOWLEDGMENTS

I would like to express my gratitude to my friend Prof. Alexandre Nicolaos Simos, who has
been supervising my scientific activities since my graduation period. I really appreciate
his dedication on following closely the development of this thesis, keeping me motivated
and confident to conduct my studies during the last years.

I gratefully acknowledge Prof. Kazuo Nishimoto for giving me numerous opportunities

and votes of confidence, which were positively decisive in my personal and professional
development. I am also very thankful to Prof. Claudio Mueller Prado Sampaio, Prof.
Eduardo Aoun Tannuri and Prof. André Luis Condino Fujarra of the laboratory TPN-
USP1 . The conditions offered me at TPN in terms of office space and resources were
excellent, providing me an adequate environment to write this thesis.

I am also indebted to Prof. Mardel Bongiovanni de Conti for teaching me the first lessons
on programming the boundary elements method and to Prof. Celso Pupo Pesce by the
relevant advices and text corrections, which certainly contributed to the achievements
reached in this work.

A special thank you goes to my colleagues Guilherme Rosetti, Felipe Ruggeri and Rodrigo
Lavieri with whom I have been learning and sharing many experiences for the last years.
It has been a real pleasure working with such a talented and successful team, in which I
truly want to continue being part and contributing for the next years.

I appreciate the valuable discussions, studies and technical publications conducted to-
gether with Pasquale Dinoi and Prof. Antonio Souto-Iglesias from ETSIN-UPM2 , who
also gave me the opportunity to conduct a set of experimental tests in their facilities, in
Madrid, in Spain, from which I obtained results that significantly increased the quality of
1
Tanque de Provas Numérico da Escola Politécnica da Universidade de São Paulo
2
Escuela Técnica Superior de Ingenieros Navales de la Universidad Politécnica de Madrid
this thesis.

3
I also acknowledge the financial support of FAPESP PhD grant (2010/08778-2).

I am grateful to my friends who have contributed directly or not to the development of this
work: Pedro Cardozo de Mello and Oygres Siqueira for their support on the conduction
and planning of the experimental tests; Fabio Tadao Matsumoto and Edgard Borges Malta
for the technical discussions, which helped me on solving many obstacles along the way;
To all my friends of TPN with whom I have been working with every day; To my dear
friends Fabio Massatoshi, Guilherme Daker, Fernando Alvarez, Kin Honda, Dante Cirillo,
Guilherme Teixeira, Lucas Pesinato, Rodrigo Seródio, Ricardo Bertoni, Daniel Soares,
Mariana Rhein, Fabiola Alvarez, Paula Leardini, Ana Paula Medeiros, Érica Sundfeld,
Débora Takemura and Carolina Oliveira by the entertaining and necessary rest times
during the weekends; To Isabel, Flavio and Bernardo Hoffmann for making me feel like I
have a second family.

I am very thankful to my wife Isadora for being always by my side, motivating and giving
me love and care every day. Thank you very much for the infinite patience and for my
undesirable delays in several evenings, which forced you to stay with me at work until
very late hours. You can count on me for everything and forever.

Finally, I am most thankful to my brother Felipe and to my parents Ariovaldo and Vania
for all the support, trust and love. Your dedication on ensuring me the best education as
possible is the most valuable lesson that I have ever learned since I was born and I will
certainly pass on this learning to your grandchildren.

3
Fundação de Amparo à Pesquisa do Estado de São Paulo
 RESUMO

Esta tese apresenta o desenvolvimento de um método de elementos de contorno (BEM) no

domı́nio do tempo baseado em fontes de Rankine para análise linear de comportamento
no mar de sistemas oceânicos. O método é formulado por dois problemas de valor inicial
de contorno definidos para os potenciais de velocidade e aceleração, sendo este último uti-
lizado para calcular de maneira acurada a derivada temporal do potencial de velocidades.
Testes de verificação são realizados para a solução dos problemas de difração, radiação e
de corpo livre para flutuar. Uma vez verificada, a ferramenta é aplicada em dois proble-
mas multicorpos considerados no estado-da-arte em termos de modelagem hidrodinâmica
utilizando BEM. O primeiro trata do problema envolvendo duas embarcações atracadas à
contrabordo. Este é um caso no qual os códigos baseados na teoria de escoamento poten-
cial são conhecidos por apresentarem dificuldades na determinação das soluções, tendendo
a superestimar as elevações de onda no vão entre as embarcações e a apresentar proble-
mas de convergência numérica associados a efeitos ressonantes de onda. O problema é
tratado por meio do método de damping lid e a convergência das séries temporais é inves-
tigada avaliando diferentes nı́veis de amortecimento. Os resultados são comparados com
dados experimentais. O segundo problema se refere à análise de sistemas multicorpos com
grandes deslocamentos relativos. Neste problema, ferramentas no domı́nio da frequência
não podem ser utilizadas, por considerarem apenas malhas fixas. Deste modo, o presente
método é estendido para considerar um gerador de malhas de painéis e um algoritmo
de interpolação de ordem alta no laço de tempo do código, possibilitando a mudança de
posições relativas entre os corpos durante a simulação. Os resultados são comparados
com dados de experimentos executados especificamente para fins de verificação do código,
apresentando uma boa concordância. De acordo com o conhecimento do autor, esta é a
primeira vez que certas questões relativas à modelagem numérica destes dois problemas
multicorpos são relatadas na literatura especializada em hidrodinâmica computacional.
Palavras-chave: comportamento no mar, método de elementos de contorno no domı́nio
do tempo, sistemas multicorpos
 ABSTRACT

The development of a time domain boundary elements method (BEM) based on Rank-
ine’s sources for linear seakeeping analysis of offshore systems is here addressed. The
method is formulated by means of two Initial Boundary Value Problems defined for the
velocity and acceleration potentials, the latter being used to ensure an accurate calcu-
lation of the time derivatives of the velocity potential. Verification tests for solving the
diffraction, radiation and free floating problems are presented. Once verified, the code
is applied for two complex multi-body problems considered to be in the state-of-the-art
for hydrodynamic modelling using BEM. The first is the seakeeping problem of two ships
arranged in side-by-side, a problem in which all potential flow codes are known to have a
poor performance, tending to provide unrealistic high wave elevations in the gap between
the vessels and to present numerical convergence problems associated to resonant effects.
The problem is here addressed by means of a damping lid method and the convergence
of the time series with different damping levels is investigated. Results are compared to
data measured in an experimental campaign. The second problem refers to the analysis of
multi-body systems composed of bodies undergoing large relative displacements. This is a
case that cannot be properly analyzed by frequency domain codes, since they only consider
fixed meshes. For this application, the present numerical method is extended to consider
a panel mesh generator in the time loop of the code, enabling the change of body relative
positions during the computations. Furthermore, a higher order interpolation algorithm
designed to recover the solutions of a previous time-step was also implemented, enabling
the calculations to progress with reasonable accuracy in time. The numerical results are
compared to data of experimental tests designed and executed for verification of the code,
and presented a very good agreement. To the author’s knowledge, this is the first time
that certain issues concerning the numerical modelling of these two complex multi-body
problems are reported in the literature specialized in hydrodynamic computations.
Keywords: seakeeping, time domain boundary elements method, multi-body systems
 List of Figures

Figure 3.1: Definition of the coordinate systems . . . . . . . . . . . . . . . . . . 46

Figure 4.1: Surface of integration for the Second Green’s Identity derivation . . 63

Figure 5.1: Loop structure of the numerical method . . . . . . . . . . . . . . . 68

Figure 5.2: Physical domain (a) and computational iso-parametric domain (b) . 73

Figure 5.3: Typical relative positions between the field point (red marker) and
panel centroid (blue marker) for a far field influence coefficient clas-
sification (L  1/d) (a) and the correspondent function values in
the parametric domain for the source (b) and dipole (c) terms . . . 76

Figure 5.4: Convergence of the source far field influence coefficient IFS ar . . . . . 76

Figure 5.5: Convergence of the dipole far field influence coefficient IFDar . . . . . 77

Figure 5.6: Typical relative positions between the field point (red marker) and
panel centroid (blue marker) for a near field influence coefficient
classification (L/d ∼ O(1)) (a) and the correspondent function val-
ues in the parametric domain for the source (b) and dipole (c) terms 77

Figure 5.7: Convergence of the source near field influence coefficient INS ear . . . 78

Figure 5.8: Convergence of the dipole near field influence coefficient INDear . . . 78

Figure 5.9: Typical relative positions between the field point (red marker) and
panel centroid (blue marker) for a self-influence coefficient classi-
fication (d = 0) (a) and the correspondent function values in the
parametric domain for the source term (b) . . . . . . . . . . . . . . 79
S
Figure 5.10: Convergence of the source self-influence coefficient ISelf . . . . . . . 79

Figure 5.11: Triangulation of the parametric space. Figure adapted from Maniar
(1995) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

Figure 5.12: Function values obtained from the new transformations: ∆(1) (a),
∆(2) (b), ∆(3) (c) and ∆(4) (d) . . . . . . . . . . . . . . . . . . . . . 81

Figure 5.13: Comparison of the direct and partitioning methods for the conver-
S
gence of the source self-influence coefficient ISelf . . . . . . . . . . . 82
Figure 6.1: Circular cylinder and hemisphere panel meshes used in the compu-
tations. Figures (a), (c) and (e) refers to the 120, 500 and 2000
circular cylinder panel meshes, respectively. Figures (b), (d) and
(f) refers to the 200, 800 and 3200 hemisphere panel meshes, re-
spectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Figure 6.2: Free surface panel meshes used in the computations. Figures (a), (b)
and (c) refers to the 289, 1225 and 4900 free surface panel meshes,
respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

Figure 6.3: Hydrodynamic horizontal force Fx induced by a wave with frequency

of ω = 4.202 rad/s. Cylinder body . . . . . . . . . . . . . . . . . . 91

Figure 6.4: Hydrodynamic horizontal force Fx induced by a wave with frequency

of ω = 6.264 rad/s. Cylinder body . . . . . . . . . . . . . . . . . . 91

Figure 6.5: Hydrodynamic horizontal force Fx induced by a wave with frequency

of ω = 9.905 rad/s. Cylinder body . . . . . . . . . . . . . . . . . . 92

Figure 6.6: Hydrodynamic vertical force Fz induced by a wave with frequency

of ω = 4.202 rad/s. Cylinder body . . . . . . . . . . . . . . . . . . 92

Figure 6.7: Hydrodynamic vertical force Fz induced by a wave with frequency

of ω = 6.264 rad/s. Cylinder body . . . . . . . . . . . . . . . . . . 93

Figure 6.8: Hydrodynamic vertical force Fz induced by a wave with frequency

of ω = 9.905 rad/s. Cylinder body . . . . . . . . . . . . . . . . . . 93

Figure 6.9: Convergence analysis of the non-dimensional horizontal hydrody-

namic force F̄x for the hemisphere body . . . . . . . . . . . . . . . 95

Figure 6.10: Convergence analysis of the non-dimensional vertical hydrodynamic

force F̄z for the hemisphere body . . . . . . . . . . . . . . . . . . . 95

Figure 6.11: Convergence analysis of the non-dimensional horizontal hydrody-

namic force F̄x for the circular cylinder body . . . . . . . . . . . . . 98

Figure 6.12: Convergence analysis of the non-dimensional vertical hydrodynamic

force F̄z for the circular cylinder body . . . . . . . . . . . . . . . . 98

Figure 6.13: Convergence analysis of the non-dimensional hydrodynamic mo-

ment M̄y for the circular cylinder body . . . . . . . . . . . . . . . . 99

Figure 6.14: Determination of A33 and B33 coefficients . . . . . . . . . . . . . . . 105

Figure 6.15: Convergence analysis of the added mass coefficient Ā11 for the hemi-
sphere body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

Figure 6.16: Convergence analysis of the radiation damping coefficient B̄11 for
the hemisphere body . . . . . . . . . . . . . . . . . . . . . . . . . . 107

Figure 6.17: Convergence analysis of the added mass coefficient Ā33 for the hemi-
sphere body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Figure 6.18: Convergence analysis of the radiation damping coefficient B̄33 for
the hemisphere body . . . . . . . . . . . . . . . . . . . . . . . . . . 110

Figure 6.19: Convergence analysis of the added mass coefficient Ā11 for the cir-
cular cylinder body . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

Figure 6.20: Convergence analysis of the radiation damping coefficient B̄11 for
the circular cylinder body . . . . . . . . . . . . . . . . . . . . . . . 113

Figure 6.21: Convergence analysis of the added mass coefficient Ā33 for the cir-
cular cylinder body . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

Figure 6.22: Convergence analysis of the radiation damping coefficient B̄33 for
the circular cylinder body . . . . . . . . . . . . . . . . . . . . . . . 116

Figure 6.23: Convergence analysis of the added mass coefficient Ā55 for the cir-
cular cylinder body . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

Figure 6.24: Convergence analysis of the radiation damping coefficient B̄55 for
the circular cylinder body . . . . . . . . . . . . . . . . . . . . . . . 119

Figure 6.25: Time history of the cylinder heave motion for the coarse, medium
and fine meshes. Incoming wave with angular frequency ω = 2.51
rad/s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

Figure 6.26: Time history of the hemisphere heave motion for the coarse, medium
and fine meshes. Incoming wave with angular frequency ω = 3.13
rad/s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

Figure 6.27: Convergence analysis of the cylinder surge RAO and comparison
with WAMIT data . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

Figure 6.28: Convergence analysis of the cylinder heave RAO and comparison
with WAMIT data . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

Figure 6.29: Convergence analysis of the cylinder pitch RAO and comparison
with WAMIT data . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

Figure 6.30: Convergence analysis of the sphere surge RAO and comparison with
WAMIT data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

Figure 6.31: Convergence analysis of the sphere heave RAO and comparison with
WAMIT data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

Figure 7.1: FPSO “Box Shaped” panel mesh . . . . . . . . . . . . . . . . . . . . 132

Figure 7.2: Free surface panel mesh for the FPSO “Box Shaped” simulations . . 133

Figure 7.3: Snapshots of the simulation at 9 different instants of time. Wave

Amplitude AI = 1 m and frequency ω = 0.370 rad/s. (a) t = 0 s,
(b) t = 20 s,(c) t = 40 s, (d) t = 60 s, (e) t = 80 s, (f) t = 100 s,
(g) t = 120 s, (h) t = 140 s and (i) t = 160 s . . . . . . . . . . . . . 134
Figure 7.4: Typical surge motion time series obtained by the present numerical
code. Motions were induced by a wave of angular frequency ω =
0.370 rad/s and incidence angle 0◦ . . . . . . . . . . . . . . . . . . 135

Figure 7.5: Typical heave motion time series obtained by the present numerical
code. Motions were induced by a wave of angular frequency ω =
0.370 rad/s and incidence angle 0◦ . . . . . . . . . . . . . . . . . . 135

Figure 7.6: Typical pitch motion time series obtained by the present numerical
code. Motions were induced by a wave of angular frequency ω =
0.370 rad/s and incidence angle 0◦ . . . . . . . . . . . . . . . . . . 136

Figure 7.7: Numerical white noise spectrum . . . . . . . . . . . . . . . . . . . . 137

Figure 7.8: Incoming wave elevation time series generated by a white noise spec-
trum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

Figure 7.9: Motion time series for sway (ξ2 ), heave (ξ3 ) and roll (ξ4 ) induced by
a wave generated by a white noise spectrum . . . . . . . . . . . . . 138

Figure 7.10: Surge response amplitude operators considering three different wave
incidence angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and
White Noise Analysis (Below) . . . . . . . . . . . . . . . . . . . . . 139

Figure 7.11: Sway response amplitude operators considering three different wave
incidence angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and
White Noise Analysis (Below) . . . . . . . . . . . . . . . . . . . . . 140

Figure 7.12: Heave response amplitude operators considering three different wave
incidence angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and
White Noise Analysis (Below) . . . . . . . . . . . . . . . . . . . . . 140

Figure 7.13: Roll response amplitude operators considering three different wave
incidence angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and
White Noise Analysis (Below) . . . . . . . . . . . . . . . . . . . . . 141

Figure 7.14: Pitch response amplitude operators considering three different wave
incidence angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and
White Noise Analysis (Below) . . . . . . . . . . . . . . . . . . . . . 141

Figure 7.15: Yaw response amplitude operators considering three different wave
incidence angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and
White Noise Analysis (Below) . . . . . . . . . . . . . . . . . . . . . 142

Figure 7.16: Hydrodynamic Calibrator of the Numerical Offshore Tank at the

University of Sao Paulo (CH-TPN-USP) . . . . . . . . . . . . . . . 143

Figure 7.17: General view of the FPSO VLCC model positioned in CH-TPN-USP144

Figure 7.18: FPSO VLCC hull panel mesh . . . . . . . . . . . . . . . . . . . . . 146

Figure 7.19: Free surface panel mesh for the FPSO simulations . . . . . . . . . . 146
Figure 7.20: Comparison between numerical results and experimental data of
sway response amplitude operator for the FPSO VLCC hull . . . . 147

Figure 7.21: Comparison between numerical results and experimental data of

heave response amplitude operator for the FPSO VLCC hull . . . . 147

Figure 7.22: Comparison between numerical results and experimental data of roll
response amplitude operator for the FPSO VLCC hull . . . . . . . 148

Figure 7.23: Comparison between numerical and experimental motion time series
for sway (ξ2 ), heave (ξ3 ) and roll (ξ4 ), considering C44 = 0.05Ccrit . 149

Figure 7.24: Comparison between numerical and experimental motion time series
for sway (ξ2 ), heave (ξ3 ) and roll (ξ4 ), considering C44 = 0.06Ccrit . 149

Figure 8.1: (b) Barge geometry and main dimensions in meters. (b) Geosim
geometry and main dimensions in meters . . . . . . . . . . . . . . . 155

Figure 8.2: Details of the geosim geometry and main dimensions in meters . . . 155

Figure 8.3: Models positions in relation to the towing tank walls. Dimensions
in meters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

Figure 8.4: General view of the models positioned in the tank . . . . . . . . . . 157

Figure 8.5: Wave probes arranged along the gap centerline . . . . . . . . . . . . 157

Figure 8.6: Body images considered in the multi-body model. Dimensions in

meters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

Figure 8.7: Symmetry line applied for the frequency domain computations. Di-
mensions in meters . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

Figure 8.8: Case 1: Comparison between WAMIT numerical data and exper-
imental values in terms of heave (top) and pitch (bottom) RAOs.
Gap width = 0.05 m . . . . . . . . . . . . . . . . . . . . . . . . . . 162

Figure 8.9: Case 1: Comparison between WAMIT numerical data and experi-
mental values in terms of wave elevations RAOs in different points
positioned along the gap centerline. Gap width = 0.05m . . . . . . 163

Figure 8.10: Case 2: Comparison between WAMIT numerical data and experi-
mental values in terms of heave (top) and pitch (below) RAOs. Gap
width = 0.1 m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

Figure 8.11: Case 2: Comparison between WAMIT numerical data and experi-
mental values in terms of wave elevations RAOs in different points
positioned along the gap centerline. Gap width = 0.1 m . . . . . . 165

Figure 8.12: Barge, geosim and gap panel meshes for Case 1 . . . . . . . . . . . 166

Figure 8.13: Free surface panel mesh for Case 1 . . . . . . . . . . . . . . . . . . 167

Figure 8.14: Case 1: Comparison between WAMIT, present values and experi-
mental data in terms of heave (top) and pitch (bottom) RAOs. Gap
width = 0.05 m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

Figure 8.15: Case 1: Comparison between WAMIT, present values and experi-
mental data in terms of wave elevations RAOs in different points
positioned along the gap centerline. Gap width = 0.05 m . . . . . . 170

Figure 8.16: Case 2: Comparison between WAMIT, present values and experi-
mental values in terms of heave (top) and pitch (bottom) RAOs.
Gap width = 0.1 m . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

Figure 8.17: Case 2: Comparison between WAMIT, present values and experi-
mental values in terms of wave elevations RAOs in different points
positioned along the gap centerline. Gap width = 0.1 m . . . . . . 172

Figure 8.18: Case 1: Gap wave elevation envelopes in spatial domain computed
with the present method. Incoming wave frequencies ω = 8.3 rad/s
(top) and ω = 8.5 rad/s (bottom). Wave propagates from positive
to negative x coordinates. Gap width = 0.05 m . . . . . . . . . . . 173

Figure 8.19: Case 2: Gap wave elevation envelopes in spatial domain computed
with the present method. Incoming wave frequencies ω = 7.7 rad/s
(top) and ω = 7.8 rad/s (bottom). Wave propagates from positive
to negative x coordinates. Gap width = 0.10 m . . . . . . . . . . . 174

Figure 8.20: Case 1: Numerical heave (top) and pitch (bottom) motions time
series for different values of damping factor . Incoming wave fre-
quency ω = 8.3 rad/s. Gap width=0.05 m . . . . . . . . . . . . . . 178

Figure 8.21: Case 1: Numerical heave (top) and pitch (bottom) motions time
series for different values of damping factor . Incoming wave fre-
quency ω = 8.5 rad/s. Gap width=0.05 m . . . . . . . . . . . . . . 179

Figure 8.22: Case 1: Numerical wave elevation time series at P2 (top), P3 (mid-
dle) and P4 (bottom) for different values of damping factors . In-
coming wave frequency ω = 8.3 rad/s. Gap width = 0.05 m . . . . 180

Figure 8.23: Case 1: Numerical wave elevation time series at P2 (top), P3 (mid-
dle) and P4 (bottom) for different values of damping factors . In-
coming wave frequency ω = 8.5 rad/s. Gap width = 0.05 m . . . . 181

Figure 8.24: Case 1: Comparison between numerical and experimental data of

the gap wave elevation envelopes in spatial domain for different
values of damping factors . (a) Incoming wave frequency ω =
8.3 rad/s. (b) Incoming wave frequency ω = 8.5 rad/s. Wave
propagates from positive to negative x coordinates. Gap width =
0.05 m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Figure 8.25: Case 1: Comparison between numerical and experimental data of
the gap wave elevation envelopes in spatial domain for different
values of damping factors . (a) Incoming wave frequency ω =
7.7 rad/s. (b) Incoming wave frequency ω = 7.8 rad/s. Wave
propagates from positive to negative x coordinates. Gap width =
0.10 m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

Figure 8.26: Case 1: Numerical and experimental comparisons in terms of heave

(top) and pitch (bottom) motions RAOs for different values of damp-
ing factors . Gap width = 0.05 m . . . . . . . . . . . . . . . . . . 184

Figure 8.27: Case 2: Numerical and experimental comparisons in terms of heave

(top) and pitch (bottom) motions RAOs for different values of damp-
ing factors . Gap width = 0.10 m . . . . . . . . . . . . . . . . . . 185

Figure 8.28: Case 1: Numerical and experimental comparisons of wave elevations

in the gap for different values of damping factors . Gap width =
0.05 m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

Figure 8.29: Case 2: Numerical and experimental comparisons of wave elevations

in the gap for different values of damping factors . Gap width =
0.10 m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

Figure 9.1: Structure of the numerical method including the re-meshing and
interpolation algorithms in the time loop . . . . . . . . . . . . . . . 193

Figure 9.2: Steps of the mesh generation algorithm applied to a circular surface
with radius R = 3 m with a circular of radius R = 1 m inside.
Figures (a), (b) and (c) refer to the first three steps of the algorithm
loop structure. Figure (d) illustrates the final mesh . . . . . . . . . 196

Figure 9.3: Four circular meshes of radius R = 5 m with two circular apertures
of radius R = 1 m in different relative distances . . . . . . . . . . . 197

Figure 9.4: Illustrative sketch of the experimental setup . . . . . . . . . . . . . 201

Figure 9.5: Lateral view of the models positioned in the tank . . . . . . . . . . 201

Figure 9.6: ATI MINI85 6 D.O.F Load Cell. Source: ATI Catalog . . . . . . . 202

Figure 9.7: Sketch of the mechanical equipment attached to Body 2 . . . . . . 202

Figure 9.8: Front view sketch of the models and wave probes arrangement . . . 203

Figure 9.9: Video frames recorded during the tests with a regular wave of fre-
quency ωI = 7.2 rad/s and Body 2 oscillation frequency ωpm = 0.48
rad/s. Instants (a) t = to s, (b) t = to + T /4 s and (c) t = to + 3T /4 s205

Figure 9.10: Free surface mesh for wave Reg 1. Wave frequency ωI = 6.4 rad/s . 206

Figure 9.11: Illustrative view of the numerical setup . . . . . . . . . . . . . . . . 207

Figure 9.12: Case 1: Comparison of numerical and experimental results in terms
of position of body 2 in time and the associated time series of the
hydrodynamic forces Fx , Fy and Fz . . . . . . . . . . . . . . . . . . 209

Figure 9.13: Case 1: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
hydrodynamic moments Mx and My . . . . . . . . . . . . . . . . . 209

Figure 9.14: Case 4: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
hydrodynamic forces Fx , Fy and Fz . . . . . . . . . . . . . . . . . . 210

Figure 9.15: Case 4: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
hydrodynamic moments Mx and My . . . . . . . . . . . . . . . . . 211

Figure 9.16: Case 7: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
hydrodynamic forces Fx , Fy and Fz . . . . . . . . . . . . . . . . . . 211

Figure 9.17: Case 7: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
hydrodynamic moments Mx and My . . . . . . . . . . . . . . . . . 212

Figure 9.18: Case 10: Comparison of numerical and experimental results in terms
of position of body 2 in time and the associated time series of the
hydrodynamic forces Fx , Fy and Fz . . . . . . . . . . . . . . . . . . 212

Figure 9.19: Case 10: Comparison of numerical and experimental results in terms
of position of body 2 in time and the associated time series of the
hydrodynamic moments Mx and My . . . . . . . . . . . . . . . . . 213

Figure 9.20: Case 3: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
hydrodynamic forces Fx , Fy and Fz . . . . . . . . . . . . . . . . . . 213

Figure 9.21: Case 3: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
hydrodynamic moments Mx and My . . . . . . . . . . . . . . . . . 214

Figure 9.22: Case 5: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
hydrodynamic forces Fx , Fy and Fz . . . . . . . . . . . . . . . . . . 214

Figure 9.23: Case 5: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
hydrodynamic moments Mx and My . . . . . . . . . . . . . . . . . 215

Figure 9.24: Case 1: Comparison between present method and experimental data
of Fx and My response spectra . . . . . . . . . . . . . . . . . . . . . 216
Figure 9.25: Case 2: Comparison between present method and experimental data
of Fx and My response spectra . . . . . . . . . . . . . . . . . . . . . 217

Figure 9.26: Case 3: Comparison between present method and experimental data
of Fx and My response spectra . . . . . . . . . . . . . . . . . . . . . 217

Figure 9.27: Case 1: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . . . . . 219

Figure 9.28: Case 4: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . . . . . 219

Figure 9.29: Case 7: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . . . . . 220

Figure 9.30: Case 10: Comparison of numerical and experimental results in terms
of position of body 2 in time and the associated time series of the
wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . . . . . 220

Figure 9.31: Case 3: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . . . . . 221

Figure 9.32: Case 8: Comparison of numerical and experimental results in terms

of position of body 2 in time and the associated time series of the
wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . . . . . 221

Figure 9.33: Case 1: Two snapshots of the simulation illustrating the maximum
wave elevation calculated at WP1 (t = 36.23) s. Perspective (a)
and front views (b). Vertical black line refers to the position of the
wave probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

Figure 9.34: Case 1: Two snapshots of the simulation illustrating the maximum
wave elevation calculated at WP2 (t = 36.72) s. Perspective (a)
and front views (b). Vertical black line refers to the position of the
wave probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

Figure A.1: Comparison of hydrodynamic vertical force F3 on a hemisphere in-

duced by a forced vertical oscillatory motion of amplitude A3 = 1.0
m, considering a damping zone length of half generated wave length
b = λ/2 and different values of dissipation intensity a . . . . . . . . 242

Figure A.2: Comparison of hydrodynamic vertical force F3 on a hemisphere in-

duced by a forced vertical oscillatory motion of amplitude A3 = 1
m, considering a damping zone length of one generated wave length
b = λ and different values of dissipation intensity a . . . . . . . . . 242
Figure A.3: Comparison of hydrodynamic vertical force F3 on a hemisphere in-
duced by a forced vertical oscillatory motion of amplitude A3 = 1
m, considering a damping zone length of one and a half generated
wave length b = 1.5λ and different values of dissipation intensity a . 243

Figure A.4: Comparison of hydrodynamic vertical force F3 on a hemisphere in-

duced by a forced vertical oscillatory motion of amplitude A3 = 1
m, considering a damping zone length of two generated wave lengths
b = 2.0λ and different values of dissipation intensity a . . . . . . . . 243

Figure A.5: Snapshots of the simulation without damping zone at 10 different

instants of time within one period. Vertical oscillatory motion am-
plitude A3 = 0.5 m (a) t = 56.98 s, (b) t = 57.24 s,(c) t = 57.50 s,
(d) t = 57.76 s, (e) t = 58.01 s, (f) t = 58.27 s, (g) t = 58.53 s, (h)
t = 58.79 s, (i) t = 59.05 s and (j) t = 59.31 s . . . . . . . . . . . . 244

Figure A.6: Snapshots of the simulation with damping zone parameters b = 2.0
and a = 1.0 at 10 different instants of time within one period.
Vertical oscillatory motion amplitude A3 = 0.5 m (a) t = 56.98 s,
(b) t = 57.24 s,(c) t = 57.50 s, (d) t = 57.76 s, (e) t = 58.01 s, (f)
t = 58.27 s, (g) t = 58.53 s, (h) t = 58.79 s, (i) t = 59.05 s and (j)
t = 59.31 s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

Figure B.1: Case 1 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 247

Figure B.2: Case 1: Comparison between numerical and experimental results in

terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 248

Figure B.3: Case 2 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 248

Figure B.4: Case 2: Comparison between numerical and experimental results in

terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 249

Figure B.5: Case 3 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 249

Figure B.6: Case 3: Comparison between numerical and experimental results in

terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 250

Figure B.7: Case 4 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 250
Figure B.8: Case 4: Comparison between numerical and experimental results in
terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 251

Figure B.9: Case 5 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 251

Figure B.10:Case 5: Comparison between numerical and experimental results in

terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 252

Figure B.11:Case 6 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 252

Figure B.12:Case 6: Comparison between numerical and experimental results in

terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 253

Figure B.13:Case 7 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 253

Figure B.14:Case 7: Comparison between numerical and experimental results in

terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 254

Figure B.15:Case 8 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 254

Figure B.16:Case 8: Comparison between numerical and experimental results in

terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 255

Figure B.17:Case 9 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 255

Figure B.18:Case 9: Comparison between numerical and experimental results in

terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 256

Figure B.19:Case 10 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 256

Figure B.20:Case 10: Comparison between numerical and experimental results

in terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 257
Figure B.21:Case 11 Comparison of numerical and experimental results in terms
of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 257

Figure B.22:Case 11: Comparison between numerical and experimental results

in terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 258

Figure B.23:Case 12 Comparison of numerical and experimental results in terms

of position of Body 2 in time and the associated time series of the
hydrodynamic loads Fx , Fy , Fz , Mx and My . . . . . . . . . . . . . 258

Figure B.24:Case 12: Comparison between numerical and experimental results

in terms of position of Body 2 in time and the associated time series
of the wave elevations at WP1, WP2 and WP3 . . . . . . . . . . . 259
 List of Tables

Table 6.1: Main numbers of the panel meshes used in the computations . . . . 88

Table 6.2: Regular incoming waves used in the simulations . . . . . . . . . . . 89

Table 6.3: Convergence analysis of the non-dimensional horizontal hydrody-

namic force F̄x for the hemisphere body. WAMIT values are used
as reference for the relative errors . . . . . . . . . . . . . . . . . . . 96

Table 6.4: Convergence analysis of the non-dimensional vertical hydrodynamic

force F̄z for the hemisphere body. WAMIT values are used as ref-
erence for the relative errors . . . . . . . . . . . . . . . . . . . . . . 97

Table 6.5: Convergence analysis of the non-dimensional horizontal hydrody-

namic force F̄x for the circular cylinder body. WAMIT values are
used as reference for the relative errors . . . . . . . . . . . . . . . . 100

Table 6.6: Convergence analysis of the non-dimensional vertical hydrodynamic

force F̄z for the circular cylinder body. WAMIT values are used as
reference for the relative errors . . . . . . . . . . . . . . . . . . . . 101

Table 6.7: Convergence analysis of the non-dimensional hydrodynamic mo-

ment M̄y for the circular cylinder body. WAMIT values are used as
reference for the relative errors . . . . . . . . . . . . . . . . . . . . 102

Table 6.8: Convergence analysis of the non-dimensional added mass Ā11 for
the hemisphere body. Hulme (1982) values are used as reference for
the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

Table 6.9: Convergence analysis of the non-dimensional radiation damping B̄11

for the hemisphere body. Hulme (1982) values are used as reference
for the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . 109

Table 6.10: Convergence analysis of the non-dimensional added mass Ā33 for
the hemisphere body. Hulme (1982) values are used as reference for
the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

Table 6.11: Convergence analysis of the non-dimensional radiation damping B̄33

for the hemisphere body. Hulme (1982) values are used as reference
for the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Table 6.12: Convergence analysis of the non-dimensional added mass Ā11 for
the circular cylinder body. WAMIT values are used as reference for
the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

Table 6.13: Convergence analysis of the non-dimensional radiation damping B̄11

for the circular cylinder body. WAMIT values are used as reference
for the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . 115

Table 6.14: Convergence analysis of the non-dimensional added mass Ā33 for
the circular cylinder body. WAMIT values are used as reference for
the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

Table 6.15: Convergence analysis of the non-dimensional radiation damping B̄33

for the circular cylinder body. WAMIT values are used as reference
for the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . 118

Table 6.16: Convergence analysis of the non-dimensional added mass Ā55 for
the circular cylinder body. WAMIT values are used as reference for
the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

Table 6.17: Convergence analysis of the non-dimensional radiation damping B̄55

for the circular cylinder body. WAMIT values are used as reference
for the relative errors . . . . . . . . . . . . . . . . . . . . . . . . . . 121

Table 6.18: Main characteristics of the cylinder and hemisphere and simulation
settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

Table 7.1: Principal characteristics of the FPSO “Box Shaped” hull and the
settings of the simulations . . . . . . . . . . . . . . . . . . . . . . . 131

Table 7.2: Regular incoming waves used in the FPSO “Box Shaped” hull sim-
ulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

Table 7.3: Main characteristics of the FPSO VLCC hull and the settings of
the simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

Table 7.4: Regular incoming waves used in the FPSO VLCC hull simulations . 145

Table 8.1: Main characteristics of the barge and geosim models . . . . . . . . 154

Table 8.2: Main characteristics of the geosim including the mechanical guiding
arms data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

Table 8.3: Relative distances between the wave probes . . . . . . . . . . . . . 156

Table 8.4: Regular waves tested for the gap width 0.05 m . . . . . . . . . . . . 158

Table 8.5: Regular waves tested for the gap width 0.1 m . . . . . . . . . . . . 159

Table 9.1: Main characteristics of the 6 D.O.F load cell . . . . . . . . . . . . . 202

Table 9.2: Regular waves considered in the tests . . . . . . . . . . . . . . . . . 204

Table 9.3: Test Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

Table 9.4: Main particulars of the panel meshes used in the computations . . . 206
 Contents

1 Introduction 31

1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

1.2 Text Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2 Literature Review 39

3 Theoretical Background 45

3.1 Problem Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.2 The Initial Boundary Value Problem . . . . . . . . . . . . . . . . . . . . . 45

3.3 The Linear Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.3.1 The Acceleration Potential . . . . . . . . . . . . . . . . . . . . . . . 57

4 Boundary Integral Equations 61

4.1 Integral Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

5 The Numerical Model 67

5.1 Boundary Elements Method . . . . . . . . . . . . . . . . . . . . . . . . . . 69

5.1.1 Integration of the Influence Coefficients . . . . . . . . . . . . . . . . 72

5.1.1.1 Near and Far-Field Influence Coefficients . . . . . . . . . . 73

5.1.1.2 Self-Influence Coefficients . . . . . . . . . . . . . . . . . . 78

5.2 Time-Marching Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

6 Verification Tests with Simple Geometries 85

6.1 Geometries Discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

6.2 Fixed Body Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

 6.3 Forced Motion Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

6.4 Free Floating Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

7 Verification Tests with Free Floating FPSOs 129

7.1 FPSO “Box Shaped” Hull . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

7.2 FPSO VLCC Hull . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

7.2.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

7.2.2 Numerical Simulations . . . . . . . . . . . . . . . . . . . . . . . . . 145

8 Multi-Body Simulations in Side-by-Side Arrangement 151

8.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154

8.1.1 Investigation of the Tank Walls Effects . . . . . . . . . . . . . . . . 159

8.2 Time Domain Numerical Model . . . . . . . . . . . . . . . . . . . . . . . . 165

8.2.1 Comparisons with WAMIT and Experimental Data . . . . . . . . . 167

8.2.2 Damping Lid Approach . . . . . . . . . . . . . . . . . . . . . . . . . 174

9 Multi-Body Simulations with Large Relative Displacements 189

9.1 The free surface re-meshing algorithm . . . . . . . . . . . . . . . . . . . . . 194

9.2 The free surface interpolation algorithm . . . . . . . . . . . . . . . . . . . 197

9.3 Verification of Numerical Computation . . . . . . . . . . . . . . . . . . . . 199

9.3.1 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

9.3.2 Numerical Grids for the Case Studies . . . . . . . . . . . . . . . . . 205

9.3.3 Comparison between Experiments and Calculations . . . . . . . . . 207

10 Conclusions 223

10.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

10.2 Proposed Improvements and Future Work . . . . . . . . . . . . . . . . . . 227

References 229

Appendix A Damping Zone Analysis 241

Appendix B Multi-Body Simulations with Large Relative Displacements 247

B.1 Comparison between Experiments and Computations . . . . . . . . . . . . 247

 31

Chapter 1

Introduction

1.1 Motivation

Seakeeping analysis is a topic of great importance for the design of floating systems, such
as ships and oil and gas production units. In fact, the design of several subsystems and
consequently the costs associated to each one may be directly impacted by the level of
the motions that the floating system may present in waves. For example, oil platform
hulls, even with appropriate responses to the waves excitation that result in low vertical
displacements may be critical for enabling the use of the so-called rigid risers, which if
applied, might reduce the costs for this type of subsystem significantly.

Current practice in the area of ocean engineering, the conduction of experimental ac-
tivities with scaled models have provided great knowledge and understanding of complex
phenomena involving the interaction among floating structures and environmental agents,
such as waves, wind and current. Through these tests, observations and measurements
can be performed in a controlled manner, partially isolating the effects of distinct external
sources present in real environments. Nevertheless, scaled model tests often present diffi-
culties in relation to physical limitations of the basins, specially when modelling systems
that operate in deep waters. Moreover, since the tests require high costs and considerable
time for its conduction, presently they are preferred for final design verifications and not
as a prediction tool for earlier stages of the design.

Another approach commonly adopted in engineering problems is the use of numerical

methods to solve the equations that describe the dynamics of the phenomenon in study.
32

Mathematical models may be solved numerically, providing approximate solutions to very

complex problems which, in most cases, are impossible to be treated analytically. Nu-
merical approaches are becoming increasingly popular in the various stages of engineering
designs, mainly due to the growth of computational power and development of new tech-
niques for numerical processing. In the context of offshore system designs, nowadays, it
is often possible to observe applications of numerical methods, reducing the demand of
experimental tests between the stages of conception and final verification.

The development of numerical tools used for solving problems of real fluid flows is one
of the major areas of research in hydrodynamics. Its application is necessary when the
phenomena that is being studied requires the inclusion of viscous effects as, for example,
the prediction of forward resistance of ships or in the design of bilge keels used to reduce the
amplitude on the roll motions of the ship. However, in spite of its increasingly widespread
use, this approach is still very unpractical for many applications, since it demands high
computational efforts and processing time to tackle the sophisticated numerical methods
that handle the Navier-Stokes equations.

Thus, for the evaluation of seakeeping problems of offshore systems, other types of methods
are applied. Due to the relatively small (and, somehow, punctual) influence of viscous
effects, the problem is commonly addressed with the use of the potential flow theory,
which allows one to determine the solutions more simply and quickly in comparison to
numerical methods that deal with real fluid flows. In this context, the use of numerical
methods that deal with potential boundary value problems, such as the Boundary Ele-
ments Method (BEM), is widely applied throughout the different design stages for solving
the hydrodynamic issues.

Presently, BEM is present in several computational tools used to solve problems related
to the hydrodynamic interaction of floating bodies with waves. Among several different
alternatives for its implementation, frequency domain and time domain approaches are two
distinct large groups that may be analyzed separately. The former approach is always used
for linear or weakly nonlinear wave theories in which the time dependence can be removed
by assuming that the solution is harmonic in time and, therefore, only steady solutions
are obtained. In this case, we may list some examples of well-known commercial softwares
as, for example, WAMIT1 , AQWA2 and WADAM3 . However, if no assumption about the
steadiness of the solution is established, a time domain approach has necessarily to be
applied. Although less efficient for problems involving single harmonic waves, time domain
codes may be extended for more generic cases in which the inclusion of nonlinear terms or
1
http://www.wamit.com
2
http://www.ansys.com/Products/Other+Products/ANSYS+Aqwa
3
http://www.dnv.com/services/software/products/sesam/sesamhydrod/wadam.asp
 33

external loads from different sources (e.g. mooring lines) is implemented straightforwardly.
However, the availability of time domain codes is fairly scarce and although there are
reports of codes, such as TIMIT (KORSMEYER; BINGHAM; NEWMAN, 1999) and
SWAN ((NAKOS, 1990), (KRING, 1994) and (HUANG, 1997)), most of them are not
commercialized and their applicability is restricted to academic researches.

In offshore applications, in which most of the problems do not involve forward speed
vessels, there is a widespread use of BEM in frequency domain due to its great efficient
in providing the steady-state hydrodynamic loads and motions with low computational
efforts. The results provided by these codes are normally presented in the form of transfer
functions, or motions Response Amplitude Operators (RAOs), which, by their turn, also
carry the assumption that the floating system dynamics is linear. Nevertheless, for some
practical cases, this treatment brings severe limitations, such as for evaluating the motions
of an oil platform under the influence of loads induced by risers and mooring lines installed
on it. With the increase of water depth, the mass and damping of mooring lines and risers
become nontrivial and the platform motions can be appreciably affected by them (YANG
et al., 2012).

For the modelling of these systems it is common to use dynamic time domain simulators,
which consider not only the effects of waves, but also those originated by wind, current and
also from the dynamics of risers and mooring lines. Specifically to the linear wave problem,
most of these simulators use potential hydrodynamic coefficients previously calculated in
frequency domain and properly transposed to time domain by the use of the so-called
memory functions. However, one should realize that some inconsistencies may appear if
the body acquires large amplitudes of motion around its mean initial position, disobeying
the original linear assumption established for the calculation of the hydrodynamic loads.
Another issue emerges from problems involving more than one body, in which the transient
hydrodynamic interaction that results from the relative displacement of the bodies must
be taken into account. The offloading operation of a Floating Production Storage and
Offloading (FPSO) unit is an example of this type of problem, in which, due to the
proximity of the FPSO and the shuttle tanker, the wave effects induced by the vessels
have an important role in the hydrodynamic loads and, consequently, in the dynamics of
such a complex operation.

The use of time domain numerical methods then arises as an alternative approach to
these problems, since the hydrodynamic calculations may be coupled to the bodies equa-
tions of motion solved by the dynamic simulator. A first benefit of such a treatment
would be the possibility to handle the multi-body transient problems more suitably, even
when applying, in a first moment, a linear method. Moreover, in comparison to models
mathematically treated in frequency domain, this approach also allows addressing future
34

extensions to nonlinear descriptions of both the hydrodynamic and the dynamic problems
in a more straightforward manner. Such extensions may enable one, for example, to con-
sider the instantaneous wetted surfaces of the bodies and to deal, in a more satisfactory
way, with viscous damping effects that, as nonlinear phenomena, are commonly modeled
by a quadratic term involving the body velocities.

Therefore, considering the facts previously pointed out, we present in this text the de-
velopment of a numerical method and, subsequently, of a computational tool written in
MATLAB4 that solves the linear transient hydrodynamic problem of floating bodies un-
der the influence of free surface waves through the use of a linear time domain boundary
elements method based on Rankine sources. Independently of its own practical interest,
which has given favorable results to many problems regarding floating body dynamics,
the robust and accurate solution of this linear problem is a very useful first stage before
attempting to handle the complete non-linear problem. As will be presented ahead, the
numerical code is firstly tested through a set of verification cases involving bodies of sim-
ple (hemisphere and circular cylinders) and realistic geometries, such as FPSO units. For
these tests, the code is employed for solving the diffraction, radiation and free floating
problems in waves, the results being compared to analytical, numerical and experimental
data.

Once verified, the code is applied for the analysis of two complex multi-body problems
that are currently in the state-of-the-art concerning the computational hydrodynamic
modelling. In the first case, the performance of the code when applied to the seakeeping
problem of two ships in side-by-side configuration is investigated. The main challenge
of this problem is related to the occurrence of wave resonant effects in the gap between
the vessels, which cause all potential flow codes to have convergence problems, since they
are unable to model the viscous effects that are important for the correct modelling of
the flow between the hulls. In order to mitigate this problem, a damping lid method
is applied in the time-domain simulations and the convergence of the time series with
different damping levels is investigated. To the author’s knowledge, this is the first time
that certain issues concerning the numerical modelling of the resonant effects in time
domain are reported in the literature specialized in side-by-side operations. Regarding
the second case study, a new method to handle the multi-body problem with bodies
undergoing large relative displacements is presented. Due to the transient nature of the
problem, this is a case that cannot be properly analyzed by frequency domain codes,
since they only consider fixed meshes. In this regard, a numerical procedure that couples
the hydrodynamic solver with a re-meshing and a free surface interpolation algorithm is
presented. This new implementation enables the change of relative positions between the
4
http://www.mathworks.com/
 35

vessels along the simulations. The numerical results are compared to data obtained in
experimental tests designed and executed for verification of the code, and present a very
good agreement.

1.2 Text Outline

Bearing the developments of a time domain code in mind we organized this text as follows:

Chapter 2 presents a literature review including the main publications which were con-
sulted throughout the doctoral work.

Chapter 3 formulates the theoretical description of the problem involving floating bod-
ies and free surface waves. The fluid is assumed incompressible and non-viscous whereas
the flow is assumed irrotational. A set of nonlinear boundary conditions describing the
dynamics and kinematics of the domain surroundings are then presented and linearized
for providing less restrictive applications in terms of the body geometries and consump-
tion of computational time in comparison to numerical methods that deal with nonlinear
problems. Special attention is given for the determination of the velocity potential time
derivative and, consequently, the pressure field. As time-domain codes require a coupled
solution of the hydrodynamic and body dynamics problems in time, an imprecise predic-
tion of this quantity may give rise to numerical instabilities and, therefore, we decided for
the use of a second integral equation using the so-called acceleration potential.

In Chapter 4 we apply Green’s second identity and transform the Laplace’s field equa-
tions for the velocity and acceleration potentials in two integral equations. This procedure
allow us to establish a numerical model which evaluates the quantities only on the bound-
aries of the fluid domain and, therefore, reduces the problem dimension to two. A brief
discussion about the choice of singularities distribution over the surfaces and about the
kind of integral equations that result from this choice is also presented. Among a few alter-
natives, we have decided for a mixed source-dipole distribution, defining all the boundary
surfaces but the free surface as Neumann boundaries.

Chapter 5 presents the numerical algorithms used to deal with the present time domain
linear wave-body formulation. The solutions of the integral equations are calculated by
means of a low order boundary elements method that transforms the continuous equations
into a set of algebraic ones that may be described as two linear systems with full matrices.
By the use of this method, the boundary surfaces are discretized in a finite number of
plane quadrilateral/triangular panels over which the potential and its normal derivative
are assumed to have constant values. The time-marching scheme is performed by a Fourth
36

Order Runge-Kutta method (RK4) that integrates the body equations of motion and also
updates the boundary conditions.

Chapter 6 presents the hydrodynamic loads obtained by the solution of the so-called
diffraction and radiation problems for an hemisphere and a vertical circular cylinder.
Moreover, free floating simulations involving these geometries are also presented. The
accuracy of the results obtained is investigated by comparisons performed with data com-
puted by software WAMIT and, for the case of the hemisphere, also analytic solutions
given by Hulme (1982). The convergence of the numerical method is checked by re-
calculating the hydrodynamic loads with different panel mesh resolutions.

Chapter 7 tests the present numerical code for a more realistic situation, presenting
results of free floating FPSOs under the influence of incoming waves, being these regular or
irregular, with different incidence angles. Results are presented in the form of time-series of
motions, free surface wave patterns and motion RAOs, which are verified by comparisons
with motion RAOs calculated by WAMIT and results obtained experimentally in the
Hydrodynamic Calibrator of the TPN-USP (CH-TPN-USP).

Chapter 8 reports the performance of the method considering a multi-body problem

involving two ships in side-by-side arrangement, as, for example, in a FLNG offload-
ing operation. The chapter also describes the experimental campaign carried out in the
ETSIN-UPM, in Spain, from which data is here used for validation purposes of the present
method. Comparison between measurements and numerical results illustrates the limi-
tation of potential flow solvers concerning the modelling of this hydrodynamic problem.
Numerical wave resonance in the gap led to wave elevations and body motions much larger
than those observed during the tests. In addition, the time domain method also presented
convergence problems for frequencies associated to the gap resonance phenomenon. In or-
der to overcome these problems, an external damping factor was introduced in the time
domain simulations, bringing a significant improvement to the numerical convergence of
the method. Moreover, despite the simplicity of the damping model adopted, the results
showed that this technique was indeed able to improve the computational predictions,
leading to a closer agreement between the experiments and the numerical results.

Chapter 9 presents a new method for dealing with multi-body hydrodynamic interactions
of bodies undergoing large relative displacements, a problem that cannot be assessed
directly with frequency domain codes. This is done by extending the numerical method
with the possibility of including a generator of unstructured triangular panel meshes in
the time loop of the code so as to account for changes of the relative positions between the
bodies during the calculations. In addition, a specific higher order interpolation algorithm
designed to recover properly the solutions of a previous time-step was also implemented,
 37

enabling the calculations to progress with reasonable accuracy in time. The numerical
results are compared with data of experimental tests designed specifically for validation
purposes, presenting a very good agreement for most of the cases.

The final conclusions and suggestions for future works are presented in Chapter 10.
38
 39

Chapter 2

Literature Review

Most of the developments presented herein arise from basic fluid dynamic equations that
state the conservation laws of mass and linear momentum. Mathematical description of
this problem may result in different governing equations which depend on the established
assumptions concerning the fluid properties and also its motion. The description of water
wave motions, in which a great part of the present study is related, allows one to assume
the fluid as Newtonian and to describe the flow as incompressible by restricting its motions
to moderate Mach number (Ma < 0.3). Based on these hypotheses, mass conservation
is mathematically guaranteed by requiring that the divergence of the fluid velocity field
is zero whereas the momentum conservation is described by the Navier-Stokes equations.
Although some of these mathematical derivations are presented further ahead, we consider
them merely as simple presentations of the equations that form the basis of our develop-
ments and for a full comprehension of this topic we refer to the widely used textbooks of
Lamb (1945), Batchelor (1967) and Thomson (1968).

The success of a mathematical model of the physical reality depends on its quality to
predict the phenomena of interest with sufficient accuracy and minimum possible effort.
Although Navier-Stokes equations are supposed to be the most exact way to describe
the flow of a Newtonian fluid, their mathematical solution is extremely complex and,
for that reason, it is only applied in problems in which the viscous effects cannot be
neglected. Fortunately, simplifications of these equations may be performed by observing
that the influence of viscosity on the seakeeping prediction of floating systems, such as oil
platform units, is limited to a thin boundary layer near the hull, which does not influence
the dominant effects induced by the oscillatory motions and incoming waves. This is
guaranteed when the Keulegan-Carpenter number KC (see equation (2.1)) is small, i.e,
40

the oscillatory motion amplitude A of the body is small compared to its characteristic
dimension D. In the seakeeping context of floating bodies, however, this is not always
true, especially, near resonant frequencies, for which even waves of small amplitude may
lead to large body displacements. In this case, empirical external damping coefficients
must be imposed in order to emulate the viscous effects acting on the body. Apart from
this exception, in general viscous effects can still be neglected and by assuming that the
vorticity throughout the fluid is zero we may treat our problem under the view of the
potential flow theory.

VT Aω 2π
ω 2πA
KC = = = (2.1)
D D D

Despite of this important simplification, a set of nonlinear boundary conditions originated

from the free surface and the floating body surface still render the problem difficult to be
solved and, therefore, further approximations are usually applied. Following the classical
work of Stokes (1847), Stoker (1957) applies a perturbation technique which proposes the
expansion of the involved quantities (e.g. velocity potential, free surface elevation etc.) in
a power series, turning the original nonlinear problem into a sequence of linked linear ones,
which allows treating the problem in different orders of magnitude. The linear solution, in
particular, forms the basis for the study of all the other wave theories. Within this scope,
besides the publications aforementioned, the works developed by Sarpkaya and Isaacson
(1981) and Mei, Stiassnie and Yue (2005a) are recommendable readings.

The mathematical formulation of problems involving the motion of floating bodies can be
addressed following the same technique. Extensive mathematical derivations and discus-
sions for the linear, or first order problem, are presented by John (1949), John (1950) and
Wehausen (1971). The second order formulation may be found in, among several others,
Pinkster (1980) and Mei, Stiassnie and Yue (2005b).

Most of the developments achieved by these studies are based on the assumption of small
amplitude waves and body displacements around a mean position, which allows a major
simplification of the equations and even analytical solutions for problems involving very
simple geometries, as those presented in Hulme (1982) and Bhatta and Rahman (2003).
These solutions have a wide range of applications, but it eventually becomes necessary to
find out alternatives in order to deal with bodies with arbitrary geometries that enable
one to analyze the complex designs currently observed in modern offshore structures. In
this sense, the use of numerical methods becomes necessary, as for example, the boundary
elements method, also known as panel method. In this method, the problem is treated by
means of an integral equation which needs to be solved for obtaining the velocity potential
induced by a continuous distribution of singularities over the domain surfaces.
 41

Hess and Smith (1967) are considered pioneers in the development of panel methods,
demonstrating its applicability for obtaining the fluid flow around three-dimensional body
geometries in infinite fluid. In their method, the body surface is discretized in a finite
number of plane quadrilateral panels with source strength considered constant over each
one of them, an assumption that simplifies the integration of the singularities considerably.
Under these assumptions the method is also known as a Low-Order Panel Method.

For applications in wave-body formulations the method can be segregated into two differ-
ent categories based on the type of singularity employed in the integral equations. The
first category makes use of the so-called free surface Green function, which automati-
cally satisfies both the radiation and the linearized free surface conditions. Wehausen
and Laitone (1960) present some of these functions derived to satisfy different types of
boundary conditions, such as the ones for linear free surface waves in finite water depth.
This approach is commonly used for linear and weakly-nonlinear wave-body theories in
which the time dependence can be removed by assuming that the solution is harmonic in
time and, therefore, the solution may be addressed by a frequency domain approach (see
for instance Newman (1992), Yang, Noblesse and Löhner (2004) and Lee and Newman
(2005) for an overview of marine hydrodynamics applications).

Firstly presented by Gadd (1975) and Dawson (1977), the second category of the method
applies the simple Rankine sources as the basic singularity and, therefore, not only the
body surface but also the free surface must be discretized. Despite this inconvenient,
Rankine source methods are much easier to handle and also accept a more flexible choice
of free surface conditions which allow, for example, the inclusion of nonlinear terms that
have to be necessarily neglected in Green function formulations and a more suitable treat-
ment for problems involving bodies with forward speed, as discussed in Nakos (1990) and
Sclavounos and Nakos (1993).

Boundary elements methods in frequency domain are widely used for solving linear prob-
lems due to its great efficiency in providing solutions with relatively low computational
efforts. This may be explained, since both the radiation and free surface conditions are
automatically satisfied by the Green function and then only the wet body surface must
be discretized. On the other hand, Green functions and its derivatives are very difficult
to compute and may only be determined for linear free surface conditions. Besides that,
the method loses efficiency when higher order effects have to be considered since the free
surface needs now to be discretized and the computational efforts associated to the calcu-
lation of second-order effects are significantly increased. In addition, as only steady-state
solutions are assessed, frequency domain approaches lead to some limitations concerning
multi-body simulations with large relative displacements, as pointed out by Tannuri et al.
(2004), Queiroz Filho and Tannuri (2009) and Bunnik (2014).
42

Bearing these constraints in mind, a method of solution in time domain using Rankine
sources as the basic singularities appears as an interesting alternative, in which the inclu-
sion of nonlinear terms and the solution of transient problems are handled more suitably.
For these reasons, Kring (1994) extended the work of Nakos (1990) and developed a lin-
ear time domain boundary elements method for the evaluation of the transient forces and
motions of floating bodies with, and without, forward speed. Giving continuity to these
works, Huang (1997) applied the so-called weak-scattering-method that does not restrict
the amplitudes of the incoming waves and the body motions to small values by assuming
that the diffracted and radiated waves from a slender body may be neglected.

An interesting approach is observed in Kim, Kring and Sclavounos (1997), in which the
linear and the second order problems are solved simultaneously in time domain by forcing
the problem with a group of waves with different frequency components. This led to
the straightforward calculation of the hydrodynamic forces in time domain, overcoming
the time consuming procedure required for the computation of the Quadratic Transfer
Function (QTF) matrices in frequency domain.

The different approaches available for nonlinear extensions is reviewed in the work of
Singh and Sen (2007). Apart from the simplifications, or not, of the boundary conditions
equations, the inclusion of nonlinear terms are basically related to corrections of the body
and free surface positions in time. In this sense, in the linear approach the mesh grid
remains fixed during the simulations whereas for the nonlinear methods a re-gridding
procedure becomes necessary.

Independently of the method or approach, time domain simulations require an accurate

prediction of the pressure field in order to generate a consistent and stable numerical algo-
rithm. A main concern is the proper evaluation of the velocity potential time derivative,
which in the original problem must be determined by a backward finite difference scheme
in time. This procedure may give rise to numerical instabilities. One way to overcome this
problem is the computation of the so-called acceleration potential, widely investigated by
Tanizawa (2000). Within different schemes, four of them are considered to be the most
consistent: (1) iterative method ((CAO; R.; SCHULTZ, 1994)), (2) modal decomposition
method ((COINTE et al., 1991) and (KOO; KIM, 2004)), (3) indirect method ((TAY-
LOR; WU, 1996)) and (4) implicit boundary condition method ((VAN DAALEN, 1993)
and (TANIZAWA, 1995)).

The application of a numerical method capable of satisfying the radiation condition and,
therefore, of avoiding that waves reflected at the numerical boundaries have an impact on
the solution is also a major concern. In practice, this condition is imposed by absorption
or dissipation schemes widely discussed in the literature, as reviewed by Romate (1992).
 43

Orlanski (1976) proposed a numerical scheme to ensure the absorption of regular waves.
This condition, however, is limited to cases in which the angular frequency is known, or
in the presence of very long waves, as observed by Clément (1996). In this sense, the use
of numerical damping zones, introduced by Israeli and Orszag (1981), is less restrictive
since it can dissipate waves within a wide frequency range. This is performed by including
extra damping terms in the free surface boundary conditions applied on a limited portion
of this surface. In practice, there are various differential schemes that may be applied
(KIM; KOO; HONG, 2014) and the relatively easy way for implementing them together
with the benefits aforementioned have led to a widespread use of this technique, which is
demonstrated by the large number of works related to different approaches as, for example,
Prins (1995), Bunnik (1999), Boo (2002), Shao (2010) and Zhen et al. (2010).

Fully nonlinear simulations by adopting a mixed Eulerian-Lagrangian approach is another

topic with an increasing number of publications. First introduced by Longuet-Higgins and
Cokelet (1976), this complex method deals with the nonlinear boundary conditions by re-
gridding the moving surfaces in time. Within three and two-dimensional developments,
detailed descriptions of this method may be found in Romate (1990), Cointe (1990), van
Daalen (1993), Beck, Cao and Lee (1994), Ferrant (1997), Liu, Xue and Yue (2001a), Liu,
Xue and Yue (2001b) and Bai and Taylor (2009). However, it is known that two main
problems arise in the implementation of this method. According to Beck and Reed (2000),
for a stable free surface time marching scheme, the method requires the use of smoothing
techniques and the inclusion of artificial damping coefficients at the free surface points.
Moreover, wave breaking phenomenon may naturally occur due to the fully nonlinear
description of the free surface, and this, unfortunately, often causes the numerical solutions
to break down.
44
 45

Chapter 3

Theoretical Background

3.1 Problem Definition

The problem of a floating body interacting with free surface gravity waves is here consid-
ered. In this situation, incoming waves that propagate on the free surface are disturbed
by the floating bodies inducing unsteady forces and moments due to the generation of
pressure changes in the fluid. Floating bodies then respond by acquiring translational
and rotational motions, which also force the fluid to move. Mathematical description of
this problem may result in different governing equations depending on the assumptions
on the motions of the body and the fluid. Moreover, it is usual to separate the equations
that describe the behavior of the fluid and the floating body, which are then linked again
by appropriate equations and matching conditions that define the wave-body formula-
tions. The following items present the theoretical formulations and their correspondent
assumptions and simplifications on which the solutions of our problem will be based.

3.2 The Initial Boundary Value Problem

Let us define three mutually perpendicular unit vectors ~e1 , ~e2 , and ~e3 which form the basis
of two coordinate systems here denoted by:

• The first system of coordinate axes O − (x, y, z) is a right-handed earth-bound axes

cartesian system with origin O, with x and y axes in the mean free surface and
positive z axis pointing upwards and out of the fluid. A point in space has position
46

or displacement vector ~δ = (x, y, z).

• The second system CoG − (X, Y, Z) is a body-fixed right-handed axes cartesian

system with origin at the body center of gravity CoG, with positive X axis in
the longitudinal direction (body bow direction) and the positive Z axis pointing
upwards. The hull surface is defined exclusively in this coordinate system, being a
point on the body surface described by the vector ~r = (X, Y, Z). The orientation
of a surface element is defined by a normal vector pointing inward of the body
~n = (nx , ny , nz ).

Figure 3.1 illustrates a body floating somewhere in the sea in which incoming gravity
waves propagate on a free surface SF S with angular frequency ω, length λ and in a di-
rection which makes an angle θ with the X-axis of the established body-bound reference
frame.

S
z
y Z
x Y
X
SB
S FS

Figure 3.1: Definition of the coordinate systems

Focusing our analysis in the fluid, we consider an arbitrary control volume Ω delimited
by a surface ∂Ω with normal vector ~n = (nx , ny , nz ) pointing outwards the fluid domain.
The flow velocity field is then defined by ~v = (u, v, w) in which u(x, y, z, t), v(x, y, z, t)
and w(x, y, z, t) are the scalar components of the velocity field at time t, at a point with
spatial coordinates (x, y, z).

We consider the flow to be irrotational and incompressible whereas the fluid is assumed
inviscid and homogeneous, allowing the velocity field to be defined by the gradient of the
potential scalar field or velocity potential Φ:

~v = ∇Φ (3.1)
 47

Under these assumptions, mass conservation is stated by the Laplace´s equation (3.2)
whereas the conservation of linear momentum is expressed by Bernoulli’s equation (3.3):

∇2 Φ = 0 (3.2)

∂Φ 1 p − p0
+ (∇Φ · ∇Φ) + gz = − + C(t) (3.3)
∂t 2 ρ

where p0 is the atmospheric pressure and C(t) is a constant parameter dependent on time,
normally omitted since it can be easily absorbed into the velocity potential by redefining
the latter without any loss of generality of the solution.

Equations (3.2) and (3.3) compose the governing equations which describe the motion of
an homogeneous and inviscid fluid under the assumptions of incompressibility and that
the vorticity throughout the fluid is zero (i.e. irrotational flow). Unique solutions for Φ
must then be calculated, particularizing the solutions of Laplace’s equation (3.2) by the
imposition of boundary conditions which, in simple words, are necessary to match the
dynamics and kinematics of the fluid with the physical boundaries of the fluid domain Ω,
such as, the wet surface of the floating body SB (t), the sea bottom surface SBO and the
free-surface SF S (t).

The free surface SF S (t), its position being not known a priori, demands the description of
a kinematic and a dynamic boundary condition. The kinematic condition states that the
velocities of the fluid and the free-surface boundary must be equal, whereas the dynamic
one imposes that, neglecting surface tensions, the pressure on the free surface must equal
the atmospheric pressure.

Thus, the dynamic boundary equation may be determined through the straightforward use
of Bernoulli’s equation (3.3), applying it at the exact free surface elevation, here denoted
by the single-valued function η(x, y, t):

∂Φ 1
+ (∇Φ · ∇Φ) + gz = 0 on z = η(x, y, t) (3.4)
∂t 2

As a consequence of the velocities compatibility on the free surface, fluid particles that
are at the free surface will remain attached to the free surface. Furthermore, considering
that these particles are vertically described by a coordinate z, a mathematical function
F (x, y, z, t) = 0 which defines the free surface can be described as F (x, y, z, t) = z −
η(x, y, t) = 0. Therefore, the kinematic condition may be expressed assuming that the
48

material derivative of the quantity z − η(x, y, t) vanishes on the free surface.

D ∂Φ ∂η ∂Φ ∂η ∂Φ ∂η
(z − η) = − − − = 0 on z = η(x, y, t) (3.5)
Dt ∂z ∂t ∂x ∂x ∂y ∂y

Description of the free surface by the function η(x, y, t) excludes the possibility of studying
the phenomena of overturning waves, since η(x, y, t) may only assume one single value
for the pair (x, y) at a given instant t. Despite such simplification, this approach is
widely used and is sufficient for the evaluation of several hydrodynamic problems, such
as the calculation of first and second order forces and motions of a floating vessel in the
sea, as may be observed, for example, in Newman (1977), Pinkster (1980) and Faltinsen
(1990). Other possibilities are also found in literature to deal with the case of very
steep waves, in which a Lagrangian or mixed Eulerian-Lagrangian descriptions is used
(see for instance Longuet-Higgins and Cokelet (1976)), but due to the difficult and time-
consuming simulations associated with the numerical schemes applied, this alternative is
still restricted to very simple applications. Although modern computers continue to push
the limits of practicability, these drawbacks are still not acceptable for many engineering
applications.

The boundary conditions associated to the floating body wet surface SB (t) and sea bottom
surface SBO are mathematically described by Neumann conditions which impose that the
fluid particles may not penetrate these surfaces. For the sea bottom and other time-
independent surfaces that may exist in the problem, this is represented by the well-known
zero-flux condition (3.6).

∂Φ
∇Φ · ~n = = 0 on SBO (3.6)
∂n

The conditions on the wet surfaces of floating bodies and other time-dependent surfaces
are treated in a very similar manner as the bottom surface. Nevertheless, the motions of
these boundaries influence the movements of the fluid at their surroundings, requiring a
compatibility condition between the fluid and the surfaces to guarantee an impermeability
condition, as presented in equation (3.7):

∂Φ ∂~δ(t)
∇Φ · ~n(t) = = · ~n(t) on SB (t) (3.7)
∂n ∂t
 49

where the time-dependent displacement ~δ of a point of the floating body relative to the
earth-bound system of axes is defined as:

~δ(t) = ξ~T (t) + ξ~R (t) × ~r (3.8)

where ξ~T (t) = (ξ1 , ξ2 , ξ3 ) and ξ~R (t) = (ξ4 , ξ5 , ξ6 ) are formed by the translation and rotation
of the body and ~r is the position vector of the point on the body in the body-bound axes
system, measured with relation to the center of gravity.

Notice that neither the surface point displacement ~δ(t) nor the surface normal vector
~n(t) are known in advance for problems involving free floating bodies and, therefore,
an additional set of equations which describe the motions of a rigid body need to be
included. Supposing that the velocity potential Φ is determined, the pressure field is
calculated through the use of equation (3.3) and, consequently, both the hydrodynamic
forces and moments acting on the body are obtained by pressure integration, as presented
in equations (3.9) and (3.10):

ZZ
F~ = p ~n dS (3.9)
SB (t)

ZZ
~ =
M p (~r × ~n) dS (3.10)
SB (t)

Applying Newton’s second law and supposing small angular displacements, which is con-
sistent with the formulation that will be presented ahead, the motion equations may be
written as:

d2 ξ~ F~
 
M 2 = ~ (3.11)
dt M

where M is the matrix of mass/inertia and ξ~ = (ξ1 , ξ2 , ξ3 , ξ4 , ξ5 , ξ6 ) is the vector containing

the motion components in six degrees of freedom (D.O.F.) of a rigid body that correspond
to surge, sway, heave, roll, pitch and yaw, respectively.

The last, but not less important, boundary condition is the radiation condition, which
in the time domain approach must enforce that the waves generated by the bodies are
outgoing waves only and do not reflect somewhere allowing these waves to interfere with
the body motions once again. In practice, this condition is imposed by absorbing or
dissipation schemes widely discussed in literature, being the numerical beach zone an
50

efficient, and maybe the most simple alternative, to be applied. Details on how this
condition is tackled in this work can be found in section 3.3.

In order to close this Initial Boundary Value Problem (IBVP), an initial condition must
be imposed at the free surface so as to determine the subsequent fluid motions. As
demonstrated by Stoker (1957), for flows starting from the rest we may set the velocity
potential at the initial instant t = 0 s as:

Φ = 0 on t = 0 s (3.12)

Summarizing, the nonlinear wave-body formulation presented is formed by an elliptic field

equation represented by Laplace’s equation with correspondent boundary conditions that
particularize its solutions to the one we are interested in. Although Laplace’s equation is
a linear homogeneous partial differential equation, the boundary conditions have several
sources of nonlinearities as, in some cases, they must be applied in time dependent sur-
faces not known in advance, conditions that make the problem very complex to be solved.
Despite this complexity, numerical algorithms that deal with this fully nonlinear problem
are becoming popular in recent years, thanks to the fast increase in computer processing
capacity. However, most of them are still limited to two-dimensional studies or very sim-
ple three-dimensional geometries and consequently are not adequate for simulating the
behavior of real ship vessels, oil platforms or the hydrodynamic interactions involved in
multi-body arrangements. Therefore, as a first step we have decided to linearize the for-
mulation and to develop a linear computational code based on the simplified formulation
presented in the remainder of this chapter.

3.3 The Linear Formulation

Linearization begins by assuming that the incoming gravity waves addressed herewith
have small amplitude AI and steepness  = kAI , in which k is the wave number.

In order to linearize the problem, we apply a classical perturbation technique using ex-
pansions in Stokes’s series, as described, for example, by Stoker (1957). Thus, to obtain a
linear formulation, we perturb all the variables using the small parameter  and substitute
them in the set of nonlinear equations already presented. With this approach, we may
identify and collect terms of order up to  for a linear problem and consequently higher
orders terms for the other problems.

As already presented, free surface condition is represented by two boundary conditions

 51

which are nonlinear equations for basically two reasons: first, there are products involving
not known variables as the velocity potential, free surface elevation and their derivatives.
Second, both conditions must be applied on a surface which is not known in advance and
is also part of the solution. The velocity potential Φ and the wave elevation η are then
perturbed using the small wave steepness parameter , as follows:

Φ = Φ̄ + Φ(1) + 2 Φ(2) + O(3 ) (3.13)

η = η̄ + η (1) + 2 η (2) + O(3 ) (3.14)

hereafter, the horizontal bar (¯) over the variables account for mean values, in time sense,
and the superscripts (i) denote the different orders of magnitude.

We now substitute the perturbed quantities (3.13) and (3.14) in the free surface conditions
(3.4) and (3.5), and group terms of the same order, obtaining:

∂Φ(1)
 
∂ Φ̄ 1
+ ∇Φ̄ · ∇Φ̄ + g η̄ +  + ∇Φ̄ · ∇Φ(1) + gη (1)
∂t 2 ∂t
+ O(2 ) = 0 on z = η̄ + η (1) + O(2 ) (3.15)

∂ Φ̄ ∂ η̄ ∂ Φ̄ ∂ η̄ ∂ Φ̄ ∂ η̄
− − −
∂z ∂t ∂x ∂x ∂y ∂y
 (1) (1)
∂ Φ̄ ∂η (1) ∂Φ(1) ∂ η̄ ∂ Φ̄ ∂η (1) ∂Φ(1) ∂ η̄

∂Φ ∂η
+ − − − − −
∂z ∂t ∂x ∂x ∂x ∂x ∂y ∂y ∂y ∂y
+ O(2 ) = 0 on z = η̄ + η (1) + O(2 ) (3.16)

Collecting terms of order up to  and dropping out the time-independent mean values,
since they are only important for problems involving bodies with forward speed, we find:

∂Φ(1)
+ gη (1) = 0 on z = η (1) (3.17)
∂t

∂Φ(1) ∂η (1)
− = 0 on z = η (1) (3.18)
∂z ∂t

One should notice that equations (3.17) and (3.18) are still being applied in unknown
positions. In order to overcome this problem, we may expand these equations in Taylor’s
52

series around the undisturbed free surface position z = 0.

∂Φ(1)
   (1) 
(1) ∂Φ (1)
+ gη |z=η = + gη |z=0
∂t ∂t
 (1) 
(1) ∂ ∂Φ (1)
+η + gη |z=0 = 0 on z = 0 (3.19)
∂z ∂t

∂Φ(1) ∂η (1)
 (1)
∂η (1)
  
∂Φ
− |z=η = − |z=0
∂z ∂t ∂z ∂t
 (1)
∂η (1)

(1) ∂ ∂Φ
+η − |z=0 = 0 on z = 0 (3.20)
∂z ∂z ∂t

Finally, if we keep terms of order up to , we find the linear dynamic (3.21) and kinematic
(3.22) free surface conditions:

∂Φ(1)
+ gη (1) = 0 on z = 0 (3.21)
∂t

∂Φ(1) ∂η (1)
− = 0 on z = 0 (3.22)
∂z ∂t

An analytic solution may be found for the boundary value problem defined by the Laplace’s
equation with the set of boundary conditions formed by the kinematic and dynamic bound-
ary conditions, and zero flux condition at the bottom surface. For example, for a constant
water depth h, the corresponding incoming regular wave field potential is defined by the
following expression:

AI g cosh k(h + z)
φI = cos(kx − ωt) on z ≤ 0 (3.23)
ω cosh kh

The linear wave dispersion relation is given by:

ω 2 = kg tanh kh (3.24)

where ω is the incoming wave angular frequency.

This solution is very useful when we are dealing with a time-domain description of the
problem because we may decompose the total potential Φ in a sum of a first order disturbed
wave field φ(1) and the incoming regular wave field φI , thus avoiding then the necessity
of modelling a numerical wave maker device to simulate the later. Since φI is a known
 53

solution we only need to include it in the zero-flux conditions accordingly.

From now on the variable of the problem Φ will be changed by φ(1) redefining also the set
of boundary conditions that must be imposed. As φI satisfies the free surface conditions,
these will be described by simply changing the variable Φ by φ(1) . For the other conditions,
however, we will need to include the incoming wave potential as will be described next.

From the previous section 3.2, it may be observed that the set of equations to be solved
are still nonlinear since the impermeability condition for moving bodies (3.7) and the
equations for forces (3.9) and moments (3.10) are also applied in a time dependent sur-
face SB (t). In order to linearize these boundary conditions, a procedure similar to the one
applied to the free surface conditions can be used by expanding the variables in Stokes’s
series and then applying Taylor’s series around a mean position S̄B , assuming that the
body presents only small relative displacements around it. Beginning with the imperme-
ability condition (3.7), we define expressions for the perturbed displacement vector ~δ(t)
and the normal vector ~n(t), as follows:

~δ = ~δ¯ + ~δ(1) + 2~δ(2) + O(3 ) (3.25)

¯ + ~n(1) + 2~n(2) + O(3 )

~n = ~n (3.26)

Substituting expressions (3.25) and (3.26) in equation (3.7) and collecting terms up to
order , results in:

~(1)
¯ = ∂ δ · ~n
∇φ(1) · ~n ¯ − ∇φI · ~n
¯ on SB (t) (3.27)
∂t

Applying the Taylor expansion around S̄B , we obtain the linear boundary condition:

~(1)
¯ = ∂ δ · ~n
∇φ(1) · ~n ¯ − ∇φI · ~n
¯ on S̄B (3.28)
∂t

To obtain the first order forces and moments on the floating body, the integration of
the pressure field p must also be performed on the mean surface S̄B . Moreover, since
the pressure is determined by the nonlinear Bernoulli’s equation (3.3), it also needs to
be linearized by using the expansion technique in order to keep the formulation coherent
with the established linear formulation. Following the same procedures presented before,
54

the next equations are then obtained:

∂φ(1) ∂φI
 
(1)
p = −ρ + (3.29)
∂t ∂t

∂φ(1) ∂φI
ZZ  
F~ (1) = −ρ + ¯ dS̄B
~n (3.30)
∂t ∂t
S̄B

∂φ(1) ∂φI
ZZ  
~ (1) =
M −ρ + (~r¯ × ~n
¯ ) dS̄B (3.31)
∂t ∂t
S̄B

The first order forces and moments (F~ (1) and M~ (1) ) will cause the floating body to present
a first order rotational and translational oscillatory motion. This motion is described by
Newton’s second law:

∂ 2 ξ~(1) ∂ ξ~(1) F~ (1)

 
M + C + Kξ~(1) = ~ (1) (3.32)
∂t2 ∂t M

(1) (1) (1) (1) (1) (1)

where ξ~(1) = (ξ1 , ξ2 , ξ3 , ξ4 , ξ5 , ξ6 ) is a vector containing the first-order translational
and rotational motions, and M is the mass matrix containing the mass of the floating body
and the moments of inertia.

One should realize that the static term of the pressure was neglected in the previous equa-
tions, assuming that the body is initially at a state of equilibrium. However, if the body
is shortly displaced from its initial position, restoring forces and moments appear, forcing
the body to oscillate around the equilibrium position. This motion can be compared with
the motion of a mass connected to a spring, where restoring force on the mass is propor-
tional to the displacement of the mass. In the linear theory approach, the restoring forces
and moments may be shifted to the left hand side of equation (3.32) through the con-
struction of a matrix K composed by time-independent restoring coefficients, which can
be expressed in the form of surface integrals over the mean wetted surface S̄B , by applying
Gauss’s divergence theorem. The matrix of hydrostatic coefficients is then determined as:

ZZ
K(3, 3) = ρg nz dS̄B (3.33)
S̄B
 55

ZZ
K(3, 4) = ρg Y nz dS̄B (3.34)
S̄B

ZZ
K(3, 5) = −ρg Xnz dS̄B (3.35)
S̄B

ZZ
K(4, 4) = ρg Y 2 nz dS̄B + ρg∀Zb (3.36)
S̄B

ZZ
K(4, 5) = −ρg XY nz dS̄B (3.37)
S̄B

K(4, 6) = −ρg∀Xb (3.38)

ZZ
K(5, 5) = ρg X 2 nz dS̄B + ρg∀Zb (3.39)
S̄B

K(5, 6) = ρg∀Yb (3.40)

where ∀ is the body displacement and (Xb , Yb , Zb ) are the body coordinates of the center
of buoyancy, which are described by:

ZZ ZZ ZZ
∀=− nx XdS̄B = − ny Y dS̄B = − nz ZdS̄B (3.41)
S̄B S̄B S̄B

ZZ
1
Xb = − nx X 2 dS̄B (3.42)
2∀
S̄
ZBZ
1
Yb = − ny Y 2 dS̄B (3.43)
2∀
S̄
ZBZ
1
Zb = − nz Z 2 dS̄B (3.44)
2∀
S̄B

One last consideration, is related to the lack of external damping coefficients which are
56

sometimes necessary since the potential flow theory does not account for viscous effects.
This is the case, for example, of the roll motion of a ship, which is usually overestimated
when applying the potential flow theory, since this D.O.F. is associated to very low energy
dissipation by wave radiation. On the other hand, for heave and pitch D.O.F. the damping
is mostly dominated by wave radiation effects and, therefore, usually there is no need to
add an external coefficient. In order to account for the modelling of both situations, an
external matrix of linear damping coefficients C will be also considered, as can be observed
in equation (3.32).

In order to obtain a unique solution we still need to consider the inclusion of a radiation
condition to ensure that the waves diffracted and radiated by the body do not reflect at
the numerical boundaries back to the body position. Indeed, since the computer memory
is finite, we are obliged to truncate the free surface somewhere and impose a numerical
scheme to dissipate or absorb the waves. Following Zhen et al. (2010), this condition is
guaranteed in our numerical model by the inclusion at these boundaries of a damping zone,
also known as numerical beach, which consists in the inclusion of an energy dissipation
term in the free surface boundary conditions, as follows:

∂η (1) ∂φ(1) p
= − ν(x, y)η (1) at z = 0 and x2 + y 2 > Ldz (3.45)
∂t ∂z
∂φ(1) p
= −gη (1) − ν(x, y)φ(1) at z = 0 and x2 + y 2 > Ldz (3.46)
∂t

Here Ldz stands for the distance from the global coordinate origin to the beginning of
the damping region and ν(x, y) is a function that defines the dissipation on this region,
described by:

p !2
x2 + y 2 − Ldz
ν(x, y) = aω (3.47)
bλ

where a defines the intensity of dissipation and b the length of the damping zone. For
practical purposes, these values must be tuned by preliminary tests in order to avoid the
occurrence of reflected waves which may spoil the solution. An investigation is this sense
is presented in Appendix A, where a study on the effects of varying the damping zone
parameters is presented. In general, it is observed that the damping zone length presents
a significant influence on the performance of the damping zone, whereas the intensity
parameter simply cannot be set with values that are too large, since it leads the method
to be unstable. In most of the cases, the combination of a damping zone length of two
wave lengths b = 2.0 and a dissipation strength of a = 1.0 leads to a good performance,
for all the practical wave frequencies.
 57

The linear wave-body formulation of the problem and the necessary assumptions to guar-
antee the applicability of this theory were presented in this section. In a few words, the
constructed Boundary Value Problem (BVP) is formed by the Laplace’s equation for the
velocity field, added by the linear boundary conditions applied at the mean surface po-
sitions of the domain. Once the velocity potential is known, pressure field around the
body surface may be calculated and integrated to find the forces and moments which are
necessary for the time progressing of the equations of motion.

This procedure is satisfactory and apparently easy to implement, except for the fact that,
by now, we do not have an exact equation for the potential time-derivative ∂φ(1) /∂t, which
would then require the use of numerical approximations for this quantity. However, it is
a well-known fact (CAO; R.; SCHULTZ, 1994) that errors on these numerical approxi-
mations for ∂φ(1) /∂t may lead to unstable schemes for the time-marching of the set of
equations and, therefore, an alternative procedure must be chosen, as will be presented
in the next section.

3.3.1 The Acceleration Potential

In a time domain simulation, it is essential to solve the equations of the fluid and floating
body motions simultaneously. This is because we need to guarantee a dynamic equilibrium
of forces between the fluid and the floating body at all times. With this in mind, an
accurate calculation of the pressure field (3.29) around the floating body is of utmost
importance.

Different numerical approximations to deal with this problem are found in the literature.
Backward difference schemes is the simplest way to follow. This method is sufficient
for studies involving fixed bodies or under prescribed motions, in which the equations of
motions do not need to be solved and the hydrodynamic forces may be post-processed.
On the other hand, free floating simulations also require the evaluation of the equations
of motions and poor estimatives of the pressure variation by finite difference schemes
normally give rise to numerical instabilities (see for instance van Daalen (1993)).

According to Koo and Kim (2004), the acceleration potential method is known to be
the most accurate and consistent way to solve ∂φ/∂t. The main idea of this method is
to set up a new IBVP by taking advantage of the facts that a linear expression for the
acceleration potential can be found and that it also satisfies Laplace’s equation. Thus, by
solving the two IBVPs for φ and ∂φ/∂t at each time-step, a simultaneous treatment of
the hydrodynamic and body dynamic equations may be conducted.

For completeness, boundary conditions must be defined for new the IBVP which require
58

a complicated treatment of the free floating body conditions. Tanizawa (2000) presents a
general review of the existing methods used to handle this problem, electing the following
ones to be the most consistent: (1) iterative method, (CAO; R.; SCHULTZ, 1994); (2)
modal decomposition method, (COINTE et al., 1991) and (KOO; KIM, 2004); (3) indirect
method, (TAYLOR; WU, 1996); and (4) implicit boundary condition method, (VAN
DAALEN, 1993) and (TANIZAWA, 1995).

In this work, we have decided to employ the method presented by van Daalen (1993) and
Tanizawa (1995) in which the equations of motions and the IBVP for the acceleration
potential are combined. As will be shown, the acceleration of the body center of gravity
may be expressed in terms of the acceleration potential ∂φ/∂t at the wet mean surface
S̄B , avoiding iterations for the matching of both values.

The acceleration potential is defined from the material derivative of fluid velocity:

∂~v
~a = + (~v · ∇)~v (3.48)
∂t
where ~a is the fluid flow acceleration. Denoting the velocity ~v by ∇φ, equation (3.48) is
re-written as:

   
∂∇φ ∂φ 1 ∂φ 1
~a = + (∇φ · ∇)∇φ = ∇ +∇ (∇φ)2 =∇ + (∇φ)2 (3.49)
∂t ∂t 2 ∂t 2

Hence, an acceleration potential Ψ may be defined as:

∂φ 1
Ψ= + (∇φ)2 (3.50)
∂t 2
and the fluid acceleration expressed by:

~a = ∇Ψ (3.51)

In order to maintain the formulation consistent with the linear theory, we combine the
perturbed velocity potential (3.13) and the perturbed acceleration potential (3.52) with
equation (3.50). Collecting terms of order up to , we find the linear acceleration equation
(3.53) .

Ψ = Ψ̄ + Ψ(1) + 2 Ψ(2) + O(3 ) (3.52)

 59

∂φ(1)
Ψ(1) = (3.53)
∂t

Since the acceleration potential Ψ is being analyzed in terms of the linear theory, it
satisfies Laplace’s equation in the whole fluid domain. Therefore, a new IBVP may be
written for the time-derivative of the velocity potential, or linear acceleration potential
Ψ(1) . With this intention, it is necessary to describe the appropriate boundary conditions
to complete the set of equations. This may be performed by applying the time-derivative
operator ∂/∂t in all the linear boundary conditions for the velocity potential. The IBVP
for the linear acceleration potential, therefore, is formed by the Laplace’s equation:

∂φ(1)
 
2
∇ = 0 in fluid domain (3.54)
∂t

and the following boundary conditions:

1. Free-surface condition:

∂φ(1)
= −gη (1) on z = 0 (3.55)
∂t
where η (1) is solved by the linear free surface dynamic condition (3.21)

2. Zero-flux condition for fixed surfaces:

∂φ(1)
   
¯ = −∇ ∂φI ¯ on SBO
∇ · ~n · ~n (3.56)
∂t ∂t
3. Zero-flux condition for moving surfaces:

2~(1)
∂φ(1)
   
∇ ¯ = ∂ δ · ~n
· ~n ¯−∇ ∂φI ¯ on S̄B
· ~n (3.57)
∂t ∂t2 ∂t

The acceleration on the body surface ∂ 2~δ(1) /∂t2 may be related to the acceleration of the
center of gravity by applying a straightforward double time differentiation of equation
(3.8). The velocities and accelerations of a point on the body surface may be written in
terms of the body center of gravity position, as presented in (3.58) and in (3.59).

(1) (1)
∂~δ (1) ∂ ξ~ T ∂ ξ~ R
= + × ~r¯ (3.58)
∂t ∂t ∂t

!
(1) (1) ~ (1) (1)
∂ 2~δ (1) ∂ 2 ξ~ T ∂ 2 ξ~ R ¯ + ∂ξ R × ∂ ξ~ R
= + × ~
r × ~r¯ (3.59)
∂t2 ∂t2 ∂t2 ∂t ∂t
60

Equation (3.59) is nonlinear as there is a double cross product of the angular velocity
(1)
vector ∂ ξ~ R /∂t. Once more, an expansion in Stokes’s series justifies the linearization of
this term, resulting in the linear expression:

(1) (1)
∂ 2~δ (1) ∂ 2 ξ~ T ∂ 2 ξ~ R
= + × ~r¯ (3.60)
∂t2 ∂t2 ∂t2

The boundary condition (3.57) may then be re-written as:

2~ (1) (1)
∂φ(1) ∂ 2 ξ~ R
   
∇ ¯ = ∂ ξ T · ~n +
· ~n ¯
× ~r · ~n − ∇
∂φI ¯
· ~n
∂t ∂t2 ∂t2 ∂t
(1) (1)
∂ 2 ξ~ T ∂ 2 ξ~ R ¯
 
∂φI ¯
= · ~n + · ~r × ~n − ∇ · ~n (3.61)
∂t2 ∂t2 ∂t

The body angular acceleration and the acceleration of the center of gravity may be isolated
in the equation of motion (3.32), as follows:

∂φ(1) ¯ dS
  RR  
" (1)
∂ 2 ξ~ T
# " (1)
∂ ξ~ T
# "
(1)
#
∂t
~n
ξ~ T
∂t2 = M−1  −C ∂t −K − ρ  RR S̄∂φ
B
(3.62)
   
(1) (1) (1) (1)
∂ 2 ξ~ R ∂ ξ~ R ξ~ R ¯ ¯
(~r × ~n) dS
 
∂t 2 ∂t ∂t
S̄B

Finalizing, the translational and rotational accelerations of equation (3.62) can be substi-
tuted in the boundary condition (3.61) in order to define a zero-flux condition for floating
bodies depending on the time-derivative of the velocity potential. As will be demonstrated
in the next chapters, these equations will allow us to define an integral equation in which
the time-derivative of the velocity potential and its normal derivative will be the only
variables to be determined and, therefore, the evaluation of the acceleration of the center
of gravity and body angular acceleration can be eliminated from the IBVP defined for the
acceleration potential ∂φ(1) /∂t.

From now on the superscript (1) denoting the first order quantities of the formulation will
be suppressed in order to simplify the notation.
 61

Chapter 4

Boundary Integral Equations

Chapter 3 presented the theoretical formulation which describes the hydrodynamic forces
and motions of floating bodies caused by the incidence of free surface gravity waves.
This formulation is summarized in two Initial Boundary Value Problems formed by the
Laplace’s equation for the velocity potential φ and for the acceleration potential ∂φ/∂t,
with the addition of appropriate boundary conditions which particularize the solutions to
the ones that we are interested in.

Analytical solutions for this problem are difficult to be found and are always related to
problems involving very simple geometries in which a technique of separation of variables
is used. The use of this technique, however, is only applicable if the boundary is a coor-
dinate surface of one of the special orthogonal coordinate systems for which the Laplace’s
equation can be separated into ordinary differential equations, which limits the type of
geometries to be analyzed, as observed by Hess and Smith (1967). Among other refer-
ences, examples are described in Hulme (1982) who presented an analytic solution for
the first order hydrodynamic forces that arise when a hemisphere undergoes prescribed
periodic oscillations in water of infinite depth and in Bhatta and Rahman (2003) in which
the diffraction and radiation problem for a circular cylinder in water of finite depth is
treated. Although restricted to very simple applications which are not normally encoun-
tered in real situations, these closed solutions may be very useful for the verification of
new numerical tools such as the one presented in this text.

Solutions of these problems for a more extensive set of geometries are found by numer-
ical approaches. In the context of solving Laplace’s equation, three of them receive
special attention: (1) Finite Difference Method (FDM), (2) Finite Elements Method
(FEM) (SERVÁN-CAMAS; GARCı́A-ESPINOSA, 2013) and Boundary Elements Method
62

(BEM). FDM and FEM, also known as field methods, discretize the differential equation
straightforwardly, generating sparse matrices with the higher values confined near the
main diagonal. Despite of the fact that these features facilitate reasonably the solutions
of the linear system, these field methods require the discretization of the entire fluid do-
main, even if we are only interested in values of the velocity or acceleration potential and
its derivatives at the boundaries. Moreover, three-dimensional grids may be quite difficult
to generate, especially near curved boundaries of, for example, the hull of a ship-shaped
vessel. Regarding the FDM, the method only allows for structured-grids, demanding a
greater control of the mesh that might not be adequate for some applications.

When dealing with the solution of BVPs for which Green functions may be obtained,
an alternative to these field methods is the use of the boundary elements method. By
applying Green’s second identity, Laplace’s equation may be re-written as an integral
equation, reducing the number of dimensions to two. Grid generation becomes easier
comparing to the field methods since only the boundary surfaces must be discretized,
enabling also a re-meshing procedure during the simulations if necessary, see for instance
van Daalen (1993) for 2D and Huang (1997) for 3D applications. A drawback of this
method, however, is related to the generation of full matrices, which in turn increase the
processing time to solve the linear systems and may also require more computer memory
than FDM and FEM.

As one may observe, advantages and disadvantages of the methods can be listed and none
of them is proved to be more efficient in relation to the others for a generic case. However,
BEM has been receiving special attention when the seakeeping problem is concerned or,
more generally, when the potential flow theory may be assumed. For this reason and
considering the other advantages described above, we chose a boundary elements method
to handle the Initial Boundary Value Problems previously presented. The next item of
this chapter, therefore, presents the mathematical procedures to transform the Laplace’s
equations for the velocity and acceleration potentials into integral ones. The mathematical
procedure for this transformation has already been presented by a vast number of authors,
as for example the textbook of Bertram (2000).

4.1 Integral Equations

Let us consider the three-dimensional fluid domain Ω0 delimited by a surface ∂Ω0 illustrated
in figure 4.1.
 63

Figure 4.1: Surface of integration for the Second Green’s Identity derivation

We define two scalar fields of class C 2 in Ω0 denoted by ϕ and ψ, which satisfy Laplace’s
equation ∇2 ϕ = ∇2 ψ = 0. Through the use of the divergence theorem, we derive the
Green’s second identity (4.1).
ZZ   ZZZ
∂ϕ ∂ψ 0
ψ −ϕ d∂Ω = ∇ · (ψ∇ϕ − ϕ∇ψ) dΩ0 =
∂n ∂n
∂Ω0 0
ZΩZ Z
= (ψ∇2 ϕ − ϕ∇2 ψ) dΩ0 = 0 (4.1)
Ω0

Hereafter we denote the scalar field ϕ by the Rankine’s source potential G(~γ ; ~x) (4.2), in
which ~x = (x, y, z) is the position of the Rankine’s source, ~γ = (ξ 0 , η 0 , ζ 0 ) is a field point
and r is the distance between ~x and ~γ .

1 1
G(~γ ; ~x) = =p (4.2)
r (x − ξ ) + (y − η 0 )2 + (z − ζ 0 )2
0 2

Thus, changing the scalar field ψ by the velocity potential φ, we obtain the following
identity:

ZZ    
∂ 1 1 ∂φ(~γ )
φ(~γ ) − d∂Ω0 = 0 (4.3)
∂n r r ∂n
∂Ω0

One should realize that the Rankine’s source (4.2) presents a singularity when r = 0 and
consequently the identity (4.3) is not valid when source and field points are coincident.
This problem can be overcome, by surrounding the source point by a small sphere of
radius a and surface ∂Ω0a . If the source point is positioned at the surface ∂Ω0 , we consider
a hemisphere with the same radius. Therefore, the sum ∂Ω0 + ∂Ω0a forms a closed surface
surrounding the fluid domain interior to ∂Ω0 but exterior to ∂Ω0a where the Rankine’s
64

source is now regular. Under these considerations, we can rewrite identity (4.3) in terms
of surface integrals on ∂Ω0 and ∂Ω0a , as follows:

ZZ    
∂ 1 1 ∂φ(~γ )
φ(~γ ) − d∂Ω0 = 0 or (4.4)
∂n r r ∂n
∂Ω0 +∂Ω0a

ZZ     ZZ    
∂ 1 1 ∂φ(~γ ) 0 ∂ 1 1 ∂φ
φ(~γ ) − d∂Ω = − φ(~γ ) − (~γ ) d∂Ω0 (4.5)
∂n r r ∂n ∂n r r ∂n
∂Ω0 ∂Ω0a

Now, we should evaluate the contribution of the right-hand term of expression (4.5) in
the limit of a → 0. For convenience, we choose a spherical coordinate system (a, α, β) in
which (0 ≤ a ≤ ∞), (0 ≤ α ≤ 2π) and (0 ≤ β ≤ π). The relations with the rectangular
coordinates are defined by:

x = a sin β cos α
y = a sin β sin α
z = a cos α (4.6)

Hence, the right-hand term of identity (4.5) may be replaced by:

Zπ Z2π    
∂ 1 1 ∂φ
− φ − a2 sin β dα dβ (4.7)
∂n r r ∂n r=a
0 0

If the source point is located inside ∂Ω0 , in the limit of a → 0 and assuming that for a
sufficiently small, φ(~γ ) and ∂φ(~γ )/∂n are constant and regular on ∂Ω0a , it follows that:

Zπ Z2π   Zπ Z2π  
1 2 ∂φ 1 2
lim −φ a sin β dα dβ + a sin β dα dβ =
a→0 a2 ∂n a
0 0 0 0
Z Z2π
π Zπ Z2π
∂φ
lim −φ [sin β] dα dβ + [a sin β] dα dβ =
a→0 ∂n
0 0 0 0
Zπ Zπ
∂φ
lim −φ [2π sin β] dβ + [2πa sin β] dβ =
a→0 ∂n
0 0
∂φ
lim −φ 4π + (~γ ) 4πa = −4πφ (4.8)
a→0 ∂n
 65

On the other hand, if the source point is located on the surface ∂Ω0 , we obtain:
π π
Z2 Z2π   Z2 Z2π  
1 2 ∂φ 1 2
lim −φ 2
a sin β dα dβ + a sin β dα dβ =
a→0 a ∂n a
0 0 0 0
π π
Z Z2π
2 Z Z2π
2
∂φ
lim −φ [sin β] dα dβ + [a sin β] dα dβ =
a→0 ∂n
0 0 0 0
π π
Z 2 Z 2
∂φ
lim −φ [2π sin β] dβ + [2πa sin β] dβ =
a→0 ∂n
0 0
∂φ
lim −φ 2π + (~γ ) 2πa = −2πφ (4.9)
a→0 ∂n

Finally, if the point source is positioned outside ∂Ω0 , identity (4.3) is valid without mod-
ifications and the integral equation for the velocity potential φ can be stated as:


ZZ   
∂ 1 1 ∂φ(~γ )
  −4πφ(~x) for ~x inside Ω0
0
φ(~γ ) − d∂Ω = −2πφ(~x) for ~x on ∂Ω0 (4.10)
∂n r r ∂n
0 for ~x outside Ω0

∂Ω0

Likewise, the integral equation for the linear acceleration potential ∂φ/∂t can be con-
structed following the same mathematical developments presented for the integral equa-
tion for velocity potential (4.10), which results in equation (4.11).

∂φ(~
x)

ZZ       −4π ∂t for ~x inside Ω0
∂φ(~γ ) ∂ 1 1 ∂ ∂φ(~γ ) 0
− d∂Ω = −2π ∂φ(~
x)
for ~x on ∂Ω0 (4.11)
∂t ∂n r r ∂t ∂n ∂t
0 for ~x outside Ω0

∂Ω0

The classification of the kind of integral equation that should be solved depends on the
type of boundary conditions which need to be imposed. For example, on the mean wet
body surface a Neumman condition must be imposed in order to guarantee the imper-
meability of body surface and, therefore, since the equations present fixed limits of inte-
gration, equations (4.10) and (4.11) are classified as Fredholm integral equations of the
second kind. van Daalen (1993) presents a brief discussion on the combinations of the
choices of singularity distributions over the surface and the type of integral equations that
lead to stable numerical algorithms. According to his work, the choice for a Fredholm
integral equation of the second kind is by far the most used option. This equation results
in matrices with large coefficients concentrated close to the main diagonal, which may be
66

solved by direct inversion or more efficiently by iterative methods when it is necessary to

solve the linear system at all time-steps.

The choice of the type of boundary condition to be imposed on the free surface is more
flexible since we can choose either a Neumann or a Dirichlet description. By choosing a
Neumann boundary condition, we maintain the system of equations with only Fredholm
equations of the second kind, keeping all the benefits aforementioned. However, if we
observe the free surface kinematic condition (3.22), it is possible to realize that after the
integral equation is solved and the velocity potential φ is known at all the boundary
surfaces, an additional numerical spatial differentiation would be necessary to define the
free surface elevation and then advance in time.

Alternatively, it is possible to apply a mixed source-dipole distribution, defining all the

boundary surfaces, but the free surface, as Neumann boundaries. An advantage is that
the normal derivative of the potential on the free surface and the velocity potential on
the body surface are now calculated straightforwardly at all the time-steps. Moreover, as
mentioned in van Daalen (1993), a choice for a mixed scheme reduces the leakage of flux
considerably and leads to better results in comparison to the source-only distribution.
For these reasons, we chose to implement a mixed source-dipole distribution, and the
mathematical implications of this choice will be described in more details in the next
chapter.
 67

Chapter 5

The Numerical Model

The mathematical problem described in the previous chapters may be decomposed in

two parts denoted by elliptical (or spatial) and time-marching problems. In a general
description, most of the numerical algorithms which deal with wave-body formulations
in time domain solves the elliptical problem at a certain time-step, whereas the time-
marching one is used to update the boundary conditions to a new time level. In this
work, the solution of the integral equations which compose the elliptical problem will be
calculated by means of a boundary elements method and the time-marching one will be
performed by a Fourth-Order Runge-Kutta method (RK4).

Describing it in more details, let us consider a free floating body under the influence of
waves in a scenario of infinite water depth in which the velocity and the acceleration
potential vanishes at the sea bottom, and so, it does not need to be discretized. The sim-
ulation begins at t = 0 with the free surface and the body at rest, leading to the consistent
initial conditions of φ = ∂φ/∂t = 0 on the free surface and ∂φ/∂n = ∂ 2 φ/∂t∂n = 0 on
the body’s surface.

The process is initialized by forcing the system with the analytical incoming wave field.
The elliptical problems are then solved providing ∂φ/∂n and ∂ 2 φ/∂t∂n on the free surface,
and φ and ∂φ/∂t on the body surface. Progressing, the first order forces and moments
acting on the body may be then obtained by pressure integration, followed by the deter-
mination of the new acceleration, velocity and position of the body center of gravity at
each stage of the RK4 scheme, which are then used for the calculation of the new body
boundary conditions for the velocity and acceleration potentials integral equations. Free
surface boundary conditions are also updated inside the RK4 scheme. At each of the
RK4 stages, the kinematic and the dynamic conditions are evaluated, providing the new
68

free surface elevation and velocity potential, which are used to define the new free surface
boundary conditions for the acceleration and velocity potential integral equations. These
steps are then followed until the desired simulation time Ts is reached. A schematic view
of this procedure is presented in figure 5.1.

Initial Conditions
∂φ ∂ ∂φ


∂n
= ∂t ∂n
= 0 on Neumann Boundaries
∂φ
φ = ∂t = 0 on Dirichlet Boundaries

Solution of the BVPs

t + ∆t
φ and ∂φ
∂t
on Neumann Boundaries
∂φ ∂ ∂φ


∂n
and ∂t ∂n
on Dirichlet Boundaries

Updating of the Dirichlet Conditions

η and φ by free surface conditions
∂φ
∂t
by dynamic condition

Updating of the Neumann Conditions

F~ and M ~ by pressure integration
2 ~ ~
, , by ξ~ by equations of motion
∂ ξ ∂ξ
∂t2 ∂t

No
t < T s? End
Yes

Figure 5.1: Loop structure of the numerical method

The algorithm briefly described may be simplified if we are dealing with simulations
considering only bodies with prescribed motions or fixed, as the accelerations, velocities
and positions of the body are known in advance. Moreover, as mentioned before, these
simulations do not require the calculation of the pressure field at each time step since the
equations of motions are not integrated in time. In this case, the pressure field may be
post-processed, calculating the time derivative of the velocity potential φ with a centered
difference scheme in time. Therefore, only the integral equation for the velocity potential
must be solved and faster algorithms are then built up.

The next items of this chapter present the numerical methods used for the solution of the
elliptical problem and for the time-marching procedures.
 69

5.1 Boundary Elements Method

The boundary elements method (also known as panel method) is here applied to solve the
integral equations (4.10) and (4.11).

In this work we apply a Low-Order Panel Method which uses plane quadrilateral/triangular
panels (surface elements) to discretize the surfaces and assume that the variable quantities
are constant over each panel, simplifying reasonably the integration of the singularities.
Hess and Smith (1967) is considered a pioneering work in this topic, applying it to a
wide variety of flow problems with bodies in infinite fluid. This method is by far the
simplest option among other possibilities, such as the Higher Order Panel Methods which
use curved panels and linear or quadratic representations of the variable quantities over
surface patches (see for instance Maniar (1995) and Qiu (2001)). The use of higher order
methods certainly would increase the accuracy of the solution if the number of unknowns
are maintained the same, but on the other hand, the integral equations would become
much more difficult to solve. The developments of higher-order methods emerged mainly
to overcome the difficulties of the low-order method to evaluate spatial derivatives that
arise from problems involving forward speed bodies and nonlinear terms. These prob-
lems, however, will not be faced in this linear version and the low-order method was then
considered sufficient to achieve the established objectives.

The boundary surfaces ∂Ω0 are discretized in Np plane quadrilateral/triangular panels in

which a unique point is selected to be a collocation point where the boundary conditions
are imposed and the variable quantities are determined. It is quite obvious that the larger
the number of panels the better is the surface domain representation and the assumption
that the variable quantities are constant over each panel.

The next step is to derive discrete representations of the integral equations (4.10) and
(4.11). Bearing this in mind, we consider a sum of Np integrals over the separated element
surfaces ∂Ω0j for all collocation points ~x, obtaining equations (5.1) and (5.2) for the velocity
and acceleration potential, respectively, as follows:

Np Z Z    
X ∂ 1 1 ∂φj (~γ )
φj (~γ ) − d∂Ω0j = −2πφi (~x) i = 1 : Np (5.1)
j=1
∂n rij rij ∂n
∂Ω0j

Np Z Z     
X ∂φj (~γ ) ∂ 1 1 ∂ ∂φj (~γ ) ∂φi (~x)
− d∂Ω0j = −2π i = 1 : Np (5.2)
j=1
∂t ∂n rij rij ∂t ∂n ∂t
∂Ω0j
70

where rij is a matrix with the distances between collocations points i and field points j
defined by:

q
rij = (xi − ξj0 )2 + (yi − ηj0 )2 + (zi − ζj0 )2 j = 1 : Np i = 1 : Np (5.3)

Assuming that the velocity and acceleration potentials and their normal derivatives are
constant over each planar panel ∂Ω0j , we may take these quantities out of the integrals,
resulting in the discretized equations (5.4) and (5.5) for the velocity and acceleration
potential, respectively.

 
Np ZZ   ZZ
X ∂ 1 ∂φj (~γ ) 1
φj (~γ ) d∂Ω0j − d∂Ω0j  = −2πφi (~x) i = 1 : Np (5.4)

j=1
∂n rij ∂n rij
∂Ω0j ∂Ω0j

 
Np ZZ     ZZ
X  ∂φj (~γ ) ∂ 1 ∂ ∂φj (~γ ) 1 ∂φi (~x)
d∂Ω0j − d∂Ω0j  = −2π i = 1 : Np

∂t ∂n rij ∂t ∂n rij ∂t

j=1
∂Ω0j ∂Ω0j
(5.5)

In a more compact notation, equations (5.4) and (5.5) may be replaced by equations (5.6)
and (5.7)

Np  
X ∂φj
φj Dij − Sij = −2πφi i = 1 : Np (5.6)
j=1
∂n

Np    
X ∂φj ∂ ∂φj ∂φi
Dij − Sij = −2π i = 1 : Np (5.7)
j=1
∂t ∂t ∂n ∂t

where Sij and Dij are known in literature as the source and dipole influence coefficients,
respectively.

Concluding, substitution of φj and ∂φj /∂t for Dirichlet boundaries (i.e free surface), and
∂φj /∂n and ∂ 2 φj /∂t∂n for Neumann boundaries (i.e body surface, bottom etc.) results
in two linear systems of N = Np equations with the following structures:

∂φ
   
  φj Sij ∂nj
(Dij + 2πδij ) −Sij ∂φj = (5.8)
Np ×Np
∂n Np ×1
−(Dij + 2πδij )φj Np ×1
 71

∂φj
" # "   #
∂ ∂φj
  ∂t  Sij ∂t ∂n
(Dij + 2πδij ) −Sij Np ×Np ∂ ∂φj = ∂φj
(5.9)
∂t ∂n
Np ×1
−(Dij + 2πδij ) ∂t Np ×1

where the δij is the Kronecker’s delta function.

The imposition of the boundary conditions for forced and fixed body simulations is rela-
tively straightforward. For fixed bodies, ∂φj /∂n and ∂ 2 φj /∂t∂n are set to zero and the
free surface boundary conditions are updated step-by-step by the integration of the kine-
matic and dynamic free-surface conditions in time. For forced body simulations, these
values are not zero, but on the other hand they are known at each time-step and must be
updated accordingly.

The simulation of free floating bodies, however, demands more attention since we need
to couple the body dynamics with the hydrodynamic problem. As the algorithm requires
the calculation of the body velocity and acceleration at each time-step, it is necessary
to relate these quantities or the equations which determine them with the systems of
equations (5.8) and (5.9).

For a better comprehension, some equations already defined in section 3.3.1 of chapter 3
will be here restated without renumbering. Thus, consider equations (3.61) and (3.62):

2~ (1) (1)
∂φ(1) ∂ 2 ξ~ R
   
∇ ¯ = ∂ ξ T · ~n +
· ~n ¯
× ~r · ~n − ∇
∂φI ¯
· ~n
∂t ∂t2 ∂t2 ∂t
(1) (1)
∂ 2 ξ~ T ∂ 2 ξ~ R ¯
 
∂φI ¯
= · ~n + · ~r × ~n − ∇ · ~n (3.61)
∂t2 ∂t2 ∂t

∂φ ¯ dS
  RR  
"
∂ 2 ξ~ T
# "
∂ ξ~ T
# "
(1)
#
∂t
~n
ξ~ T
∂t2 = M−1  −C ∂t −K − ρ  RR S̄∂φ
B
(3.62)
   
∂ 2 ξ~ R ∂ ξ~ R (1)
ξ~ R (~
r¯ × ~n
¯ ) dS  
∂t2 ∂t ∂t
S̄B

which express the body boundary condition for the acceleration potential and the accel-
eration of the body center of gravity isolated in the equations of motion, respectively.

A first alternative to couple the body acceleration with the linear system (5.9) is the
direct substitution of (3.61) into the first line of the right-hand term. Nevertheless, an
iterative procedure is necessary to solve these equations since neither the body acceler-
ations (∂ 2 ξ~ T /∂t2 , ∂ 2 ξ~ R /∂t2 ) nor the acceleration potential ∂φ/∂t are known. Although
the simplicity of this method for programming and the excellent convergence showed by
72

Cao, R. and Schultz (1994), this procedure may lead to very time-consuming algorithms
and, therefore, was avoided in our developments.

Following the works of van Daalen (1993) and Tanizawa (1995), the approach applied
here combines equation (3.62) and the boundary condition (3.61) to eliminate the body
acceleration of the equations, the acceleration potential ∂φ/∂t remaining as the only
unknown.

5.1.1 Integration of the Influence Coefficients

Once the numerical method is established, we need to assess the behaviour of the spatial
integrals represented by the source and dipole coefficients in order to choose an appropri-
ate numerical integration method. In fact, their behaviours may vary significantly in a
range from smooth, quasi-singular to singular, depending on the variation of the element
surface, collocation and field points positions. Maniar (1995) presents an empirical clas-
sification of each of these integrand shapes by comparing a characteristic length scale (L)
of the element surface to the distance (d) between the field point and the panel centroid.
According to his work, when L/d  1, the variation of the integrands is small and their
behaviours are smooth, defining the often called far-field influence coefficients. When
L/d ∼ O(1), the variations are rapid and the integrands are nearly-singular, categorizing
the integrals as near-field influence coefficients. Finally, when the field point lies on the
surface element, more specifically at the panel centroid position, both integrands become
singular and the integrals are called self-influence coefficients.

Maniar (1995) also states that the integrals categorized as near-field and self-influence
coefficients contribute dominantly to the global matrices and shall be carefully evaluated.
Bearing this in mind, we dedicate the following discussions for their evaluations, presenting
the different approaches applied for the calculation of each integral. All the coefficients
are calculated with an absolute error of 10−5 .

As will be demonstrated, the integration of the far-field influence coefficients by gaussian

quadratures present an easy and fast convergence option, using a small number of gaussian
points. Regarding the evaluation of the near-field influence coefficients, although several
authors apply the partitioning technique of Maniar (1995) (such as Danmeier (1999)
and Kim, Lee and Kerwin (2007)), in this work the coefficients are calculated by simply
increasing the number of gaussian points until the desired convergence is reached.

The evaluation of the self-influence coefficients, however, demands a more sophisticated

approach in which a subdivision of the element surfaces in triangles become essential for
an accurate calculation. In fact, convergence of the results are not achieved even using a
 73

very large number of gaussian points. It is worth mentioning that only the source self-
influence coefficients evaluation must be assessed. Since we are dealing exclusively with
planar panels, the dipole coefficients vanish if the collocation point is located at the same
plane of the panel.

5.1.1.1 Near and Far-Field Influence Coefficients

The integrals representing the near and far-field influence coefficients are calculated in a
straightforward manner. This is performed in an iso-parametric domain and numerically
calculated by a Gauss-Legendre quadrature method. The surface elements (or panels)
in the physical domain (x, y, z) are parameterized by a vectorial function ~r(u, v) of two
parameters u and v, defined over a quadrilateral unitary domain {(u, v)|−1 ≤ u ≤ 1, −1 ≤
v ≤ 1} at the plane uv, as follows:

~r(u, v) = x(u, v)~e1 + y(u, v)~e2 + z(u, v)~e3 (5.10)

in which x(u, v), y(u, v) and z(u, v) are denoted as parametric equations.

In order to exemplify how the parametric equations are derived, consider a quadrilateral
surface element with vertices (x1 , y1 , z1 ), (x2 , y2 , z2 ), (x3 , y3 , z3 ) and (x4 , y4 , z4 ), and the iso-
parametric computational domain, as illustrated in Figure 5.2(a) and 5.2(b), respectively.

1
(x1 , y1 , z1 )
5

0.5
4
z (m)

(x2 , y2 , z2 )
0
v

2
−1 (x4 , y4 , z4 ) −0.5
(x3 , y3 , z3 )
−1.5
2 3
0 1 −1
y (m) −2 −2 −1 −1 −0.5 0 0.5 1
x (m) u

(a) (b)

Figure 5.2: Physical domain (a) and computational iso-parametric domain (b)

Taking the coordinate x, for example, we first define two spatial curves αx (u) and βx (u)
which parameterize the paths x1 → x2 and x4 → x3 in terms of u, respectively.

(u + 1)(x2 − x1 )
βx (u) = + x1 (5.11)
2
74

(u + 1)(x3 − x4 )
αx (u) = + x4 (5.12)
2

Once the coordinates in the direction of u are defined, now, we need to parameterize
the region in between these spatial curves αx (u) and βx (u) in terms of v, obtaining the
parametric equation x(u, v):

(v + 1) (u + 1)(x3 − x4 ) (u + 1)(x2 − x1 )
    
(v + 1)
x(u, v) = + x4 + − +1 + x1 (5.13)
2 2 2 2

The parametric equations y(u, v) and z(u, v) are then defined, analogously, and described
by:

(v + 1) (u + 1)(y3 − y4 ) (u + 1)(y2 − y1 )
    
(v + 1)
y(u, v) = + y4 + − +1 + y1 (5.14)
2 2 2 2

(v + 1) (u + 1)(z3 − z4 ) (u + 1)(z2 − z1 )
    
(v + 1)
z(u, v) = + z4 + − +1 + z1 (5.15)
2 2 2 2

Thus, the integrations of an arbitrary function f (x, y, z) over each panel surface may be
performed by:

ZZ Z1 Z1
f (x, y, z)dΩ0j = f (x(u, v), y(u, v), z(u, v))||J(u, v)||dudv (5.16)
∂Ω0j −1 −1

where J(u, v) is the Jacobian of the transformation, which is calculated by the following
cross product

∂~r ∂~r
J(u, v) = × (5.17)
∂u ∂v

Further on, by applying the Gauss-Legendre quadrature, we may approximate the integral
by a weighted sum of the integrand values, obtaining

Z1 Z1 n X
X n
f (x(u, v), y(u, v), z(u, v))||J(u, v)||dudv ≈ wi wk f (ui , vk )||J(ui , vk )|| (5.18)
−1 −1 i=1 k=1
 75

where wi and wk are the weights and ui and vi are the ith roots of the Legendre’s polynomial
Pn (x).

Figure 5.3 and 5.6 present typical relations between the characteristic surface element scale
(L) and distance between the panel centroid and the field point (d) for far and near fields
classification of the integrals, respectively. Moreover, the correspondent function values
multiplied by the Jacobian of the transformation are also presented. In the figures, the
source and dipole terms are denoted as IFS ar and IFDar for the far field influence coefficients,
and INS ear and INDear for the near field ones. Notice that in a far field condition the integrand
are very smooth whereas in the near field a rapid increase of the integrand (in this case
at v = 1 edge) is clearly perceivable.

As stated before, this behavior influences directly the evaluation of the integrals, de-
manding a specific analysis concerning the number of gaussian points necessary for the
convergence achievement. Concerning the far field influence coefficients, the convergence
analysis of the source and dipole terms are presented in Figures 5.4 and 5.5, respectively.
In accordance to the smooth function behavior observed in Figure 5.3, a very fast con-
vergence of the integrals is observed for both terms, requiring only n = 3 gaussian points
to achieve an absolute error of 10−5 , between two successive evaluations (n and n + 1
gaussian points). Therefore, any far field influence coefficients are evaluated in the code
using n = 4 gaussian points aiming at guaranteeing the convergence.

The same analysis for the near field influence coefficients is presented in Figures 5.7 and
5.8 for the source and dipole terms, respectively. Firstly, we observe that the convergence
is reached with a number of gaussian points much larger than the ones used in the far field
analysis, which is strictly related to the rapid variation of the function at the edge v = 0.
Moreover, the source and dipole coefficients present different behaviors and consequently
required distinct number of gaussian points, which turns more difficult to set a unique
value using as reference only the ratio L/d. In order to overcome this problem, each of the
near field influence coefficients are evaluated in the code until the error of two successive
calculations present an absolute error lower than 10−5 , which in general occurs in a range
of 10 ≤ n ≤ 30 gaussian points.
76

0.5

0.4

0.3

z (m)
d
0.2

0.1

0
3
2 15
L 10
1 5
y (m) 0 0 x (m)

(a)

0.4 0.015
f (u, v)||J(u, v)||

f (u, v)||J(u, v)||

0.3
0.01
0.2
0.005
0.1

0 0
1 1
1 1
0 0
0 0
v −1 −1 u v −1 −1 u

(b) (c)

Figure 5.3: Typical relative positions between the field point (red marker) and panel
centroid (blue marker) for a far field influence coefficient classification (L  1/d) (a) and
the correspondent function values in the parametric domain for the source (b) and dipole
(c) terms

−4
x 10
0.5219 1.5

0.5218

0.5218
||IFS ar (n + 1) − IFS ar (n)||

0.5218 1

0.5218
IFS ar

0.5218

0.5217 0.5

0.5217

0.5217

0.5217 0
1 2 3 2 3
Gaussian Points (n) Gaussian Points (n)

(a) (b)

Figure 5.4: Convergence of the source far field influence coefficient IFS ar
 77

−4
x 10
0.024

0.0239

||IFDar (n + 1) − IFDar (n)||

0.0239
2
0.0238
IFDar

0.0238
1
0.0237

0.0237

0.0236 0
1 2 3 2 3
Gaussian Points (n) Gaussian Points (n)

(a) (b)

Figure 5.5: Convergence of the dipole far field influence coefficient IFDar

0.5

0.4

0.3 d
z (m)

0.2

0.1

0
3
L 3
2
2
1 1
y (m) 0 0 x (m)

(a)

3 50

40
f (u, v)||J(u, v)||

f (u, v)||J(u, v)||

2
30

20
1
10

0 0
1 1
1 1
0 0.5 0 0.5
0 0
−0.5 −0.5
v −1 −1 u v −1 −1 u

(b) (c)

Figure 5.6: Typical relative positions between the field point (red marker) and panel
centroid (blue marker) for a near field influence coefficient classification (L/d ∼ O(1)) (a)
and the correspondent function values in the parametric domain for the source (b) and
dipole (c) terms
78

4.2
0.4
−4

(n)||
4.15 x 10
1
0.35

(n + 1) − INear
4.1

(n)||

S
0.3
4.05

(n + 1) − INear
0.25 0.5

S
4
INear

0.2
S

3.95

||INear
S
3.9 0.15 0
8 10 12 14

||INear
Gaussian Points (n)

S
3.85 0.1

3.8 0.05

3.75 0
0 2 4 6 8 10 12 14 2 4 6 8 10 12 14
Gaussian Points (n) Gaussian Points (n)

(a) (b)

Figure 5.7: Convergence of the source near field influence coefficient INS ear

6
17 −4

(n)||
x 10
1
5 (n + 1) − INear
16
(n)||

D
(n + 1) − INear

4 0.5
15
D
INear

3
||INear

14
D

0
17 18 19 20 21
2 Gaussian Points (n)
||INear

13
D

12 1

11 0
0 5 10 15 20 25 0 5 10 15 20 25
Gaussian Points (n) Gaussian Points (n)

(a) (b)

Figure 5.8: Convergence of the dipole near field influence coefficient INDear

5.1.1.2 Self-Influence Coefficients

Now we turn our attention for the evaluation of the self-influence coefficients that is
related to the situation in which the field point coincides with the panel centroid. In
order to elucidate the complexity of such a problem, Figure 5.9 presents the singular
S
function behavior of the source term ISelf , which presents a notable and undetermined
peak value at the center of the parametric domain. In addition, Figure 5.10 presents the
convergence analysis of this coefficient using the same integration approach followed for the
evaluation of the far and near field coefficients. Notice that even using n = 150 gaussian
points the error remained higher than the established criteria of 10−5 . Therefore, a more
sophisticated approach must be used aiming at evaluating these terms. The analysis of
D
the dipole coefficient ISelf is neglected, since by definition, the scalar product ~r · ~n equals
 79

zero if the field point is located at the same plane of the panel, and then, the integral
vanishes in this situation.

1
150
0.5

f (u, v)||J(u, v)||

100
z (m)

−0.5 50

−1
3 0
1
2 3 1
2 0
1 1 0
y (m) 0 0 x (m) v −1 −1 u

(a) (b)

Figure 5.9: Typical relative positions between the field point (red marker) and panel
centroid (blue marker) for a self-influence coefficient classification (d = 0) (a) and the
correspondent function values in the parametric domain for the source term (b)

8.5 1.4
−4
x 10
(n)||

14
8 1.2
(n + 1) − ISelf

12
(n)||

1
7.5 10
(n + 1) − ISelf
S

0.8 8
ISelf

7
||ISelf
S

6
S

0.6
6.5 50 100 150 200
||ISelf

0.4 Gaussian Points (n)

S

6 0.2

5.5 0
0 50 100 150 200 0 50 100 150 200
Gaussian Points (n) Gaussian Points (n)

(a) (b)

S
Figure 5.10: Convergence of the source self-influence coefficient ISelf

The procedure here applied to evaluate the self-influence coefficients is based on the works
of Maniar (1995) and Kim, Lee and Kerwin (2007), which remove the singularity of the
integrand by partitioning the quadrilateral parametric space into four triangles. Following
their works, we introduce the quadratic transformations (5.19) to (5.22) for the triangles
∆(1) , ∆(2) , ∆(3) and ∆(4) , respectively. In this method, the apex of each triangle is located
at the panel centroid and their bases at the sides of the square, as illustrated in Figure
5.11, in which ut and vt are two parameters defined in a new parametric quadrilateral
domain {(ut , vt )|0 ≤ ut ≤ 1, −1 ≤ vt ≤ 1}.
80

0.5 ∆(2)

0 ∆(3) ∆(1)
v

−0.5 ∆(4)

−1
−1 −0.5 0 0.5 1
u

Figure 5.11: Triangulation of the parametric space. Figure adapted from Maniar (1995)

u = ut
∆(1) : (5.19)
v = ut vt

u = ut vt
∆(2) : (5.20)
v = ut

u = −ut
∆(3) : (5.21)
v = −ut vt

u = −ut vt
∆(4) : (5.22)
v = −ut

S
The integral ISelf may then be evaluated by summing the contribution of each of the
four triangles, as presented in expression (5.23). Notice that the four integrals must be
evaluated with the associated quadratic transformations (5.19) to (5.22).

4 ZZ
X
S
ISelf = f (ut , vt )||J(u, v)||||Jt (ut , vt )||dut dvt (5.23)
m=1 (m)
∆
 81

in which Jt (ut , vt ) is the Jacobian of the new transformation.

Figure 5.12 illustrates the four integrands obtained from the new transformations, where
it is possible to observe a considerable enhancement in terms of the functions smoothness.
As a consequence, these new surfaces can easily be integrated by using the standard
gaussian integration method presented before.

f (ut , vt )||J(u, v)||||J(ut , vt )||

f (ut , vt )||J(u, v)||||J(ut , vt )||

1.5 1

1 0.8

0.5 0.6

0 0.4
1 1
1 1
0 0
0.5 0.5
vt −1 0 ut vt −1 0 ut

(a) (b)
f (ut , vt )||J(u, v)||||J(ut , vt )||

f (ut , vt )||J(u, v)||||J(ut , vt )||

1.5 1.5

1 1

0.5 0.5

0 0
1 1
1 1
0 0
0.5 0.5
vt −1 0 ut vt −1 0 ut

(c) (d)

Figure 5.12: Function values obtained from the new transformations: ∆(1) (a), ∆(2) (b),
∆(3) (c) and ∆(4) (d)

The improvements brought through the use of this partitioning approach are demonstrated
in Figure 5.13, in which the convergence analysis presented in Figure 5.10 is restated with
the inclusion of the new reached results. It is evident that using the partitioning method
a tremendous improvement in the convergence of the integral is achieved, in which an
absolute error of 10−5 is rapidly obtained with much less gaussian points than the direct
method. Typically, by using 10 gaussian points the convergence of the integral is reached.
82

10 2
Direct Method Direct Method
Partitioning Method Partitioning Method
9.5 −4
x 10

(n)||
1

(n)||
9 1.5

(n + 1) − ISelf
(n + 1) − ISelf
8.5

S
S
8 0.5
ISelf

1
S

7.5

||ISelf
7

||ISelf

S
0

S
0.5 7.5 8 8.5 9
6.5 Gaussian Points (n)
6

5.5 0
0 50 100 150 200 0 50 100 150 200
Gaussian Points (n) Gaussian Points (n)

(a) (b)

Figure 5.13: Comparison of the direct and partitioning methods for the convergence of
S
the source self-influence coefficient ISelf

5.2 Time-Marching Scheme

As mentioned before, the time-marching of both fluid and body motion equations are
performed by a Fourth Order Runge-Kutta method which provides high accuracy and
stability to numerical methods, such as the one we are dealing with, as already pointed
out by Tanizawa (2000). Particularly, the method demands four intermediate calculations
to advance the simulation single time step ∆t. The method was implemented as follows:

k1 = h(tn , y) (5.24)
k2 = h(tn + h/2, y + hk1 /2) (5.25)
k3 = h(tn + h/2, y + hk2 /2) (5.26)
k4 = h(tn + h, y + hk3 ) (5.27)
h
yn+1 = yn + (k1 + 2k2 + 2k3 + k4 ) (5.28)
6

In the equations above, y denotes generically all the variables that must be calculated
within a time-step ∆t, such as the body position ξ, ~ free surface elevation η and the
velocity potential at the undisturbed water surface. For the calculation of the body
motions, the set of six second order differential equations (one for each degree of freedom)
are transformed into a set of twelve first order equations prior to their evaluations in RK4
scheme.

One last discussion, but also very important for the consistency of the method, is related
to two numerical conditions which must be satisfied at the beginning of the simulations.
 83

First, we need to enforce a rest condition over all the boundary surfaces to respect the
initial conditions of the IBVPs and then guarantee a correct determination of the sub-
sequently fluid and body motions. Second, the numerical model should avoid impulsive
responses in the domain since it may induce long transient periods with no physical in-
terest to our analysis. In our numerical model, these conditions are satisfied by the use of
a ramp function fr (t) that multiplies the boundary velocities and accelerations, guaran-
teeing a smooth and slow increase of the variables until the achievement of a steady-state
solution. The ramp function is defined by:

( h i
1 πt
2
1 − cos( Tr ) if t ≤ Tr
fr (t) = (5.29)
1 if t > Tr

where Tr is the ramp time which is normally set as a multiple of a characteristic wave
period involved in the simulations. If one needs to obtain the transient solution, such in
a numerical decay test, very small values of Tr must be used.
84
 85

Chapter 6

Verification Tests with Simple

Geometries

Since one of the main goals of the numerical method described in this text is the evaluation
of the body motions caused by the action of free surface gravity waves, we need to check
its capability to predict the hydrodynamic loads involved in such a problem. Upon the
already established linear theory assumption, the superposition of solutions is valid and
the hydrodynamic loads, induced by the free floating body motions under waves, may be
assumed as a sum of two different components, which may be calculated by the well known
diffraction and radiation problems. The diffraction problem consist on the calculation
of the hydrodynamic forces and moments, or exciting forces and moments, induced by
the interaction of the wave field with a fixed body. The radiation problem refers to
the calculation of the forces induced by harmonic oscillations of the body in still water
and in the absence of incoming waves. When presenting numerical results for the forces
exerted on oscillating bodies it is usual to quote values for the added mass and damping
coefficients, which measure the components in phase with acceleration and velocity of the
body, respectively, and this convention will be applied here.

This chapter presents the computations of the hydrodynamic loads resulting from the
diffraction and radiation problems in infinite water depth, considering simple geometries
such as an hemisphere and a circular cylinder. Ultimately, the response of these bodies
in regular waves is also presented. Benchmark results were obtained from calculations
performed with the software WAMIT which solves the same Boundary Value Problem
here defined, but in the frequency domain. Lee and Newman (2005) presented a review
about case studies already analyzed with WAMIT, which includes both a Low-Order and a
86

Higher-Order numerical approach. It is important to emphasize that the objective of this

analysis is to validate the present numerical results and not to compare the performance
of both Low-Order models. In fact, they would not be comparable to each other, since
one treats the problem in time domain and the other in frequency domain. Therefore,
for comparison purposes, all the numerical results obtained by WAMIT were run using
the Higher-Order method, since for the type of geometries here assessed, the convergence
of the results are reached faster, with very low computational efforts and errors that are
insignificant when comparing the results to analytical solutions, ensuring the accuracy
of this benchmark data. For the hemisphere radiation problem, the added mass and
potential damping coefficients are compared to analytical solutions presented by Hulme
(1982).

Convergence analyses were performed testing three panel grid resolutions for the free
surface and the body wet surface. Aiming at maintaining the results dependent only on
the number of panels, the mesh number of panels was increased using a constant factor
that multiplies both sides of the quadrilateral panels, thus, keeping constant the panels
aspect ratio.

6.1 Geometries Discretization

An hemisphere surface of unitary radius rh = 1 m was discretized in three different grids

with 200, 800 and 3200 panels and a circular cylinder, of radius rc = 1 m and draught
Tc = 1 m, in 120, 500 and 2000 panels. Notice that we tried to set the scale factor value to
four by duplicating the discretization parameters in each direction by two. In the circular
cylinder case the panels are more concentrated near the intersection region formed by
the encounter of the free surface and the body surface. This was done to guarantee a
minimum resolution of panels for cases involving high frequency waves in which their
lengths are small and the velocity profile confined to this region. Moreover, as observed
by Hess and Smith (1967), if several small elements are in the vicinity of a larger one, the
accuracy of the solution is associated with the larger elements. For the hemisphere this
was guaranteed by increasing the total number of panels. The six body meshes considered
in the calculations are displayed in Figure 6.1.
 87

(a) (b)

(c) (d)

(e) (f)

Figure 6.1: Circular cylinder and hemisphere panel meshes used in the computations.
Figures (a), (c) and (e) refers to the 120, 500 and 2000 circular cylinder panel meshes,
respectively. Figures (b), (d) and (f) refers to the 200, 800 and 3200 hemisphere panel
meshes, respectively.

The free surface length depends directly on the wave lengths that will be simulated and
due to the limitations of computer memory, we need to truncate the free surface within
reasonable numbers, bearing also in mind that a percentage of this area must be desig-
nated for energy dissipation since the radiation condition must be satisfied. For these
computations, we considered circular free surfaces meshes with radius rf s = 30 m, pro-
viding sufficient space for the propagation of the waves considered herein. Similar to the
body surfaces, three different panel meshes are considered containing 289, 1225 and 4900
panels and, again, an increasing factor value of approximate four was used. Figure 6.2
displays the meshes used in the computations where it is possible to observe that a high
number of panels is situated near the body surface. Once again, a minimum number of
panels is needed to ensure that the waves are being properly considered.
88

(a) (b)

(c)

Figure 6.2: Free surface panel meshes used in the computations. Figures (a), (b) and (c)
refers to the 289, 1225 and 4900 free surface panel meshes, respectively.

A set of six panel meshes are then constructed by grouping the body and free surface
meshes in pairs. Denoting the meshes with the lowest number of panels as coarse meshes,
the intermediates as medium meshes and the ones with greatest resolution as fine meshes,
Table 6.1 presents the main numbers of the panel meshes that will be simulated.

Table 6.1: Main numbers of the panel meshes used in the computations
Hemisphere Circular Cylinder
Coarse Medium Fine Coarse Medium Fine
Body Mesh 200 800 3200 120 500 2000
Free Surface Mesh 289 1225 4900 289 1225 4900
Final Mesh 489 2025 8100 409 1725 6900
 89

6.2 Fixed Body Simulations

We begin our numerical results presentation simulating cases of fixed bodies interacting
with incoming regular waves. Table 6.2 presents the main features of the set of 24 regular
waves in a frequency range between 1.716 rad/s and 9.905 rad/s, which were tested in
our numerical model. All the waves present unitary amplitudes AI = 1 m. For all the
simulations the time step was set to ∆t = T /60 s and the numerical beach coefficients
were set to a = 1.0 and b = 2.0, with the exception of waves 1 to 3, which were simulated
considering b = 1.0, since their lengths were such that the damping zone would not fit
into the domain.

It is important to emphasize that despite the fact that the wave amplitudes present the
same magnitude order of the body geometry dimensions, the numerical method here de-
veloped treats the problem upon the linear theory assumption and, therefore, its solutions
have a linear dependence on the wave amplitude. Thus, if not specified, all the simulations
that will be presented in this thesis consider waves of unitary amplitudes so as to make
possible a direct comparison of the solution quantities in terms of RAOs calculated by
other numerical tools or experimental activities.

Table 6.2: Regular incoming waves used in the simulations

ID ω (rad/s) T(s) λ(m) ID ω (rad/s) T(s) λ(m)
1 1.716 3.662 20.932 13 4.429 1.419 3.142
2 1.981 3.172 15.707 14 4.952 1.269 2.514
3 2.215 2.837 12.563 15 5.425 1.158 2.094
4 2.426 2.590 10.473 16 5.860 1.072 1.795
5 2.620 2.398 8.979 17 6.264 1.003 1.571
6 2.801 2.243 7.856 18 6.644 0.946 1.396
7 2.971 2.115 6.983 19 7.004 0.897 1.256
8 3.132 2.006 6.284 20 7.672 0.819 1.047
9 3.431 1.831 5.236 21 8.287 0.758 0.898
10 3.706 1.695 4.488 22 8.859 0.709 0.785
11 3.962 1.586 3.927 23 9.396 0.669 0.698
12 4.202 1.495 3.491 24 9.905 0.634 0.628

As mentioned before, simulations start from a rest condition and during a pre-specified
time is controlled by a ramp function in order to provide a smooth transition to the
steady-state solution. The regular incoming wave is imposed in the problem as a boundary
condition. As previously discussed, we could decompose the total potential in a sum of an
analytic potential of regular waves and a disturbed wave part which enabled us to change
the problem variable. This approach brings several benefits to our model if we do not
need to make comparisons which involve the wave generation itself. For example, once the
90

waves are generated at the body surroundings, we do not need to have great concentrations
of panels at the whole free surface allowing us to sparse the free surface panels from the
body towards the free surface edge. Moreover, due to energy conservation, as the waves
propagate away from the body their amplitudes decrease and, therefore, less effort to
dissipate the waves are necessary in comparison to a situation in which is necessary to
damp the total potential (incoming wave + disturbed waves). Both of these examples
reduce the number of necessary panels to discretize the free surface and, consequently,
improve the code performance to reproduce the results in relation to wave generation
approaches.

Despite of the benefits pointed out above, it is important to emphasize that even concen-
trating small panels near the body surface and using numerical damping zones, a minimum
number of panels per wave length is required to correctly satisfy the dispersion relation of
waves in infinite water depth and, also, avoid signal modulations induced by the lack of
panels. In order to illustrate it, we perform simulations with the cylinder body considering
three wave frequencies of ω = 4.202 rad/s, ω = 6.264 rad/s and ω = 9.905 rad/s with
respective wave lengths of λ = 3.491 m, λ = 1.571 m and λ = 0.628 m. The time series of
the horizontal Fx and vertical Fz hydrodynamic forces are presented in Figures 6.3, 6.4,
6.5 and 6.6, 6.7, 6.8, respectively.

It is possible to realize that as the angular frequency increases larger differences in terms
of amplitude and phase between the hydrodynamic forces calculated with each mesh are
observed. Wave reflections due to the lack of panels are clearly observed for the coarse
mesh when simulating waves with frequency higher than ω = 6.264 rad/s. For the highest
frequency, for example, the vertical forces calculated by the coarse mesh present non
physical results with order of magnitude much larger than those obtained with the other
meshes. Moreover, a slightly variation of phase may also be observed. Results for the
medium and fine meshes present better agreement for all the frequencies, but the highest
one, what indicates that only the fine mesh is capable to correctly propagate waves of
such small length.

The results for the hemisphere body are not presented since the same behavior was ob-
served. This was expected since the propagation of waves is related to the free surface
meshes which are the same for both geometries.
 91

4
x 10
Coarse Medium Fine
2

1
Fx (N)

−1

−2

0 5 10 15
Time (s)

Figure 6.3: Hydrodynamic horizontal force Fx induced by a wave with frequency of ω =

4.202 rad/s. Cylinder body

4
x 10
1.5
Coarse Medium Fine

0.5
Fx (N)

−0.5

−1

−1.5
0 2 4 6 8 10
Time (s)

Figure 6.4: Hydrodynamic horizontal force Fx induced by a wave with frequency of ω =

6.264 rad/s. Cylinder body
92

2000
Coarse Medium Fine
1500

1000

500
Fx (N)

−500

−1000

−1500

−2000
0 1 2 3 4 5 6 7
Time (s)

Figure 6.5: Hydrodynamic horizontal force Fx induced by a wave with frequency of ω =

9.905 rad/s. Cylinder body

2000 Coarse Medium Fine

1500

1000

500
Fz (N)

−500

−1000

−1500

−2000
0 5 10 15
Time (s)

Figure 6.6: Hydrodynamic vertical force Fz induced by a wave with frequency of ω = 4.202
rad/s. Cylinder body
 93

Coarse Medium Fine

100

50
Fz (N)

−50

−100
0 2 4 6 8 10
Time (s)

Figure 6.7: Hydrodynamic vertical force Fz induced by a wave with frequency of ω = 6.264
rad/s. Cylinder body

150
Coarse Medium Fine

100

50
Fz (N)

−50

−100

−150
0 1 2 3 4 5 6 7
Time (s)

Figure 6.8: Hydrodynamic vertical force Fz induced by a wave with frequency of ω = 9.905
rad/s. Cylinder body

In order to check the convergence of the results we compare the horizontal and vertical non-
94

dimensional forces for all panel meshes and wave frequencies presented in Table 6.2. For
the circular cylinder, the non-dimensional moment My around the y-axis is also presented.
The non-dimensional forces and moments are calculated using the following definitions:

Hemisphere Circular Cylinder

fx fx
F̄x = 2 3 F̄x =
3
πrh ρg ρgπrc2 Tc
fz fz (6.1)
F̄z = 2 3 F̄z =
3
πrh ρg ρgπrc2 Tc
my
M̄y =
ρgπrc3 Tc

where fx , fz and my are the, respective, amplitudes of the forces Fx and Fz and moment
My . Note that for the medium and fine mesh, the amplitude values are easily determined
by the forces and moments time series, since a steady-state solution is available. For the
coarse mesh, however, only a short sample of the time series may be used because the
solution is spoiled with numerical reflections.

The convergence analysis for the non-dimensional horizontal and vertical hydrodynamic
forces for the hemisphere body is presented in Tables 6.3 and 6.4, respectively. In order
to better visualize the results, the data are also illustrated in Figures 6.9 and 6.10. The
WAMIT results used as benchmark data was obtained using the Higher Order approach
with a panel size parameter of 0.05, which was defined by a prior grid convergence analysis.
From the results, it may be observed that the relative errors between both calculations
tend to reduce with the increase of the number of panels, achieving very low values
for the mesh with finest resolution. One should notice that, even considering the fine
mesh, some relative errors are still higher, especially for the vertical hydrodynamic forces.
Nevertheless, we may also realize that these errors are being obtained with very low values
which any insignificant difference may result in very large relative errors.

The simulations with the circular cylinder presented very similar trends in comparison to
the hemisphere ones. Again, convergence of the results is confirmed increasing the number
of panels and very low relative errors are observed for most of the wave frequencies. Tables
6.5, 6.6 and 6.7 present the convergence analyses performed with the horizontal force,
vertical force and the moment in relation to the y-axis, respectively. Once more, the
data are illustrated in Figures 6.11, 6.12 and 6.13. In general, a good agreement between
WAMIT results and the present values is observed.
 95

1
Wamit Coarse Medium Fine

0.8

0.6
F̄x

0.4

0.2

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.9: Convergence analysis of the non-dimensional horizontal hydrodynamic force

F̄x for the hemisphere body

1.4
Wamit Coarse Medium Fine
1.2

0.8
F̄z

0.6

0.4

0.2

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.10: Convergence analysis of the non-dimensional vertical hydrodynamic force F̄z
for the hemisphere body
96

Table 6.3: Convergence analysis of the non-dimensional horizontal hydrodynamic force

F̄x for the hemisphere body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|F̄x −WAMIT| |F̄x −WAMIT| |F̄x −WAMIT|
ID ω WAMIT F̄x WAMIT
F̄x WAMIT
F̄x WAMIT
1 1.716 0.403 0.394 0.023 0.400 0.009 0.402 0.003
2 1.981 0.516 0.501 0.029 0.510 0.012 0.513 0.005
3 2.215 0.614 0.595 0.031 0.606 0.014 0.609 0.008
4 2.426 0.694 0.678 0.023 0.689 0.007 0.693 0.002
5 2.620 0.755 0.738 0.023 0.748 0.010 0.752 0.005
6 2.801 0.796 0.776 0.025 0.789 0.009 0.793 0.004
7 2.971 0.817 0.800 0.020 0.810 0.009 0.813 0.004
8 3.132 0.821 0.804 0.021 0.815 0.008 0.819 0.003
9 3.431 0.795 0.780 0.019 0.791 0.006 0.794 0.002
10 3.706 0.744 0.734 0.014 0.739 0.007 0.741 0.004
11 3.962 0.685 0.675 0.014 0.681 0.005 0.683 0.002
12 4.202 0.626 0.616 0.016 0.621 0.007 0.623 0.005
13 4.429 0.572 0.565 0.013 0.566 0.010 0.567 0.009
14 4.952 0.460 0.448 0.026 0.455 0.011 0.456 0.010
15 5.425 0.378 0.369 0.023 0.374 0.009 0.376 0.004
16 5.860 0.316 0.313 0.012 0.313 0.011 0.317 0.003
17 6.264 0.265 0.246 0.072 0.268 0.010 0.272 0.028
18 6.644 0.232 0.221 0.048 0.231 0.001 0.235 0.014
19 7.004 0.203 0.199 0.016 0.202 0.002 0.203 0.003
20 7.672 0.160 0.143 0.101 0.154 0.035 0.157 0.014
21 8.287 0.134 0.110 0.180 0.122 0.086 0.129 0.033
22 8.859 0.107 0.092 0.141 0.106 0.011 0.108 0.004
23 9.396 0.091 0.077 0.156 0.085 0.067 0.089 0.029
24 9.905 0.080 0.042 0.470 0.068 0.142 0.077 0.030
 97

Table 6.4: Convergence analysis of the non-dimensional vertical hydrodynamic force F̄z
for the hemisphere body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|F̄z −WAMIT| |F̄z −WAMIT| |F̄z −WAMIT|
ID ω WAMIT F̄z WAMIT
F̄z WAMIT
F̄z WAMIT
1 1.716 1.017 1.003 0.015 0.993 0.024 0.998 0.019
2 1.981 0.902 0.888 0.016 0.897 0.006 0.903 0.000
3 2.215 0.805 0.796 0.011 0.803 0.002 0.805 0.001
4 2.426 0.722 0.711 0.015 0.715 0.009 0.717 0.007
5 2.620 0.650 0.642 0.013 0.646 0.007 0.647 0.005
6 2.801 0.589 0.581 0.013 0.584 0.008 0.585 0.006
7 2.971 0.535 0.530 0.009 0.530 0.008 0.531 0.007
8 3.132 0.487 0.481 0.012 0.483 0.008 0.484 0.007
9 3.431 0.409 0.408 0.000 0.405 0.008 0.405 0.008
10 3.706 0.346 0.342 0.013 0.343 0.010 0.343 0.010
11 3.962 0.296 0.297 0.004 0.296 0.002 0.296 0.003
12 4.202 0.256 0.253 0.010 0.256 0.002 0.256 0.001
13 4.429 0.222 0.229 0.031 0.223 0.005 0.223 0.004
14 4.952 0.162 0.167 0.031 0.164 0.010 0.164 0.009
15 5.425 0.120 0.118 0.018 0.123 0.029 0.123 0.024
16 5.860 0.092 0.096 0.039 0.094 0.017 0.093 0.009
17 6.264 0.072 0.074 0.021 0.073 0.002 0.072 0.003
18 6.644 0.058 0.054 0.072 0.058 0.007 0.058 0.002
19 7.004 0.048 0.048 0.011 0.048 0.004 0.048 0.016
20 7.672 0.033 0.031 0.041 0.035 0.077 0.034 0.057
21 8.287 0.024 0.025 0.048 0.024 0.000 0.023 0.016
22 8.859 0.018 0.016 0.119 0.017 0.056 0.018 0.002
23 9.396 0.013 0.014 0.026 0.016 0.150 0.015 0.114
24 9.905 0.011 0.014 0.246 0.012 0.073 0.011 0.007
98

0.8

0.7

0.6

0.5
F̄x

0.4

0.3

0.2

0.1
Wamit Coarse Medium Fine
0
0 2 4 6 8 10
ω (rad/s)

Figure 6.11: Convergence analysis of the non-dimensional horizontal hydrodynamic force

F̄x for the circular cylinder body

Wamit Coarse Medium Fine

0.8

0.7

0.6

0.5
F̄z

0.4

0.3

0.2

0.1

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.12: Convergence analysis of the non-dimensional vertical hydrodynamic force F̄z
for the circular cylinder body
 99

0.35
Wamit Coarse Medium Fine
0.3

0.25

0.2
M̄y

0.15

0.1

0.05

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.13: Convergence analysis of the non-dimensional hydrodynamic moment M̄y for
the circular cylinder body
100

Table 6.5: Convergence analysis of the non-dimensional horizontal hydrodynamic force

F̄x for the circular cylinder body. WAMIT values are used as reference for the relative
errors
Coarse Medium Fine
|F̄x −WAMIT| |F̄x −WAMIT| |F̄x −WAMIT|
ID ω WAMIT F̄x WAMIT
F̄x WAMIT
F̄x WAMIT
1 1.716 0.432 0.412 0.048 0.425 0.017 0.430 0.005
2 1.981 0.557 0.525 0.056 0.546 0.018 0.553 0.006
3 2.215 0.665 0.625 0.060 0.651 0.021 0.659 0.009
4 2.426 0.751 0.711 0.053 0.739 0.016 0.748 0.004
5 2.620 0.810 0.768 0.052 0.796 0.018 0.805 0.007
6 2.801 0.841 0.798 0.052 0.828 0.016 0.836 0.006
7 2.971 0.847 0.811 0.042 0.834 0.014 0.842 0.005
8 3.132 0.832 0.800 0.039 0.823 0.011 0.829 0.003
9 3.431 0.768 0.750 0.023 0.764 0.005 0.768 0.000
10 3.706 0.687 0.682 0.008 0.685 0.003 0.687 0.001
11 3.962 0.610 0.604 0.009 0.610 0.001 0.611 0.002
12 4.202 0.541 0.536 0.008 0.539 0.003 0.540 0.002
13 4.429 0.481 0.471 0.021 0.477 0.009 0.478 0.007
14 4.952 0.368 0.344 0.067 0.360 0.021 0.364 0.010
15 5.425 0.291 0.267 0.082 0.283 0.025 0.289 0.005
16 5.860 0.236 0.222 0.060 0.230 0.023 0.237 0.005
17 6.264 0.196 0.182 0.072 0.194 0.010 0.199 0.016
18 6.644 0.165 0.161 0.028 0.166 0.003 0.168 0.016
19 7.004 0.142 0.144 0.017 0.143 0.006 0.143 0.006
20 7.672 0.108 0.099 0.087 0.104 0.039 0.107 0.010
21 8.287 0.086 0.077 0.102 0.081 0.065 0.087 0.005
22 8.859 0.071 0.068 0.041 0.071 0.000 0.071 0.004
23 9.396 0.059 0.063 0.072 0.056 0.048 0.057 0.028
24 9.905 0.050 0.035 0.315 0.044 0.126 0.050 0.007
 101

Table 6.6: Convergence analysis of the non-dimensional vertical hydrodynamic force F̄z
for the circular cylinder body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|F̄z −WAMIT| |F̄z −WAMIT| |F̄z −WAMIT|
ID ω WAMIT F̄z WAMIT
F̄z WAMIT
F̄z WAMIT
1 1.716 0.583 0.555 0.047 0.564 0.032 0.570 0.021
2 1.981 0.490 0.466 0.049 0.482 0.015 0.488 0.003
3 2.215 0.412 0.393 0.047 0.407 0.012 0.412 0.002
4 2.426 0.349 0.328 0.058 0.340 0.023 0.344 0.012
5 2.620 0.295 0.277 0.061 0.289 0.022 0.292 0.009
6 2.801 0.251 0.234 0.066 0.244 0.025 0.248 0.011
7 2.971 0.213 0.197 0.074 0.207 0.028 0.210 0.012
8 3.132 0.182 0.167 0.082 0.176 0.031 0.179 0.013
9 3.431 0.133 0.120 0.097 0.128 0.038 0.131 0.014
10 3.706 0.098 0.085 0.126 0.093 0.046 0.096 0.017
11 3.962 0.072 0.064 0.114 0.070 0.035 0.072 0.002
12 4.202 0.054 0.048 0.115 0.053 0.024 0.055 0.009
13 4.429 0.041 0.038 0.057 0.040 0.005 0.042 0.022
14 4.952 0.020 0.023 0.141 0.022 0.090 0.022 0.074
15 5.425 0.010 0.014 0.362 0.012 0.186 0.012 0.105
16 5.860 0.006 0.008 0.462 0.006 0.156 0.006 0.057
17 6.264 0.003 0.003 0.044 0.003 0.063 0.003 0.018
18 6.644 0.002 0.001 0.293 0.001 0.164 0.002 0.063
191 7.004 0.001 0.003 − 0.002 − 0.001 −
201 7.672 0.000 0.004 − 0.002 − 0.001 −
211 8.287 0.000 0.001 − 0.000 − 0.000 −
221 8.859 0.000 0.002 − 0.001 − 0.001 −
231 9.396 0.000 0.004 − 0.002 − 0.000 −
241 9.905 0.000 0.003 − 0.000 − 0.000 −

1
The relative errors are omitted because the comparison involves very low values for which any in-
significant differences result in very large relative errors
102

Table 6.7: Convergence analysis of the non-dimensional hydrodynamic moment M̄y for
the circular cylinder body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|M̄y −WAMIT| |M̄y −WAMIT| |M̄y −WAMIT|
ID ω WAMIT M̄y WAMIT
M̄y WAMIT
M̄y WAMIT
1 1.716 0.133 0.139 0.044 0.135 0.015 0.134 0.005
2 1.981 0.173 0.179 0.034 0.176 0.013 0.174 0.004
3 2.215 0.209 0.215 0.029 0.211 0.010 0.209 0.001
4 2.426 0.237 0.246 0.036 0.241 0.016 0.238 0.007
5 2.620 0.256 0.266 0.037 0.260 0.013 0.257 0.004
6 2.801 0.267 0.276 0.036 0.271 0.015 0.268 0.005
7 2.971 0.268 0.281 0.046 0.273 0.017 0.270 0.005
8 3.132 0.263 0.276 0.048 0.268 0.019 0.265 0.007
9 3.431 0.241 0.256 0.061 0.247 0.024 0.244 0.010
10 3.706 0.214 0.229 0.070 0.219 0.023 0.215 0.008
11 3.962 0.187 0.198 0.063 0.191 0.024 0.188 0.010
12 4.202 0.163 0.172 0.055 0.165 0.017 0.163 0.005
13 4.429 0.142 0.147 0.034 0.143 0.008 0.142 0.001
14 4.952 0.102 0.101 0.010 0.102 0.001 0.102 0.003
15 5.425 0.075 0.077 0.027 0.076 0.016 0.076 0.013
16 5.860 0.056 0.063 0.124 0.059 0.050 0.059 0.037
17 6.264 0.043 0.051 0.176 0.047 0.084 0.046 0.057
18 6.644 0.034 0.040 0.185 0.037 0.088 0.035 0.049
19 7.004 0.027 0.031 0.172 0.028 0.050 0.027 0.020
20 7.672 0.018 0.017 0.050 0.017 0.034 0.018 0.004
21 8.287 0.012 0.016 0.348 0.014 0.117 0.013 0.082
221 8.859 0.009 0.013 − 0.010 − 0.009 −
231 9.396 0.007 0.007 − 0.006 − 0.006 −
241 9.905 0.005 0.004 − 0.006 − 0.006 −

6.3 Forced Motion Simulations

This item presents the simulations of forced harmonic oscillations imposed on both the
hemisphere and the circular cylinder body. Unlike the fixed body simulations, this study
does not contain incident waves and the hydrodynamic loads are generated only by body
oscillatory motions with unitary amplitude. Again, all the simulations were run with a
time step of ∆t = T /60 s and with the numerical beach coefficients set to a = 1.0 and
b = 1.0 or b = 2.0, depending on the wave length.

In linear theory it is quite usual to decompose the total hydrodynamic force induced by
1
The relative errors are omitted because the comparison involves very low values for which any in-
significant differences result in very large relative errors
 103

harmonic oscillations in two different components associated to added mass and radiation
damping, which are in phase with the body acceleration and velocity, respectively. To de-
rive these quantities let us consider, for example, a pure heave forced harmonic oscillation
with amplitude A3 and angular frequency ω, as follows:

ξ3 (t) = A3 sin(ωt) (6.2)

The body velocity and acceleration are then easily found by:

∂ξ3 (t)
= ωA3 cos(ωt) (6.3)
∂t

∂ 2 ξ3 (t)
= −ω 2 A3 sin(ωt) (6.4)
∂t2

We now assume that the hydrodynamic forces induced by this motion may be separated
into a component in phase with the body acceleration and another with the body velocity,
as described next:

∂ 2 ξ3 (t) ∂ξ3 (t)

F3 (t) = A33 2
+ B33 (6.5)
∂t ∂t
where the coefficients, Aij and Bij , refer to the added mass and radiation damping, re-
spectively, on the ith degree of freedom induced by a motion in the j th degree of freedom.

Combining equations (6.3), (6.4) and (6.5), it follows that:

F3 (t) = −ω 2 A3 A33 sin(ωt) + ωA3 B33 cos(ωt) (6.6)

By using Fourier’s formula we can determine the coefficients that multiply the sine and
cosine functions and consequently define the added mass and radiation damping coeffi-
cients. Therefore, multiplying equation (6.6) by sin(ωt) and integrating over one period
T = 2π/ω, the coefficient A33 is calculated by:

t+ T2
Z
1
A33 =− F3 (t) sin(ωt)dt (6.7)
A3 πω
t− T2

Analogously, we multiply equation (6.6) by cos(ωt) and integrating over one period T =
104

2π/ω, the coefficient B33 is calculated by:

t+ T2
Z
1
B33 = F3 (t) cos(ωt)dt (6.8)
A3 π
t− T2

Numerically, the integral terms in equations (6.7) and (6.8) are calculated by the trape-
zoidal rule. This must be performed marching in time and consequently we define a time
series for the coefficients A33 and B33 which tends to assume a constant value if the hydro-
dynamic force is in steady state. In other words, we are applying a moving window Fourier
analysis of the temporal series with a window width equal to one oscillation period.

In order to illustrate it, Figure 6.14 presents the time series for the hydrodynamic force
F3 , for the added mass A33 and for the radiation damping B33 . Notice that the coefficient
curves present some oscillations at the beginning of the simulation, warning us that the
hydrodynamic forces are still in a transient period. Going further, we observe that after
a certain instant of time, the signals become practically constant, indicating that the
force is in steady state. After a constant behavior of the curve is observed, the added
mass and radiation damping coefficients are then determined by an average of its values.
The calculation of the added mass and radiation damping coefficients for other degrees of
freedom follows the same approach.
 105

5000
F3 (N )

−5000
0 5 10 15 20 25 30 35 40
Time (s)
2000
A33 (kg)

1000

0
0 5 10 15 20 25 30 35
Time (s)
1500
B33 (kg/s)

1000

500

0
0 5 10 15 20 25 30 35
Time (s)

Figure 6.14: Determination of A33 and B33 coefficients

The calculations of the added mass and radiation damping coefficients are then repeated
using the same angular frequencies described in Table 6.2. Convergence analysis of the
results are performed simulating each case for the coarse, medium and fine meshes pre-
sented in Table 6.1. Aiming at verifying the results, the coefficients obtained with the
hemisphere body are compared to the analytical solution presented by Hulme (1982),
whereas the ones for the circular cylinder body are compared to data calculated by the
software WAMIT. Again, these values are used as a reference for determination of relative
errors. The added mass and radiation damping coefficients Āij and B̄ij are normalized as
follows:

Hemisphere Circular Cylinder

Aij Aij
Āij = 2 3 Āij = for i,j ≤ 3
3
πrh ρ ρπrc2 Tc
Bij Bij
B̄ij = 2 3 B̄ij = for i,j ≤ 3
3
πrh ρω ρπrc2 Tc ω (6.9)
Aij
Āij = for i,j ≥ 4
ρπrc3 Tc
Bij
B̄ij = for i,j ≥ 4
ρπrc3 Tc ω
106

The non-dimensional added mass and radiation damping coefficients for heave and surge
modes of the hemisphere body are presented in Tables 6.8, 6.9, 6.10 and 6.11 with the
respective plots illustrated in Figures 6.15, 6.16, 6.17 and 6.18. Notice that the present
calculations agree very well with the analytical solutions derived by Hulme (1982) for all
the angular frequencies analyzed. In addition, differences between the results obtained
with each panel mesh are very small, pointing out that for this body geometry very low
computational costs are necessary for a reasonable numerical prediction of these coeffi-
cients.

The same behaviour is not observed in the circular cylinder results, presented in Tables
6.12, 6.13, 6.14 and 6.15 and displayed in the plots 6.19, 6.20, 6.21 and 6.22. Although a
fine agreement is observed for the medium and fine meshes, calculations performed with
the coarse mesh does not provide accurate results as, for example, the added mass coeffi-
cients for the heave mode presented in Figure 6.21. Looking for the pitch mode coefficients
Ā55 and B̄55 we conclude that even the fine mesh is insufficient to predict accurately the
same values calculated by the WAMIT higher order scheme. This may be justified by
the presence of an edge near the cylinder bottom which renders the convergence of the
Low-Order method much more difficult, specially because the meshes were constructed
concentrating the panels near the free surface and scattering them towards the bottom
edge of the cylinder.
 107

0.8
Hulme(1982) Coarse Medium Fine
0.7

0.6

0.5
Ā11

0.4

0.3

0.2

0.1
0 2 4 6 8 10
ω (rad/s)

Figure 6.15: Convergence analysis of the added mass coefficient Ā11 for the hemisphere
body

0.5
Hulme(1982) Coarse Medium Fine

0.4

0.3
B̄11

0.2

0.1

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.16: Convergence analysis of the radiation damping coefficient B̄11 for the hemi-
sphere body
108

Table 6.8: Convergence analysis of the non-dimensional added mass Ā11 for the hemisphere
body. Hulme (1982) values are used as reference for the relative errors
Coarse Medium Fine
|Ā11 −HULME| |Ā11 −HULME| |Ā11 −HULME|
ID ω (HULME, 1982) Ā11 HULME
Ā11 HULME
Ā11 HULME
1 1.716 0.585 0.567 0.030 0.577 0.014 0.581 0.006
2 1.981 0.618 0.600 0.028 0.608 0.016 0.612 0.009
3 2.215 0.644 0.627 0.027 0.635 0.014 0.638 0.009
4 2.426 0.659 0.642 0.025 0.653 0.008 0.656 0.005
5 2.620 0.658 0.645 0.021 0.653 0.008 0.657 0.002
6 2.801 0.642 0.633 0.014 0.639 0.004 0.643 0.001
7 2.971 0.613 0.606 0.011 0.612 0.001 0.615 0.004
8 3.132 0.574 0.573 0.002 0.576 0.003 0.578 0.007
9 3.431 0.486 0.490 0.009 0.491 0.009 0.491 0.011
10 3.706 0.404 0.410 0.016 0.408 0.011 0.409 0.012
11 3.962 0.337 0.344 0.020 0.343 0.018 0.343 0.019
12 4.202 0.287 0.290 0.013 0.292 0.020 0.293 0.021
13 4.429 0.249 0.252 0.009 0.254 0.019 0.255 0.021
14 4.952 0.195 0.193 0.011 0.199 0.018 0.200 0.023
15 5.425 0.172 0.168 0.022 0.174 0.013 0.175 0.019
16 5.860 0.163 0.156 0.045 0.164 0.007 0.166 0.013
17 6.264 0.162 0.156 0.036 0.162 0.002 0.163 0.009
18 6.644 0.164 0.158 0.037 0.164 0.003 0.165 0.004
19 7.004 0.168 0.162 0.036 0.167 0.006 0.168 0.002
20 7.672 0.177 0.170 0.038 0.175 0.010 0.177 0.002
21 8.287 0.187 0.179 0.039 0.184 0.015 0.185 0.006
22 8.859 0.195 0.186 0.047 0.192 0.017 0.194 0.007
23 9.396 0.202 0.193 0.043 0.198 0.020 0.200 0.010
24 9.905 0.209 0.197 0.056 0.204 0.023 0.206 0.012
 109

Table 6.9: Convergence analysis of the non-dimensional radiation damping B̄11 for the
hemisphere body. Hulme (1982) values are used as reference for the relative errors
Coarse Medium Fine
|B̄11 −HULME| |B̄11 −HULME| |B̄11 −HULME|
ID ω (HULME, 1982) B̄11 HULME
B̄11 HULME
B̄11 HULME
1 1.716 0.026 0.019 0.258 0.019 0.257 0.019 0.252
2 1.981 0.056 0.046 0.172 0.048 0.129 0.048 0.135
3 2.215 0.099 0.084 0.145 0.091 0.082 0.090 0.083
4 2.426 0.152 0.136 0.101 0.144 0.052 0.145 0.042
5 2.620 0.209 0.189 0.097 0.197 0.060 0.200 0.046
6 2.801 0.265 0.245 0.078 0.252 0.051 0.255 0.038
7 2.971 0.315 0.292 0.073 0.301 0.042 0.305 0.032
8 3.132 0.354 0.331 0.065 0.341 0.037 0.344 0.027
9 3.431 0.398 0.380 0.044 0.388 0.026 0.391 0.018
10 3.706 0.406 0.396 0.025 0.399 0.018 0.401 0.013
11 3.962 0.393 0.386 0.017 0.388 0.013 0.390 0.009
12 4.202 0.370 0.367 0.006 0.366 0.009 0.367 0.005
13 4.429 0.342 0.341 0.003 0.340 0.007 0.341 0.004
14 4.952 0.277 0.277 0.001 0.277 0.001 0.278 0.004
15 5.425 0.224 0.223 0.002 0.225 0.008 0.226 0.009
16 5.860 0.183 0.183 0.005 0.184 0.008 0.184 0.010
17 6.264 0.151 0.150 0.007 0.153 0.016 0.154 0.017
18 6.644 0.127 0.125 0.012 0.129 0.020 0.129 0.020
19 7.004 0.107 0.105 0.019 0.110 0.024 0.110 0.024
20 7.672 0.079 0.079 0.001 0.082 0.034 0.082 0.033
21 8.287 0.061 0.060 0.019 0.063 0.042 0.063 0.039
22 8.859 0.048 0.044 0.080 0.050 0.053 0.050 0.049
23 9.396 0.039 0.036 0.070 0.041 0.059 0.041 0.055
24 9.905 0.032 0.028 0.107 0.034 0.070 0.034 0.067
110

1
Hulme(1982) Coarse Medium Fine
0.9

0.8

0.7
Ā33

0.6

0.5

0.4

0 2 4 6 8 10
ω (rad/s)

Figure 6.17: Convergence analysis of the added mass coefficient Ā33 for the hemisphere
body

0.4
Hulme(1982) Coarse Medium Fine
0.35

0.3

0.25
B̄33

0.2

0.15

0.1

0.05

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.18: Convergence analysis of the radiation damping coefficient B̄33 for the hemi-
sphere body
 111

Table 6.10: Convergence analysis of the non-dimensional added mass Ā33 for the hemi-
sphere body. Hulme (1982) values are used as reference for the relative errors
Coarse Medium Fine
|Ā33 −HULME| |Ā33 −HULME| |Ā33 −HULME|
ID ω (HULME, 1982) Ā33 HULME
Ā33 HULME
Ā33 HULME
1 1.716 0.716 0.715 0.001 0.726 0.014 0.725 0.014
2 1.981 0.645 0.630 0.023 0.652 0.010 0.660 0.024
3 2.215 0.586 0.579 0.012 0.586 0.000 0.594 0.014
4 2.426 0.538 0.530 0.015 0.533 0.010 0.538 0.001
5 2.620 0.500 0.493 0.014 0.501 0.002 0.503 0.007
6 2.801 0.470 0.462 0.017 0.471 0.003 0.473 0.008
7 2.971 0.446 0.440 0.015 0.447 0.002 0.449 0.007
8 3.132 0.428 0.420 0.018 0.429 0.001 0.431 0.007
9 3.431 0.405 0.395 0.023 0.404 0.002 0.407 0.005
10 3.706 0.392 0.382 0.026 0.391 0.004 0.393 0.002
11 3.962 0.387 0.375 0.031 0.385 0.006 0.387 0.001
12 4.202 0.386 0.373 0.034 0.383 0.008 0.386 0.001
13 4.429 0.388 0.375 0.035 0.384 0.010 0.388 0.002
14 4.952 0.399 0.383 0.039 0.394 0.013 0.397 0.004
15 5.425 0.411 0.395 0.040 0.405 0.015 0.409 0.006
16 1 5.860 − − − − − − −
17 6.264 0.432 0.413 0.044 0.425 0.017 0.429 0.007
18 1 6.644 − − − − − − −
19 7.004 0.447 0.427 0.046 0.439 0.017 0.443 0.008
20 7.672 0.457 0.437 0.044 0.450 0.017 0.454 0.008
21 8.287 0.465 0.444 0.045 0.456 0.018 0.461 0.009
22 8.859 0.470 0.449 0.045 0.462 0.017 0.466 0.008
23 9.396 0.474 0.452 0.047 0.465 0.019 0.469 0.010
24 9.905 0.477 0.455 0.046 0.468 0.020 0.472 0.011

1
These values were not calculated since the analytical results were not presented in (HULME, 1982)
112

Table 6.11: Convergence analysis of the non-dimensional radiation damping B̄33 for the
hemisphere body. Hulme (1982) values are used as reference for the relative errors
Coarse Medium Fine
|B̄33 −HULME| |B̄33 −HULME| |B̄33 −HULME|
ID ω (HULME, 1982) B̄33 HULME
B̄33 HULME
B̄33 HULME
1 1.716 0.325 0.311 0.046 0.327 0.004 0.332 0.020
2 1.981 0.341 0.327 0.040 0.328 0.039 0.338 0.009
3 2.215 0.339 0.327 0.036 0.328 0.032 0.330 0.026
4 2.426 0.327 0.319 0.024 0.324 0.009 0.321 0.018
5 2.620 0.310 0.304 0.017 0.309 0.004 0.309 0.003
6 2.801 0.290 0.287 0.012 0.289 0.001 0.290 0.000
7 2.971 0.269 0.268 0.005 0.270 0.002 0.270 0.002
8 3.132 0.248 0.248 0.001 0.250 0.005 0.250 0.005
9 3.431 0.210 0.212 0.011 0.212 0.010 0.211 0.009
10 3.706 0.176 0.178 0.015 0.178 0.013 0.177 0.010
11 3.962 0.147 0.151 0.028 0.150 0.022 0.149 0.017
12 4.202 0.123 0.128 0.038 0.127 0.030 0.126 0.022
13 4.429 0.103 0.108 0.048 0.107 0.034 0.106 0.025
14 4.952 0.067 0.073 0.077 0.071 0.056 0.070 0.040
15 5.425 0.045 0.051 0.122 0.049 0.080 0.048 0.057
161 5.860 − − − − − − −
17 6.264 0.022 0.027 0.215 0.025 0.135 0.024 0.092
181 6.644 − − − − − − −
19 7.004 0.012 0.015 − 0.014 − 0.013 −
202 7.672 0.007 0.010 − 0.009 − 0.008 −
212 8.287 0.004 0.007 − 0.006 − 0.005 −
222 8.859 0.003 0.004 − 0.004 − 0.003 −
232 9.396 0.002 0.003 − 0.003 − 0.002 −
242 9.905 0.001 0.002 − 0.002 − 0.002 −

1
These values were not calculated since the analytical results were not presented in Hulme (1982)
2
The relative errors are omitted because the comparison involves very low values which any insignificant
differences result in very large relative errors
 113

0.9
Wamit Coarse Medium Fine
0.8

0.7

0.6
Ā11

0.5

0.4

0.3

0.2

0.1
0 2 4 6 8 10
ω (rad/s)

Figure 6.19: Convergence analysis of the added mass coefficient Ā11 for the circular cylin-
der body

Wamit Coarse Medium Fine

2

1.5
B̄11

0.5

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.20: Convergence analysis of the radiation damping coefficient B̄11 for the circular
cylinder body
114

Table 6.12: Convergence analysis of the non-dimensional added mass Ā11 for the circular
cylinder body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|Ā11 −WAMIT| |Ā11 −WAMIT| |Ā11 −WAMIT|
ID ω WAMIT Ā11 WAMIT
Ā11 WAMIT
Ā11 WAMIT
1 1.716 0.709 0.669 0.056 0.694 0.021 0.704 0.008
2 1.981 0.758 0.714 0.058 0.738 0.027 0.749 0.012
3 2.215 0.794 0.746 0.060 0.777 0.021 0.783 0.014
4 2.426 0.803 0.758 0.057 0.790 0.016 0.798 0.007
5 2.620 0.782 0.743 0.049 0.770 0.015 0.781 0.001
6 2.801 0.731 0.704 0.036 0.726 0.006 0.733 0.004
7 2.971 0.660 0.653 0.010 0.661 0.003 0.665 0.009
8 3.132 0.580 0.589 0.015 0.589 0.014 0.589 0.015
9 3.431 0.432 0.463 0.074 0.447 0.036 0.442 0.025
10 3.706 0.320 0.359 0.121 0.337 0.052 0.329 0.030
11 3.962 0.245 0.286 0.166 0.264 0.076 0.256 0.044
12 4.202 0.198 0.237 0.200 0.215 0.089 0.207 0.049
13 4.429 0.168 0.201 0.198 0.184 0.094 0.176 0.048
14 4.952 0.138 0.166 0.201 0.150 0.088 0.144 0.042
15 5.425 0.136 0.159 0.171 0.145 0.064 0.140 0.028
16 5.860 0.144 0.164 0.142 0.150 0.044 0.146 0.016
17 6.264 0.154 0.169 0.094 0.159 0.029 0.156 0.008
18 6.644 0.166 0.177 0.069 0.168 0.017 0.166 0.002
19 7.004 0.176 0.185 0.049 0.178 0.012 0.176 0.001
20 7.672 0.194 0.201 0.036 0.195 0.004 0.193 0.004
21 8.287 0.209 0.211 0.009 0.208 0.002 0.207 0.007
22 8.859 0.220 0.222 0.008 0.219 0.004 0.218 0.008
23 9.396 0.229 0.231 0.009 0.227 0.008 0.227 0.010
24 9.905 0.236 0.237 0.004 0.234 0.009 0.234 0.010
 115

Table 6.13: Convergence analysis of the non-dimensional radiation damping B̄11 for the
circular cylinder body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|B̄11 −WAMIT| |B̄11 −WAMIT| |B̄11 −WAMIT|
ID ω WAMIT B̄11 WAMIT
B̄11 WAMIT
B̄11 WAMIT
1 1.716 0.076 0.056 0.265 0.059 0.223 0.060 0.205
2 1.981 0.193 0.151 0.218 0.169 0.126 0.171 0.113
3 2.215 0.385 0.316 0.179 0.358 0.069 0.363 0.056
4 2.426 0.645 0.533 0.173 0.604 0.064 0.623 0.035
5 2.620 0.946 0.786 0.168 0.874 0.076 0.906 0.042
6 2.801 1.245 1.043 0.162 1.159 0.069 1.201 0.036
7 2.971 1.505 1.273 0.154 1.412 0.062 1.460 0.030
8 3.132 1.703 1.459 0.143 1.608 0.056 1.659 0.026
9 3.431 1.907 1.688 0.114 1.827 0.042 1.874 0.017
10 3.706 1.926 1.763 0.085 1.866 0.032 1.905 0.011
11 3.962 1.851 1.729 0.066 1.809 0.023 1.840 0.006
12 4.202 1.737 1.655 0.047 1.711 0.015 1.735 0.001
13 4.429 1.612 1.546 0.041 1.597 0.009 1.614 0.001
14 4.952 1.318 1.297 0.016 1.326 0.006 1.332 0.010
15 5.425 1.081 1.066 0.013 1.097 0.015 1.096 0.014
16 5.860 0.896 0.902 0.007 0.918 0.025 0.915 0.021
17 6.264 0.753 0.782 0.037 0.781 0.037 0.776 0.030
18 6.644 0.642 0.664 0.034 0.670 0.044 0.663 0.034
19 7.004 0.553 0.582 0.052 0.582 0.053 0.575 0.040
20 7.672 0.425 0.458 0.079 0.453 0.066 0.446 0.050
21 8.287 0.338 0.350 0.035 0.364 0.075 0.357 0.055
22 8.859 0.277 0.279 0.008 0.302 0.092 0.295 0.066
23 9.396 0.232 0.216 0.071 0.255 0.096 0.249 0.070
24 9.905 0.198 0.186 0.060 0.218 0.098 0.214 0.079
116

0.75
Wamit Coarse Medium Fine

0.7

0.65
Ā33

0.6

0.55

0.5
0 2 4 6 8 10
ω (rad/s)

Figure 6.21: Convergence analysis of the added mass coefficient Ā33 for the circular cylin-
der body

0.35
Wamit Coarse Medium Fine
0.3

0.25

0.2
B̄33

0.15

0.1

0.05

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.22: Convergence analysis of the radiation damping coefficient B̄33 for the circular
cylinder body
 117

Table 6.14: Convergence analysis of the non-dimensional added mass Ā33 for the circular
cylinder body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|Ā33 −WAMIT| |Ā33 −WAMIT| |Ā33 −WAMIT|
ID ω WAMIT Ā33 WAMIT
Ā33 WAMIT
Ā33 WAMIT
1 1.716 0.623 0.611 0.020 0.630 0.012 0.630 0.011
2 1.981 0.584 0.568 0.028 0.592 0.013 0.594 0.017
3 2.215 0.557 0.545 0.020 0.562 0.009 0.560 0.005
4 2.426 0.539 0.530 0.017 0.538 0.002 0.542 0.006
5 2.620 0.529 0.520 0.016 0.533 0.008 0.532 0.007
6 2.801 0.523 0.515 0.016 0.527 0.007 0.526 0.006
7 2.971 0.522 0.513 0.016 0.526 0.008 0.525 0.006
8 3.132 0.522 0.514 0.017 0.526 0.007 0.525 0.006
9 3.431 0.527 0.518 0.019 0.531 0.006 0.530 0.005
10 3.706 0.534 0.523 0.020 0.537 0.005 0.537 0.004
11 3.962 0.541 0.529 0.023 0.543 0.003 0.543 0.003
12 4.202 0.547 0.534 0.024 0.549 0.003 0.549 0.003
13 4.429 0.552 0.539 0.025 0.554 0.003 0.554 0.002
14 4.952 0.561 0.547 0.026 0.563 0.002 0.563 0.002
15 5.425 0.567 0.552 0.026 0.568 0.002 0.568 0.003
16 5.860 0.571 0.555 0.028 0.571 0.001 0.571 0.001
17 6.264 0.573 0.557 0.028 0.574 0.001 0.574 0.002
18 6.644 0.575 0.558 0.029 0.575 0.000 0.575 0.001
19 7.004 0.576 0.560 0.028 0.577 0.001 0.577 0.002
20 7.672 0.578 0.562 0.028 0.579 0.001 0.579 0.002
21 8.287 0.579 0.562 0.029 0.579 0.000 0.580 0.001
22 8.859 0.580 0.564 0.028 0.581 0.001 0.581 0.001
23 9.396 0.581 0.563 0.030 0.580 0.001 0.581 0.000
24 9.905 0.581 0.564 0.030 0.581 0.001 0.581 0.000
118

Table 6.15: Convergence analysis of the non-dimensional radiation damping B̄33 for the
circular cylinder body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|B̄33 −WAMIT| |B̄33 −WAMIT| |B̄33 −WAMIT|
ID ω WAMIT B̄33 WAMIT
B̄33 WAMIT
B̄33 WAMIT
1 1.716 0.275 0.246 0.105 0.274 0.004 0.282 0.026
2 1.981 0.298 0.257 0.138 0.285 0.046 0.296 0.007
3 2.215 0.296 0.267 0.097 0.276 0.068 0.294 0.006
4 2.426 0.278 0.252 0.093 0.270 0.027 0.277 0.002
5 2.620 0.251 0.228 0.091 0.245 0.022 0.251 0.000
6 2.801 0.221 0.202 0.084 0.217 0.015 0.222 0.005
7 2.971 0.191 0.175 0.079 0.188 0.013 0.192 0.009
8 3.132 0.162 0.150 0.071 0.161 0.007 0.164 0.014
9 3.431 0.114 0.108 0.047 0.114 0.007 0.116 0.024
10 3.706 0.078 0.075 0.038 0.079 0.019 0.080 0.034
11 3.962 0.052 0.054 0.033 0.055 0.050 0.055 0.060
121 4.202 0.035 0.037 − 0.038 − 0.038 −
131 4.429 0.023 0.026 − 0.026 − 0.026 −
141 4.952 0.008 0.011 − 0.011 − 0.010 −
151 5.425 0.003 0.006 − 0.005 − 0.005 −
161 5.860 0.001 0.003 − 0.003 − 0.003 −
171 6.264 0.000 0.002 − 0.002 − 0.002 −
181 6.644 0.000 0.002 − 0.002 − 0.002 −
191 7.004 0.000 0.001 − 0.001 − 0.001 −
201 7.672 0.000 0.001 − 0.001 − 0.001 −
211 8.287 0.000 0.001 − 0.001 − 0.001 −
221 8.859 0.000 0.001 − 0.001 − 0.001 −
231 9.396 0.000 0.000 − 0.001 − 0.001 −
241 9.905 0.000 0.000 − 0.000 − 0.001 −

1
The relative errors are omitted because the comparison involves very low values for which any in-
significant differences result in very large relative errors
 119

0.2
Wamit Coarse Medium Fine

0.18

0.16
Ā55

0.14

0.12

0.1
0 2 4 6 8 10
ω (rad/s)

Figure 6.23: Convergence analysis of the added mass coefficient Ā55 for the circular cylin-
der body

0.25
Wamit Coarse Medium Fine

0.2

0.15
B̄55

0.1

0.05

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.24: Convergence analysis of the radiation damping coefficient B̄55 for the circular
cylinder body
120

Table 6.16: Convergence analysis of the non-dimensional added mass Ā55 for the circular
cylinder body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|Ā55 −WAMIT| |Ā55 −WAMIT| |Ā55 −WAMIT|
ID ω WAMIT Ā55 WAMIT
Ā55 WAMIT
Ā55 WAMIT
1 1.716 0.166 0.177 0.064 0.174 0.045 0.170 0.020
2 1.981 0.171 0.182 0.064 0.179 0.044 0.175 0.019
3 2.215 0.175 0.186 0.066 0.183 0.044 0.178 0.019
4 2.426 0.176 0.188 0.069 0.184 0.048 0.179 0.021
5 2.620 0.173 0.186 0.073 0.182 0.049 0.177 0.023
6 2.801 0.168 0.181 0.078 0.177 0.053 0.172 0.025
7 2.971 0.160 0.174 0.088 0.169 0.057 0.165 0.028
8 3.132 0.152 0.166 0.094 0.161 0.061 0.157 0.031
9 3.431 0.136 0.150 0.104 0.145 0.066 0.141 0.033
10 3.706 0.125 0.137 0.102 0.133 0.065 0.129 0.032
11 3.962 0.117 0.128 0.093 0.125 0.065 0.121 0.033
12 4.202 0.113 0.122 0.084 0.120 0.062 0.116 0.032
13 4.429 0.110 0.118 0.074 0.117 0.059 0.113 0.029
14 4.952 0.109 0.115 0.057 0.114 0.052 0.112 0.027
15 5.425 0.110 0.115 0.044 0.115 0.047 0.113 0.023
16 5.860 0.112 0.116 0.036 0.117 0.043 0.115 0.021
17 6.264 0.114 0.118 0.035 0.119 0.041 0.117 0.020
18 6.644 0.116 0.120 0.033 0.121 0.040 0.118 0.019
19 7.004 0.118 0.122 0.033 0.123 0.040 0.120 0.019
20 7.672 0.120 0.124 0.034 0.125 0.040 0.123 0.019
21 8.287 0.122 0.126 0.034 0.127 0.040 0.124 0.018
22 8.859 0.123 0.128 0.036 0.128 0.041 0.125 0.019
23 9.396 0.124 0.128 0.033 0.129 0.039 0.126 0.017
24 9.905 0.125 0.129 0.038 0.129 0.038 0.127 0.016
 121

Table 6.17: Convergence analysis of the non-dimensional radiation damping B̄55 for the
circular cylinder body. WAMIT values are used as reference for the relative errors
Coarse Medium Fine
|B̄55 −WAMIT| |B̄55 −WAMIT| |B̄55 −WAMIT|
ID ω WAMIT B̄55 WAMIT
B̄55 WAMIT
B̄55 WAMIT
1 1.716 0.007 0.006 0.108 0.006 0.163 0.006 0.179
2 1.981 0.019 0.017 0.067 0.018 0.064 0.017 0.094
3 2.215 0.038 0.036 0.039 0.037 0.028 0.036 0.054
4 2.426 0.064 0.064 0.001 0.064 0.005 0.063 0.013
5 2.620 0.095 0.095 0.002 0.094 0.007 0.093 0.018
6 2.801 0.125 0.126 0.011 0.125 0.000 0.123 0.013
7 2.971 0.151 0.154 0.022 0.152 0.009 0.150 0.007
8 3.132 0.170 0.177 0.039 0.173 0.018 0.170 0.002
9 3.431 0.188 0.205 0.091 0.195 0.036 0.190 0.010
10 3.706 0.186 0.211 0.136 0.196 0.052 0.189 0.018
11 3.962 0.174 0.203 0.171 0.185 0.066 0.178 0.026
12 4.202 0.157 0.190 0.207 0.170 0.081 0.162 0.033
13 4.429 0.140 0.174 0.247 0.153 0.092 0.145 0.038
14 4.952 0.101 0.134 0.323 0.114 0.129 0.107 0.059
15 5.425 0.072 0.105 0.455 0.084 0.168 0.078 0.080
16 5.860 0.051 0.078 0.527 0.062 0.203 0.056 0.097
17 6.264 0.037 0.059 0.607 0.046 0.254 0.042 0.126
18 6.644 0.027 0.048 0.797 0.035 0.300 0.031 0.150
19 7.004 0.020 0.039 0.987 0.027 0.352 0.023 0.179
20 7.672 0.011 0.022 1.009 0.016 0.468 0.014 0.247
211 8.287 − 0.017 − 0.011 − 0.009 −
221 8.859 − 0.012 − 0.008 − 0.006 −
231 9.396 − 0.007 − 0.006 − 0.004 −
241 9.905 − 0.005 − 0.004 − 0.003 −

6.4 Free Floating Simulations

The results presented so far still did not involve calculations considering the body equa-
tions of motions coupled with the integral equations derived from the hydrodynamic prob-
lem and, therefore, they cannot be used to evaluate whether the coupling scheme, pre-
viously proposed by van Daalen (1993) and Tanizawa (1995), is implemented correctly
or not. With this objective, next we apply our numerical code for the evaluation of free
motions of the hemisphere and cylinder bodies. The calculations are performed with the
panel meshes presented in Table 6.1. In the simulations, the incident regular waves with
1
The relative errors are omitted because the comparison involves very low values which any insignificant
differences result in very large relative errors
122

unitary amplitude AI = 1 m propagate in the x positive direction with the angular fre-
quencies presented in Table 6.2. The time-step and the numerical beach zone coefficients
were set with the same values used for the fixed and forced motion simulations.

For the calculation of the body motions in waves, we had also to define matrices of
mass/inertias for each one of the geometries evaluated. Moreover, for the calculations
involving the cylinder, in which the pitch D.O.F was also evaluated, an arbitrary lin-
ear external damping coefficient C55 was also considered. This external coefficient was
included, since the pitch motion of the cylinder presents very low damping by wave ra-
diation, which renders the time convergence of the method in the resonance frequency
very difficult. For comparison purposes, the same external damping value was applied in
the WAMIT model. The main parameters considered in the simulations are presented in
Table 6.18.

Typical time histories of the heave motions of the cylinder and hemisphere are exemplified
in Figures 6.25 and 6.26, respectively. These simulations were carried out with incoming
waves with frequencies equal to the heave natural frequencies of each body. For a bet-
ter visualization of the curves only a part of the steady-state portion of the signals are
presented. It is worth mentioning that the time histories presented a regular behaviour
even simulating the body responses for more than a hundred wave cycles, attesting the
stability of the code. In fact, the simulation could be continued for much longer, without
compromising the quality of the results.

Table 6.18: Main characteristics of the cylinder and hemisphere and simulation settings
Parameters Cylinder Hemisphere
Radius (m) 1.00 1.00
Draught (m) 1.00 1.00
Mass (kg) 3.14E+3 2.09E+3
Pitch Inertia1 (kg.m2 ) 1.57E+3 −
COG X coordinate2 (m) 0.00 0.00
COG Y coordinate2 (m) 0.00 0.00
COG Z coordinate2 (m) -0.50 -0.50
C55 (kg.m2 /s) 4.35E+2 −

1
Values were calculated in relation to the body center of gravity
2
Values are described in relation to the global coordinate system (x,y,z)=(0,0,0)
 123

Coarse Medium Fine

6

2
ξ3 (m)

−2

−4

−6
140 142 144 146 148 150
Time (s)

Figure 6.25: Time history of the cylinder heave motion for the coarse, medium and fine
meshes. Incoming wave with angular frequency ω = 2.51 rad/s

2.5
Coarse Medium Fine
2

1.5

1
0.5
ξ3 (m)

0
−0.5

−1

−1.5

−2
112 114 116 118 120
Time (s)

Figure 6.26: Time history of the hemisphere heave motion for the coarse, medium and
fine meshes. Incoming wave with angular frequency ω = 3.13 rad/s
124

The convergence of the bodies motions results for each wave frequency described in Table
6.2 is evaluated next. Furthermore, the results are compared with data provided by the
software WAMIT. This evaluation is done in frequency domain by calculating the motions
RAOs taking the rms (root mean squared) of a steady-state portion of each motion time
series.

The cylinder RAOS of surge, heave and pitch motions obtained with the present method
and WAMIT are plotted in Figures 6.27, 6.28 and 6.29, respectively. Overall, a good
agreement is observed with the WAMIT results. As can be seen, the results converge with
the increasing number of panels, which is attested by the very similar results obtained
with the medium and fine meshes. Notice, however, that by applying the coarse mesh, the
curves tend to present an oscillatory behaviour, which is intensified with the increasing
wave frequency. As discussed before, this occurs due to the low resolution of panels per
wave length that causes the appearance of an amplitude modulation in the signal.

The analogous results for the hemisphere body are presented in Figures 6.30 and 6.31 for
surge and heave motions, respectively. In general, the same conclusions pointed out for
the cylinder are maintained, in which the medium and fine mesh results agreed very well
with the WAMIT data. Furthermore, these results demonstrate the capability of our code
to predict the motions of floating bodies under no forward speed in waves, confirming that
the second integral equation defined for the acceleration potential is correctly implemented
and the equilibrium between the dynamic and hydrodynamic forces was conserved during
the whole simulation.
 125

1
Wamit Coarse Medium Fine

0.8

0.6
ξ¯1 /AI

0.4

0.2

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.27: Convergence analysis of the cylinder surge RAO and comparison with
WAMIT data

6
Wamit Coarse Medium Fine

4
ξ¯3 /AI

0
0 2 4 6 8 10
ω (rad/s)

Figure 6.28: Convergence analysis of the cylinder heave RAO and comparison with
WAMIT data
126

2.5
Wamit Coarse Medium Fine

2
ξ¯5 /AI (o /m)

1.5

0.5

0
0 2 4 6 8 10 12
ω (rad/s)

Figure 6.29: Convergence analysis of the cylinder pitch RAO and comparison with
WAMIT data

1
Wamit Coarse Medium Fine

0.8

0.6
ξ¯1 /AI

0.4

0.2

0
0 2 4 6 8 10 12
ω (rad/s)

Figure 6.30: Convergence analysis of the sphere surge RAO and comparison with WAMIT
data
 127

2
Wamit Coarse Medium Fine

1.5
ξ¯3 /AI

0.5

0
0 2 4 6 8 10 12
ω (rad/s)

Figure 6.31: Convergence analysis of the sphere heave RAO and comparison with WAMIT
data
128
 129

Chapter 7

Verification Tests with Free Floating

FPSOs

Floating Production Storage and Offloading (FPSO) units are floating vessels widely used
by the offshore oil and gas industry, being also considered as the principal exploration
solution for the recent oil discoveries in the pre-salt layer, in Brazil. One of the reasons
for that is related to positive features of this kind of platform, such as the possibility of
bypassing the construction stages using the hull of a converted oil tanker vessel, the large
space to allocate the process plant and also its capability to store considerable amounts
of oil which avoids the laying of extensive long distance pipelines from the oil well to an
onshore terminal.

Designed to operate with forward speed, an anchored ship hull may suffer with the inci-
dence of waves, presenting large motions and accelerations. For beam seas, for example,
even a moderate sea state condition may imply in severe resonant roll motion, which are
further aggravated by the very low energy dissipation by wave generation. In a tentative
to avoid this inconvenience, the FPSO may be anchored in a single point mooring arrange-
ment (SPM) allowing the vessel to rotate freely to best respond to weather conditions or
in a spread mooring system (SMS) which is designed to maintain an appropriate vessel
heading for the most severe waves, facilitating the offloading operations.

Independently on the mooring arrangement, several aspects must be assessed during the
design stages. Specifically on the hydrodynamic task, extensive studies of topics such
as seakeeping, definition of heading (only for SMS) and offloading operations must be
performed, requiring both the conduction of experimental tests and numerical predictions
using reliable and validated tools.
130

In this sense, we test our numerical model for two FPSO hull types aiming at verifying its
capability to evaluate the free floating motions of real hull shapes in waves. The first is a
349,000 m3 FPSO “box shaped” hull, in loaded draught, which is constructed intentionally
for oil and gas exploration. The second is a 311,110 m3 Very Large Crude Carrier (VLCC)
vessel converted to perform the tasks of an FPSO. In the former case, the results of our
code is confronted with the ones provided by WAMIT for three wave incidence angles and a
wide range of frequencies. For the latter, the results are compared to experimental results
previously conducted at the Hydrodynamic Calibrator of the Numerical Offshore Tank of
the University of Sao Paulo (CH-TPN-USP). This experiment were conceptualized in the
context of another research project that aimed at studying the FPSO roll motion and,
therefore, our comparisons are restricted to beam waves only.

7.1 FPSO “Box Shaped” Hull

The main characteristics of the FPSO “Box Shaped” hull and the settings of the simu-
lations are listed in Table 7.1. It can be observed that we have included external linear
restoring coefficients for the surge, sway and yaw D.O.F. aiming at maintaining the so-
lutions with zero mean. Although our numerical code does not calculate drift forces,
this was necessary since a slight non-symmetrical topology of the panel mesh induces
non-physical numerical drifts, as was observed during preliminary simulation tests. In
addition, external damping coefficients associated to these D.O.F. were simply tuned to
avoid long transient periods. Another remark is the inclusion of an empirical linear ex-
ternal roll damping coefficient of 5% of its critical value, so as to avoid very large roll
motions and provide a faster numerical convergence, since our numerical code does not
consider viscous effects.
 131

Table 7.1: Principal characteristics of the FPSO “Box Shaped” hull and the settings of
the simulations
Length over all 306.00 m
Beam moulded 54.00 m
Depth moulded 31.50 m
Draught 23.20 m
Mass 3.49E+08 kg
Roll Inertia1 1.16E+11 kg.m2
Pitch Inertia1 2.09E+12 kg.m2
Yaw Inertia 1 2.12E+12 kg.m2
COG X coordinate2 0.93 m
COG Y coordinate2 0.00 m
COG Z coordinate2 -5.14 m
K11 1.00E+05 kg/s2
K22 1.00E+05 kg/s2
K66 1.00E+09 kg/s2
C11 1.00E+07 kg/s
C22 1.00E+07 kg/s
C44 6.03E+09 kg.m/s
C66 6.00E+09 kg.m/s

Calculations were done for wave incidences of 0◦ , 45◦ and 90◦ , which represents stern,
stern-quartering and beam waves, respectively. The simulations were conducted consid-
ering two distinct approaches. In the first one, the motions of the FPSO are calculated
for a set of 13 regular waves, whose main characteristics are presented in Table 7.2. In
the second approach, we apply a white noise spectrum, in which not only one wave, but a
simultaneous package of different wave components with the same amplitude is taken into
account. In this work, we considered a discrete number of wave frequencies in the range
0.18-0.9 rad/s with an interval of 0.005 rad/s. As will be presented ahead, through the
use of this procedure we may obtain the RAOs of the dynamic system performing only
a single code run for each incidence angle, which is much more efficient than using the
standard procedure of choosing several regular waves.
1
Values were calculated in relation to the body center of gravity
2
Values are described in relation to the global coordinate system (x,y,z)=(0,0,0)
132

Table 7.2: Regular incoming waves used in the FPSO “Box Shaped” hull simulations
ID ω (rad/s) T(s) λ(m)
1 0.898 7.000 76.504
2 0.698 9.000 126.466
3 0.571 11.000 188.919
4 0.483 13.000 263.861
5 0.419 15.000 351.295
6 0.370 17.000 451.219
7 0.331 19.000 563.633
8 0.299 21.000 688.538
9 0.273 23.000 825.933
10 0.251 25.000 975.819
11 0.233 27.000 1138.195
12 0.217 29.000 1313.062
13 0.203 31.000 1500.419

Figures 7.1 and 7.2 display the FPSO body and the free surface meshes used in the
simulations. Convergence was reached using 1764 panels on the body surface and 5000
panels on the free surface. For the analysis considering regular waves, the time-step was
set to ∆t = T /60 s and for the transient wave package ∆t = min(T )/60, in which min(T )
is the lowest period within the set of wave components.

Figure 7.1: FPSO “Box Shaped” panel mesh

 133

Figure 7.2: Free surface panel mesh for the FPSO “Box Shaped” simulations

We first present, in Figure 7.3, nine snapshots of the free surface elevation around the
FPSO at different instants of the simulation, considering an incoming stern regular wave
of angular frequency ω = 0.370 rad/s. The amplitude of the incoming regular wave is
scaled so as to better visualize the disturbed wave patterns. One should realize that
during the first 100 seconds, approximately, the free surface presents a transient behavior
in which we cannot recognize a clear wave pattern around the hull, which is then followed
by a steady state wave pattern in which the incoming waves are less disturbed by the
FPSO.

The sequence of events described above may also be observed by the FPSO motion time
series, presented in Figures 7.4, 7.5 and 7.6 where we clearly realize a transient period
during the first wave cycles. Moreover, we also realize that the motion in surge presents
a longer transient due to the additional external restoring.

The group of disturbed waves generated during the transient period must be dissipated by
the damping zone in order to ensure that the solution reaches a steady state. Therefore,
a precise tuning of the damping zone parameters considering the characteristics of the
regular waves in analysis must be assessed. This was performed by preliminary test runs
that led to a numerical beach of two wave lengths long (b = 2.0) and intensity a = 1.0.
134

(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

Figure 7.3: Snapshots of the simulation at 9 different instants of time. Wave Amplitude
AI = 1 m and frequency ω = 0.370 rad/s. (a) t = 0 s, (b) t = 20 s,(c) t = 40 s, (d) t = 60
s, (e) t = 80 s, (f) t = 100 s, (g) t = 120 s, (h) t = 140 s and (i) t = 160 s
 135

0.4

0.2
ξ1 (m)

−0.2

−0.4

0 100 200 300 400

Time (s)

Figure 7.4: Typical surge motion time series obtained by the present numerical code.
Motions were induced by a wave of angular frequency ω = 0.370 rad/s and incidence
angle 0◦

0.6

0.4

0.2
ξ3 (m)

−0.2

−0.4

−0.6
0 100 200 300 400
Time (s)

Figure 7.5: Typical heave motion time series obtained by the present numerical code.
Motions were induced by a wave of angular frequency ω = 0.370 rad/s and incidence
angle 0◦
136

0.015

0.01

0.005
ξ5 (rad)

−0.005

−0.01

−0.015
0 100 200 300 400
Time (s)

Figure 7.6: Typical pitch motion time series obtained by the present numerical code.
Motions were induced by a wave of angular frequency ω = 0.370 rad/s and incidence
angle 0◦

Next, we present examples of output time series obtained through the simulations using
the white noise spectrum (Swn ), presented in Figure 7.7, for a wave incidence angle of
0◦ . Figure 7.8 presents the undisturbed incoming wave time history at the position of the
FPSO “Box Shaped” CoG, which is obtained from the realization of the white noise energy
spectrum into a time series representation. One should realize that at a specific time a
constructive wave is formed exactly at the body position, resulting in a very large free
surface elevation that occurs by the fact that the time series was generated considering all
the wave components with zero phase values. The correspondent motion time histories of
surge, heave and pitch are presented in Figure 7.9.
 137

80

70

60

50
Swn (m2 s)

40

30

20

10

0
0.2 0.4 0.6 0.8 1
ω (rad/s)

Figure 7.7: Numerical white noise spectrum

150

100

50
ηI (m)

−50

−100

−150
0 500 1000 1500 2000 2500 3000 3500
t(s)

Figure 7.8: Incoming wave elevation time series generated by a white noise spectrum
138

50
ξ1 (m)

−50
0 500 1000 1500 2000 2500 3000 3500
t (s)
50
ξ3 (m)

−50
0 500 1000 1500 2000 2500 3000 3500
t (s)
50
ξ4 ( ◦ )

−50
0 500 1000 1500 2000 2500 3000 3500
t (s)

Figure 7.9: Motion time series for sway (ξ2 ), heave (ξ3 ) and roll (ξ4 ) induced by a wave
generated by a white noise spectrum

In order to verify the results, we compare the present values with RAOs obtained with
the frequency domain software WAMIT. For regular waves, this is done by taking the root
mean square (rms) of a steady-state portion of each motion time series. For the analysis
of the white-noise waves a more sophisticated approach is applied using the cross power
spectral density (cpsd)1 between the motions and the incoming wave time series. The
results are compared by the RAOs for each D.O.F and incoming wave incidence angles,
and are presented in Figures 7.10, 7.11, 7.12, 7.13, 7.14 and 7.15 for surge, sway, heave,
roll, pitch and yaw D.O.F., respectively.

As can be seen, except for some small deviations, specially for values near the peaks,
a very good agreement between both codes is observed for all D.O.F. and wave angles
analyzed, demonstrating, therefore, the capability of our code to predict the motions of a
floating body under no forward speed with more realistic geometries than the hemisphere
and cylinder ones. Moreover, the use of the white noise spectrum for the determination
of the RAOs also presents a very good agreement with WAMIT with the advantage of
rendering the calculations much more efficient, since only one code run per wave incidence
angle must be executed. These results also point out that standard sea spectra, such as
1
Analysis was performed with the cpsd function available in MATLAB version 7.10.0
 139

JONSWAP and Pierson-Moskowitz, may also be applied for the evaluation of motions in
sea conditions, which may be performed simply by changing the white noise spectrum
for a sea one. In this case, however, the realization in time of the sea spectra is often
performed considering random phases between the frequencies.

1
Present values 0◦ (Regular Wave)
Present values 45◦ (Regular Wave)
0.8 Present values 90◦ (Regular Wave)
ξ̄1 /AI (m/m)

Wamit 0◦ ◦
0.6 Wamit 45◦
Wamit 90
0.4

0.2

0
0 0.5 1 1.5
ω (rad/s)

1
Present values 0◦ (White Noise)
Present values 45◦ (White Noise)
0.8 Present values 90◦ (White Noise)
ξ̄1 /AI (m/m)

Wamit 0◦ ◦
0.6 Wamit 45◦
Wamit 90
0.4

0.2

0
0 0.5 1 1.5
ω (rad/s)

Figure 7.10: Surge response amplitude operators considering three different wave incidence
angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and White Noise Analysis (Below)
140

1
Present values 0◦ (Regular Wave)
Present values 45◦ (Regular Wave)
0.8 Present values 90◦ (Regular Wave)
ξ¯2 /AI (m/m)

Wamit 0◦ ◦
0.6 Wamit 45◦
Wamit 90
0.4

0.2

0
0 0.5 1 1.5
ω (rad/s)

1
Present values 0◦ (White Noise)
Present values 45◦ (White Noise)
0.8 Present values 90◦ (White Noise)
ξ¯2 /AI (m/m)

Wamit 0◦ ◦
0.6 Wamit 45◦
Wamit 90
0.4

0.2

0
0 0.5 1 1.5
ω (rad/s)

Figure 7.11: Sway response amplitude operators considering three different wave incidence
angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and White Noise Analysis (Below)

2.5
Present values 0◦ (Regular Wave)
Present values 45◦ (Regular Wave)
2 Present values 90◦ (Regular Wave)
ξ¯3 /AI (m/m)

Wamit 0◦ ◦
1.5 Wamit 45◦
Wamit 90
1

0.5

0
0 0.5 1 1.5
ω (rad/s)

2.5
Present values 0◦ (White Noise)
Present values 45◦ (White Noise)
2 Present values 90◦ (White Noise)
ξ¯3 /AI (m/m)

Wamit 0◦ ◦
1.5 Wamit 45◦
Wamit 90
1

0.5

0
0 0.5 1 1.5
ω (rad/s)

Figure 7.12: Heave response amplitude operators considering three different wave inci-
dence angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and White Noise Analysis
(Below)
 141

3
Present values 0◦ (Regular Wave)
Present values 45◦ (Regular Wave)
Present values 90◦ (Regular Wave)

ξ¯4 /AI (◦ /m)

2 Wamit 0◦ ◦
Wamit 45◦
Wamit 90

0
0 0.5 1 1.5
ω (rad/s)

3
Present values 0◦ (White Noise)
Present values 45◦ (White Noise)
Present values 90◦ (White Noise)
ξ¯4 /AI (◦ /m)

2 Wamit 0◦ ◦
Wamit 45◦
Wamit 90

0
0 0.5 1 1.5
ω (rad/s)

Figure 7.13: Roll response amplitude operators considering three different wave incidence
angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and White Noise Analysis (Below)

1
Present values 0◦ (Regular Wave)
Present values 45◦ (Regular Wave)
0.8 Present values 90◦ (Regular Wave)
ξ¯5 /AI (◦ /m)

Wamit 0◦ ◦
0.6 Wamit 45◦
Wamit 90
0.4

0.2

0
0 0.5 1 1.5
ω (rad/s)

1
Present values 0◦ (White Noise)
Present values 45◦ (White Noise)
0.8 Present values 90◦ (White Noise)
ξ¯5 /AI (◦ /m)

Wamit 0◦ ◦
0.6 Wamit 45◦
Wamit 90
0.4

0.2

0
0 0.5 1 1.5
ω (rad/s)

Figure 7.14: Pitch response amplitude operators considering three different wave incidence
angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and White Noise Analysis (Below)
142

0.4
Present values 0◦ (Regular Wave)
Present values 45◦ (Regular Wave)
0.3 Present values 90◦ (Regular Wave)
ξ¯6 /AI (◦ /m)

Wamit 0◦ ◦
Wamit 45◦
0.2 Wamit 90

0.1

0
0 0.5 1 1.5
ω (rad/s)

0.4
Present values 0◦ (White Noise)
Present values 45◦ (White Noise)
0.3 Present values 90◦ (White Noise)
ξ¯6 /AI (◦ /m)

Wamit 0◦ ◦
Wamit 45◦
0.2 Wamit 90

0.1

0
0 0.5 1 1.5
ω (rad/s)

Figure 7.15: Yaw response amplitude operators considering three different wave incidence
angle 0◦ , 45◦ , 90◦ . Regular Wave Analysis (Above) and White Noise Analysis (Below)

7.2 FPSO VLCC Hull

Numerical simulations were also conducted for a converted FPSO VLCC hull type. The
results are compared to experimental data previously obtained in wave tests carried out
at the CH-TPN-USP, in Sao Paulo, Brazil, which is a wave basin with dimensions of 14
m×14 m and water depth of 4 m (MELLO et al., 2010). This tank is equipped with a set
of 148 independent flaps for the generation and active absorption of waves, which provides
good precision and stability for the wave field during the tests. A perspective view of the
wave basin is presented in Figure 7.16.
 143

Figure 7.16: Hydrodynamic Calibrator of the Numerical Offshore Tank at the University
of Sao Paulo (CH-TPN-USP)

7.2.1 Experimental Setup

For the execution of the tests, a 1 : 90 small-scale model of the VLCC hull was positioned
at the center of tank and equipped with reflective targets used by an optical tracking
system which was mounted on the carriage for the measurement of 6 D.O.F motions of
the model. In addition, a set of four soft springs were attached to the model and fixed to
four vertical bars positioned at the corners of the wave basin in order to restrain its drift.
The equivalent restoring coefficients in the horizontal plane were determined by pull-out
tests. As the main goal of the tests was the study of roll motions, only beam waves were
considered in the test matrix, which included regular, transient and sea wave conditions.
Figure 7.17 presents a view of the model during the tests.

Concerning the regular waves, a study increasing the incoming wave amplitude for a fixed
wave frequency ωI = 0.443 rad/s (near to the roll resonance frequency) was conducted
aiming at investigating the nonlinear roll motions response and its associated damping
coefficients. For this study, however, only the regular wave of smallest steepness will be
considered, being in accordance with the linear wave theory here adopted. The transient
wave, by its turn, was used for a fast RAO determination, in which a wave package with
a constant amplitude of 3 meters and frequencies in between 0.349 rad/s and 0.785 rad/s,
in full scale, was applied. In this wave package, the maximum wave steepness is around
3%, being also within the scope of the linear wave theory.

It is important to mention that, in this study, irregular waves were not used for compar-
ison purposes. Although this is not being evaluated, the simulations with irregular seas
would be carried out in a similar manner as those conducted for the white-noise spectrum
144

presented in section 7.1, which is a good indication that the code is capable to properly
conduct this kind of computation with reasonable accuracy. Indeed, since the problem is
being treated based on the linear theory, the numerical analysis of irregular waves is merely
treated as a superposition of different regular wave components of distinct amplitudes,
frequencies and phases. In practice, this is performed by simply changing the white-noise
spectrum, applied in section 7.1, for a representative sea-wave energy spectrum, such as
JONSWAP, Pierson-Moskowitz etc.

Figure 7.17: General view of the FPSO VLCC model positioned in CH-TPN-USP

The main characteristics of the FPSO model (in full scale) are listed in Table 7.3. Model
center of gravity and inertia of roll, pitch and yaw were calibrated using ballast weights.

Table 7.3: Main characteristics of the FPSO VLCC hull and the settings of the simulations
Length over all 334.44 m
Beam moulded 54.72 m
Depth moulded 21.51 m
Draught 21.51 m
Mass 3.09E+08 kg
Roll Inertia1 1.05+11 kg.m2
Pitch Inertia1 1.79E+12 kg.m2
Yaw Inertia 1 1.83E+12 kg.m2
COG X coordinate2 10.10 m
COG Y coordinate2 0.00 m
COG Z coordinate2 -6.76 m
K22 8.67E+05 kg/s2
C44(5%Ccrit ) 5.48E+09 kg.m/s
C44(6%Ccrit ) 6.03E+09 kg.m/s

1
Values were calculated in relation to the body center of gravity
2
Values are described in relation to the global coordinate system (x,y,z)=(0,0,0)
 145

7.2.2 Numerical Simulations

The numerical simulations for the RAO determination were conducted considering a set
of 45 regular waves, which covered a frequency range between 0.307 to 0.873 rad/s, as
may be observed in Table 7.4. As viscous damping effects may not be neglected when
evaluating roll motions, the numerical results were tested with two different external roll
damping coefficients C44, also presented in Table 7.3, which were calculated as 5% and
6% of the critical damping Ccrit .

Table 7.4: Regular incoming waves used in the FPSO VLCC hull simulations
ID ω (rad/s) T(s) λ(m) ID ω (rad/s) T(s) λ(m)
1 0.307 20.466 653.992 23 0.517 12.153 230.605
2 0.323 19.453 590.805 24 0.534 11.766 216.156
3 0.340 18.480 533.201 25 0.550 11.424 203.762
4 0.356 17.649 486.350 26 0.566 11.101 192.405
5 0.372 16.890 445.413 27 0.582 10.796 181.971
6 0.388 16.194 409.435 28 0.598 10.507 172.364
7 0.390 16.111 405.247 29 0.614 10.233 163.498
8 0.395 15.907 395.052 30 0.631 9.958 154.807
9 0.400 15.708 385.238 31 0.647 9.711 147.245
10 0.404 15.552 377.647 32 0.663 9.477 140.224
11 0.405 15.514 375.784 33 0.679 9.254 133.693
12 0.415 15.140 357.893 34 0.695 9.041 127.608
13 0.420 14.960 349.422 35 0.711 8.837 121.930
14 0.425 14.784 341.249 36 0.728 8.631 116.302
15 0.430 14.612 333.359 37 0.744 8.445 111.353
16 0.435 14.444 325.739 38 0.760 8.267 106.714
17 0.437 14.378 322.765 39 0.776 8.097 102.359
18 0.443 14.183 314.081 40 0.792 7.933 98.265
19 0.453 13.870 300.367 41 0.808 7.776 94.412
20 0.469 13.397 280.223 42 0.825 7.616 90.561
21 0.485 12.955 262.039 43 0.841 7.471 87.148
22 0.501 12.541 245.569 44 0.857 7.332 83.924
45 0.873 7.197 80.876

The body and free surface meshes used in the simulations are illustrated in Figures 7.18
and 7.19, respectively. The meshes designed for the FPSO and the free surface were
defined after a convergence test and were composed of 1088 and 3600 panels, respectively.
The free surface mesh was constructed with radius rf s = 2000 m and with a high panel
concentration near the FPSO position. Once more, the time-step was set to ∆t = T /60 s
and the damping zone coefficients considered b = 2.0 and a = 1.0.
146

Figure 7.18: FPSO VLCC hull panel mesh

Figure 7.19: Free surface panel mesh for the FPSO simulations

Figures 7.20, 7.21 and 7.22 present the comparisons between numerical results and ex-
perimental data for sway, heave and roll D.O.F., respectively. As expected, the change
of the external roll damping coefficients brought higher deviations at frequencies near to
the natural frequency. In addition, this variation did not influence the heave motions and
had only a slight impact in the sway ones, which is justified by the existence of a nonzero
hydrodynamic crossing term coefficient, originated from the radiation problem, coupling
these two degrees of freedom. In general, the numerical method recovered well the ex-
perimental curves, being capable of predicting the motion amplitudes with reasonable
accuracy.
 147

Present values C44 = 0.05Ccrit

0.9 Present values C44 = 0.06Ccrit
Experiment (Transient Wave)
0.8
Experiment (Regular Wave)
0.7

0.6
ξ¯2 /AI

0.5

0.4

0.3

0.2

0.1
0.2 0.4 0.6 0.8 1
ω (rad/s)

Figure 7.20: Comparison between numerical results and experimental data of sway re-
sponse amplitude operator for the FPSO VLCC hull

1.6
Present values C44 = 0.05Ccrit
1.4 Present values C44 = 0.06Ccrit
Experiment (Transient Wave)
1.2 Experiment (Regular Wave)

1
ξ¯3 /AI

0.8

0.6

0.4

0.2

0
0.2 0.4 0.6 0.8 1 1.2
ω (rad/s)

Figure 7.21: Comparison between numerical results and experimental data of heave re-
sponse amplitude operator for the FPSO VLCC hull
148

6
Present values C44 = 0.05Ccrit
Present values C44 = 0.06Ccrit
5 Experiment (Transient Wave)
Experiment (Regular Wave)
4
ξ¯4 /AI

0
0.2 0.4 0.6 0.8 1
ω (rad/s)

Figure 7.22: Comparison between numerical results and experimental data of roll response
amplitude operator for the FPSO VLCC hull

Direct comparisons between numerical and experimental time series of motion for sway
(ξ2 ), heave (ξ3 ) and roll (ξ4 ), considering C44 = 0.05Ccrit and C44 = 0.06Ccrit are presented
in Figures 7.23 and 7.24, respectively. These motions arise from the FPSO interaction
with a regular wave of frequency ω = 0.443 rad/s and amplitude AI = 0.92 m, both
in full scale. The time series are synchronized considering only the motion data of one
D.O.F., in this case heave, so as to preserve the phase information with the other D.O.F..
A good agreement is observed for the heave motion time series, in which the experimental
and numerical curves are practically coincident. On the other hand, although the roll
and sway motion phases were predicted accurately, slightly discrepancies in amplitude
may be noticed, specially in Figure 7.24, where an external coefficient of C44 = 0.06Ccrit
is being applied. This result is contradictory with the comparisons made for the roll
RAO, see Figure 7.22, obtained with the transient wave, in which the consideration of
the higher damping coefficient clearly improved the agreement. However, as could also
be observed in the figure, the roll RAO measured in the experiment considering transient
and regular waves presented different amplitudes, which do not allow a tuning of the
linear damping coefficient that is equally accurate for both cases. In spite of this fact,
good predictions were obtained with the numerical method, attesting once again that
the equilibrium between the dynamic and hydrodynamic forces was preserved during the
whole simulation.
 149

Experiment Present values

1

ξ2 (m)
0

−1
0 50 100 150 200 250 300 350 400 450 500
t(s)
2
ξ3 (m)

−2
0 50 100 150 200 250 300 350 400 450 500
t(s)
5
ξ4 ( ◦ )

−5
0 50 100 150 200 250 300 350 400 450 500
t(s)

Figure 7.23: Comparison between numerical and experimental motion time series for sway
(ξ2 ), heave (ξ3 ) and roll (ξ4 ), considering C44 = 0.05Ccrit

Experiment Present values

1
ξ2 (m)

−1
0 50 100 150 200 250 300 350 400 450 500
t(s)
2
ξ3 (m)

−2
0 50 100 150 200 250 300 350 400 450 500
t(s)
5
ξ4 ( ◦ )

−5
0 50 100 150 200 250 300 350 400 450 500
t(s)

Figure 7.24: Comparison between numerical and experimental motion time series for sway
(ξ2 ), heave (ξ3 ) and roll (ξ4 ), considering C44 = 0.06Ccrit
150
 151

Chapter 8

Multi-Body Simulations in
Side-by-Side Arrangement

The hydrodynamic interaction of two vessels in a side-by-side arrangement is currently

receiving substantial attention due to its practical application in the offloading process
involving the so-called Floating Liquefied Natural Gas (FLNG) units and Liquefied Nat-
ural Gas (LNG) carriers. As a result, much attention has been given to the analysis of
multi-body interaction effects with regard to the prediction of risk of collision between
the vessels (ZHAO et al., 2011). In this regard, the correct modelling of the fully coupled
dynamics of this multi-body problem and also the prediction of the vessels relative mo-
tions are important aspects when planning such a complex operation (KOO; KIM, 2005).
Indeed, this analysis must also be taken in to account in the design of a mooring system
for this kind of operation (KASHIWAGI; ENDO; YAMAGUCHI, 2005).

Wave resonant effects in the gap is one of the challenging problems in the hydrodynamic
modelling of two bodies arranged in a side-by-side configuration. These resonances create
different mode shapes of wave elevation in the gap at each associated resonant frequency,
which is a behaviour quite similar to the one that takes place in moonpools (MOLIN,
2001). The three basic modes are normally referred to as the piston mode, longitudinal
and transversal sloshing. An approximation formula for the estimation of these frequencies
in open boundaries may be found in Molin et al. (2002). A comprehensive numerical
investigation with the purpose of studying the different resonant frequencies and modes
is given by Sun, Taylor and Taylor (2010).

Although, in reality, resonant effects in the wave height may occur in the gap, they
are largely dampened by viscous effects and, therefore, the conventional potential flow
152

methods are known to overestimate the hydrodynamic forces, the wave elevation in the
gap and consequently the body motions. This occurs because these methods are unable
to model the viscous effects, namely skin friction and flow separation on the hull side,
that are deemed important, especially the latter, for the flow in the small gap between
the hulls ((MOLIN et al., 2009) and (KRISTIANSEN; FALTINSEN, 2010)).

CFD applications for dealing with the viscous effects in this resonant problem are certainly
envisaged, but, to this moment, the high computational effort they demand still renders
these applications infeasible for practical engineering purposes. For this reason, mixed
approaches combining the benefits provided by viscous and potential flow solvers may be
considered as a promising alternative to handle the problem. Kristiansen and Faltinsen
(2012) applied this methodology in a two-dimensional numerical wave tank with a floating
body. Elie et al. (2013) presented numerical results computed with their three-dimensional
SWENSE (Spectral Wave Explicit Navier-Stokes Equations) numerical method for two
side-by-side fixed barges in different regular waves incident angles.

However, for the time being, computational methods most often applied to model the side-
by-side problem are based on the potential flow theory, in which suppression methods to
deal with the resonant problems are used as an attempt to better reproduce the physics of
the phenomenon. In the context of linear frequency domain diffraction/radiation codes,
Huijsmans, Pinkster and de Wilde (2001) imposed a no flux vertical condition by applying
a rigid lid along the gap length. Newman (2003) improved the lid method using the gener-
alized mode technique with a set of basis functions composed of Chebyshev polynomials.
This method allows wave motion in the gap, in which the wave elevation is controlled by
imposing damping factors for each generalized mode.

In a different approach, Chen (2005) formulated a suppression method, namely the damp-
ing lid method, introducing a damping force directly into the conventional free surface
boundary conditions. A major advantage of this method is that only one value of damp-
ing factor is needed. A first attempt for applying this method is presented in Fournier,
Naciri and Chen (2006), where comparisons between numerical and experimental results
for a Floating Storage and Regasification Unit (FSRU) and a LNG carrier positioned
in side-by-side attested that the method was effective in attenuating the wave elevation
in the confined zone, which consequently led to a better reproduction of the first order
motions and drift forces. Following the same approach, Pauw, Huijsmans and Voogt
(2007) applied the damping lid method for the investigation of resonant effects for LNG
carriers. Experimental tests were conducted considering only one ship model, which was
positioned at half-gap witdh from the lateral tank wall. Results were discussed in terms of
wave elevation in the gap, motion RAOs (Response Amplitude Operators) and wave drift
Quadratic Transfer Functions (QTFs). The authors concluded that the damping factor
 153

has a much greater effect on the drift forces than on the first order quantities and that an
unique value of damping parameter was not enough to provide a full agreement with the
test data. Furthermore, a slight frequency shift in wave elevations and vessels motions
between measurements and diffraction results was observed. Bunnik, Pauw and Voogt
(2009) overcame the frequency shift problem by applying the damping lid technique not
only in the gap but also inside the hulls, in order to remove the so-called irregular frequen-
cies (those generated by the spurious numerical solution of internal problems inside the
hulls). In fact, the authors have shown that such inconsistence of the numerical scheme
was related to the rigid lid imposed inside the bodies to remove irregular frequencies of
the solution, which was introducing a strong grid dependence into the computations.

Clauss, Dudek and Testa (2013) compared numerical and experimental results considering
a barge-LNG carrier system and found numerical resonance peaks in the heave RAO, but
not in the wave elevation inside the gap. In contrast with the majority of the previous
works, they concluded that no external damping was necessary for modelling the wave
elevations observed in the gap. This sort of discrepancy found among recent works on the
theme emphasizes that additional systematic studies are still needed for a more compre-
hensive understanding of the flow phenomena and the performance of the computational
techniques available for modelling them.

The analysis of multi-body hydrodynamic interactions have also being studied in time
domain ((BUCHNER; DIJK; DE WILDE, 2011), (KOO; KIM, 2005), (NACIRI; WAALS;
DE WILDE, 2007) and (ZHAO et al., 2013)). This approach is often used when the fully-
coupled analysis involving not only the bodies, but also the mooring lines and fenders must
be assessed, a situation in which a time domain approach must be necessarily applied. All
these works, however, were based on the use of Cummins’s equations (CUMMINS, 1962),
in which frequency domain hydrodynamic coefficients are used as input data for the time
domain calculations.

An alternative approach is to treat the hydrodynamic problem directly in time domain.

Kim and Kim (2008) applied a time domain boundary elements method for studying
the motion responses of adjacent vessels. They reproduced the resonant modes but did
not make any attempt to suppress the resonant effects. Although the authors presented
only linear results, the method can be extended to nonlinear problems by considering the
nonlinear restoring and Froude-Krylov forces. Applications of time domain codes to side-
by-side problems, including nonlinear results, were also presented by Yan, Ma and Cheng
(2009) and Hong and Nam (2011). In the latter, a poor convergence on the time series of
the second order sway forces was observed for resonant frequencies. These studies were
based on the finite element method technique and also disregarded any sort of artificial
damping in the gap.
154

Although the number of studies related to the application of time domain hydrodynamic
solvers is still small in comparison to the conventional frequency domain method, it does
represent a promising approach for the seakeeping analysis of fully coupled multi-body
systems comprised by floating bodies, mooring lines, fenders etc. Nevertheless, the per-
formance of suppression methods modeled in time domain potential flow solvers is a topic
that still requires further investigation.

With this in mind, this chapter discusses the performance of the present time domain
code applied to a multi-body system in side-by-side configuration. In order to provide
benchmark data for the TDRPM numerical simulations, fundamental tests with simplified
geometries were carried out at the model basin of CEHINAV-Technical University of
Madrid (UPM). Moreover, the results were also compared to the ones calculated by the
software WAMIT, which was also used for the conceptual planning of the experiments.

8.1 Experimental Setup

The experimental tests considered a multi-body system comprising two bodies of canonical
geometries, namely a barge and a geosim, arranged in two different side-by-side configu-
rations. The main characteristics of the models are presented in Table 8.1 and illustrated
in Figures 8.1 and 8.2, the latter presenting the geosim geometry in detail. The tests were
conducted at the towing tank of the CEHINAV-Technical University of Madrid (UPM).
The tank is 100 m long, 3.8 m wide and the water depth is 2.5 m.

Table 8.1: Main characteristics of the barge and geosim models

Parameters Barge Geosim
Length over all 1.67 m 2.00 m
Beam 0.665 m 0.40 m
Depth 0.205 m 0.32 m
Draught 0.12 m 0.18 m
Displacement 133.26 kg 83.30 kg
Pitch Inertia1 − 31.84 kg.m2
COG X coordinate2 − 0.00 m
COG Y coordinate2,3 − -0.25/-0.5 m
COG Z coordinate2 − 0.00 m

1
Values were calculated in relation to the body center of gravity
2
Values are described in relation to the global coordinate system (x,y,z)=(0,0,0)
3
The Y coordinate of the COG is half of the gap width
 155

1.67

0.205

Figure 8.1: (b) Barge geometry and main dimensions in meters. (b) Geosim geometry
0.665

and main dimensions in meters

0.20 0.20
0.40

0.20
0.22 0.22
0.10 0.22
0.32

1.50 0.50

Figure 8.2: Details of the geosim geometry and main dimensions in meters

The test conditions were restricted to very simple configurations, aiming at providing
benchmark data for the numerical results. Bearing this in mind, during the tests the barge
was kept fixed and the geosim was attached to mechanical guiding arms, which restrained
its sway, roll and yaw motions. Moreover, the geosim drift in surge was controlled by
two springs linking the mechanical device to the model. This procedure enabled one to
keep the gap width and length practically constant during the measurements, providing
a convenient configuration for numerical modelling. Also for the purpose of calibration
of the numerical method, the mass of the guiding arms and the total mass and inertia of
the complete system were determined and are presented in Table 8.2.

Table 8.2: Main characteristics of the geosim including the mechanical guiding arms data
Feature Geosim
Guiding arms mass 5.5 kg
Total Mass 88.80 kg
Pitch radius of gyration 0.598 m
Heave Period 0.95 s
Pitch Period 1.26 s
156

The geosim model was positioned in the tank with its longitudinal axis coincident with the
longitudinal axis of the tank and the barge was located aside of the geosim, as represented
in the sketch shown in Figure 8.3.

Figure 8.3: Models positions in relation to the towing tank walls. Dimensions in meters

Geosim heave and pitch motions were tracked by means of a laser system installed in the
towing tank carriage. Additionally, four wave probes (P1, P2, P3 and P4, see Figure 8.3)
were used for monitoring the wave elevation in different locations, three of them being
positioned inside the gap. The relative distances between the wave probes are presented
in Table 8.3. Figures 8.4 and 8.5 present two photographs of the experimental setup,
including the probes arrangement.

Table 8.3: Relative distances between the wave probes

L(P1-P2) 0.48 m
L(P2-P3) 0.475 m
L(P3-P4) 0.485 m
L(P4-Stern Geosim) 0.28 m
 157

Figure 8.4: General view of the models positioned in the tank

Figure 8.5: Wave probes arranged along the gap centerline

Since the experiments were conducted in a towing tank, special care had to be taken in
order to minimize the influence of reflected waves from the tank walls within the frequency
range of interest (namely, the range where the gap resonant modes are expected to oc-
cur). In fact, the distance between the models was defined considering previous numerical
results calculated by the software WAMIT, which aimed at evaluating the expected gap
resonant frequencies and whether the presence of the tank walls would interfere with the
158

experimental measurements at these frequencies. As will be demonstrated in the next

section, by adopting proper gap widths, it was possible to restrict the influence of the
tank walls to a range of frequencies different to the one in which we were interested in.

Once these preliminar investigations were performed, two experimental setups of gap
widths 0.05 m and 0.10 m were selected and denoted as Cases 1 and 2, respectively.
These tests were conducted focusing on the study of linear effects and included a set of
49 and 32 regular waves of small amplitude for Cases 1 and 2, respectively. The main
characteristics of the waves may be observed in Tables 8.4 and 8.5. The wave height
for every frequency was determined considering a wave steepness smaller or equal to
H/λ ≤ 3% (limit established for the highest wave frequencies) and the maximum allowed
wave height was 0.04 m in order to guarantee a stable propagation of the waves along the
tank. For Case 1, which was tested first, the number of waves was larger because some
of them were used with the intent of evaluating the influence of the tank walls and also
checking whether the preliminary numerical models were providing reasonable results. For
Case 2 the test matrix was reduced to 32 waves, concentrating most of the waves within
the frequency range where gap resonance had been predicted by the numerical models.
Due to the limitations concerning the width of the towing tank, only head waves were
considered.

Table 8.4: Regular waves tested for the gap width 0.05 m
ID ω (rad/s) T(s) H(m) ID ω (rad/s) T(s) H(m)
1 4.000 1.571 0.040 18 7.400 0.849 0.034
2 4.250 1.478 0.040 19 7.600 0.827 0.032
3 4.500 1.396 0.040 20 7.700 0.816 0.031
4 4.750 1.323 0.040 21 7.800 0.806 0.030
5 5.000 1.257 0.040 22 7.900 0.795 0.030
6 5.250 1.197 0.040 23 8.000 0.785 0.029
7 5.400 1.164 0.040 24 8.125 0.773 0.028
8 5.500 1.142 0.040 25 8.250 0.762 0.027
9 5.600 1.122 0.040 26 8.375 0.750 0.026
10 5.750 1.093 0.040 27 8.500 0.739 0.026
11 6.000 1.047 0.040 28 8.625 0.728 0.025
12 6.200 1.013 0.040 29 8.750 0.718 0.024
13 6.400 0.982 0.040 30 8.875 0.708 0.023
14 6.600 0.952 0.040 31 9.000 0.698 0.023
15 6.800 0.924 0.040 32 9.250 0.679 0.022
16 7.000 0.898 0.038 33 9.500 0.661 0.020
17 7.200 0.873 0.036 34 9.750 0.644 0.019
35 10.000 0.628 0.018
 159

Table 8.5: Regular waves tested for the gap width 0.1 m
ID ω (rad/s) T(s) H(m) ID ω (rad/s) T(s) H(m)
1 4.000 1.571 0.040 14 7.750 0.811 0.031
2 4.500 1.396 0.040 15 7.875 0.798 0.030
3 5.000 1.257 0.040 16 8.000 0.785 0.029
4 5.500 1.142 0.040 17 8.125 0.773 0.028
5 6.000 1.047 0.040 18 8.250 0.762 0.027
6 6.500 0.967 0.040 19 8.375 0.750 0.026
7 6.750 0.931 0.040 20 8.500 0.739 0.026
8 7.000 0.898 0.038 21 8.750 0.718 0.024
9 7.125 0.882 0.036 22 9.000 0.698 0.023
10 7.250 0.867 0.035 23 9.250 0.679 0.022
11 7.375 0.852 0.034 24 9.500 0.661 0.020
12 7.500 0.838 0.033 25 9.750 0.644 0.019
13 7.625 0.824 0.032 26 10.000 0.628 0.018

8.1.1 Investigation of the Tank Walls Effects

The higher order WAMIT module was applied with the intent of investigating the influence
of the tank walls on the experimental results. Since WAMIT is formulated in frequency
domain and applies the well known free surface Green function, which eliminates the
necessity of discretizing the free surface, the code is more suitable to be used for exhaustive
variations of important parameters that must be assessed in the problem, if compared to
the present time domain solver.

For this analysis, numerical models with and without tank walls were computed in WAMIT.
The numerical models with the tank walls were emulated for both cases (1 and 2) by intro-
ducing sets of bodies images (barge + geosim) with respect to the tank walls. Moreover,
since two tank walls should be considered and the exact solution of the problem would
only be reached by using an infinite series of images, a convergence analysis was performed
in order to determine the minimum number of images above which the results would not
be significantly affected. In total, 5 sets of body images had to be considered for achieving
a reasonable convergence, resulting in a model composed of 12 bodies, as illustrated in
Figure 8.6. In order to reduce the computational effort of the WAMIT computations, a
symmetry plane, illustrated in Figure 8.7, was also considered, resulting in a model of 6
bodies plus a symmetry plane.

Concerning the bodies meshes, the panel size parameter was set to 0.1 and 0.2 for Cases
1 and 2, respectively, values that were also defined after a convergence analysis of the
mesh. One should notice that a smaller value of panel size was required for Case 1, since
160

3.80 0.985 0.985 1.90 1.90 0.985 0.985 3.80

Tank Tank
Wall Wall

Figure 8.6: Body images considered in the multi-body model. Dimensions in meters

3.80 1.97 1.90

Figure 8.7: Symmetry line applied for the frequency domain computations. Dimensions
in meters

the gap width was smaller.

Figure 8.8 presents a comparison between RAO calculations with the WAMIT model and
the measured data for Case 1. The comparison refers to the geosim heave and pitch RAOs
and shows that the heave motion amplifications observed in the experiment for frequencies
between 3.5 rad/s and 7.0 rad/s are caused by waves reflected by the tank walls. The
evidence for this comes from the fact that the WAMIT numerical model with walls was
indeed able to capture the same trends of the experimental curve. Moreover, one should
also notice that the influence of the walls tends to be minimized for higher frequencies
(> 7rad/s), for which both WAMIT numerical models present similar results. It is also
worth mentioning that the pitch motion is not significantly affected by the presence of
the walls.

The influence of the tank walls on the wave elevations at the gap centerline is investigated
in Figure 8.9, from which it can be inferred that the wall effects are of minor degree.
Both WAMIT numerical models (with and without walls) present good agreement with
the experimental data for wave frequencies lower than 8 rad/s, in the same range where
 161

the heave motions are significantly influenced by the tank walls. Nonetheless, one should
realize that the numerical results provided by both models (with and without walls)
present significant wave amplifications for a range of frequencies between 8.0 and 9.0
rad/s, which are in general much larger than the observed experimental values.

As aforementioned, potential flow solvers present some difficulties when dealing with phys-
ical problems that contain narrow gaps, tending to provide unrealistic wave elevations and
body motions for the associated gap resonant frequencies, where the potential flow physics
is not enough to represent the hydrodynamic phenomenon. A more complete discussion
about this issue will be presented in the next sections, along with the description of the
numerical modelling based on the time domain solver.

Analogous results for Case 2 are presented in Figures 8.10 and 8.11. By analyzing both
figures, it is possible to observe that the results present a behavior similar to the ones
obtained for Case 1, once again emphasizing that the influence of the tank walls is more
pronounced for wave frequencies lower than 7.0 rad/s.

Based on these analyses, it was possible to conclude that although the experimental
data was conducted in a towing tank with walls relatively close to the model bodies, the
influence of reflected waves was restricted to frequencies below the range of main interest
to this study, namely the range where gap resonant effects are caused by the hydrodynamic
interaction of the models in side-by-side configuration.
162

1.4
WAMIT with walls
1.2 WAMIT without walls
Experimental
1
ξ /A [m/m]

0.8
I

0.6
3

0.4

0.2

0
3 4 5 6 7 8 9 10 11 12
ω [rad/s]

0.25

0.2
ξ /A [deg/mm]

0.15
I

0.1
5

0.05

0
3 4 5 6 7 8 9 10 11 12
ω [rad/s]

Figure 8.8: Case 1: Comparison between WAMIT numerical data and experimental values
in terms of heave (top) and pitch (bottom) RAOs. Gap width = 0.05 m
 163

P1
ηP1/AI [m/m]
10 WAMIT with walls
WAMIT without walls
Experimental
5

0
4 5 6 7 8 9 10
ω [rad/s]
P2
ηP2/AI [m/m]

10

0
4 5 6 7 8 9 10
ω [rad/s]
P3
ηP3/AI [m/m]

10

0
4 5 6 7 8 9 10
ω [rad/s]
P4
ηP4/AI [m/m]

10

0
4 5 6 7 8 9 10
ω [rad/s]

Figure 8.9: Case 1: Comparison between WAMIT numerical data and experimental values
in terms of wave elevations RAOs in different points positioned along the gap centerline.
Gap width = 0.05m
164

1.4
WAMIT with walls
1.2 WAMIT without walls
Experimental
1
ξ3/AI [m/m]

0.8

0.6

0.4

0.2

0
3 4 5 6 7 8 9 10 11 12
ω [rad/s]

0.25

0.2
ξ5/AI [deg/mm]

0.15

0.1

0.05

0
3 4 5 6 7 8 9 10 11 12
ω [rad/s]

Figure 8.10: Case 2: Comparison between WAMIT numerical data and experimental
values in terms of heave (top) and pitch (below) RAOs. Gap width = 0.1 m
 165

P1
6
WAMIT with walls
ηP1/AI [m/m]

4 WAMIT without walls

Experimental
2

0
4 5 6 7 8 9 10
ω [rad/s]
P2
6
ηP2/AI [m/m]

0
4 5 6 7 8 9 10
ω [rad/s]
P3
6
ηP3/AI [m/m]

0
4 5 6 7 8 9 10
ω [rad/s]
P4
6
ηP4/AI [m/m]

0
4 5 6 7 8 9 10
ω [rad/s]

Figure 8.11: Case 2: Comparison between WAMIT numerical data and experimental
values in terms of wave elevations RAOs in different points positioned along the gap
centerline. Gap width = 0.1 m

8.2 Time Domain Numerical Model

The barge and geosim hulls used in the time domain simulations were modeled with 1584
and 1504 quadrangular panels, respectively. For the free surface meshes, 4797 panels were
applied in Case 1 and 5751 panels in Case 2, the difference being associated exclusively
166

to the number of panels required to model the gap surface (750 panels for Cases 1 and
1500 for Case 2). These values were obtained after a convergence analysis in terms of the
geosim motions and wave elevations.

Since the analysis conducted with WAMIT did not indicate significant wall effects in the
gap resonant frequencies, the walls have been neglected in the time-domain simulations,
thus reducing significantly the number of panels and, consequently, the total processing
time.

By taking Case 1 (gap equal to 0.05 m) as an example, Figures 8.12 and 8.13 present
the bodies (with the gap) and free surface meshes used in the computations, respectively.
Notice that the meshes have been refined towards the bodies edges, with a larger concen-
tration of panels near the gap region in order to improve the numerical convergence. In
addition, the free surface mesh has been designed to allow a smooth transition from the
gap region to the outer free surface. The same mesh topology has been applied in Case
2.

Figure 8.12: Barge, geosim and gap panel meshes for Case 1

The present method was run for several regular wave frequencies within the range of fre-
quencies tested experimentally. Each simulation has provided a set of time series describ-
ing the geosim motions in the 3 D.O.F. (surge, heave and pitch) and the wave elevation at
the wave probes locations (Table 8.3), from which the signal amplitudes have been char-
acterized with the root mean square (rms) from a steady state interval. The time-step of
each simulation has been set to ∆t = T /30 s, whereas the numerical damping zone has
been set with two wave lengths (b = 2.0) and intensity a = 1.0 for all cases.
 167

Figure 8.13: Free surface panel mesh for Case 1

8.2.1 Comparisons with WAMIT and Experimental Data

The comparison between the present method, WAMIT and experimental results in terms
of geosim motions and wave elevations for Case 1 are presented in Figures 8.14 and
8.15, whereas for Case 2 they are presented in Figures 8.16 and 8.17. One should notice
that only the WAMIT model without walls is being considered, since the results are
focused exclusively on the range of frequencies in which the influence of tank walls can be
disregarded (see section 8.1.1).

The overall agreement between the time-domain and WAMIT results is good for all the
RAO curves. Even considering the fact that the present method is formulated in time
domain, such an agreement was indeed expected since the boundary value problem solved
by both computational codes is exactly the same. Also, it is clear that the numerical
results tend to overestimate, in both cases, the experimental data for some frequency
intervals.

For Case 1 (Figure 8.14), one should notice that in the frequency range 8 − 9 rad/s there
is a numerical resonance, which may be identified by the spurious peaks in the heave
and pitch RAOs at approximately 8.3 and 8.5 rad/s, respectively. The same trends are
observed for the wave elevations RAOs (Figure 8.15), especially for the measurement
points that are in fact inside the gap between the bodies (P2, P3 and P4). Since the
wave probe P1 was positioned outside the gap, it does not present significant influence
from the resonant effects, resulting in a smoother RAO curve when compared to the other
probes. At this location, a perfect agreement between numerical and experimental results
was observed.

The wave modes associated to the resonant frequencies of 8.3 and 8.5 rad/s are presented
168

in Figure 8.18, which illustrates the envelopes of wave elevation inside the gap in spatial
domain calculated by the time domain solver for Case 1. For the resonant wave frequency
8.3 rad/s, it is possible to observe that the wave elevation in the gap presents a piston type
resonant mode, in which the wave behaves like a column of water moving up and down
in the region between the vessels (MOLIN, 2001). Regarding the resonant frequency 8.5
rad/s, one may realize that instead of a piston mode, a second longitudinal mode is visu-
alized. It is interesting to observe that this mode is related to the spurious amplification
of the pitch motion at this same frequency (Figure 8.14).

The same qualitative behavior observed for Case 1 may be stated for Case 2 (Figures
8.16 and 8.17). In this case, however, since the gap width was increased to 0.10 m, the
numerical resonant frequencies are slightly shifted towards lower frequencies 7.5 − 8.5
rad/s. In this configuration, the numerical resonant peaks for heave and pitch RAOs are
approximately 7.7 and 7.8 rad/s. Once again, these resonant frequencies coincide with
the piston and second longitudinal modes of the gap, respectively, as may be observed in
Figure 8.19. It is also interesting to observe that, for waves that are outside of the gap
resonant frequency range, the numerical results agree with the experimental data very
well.
 169

0.5
Experimental
WAMIT without walls
0.4 Present Values
ξ /A [m/m]

0.3
I

0.2
3

0.1

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s]

50

40
ξ /A [deg/m]

30
I

20
5

10

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s]

Figure 8.14: Case 1: Comparison between WAMIT, present values and experimental data
in terms of heave (top) and pitch (bottom) RAOs. Gap width = 0.05 m
170

P1
η /A [m/m]

4
I

2
P1

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s]
P2
η /A [m/m]

6
WAMIT without walls
4 Experimental
I

TDRPM
2
P2

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s]
P3
η /A [m/m]

6
4
I

2
P3

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s]
P4
η /A [m/m]

4
I

2
P4

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s]

Figure 8.15: Case 1: Comparison between WAMIT, present values and experimental data
in terms of wave elevations RAOs in different points positioned along the gap centerline.
Gap width = 0.05 m
 171

WAMIT without walls

Experimental
0.5 Present Values

0.4
ξ3/AI [m/m]

0.3

0.2

0.1

0
6.5 7 7.5 8 8.5 9 9.5 10
ω [rad/s]

50

40
ξ5/AI [deg/m]

30

20

10

0
6.5 7 7.5 8 8.5 9 9.5 10
ω [rad/s]

Figure 8.16: Case 2: Comparison between WAMIT, present values and experimental
values in terms of heave (top) and pitch (bottom) RAOs. Gap width = 0.1 m
172

P1
6
WAMIT without walls
η /A [m/m]

4 Experimental
Present Values
I

2
P1

0
6.5 7 7.5 8 8.5 9 9.5 10
ω [rad/s]
P2
6
η /A [m/m]

4
I

2
P2

0
6.5 7 7.5 8 8.5 9 9.5 10
ω [rad/s]
P3
6
η /A [m/m]

4
I

2
P3

0
6.5 7 7.5 8 8.5 9 9.5 10
ω [rad/s]
P4
6
η /A [m/m]

4
I

2
P4

0
6.5 7 7.5 8 8.5 9 9.5 10
ω [rad/s]

Figure 8.17: Case 2: Comparison between WAMIT, present values and experimental
values in terms of wave elevations RAOs in different points positioned along the gap
centerline. Gap width = 0.1 m
 173

ω = 8.3 rad/s
8

4
ηgap/AI [m/m]

−2

−4

−6

−8
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
x (m)

ω = 8.5 rad/s
8

4
ηgap/AI [m/m]

−2

−4

−6

−8
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
x (m)

Figure 8.18: Case 1: Gap wave elevation envelopes in spatial domain computed with
the present method. Incoming wave frequencies ω = 8.3 rad/s (top) and ω = 8.5 rad/s
(bottom). Wave propagates from positive to negative x coordinates. Gap width = 0.05
m
174

ω = 7.7 rad/s
3

2
ηgap/AI [m/m]

−1

−2

−3
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
x (m)
ω = 7.8 rad/s
5
ηgap/AI [m/m]

−5
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
x (m)

Figure 8.19: Case 2: Gap wave elevation envelopes in spatial domain computed with
the present method. Incoming wave frequencies ω = 7.7 rad/s (top) and ω = 7.8 rad/s
(bottom). Wave propagates from positive to negative x coordinates. Gap width = 0.10
m

8.2.2 Damping Lid Approach

Numerical methods based on the potential flow theory are known to have a poor perfor-
mance when dealing with multiple bodies arranged in a side-by-side configuration, tending
to provide unrealistically high wave elevations in the gap between the vessels, which lead
to poor estimations of forces and motions. In practice, the wave elevations would be lim-
ited by viscous effects that are not accounted for by potential flow solvers. In this regard,
the application of suppression methods to handle these gap resonant problem becomes an
interesting alternative for a better representation of the physical problem.

In addition to the unrealistic values observed when the mathematical problem is solved in
 175

time domain, a very slow numerical convergence for frequencies near the resonant ones is
also observed. In these cases, a considerable number of wave cycles is necessary to reach
a steady state.

In order to deal with these gap resonant effects, a new development was implemented
in the numerical method, in which the free surface boundary conditions (kinematic and
dynamic) are reformulated by means of the damping lid technique used in frequency
domain by Chen (2005), Fournier, Naciri and Chen (2006), Pauw, Huijsmans and Voogt
(2007) and Bunnik, Pauw and Voogt (2009). In this method, a constant damping factor
 is included in the free surface boundary conditions ((3.21) and (3.22)), as presented in
equations (8.1) and (8.2):

∂η ∂φ
= − η on z = 0 in the gap, (8.1)
∂t ∂z

∂φ
= −ηg − φ on z = 0 in the gap. (8.2)
∂t

One should notice that the undamped free surface elevation is recovered by setting the
damping factor  equal to zero. As will be heuristically demonstrated later, the imposition
of the damping parameter attenuates the numerical convergence problems, stabilizing the
solutions much faster. Another characteristic of the method is that the damping parameter
influences the results only in the resonant frequency range. Moreover, one of the main
advantages of this method is that it does not require the use of additional degrees of
freedom in the model, as the technique applied by Newman (2003). Instead, only one
value of damping factor is included in the free surface boundary conditions.

On the other hand, it is important to mention that the  value is, for the moment, not
defined rationally, but it is merely “tuned” with the experimental data considering that
the coefficient depends neither on the wave amplitude or frequency. Thus, for comparison
purposes only three different damping factors have been used  = (0.0, 0.0625, 0.125) 1/s.
Despite the simplicity of this technique, it will be shown that reasonable results can be
obtained in terms of wave elevations and body motions in most of the cases.

First, a sensitivity analysis concerning the behavior of the time series of motions and
wave elevations with respect to the varying damping factor  is presented. Considering
as an example the results of Case 1, Figures 8.20 and 8.21 display the time histories
for heave and pitch motions considering waves with frequencies 8.3 and 8.5 rad/s, which
correspond to the gap resonant frequencies for heave and pitch motions, respectively. The
results show that the imposition of a damping factor on the gap is effective in reducing
176

the amplitudes of heave and pitch motions at their respective resonant frequencies. Notice
that, for the wave frequency of 8.3 rad/s (heave resonance frequency), the amplitudes of
the pitch motions are not affected by the variation of  values, while the opposite occurs
for the heave motion when the wave frequency of 8.5 rad/s (pitch resonance frequency) is
considered.

The time series of wave elevations at points P2, P3 and P4 are presented in Figures
8.22 and 8.23 for the wave frequencies 8.3 and 8.5 rad/s, respectively. As for the geosim
motions, the use of the damping lid technique reduces the wave amplitudes in the gap,
which is a good indication that the method may be used in order to emulate the wave
behavior observed in the experiments more closely.

One should also realize that the inclusion of the damping factor in the formulation clearly
favors the convergence of the time series (Figures 8.22 and 8.23). It is possible to observe
that the time series obtained with non-zero  values reached a steady state much faster
than the cases with zero . In fact, with zero  more than 150 wave cycles in the simulation
were necessary to stabilize the motions and wave elevations. This behavior is analogous
for the motions and wave elevations obtained in Case 2.

The associated envelopes of wave elevation in the gap for Case 1 are presented in Figures
8.24(a) and 8.24(b). In these figures, the experimental amplitudes of the three monitored
points in the gap are also included for comparison purposes. It can be noticed that the
damping lid technique implemented in this work greatly improved the results, presenting
a much better agreement between the computed wave elevations and the experimental
data, especially for the damping factor value of  = 0.125 1/s, which provided the best fit
with the measured data.

Figures 8.25(a) and 8.25(b) present the wave elevation envelopes for Case 2, considering
the resonant frequencies ω = 7.7 and ω = 7.8 rad/s. Once more, the damping technique
provides a considerable improvement to the wave profile patterns in the gap, reproducing
the test data reasonably well. In this case, the damping factor  = 0.0625 1/s presented
a slightly better performance in relation to  = 0.125 1/s.

The influence of the damping factors on the motions RAOs is presented in Figures 8.26
and 8.27 for Cases 1 and 2, respectively. In general, the damping factor decreases the res-
onant values in the frequency range of interest, eliminating the spurious characteristic of
the curves. Nevertheless, in both cases, the numerical results of heave motion present dif-
ferent trends when compared to the experimental data. Regarding the pitch motions, the
inclusion of the damping coefficient improves the matching with the test data, recovering
the trends of the RAOs.
 177

For the wave elevations RAOs in the gap, the use of the damping factor also improved the
predicted wave amplitudes in comparison to the measured data, eliminating most of the
irregular oscillations observed for  = 0.0 1/s. For probes P2 and P4, the application of the
method enabled to recover the wave amplitudes inside the gap reasonably well. Concerning
P3, a larger discrepancy between the results is observed, since the experimental data
presented higher elevations in comparison to the numerical predictions even without the
inclusion of a damping factor. It is also noticeable that all the numerical results tend to
the same asymptotic value when the waves are outside of the resonant frequency range.

One should notice, however, that although the damping lid method applied does not
intend to capture the physics of the flow in the gap, the use of this simplified technique,
which incorporates damping factors independent on the wave amplitudes and frequencies,
has provided reasonable numerical results with a fair agreement with the experimental
data for most of the cases. Overall, in terms of the motions and wave elevations, the
value of damping of  = 0.125 1/s has provided the best agreement with the test results
in general.
178

ω=8.3 rad/s
0.5
ε = 0.0 [1/s]
ε = 0.0625 [1/s]
ε = 0.125 [1/s]
ξ3(t)/AI [m/m]

−0.5
0 10 20 30 40 50 60 70 80 90 100 110
t [s]

ω=8.3 rad/s
0.4

0.3

0.2
ξ5(t)/AI [rad/m]

0.1

−0.1

−0.2

−0.3

−0.4
0 10 20 30 40 50 60 70 80 90 100 110
t [s]

Figure 8.20: Case 1: Numerical heave (top) and pitch (bottom) motions time series for
different values of damping factor . Incoming wave frequency ω = 8.3 rad/s. Gap
width=0.05 m
 179

ω=8.5 rad/s
0.5
ε = 0.0 [1/s]
ε = 0.0625 [1/s]
ε = 0.125 [1/s]
ξ3(t)/AI [m/m]

−0.5
0 10 20 30 40 50 60 70 80 90 100 110
t [s]

ω=8.5 rad/s
0.6

0.4

0.2
ξ5(t)/AI [rad/m]

−0.2

−0.4

−0.6

−0.8
0 10 20 30 40 50 60 70 80 90 100 110
t [s]

Figure 8.21: Case 1: Numerical heave (top) and pitch (bottom) motions time series for
different values of damping factor . Incoming wave frequency ω = 8.5 rad/s. Gap
width=0.05 m
180

P2: ω=8.3 rad/s

5
ηP2(t)/AI [m/m]

ε = 0.0 [1/s]
ε = 0.0625 [1/s]
−5
ε = 0.125 [1/s]
0 10 20 30 40 50 60 70 80 90 100 110
t [s]
P3: ω=8.3 rad/s

5
ηP3(t)/AI [m/m]

−5

0 10 20 30 40 50 60 70 80 90 100 110
t [s]
P4: ω=8.3 rad/s

5
ηP4(t)/AI [m/m]

−5

0 10 20 30 40 50 60 70 80 90 100 110
t [s]

Figure 8.22: Case 1: Numerical wave elevation time series at P2 (top), P3 (middle) and
P4 (bottom) for different values of damping factors . Incoming wave frequency ω = 8.3
rad/s. Gap width = 0.05 m
 181

P2: ω=8.5 rad/s

5
ηP2(t)/AI [m/m]

−5

0 10 20 30 40 50 60 70 80 90 100 110
t [s]
P3: ω=8.5 rad/s

5
ηP3(t)/AI [m/m]

ε = 0.0 [1/s]
ε = 0.0625 [1/s]
−5
ε = 0.125 [1/s]
0 10 20 30 40 50 60 70 80 90 100 110
t [s]
P4: ω=8.5 rad/s

5
ηP4(t)/AI [m/m]

−5

0 10 20 30 40 50 60 70 80 90 100 110
t [s]

Figure 8.23: Case 1: Numerical wave elevation time series at P2 (top), P3 (middle) and
P4 (bottom) for different values of damping factors . Incoming wave frequency ω = 8.5
rad/s. Gap width = 0.05 m
182

ω=8.3 rad/s
8

4
ηgap/AI [m/m]

−2

−4 ε = 0.0 [1/s]
ε = 0.0625 [1/s]
−6 ε = 0.125 [1/s]
Experimental
−8
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
x [m]

(a)

ω=8.5 rad/s
8

4
ηgap/AI [m/m]

−2

−4 ε = 0.0 [1/s]
ε = 0.0625 [1/s]
−6 ε = 0.125 [1/s]
Experimental
−8
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
x [m]

(b)

Figure 8.24: Case 1: Comparison between numerical and experimental data of the gap
wave elevation envelopes in spatial domain for different values of damping factors . (a)
Incoming wave frequency ω = 8.3 rad/s. (b) Incoming wave frequency ω = 8.5 rad/s.
Wave propagates from positive to negative x coordinates. Gap width = 0.05 m
 183

ω=7.7 rad/s
3

1
ηgap/AI [m/m]

ε = 0.0 [1/s]
ε = 0.0625 [1/s]
0 ε = 0.125 [1/s]
Experimental
−1

−2

−3
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
x [m]

(a)

ω=7.8 rad/s
5
ε = 0.0 [1/s]
ε = 0.0625 [1/s]
ε = 0.125 [1/s]
Experimental
ηgap/AI [m/m]

−5
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
x [m]

(b)

Figure 8.25: Case 1: Comparison between numerical and experimental data of the gap
wave elevation envelopes in spatial domain for different values of damping factors . (a)
Incoming wave frequency ω = 7.7 rad/s. (b) Incoming wave frequency ω = 7.8 rad/s.
Wave propagates from positive to negative x coordinates. Gap width = 0.10 m
184

0.5
ε = 0.0 [1/s]
ε=0.0625 [1/s]
0.4
ε=0.125 [1/s]
Experimental
ξ /A [m/m]

0.3
I

0.2
3

0.1

0
8 8.2 8.4 8.6 8.8 9 9.2 9.4 9.6 9.8 10
ω [rad/s]

50

40
ξ /A [deg/m]

30
I

20
5

10

0
8 8.2 8.4 8.6 8.8 9 9.2 9.4 9.6 9.8 10
ω [rad/s]

Figure 8.26: Case 1: Numerical and experimental comparisons in terms of heave (top)
and pitch (bottom) motions RAOs for different values of damping factors . Gap width
= 0.05 m
 185

ε=0.0 [1/s]
ε=0.0625 [1/s]
0.5
ε=0.125 [1/s]
Experimental
0.4
ξ3/AI [m/m]

0.3

0.2

0.1

0
6.5 7 7.5 8 8.5 9 9.5 10
ω [rad/s]

50

40
ξ5/AI [deg/m]

30

20

10

0
6.5 7 7.5 8 8.5 9 9.5 10
ω [rad/s]

Figure 8.27: Case 2: Numerical and experimental comparisons in terms of heave (top)
and pitch (bottom) motions RAOs for different values of damping factors . Gap width
= 0.10 m
186

P1
η /A [m/m]

4
I

2
P1

0
8 8.2 8.4 8.6 8.8 9 9.2 9.4 9.6 9.8 10
ω [rad/s]
P2 ε = 0.0 [1/s]
ε = 0.0625 [1/s]
ε = 0.125 [1/s]
η /A [m/m]

6
Experimental
4
I

2
P2

0
8 8.2 8.4 8.6 8.8 9 9.2 9.4 9.6 9.8 10
ω [rad/s]
P3
η /A [m/m]

6
4
I

2
P3

0
8 8.2 8.4 8.6 8.8 9 9.2 9.4 9.6 9.8 10
ω [rad/s]
P4
η /A [m/m]

4
I

2
P4

0
8 8.2 8.4 8.6 8.8 9 9.2 9.4 9.6 9.8 10
ω [rad/s]

Figure 8.28: Case 1: Numerical and experimental comparisons of wave elevations in the
gap for different values of damping factors . Gap width = 0.05 m
 187

P1
η /A [m/m]
4
I

2
P1

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s] ε = 0.0 [1/s]
P2
ε = 0.0625 [1/s]
ε = 0.125 [1/s]
η /A [m/m]

4
Experimental
I

2
P2

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s]
P3
η /A [m/m]

4
I

2
P3

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s]
P4
η /A [m/m]

4
I

2
P4

0
7 7.5 8 8.5 9 9.5 10
ω [rad/s]

Figure 8.29: Case 2: Numerical and experimental comparisons of wave elevations in the
gap for different values of damping factors . Gap width = 0.10 m
188
 189

Chapter 9

Multi-Body Simulations with Large

Relative Displacements

The seakeeping analysis of offshore systems is commonly assessed by the use of dynamic
time domain simulators, which consider not only the loads induced by the waves, but
also those originated by wind, current and from the dynamics of risers and mooring lines.
Almost all of these simulators evaluate the wave forces by using potential hydrodynamic
coefficients previously calculated in frequency domain, which are then post-processed and
transformed to time domain by writing the equation of the floating body in terms of
convolutions of its motions with the so-called impulsive response functions. This equation
is often referred to as the Cummins’s equations (CUMMINS, 1962):

~
∂ 2 ξ(t)
Z ∞ ~
∂ ξ(t)

F~ (t)

(M + A(ω∞ )) + R(τ ) ~ =
(t − τ )dτ + Kξ(t) (9.1)
∂t2 0 ∂t M~ (t)

where M is the matrix of mass/inertia, A(ω∞ ) is the matrix of added mass/inertia co-
~ is the translational/rotational displacement vector, K
efficients in infinite frequency, ξ(t)
is the matrix of restoring coefficients, F~ (t) and M
~ (t) are vectors containing the forces
and moments from external loads, and R(τ ) is a matrix with the retardation functions
or ”memory functions”, which is calculated in terms of the radiation damping coefficients
B(ω) from a cosine Fourier transform, as follows:

Z ∞
2
R(τ ) = B(ω) cos(ωτ )dω (9.2)
π 0

In this approach, the frequency domain code is previously executed considering several
190

wave incidence angles and frequencies, which are used for the construction of a database
containing exciting forces, added mass, potential damping and force drift coefficients.
Therefore, in order to account for changes in the heading of the floating system in rela-
tion to the incoming wave angle, the simulator simply performs a search on the database,
updating the exciting and drift forces. An example of a simulator that applies this method-
ology is the Numerical Offshore Tank (TPN), a code that has been developed at the Uni-
versity of Sao Paulo since 2001, in cooperation with the oil state company Petrobras (see
for instance Nishimoto et al. (2003)). In this case, the software WAMIT is used for the
hydrodynamic coefficients calculations.

However, in the context of simulations considering the analysis of multi-body systems this
approach cannot be extended in a straightforward manner, since the relative positions of
the floating bodies may change during the simulations, thus modifying the wave field
surrounding them and also the mutual effects caused by each one of the bodies. As
frequency domain codes consider only fixed meshes, the aforementioned database had also
to consider the changes of the relative positions between the bodies, so as to make the
updating of the hydrodynamic coefficients possible during the simulations. Examples of
practical operations in which this kind of problem may be observed are: tandem offloading
operations in which the direction of wind, waves or current is suddenly changing, or the
shuttle is moving due to fishtailing; floatover operations when a barge with topsides is
docking between the columns of a semisubmersible unit; and the berthing manoeuver of a
LNG carrier towards an offshore LNG terminal, supported by tugboats (BUNNIK, 2014).

Tannuri et al. (2004) developed a numerical scheme that tries to deal with this problem by
coupling the frequency domain code WAMIT and the time domain simulator TPN. This
version of the code was developed for the analysis of tandem type offloading operations
involving a shuttle tanker and an FPSO. The procedure involved monitoring the relative
positions of the bodies during the simulations and, when a significant change was detected,
the calculations were interrupted and the software WAMIT was executed once again for
updating the hydrodynamic coefficients. Analysis of the relevance of such update was done
by comparing the mean offsets and the first order motions of the shuttle tanker arising
from the simulation using the original procedure, which does not consider the update of
relative positions, and the ones computed with the new methodology, which incorporates
the WAMIT code in the simulation loop. The authors identified significant differences
in the results obtained with the two different approaches, emphasizing the importance of
such consideration.

Still with respect to this development in the TPN code, Queiroz Filho and Tannuri (2009)
investigated the wave shielding effect induced by the FPSO in the shuttle tanker by
analyzing the wave forces experienced by the latter and the consequences in terms of the
 191

power demanded by a dynamic positioning (DP) system designed to perform a station-

keeping task. Again, the importance of incorporating the hydrodynamic update was
demonstrated, showing that errors up to 27% in the DP power were found if the shadow
effects and changes of the relative position were not considered.

More recently, Bunnik (2014) presented results concerning a berthing operation in which
an LNG carrier was pushed towards an LNG FPSO in sea waves of different incidence
angles. In his work, an approach similar to the one presented by Tannuri et al. (2004)
was applied. The results showed that, depending on the wave direction, the effects of the
changes in hydrodynamic interaction can be significant, resulting in a considerable change
in the time needed to complete the berthing. The author also states that, although the
results are apparently logical and realistic, a validation procedure is still required to
quantify the accuracy of the developed model.

Although these works present a promising approximation for such a complex problem,
some aspects of this approach are still questionable. For example, due to the large com-
putational effort necessary to perform the updating of the hydrodynamic coefficients with
a small interval, the numerical scheme needs to consider an arbitrary criterion established
in terms of the variation of the relative positions between the bodies. This criterion then
determines whether the simulator is interrupted and the coefficients are updated or not.
In addition, this procedure may introduce spurious and impulsive hydrodynamic forces
as a result of the discrete aspect of the method and, thus, still requires further investiga-
tions. Another observation refers to the fact that a frequency domain analysis represents a
multi-body system which oscillates steadily at one particular frequency around the mean
linearized position. In this sense, as also emphasized by Bunnik (2014), it is quite incon-
sistent to use the results in a simulation where the relative motions between the bodies
are indeed changing. As a consequence, the effects of the flow memory are not treated in
a consistent manner, since the retardation functions are convoluted with velocities that
are related to a different relative position that occurred in the past.

Aiming at partially overcoming these problems and provide a first step towards the com-
plete solution of such a complex analysis, we propose in this chapter a new development
that accounts for large relative displacements between two floating bodies under the action
of incoming sea waves by solving the boundary value problems together with a re-meshing
scheme that defines new free surface panel meshes as the bodies displace from their origi-
nal positions. Moreover, an interpolation algorithm used to determine the wave elevation
and the velocity potential distribution on the new free surface points is also included. The
loop structure of the method is presented in Figure 9.1.

Obviously, the method does only consider linear wave effects, therefore neglecting the
192

loads induced by wind, current or drift forces. In fact, it is known that in practice these
are the forces that may give rise to large horizontal relative displacements between the
bodies, but the calculation of these quantities would require other numerical developments
that are beyond the main objective of this thesis.

Therefore, in a first assessment, we consider that a large relative displacement occurs

somehow and offsets one of the bodies from its initial mean position to another one. In
such way, the trajectory of the body is chosen arbitrarily and prior to the beginning of the
time domain calculations, being input into the model as a prescribed motion. In addition,
we consider that the large body horizontal displacement is slow and, therefore, do not
generate waves themselves. As a consequence, we may treat the problem as a quasi-static
problem regarding the low frequency displacement.

Upon the considerations aforementioned, we maintain the linear characteristic of our

code and so the theoretical basis established in section 3.3 of this text. In this sense,
the following sections present the developments carried out for the new method, giving
emphasis to the algorithm that performs the re-meshing of the free surface grid and the
interpolation schemes used for the determination of the quantities at the new points of
the surface after a new mesh is generated.
 193

Free Surface Mesh Generation

(Algorithm described in section 9.1)

Initial Conditions
∂φ ∂ ∂φ


∂n
= ∂t ∂n
= 0 on Neumann Boundaries
∂φ
φ = ∂t = 0 on Dirichlet Boundaries

Solution of the BVPs

t + ∆t
φ and ∂φ
∂t
on Neumann Boundaries
∂φ ∂ ∂φ


∂n
and ∂t ∂n
on Dirichlet Boundaries

Updating of the Dirichlet Conditions

η and φ by free surface conditions
∂φ
∂t
by dynamic condition

Updating of the Neumann Conditions

F~ and M ~ by pressure integration
∂ ξ~ ∂ ξ~
2
, , by ξ~ by equations of motion
∂t2 ∂t

Updating of the bodies positions

Prescribed Motion

Free Surface Re-meshing Algorithm

Generation of a new panel mesh
Recomputation of the influence coeffi-
cients Dij and Sij , presented in section 5.1

Free Surface Interpolation Algorithm

Interpolation of η and φ for the
new free surface mesh points

No
t < T s? End
Yes

Figure 9.1: Structure of the numerical method including the re-meshing and interpolation
algorithms in the time loop
194

9.1 The free surface re-meshing algorithm

The re-meshing algorithm developed in this work is merely an adaptation of the 2D mesh
generator elaborated by Persson (2005), entitled in his work as “A Simple Mesh Generator
in MATLAB”, for its inclusion in the time domain loop of the boundary elements method
presented in this thesis. This method was chosen by its simplicity for being implemented
and integrated in our code, as it is explained in a few lines, going in the opposite way of
the standard complex meshing softwares that are nearly inaccessible and unfeasible to be
coupled in other tools.

The mesh generator algorithm and how it is implemented computationally is fully de-
scribed in Persson (2005). In this context, and for the sake of completeness, we reproduce
here only the theoretical basis necessary for the comprehension of the method.

A simple mechanical analogy between a planar triangular mesh and a 2D truss structure
forms the basis of this mesh generator. Accordingly, the triangles edges correspond to bars
and the vertexes correspond to joints of the truss, in such a way that each bar has a force-
displacement relationship f (l, lo ), which depends on its current l and desired l0 lengths.
From a pre-determined surface defined by the user, there is an external reaction normal
force acting at each boundary node, whose magnitude is just large enough to restrain
the nodes to move outside this boundary. Therefore, the principal unknowns, which are
the x and y coordinates of the joints, are found by solving for a static equilibrium in the
structure.

The Nmp mesh points may be arranged into a Nmp × 2 array p, containing their corre-
spondent x and y coordinates:

 
p= xy Nmp ×2
(9.3)

The force vector F (p) acting at each mesh point is decomposed in horizonal and vertical
components, containing also the internal forces Fint from the bars and the external reaction
forces from the boundaries Fext , as follows:

   
F (p) = Fint,x (p) Fint,y (p) Nmp ×2
+ Fext,x (p) Fext,y (p) Nmp ×2
(9.4)

As mentioned by Persson (2005), F (p) depends on the current positions of the bars con-
necting the joints, which by its turn is adjusted by Delaunay triangulation of the mesh
points, deciding the edges. Therefore, the force vector F (p) is not a continuous function of
p, since the truss topology is changed by the Delaunay triangulation as the points move.
 195

This imposes a complicated problem for solving for a static equilibrium in the structure,
represented by the system F (p) = 0.

In order to overcome this problem, an artificial time dependence is introduced to the

problem, transforming it in a system of Ordinary Differential Equations (ODEs) with non-
physical units, in which if a stationary solution is found (within a established criteria),
we assume that the system F (p) = 0 is satisfied.

dp
= F (p) t ≥ 0 (9.5)
dt

The system of ODEs is approximated by a simple forward Euler method that in a dis-
cretized and artificial time tn = n∆t approximates the current solution pn ≈ p(tn ) by

pn+1 = pn + ∆tF (pn ) (9.6)

In this way, when evaluating the force function, the positions pn are known and also the
structure topology of the current step. Furthermore, the external reaction forces enter
in the problem by moving back, to the closest boundary point, all the points that went
outside the surface boundaries during the update pn to pn+1 .

In these developments, the force function assumes the form f (l, l0 ) = k(l0 − l), considering
that only repulsive forces (this ensures that the points spread out across the whole surface)
are allowed:


k(l0 − l) if l < l0
f (l, l0 ) = (9.7)
0 if l ≥ l0

in which k is included to provide correct units and set to k = 1.

The simplified loop structure of the mesh generator is described in the following items:

1. Creation of a uniform distribution of mesh points within a bounding box that covers
a pre-specified surface;

2. Removal of all points outside the desired geometry surface;

3. Determination of the truss topology by a Delaunay triangulation and calculation of

the structure bars lengths l;

4. Calculation of the force magnitude at the joints of each structure bars by using
equation (9.4);
196

5. Calculation of the force components at the joints of each structure bars;

6. Calculation of F (p) by summing the force vector components from all bars meeting
at a node.

7. Updating of the node positions p by using equation (9.6);

8. If a point ends up outside the surface after the update of p, it is moved back to the
closest point on the boundary.

9. The last positions pn+1 are compared with the previous ones pn and if non “large
movements” are detected (specified criteria) the algorithms is stopped, otherwise
the algorithm is restarted in item 3, continuing until the criteria is satisfied.

We exemplify the algorithm by applying it in the generation of a circular mesh of radius

R = 3 m with a circular aperture of radius R = 1 m inside. This surface is equivalent
to the free surface meshes presented in chapter 6, but with a smaller diameter. Figures
9.2(a), 9.2(b) and 9.2(c) illustrate the first three steps of the algorithm, whereas Figure
9.2(d) presents the final mesh, it means, when the system F (p) = 0 is satisfied.

(a) (b)

(c) (d)

Figure 9.2: Steps of the mesh generation algorithm applied to a circular surface with
radius R = 3 m with a circular of radius R = 1 m inside. Figures (a), (b) and (c) refer to
the first three steps of the algorithm loop structure. Figure (d) illustrates the final mesh
 197

We now present, in Figure 9.3, four circular meshes of radius R = 5 m with two circular
apertures of radius R = 1 m in different relative distances, illustrating the algorithm
capability to deal with the generation of free surface meshes that may be applied in
hydrodynamic problems involving two bodies. These meshes represent four instants of a
simulation involving two circular cylinders, in which the trajectory of one of them was
prescribed to perform a large horizontal displacement. It is worth mentioning that these
meshes are not restricted to cases considering only bodies with circular sections, since
ship-shaped vessels, for example, could be surrounded by circular meshes that fit right
into the circular apertures.

(a) (b)

(c) (d)

Figure 9.3: Four circular meshes of radius R = 5 m with two circular apertures of radius
R = 1 m in different relative distances

9.2 The free surface interpolation algorithm

A simulation involving relative large displacements between two or more bodies requires
the re-generation of the free surface mesh at each time-step. As could be observed in
the latter section, as the free surface is re-meshed, new collocation points (centroid of
each triangle) are generated, but, unfortunately, neither the free surface elevation nor the
198

velocity potential are known at these new points, and the simulation cannot continue.

In order to overcome this problem, it becomes necessary to apply an interpolation scheme

to link the mesh generator with the flow solver, so as to allow the simulation to be
restarted from a previous free surface state. More precisely, after generating a new mesh,
the interpolation scheme must recover, as accurately as possible, the previous solution
field defined on an old mesh to proceed with the computation.

The choice of the interpolation scheme must be done carefully, since an inadequate selec-
tion may be a source of error that, in time-dependent problems, accumulate throughout
the simulations. Therefore, linear interpolation methods are normally avoided, in favor of
higher order schemes. With this in mind, we have implemented an interpolation scheme
that couples a second order polynomial with a weighted moving least-squares method.
This method is similar to the one applied by Wang (2005).

In this scheme, the shape of the free surface elevation and the velocity potential distribu-
tion is represented by the second order polynomials presented in equations (9.8)

η = Fη (x, y) = a1 + a2 x + a3 y + a4 x2 + a5 xy + a6 y 2 (9.8a)
φ = Fφ (x, y) = b1 + b2 x + b3 y + b4 x2 + b5 xy + b6 y 2 (9.8b)

in which, fixing to one point in space p0 , the twelve coefficients ai , bi , i = 1, 2, 3..., 6 of the
polynomials are calculated considering only the nearest points within a distance of 2l0 ,
where l0 is the local mesh size. Denote these nearest points as pk , k = 1, 2, 3..., Nnp , in
which Nnp is the number of points.

Thus, the coefficients are determined by the use of a weighted moving least-squares method
with the error function (9.9) given as:

Nnp
X
σ(a1 , a2 , a3 , a4 , a5 , a6 ) = Wk [Fη (xk , yk ) − ηk ]2 Nnp ≥ 6 (9.9a)
j=1
Nnp
X
σ(b1 , b2 , b3 , b4 , b5 , b6 ) = Wk [Fφ (xk , yk ) − φk ]2 Nnp ≥ 6 (9.9b)
j=1

in which Wk , see expression (9.10), is the weight function for the nearest points pk . Notice
that the weight function decreases exponentially with the distance between pk and the
point p0 .

|pk −p0 |
−
Wk = e 2l0
(9.10)
 199

In accordance to the least-squares method, let ∂σ/∂aj = 0 and ∂σ/∂bj = 0, obtaining the
linear systems of equations (9.11) for aj and bj .

6
X
Aη ij aj = Bη i i = 1, 2, 3, ..., 6 (9.11a)
j=1
6
X
Aφ ij bj = Bφ i i = 1, 2, 3, ..., 6 (9.11b)
j=1

in which Aη , Aφ , Bη and Bφ are defined by the expressions (9.12).

Nnp
X
Aη ij = Aφ ij = Wk βkj βki (9.12a)
k=1
Nnp
X
Bη ij = Wk ηk βki (9.12b)
k=1
Nnp
X
Bφ ij = Wk φk βki (9.12c)
k=1

Finalizing, the coefficients β are calculated by the following expressions:

βk1 = 1 βk2 = xk βk3 = yk (9.13)

βk4 = x2k βk2 = xk yk βk3 = yk2 (9.14)

in which k = 1, 2, 3..., Nnp .

The coefficients a and b that result from this interpolation scheme are determined from a
mesh, here denoted as current mesh, in which the free surface elevation and the velocity
potential distribution are known. Hence, after the positions of the bodies are changed
and the re-meshing of the free surface is performed, the polynomials (9.8) are used to
determine the desired quantities at the new points of interest.

9.3 Verification of Numerical Computation

Numerical simulations considering two bodies with one of them undergoing large displace-
ments during the time domain calculations are presented in the following sections. The
main objective of this study is to evaluate the hydrodynamic interaction between the
bodies in a scenario in which the horizontal relative positions of the bodies vary in time.
With this approach, the wave shielding effects caused by one of the bodies on the other
200

is also taken into account and its effects in terms of forces can be quantified.

Once the objectives are established, the study is performed with a set of simulations
considering only diffraction effects, it means, the bodies are restrained to oscillate in their
six degrees of freedom. In this case, one of the bodies is displaced horizontally from its
initial position, performing a slow oscillatory motion with large horizontal amplitude.

However, the verification of the results is not a simple task, since neither numerical nor
experimental results considering such scenario could be found in the literature. Moreover,
the software WAMIT, which was being used for the verification of the results involving
single body simulations, cannot be directly applied for this problem, since it considers only
frequency domain calculations that assume, as mentioned before, that the body oscillates
an infinite long time around its mean linearised position.

Bearing this in mind, a set of fundamental experimental tests was designed and conducted
aiming specifically at verifying the performance of the present numerical method for a
multi-body system with bodies undergoing large relative displacements. The tests were
carried out at the CH-TPN-USP (main particulars of the basin are described in section
7.2). The main characteristics of the experimental setup are described in the next section.

9.3.1 Experiments

The experimental tests considered a multi-body system comprising two identical alu-
minium circular cylinders, namely Body 1 and Body 2, which have 0.40 m of diameter,
0.36 m of height and were tested with a draught of 0.20 m. Figures 9.4 and 9.5 present
an illustrative sketch and a photograph of the experimental setup, respectively.
 201

WAVES

y
Body Forced
to Move
WP1
x2(t)=Apmsin(ωpmt)

14 m
WP2 x

WP3 Fixed Body

14 m

Figure 9.4: Illustrative sketch of the experimental setup

Figure 9.5: Lateral view of the models positioned in the tank

These tests were conceived in a very fundamental configuration with the main goal of
providing benchmark data for the present numerical method. Thus, during the tests the
Body 1 was kept fixed and connected to a 6 D.O.F load cell (see Figure 9.6), which was
used to measure the hydrodynamic forces and moments induced by the waves. In addition,
this load cell was properly positioned in the model in order to follow the sign convention
of the coordinate system presented in Figure 9.4. The main particulars of the load cell
are presented in Table 9.1.
202

Figure 9.6: ATI MINI85 6 D.O.F Load Cell. Source: ATI Catalog

Table 9.1: Main characteristics of the 6 D.O.F load cell

Calibration SI-1900-80
Forces/Moments Fx , Fy (N) Fz (N) Mx , My (Nm) Mz (Nm)
Sensing Ranges 1900 3800 80 80
Single-Axis-Overload 13000 27000 500 610

The Body 2 was positioned upstream of Body 1 and was attached to a ball screw shaft
driven by a servo motor, this system being used to impose a large prescribed oscillatory
motion on the body during the measurements, as illustrated in Figure 9.7. Moreover, this
mechanical device was also equipped with a resistive potentiometer, which enabled one
to monitor the Body 2 position in a synchronized manner with the forces and moments
measured in Body 1. These procedures provided a convenient and controlled multi-body
system setup for the sake of the numerical modelling of the problem.

Oscillatory Motion

Servo Motor Ball Screw Shaft

Body 2

Figure 9.7: Sketch of the mechanical equipment attached to Body 2

 203

Besides the load cell, three wave probes, namely WP1, WP2 and WP3, were used to mea-
sure the wave elevations at different locations near the bodies. The WP2 was positioned
at the mean position between the bodies and WP1 and WP3 were located upstream and
downstream, respectively. The relative distances between the wave probes are presented
in Figure 9.8.

WAVES

WP3 WP2 WP1

WL Body 1 Body 2

0.70 m 0.70 m

Figure 9.8: Front view sketch of the models and wave probes arrangement

Special care had to be taken into account for the definition of the wave frequencies to be
considered in this test. One must keep in mind the difficulties associated with measur-
ing interaction forces induced by waves diffracted from the bodies, which are commonly
much lower than those caused by the incident wave itself. Thus, small interferences from
reflected waves coming from the tank walls may disturb the results, demanding a good
performance of the flaps active control system on absorbing the waves. As presented in
Mello et al. (2010), the CH-TPN-USP has a better performance when absorbing waves
in the range of frequencies between 3.14 rad/s and 7.5 rad/s and, therefore, the wave
frequencies were selected within this range. In addition, since this experimental campaign
was conducted focusing on the verification of the present linear numerical method, the
tests considered only regular waves of small amplitude and steepness (H/λ ≤ 2%), which
were previously calibrated without the models. Only one wave incident angle was consid-
ered (θ = 270o ) due to the restrictions regarding an appropriate space for the positioning
of the mechanical device in the tank bridge. In total, 4 regular waves were selected for
this test, and their main particulars are presented in Table 9.2.
204

Table 9.2: Regular waves considered in the tests

ID ω (rad/s) T(s) H(m)
Reg 1 6.400 0.982 0.023
Reg 2 6.800 0.924 0.020
Reg 3 7.000 0.898 0.019
Reg 4 7.200 0.873 0.018

The tests begin with the cylinders aligned with respect to the y axis and with a distance
of 0.6 m (center-to-center). Once the wave flaps start to move, the servo motor is also
activated and the Body 2 is horizontally displaced from its initial position with a known
prescribed oscillatory motion of frequency ωpm and amplitude Apm = 0.37 m, this value
being the maximum stroke of the ball screw shaft of the mechanical equipment. The
oscillation frequency ωpm was defined as a ratio of the incoming wave frequency ωI . For
each regular wave three different oscillation frequencies of Body 2 were considered, being
these ωpm = ωI /15, ωpm = ωI /30 and ωpm = ωI /60. Therefore, the test matrix considered
12 different cases, which are described in Table 9.3.

Table 9.3: Test Matrix

Oscillator Parameters
Case Wave ωI /ωpm ωpm (rad/s) Tpm (s) Apm (m)
1 Reg 1 15 0.427 14.73 0.37
2 Reg 1 30 0.213 29.45 0.37
3 Reg 1 60 0.107 58.90 0.37
4 Reg 2 15 0.453 13.86 0.37
5 Reg 2 30 0.227 27.72 0.37
6 Reg 2 60 0.113 55.44 0.37
7 Reg 3 15 0.467 13.46 0.37
8 Reg 3 30 0.233 26.93 0.37
9 Reg 3 60 0.117 53.86 0.37
10 Reg 4 15 0.480 13.09 0.37
11 Reg 4 30 0.240 26.18 0.37
12 Reg 4 60 0.120 52.36 0.37

A better understanding of the experimental setup may be achieved by observing the

pictures in Figure 9.9, which displays three video frames recorded during the tests for
Case 10. Observe that at instant t = to s (Figure 9.9(a)), the two bodies are aligned with
respect to the y axis; further on, at t = to + T /4 s (Figure 9.9(b)), the Body 2 is displaced
0.37 m to right. Finally, at instant t = to + 3T /4 s (Figure 9.9(c)), the Body 2 is situated
0.37 m left from its initial position. This sequence of events was repeated four times for
each case described in Table 9.3.
 205

(a) (b) (c)

Figure 9.9: Video frames recorded during the tests with a regular wave of frequency
ωI = 7.2 rad/s and Body 2 oscillation frequency ωpm = 0.48 rad/s. Instants (a) t = to s,
(b) t = to + T /4 s and (c) t = to + 3T /4 s

9.3.2 Numerical Grids for the Case Studies

The simulations are performed with two circular cylinders discretized in 500 quadrilateral
panels. These panel meshes follow the same topology and approximate panel resolution
of the ones used for the calculations presented before in chapter 6, in which a convergence
analysis concerning the verification of the grids was performed.

As in the previous cases, circular free surface meshes were used for the computations shown
next. Nevertheless, since the processing times for the present cases are considerably larger
in comparison to the simulations involving only fixed meshes, different free surface meshes
were applied for each regular wave frequency. As mentioned before, the radius of the free
surface has a direct dependence on the wave length and, therefore, a significant number
of panels can be saved if the meshes are generated for each specific wave frequency.

For the present calculations, the free surface meshes were constructed with radii of two
wave lengths (2λ), in which one wave length was used for damping the waves through the
application of a numerical beach zone set with b = 1.0 and a = 1.0. In this part of the
mesh, a stretching factor was also applied so as to decrease the panel resolution at regions
far away from the bodies positions. Indeed, this procedure also helped on damping the
waves, since the numerical dissipation was intensified by the coarse characteristics of the
region. Moreover, this part of the mesh was also kept fixed along the simulations in order
to decrease the computational time dedicated to the construction of new meshes at each
time step. An example of free surface mesh applied in the computations is presented in
Figure 9.10.
206

Figure 9.10: Free surface mesh for wave Reg 1. Wave frequency ωI = 6.4 rad/s

The main characteristics of each free surface mesh used are presented in Table 9.4. It is
important to mention that the number of panels represents a reference value, since the to-
tal number is not maintained fixed during the simulations, as the meshes are reconstructed
at each time step.

Table 9.4: Main particulars of the panel meshes used in the computations
Wave ωI (rad/s) rf s (m) Free Surface Panels Cylinders Panels Total
Reg 1 6.4 3.01 1581 1000 2581
Reg 2 6.8 2.67 1417 1000 2417
Reg 3 7.0 2.52 1314 1000 2314
Reg 4 7.2 2.38 1246 1000 2246

The coordinate system of the numerical computations follows the same one defined for
the experimental tests. For a wave angle of θ = 270o , the waves propagate in the negative
direction of axis y, as presented in Figure 9.11. Once again, a time step of ∆t = T /60
was considered for the simulations.
 207

WAVES

x2(t)=Apmsin(ωpmt) x
Body Forced rfs
to Move
(Body 2)
Fixed Body
(Body 1)

Figure 9.11: Illustrative view of the numerical setup

Similar to the previous calculations, the simulation begins with a ramp time of Tr = T
s in order to avoid the generation of spurious waves into the domain as well as non-
physical long transient periods in the solution. Moreover, the simulations were run with
the amplitudes of the regular waves considered in the experiments.

9.3.3 Comparison between Experiments and Calculations

This section presents the comparison between the numerical computations and the exper-
imental data for the case studies involving a multi-body system with one of the bodies
undergoing large horizontal relative displacements. Verification of the numerical results
is performed in terms of hydrodynamic forces, moments and wave elevations at three dif-
ferent locations near the bodies. For the sake of conciseness, only the main results, which
represents most of the characteristics observed in both experiment and computations is
presented in this section. The comparison between numerical and experimental results
for all the 12 cases tested (Table 9.3) is presented in Appendix B.

Besides the comparison between the present computations and the experimental measure-
ments, the results presented ahead also include steady-state solutions of several relative
positions between the bodies calculated by the frequency domain software WAMIT. This
approach does not provide forces and motions time series, but their amplitudes according
to the bodies relative positions, which may be used, in its turn, for the construction of
envelope curves. In this regard, the time series calculated by the present method may be
compared to these envelopes in order to evaluate whether the flow memory and transient
208

effects, which is not considered by WAMIT, influence the results or not.

Considering Case 1 as an example, Figures 9.12 and 9.13 present the comparisons between
the present calculations, experimental data and also the envelope amplitude curves pro-
vided by the software WAMIT for the hydrodynamic forces (Fx , Fy and Fz ) and moments
(Mx and My ) in Body 1, respectively. This case considers an incident wave frequency of
ωI = 6.4 rad/s and a Body 2 oscillation frequency of ωpm = ωI /15 = 0.427 rad/s. These
figures also present the position of Body 2 in time, which was used for the synchroniza-
tion between the numerical and experimental results. The circle markers on the envelope
curves indicate the positions of Body 2, for which the WAMIT software was run.

The results show that the hydrodynamic forces and moments Fy , Fz and Mx do not
present significant amplitude variations with the change of relative positions between the
cylinders. Even though, one may realize that the small modulations observed in the
experimental time series were very well captured by the present method. It may also be
visualized that the time domain solver provided slightly better results in relation to the
software WAMIT, specially for the force Fy , where one may observe that the frequency
domain code tend to over-predict the results when Body 2 is located at x2 = 0. This was
somewhat expected, since WAMIT considers that the bodies remain at the same relative
distance for an infinite time.

Larger modulations of the signals amplitudes are clearly observed for the hydrodynamic
force Fx and moment My . As expected, WAMIT provides null values for Fx and My when
the cylinders are in a tandem configuration in x2 = 0 m. This occurs because its solutions
neglect the influence of the wave flow arising from the interaction of the incoming wave
with Body 2 in its previous positions. As a consequence, the WAMIT envelopes also
present a slight shift in time, predicting the minimum and maximum values at different
instants than those observed in the measurements. As can also be observed, the present
time domain code brought a significant improvement to the results, predicting these in-
stants more accurately, since it considers the flow memory during the computations.
 209

Experiment Present Values Wamit

0.5

x2 (m)
0

−0.5
15 20 25 t(s) 30 35 40

2
Fx (N)

0
−2
15 20 25 t(s) 30 35 40

10
Fy (N)

0
−10
15 20 25 t(s) 30 35 40
5
Fz (N)

−5
15 20 25 t(s) 30 35 40

Figure 9.12: Case 1: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic forces Fx , Fy
and Fz

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
15 20 25 t(s) 30 35 40

5
Mx (N.m)

−5
15 20 25 t(s) 30 35 40

0.5
My (N.m)

−0.5
15 20 25 t(s) 30 35 40

Figure 9.13: Case 1: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic moments Mx
and My
210

Analogous results for Cases 4, 7 and 10 are presented in figures 9.14 to 9.19. In all these
cases, the oscillation frequency of Body 2 was also set with ωpm = ωI /15 rad/s. Although
in some of these cases the present numerical method and experimental results exhibit
small deviations, a good agreement is observed overall. An interesting trend is observed
in the time series for Fx and My , in which the present method and the experimental
measurements show that the signal amplitudes are not the same for equal relative dis-
tances. Considering, for instance, the My time series presented in Figure 9.13, one should
notice that the hydrodynamic moment My is larger when Body 2 is approaching Body 1
in comparison to the situation in which the Body 2 is moving away. Again, this behaviour
indicates that the flow memory of the wave field, which is not accounted for by frequency
domain codes, plays an important role for the proper computations of the hydrodynamic
loads, specially when the relative positions of the bodies changes relatively fast along the
simulations. For example, if one observes the results for Case 3 (Figures 9.20 and 9.21)
and Case 5 (Figures 9.22 and 9.23), in which the oscillation frequencies of Body 2 for
Cases 1 and 4 were reduced to ωpm = 6.4/60 = 0.107 rad/s and ωpm = 6.8/30 = 0.2267
rad/s, respectively, it is possible to realize that the signals become more symmetrical with
respect to the mean point of each envelope, presenting approximately the same values
for the same relative distances between the bodies. In addition, as these scenarios are
closer to the one assumed by the software WAMIT, since Body 2 moves more slowly, both
numerical results then exhibit a good agreement with the test data.

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
15 20 25 t(s) 30 35 40

2
Fx (N)

0
−2
15 20 25 t(s) 30 35 40

10
Fy (N)

0
−10
15 20 25 t(s) 30 35 40
5
Fz (N)

−5
15 20 25 t(s) 30 35 40

Figure 9.14: Case 4: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic forces Fx , Fy
and Fz
 211

Experiment Present Values Wamit

0.5

x2 (m)
0

−0.5
15 20 25 t(s) 30 35 40

5
Mx (N.m)

−5
15 20 25 t(s) 30 35 40

0.5
My (N.m)

−0.5
15 20 25 t(s) 30 35 40

Figure 9.15: Case 4: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic moments Mx
and My

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
15 20 25 t(s) 30 35

2
Fx (N)

0
−2
15 20 25 t(s) 30 35
10
Fy (N)

−10
15 20 25 t(s) 30 35
5
Fz (N)

−5
15 20 25 t(s) 30 35

Figure 9.16: Case 7: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic forces Fx , Fy
and Fz
212

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
15 20 25 30 35
t(s)
5
Mx (N.m)

−5
15 20 25 30 35
t(s)

0.5
My (N.m)

−0.5
15 20 25 30 35
t(s)

Figure 9.17: Case 7: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic moments Mx
and My

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
15 20 t(s)25 30 35

2
Fx (N)

0
−2
15 20 t(s) 25 30 35
10
Fy (N)

−10
15 20 t(s) 25 30 35
5
Fz (N)

−5
15 20 t(s) 25 30 35

Figure 9.18: Case 10: Comparison of numerical and experimental results in terms of
position of body 2 in time and the associated time series of the hydrodynamic forces Fx ,
Fy and Fz
 213

Experiment Present Values Wamit

0.5

x2 (m)
0

−0.5
15 20 25 30 35
t(s)
2
Mx (N.m)

−2
15 20 25 30 35
t(s)
0.5
My (N.m)

−0.5
15 20 25 30 35
t(s)

Figure 9.19: Case 10: Comparison of numerical and experimental results in terms of
position of body 2 in time and the associated time series of the hydrodynamic moments
Mx and My

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
60 80 100 t(s)120 140 160
5
Fx (N)

−5
60 80 100 t(s)120 140 160
20
Fy (N)

−20
60 80 100 120 140 160
t(s)
5
Fz (N)

−5
60 80 100 120 140 160
t(s)

Figure 9.20: Case 3: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic forces Fx , Fy
and Fz
214

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
60 80 100 120 140 160
t(s)
5
Mx (N.m)

−5
60 80 100 120 140 160
t(s)
1
My (N.m)

−1
60 80 100 120 140 160
t(s)

Figure 9.21: Case 3: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic moments Mx
and My

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
30 35 40 45 50 t(s) 55 60 65 70 75 80

2
Fx (N)

0
−2
30 35 40 45 50 t(s)55 60 65 70 75 80
10
Fy (N)

−10
30 35 40 45 50 t(s)55 60 65 70 75 80
5
Fz (N)

−5
30 35 40 45 50 t(s)55 60 65 70 75 80

Figure 9.22: Case 5: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic forces Fx , Fy
and Fz
 215

Experiment Present Values Wamit

0.5

x2 (m)
0

−0.5
30 35 40 45 50 55 60 65 70 75 80
t(s)
5
Mx (N.m)

−5
30 35 40 45 50 55 60 65 70 75 80
t(s)

0.5
My (N.m)

−0.5
30 35 40 45 50 55 60 65 70 75 80
t(s)

Figure 9.23: Case 5: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the hydrodynamic moments Mx
and My

Considering again the results for Case 1, Figure 9.24 presents comparisons in terms of
numerical and experimental energy response spectra for Fx and My , from which one may
realize the presence of energy in two other frequency bands besides the incident wave one.
This fact reveals that the oscillatory motion of Body 2 induces a Doppler effect in the
Body 1 measured signals, which increases and reduces the wave frequency when Body 2
is approaching and moving away from Body 1, respectively. In fact, since Body 1 is a
circular cylinder, the force Fx and moment My induced directly by the interaction with
the incoming wave tend to vanish, remaining only the force/moment components induced
by the waves diffracted from Body 2. Thus, in this problem, Body 2 works as a moving
source that emits free surface waves at a constant frequency ωI . As a result, when Body
2 is approaching Body 1 from negative x coordinates, the wave fronts begin to cluster on
the right side and spread further apart on the left side of Body 2, which modifies the wave
frequency perceived by Body 1.

As also indicated in Figure 9.24, these frequency components are calculated by the sum
and subtraction of ωpm with the incident wave frequency ωI , which was verified by observ-
ing also the Fx and My energy spectra for other cases, such as Cases 2 and 3, in which the
oscillatory motion frequencies of Body 2 were reduced to ωpm = 6.4/30 = 0.213 rad/s and
ωpm = 6.4/60 = 0.107 rad/s, respectively. Once more, the results provided by the present
216

method recovered very well the experimental data, capturing with reasonable accuracy
the physics of the problem. The same behavior was also observed for the other cases.

Experiment Present Values

5

4
ω+ ≈ ωI + ωpm ≈ 6.83 rad/s
SFx (N 2 s)

2 ωI = 6.4 rad/s
ω− ≈ ωI − ωpm ≈ 5.97 rad/s
1

0
4.5 5 5.5 6 6.5 7 7.5 8
ω (rad/s)

0.4

0.3
SMy (N 2 m2 s)

ω+ ≈ ωI + ωpm ≈ 6.83 rad/s

0.2
ωI = 6.4 rad/s
0.1 ω− ≈ ωI − ωpm ≈ 5.97 rad/s

0
4.5 5 5.5 6 6.5 7 7.5 8
ω (rad/s)

Figure 9.24: Case 1: Comparison between present method and experimental data of Fx
and My response spectra
 217

Experiment Present Values

8

6
SFx (N 2 s)
ω+ ≈ ωI + ωpm ≈ 6.61 rad/s
ω− ≈ ωI − ωpm ≈ 6.19 rad/s
4

2 ωI = 6.4 rad/s

0
4.5 5 5.5 6 6.5 7 7.5 8
ω (rad/s)

0.4

0.3
ω+ ≈ ωI + ωpm ≈ 6.61 rad/s
SM y (N 2 m2 s)

ω− ≈ ωI − ωpm ≈ 6.19 rad/s

0.2

0.1 ωI = 6.4 rad/s

0
4.5 5 5.5 6 6.5 7 7.5 8
ω (rad/s)

Figure 9.25: Case 2: Comparison between present method and experimental data of Fx
and My response spectra

20
Experiment Present Values
15
SFx (N 2 s)

10 ω+ ≈ ωI + ωpm ≈ 6.51rad/s
ω− ≈ ωI − ωpm ≈ 6.29 rad/s

5
ωI = 6.4 rad/s

0
4.5 5 5.5 6 6.5 7 7.5 8
ω (rad/s)

0.8

0.6
ω+ ≈ ωI + ωpm ≈ 6.51rad/s
SMy (N 2 m2 s)

0.4
ω− ≈ ωI − ωpm ≈ 6.29 rad/s
0.2 ωI = 6.4 rad/s

0
4.5 5 5.5 6 6.5 7 7.5 8
ω (rad/s)

Figure 9.26: Case 3: Comparison between present method and experimental data of Fx
and My response spectra
218

Examples of comparisons between numerical and experimental wave elevations for wave
probes WP1, WP2 and WP3 (see WPs positions in Figure 9.4 and 9.8) for Cases 1, 4, 7
and 10 are presented in Figures 9.27, 9.28, 9.29 and 9.30, respectively. In all these cases
the oscillation frequencies of Body 2 were set to ωpm = ωI /15 rad/s. One may realize that
the wave elevations measured at the WP positioned behind Body 1 (WP3 - ηP 3 ) present
a very constant behavior, showing to be independent on the translational movements of
Body 2. In fact, this wave probe is located at a sheltered area provided by Body 1, which
always results in low wave amplitude values for all cases tested.

On the other hand, small amplitude modulations may be observed for WP1 and WP2,
in which again very good agreements are observed between the time series computed by
the present method and the experimental results. Indeed, the present method presented
a slightly better performance in relation to the WAMIT results, predicting with more
accuracy the wave elevations measured by WP2 (ηP 2 ). As one may observe, for this wave
probe the frequency domain results present higher wave amplitudes in relation to the time
domain and experimental measurements when Body 2 is approaching Body 1. Once more,
this behaviour tends to be minimized with the increasing of Body 2 oscillation period as
exemplified by the wave elevations for Cases 3 and 8 displayed in Figures 9.31 and 9.32,
respectively. For these cases the translational oscillation frequencies of Body 2 of cases
1 and 7 were reduced to ωpm = 6.0/60 = 0.107 rad/s and ωpm = 7.0/30 = 0.233 rad/s,
respectively.
 219

Experiment Present Values Wamit

0.5

x2 (m)
0

−0.5
15 20 25 t(s) 30 35 40
0.02
ηP 1 (m)

−0.02
15 20 25 t(s) 30 35 40
0.02
ηP 2 (m)

−0.02
15 20 25 t(s) 30 35 40
0.02
ηP 3 (m)

−0.02
15 20 25 t(s) 30 35 40

Figure 9.27: Case 1: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the wave elevations at WP1, WP2
and WP3

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
15 20 25 t(s) 30 35 40
0.02
ηP 1 (m)

−0.02
15 20 25 t(s) 30 35 40

0.02
ηP 2 (m)

−0.02
15 20 25 t(s) 30 35 40
0.01
ηP 3 (m)

−0.01
15 20 25 t(s) 30 35 40

Figure 9.28: Case 4: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the wave elevations at WP1, WP2
and WP3
220

Experiment Present Values Wamit

0.5
x2 (m)
0

−0.5
15 20 25 t(s) 30 35
0.02
ηP 1 (m)

−0.02
15 20 25 t(s) 30 35
0.02
ηP 2 (m)

−0.02
15 20 25 t(s) 30 35
0.01
ηP 3 (m)

−0.01
15 20 25 t(s) 30 35

Figure 9.29: Case 7: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the wave elevations at WP1, WP2
and WP3

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
15 20 25 t(s) 30 35
0.02
ηP 1 (m)

−0.02
15 20 25 t(s) 30 35
0.02
ηP 2 (m)

−0.02
15 20 25 t(s) 30 35
0.01
ηP 3 (m)

−0.01
15 20 25 t(s) 30 35

Figure 9.30: Case 10: Comparison of numerical and experimental results in terms of
position of body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3
 221

Experiment Present Values Wamit

0.5

x2 (m)
0

−0.5
60 80 100 t(s)120 140 160
0.02
ηP 1 (m)

−0.02
60 80 100 t(s)120 140 160
0.05
ηP 2 (m)

−0.05
60 80 100 120 140 160
t(s)
0.02
ηP 3 (m)

−0.02
60 80 100 t(s)120 140 160

Figure 9.31: Case 3: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the wave elevations at WP1, WP2
and WP3

Experiment Present Values Wamit

0.5
x2 (m)

−0.5
30 35 40 45 50 55 60 65 70 75 80
t(s)
0.02
ηP 1 (m)

−0.02
30 35 40 45 50 55 60 65 70 75 80
t(s)
0.02
ηP 2 (m)

−0.02
30 35 40 45 50 55 60 65 70 75 80
t(s)
0.01
ηP 3 (m)

−0.01
30 35 40 45 50 55 60 65 70 75 80
t(s)

Figure 9.32: Case 8: Comparison of numerical and experimental results in terms of posi-
tion of body 2 in time and the associated time series of the wave elevations at WP1, WP2
and WP3
222

Finally, examples of wave elevation patterns for Case 1 at two different instants asso-
ciated to the maximum wave amplitudes at WP1 and WP2 are illustrated in Figures
9.33 and 9.34, respectively. One should notice that, although the maximum wave heights
of these wave probes occur at different instants, in both cases the Body 2 is positioned
approximately in front of Body 1.

(a) (b)

Figure 9.33: Case 1: Two snapshots of the simulation illustrating the maximum wave
elevation calculated at WP1 (t = 36.23) s. Perspective (a) and front views (b). Vertical
black line refers to the position of the wave probe

(a) (b)

Figure 9.34: Case 1: Two snapshots of the simulation illustrating the maximum wave
elevation calculated at WP2 (t = 36.72) s. Perspective (a) and front views (b). Vertical
black line refers to the position of the wave probe
 223

Chapter 10

Conclusions

10.1 Conclusions

This work presented the development and verification of a numerical code that deals with
wave-body interactions in time domain under the assumption of potential flow theory. The
original problem is nonlinear and requires the imposition of boundary conditions on time-
dependent surfaces which are not known prior to the problem solution as, for example,
the free surface elevation and the position of the floating body surface at each instant of
time. These features make the problem very difficult to be solved, being the focus of a
vast number of articles which describe methods that deal with this nonlinear problem.
Nevertheless, their applicabilities are still restricted to bi-dimensional problems or very
simple three-dimensional geometries, which motivated us to apply a linear wave-body
formulation as a first stage of the research.

Under the assumption of waves with small amplitude and steepness, and that the bodies do
not present large motions around their mean positions, the set of equations were linearized
by expansions using Stoke’s and Taylor’s series. Consequently, the total velocity potential
could be decomposed in a sum of a disturbed wave potential and an analytic solution of
an incoming wave field, avoiding the necessity of including a numerical wave-maker to
generate the waves and making possible to reduce the number of panels by concentrating
them around the bodies.

Because of the accuracy required for the calculation of the velocity potential time deriva-
tive, the inclusion of a second Initial Boundary Value Problem for the so-called acceleration
224

potential was made. After a deep research in the literature, we concluded that this method
is the most accurate scheme for the calculation of the velocity potential time derivative
and consequently for the pressure field which is essential for a consistent formulation in
time domain. Although this method is often mentioned in the literature, results with its
application in three-dimensional problems are rarely reported.

A Low Order BEM solver formulated in terms of Green’s second identity was then de-
veloped transforming the original integral equations in two linear systems of algebraic
equations. The integration of the influence coefficients received special attention along
this topic, illustrating the difficult convergence of the integrals when field and collocation
points are close to each other. In order to overcome this problem, a method that split
up the quadrilateral panels into triangles was implemented, so as to desingularize the
integrand function, transforming it into a smoother function. With this approach, the
integrals convergence were reached with a significant reduction of gaussian points in com-
parison to the direct integration method. The method also holds for triangular panels,
treating the problem as a quadrilateral panel with two vertices collapsed.

Additional numerical schemes which are important for the performance of the numeri-
cal method were also discussed as, for example, the inclusion of dissipation terms on a
portion of the free surface to avoid wave reflections from the domain boundaries and the
inclusion of a ramp function that acts in the first seconds of the simulation in order to
provide the solution with a gradual increase, to reduce the transient periods and to avoid
impulsive responses of the dynamic system. The time scheme and the updating of the
boundary conditions were conducted with the widely used RK4 scheme, which provides
high accuracy and large stability regions.

The numerical code was first tested by calculating the hydrodynamic loads on a hemi-
sphere and a circular cylinder, solving the well-known problems of diffraction and radia-
tion in waves. In addition, free floating simulations were also conducted in order to verify
whether the procedures adopted to couple the fluid and body equations were correctly
implemented. Convergence of the results when increasing the number of panels was also
verified by performing several simulations with meshes of different panel resolutions. As
expected, the larger the number of panels the lower were the relative errors between the
present calculations and benchmark values, which were provided by analytic solutions and
by the software WAMIT. As could be observed, these comparisons presented very good
agreement, attesting that the numerical code is indeed capable of predicting the first-order
hydrodynamic loads and motions associated to the wave-body interaction problem.

One step forward, free floating simulations with two different FPSO hull types were also
conducted. The present computations were verified through comparisons of motion RAOs
 225

obtained numerically with WAMIT or experimentally in the CH-TPN-USP. In the sim-

ulations, the RAOs were obtained considering both monochromatic regular waves and
a white-noise spectrum technique that made possible to calculate the RAOs in a very
efficient way, performing only one simulation per wave incidence angle. Despite some
small deviations, specially concerning resonance peak values, the results presented good
agreement with both WAMIT and the experimental data. Moreover, a direct compari-
son of numerical and experimental time series was also presented, in which not only the
amplitudes but also the phases between the motions could be verified.

The performance of the method when dealing with a multi-body problem involving two
bodies in side-by-side arrangement, as, for example, in a FLNG offloading operation, was
also reported. Aiming at obtaining benchmark data, an experimental campaign was car-
ried out at ETSIN, in Spain. The tests were conceived with a very fundamental setup,
considering only regular waves and just one wave direction. In addition, the setup con-
sidered one of the bodies fixed (barge) and the other (geosim) restrained in three D.O.F.
(sway, roll and yaw) in order to maintain the gap width fixed during the measurements.
Results were discussed in terms of the geosim motions and wave amplitudes in the gap.
The numerical results presented undesirable overestimations of the experimental values
for a range of frequencies, which showed to be dependent on the gap width. In fact, the
numerical and experimental results have shown that the resonant frequencies were lower
for the system with the largest gap width. Moreover, different resonant frequencies were
observed for heave and pitch motions. As demonstrated by the wave envelopes inside
the gap, the heave motion is significantly amplified due to the presence of a piston-type
mode inside the gap, whereas the pitch motion is influenced by the occurrence of a second
longitudinal mode. The same conclusions could be drawn in this respect for both gap
widths.

Still concerning the side-by-side problem, it was possible to observe that the numerical
model presented numerical convergence problems when simulating wave frequencies near
to the gap resonant ones. In these cases, the time series for motions and wave elevations
showed a long transient period, attesting the numerical problems to reach the steady-state.
This problem was solved by the application of a damping lid method, which incorporates
a damping factor in the free surface boundary conditions. By considering this method,
the time series reached the steady state much faster. This is indeed a positive indication
regarding future applications of the method to the analysis of multi-vessels in irregular
wave conditions. In addition, despite the simplicity of the damping model, the use of the
damping lid technique has also improved the numerical results, reducing the discrepancies
observed with the experimental data. Nevertheless, it should be noticed that the choice
of the damping parameter has been done heuristically and for a better insight about the
influence of this parameter on the numerical solutions, further analysis is still required.
226

Finally, the text is concluded with the presentation of a new method to care with multi-
body hydrodynamic interactions of bodies undergoing large relative displacements, a prob-
lem that cannot be assessed directly with frequency domain codes. As an important part
of this development, a generator of unstructured triangular panel meshes that could be
integrated in the time-loop of the code was implemented, so as to account for changes of
the relative positions between the bodies during the calculations. Moreover, an specific
higher order interpolation algorithm had to be considered to proper recover the solutions
of a previous time-step and to enable the progressing of the calculations with reasonable
accuracy.

Benchmark data was obtained through the conduction of fundamental experimental tests
in the CH-TPN-USP dedicated to the evaluation of this specific problem. The tests
considered two circular cylinders, being one fixed and attached to a 6 D.O.F load cell,
and the other coupled to a mechanical device that was used to impose large and slow
horizontal movements on this body. In order to evaluate the influence of the prescribed
oscillatory motion frequency on the results, for each regular wave three different oscillation
frequencies were considered, these being defined as a ratio of the incident wave frequency.

Despite the fundamental configuration of the tests, the experiments presented a relatively
high degree of complexity. Since the main objective was to evaluate the hydrodynamic
interaction loads between the bodies, which are indeed very small in relation to the forces
induced by the incident wave itself, special care had to be taken in order to minimize the
influence of wave reflections from the tank walls. In fact, this was the main difficulty of
the experimental campaign, since each test case demanded a long period of acquisition,
which, by its turn, increased the possibility of wave reflections effects on the results. For
this reason, the tests were conducted considering regular waves in a frequency range, for
which the CH-TPN-USP active absorption system presents its best performance.

The new numerical implementation was then verified by reproducing the cases tested
experimentally. Results were compared in terms of hydrodynamic loads calculated in
the fixed body and wave elevations on three wave probes that were positioned upstream,
in between and downstream of the bodies. Through these comparisons, it was possible
to verify that when the most rapid prescribed oscillatory motion tested was considered,
the time series for transversal forces/moments (in relation to the incident wave angle)
were not symmetric, meaning that the signals presented different amplitudes for the same
relative distances between the bodies. In fact, the analysis showed that this was occurring
due to a Doppler effect induced by the large oscillatory motion of the second body, a
physical behaviour that was very well captured by the time domain code. Moreover, this
result also emphasized the importance of considering a transient solution of the problem,
which cannot be calculated directly by a frequency domain code. In general, as could be
 227

observed, the present method was able to reproduce the experimental data for all cases
tested, illustrating that the variation of the relative positions of the bodies in time domain
are being correctly modeled.

10.2 Proposed Improvements and Future Work

Based on the conclusions of this thesis, the following tasks are considered as important
future works for the improving of the method and its applicability:

• Development of a higher order method, in which both the geometry and the solutions
are described in terms of Non-Uniform Rational B-Spline (NURBS) surfaces. With
this approach, the study of wave-current interaction effects and the second order
problem become more suitable to be tackled, since the spatial derivatives of the
velocity potential may be calculated straightforwardly.

• Investigation of other techniques that consider the inclusion of damping coefficients

to emulate the viscous effects in the side-by-side problem, as well as different pos-
sibilities to calibrate these coefficients. In the method applied in this thesis, a
preliminary adjustment of the damping coefficient was performed by matching the
numerical results with the experimental data. However, a consistent calibration
of these coefficients still demand more investigations of the problem, which may
require, for example, the conduction of additional experimental campaigns.

• Implementation of drift forces into the numerical method that deals with multi-body
systems with large relative displacements. This development would make possible
the simulation of real operations, involving real body geometries, such as FPSOs,
Semi-submersible etc. Validation of the results, however, would be a critical task,
requiring a very specific and complex experimental test.

• Integration of the present method to a mooring line code in order to perform simu-
lations of moored, single or multi-body, systems situated on deep sea waters, where
the inertial and damping effects of mooring lines and risers become nontrivial and
the body dynamics can be appreciably affected by them. For instance, this develop-
ment could be performed in a dynamic simulator, such as the TPN of the University
of Sao Paulo, in which, currently, the wave effects are considered decoupled of the
equations of motion by using linear hydrodynamic coefficients previously calculated
in the frequency domain software WAMIT. As previously discussed in this thesis,
this approach brings limitations for solving multi-body transient problems, which
228

could be partially solved by coupling the hydrodynamic calculations to the bodies

equations of motion solved by the dynamic simulator.

• Development of a combined BEM-FEM scheme for the solution of hydroelastic re-

sponse of flexible seagoing ultra large vessels. As stated by Kim, Kim and Kim
(2009), as the sizing of ships is getting larger, their structural natural frequencies
generally tends to move down to a smaller frequency range. In addition, forward
speed makes the encounter excitation frequency of waves to move much closer to
their natural frequency range, leading to higher chance of resonance between the
two even under linear wave regime.
 229

References

BAI, W.; TAYLOR, R. E. Fully nonlinear simulation of wave interaction with fixed and
floating flared structures. Ocean Engineering, v. 36, p. 223–236, 2009.

BATCHELOR, G. K. An Introduction to Fluid Dynamics. : Cambridge University

Press, 1967.

BECK, R. F.; CAO, Y.; LEE, T.-H. Fully nonlinear water wave computations using the
desingularized method. In: Proceedings of the Sixth International Conference on
Numerical Ship Hydrodynamics. 1994.

BECK, R. F.; REED, A. M. Modern seakeeping computations for ships. In: Proceedings
23rd ONR Symposium on Naval Hydrodynamics. 2000.

BERTRAM, V. Practical Ship Hydrodynamics. : Butterworth-Heinemann, 2000.

BHATTA, D.; RAHMAN, M. On scattering and radiation problem for a cylinder in water
of finite depth. International Journal of Engineering Science, v. 41, p. 931–967,
2003.

BOO, S. Y. Linear and nonlinear irregular waves and forces in a numerical wave tank.
Ocean Engineering, v. 29, p. 475–493, 2002.

BUCHNER, B.; DIJK, A. V.; DE WILDE, . J. J. Numerical multiple-body simulations of

side-by-side mooring to a FPSO. In: In Proceedings of the 21st ISOPE Conference.
p. 343–353, 2011.

BUNNIK, T. A simulation approach for large relative motions of multi-body offshore op-
erations in waves. In: Proceedings of the ASME 2014 33rd International Confer-
ence on Ocean, Offshore and Arctic Engineering, OMAE2014, San Francisco,
California, USA. 2014.
230

BUNNIK, T.; PAUW, W.; VOOGT, A. Hydrodynamic analysis for side-by-side offloading.
In: In Proceedings of the 19th ISOPE Conference. 2009.

BUNNIK, T. H. J. Seakeeping Calculations for Ships, taking into account the

Non-linear Steady Waves. 1999. Thesis (PhD) — Technische Universiteit Delft, 1999.

CAO, Y.; R., B.; SCHULTZ, W. Non-linear motions of floating bodies in incident waves.
In: 9th Workshop on Water Waves and Floating Bodies. 1994.

CHEN, C.-B. Hydrodynamic analysis for offshore LNG terminals. In: In Proceedings
of the 15th ISOPE Conference. 2005.

CLAUSS, G. F.; DUDEK, M.; TESTA, D. Gap effects at side-by-side LNG-transfer oper-
ations. In: In OMAE 2013-10749, Proceedings of the 32nd International Con-
ference on Ocean, Offshore and Arctic Engineering. 2013.

CLéMENT, A. Coupling of two absorbing boundary conditions for 2d time-domain sim-

ulations of free surface gravity waves. Journal of Computational Physics, v. 126, p.
139–151, 1996.

COINTE, R. Nonlinear simulation of transient free surface flows. In: Fifth International
Conference on Numerical Ship Hydrodynamics. 1990.

COINTE, R.; GEYER, P.; KING, B.; MOLIN, B.; TRAMONI. Nonlinear and linear
motions of a rectangular barge in a perfect fluid. In: Eighteenth Symposium on Naval
Hydrodynamics. 1991.

CUMMINS, W. The Impulsive Response Function and Ship Motions. : Depart-

ment of the Navy David Taylor Model Basin, (1962).

DANMEIER, D. G. A Higher-Order Panel Method for Large-Amplitude Simula-

tions of Bodies in Waves. 1999. Thesis (PhD) — Massachusetts Institute of Technology,
1999.

DAWSON, C. W. A practical computer method for solving ship-wave problems. In: Pro-
ceedings of the Second International Conference on Numerical Ship Hydro-
dynamics. 1977.
 231

ELIE, B.; RELIQUET, G.; P.-E., G.; THILLEUL, O.; FERRANT, P.; GENTAZ, L.;
LEDOUX, A. Simulation of the gap resonance between two rectangular barges in regular
waves by a free surface viscous flow solver. In: Proceedings of the ASME 2013
32nd International Conference on Ocean, Offshore and Artic Engineering,
OMAE2013, Nantes, France. 2013.

FALTINSEN, O. M. Sea Loads on Ships and Offshore Structures. : Press Syndicate

of the University of Cambridge, 1990.

FERRANT, P. Simulation of strongly nonlinear wave generation and wave-body interac-

tions using a 3-D MEL model. In: Twenty-First Symposium on Naval Hydrody-
namics. 1997.

FOURNIER, J. R.; NACIRI, M.; CHEN, X. B. Hydrodynamics of two side-by-side ves-

sels. Experiments and numerical simulations. In: In Proceedings of the 16th ISOPE
Conference. 2006.

GADD, G. E. A method of computing the flow and surface wave pattern around full
forms. In: The Royal Institution of Naval Architects. 1975.

HESS, J. L.; SMITH, A. M. O. Calculation of potential flow about arbitrary bodies.

Progress in Aerospace Sciences, p. 1–138, 1967.

HONG, S. Y.; NAM, B. Analysis of 2nd-order wave force on floating bodies using FEM
in time domain. International Journal of Offshore and Polar Engineering, v. 21,
p. 22–28, 2011.

HUANG, Y. Nonlinear Ship Motions by a Rankine Panel Method. 1997. Thesis

(PhD) — Massachusetts Institute of Technology, 1997.

HUIJSMANS, R. H. M.; PINKSTER, J. A.; DE WILDE, . J. J. Diffraction and radiation

of waves around side-by-side moored vessels. In: In Proceedings of the 11th ISOPE
Conference. p. 406–412, 2001.

HULME, A. The wave forces acting on a floating hemisphere undergoing forced periodic
oscillations. Journal of Fluid Mechanics, v. 121, p. 443–463, 1982.
232

ISRAELI, M.; ORSZAG, S. A. Approximation of radiation boundary conditions. Journal

of Computational Physics, v. 41, p. 115–135, 1981.

JOHN, F. On the motion of floating bodies, I. Communications on Pure and Applied

Mathematics, v. 2, p. 13–57, 1949.

JOHN, F. On the motion of floating bodies, II. Communications on Pure and Ap-
plied Mathematics, v. 3, p. 45–101, 1950.

KASHIWAGI, M.; ENDO, K.; YAMAGUCHI, H. Wave drift forces and moments on two
ships arranged side by side in waves. Ocean Engineering, v. 32, p. 529–555, 2005.

KIM, G.-H.; LEE, C.-S.; KERWIN, J. A B-spline based higher order panel method for
analysis of steady flow around marine propellers. Ocean Engineering, v. 34, p. 2045–
2060, 2007.

KIM, K.-H.; KIM, Y. Time-domain analysis of motion responses of adjacent multiple

floating bodies in waves. In: In Proceedings of the 18th ISOPE Conference. 2008.

KIM, M. W.; KOO, W.; HONG, S. Y. Numerical analysis of various artificial damping
schemes in a three-dimensional numerical wave tank. Ocean Engineering, v. 75, p.
165–173, 2014.

KIM, Y.; KIM, K.-H.; KIM, Y. Springing analysis of a seagoing vessel using fully coupled
bem-fem in the time domain. Ocean Engineering, v. 36, p. 785–796, 2009.

KIM, Y.; KRING, D. C.; SCLAVOUNOS, P. D. Linear and nonlinear interactions of

surface waves with bodies by a three-dimensional rankine panel method. Applied Ocean
Research, v. 19, p. 235 – 249, 1997.

KOO, B.; KIM, M. Hydrodynamic interactions and relative motions of two floating plat-
forms with mooring lines in side-by-side offloading operation. Applied Ocean Research,
v. 27, p. 292–310, 2005.

KOO, W.; KIM, M.-H. Freely floating-body simulation by a 2D fully nonlinear numerical
wave tank. Ocean Engineering, v. 31, p. 2011–2046, 2004.
 233

KORSMEYER, F.; BINGHAM, H.; NEWMAN, J. TiMIT - A Panel Method for

Transient Wave-Body Interactions. : Massachusetts Institute of Technology, (1999).

KRING, D. C. Time Domain Ship Motions by a Three-Dimensional Rankine

Panel Method. 1994. Thesis (PhD) — Massachusetss Institute of Technology, 1994.

KRISTIANSEN, T.; FALTINSEN, O. M. A two-dimensional numerical and experimental

study of resonant coupled ship and piston-mode motion. Applied Ocean Research,
v. 32, p. 158–176, 2010.

KRISTIANSEN, T.; FALTINSEN, O. M. Gap resonance analyzed by a new domain-

decomposition method combining potential and viscous flow draft. Applied Ocean Re-
search, v. 34, p. 198–208, 2012.

LAMB, H. Hydrodynamics. : Dover, 1945.

LEE, C.-H.; NEWMAN, J. N. Computation of wave effects using the panel method. In:
CHAKRABARTI, S. K. (Ed.). Numerical Models in Fluid-Structure Interaction.
: WIT Press, 2005.

LIU, W.-M.; XUE, M.; YUE, D. K.-P. Computations of fully nonlinear three-dimensional
wave-wave and wave-body interactions. Part 2. nonlinear waves and forces on a body.
Journal of Fluid Mechanics, v. 438, p. 41–66, 2001.

LIU, Y.; XUE, M.; YUE, D. K.-P. Computations of fully nonlinear three-dimensional
wave-wave and wave-body interactions. Part 1. dynamics of steep three-dimensional waves.
Journal of Fluid Mechanics, v. 438, p. 11–39, 2001.

LONGUET-HIGGINS, M. S.; COKELET, E. D. The deformation of steep surface waves

on water. I. A numerical method of computation. Proceedings of the Royal Society
of London, v. 350, p. 1–26, 1976.

MANIAR, H. D. A three dimensional higher order panel mehod based on B-

splines. 1995. Thesis (PhD) — Massachussetts Institute of Technology, 1995.

MEI, C. C.; STIASSNIE, M.; YUE, D. K.-P. Theory and Applications of Ocean
Surface Waves. Part 1: Linear Aspects. : World Scientific Publishing Co. Pte. Ltd,
2005.
234

MEI, C. C.; STIASSNIE, M.; YUE, D. K.-P. Theory and Applications of Ocean
Surface Waves. Part 2: Nonlinear Aspects. : World Scientific Publishing Co. Pte.
Ltd, 2005.

MELLO, P. C.; CARNEIRO, M. L.; TANNURI, E. A.; NISHIMOTO, K. USP active

absorption wave basin: From conception to commissioning. In: Proceedings of the
ASME 2010 29th International Conference on Ocean, Offshore and Artic En-
gineering, OMAE2010, Shangai, China. 2010.

MOLIN, B. On the piston and sloshing modes in moonpools. Journal of Fluid Me-
chanics, v. 430, p. 27–50, 2001.

MOLIN, B.; REMY, F.; CAMHI, A.; LEDOUX, A. Experimental and numerical study
of the gap resonances in-between two rectangular barges. In: 13th Congreess of Intl
Maritime Association of Mediterranean IMAM 2009, Instanbul, Turkey. 2009.

MOLIN, B.; REMY, F.; KIMMOUN, O.; STASSEN, Y. Experimental study on the wave
propagation and decay in a channel through a rigid ice-sheet. Applied Ocean Research,
v. 24, p. 247–260, 2002.

NACIRI, M.; WAALS, O.; DE WILDE, . de J. J. Time domain simulations of side-

by-side moored vessels lessons learnt from a benchmark test. In: Proceedings of the
OMAE2007 26th International Conference on Offshore Mechanics and Artic
Engineering. 2007.

NAKOS, D. E. Ship Wave Patterns and Motions by a Three Dimensional Rank-

ine Panel Method. 1990. Thesis (PhD) — Massachusetts Institute of Technology, 1990.

NEWMAN, J. N. Marine Hydrodynamics. : The MIT Press, 1977.

NEWMAN, J. N. Panel methods in marine hydrodynamics. In: 11th Australian Fluid

Mechanics Conference. 1992.

NEWMAN, J. N. Application of generalized modes for the simulation of free

surface patches in multi body hydrodynamics. : Wamit Consortium Report, (2003).

NISHIMOTO, K.; FERREIRA, M. D.; MARTINS, M. R.; MASETTI, I. Q.; FILHO, P. D.

M.; RUSSO, A. A.; CALDO, J.; SILVEIRA, S. S. Numerical offshore tank: Development
 235

of numerical offshore tank for ultra deep water oil production systems. In: Proceedings
of the 22nd International Conference on Offshore Mechanics Artic Engineer-
ing. 2003.

ORLANSKI, J. A simple boundary condition for unbounded hyperbolic flows. Journal

of Computational Physics, v. 21, p. 251–269, 1976.

PAUW, W. H.; HUIJSMANS, R. H. M.; VOOGT, A. Advances in the hydrodynamics

of side-by-side moored vessels. In: Proceedings of the 26th International Confer-
ence on Offshore Mechanics and Artic Engineering, OMAE 2007, San Diego,
California, USA. 2007.

PERSSON, P.-O. Mesh Generation for Implicit Geometries. 2005. Thesis (PhD) —
Massachusetts Institute of Technology, 2005.

PINKSTER, J. A. Low Frequency Second Order Wave Exciting Forces on Float-

ing Structures. 1980. Thesis (PhD) — Technische Universiteit Delft, 1980.

PRINS, H. Time-Domain Calculations of Drift Forces and Moments. 1995. Thesis

(PhD) — Tecnische Universiteit Delft, 1995.

QIU, W. A Panel-Free Method for Time-Domain Analysis of Floating Bodies

in Waves. 2001. Thesis (PhD) — Dalhousie University, 2001.

QUEIROZ FILHO, . A.; TANNURI, E. DP offloading operation: A numerical evaluation

of wave shielding effect. In: 8th Conference on Manouvering and Control Marine
Craft (MCMC2009). 2009.

ROMATE, J. E. The numerical simulation of nonlinear gravity waves. Engineering

Analysis with Boundary Elements, v. 7, p. 156–166, 1990.

ROMATE, J. E. Absorbing boundary conditions for free surface waves. Journal of Com-
putational Physics, v. 99, p. 135–145, 1992.

SARPKAYA, T.; ISAACSON, M. Mechanical of Waves Forces on Offshore Struc-

tures. : Van Nostrand Reinhold Company, 1981.
236

SCLAVOUNOS, P. D.; NAKOS, D. E. Stability analysis of panel methods for free-surface

flows with forward speed. In: 17th Symposiumn on Naval Hydrodynamics. 1993.

SERVÁN-CAMAS, B.; GARCı́A-ESPINOSA, J. Accelerated 3d multi-body seakeeping

simulations using unstructured finite elements. Journal of Computational Physics,
v. 252, p. 382–403, 2013.

SHAO, Y.-L. Numerical Potential-Flow Studies on Weakly-Nonlinear Wave-

Body Interactions with/without small Forwar Speeds. 2010. Thesis (PhD) —
Norwegian University of Science and Technology, 2010.

SINGH, S. P.; SEN, D. A comparative linear and nonlinear ship motion study using a
3-D time domain methods. Ocean Engineering, v. 34, p. 1863–1881, 2007.

STOKER, J. J. Water Waves. : Interscience Publishers, 1957.

STOKES, G. On the theory of oscillatory waves. Mathematical and Physical Papers

- Reprinted from the Original Journals and Transactions, v. 8, p. 441–455, 1847.

SUN, L.; TAYLOR, R. E.; TAYLOR, P. First and second-order analysis of resonant waves
between adjacent barges. Journal of Fluid Strucutures, v. 26, p. 954–978, 2010.

TANIZAWA, K. A nonlinear simulation method of 3-D body motions in waves. Journal

of the Society of Naval Architects of Japan, 1995.

TANIZAWA, K. The state of the art on numerical wave tank. In: Proceeding of 4th
Osaka Colloquium on Seakeeping Performance of Ships. p. 95–114, 2000.

TANNURI, E. A.; BRAVIN, T. T.; SIMOS, A. N.; ALVES, K. H.; NISHIMOTO, K.;
FERREIRA, M. D. Dynamic simulation of offloading operation considering wave interac-
tions between vessels. In: Proceedings of OMAE04 23nd International Conference
on Offshore Mechanics and Artic Engineering. 2004.

TAYLOR, R. E.; WU, G. Transient motion of a floating body in steep water waves. In:
Proceedings of the 11th International Workshop on Water Waves and Floating
Bodies. 1996.
 237

THOMSON, L. N. M. Theoretical Hydrodynamics. : The Macmillan and Company,

1968.

VAN DAALEN, . E. F. G. Numerical and Theoretical Studies of Water Waves

and Floating Bodies. 1993. Thesis (PhD) — University of Twente, 1993.

WANG, Q. X. Unstructured MEL modelling of nonlinear unsteady ship waves. Journal

of Computational Physics, v. 210, p. 368–385, 2005.

WEHAUSEN, J. V. The motion of floating bodies. Annual Review of Fluid Mechan-

ics, v. 3, p. 237–268, 1971.

WEHAUSEN, J. V.; LAITONE, E. V. Surface waves. Handbuch der Physics, v. 9, p.

446–815, 1960.

YAN, S.; MA, Q. W.; CHENG, X. Fully nonlinear hydrodynamic interaction between
two 3D floating structures in close proximity. In: In Proceedings of the 19th ISOPE
Conference. 2009.

YANG, C.; NOBLESSE, F.; LöHNER, R. Comparison of classical and simple free-surface
green functions. In: Proceedings of The Fourteenth International Offshore and
Polar Engineering Conference. 2004.

YANG, M.; TENG, B.; NING, D.; SHI, Z. Coupled dynamic analysis for wave interaction
with truss spar and its mooring line/riser system in time domain. Ocean Engineering,
v. 39, p. 72–87, 2012.

ZHAO, W.; YANG, J.; HU, Z.; TAO, L. Prediction of hydrodynamic performance of an
FLNG system in side-by-side offloading operation. Journal of Fluid and Structures,
v. 46, p. 89–110, 2013.

ZHAO, W. H.; YANG, J. M.; HU, Z. Q.; WEI, Y. F. Recent developments on the hydrody-
namics of floating liquid natural gas (FLNG). Ocean Engineering, v. 38, p. 1555–1567,
2011.

ZHEN, L.; BIN, T.; DE-ZHI, N.; YING, G. Wave-current interactions with three-
dimensional floating bodies. Journal of Hydrodynamics, v. 22, p. 229–240, 2010.
238
APPENDIX
 241

Appendix A

Damping Zone Analysis

This section presents an empirical study concerning the performance of the damping
zone on absorbing the waves, varying its parameters b and a. The study considers an
hemisphere body and a circular free surface discretized with 800 and 1225 quadrilateral
panels, respectively. The details of the meshes may be observed in section 6.1. The
comparison of the damping zone performance for different parameter values (b and a) is
performed in terms of the hydrodynamic vertical force F3 induced by a imposed vertical
oscillatory motion of unitary amplitude A3 = 1.0 m and an arbitrary frequency ω = 2.426
rad/s. The time-step applied on the simulations is ∆t = T /60 s, being T the oscillatory
motion period.

Figures A.1, A.2, A.3 and A.4 present the comparisons of hydrodynamic vertical forces
F3 for damping zones of lengths b = 2, b = 1.0, b = 1.5 and b = 2.0, respectively. Notice
that each figure presents five forces time series, which are associated to different a values
(a = 0.0, a = 0.5, a = 1.0, a = 1.5 and a = 2.0. Among these time series, the one for
a = 0.0 represents the case, in which no damping zone was applied in the simulations.
One may observe that the parameter b presents a larger influence on the results then
parameter a. In fact, for damping zone lengths lower than one generated wave length
(b < 1.0), significant amplitude modulations, induced by wave reflections, on the signals
are observed, irrespective of the damping zone intensity value a (Figure A.1). In addition,
the results demonstrate that as the damping zone length is increased, these modulations
tend to vanish even considering the smallest tested value of a.

Figures A.5 and A.6 present the wave field patterns resulted from a simulation disregarding
the damping zone and one considering b = 2.0λ and a = 1.0, respectively. Notice that the
case without the damping zone present a constructive and destructive pattern, resulted
from reflections on the domain boundaries, whereas the case with damping zone present
a regular progressive wave pattern field.
242

4
x 10
1 a=0 a=0.5 a=1.0 a=1.5 a=2.0

0.5
Fz (N)

−0.5

−1
0 10 20 30 40 50
t (s)

Figure A.1: Comparison of hydrodynamic vertical force F3 on a hemisphere induced by a

forced vertical oscillatory motion of amplitude A3 = 1.0 m, considering a damping zone
length of half generated wave length b = λ/2 and different values of dissipation intensity
a

4
x 10
1 a=0 a=0.5 a=1.0 a=1.5 a=2.0

0.5
Fz (N)

−0.5

−1
0 10 20 30 40 50
t (s)

Figure A.2: Comparison of hydrodynamic vertical force F3 on a hemisphere induced by

a forced vertical oscillatory motion of amplitude A3 = 1 m, considering a damping zone
length of one generated wave length b = λ and different values of dissipation intensity a
 243

4
x 10
1 a=0 a=0.5 a=1.0 a=1.5 a=2.0

Fz (N) 0.5

−0.5

−1
0 10 20 30 40 50
t (s)

Figure A.3: Comparison of hydrodynamic vertical force F3 on a hemisphere induced by

a forced vertical oscillatory motion of amplitude A3 = 1 m, considering a damping zone
length of one and a half generated wave length b = 1.5λ and different values of dissipation
intensity a

4
x 10
1 a=0 a=0.5 a=1.0 a=1.5 a=2.0

0.5
Fz (N)

−0.5

−1
0 10 20 30 40 50
t (s)

Figure A.4: Comparison of hydrodynamic vertical force F3 on a hemisphere induced by

a forced vertical oscillatory motion of amplitude A3 = 1 m, considering a damping zone
length of two generated wave lengths b = 2.0λ and different values of dissipation intensity
a
244

(a) (b)

(c) (d)

(e) (f)

(g) (h)

(i) (j)

Figure A.5: Snapshots of the simulation without damping zone at 10 different instants of
time within one period. Vertical oscillatory motion amplitude A3 = 0.5 m (a) t = 56.98
s, (b) t = 57.24 s,(c) t = 57.50 s, (d) t = 57.76 s, (e) t = 58.01 s, (f) t = 58.27 s, (g)
t = 58.53 s, (h) t = 58.79 s, (i) t = 59.05 s and (j) t = 59.31 s
 245

(a) (b)

(c) (d)

(e) (f)

(g) (h)

(i) (j)

Figure A.6: Snapshots of the simulation with damping zone parameters b = 2.0 and
a = 1.0 at 10 different instants of time within one period. Vertical oscillatory motion
amplitude A3 = 0.5 m (a) t = 56.98 s, (b) t = 57.24 s,(c) t = 57.50 s, (d) t = 57.76 s,
(e) t = 58.01 s, (f) t = 58.27 s, (g) t = 58.53 s, (h) t = 58.79 s, (i) t = 59.05 s and (j)
t = 59.31 s
246
 247

Appendix B

Multi-Body Simulations with Large

Relative Displacements

B.1 Comparison between Experiments and Compu-

tations

Experiment Present Values Wamit

5
Fx (N)

0
−5
15 20 25 30 35 40
t(s)
20
Fy (N)

0
−20
15 20 25 30 35 40
t(s)
5
Fz (N)

0
−5
15 20 25 30 35 40
t(s)
Mx (N.m)

5
0
−5
15 20 25 30 35 40
t(s)
My (N.m)

1
0
−1
15 20 25 30 35 40
t(s)

Figure B.1: Case 1 Comparison of numerical and experimental results in terms of position
of Body 2 in time and the associated time series of the hydrodynamic loads Fx , Fy , Fz ,
Mx and My
248

Experiment Present Values Wamit

0.02
ηP 1 (m)

−0.02
15 20 25 30 35 40
t(s)
0.05
ηP 2 (m)

−0.05
15 20 25 30 35 40
t(s)
0.02
ηP 3 (m)

−0.02
15 20 25 30 35 40
t(s)

Figure B.2: Case 1: Comparison between numerical and experimental results in terms of
position of Body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
35 40 45 50 55 60 65 70 75 80 85
t(s)
20
Fy (N)

0
−20
35 40 45 50 55 60 65 70 75 80 85
t(s)
5
Fz (N)

0
−5
35 40 45 50 55 60 65 70 75 80 85
t(s)
Mx (N.m)

5
0
−5
35 40 45 50 55 60 65 70 75 80 85
t(s)
My (N.m)

1
0
−1
35 40 45 50 55 60 65 70 75 80 85
t(s)

Figure B.3: Case 2 Comparison of numerical and experimental results in terms of position
of Body 2 in time and the associated time series of the hydrodynamic loads Fx , Fy , Fz ,
Mx and My
 249

Experiment Present Values Wamit

0.02

ηP 1 (m)
0

−0.02
35 40 45 50 55 60 65 70 75 80 85
t(s)
0.05
ηP 2 (m)

−0.05
35 40 45 50 55 60 65 70 75 80 85
t(s)
0.02
ηP 3 (m)

−0.02
35 40 45 50 55 60 65 70 75 80 85
t(s)

Figure B.4: Case 2: Comparison between numerical and experimental results in terms of
position of Body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
60 80 100 120 140 160
t(s)
20
Fy (N)

0
−20
60 80 100 120 140 160
t(s)
5
Fz (N)

0
−5
60 80 100 120 140 160
t(s)
Mx (N.m)

5
0
−5
60 80 100 120 140 160
t(s)
My (N.m)

1
0
−1
60 80 100 120 140 160
t(s)

Figure B.5: Case 3 Comparison of numerical and experimental results in terms of position
of Body 2 in time and the associated time series of the hydrodynamic loads Fx , Fy , Fz ,
Mx and My
250

Experiment Present Values Wamit

0.02
ηP 1 (m)

−0.02
60 80 100 120 140 160
t(s)
0.05
ηP 2 (m)

−0.05
60 80 100 120 140 160
t(s)
0.02
ηP 3 (m)

−0.02
60 80 100 120 140 160
t(s)

Figure B.6: Case 3: Comparison between numerical and experimental results in terms of
position of Body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
15 20 25 30 35 40
t(s)
20
Fy (N)

0
−20
15 20 25 30 35 40
t(s)
5
Fz (N)

0
−5
15 20 25 30 35 40
t(s)
5
Mx (N.m)

0
−5
15 20 25 t(s) 30 35 40
1
My (N.m)

0
−1
15 20 25 30 35 40
t(s)

Figure B.7: Case 4 Comparison of numerical and experimental results in terms of position
of Body 2 in time and the associated time series of the hydrodynamic loads Fx , Fy , Fz ,
Mx and My
 251

Experiment Present Values Wamit

0.02

ηP 1 (m)
0

−0.02
15 20 25 30 35 40
t(s)
0.02

0
ηP 2 (m)

−0.02

−0.04
15 20 25 30 35 40
t(s)
0.01
ηP 3 (m)

−0.01
15 20 25 30 35 40
t(s)

Figure B.8: Case 4: Comparison between numerical and experimental results in terms of
position of Body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
30 40 50 t(s) 60 70 80
20
Fy (N)

0
−20
30 40 50 t(s) 60 70 80
5
Fz (N)

0
−5
30 40 50 t(s) 60 70 80
Mx (N.m)

5
0
−5
30 40 50 t(s) 60 70 80
My (N.m)

1
0
−1
30 40 50 t(s) 60 70 80

Figure B.9: Case 5 Comparison of numerical and experimental results in terms of position
of Body 2 in time and the associated time series of the hydrodynamic loads Fx , Fy , Fz ,
Mx and My
252

Experiment Present Values Wamit

0.02
ηP 1 (m)

−0.02
30 35 40 45 50 55 60 65 70 75 80
t(s)
0.02
ηP 2 (m)

−0.02
30 35 40 45 50 55 60 65 70 75 80
t(s)
0.01
ηP 3 (m)

−0.01
30 35 40 45 50 55 60 65 70 75 80
t(s)

Figure B.10: Case 5: Comparison between numerical and experimental results in terms of
position of Body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
60 80 100 t(s) 120 140 160
20
Fy (N)

0
−20
60 80 100 t(s) 120 140 160
5
Fz (N)

0
−5
60 80 100 t(s) 120 140 160
Mx (N.m)

5
0
−5
60 80 100 t(s) 120 140 160
My (N.m)

1
0
−1
60 80 100 t(s) 120 140 160

Figure B.11: Case 6 Comparison of numerical and experimental results in terms of position
of Body 2 in time and the associated time series of the hydrodynamic loads Fx , Fy , Fz ,
Mx and My
 253

Experiment Present Values Wamit

0.02

ηP 1 (m)
0

−0.02
60 70 80 90 100 110 120 130 140 150 160
t(s)
0.02
ηP 2 (m)

−0.02
60 70 80 90 100 110 120 130 140 150 160
t(s)
0.01
ηP 3 (m)

−0.01
60 70 80 90 100 110 120 130 140 150 160
t(s)

Figure B.12: Case 6: Comparison between numerical and experimental results in terms of
position of Body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
15 20 25 30 35
t(s)
10
Fy (N)

0
−10
15 20 25 30 35
t(s)
5
Fz (N)

0
−5
15 20 25 t(s) 30 35
5
Mx (N.m)

0
−5
15 20 25 t(s) 30 35
1
My (N.m)

0
−1
15 20 25 t(s) 30 35

Figure B.13: Case 7 Comparison of numerical and experimental results in terms of position
of Body 2 in time and the associated time series of the hydrodynamic loads Fx , Fy , Fz ,
Mx and My
254

Experiment Present Values Wamit

0.02
ηP 1 (m)

−0.02
15 20 25 30 35
t(s)
0.02
ηP 2 (m)

−0.02
15 20 25 30 35
t(s)
0.01
ηP 3 (m)

−0.01
15 20 25 30 35
t(s)

Figure B.14: Case 7: Comparison between numerical and experimental results in terms of
position of Body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
30 40 50 t(s) 60 70 80
10
Fy (N)

0
−10
30 40 50 t(s) 60 70 80
5
Fz (N)

0
−5
30 40 50 t(s) 60 70 80
5
Mx (N.m)

0
−5
30 40 50 t(s) 60 70 80
1
My (N.m)

0
−1
30 40 50 t(s) 60 70 80

Figure B.15: Case 8 Comparison of numerical and experimental results in terms of position
of Body 2 in time and the associated time series of the hydrodynamic loads Fx , Fy , Fz ,
Mx and My
 255

Experiment Present Values Wamit

0.02

ηP 1 (m)
0

−0.02
30 35 40 45 50 55 60 65 70 75 80
t(s)
0.02
ηP 2 (m)

−0.02
30 35 40 45 50 55 60 65 70 75 80
t(s)
0.01
ηP 3 (m)

−0.01
30 35 40 45 50 55 60 65 70 75 80
t(s)

Figure B.16: Case 8: Comparison between numerical and experimental results in terms of
position of Body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
60 80 100 t(s) 120 140 160
10
Fy (N)

0
−10
60 80 100 t(s) 120 140 160
5
Fz (N)

0
−5
60 80 100 t(s) 120 140 160
Mx (N.m)

5
0
−5
60 80 100 t(s) 120 140 160
My (N.m)

1
0
−1
60 80 100 t(s) 120 140 160

Figure B.17: Case 9 Comparison of numerical and experimental results in terms of position
of Body 2 in time and the associated time series of the hydrodynamic loads Fx , Fy , Fz ,
Mx and My
256

Experiment Present Values Wamit

0.02
ηP 1 (m)

−0.02
60 70 80 90 100 110 120 130 140 150 160
t(s)
0.02
ηP 2 (m)

−0.02
60 70 80 90 100 110 120 130 140 150 160
t(s)
0.01
ηP 3 (m)

−0.01
60 70 80 90 100 110 120 130 140 150 160
t(s)

Figure B.18: Case 9: Comparison between numerical and experimental results in terms of
position of Body 2 in time and the associated time series of the wave elevations at WP1,
WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
15 20 25 30 35
t(s)
10
Fy (N)

0
−10
15 20 25 30 35
t(s)
5
Fz (N)

0
−5
15 20 25 30 35
t(s)
Mx (N.m)

2
0
−2
15 20 25 30 35
t(s)
My (N.m)

1
0
−1
15 20 25 30 35
t(s)

Figure B.19: Case 10 Comparison of numerical and experimental results in terms of

position of Body 2 in time and the associated time series of the hydrodynamic loads Fx ,
Fy , Fz , Mx and My
 257

Experiment Present Values Wamit

0.02

ηP 1 (m)
0

−0.02
15 20 25 30 35
t(s)
0.02
ηP 2 (m)

−0.02
15 20 25 30 35
t(s)
0.01
ηP 3 (m)

−0.01
15 20 25 30 35
t(s)

Figure B.20: Case 10: Comparison between numerical and experimental results in terms
of position of Body 2 in time and the associated time series of the wave elevations at
WP1, WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
30 35 40 45 50 55 60 65 70 75
t(s)
10
Fy (N)

0
−10
30 35 40 45 50 55 60 65 70 75
t(s)
5
Fz (N)

0
−5
30 35 40 45 50 55 60 65 70 75
t(s)
Mx (N.m)

2
0
−2
30 35 40 45 50 55 60 65 70 75
t(s)
My (N.m)

1
0
−1
30 35 40 45 50 55 60 65 70 75
t(s)

Figure B.21: Case 11 Comparison of numerical and experimental results in terms of

position of Body 2 in time and the associated time series of the hydrodynamic loads Fx ,
Fy , Fz , Mx and My
258

Experiment Present Values Wamit

0.02
ηP 1 (m)

−0.02
30 35 40 45 50 55 60 65 70 75
t(s)
0.02
ηP 2 (m)

−0.02
30 35 40 45 50 55 60 65 70 75
t(s)
0.01
ηP 3 (m)

−0.01
30 35 40 45 50 55 60 65 70 75
t(s)

Figure B.22: Case 11: Comparison between numerical and experimental results in terms
of position of Body 2 in time and the associated time series of the wave elevations at
WP1, WP2 and WP3

Experiment Present Values Wamit

5
Fx (N)

0
−5
50 60 70 80 90 100 110 120 130 140 150
t(s)
10
Fy (N)

0
−10
50 60 70 80 90 100 110 120 130 140 150
t(s)
2
Fz (N)

0
−2
50 60 70 80 90 100 110 120 130 140 150
t(s)
Mx (N.m)

2
0
−2
50 60 70 80 90 100 110 120 130 140 150
t(s)
My (N.m)

1
0
−1
50 60 70 80 90 100 110 120 130 140 150
t(s)

Figure B.23: Case 12 Comparison of numerical and experimental results in terms of

position of Body 2 in time and the associated time series of the hydrodynamic loads Fx ,
Fy , Fz , Mx and My
 259

Experiment Present Values Wamit

0.02

ηP 1 (m)
0

−0.02
50 60 70 80 90 100 110 120 130 140 150
t(s)
0.02
ηP 2 (m)

−0.02
50 60 70 80 90 100 110 120 130 140 150
t(s)
0.01
ηP 3 (m)

−0.01
50 60 70 80 90 100 110 120 130 140 150
t(s)

Figure B.24: Case 12: Comparison between numerical and experimental results in terms
of position of Body 2 in time and the associated time series of the wave elevations at
WP1, WP2 and WP3