Bulletin Technique
Couv BT 2008:Couv BT 2008 14/10/09 11:01 Page1
Introduction
Welcome to this edition of Bulletin Technique, a unique resource
of technical papers delivered by experts from Bureau Veritas
during 2008.
It is divided into five sections, grouping the papers under
Hydrodynamics and HydroStructure interaction, Structure,
Offshore Engineering & Mooring, Rules and Classification and
Risk Engineering.
Each section ranges over a wide range of topics which experts
have tackled during the year. The variety and depth of the papers
are a testimony to the wide range of expertise and resources
committed to research by Bureau Veritas. The papers give full
details of the topics and are fully referenced, and the authors
would be very happy to discuss their work further with interested
parties.
Pierre Besse
Research Director
Bureau Veritas
Bulletin Technique 2008
Editorial BT2008:Editorial BT2008 05/10/09 20:54 Page1
Hydrodynamics & Hydro Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5
Computations of lowfrequency wave loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7
X. B. Chen & F. Rezende
Simulation of secondorder roll motions of a FPSO . . . . . . . . . . . . . . . . . . . . . . . .11
F. Rezende, X. B. Chen & M. D. Ferreira
Consistent hydrostructure interface for evaluation of global structural
responses in linear seakeeping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19
S. Malenica, E. Stumpf, F. X. Sireta & X. B. Chen
Effects on sloshing pressure due to the coupling between
seakeeping and tank liquid motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
L. Diebold, E. Baudin, J. Henry & M. Zalar
Steep wave impact onto a complex 3D structure . . . . . . . . . . . . . . . . . . . . . . . . . .31
I. Ten, A. Korobkin, S. Malenica ,J. de Lauzon & Z. Mravak
Secondorder wave loads on a LNG carrier in multidirectional waves . . . . . . .35
M. Renaud, F. Rezende, O. Waals, X. B. Chen & R. van Dijk
Diffraction and radiation with the effect of bathymetric variations . . . . . . . . . . .43
G. De Hauteclocque, F Rezende & X. B. Chen
Threedimensional hydroelastic Wagner impact using variational inequalities 51
T. Gazzola & J. de Lauzon
Some aspects of 3D linear hydroelastic models of springing . . . . . . . . . . . . . . . . .59
S. Malenica, J.T. Tuitman, F. Bigot & F. X. Sireta
Advanced computations of mooring systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . .69
C. Brun, D. Coache & F. Rezende
3DFEM3DBEM model for springing and whipping analyses of ships . . . . . . . . .77
S.Malenica & J. T. Tuitman
Fluidstructure interaction modeling, relating to membrane LNG ship
cargo containment system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .87
W. S. Kim, B. J. Noh, H. Lee, Z. Mravak, J. de Lauzon, J. R. Maguire, D. Radosavljevic,
S. H. Kwon & J. Y. Chung
Hydroelastic aspects of large container ships . . . . . . . . . . . . . . . . . . . . . . . . . . . .97
I. Senjanovic, S. Malenica, S. Tomaevic & M. Tomic
Some aspects of whipping response of container ships . . . . . . . . . . . . . . . . . . .107
J. Tuitman & S. Malenica
A Comprehensive and Pratical Strength Assessment Methodology for
Container Ships Taking Into Account Non Linear and HydroElastic Loading .111
G. de Jong & M. Huther
Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .121
Simulation of the behaviour of fatigue cracks: a tool for inspection
decision making on a ships deck beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .123
V. Boutillier, M. Serror & G. Parmentier
Quelques aspects hydrodynamique/structure en conception
des porte conteneurs geants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .135
M. Huther, S. Malenica & F. Mauduit
Bulletin Technique 2008
Contents
2
). In Chen & Duan (2007), it is denoted that
=
1
2
and = (
1
+
2
)/2 then (
1
,
2
) = (+/2, /2) (2)
By assuming 1, the quadratic transfer function (QTF) is developed as an expansion :
F(
1
,
2
) = F
0
(, ) + F
1
(, ) + O[()
2
] (3)
with the zerothorder term contributed by the quadratic products of rstorder wave elds and formulated :
F
0
= F
0
q
=
2
__
H
ds
_
(
)n
n
n
2
2g
_
d (
)n (4)
as integration over the hull H and along the waterline in their mean position. In (4), stands for the
rstorder velocity potential and
n
= n the normal derivative of on H. The superscript
indicates
the complex conjugate. The expression (4) shows that F
0
is a pure real function dependent on the wave
frequency , which is nothing else than the formulation of drift loads (with a factor 2 of that usually used
due to the convention here).
The O() order term in (3) is composed of four components :
F
1
= F
1
q
+ F
1
p1
+F
1
p2
+F
1
p3
(5)
with one due to rstorder wave elds :
F
1
q
= i
_
2
__
H
ds
_
(
)n (2/)
n
+
n
2
2g
_
d (
)n
_
(6)
a pure imaginary function of . In (6) for F
1
q
, we have involved the terms (, ) which are dened as the
derivative of (, ) with respect to . The contribution of secondorder incoming waves and diraction
waves of free wave type is given analytically :
F
1
p1
= ig4CS(k/)[(1+m
xx
) cos +m
yx
sin , (1+m
yy
) sin +m
xy
cos ] (7)
in which stands for the buoyant volume, C and S dependent on wavenumber k and waterdepth h are given
in Chen (2006b), while (m
xx
, m
xy
, m
yx
, m
yy
) addedmass coecients in doublebody ow, dependent only
on the hull geometry. To note (, k, ) are wave heading, wavenumber and wave frequency, respectively.
The component F
1
p1
expressed by (7) is pure imaginary dependent on the hull geometry and wave heading.
The component F
1
p2
is associated with the secondorder correction of the boundary condition on H and
written as :
F
1
p2
= i
__
H
ds {(ix)(R
n) (x)
n}[]
0
(8)
with (x, R) as the displacement vector and rotation vector, respectively. In (8), the real functions []
0
are
dened as the radiation potentials at zero frequency (doublebody ow) associated with the components of
the normal vector on H. Again, F
1
p2
given by (8) is a pure imaginary function.
Finally, the component representing the eect of forcing pressure over the free surface F (secondorder
correction of the boundary condition on F) is expressed by :
F
1
p3
= i(/g)
__
F
ds
_
2
zz
P
+ k
2
I
_
[]
0
(9)
in which (
P
,
I
) represent the diractionradiation potential and incoming wave potential, two components
of the rstorder potential =
I
+
P
.
The secondorder lowfrequency wave loads F(
1
,
2
) dened by (3) in bichromatic waves of frequencies
(
1
,
2
) are composed of one component F
0
() depending on = (
1
+
2
)/2 and another F
1
() linearly
proportional to =
1
2
. The striking fact is that F
0
() is a pure real function while F
1
() a
pure imaginary function. The usual approximation proposed by Newman (1974) largely used in practice
is based on the use of F
0
so that it is O(1) approximation. This study conrms that not only the O(1)
approximation can underestimate largely the secondorder wave loads but also it provides wrong phase
dierences with respect to incoming waves since the complete QTF is a complex function while that by the
O(1) approximation is purely real.
2. Timeseries reconstruction of lowfrequency loading
In irregular waves represented by wave energy spectrum S
j=1
a
j
exp(i
j
t) (10)
and the complex amplitude :
a
j
= exp[ik
j
(xcos + y sin) + i
j
]
_
2S
(
j
)d
j
(11)
associated with (
j
, k
j
, ,
j
) the wave frequency, wave number, heading and random phase, and dependent
on the position (x, y) with respect to the reference point (0, 0) and the sampling space d
j
of the spectrum.
The lowfrequency wave loading in temporal domain is dened by a double summation :
F(t) =
_
_
_
N
i=1
N
j=1
F
ij
a
i
a
j
exp[i(
i
j
)t]
_
_
_
(12)
with a
j
being the complex conjugate of a
j
, and F
ij
= F(
i
,
j
) the QTF representing the component in
one degree of freedoms or a vector of wave load QTF. At each time step, the lowfrequency wave loading
is evaluated to perform the time simulation of motions. Due to the double summation, the expression (12)
is timeconsuming. One approximation largely used in practice is called Newmans approximation in Molin
(2002) which is based on :
F
ij
= sign(F)
_
F
ii
F
jj
(13)
with sign(F) the sign of wave loading which is assumed to remain the same when
j
varies. Introducing
(13) into (12), the timeseries reconstruction of lowfrequency wave loading becomes
F(t) =
j=1
_
F
jj
a
j
exp(i
j
t)
2
sign(F) (14)
the square of a single summation which is much more economic than the double summation (12).
Considering the fact that
F
0,1
(, ) =
_
F
0,1
(
i
,
i
) +F
0,1
(
j
,
j
)
/2 + O
_
(
i
j
)
2
with = (
i
+
j
)/2 (15)
the approximation (3) can be rewritten as
F
ij
= (F
0
ii
+F
0
jj
)/2 + i(
i
j
)(F
1
ii
+F
1
jj
)/2 (16)
with F
0
jj
= F
0
(
j
,
j
) and F
1
jj
= i F
1
(
j
,
j
) for j = 1, 2, , N so that both F
0
jj
and F
1
jj
are real
functions.
Introducing above expression (16) into (12), the timeseries reconstruction of lowfrequency wave loading
can be obtained by :
F(t) =
_
_
_
_
N
j=1
F
0
jj
a
j
exp(i
j
t)
_
E(t)
_
_
_
d
dt
_
_
_
_
N
j=1
F
1
jj
a
j
exp(i
j
t)
_
E(t)
_
_
_
(17)
involving only single summations. The derivative d()/dt in the compact form (17) is understood to apply
only to the factor exp(i
j
t). The rst term in (17) can be considered as the variant of O(1) approximation
of QTF, which is, unlike (14), not restricted by the assumption of a unique sign for QTF.
The new formulation (3) of QTF by the O() approximation provides a novel method to evaluate the
lowfrequency secondorder wave loads in a more accurate than O(1) approximation and more ecient way
comparing to the computation of complete QTF. All the more interesting is that the reconstruction of time
series of wave loads is shown to be single summations (17) by using (3) instead of a double summation (12), if
the complete QTF (1) is used, to account all pairs of wave interactions which is much more timeconsuming.
3. Numerical results and discussions
Among components of F
1
in (3), the component F
1
q
dened by (6) can be evaluated directly when the
terms as the derivative of rstorder potential with respect to wave frequency are obtained. An indirect
way consisting to evaluate the nite dierence after having obtained F
q
for bichromatic waves and drift
component F
0
q
by (4) can be implemented. The term F
1
p1
dened by (7) depending on the secondorder
potential of incoming waves and hull geometry can be a dominant one for a body of large volume () and
in water of small depth since C (3/4)ka
2
/(kh)
3
for kh 0 as shown in Chen (2006b). F
1
p2
given by (8)
depending on rstorder motions is simple and easy to evaluate, negligible for large wavenumber. Finally,
F
1
p3
expressed by (9) is represented by an integral over the free surface which can be performed in a limited
area around the body since the integrand decreases rapidly for the radial distance R , as conrmed by
the numerical computations here.
E
T
0
2
4
6
8
10
0 50 100 150 200 250
0
45
90
90
45
Figure 1: Mesh on LNG hull & the free surface (left) and the integrand of F
1
p3
(9) in function of radial
distances and polar angles (right)
Numerical computations are performed for a standard 138 Km
3
LNG vessel with the main dimensions
(length, width and draft) =(274m, 44.2m and 11m), respectively. The half hull is represented by a mesh
composed of 2204 planar panels and the area of free surface around the body is meshed in an elliptic form
with a ratio of the major axis (along the axis ox) over the minor axis (along the axis oy) equal to 2. About
4000 to 8000 panels are used for a major axis equal to 200m to 500m. A zoom of the hull mesh and the area
on the free surface is shown on the left part of Figure 1. Along three radial lines corresponding the polar
angles (0
, 45
and 90
Px
+
y
Py
)+(
x
Px
+
y
Py
)k
2
_
+
_
d (
Px
n
x
+
Py
n
y
)
_
(18)
by applying Stokes theorem. In (9), the line includes the waterline and the exterior border of the area F.
The component F
1
p3
is evaluated in this way and compared with other components.
80000
60000
40000
20000
0
20000
40000
60000
80000
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
2.5e+006
2e+006
1.5e+006
1e+006
500000
0
500000
1e+006
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
F
0
q
iF
1
q
iF
1
p1
{Fp1 +Fp2}
iF
1
p3
{Fq}
{Fq}
Figure 2: Dierent components of lowfrequency wave loading
The components in (4) and (5) of lowfrequency wave load (3) on the LNG vessel are evaluated for
a series of wave frequencies associated with = 0.3 rad/s and varying from 0 to 0.16 rad/s in water
of 15m in depth. They are depicted on Figure 2. The rst part dependent on the quadratic products of
rstorder wave elds is shown on the left while the second part contributed by the secondorder incoming
and diraction waves on the right. The real part of QTF includes only F
0
q
(solid line) which remains constant
for all and is a very good approximation to the exact values  the real part of F
q
represented by the
crossdashed line. The imaginary part of QTF contains 4 components. The component F
1
q
(dashed line)
is shown on the left together with the exact values  the imaginary part of F
q
(squaredashed line). The
component F
1
p1
(dashed line) are depicted on the right with the exact values F
p1
+F
p2
(crossdashed line).
This indicates that F
1
p2
is negligible. Finally, the component F
1
p3
is shown on the right by the squaredashed
line.
The quadratic contribution F
1
q
is positive while the contribution of secondorder potentials F
1
p1
is
negative. The component associated with the forcing eect on the free surface F
1
p3
is small comparing to
others. Furthermore, the component F
1
p1
given by the analytical expression (7) is largely dominant for
> 0.01 rad/s. Globally, the O() approximation of QTF is excellent up to = 0.05 rad/s and still
very good for 0.10 rad/s. The O() approximation of QTF is so believed to give much better values
of secondorder lowfrequency wave loading in an ecient way as described by (17) involving only single
summations.
References
[1] Chen X.B. & Duan W.Y. (2007) Formulation of lowfrequency QTF by O() approximation, Proc
22nd IWWWFB, Plitvice (Croatia).
[2] Chen X.B. (2006a) Middleeld formulation for the computation of wavedrift loads, J. Engineering
Math. DOI 10.1007/s106650069074x. 82, 59:6182. Published online: Sep 2006.
[3] Molin B. (1979) Secondorder diraction loads upon threedimensional bodies, Appl. Ocean Res., 1,
197202.
[4] Chen X.B. (2006b) Setdown in the secondorder Stokes waves, Proc. 6th Intl Conf. on HydroDynam
ics, Ischia (Italy), 17985.
[5] Newman J.N. (1974) Secondorder, slowlyvarying forces on vessels in irregular waves, Proc. Intl Symp.
Dyn. Marine Vehicle & Struc. in Waves, Mech. Engng. Pub., London (UK), 19397.
[6] Molin B. (2002) Hydrodynamique des structures oshore Editions Technip.
SIMULATION OF SECONDORDER ROLL MOTIONS OF A FPSO
Fl avia Rezende
Research Department
Bureau Veritas (France)
XiaoBo Chen
Research Department
Bureau Veritas (France)
Professorship, HEU (China)
Marcos D. Ferreira
CENPES
PETROBRAS
(Brazil)
ABSTRACT
The roll motions are a key parameter on the design of FPSOs
that operate in moderate and severe environmental conditions.
To reduce the magnitude of roll motions, some techniques based
on changing the vertical position of gravity center are used to
put the roll natural period outside of the frequency range of the
linear waves. However, recent model tests and also numerical
calculations have shown that the vessel may still experience large
roll motions which are considered to be induced by secondorder
wave loads.
Further to the work in Rezende et al. (2007) to compute
the roll response in frequency domain, new developments to per
form simulations in time domain are presented here. In this new
method, variations of secondorder roll moments dependent on
the roll and heave motions are taken into account consistently. It
is shown that, unlike the horizontal loads, the quadratic transfer
functions of the vertical loads depend on the instantaneous po
sition of the vessel. The variation of the roll moment with the
heave position of the vessel has been considered more important
than the variation obtained only with the inclination of the vessel.
Furthermore, numerical results of roll simulations are compared
with model tests results and presented in the paper.
NOMENCLATURE
FPSO Floating Production Storage and Ofoading Unit
QTF Quadratic Transfer Function
INTRODUCTION
It is observed that several FPSOs operating in different ar
eas around the world experience large roll motions, resulting in
delays and production down time. For the new constructions,
the optimization of the hull geometry in order to achieve good
motions characteristics has been an issue of major concern. One
technique used to limit the roll response is to design the unit such
that the roll natural period is outside of the range of linear wave
energy (roughly from3s to 20s). However, recent tests for FPSOs
with roll resonant periods larger than the maximal linear wave
period have demonstrated the presence of roll response at their
natural periods, what is attributed to nonlinear mechanisms.
We consider the theory of secondorder wave diffraction and
radiation within which the wave loads occurring at the differ
ence of wave frequencies can be evaluated. These lowfrequency
wave loads are considered to be the main source of large pe
riod roll motion. Much work has been done for the prediction of
the horizontal components of secondorder loads and two classes
of theories have been developed: fareld formulation based on
the momentum principle and the neareld based on the direct
secondorder pressure integration on the hull surface. The fruit
ful results have been obtained in the application to the design
of mooring system. However, few works as in Pinkster et Dijk
(1985) and in Chen et Molin (1989) have been pursued on the
computation of the vertical components of secondorder loads
due to two critical issues associated with their numerical evalua
tion.
One concerns the accurate prediction of secondorder wave
moments and another, the coupling of linear and secondorder
motions. Concerning the vertical components (force in the ver
tical direction and moments around the horizontal axis), the far
1 Copyright c 2008 by ASME
Proceedings of the ASME 27th International Conference on Offshore Mechanics and Arctic Engineering
OMAE2008
June 1520, 2008, Estoril, Portugal
OMAE200857405
eld formulation cannot be applied. Furthermore, since diffrac
tion and radiation velocities are singular in the zone of hull close
to sharp corners, the neareld formulation, in most cases, cannot
give convergent results. Recently, a new formulation to evaluate
the second order loads was proposed by Chen (2004), which is
obtained by applying the variants of Stokes theorem and Green
theorem to a domain limited by a control surface. As this formu
lation is written on a control surface at a certain distance of the
body it is so called middleeld formulation and applicable to the
evaluation of both horizontal and vertical low frequency loads.
Further to the work by Rezende et al. (2007) to compute
the roll response in frequency domain, new developments to per
form simulations in time domain are presented here. In this new
method, variations of secondorder roll moments dependent on
the roll motion are taken into account consistently. It is shown
that, unlike the horizontal loads, the quadratic transfer functions
of the vertical loads depend on the instantaneous position of the
vessel. The variation of the roll moment with the heave posi
tion of the vessel has been considered more important than the
variation obtained only with the inclination of the vessel.
FORMULATION OF LOWFREQUENCY LOADS
The lowfrequency quadratic transfer function (QTF) is de
ned as the secondorder wave loads occurring at the frequency
equal to the difference (
1
2
) of two wave frequencies
(
1
,
2
) of bichromatic waves. QTF is composed of two distinct
parts : one dependent only on the quadratic products of rst
order wave elds and another contributed by the secondorder
potentials of the incoming and diffracted waves. The formula
tion for the secondorder forces is presented below. Its extension
to secondorder moments is easy and omitted here for the sake of
space.
F(
1
,
2
) = F
q
(
1
,
2
) +F
p
(
1
,
2
) (1)
The rst part F
q
can be written in the way presented in [11] :
F
q
=
2
ZZ
H
ds
_
(
1
2
)n
n2
n1
2
_
2
2g
I
d(
1
2
)n/ cos +F
S
q
(2)
as integration over the hull H and along the waterline in their
mean position. In (2), stands for the rstorder velocity poten
tial and
n
= n the normal derivative of on H. The angle
is dened as that between the normal vector n = (n
x
, n
y
, n
z
) and
the horizontal plane. The subscripts (
1
,
2
) represent the quanti
ties associated with the wave frequencies (
1
,
2
), respectively,
while the superscript
indicates the complex conjugate. In (2),
the term F
S
q
is a special component for vertical forces and mo
ments around the horizontal axes which is written as :
F
S
q
=
g
2
I
d
_
(
1
1
Z
2
Z
1
)tan
[
1
(X
2
n
x
+Y
2
n
y
) +
2
(X
1
n
x
+Y
1
n
y
)]/cos
_
k
K
33
(
1
2
+
1
2
)Z
g
/2
K
34
(
1
2
+
1
)/4 +K
35
(
1
2
+
1
)/4 (3)
in which is the wave elevation and (X,Y, Z) are displacements
along the waterline. (, , ) are the rotations and (K
33
, K
34
, K
35
)
the components of hydrostatic stiffness matrix. Finally, Z
g
is the
vertical coordinate of the gravity center.
The formulation (2) derived from Eq.27 in [11] obtained by
applying the two variants of Stokes theorem given in [8] to the
classical neareld (pressureintegration) formulation as in [9],
is compact and used here. It is directly applicable to force com
ponents in horizontal directions. The extension to other compo
nents is direct and omitted here. In (2), we involve the gradient of
velocity potentials which is sensitive to the singularities present
in the velocity eld at sharp corners. In particular, the integration
of terms (
1
2
) converges slowly or in the worst cases, may
be nonconvergent.
In [11], after having the new neareld formulation by ap
plying the variants of Stokess theorem, we considered a uid
volume enclosed by the hull, a control surface at a distance from
the body and the mean free surface limited by the waterline and
the intersection of the control surface with free surface, and have
obtained the general formulation (Eq.8a & 8b in [11]) of second
order loads by using Gausss theorem. This formulation can be
simplied if we construct a control surface surrounding the hull
touching the free surface only along the waterline :
F
q
= (
1
2
)
2
ZZ
H
ds
_
n1
2
/
1
n2
1
/
2
_
2
2g
I
d(
1
2
)n
2
ZZ
C
ds
_
(
1
2
)n
n2
n1
2
_
+F
S
q
(4)
in which C stands for the control surface dened as an arbitrary
one surround the body.
The integration of terms (
1
2
) in (4) now performed
on the control surface C converges rapidly since C is at some dis
tance from the hull where the velocity eld does not present any
singularity. The integration on the hull H is of order O(
1
2
)
and as small as
n
which tends to zero for large wave frequencies.
The middleeld formulation (4) is used for the computation of
the rstpart of secondorder loads F
q
in the following.
2 Copyright c 2008 by ASME
The second part F
p
is expressed in the way [7] :
F
p
=i(
1
2
)
ZZ
H
ds
_
(2)
I
n(
(2)
In
N
H
)[]
_
+i(
1
2
)
g
ZZ
F
dsN
F
[] (5)
in which the rst term in the hull integral corresponds to the
secondorder FroudeKrylov component contributed by the in
coming wave potential
(2)
I
dened by :
(2)
I
=ia
1
a
2
Ag
2
coshk
m
(z+h)
coshk
m
h
e
ik
m
(xcos+ysin)+i(
1
2
)
gk
m
tanhk
m
h (
1
2
)
2
(6)
with k
m
= k
1
k
2
and (
1
,
2
) being the phases of rstorder in
coming waves associated with (
1
,
2
), respectively. In (6), A is
written :
A =
1
2
k
1
k
2
[1 +tanhk
1
htanhk
2
h]
+
1
2
_
k
2
1
/
1
cosh
2
k
1
h
k
2
2
/
2
cosh
2
k
2
h
_
(7)
where we have used the notations (a
1
, a
2
) and (k
1
, k
2
) stand
ing for the wave amplitudes and wavenumbers associated with
(
1
,
2
) via the dispersion equation k
1,2
tanhk
1,2
h =
2
1,2
/g with
the waterdepth h, respectively, while the wave heading with re
spect to the positive xaxis is denoted by .
The second term in the hull integral of (5) and the term de
ned by the integral over mean free surface F come from the
application of Haskind relation and represent the contribution of
the secondorder diffraction potential, as shown in [7]. The terms
(N
H
, N
F
) are the second members of the boundary conditions
satised by the secondorder diffraction potential on the hull H
and the mean free surface F, respectively. They are written as :
2N
H
=(x
1
)
2
n(x
2
)
1
n
(i
2
x
2
) (R
1
n) (i
1
x
1
+
1
) (R
2
n) (8)
and
N
F
= i(
1
2
)(
1
P2
+
P
I2
)
i
1
2g
_
1
(
2
2
z
+g
2
zz
)
P2
+gk
2
2
(1tanh
2
k
2
h)
P1
I2
)
+
i
2
2g
_
2
(
2
1
z
+g
2
zz
)
P1
+gk
2
1
(1tanh
2
k
1
h)
P2
I1
)
(9)
in which x is the displacement vector at a point on H and R the
vector of rotations. In (9),
I
represents the rstorder potential
of incoming waves while
P
=
I
stands for that of perturba
tion including the diffraction and radiation components. Finally,
[] in (5) represents a vector of rstorder radiation potentials os
cillating at the difference frequency (
1
2
). They satisfy the
homogeneous condition :
_
(
1
2
)
2
+g
z
[] = 0 (10)
on the mean free surface F and
n
[] = n (11)
on the hull H.
The full QTF (1) is composed of two parts (F
q
, F
p
) given
by the formulations (2) and (5), respectively. The formulation
(2) for F
q
derived in [11] is simpler than that in [9]. The for
mulation (5) for F
p
by [7] is often called indirect method since
it provides a way to evaluate the contribution from the second
order diffraction potential through the Haskind relation such that
the secondorder diffraction potential is not explicitly computed.
SECONDORDER ROLL MOTIONS
We are limited to study the roll motion of FPSO in beam
sea and assume that the coupling with other motions affects only
the excitation moments. The equation of roll motions under an
arbitrary excitation containing a convolution of its impulsive re
sponse for taking account of memory effect is written as :
[I +I
a
()]
+
Z
t
0
(t )
d +B
v
+K = M(t) (12)
with the memory function (t) dened by :
(t) =
2
Z
0
[I
a
() I
a
()] sin(t)d (13)
involving the added inertial coefcient I
a
() due to radiation
wave elds. The terms (I, B
v
, K) are roll inertia in air, viscous
damping and hydrostatic stiffness, respectively.
In (12), the excitation roll moment M(t) is composed
of the rstorder moment M
(1)
(t) and secondorder moment
M
(2)
(t, , Z
0
+) which is sensible to the mean position of FPSO,
i.e. the roll angle and the vertical position including the heave
motion :
M(t) = M
(1)
(t) +M
(2)
(t, , Z
0
+) (14)
3 Copyright c 2008 by ASME
with Z
0
the mean vertical position. Since the wave frequencies
are much higher than the roll resonant frequency, the roll motion
is not sensible to the rstorder excitation moment. We may take
only the secondorder moment in (14) and the motion equation
can be approximated by :
[I +I
a
()]
+ [B
r
()+B
v
]
+K = M
(2)
(t, , Z
0
+) (15)
with equal to the frequency of roll resonance. The term B
r
()
is the radiation damping in roll.
Further to the work in frequency domain by Rezende et al.
(2007), we simulate here the roll motion in time domain. In irreg
ular waves represented by wave energy spectrum S
() charac
terized by the parameters like signicant heights, peak periods
and form coefcients, the elevation of free surface is written as a
Fourier series :
(t) =
_ N
j=1
a
j
exp(i
j
t)
_
(16)
and the complex amplitude :
a
j
= exp[ik
j
(xcos+ysin) +i
j
]
_
2S
(
j
)d
j
(17)
associated with (
j
, k
j
, ,
j
) the wave frequency, wave number,
heading and random phase, and dependent on the position (x, y)
with respect to the reference point (0, 0) and the sampling space
d
j
of the spectrum.
The heave motion (t) involving in the vertical mean posi
tion can be obtained by the Fourier series :
(t) =
_ N
j=1
X
3
(
j
)a
j
exp(i
j
t)
_
(18)
with X
3
() the heave RAOs.
The lowfrequency roll moment in temporal domain is de
ned by a double summation :
M
(2)
(t, ) =
_ N
i=1
N
j=1
F(
i
,
j
, , Z
0
+)a
i
a
j
e
i(
i
j
)t
_
(19)
with a
j
being the complex conjugate of a
j
, and F(
i
,
j
, , Z
0
+
) the QTF of roll moments evaluated by the formulations (4) and
(5). In fact, the QTF is precomputed at several mean positions
(
k
, Z
m
) with k, m = 1, 2, . in a range of values of
min
k
max
and Z
min
Z
m
Z
max
. At each time step, the lowfrequency
QTF depending on (, Z
0
+) is obtained by interpolation from
the table of F(
i
,
j
,
k
, Z
m
).
NUMERICAL RESULTS AND COMPARISONS
In the following a comparison between the numerical com
putations and the model tests results is presented for one of the
concept designs of FPSOBR from Petrobras. The model tests
were performed at the University of S ao PauloBrazil.
The table 1 presents the dimensions and hydrostatic charac
teristics of the FPSO. The scale of the model was 1:90 and it was
connected to the basin by means of a horizontal mooring system,
designed not to inuence the behavior of the unit. Figure 1 below
represents the positioning of the model and of the sensors in the
basin, where A and B are wave probes located close to the wave
generator and to the model respectively; C are load cells located
at the extremity of the mooring lines and D are the measurer of
the model motions.
For the purpose of numerical solution, a frequency domain
analysis has been performed for different positions of the ves
sel, in order to obtain the quadratic transfer functions of roll mo
ment. The lowfrequency roll motions of the vessel have been
calculated in time domain by considering the interpolated QTF
at each time step depending on the instantaneous position of the
vessel. First, only the inclinations around the mean heave po
sition have been taken into account. Then, the inuence of the
heave motion on the second order roll moment has been included
by considering inclinations around different heave positions. In
a total, fteen positions have been considered (ve inclinations
around three drafts).
For the diffractionradiation calculation, the hull wetted sur
face has been meshed by quadrilateral panels. The total number
of panels used is 4028. A mesh has been generated for each po
sition. Three drafts have been used corresponding to three heave
positions (18.94m, 21.94m and 25.94m), and ve inclinations
varying from 8.0deg to 8.0deg with a step of 4deg. The g
ure 2 represents the mesh for a given position of the vessel. For
the calculation of the second order roll moment, the middle eld
formulation has been applied and a control surface has been gen
erated, which is illustrated on Figure 3.
Figures 4 and 5 present the comparison between the RAOs
of heave and roll derived from numerical calculations and model
tests.
The gures 6 and 7 present the second order roll moments,
real and imaginary parts, calculated at the upright position of the
FPSO (solid line) and at inclined positions, 4.0deg and 8.0deg
(dashed and dotdashed lines,respectively), for a draft corre
sponding to the mean position of heave (0.4m). The gures
8 and 9 present the second order roll moments calculated at
three different drafts corresponding to different heave positions
(4.0m, 0.4m and 3.0m), being 0.4m the mean heave position
of the vessel obtained from model tests.
A time domain analysis has been performed in order to cal
culate the second order roll motions of the FPSO. The time se
ries of roll motions with and without considering the effects of
the varying position of the FPSO are presented on gure 10
4 Copyright c 2008 by ASME
by dashed and solid line, respectively. The sea state used in
the calculations is represented by a JONSWAP spectrum, with
Hs=5.5m, Tp=10.76s and = 1.54.
The statistical parameters in table 2 have been derived from
the computed time series of roll motions and compared to those
obtained from the model tests.
CONCLUSIONS
In this paper the formulations used to compute the quadratic
transfer functions of second order roll moment in frequency do
main and roll motions in time domain have been presented.
The application to FPSOBR illustrates well the phe
nomenum of large roll motions ocurring in irregular waves, al
though the roll resonant period is larger than the maximum pe
riod of wave energy such that the linear excitation is negligible.
Additionally, it has been shown that, unlike the horizontal loads,
the quadratic transfer functions of the vertical loads depend on
the instantaneous position of the vessel. For the specic case of
FPSOBR, the variation of the roll moment with the heave posi
tion of the vessel has been considered more important than the
variation obtained only with the inclination of the vessel. The
agreement between the numerical results with the model tests is
generally good. Including the effects of the heave motions on the
roll motions calculations have signicantly improved the results.
REFERENCES
[1] REZENDE F., CHEN X.B. & FERREIRA M.D. Second Or
der Roll Motions for FPSOs operating in Severe Environ
mental Conditions. Proc. OTC Conf, Houston, USA, paper
No.18906.
[2] PINKSTER J.A.& VAN DIJK A.W (1985) Wave drift
forces on large semisubmersibles Joint Industry Projects
NSMB,
[3] CHEN X.B. & MOLIN B. (1989) Numerical prediction of
semisubmersible nonlinear motions Intl Symp. on Num.
Ship Hydrodynamics, Hiroshima (Japan).
[4] CHEN X.B. (2004) Hydrodynamics in Offshore and Naval
Applications  Part I. Keynote lecture at the 6th Intl Confer
ence on Hydrodynamics, Perth (Australia).
[5] MOLIN B. (1979a) Computations of wave drift forces.
Proc. OTC Conf., Houston (USA), paper No.3627.
[6] LIU Y.H. (2003) On secondorder roll motions of ships.
Proc. OMAE03, Cancun (Mexico).
[7] MOLIN B. (1979b) Secondorder drift forces upon large
bodies in regular waves. Proc. BOSS79., London (UK),
36370.
[8] DAI Y.S. (1998) Potential ow theory of ship motions in
waves in frequency and time domain (in Chinese). Press of
the National Defense Industries, Beijing (China).
[9] PINKSTER J.A. (1980) Low frequency second order wave
exciting forces on oating structures. H. Veenman En Zonen
B.V.  Wageningen (The Netherlands).
[10] FERREIRA M.D. & LEE C.H. (1994) Computation of
secondorder mean wave forces and moments in multibody
interaction. Proc. 7th Intl Conf. Behaviour Off. Structures,
BOSS94, Boston (USA) 2, 30313.
[11] CHEN X.B. (2006) Middleeld formulation for the com
putation of wavedrift loads Journal of Engineering Math
ematics.
5 Copyright c 2008 by ASME
Table 1. Main Characteristis of FPSOBR
Description Symbol Unit Magnitude
Length L m 320.00
Breadth B m 51.00
Draft mean T m 21.50
Displacement tf 308000
Metacentric height
including correction FS
GM m 4.50
Radius of Gyration
transverse Rxx m 19.66
longitudinal Ryy m 80.38
vertical Rzz m 82.57
Figure 1. Sketch of the basin
Figure 2. Mesh: Draft 21.94m  Upright
Figure 3. Mesh: Control Surface Mesh for Middle Field Computation
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
8 10 12 14 16 18 20 22 24 26
R
A
O
o
f
h
e
a
v
e
(
m
/
m
)
Wave Period (s)
Numerical
Model Tests
Figure 4. RAO of heave
0
1
2
3
4
5
6
7
8
8 10 12 14 16 18 20 22 24 26
R
A
O
o
f
r
o
l
l
(
d
e
g
/
m
)
Wave Period (s)
Numerical
Model Tests
Figure 5. RAO of roll
6 Copyright c 2008 by ASME
30000
20000
10000
0
10000
20000
30000
0 0.2 0.4 0.6 0.8 1
R
o
l
l
M
o
m
e
n
t

r
e
a
l
p
a
r
t
(
k
N
.
m
/
m
2
)
Wave Frequency (rad/s)
Upright Condition
Inclination=4.0deg
Inclination=8.0deg
Figure 6. Second order roll moment  real part: effect of inclination
40000
30000
20000
10000
0
10000
0 0.2 0.4 0.6 0.8 1
R
o
l
l
M
o
m
e
n
t

i
m
a
g
i
n
a
r
y
p
a
r
t
(
k
N
.
m
/
m
2
)
Wave Frequency (rad/s)
Upright Condition
Inclination=4.0deg
Inclination=8.0deg
Figure 7. Second order roll moment  imaginary part: effect of inclination
40000
30000
20000
10000
0
10000
20000
30000
0 0.2 0.4 0.6 0.8 1
R
o
l
l
M
o
m
e
n
t

r
e
a
l
p
a
r
t
(
k
N
.
m
/
m
2
)
Wave Frequency (rad/s)
Upright Condition;Heave=0.4m
Upright Condition;Heave=4.0m
Upright Condition;Heave=3.0m
Figure 8. Second order roll moment  real part: effect of heave
40000
30000
20000
10000
0
10000
0 0.2 0.4 0.6 0.8 1
R
o
l
l
M
o
m
e
n
t

i
m
a
g
i
n
a
r
y
p
a
r
t
(
k
N
.
m
/
m
2
)
Wave Frequency (rad/s)
Upright Condition;Heave=0.4m
Upright Condition;Heave=4.0m
Upright Condition;Heave=3.0m
Figure 9. Second order roll moment  imaginary part: effect of heave
5
4
3
2
1
0
1
2
3
4
5
1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
R
o
l
l
(
d
e
g
)
Time (s)
Without effect of instantaneous position
With effect of instantaneous position (roll and heave)
Figure 10. Time series of roll motions
Table 2. Statistical values of roll motions
Parameter Without
position
effects
With
Effects
of roll
With
Effects
of roll +
Heave
Model
Tests
Mean
(deg)
0.002 0.002 0.116 0.39
Stand.
Dev.
(deg)
0.990 0.985 1.146 1.280
Signif.
Value
(deg)
3.944 3.922 4.540 4.690
Tz (s) 21.20 21.52 21.18 21.81
7 Copyright c 2008 by ASME
CONSISTENT HYDROSTRUCTURE INTERFACE FOR EVALUATION OF GLOBAL
STRUCTURAL RESPONSES IN LINEAR SEAKEEPING
Sime Malenica
Research Department
Bureau Veritas (France)
Estelle Stumpf
Research Department
Bureau Veritas (France)
Franc oisXavier Sireta
Research Department
Bureau Veritas (France)
XiaoBo Chen
Research Department
Bureau Veritas (France)
Professorship, HEU (China)
INTRODUCTION
The difculties related to the equilibration of the 3D FEM
structural model, in the context of hydrostructure interactions in
linear seakeeping are discussed. Different philosophies in mod
eling the structural and hydrodynamic parts of the problem, usu
ally lead to very different meshes (hydro and structure) which re
sults in unbalanced structural model and consequently in doubt
ful results for structural responses. The procedure usually em
ployed consists in using different kinds of interpolation schemes
to transfer the total hydrodynamic pressure from hydrodynamic
panels to the centroids of the structural nite elements. This ap
proach is both, very complex for complicated geometries, but
also rather inaccurate. The method that we propose here is based
on two main ideas:
1. Pressure recalculation instead of interpolation
2. Separate transfer of different pressure components (incident,
diffraction, radiation & hydrostatic variation)
The rst point removes the difculties related to the interpola
tion techniques, and allows for a very robust method of pressure
transfer. The second point ensures the perfect equilibrium be
cause the body motions are calculated after integration over the
structural mesh, which means that the equilibrium is implicitly
imposed. It should be noted that this procedure is not completely
straightforward and several numerical tricks need to be intro
duced. However, once these difculties are solved, the nal nu
merical code is extremely robust and can be easily coupled with
any of the general 3D FEM packages.
LINEAR SEAKEEPING ANALYSIS
Before considering the hydrostructure interaction problem
in more details, and for the sake of clarity, rst we briey de
scribe the basics of the seakeeping model which is used in most
of the seakeeping tools based on Boundary Integral Equation
techniques.
The problem is formulated in frequency domain under the poten
tial ow assumptions. The total velocity potential is decomposed
into the incident, diffracted and 6 radiated components:
=
I
+
D
i
6
j=1
Rj
(1)
where :
1 Copyright c 2008 by ASME
Proceedings of the ASME 27th International Conference on Offshore Mechanics and Arctic Engineering
OMAE2008
June 1520, 2008, Estoril, Portugal
OMAE200857077
I
 incident potential
D
 diffraction potential
Rj
 radiation potential
j
 rigid body motions
At the same time, the corresponding dynamic pressure is found
from the linear Bernoulli equation, and the similar decomposi
tion is adopted:
p = i = p
I
+ p
D
+
6
j=1
j
p
Rj
(2)
In order to obtain the the total hydrodynamic pressure, the dy
namic variation of the hydrostatic pressure should also be added
to the above expression:
p
hs
=g[
3
+
4
(Y Y
G
)
5
(X X
G
)] (3)
where the subscript
G
denotes the position of the center of
gravity, with respect to which the motion equation is written.
It is important to note that the motion equation is written
in the so called earth xed reference system, or in the system
parallel to it, if the body is animated with forward speed. For that
reason the restoring matrix is not obtained directly by integration
of the hydrostatic pressure (3), but also the change of the normal
vector should be taken into account.
F
hs
= [ C ]{} =
S
H
B
[p
hs
ngZndS] (4)
where denotes the rotational component of the motion vec
tor = (
4
,
5
,
6
), and S
H
B
denotes the hydrodynamic mesh of
the wetted body surface. Note that the compact notation is used
throughout whole the paper, so that the normal vector n denotes
(n
x
, n
y
, n
z
) for i = 1, 3 , and (RR
G
) n for i = 4, 6.
After integrating the pressure over the wetted body surface, the
corresponding forces are obtained and the rigid body motion
equation, in frequency domain, is usually written in the following
form:
2
([ M] +[ A ]) i[ B ] +[ C ]
{} ={F
DI
} (5)
where:
[ M]  genuine mass matrix
[ A ]  added mass matrix
[ B ]  damping matrix
[ C ]  hydrostatic restoring matrix
{F
DI
}  excitation force vector
The nal expressions for the excitation, added mass and damping
coefcients are:
F
DI
i
= i
S
H
B
(
I
+
D
)n
i
dS (6)
2
A
i j
+iB
i j
=
2
S
H
B
Rj
n
i
dS (7)
Note also that, in the general case, the total restoring matrix is
a sum of the pressure part (4) and the gravity part which is zero
in the present case because the motion equation is written with
respect to the center of gravity.
Solution of the boundary value problems
Within the Bureau Veritass numerical code HYDROSTAR,
the Boundary Integral Equation (BIE) method based on source
formulation, is used to solve the Boundary Value Problems
(BVP) for different potentials (see Chen(2004)).
In the case of zero forward speed, the general form of the BVP
is:
= 0 in the uid
+
z
= 0 z = 0
n
=V
n
on S
b
lim
R(
R
i)
= 0 R
(8)
where V
n
denotes the body boundary condition which depends
on the considered potential:
D
n
=
I
n
,
Rj
n
= n
j
(9)
Within the source formulation, the potential at any point in
the uid is expressed in the following form:
=
S
H
B
GdS (10)
where G stands for the Green function, and is the unknown
source strength which is found after solving the following inte
gral equation:
1
2
+
S
H
B
G
n
dS =V
n
, on S
H
B
(11)
2 Copyright c 2008 by ASME
This equation is solved numerically, after discretizing the wet
ted part of the body into a number of at panels over which the
constant source distribution is assumed.
LOADING OF THE STRUCTURAL MODEL
The loading of the structural model is composed of two
parts:
1. Inertia loads
2. External pressure loads
Inertia loads can be included straightforwardly by associating the
acceleration vector to each nite element. Concerning the pres
sure loading, most of the methods nowadays use the different in
terpolation schemes in order to transfer the total hydrodynamic
pressure (2,3) from hydro model (centroids of the hydro panels)
to the structural model (centroids or nodes of nite elements).
Besides the problems of interpolation, it is important to note that
the motion amplitudes, which are present in the denition of the
total pressure, were calculated after integration over the hydrody
namic mesh. For that reason it is impossible to obtain the com
pletely equilibrated structural model. Indeed, the FEM model
has its own integration procedure which is usually different.
As briey stated in the introduction, in order to obtain the perfect
equilibrium of the structural model we introduce two main ideas:
1. Recalculation of pressure in structural points, instead of in
terpolation
2. Separate transfer of pressure components, and calculation
of hydrodynamic coefcients (added mass, damping, hydro
statics & excitation) by integration over the structural mesh
Here below we discuss these two points in more details.
Hydrodynamic pressure
What we propose here for transfer of the pressure from hy
drodynamic model to the structural model, is to recalculate the
pressure at the required locations instead of interpolating it from
hydrodynamic model (see Blandeau et al.(1999)). This becomes
possible thanks to the particularities of the BIE method which
gives the continuous representation of the potential through the
whole uid domain Z < 0. In this way the communication be
tween the hydrodynamic and structural codes is extremely sim
plied. Indeed, it is enough for the structural code to give the co
ordinates of the points where the potential is required and the hy
drodynamic code just evaluates the corresponding potential by:
(x
s
) =
S
H
B
(x
h
)G(x
h
; x
s
)dS (12)
where x
s
= (x
s
, y
s
, z
s
) denotes the structural point and x
h
=
(x
h
, y
h
, z
h
) the hydrodynamic point.
In the case of linear seakeeping without forward speed, this
operation is sufcient because the pressure is directly propor
tional to the velocity potential and, within the source formula
tion, the potential is continuous across the body wetted surface.
This is very important point because, due to the differences in the
hydrodynamic and structural mesh, the structural points might
fall inside the hydrodynamic meshes.
Once each pressure component has been transfered onto the
structural mesh, the new hydrodynamic coefcients are cal
culated by integration over the structural mesh:
F
DI
S
i
= i
S
S
B
(
S
I
+
S
D
)n
i
dS (13)
2
A
S
i j
+iB
S
i j
=
2
S
S
B
S
Rj
n
i
dS (14)
where the superscript
S
indicates that the quantities are taken
on the structural mesh.
Hydrostatic pressure variations
Here we concentrate on the calculation of the hydrostatic
restoring matrix which is obtained after the integration of the
hydrostatic pressure variations due to the body motions (3). The
procedure is rather similar to the hydrodynamic pressure, and we
just need to integrate the expression (3) over the structural mesh.
For the sake of clarity, let us rst rewrite the hydrostatic pressure
variations in the following compact form:
p
hs
=
6
j=1
j
p
hs
j
(15)
where:
p
hs
1
= 0 , p
hs
2
= 0 , p
hs
3
=g (16)
p
hs
4
=g(Y Y
G
) , p
hs
5
=g(X X
G
) , p
hs
6
= 0 (17)
With these notations, the rst part of the hydrostatic restor
ing matrix becomes:
C
p
i j
=
S
S
B
p
hs
j
n
i
dS (18)
3 Copyright c 2008 by ASME
In order to obtain the complete hydrostatic restoring matrix, one
additional term accounting for the change of coordinate system
should be added to the above expression as shown in equation
(4). In the earth xed coordinate system, this additional term
is accounted for by the change of the normal vector (4). How
ever, the structural response is calculated in the body xed coor
dinate system in which the normal vector do not change. It can
be shown (e.g. see Malenica(2003)) that, in the body xed co
ordinate system, the change of normal vector is equivalent to the
change of the gravity action, so that we can write:
F
g
=mgk = [ C ]
g
{} (19)
where the only non zero elements of the matrix [ C ]
g
are C
g
24
and C
g
15
which will be canceled by the contributions implicitly
present in [ C ]
p
.
The total restoring matrix becomes:
[ C ]
S
= [ C ]
p
+[ C ]
g
(20)
where the superscript
S
indicates that the pressure related part
was calculated by integration over the structural mesh.
Motion equation and nal loading of the structural
model
The nal motion equation can now be rewritten in the form:
2
([ M] +[ A ]
S
) i[ B ]
S
+[ C ]
S
{}
S
={F
DI
}
S
(21)
Solution of this equation gives the body motions {}
S
so that the
total linear pressure can be written in the form
p
S
= p
S
I
+ p
S
D
+
6
j=1
S
j
(p
S
Rj
+ p
hs
j
) (22)
In summary the nal loading of the structural model will be com
posed of the following 3 parts:
2
m
i
S
i
 Inertial loading
p
S
i
 Pressure loading
m
i
g
S
k  Gravity term
The inertial loading and the gravity term are to be applied on
each nite element. The pressure loading is to be applied only
on wetted nite elements. It is clear that the above structural
loading will be in perfect equilibrium because this equilibrium is
implicitly imposed by the solution of the motion equation (21) in
which all different coefcients were calculated by using directly
the information from the structural FEM model.
Seakeeping problem with forward speed
In the case of the body advancing with forward speed in
waves, the situation is slightly more complicated because the hy
drodynamic pressure is no more directly proportional to the po
tential but also to its gradient. There are several methods in use
to solve the linear seakeeping problem with forward speed, but
it seems that, up to now, none of them is fully consistent. Any
way, regardless of the approximations and the method which is
used to solve the BVP with forward speed, the main difculties
related to pressure transfer are the same. In order to illustrate
them, we choose the simplest case of the so called encounter fre
quency approximation. According to this method the effects of
forward speed are mainly taken into account through the change
of excitation frequency. The body moving with constant speed U
in waves with frequency coming with incidence , will expe
rience the excitation at the encounter frequency
e
:
e
= U cos (23)
where represents the wave number =
2
/g.
At the same time, since the problem is described in the coor
dinate system steadily moving with the body speed U, the time
derivative in the Bernoulli equation, changes to:
t
D
Dt
=
t
U
x
(24)
so that the expression for linear hydrodynamic pressure be
comes:
p = i
e
+U
x
(25)
The problem with the gradient of the potential, in the context
of source formulation of BIE, is that the gradient is discontinu
ous across the hydrodynamic mesh, so that the recalculation of
the pressure on structural points will likely lead to discontinu
ities in the pressure distribution. This means that a preprocessing
of the structural points is necessary i.e. these points needs to
be articially moved to the exterior of the hydrodynamic model.
Strictly speaking, structural points do not change their position
4 Copyright c 2008 by ASME
but the pressure associated to them is calculated in the points
which are slightly moved off their original position. The error
introduced by these manipulations remains negligible within the
limits of the linear theory considered here.
NUMERICAL IMPLEMENTATION
The calculation of the hydrodynamic coefcients (added
mass, damping, excitation & restoring) by integration over the
structural mesh (13,14,18) can be done in different ways and
usually depends on the type of nite elements that are used by
FEM solver. Here below we propose two methods of integra
tion, the rst one which is completely independent of the type of
the FEM package and the second one which apply for the most
typical shell elements.
Fully independent numerical integration over struc
tural mesh
In this method, the hydrodynamic coefcients are obtained
after solving the additional structural problems for different
pressure loadings. The general scheme is shown in Figure 1.
Different modules and associated interface les are enumerated
below:
AMG
Hydrodynamic mesh generator.
HYDROSTAR
Hydrodynamic seakeeping code.
NASTRAN
Structural 3D FEM code.
LISA
Code for preprocessing of the structural points.
MARGE
Code for solving the motion equation and for denition of
the nal loading.
<>H.MSH
File containing the hydrodynamic mesh data.
<>S.MSH
File containing the structural mesh data.
<>.PPI
File containing the preprocessed structural points.
<>.PRS
File containing the different pressure components for each
structural point.
<>.LOA
File containing the nal structural loading.
<>.HSI
General input le for HYDROSTAR (frequencies, inci
dences, speed ...).
<>.RFO
File containing the reaction forces.
<>.OUT
NASTRAN output les.
Figure 1. General coupling scheme for fully independent solution.
Within this method, the structural model is put onto articial sup
ports and loaded by the external loads corresponding to the dif
ferent pressure parts (diffraction, radiation, hydrostatics). The
hydrodynamic coefcients are deduced from the reaction forces
after solving corresponding structural response. In total 17 load
ing cases need to be considered: 2 for diffraction and incident
pressure, 12 for radiation pressures and 3 for hydrostatic pres
sure variations. The boundary conditions at the supports can be
chosen rather arbitrarily, as far as they dene an isostatic system
i.e. only 6 additional unknowns are introduced. For example,
referring to the Figure 2 and assuming that the structural model
was loaded by the real part of the pitch pressure {p
R5
}, the
added mass coefcient A
35
will be A
35
= R1Z +R2Z +R3Z.
5 Copyright c 2008 by ASME
Z
Y
X
R1Z
R3Y
R2X
R3Z
R1Y
R2Z
Figure 2. Possible supports denition.
The obvious disadvantage of this approach is the necessity to
solve for additional 17 loading cases by FEM solver. The main
advantage is the perfect equilibrium of the model regardless of
the FEM package which is used.
Typical numerical integration over the structural mesh.
Within this method, the pressure integration is still per
formed over the structural mesh, but only the most typical 
nite elements are considered. The advantage of the method is
that there is no necessity to solve for additional structural load
ing cases and hydrodynamic coefcients are calculated outside
the FEM package (MARGE module in this case). The disadvan
tage is that the small disequilibrium might persist for the FEM
codes which use a different integration scheme. However, this
disequilibrium is likely to be very minor so that probably this
method will remain as the most efcient one. The general cou
pling scheme is shown in Figure 3. As we can see the procedure
is signicantly simplied and can easily be adapted to any par
ticular FEM package.
NUMERICAL RESULTS AND DISCUSSION
Here below, we present few numerical results showing the
efciency of two proposed approaches. The example which was
chosen is the 7800 TEU Container ship. The corresponding
structural and hydrodynamic meshes are shown in Figure 4. In
Figures 5 and 6, the added mass coefcients, damping coef
cients, excitations and motion RAOs are presented. FEMM1
denotes the results obtained after integration over the structural
Figure 3. General coupling scheme for typical FEM package.
mesh using the fully independent method, and FEMM2 those
obtained using the typical integration method, as explained in
the previous sections.
Figure 4. Hydrodynamic and structural mesh of container ship.
6 Copyright c 2008 by ASME
0
1e+008
2e+008
3e+008
4e+008
5e+008
0 0.2 0.4 0.6 0.8 1 1.2 1.4
A
_
3
3
[
k
g
]
omega [rad/s]
Hydrostar
FEM_M1
FEM_M2
0
1e+009
2e+009
3e+009
4e+009
5e+009
6e+009
7e+009
8e+009
9e+009
1e+010
0 0.2 0.4 0.6 0.8 1 1.2 1.4
A
_
4
4
[
k
g
m
^
2
]
omega [rad/s]
Hydrostar
FEM_M1
FEM_M2
0
1e+007
2e+007
3e+007
4e+007
5e+007
6e+007
7e+007
8e+007
9e+007
0 0.2 0.4 0.6 0.8 1 1.2 1.4
B
_
3
3
[
k
g
/
s
]
omega [rad/s]
Hydrostar
FEM_M1
FEM_M2
0
5e+008
1e+009
1.5e+009
2e+009
2.5e+009
0 0.2 0.4 0.6 0.8 1 1.2 1.4
B
_
4
4
[
k
g
m
^
2
/
s
]
omega [rad/s]
Hydrostar
FEM_M1
FEM_M2
Figure 5. Added mass and damping coefcients.
0
2e+007
4e+007
6e+007
8e+007
1e+008
1.2e+008
1.4e+008
0 0.2 0.4 0.6 0.8 1 1.2 1.4
F
_
3
[
N
]
omega [rad/s]
Hydrostar
FEM_M1
FEM_M2
0
5e+007
1e+008
1.5e+008
2e+008
2.5e+008
3e+008
0 0.2 0.4 0.6 0.8 1 1.2 1.4
F
_
4
[
N
m
]
omega [rad/s]
Hydrostar
FEM_M1
FEM_M2
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2 1.4
x
i_
3
[
m
/
m
]
omega [rad/s]
Hydrostar
FEM_M1
FEM_M2
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0 0.2 0.4 0.6 0.8 1 1.2 1.4
x
i_
4
[
r
a
d
/
m
]
omega [rad/s]
Hydrostar
FEM_M1
FEM_M2
Figure 6. Excitation forces and RAOs for = 45
o
.
7 Copyright c 2008 by ASME
As we can see the agreement between two method of pres
sure transfer is excellent and the data are on top of each other.
This is due to the fact that, in the case of NASTRAN FEM solver,
the integration over the structural mesh is performed exactly in
the same way by MARGE and by NASTRAN itself. The dif
ferences which exist between FEM and Hydrostar results are ex
pected and are due to the differences in two meshes (hydro and
structure). These differences are exactly the ones which ensure
the equilibrium of the structural model! The fact that these dif
ferences are not very important shows that the overall coupling
procedure is very efcient.
CONCLUSIONS
As far as the linear seakeeping analysis is concerned, the
fully consistent transfer of hydrodynamic loading from hydrody
namic model to the 3D FEM structural model is never perfect,
even if the two meshes coincide. However, depending on the
method of pressure transfer, the coupling can be more or less ef
cient/consistent. The main difculty lies in the fact that two
meshes (hydro and structural) are usually very different, which
ends up with the unbalanced structural model if the coupling pro
cedure is not correctly performed. In this paper we presented
two methods which both lead to completely equilibrated struc
tural model. They can easily be adapted to any of the structural
FEM packages.
ACKNOWLEDGEMENTS
Important part of this work was done under the
MARSTRUCT European project.
REFERENCES
[1] BLANDEAU F., FRANCOIS M., MALENICA
S. & CHEN
X.B. 1999. : Linear and nonlinear wave loads on FPSO
s., In Proc. of ISOPE99, Brest, France.
[2] CHEN X.B. 2004. : Hydrodynamics in Offshore and
Naval applications, Proc. Int. Conf. on Hydrodynamics,
Perth, Australia.
[3] MALENICA
S. 2003. : Some aspects of hydrostatic cal
culations in linear seakeeping, In Proc. of NAV Conf.,
Palermo, Italy.
[4] MALENICA
S., STUMPF E., DELAFOSSE V., CHEN X.B.
& SENJANOVI C I. 2006. : Some aspects of hydro
structure interfacing in seakeeping., In Proc. of ISOPE06,
San Francisco, US.
8 Copyright c 2008 by ASME
23
rd
IWWWFB, Jeju, Korea 2008
EFFECTS ON SLOSHING PRESSURE
DUE TO THE COUPLING BETWEEN
SEAKEEPING AND TANK LIQUID MOTION
Louis Diebold, Eric Baudin, Jacqueline Henry, Mirela Zalar
Bureau Veritas  Marine Division, France
INTRODUCTION
The influence of dynamic coupling due to the interactions
between ship motions and tank liquid motion on the pressure
levels on tank boundaries is here investigated by experimental
means, using 6 d.o.f. test rig and 1/70 scaled tank model of
standard LNG Carrier.
Vessel response, given as traditional noncoupled and
realistic coupled motions are obtained by numerical
calculations performed with Bureau Veritas hydrodynamic
software HydroStar
(2)
Coupling
We can now write the motion equation of the coupled system:
[ ] [ ] [ ] [ ] ( ) ( [ ]
[ ] [ ] [ ] ( ) ){ } { }
DI
Q Q TQ T Q
Q TQ T Q Q
i
F C C C
B A A A M
= + +
+ + + +
2
(3)
Numerical Results and Calibration of
The calibration of the above described parameter is
performed through comparisons with experimental results [4].
The numerical calculations are performed with HydroStar. In
the selected configuration, two separated prismatic LNG cargo
tanks were modeled, located at the fore and aft parts of the
vessel. Among the different configurations tested, the
following one is of particular interest for our case.
The filling ratio is 30% of the height for the both tanks. The
mesh used for our hydrodynamic computations is shown
below:
Fig. 1: Hydrodynamic mesh with two tanks filled at 30%H
RAOs in roll (=90) is presented for two values of the
parameter (=0.02 and =0.1).
Fig. 2: Roll RAO for =90
For roll motion, we can observe the two characteristic peaks of
a coupled system (ship + tanks). The first peak is associated
with the motions of the ship and the second one with the liquid
motions in the tanks. The value =0.02 gives the best results
and will be considered hereafter.
APPLICATION TO A REALISTIC CASE
Environmental Conditions
Environmental conditions applied in this study corresponds to
realistic site specific alldirections wave scatter diagram, with
5 m of maximum recorded significant wave height. Wave
energy spectrum is generated according to JONSWAP model
(derived for seas with limited fetch), with spectral peak
parameter assumed 3.0.
Hydrodynamic Model
A LNG carrier with four tanks is here considered. The filling
ratios are 90% of the height for the tanks (1) and (3) and 30%
of the height for the tanks (2) and (4). The mesh used for the
hydrodynamic computations is presented below:
Fig. 3: Hydrodynamic mesh of the LNG Carrier filled at 90%H in
tanks (1) & (3) and 30%H in tanks (2) & (4).
Non Coupled Coupled Transfer Function
In this section, results of hydrodynamic computation are
displayed in form of Response Amplitude Operators (RAOs).
Moreover, we present the comparison of sway and roll (the
most affected degree of freedom due to coupling) RAOs
between non coupled and coupled motions. The motions
affected by the coupling are the surge, sway, roll and yaw
motions. Particularly, the roll motion is strongly affected with
the presence of the two peaks described above instead of one
peak in the case of the non coupled motion.
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
S
W
A
Y
(
m
/
m
)
Wave frequency(rad/s)
NONCOUPLED  SWAY RAO at COG
180.0
195.0
210.0
225.0
240.0
255.0
270.0
270.0
285.0
300.0
315.0
330.0
345.0
360.0
0.00
0.20
0.40
0.60
0.80
1.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
S
W
A
Y
(
m
/
m
)
Wave frequency(rad/s)
COUPLED  SWAY RAO at COG
180
195
210
225
240
255
270
285
300
315
330
345
360
Fig. 4: RAO in Sway for Non Coupled / Coupled Motion
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
R
O
L
L
(
/
m
)
Wave frequency(rad/s)
NONCOUPLED  ROLL RAO at COG
180.0
195.0
210.0
225.0
240.0
255.0
270.0
270.0
285.0
300.0
315.0
330.0
345.0
360.0
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
R
O
L
L
(
/
m
)
Wave frequency(rad/s)
COUPLED  ROLL RAO at COG
180
195
210
225
240
255
270
285
300
315
330
345
360
Fig. 5: RAO in Roll for Non Coupled / Coupled Motion
Then, spectral analysis has been performed for each
combination of associated conditions (Hs, Tp, and heading)
using JONSWAP spectrum for sitespecific environmental
conditions. The amplitudes of 1/10
th
significant level and
response zerocrossing periods for non coupled and coupled
motions are detailed. These figures highlight for sway and roll
the operational case (Tp, heading) corresponding to the worst
motions.
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
4 5 6 7 8 9 10 11 12 13
180
195
210
225
240
255
270
285
300
315
330
345
360
1.60
1.44
1.28
1.12
0.96
0.80
0.64
0.48
0.32
0.16
0.00
NON COUPLED  SWAY
AMPLITUDE
(m)
5.8
5
.8
5
. 8
5.8
6.4
6
. 4
6.4
6.4
7.1
7
.1
7
.1
7.1
7.5
7
.5
7
.5
7.5
7.8
7
.8
7
.8
7.8
8
. 3
8
.3
8.3
8.3
8
.3
8.3
1
0
.8
1
0
.8
1
4
. 1
1
4
. 1
9 14.1
8 10.8
7 8.3
6 8.3
5 7.8
4 7.5
3 7.1
2 6.4
1 5.8
TRANSVERSE
RESONANCE
FILLING TR (s)
10%H
20%H
30%H
40%H
50%H
60%H
70%H
80%H
90%H
9.1 30%H
9
. 1
9
.1
9.1
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
4 5 6 7 8 9 10 11 12 13
180
195
210
225
240
255
270
285
300
315
330
345
360
1.60
1.44
1.28
1.12
0.96
0.80
0.64
0.48
0.32
0.16
0.00
COUPLED  SWAY
AMPLITUDE
(m)
5.8
5
.8
5.8
6.4
6
.4
6.4
7.1
7
.1
7
.1
7.1
7.5
7
.5
7
.5
7.5
7.8
7
.8
7
.8
7.8
8.3
8
.3
8.3
8.3
8
.3
8
.3
8.3
10.8
1
0
.8
1
4
.1
9 14.1
8 10.8
7 8.3
6 8.3
5 7.8
4 7.5
3 7.1
2 6.4
1 5.8
TRANSVERSE
RESONANCE
FILLING TR (s)
10%H
20%H
30%H
40%H
50%H
60%H
70%H
80%H
90%H
9.1 30%H
9.1
9
.1
9
.1
Fig. 6: Sway A1/10 & RTZ for Non Coupled / Coupled motion
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
4 5 6 7 8 9 10 11 12 13
180
195
210
225
240
255
270
285
300
315
330
345
360
3.00
2.70
2.40
2.10
1.80
1.51
1.21
0.91
0.61
0.31
0.01
NON COUPLED  ROLL
AMPLITUDE
(dg)
5.8
5.8
5
.8
5.8
6.4
6
.4
6
.4
6.4
7.1
7
.1
7
.1
7.1
7.5
7
.5
7
.5
7.5
7.8
7
.8
7
.8
7.8
8.3
8
.3
8.3
8.3
8
.3
8
. 3
8.3
1
0
.8
1
0
.8
10.8
1
4
.1
1
4
.1
9 14.1
8 10.8
7 8.3
6 8.3
5 7.8
4 7.5
3 7.1
2 6.4
1 5.8
TRANSVERSE
RESONANCE
FILLING TR (s)
10%H
20%H
30%H
40%H
50%H
60%H
70%H
80%H
90%H
9.1 30%H
9.1
9
.1
9
.1
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
4 5 6 7 8 9 10 11 12 13
180
195
210
225
240
255
270
285
300
315
330
345
360
3.00
2.70
2.40
2.10
1.80
1.50
1.20
0.90
0.60
0.30
0.00
COUPLED  ROLL
AMPLITUDE
(dg)
5.8
5
.8
5
. 8
5.8
6.4
6
.4
6
. 4
6.4
7.1
7
.1
7
.1
7.1
7
.5
7
.5
7.5
7
.8
7.8
7.8
8
.3
8.3
8.3
8.3
8
.3
8
.3
8.3
1
0
.8
1
0
.8
10.8
9 14.1
8 10.8
7 8.3
6 8.3
5 7.8
4 7.5
3 7.1
2 6.4
1 5.8
TRANSVERSE
RESONANCE
FILLING TR (s)
10%H
20%H
30%H
40%H
50%H
60%H
70%H
80%H
90%H
9.1 30%H
9
.1
9
.1
9.1
Fig. 7: Roll A1/10 & RTZ for Non Coupled / Coupled motion
MODEL TEST DESCRIPTION
Sloshing model test practice is based on the measurement of
fluid impact pressure on the tank walls.
The model corresponds to the tank N2 of BV reference vessel
with standard cargo capacity of 138 000 m
3
, scaled to 1/70 and
made of a 20 mm thick Plexiglas
.
The impact pressures are measured by dynamic ICP
pressure
sensors which natural frequency is above 100 kHz. Static
pressure is not taken into account. A total of 54 points around
the model could be used to locate the pressure sensors. A total
of 16 pressure sensors is used for the tests. The sampling rate
used is 20 KHz on each channel. The acquisition control
program has been developed by Bureau Veritas. Several
VBA
), aimed to
complete the assessment procedure from the impact pressure
point of view.
EXPERIMENTAL RESULTS
Presentation of Results
In this section are detailed our experimental results concerning
the pressure results for the tank No. (2) described above. The
duration of the tests is 5hours at full scale [5].
The Fig. 8 represents the two sensor configuration used during
our experiments. On the left side, the sensors location for
=270 is represented. On the right side, the sensors location
for the other headings considered here is represented.
Fig. 8: Sensors Location for =270 / other headings
First, we consider harmonic excitations obtained after spectral
analysis for the case of zero forward speed. The pressure
levels caused by non coupled and coupled motions are shown
just below and can be compared.
For instance, the Fig. 9 shows the maximum pressure recorded
among our pressure sensors for Non Coupled/Coupled
motions. The Fig. 10 represents the highest average recorded
among all the sensors of the ten highest pressure peaks
(N.C./C. motions). The Fig. 11 shows the maximum number
of impacts recorded among all the pressure sensors (N.C./C.
motions). The Fig. 12 shows the statistical pressure associated
at 3 hourreturn period (named P
stat
in the whole paper),
calculated by the probability law which fits the best among
Weibull3 parameters, LogNormal 3 parameters (N.C./C.
motions). For each result, a unique scale is used, selected as
the one giving the highest values between (non coupled,
coupled) / (harmonic) excitations. Results for random
excitations will be presented during the Workshop.
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
7.5 8.5 9.5 10.5 11.5 12.5
225
240
255
270
285
300
1400
1260
1120
980
840
700
560
420
280
140
0
HARMONIC  NON COUPLED  Pmax
Pmax
(mbar)
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
7.5 8.5 9.5 10.5 11.5 12.5
225
240
255
270
285
300
1400
1260
1120
980
840
700
560
420
280
140
0
HARMONIC  COUPLED  Pmax
Pmax
(mbar)
Fig. 9: Harmonic Pmax : Non Coupled / Coupled
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
7.5 8.5 9.5 10.5 11.5 12.5
225
240
255
270
285
300
700
630
560
490
420
350
280
210
140
70
0
HARMONIC  NON COUPLED  10 Pmax
10 Pmax
(mbar)
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
7.5 8.5 9.5 10.5 11.5 12.5
225
240
255
270
285
300
700
630
560
490
420
350
280
210
140
70
0
HARMONIC  COUPLED  10 Pmax
10 Pmax
(mbar)
Fig. 10: Harmonic 10 Pmax : Non Coupled / Coupled
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
7.5 8.5 9.5 10.5 11.5 12.5
225
240
255
270
285
300
1650
1485
1320
1155
990
825
660
495
330
165
0
HARMONIC NON COUPLED  N impacts
N impacts
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
7.5 8.5 9.5 10.5 11.5 12.5
225
240
255
270
285
300
1650
1485
1320
1155
990
825
660
495
330
165
0
HARMONIC COUPLED  N impacts
N impacts
Fig. 11: Harmonic N impacts : Non Coupled / Coupled
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
7.5 8.5 9.5 10.5 11.5 12.5
225
240
255
270
285
300
1000
900
800
700
600
500
400
300
200
100
0
HARMONIC  NON COUPLED  Pstat
Pstat
(mbar)
WAVE PERIOD Tp (s)
R
E
L
A
T
IV
E
W
A
V
E
H
E
A
D
IN
G
(
)
7.5 8.5 9.5 10.5 11.5 12.5
225
240
255
270
285
300
1000
900
800
700
600
500
400
300
200
100
0
HARMONIC  COUPLED  Pstat
Pstat
(mbar)
Fig. 12: Harmonic Pstat : Non Coupled / Coupled
Overall Analysis: 5hour FullScale Duration
Even if all headings have not been represented on the above
graphs, the most prevailing cases are displayed. Indeed, the
pressure levels for the headings (=180, 195, 210) are low
compared to those detailed hereunder.
For the harmonic excitation, noncoupled motions represent
the critical case for all kind of results considered (P
max
, 10
P
max
, number of impacts, P
stat
).
Even if the most critical case is difficult to identify since it
differs from the parameter studied (P
max
, 10 P
max
, number of
impacts or P
stat
), the 4 following cases appear to be the most
relevant:
(Tp, )=(10.5 s, 285)
(Tp, )=(9.5 s and 10.5s, 270)
(Tp, )=(10.5 s, 255)
Pressure levels for coupled motions are low compared with
those obtained with noncoupled motions.
As it will be shown during the Workshop, concerning random
excitations, the same tendency between noncoupled and
coupled motions is observed. Indeed, noncoupled motions
appear to be more critical in terms of pressure levels, except
for the number of impacts.
Finally, for 5hour fullscale duration analysis among all the
hydrodynamic configurations (noncoupled/coupled;
harmonic/random) considered, the most critical one appears to
be the noncoupled motion with harmonic excitation.
Critical Cases: 30hour FullScale Duration
Following these results, a focused analysis from fivehour to
sixtyhour full scale has been carried out on four the most
severe cases listed here above.
Our main concern was to assess statistical pressures P
stat
using
additional statistical distributions: Pareto and Generalized
Extreme Value, selected threshold level, 95% confidence
intervals, and the test duration required to collect sufficient
size of the data sample allowing the convergence of statistical
results.
In that regard, our observation is that 30hours experiment is
giving satisfactory steadiness of P
stat
(for 3hours return
period) and confidence intervals. This test duration has been
selected for all comparisons of critical cases using 4 statistical
laws. This detailed analysis of 30hours full scale leads to the
same conclusions as 5hour full scale cases elaborated here
above.
CONCLUSION
The influence of dynamic coupling due to the interactions
between ship motions and tank liquid motion on the pressure
levels on tank boundaries is here investigated by experimental
means, using 6 d.o.f. test rig and 1/70 scaled tank model of
standard LNG Carrier.
Among all considered hydrodynamic configurations (non
coupled/coupled; harmonic/random), the most critical one for
the sloshing pressure appears to be noncoupled motion with
harmonic excitation. This first conclusion of the study is very
interesting since nowadays, due to the restricted computer
resources, the major part of numerical calculations using CFD
tools are performed with harmonic excitations. In addition, the
state of the art of hydrodynamic computation is based on
classical assumptions of rigid body motion without dynamic
effects of free surfaces in tanks. Comparing sloshing effects
induced by traditional sloshing excitation and the realistic
one which is the random motion accounting for dynamic
coupling with liquid motion in the tanks, it appears that
currently used numerical model seems to be conservative, at
least for the cases studied herein.
Further numerical sloshing analysis are envisaged to be
carried in order to verify and confirm the observations from
experimental study presented in this paper. For instance, the
same kind of results will be presented at the Workshop for
two partial fillings of 50%H instead of 30%H considered here.
Comparisons between these experimental results and
numerical calculations will be presented at the Workshop.
In addition, due attention should be given to the statistical
adjustment, particularly related to the proper selection of
applicable statistical distribution and relevant acceptance
criteria. Finally, it is to be underlined that conclusions drawn
out from this study should remain restricted only to the
assumed operational case and any extrapolation to other
configuration (as other filling or other tank capacity, for
instance) may mislead to the erroneous recommendation.
REFERENCES
[1] MALENICA, ., ZALAR, M. & CHEN, X.B., Dynamic
coupling of seakeeping and sloshing, 13
th
ISOPE
Conference, Honolulu, USA, 2003.
[2] ZALAR, M., CAMBOS, P., BESSE, P., LE GALLO, B.
& MRAVAK, Z., Partial Fillings of Membrane Type
LNG Carriers, 21
st
GASTECH Conference, Bilbao,
Spain, 2005.
[3] ZALAR, M., MALENICA, ., & DIEBOLD L.:
"Selected Hydrodynamic Issues in Design of Large LNG
Carriers", RINA ICSOT Conference, Busan, Korea, 2006.
[4] GAILLARDE, G., LEDOUX, A & LYNCH, M.
Coupling Between Liquefied Gas and Vessels Motion
for Partially Filled Tanks: Effects on Seakeeping, RINA
Conference, London, UK, September 2004.
[5] ZALAR M., DIEBOLD L., BAUDIN E. & HENRY J.:
"Sloshing Effects Accounting for Dynamic Coupling
between Vessel and Tank Liquid Motion ", 26
th
OMAE
Conference, San Diego, USA, 2007.
[6] NEWMAN, J.N., Wave effects on vessels with internal
tanks, 20
th
Workshop on Water Waves and Floating
Bodies, Spitsbergen, Norway, 2005.
[7] KIM, Y., Numerical simulation of sloshing flows with
impact load, Applied Ocean Research, 23, 2001.
[8] ROGNEBAKKE, O.F., FALTINSEN, O.M., Coupling
of sloshing and ship motions, Journal of Ship Research
47, 2003.
Abstract for 23rd IWWWFB, Jeju, Korea, 2008
Steep wave impact onto a complex 3D structure
Ten I.
1
, Korobkin A.
2
, Malenica
S.
1
, De Lauzon J.
1
& Mravak Z.
1
(1) BUREAU VERITAS  DR, Paris, France (igor.ten@bureauveritas.com)
(2) School of Mathematics, University of East Anglia, Norwich, UK
Introduction
The hydroelastic interactions during the impact of steep wave onto a complex structures are discussed.
This problem is relevant for sloshing impacts apearing in the tanks of the LNG ships. The overall problem
of sloshing impacts being extremely complex, some rational approximations were presented in [2], and
this work represents the specic part related to the steep wave impact. First results for this type of
impact were presented in [1] for purely 2D case and here we extend it to the impact onto a 3D complex
structures, such as the NO96 boxes of the containement system of the LNG tanks. Indeed, even if the
uid ow can reasonably be approximated as a 2D, the complexity of the containement system structure
need to be considered fully 3D and solved by complex numerical solvers such as Abaqus, Nastran ... The
method which is presented here uses the so called hybrid approach which means that the uid ow is
modelled by the 2D strip approach, while the structural behavior is solved using the 3D FEM model.
Mathematical formulation
The basic conguration before impact is shown in Figure 1. The uid of height H and width L occupies
a region x < 0, 0 < y < L, and 0 < z < H, where the plane z = 0 corresponds to the at rigid bottom,
and the vertical zaxis is directed upward. Before the impact, t < 0, a part of the liquid boundary x = 0,
0 < z < H H
w
is in contact with the vertical wall. The boundary part x = 0, H H
w
< z < H
corresponds to the vertical face of the wave (hydraulic jump), which approaches the wall at constant
speed U and hits the wall at t = 0. Only one part of the vertical wall is elastic, S [y
1
, y
2
] [z
1
, z
2
], and
the rest is rigid.
U
H
Hw
L
U
c =
0
H
Hw
2
=0
y
=0
=0
x
=w
x t
=Uw
x t
=U
x
z1
z2
y1 y2
Figure 1. Formulation of the problem
The uid ow is studied within the acoustic approximation. At the initial stage of the impact, which
is of short duration, the problem is linearized and the boundary value problem shown in Figure 1 is
formulated. The problem is solved in nondimensional variables which relate to dimensional one, denoted
by a prime, as follow
x
= Hx, y
= Hy, z
= Hz,
= UH, H
w
= Hh
w
, t
=
H
c
0
t, p
= Uc
0
p, w
=
HU
c
0
w, L = Hl. (1)
where, c
0
, , , p, and w are sound speed in the uid at rest, the uid density, velocity potential, pressure
distribution, and deection of the structure, respectively.
Fluid ow
Within the acoustic approximation the ow is potential. Velocity potential (x, y, z, t) satises in non
dimensional variables (1) the wave equation
=
tt
(x, y, z) (, 0] [0, l] [0, 1], (2)
boundary conditions
y
= 0, y = 0 and y = l, (3)
= 0, z = 1, (4)
z
= 0, z = 0, (5)
x
=
1
+ w
t
2
, x = 0, (6)
where
1
(y, z) =
1, 0 < y < L, 1 h
w
= h
1
< z < 1,
0, 0 < y < L, 0 < z < h
1
= 1 h
w
,
2
(y, z) =
1, (y, z) S,
0, (y, z) / S,
(7)
and initial conditions
(x, y, z, 0) =
t
(x, y, z, 0) = 0. (8)
The hydrodynamic pressure acting on the wall is given by the linearized CauchyLagrange integral
p(y, z, t) =
t
(0, y, z, t). (9)
The solution of the problem (2)(8) is sought in the form of eigenfunction expansion:
(x, y, z, t) =
n=1
X
n
(x, t)V
n
(y, z). (10)
V
n
=
2
l
cos [
i(n)
y] cos [
j(n)
z],
i
=
i
l
(i = 0, 1, 2, ...),
j
=
2
(2j 1) (j = 1, 2, ...), (11)
nm
= 1 if n = m and zero otherwise. The eigenfunctions V
n
(y, z) satisfy the boundary conditions (3)(5).
Substituting (10) into (2) we obtain equations for the coecients X
n
(x, t) which form boundaryvalue
problem with initial conditions provided by (8). The obtained problem is solved with the help of Laplace
transform. Finally, after few algebra, the solution at x = 0 can be written as the following
X
n
(0, t) =
t
0
F
n
()J
0
(r
n
(t ))d, F
n
(t) =
1
0
l
0
[
1
(y, z) + w
t
(y, z, t)
2
(y, z)] V
n
(y, z)dydz, (12)
where J
0
is the Bessel function.
The pressure distribution p(y, z, t) on the wall, x = 0, is :
p(y, z, t) =
n=1
d
dt
t
0
F
n
()J
0
(r
n
(t ))d
V
n
(y, z). (13)
Structural dynamics
The structural deection at the wall w(y, z, t) is the solution of the structural dynamic equation subjected
to the following initial conditions:
w(y, z, 0) = w
t
(y, z, 0) = 0. (14)
Regardless of the method (analytical or numerical) which is employed to solve the structural dynamics,
the wall deection is developed in the following series:
w(y, z, t) =
n=1
a
n
(t)
n
(y, z), (15)
where
n
(y, z) are the adequate shape functions.
The most natural choice for the shape functions are the structural eigenmodes because they satisfy the
boundary conditions by denition, and in addition they are orthogonal. The orthogonal property of the
eigenmodes allows the reduction of the structural dynamic problem to the evolution equation for the
modal amplitudes a
n
(t):
n
d
2
a
n
dt
2
+ d
n
a
n
+
n
da
n
dt
S
p(y, z, t)
n
(y, z)dS, (16)
with the following initial conditions:
a
n
(0) = 0, a
n
(0) = 0. (17)
The coecients
n
, d
n
and
n
are the normalized mass, stiness and damping coecients respectively.
In the case of simple elastic structures (uniform beam, plate, ...) these coecients can be obtained
analytically, but in the general case the numerical methods are usually employed. The most common
method is the nite element method (FEM) which will be used here in the context of the commercial
code Abaqus.
Coupling
In order to solve the coupled hydroelastic problem, we need to express the righthand side in (16), in terms
of the modal coecients a
n
(t). After some algebra we arrive at the following system of integrodierential
equations with the unknown coecients a
n
(t):
n
d
2
a
n
dt
2
+ d
n
a
n
+
n
da
n
dt
= P
n
(t)
d
2
dt
2
m=1
t
0
a
m
()K
nm
(t )d. (18)
Here
P
n
(t) =
k=1
v
k
T
kn
J
0
(r
k
t), K
nm
(t) =
k=1
T
km
T
kn
J
0
(r
k
t), (19)
T
kn
=
S
V
k
(y, z)
n
(y, z)dS, v
k
1
1h
w
l
0
V
k
(y, z)dydz. (20)
Let us dene new unknown function b
n
(t) as
b
n
n
a
n
+
m=1
t
0
a
m
()K
nm
(t )d. (21)
Then equation (18) takes the form
d
2
b
n
dt
2
+ d
n
a
n
+
n
da
n
dt
= P
n
(t), n = 1, 2, ... . (22)
The system (21) and (22) is solved numerically with the following initial conditions (n > 1):
b
n
(0) = 0,
b
n
(0) = a
n
(0) = 0. (23)
Numerical Results
In order to validate the coupling procedure for FEM structural modelling, we chose the case of the uniform
plate for which the analytical solution is available. In the case of the FEM method, the integral (20)
must be evaluated numerically. This is done by tting
(l)
n
, where superscript l denotes the
n
value at
the lth node, by BSpline tting surface
n
(y, z) using the IMSL standard Fortran subroutines.
We chose the case where the plate occupies the whole wall and the following basic parameters are used:
sound speed c
0
= 1500m/s, water density
w
= 1000kg/m
3
, impact velocity U = 1m/s, structural
damping = 0.001s, water height H = 2.0m, wave height H
w
= 0.5m Poissons ratio = 0.3, plate
density
b
= 7800kg/m
3
, Youngs module E = 0.207e12N/m
2
, width of the wall L = 2.0m and plate
thickness h = 0.02m.
In Figure 2, the time history of the plate deection at few representative points is presented. As we can
see the agreement between two class of results is almost perfect which concludes the validation of the
numerical model.
Now we chose more complex case of 3D structure representing the rectangular box of 1m width, 1m
height and 0.3m depth. The center of the box is placed at y = 1.0m, z = 1.5m and the channel width is
L = 2m. All other parameters are the same. Few snapshots during the impact are presented in Figure 3.
There is no results for comparisons in this case, and we can just mention that the calculations are stable
both in space and in time.
Commercial code
Analytical
z = 1.759 m
z = 1.241 m
z = 1.500 m
0.003
0.002
0.001
0.000
0.001
0.002
0.003
d
e
e
c
t
i
o
n
s
,
m
time, sec
0.00 0.01 0.02 0.03 0.04
Commercial code
Analytical
z = 1.759 m
z = 1.241 m
z = 1.500 m 0.003
0.002
0.001
0.000
0.001
0.002
0.003
d
e
e
c
t
i
o
n
s
,
m
time, sec
0.00 0.01 0.02 0.03 0.04
Figure 2. Time history of plate deection for few representative points at y = 1.0m (left) and y = 1.5m (right).
Figure 3. Snapshots during steep wave impact onto rectangular box.
Conclusions
We presented here the semi numerical method able to simulate the steep wave impact onto a complex
structures modelled by the general 3D FEM numerical codes such as Abaqus. The model was validated
on the case of elastic plate for which the analytical solution is available. The future work consist in
applying and validating the method on the real NO96 boxes used in the tanks of LNG carriers.
References
[1] Korobkin, A. & Malenica
S., 2007. : Steep Wave Impact Onto Elastic Wall, 22nd IWWWFB,
Plitvice, Croatia.
[2] Malenica
S, Korobkin A.A., Scolan Y.M., Gueret R., Delafosse V., Gazzola T.,
Mravak Z., Chen X.B. & Zalar M. 2006. : Hydroelastic impacts in the tanks of LNG carriers.,
4th Int. Conference on Hydroelasticity, Wuxi, China.
SECONDORDER WAVE LOADS ON A LNG CARRIER
IN MULTIDIRECTIONAL WAVES
Mathieu Renaud
Research Department
Bureau Veritas (France)
Fl avia Rezende
Research Department
Bureau Veritas (France)
Olaf Waals
MARIN
(The Netherlands)
Xiaobo Chen
Research Department
Bureau Veritas (France)
Professorship, HEU (China)
Radboud van Dijk
MARIN
(The Netherlands)
ABSTRACT
Due to the installation of LNG terminals moored in prox
imity to the coast, the wave kinematics in shallow water and the
consequence on the behavior of those terminals have recently be
came a major concern of the offshore industry. One key issue is
the accurate simulation of the lowfrequency motions of LNG
carriers, specially the surge, for which the vessel presents low
damping, in order to perform the design of the mooring system.
The present paper focuses on the effect of wave directionality
on secondorder slowdrift loads and the related response of the
vessel.
The paper describes results of model tests in regular cross
waves  monochromatic but coming from two directions sepa
rated by 90 degrees, as well as bichromatic cross waves. The
new middle eld formulation extended to the case of cross
waves, is used to compute the wave drift loads and lowfrequency
Quadratic Transfer Function (QTF). The results are compared
with those from the model tests.
NOMENCLATURE
LNG Liqueed Natural Gas
QTF Quadratic Transfer Function
i
Pulsation of the incoming wave coming from direction
i
Firstorder velocity potential
i
Phases of rstorder incoming waves associated with
(
i
,
i
)
a
i
Wave amplitude of incoming waves associated with (
i
,
i
)
k
i
Wavenumber of incoming waves associated with (
i
,
i
)
Elevation of free surface
E(
i
,
j
) Phase function : representing the phase difference of
two cross waves
INTRODUCTION
Large LNG terminals have been designed to operate in off
shore areas approximate to harbors, where water is of nite depth
and waves are multidirectional. In the design of such mooring
systems of LNG terminals in a zone of shallow water, one im
portant concern is the accurate simulation of the lowfrequency
motions to which the secondorder wave loading is well known
as the main source of excitation. In the last decade, the focus of
the hydrodynamic research has been on deep and ultra deep wa
ter. Although there have been many works on the estimation of
secondorder loads for the case of long crested waves, there are
relatively few works on the estimation of secondorder loads in
directional waves including the works in [2], [3] and [5].
In [6], the formulations of secondorder wave loads were
1 Copyright c 2008 by ASME
Proceedings of the ASME 27th International Conference on Offshore Mechanics and Arctic Engineering
OMAE2008
June 1520, 2008, Estoril, Portugal
OMAE200857409
presented, in particular, the new formulation based on the use of
control surface was shown to provide results as accurate as the
fareld formulation and as general as the pressureintegration
formulation. The excellent agreement of secondorder drift loads
in cross waves with the semianalytical results validates the de
velopment.
Further to the work in [6], the formulations of secondorder
lowfrequency loads in cross waves are presented here. The re
sults of model tests of a LNG carrier in regular cross waves
and bichromatic cross waves and the comparison with numeri
cal computations are illustrated and analyzed.
FORMULATION OF LOWFREQUENCY LOADS
The lowfrequency quadratic transfer function (QTF) is de
ned as the secondorder wave loads occurring at the frequency
equal to the difference (
1
2
) of two wave frequencies
(
1
,
2
) of bichromatic waves. Furthermore, the wave directions
relative to the positive xaxis are denoted by (
1
,
2
). It is under
stood that the bidirectional bichromatic wave is characterized
by, at the rst order, one regular wave of (
1
,
1
) and another of
(
2
,
2
), without loss of generality.
QTF is composed of two distinct parts : one dependent only
on the quadratic products of rstorder wave elds and another
contributed by the secondorder potentials of the incoming and
diffracted waves.
F(
1
,
2
,
1
,
2
) = F
q
(
1
,
2
,
1
,
2
) +F
p
(
1
,
2
,
1
,
2
) (1)
The rst part F
q
can be written in the way presented in [10] :
F
q
=
2
ZZ
H
ds
_
(
1
2
)n
1
n2
n1
2
_
2
2g
I
d(
1
2
)n (2)
as integration over the hull H and along the waterline in their
mean position. In (2), stands for the rstorder velocity poten
tial and
n
= n the normal derivative of on H. The sub
scripts (
1
,
2
) represent the quantities associated with the wave
frequencies and directions (
1
,
1
) and (
2
,
2
), respectively,
while the superscript
indicates the complex conjugate.
The formulation (2) derived from Eq.27 in [10] obtained by
applying the two variants of Stokes theorem given in [7] to the
classical neareld (pressureintegration) formulation as in [8],
is compact and used here. It is directly applicable to force com
ponents in horizontal directions. The extension to other compo
nents is direct and omitted here. In (2), we involve the gradient of
velocity potentials which is sensitive to the singularities present
in the velocity eld at sharp corners. In particular, the integration
of terms (
1
2
) converges slowly or in the worst cases, may
be nonconvergent.
In [10], after having the new neareld formulation by ap
plying the variants of Stokess theorem, we considered a uid
volume enclosed by the hull, a control surface at a distance from
the body and the mean free surface limited by the waterline and
the intersection of the control surface with free surface, and have
obtained the general formulation (Eq.8a & 8b in [10]) of second
order loads by using Gausss theorem. This formulation can be
simplied if we construct a control surface surrounding the hull
touching the free surface only along the waterline :
F
q
= (
1
2
)
2
ZZ
H
ds
_
n1
2
/
1
n2
1
/
2
_
2
2g
I
d(
1
2
)n
2
ZZ
C
ds
_
(
1
2
)n
n2
n1
2
_
(3)
in which C stands for the control surface dened as an arbitrary
one surround the body.
The integration of terms (
1
2
) in (3) now performed
on the control surface C converges rapidly since C is at some dis
tance from the hull where the velocity eld does not present any
singularity. The integration on the hull H is of order O(
1
2
)
and as small as
n
which tends to zero for large wave frequencies.
The middleeld formulation (3) is used for the computation of
the rstpart of secondorder loads F
q
in the following.
The second part F
p
is expressed in the way [4] :
F
p
=i(
1
2
)
ZZ
H
ds
_
(2)
I
n(
(2)
In
N
H
)[]
_
+i(
1
2
)
g
ZZ
F
dsN
F
[] (4)
in which the rst term in the hull integral corresponds to the
secondorder FroudeKrylov component contributed by the in
coming wave potential
(2)
I
dened by :
(2)
I
=ia
1
a
2
Ag
2
coshk
m
(z+h)
coshk
m
h
exp[ik
m
x +i(
1
2
)]
gk
m
tanhk
m
h (
1
2
)
2
(5)
with
k
m
x = (k
1
cos
1
k
2
cos
2
)x +(k
1
sin
1
k
2
sin
2
)y (6)
(
1
,
2
) being the phases of rstorder incoming waves associated
with (
1
,
1
) and (
2
,
2
), respectively.
2 Copyright c 2008 by ASME
In (5), k
m
is given by :
k
m
=
_
k
2
1
+k
2
2
2k
1
k
2
cos(
1
2
) (7)
and A written :
A =
1
2
k
1
k
2
[cos(
1
2
) +tanhk
1
htanhk
2
h]
+
1
2
_
k
2
1
/
1
cosh
2
k
1
h
k
2
2
/
2
cosh
2
k
2
h
_
(8)
where we have used the notations (a
1
, a
2
) and (k
1
, k
2
) stand
ing for the wave amplitudes and wavenumbers associated with
(
1
,
2
) via the dispersion equation k
1,2
tanhk
1,2
h =
2
1,2
/g with
the waterdepth h, respectively, while the wave heading with re
spect to the positive xaxis is denoted by .
The second term in the hull integral of (4) and the term de
ned by the integral over mean free surface F come from the
application of Haskind relation and represent the contribution of
the secondorder diffraction potential, as shown in [4]. The terms
(N
H
, N
F
) are the second members of the boundary conditions
satised by the secondorder diffraction potential on the hull H
and the mean free surface F, respectively. They are written as :
2N
H
=(x
1
)
2
n(x
2
)
1
n
(i
2
x
2
) (R
1
n) (i
1
x
1
+
1
) (R
2
n) (9)
and
N
F
= i(
1
2
)(
1
P2
+
P
I2
)
i
1
2g
_
1
(
2
2
z
+g
2
zz
)
P2
+gk
2
2
(1tanh
2
k
2
h)
P1
I2
)
+
i
2
2g
_
2
(
2
1
z
+g
2
zz
)
P1
+gk
2
1
(1tanh
2
k
1
h)
P2
I1
)
(10)
in which x is the displacement vector at a point on H and R the
vector of rotations. In (10),
I
represents the rstorder potential
of incoming waves while
P
=
I
stands for that of perturba
tion including the diffraction and radiation components. Finally,
[] in (4) represents a vector of rstorder radiation potentials os
cillating at the difference frequency (
1
2
). They satisfy the
homogeneous condition :
_
(
1
2
)
2
+g
z
[] = 0 (11)
on the mean free surface F and
n
[] = n (12)
on the hull H.
The full QTF (1) is composed of two parts (F
q
, F
p
) given
by the formulations (2) and (4), respectively. The formulation
(2) for F
q
derived in [10] is simpler than that in [8]. The for
mulation (4) for F
p
by [4] is often called indirect method since
it provides a way to evaluate the contribution from the second
order diffraction potential through the Haskind relation such that
the secondorder diffraction potential is not explicitly computed.
SECONDORDER LOADS IN CROSS WAVES
In cross waves of two frequencies (
1
,
2
) with two head
ings (
1
,
2
), the rstorder elevation of free surface for regular
waves can be expressed by :
(t) ={
1
e
i
1
t
+
2
e
i
2
t
} (13)
with the complex amplitudes :
1
= a
1
e
ik
1
(xcos
1
+ysin
1
)+i
1
(14a)
2
= a
2
e
ik
2
(xcos
2
+ysin
2
)+i
2
(14b)
associated with the real amplitude (a
1
, a
2
), wave numbers
(k
1
, k
2
), coordinates (x, y) of the reference point and the initial
phases (
1
,
2
) of each wave component.
The time series of secondorder lowfrequency wave loads
are then written by :
F(t) =
_
1
2
1
F(
1
,
1
,
1
,
1
) +
1
2
2
F(
2
,
2
,
2
,
2
)
+
1
2
F(
1
,
2
,
1
,
2
)e
i(
1
2
)t
_
(15)
in which F(
1
,
2
) are the QTF given by (1) and computed by
(3) for the rst part and (4) for the second part.
In the cross waves of unique wave frequency (
1
=
2
), the
lowfrequency load is reduced to the drift loads expressed by :
F(t) =
_
1
2
a
2
1
F(
1
,
1
,
1
,
1
) +
1
2
a
2
2
F(
1
,
1
,
2
,
2
)
+a
1
a
2
E(
1
,
2
)F(
1
,
1
,
1
,
2
)
_
(16)
in which (a
1
, a
2
) are the real amplitudes of cross waves dene in
(14) and E(
1
,
2
) is the phase function :
E(
1
,
2
) = e
ik
1
x(cos
1
cos
2
)+ik
1
y(sin
1
sin
2
)+i(
1
2
)
(17)
3 Copyright c 2008 by ASME
representing the phase difference of two cross waves (14).
The rst two terms in (16) are the drift forces associated
with each regular wave while the third one is contributed by the
interaction between two waves of different headings. From (17),
the interaction term varies with the phase function depending on
the reference position and initial phases of two regular waves.
NUMERICAL AND MODEL TEST RESULTS
Numerical computations and model tests are performed for a
standard 138000 m
3
LNG vessel moored in the waterdepth equal
to 15m. The main particulars of the LNG vessel are shown in
Table 1. The vessel was moored by a soft spring mooring system
The mooring stiffnesses given in Table 1 were adjusted to obtain
a realistic natural period for surge with a target value of 124s.
Table 1. MAIN PARTICULARS OF LNG VESSEL
Designation Sym. Unit Value
Length between perpendiculars Lpp m 274.000
Width B m 44.200
Depth D m 25.000
Draft T m 11.000
Displacement m
3
97120
COG above keel KG m 16.300
COB above keel KB m 5.857
COB from Aft perpendicular LCB m 135.676
Metacentric height GM m 4.770
Transverse gyration radius Kxx m 15.200
Longitudinal gyration radius Kyy m 68.500
Longitudinal mooring stiffness Cx KN/m 281.000
Transverse mooring stiffness Cy KN/m 254.000
Several series tests of cross waves have been performed in
the Offshore Basin of MARIN. The basin has a movable oor
which is used to adjust the water depth to 15m. In all cases, two
regular waves of headings 135
and 225
on the port
side/starboard sides. The cross waves are also computed and also
illustrated on Figure 2 below the picture of model tests.
The mesh used in Hydrostar is composed of 4408 at panels
on the wetted part of LNGvessel. For the computation of second
order wave loads by using the middleeld formulation, a control
surface composed of 3330 at panels is automatically generated
and presented on Figure 3 together with the hull mesh.
First, the mean drift forces F
x
in cross waves of headings
(
1
= 135
and
2
= 225
, 225
)
the lowfrequency loads F
x
are depicted on Figure 7. The real and
imaginary parts of numerical computations are presented by solid
and dotdashed lines, respectively while the test measurements
by the symbols (circle and square) for the real and imaginary
parts, respectively.
Finally, we present some results of numerical computations.
The real and imaginary parts of drift forces F
x
in regular cross
waves are depicted on Figures 8 and 9. The main heading of reg
ular waves is 180
to 165
are presented.
The lowfrequency loads F
x
oscillating at =0.05 rad/s in
bichromatic cross waves are presented on Figures 10 and 11 for
the real and imaginary parts, respectively. The main heading of
waves is 180
, )
The wave frequency varies from 0.3 to 1.6 rad/s while the sec
ond wave heading from 120
to 165
.
For two xed headings (135
, 225
), lowfrequency wave
forces for different differencefrequencies varying from 0.0 to
0.075 rad/s are presented on Figures 12 and 13 for the real and
imaginary parts, respectively.
150000
100000
50000
0
50000
100000
150000
200000
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Real part (165 deg)
Real part (150 deg)
Real part (135 deg)
Real part (120 deg)
Figure 8. REAL PART OF DRIFT LOAD IN REGULAR CROSS WAVES
150000
100000
50000
0
50000
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Imag. part (165 deg)
Imag. part (150 deg)
Imag. part (135 deg)
Imag. part (120 deg)
Figure 9. IMAGINARY PART OF DRIFT LOAD IN REGULAR CROSS
WAVES
CONCLUSIONS
The formulationof quadratic transfer function (QTF) of low
frequency wave loads in bichromatic cross waves is given in
the paper with the extension of middleeld formulation for the
rst part depending on the quadratic products of rstorder wave
elds. The results of numerical computations and experimen
tal measurements are presented. The comparison between nu
merical computations and model tests is globally good. How
ever, more results from model tests are needed for fully validate
the numerical developments. Although the further analyses are
needed to quantify the impact to practices, it is already shown
that the interaction effect between waves from different headings
is very important and has to be taken into account in the estima
tion of excitation loads to mooring systems.
6 Copyright c 2008 by ASME
150000
100000
50000
0
50000
100000
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Real part (165 deg)
Real part (150 deg)
Real part (135 deg)
Real part (120 deg)
Figure 10. REAL PART OF LOWFREQUENCY LOAD IN BICHRO
MATIC CROSS WAVES WITH = 0.05 rad/s
200000
150000
100000
50000
0
50000
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Imag. part (165 deg)
Imag. part (150 deg)
Imag. part (135 deg)
Imag. part (120 deg)
Figure 11. IMAGINARY PART OF LOWFREQUENCY LOAD IN
BICHROMATIC CROSS WAVES WITH = 0.05 rad/s
REFERENCES
[1] CHEN X.B. (2004) Hydrodynamics in Offshore and Naval
Applications  Part I. Keynote lecture at the 6th Intl Confer
ence on Hydrodynamics, Perth (Australia).
[2] PINKSTER J.A. (1988) The inuence of directional spread
ing of waves on mooring forces. Proc. OTC Conf., Houston
(USA), paper No.5629.
[3] KROKSTAD J.R. (1991) Secondorder loads in multidirec
tional seas. Ph.D. Phesis, The Norwegian Institute of Tech
nology.
[4] MOLIN B. (1979b) Secondorder drift forces upon large
bodies in regular waves. Proc. BOSS79., London (UK),
36370.
[5] MOLIN B. & FAUVEAU (1984) Effect of wave
directionality on secondorder loads induced by the set
100000
50000
0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Real. part (Dw =0.)
Real part (Dw =0.025)
Real part (Dw =0.05)
Real part (Dw =0.075)
Figure 12. REAL PART OF LOWFREQUENCY LOAD IN BICHRO
MATIC CROSS WAVES WITH (
1
,
2
) = (135
, 225
)
0
50000
100000
150000
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Imag part (Dw =0.025)
Imag part (Dw =0.05)
Imag part (Dw =0.075)
Figure 13. IMAGINARY PART OF LOWFREQUENCY LOAD IN
BICHROMATIC CROSS WAVES WITH (
1
,
2
) = (135
, 225
)
down. J. Appl. Ocean Research. Vol.6, No.2, 6672.
[6] REZENDE F., LI X. & CHEN X.B. (2007) Secondorder
loads on LNG terminals in multidirectional sea in water
of nite depth. Proc. 26th Intl. Conf. OMAE, San Diego
(USA).
[7] DAI Y.S. (1998) Potential ow theory of ship motions in
waves in frequency and time domain (in Chinese). Press of
the National Defense Industries, Beijing (China).
[8] PINKSTER J.A. (1980) Low frequency second order wave
exciting forces on oating structures. H. Veenman En Zonen
B.V.  Wageningen (The Netherlands).
[9] FERREIRA M.D. & LEE C.H. (1994) Computation of
secondorder mean wave forces and moments in multibody
interaction. Proc. 7th Intl Conf. Behaviour Off. Structures,
BOSS94, Boston (USA) 2, 30313.
7 Copyright c 2008 by ASME
[10] CHEN X.B. (2006) Middleeld formulation for the com
putation of wavedrift loads Journal of Engineering Math
ematics. Published online: 27 Sep 2006, Vol.59, pp6182.
8 Copyright c 2008 by ASME
Diraction and radiation with the eect of bathymetric
variations
G. de Hauteclocque, F. Rezende & XB Chen
Research Department, Bureau Veritas
Courbevoie, Paris La Defense
email: guillaume.dehauteclocque@bureauveritas.com
ABSTRACT: New developments have been performed to study the wave kinematics and
seakeeping of oating bodies in an area of variable water depth. It consists of extending
the classical diraction/radiation method applicable to the case of constant water depth to
be able to capture the major eects of the bathymetry. The part of sea bottom under the
oating body is modeled as a second xed body. The formulation of partial transparency
is applied to an extended zone outside the opaque area to eliminate the unexpected dis
turbance due to the otherwise abrupt truncation of sea bottom along the edge of opaque
bottom. The good comparison with semianalytical solutions (GNWave) of wave kinemat
ics along a slopped seabed has been obtained. Numerical results for the case of a barge
oating over a horizontal bathymetry and for a LNG over a slopped seabed show that
the interference eect of variable bottom on the wave loads is important. The method
can then be used in the context of LNG terminals in coastal zones where the bathymetric
variation cannot be ignored.
1 INTRODUCTION
The recent development of LNG terminals has shown the need to extend the capability
of seakeeping computations to relatively shallow water area. Among the dierent issues
associated with the shallow water, there is the inuence of a varying bathymetry which
might induce signicant changes on kinematics of waves and on the behavior of the ship.
A classical way to take into account varying bathymetry in diraction/radiation sea
keeping code is to model the bottom as a xed second body. However, it has been shown
that this method presents diculties when the bottom has to be truncated at dierent
water depths (Buchner 2006). Indeed, the abrupt truncation of the bathymetry induces
unwanted and strong reections. A smoothly sloped extension of the boundary does re
duce these reections but not as much as expected (Buchner 2006). This paper presents a
new way to smooth more eciently the truncation by introducing a partially transparent
extension. The theoretical formulation of the transparency is explained in a rst part;
next the eciency of the method is tested and discussed on dierent congurations.
2 THEORY
The potential theory of wave diraction and radiation around oating or xed bodies is
adopted. A summary of the wave diraction/radiation theory is given in Chen (2004).
The modication on the boundary condition over part of sea bottom is explained here.
2.1 Diraction and radiation over an uneven bottom
A oating body above a globally at bottom but uneven in a restricted area is considered:
z =
h (x, y) B
h(x, y) (x, y)
B
(1)
where B represents the unlimited at bottom of depth h and
B the limited area where the
depth is variable. In the uid domain D limited by the mean free surface F (at z = 0),
the hull of the oating body H and bathymetry B
B, the velocity potential can be
decomposed as:
=
I
+
D
+
R
(2)
Where
I
is the incident wave eld,
D
the diraction potential and
R
the radiated
potential induced by the hull motion. The problem can now be written as:
D,R
= 0 P D
D,R
z
(
2
/g)
D,R
= 0 P F(z = 0)
D
n
=
I
n
and
R
jn
= n
j
P H
D,R
z
= 0 P B
D
n
=
I
n
and
R
jn
= 0 P
B
(3)
In which
R
j
is the j
th
component of the radiation potential expressed by
R
= i
6
j=1
R
j
with
j
the j
th
mode of motion.
2.2 Diraction over an uneven partially transparent bottom
Without loss of generality, we consider the diraction problem associated with a partially
transparent uneven bottom, without presence of body. The perturbation potential
P
is
dened as
P
=
D
+
T
(4)
so that
=
I
+
P
(5)
Where
D
is associated to the at bottom B and
T
to the partially transparent bottom
D
n
=
T
n
P B
T
n
= (1 )(
I
n
+
D
n
) P
B
(6)
Where is the transparency coecient. Thus, if = 0 we have
n
= 0 on
B, the uneven
bottom is thus fully opaque. If = 1, it can be shown that
D
=
T
= 0 and the uneven
bottom is not taken into account at all, i.e. fully transparent.
For value of between 0 and 1, it can be considered that the bottom is partially trans
parent. However, the decomposition of
P
in
D
and
T
is not unique, one way to close
the problem with a source method is to dene
T
only as the potential associated with
the source at P(x, y, z) while
D
comes from the contribution of all the source except of
those on P(x, y, z).
2.3 Boundary element method
Considering a source distribution (Q) on the hull H and on the bottom
B, the pertur
bation potential and its gradient are expressed by:
P
=
B
ds(Q)(Q)G(P, Q) and
P
=
B
ds(Q)(Q)G(P, Q) (7)
G(P, Q), which is the Green function representing a potential ow at a eld point P
induced by a source located at Q, satises :
G
z
(
2
/g)G = 0 P F(z = 0)
G
z
= 0 P B
2
G(P, G) = 4(P Q) P D
(8)
The source distribution is determined by the integral equation :
2(P) + PV
B
ds(Q)(Q)G
n
(P, Q) = V
n
P H
B (9)
in which the word PV before the integral means to interpret the integral as a Cauchy
principal value, and :
V
n
=
I
n
for =
D
P H
n
j
for =
R
j
P H
I
n
for =
D
P
B
0 for =
R
j
P
B
(10)
If the bottom
B is partially transparent ( coecient), the perturbation potential is de
composed as :
P
=
T
+
D
with
D
=
P
T
(11)
Where we choose to associate the potential
T
with the source at the point P. If we note
d(Q P) as the disc centered at P and with a radius tending to zero, this can be written
as :
T
(P) =
d(QP)
ds(Q)(Q)G
n
(Q) (12)
D
(P) =
Bd
ds(Q)(Q)G
n
(Q) (13)
From equation (6) we have
T
n
+ (1 )
D
n
= +(1 )V
n
P
B (14)
The coecient can now be introduced in the integral equation:
2(P) + (1 )
Bd
ds(Q)(Q)G
n
(Q) = (1 )V
n
P
B (15)
It can be veried that if = 1, the panel is thus transparent and if = 0, the integral
equation is identical to the one obtained with a normal bottom.
3 RESULTS
Three congurations are presented here to validate the eciency of a partially transparent
area to remove the edge eect. The rst one involves a horizontal bathymetry, the second
one the study of waves kinematics over a varying bathymetry, and the calculation of a
standard LNG RAOs over a sloped bottom.
3.1 Submerged horizontal bathymetry
A barge oating above a circular horizontal bathymetry is considered. The dimension of
the barge are 80m20m4m, the bathymetry diameter is 160m and its depth is set to
15m; the water depth is 20m. This case is chosen because of the availability of a reference
computation which is a classical computation at uniform water depth 15m.
(a) Side view (b) Top view
Figure 1: Barge over a submerged bottom, description of the geometry.
Without partially transparent area, the added mass and damping oscillate around the
exact solution. Increasing the size of the bathymetry makes the number of oscillations
decreases with no change on the amplitude.
With a partially transparent area, where the panels smoothly disappear (cos
2
distribu
tion), the amplitude of the oscillations is signicantly reduced, especially for high frequen
cies. The distribution of the transparency parameter has to be as smooth as possible, a
cos
2
distribution has been found to be particularly ecient. The next step is to dene
how long the transition area should be. Computations with dierent beach size (80, 160,
240, 320 and 400m) are thus performed. As expected, greater the partially transparent
area is, better the results are (Figure 2a). Moreover, as the beach increases, the result
converges to the reference.
To dene a relation between the wavelength and the eciency of the partially transpar
ent area size, the relative error is plot against the ratio Beach size / Wavelength (Figure
2b). Based on this graph, the beach size of one wave length seems to be a criterion (which
seems physically consistent) to reduce the reection on the edge of the bathymetry to an
acceptable level. The result for other added mass and damping terms are similar.
0.0E+00
2.0E+06
4.0E+06
6.0E+06
8.0E+06
1.0E+07
1.2E+07
0.1 0.3 0.5 0.7 0.9
Frequency (rad/s)
C
A
3
3
0m
80m
160m
240m
reference
(a) Radiation damping
0
1
2
3
4
5
6
7
8
0 1 2 3 4 5
Ratio wavelength / Beach length
C
A
3
3
r
e
l
a
t
i
v
e
e
r
r
o
r
(
%
)
260m
152m
99m
80m
(b) Relative error on the radiation damping
Figure 2: Barge over a submerged bottom, Radiation damping
3.2 Comparison with a shallow water code GNWave
As shown in Buchner (2006) the wave kinematics computed without any special treatment
is not realistic, due to strong reection at the edge of the bathymetry. The purpose of
this section is to check if the partially transparent area improves the wave kinematics
computation. The results obtained are compared with a 2D shallow water code, GNWave,
based on the GreenNaghdi theory. This theory is very dierent from the perturbation
approach. Instead of limiting the order of the solution, the assumption on the vertical
velocity shape and the momentum equation is then depth integrated. The calculation is
made in the temporal domain and allows a fully non linear solution of the problem. Details
on this formulation are provided by Demirbilek & Webster (1992).
As nonlinearity is not the topics of this paper, for further comparison with HydroStar,
which is in frequency domain, the temporal signal from GNWave is treated by a Fourier
transformation and only the rst harmonic is studied (Very low wave amplitudes are used
to avoid second order component). In practice, time evolutions of the wave elevation are
measured at several points of the numerical water channel and an FFT is then performed
on each signal. The time selected has to be long enough to reach the established behavior
and short enough to avoid parasite reected waves (reection on the wave maker and on
the imperfect opening condition).
The bathymetry used for this test is represented on gure 3a. In order to enlarge the
bathymetry impact, the slope is chosen quite steep.
(a) Bathymetry prole
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
500 700 900 1100 1300
X (m)
A
m
p
l
i
t
u
d
e
(
m
/
m
)
GNwave
Hydrostar
(b) Wave amplitude
Figure 3: 2D calculation
The comparison obtained in 2D is very satisfactory, two very dierent theories agree
quite well on the test (Figure 3b). The dispersion relationship is well satised upstream
with both codes. The reected waves computed by HydroStar and GNWave are compara
ble and quite weak. The smoothing area chosen for the HydroStar computation is about
2 wavelengths.
(a) 3D mesh with smoothing area
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
500 700 900 1100 1300
X (m)
A
m
p
l
i
t
u
d
e
(
m
/
m
)
GNwave
Hydrostar 3D average
Hydrostar y=0
Hydrostar Y=40
(b) Wave amplitude
Figure 4: 3D calculation
Compared to 2D computation which has shown good results, 3D computation should
not rise any other problem than mesh size. The mesh has to be wide enough, then, there
is the need of the smoothing area upstream, but also on the sides. It is thus hard to get
results as good as in 2D, because compromises have to be done (panel size, smoothing
area length or width of the opaque bathymetry) to keep an acceptable computation cost
(keeping in mind that the aim is to add a oating body to the computation).
The result obtained without using partial transparency presents a strong reected wave
in X direction and also signicant oscillations along the Y axis. The wave kinematics
pattern is obviously unrealistic (gure 5a) and would induce wrong loads when computing
a ship motion above this bottom. The smoothing areas in both X and Y direction allows
the computation of the wave refraction on the bathymetry almost without unphysical
reection.
The results with the incidence of 0
If it is incontestable that the depth has an impact on the motion of the vessel, it is not
obvious to predict whether or not the bathymetry variation has a signicant impact on the
behavior of the ship. The RAOs of the ship are compared with the modeled bathymetry
and at the mean depth (Figure 8a). For the surge motion, the variation of the bathymetry
does not change the response, however, for heave motion; the slope does have an inuence,
especially if the incidence diers from 0
(Figure 9).
This change of motion is not only caused by the modied kinematics of the incident
wave, but also from the modication of the added mass and damping of the ship (see
Figure 8b). In this conguration, that is to say with the LNG perpendicular to the coast,
the added mass and damping term of heave and roll are the most sensible to the sloped
bottom.
0.0E+00
1.0E01
2.0E01
3.0E01
4.0E01
5.0E01
6.0E01
7.0E01
8.0E01
9.0E01
1.0E+00
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Frequency (rad/s)
S
u
r
g
e
(
m
/
m
)
Water depth 22.5m
Sloped bottom
(a) Surge 0
7.6E+09
7.8E+09
8.0E+09
8.2E+09
8.4E+09
8.6E+09
8.8E+09
9.0E+09
9.2E+09
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Frequency (rad/s)
C
A
4
4
Water depth 22.5m
Sloped bottom
(b) Roll radiation damping
Figure 7: RAOs with and without sloped bottom
0.0E+00
1.0E01
2.0E01
3.0E01
4.0E01
5.0E01
6.0E01
7.0E01
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Frequency (rad/s)
H
e
a
v
e
(
m
/
m
)
Water depth 22.5m
Sloped bottom
(a) Heave 0
0.0E+00
1.0E01
2.0E01
3.0E01
4.0E01
5.0E01
6.0E01
7.0E01
8.0E01
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Frequency (rad/s)
H
e
a
v
e
(
m
/
m
)
Water depth 22.5m
Sloped bottom
(b) Heave 15
t
0
d. (1)
The Wagner approximation linearises the wet and free surfaces on the plan {z = 0}, assuming
uid and structure make a at impact.
The problem was rst solved for rigid body impacts (see [GKMS05]). For the issue of exible
structure impact, it is necessary to take into account the deection of the structure (see gure 1).
u
z
(x, y, t)
x
y
z
Figure 1: Water impact for a exible structure. The vertical deformation of the structure uz must be taken into account.
The vertical derivative
z
of the displacement potential on the plan {z = 0} characterises the
elevation of water (see e.g. [GKMS05]). The position of the structure on the wet surface
w
is
f h + u
z
, where f describes the structure shape, h is the penetration depth and u
z
is the vertical
deformation of the structure.
On the free surface
f
, the Wagner approximation leads to = 0. The boundary value problem
satied by is
= 0 in = {z < 0}
= 0 on
f
z
= f(x) h(t) +u
z
(x, t) on
w
0 in the far eld.
(2)
f
and
w
are unknown, and must be determined as part of the solution.
The boundary value problem (2) can be transformed into variational inequalities (see [GKMS05]).
a(, v ) l
0
f
(v ) +l
1
f
(u, v ), v K. (3)
The proper denition of functional space K can be found in [GKMS05]. a, l
0
f
and l
1
f
are
l
0
f
(v) =
{z=0}
(f(x, y) h(t)) v dx dy, (4)
l
1
f
(u, v) =
{z=0}
e
z
u(x, y, t) v dx dy, (5)
and
a(u, v) =
1
2
(x,y)D
(x
0
,y
0
)D
2
u(x, y)
2
v(x
0
, y
0
)
dx
0
dy
0
dx dy
(x x
0
)
2
+ (y y
0
)
2
. (6)
Bilinear form a is the Laplacian bilinear form, and is transformed using boundary element method.
The integration domain D is a part of {z = 0} large enough, such that we are sure that it contains the
wet surface (see [GKMS05] for details).
The uid variational formulation of the hydroelastic Wagner impact is very similar to its formula
tion for the uncoupled case. The only difference is the term l
1
f
(u, v) which translates the exibility of
the structure into the variational problem. Anyway, for a given deection u, the mathematical formu
lation for the coupled and uncoupled uid problems are equivalent, and the hydroelastic inequality
can be transformed into a constrained minimisation question.
= arg min
vK
1
2
a(v, v) l
0
f
(v) l
1
f
(u, v)
. (7)
1.2 Structure model
Writing the variational formulation of a linear elastic structure under the small displacement assump
tion is classical; details may be found in [Aub00]. The behavior law is
ij
= R
ijkl
kl
, with , R
and respectively the stress, Hooke and strain tensors.
The principle of virtual power (within the small perturbations approximation) is
2
u
t
2
v d +
s
R
ijkl
kl
(u)
ij
(v) d
=
{z=0}
(p n) v d, v H
1
(
s
)
3
. (8)
where u is the displacement and p is the uid pressure.
In order to keep the symmetry with the uid problem, we divide the classical mass and stiffness
bilinear forms by the uid volumic mass
f
. This leads to the denition of bilinear forms m (mass), k
(stiffness) and l
s
(right hand side).
m(u, v) =
f
u v d,
k(u, v) =
s
1
f
R
ijkl
kl
(u)
ij
(v) d,
l
s
(
f, v) =
{z=0}
f v d.
(9)
1.3 Hydroelastic system of equations
Bilinear forms l
s
and l
1
f
are adjoint of each other. l
1
f
is then linked to operator
t
R, and l
s
will be
associated to operator R.
l
s
and l
1
f
will also play the crucial role in the information transfert between incompatible uid and
structure meshes.
In the end, we dene
1
2
t
A
t
L
0
t
L
1
. (11)
Hence the following relation between the structure deection and the displacement potential:
= H
t
R(X). (12)
The pressure is linearised as p =
2
t
. The hydroelastic is then written as the following system
2
X
t
2
+KX = R
t
2
= H
t
R(X)
(13)
2 TIME DISCRETISATION
The hydroelastic system (13) is discretised in time using a second order central nite difference
scheme, with a constant time step t.
We note as an exponent the time step number n (with t = n t). For example, X
n1
is the
structure displacement at time t = (n 1)t.
M
X
n+1
2X
n
+X
n1
t
2
+K
X
n+1
+X
n1
2
= R
n+1
2
n
+
n1
t
2
+O(t
2
),
n+1
= H
t
R(X
n+1
).
(14)
At each time step, we are searching for X
n+1
and
n+1
; the vectors X
n1
, X
n
,
n1
and
n
begin
known and written on the right hand side.
n+1
= H
t
R(X
n+1
),
(M+
1
2
t
2
K)X
n+1
+R
n+1
= G
n
,
(15)
where G
n
= M(2X
n
X
n1
) +R(2
n
n1
)
1
2
t
2
KX
n1
. We dene the structure operator
S =
M+
1
2
t
2
K
1
. (16)
The discretised hydroelastic system (14) is rewritten as
S
1
X
n+1
+R
n+1
= G
n
,
n+1
= H
t
R(X
n+1
).
(17)
This system of two equations can be reduced to one single equation, by substituting the second relation
into the rst one.
S
1
X
n+1
+FX
n+1
G
n
= 0, (18)
where the uid operator F was dened as
F = R H
t
R. (19)
3 SOLVING THE HYDROELASTIC PROBLEM
The nonlinear hydroelastic equation (18) is considered under the from
N(X
n+1
) = 0. (20)
Equation (20) can be seen as a multidimensional rootnding problem. Using the Newton method to
solve this equation leads to the iterative process
X
n+1
p+1
= X
n+1
p
J
N
(X
n+1
p
)
1
N(X
n+1
p
), (21)
where J
N
(X
n+1
p
) is the Jacobian of operator N at point (X
n+1
p
).
The main issue for the Newton resolution method is the computation of the Jacobian J
N
of the
considered operator N.
N is dened as the sum of three operators: X S
1
(X), X F(X) and the constant operator
X G. Since the derivation operation is linear, the Jacobian of the sum is the sum of the Jacobians.
First, the operator X G is constant, and thus its Jacobian is zero.
In a second time, we need to compute the Jacobian of the structure operator. A linear elastic
structure is considered. In this precise case, the operator S
1
dened in equation (16) is linear. And
then, its Jacobian is S
1
itself.
J
S
1 = S
1
. (22)
In the end, the Jacobian of the uid operator F must be computed too. F is dened in equation
(19). The Jacobian of this composed function at point X is
J
F
(X) = J
R
(H
t
R(X)) J
H
(
t
R(X)) Jt
R
(X) (23)
Hence, the aim is to compute the Jacobians of R,
t
R and H. R and
t
R are linear; their Jacobians are
R and
t
R themselves J
R
= R and Jt
R
=
t
R. The crucial issue is then to compute the derivative of
operator H dened in equation (11).
The main difculty is that His not derivable everywhere, due to the constraints in the minimisation
process.
Let us suppose that the optimisation process in the operator H is unconstrained. Then operator H
would be the afne function L
1
A
1
L
1
+A
1
L
0
. The Jacobian of this function is A
1
.
In our case, the constrained optimisation is equivalent to an unconstrained minimisation inside the
cone of constraints C.Of course, the C is unknown, and has to be determined as a part of the solution.
Assuming the cone known, then the Jacobian of H could be computed, and would be the inverse
of the restriction of A to the cone C, i.e. J
H
= (A
C
)
1
.
For every structure deformation X
n+1
p
, it is possible to compute the displacement potential, and
to deduce the cone of constraints C
p
(which is where the displacement potential is not zero).
As a conclusion, the Jacobian of operator N is
J
N
= S
1
+R (A
C
)
1
t
R. (24)
where the term R (A
C
)
1
t
R can be considered as an added mass matrix for the uid structure
interface C.
By dening the Jacobian in this way, we skip the problem of derivating along the constraints.
4 VALIDATION CASE
The Wagner problem for an elastic cone has been solved in [Sco04]. The cone has a radius R =
0.128 m, and is made of aluminium (E = 1.2 10
2
N/m
2
, = 0.3, = 2700 kg/m
3
). The thickness
of the plate is 1.5 mm, and it is clamped in the center and on the boundaries. The dead rise angle is
6
.
Comparison is made with an uncoupled simulation (comparing displacement and von Mises stresses
at the center of the plate).
CONCLUSIONS
A threedimensionnal hydroelastic impact model has been described. Numerical resolution has been
carefully studied, and the process has been validated with respect to semi analytical solutions. Com
parisons with respect to the experiments are now needed to validate the theoretical model.
4
2
0
2
4
6
8
10
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008
d
i
s
p
l
a
c
e
m
e
n
t
(
m
m
)
time (sec)
Coupled
Uncoupled
Figure 3: Displacement at the center of the plate, for both coupled and uncoupled cases.
0
50
100
150
200
250
300
350
400
450
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008
S
t
r
e
s
s
(
M
P
a
)
time (sec)
Coupled
Uncoupled
Figure 4: Von Mises stress at the center of the plate, for both coupled and uncoupled cases.
References
[Aub00] D. Aubry. Mcanique des milieux continus. cole Centrale Paris, 2000.
[GKMS05] T. Gazzola, A.A. Korobkin, S. Malenica, and Y.M. Scolan. Threedimensional Wagner
problem using variational inequalities. In proceedings of the International Workshop on
Water Waves and Floating bodies, volume 20, 2005.
[Kor82] A.A. Korobkin. Formulation of penetration problem as a variational inequality. Din.
Sploshnoi Sredy, 58:7379, 1982.
[Sco04] Y.M. Scolan. Hydroelastic behaviour of a conical shell impacting on a quiescentfree
surface of an incompressible liquid. Journal of Sound and Vibration, 277, Issues 1
2:163203, 2004.
[Wag32] H. Wagner. ber Stound Gleitvorgnge an der Oberche von Flssigkieten. Zeitschrift
fr Angewandte Mathematik und Mechanik, 12:192215, 1932.
SOME ASPECTS OF 3D LINEAR HYDROELASTIC MODELS OF
SPRINGING
Malenica
S.
1
, Tuitman J.T.
2
, Bigot F.
1
& Sireta F.X.
1
(1) BUREAU VERITAS  DR, Paris, France (sime.malenica@bureauveritas.com)
(2) Delft University of Technology, Netherlands.
ABSTRACT
The paper deals with the modeling of the linear wave induced ship vibrations called springing.
The presented method is based on the full coupling of 3DFEM structural model and 3DBEM
hydrodynamic model. Only linear springing in frequency domain is considered. The so called
modal approach is used, which means that the total structural response is presented as a series of
the dry structural modes precalculated by the 3DFEM structural code. The coupling with 3DBEM
hydrodynamic model is performed by dening the additional boundary value problems for radiation
potentials associated with those structural modes. Finally the dynamic modal equation is solved
which makes the strains and stresses RAOs available in any part of the ship structure.
The total structural response is decomposed into quasi static and dynamic part, in order to clearly
evaluate the inuence of springing on the overall ship structural response. The decomposition into
the quasi static and dynamic part also allows superimposing the springing response to the existing
quasi static fatigue calculation methods and ensures the proper convergence of the results.
1 Introduction
Springing is usually dened as the entertained global ship structural vibrations induced by water
waves. Springing is a resonant phenomenon in contrast to the whipping which is the transient ship
vibrational response induced by impulsive loading (slamming, green water, underwater explosion,
...). A typical springing and whipping responses are shown in Figure 1, where we can clearly
observe the fundamental dierence between the two phenomena.
Figure 1: Typical springing (left) and whipping (right) ship structural response. Top  total signal, bottom
 ltered signal.
The springing type of ship hydroelastic structural response is usually not considered in the design
process of the conventional seagoing ships. The main reason is the signicant gap between the
ship wet natural frequencies and the wave frequencies encountered by the ship in the common
sea states. This does not means that the conventional ships are not experiencing the springing
during their service, but springing amplitudes are believed to be quite small. However, this is not
necessarily true and the evidences of the important nonlinearly induced springing response were
recently reported for some bulk carriers (e.g. [7]). Non linear hydroelastic springing modeling is
extremely dicult and, it is fair to say that, no ecient numerical model exists up to now. This
is mainly due to the complicated non linear hydrodynamics which, in the case of ship advancing
in waves with a forward speed, is still not solved properly and that even in the linear case.
In the present context, it is also important to recall the case of the springing response of the
tendons of the Tension Leg Platforms (TLP), e.g. [5], where the non linear wave excitation was
more or less successfully explained by the so called second order diraction theories, which were
numerically possible thanks to the fact that the TLP is stationary oating body (no forward speed).
In principle, the same kind of methods should be applied in the case of the ships but, as already
explained, this seems to be beyond the state of the art of the nowadays numerical methods.
Figure 2: Ultra large container ship and its typical midship section.
The springing type of ship structural response was reactualized recently, in the context of the
Ultra Large Container Ships (Fig. 2). Indeed, due to their huge dimensions (length close to 400
m) which reduce the structural natural frequencies and particular operating conditions (speed up
to 27 knots) which increase the excitation frequencies, a linearly induced springing becomes pos
sible. On the other hand, due to their complex structure with open midship section, these ships
can hardly be modeled by the equivalent beam models, which was the usual practice for spring
ing/whipping calculations, and complete 3DFEM model is necessary. All these facts represent the
main motivation for the present study.
2 Linear hydroelastic model in frequency domain
The general methodology for hydroelastic seakeeping model is rather well known and the rst
developments can be attributed to Bishop & Price [1]. In their work they used the simplied, so
called Timoshenko beam model for structural modeling and strip theory for seakeeping part. Since
then several more or less sophisticated models were proposed (e.g. [2, 8, 9, 10]).
Below we briey recall the basic principles of the model used in this study. The 3DBEM model for
the seakeeping is coupled to a 3DFEM model of the ship structure. A more detailed description
of the applied 3DBEM model can be found in [3] and [6].
In contrast to the well known rigid body seakeeping model, the hydroelastic model basically extends
the motion representation with the additional modes of motion/deformation chosen as a series of
the dry structural natural modes. We write:
H(x, y, z, t) =
N
i=1
i
(t)h
i
(x, y, z) =
N
i=1
i
(t)[h
i
x
(x, y, z)i +h
i
y
(x, y, z)j +h
i
z
(x, y, z)k] (1)
where h
i
(x, y, z) denotes the general motion/deformation mode which can be either rigid or elastic.
The above decomposition leads to the additional radiation boundary value problems (BVP) for
elastic modes, with the following change in the body boundary condition:
Rj
n
= h
j
n (2)
After solving the dierent BVPs the resulting pressure is calculated using Bernoullis equation
and integrated over the wetted surface in order to obtain the corresponding forces, so that the
following coupled dynamic equation can be written:
_
2
e
([ m ] + [ A ]) i
e
[ B ]+ [ k ] + [ C ]
_
{} = {F
DI
} (3)
where:
[ m ]  modal genuine mass
[ k ]  modal structural stiness
[ A ]  hydrodynamic added mass
[ B ]  hydrodynamic damping
[ C ]  hydrostatic stiness
{}  modal amplitudes
{F
DI
}  modal excitation
The solution of the above equation gives the motion amplitudes and phase angles
i
and the
problem is formally solved. Note that the motion equation includes both 6 rigid body modes and
a certain number of elastic modes.
Several technical diculties need to be solved before arriving to the above motion equation (3).
Certainly the most dicult one is the solution of the corresponding hydrodynamic BVP. In this
paper we do not enter into the detailed description of the methods used to solve the seakeeping
problem at forward speed and we just mention that these diculties remain the same, both for
the rigid and elastic body. It is fair to say that the numerical methods which are used to solve
the seakeeping problem nowadays, are not fully ready yet for a general combination of speed,
heading and frequency. However, most of the methods have approximate solutions to account
for the forward velocity. The method used in this paper is the so called encounter frequency
approximation which was reasonably well validated for rigid body case.
The second technical diculty is related to the evaluation of the hydrostatic restoring matrix
which, in the most general case, takes quite complicated form:
C
ij
= C
H
ij
+C
m
ij
(4)
C
H
ij
= g
_ _
S
B
{(Z h
j
+h
j
z
)h
i
n +Z[(h
i
)h
j
(h
j
)h
i
]n}dS (5)
C
m
ij
= g
___
V
(h
j
)h
i
z
dm (6)
As we can see from these expressions, the calculation of the restoring matrix requires the knowledge
of the dierent gradients of the modal shape functions which are not trivial to evaluate using the
3DFEM structural models. However, some simplications of the above expressions seem to be
possible and they are under investigation.
The third important technical diculty concerns the application of the body boundary condition
(2) for the general mode of motion. In the next section we briey explain how this was done in
the present work.
2.1 Body boundary condition
In the case of the simplied beam structural model, the transfer of the modal displacements of the
beam onto a hydrodynamic mesh can be done relatively easy, but in the case of 3DFEM structural
model this transfer requires very careful attention. Indeed, the modal displacements are known on
the structural 3DFEM mesh which is usually completely dierent from the hydrodynamic mesh,
so that a special interpolation procedure is necessary. In the present work, the following procedure
is adopted.
For each hydrodynamic point (panel center) the following steps are performed, these steps are
illustrated in gure 3:
yes
h(p) = h(s
1
)
p project on elements
s
1
s
3
s
2
p
d
no
s
1
s
3
s
2
yes
Select closets
h(p) =
f
intp
(h(s
1
), h(s
2
))
p
p project on lines
no
p
s
4
s
5
s
6
select closets point
h(p) = h(s
5
)
no
p
s
1
s
3
s
2
yes
Select closets
h(p) =
f
intp
(h(s
1
), h(s
2
), h(s
3
))
d <
Figure 3: Mapping procedure
1. Looking for the 3 closest structural points. In the case that one of the 3 points is within
small enough distance () from the considered hydrodynamic point, we retain that structural
point for interpolation.
2. The list of the structural nite elements containing at least one of the above dened points
is created.
3. The hydro point is projected onto the surfaces created by the retained structural elements. If
the projection falls inside the element, the corresponding distance is calculated. The element
with smallest distance from the hydro point is retained for interpolation.
4. In the case when the hydro point do not project on any element from the list, the projection
on the sides of the considered elements is performed. If the projection falls onto the side,
the corresponding distance is calculated. The side with the smallest distance from the hydro
point is retained for interpolation.
5. In the case when there is neither point at distance, nor element nor the element side on which
point project, the structural point closest to the hydro point is retained for interpolation.
6. The interpolation using the shape functions of the retained nite element, of the structural
displacements is performed on the projection of the hydro point and the calculated displace
ments are associated to the hydro point.
This procedure was veried on several ship types and showed to be very ecient. In Figure 4 we
present one example for the rst torsional mode of typical ULCS.
Figure 4: First torsional mode of ULCS and the transfer of modal displacements from structural to
hydrodynamic model.
2.2 Calculation of the stresses in frequency domain
The solution of the hydroelastic motion equation (3) includes the linear springing response auto
matically. Indeed, once this motion equation solved, all necessary quantities (motions, accelera
tions, stresses, ...) can be easily calculated because the modal decomposition remains valid for
any particular quantity. This means that, for example, we can write for the stress distribution
(x, y, z, ):
(x, y, z, ) =
N
i=1
i
()
i
(x, y, z) (7)
where
i
(x, y, z) represents the spatial distribution of the stresses corresponding to each mode of
motion/deformation and
i
() are the modal amplitudes i.e. solution of the motion equation (3).
2.3 Separation of the quasi static and dynamic contributions
In the present context, it is important to note that the rigid body modes do not contribute to any
stress, and that the above expression (7) is slowly convergent in general. This is especially true for
the structural details which are aected by the local structural eects. A good example of such
detail is a hatch corner of the container ship. The longitudinal response is well described by the rst
few longitudinal exible modes. However, in order to obtain any contribution of the transversal
(side shell) loading, one needs to include natural mode shapes which describe the deformation in
transversal direction. The 3DFEM methods calculate the structural natural frequencies of the
ship from the lowest natural frequency. By using a proper 3DFEM model of ship it is relatively
easy to obtain the rst few torsional and longitudinal modes. However, after these lowest global
structural modes, the numerous local modes with similar natural frequencies will quickly pollute
the solution and it will be almost impossible to obtain the subset of modes necessary for accurate
representation of the transversal stresses of the hatch corner. Fortunately, the global structural
dynamic response is well described by the rst few lowest modes. In order to obtain the converged
stress distribution, the total structural response should be properly separated into the quasi static
and dynamic parts. The stresses due to the quasi static ship response will be calculated, after
separation, by the so called direct method which calculates the structural response after the rigid
body hydrodynamic pressure is transferred on the wetted nite elements (see section 2.4). It is
also important to note that the decomposition into the quasi static and dynamic response is made
with respect to the structural response, which means that the quasi static structural response does
also includes the dynamic rigid body response and the dynamic response includes the dynamic
structural response only.
In the present approach the decomposition of the dierent parts of the response is done by rst
schematically rewriting the motion equation (3) in the following form:
__
[ RR ] [ RE ]
[ ER ] [ EE ]
_
+
_
[ 0 ] [ 0 ]
[ 0 ] [ k ]
___
E
_
=
_
F
R
F
E
_
(8)
where R stands for the rigid body parts, E for the elastic ones and k is the structural modal
stiness matrix.
At the same time we separate the total response amplitudes into the quasi static and dynamic
parts:
R
=
R
0
+
R
d
,
E
=
E
0
+
E
d
(9)
The quasi static part of the responses is dened by the following equations:
[ RR ]{
R
0
} = {F
R
} (10)
[ k ]{
E
0
} = {F
E
} [ ER ]{
R
0
} (11)
After inserting (9), (10) and (11) into (8), the following linear system of equations for dynamic
parts is obtained:
__
[ RR ] [ RE ]
[ ER ] [ EE ]
_
+
_
[ 0 ] [ 0 ]
[ 0 ] [ k ]
___
R
d
E
d
_
=
_
[ RE ]
R
0
[ EE ]
E
0
_
(12)
As mentioned before, the above procedure was adopted in order to be able to keep the classical
direct approach for the quasi static part [4] and to clearly identify the dynamic part as a correction
of the quasi static one. Anyhow, the proposed decomposition completely removes the convergence
problems discussed before.
2.4 Practical evaluation of the quasi static contribution
The direct approach is used to obtain the converged quasi static response. The hydrodynamic
pressures are applied to the structural model and the response is calculated using FEM. The
direct approach described in this section ensures perfect balance between the hydrodynamic and
structural models.
The nal loading of the structural model is composed of two parts:
1. Inertia loads
2. External pressure loads
Inertia loads can be included straightforwardly by associating the acceleration vector to each nite
element. Concerning the pressure loading, most of the methods nowadays use the dierent inter
polation schemes in order to transfer the total hydrodynamic pressure from hydro model (centroids
of the hydro panels) to the structural model (centroids or nodes of nite elements). Besides the
problems of interpolation, it is important to note that the motion amplitudes, which are present in
the denition of the total pressure, were calculated after integration over the hydrodynamic mesh.
For that reason it is impossible to obtain a completely balanced structural model. Indeed, the
FEM model has its own integration procedure which is usually dierent.
In order to obtain the perfect equilibrium of the structural model we use two main ideas [4]:
Recalculation of the pressure in structural points, instead of interpolation
Separate transfer of pressure components, and calculation of hydrodynamic coecients (added
mass, damping, hydrostatics & excitation) by integration over the structural mesh
The recalculation of the hydrodynamic pressure on the structural points, is possible thanks to
the particularities of the BIE method which gives the continuous representation of the potential
through the whole uid domain Z < 0. In this way the communication between the hydrodynamic
and structural codes is extremely simplied. Indeed, it is enough for the structural code to give the
coordinates of the points where the potential is required and the hydrodynamic code just evaluates
the corresponding potential by:
(x
s
) =
_ _
S
H
B
(x
h
)G(x
h
; x
s
)dS (13)
where x
s
= (x
s
, y
s
, z
s
) denotes the structural point and x
h
= (x
h
, y
h
, z
h
) the hydrodynamic point.
In the case of linear seakeeping without forward speed, this operation is sucient because the
pressure is directly proportional to the velocity potential and, within the source formulation, the
potential is continuous across the body wetted surface. This is very important point because, due
to the dierences in the hydrodynamic and structural mesh, the structural points might fall inside
the hydrodynamic meshes.
In the case of the body advancing with forward speed in waves, the situation is slightly more
complicated because the hydrodynamic pressure is no more directly proportional to the potential
but also to its gradient. The problem with the gradient of the potential, in the context of source
formulation of BIE, is that the gradient is discontinuous across the hydrodynamic mesh, so that the
recalculation of the pressure on structural points will likely lead to discontinuities in the pressure
distribution. This means that a preprocessing of the structural points is necessary i.e. these points
needs to be articially moved to the exterior of the hydrodynamic model. The procedure similar to
the one presented in Figure 3 is used to obtain these projection points. Strictly speaking, structural
points do not change their position but the pressure associated to them is calculated in the points
which are slightly moved o their original position. The error introduced by these manipulations
remains negligible within the limits of the linear theory considered here.
Anyhow, once each pressure component has been transfered onto the structural mesh, the new
hydrodynamic coecients are calculated by integration over the structural mesh:
F
DI
S
i
= i
_ _
S
S
B
(
S
I
+
S
D
)n
i
dS (14)
2
A
S
ij
+iB
S
ij
=
2
_ _
S
S
B
S
Rj
n
i
dS (15)
where the superscript
S
indicates that the quantities are taken on the structural mesh.
A similar procedure is applied for the variation of the hydrostatic pressure in order to obtain the
restoring matrix and after that the new motion equation is written:
_
2
([ M ] + [ A ]
S
) i[ B ]
S
+ [ C ]
S
_
{
0
}
S
= {F
DI
}
S
(16)
Solution of this equation gives the rigid body body motions {
0
}
S
so that the total linear pressure
can be calculated. In summary the nal loading of the structural model will be composed of the
following 3 parts:
2
m
i
S
i0
 Inertial loading (to be applied on each nite element)
p
S
i
 Pressure loading (to be applied only on wetted nite elements)
m
i
g
S
k  Gravity term (to be applied on each nite element)
It is clear that the above structural loading will be in perfect equilibrium because this equilibrium
is implicitly imposed by the solution of the motion equation (16) in which all dierent coecients
were calculated by using directly the information from the structural FEM model. In that respect
and in order to ensure full consistence between the results of the direct approach as described in
this section and the quasi static part of the modal approach described in the previous sections, the
hydrodynamic coecients in equation (3) should also be obtained by integration over the structural
mesh as explained in this section.
Let us also mention that, in practice, the pressure part of the loading can be applied in terms of
the nodal forces instead of the pressure which proved to be more ecient.
2.5 Top down analysis of structural details
Springing hydroelastic analysis is usually performed using the relatively coarse structural mesh
and some additional manipulations are necessary in order to obtain the local stresses in the critical
structural details. Usually the so called top down analysis is applied. The top down analysis is well
developed for the classical quasi static structural responses. The principles are relatively simple
Figure 5: Typical structural details and their position on the ship structure.
and the calculations are performed in two steps:
1. Calculation of the global structural response on the coarse mesh
2. Application of the structural deformations of the coarse mesh on the ne mesh and calculation
of the local stresses
One typical situation is shown in Figure 5. As already mentioned before, in the case of the quasi
static structural analysis, the loading case for the structural FEM model is built by considering
the ship as a rigid body. After solving the rigid body seakeeping boundary value problems, the
resulting pressure is applied on the FEM mesh together with the inertia loads resulting from the
rigid body accelerations. Since, in the rigid body case, the hydrodynamic and structural parts of
the problem are independent they can be performed separately and the transfer, of the resulting
deformations of the coarse mesh, onto the ne mesh is relatively straightforward. In the case of
dynamic structural analysis, which is of main concern here, the additional loading cases for ne
meshes need to be created. These additional loading cases correspond to each structural natural
mode deformations. The resulting local stresses should be added to the quasi static ones, after
being previously multiplied by the associated dynamic modal amplitudes
E
d
.
3 Numerical results and discussions
First we present few results related to the convergence of the modal representations for the struc
tural response in terms of the global internal loads and the local structural stresses. Figure 6 shows
the RAOs for these quantities calculated using dierent number of mode shapes and the direct
approach. Note that only the static response is considered here as it is very dicult to calculate
the real dynamic response by the direct approach. As is clearly visible in gure 6 and true for most
of the cases, the convergence in terms of the global quantities, such as vertical bending moment
or sectional shear forces is much better, than the convergence of the local stresses. This result
appears to be in line with the intuition because the local stresses are likely to be inuenced, much
more, by the higher structural modes while the global quantities will normally lter these local
contributions in most of the cases. All this implies that, in order to ensure proper convergence, the
quasi static part of the structural response should be calculated using the classical direct approach
for rigid body hydrodynamics, and only the dynamic part should be calculated using the modal
approach. In this way, the number of the dry structural modes which are necessary for dynamic
analysis is reducing to a few global modes only and these modes can be calculated relatively easily
by any commercial 3DFEM software.
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
M
_
y
[
G
N
m
]
Omega_e [rad/s]
Nf=8
Nf=16
Nf=32
Nf=48
Nf=64
Direct
0
10
20
30
40
50
60
70
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
S
ig
m
a
[
M
P
a
]
Omega_e [rad/s]
Nf=8
Nf=16
Nf=32
Nf=48
Nf=56
Nf=64
Direct
Figure 6: Convergence of the modal representation for vertical bending moment (left) and local structural
stresses (right).
In Figure 7, the results for the local stresses in the hatch corner of the ULCS are presented
and decomposed into quasi static and dynamic parts, and in Figure 8 the local stresses for two
dierent ship speeds are illustrated. It is interesting to observe that the dynamic part of the
0
10
20
30
40
50
60
70
0 0.5 1 1.5 2 2.5
S
i
g
m
a
[
M
P
a
]
Omega_e [rad/s]
Quasi static
Dynamic
Total
Figure 7: Typical stress RAO decomposed into quasi static and dynamic parts.
response contributes to the stresses even for the frequencies well away from the structural natural
frequency. This might seem a bit strange, but one should remind that these are very exible
ships. The static deformation of the structure, due to the still water bending, between midship
and the fore perpendicular is of the order of 1m for the ULCS. The wave bending moment is of the
same order as the still water bending moment and will cause the same order of deections. This
deection will inuence the heave response and the total bending moment.
On the other hand, the inuence of ship speed on dynamic springing response is evident. Indeed
higher speed will induce the higher stresses because the encounter frequency will be increased so
that the corresponding wave lengths will increase too, and the modal excitation will become more
important.
4 Conclusions
We presented here an ecient linear model for springing calculations. The main advantage of
the model, as compared to more classical equivalent beam models, lies in the fact that a 3DFEM
structural model is used. This allows for an easy and fast access to the local stresses at any position
0
20
40
60
80
100
120
0 0.5 1 1.5 2 2.5 3
S
i
g
m
a
[
M
P
a
]
Omega_e [rad/s]
v=0.6Vmax
v=Vmax
Figure 8: Typical stress RAOs for dierent ship speeds.
of the ship. Furthermore, the response decomposition into the quasi static and dynamic parts is
proposed in order to ensure the proper convergence of the results.
Some preliminary results which are presented here, stress the importance of utilizing a hydroelastic
analysis for ULCSs even when springing is not explicitly considered.
The remaining work will include more thorough validation of the model by comparisons with model
tests and full scale measurements, as well as the proper integration of the springing responses into
the design procedures.
ACKNOWLEDGEMENTS
Authors acknowledge the support of the MARSTRUCT European project.
References
[1] Bishop, R.E.D. & Price, W.G., 1979. : Hydroelasticity of ships, CUP.
[2] Jensen J.J. & Wang Z., 1999. : Wave induced hydroelastic response of high speed monohull
displacement ship., 2nd. Int. Conf. on Hydroelasticity, Fukuoka, Japan.
[3] Malenica
S., Molin B., Remy F. & Senjanovi c I., 2003. : Hydroelastic response of a
barge to impulsive and non impulsive wave loads., 3rd. Int. Conf. on Hydroelasticity, Oxford.
[4] Malenica
S., Stumpf E., Sireta F.X. & Chen X.B., 2008. : Consistent hydrostructure
interface for evaluation of global structural responses in linear seakeeping., 26th OMAE Conf.
[5] Molin B., 2003. : Hydrodynamique O shore., Technip.
[6] Newman, J.N., 1994. : Wave eects on deformable bodies, Applied Ocean Research, Vol.
16, pp.4759.
[7] Perunovic J.V. & Jensen J.J., 2005. : Non linear springing excitation due to a bidirec
tional wave eld., Marine Structures 18, pp. 332358.
[8] Wu M.K. & Moan T., 1996. : Linear and nonlinear hydroelastic analysis of high speed
vessels., J. of Ship Res., Vol.40/2, pp.149163.
[9] Wu Y. & Price W.G., 1986. : A general form of the interface boundary condition of uid
structure interaction and its applications., CSSRC Rep. 86010.
[10] Xia J. & Wang Z., 1997. : Time domain hydroelasticity theory of ships responding to
waves., J. of Ship Res., Vol.41/4, pp.286300.
Advanced computations of mooring systems
C. Brun, D. Coache & F. Rezende
Research Department, Bureau Veritas
56, Place de lIris, 92400 Courbevoie, Paris La Defense
email: ariane.veristar@bureauveritas.com
ABSTRACT: The analysis of mooring systems is becoming more and more complex with
the emergence of large oating units like LNG terminals anchored in deepwater or in water
areas of relatively small depth. Depending on the site characteristics and environmental
conditions, the design of mooring systems needs more and more aides from advanced
computations to simulate the motion of oating units and the variation of mooring lines
tensions. Based on the classical quasidynamic formulation developed since more than 20
years in Bureau Veritas, new developments have been recently performed. They consist of
the accurate simulation of secondorder wave loading by involving full QTF, the interaction
between two vessels and the modelling of their connections of dierent types. Furthermore,
the optimisation of temporal simulations and the development of user friendly interface
make more ecient the practical applications. To illustrate the powerful features of our
model, an example of detailed analysis will be included in the paper.
1 INTRODUCTION
Floating units are used increasingly to drill and store the oshore oil and gas. Because
of the oshore environment diversity depending on location and time, the design of these
moored structures has to be studied intensively and in an accurate way.
Nowadays, oating oshore units are installed in areas of dierent water depths. Some
FPSO are anchored by more than 1500m of water depth and recent examples of Liqueed
Natural Gas (LNG) terminals anchored in shallow water depth (around 30m). Both,
deep and shallow water operations, present dierent particularities to the mooring system
design and the calculation methods shall be chosen accordingly.
In addition, sidebyside operations are starting to be routinely used for both oil and
gas ooading. Normally, limiting weather conditions are dened based on operational
experience. However, it is very important to access those values by means of numerical
computations. Moreover, the hydrodynamic and mechanical interactions of two bodies in
close proximity are not trivial and need to be evaluated carefully.
Furthermore, considering that more and more environmental data are available, exten
sive calculations may be now performed based on hindcast model results or measurements.
Taking into account the improvement of the computers eciency, computation time is still
a limiting parameter, especially for models based on a full coupling between mooring and
structure.
It is important in Bureau Veritas activities to be able to perform mooring analysis as
part of an independent verication of engineering designs. In this context, the ARIANE
mooring software is in continuous development. This paper deals with aspects of multi
body modelling and wave lowfrequency loads computations, that both constitute one of
the latest research topics. As an example of application, a multibody system during
ooading operations in sidebyside is studied and some numerical results are presented.
2 GENERAL METHODOLOGY
There are basically two types of analyses that are generally accepted in the common
practice:
Fully coupled analyses: the lines are modelled using the nite element method. The
motions of the vessel are calculated accounting for the contribution of the lines (drag
force, damping and stiness). The problem may be solved either in frequency domain
or time domain.
Weakly coupled analyses: For these analyses, two models are necessary. In the rst
model the lines are modelled as nonlinear springs without mass. The drag and
damping forces coming from the mooring lines shall be precalculated and applied
to the vessel as external data. The motions of the vessel are calculated either in
frequency or time domain and the tensions in the mooring lines are given by the
displacement at the fairlead point (quasidynamic analysis). In the second model
the dynamic behaviour of the mooring lines are calculated by nite element and the
displacements at the fairlead are prescribed.
Fully coupled analyses (lines + vessel) are very time consuming. Thats why its inter
esting to perform a weakly coupled analysis where the dynamics of the mooring lines are
evaluated only for the critical cases identied in the quasidynamic analysis.
3 MULTIBODY CALCULATIONS
More and more studies are dedicated to multibody congurations. Thus, the interactions
between multiple bodies has become a key issue for the analysis of complex mooring
systems or ooading operations, which main congurations are in tandem (two vessels
located one behind the other) and in side by side (two vessels parallel to each other).
Another multibody application could be turrets modelling. Indeed, instead of modelling
the turret by one single point where all lines are connected, the turret and the vessel can
be modelled as two dierent bodies. Thus, mooring lines between the seabed and the
turret can be correctly located around the turret and loads between the turret and the
vessel estimated.
The interactions between two bodies can be divided into two classes: mechanical (due
to the connection system) and hydrodynamic/aerodynamic (waves, wind and current). In
case the vessels are suciently far from each other (for example in most part of tandem
congurations), the hydrodynamic/aerodynamic interactions may be neglected. However,
when the vessels are close from each other, those interactions shall be taken into account
properly. In the present paper, only the aspects regarding the mechanical interactions
between two vessels in sidebyside conguration are considered.
3.1 Multi body specic elements
In order to simulate correctly multibody situations, some particular elements should be
added. In side by side congurations for example, hulls are protected by fenders.
The characteristic of a fender is highly nonlinear and the contact point between a fender
and a vessel is not constant during the simulation. One way to calculate fender loads is
to model a fender attached to one vessel and acting with another vessel. In this case, the
point of application of the fender loads on the second vessel is calculated at each time
step. In the same way as for a mooring line, the characteristic of the fender should also
be given as a curve representing horizontal tension in function of the eective distance,
being the eective distance equal to the diameter of the fender.
Figure 1: ARIANE fender module a) Fender display b) Fender modelling
3.2 Mooring stiness matrix
A main parameter in multibody simulations is the global stiness matrix which is com
posed of that of mooring lines linked to the sea bottom and that due to the connection
between two vessels. This matrix gives valuable information of the resonant modes of
the system, for a given position of the vessels, in particular at the equilibrium position of
the moored system. Additionally, one may be interested in investigating how this matrix
varies as the vessels positions change. An example of stiness matrix for a sidebyside
conguration is given in section 5.
4 WAVES LOWFREQUENCY LOADS
The wave lowfrequency loads are the secondorder loads that excite the vessel at a fre
quency equal to the dierence of two wave frequencies (
1
2
). As the resonant frequency
of the mooring systems is often very low ( < 0.05rad/s), the waves lowfrequency loads
represent one of the main sources of the system excitation. The accurate estimation of
those loads is crucial for the correct estimation of the lines tension.
In the common practice, the design of moored systems is made using the Newman
approximation, which is based on a very low resonance frequency of the system (
(
1
+
2
)/2) and small contribution of the second order wave elds. The advantage of
this formulation is that it only requires the computation of the diagonal terms of the QTF
(Quadratic Transfer Function) matrix and the time series reconstruction is made using
single sums only. This type of approximation is satisfactory for systems moored in deep
water but can underestimate the drift loads in shallow water or for sti mooring systems.
In these cases, in order to estimate correctly the second order loads, the full QTF matrix
shall be computed. The disadvantage of the FullQTF method concerns the time series
reconstruction by means of double sums which increases signicantly the computation
time.
Recently, a new formulation has been presented in [1] and [2]. Considering 1, the
QTF can be developed in an expansion depending on . The quadratic transfer function
T(
1
,
2
) is then composed of one component T
0
() depending on = (
1
+
2
)/2 and
another component T
1
() linearly proportional to = (
1
2
). This formulation,
hereafter called BV approximation, presents very good results for systems which resonance
frequencies are below 0.05 rad/s and for any water depth. The main advantage of this
formulation concerns the time series reconstruction of loads that can be done by means of
single sums in the same way as Newman approximation.
In gure 2 the comparison between the three methods of computation of waves second
order loads is made for a LNG vessel at a water depth of 15m in head waves condition
(PiersonMoscowitz, Hs = 3m, Tp = 10s). It can be noted that the results given by the
BV approximation match quite well with the ones given by the full QTF, while Newman
approximation dramatically underestimates the loads.
Figure 2: Comparison of the vessels surge motions according to dierent calculation
methods in shallow water
5 NUMERICAL EXAMPLE AND DISCUSSIONS
A multibody system has been chosen as case study, where one of the vessels is anchored
to the ground and the other is moored to the rst one in sidebyside conguration. The
behaviour of both vessels in such conguration is discussed and some numerical results
are presented.
5.1 Description of the mooring system
The system is composed of two bodies linked together in a sidebyside conguration. One
of the two vessels is moored to the seabed by means of a spread mooring system, while
the other is linked to the rst one using synthetic ropes (see gure 3). The anchoring lines
are composed by chains in catenary conguration at a water depth of 500m.
Figure 3: Representation of the mooring system used in the example
In addition to the lines between the two vessels, two fenders are modelled between the
vessels. Their characteristic is given in gure 4.
Figure 4: Representation of the fenders reaction force against the eective distance be
tween the vessels
5.2 Resonant periods of the system
Before performing the mooring analysis in presence of all the environmental loads, its
interesting to evaluate the resonant periods of the system and if/how those periods change
with respect to the position of the vessels. This analysis can bring valuable information to
the choice of the calculation method. It can be performed in both frequency domain, using
the stiness matrix calculated at dierent positions of the vessels, and in time domain,
using decay tests performed around dierent mean positions.
5.2.1 Stiness matrix and resonant modes
The stiness matrix of a multibody system is a nn matrix with n equal to the number
of modes times the number of bodies. The table 1 presents an example of stiness matrix
calculated at the equilibrium position (without loads) of the above described mooring
system.
surge b1 sway b1 yaw b1 surge b2 sway b2 yaw b2
surge b1 5.36E+02 9.44E02 1.37E+04 4.44E+02 9.44E02 1.37E+04
sway b1 8.98E02 2.76E+03 3.67E+00 8.98E02 2.67E+03 3.67E+00
yaw b1 1.36E+04 3.67E+00 2.33E+07 1.20E+04 5.12E01 1.96E+07
surge b2 4.44E+02 9.44E02 1.20E+04 4.44E+02 9.44E02 1.20E+04
sway b2 8.98E02 2.67E+03 6.73E01 8.98E02 2.67E+03 6.73E01
yaw b2 1.36E+04 3.67E+00 1.96E+07 1.20E+04 5.12E01 2.03E+07
Table 1: Stiness matrix at the equilibrium position of the multibody system (given in
kN/m, kN, kN.m)
The terms in the matrix represent the force/moment on the body i given by the dis
placement of the body j. When i = j, then we have the mechanical interaction between
the bodies. In this example its interesting to notice that the term b1 b1 correspond
to the stiness given by both the mooring system between the vessels and the anchoring
system, while all the other terms are only contributed by the mooring between the vessels.
The stiness of the anchoring system only is given by the dierence between the total
stiness presented for body 1 (anchoring + mooring) and the stiness presented for body
2 (mooring) for a given mode.
With the stiness matrix, the Eigen modes of the system can be calculated in frequency
domain. For a system composed by two vessels, twelve eigen modes are calculated. Dier
ently from the single body analysis, its dicult to attribute one mode to a single motion.
Sometimes a mode is contributed by a combination of motions of the two vessels. The
table 2 presents the twelve resonant modes of the system. The rst six modes are given
by the hydrostatic stiness of the vessels and the last six modes are given by the mooring
stiness of the system. The DoFs presented are the motions that contribute the most to
the mode. It can be noticed that for very small displacements around the equilibrium
position, all the modes due to the mooring/anchoring system are outside the range of the
waves periods. However the outofphase modes are excited by dierence frequences ()
considerably high comparing to the inphase modes. In that case, the full QTF matrix is
required for the accurate calculation of the relative motions of the vessels.
Mode Frequency (rad/s) Period (s) DoF
1 0.63 9.95 pitch b 2
2 0.61 10.29 heave b 2
3 0.53 11.82 pitch b 1
4 0.491 12.792 heave b 1
5 0.456 13.788 roll b 1
6 0.388 16.209 roll b 2
7 0.177 35.503 outofphase sway/yaw
8 0.136 46.031 outofphase sway
9 0.082 76.265 outofphase surge
10 0.03 209.439 inphase yaw
11 0.015 407.671 inphase surge
12 0.014 460.071 inphase sway
Table 2: Resonant modes of the system at the equilibrium position
The modal analysis may be repeated for stiness matrices calculated at dierent po
sitions of the vessels in order to obtain the variation of the modes with respect to the
position.
5.2.1 Decay tests
Another way to evaluate the resonant periods of the system is by means of decay tests.
The decay tests are performed by applying an initial oset to one of the vessels and letting
the vessels move until they achieve the equilibrium. In the same way as for the frequency
domain, where we calculate the stiness matrices for dierent positions, we may also
perform decay tests around dierent equilibrium positions by applying constant loads to
one of the vessels all along the simulation. Doing so, dierent periods are found if the
mooring system stiness is nonlinear.
The gure 5 presents the sway decay tests for dierent mean positions of the body 2,
while the body 1 is kept xed all during the simulation. It can be noticed that the outof
phase sway period changes signicantly with the relative position of the two vessels. When
the relative motion is signicant, the period starts to decrease and may even come to range
of the waves periods in case of very large relative motions (order of 2m). This is due to
the nonlinear characteristics of the lines and fenders. This means that when large relative
motions are expected an analysis where both rst and secondorder hydrodynamics are
coupled properly in timedomain may be required.
2.5
2
1.5
1
0.5
0
0.5
1
1.5
2
2.5
0 100 200 300 400 500 600 700 800 900 1000
S
w
a
y
m
o
t
i
o
n

B
o
d
y
2
Time (s)
Mean offset= 0.0m / Tn=52s
Mean offset=0.5m / Tn=55s
Mean offset=1.0m / Tn=36s
Mean offset= 0.4m / Tn=49s
Mean offset= 1.0m / Tn=34s
Figure 5: Sway decay tests
5.2 Time domain simulations
In order to evaluate qualitatively the behaviour of the vessels in presence of waves, a
time domain analysis has been performed for a quartering sea state with H
S
= 3.0m and
T
p
= 10.0s.
In gure 6 a) it can be observed that the vessels move globally together oscillating
at the natural period of the anchoring system between the body 1 and the ground (in
phase motions). The mooring lines connecting the two vessels are not sensitive to the
inphase motions but to the outofphase motions. Figure 6 b) presents the outofphase
sway motions of the vessels for a time window of 1000s. Its interesting to observe that
those motions are given by a combination of rst and second order motions. The second
order motions however can have very low natural period (around 40s) which requires the
calculation of the lowfrequency wave loads for up to 0.20.
5
6
7
8
9
10
11
12
13
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
S
w
a
y
m
o
t
i
o
n
(
m
)
Time (s)
Sway motions  Body 1
Sway motions  Body 2
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
O
u
t

o
f

p
h
a
s
e
s
w
a
y
(
m
)
Time (s)
outofphase sway  lowfrequency only
outofphase sway  lowfrequency + wave frequency
Figure 6: a) Vessels sway motions b) Outofphase motions
6 CONCLUSIONS
The latest developments made in Bureau Veritas on mooring system analysis have been
described, which are mainly linked to shallow water applications and/or multibody con
gurations. Those developments have been implemented in the new version of the software
Ariane that has been used for the analysis in this paper.
For the evaluation of second order loads, dierent methods have been discussed. It
has been shown that the Newman approximation, commonly used, does not provide an
accurate estimation of the lowfrequency wave loads for shallow water and/or sti mooring
system. The new O() approximation may be a good alternative for mooring systems
with resonant frequencies up to 0.06rad/s at any water depth, presenting as advantage
the same computation time as Newman approximation. For resonant periods greater than
0.06rad/s, no choice is currently available other than the full QTF computation.
For the multibody congurations, the focus has been on the hydrodynamic and mehan
ical interactions between the vessels. As an example of application a sidebyside mooring
system has been presented. It has been shown the importance of performing a modal
analysis of the system at dierent positions prior to the denition of the methodology
to be employed, for example, for the evaluation of the second order loads (Newman,
O() approximation or full QTF). The timedomain simulations with environmental
loads demonstrate that in sidebyside conguration, the relative motions between the
vessels are due to a combination of rst and second order loads. The second order mo
tions however can have very low natural period (around 40s) which requires the calculation
of the lowfrequency wave loads for up to 0.20rad/s.
References
[1] Chen X.B & Duan W.Y. (2007) Formulation of lowfrequency QTF by O() ap
proximation, 22
nd
IWWWFB,Plitvice (Croatia).
[2] Chen X.B & Rezende F. (2008) Computations of lowfrequency wave loading, 23
rd
IWWWFB, Jeju (Korea).
[3] Brun C.,Rezende F.,Coache D., Mombaerts J. (2008) Impact of the use of Full
QTF on LNGC Moored in Shallow Water Studies, OTC.
[4] Chen X.B. (1994) Approximation on the Quadratic Transfer Function of Low
Frequency Loads, BOSS.
[5] Newman J.N. (1974) Second Order, slowlyvarying forces on vessels in irregular waves,
Intl. Symp. Dyn. Marine Vehicle & Struc. In Waves, Mech. Engng. Pub., London (UK)
[6] Chen X.B. (2006) Hydrodynamic analysis for oshore LNG terminals 2
nd
International
Workshop on Applied Oshore Hydrodynamics
[7] Naciri M. (2007) Time domain simulations of sidebyside moored vessels; Lessons
learnt from a benchmark test, 26
th
OMAE.
[8] Naciri M., Buchner B., Huijsmans R., Andrews J. (2004) Low Frequency Motions
of LNG Carriers Moored in Shallow Water, 23
rd
OMAE.
3DFEM3DBEM MODEL FOR SPRINGING AND WHIPPING ANALYSES OF
SHIPS
i
(t)h
i
(x, y, z) (1)
=
N
X
i=1
i
(t)[h
i
x
(x, y, z)i + h
i
y
(x, y, z)j + h
i
z
(x, y, z)k]
where h
i
(x, y, z) denotes the general motion/deformation
mode which can be either rigid or elastic.
The above decomposition leads to the additional radiation
boundary value problems (BVP) for elastic modes, with the
following change in the body boundary condition:
Rj
n
= h
j
n (2)
After solving the dierent BVPs the resulting pressure is
calculated using Bernoullis equation and integrated over
the wetted surface in order to obtain the corresponding
forces, so that the following coupled dynamic equation can
be written:
2
e
([ m ] + [ A ]) i
e
[ B ]+ [ k ] + [ C ]
{} = {F
DI
}
(3)
where:
[ m ]  modal genuine mass
[ k ]  modal structural stiness
[ A ]  hydrodynamic added mass
[ B ]  hydrodynamic damping
[ C ]  hydrostatic stiness
{}  modal amplitudes
{F
DI
}  modal excitation
The solution of the above equation gives the motion am
plitudes and phase angles
i
and the problem is formally
solved. Note that the motion equation includes both 6 rigid
body modes and a certain number of elastic modes.
Several technical diculties need to be solved before arriv
ing to the above motion equation (3). Certainly the most
dicult one is the solution of the corresponding hydrody
namic BVP. In this paper we do not enter into the detailed
description of the methods used to solve the seakeeping
problem at forward speed and we just mention that these
diculties remain the same, both for the rigid and elastic
body. It is fair to say that the numerical methods which
are used to solve the seakeeping problem nowadays, are not
fully ready yet for a general combination of speed, heading
and frequency. However, most of the methods have approx
imate solutions to account for the forward velocity. The
method used in this paper is the so called encounter fre
quency approximation which was reasonably well validated
for rigid body case.
The second technical diculty concerns the application of
the body boundary condition (2) for the general mode of
motion. In the next section we briey explain how this was
done in the present work.
2.2 Body boundary condition
In the case of the simplied beam structural model,
the transfer of the modal displacements of the beam onto
a hydrodynamic mesh can be done relatively easy, but in
the case of 3DFEM structural model this transfer requires
very careful attention. Indeed, the modal displacements
are known on the structural 3DFEM mesh which is usu
ally completely dierent from the hydrodynamic mesh, so
that the special interpolation procedure is necessary. In the
present work, the following procedure is adopted.
For each hydrodynamic point (panel center) the following
steps are performed, these steps are illustrated in gure 1:
1. Looking for the 3 closest structural points. In the
case that one of the 3 points is within small enough
distance () from the considered hydrodynamic point,
we retain that structural point for interpolation.
2. The list of the structural nite elements containing
at least one of the above dened points is created.
3. The hydro point is projected onto the surfaces created
by the retained structural elements. If the projection
falls inside the element, the corresponding distance is
calculated. The element with smallest distance from
the hydro point is retained for interpolation.
4. In the case when the hydro point do not project on
any element from the list, the projection on the sides
of the considered elements is performed. If the pro
jection falls onto the side, the corresponding distance
is calculated. The side with the smallest distance
from the hydro point is retained for interpolation.
5. In the case when there is neither point at dis
tance, nor element nor the element side on which
point project, the structural point closest to the hy
dro point is retained for interpolation.
6. The interpolation using the shape functions of the re
tained nite element, of the structural displacements
is performed on the projection of the hydro point and
the calculated displacements are associated to the hy
dro point.
yes
h(p) = h(s
1
)
p project on elements
s
1
s
3
s
2
p
d
no
s
1
s
3
s
2
yes
Select closets
h(p) =
f
intp
(h(s
1
), h(s
2
))
p
p project on lines
no
p
s
4
s
5
s
6
select closets point
h(p) = h(s
5
)
no
p
s
1
s
3
s
2
yes
Select closets
h(p) =
f
intp
(h(s
1
), h(s
2
), h(s
3
))
d <
Fig. 1: Mapping procedure
This procedure was veried on several ship types and
showed to be very ecient. In Figure 2 we present one
example for the rst torsional mode of typical ULCS.
2.3 Springing
The solution of the hydroelastic motion equation (3)
includes the linear springing response automatically. How
ever, there is no clear justication for considering springing
as a linear phenomena, at least as far as the wave excita
tion part is considered. Indeed, experience shows that many
type of ship suer from more or less important structural re
sponse around their rst structural natural frequencies even
if there is no important wave energy around these frequen
cies. The explication for existence of the structural response
can be explained only by introducing the nonlinearity into
the hydrodynamic model. The well known example is the
springing of the tendons of the TLP platforms where the hy
drodynamic excitation was successfully explained using the
weakly nonlinear second order theory. In principle similar
Fig. 2: First torsional mode of ULCS and the transfer
of modal displacements from structural to hydrodynamic
model.
kind of methods should also be applied in the case of the
ships, but unfortunately the state of the art in ship hydro
dynamics do not allow for that so that only approximate
and empirical methods can be used.
However, due to their huge dimensions and particular struc
tural properties, the case of ULCS is quite particular be
cause their rst structural natural frequencies become quite
low so that it can be excited in a linear sense too! This
means that the linearity excited springing might become
dominant which justify the linear model for its assessment.
The nonlinear hydroelastic model in time domain which
will be presented in section 3 does also include the linear
springing response. The Froude Krylov loading is the only
nonlinear wave excitation in this model. Some nonlinear
springing due to this Froude Krylov is observed in the re
sults of the time domain calculations, but to the authors
opinion a more sophisticated non linear model is necessary
to correctly calculate the nonlinear springing response.
2.4 Calculation of the stresses in frequency domain
Once the motion equation (3) solved, all necessary
quantities (motions, accelerations, stresses, ...) can now
be calculated. Indeed, the modal decomposition remains
valid for any quantity and, in particular, we can write for
the stress distribution (x, y, z, ):
(x, y, z, ) =
N
X
i=1
i
()
i
(x, y, z) (4)
where
i
(x, y, z) represents the spatial distribution
of the stresses corresponding to each mode of mo
tion/deformation. Equation similar to (4) remains valid in
the time domain model too, (10). The only change concerns
the replacement of by t.
2.5 Separation of the quasi static and dynamic con
tributions
In the present context, it is important to note that the
rigid body modes do not contribute to any stress, and that
the above expression (4) is slowly convergent in general.
This is especially true for the structural details which are
aected by the local structural eects. A good example of
such detail is a hatch corner of the container ship. The
longitudinal response is well described by the rst few lon
gitudinal exible modes. However, in order to obtain any
contribution of the transversal (side shell) loading, one need
to include natural mode shapes which describes the defor
mation in transversal direction. The 3DFEM methods cal
culates the structural natural frequencies of the ship from
the lowest natural frequency. By using a proper 3DFEM
model of ship it is relatively easy to obtain the rst few tor
sional and longitudinal modes. However, after these lowest
global structural modes, the numerous local modes with
the similar natural frequencies will quickly pollute the
solution and it will be almost impossible to obtain a sub
set of modes necessary for accurate representation of the
transversal stresses of the hatch corner. Fortunately, the
global structural dynamic response is well described by the
rst few lowest modes. In order to obtain the converged
stress distribution, the total structural response should be
properly separated into the quasi static and dynamic parts.
The stresses due to the quasi static ship response will be
calculated, after separation, by the so called direct method
which calculates the structural response after the rigid
body hydrodynamic pressure is transferred on the wetted
nite elements
8)
. It is also important to note that the de
composition into the quasi static and dynamic response is
made with respect to the structural response, which means
that the quasi static structural response does also includes
the dynamic rigid body response and the dynamic response
includes the dynamic structural response only.
In the present approach the decomposition of the dierent
parts of the response is done by rst schematically rewriting
the motion equation (3) in the following form:
[ RR ] [ RE ]
[ ER ] [ EE ]
[ 0 ] [ 0 ]
[ 0 ] [ k ]
F
R
F
E
(5)
where R stands for the rigid body parts, E for the elastic
ones and k is the structural modal stiness matrix.
At the same time we separate the total response amplitudes
into the quasi static and dynamic parts:
R
=
R
0
+
R
d
,
E
=
E
0
+
E
d
(6)
The quasi static part of the responses is dened by the
following equations:
[ RR ]{
R
0
} = {F
R
} (7)
[ k ]{
E
0
} = {F
E
} [ ER ]{
R
0
} (8)
After inserting (6), (7) and (8) into (5), the following linear
system of equations for dynamic parts is obtained:
[ RR ] [ RE ]
[ ER ] [ EE ]
[ 0 ] [ 0 ]
[ 0 ] [ k ]
R
d
E
d
[ RE ]
R
0
[ EE ]
E
0
(9)
As mentioned before, the above procedure was adopted in
order to be able to keep the classical direct approach for the
quasi static part
8)
and to clearly identify the dynamic part
as a correction of the quasi static one. Anyhow, the pro
posed decomposition completely removes the convergence
problems mentioned before.
2.6 Top down analysis of structural details
Springing hydroelastic analysis is usually performed us
ing the relatively coarse structural mesh and some addi
tional manipulations are necessary in order to obtain the
local stresses in the critical structural details. Usually the
so called top down analysis is applied. The top down analy
sis is well developed for the classical quasi static structural
responses. The principles are relatively simple and the cal
culations are performed in two steps:
1. Calculation of the global structural response on the
coarse mesh
2. Application of the structural deformations of the
coarse mesh on the ne mesh and calculation of the
local stresses
One typical situation is shown in Figure 3. As already
Fig. 3: Typical structural details and their position on the
ship structure.
mentioned before, in the case of the quasi static structural
analysis, the loading case for the structural FEM model is
built by considering the ship as a rigid body. After solving
0
10
20
30
40
50
60
70
0 0.5 1 1.5 2 2.5
S
i
g
m
a
[
M
P
a
]
Omega_e [rad/s]
Quasi static
Dynamic
Total
Fig. 4: Typical stress RAOs.
the rigid body seakeeping boundary value problems, the re
sulting pressure is applied on the FEM mesh together with
the inertia loads resulting from the rigid body accelerations.
Since, in the rigid body case, the hydrodynamic and struc
tural parts of the problem are independent they can be per
formed separately and the transfer, of the resulting defor
mations of the coarse mesh, onto the ne mesh is relatively
straightforward. In the case of dynamic structural analy
sis, which is of main concern here, the additional loading
cases for ne meshes need to be created. These additional
loading cases corresponds to each structural natural mode
deformations. The resulting local stresses should be added
to the quasi static ones, after being previously multiplied
by the associated dynamic modal amplitudes
E
d
. One typ
ical example of the local stresses in the hatch corner of the
classical ULCS are shown in Figure 4.
2.7 Hydroelasticity
It is interesting to observe that the dynamic part of
the response contributes to the stresses even for the fre
quencies well away from the natural frequency. This might
seems a bit strange, but one should remind that these are
very exible ships. The static deformation of the structure,
due to the still water bending, between midship and the
fore perpendicular is of the order of 1m for the ULCS. The
wave bending moment is of the same order as the still water
bending moment and will cause the same order of deec
tions. This deection will inuence the heave response and
the total bending moment.
All this suggests the importance to utilize a hydroelastic
analysis for ULCSs even when springing and whipping are
not explicitly considered.
3 NON LINEAR TIME DOMAIN SIMULA
TIONS
In contrast to the linear springing calculations which
can be performed in frequency domain, the nonlinear
springing and whipping simulations require the time do
main simulations. The procedure used in this paper is based
on the method proposed in Cummins
3)
and elaborated in
Ogilvie
10)
. This method uses the frequency domain hydro
dynamic solution and transfers it to the time domain using
the inverse Fourier transform. In this way the following
time domain motion equation is obtained:
([ m ]+[ A
]){
()}d = {F
DI
(t)} +{Q(t)}
where overdots denote the time derivatives and:
[ A
Z
0
B
ij
() cos t d (11)
After the impulse response functions K
ij
have been calcu
lated, the motion equation (10) is integrated in time using
the Runge Kutta 4th order scheme.
The main advantage of the time domain method lies in the
fact that we can introduce nonlinear components in the
excitation forces Q(t). There are many non linear eects
which are missing in the linear model and the state of the
art in the numerical seakeeping, do not allows for the con
sistent inclusion of all of them. In this paper we include the
so called Froude Krylov correction for the wave excitation
and the non linear slamming forces. Here below we briey
explain the way in which the Froude Krylov and slamming
loads are calculated.
3.1 Froude Krylov approximation
It is well known that the linear hydrodynamic model
evaluate the pressure up to z = 0 only so that we can end
up with the negative pressure close to the waterline, as well
as with zero pressure above the waterline and below the
wave crest which is unphysical. The so called Froude Krylov
approximation is usually employed to correct this pressure
distribution close to the waterline. The simplest Froude
Krylov variant is rather intuitive and consists in adding
the hydrostatic part of the pressure (gz) below the wave
crest (in the linear sense) and by putting zero total pressure
above the wave trough. Even if the methodology might ap
pear quite simple, the correct practical implementation in
the numerical codes might become very tricky. To the au
thors knowledge, the details of the numerical implementa
tion of the Froude Krylov method are usually not discussed
in the literature and that is why we are briey elaborating
on it here below. It is also important to note that some
other extrapolation procedures, like Wheeler stretching, for
the pressure close to the waterline are sometimes employed,
but we do not discuss them here.
According to the linear theory the linear hydrodynamic
pressure at any point below z = 0, is composed of two
parts:
pure hydrodynamic part associated with the velocity
potential and its derivatives
hydrostatic pressure variation due to ship motions
For the wave of unit amplitude we can write:
p(x, y, z, t) = {[i
e
U( x)]e
i
e
t
}
+{g
v
e
i
e
t
} (12)
where (x, y, z) is the linear unsteady potential, (x, y, z)
is the steady potential due to the forward speed in calm
water, and
v
is the vertical displacement of the considered
point on the ship surface.
In order to simplify the notations, it is useful to express the
pressure in meters of water height and write:
p(x, y, z, t) =
Ap(x, y, z, t)
g
= p
z
(13)
In this way the relative wave elevation (usually denoted by
R
) is equal to p at z = 0
R
(x, y, t) = p
0
(14)
Now we consider separately the case p
0
> 0 and p
0
< 0. In
Wave profile
Total pressure (>0)
z=0 z=0
Hydrostatic pressure (>0)
Hydrodynamic pressure (>0)
(>0)
Fig. 5: Modied pressure distribution for p > 0.
the case of p > 0 (Fig. 5) we can identify two dierent wet
ted regions with the following total pressure distribution:
P(z, t) =
(
p
0
z for 0 < z <
p
z
z for z < 0
(15)
The choice of the maximal wetted point is quite natural:
= p
0
(16)
In the case of p < 0 (Fig. 6) the situation is more com
plicated and we need to be careful. As far as the pressure
z=0
Wave profile
Hydrostatic pressure (>0)
Hydrodynamic pressure (<0)
(<0)
Total pressure (>0)
z=0
Fig. 6: Modied pressure distribution for p < 0.
distribution is concerned, the situation is similar to the pre
vious case, and the choice is quite natural (even if, strictly
speaking, not fully consistent):
P(z, t) =
(
0 for 0 > z >
p
z
z for z <
(17)
The main problem appears to be the choice of the minimal
wetted point . The usual choice for is the same as in
the previous case:
= p
0
(18)
The problem is that this choice introduce the jump in the
pressure distribution because:
p
0
= p
z
, for z = (19)
which means that the pressure at the point imediatly below
the wave prole will not be zero but:
P(, t) = p
= p
p
0
(20)
The reason for this is the inconsistency of the adopted ap
proach which do not include all nonlinear terms, so that all
necessary conditions can not be satised at the same time
and we need to sacrify something.
Let us now choose the value of as a solution of the fol
lowing equation:
= p
(21)
It is important to note that an iterative procedure should
be used to obtain the solution of this equation. It is easy
to see that this choice eliminates the jump in the pressure
distribution. The disadvantage is that we have to solve the
equation (21). The physical consequence of the above
choice is that the wave prole is changed from the assumed
one (
R
). However, this slight move of the free surface is
within the order of the approximations we adopted at the
begining. Strictly speaking both choices are not fully con
sistent but the second one has the advantage to allow for
more physical pressure distribution.
In summary, we can say that the Froude Krylov approxi
mation is not so clear and we need to be very careful in the
implementation and in the interpretation of the results. In
deed, as we have seen, several approximations are involved
in the above described procedure and, in spite of the quite
complicated considerations, it is impossible to remain fully
consistent. It is also important to note that several variants
of the Froude Krylov method are in use and the simplest
one consists in taking only the incident wave part of ele
vation in the denition of the relative wave elevation
R
.
This choice highly simplify the numerical implementation
and that is the reason why it is most often employed. In
that case we need to be extremely careful when interpreting
the results. This, lets call it incident wave Froude Krylov
approximation, is likely to be valid only in the case of rather
long waves where the diraction and radiation parts play
less important role in the overall contribution to relative
wave elevation
R
. However, in the case of ULCS which
have quite a low block coecient i.e. rather slender form
the incident wave Froude Krylov approximation can be rea
sonably justied for head or almost head wave conditions
for which the springing and especially whipping are most
likely to occur.
3.2 Course keeping
The nonlinear part of the force in equation (10) will
cause a mean loading at the ship. The ship will not hold
course and speed due to this mean force. This mean force
is not an error in the calculation. The mean force by the
Froude Krylov calculation is a part of the added resistance
in waves (RAW) and slamming forces cause also a loading
in the horizontal direction.
The drift of the ship in the time domain calculation should
be compensated in order to obtain a proper calculation.
Springs and dampers are often added to keep the ship at
the correct positions. An other solution is to model the
compete resistance, propulsion and maneuvering system of
the ship. Both method require quite some tuning to be able
to make a realistic simulation. To the authors opinion the
correct nonlinear surge, sway and yaw response will not
aect the structural loading much. Also one could ques
tion if using a spring, damper system would give the cor
rect nonlinear horizontal motions. In the present approach
Lagrange multipliers are added to equation (10). These
Lagrange multipliers impose the surge, sway and yaw mo
tions as is calculated using the frequency domain solution
(3). This gives a reasonable horizontal motions without any
tuning.
3.3 Slamming induced whipping
The time domain calculation will automatically include
the linear and non linear springing. In order to calculate
the slamming induced whipping response, the slamming
forces should be calculated during the seakeeping calcula
tion. This forces are added to the force vector {Q(t)} of
equation (10).
3.3.1 Slamming force
The hydrodynamic modeling of slamming is extremely
complex and still no fully satisfactory slamming model ex
ists. However, the 2D modeling of slamming is well mas
tered today and 2D models are usually employed to assess
the slamming loads on ships. Within the potential ow
approach, which is of concern here, several more or less
complicated 2D slamming models exist, starting from sim
ple vonKarman model and ending by the fully nonlinear
model. In between these two models there are several in
termediate ones such as Generalized Wagner Model (GWM)
11)
and Modied Logvinovich Model (MLM)
5)
which are
implemented in the presented method. We will not go into
too much details of these models and we refer to
7)
. The
advantage of the MLM method when compared to GWM
method lies in the lower requirements of CPU time. Most
often, at least when rigid body impact is concerned, two
methods give comparable results as shown in Figure 7 The
GWM method is used for the presented examples because
this method has a wider range of validity and the number
of slamming calculations remains limited. If many calcula
tions, for example a complete scatter diagram, have to be
preformed the MLM method might become more attrac
tive. The validity of the MLM calculation can be checked
0
1000
2000
3000
4000
5000
6000
0 0.2 0.4 0.6 0.8 1 1.2 1.4
F
_
z
[
K
N
]
t [s]
section_3
GWM
MLM
Fig. 7: Slamming forces on typical 2D section obtained by
GWM and MLM method.
using the GWM results for the some cases.
The use of 2D methods implicitly requires the employment
of the so called strip approach for 3D simulations. The usual
procedure consist in cutting the 3D mesh into a certain
number of 2D sections as shown in Figure 8. For each par
Fig. 8: 3D hydrodynamic mesh and corresponding 2D
sections.
ticular section, independent calculation of slamming loading
are performed and the overall slamming is obtained by sum
ming up dierent contributions. The implementation of this
procedure is dierent for, the most often used, beam struc
tural model and for 3DFEM model. Indeed, in the case of
the nonuniform Timoshenko beam model, only the overall
slamming force on the slamming sections is required, while
in the case of the 3DFEM model the transfer is slightly more
complicated. Here below we briey describe the method
which is likely to be very ecient for 3DFEM case.
The general expression for pressure excitation force pro
jected on the motion/deformation mode h
i
, can be written
in the following form:
F
i
=
ZZ
S
ph
i
ndS (22)
where S denotes the part of the ship surface on which the
pressure is applied and that regardless if the pressure is
calculated using the 2D or 3D methods. In the present case,
and since the slamming pressure is calculated on the 2D
sections, the above expression is rewritten in the following
form:
F
i
=
N
s
X
i=1
ZZ
S
s
p
s
h
i
s
n
s
dS
s
(23)
where S
s
denotes the surface associated with the corre
sponding slamming section, n
s
the associated normal vec
tor, p
s
the calculated 2D slamming pressure and h
i
s
the
projection of the 3D deformation mode h
i
onto 2D slam
ming section. The dierential element of the surface of the
slamming section can be rewritten as:
dS
s
= b
s
(l)dl
s
(24)
where b(l) is the normalized width of the section:
b
s
(l) =
dS
s
dl
s
(25)
which can be easily calculated from the 3D sectional mesh
extracted from 3D mesh (Figure 8). The same is true for
the associated normal vector, so that the following nal ex
pression, for the slamming induced modal excitation force,
can be written:
F
i
=
N
s
X
i=1
Z
L
s
p
s
h
i
s
n
s
b
s
dl
s
(26)
This expression can be easily implemented into the existing
rigid body slamming 2D module based on GWM or MLM
method.
We note here that, due to the lack of time, for the examples
presented in this paper a simplied method is used. The
modal excitation is calculated by multiplying the 2D slam
force with the area of the 3D section and the average mode
shape over the sections. This methods works well for the
rst longitudinal modes
3.3.2 Whipping
One typical whipping event from full scale measure
ments is shown in Figure 9 (taken from
1)
). The curve
represents the time history of the vertical bending moment
after slamming event. Two important points should be ob
served: whipping can increase the wave bending moment
signicantly and the transient whipping vibrations can last
for quite a long time because of low damping. This is very
important because it implies that whipping will be rele
vant both for extreme structural response as well as for
the fatigue loading. Anyway, once the above procedure im
plemented, it is possible to calculate the time history of
the stresses at any particular nite element. The decom
position into quasi static and dynamic contribution do not
seems to be straightforward. The most practical way is to
perform two separate simulations: one purely quasi static
and another with the structural dynamics included. The
above described procedure is applied on 11400 TEU ship
sailing with speed U = 10m/s in the irregular head waves
Fig. 9: Whipping from full scale measurements.
(Jonswap spectrum H
s
= 5m and T
p
= 16s). The typical
situation at one particular time instant is shown in Figure
10 and the typical stress time history around a slamming
event in Figure 11. Typical transitory whipping behavior
can be observed. It should be noted that only modal super
position is used for this gure and no separation into quasi
static and dynamic parts as only the longitudinal loading
will contribute to the stress for this element. In the present
case the local structural stress is presented but the exactly
same procedure can be applied for the global quantities as
well. In particular the vertical bending moment at a partic
ular ship section can be evaluated. Calculation of the ver
tical bending moment is usually required in order to have
quick idea of the inuence of whipping on the overall ship
design. In Figure 12 one typical simulation result for ver
tical bending moment is presented. This example is very
illustrative and shows that, even if the whipping usually
appears in sagging conditions, the vertical bending moment
in hogging conditions is also aected due to relatively low
structural damping. The stress in the structural detail is
Fig. 10: Stress distribution in the ship structure at one
particular time instant. The total stress includes the linear
and non linear quasi static, springing and whipping contri
butions at the same time.
calculated for 2000 second in the same sea state to obtain
a better quantication of the eect of whipping. The non
linear calculation is run with and without slamming using
the same wave train. Figure 13 shows the probability distri
bution of the stress in the structure. The rainow cycles and
the fatigue damage calculated using a standard SNcurve is
shown in gure 14. The slamming induced whipping clearly
increase the amplitude of the stress and reduces the fatigue
life of the structural detail.
20
25
30
35
40
45
50
55
60
65
70
75
415 420 425 430 435 440 445
S
i
g
m
a
[
M
P
a
]
Time [s]
With slamming
Without slamming
8
6
4
2
0
2
4
6
415 420 425 430 435 440 445
S
i
g
m
a
[
M
P
a
]
Time [s]
Fig. 11: Typical stress time history during the whipping
event (top  total, bottom  ltered).
500000
0
500000
1e+006
1.5e+006
2e+006
2.5e+006
3e+006
3.5e+006
4e+006
8160 8180 8200 8220 8240
V
e
r
t
i
c
a
l
b
e
n
d
i
n
g
m
o
m
e
n
t
a
t
f
r
a
m
e
8
8
[
K
N
m
]
Time [s]
Elastic  with slamming
Rigid  no slamming
Fig. 12: Typical time history of the vertical bending mo
ment during the whipping event.
4 CONCLUSION
The method for hydroelastic analysis using the 3DFEM
3DBEM coupling procedure is presented. Both frequency
and time domain approaches are discussed. The overall
problem is extremely dicult and only approximate solu
tion is possible up to now. In spite of numerous technical
diculties the present model seems to produce quite real
istic results. More detailed validations and calibrations are
necessary before employing the method in practice.
REFERENCES
1. Aalberts P.J. & Nieuwenhuijs M., 2006. : Full
scale wave and whipping induced hull girder loads.,
0.001
0.01
0.1
1
50 55 60 65 70 75 80 85
P
r
o
b
a
b
ilit
y
o
f
e
x
c
e
e
d
e
n
c
e
Stress [MPa]
With slam
No slam
Fig. 13: Probability of exceedence of stress in the struc
tural detail
1
10
100
1000
0 10 20 30 40 50 60 70 80
0
2e06
4e06
6e06
8e06
1e05
1.2e05
N
u
m
b
e
r
o
f
c
y
c
le
s
F
a
t
ig
u
e
d
a
m
a
g
e
p
e
r
h
o
u
r
Stress [MPa]
cycles with slam
cycles no slam
damage with slam
damage no slam
Fig. 14: Rainow count and fatigue damage of the struc
tural detail
4th. Int. Conf. on Hydroelasticity, Wuxi, China.
2. Bishop, R.E.D. & Price, W.G., 1979. : Hydroe
lasticity of ships, Cambridge University Press.
3. Cummins W.E., 1962. : The impulse response func
tion and the ship motions., Schisctechnik, Vol. 47,
pp.101109.
4. Jensen J.J. & Wang Z., 1999. : Wave induced
hydroelastic response of high speed monohull dis
placement ship., 2nd. Int. Conf. on Hydroelasticity,
Fukuoka, Japan.
5. Korobkin A.A., 2005. : Analytical models of water
impact., Euro Journal on Applied Mathematics.
6. Malenica
S., Molin B., Remy F. & Senjanovi c I.,
2003. : Hydroelastic response of a barge to impulsive
and non impulsive wave loads., 3rd. Int. Conf. on
Hydroelasticity, Oxford, UK.
7. Malenica
S. & Korobkin A.A., 2007. : Some as
pects of slamming simulations in seakeeping., 9th.
Int. Conf. on Numerical Ship Hydrodynamics, Ann
Arbor, USA.
8. Malenica
S., Stumpf E., Sireta F.X. & Chen
X.B., 2008. : Consistent hydrostructure interface
for evaluation of global structural responses in linear
seakeeping., 26th OMAE Conf.
9. Newman, J.N., 1994. : Wave eects on deformable
bodies, Applied Ocean Research, Vol. 16, pp.4759.
10. Ogilvie T.F., 1964. : Recent progress toward the
understanding and prediction of ship motions., 5th
Symposium on Naval Hydrodynamics.
11. Zhao R., Faltinsen O.M. & Aarsnes J.V., 1996.
: Water entry of arbitrary two dimensional section
with and without ow separation., 21st Symposium
on Naval Hydrodynamics.
12. Wu M.K. & Moan T., 1996. : Linear and nonlinear
hydroelastic analysis of high speed vessels., J. of Ship
Res., Vol.40/2, pp.149163.
13. Wu Y. & Price W.G., 1986. : A general form of
the interface boundary condition of uid structure in
teraction and its applications., CSSRC Rep. 86010.
14. Xia J. & Wang Z., 1997. : Time domain hydroelas
ticity theory of ships responding to waves., J. of Ship
Res., Vol.41/4, pp.286300.
Paper No. ISOPE2008TPC269 First authors last (family) name: Kim Total number of pages: 10
Fluidstructure interaction modeling, relating to membrane LNG ship cargo containment system
W S Kim
1
, B J Noh
1
, H Lee
2
, Z Mravak
3
, J de Lauzon
3
, J R Maguire
4
, D Radosavljevic
4
, S H Kwon
5
, J Y Chung
5
1
Shipbuilding Division, Hyundai Heavy Industries Co., Ltd., Ulsan, Korea
2
Technology, Research and Product development Group, American Bureau of Shipping, Houston, USA
3
Marine Division, Bureau Veritas, Paris, France
4
Technical Investigations, Lloyds Register EMEA, London, England
5
Department of Naval Architecture & Ocean Engineering, Pusan National University, Busan, Korea
ABSTRACT
Over the period 200607, a J DP (joint development project) for LNG
cargo containment system(CCS) was carried out by HHI, ABS, BV,
LR and PNU. The aim of the project was to develop best current
practice for the analysis of fluidstructure interaction (FSI) events, in
particular relating to sloshing impact loads in membrane type LNG
CCS, which could lead to the improved evaluation of the structural
safety of GTT Mark III type CCS.
KEY WORDS: Liquefied Natural Gas (LNG); Cargo Containment
System(CCS); FluidStructure Interaction (FSI); Wet Drop Test
INTRODUCTION
In recent years, the needs for natural gas (NG) have been quickly and
steadily increasing, due to the fast growth of world energy usage and
NGs environmentfriendly nature. The demand for building of LNG
(liquefied natural gas) carriers with bigger cargo capacity has been also
increasing to economically transport LNG. In cargo tanks of membrane
type LNG carriers, because there is no internal supporting structure
such as partial bulkheads, sloshing loads by LNG is one of the most
important factors for the safety of insulation system and supporting hull
structure. Therefore, to design a newsize LNG carrier safely, it is
needed evaluating the sloshing impact load and assessing the structural
response to that impact load. A dynamic structural analysis considering
fluidstructure interaction (FSI) under sloshing impact load is the most
essential for assessment of the structural safety of cargo containment
system (CCS) in a membrane type LNG carrier (Namet al., 2005 and
2006). Till now, however, a single method to predict the structural
response of CCS on absolute basis has not been validated yet.
In early 2006, Hyundai Heavy Industries (HHI) commissioned Pusan
National University (PNU) to carry out physical wet drop tests of Mark
III type LNG tank insulation specimens, to clarify the impact pressure
acting on the insulation membrane and the strains developed in the
insulation system. In parallel, HHI proposed a joint development
project (J DP) to major classification societies to carry out the numerical
simulation of the wet drop test by each own manner. The main purpose
of the J DP was to develop the best practice of analytical solution for
the evaluation of structural safety of CCS in membrane type LNG
carriers in consideration of FSI under sloshing impact loads. Three
classification societies of American Bureau of Shipping (ABS), Bureau
Veritas (BV) and Lloyds Register Asia (LR) participated in the J DP of
Wet Drop Test Simulation initiated by HHI.
This paper summarizes the simulation works carried out by four
participants of HHI, ABS, BV and LR, including wet drop tests carried
out by PNU.
WET DROP TEST
The tests were carried out at Slamming research laboratory in PNU.
Figure 1 illustrates the test facility. The test specimen is guided by four
linear motion (LM) sliders installed at main vertical columns.
Transverse and rotational motions of the specimen are restrained for
strict vertical drop maintaining the inclined (incident) angle. The test
rig is placed on a water basin of 3 meter depth and the maximum drop
height is 4.3 meter (Chung et al., 2006). The drop unit where the test
specimen is to be attached is shown in Fig. 2. The drop unit consisted
of steel housing for the test specimen, supporters and jigs connected to
LM guide sliders.
Fig. 1 Test rig for wet drop test
Total three types of specimens were tested; flat membrane (Flat),
corrugated membrane (Light) and corrugated membrane with added
weight (Heavy). The distinction of types comes mainly from shape of
the membrane and weight of the specimen. Test specimens of Flat and
Light are shown in Fig. 3. Both specimens consisted of plywood,
reinforced polyurethane foam (RPUF), rigid triplex, mastics and
stainlesssteel membrane. For Heavy specimen, additional weight of
500 Kg was welded inside of steel housing as shown in Fig. 4. Total
weight of the assembled specimen was about 1,000 Kg for Flat and
Light, and 1,500 Kg for Heavy.
Fig. 2 Configuration of drop unit
Housi ng
Fl at & Cor r ugat ed
membr ane
R  PUF & mast i c
Fig. 3 Drop specimens (Flat and Light)
Li ght c ondi t i on Heav y c ondi t i on
Fig. 4 Inside of housing for Light and Heavy
Three pressure sensors were located on the membrane surface along the
flow direction, which were denoted by P1, P2 and P3, as shown in Fig.
5. Uniaxial strain gauges of 5 mm length were placed on the mastic
and the plywood, whose locations were denoted by SM1, SM2, SP1,
SP2, , SP10. The sampling rate of the pressure and strain signal was
20 kHz. Motion of the dropped unit was monitored by a highspeed
camera, and dropping velocity was obtained by digitizing captured
images.
As given in Table 1, total 20 cases of the drop test (9 cases for Flat and
Light, and 2 cases for Heavy) were carried out. For each case, two tests
were performed to confirm the repeatability of measured data.
(a) Sensor layout
MARK III Corrugation membrane
340
3
4
0
940
9
4
0
9
4
0
120 120
Flow direction
Diameter = 6mm
Thickness = 2mm
(b) Pressure sensors
(c) Strain gauges
Fig. 5 Sensor layout and locations
Table 1. Test case matrix for wet drop
Incident Angle (degree) Drop Height (meter)
Item
Model
0 4 8 2 3 4
Flat
Light
Heavy
SIMULATION
In this section, the method how to simulate the wet drop tests
numerically by each participant is introduced in brief.
Methodology
HHI
MSC/DYTRAN was used for both of fluid and structure models
adopting General coupling. FSI was carried out by solving equations of
motion for the structure and the fluid simultaneously at each time step.
ABS
At the first stage, pressure distribution history on the drop specimen
Inner hull
Mastic
Strain gauge
(plywood)
Strain gauge
(mastic)
Pressure gauge
RPUF
Primary barrier (STS 304L)
was computed by using FLUENT under the assumption of the rigid
body. The computed impact pressure, which was subdivided into 4 by 4
patch loads, was used as the input loads on the structural model
coupled with the fluid element, to consider FSI effect in indirect way,
for the dynamic structure analysis (ABS, 2006). MSC/DYTRAN was
used for the dynamic structure analysis adopting Euler coupling
between the structure and fluid elements.
BV
A modal method was used to perform hydroelastic coupled analysis of
impact calculation. At the first stage, eigenmodes of the structure were
calculated using ABAQUS to be transferred to BVs inhouse program
for hydroelastic coupling, which was based on the Wagners impact
theory and variational inequality method (Gazzola et al., 2006 and
Malenica et al., 2006). Then for each time step of the simulation, BVs
inhouse solver iterated to find the mutually dependent hydrodynamic
load and structural response; convergence was achieved using
Newtons algorithm. Using this technique and programs the required
CPU time and data storage capacity necessary for the calculation of
hydroelastic problems were significantly reduced
LR
Three approaches were examined. The first one was FEAonly
approach by using ABAQUS. The fluid and the structure were entirely
modeled by finite elements and calculated simultaneously. The second
one was 1way coupled approach by linking STARCD to ABAQUS
via MpCCI. STARCD calculated pressure distribution on the drop
specimen surface at each time step under assumption of the rigid body,
and ABAQUS calculated dynamic structure response under the
transmitted pressure history. The third one was 2way coupled
approach by linking STARCD and ABAQUS each other via MpCCI.
STARCD calculated pressure distribution on the drop specimen
surface at certain time step under the assumption of the rigid body, and
ABAQUS calculated deformation of structural body for calculation of
STARCD at the next time step. Then, STARCD calculated pressure
distribution on the drop specimen surface again at the next time step
with the deformed rigid body, and this calculated pressure distribution
was transmitted to ABAQUS for calculation of structure response at the
next time step. This procedure of calculation and transmission was
repeated during the whole calculation period. Only two cases of drop
tests were selected to illustrate the effect of coupling methods, namely
Flat specimen cases dropped at 4 meter height with incident angles of 0
and 4.
Model
HHI
Figure 6 shows the drop structure model of Light specimen, partly cut
for visibility, with the membrane up and the housing down (a) and in
the drop position with the membrane down and the housing up (b). The
exact complex corrugation geometry was replaced by simplified
corrugation geometry. A contact was defined between the membrane
and the top plywood. Rigid connections were defined between the
secondary barrier and the primary insulation foam. Other bonding and
welding connections were modeled by having common nodes. A
contact was defined between the primary/secondary insulation foam
and the housing. All 16 hinge lugs, which were welded to the eight
corners on two sides of the steel housing, were modeled. Supporting
brackets and LM guide sliders were not directly modeled, but their
mass was divided into 16 lugs as concentrated mass. The restraining
effect of supporting brackets and LM guides was modeled by
restraining the motion in X and Y global directions at the hinge axis of
each lug. Figure 7 shows the fluid model. A special consideration for
air flow between membrane corrugation and water surface, as well as
water behavior in the impact area was taken account. Fluid domain
dimensions were taken 5 x 5 x 4 meters in breadth x width x height.
The lower 3 meters were filled with water as in the actual basin, and the
upper 1 meter was filled with air. To keep the possible minimum
number of elements for calculation efficiency, mesh size was gradually
increased (biased) from the center towards corners. 1 bar boundary
condition was applied at out 5 faces of air region. Outer boundaries of
water region were closed. Air was assumed to be an ideal gas with a
gas constant of 1.4 and density of 1.18 kg/m
3
. Water was modeled as
compressible fluid using a polynomial equation of state. Density was
taken as 1,000 kg/m
3
and bulk modulus was taken as 2.2 GPa. The total
number of fluid elements was about 370,000.
(a)
(b)
Fig. 6 Drop structure model of HHI
Pressure=1bar
Rigid wall
Fig. 7 Fluid model of HHI
ABS
Figure 8 shows a typical CFD model of FLUENT to calculate the
pressure time history of Light/Heavy specimen with 0 incident angle.
In consideration of symmetry, half or oneforth of the fluid and the
structure was modeled. The structure was a rigid body. Water was
modeled as incompressible fluid and air was modeled as compressible
ideal gas. Material properties of air and water were provided within
FLUENT material database. Grid sizes were 6.25 ~20 mm depending
on the type of specimens and incident angles. Figure 9 shows the FE
model. The top surface of the structure was contacting with the fluid,
which was modeled by isotropic elastic fluid element. Corrugation of
the membrane was not considered for the structure analysis.
Fig. 8 CFD model of ABS (corrugated membrane with 0 indent angle)
(a) Steel casing (b) Mastics
(c) RPUF & Plywood (d) Membrane
Fig. 9 FE model of ABS
BV
Figure 10 shows the FE model of Light/Heavy specimens. In
consideration of the structure and load symmetry, only one half of the
specimen was modeled and analyzed. The interaction of the analyzed
part with the test rig structure was defined as fix boundary conditions at
four specimen hinge points. The rig structure was assumed as rigid
during the impact phase. The mutual interaction between the different
structural parts could be characterized as tie and therefore the following
pairs were defined in FE model: mastic ropehousing, mastic rope
secondary plywood, secondary plywoodsecondary foam, primary
foamprimary plywood, primary membrane edgeshousing, vertical
housing platesRPUF vertical surfaces and primary membraneprimary
plywood surface. Different mesh size was tested to determine the
proper finite elements dimension which allowed enough stabile and
precise results and also reasonable computing time and required data
storage capacity. The utilized FE mesh size for the insulation structure
was about 30 x 30 mm and total number of elements in FE model was
about 15,000.
Fig. 10 FE model of BV
For the discretization and analysis of the fluid domain, the finite
boundary elements method was used. Therefore only the contact
surface of the fluid domain had to be meshed. To properly calculate
pressure distribution on the impact surface a very fine mesh should be
defined in the proximity of contact line during the whole duration of
impact. To successfully follow the contact line displacement over the
impact surface, extremely fine mesh should be defined over the whole
surface. This would require a significant number of elements, as well as
computing time and data storage capacity. To avoid these difficulties
an adaptive meshing technique could be applied. At each calculation
time step, a new mesh for the fluid domain was generated having the
smallest element size in the proximity of contact line. For the
illustration of applied adaptive meshing technique, the mesh of impact
surface for two instant t
1
and t
2
(with t
1
<t
2
) is given in Fig. 11. It could
be noted that the mesh refinement follows the evolution of the contact
line.
Fig. 11 Mesh on the impacted surface at instant t
1
and instant t
2
LR
Figure 12 shows the FE model, which was used for FEAonly
approach. Characteristics of the dropping unit model are similar to
those of ABS, except the corrugated membrane. Geometry of the
corrugated membrane was modeled precisely by using very fine mesh.
Water was modeled by elements with very small shear viscosity value
of 1.0E10 Ns/mm
2
. Air was not modeled. Figure 13 shows CFD model
of STARCD for 1way and 2way coupled approach. A pressure
boundary condition was applied to the upper most part of the domain.
All other boundaries were no slip walls. In CFD model, compressibility
of both water and air was taken into account.
Fig. 12 FE model of LR for FEAonly approach
Computational domain with Local mesh refinement around
moving mesh setup the box and corrugation surface
Fig. 13 CFD model of LR
Material Properties
All participants agreed to use exactly the same material data. Material
properties used in simulations are listed in Table 2.
Table 2. Material properties
Unit: MPa, ton/mm
3
Plywood RPUF Triplex Mastics Membrane Steel
En 8900 142 13133 2934 200000 206000
E
s
7500 142  2934 200000 206000
E
t
520 84 9100 2934 200000 206000
n
ns
0.17 0.24 0.3 0.3 0.27 0.3
nnt 0.17 0.18  0.3 0.27 0.3
nst 0.17 0.18  0.3 0.27 0.3
G
ns
196 12.2    
G
nt
196 12.2    
Gst 196 12.2    
Density 7.10E10 1.25E10 2.50E09 1.50E09 7.85E09 7.85E09
Model Ortho Ortho Iso Iso Iso Iso
Simulation Cases
HHI
All 20 test cases of wet drop given in Table 1 and the static load test
were simulated.
ABS
Four cases of following drop tests were simulated.
 Flat and Light specimens with 0 incident angle at 2 meter drop height
(Pressure only)
 Flat specimen with 4 incident angle at 2 meter drop height (Pressure
only)
 Heavy specimen with 0 incident angle at 2 meter drop height
BV
Twelve cases of following drop tests and the static load test were
simulated. The further development of hydrodynamic code which is in
progress needs to cover impact with 0 incident angle.
 Flat specimen with 4 and 8 incident angles at drop height equal to 2
,3 and 4 meters
 Light specimen with 4 and 8 incident angles at drop height equal to
2, 3 and 4 meters
LR
Nine cases of following drop tests and the static load test were
simulated.
 Flat specimen with 0 incident angle at 3 meter drop height
 Flat specimen with 0 incident angle at 4 meter drop height
(including 2way coupled approach)
 Flat specimen with 4 incident angle at 3 meter drop height
 Flat specimen with 4 incident angle at 4 meter drop height
(including 2way coupled approach)
 Light specimen with 0 and 4 incident angles at 3 and 4 meters drop
height
 Heavy specimen with 0 incident angle at 2 meter drop height
CUTUP OF SPECIMENS
After drop tests, internal damages and fabrication quality of all
specimens were checked by dismantling steel housing and slicing CCS
by 50 mm interval, as shown in Fig. 14. Some damages were found at
corrugated membranes as shown in Fig. 15. In case of Light specimen,
large corrugations were buckled due to pressure and small corrugations
were bent due to jet flow. In case of Heavy specimen, however, large
corrugations only were buckled and there was no damage in small
corrugations because the drop test with an incident angle was not
carried out. Some cracks on RPUF underneath mastics were found, as
shown in Fig. 16, in case of Flat specimen. The main cause of the crack
is assumed to be stress concentration under pressure loading and
repeated drop tests.
Figure 17 shows bonding status of the mastic to the plywood at the
middle part of three specimens. Due to poor craftsmanship, there were
a lot of cavities in the mastic. In case of Flat specimen, bonding
between the mastic and the bottom plywood was protruded quite a lot,
which was expected to affect the bending response of the plywood near
to the mastic. In case of Light specimen, the bonding status was worse
than the one of Flat specimen. There were noticeable defects of
bonding between the mastic and the plywood. Among three specimens,
the bonding in Heavy specimen was comparatively quite good.
EVALUATION OF SIMULATION RESULTS
Static Load
Fig. 14 Slicing CCS
(a) Light specimen (b) Heavy specimen
Fig. 15 Damaged membrane of Light and Heavy specimens
Fig. 16 Cracks on RPUF of Flat specimen
(a) Flat specimen
(b) Light specimen
(c) Heavy specimen
Fig. 17 Bonding status of mastic to plywood
After all drop tests, static load tests were carried out and strain
distributions on the bottom plywood of specimens were measured to
check the integrity of the physical specimens and measuring gauges.
Figure 18 shows how to carry out the static load test. The static load
tests with Flat and Heavy specimens only were carried out. In case of
Light specimen, the static load test could not be carried out because the
signal cables were already removed after the drop tests. HHI, BV and
LR simulated the static load test by using each own method and
structure model. Since the numerical models were identical except for
the membrane and the mass, which should not have a large effect on
results, each partys simulation was carried out for only one model of
either Flat or Heavy.
Figure 19 shows the comparison of measurements and simulations at
the static load weight of 1,440 Kg. The measurement result of Flat
specimen shows very strange strain distribution comparing to those of
Heavy specimen and simulations. This seems mainly due to poor
bonding and quality of the mastic, as described previously. In Heavy
specimen, the measured strain values and distribution closely
correspond to those obtained in simulations, although they are not
symmetry. These results, in line with those obtained by cutup
specimens, indicate that comparisons of simulated strains with
measured results during drop tests may be meaningful only in case of
Heavy specimen.
(a) 0 Kg (b) 40 Kg (c) 540 Kg
(d) 1,040 Kg (e) 1,440 Kg
Fig. 18 Weight set up for static load test (Heavy specimen)
 300
 200
 100
0
100
200
300
 200  160  120  80  40 0 40 80 120 160 200
Distance(mm)
S
t
r
a
in
(
)
Flat measured
Heavy measured
HHI simulation
LR simulation
BV simulation
Position
of
mastic
Fig. 19 Comparison of strains at plywood under static load of 1,440 Kg
Wet Drop
During drop tests, velocity of drop specimens was calculated using
tracking points in pictures taken by a high speed camera and bundled
software. To get the drop accelerations in drop tests, the slope of the
obtained velocity time histories were calculated. The averaged
acceleration was 8.787 m/s
2
and the averaged final entry velocity was
94.6% of the free fall theoretical value calculated from the gravitational
acceleration assuming no resistance or friction. Accordingly, it was
decided to take the velocity as 95% of the free fall theoretical velocity
for simulations.
Pressure Time History
Figures 20 and 21 show typical comparisons of pressure time histories
at the membrane center (P2) of Flat and Light specimens with the
incident angle of 4 at drop height of 2 and 4 meters, respectively. Due
to differences between time zero in simulations of each party and time
zero in drop tests, all results are shifted on the time axis such that the
time of maximum pressure at P2 in measurements is matched with the
time of that pressure in simulations.
HHI
Peak pressures in simulations are generally close to or a little lower
than those in measurements. Pressure durations in simulations are about
20% longer in general than those in measurements. Especially, quick
rising of pressure is not as apparent as that of measurements. These
differences between the results of simulations and measurements seem
to be mainly owing to the size of meshes in the simulation model. The
mesh size in simulations was 20 x 20 mmfor the structure model and
30 x 30 mm for the fluid model, whereas the pressure gauge in
measurements was 6 mm diameter. Further research work will be
carried out to verify the effect of the mesh size.
ABS
The simulated pressure shows good agreement in magnitude and shape
with the measured pressure for the most of the cases. It is of interest to
note that simulated pressure agrees well with the experimental value
although fluidstructure interaction was not considered in the CFD
simulation.
BV
When the problem is strongly hydroelastic, measured/calculated
pressure seems not proper value for the evaluation and comparison.
Structural response during the impact in hydroelastic coupled analysis
has to be expressed and compared based on the structural deformation
and induced stresses. The load, i.e. pressure is the value which doesnt
have particular meaning in this kind of analysis. Even measurements of
pressure, in particular pressure peak value is extremely difficult and
could be influenced by many local phenomena. For the purpose of this
project and for the demanded comparison, pressure as one intermediate
result was extracted from calculations. Comparison of calculated results
with measurements is shown in Fig. 20. For the cases with flat primary
membrane calculated values have fairly good agreement with
measurements. In general, calculated peak values are higher than
measured. For the cases with corrugated primary membrane, calculated
peak values were much higher than measured.
LR
It is clear from Figs. 20 and 21 that, as expected, CFD prediction of the
flow field and the resulting pressures are much more accurate than the
FEA simplified approach. Furthermore it was not possible to obtain
direct reading of pressure within the FEA code and therefore it had to
be obtained indirectly via stresses in the vertices nearest to the position
of the pressure gauges during experiment. This was furthermore
complicated by the noise in the results from the explicit solver which
required filtering. The resulting pressure plots exhibit several spikes in
the curve which are not observed in the measurements. Taking all
results into consideration, it is clear that the ABAQUS simplified water
model is not very accurate in reproducing local fluid behavior in any
great detail. CFD, on the other hand, generally shows excellent
agreement between predictions and measurements, both in terms of the
magnitude and in terms of the behavior of pressure. This accuracy was
achieved by taking all elements of the drop into consideration: the drop
started at the appropriate height (3 or 4 meters) and was fully tracked
until the point of impact when coupling with ABAQUS was engaged.
Compressibility of both water and air was taken into account. Fine
mesh in the vicinity of the membrane allowed for high resolution of
water surface deformation during the impact and accurate pressure
distribution across the whole of the surface. The drawback in achieving
this accuracy was the relatively long runtime. Some compromises have
therefore to be made to keep runtimes within practical limits, but, as
figures demonstrate, little is sacrificed in terms of accuracy.
Strain Time History
As described previously, it was hard to compare the structural
responses one by one between simulations and measurements, due to
poor bonding and quality of the mastics in specimens. Among signals
of gauges, values of SP8 rather seemed to show close relation to results
of simulations, because it was located between mastics. In this regard,
comparisons of strain time histories at SP8 position only are presented
in this paper. Figures 22 and 23 show typical comparisons of strain
0.01 0.02 0.03 0.04
time [sec]
4
0
4
8
12
16
20
24
28
32
P
r
e
s
s
u
r
e
[
b
a
r
]
Pressure P2 flat, 2m, 4d
PNU
HHI
ABS
BV
(a) Drop height 2 meter
(b) Drop height 4 meter
Fig. 20 Pressure time history at P2 of Flat specimen (incident angle 4)
time histories at SP8 position of the bottom plywood of Flat and Light
specimens with the incident angle of 4 at drop height of 2 and 4
meters, respectively. Figure 24 shows comparison of strain time
histories at SP8 position of the bottom plywood in Heavy specimen at
the drop height of 2 meter with 0 incident angle. As the case of pressure
time history, all results are shifted on the time axis such that the time of
maximum pressure at P2 in measurements is matched with the time of
that pressure in simulations.
HHI
All results of simulations show good agreement with those of
measurements in general according to comparisons of strain time
histories at SP8.
ABS
Both magnitude and shape of the simulated strain response show good
agreement with the measured results.
BV
The strain results of simulations show good agreement with those of
measurements. Calculated strains are in general slightly higher than
measured values; this might be due to the structural FE model which
doesnt have defined damping materials property.
0.01 0.02 0.03 0.04
time [sec]
4
0
4
8
12
16
20
24
28
32
P
r
e
s
s
u
r
e
[
b
a
r
]
Pressure P2 light, 2m, 4d
PNU
HHI
(a) Drop height 2 meter
0.01 0.02 0.03 0.04
time [sec]
4
0
4
8
12
16
20
24
28
32
P
r
e
s
s
u
r
e
[
b
a
r
]
Pressure P2 light, 4m, 4d
PNU
HHI
LR(FEAonly)
LR(CFD1way)
(b) Drop height 4 meter
0.01 0.02 0.03 0.04
time [sec]
4
0
4
8
12
16
20
24
28
32
P
r
e
s
s
u
r
e
[
b
a
r
]
Pressure P2 flat, 4m, 4d
PNU
HHI
LR(FEAonly)
LR(CFD1way)
LR(CFD2way)
BV
Fig. 21 Pressure time history at P2 of Light specimen (incident angle 4)
LR
What the results show for the cases with the incident angle of 0 is the
following:
 Both FEAonly (simplified water model) and CFD 1way coupling
overpredict the measured strains.
 The simplified water model in FEA does not cope well with the 0
degree impact and produces unrealistic spikes in the filtered data
which are visible both in pressure results and resulting strains.
 CFD 1way predicts pressure well, providing the correct input to
stress calculations which results in the smooth strain curves similar in
character to those measured. However, strain magnitude is over
predicted compared with measurements to about the same level as the
FEAonly model. This suggests that spikes generated by the FEA
water model do not have an important influence on the maximum
strain values. It also suggests that the structural response is less
sensitive to the prediction accuracy of local pressure variations.
 Calculated strains with CFD 2way coupling are much closer to those
measured. For point SP8, predicted strain magnitudes match
measurements quite well, although the measured values show sharper
rise and fall. This is an encouraging result and the outcome, which
one would hope to see when the full feedback of 2way coupling is
introduced. Unfortunately, 2way coupling has a dramatic effect on
the predicted pressures by reducing the peak magnitude by about
0.01 0.02 0.03 0.04
time [sec]
3000
2000
1000
0
1000
2000
3000
4000
5000
6000
7000
8000
M
i
c
r
o

S
t
r
a
i
n
Strain of backplywood  SP8, flat, 2m, 4d
PNU
HHI
BV
(a) Drop height 2 meter
0.01 0.02 0.03 0.04
time [sec]
3000
2000
1000
0
1000
2000
3000
4000
5000
6000
7000
8000
M
i
c
r
o

S
t
r
a
i
n
Strain of backplywood  SP8, flat, 4m, 4d
PNU
HHI
LR(FEAonly)
LR(CFD1way)
LR(CFD2way)
BV
(b) Drop height 4 meter
Fig. 22 Strain time history at SP8 of Flat specimen (incident angle 4)
40%. This in effect means that in order to match measurements, the
solid stress model has to see much lower pressures than the
experimental measurements show.
To assess whether the same findings are applicable to any of other
cases, the cases with the incident angle of 4 are run 2way coupled
(Fig. 22). While the outcome agrees with the previous case in most
points, one significant difference is observed: strains are overpredicted
compared to measurements even with a 2way coupling although
pressures again show a clear drop to below measured values. These
findings indicate that there may be a mismatch between the stress
model and the tested physical sample. Due to the complex structural
design of the Mark III corrugated section, it is difficult to ascertain
what has the biggest influence, but the computed strains suggest that
the models need to be generally stiffer in order to predict lower strains
in response to recorded pressures. A stiffer model would results in the
lower deformations being passed over to CFD model and, one would
expect, a smaller drop in the 2way coupled pressure prediction.
Potential issues with material properties could also be linked to static
tests. From Fig. 19, it is clear that strain overprediction can be
observed for a static load condition although this is much less
pronounced than for the actual drop case (dynamic impact). All this ties
0.01 0.02 0.03 0.04
time [sec]
3000
2000
1000
0
1000
2000
3000
4000
5000
M
i
c
r
o

S
t
r
a
i
n
Strain of backplywood  SP8, light, 2m, 4d
PNU
HHI
BV
(a) Drop height 2 meter
0.01 0.02 0.03 0.04
time [sec]
3000
2000
1000
0
1000
2000
3000
4000
5000
M
i
c
r
o

S
t
r
a
i
n
Strain of backplywood  SP8, light, 4m, 4d
PNU
HHI
LR(FEAonly)
LR(CFD1way)
BV
(b) Drop height 4 meter
Fig. 23 Strain time history at SP8 of Light specimen (incident angle 4)
0.01 0.02 0.03 0.04
time [sec]
3000
2000
1000
0
1000
2000
3000
4000
5000
M
i
c
r
o

S
t
r
a
i
n
Strain of backplywood  SP8, heavy, 2m, 0d
PNU
HHI
ABS
LR(FEAonly)
LR(CFD1way)
Fig. 24 Strain time history at SP8 of Heavy specimen
(drop height 2 meter, no incident angle)
in with the overall impression that the complexity of the Mark III
section and inherent difficulty in assigning accurate material properties
could result in the solid model not being accurate enough to allow good
prediction of strains.
CONCLUDING REMARKS
To establish a reliable prediction method of LNG CCS response to
sloshing loads, physical wet drop tests of CCS specimens and their
numerical simulations were carried out by four parties of HHI, ABS,
BV and LR using each partys own method, under the name of J DP for
Wet Drop Test Simulation. Three types of the specimen were fabricated
and droptested by PNU. Flat and Light specimens were tested at drop
heights of 2, 3 and 4 meter in combination with incident angles of 0, 4
and 8. Heavy specimen was tested at drop heights of 2 and 3 meter
only without an incident angle. Comparison works of pressure and
strain time histories between the results of measurements and
simulations were performed.
According to comparisons of pressure time histories on the membrane,
results of each partys simulation show good agreement with those of
measurements in general, even though some parties results require
some further consideration. In case of strain time histories, it was
verified through static load tests and cuttingup specimens that results
of measurements were not so reliable due to poor craftsmanship in
mixing and bonding mastics during fabrication of specimens.
Notwithstanding errors in measurements, comparisons of strain time
histories at a typical location of the bottom plywood between mastics,
where strain signals in static load tests showed relatively close relation
to results of simulations, were performed. All results of each partys
simulations present reasonable response of strain time histories.
In line with findings and outcomes through this J DP, it is highly
recommended to proceed with next stage of a development project to
establish more practical and rational assessment procedure for CCS and
supporting hull structures of LNG carriers under sloshing loads.
REFERENCES
American Bureau of Shipping (2006). Guidance Notes on Strength
assessment of MembraneType LNG Containment Systems under
Sloshing Loads
Chung, J Y, Lee, J H, Kwon, SH, Ha, MK, Bang, CS, Lee, J N, and Kim,
J J (2006). Wet Drop Test for LNG Insulation System, Proc 16
th
(2006) Int Offshore and Polar Eng Conf, Vol III, pp 199204, San
Francisco
Gazzola, T, Korobkin, A, and Malenica, S (2006). Hydroelastic
Wagner Impact using Variational Inequalities, Proc Int Workshop
on Water Waves and Floating Bodies, Loughborough
Malenica, S, Korobkin, A, Gueret, R, Delafosse, V, Gazzola, T,
Mravak, Z, and Zalar, M (2006). Hydroelastic Impacts in the
Tanks of LNG Carriers, Proc 4
th
(2006) Int Conf on
Hydroelasticity, Wuxi
Nam, SK, Kim, WS, Noh, BJ , Shin, HC, Choe, IH, Park, KH, Kim, DE,
and Rashed S (2005). Structural Response of Membrane Tanks to
Sloshing Load in a Mark III Type LNG Carrier, Proc 19
th
(2005)
TEAM Conf, pp 347352, Singapore
Nam, SK, Kim, WS, Noh, BJ , Shin, HC and Choe, IH, (2006). The
Parametric Study on the Response of Membrane Tanks in a Mark
III Type LNG Carrier Using Fully Coupled Hydroelastic Model,
Proc Int Conf on ICSOT 2006: Design, Construction & Operation
of Natural Gas Carriers & Offshore Systems, pp 147154, Busan
Copyright 2008 by ASME 1
HYDROELASTIC ASPECTS OF LARGE CONTAINER SHIPS
Ivo Senjanovi
University of Zagreb
Faculty of Mechanical Engineering
and Naval Architecture
I. Lucica 5, 10000 Zagreb
Croatia
ime Malenica
Bureau Veritas
17 bis, Place des Reflets, La
Dfense 2
Paris La Defense Cedex 92077
France
Stipe Tomaevi, Marko Tomi
University of Zagreb
Faculty of Mechanical Engineering
and Naval Architecture
I. Lucica 5, 10000 Zagreb
Croatia
ABSTRACT
The importance of hydroelastic analysis of large and flexible
container ships of today is pointed out. A methodology for
investigation of this challenging phenomenon is drawn up and a
mathematical model is worked out. It includes definition of ship
geometry, mass distribution, structure stiffness, and combines ship
hydrostatics, hydrodynamics, wave load, ship motion and vibrations.
Based on the presented theory, a computer program is developed and
applied for hydroelastic analysis of a flexible segmented barge for
which model test results of motion and distortion in waves have been
available. A correlation analysis of numerical simulation and
measured response shows quite good agreement of the transfer
functions for heave, pitch, roll, vertical and horizontal bending and
torsion. The developed tool is furthermore used for hydroelastic
analysis of a large container ship.
1. INTRODUCTION
Large container ships are very flexible and their structural natural
frequencies can fall into the range of the encounter frequencies in an
ordinary sea spectrum. Therefore, the wave induced hydroelastic
response becomes an important issue especially for improving the
classification rules and ensuring ship safety. For container ships, due
to open crosssection, the lowest elastic natural modes are those of
coupled horizontal and torsional vibrations. This coupling is highly
pronounced due to the fact that the shear (torsional) centre is below
the keel, i.e. far from the centre of gravity. The classical approach to
determine ship motions and wave loads is based on the assumption
that the ship hull acts as a rigid body [1]. That approach is not
reliable enough for ultra large ships due to mutual influence of the
wave load and structure response. Therefore, a reliable solution
requires analysis of wave load and ship vibrations (springing and
whipping) as a coupled hydroelastic problem.
The hydroelastic theory has been established by Bishop and Price as
the unified stripbeam theory [2]. The hull model is represented by
the Timoshenko beam and the fluid action by the strip theory [3]. The
hydroelastic theory is extended to general floating structures in such a
way that the space structural modes are obtained by the finite element
method and the fluid action is determined by a threedimensional
potential theory [4].
One of the most recent detailed investigations of ship hydroelasticity
in case of bulk carrier is elaborated in [5]. Timoshenko beam
idealization and 3D FEM model combined with potential flow theory
are applied. The vertical bending responses obtained from one and
threedimensional models are in good agreement, while differences
are observed for the horizontal bending and twisting responses. This
may be due to inadequate modelling of the highly nonprismatic hull
girder, discontinuities between open and closed crosssection
segments, and warping effects [6].
The above stateoftheart gave incentive to further investigation of
this challenging problem. The present paper is intended to propose
the consistent theory for hydroelastic coupling between the structural
model, represented by either 1D beam model or 3D FEM model, and
3D hydrodynamic model. The well known principles of the modal
superposition method are employed. However, the proposed model is
slightly different: namely in the way in which hydrostatics is taken
into account and in the way the coupling with 3D hydrodynamic
model is performed.
The methodology of hydroelastic analysis, which includes definition
of the structural model, ship and cargo mass distributions, and
geometrical model of ship surface, is shown in Fig. 1 according
to [7]. First, dry natural vibrations are calculated, and then modal
hydrostatic stiffness, modal added mass, damping and modal wave
load are determined. Finally, wet natural vibrations are obtained as
well as the transfer functions (RAOresponse amplitude operator) for
determining ship structural response to wave excitation.
Proceedings of the ASME 27th International Conference on Offshore Mechanics and Arctic Engineering
OMAE2008
June 1520, 2008, Estoril, Portugal
OMAE200857111
Copyright 2008 by ASME 2
FIGURE 1. METHODOLOGY OF HYDROELASTIC
ANALYSIS
2. STRUCTURAL MODEL
The hydroelastic problem can be solved at different levels of
complexity and accuracy. The best, but highly timeconsuming way,
is to consider 3D FEM structural model and 3D hydrodynamic model
based on the radiationdiffraction theory. Such an approach is
recommended only for the final strength analysis of ship structures.
However, in the preliminary strength analysis, it is more rational and
convenient to couple 1D FEM model of ship hull with 3D
hydrodynamic model.
1D FEM procedure for vertical ship hull vibrations is well known in
literature [1]. Coupled horizontal and torsional vibrations are a more
complex problem. The matrix finite element equation for coupled
natural vibrations yields [8]
{ } [ ] { } [ ] { }
e
e e e e
f k m = +
, (1)
where
{f}
e
nodal forces vector
{}
e
nodal displacements vector
[k]
e
stiffness matrix
[m]
e
mass matrix.
These quantities consist of flexural and torsional parts
{ } { }
e e
e e
,
P U
f
R V
= =
` `
) )
(2)
[ ] [ ]
e e
0
, .
0
bs sb st
wt ts tw
k m m
k m
k m m
( (
= =
( (
(3)
Vectors of nodal forces and displacements are
{ } { }
(0) (0)
(0) (0)
, ,
( ) ( )
( ) ( )
w
w
T Q
B M
P R
T l Q l
B l M l
= =
` `
) )
(4)
{ } { }
(0) (0)
(0) (0)
, .
( ) ( )
( ) ( )
w
U V
w l l
l l
= =
` `
) )
(5)
In the above formulae symbols Q, M, T and B
w
denote shear force,
bending moment, torque and warping bimoment, respectively. Also,
, , and w are deflection, rotation of crosssection, twist angle
and its variation, respectively. The submatrices have the following
meaning:
[k
bs
] bending  shear stiffness matrix
[k
wt
] warping  torsion stiffness matrix
[m
sb
] shear  bending mass matrix
[m
tw
] torsion  warping mass matrix
[m
st
] = [m
ts
]
T
shear  torsion mass matrix.
It is evident that coupling between horizontal and torsional vibrations
is realised through the mass matrix due to eccentricity of the centre of
gravity and shear centre.
Before assembling of finite elements it is necessary to transform Eq.
(1) in such a way that all the nodal forces as well as nodal
displacement, Eqs. (4) and (5), are related to the first and then to the
second node. Furthermore, Eq. (1) has to be transformed from local
to global coordinate system. The origin of the former is located at the
shear centre, and of the latter at the base line.
Regardless of the FEM approach for the structural model, the
governing matrix equation of dry natural hull vibrations yields [9]
[ ] [ ]
( )
{ } { }
2
0 = K M , (6)
where
[K]  stiffness matrix
[M]  mass matrix
 dry natural frequency
{}  dry natural mode.
As solution of the eigenvalue problem (6),
i
and {}
i
are obtained
for each ith dry mode, where i = 1,2...N, N is total number of degrees
of freedom. Now the natural modes matrix can be constituted
[ ] { } { } { } { } , ... ...
1 2 i N
(
=
, (7)
and the modal stiffness and mass can be determined [9]
[ ] [ ] [ ][ ] [ ] [ ] [ ][ ] ,
T T
= = k K m M . (8)
Since the dry natural vectors are mutually orthogonal, matrices [k]
and [m] are diagonal. Terms k
i
and
2
i i
m represent deformation
energy and kinetic energy of the ith mode, respectively.
Note that generally the first six natural frequencies
i
are zero with
corresponding eigenvectors representing the rigid body modes. As a
result, the first six diagonal elements of [k] are also zero, while the
first three elements in [m] are equal to structure mass, the same in all
directions x, y, z, and the next three elements represent the mass
moment of inertia around the coordinate axes.
Copyright 2008 by ASME 3
3. GEOMETRICAL MODEL OF WETTED SURFACE
For determining pressure forces acting on the wetted surface it is
necessary to specify panels and their position in space. Wet surface is
given by offsets of waterline ordinates at body plan stations. If the
strip method is used for pressure calculation, then the panels bounded
with two close stations can be used.
If 3D radiationdiffraction theory is used for hydrodynamic pressure
determination, the wetted surface mesh can be created in such a way
that panels of more regular shape and refined subdivision in the area
of the free surface are achieved [10]. Such a rational mesh is shown
in Fig. 2, where the panel rows follow the ship body diagonals
similarly to the structural elements of ship outer shell. In this way, the
efficiency of hydrodynamic calculation is increased.
FIGURE 2. RATIONAL MESH OF WETTED SURFACE
4. DRY MODES OF WETTED SURFACE
As mentioned in Section 2, structural dry modes can be determined
by 1D or 3D FEM analysis. If 1D analysis is used, the beam modes
are spread to the ship wetted surface as follows.
Vertical vibration
d
( )
d
w
v
Z z w
v N v
x
= + h i k , (9)
horizontal vibration
d
d
w
h
Y w
h h
x
= + h i j , (10)
torsional vibration
( ) Z z Y
t S
= h j k , (11)
where w is hull deflection, is twist angle, Y and Z are coordinates of
the point on ship surface, and z
N
and z
S
are coordinates of neutral line
and shear centre respectively.
If strong coupling between horizontal and torsional vibration occurs,
as in the case of container ships, the coupled mode yields
d
d
( )
d d
w
h
Y u w Z z Y
ht h S
x x
 
(
 = + + +

\
h i j k , (12)
where ( , , ) u u x Y Z = is the crosssection warping function reduced
to the wetted surface [11].
5. HYDRODYNAMIC MODEL
5.1. General
The fully consistent and efficient seakeeping model with forward
speed, even for rigid body, does not exist yet and only the
approximate models are employed for the time being. These
approximated models spread from the so called 2D strip theories,
over so called encounter frequency approximation (main effect is put
on the change of frequency due to speed) and different 3D
approximations based on the exact forward speed Green function
with or without coupling with the ship steady flow, to nonlinear time
domain seakeeping models. However, according to the authors
knowledge, none of the above methods has been validated enough,
and the problem is still open.
The most widely used method, the one based on the encounter
frequency approximation, is applied in this study, even if the method
utilizing the exact forward speed Green function (together with so
called NeumanKelvin approximation) is also implemented in the
software used [12].
The choice of hydrodynamic model is not likely to change the
general conclusions of this paper, because the experience shows that,
in the case of a rigid body, the encounter frequency method agrees
quite well with more complicated methods, at least for the global ship
loadings. This, however, does not mean that we should forget this
important assumption, and that point should be kept in mind when
more detailed correlations with experiments and full scale results will
be undertaken.
Anyhow, the coupling procedure does not depend on the used
hydrodynamic model, and is therefore described here for the zero
speed case, as the simplest one.
5.2. Theoretical Background
The harmonic hydroelastic problem is considered in frequency
domain and therefore we operate with amplitudes of forces and
displacement. In order to perform coupling of the structural and
hydrodynamic models, it is necessary to subdivide the external
pressure forces in a convenient manner [13]. First, the total
hydrodynamic force F
h
has to be split into two parts: the first part F
R
depending on the structural deformations, and the second one F
DI
representing pure excitation
h R DI
F F F = + . (13)
Furthermore, the modal superposition method can be used. The
vector of the wetted surface deformations H
(x, y, z) can be presented
as a series of dry natural modes h
i
(x, y, z)
( , , ) ( , , )
1
( , , ) ( , , ) ( , , ) ,
1
N
x y z x y z
i i
i
N
i i i
h x y z h x y z h x y z
i x y z
i
=
=
(
= + +
(
=
H h
i j k
(14)
where
i
are unknown modal coefficients. Vectors h
i
(x, y, z) related
to wetted surface are obtained from the structural dry modes as
explained in the previous section.
As far as the hydrodynamic part of the hydroelastic problem is
concerned, the potential flow theory is adopted. Within this approach
the fluid is assumed inviscid and fluid flow irrotational, so that the
velocity potential can be defined and the corresponding boundary
Copyright 2008 by ASME 4
value problem can be formulated. The general seakeeping problem
for ship advancing with forward speed in waves is an extremely
difficult problem and only approximate solutions exist today. In this
paper we do not enter into the details of the pure hydrodynamic
analysis and we concentrate on the coupling principles only. Indeed,
the coupling procedure remains the same regardless of the
complexity of the boundary value problems for velocity potentials.
Thus, this coupling procedure is illustrated for the seakeeping
problem without forward speed. The total velocity potential , is
defined with the Laplace differential equation and the given boundary
values
0 within the fluid
0 at the free surface, 0
on the wetted body surface, ,
z
z
i S
n
+ = =
Hn
(15)
where is the wave number,
2
/ g = , is wave frequency, n is
the wetted surface normal vector, and i is the imaginary unit.
Furthermore, the linear wave theory enables the following
decomposition of the total potential
1
N
i
I D j Rj
j
= +
=
, (16)
where
( )
e
gA z ix
i
I
+
= (17)
 incident wave potential
 diffraction wave potential
 radiation wave potential
A  wave amplitude
I
D
Rj
Now, the body boundary conditions (15) can be deduced for each
potential
,
Rj
D I
j
n n n
= =
h n . (18)
It is necessary to point out that the diffraction and radiation potentials
should also satisfy the radiation condition at infinity.
Once the potentials are determined, the modal hydrodynamic forces
are calculated by pressure work integration over the wetted surface.
The total linearised pressure can be found from Bernoulli's equation
p i gz = . (19)
First, the term associated with the potential is considered and
subdivided into excitation and radiation parts (16)
( ) d
DI
F i S
i I D i
S
= +
h n , (20)
2
d
1
N
R
F S
i j Rj i
j S
=
=
h n . (21)
Thus, (20) represents the modal pressure excitation. Now one can
decompose (21) into the modal inertia force and the damping force
associated with acceleration and velocity respectively
2
Re( ) , Re d
1
N
a R
F F A A S
i i j ij ij Rj i
j S
= = =
=
h n , (22)
Im( ) , Im d
1
N
v R
F F B B S
i i j ij ij Rj i
j S
= = =
=
h n , (23)
where A
ij
and B
ij
are elements of added mass and damping matrices,
respectively.
Determination of added mass and damping for rigid body modes is a
wellknown procedure in ship hydrodynamics [1]. Now the same
procedure is extended to the calculation of these quantities for elastic
modes.
The hydrostatic part of the total pressure, gz in (19) is considered
within the hydrostatic model.
6. HYDROSTATIC MODEL
6.1. Pressure Forces
Hydroelastic analysis is performed by the modal superposition
method. Modal forces represent work of actual static and dynamic
forces on rigid body and elastic mode displacements. Thus, modal
restoring forces consist of timedependent modal pressure forces and
modal gravity forces. In order to specify the contribution of the
hydrostatic part of the total pressure, gz in (19), to hydrostatic
stiffness it is necessary to determine the change of the modal
hydrostatic force as the difference between its instantaneous value
and the initial value for the vibration mode h
i
of the body wetted
surface Z = Z (x, y) [14]
d d
H
F g Z S g Z S
i i i
S S
= +
h n h n
. (24)
Each of the above quantities can be presented in the form,
( )
h n h n h n (25)
Variation of the particular quantity can be determined by applying the
notion of directional derivative H , where H is given with (14)
and is Hamilton differential operator
1 1
N N
j j j
h h h
j j j x y z
x y z
j j
 
= = + +

\
= =
H h . (26)
Copyright 2008 by ASME 5
As a result
( ) , ( ) , ( ) Z Z
i i
= = = = H Hk h H h n H n . (27)
Determining the variation of the body surface normal vector, n,
according to (27) is a rather difficult task. Therefore, a relatively
simpler procedure, taken from [14], is applied. By using Eqs. (26),
(27) and n from [14], the modal hydrostatic force (25) can be
presented in the following form:
1
N
H H
F C
i j ij
j
=
=
, (28)
where
Hp H Hh Hn
C C C C
ij ij ij ij
= + + (29)
is the i,jth element of the hydrostatic stiffness matrix, composed of
static pressure, surface mode and normal vector contributions,
respectively
( )d
Hp j i i i
C g h h n h n h n S
ij z x x y y z z
S
= + +
, (30)
d ,
i i i
h h h
j j j Hh x x x
C g Z h h h n
ij x y z x
x y z
S
i i i
h h h
y y y j j j
h h h n
x y z y
x y z
i i i
h h h
j j j
z z z
h h h n S
x y z z
x y z
 

= + + +


\
 

+ + +


\
(  
( 
+ +
( 

(
\
(31)
d
j j
j j
h h
h h
y y
Hn i z z
C g Z n n n h
ij x y z x
y z x x
S
j j j j
h h h h
i x x z z
n n n h
x y z y
y x z y
j j
j j
h h
h h
y y
i x x
n n n h S
x y z z
z z x y
(  
( 
= + +
( 
( 
\
(  
( 
+ +
( 

(
\
(  
( 
+
` ( 
( 
\ )
.
(32)
Note that the wetted surface coordinate Z is measured from the
waterplane. Based on the constitution of the above coefficients, it is
evident that the hydrostatic stiffness matrix is not diagonal.
In literature, various definitions of hydrostatic stiffness matrix can be
found [15], [16]. The advantage of the present method is that the
derived formulae are general and applicable for a complex body, as
well as for its parts. This is important for determining the local
hydrostatic action as internal loads, transfer of load to a FEM
structural model, etc.
6.2. Gravity forces
The above expressions represent only the action of the hydrostatic
pressure, and the gravity part has to be added in order to complete the
total restoring coefficients. Similarly to the pressure part, Eqs. (24)
and (25), a change of the generalised modal gravity force associated
with a particular mode yields [14]
d
m
F g m
i i
V
=
h k , (33)
where V is body volume and dm is differential mass.
By employing (27) and further (26) one can write
( ) d
1
N
m i m
F g h m C
i z j ij
j V
= =
=
H , (34)
where
d
i i i
h h h
j j j m z z z
C g h h h m
ij x y z
x y z
V
 

= + +


\
. (35)
6.3. Restoring Stiffness
Finally, the complete restoring coefficients read
H m
C C C
ij ij ij
= + . (36)
It is important to point out that the above expressions for the
hydrostatic and gravity coefficients are general and therefore valid
not only for the elastic modes but also for the rigid body modes as
well as for their coupling [14].
7. HYDROELASTIC MODEL
After the structural, hydrostatic and hydrodynamic models have been
determined, the hydroelastic model can be constituted. For that
purpose, let us impose modal hydrodynamic forces (21), (22) and
(23) and hydrostatic and gravity forces (28), (33) to the modal
structural model, Section 2
[ ] [ ]
( )
{ } { } { } { } { } { }
2 DI a v H m
= + + + + k m F F F F F . (37)
Furthermore, all terms dependent on unknown modal amplitudes,
i
,
can be separated on the left hand side. Thus, the governing matrix
differential equation, which combines rigid body motions and
vibrations, is deduced
[ ] [ ] [ ] [ ] ( ) [ ] [ ] ( ) { }{ } { }
2
( ) ( ) . i + + + = k C d B m A F (38)
where all quantities are related to the dry modes:
[k]  structural stiffness
[d]  structural damping
[m]  structural mass
[C]  restoring stiffness
[B()]  hydrodynamic damping
[A()]  added mass
{}  modal amplitudes
{F}  wave excitation
 encounter frequency.
Copyright 2008 by ASME 6
Structural damping can be given as a percentage of the critical value
based on experience with similar ships. As it is wellknown, added
mass and hydrodynamic damping depends on the frequency. The
solution of (38) gives the modal amplitudes
i
and displacement of
any point of the structure obtained by retracking to (14).
The wet natural modes can also be determined by solving the
eigenvalue problem extracted from (38)
[ ] [ ] [ ] [ ] ( )
{ }
{ } { }
2
( ) 0 + + = k C m A . (39)
Now damping is neglected since its influence on the eigenpair is very
small. The solution of (39) gives natural frequencies of ship motion
and vibration in water and the corresponding socalled wet natural
modes. Since added mass is a frequency dependent function, it is
evident that an iteration procedure has to be employed to solve (39).
Therefore, the wet modes are not orthogonal. Also, there are no more
zero natural frequencies and pure rigid body modes due to their
coupling with the elastic modes [13].
8. HYDROELASTICITY OF A FLEXIBLE BARGE
The implementation of the structural model in hydroelastic analysis is
checked in case of a flexible barge, for which test results are
available [7], [13]. The barge consists of 12 pontoons, which are
connected by a steel rod above the deck level, Figure 3.
FIGURE 3. BARGE IN WAVES
The deformation centre is above the gravity centre and strong
coupling between horizontal and torsional vibrations is achieved as in
case of container ships. For illustration, Figures 4 and 5 show
correlation of the calculated and the measured transfer function of
horizontal bending moment and torque, respectively, as function of
wave period T, [13], [17].
FIGURE 4. TRANSFER FUNCTION OF BARGE
HORIZONTAL BENDING MOMENT, c cc c=60
FIGURE 5. TRANSFER FUNCTION OF BARGE TORQUE,
c cc c =60
Also, convergence of the applied modal superposition method in
hydroelastic analysis is confirmed in the case of the same flexible
barge with no forward speed. Figure 6 shows absolute amplitude
values
i
of normalized modes. The first three modes are related to
sway, roll and yaw, while the remaining modes are elastic. All
diagrams are determined for heading angle =120
o
, and each of them
is a result of single wave length in the range /L=0.52.
FIGURE 6. MODAL AMPLITUDES OF COUPLED
HORIZONTAL AND TORSIONAL VIBRATIONS OF
FLEXIBLE BARGE, U=0 kn, c cc c=120
9. HYDROELASTICITY ANALYSIS OF CONTAINER
SHIP
9.1. Ship Particulars
The application of developed theory is illustrated in case of a 7800
TEU VLCS (Very Large Container Ship), Figure 7. The main vessel
particulars are the following:
Length overall L
oa
= 334 m
Length between perpendiculars L
pp
= 319 m
Breadth B = 42.8 m
Depth H = 24.6 m
Draught T = 14.5 m
Displacement, full load
f
= 135336 t
Displacement, ballast
b
= 68387 t
Engine power P = 69620 kW
Ship speed v = 25.4 kn .
Copyright 2008 by ASME 7
FIGURE 7. 7800 TEU CONTAINER SHIP
The ship hull stiffness properties are calculated by program STIFF,
based on the theory of thinwalled girders [11]. The geometrical
properties rapidly change values in the engine and superstructure area
due to closed ship crosssection. This is especially pronounced in
case of torsional modulus, which takes quite small values for open
crosssection and rather high for the closed one [18].
9.2. Natural Vibrations
Dry natural vibrations are calculated by program DYANA [18]. The
ship hull is divided into 50 beam finite elements. Finite elements of
closed crosssection (6 d.o.f.) are used in the ship bow, ship aft and in
the engine room area. Dry natural frequencies for vertical vibrations,
and coupled horizontal and torsional vibrations for full load and
ballast conditions are listed in Table 1. Their values for ballast
condition are higher due to the lower mass. The lowest frequency
value, which plays the main role in wave excitation, is detected for
coupled vibrations. As a result of rather low torsional stiffness, this
value belongs to primarily torsional mode.
TABLE 1 . DRY NATURAL FREQUENCIES, i [rad/s]
Full load Ballast
Mode
no.
Vertical
Horizontal +
torsional
Vertical
Horizontal +
torsional
1 4 2.18 5.59 3.98
2 8.41 4.23 11.6 6.79
3 13.22 7.08 17.85 11.55
4 18.07 9.23 25.02 14.9
5 23.04 13.19 32.52 20.59
6 28.09 15.37 39.21 23.03
7 32.77 18.22 46.29 27.98
8 37.22 22.65 53.25 32.64
9 41.73 23.75 60.4 36.33
10 42.27 28.38 66.96 40.38
Once the dry natural modes of ship hull are determined it is possible
to transfer the beam node displacements to the ship wetted surface for
the hydrodynamic calculation, Eqs. (9) and (12). The first two dry
natural modes of the ship wetted surface in case of vertical and
coupled horizontal and torsional vibrations for full load are shown in
Figures 8, 9, 10 and 11.
FIGURE 8. THE FIRST NATURAL MODE OF VERTICAL
VIBRATIONS, 1=4 rad/s
FIGURE 9. THE SECOND NATURAL MODE OF VERTICAL
VIBRATIONS, 2=8.41 rad/s
In addition, the values of natural frequencies of wet ship in full load
condition are listed in Table 2. By comparing them to those of dry
ship in Table 1, one sees that vertical vibration frequencies are more
reduced than those of coupled horizontal and torsional vibrations.
TABLE 2. WET NATURAL FREQUENCIES, i [rad/s]
Full load
Mode no.
Vertical
Horizontal +
torsional
1 3.07 2.04
2 6.33 3.65
3 10 6.57
4 13.96 8.28
5 18.09 11.73
6 22.18 13.46
7 26.23 16.34
8 30.48 20.38
9 35.01 21.51
10 41.23 25.88
FIGURE 10. THE FIRST NATURAL MODE OF COUPLED
HORIZONTAL AND TORSIONAL VIBRATIONS,
1 = 2.18 rad/s
Copyright 2008 by ASME 8
FIGURE 11. THE SECOND NATURAL MODE OF COUPLED
HORIZONTAL AND TORSIONAL VIBRATIONS,
2 = 4.23 rad/s
9.3. Ship Response
Numerical calculation of ship response to waves is performed for full
load and ballast condition, unit harmonic wave amplitude, and set of
heading angles, ship speed and wave length [18]. Here, only some
selected results for full load and ship speed of U = 25 kn are
presented. Vertical response is shown for head sea, c = 180, and
coupled horizontal and torsional response for quartering sea,
c = 120. In all figures hydroelastic response is correlated to rigid
body response determined by program HYDROSTAR [12], where
the so called encounter frequency approximation was used for the
solution of the seakeeping problem with forward speed. Due to the
lack of measured data on large container vessels, modal damping that
is usually used in global vibration calculations of merchant ships, i.e.
2% of the critical value, is applied in the analysis [19]. Figure 12
shows the transfer function of the vertical bending moment at the
midship section. Heave resonance occurs at encounter frequency
e
= 0.83 rad/s, while bending resonance is achieved at
e
= 3.74 rad/s. In the former case, bending moments determined by
hydroelasticity and rigid body motion are almost the same, Figure 13.
In the latter case, the hydroelastic bending moment takes very high
values compared to the rigid body one, Figure 14.
FIGURE 12. TRANSFER FUNCTION OF VERTICAL
BENDING MOMENT, x=161.35 m, c cc c=180, U=25 kn
FIGURE 13. VERTICAL BENDING MOMENT,
e=0.83 rad/s, =246.55 m, c cc c=180, U=25 kn
FIGURE 14. VERTICAL BENDING MOMENT,
e=3.74 rad/s, =33.82 m, c cc c=180, U=25 kn
The results of coupled horizontal and torsional ship response are
presented in Figures 15 to 20. Figures 15 and 16 show transfer
functions of horizontal bending moment and torque referred to the
shear (torsional) centre, at the midship section, respectively.
Hydroelastic effect is more pronounced in torsional than in bending
response.
FIGURE 15. TRANSFER FUNCTION OF HORIZONTAL
BENDING MOMENT, x=168.57 m, c cc c=120, U=25 kn
Copyright 2008 by ASME 9
FIGURE 16. TRANSFER FUNCTION OF TORQUE,
x=168.57 m, c cc c=120, U=25 kn
Large relative discrepancies between the hydroelastic and rigid body
torque in low frequency domain in Figure 16 might be the result of
illconditioning of the former. In hydrorigidbody analysis, namely,
sectional forces are determined by pressure integration over wetted
surface, while in hydroelastic analysis sectional forces are obtained
by derivation of hull displacements, which are predominantly rigid
body ones in low frequency domain. Looking through the modal
decomposition, derivatives of rigid body modes are zero, while those
of elastic body are quite small with decreased accuracy.
FIGURE 17. HORIZONTAL BENDING MOMENT,
e=1.0212 rad/s, =125.79 m, c cc c=120, U=25 kn
Distributions of bending moment in case of rigid body and elastic
resonance are shown in Figures 17 and 18. In the former case
hydroelastic and rigid body bending moments are quite close, while
in the latter case the elastic bending moment is much higher than the
rigid body one. Similar situation occurs with torques, Figures 19 and
20. When elastic response is in resonance with wave excitation the
hydroelastic torque takes very high values.
FIGURE 18. HORIZONTAL BENDING MOMENT,
e=2.0169 rad/s, =46.61 m, c cc c=120, U=25 kn
FIGURE 19. TORQUE, e=1.3236 rad/s, =85.31 m, c cc c=120,
U=25 kn
FIGURE 20. TORQUE, e=2.0169 rad/s, =46.61 m, c cc c=120,
U=25 kn
9.4. Validation Of 1D FEM Model
For this purpose, the light weight loading condition of dry ship with
displacement =33692 t is considered. The first dominantly torsional
mode and the second dominantly horizontal flexural mode of the
wetted surface are quite similar to those shown in Figures 10 and 11,
obtained for the full load. The reliability of 1D FEM analysis is
verified by 3D FEM analysis of the considered ship [17]. The first
two 3D dry natural modes are similar to those of 1D analysis, Figures
10 and 11. The corresponding natural frequencies obtained by 1D and
3D analyses are compared in Table 3. Quite good agreement is
achieved. Values of natural frequencies for higher modes are difficult
to correlate, since strong coupling occurs between global hull modes
and local substructure modes of 3D analysis.
TABLE 3. NATURAL FREQUENCIES OF CONTAINER SHIP
IN LIGHT WEIGHT CONDITION,
[rad/s]
Mode no. 1 (T+HB) 2(HB+T)
1D FEM 5.39 9.23
3D FEM 5.41 9.42
Copyright 2008 by ASME 10
10. CONCLUSION
Ultra large container ships are quite flexible and stretch the bounds of
present classification rules for reliable structure design. Therefore,
hydroelastic analysis has to be performed. One of the basic steps is
dry natural vibration analysis of ship hull. Vertical vibration
calculation is performed in a standard way, while the coupled
horizontal and torsional vibration is rather complex.
The performed vibration analysis of the 7800 TEU Container ship
shows that the developed hydroelasticity theory, utilizing 1D FEM
structural model and 3D hydrodynamic model, is an efficient tool for
application in ship hydroelastic analyses. The obtained results point
out that the transfer functions of hull sectional forces in case of
resonant vibration (springing) are much higher than in resonant ship
motion. This is the main issue of the paper, which requires further
investigation.
In order to complete hydroelastic analysis of container ships and
confirm its importance for ship safety, it is necessary to proceed
further to ship motion calculation in irregular waves for different sea
states, based on the known transfer functions. This includes
determination of global wave loads, i.e. bending and torsional
moments and their conversion into stresses, stress concentration in
critical areas of ship structures, especially in hatch corners due to
suspended warping, and fatigue of structural details. The same
numerical procedures, which are used in hydrorigidbody analysis,
are at disposal and are applicable in this case.
The used beam model of ship hull is a reasonable choice for
determining wave load. However, stress concentration in hatch
corners calculated directly by the beam model is underestimated
[20,21]. This problem can be overcome by applying substructure
approach, i.e. 3D FEM model of substructure with imposed boundary
conditions from beam response. In any case, 3D FEM model of
complete ship is preferable from the viewpoint of determining stress
concentration.
Another aspect of hydroelasticity is slamming and whipping, which
are related to vertical bending. This impulsive load is considered in
time domain and is outside the scope of this paper. However, it must
be considered within general container ship problematic.
At the end of a complete investigation, which also has to include
model tests and full scale measurements, it will be possible to decide
on the extent of the revision of Classification Rules for the design and
construction of ultra large container ships. Also, operational
conditions for this type of ships, i.e. reduction of ship speed and
change of heading angle for some sea states due to resonant flexural
and torsional response (as well as parametric rolling) have to be
reconsidered.
ACKNOWLEDGMENT
The authors would like to express their gratitude to Ms. Estelle
Stumpf, research engineer of Bureau Veritas, Marine Division
Research Department, for performing the 3D FEM vibration analysis
of container ship and kind permission to publish the valuable results.
REFERENCES
[1] Principles of Naval Architecture. SNAME, 1988.
[2] Bishop RED, Price WG. Hydroelasticity of Ships.
Cambridge University Press, 1979.
[3] Salvensen N, Tuck EO, Faltinsen O. Ship motion and sea
loads. Transactions, SNAME 1970; Vol. 70, p. 250287.
[4] Price WG, Wu Y. Hydroelasticity of Marine Structures,
in Theoretical and Applied Mechanics, Eds. Niordson FI
and Olhoff N. Elsevier Science Publishers B.V. (North
Holland), 1985.
[5] Hirdaris SE, Price WG, Temarel P. Two and three
dimensional hydroelastic modelling of a bulker in regular
waves. Marine Structures 2003; 16, p. 627658.
[6] ISSC, Technical Committee II.2. Proceedings of the 16
th
ISSC, 2006.
[7] Remy F, Molin B, Ledoux A. Experimental and
numerical study of the wave response of a flexible barge.
Hydroelasticity in Marine Technology. Wuxi, China,
2006. p. 255264.
[8] Senjanovi I, Grubii R. Coupled horizontal and
torsional vibration of a ship hull with large hatch
openings. Computers & Structures 1991; Vol. 41, No. 2,
p. 213226.
[9] Bathe KJ. Finite Element Procedures. Prentice Hall,
1996.
[10] Malenica , Senjanovi I, Chen XB. Automatic mesh
generation for naval and offshore hydrodynamic
simulations. SORTA'04, Plitvice Lakes, Croatia, 2004.
[11] Senjanovi I, Fan Y. A higherorder theory of thinwalled
girders with application to ship structures. Computers &
Structures 1992; 43(1), p. 3152.
[12] Hydrostar, User's manual. Bureau Veritas, Paris, 2006.
[13] Malenica , Molin B, Remy F, Senjanovi I. Hydroelastic
response of a barge to impulsive and nonimpulsive wave
load. Hydroelasticity in Marine Technology. Oxford, UK,
2003, p. 107115.
[14] Malenica . Some aspects of hydrostatic calculations in
linear seakeeping. The 14
th
NAV Conference, Palermo,
Italy, 2003.
[15] Newman JN. Wave effects on deformable bodies.
Applied Ocean Research 1994; 16, p. 4759.
[16] Huang LL, Riggs RR. The hydrostatic stiffness of
flexible floating structure for linear hydroelasticity.
Marine Structures 2000; 13, p. 91106.
[17] Malenica , Senjanovi I, Tomaevi S, Stumpf E. Some
aspects of hydroelastic issues in the design of ultra large
container ships. The 22
nd
IWWWFB, Plitvice Lakes,
Croatia, 2007.
[18] Tomaevi S. Hydroelastic model of dynamic response of
Container ships in waves. Ph. D. Thesis, FSB, Zagreb, (in
Croatian), 2007.
[19] Kumai T. Damping factors in the higher modes of ship
vibrations. European Shipbuilding 1958; Vol. 7, No. 1, p.
2934.
[20] Pedersen PT. Torsional response of containerships.
Journal of Ship Research 1985, Vol. 29, No. 3, p. 194
205.
[21] Senjanovi I, Fan Y. Pontoon torsional strength analysis
related to ship with large deck openings. Journal of Ship
Research 1991; Vol. 35, No. 4, p. 339351.
Abstract for 23rd IWWWFB, Jeju, Korea, 2008
Some aspects of whipping response of container ships
Tuitman J
1
& Malenica
S
2
(1) Delft University of Technology, Netherlands (johan.tuitman@tno.nl)
(2) BUREAU VERITAS  DR, Paris, France (sime.malenica@bureauveritas.com)
Introduction
Whipping is usually dened as a transient hydroelastic ship structural response due to impulsive loading
such as slamming, green water, underwater explosion, etc. Here we concentrate on the slamming induced
whipping. Slamming induced whipping is observed both in experiments and in full scale measurements
for any kind of ships as far as they encounter heavy seas in which the slamming type of loading is likely
to occur. One example of the typical whipping event is shown in gure 1 (taken from [1]). This gure
represents the time evolution of the vertical bending moment, following severe slamming event, at the
midship of the relatively small (L
pp
= 124m) general cargo/container vessel. As we can see, the whipping
contribution to the overall vertical bending moment is not only very important but it also last for a
relatively long time due to the low structural damping. One slam event increases multiple extremes in
the bending moment which makes the whipping phenomena to be relevant both for extreme and fatigue
loading of the ship structure.
However, up to the authors knowledge, the hydroelastic whipping eects are not properly accounted
for in the ship design. This is mainly due to the diculties in the correct modelling of whipping and the
needed calculation time. Indeed, in order to calculate the whipping response, one should combine several
aspects (seakeeping in large waves at forward speed, slamming, hydroelasticity, etc.) which are dicult
to model even independently. In spite of all these diculties, there is a lot of research on whipping going
on nowadays, especially in the context of the ultra large container ships. There are several reasons why
the whipping is likely to be more important for the large container ships: high ship speed (close to 30
knots) and large bow are induce higher slamming loads, large ship size reduce the natural frequencies.
In this paper we present the recent numerical developments related to whipping and we apply them to
two container ships dierent in size (L
pp
260m and L
pp
360m). Both extreme loading and fatigue is
discussed.
Figure 1: Typical whipping event.
Numerical model
Numerical model that we use is based on the coupling in between the 3D diraction radiation seakeeping
code for hydrodynamic part and the Timoshenko beam model for structural part. The so called modal
approach is adopted which means that the ship motions and deformations are represented in a series of
6 rigid body modes and several (5 to 10) dry structural modes. The basics of the theory are presented
in [5] and [8] and here below we present just the nal dynamic modal equation:
([ m ]+[ A
]){
(t)}+[ b ]{
(t)}+([ k ]+[ C ]){(t)}+
t
0
[ K(t ) ]{
] is the innite
frequency added mass matrix, [ b ] is the damping matrix, [ k ] is the structural stiness matrix, [ C ] is
the hydrostatic restoring matrix, [ K(t) ] is the matrix of the hydrodynamic memory functions, F is the
vector of the nonimpulsive wave loading (linear and nonlinear) and Q is the vector of impulsive loads.
The dimension of all the above matrices is (6 +N
f
) (6 +N
f
) where N
f
represents the number of dry
structural modes. This equation is integrated in time using the 4th order Runge Kutta method.
It is important to note that the linear hydrodynamic coecients in the above equation are derived
from the frequency domain calculations, using the well known method proposed in [3, 7]. The nonlinear
part of the loading includes nonimpulsive and impulsive parts. The non impulsive part represents the
so called Froude Krylov loads which basically corrects the hydrostatics and pressure distribution around
the waterline while the impulsive part represents the slamming loads. The slamming part is probably
the most dicult part to evaluate numerically. A very robust method is needed and the calculation
of the slamming force should not take to much CPU time to be able to preform long time whipping
simulations. The only fast and reliable methods for evaluation of the slamming forces which are available
today are based on the so called 2D strip approach. The bow is modeled by multiple 2D sections and the
slamming force is evaluated at these sections. In this paper we use two dierent methods for evaluation
of the slamming loads at each 2D section: the Generalized Wagner Model (GWM) [9] and the Modied
Logvinovich Model (MLM) [4]. The advantage of the second method is much lower CPU time, but the
disadvantages is the lower domain of validity. We do not discuss these methods in details here and we
refer to [6].
The above described procedure was integrated into a single numerical tool able to perform the long time
whipping simulations for any prescribed irregular sea state. These simulations allows for determination
of the probability of exceedence of the maxima and for determination of fatigue damage.
Numerical results
As already mentioned, two container ships of dierent size were chosen for illustration of the overall
procedure and for demonstration of the importance of whipping in container ship design. The rst ship
(S1) has the length between perpendiculars of 360m and the second one (S2) 260m. Only the case of head
waves and zero forward speed is investigated which means that the vertical modes only will participate.
The rst few structural natural frequencies in vertical plane, corresponding natural mode shapes for
Mode S1 S2
1 0.432 0.600
2 0.905 1.225
3 1.445 1.950
4 2.008 2.738
5 2.606 3.631
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
w
Time [s]
1st
2nd
3rd
4th
5th
Figure 2: Structural natural frequencies and mode shapes.
S1 and transfer of the rst mode onto hydrodynamic mesh, are shown in Figure 2. Only ve natural
modes are used in the calculations because the tests showed that the higher modes do not participate
signicantly, and because the mode with the highest frequency determines the stable time step.
The midship bending moment is calculated using dierent approaches to investigate the eect of
whipping. The same wave trains are used for the dierent calculations for the same sea condition to
make comparison more valid. The st approach is using linear theory only with a rigid ship. The
contribution of the nonlinear Froude Krylov forces is investigated using a rigid model with the non
linear seakeeping code without slamming. The last approach is a exible ship using nonlinear seakeeping
code and applying the slamming loads. For every sea condition six halve hour runs using 250 wave
components are glued to obtain enough statistical information.
Example of typical output is presented in Figure 3 where the zoom on the typical whipping event is
also shown. As we can see from this gure, the slamming usually occurs in sagging conditions but it last
long enough to inuence the maximum hogging moment too. These signals are rich of informations and
can be used for determination of the maximum expected values as well as for the determination of the
fatigue damage by using the rainow counting method.
4
2
0
2
4
6
8
10
12
14
16
0 2000 4000 6000 8000 10000
M
y
[
G
N
m
]
Time [s]
4
2
0
2
4
6
8
10
12
14
16
8840 8860 8880 8900 8920
M
y
[G
N
m
]
Time [s]
Elastic  with slamming
Rigid non linear
Rigid linear
Figure 3: Typical time history of the vertical bending moment and zoom around a whipping event.
The probability of exceedence of the midship bending moment is shown in gure 4 for the dierent
approaches. Even if the relatively mild sea states were chosen, (H
s
= 9m, T
z
= 11s for S1, and H
s
= 10m,
T
z
= 13.95s for S2), the inuence of whipping is clearly visible. The non linear Froude Krylov forces
increase the sagging moment signicantly and the whipping response does increase it even more. Figure
5 shows the cycle count of the bending moment using the rainow method and the calculated fatigue
damage of the deck. The whipping increase the fatigue damage signicantly by increasing the amplitude
of the large cycles.
0.001
0.01
0.1
1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
P
r
o
b
a
b
ilit
y
o
f
e
x
c
e
e
d
e
n
c
e
My [GNm]
Elastic  with slamming  Hogging
Rigid nonlinear  no slamming  Hogging
Elastic  with slamming  Sagging
Rigid nonlinear  no slamming  Sagging
Rigid linear  no slamming
0.001
0.01
0.1
1
0 2 4 6 8 10 12
P
r
o
b
a
b
ilit
y
o
f
e
x
c
e
e
d
e
n
c
e
My [GNm]
Elastic  with slamming  Hogging
Rigid nonlinear  no slamming  Hogging
Elastic  with slamming  Sagging
Rigid nonlinear  no slamming  Sagging
Rigid linear  no slamming
Figure 4: Probability of exceedence of the midship bending moment (S1left, S2right).
1
10
100
1000
10000
0 2 4 6 8 10 12
N
u
m
b
e
r
o
f
c
y
c
l
e
s
Bending moment [GNm]
Elastic  with slamming
Rigid nonlinear  no slamming
Rigid linear  no slamming
0
2e06
4e06
6e06
8e06
1e05
1.2e05
1.4e05
0 2 4 6 8 10 12
F
a
t
i
g
u
e
d
a
m
a
g
e
Bending moment [GNm]
Elastic  with slamming
Rigid nonlinear  no slamming
Rigid linear  no slamming
Figure 5: Rainow count of bending moment and fatigue damage of the deck for S1.
To obtain design values many sea states have to be evaluated. The sagging moment with a probability
of 10
5
calculated using a Weibull extrapolation and the fatigue damage per hour are shown in gure
6 for a limited number of sea states. In this case the slamming and whipping occurs only in the very
severe sea states. When a non zero velocity would be used, the number of cells where whipping occurs
will increase.
6 8 10 12 14 16 18 6
7
8
9
10
11
12
13
14
1e+06
2e+06
3e+06
4e+06
5e+06
6e+06
7e+06
8e+06
9e+06
1e+07
Rigid linear
Rigid nonlinear
Elastic  with slamming
Tz [s]
Hs [m]
6 8 10 12 14 16 18 6
7
8
9
10
11
12
13
14
0
5e05
0.0001
0.00015
0.0002
0.00025
0.0003
0.00035
0.0004
0.00045
Rigid linear
Rigid nonlinear
Elastic  with slamming
Tz [s]
Hs [m]
Figure 6: The 10
5
probability sagging moment (left) and the fatigue damage per hour (right).
Conclusions
We presented here the numerical method which can be used in ship design for determination of the
inuence of whipping on wave loadings and ship structural responses. The method was demonstrated on
two container ships of dierent size and in both cases the inuence of whipping was found to be important
not only for the maximum values but also for fatigue. It should be mentioned that the examples which
were chosen are just demonstrative ones, and the general methodology should include more complex set
of calculations namely the model should include the eects of forward speed, heading, dierent loading
conditions (full, ballast, etc.) and dierent sea states. At the same time the sensitivity to some other
parameters such as damping, direction of 2D slamming strips, aft slamming, l ength of runs, number of
wave components, etc. should be properly investigated. Maybe the most critical point in the analysis is
the determination of slamming loads which are extremely dicult to evaluate. In this work we used two
dierent methods GWM and MLM and even if there are some dierences in the evaluation of the local
forces the nal results in terms of maximum values and fatigue seem to be in good agreement. This is
important because the MLM method requires much less CPU time. It should also be kept in mind that
both methods are limited to the 2D calculations and some 3D correction coecients need to be employed.
How this should be done is not clear yet.
References
[1] Aalberts P.J. & Nieuwenhuijs M. 2006. : Full scale wave and whipping induced hull girder
loads., 4th. Int. Conf. on Hydroelasticity, Wuxi, China.
[2] Bishop, R.E.D. & Price, W.G., 1979. : Hydroelasticity of ships, Cambridge University Press.
[3] Cummins W.E., 1962. : The impulse response function and ship motions., Schistecknik
[4] Korobkin A.A., 2005.: Analytical models of water impact., Euro. J. Applied Mathematics, 16,
pp. 118.
[5] Malenica
S., Molin B., Remy F. & Senjanovi c I., 2003. : Hydroelastic response of a barge
to impulsive and non impulsive wave loads., 3rd. Int. Conf. on Hydroelasticity, Oxford, UK.
[6] Malenica
S. & Korobkin A.A., 2007. : Some aspects of slamming calculations in seakeeping.,
9th. Int. Conf. on Numerical Ship Hydrodynamics, Ann Arbor, USA.
[7] Ogilvie T.F., 1964. : Recent progress toward the understanding and prediction of ship motions.,
5th. Symp. on Naval Hydrodynamics.
[8] Tuitman J. Aanhold H 2007: Using generalized modes for time domain seakeeping calculations,
22th IWWWFB, Croatia
[9] Zhao R., Faltinsen O.M. & Aarsnes J.V., 1996. : Water entry of arbitrary two dimensional
sections with and without ow separation., 21st Symp. on Naval Hydrodynamics, Trondheim, Nor
way.
A COMPREHENSIVE AND PRACTICAL STRENGTH ASSESSMENT METHODOLOGY
FOR CONTAINER SHIPS TAKING INTO ACCOUNT NON LINEAR AND HYDRO
ELASTIC LOADING
G de Jong and M Huther, Bureau Veritas, France
SUMMARY
The drive for economy of scale effects has fuelled the strong growth in size and capacity of new container ships. The
combination of relatively low hull girder stiffness and high speed (decreased natural frequency of hull girder vibrations
and increased wave encounter frequency) may cause phenomena which are of second order for average size vessels to
become of high importance for these large vessels. This is particularly the case for vibratory structural response and
associated fatigue damage caused by whipping and springing.
Making use of experience feedback with calculations and full scale measurements this paper presents a methodology to
estimate the consequences of these phenomena during the design stage of the vessel. Practical application of the
methodology has been made possible by further developing the Bureau Veritas inhouse hydrodynamic suite
HydroSTAR to take into account hydroelasticity and non linear time domain simulations. The software has been
validated against model tests, while a fullscale measurement campaign is ongoing.
Application of the methodology to ultra large container ships confirms that stress levels are indeed increasing. As a
consequence, a significant increase both hull girder loads and fatigue damage accumulation is found, which can no
longer be ignored when analysing the ship structure. When considering future designs for even larger container ships of
400 m in length and beyond, the effects of whipping and springing will become even more important.
1. INTRODUCTION
Over the past five years the maximum size of Post
Panamax container ships has increased dramatically to
achieve economy of scale effects. Together with the
rapid growth in trade volume and mileage between the
emerging production economies in the Far East and the
established economies (mainly Europe), as well as a firm
belief that market share is the key for future success, an
unprecedented surge in new orders for ultra large
container ships (ULCS, 10 000+TEU) has been created.
According to data published by Clarkson, the world
orderbook for this segment stood at 187 vessels at the
end of the first quarter of 2008. The majority of these
giant boxships are expected to be delivered in 2009
(17%), 2010 (29%) and 2011 (43%).
Although the turmoil in the financial markets and
uncertainty regarding economic growth (possible
slowdown in China, fear for recession in the US, high oil
price) have drastically reduced the amount of new orders
in the first quarter of 2008, there is no reason to believe
that the drive for increasing the vessel size will come to a
halt at the present maximum of 13 300 TEU for a ship
overall length of about 370 m
1
. Therefore, realizing that a
new upheave in container ship ordering would probably
produce designs of well over 400 m in length, it is of key
importance to reflect on the technical challenges
associated with scale enlargement of ultra large boxships.
1
The only exemption is the Eclass series of Maersk,
with an estimated capacity of 15 200 TEU and a length
of 397 m according to data from AXS Alphaliner.
In fact, if the increase in size is achieved through
extrapolation of existing ship dimensions and application
of environmental loads estimated from dimensional
analysis, it may be expected that phenomena which are
of second order for average size vessels can become
important for large vessels. This is particularly the case
for vibratory structural response and associated fatigue.
Experience feedback shows that two dynamic
phenomena affecting the hull girder require specific
attention when considering large container ships:
whipping (transitory hull girder vibrations caused by
hydrodynamic impact at the bow) and springing
(excitation of the first natural modes of vibration of the
hull girder). Indeed, increasing the size of ships with
large deck openings will decrease the natural frequencies
of hull girder vibrations, while an increase in service
speed will result in a higher wave encounter frequency.
Consequently the wave encounter frequency may be in
the same range as the lowest hull girder natural
frequencies, causing important dynamic amplification
effects.
As it is not fully clear to what extent these phenomena
are covered by the implicit and explicit safety margins of
classical rules and regulations, the effect of the dynamic
behaviour of large container ships in a seaway is to be
studied by direct calculations.
In this paper the two phenomena and their impact on ship
structures are analysed. In addition, the methodology
recently developed by Bureau Veritas to assess whipping
and springing effects in large container ship is presented,
as well as the implementation into the inhouse
developed hydrodynamic simulation suite HydroSTAR.
2. WHIPPING
Whipping can be characterised as transitory response of
the hull girder in the first natural mode caused by
hydrodynamic impact at the bow due to reentry of the
foreship into the water (slamming) or violent wave
impact at the stem (slapping), see figure 1.
Figure 1: Container ship experiencing slamming
The first natural mode of the hull girder could be
simulated by a system with one degree of freedom
damped at the corresponding frequency. The associated
forces are extremely high but of short time duration,
similar to an impulse [1] and consequently infer
transitory damped vibrations.
2.1 SHIP STRUCTURAL RESPONSE
Slamming and slapping are known to occur in head seas
when the vertical (relative) motions are highest. The
deformations associated with the hull girder vibration
caused by hydrodynamic impact are superimposed on the
wave induced hull girder deformations, see figure 2.
Figure 2: Full scale measurements of stresses in the deck
of the vessel Marcel Bayard
It has been known for several years that ships with large
deck openings are prone to sustaining damages as a result
of whipping. First of all, the design value of the vertical
bending moment can be exceeded, causing permanent
deformation of the hull girder. Secondly, high frequent
stress cycles can be generated, causing premature fatigue
cracks in structural details [2, 3].
The understanding and analysis of the phenomenon of
exceeding the acceptable bending moment is relatively
straightforward, while the prediction of the occurrence of
fatigue cracks is much more difficult. However, the risk
for fatigue damages has been clearly demonstrated on the
vessel Marcel Bayard. In fact, the full scale measurement
campaign [2] was initiated after the appearance of fatigue
cracks when the vessel was only three years in service.
The study showed that, although the vessel length is only
110 m, the cracks could be explained by frequent
occurrence of whipping as a result of particular operating
conditions for this type of vessel and large deck openings
causing high stress concentrations in the deck.
Therefore, it is important to be able to estimate effects of
whipping on the stresses in the deck during the design
stage. This requires the combined simulation of the
behaviour of the vessel under wave loading and
hydrodynamic impact caused by slamming and slapping.
2.2 CALCULATION METHODS
Direct calculations on a ship in a seaway in order to
determine the extreme hull girder loads and the fatigue
life of structural details have been common practice for
some decades. The methods used are based on the fast
spectral approach and are, after calibration, sufficiently
accurate for the classical checks. As whipping is a non
linear impulsive phenomenon, the (linear) spectral
approach does not longer hold. Instead, time domain
methods have been developed which can solve the
coupling problem between the 3D hydrodynamic
diffractionradiation calculations and the dynamic
deformation of the hull girder modelled as a Timoshenko
beam or a 3D finite element model. For this modal
approach the motions and deformations of the vessel are
represented by a series of six rigid modes and five to ten
dry structural modes [4]. It is to be noted that the linear
hydrodynamic coefficients in the time step equations are
obtained from the classical frequency domain
calculations [5, 6].
There are two methods available for the determination of
the slamming and slapping loads (dynamically exciting
the hull girder), which are essentially 2dimensional: the
Generalized Wagner Model and Modified
Logvinovich Model.
The determination of the structural response to whipping
is possible by combined application of calculation tools
for the rigid ship in the frequency domain and for the
elastic ship in the time domain (non linear). This set of
tools was successfully verified on vessels of 260 m and
360 m of length [7], see figures 3 and 4.
Figure 3: Example of time history of non linear bending
moment time including whipping
Figure 4: Zoom of figure 3 at the instant of hydro
dynamic impact (response types: elastic non
linear, rigid non linear and rigid linear)
The calculations are performed according to the scheme
shown in table 1.
Table 1: Stepbystep procedure for estimating the
extreme hull girder loads
1
Selection of the sea states and their probability of
occurrence for the relevant operational profile of
the vessel (standard: North Atlantic)
2
Calculation of the linear response and
determination of the possible extreme navigation
conditions for the incident waves (ship speed,
wave height)
3
Determination of navigation conditions for which
slamming could occur (wave incidence, ship
speed, wave height)
4
Calculation of the non linear behaviour for
navigation conditions with high risk of slamming
CwpSw CwpSw
=
+
2.3 EXTREME HULL GIRDER LOADS
Observation of the elastic response shows that whipping
increases the extreme values of the vertical bending
moment. Therefore, it is necessary to take this effect into
account for checking the ultimate strength of the hull
girder. The performed calculations [7] and full scale
measurements [2] show that whipping occurs when the
ship is in sagging condition. Therefore, for the most
severe navigation conditions the extreme slamming loads
and consequently the increase of the sagging wave
bending moment are calculated. Full scale measurements
show that the damping of the vibratory response is small
(of the order of =1.5%), which means that whipping is
still important half a wave period later when the vessel is
in hogging condition and that the ultimate strength check
is also to be carried out for this condition
2
. This check
can be done in analogy with the class rules for the
bending moment corrected for the whipping effect as
described [8].
2
Especially since container ships normally operate in
hogging condition, thus providing the extreme hull girder
bending moment.
Application on a 250 m container ship shows that the
increase of the total bending moment can reach 20 to
30% for severe navigation conditions, which is in
agreement with full scale measurements [2]. The
calculations performed for ultra large container ships
show that a similar percentage of increase may be
expected for these ships.
2.4 FATIGUE
Observation of the whipping response (figures 3 and 4)
reveals two effects regarding fatigue:
Increase of amplitude of the wave frequent hull girder
stresses
Creation of high frequent damped stress cycles at the
first natural frequency of hull girder bending
For welded details it is considered that the residual
stresses due to the heat input create a high level of
tension in zones where cracks may be expected, resulting
in a high R coefficient (minimum stress divided by
maximum stress, R > 0.5). For these high R values
variation of the average stress level has little influence on
the fatigue lifetime.
For details without welding there are no residual stresses
and analysis becomes somewhat more delicate. In order
to simplify matters, in first instance the same approach as
for welded structures is applied. As this means that a
high level of residual stress is presupposed, this is a
conservative simplification.
Application of Miners law, which is linear by nature,
allows for separating the low frequency (wave) and high
frequency (whipping) contributions and add both
individual effects to obtain the total result, see figure 5.
Figure 5: Separation of whipping effects for fatigue
calculation
Knowing the stress range of the first whipping cycle and
the damping it is possible to determine the distribution of
cycles and their associated stress ranges as depicted in
figure 6.
Figure 6: Distribution of high frequent, whipping
induced, stress cycles
The associated equations are
( )
=
2
3 j 4
2
1
exp A
j
Swp
(1)
with j designating the n
th
whipping cycle
=
2
1
exp
Sw Cwp
A
2
(2)
The short term distribution is obtained by noticing that
for every single whipping response there is only one
cycle at every level and that there will be a whipping
response associated to every wave cycle, i.e., at each
level the number of cycles is equal to the total number of
waves, see figure 7.
Figure 7: Short term distribution of whipping stresses for
a given wave amplitude
Miners sum can be computed by superimposing the
short term distributions of the stress ranges. Every cycle
can be considered as a response to a regular wave of
constant amplitude, while non linear response
calculations to regular waves can be utilised as well.
For a given wave frequency in head seas (being the
dominant wave condition for slamming to occur) it is
possible to determine the response function for a wave
height H and a threshold wave height H
th
below which
there is no risk of slamming:
( ) H f Sw Cwp = (3)
th
H H >
An overview of the stepbystep calculation method is
presented in table 2.
A study performed for an ultra large container ship
shows that whipping effectively causes amplification of
the hull girder bending stresses, as well as a measurable
contribution to the accumulation of fatigue damage. In
short term head sea condition the correction of the low
frequency stress range can yield a very significant
increase of up to 100% for some navigation conditions.
Considering that whipping disappears for nonsevere sea
states, which are far more numerous than severe sea
states, and that there is no whipping for transverse and
following seas, the total long term increase of the Miner
sum due to whipping is estimated to be of the order of 3
to 5%.
Table 2: Stepbystep procedure for estimating the
whipping induced fatigue damage
1
Determination of the maximum speed of the vessel
as function of the wave height for the considered
wave frequencies and incidence angles (maximum
bearable navigation condition on the bridge)
2
Calculation of the transfer functions and associated
short term response spectra without whipping
3
Determination of the whipping response in head
seas as function of the wave height according to
the wave frequencies of the considered sea states
and taking into account the whipping threshold
4
Approximation of a linear rule giving the
amplitude of the first whipping cycle as a function
of the amplitude of the rigid body response
5
Approximation of the variation on the linear rule
as a function of the wave incidence angle
6
Identification of the short term sea states for which
whipping is likely to occur
7
Correction of the low frequent (wave) stress range
distributions for the whipping effect
8
Calculation of the Miner sum for all short term
low frequent distributions
9
Determination of the high frequent whipping
responses and associated Miner sums for every
short term distribution for which whipping occurs
10
Calculation of total Miner sum taking into account
low frequency contributions corrected for
whipping and high frequency contributions
Following the analysis it can be concluded that the Miner
sum associated to the high frequent part has a relatively
small contribution for the following reasons:
Whipping occurs only in a limited number of sea
states (head seas, wave height above threshold value)
The reduction of the stress amplitudes due to
damping puts the majority of the cycles under the
changing point of the slope of the SN curve used
(due to the random character of the phenomenon a 2
slope SN curve is used for calculating the Miner
sum).
3. SPRINGING
Springing can be described as enforced wave induced
hull girder vibrations. The calculation of the natural
frequency of the two node hull girder bending mode for
vessels with a length up to 350 m shows values above the
frequency of the shortest observed waves. For instance,
North Atlantic wave tables give a lower period Tz=3.5 s
(0.28 Hz), while the natural frequency including added
mass of a 300,000 dwt tanker has been calculated at 0.5
Hz (T=2.0 s). Springing was identified in 1972 during a
full scale study on board a 340 m tanker [9], see figure 8.
The response of the first natural mode of hull girder
vibration, in spite of the high frequency, can be explained
by the fact that for a given wave period the energy is
distributed over a broad bandwidth. Using a Pierson
Moskowitz spectrum, the energy associated to a wave
period of 3.5 s is distributed over a bandwidth of 0.12 to
0.65 Hz, while for a wave period of 8.5 s this is 0.05 to
1.7 Hz.
In addition, for head seas the encounter frequency is
increased by the ship speed as follows:
V
g
2
H
H e
+ = (4)
Figure 8: Response spectrum deck longitudinal stresses
on a tanker in head seas for rigid mode and
including springing [9]
For the referenced tanker, sailing at a speed of 15 kn (7.7
m/s), the energy for a wave period of 3.5 s will be
distributed over a bandwidth of 0.19 to 2.73 Hz, while
for a wave period of 8.5 s this bandwidth is 0.06 to 16 Hz.
These data clearly show that it the natural mode at 0.5 Hz
is effectively excited.
On vessels ranging from 300 to 350 m in length, if
springing is at all visible from the spectral analysis, the
level will be very weak and without effect on the
structural resistance. When the length of the vessel
increases or when the deck contains large openings the
structural rigidity and therefore the first natural
frequency mode decrease. When the ship speed increases,
the encounter frequency increases as well. In both cases
the frequency band associated with springing moves
towards the region where the spectral density of wave
energy is higher, thus amplifying the springing response
(the springing associated peak in figure 8 moves to the
left). This effect was noticed when performing design
analysis for tankers of 550,000 dwt (414 m in length) and
later confirmed during a four year full scale measurement
campaign [10, 11].
When considering the characteristics of a typical ultra
large container ship, with a length of over 360 m, speed
of about 25 kn and large open deck structure, it becomes
clear that springing is a phenomenon to be considered
during the design of the ship structure.
3.1 SHIP STRUCTURAL RESPONSE
A typical time series of measured longitudinal stress in
the deck is presented in figure 9, which shows a high
frequent springing signal of nearly constant amplitude
superimposed on the low frequent (wave) component.
Figure 9: Measurement of deck longitudinal stress with
visible springing effect
Studies and measurements performed on 550,000 dwt
tankers show that springing is sensitive for sea states
with short wave periods. For these periods the maximum
wave height is generally low, see table 3.
Table 3: Maximum significant wave heights for North
Atlantic wave environment according to IACS
Tz (s) 3.5 4.5 5.5 6.5 7.5 8.5
Hs (m) 1.5 4.5 9.5 13.5 14.5 14.5
Due to the weak damping, =1.5% obtained from full
scale measurements [2, 11], the available energy for the
whipping response is limited and the amplitude remains
small compared to the wave frequent response.
Therefore, the influence of springing on the extreme hull
girder loads, corresponding to severe sea states (higher
natural period and very high significant wave height), is
not significant.
On the other hand, the high frequency and character of
the enforced vibration (a quasi permanent phenomenon
in sea states with relatively small waves, which are the
most frequent during the life of the vessel) yield a non
negligible springing induced contribution to the
accumulated fatigue damage. This is confirmed by
experience feedback of the 550,000 dwt tankers [11].
3.2 CALCULATION METHODS
Taking into consideration that the springing induced
deformations are much smaller than the rigid body
motions of the hull girder, it can be assumed that the
problem can be treated in a linear sense and that the total
response can be obtained by summation of the separately
calculated rigid and elastic contributions [10]. Therefore
it is possible to apply the linear tools developed for
hydrodynamic and structural response calculations [13].
First of all, a classical calculation can be performed to
determine the linear hydrodynamic coefficients, in
particular the added mass and the transfer functions of
the rigid body responses (figure 10). Secondly, the
natural frequencies for the elastic hull and the associated
transfer functions for springing are calculated (figure 11).
Figure 10: Calculation of rigid body transfer functions
Figure 11: Calculation of transfer function for springing
in torsional mode (elastic hull)
In this way the transfer functions for the quasi static,
springing and total response are obtained for the selected
navigation conditions (figure 12).
=
+
=
+
Figure 12: Transfer function of peak stress in hot spot
(welded detail) in a container ship (springing,
wave and total)
The calculations are performed according to the scheme
shown in table 4.
Table 4: Stepbystep procedure for springing analysis
1
Selection of the sea states and their probability of
occurrence for the relevant operational profile of
the vessel (standard: North Atlantic)
2
Calculation of the linear response and
determination of the possible extreme navigation
conditions for the incident waves (ship speed,
wave height)
3
Determination of navigation conditions for which
springing could occur (wave incidence, ship speed,
wave height)
4
Calculation of the linear elastic behaviour for wave
incidences where springing occurs
3.3 FATIGUE
Observation of the springing response for a given sea
state (figure 9) shows that, on top of the wave frequent
response, a time series of stress cycles with nearly
constant amplitude at the first natural frequency of the
hull girder is generated.
The calculation of the short term fatigue damage
accumulation requires the determination of the
distribution of the high frequent stress cycles. The exact
method for achieving this would be Rainflow counting.
In order to do that a time history needs to be available,
which is impractical for the spectral approach. By
considering the same hypothesis as for the whipping and
the using the linearity of Miners rule, it is possible to
separate both the high frequent and low frequent
contributions as depicted in figure 13 superimpose them
to obtain the total result.
As shown for whipping, the range of the wave frequent
stress cycles needs to be corrected for the springing
effect as well. As only the navigation conditions with a
risk for the occurrence of springing need to be considered,
it is important to take into account speed reduction in
case the wave height increases, because this can cancel
the springing effect altogether.
Figure 13: Separation of springing effects for fatigue
calculation
For each of the contributions corresponding to a signal
with narrow banded spectrum, the fatigue accumulation
can be computed according the methodology for wave
frequent fatigue assessment [12]. The short term
distribution of the stress range S
i
can be described by a
Rayleigh distribution, where the coefficient m
0
is
determined by the area under the response spectrum:
( )
= >
0
2
i
i
m 8
S
exp S S prob (5)
For a given structural detail with known SN curve it is
possible to calculate the Miner sums separately and add
them, taking into account the probability of occurrence of
each short term navigation condition (wave height and
period, vessel speed and wave incidence angle):
springing corrected , wave total
D D D + = (6)
An overview of the stepbystep calculation method is
presented in table 5.
Table 5: Stepbystep procedure for estimating the
springing induced fatigue damage
1
Determination of the maximum speed of the vessel
as function of the wave height for the considered
wave frequencies and incidence angles (maximum
bearable navigation condition on the bridge)
2
Calculation of the transfer functions and associated
short term response spectra without springing
3
Determination of the frequency bands with high
energy density within the wave spectra for the
considered navigation conditions
4
Calculation of the natural frequencies of the hull
girder, taking into account the added mass
5
Identification of the short term sea states for which
springing is likely to occur
6
Calculation of the transfer functions for springing
and associated short term response spectra for the
considered sea states
7
Calculation of total Miner sum taking into account
low frequency contributions corrected for
springing and high frequency contributions
The method has been checked at short term level against
the considered exact method, i.e. the Rainflow counting
of the stress time history. To do so, a time history has
been generated from the short term 2peak response (see
figure 12) using specific software. Then a Rainflow
counting according to the AFNOR standard procedure
has been performed and the Miner sum calculated.
Two typical cases have been identified. For sea states
with Tz above 5 s the low frequent wave stress is
dominant over the high frequent springing, while for
smaller Tz the situation is reversed, see figure 14.
Figure 14: Stress time histories for Tz =6.5 s (above)
and Tz =3.5 s (below) (First modal period T
1
=2.15 s)
The comparison between the methods confirms that the
proposed procedure provides acceptable short term
Miner sum values. However, the correction factor for the
low frequent wave stress range has to be adjusted to the
typical cases to prevent underestimations.
A study performed for an ultra large container ship
shows that, due to the large deck openings, the first
natural mode of the hull girder is a torsional mode, which
is followed by a number of vertical bending and torsional
modes, see table 6.
Table 6: Wet natural hull girder modes and frequencies
ultra large container carrier
Hull girder vibration mode f (Hz)
1node torsional 0.37
2node vertical bending 0.47
2node torsional 0.50
3node vertical bending 0.90
An investigation of the navigation conditions with a risk
for springing occurrence showed that the torsional modes
are to be considered for wave incidence angles of 120
and 90 degrees, while the flexural modes are to be
considered for wave incidence angles of 180 and 120
degrees (180 degrees corresponding to head seas).
The fatigue lifetime of a structural detail in the deck
experiencing high dynamic loading has been calculated
according applying North Atlantic scatter diagram, see
figure 15.
Figure 15: Fatigue analysis of hatch corner detail of an
ultra large container ship
The springing response has been calculated for all wave
heights and periods of the wave data table until a
significant wave height of 5 m. For greater wave heights
it is assumed that the vessel speed will be reduced and,
consequently, that the shifted encounter frequency will
yield only irrelevant springing response.
The calculation of the short term Miner sum including
springing shows that, compared to the situation
excluding high frequent springing part, an increase in
fatigue damage of up to a factor 3 can be reached for
small period sea states. However, as small period sea
states represent only a limited part of the long term total
Miner sum, and taking into account that springing does
not occur for transverse and following seas, the total long
term increase of the Miner sum due to springing is
estimated to be of the order of 4 to 10%.
A sensitivity study shows that if the maximum
significant wave height is increased from 5 to 6 m, the
springing related Miner sum increases by 20%. However,
the effect on the total Miner sum is of second order.
4. HYDROSTAR
The described methodology to assess the effects of
whipping and springing during the design stage of ultra
large container ships has been made practically
applicable within the Bureau Veritas inhouse
hydrodynamic simulation suite HydroSTAR, which has
been developed and validated for more than 20 years to
support new technological challenges. Originally started
as a linear 3D diffractionradiation scheme for floating
offshore units, today HydroSTAR is a multi disciplinary
hydrodynamic package widely used in the marine and
offshore industry. Typical issues like irregular
frequencies, forward speed effect, roll damping,
automatic mesh transfer, second order wave loads and
multi body interactions are accounted for.
In order to simulate whipping and springing hydroelastic
and non linear effects are to be accounted for. Hydro
elastic effects are handled using the modal approach, for
which the equation of motion can be written as [13]:
[ ] [ ] ( ) [ ] [ ] ( ) [ ] [ ] ( ) { }{ } {
DI
F C k b B A m = + + + + i
2
} (7)
where [m] is the structural mass, [b] the structural
damping, [k] the structural stiffness, [A] the
hydrodynamic added mass, [B] the hydrodynamic
damping, [C] the hydrodynamic restoring, {} the modal
amplitudes and {F
DI
} the modal hydrodynamic excitation.
The hydroelastic coupling is visualised in figure 16.
Figure 16: Hydrodynamic mesh following the structural
deformation of the hull (first modal period)
The non linear time domain simulations make use of the
procedure proposed by Cummins (1962), for which the
equation of motion can be written as follows [13]:
[ ] [ ] ( ) ( ) { } [ ] [ ] ( ) ( ) { }
( ) [ ] ( ) { } ( ) ( ) { t t d t t
t t
t
0
Q F K
C k A m
+ = +
+ + +
&
& &
}
(8)
where [A
d t cos B
2
t K
0
ij ij
= (9)
This method enables the introduction of non linear
components in the excitation forces F(t) and Q(t) [13].
As an example, for the calculation of the extreme hull
girder bending moment including whipping effect a
(weakly) non linear FroudeKrylov model is applied, see
figure 17.
Figure 17: Non linear time domain simulation using
FroudeKrylov approach (loads on finite
element model)
The resulting overall computation scheme of the software
called HydroSTAR++is depicted in figure 18.
Figure 18: Computation scheme of HydroSTAR++for
hydrostructural simulation of container ships
4.1 MODEL TESTING
In order to validate and calibrate the methodology for the
hydroelastic and time domain simulations, several
model tests have been carried out and compared to the
calculation results. In figure 19 depicts the torsional
response of an elastic barge, showing good agreement
between measurements and calculation results [13].
Figure 19: Elastic barge in torsion (impression of test left,
RAO of torsional angle right): comparison
between experiments and calculation results
In figure 20 the initial deformation and the time history
of the vertical displacement after releasing it from the
original position are depicted. Again, there is good
agreement between calculations and experiments.
Figure 20: Elastic decay test (initial position left, time
history right): comparison between
experiments and calculation results
4.2 FULL SCALE MEASUREMENTS
Within the scope of the J oint Industry Project
Lashing@Sea full scale measurements are carried out on
board the BV class 9,400 TEU container ship CMA
CGM Rigoletto. Motion and acceleration sensors have
been distributed over the hull and two cross sections have
been equipped with strain sensors to record the dynamic
response in terms of accelerations and internal hull girder
loads.
In order to study the behaviour of stacked containers and
lashing equipment a dedicated measurement container
has been installed above the aft deck. An example of a
time series of measured accelerations in the container is
shown in figure 21.
Figure 21: Time series of full scale measurement in
transverse accelerations in container
(normalised)
The results of the measurement campaign can be used to
study the real life behaviour of large container ships
and to further validate and calibrate the developed
numerical tools.
5. CONCLUSIONS
An increase in the size and speed of container ships
importantly changes the behaviour in a seaway.
Phenomena already known for smaller size vessels can
be amplified to reach levels high enough to have an
impact on the design of the hull. This is particularly true
for the cases of whipping and springing.
The springing phenomenon was first studied during the
construction of the 550,000 dwt tankers in 1975, for
which time domain calculations have been performed. At
that time the whipping phenomenon was also known,
specifically in relation to ships with large deck openings,
but computer simulation was not yet feasible. Recent
developments in computer computation power and
efficient algorithms allow performing non linear hydro
structural calculations in the time domain.
A methodology has been proposed to estimate the effects
of whipping and springing on the extreme hull girder
loads and fatigue lifetime of structural details. In order to
be able to study the fatigue related effects during the
design stage of a vessel, it is assumed that the low
frequent and high frequent contributions can be dealt
with separately. This simplification importantly reduces
the required amount of non linear time domain
simulations.
The application of the methodology has been made
possible by further development of the Bureau Veritas
inhouse hydrodynamic simulation suite HydroSTAR to
solve hydroelastic and non linear time domain problems.
Extensive validation has been done by model testing,
while the ongoing full scale measurement campaign will
provide additional valuable feedback.
Application of the methodology to ultra large container
ships confirms that the stress levels are indeed increasing.
As a consequence, a significant increase both hull girder
loads and fatigue damage accumulation is found, which
can no longer be ignored when analysing the ship
structure.
Whipping effects can cause an increase of about 20% in
total vertical bending moment, for which the hull girder
ultimate strength is to be checked.
For current container ship designs of about 350 m with
extremely large open deck structure, whipping can cause
an increase in short term fatigue damage accumulation
by a factor of up to 100% for some head sea navigation
conditions. The increasing effect on the long term
whipping induced fatigue damage accumulation is
estimated to be in the range of 3 to 5%.
With regard to springing, as a consequence of the
flexibility of the hull girder and high vessel speed,
current designs of about 350 m in length show an
increase in short term fatigue damage accumulation by a
factor of up to 3 for small period sea states, which is
superior to the referenced tankers of 414 m. The
increasing effect on the long term springing induced
fatigue damage accumulation is estimated to be in the
range of 4 to 10%.
It is expected that full scale measurements results will
provide valuable insight into the real dynamic behaviour
of large container ships, as well as the accuracy of the
calculations. In addition, further development of the
methodology and hydrostructural computation schemes
are ongoing and are expected to further enhance
prediction capabilities.
When considering future designs for even larger
container ships of 400 m in length and beyond, it is
certain that with similar hull structural design the
increase of fatigue damage accumulation due to
whipping and springing will become even more
important. By realizing that springing is a forced
vibration phenomenon with little damping and that the
whipping response is of non linear character, the
dynamic phenomena and associated consequences will
have to be analysed with greater attention.
6. ACKNOWLEDGEMENTS
The models and results presented in this paper require
strong interaction of many technical competencies.
Therefore, the authors would like to thank all the people
involved in the research and application of the developed
methodologies, in particular . Malenica, F. Mauduit, N.
Germain, F. Bigot, F.X. Sireta, S. Maherault, E. Stumpf,
V. Bouitillier, J . Henry and G. Parmentier.
7. REFERENCES
1. Huther M, Ket N, Water impact risk estimation
during ship design, Ship Behaviour at Sea,
International Symposium on Ship Hydrodynamics
and Energy Saving, ISSHES, Madrid, September
1983
2. Osouf J , Etude exprimentale du comportement
dynamique dur houle du navire cblier Marcel
Bayard, Revue NTM, 1973 (in French)
3. Drummen I, Moan T, Storhaug G, Moe E,
Experimental and full scale investigation of the
importance of fatigue damage due to waveinduced
vibration stress in a container vessel, Design &
Operation of Container Ships, RINA, London,
November 2006
4. Malenica , Molin B, Remy F, Senjanovic I,
Hydroelastic response of a barge to impulsive wave
loads, 3
rd
International Conference on
Hydroelasticity, Oxford, 2003
5. Cummins WE, The impulsive response function
and ship motions, Schiffstechnik, 1962
6. Ogilvie TF, Recent progress toward the
understanding and prediction of ship motions, 9
th
International Conference on Numerical Ship
Hydrodynamics, Ann Arbor (USA), 2007
7. Tuitman J , Malenica , Some aspects of whipping
response of container ships, 23
rd
IWWWFV, J egu
(Korea), 2008
8. Bureau Veritas, Rules for the Classification of Steel
Ships, November 2007
9. Planeix J M, Wave loads, A correlation between
calculations and measurements, Revue International
Shipbuilding Progress, August 1972
10. dHautefeuille B, Huther M, Baudin M, Osouf J ,
Calcul de springing par quations intgrales, Revue
NTM, 1977
11. Huther M, Osouf J , SEFACO, Four years of
experience in ship hull girder stress monitoring,
MARINTEC, Shanghai, October 1983
12. Huther M, Beghin D, Mahrault S, A fatigue
strength guidance note for welded ship structure
assessments, IIW/IIS doc XIII11174298/XV992
98, 1998
13. Malenica , Hydro structure interaction in
seakeeping, International Workshop on Coupled
methods in numerical dynamics IUC, Dubrovnik,
September 2007
8. AUTHORS BIOGRAPHIES
Gijsbert de Jong holds the current position of product
manager at Bureau Veritas. He is responsible for the
international business development in the field of
container ships. Gijsbert joined Bureau Veritas in 2001
after obtaining an MSc in Naval Architecture & Marine
Engineering from Delft University of Technology.
Before moving to the sales and marketing management
team in Paris, he has worked as hull surveyor and
department manager in the plan approval office in
Rotterdam.
Michel Huther is a graduate Mechanical Engineer
(Ecole Centrale de Lyon, 1965) and Naval Architect &
Marine Engineer (Ingnieur Civil du Gnie Maritime,
1967). Michel joined Bureau Veritas in 1969 and has
worked as research engineer and department head for the
ship research and rule development department, with
many contributions to European R&D projects. In 2000
he was promoted deputy to the director of the marine
technical management. After his retirement in 2003
Michel is holding the position of technical advisor to the
Bureau Veritas marine technical management. He is a
member of SNAME, ATMA & SF2M and has published
over 200 scientific and technical papers on ship structural
behaviour and safety.
STRUCTURE
Managing structural fatigue onboard ships requires either fitted
approaches for existing fleet and new tools to prevent the emergence of
new fatigue sources such as whipping or springing. The papers in this
section focus on risk based maintenance versus crack propagation and
prediction of fatigue loads resulting from hydrodynamic structural
interaction in large containerships.
Bulletin Technique  Bureau Veritas 2008
ChapitresBT2008:ChapitresBT2008 05/10/09 20:54 Page2
1
SIMULATION OF THE BEHAVIOUR OF FATIGUE CRACKS: A TOOL FOR
INSPECTION DECISION MAKING ON A SHIPS DECK BEAM
V. Boutillier, M. Serror, G. Parmentier, Bureau Veritas, 17 bis place des Reflets,
La Dfense 2, 92400 COURBEVOIE
ABSTRACT
To assess the condition of ships and FPSOs, it is necessary to fit the inspection process to the safety
targets. By using the knowledge gained from the inspection data, failure of the ships structural
components may be avoided. Crack propagation is an important issue to consider when devising an
appropriate inspection plan; this is particularly true for ships and FPSOs that operate in regions with high
sea states. On some types of ships, it can be very difficult to perform inspections due to certain structural
arrangements, for example wall fascias that hide the structural members, and the time required to perform
this task leads to heavy costs. Thus, it is important to know the exact behaviour of how cracks propagate
in order to determine, as far as possible, if the inspection process may be delayed.
The original feature of the methodology is the use of Line Spring Method to calculate the stress intensity
factors. Based on this approach, we have developed a probabilistic analysis using Monte Carlo
simulation.
The main benefit of this assessment technique is the added flexibility in the development of an inspection
plan that best suits the ship or FPSO considered. This approach may help operators and engineers to
make decisions for inspections, repairs and maintenance processes, in order to have the lowest cost
whilst complying with the required safety targets. Finally, an example is provided on a ships deck beam,
which shows us the effects of inspection parameters on the probability of failure and probability of
detection.
1. INTRODUCTION
The aim of a fatigue analysis on ships or offshore
structures is to evaluate the damage due to cracks
propagation, leading sometimes to the failure of
parts of the structure. Ship structures are submitted
to cyclical forces, such as wave actions, as well as
loading and unloading sequences. On some types
of ships, it can be very difficult to perform
inspections due to certain structural arrangements,
(for example wall fascias that hide the structural
members), and the time required to perform this
task leads to heavy costs.
Therefore, in order to help operators and engineers
to make decision for inspections, maintenance and
repairs, a probabilistic crack propagation approach
based on a new crack propagation method has
been developed.
This probabilistic approach is based on SAPHIRS
[1] which evaluates crack propagation and a
Monte Carlo simulation for the probabilistic part.
That methodology gives us probability of failure.
2. SAPHIRS
2.1 METHODOLOGY
The aim of the methodology is to forecast as
accurately as possible the fatigue crack initiation
and the fatigue crack growth in ships or offshore
structures components that have been submitted to
fluctuated and repeated loads. These loads can be
represented by a combination of unitary loads or
modes if the response of the component is linear
elastic. To take into account the discontinuity of
the presence of a crack which induces non
linearity effects, the mathematic models generally
use the contour integral evaluation related to the
energy release rate. That action implies re
meshing after each propagation step. Remeshing
is very time consuming. The stress intensity factor
is one of the most important parameter
responsibilities of the crack propagation. It
depends on the notch geometry and on the stresses
on the crack lips. It is well known that the fatigue
crack propagation varies exponentially with it.
2
Therefore, with the use of both LineSpring and
the stress intensity factors to determine the stress
field, success is achieved accompanied with very
easy implementation. This method allows taking
into account the stiffness variation during crack
propagation, impossible with analytical solutions
or finite element methods without remeshing.
2.2 POSITION OF THE CRACKS
The finite element model must be carefully
performed with shell elements including all the
directions of loads that the structure encounters.
This first and very important step establishes stress
cartography. From these results with the use of the
concept of hot spot, we are able to define the
position of one or several cracks. The experience
showed that the position of the cracks is more
commensurate with the structural geometry, rather
than the forces directions. However, the software
leaves open the ability for a crack possibly
growing due to a forces direction, and somewhere
else, growing from another forces direction. But
most importantly is the software which has the
ability to consider the interactions between several
cracks (Figure n1).
Figure n1: Two cracks
Our process needs to know a priori the crack path.
So it has to be determined [2], and the simplest
method is to use the principal stresses directions
(Figure n2). The main assumption consists of
setting the crack path in a direction perpendicular
to the maximum principal stresses whose
directions are the eigenvectors of the stress tensor.
So the first step is to determine the stress field of
the component to be able to draw the path of the
hot spots.
Figure n2: Principal stress directions
Often, the hot spots appear at an intersection of
two sheets, and the crack must be defined at the
weld toe position, even if no weld has been
modeled. Wherever the hot spots take place, we
build one or several partthrough cracks, by one or
several lines of coupled nodes.
2.3 THE LINE SPRING METHODOLOGY
The Line Spring model was introduced in 1972 by
Rice and Levy [3] to estimate stress intensity
factors due to tension and bending in large plates
containing partthrough surface cracks. He
essentially reduced the threedimensional problem
to a twodimensional one, in mode I. This model
has been extended to the modes II and III by
Desvaux [4]. The idea was to analyze the part
through crack as a several single edged specimen,
the cracked structure part (Figure n3).
Figure n3: A several single edge specimen
The crack is defined as a line of couple of nodes,
with same coordinates, and separated. The both
nodes simulate the crack, each node located on
each crack lip.
The LineSpring model is based on the fact that
there is a relationship at each point along the cut,
Two cracks
Principal stress
directions
3
between local displacements and the loadings, the
compliance coefficients. Then, from the point of
view of the plate, the surface crack is represented
by a throughcrack with a continuous distribution
of generalized springs connected across the line of
discontinuity, the Line Spring (Figure n4).
Figure n4: LineSpring
After the crack has been defined as upper, the
keystone of the software can be set up. The
compliance coefficients, the S
ij
pq
matrix, are
defined by the additional relative displacements
and rotations due to the cracks presence. Thanks
to a finite element calculation, the second ones
being the displacements of the crack lips, d
i
p
are
determined for each couple of nodes i, in the
direction p, resulting in an unitary force F
j
q
,
exerted on the couple node j in the direction q, in
order to build the compliance matrix. It is the
same to evaluate the displacements of the crack
lips on each couple of node i, d
i
p
due to external
loads, F
ext
, which are applied on the fullycracked
component. These loads may be unitary and
exerted on any point on the structure and
according to any direction. The displacements will
be then worked out by an easy linear combination.
Thus, for the desired crack shape, the resolution
kernel of the software SAPHIRS, is to find loads
{F}, and {d
i
p
} to solve:
d
i
p
 d
i
=  S
ij
pq .
F
j
q
(1)
This first equation lets us verify the loads along
the virtual crack, when there are no cracks yet, i.e.,
d
i
p
=0.
The stresses on the intact structure are again
evaluated from the equation:
d
i
= S
ij
pq .
F
j
q
(2)
At the end of this step, all the behavior of the
structure followed by the behavior of the crack is
condensed in the S
ij
matrix.
2.4 CRACK INITIATION
The propagation crack process always begins by
the crack initiation. This step is fundamental for
the next propagation step. The shape of the crack
front is drawn by the stress field distribution. If the
stress gradient is important, the initiation and then
the propagation will occur in the hot spot and its
neighborhood. If the stress gradient is small, the
initiation will go off gradually, and the crack will
extend first along the surface. It leads to a multi
initiation of the crack. All these phenomena are
taken into account in the software. In this phase,
the damage is cumulated cycle after cycle, either
by a local approach or by an SN curve approach.
2.4 (a) Local approach
The module CALEND stems from Petitpass [5]
works. It uses the local stresses determined by the
precedent step. For designing welded components,
it allows you to take into account stress
concentration factors K
t
to evaluate the stress
tensor at the weld toe. Residual stresses can be
included, too. As the local stresses are often
greater then the yield stress, a plastic correction
has to be done. Then a fatigue stress criterion is
calculated on different planes of the stress tensor
(using Dang Van criterion [6]) and damage is
accumulated (using Miners rule and Basquins
law) on extracted cycles (using Rainflow
counting). Finally, damage is taken as the
maximum damage obtained on all planes of the
stress tensor, and the crack initiation occurs when
damage at a node is given and overestimated
critical value.
In each point where the damage is an upper
critical damage, it is assessed as an initial flaw,
generally the 1/10
th
of the sheet thickness. After
that, the propagation may occur.
2.4 (b) SN curve approach
This treatment of the initiation phase is built on
the initiation SN curves, which supply the
probability to get an initial flaw, after N
i
solicitations under the stresses, S
ij
, on the couple j.
This probability is supposed following a Weibull
Lower crack tip
Line Spring Upper crack tip
d dd d
i
p
4
distribution, with two parameters a and b.
) ( exp 1
1 ij
n i
i
j
KS
N
P
=
=
=
(3)
The crack initiation occurs when the probability at
a node is over a given critical value.
Those critical values may be different if the non
homogenous surface weld toe is taken into account
This step allows initiating the crack by laying an
initial flaw with a given size, a
0
(Figure
n5).While the crack growths on this node (Figure
n6), the other nodes will initiate at their turn. So,
this phase allows the crack propagating through
the thickness of the sheet, but also, in the direction
parallel to the sheet. The crack propagation
diagram may be summarized as follows:
Figure n5: MultiInitiation
Figure n6: Propagation through thickness
2.5 CRACK PROPAGATION: PART
THROUGH THICKNESS
The crack grows from each point where the
initiation has occurred. The Tada coefficients A
ij
,
depend only on the ratio l/h, (l: crack depth and h:
plate thickness), and vanish when l=0.These
coefficients set up relations between the forces
acting upon the crack lips and the crack lips
opening displacements.
d
i
p
= A
ij
pq .
F
j
q
(4)
Then, for any crack depth, it is easy to find the
displacements d
i
, and the forces F
j
q
, by solving
both (1) and (4). Now the Irwins (1967) relation
in potential energy accompanying a variation of
crack depth and the stress intensity factor are
employed as a basis for calculating the stress
intensity. The stress intensity factors, first in mode
I, K
I
from Tada (1973) [7] and then extended to
the modes II and III, K
II,
K
III
by Desvaux (1983)
may be expressed as follows, in every cracked
point, j:
)
6
( ) (
2
, ,
h
g M
h
g N
h j K
M I j N I j
I
+ =
(5)
) , ( ) , ( ) (
)) , ( (
2
) 2 / tan(
/ 24 ) , (
)) , ( (
2
) 2 / tan(
2 ) , (
T j K Q j K j K
h l f
h
h l
h T T j K
h l f
h
h l
Q Q j K
II II II
II
II
+ =
=
=
(6)
,
) (
II III
g h j K = (7)
) 1983 (
,
), , (
:
:
:
:
:
:
5
3
1
4
2
Desvaux by given
x II
g h l f with
F moment twisting antiplane T
F shearforce antiplane Q
F s shearstres transverse
F moment bending M
F tension N
with
Then, for every crack depth on each node
belonging to the crack line, the three stress
intensity factors are evaluated according to the
loads acting on the structure. In the structures, we
study the mode I because it is the most
preponderant crack propagation. Then, the crack
propagation rate, in the software SAPHIRS
follows the Paris law:
, ) ( /
m
s I
K K C dN dl = (8)
With Ks: the threshold stress intensity factors,
and
K
I
=K
I,max
 K
I,min
The crack propagation through the plate thickness
goes cycle after cycle until a failure criterion
occurs. After every propagation step, an update of
the defect size is performed, and the equations (3)
and (4) are solved again.
The software proposes three failure criteria:
Ductile criterion
Brittle criterion
Critical size
When one of these criteria is realized, the ligament
Plate
thickness
Flaws on the surface
Plate
thickness
5
breaks and the crack becomes a through thickness
crack.
2.6 THROUGH CRACK BOX: THE CRACK
BOX TECHNIQUE
It is important to emphasize that the stress
intensity factors calculated with Tada coefficients
are available for cracks propagation in the depth of
a plate. When the ligament rupture occurs, the
crack propagation extends in a longitudinal
direction parallel to the plate (Figure n7).
Figure n7: Three crack configurations
The Crack Box idea, developed by Lebaillif
(2005) [8], is set to use the technique of sub
modeling: the displacements of nodes located in
the neighborhood of the cracks, determined in the
same way of the crack lips displacements, are used
to drive a 3D submodel of the structure, including
the crack tip which is meshed using fracture
mechanics elements (Figure n8).
Figure n8: Crack Box Technique
Thus, the stress intensity factors are previously
calculated at all the crack tip nodes of the sub
model, with unitary displacements applied at its
border. Next, the three stress intensity factors are
calculated using a linear combination with the
displacements in each mode.
3. PROBABILISTIC APPROACH
3.1 METHODOLOGY
A sensitivity study has helped us to determine the
random parameters (such as material
characteristics, initial crack length, etc.) which
govern crack propagation. These parameters are
described here after respectively with their density
function [9].
Random
parameter
Density function
Yield stress
minimum Weibull law
 355 =
threshold
MPa
 411 = u
 1 . 2 = k
Ultimate
strength
minimum Weibull law
 490 =
threshold
MPa
 558 = u
 1 . 2 = k
Paris law
parameter:
Kthreshold
normal law
 Mean = 10 MPa m
Standard deviation=2.5 MPa m
Toughness
minimum Weibull law
 40 =
threshold
MPa m
 106 = u
 3 . 3 = k
We have to notice that we didnt take loads as
random parameter.
For each simulation, we observed the detail
behaviour, considering the ship used 80% of 30
years in navigation. During this span time, we
inspect the ship considering one inspection plan.
Each plan is described by its type and the time
interval between each inspection.
SAPHIRS evaluates the crack propagation and
gives the crack length at each step.
At the end of each simulated inspection, two
possibilities exist. The first one, no crack is
detected, so the simulation goes on. The other
possibility is, a crack is detected. Therefore, we
repaired the structure (without redesign). Thereby,
a new detail is done, so we make a new Monte
Carlo simulation in order to give new material
Plate
thickness
A through thickness crack
u
j
6
characteristics for the considered detail. If a
failure appends during the simulation, we do not
repair, we stop the simulation.
Inspection type is defined by a probability of
detection (pod). This pod helps us to know either
the surveyor detect the crack or not during the
inspection. In fact, thanks to this pod, we were
able to determine the probability the surveyor had
to detect this crack length. Next, we choose a
random value between 0 and 1, if this value is
lower than the probability to detect the crack, that
means we detect it, however if the random value is
greater than the probability to detect the flaw, the
crack propagation goes on until the next inspection
or a failure.
We have decided to look at two different
inspection types. Therefore, we defined two pod
represented by two lognormal laws whose
parameters are defined below:
Visual inspection:
 mean = 20
 standard deviation = 10
NDT (Non Destructive Testing)
inspection:
 mean = 2.87
 standard deviation = 1.03
The total number of simulation depends on the
wanted accuracy.
After each inspection, probability of failure and
confidence interval are calculated.
Probability of failure, p, is evaluated by:
With: n: the total number of simulations
k: the total number of failure
Confidence interval is calculated by:
( )
2 1 2 / 1
,
F
: is a FisherSnedecor law
1 : is a confidence interval
This confidence interval is used to limit the
parameter estimation. In our case, this interval
corresponds to 95%.
The following figure shows us the program
mechanism.
Figure n9: program mechanism
4. A SHIPS DECK BEAM SIMULATION
4.1 HYPOTHESIS
The loading histories are all the sea states
significant heights encountered by the ship during
its life. Each sea cycles is supposed to last 6 hours,
i.e. 1 800 cycles. The diagram below shows the
significant height distribution state, HS1.
New random
simulation
Monte Carlo
simulation
Planning inspection
Detail description
Crack propagation
SAPHIRS
Monte Carlo
simulation
Visual Crack
Repair required
No Visual Crack
The crack keeps going
Yard inspection
Inspection
Critical length crack
reached
Failure
n
k
p =
n k si p = =1
sup
( ) ( ) ( ) ( )
( ) ( ) ) ( 2 ), 1 ( 2 1
2 , 1 2 1
5 . 97
5 . 97
sup
k n k F k k n
k n k F k
p
+ + +
+
=
0 0
inf
= = k si p
( ) ( ) ) ( 2 ), 1 ( 2 1
5 . 97
inf
k k n F k n k
k
p
+ + +
=
7
Sea states HS1
0
2
4
6
8
10
12
1 32 63 94 125 156 187 218 249 280 311 342 373 404 435 466 497
Hs
n
u
m
e
r
o
u
s
o
f
t
h
e
s
e
a
s
t
a
t
e
s
Initiation after the step
61
73200 cycles
Figure n10. Significant heights of loading histories
4.2 EFFICIENCY OF INSPECTION PLAN
ON FAILURE PROBABILITIES
Time interval influence
On these simulations, we show the time interval
influence on the failure probabilities.
The first one is visual inspection every 6 year, the
second one is annual visual inspection and finally
the last one is visual inspection every 6 month.
Table 1: failure probabilities for visual
inspection plan
Failure
Probabilities
Lower
boundary
Mean
estimation
Upper
boundary
Visual
Inspection
Every 6 year
3.15*10
1
3.93*10
1
4.76*10
1
Visual
Annual
Inspection
4.71*10
3
1.17*10
2
2.39*10
2
Visual
Inspection
Every 6 month
1.41*10
2
2.5*10
2
4.09*10
2
The following diagrams show the failure
probabilities evolution after each simulation of
80% of 30 years.
1.00E03
1.00E02
1.00E01
1.00E+00
3 24 45 66 87 108 129 150
Figure n11. failure probabilities  visual inspection every 6
year
1.00E03
1.00E02
1.00E01
1.00E+00
3 24 45 66 87 108 129 150 171 192 213 234 255 276 297 318 339 360 381 402 423 444 465 486 507 528 549 570 591
Figure n12. failure probabilities  visual annual inspection
1.00E03
1.00E02
1.00E01
1.00E+00
3 54 105 156 207 258 309 360 411 462 513 564
Figure n13. failure probabilities  visual inspections every 6
month
The upper diagrams show that we run only 150
simulations for visual inspection every 6 year
against 600 simulations for visual annual
inspections and visual inspection every 6 month.
Nevertheless, we may notice that the confidence
interval is rather large for inspection every 6
month or every year than for inspection every 6
year.
8
The reason is that during annual inspection, the
number of failure decreases and therefore the
confidence interval is more difficult to evaluate. It
is the reason why we carry out more simulations
when inspections are close together than for every
6 year inspections.
Inspection type influence
On this calculation, we show the influence of
inspection type. We changed visual inspections by
NDT inspections.
As we may predict, NDT inspections give better
results than visual inspections.
Table 2: failure probabilities for NDT
inspection plan
Failure
Probabilities
Lower
boundary
Mean
estimation
Upper
boundary
NDT
inspection
Every 6 year
2.34*10
1
3.07*10
1
3.87*10
1
NDT
Annual
inspection
6.89*10
3
1.50*10
2
2.83*10
2
NDT
inspection
Every 6 month
1.82*10
3
6.67*10
3
1.70*10
2
Here again, it can be observed, as for visual
method, that if we carry out every 6 month
inspections, it didnt provided better results than
annual inspections.
Failure results depend either on the probability of
the structure to resist to fatigue and crack
propagation and on the performance of inspection
process.
The selected detail was known as ill designed for
the applied loads. Therefore, the question was how
a reinforced inspection planning would improve or
not, the structural reliability, and show the limit of
inspection versus inspection methods and
inspection periods.
These failure probabilities show us that the time
interval has much influence than the type. In fact,
reducing the time interval will supply better results
than changing the inspection type.
4.3 EFFICIENCY OF INSPECTION PLAN
FOR DETECTING CRACKS BEFORE
FAILURE
We evaluated two different probabilities of
detection.
The first one is the probability to detect a flaw
leading to a repair (table 3 & 4).
The second one is the probability to detect a crack
before a failure (table 5 & 6).
Table 3: Probabilities to detect a flaw leading to
a repair for visual inspection plan
Probability to
detect a flaw
leading to a
repair
Lower
boundary
Mean
estimation
Upper
boundary
Visual
inspection
Every 6 years
1.98*10
1
2.31*10
1
2.66*10
1
Visual
inspection
Annual
8.22*10
2
8.65*10
2
9.10*10
2
Visual
inspection
Every 6 month
4.86*10
2
5.11*10
2
5.36*10
2
Table 4: Probabilities to detect a flaw leading to
a repair for NDT inspection plan
Probability
to detect a
flaw leading
to a repair
Lower
boundary
Mean
estimation
Upper
boundary
NDT
inspection
Every 6 years
2.51*10
1
2.48*10
1
3.19*10
1
NDT
inspection
Annual
1.08*10
1
1.13*10
1
1.18*10
1
NDT
inspection
Every 6 month
5.13*10
2
5.37*10
2
5.63*10
2
Table 3 and 4 show no significant discrepancies
between visual inspection and inspection carried
out with the help of NDT methods.
9
This is due to the fact that the probability to detect
a flaw depends mainly on the probability that a
flaw exists.
Table 5: Probabilities to detect a crack before a
failure for visual inspection plan
Probability
to detect a
crack before
a failure
Lower
boundary
Mean
estimation
Upper
boundary
Visual
inspection
Every 6 years
5.3*10
1
6.13*10
1
6.92*10
1
Visual
inspection
Annual
6.72*10
1
7.1*10
1
7.46*10
1
Visual
inspection
Every 6 month
7.53*10
1
7.88*10
1
8.20*10
1
Table 6: Probabilities to detect a crack before a
failure for NDT inspection plan
Probability
to detect a
crack before
a failure
Lower
boundary
Mean
estimation
Upper
boundary
NDT
inspection
Every 6 years
5.44*10
1
6.27*10
1
7.04*10
1
NDT
inspection
Annual
7.16*10
1
7.53*10
1
7.87*10
1
NDT
inspection
Every 6 month
6.87*10
1
7.25*10
1
7.60*10
1
Table 5 and 6 show us that increasing the number
of inspection (for example performing inspection
every 6 month instead of every 6 year) increases
the probability of crack detection before failure.
In the same way, help of NDT methods increases
the probability of crack detection before failure.
Increasing the number of simulation would
probably shows that inspection every 6 month
would be better that every year.
Nevertheless, the lower boundary of simulation
shows that the efficiency to this improvement is
limited. This is mainly due to the probability to
miss a crack of a large size. In fact, failure
happens because we repair identically after we
detect it. Therefore, the crack will propagate as the
same as before. Furthermore, the surveyor could
miss a crack during inspection (due to the chosen
pod), and this is that crack which is going to fail.
4.4 CRITICAL LENGTH INFLUENCE ON
STRUCTURAL RELIABILITY
For previous simulations the critical length was
equal to 114 mm, but for these simulations, it was
taken equal to 54 mm.
Considering a visual inspection every 6 year and
this new critical length, we may compare results
with those obtain for the former one.
Table 7: probabilities for visual inspection plan
function of the critical length
Visual
inspection
every 6 year
ac = 54mm
Lower
boundary
Mean
estimation
Upper
boundary
Probability of
failure
5.71*10
1
6.53*10
1
7.29*10
1
Probability to
detect a flaw
leading to a
repair
1.66*10
1
2.02*10
1
2.41*10
1
Probability to
detect a crack
before a
failure
3.53*10
1
4.33*10
1
5.17 *10
1
Regarding the probability of failure, if we consider
that the failure is happening with a smaller critical
length (54 mm instead of 114mm), the probability
of failure will increase.
However, concerning the both probabilities of
detection, we can see that the critical length has no
influence.
10
Table 8: probabilities for visual inspection plan
function of the critical length
Annual
visual
inspection
ac = 54mm
Lower
boundary
Mean
estimation
Upper
boundary
Probability
of failure
1.48*10
2
4.00*10
2
8.50*10
2
Probability
to detect a
flaw leading
to a repair
9.16*10
2
1.01*10
1
1.11*10
1
Probability
to detect a
crack before
a failure
7.86*10
1
8.53*10
1
9.06*10
1
The same conclusions may be done for annual
inspection with (ac = 54mm) (table 8).
We may explain this because the crack grows up
quickly when it is possible to see it. In other terms,
the crack grows slowly at the beginning and will
grow very quickly after.
In fact, probabilities are almost the same, either
with a critical length of 54mm or 114mm. It
means that we consume a lot of cycles to arrive at
54 mm.
5. CONCLUSION
For the considered ship, wall fascias lead to high
inspection costs. Relative improvement of
inspection every 6 month is not sufficient to
justify additional cost and time lost of operating
availability resulting from this additional
inspection.
Monte Carlo approach is very effective to simulate
successive inspection and repair due to a poor
detail design.
To justify the reliability and inspection planning of
a well design detail, Monte Carlo approach will
lead to a very large number of simulations.
Nevertheless, alternative approach based on
FORMSORM methods will better suit for design
stage, due to the fact that the design is made in
order to avoid failure.
An overall methodology has been introduced to
forecast crack initiation and propagation without
remeshing. This method rounds up a lot of strong
points: realistic stresses, local approach initiation,
easily calculable stress intensity factors, several
cracks with their interactions. It has been applied
on a ship deck beam, and we have seen that lots of
data are accessible to monitor the crack
propagation. The fundamental idea is that the
perfect knowledge of the crack lips displacements
allows for determining the stress intensity factors
which govern the crack propagation. Taking into
account realistically the propagation phenomena
seems the right way to lead to a life span close to
reality.
These results performed with this probabilistic
crack propagation approach have shown that the
code is efficient and easy to carry out.
It proved that the tool may helps operators to
make decisions for inspection, repair and the
maintenance process. In fact, we can determine,
function of how cracks propagate if the inspection
process may be delayed or not.
REFERENCES
1. Serror M., Lebaillif D., Huther I., 2007,
Simulation of behavior of fatigue cracks: A
complete industrial process on a ship deck beam,
PRADS 2007
2. Lebaillif D., Ma S., Huther M., Lieurade H.P,
Petitpas E., Recho N., 2003, Fatigue Crack
Propagation and Path Assessment in Industrial
Structures, International Conference on Fatigue
Crack Paths / FCP 20031820 Sept. 2003, Parma
(Italy)
3. Rice J.R. and Levy N, 1972 The PartThrough
Surface Crack in an Elastic Plate, Journal of
Applied Mechanics, March 1972, 185184
4. Desvaux G.J., 1985 The Line Spring Model for
Surface Flaw, an extension to mode II and mode
III, Master of Science in Mechanical Engineering
at the MIT, June 1985
5. Petitpas E., Lebrun B., de Meerleer S., Fournier
P., Charrier A., Pollet F., 2000, Use of the local
11
approach for the calculation of fatigue strength of
welds. IIW, Florence, 914/07/2000
6. Recho N., 2008, Calculation tools for fatigue
in french Outils de calcul pour la fatigue, EPF
etude bibliographique, Feb 2008
7. Tada H., Paris P.C. & Irwin G.R., The stress
analysis of cracks handbook DEL Research
Corporation
8. Lebaillif D., Huther I., Serror M., Recho N.,
2005, Fatigue Crack initiation and propagation: a
complete industrial process compared with
experiments on industrial welded structure,
Fatigue Design. Cetim, Senlis, Nov. 2005
9. Parmentier G., 2007, Internal Bureau Veritas
report
ATMA 2008
QUELQUES ASPECTS HYDRODYNAMI QUE/STRUCTURE
EN CONCEPTI ON DES PORTECONTENEURS GEANTS
ime MALENI CA, Florence MAUDUI T, Michel HUTHER
BUREAU VERITAS Division Marine Paris (France)
SOMMAIRE
La nouvelle tendance accrotre la taille des porteconteneurs (jusqu' 400m de longueur) fait
apparatre en conception et en procdures de vrification des thmes nouveaux en calculs
hydrodynamiques et de structures concernant les vrifications en rsistance ultime et en fatigue.
Certains thmes concernent la rponse hydrolastique qui devient importante en raison des priodes
propres structurelles relativement basses des ces navires et des fortes contraintes oprationnelles
(vitesse maximale de l'ordre de 27 nuds). Effectivement, la combinaison de la diminution des
frquences propres et l'augmentation des frquences d'excitation peut conduire l'excitation force des
vibrations de la poutre navire par la houle appel "springing", vibrations qui peuvent affecter de
manire significative la dure de vie du navire. En sus de ce phnomne, la vibration transitoire
engendre par le "slamming" ("whipping") peut affecter aussi bien la rsistance ultime que la fatigue.
Dans ces conditions, les rgles classiques des socits de classification atteignent leurs limites pour ces
navires et ainsi il est ncessaire de faire appel une approche dite calculs directs. Cette
communication prsente et discute la procdure complte du Bureau Veritas base sur des calculs
directs
SUMMARY
SOME ASPECTS OF HYDRO STRUCTURAL ISSUES
IN THE DESIGN OF ULTRA LARGE CONTAINER SHIPS
Recent trends in increasing the size of the container ships (up to 400m in length), raise the new hydro
structural issues in their design and verification process, both from ultimate strength and fatigue points
of view. Some of these issues are related to the hydroelastic structural responses which become
important due to the relatively low structural natural frequencies of these ships, and due to the strong
operational requirements (maximum speed around 27 knots). Indeed, the combination of the reduced
natural frequencies and increased excitation frequency can lead to the forced wave induced hull girder
ship vibrations called springing, which might significantly affect the ship fatigue life. In addition to
this, the slamming induced transient vibration (whipping) can affect both the ultimate strength and
fatigue. In any case, the classical classification societys rules are reaching their limits for these ships
and the so called direct calculation approach is needed. In this paper the overall Bureau Veritas
procedure based on direct calculations is presented and discussed.
1. I NTRODUCTI ON
Plus grand semble aujourd'hui synonyme de
plus rentable, mais en fait le gigantisme a
toujours fascin l'homme. L'histoire nous
apporte de nombreux exemples depuis la plus
haute antiquit et la construction navale de
notre poque n'chappe pas ce phnomne.
Les 40 dernires annes ont t fertiles en
projets et constructions de navires gants. Dans
les annes 60 les ptroliers ont doubls de
taille tous les 2 ans jusqu' atteindre en 1975
550.000 tdw pour une longueur de 414m [1],
puis ce furent les navires de croisires qui
passrent de 2000 passagers dans les annes 60
5000 passagers pour le dernier navire en
construction [2].
Aujourd'hui c'est au tour des porteconteneurs
qui de 6.000 TEU dans les annes 60 ont
atteint 11.000 et 11.400 TEU en 2007 [3] avec
aujourd'hui des projets de 14.000 TEU et une
longueur de l'ordre de 400m.
Si l'augmentation de taille peut tre obtenue
par extrapolation des dimensions de navires
existants et les sollicitations de houle values
par analyse dimensionnelle, des phnomnes
dynamiques de second ordre pour des tailles
moyennes peuvent devenir importants pour de
grandes dimensions. C'est en particulier le cas
pour les rponses structurelles vibratoires et la
fatigue.
Il suffit de regarder le pass pour s'en
convaincre. Deux exemples peuvent illustrer
l'apparition au premier plan de phnomnes
antrieurement secondaires, l'avion Comet
(1954), premier grand avion commercial
raction, qui a fait apparatre l'importance de la
fatigue en aronautique et le ptrolier Magdala
(1968), premier 250.000 tdw, qui a mis en
lumire les risques de flambement des
composants primaires[4].
Le retour d'exprience conduit considrer
deux phnomnes dynamiques affectant la
poutre navire avec une attention toute
particulire pour les porteconteneurs gants,
le
whipping
1
(mise en vibration transitoire due
aux chocs hydrodynamiques l'avant
slamming et slapping), le springing (excitation
entretenue du premier mode de vibration libre).
En effet, la taille croissante des navires et leur
pont ouvert diminuent les frquences propres
de la poutre navire alors que la vitesse
augmente les frquences de rencontre avec les
houles les amenant dans la mme gamme avec
des effets dynamiques amplifis.
Il apparat alors que les rgles classiques des
socits de classification atteignent leurs
limites rendant ncessaire des approches
faisant appel des calculs directs de
comportement dynamique sur houle.
Nous allons analyser ces deux phnomnes,
leur impact sur les structures des navires, les
moyens disponibles pour les aborder et leur
incidence sur la conception des navires.
2. WHI PPI NG
Le whipping est la rponse transitoire du
premier mode de la poutre navire aux chocs
hydrodynamiques sur les fonds l'avant du
navire aprs djaugeage (slamming) ou sur un
fort dvers d'trave (slapping) (figure 1).
Figure 1: Retour violent dans l'eau de l'avant
d'un navire aprs djaugeage
Le comportement du premier mode de la
poutre navire peut tre simul par un systme
un degr de libert amorti de mme frquence.
Les efforts sont de courte dure mais trs
1
Les termes tossage et fouettement sont utiliss en
franais pour slamming et whipping mais slapping
et springing n'ayant pas d'quivalent franais, nous
utiliserons la terminologie anglaise dans le texte
levs, similaires une impulsion [5] et
induisent ainsi des vibrations transitoires
amorties.
2.1. I ncidence sur la tenue de la structure
Les slamming et slapping sont connus pour se
produire par mer de l'avant, lorsque les
mouvements verticaux sont les plus forts.
La dformation de vibration de la poutre navire
engendre par le choc se superpose celle
engendre par la houle (figure 2).
Figure 2: Navire Marcel Bayard, mesures la
mer des contraintes au pont,
houle plus whipping
Les risques de dommage dus au whipping sur
des navires pont ouvert ont t identifis
depuis de nombreuses annes, le premier effet
tant un moment vertical suprieur celui
d'chantillonnage et engendrant une
dformation permanente de la poutre navire, le
second tant la gnration de cycles de
contraintes pouvant conduire des fissurations
de fatigue prmatures[6 , 7].
Si le dpassement du moment maximum de
flexion admissible entranant la dformation de
la poutre navire est ais comprendre,
l'apparition de fissure de fatigue est plus
difficile admettre.
Cependant, le navire Marcel Bayard illustre
bien ce risque. Les mesures la mer rapportes
en [6] ont t dcides suite l'apparition de
fissures au bout d'un temps anormalement
court, 3 ans. Les mesures et l'tude ont montr
que malgr sa taille moyenne, 110m longueur,
les avaries pouvaient s'expliquer par le
whipping frquent provenant des conditions
d'opration particulires ce type de navire et
les grandes coutilles entranant de fortes
concentrations de contraintes au pont.
Au niveau du projet d'un navire il est donc
important de pouvoir valuer l'incidence du
whipping sur les contraintes au pont, ce qui
requiert le calcul du comportement du navire
sur houle puis celui de la structure aux efforts
impulsifs engendrs par les slamming et
slapping.
2.2. Moyens de calcul et mthode
Depuis quelques dcennies le calcul direct sur
houle d'un navire est devenu courant pour les
analyses en effort extrme et de cumul de
fatigue.
Les mthodes utilises par les bureaux d'tudes
reposent sur l'approche spectrale, rapide
mettre en uvre, et ayant aprs calibration une
prcision suffisante pour les vrifications
classiques.
Par contre le whipping tant un phnomne
impulsif non linaire, l'approche spectrale
devient insuffisante. Il a donc t dvelopp
des mthodes de calculs pas de temps
permettant de rsoudre le problme de
couplage entre la diffraction/radiation 3D
hydrodynamique et la dformation dynamique
du navire modlis par une poutre suivant
Timoshenko.
Dans cette approche modale, les mouvements
et dformations du navire sont reprsents par
une srie de 6 modes de corps rigides plus 5
10 modes structurels sans masse d'eau ajoute
[8]. Mais il est noter que les coefficients
hydrodynamiques linaires dans les quations
pas de temps sont obtenus partir de calculs
dans le domaine frquentiel classique [9 , 10].
Les mthodes de dtermination des efforts de
slamming/slapping, excitations de la rponse
dynamique, correspondent aux deux mthodes
disponibles aujourd'hui "Generalized Wagner
Model" et "Modified Logvinovich Model",
toutes deux bases sur des approches 2D.
Les calculs de rponse de whipping sont ainsi
possibles par la mise en uvre de la chane de
logiciels de calcul de comportement du navire
rigide dans le domaine frquentiel et un calcul
pas de temps non linaire de la poutre navire
lastique. L'ensemble a t vrifi avec succs
sur des navires de 260 et 360 m de longueur
[11] (figure 3 et 4).
Figure 3:Exemple de moment de flexion non
linaire avec whipping sur un tat de mer fix
Figure 4: Zoom l'instant d'un choc de
slamming (rponses: lastique, rigide non
linaire, rigide linaire)
La mthode de calcul se droule alors suivant
les tapes cidessous:
1. slection des tats de mer et de leur
frquence d'existence suivant les zones
d'opration du navire (standard: Atlantique
Nord)
2. calcul de comportement linaire avec
dtermination des conditions extrmes
possibles suivant l'incidence de la houle
(vitesse du navire, hauteur de houle)
3. dtermination des conditions de navigation
avec possibilit de slamming (incidence de
houle, vitesse, hauteur de houle)
4. calcul de comportement non linaire pour
les conditions de navigation avec
slamming
2.3. Effets sur efforts extrmes
L'observation des rponses lastiques montrent
que le whipping augmente les valeurs extrmes
du moment de flexion vertical de la poutre
navire. Il est donc ncessaire de prendre en
compte cet accroissement pour les vrifications
de la structure l'tat ultime.
Les calculs effectus [11] et les mesures la
mer [6] montrent que le whipping apparat
lorsque le moment de houle est en contrearc.
A partir de la condition de navigation la plus
svre, force de slamming extrme, il est
dtermin l'augmentation du moment
rglementaire maximal en contrearc.
Les mesures la mer montrent que
l'amortissement de la vibration transitoire est
faible, de l'ordre de =1,5%, aussi la vibration
de whipping est encore sensible lorsque le
moment de houle passe en arc (figure 4). La
valeur du moment rglementaire en arc peut
donc aussi ncessiter une augmentation.
Les calculs de structure peuvent alors tre
effectus suivant le rglement de classification
en utilisant les moments rglementaires
corrigs des accroissements dus au whipping.
L'application sur un porte conteneurs de 250m
de longueur montre que l'accroissement du
moment de flexion maximal total (houle plus
eau calme) pouvait atteindre 20 30 % pour
des conditions de navigation svres, en accord
avec les mesures la mer [6].
Les calculs effectus sur des porteconteneurs
gants conduisent considrer que cet ordre de
grandeur reste valable pour ces navires
2.4. Effets sur la fatigue
L'observation de la rponse en whipping
(figure 4) montrent deux effets en fatigue:
accroissement de l'tendue de contrainte
poutre navire rigide la frquence relative
de la houle
gnration de cycles d'amplitudes
dcroissantes la frquence du premier
mode de la poutre navire lastique.
Pour les dtails souds, on considre que les
contraintes rsiduelles de soudage entranent
un niveau de traction lev dans les zones
d'apparition des fissures permettant d'admettre
pour les calculs un rapport R (contrainte min
divise par contrainte max) lev (R>0.5).
Les essais de fatigue montrent que ce rapport
lev efface l'influence des variations de
contrainte moyenne sur l'estimation de la dure
de vie en fatigue.
Pour des dtails sans soudure il n'y a pas de
contraintes rsiduelles, aussi le traitement
devient trs dlicat. Pour simplifier on
applique cependant la mme approche que
pour les dtails souds ce qui revient
considrer qu'il existe un niveau lev de
contrainte moyenne en traction, en fait
hypothse du ct de la scurit, donc
acceptable.
L'utilisation de la rgle linaire de Miner
permet alors de sparer les deux composantes,
basse frquence (houle) et haute frquence
(whipping) et de calculer les dommages
sparment, puis de les sommer (figure 5).
=
+
Figure 5: Sparation des effets de whipping en
vu du calcul de fatigue
Ayant dtermin par un calcul de rponse de la
poutre navire l'tendue du premier cycle d'une
rponse de whipping (not CwpSw sur la
figure 6), il est possible de dterminer la
distribution transitoire amortie des tendues
(figure 6):
CwpSw CwpSw
Figure 6: Composante haute frquence
transitoire de whipping
Les quations sont alors:
Swp
j
=A
2
) 3 4 (
1
exp
2
j (1)
avec j n d'ordre du cycle
A =
2
1
exp
2
Sw Cwp
(2)
La distribution court terme est obtenue en
remarquant que pour une rponse il n'y a que 1
cycle chaque niveau et qu'il y a une rponse
chaque cycle de houle (figure 7).
n
Swp
CwpSw
n
Swp
CwpSw
Figure 7: Distribution des contraintes de
whipping pour une amplitude de houle donne
Le calcul de la somme de Miner peut
s'effectuer partir des distributions court
terme des tendues de contraintes. Chaque
palier d'une distribution peut tre assimil la
rponse sur une houle rgulire d'amplitude
constante, aussi on peut utiliser des calculs de
rponses non linaire sur houle rgulire.
Pour une houle de frquence donne mer de
l'avant, puisque le slamming ne se produit
pratiquement que mer de l'avant, il est possible
de dterminer les rponses fonction de la
hauteur de houle H et la hauteur de houle seuil
H
th
au dessous de laquelle il n'y a plus de
risque de slamming:
CwpSw =f(H) H >H
th
(3)
La mthode de calcul devient alors la suivante:
1. dtermination, pour les frquences de
houle de la table des tats de mer et les
incidences retenues, de la vitesse du navire
maximale suivant la hauteur de houle
(conditions de navigation extrmes
supportables la passerelle)
2. calcul des fonctions de transfert et des
rponses court terme sans whipping
3. dtermination, mer de l'avant et pour les
frquences de houle de la table des tats de
mer retenue de la hauteur de houle seuil et
des rponses en whipping en fonction de la
hauteur de houle
4. approximation d'une loi linaire donnant
l'amplitude du premier cycle de whipping
fonction de l'amplitude de la rponse en
mode rigide
5. approximation de la variation de la loi
linaire en fonction de l'angle d'incidence
de la houle
6. identificatioon des tats court terme avec
possibilit de whipping
7. correction de whipping de l'tendue de
contrainte des distributions court terme
basse frquence concernes
8. calcul des sommes de Miner pour tous les
court termes (avec et sans whipping)
9. pour chaque court terme avec whipping,
dtermination des distributions de la
composante haute frquence de whipping
et somme de Miner
10. somme de Miner totale, basse frquence
corrige du whipping et haute frquence de
whipping.
Une valuation effectue pour un
porteconteneurs gant montre que le whipping
concerne essentiellement les contraintes
engendres par le moment de flexion vertical
et a une incidence mesurable sur le cumul de
fatigue.
La correction d'tendue de contrainte basse
frquence entrane une augmentation de l'ordre
de 3% de la somme de Miner totale, avec
100% d'accroissement pour les mers de l'avant
seules.
La somme de Miner de la composante haute
frquence par contre est ngligeable,
principalement du aux faits que:
le whipping n'apparat que sur un nombre
limit d'tat de mer (mer de l'avant, houles
de hauteur suprieur au seuil)
la dcroissance des amplitudes des cycles
entranent la majorit des cycles sous le
niveau de changement de pente de la
courbe SN utilise en raison du caractre
alatoire des phnomnes.
3. SPRI NGI NG
Le springing est une vibration de rponse
force de la poutre navire excite par la houle.
Des calculs de frquence propre du mode
deux nuds de la poutre navire effectu sur
des navires de taille jusqu' 350m de long
conduisent des valeurs bien suprieures
celles des plus courtes houles observes. Les
tables de l'Atlantique Nord donnent comme
plus basse priode Tz=3,5s (0,28Hz) et le
calcul pour un ptrolier de 300.000tdw donne
comme frquence propre avec masse d'eau
ajoute 0,5Hz (T=2,0s).
Cependant le springing a t identifi ds 1972
lors du dpouillement de mesures la mer sur
un ptrolier de 340m [12] (figure 8).
La rponse du 1er mode de vibration de la
poutre navire, malgr sa frquence leve,
s'explique par le fait que pour une priode de
houle donne, l'nergie est rpartie sur une
large bande de frquence. En utilisant la
reprsentation de PiersonMoskowitz, pour une
houle de 3,5s on peut considrer une nergie
dans la bande de [0,120,65]Hz et pour une
houle de 8,5s dans la bande de [0,051,7]Hz.
Figure 8:Spectre de rponse de la contrainte de
flexion au pont d'un ptrolier, mer de l'avant,
mode rigide et springing [12]
En outre mer de l'avant, en raison de la vitesse
du navire, la frquence d'excitation
e
est
suprieure celle de la houle
H
. Cette
frquence d'excitation (frquence de rencontre)
est donne par la formule:
e
=
H
+
V
g
H
2
(4)
ce qui pour le ptrolier mentionn, avec une
vitesse de 15nd (7,7m/s), donne l'nergie pour
une houle de 3,5s dans la bande de [0,19
2,73]Hz et pour une houle de 8,5s dans la
bande de [0,0616,0]Hz. Comme on le
constate, le mode 0,5Hz est effectivement
excit.
Sur les navires de 300 350m de longueur, si
le springing est visible sur une analyse
spectrale, son niveau est trs faible et sans
incidence sur la rsistance de la structure.
Lorsque la longueur du navire augmente ou si
le pont principal comporte de grandes
ouvertures, sa rigidit diminue et la frquence
propre du 1er mode diminue. Lorsque la
vitesse augmente, les frquences d'excitation
augmentent. Dans les deux cas la frquence de
springing se dplace vers la rgion des spectres
plus forte densit d'nergie accroissant la
rponse (figure 8, le pic de springing se
dplace vers la gauche).
Cet effet a t constat lors des tudes des
projets des ptroliers de 550.000tdw (longueur
414m) puis confirm par 4 annes de mesures
la mer [13,14].
Lorsque l'on considre les caractristiques des
porteconteneurs gants, longueur 363m,
vitesse 24nd, pont trs largement ouvert [15], il
apparat clairement que le springing est un
phnomne prendre en considration pour la
conception de la coque.
3.1. I ncidence sur la tenue de la structure
La rponse temporelle de la contrainte au pont
a la forme de la figure 9, un signal haute
frquence d'amplitude quasi constant
(springing) sur une porteuse basse frquence
(houle)
Figure 9: Mesure de la contrainte au pont avec
springing apparent
Les tudes et mesures effectues sur les
ptroliers de 550.000tdw montrent que le
springing est sensible pour les tats de mer
plus faible priode. Pour ces priodes, les
tables d'observations montrent que les hauteurs
de houle maximales sont faibles (tableau 1).
Tz (s) 3.5 4.5 5.5 6.5 7.5 8.5
Hs (m) 1,5 4,5 9,5 13,5 14,5 14,5
Tableau 1: Hauteurs significatives maximales
pour l'Atlantique Nord (table IACS)
Due au trs faible amortissement, =1.5%
obtenu par mesures la mer [6,14], l'nergie
totale de la rponse reste trs limite, et
l'amplitude est faible par rapport celle de la
houle (figure 9).
Ainsi l'influence du springing ne sera pas
significative sur les valeurs extrmes qui
correspondent aux fortes houles de priode
importante.
Par contre la haute frquence et le caractre
vibration force, donc quasi permanente du
phnomne sur les faibles houles qui sont les
plus frquentes au cours de la vie du navire,
entrane un cumul de fatigue non ngligeable,
confirm par le retour d'exprience des
ptroliers de 550.000tdw [14].
3.2. Moyens de calcul et mthode
Si l'on considre les mouvements verticaux de
la poutre navire et sa dformation due au
springing, il est possible de faire l'hypothse
des petits mouvements pour le springing ce qui
permet d'effectuer un calcul linaire en
sparant les deux phnomnes, rponse rigide,
rponse lastique [13].
Il est ainsi possible d'utiliser les outils linaires
dvelopps pour les calculs hydrodynamiques
et de comportement des structures [17].
Un premier calcul, classique de comportement
sur houle, permet de dterminer en linaire les
coefficients hydrodynamiques, en particulier
les masses d'eau ajoutes, et la fonction de
transfert de la rponse de la poutre navire
rigide (figure 10).
Figure 10: Calcul de la fonction de transfert
poutre navire rigide
Un deuxime calcul de poutre navire lastique
permet de dterminer les frquences propres
poutre navire, et la fonction de transfert de
springing (figure 11).
Figure 11: Calcul de la fonction de transfert de
springing en torsion de la poutre
navire lastique
On obtient ainsi les fonctions de transfert
quasistatiques, springing et totales pour les
conditions de navigation slectionnes pour les
calculs de structure (figure 12)
La dtermination des fonctions de transfert
permet de calculer les rponses sur spectres de
houle et les distributions court terme sur
chaque tat de mer correspondant aux
conditions de navigations retenues pour le
projet avec les outils classiques mis au point
pour les effets de houle navire rigide.
Figure 12: Fonctions de transfert de la
contrainte au point chaud d'un dtail soud de
porteconteneurs (springing, houle, totale)
La mthode de calcul se droule alors suivant
les tapes cidessous:
1. slection des tats de mer et de leur
frquence d'existence suivant les zones
d'opration du navire (standard: Atlantique
Nord)
2. calcul de comportement linaire avec
dtermination des conditions extrmes
possible suivant l'incidence de la houle
(vitesse du navire, hauteur de houle)
3. dtermination des conditions de navigation
avec possibilit de springing (incidence de
houle, vitesse, priode de houle)
4. calcul de comportement linaire lastique
pour les incidences de houle avec
springing
3.3. Effets sur la fatigue
L'observation de la rponse en springing sur un
tat de mer donn (figure 9) montrent une
gnration de cycles d'amplitude quasi
constante la frquence du premier mode de la
poutre navire lastique sur la rponse la
frquence de houle.
Le calcul de cumul de fatigue court terme
requiert la dtermination de la distribution des
tendues de contrainte. La mthode exacte
serait un comptage Rainflow, mais pour ce
faire il est ncessaire d'avoir un historique
temporel, ce qui est peu pratique partir d'une
approche spectrale.
En considrant les mmes hypothses que dans
le cas du whipping et la rgle de cumul linaire
de Miner, il est possible de sparer les deux
composantes (figure 13), basse frquence
(houle) et haute frquence (springing), et de
calculer le dommage par somme de Miner.
Comme pour le whipping, l'tendue de
contrainte de vague doit tre corrige de
l'amplitude de springing. Seules les conditions
de navigation avec risque de springing sont
considrer, il est donc important de prendre en
compte les rductions de vitesse lorsque la
hauteur de houle augmente car elle rduit ou
mme supprime le phnomne de springing.
=
+
=
+
Figure 13: Sparation des effets de springing
en vu du calcul de fatigue
Chacune des composantes, houle seule et
springing seul, correspondant un signal
spectre troit il peut tre appliqu la
mthodologie du calcul de fatigue sur houle
[16]. La distribution de l'tendue de contrainte
S
i
court terme est reprsente par une
distribution de Rayleigh dont le coefficient est
dtermin partir de l'aire du spectre de
rponse m
0
:
prob{S >S
i
}=
0
2
8
exp
m
S
i
(4)
Lorsque le nombre de cycles total du court
terme est connu N
t
(dure du court terme
divis par la priode moyenne de passage par
zro), pour chaque S
i
il peut tre calcul le
nombre de cycles n
i
:
n
i
=(p
i
p
i1
)N
t
en notant p
i
=prob{S >S
i
}
Pour un dtail de structure donn, connaissant
le courbe SN, il est possible de calculer
sparment les sommes de Miner puis de les
sommer en tenant compte des probabilits
d'existence de chaque condition court terme
(priode et hauteur de houle, vitesse du navire,
incidence de la houle):
D
total
=D
houle/corrige
+D
springing
La mthode de calcul devient alors la suivante:
1. dtermination, pour les frquences de
houle de la table des tats de mer et les
incidences retenues, de la vitesse du navire
suivant la hauteur de houle (conditions de
navigation extrmes supportables la
passerelle)
2. calcul des fonctions de transfert sans
springing et des rponses court terme
3. dtermination des domaines de frquences
des spectres de densit d'nergie pour les
conditions de navigation de calcul
4. calcul des frquences propres de la poutre
navire avec masses d'eau ajoutes
5. dtermination des tats court terme avec
risque de springing
6. calcul des fonctions de transfert de
springing et des rponses court terme
pour les tats court terme concerns
7. somme de Miner totale, basse frquence de
houle corrige du springing et haute
frquence de springing.
Une valuation effectue pour un
porteconteneurs gant montre que, en raison
du pont trs largement ouvert, la frquence
propre de poutre navire la plus basse est celle
en torsion, suivi par celle de flexion vertical
(tableau 2).
Mode de vibration de la poutre f (Hz)
mode en torsion 1 noeuds 0,37
mode en flexion 2 noeuds 0,49
mode en torsion 2 noeuds 0,50
mode en flexion 3 noeuds 0,90
Tableau 2: Frquences propres de la poutre
navire avec masses d'eau entrane
La dtermination des conditions de navigation
avec risque de springing a conduit considrer
le springing de torsion pour les incidences de
120 et 90 et le springing de flexion pour les
incidences de 180 et 120 (180 correspond
une mer de l'avant).
Sur un dtail de structure au pont (figure 14),
rgion la plus sollicite en dynamique
d'ensemble, une analyse de la dure de vie en
fatigue a t effectue suivant la mthode
dcrite.
Figure 14: Dtail de structure de pont d'un
porteconteneurs gant: coin d'coutille
La rponse de springing a t calcule pour
toutes les priodes de houle et les hauteurs de
la table d'observation jusqu' 5m significatif.
Pour les hauteurs suprieures il a t considr
qu'il y avait rduction de vitesse et que la
modification des frquences de rencontre
rendait le springing ngligeable.
Les sommes de Miner ont t calcules court
terme. Le springing entrane une augmentation
de l'ordre de 4% 10% de la somme de Miner
totale.
Une analyse de la sensibilit la hauteur de
houle maximale montre que le passage de 5m
6m significatif augmente le cumul de Miner de
springing de 20%, mais a un impact du second
ordre sur la somme totale.
Une tude en cours analyse une alternative
plus prcise l'approche cidessus. Pour ce
faire un signal temporel de la contrainte est
gnr par transforme de Fourrier inverse
partir du spectre de rponse global (houle plus
springing) sur lequel le comptage des tendues
de contraintes est effectu par la mthode
Rainflow. Le cumul de fatigue par somme de
Miner est alors considr plus prcis.
Les rsultats obtenus montrent des carts entre
les deux approches mais confirment la validit
de l'approche simplifie utilise.
4. CONCLUSI ON
L'augmentation des tailles et vitesses des
navires, en particulier des porteconteneurs,
modifie les comportements sur houle.
Des phnomnes connus sur les navires de plus
petites tailles peuvent tre amplifis jusqu'
atteindre des niveaux impactant la conception
des coques. C'est en particulier, pour la poutre
navire, le cas du whipping, rponse vibratoire
transitoire suite un slamming ou slapping, et
du springing, vibration force par la houle
irrgulire.
Le phnomne de springing avait t analys
puis vrifi la mer lors de la construction des
ptroliers de 550.000tdw en 1975, avec les
moyens de calcul de l'poque.
Le whipping est aussi connu depuis cette
mme poque avec des consquences gnantes
sur des navires pont ouvert, mais alors sans
possibilit de calcul en raison de la faiblesse
des moyens informatiques.
Le dveloppement des outils informatiques
ayant permis la mise en uvre de mthodes
complexes d'hydrodynamique navale dans le
domaine non linaire et avec couplages
fluide/structure, il a t possible de dvelopper
une approche d'estimation de ces phnomnes
et de leurs consquences sur les contraintes
extrmes et la fatigue en service.
La mise en application pour des projets de
porteconteneurs gants a confirm que leur
taille et disposition gnrale conduisent des
accroissement des niveaux de contrainte et du
cumul de fatigue significatif, ne permettant
plus de les ignorer pour l'chantillonnage des
structures.
Les effets de whipping peuvent entraner un
accroissement de l'ordre de 20% du moment de
flexion vertical total, important pour les
vrification l'tat ultime.
Concernant le springing, les projets actuels de
350m de long, en raison de la flexibilit due
un pont trs ouvert et de la vitesse leve,
montrent un niveau d'accroissement en fatigue
suprieur ceux des ptroliers de 414m,
restant cependant limit et se mesurant en pour
cents.
Pour des projets de plus grande taille, 400m et
audel, il est certain qu'avec des dispositions
de coques similaire, l'accroissement de fatigue
deviendra plus important. En se rappelant qu'il
s'agit d'un phnomne de vibration force trs
peu amorti, les rponses ne sont pas linaires et
devront donc tre analyses avec une plus
grande attention.
5. REFERENCES
[1] Bureau Veritas 1828/1978  A record of
150 years  PEMA 2B Publishing, 1978
[2] Another Genesis for Aker  The
MotorShip review  May 2007, p 6
[3] H.J . REUSS et J . BARNES, Maersk takes
world biggest and fastest  The
MotorShip review  November 2006,
p 1617
[4] P. de Livois et B. Parizot Le gigantisme
en construction navale, ses consquences
sur la scurit des navires Revue
Navigation Avril et juillet 2007
[5] M. Huther et N. Ket Water impact risk
estimation during ship design Ship
Behaviour at Sea, International
Symposium on Ship Hydrodynamics and
Energy Saving, ISSHES 83 Madrid,
September 1983
[6] J . Osouf Etude exprimentale du
comportement dynamique dur houle du
navire cblier "Marcel Bayard" Revue
NTM 1973
[7] I. Drummen, T. Moan, G. Storhaug, E.
Moe Experimental and full scale
investigation of the importance of fatigue
damage due to waveinduced vibration
stress in a container vessel Design &
Operation of Container Ships, Royal
Institution of Naval Architects, London,
November 2006
[8] .Malenica, B. Molin, F. Remy, I.
Senjanovic Hydroelastic response of a
barge to impulsive and non impulsive
wave loads 3rd International
Conference on Hydroelasticity Oxford,
2003
[9] W.E. Cummins The impulse response
function and ship motions
Schiffstecknik, 1962
[10] T.F. Ogilvie Recent progress toward the
understanding and prediction of ship
motions 9th International Conference
on Numerical Ship Hydrodynamics
Ann Arbor (USA), 2007
[11] J . Tuitman,S. Malenica Some aspects of
whipping response of container ships
23rd IWWWFV J egu (Korea), 2008
[12] J .M. Planeix Wave loads, A correlation
between calculations and measurements
Revue International Shipbuilding
Progress, August 1972
[13] B. d'Hautefeuille, M. Huther, M. Baudin,
J . Osouf Calcul de springing par
quations intgrales Revue NTM, 1977
[14] M. Huther et J . Osouf SEFACO, Four
years of experience in ship hull girder
stress monitoring MARINTEC
Shangai, October 1983
[15] L. Grard  De 6 500EVP 11 400EVP
sur un mme concept, jusquo seratil
possible daller ? ATMA, juin 2007
[16] M. Huther, D. Beghin, S. Mahrault A
fatigue strength guidance note for welded
ship structure assessment IIW/IIS doc
XIII11174298 / XV99298, 1998
[17] S. Malenica Hydro structure interaction
in seakeeping International Workshop
on Coupled methods in numerical
dynamics IUC Dubrovnik (Croatia),
September 2007
OFFSHORE ENGINEERING
& MOORING
Finding the right site for an offshore unit, understanding the local
conditions and then designing moorings which will hold it in place are the
three apparently simple but in practice complex problems which many of
the papers in this section address. There is a strong focus on moorings,
and also papers which summarise the long experience of Bureau Veritas
with offshore concrete floating structures.
Bulletin Technique  Bureau Veritas 2008
ChapitresBT2008:ChapitresBT2008 05/10/09 20:54 Page3
1 Copyright 2008 by ASME
CHARACTERIZATION OF POLYESTER MOORING LINES
Michel FRANOIS
Bureau Veritas, Paris
Peter DAVIES
IFREMER Brest Centre
ABSTRACT
Fibre ropes are extensively used in marine applications.
One critical area of interest is their application as mooring lines
for floating offshore platforms, for which primarily polyester is
now employed in various regions (offshore Brazil  now for 10
years, West Africa, Gulf of Mexico). Evaluating the response of
the system requires a description of the loadelongation
properties of the rope.
A practical model involving two sets of stiffness data is
currently used, and procedures for their measurement are
available. This paper presents an overview of this model, then
focuses on recent work on the quasistatic stiffness of polyester
ropes. This is addressing the variations of the mean tension in
the lines, at a very slow rate, under changing weather
conditions.
Extensive tests were performed, principally on polyester
subrope samples. Some tests were also performed on a full
size 800ton MBS rope. Besides standard tests, specific tests
were performed over an extended range of loading, to cover the
situations that may be found in a wide range of systems and
design conditions. The factors (measurement accuracy, test
conditions, etc) affecting the values are discussed along with
the presentation of tests and results.
Results are interpreted to provide practical data for
mooring analysis, in the form of a quasistatic loadelongation
characteristic. These results also give a better insight into the
viscoelastoplastic response of polyester fibre ropes.
For the dynamic stiffness of polyester ropes, an overview
of recent and earlier test data is presented. The dependence of
dynamic stiffness on testing parameters is discussed,
highlighting mean load as the principal parameter under real
stochastic loading, and confirming the current practice for
modelling dynamic stiffness in design.
1 INTRODUCTION
Fibre ropes are extensively used in a number of marine
applications. One critical area of interest is their application as
mooring (anchoring) lines for the stationkeeping of floating
offshore platforms in deepwater. Following a development
period, and now ten years after the first installations of floating
production systems by Petrobras in Brazil, this technology has
reached a stage of maturity: fibre rope station keeping systems
are now employed commonly offshore Brazil, and also in other
regions around the world (in West African waters, in the Gulf
of Mexico, ). Polyester, the material primarily used in this
application, is addressed in this paper.
Evaluating the response of the system, and then the
adequacy of maximum offset and line tensions with the
relevant acceptance criteria, requires a description of the load
elongation properties of the rope. However, these properties are
rather complex to evaluate and specify, in comparison with the
linear elastic behaviour of equivalent steel components, as they
are nonlinear and time dependent. Besides, the loading
regimes of a rope in an anchoring line are quite specific with
respect to usual service conditions of fibre ropes.
A first insight into the loadelongation properties of fibre
rope lines was given in [1]. Within the French CLAROM fibre
rope projects, since 1997 [2], extensive testing has been carried
out. This resulted, in 2000, in a practical (Engineering) model
for the loadelongation characteristics of polyester ropes, that
was presented and documented in [3] and [4], and is currently
in use within several State of the Art mooring analysis and line
dynamics software packages. Tests performed since on other
materials [5] have generally confirmed the applicability of this
model.
In the meantime, testing procedures have been developed
for rope qualification (see [6]), that are now available with the
ISO standard for rope qualification [7] and the BV Guidance
Note [8]. The model for engineering and analysis of fibre rope
mooring systems and the testing procedures in these documents
define rope properties in a consistent manner. This model
Proceedings of the ASME 27th International Conference on Offshore Mechanics and Arctic Engineering
OMAE2008
June 1520, 2008, Estoril, Portugal
OMAE200857136
2 Copyright 2008 by ASME
involves a separation into several terms. One of these terms is
the Quasistatic stiffness, that is the focus of the present
paper.
The tests were performed principally on polyester subrope
samples. Some tests were also performed on a full size parallel
construction 800ton MBS rope. In addition to the standard
tests, an extended series of test sequences was performed, to
get an understanding of rope behaviour under a wider range of
load and elongation, and to cover the particular situations that
may be found in some systems or design conditions.
Besides the interpretation of results to provide practical
data for analysis, these results also gave a better insight into the
viscoelastoplastic response of polyester fibre ropes. As
correct measurement and interpretation of test data are essential
for platform safety, the factors affecting the values are
discussed (measurement accuracy, test conditions, ) along
with the presentation of tests.
For the dynamic stiffness, a discussion on the dependence
of dynamic stiffness on testing parameters is made, based on
the comprehensive data from recent tests and on earlier data.
2 FIBRE ROPES LOADELONGATION PROPERTIES
AND MODEL
Issues
The loadelongation properties of fibres and fibre ropes
exhibit a nonlinear and time dependent (viscoelastoplastic)
behaviour. This is primarily the result of the load and time
dependence of the materials forming the filaments (a complex
assembly of long chains of organic molecules), and to a lesser
extent from the effects of rope construction [9]. As a result
these properties cannot be reduced to a loadelongation
characteristic, even a non linear one, and particularly, NOT
to the load elongation curve of a new sample under monotonic
loading (as obtained from a standard breaking test).
In the lines of a stationkeeping system, once the system is
deployed/installed and set under tension, the rope will be
maintained under a sustained tension for months, even years or
decades in the case of a deepwater permanent mooring, then
subjected to the random loads induced by the environment
(wind, waves, current). It is then convenient for the evaluation
of system response, as proposed in [8], to separate the response
into three terms, related to the time scale of actions, and
matching the typical steps of a mooring analysis (be it
frequency or time domain) :
Mean elongation (system pretension, permanent load),
Viscoelastic response to slow variations of mean load
under changing weather, modelled by the quasistatic
stiffness, as further discussed in sections 3 and 4 below,
Response to dynamic actions (both low frequency and wave
frequency) modelled by the dynamic stiffness (see section 5
below).
A particularly important aspect, quite specific to fibre
ropes, is the modification of the properties of a rope during the
first loading(s) and during the early stages of rope service. This
process, called bedding in, is primarily due to changes at a
molecular level within fibres. It results in a stabilisation of the
rheological properties of the rope, and in an essentially
permanent  not recoverable  elongation with respect to the
rope initial length at the time of manufacturing. The length of a
finished rope is thus defined in [7] as a beddedin length at a
specified tension.
Besides, as a noteworthy consequence of time dependence,
it is important to consider in the definition of test sequences
and the derivation of engineering data that the time scale of
actions on a test bench is much smaller than for a line insitu:
this time dependence (also beddingin) will thus affect
somewhat differently the response.
QuasiStatic stiffness
Following the concept proposed in [1], a quasistatic
stiffness was defined, ten years ago, in order to model the
viscoelastic response of ropes to slow variations of mean load,
under the effect of changing weather conditions, i.e. a time
scale of several hours or days, where the tension in the line,
initially the line pretension, increases (in the "windward"
lines) or decreases (in the "leeward" lines), at a very slow rate.
In this test (see [7], [8] and below), after a proper bedding
in, the rope is cycled between two tension levels, with a
constant load plateau at each level, during which creep or
recovery is measured. Several cycles are needed to get rid of
the initial condition of the rope on the test bench (which is not
representative of actual conditions), and to obtain stabilised
results. Typically, 3 cycles of twice hour each are used,
keeping the duration of the test within a practical time frame.
In order to get a stiffness that is more representative of the
real loading duration of the events intended to be modelled (see
below), cycles of longer duration can be simulated as follows
from test results (load and elongation versus time).
1) The elongation L( t ) along each creep (or recovery)
plateau can be written, by a three parameter fit (see [8] ), as :
L( t ) = L( t
p
) + a
c
. log [ 1 + ( t  t
p
) / t
a
] (1)
This fit (on a
c
, t
a
, and L(t
p
)) is independent of time unit
and origin, and the result does not depend on the selection of t
p
,
the time at any point along the load plateau.
2) From L( t ) , the elongation L at the end of each
cycle of duration can be obtained as :
L L( t
p
) + a
c
. log [ / t
a
] (2)
3) The (linearised) quasistiffness is then taken as a secant
stiffness between the end points of the last successive cycles
of duration . Normalising loads by rope MBS, the non
dimensional quasistatic stiffness KrS is obtained.
The two load levels are normally taken as 10% and 30% of
MBS, and the duration is normally taken as 12 h, providing
values for typical storm conditions. Different levels or
durations could apply to some design or metocean conditions
(e.g. a damaged system, a loop current event). The tests
presented in this paper were performed to get an understanding
of rope behaviour under a wider range of load and elongation,
and in order to provide data for such conditions.
3 Copyright 2008 by ASME
3 ROPE TESTING
General :
The CLAROM French Mooring Line project has been
working on fibre rope moorings over the last ten years and has
generated a large database of material properties. Tests have
been performed at scales from single filament [9] up to 800ton
break load ropes [10], including the intermediate yarn,
assembled yarn and subrope levels. Here only results from the
subrope and full scale ropes will be discussed, but it is
important to underline that by working on filaments and yarns
the material and rope construction contributions to overall rope
behaviour can be quantified and modelled. This provided also
some light on the underlying mechanisms at molecular level, if
not yet a definitive interpretation.
Experimental setup
Subrope tests were performed at the IFREMER test
facilities in Brest. A 9meter long 100ton capacity test frame
was used, equipped with a 1.5meter course hydraulic piston
(see Figure 1). A programmable controller enables loading
sequences to be preprogrammed, so that long, complex test
sequences lasting several days can be defined and run. Full size
rope tests were run at the LCPC test laboratory in Nantes, on
the 2400ton test frame (see e.g. [10]).
Figure 1 . Subrope on test frame
Both series of tests used the same instrumentation to
measure strains in the central rope section away from the
splices : wire transducers clamped to the rope for stiffness
measurements and digital cameras linked to an image analysis
system for break tests and to check wire transducer
measurements. The tension in the rope is measured by the
straingages load cell system of the test frame. All
measurements are recorded, at a high data acquisition rate (up
to 5 Hz for dynamic stiffness), and stored on a PC for
subsequent analysis. Accurate and continuous time series of
load and elongation are thus available for analysis.
Materials
The tests described here were performed on 8strand
braided polyester subrope samples, with a 70 t breaking
strength, and on test lengths of a full size (800 t MBS) parallel
construction rope based on the same subropes. The fibre is a
standard high tenacity grade with marine finish. Oliveira S,
Portugal supplied all samples.
Test sequences
The standard test to measure quasistatic stiffness
described in [7] consists of three 1hour loadunload cycles
between 10% and 30% (of the break strength), applied to a
rope which has been fully beddedin (i.e. an initial loading to
50% , with the load maintained constant for 30 minutes, then
100 cycles between 10 and 30%).
In order to widen the scope of this characterisation, an
extended quasistatic stiffness test was developed : As shown in
Figure 2, a series of loadunload cycles is applied, similar to
those of the standard test, all with the same 20% range but at
different sets of load levels, starting and ending with a set of
cycles at 1030%. Four cycles are applied at each load
condition, in order to verify stabilisation.
0
100
200
300
400
500
600
0 5 10 15 20 25 30 35 40 45 50
Time, h
F
o
r
c
e
,
k
N
Linear
density
Figure 2 . Extended quasistatic stiffness sequence
Force versus time applied to subrope samples 3 and 4
On two samples , an extended sequence was applied to a
rope that has been previously subjected to beddingin cycles
following the standard procedure quoted before, or equivalent
(on one of them). On two other samples, a sequence was
applied to a rope with a lower degree of beddingin cycles,
obtained by applying instead the procedure defined in [7] for
the linear density test (i.e. an initial loading to 20%, then 100
cycles between 15 and 25%, ending at a 20% load). Samples
were then loaded to failure. A fifth sample was used for
dynamic stiffness measurements (see section 5 below).
4 Copyright 2008 by ASME
Test results  stiffness
Figure 3 shows an example of the resulting forcestrain
plot during the extended quasistatic stiffness test (continuous
record over 48 h).
0
100
200
300
400
500
600
0 1 2 3 4 5 6 7 8 9 10
Strain, %
F
o
r
c
e
,
k
N
Figure 3 . Extended quasistatic stiffness sequence, sub
rope samples 3 and 4.
From the elongations at the end of the last two half cycles,
a secant stiffness is obtained (without extrapolation at this
stage). The quasistatic stiffness is found to be increasing with
mean load, with a trend similar to that of the dynamic stiffness
(see Figure 5). As shown in Figure 4, there is a close match
between the results of all four samples, and a very limited
effect of previous beddingin : stiffness for the samples with
mild beddingin are only marginally lower than for well
beddedin samples, the lowest point corresponding to the very
first load set after measurement of L
20
.
0
8
16
24
32
0 10 20 30 40 50 60 70 80
Mean Load (% of MBS)
K
r
S
Figure 4 . Quasistatic stiffness versus mean load,
all four samples.
Comparing (see Figure 5) the results of the measurements
performed on subropes with those from full size 800 ton break
load ropes, an excellent agreement is found for both the quasi
static and the dynamic stiffness when expressed in tenacity unit
(N/Tex), as on Figure 5. Same result would be found if these
stiffnesses are expressed as a nondimensional Kr, using a
normalisation by MBS of the full size rope (see [8]).
Influence of mean load
0
2
4
6
8
10
12
14
16
18
0 10 20 30 40 50 60 70
Mean load, %
S
t
i
f
f
n
e
s
s
,
N
/
t
e
x
Rope DYN
Subrope DYN
Rope QS
Subrope QS
Figure 5 . Correlation between results for subrope and full
size ropes; quasistatic and dynamic stiffness.
Test results  mean elongation
The tests also provided information on the mean
elongation of the rope, and its dependence on the loading
history.
On the samples to which the linear density test sequence
was applied, the elongation at a 20% load, at the end of the
sequence provides the reference length L
20
for rope purchasing,
according to the length measurement method in the
standard [6].
Besides, a mean elongation at 20% can be obtained from
each set of 1030 cycles (indeed taken as the mean between the
elongation at the end of the last 10% and 30% cycles), or
inferred from the fully relaxed condition (see 4 below). As
shown in Figure 6, a very good agreement between the results
of each group of two ropes is found.
20
30
40
50
60
70
80
90
0.0 0.5 1.0 1.5 2.0 2.5
elongation (% ) above L
20
F
m
a
x
(
%
o
f
B
S
)
OL1 OL2
OL3 OL4
Figure 6 . Mean elongation at a 20% load
all four samples.
The mean elongation of wellbeddedin samples is stable,
about 1.6% higher than L
20
, for all loads up to 40%. For higher
maximum loads, further (delayed) permanent elongation is
observed, increasing almost linearly until 70% (the maximum
load for these two samples).
5 Copyright 2008 by ASME
For the samples with a milder beddingin, the elongation
increases quickly when the load exceeds the maximum seen
during beddingin, confirming that L
20
is a lowerbound of the
installed length, and is always lower than for well beddedin
ropes, up to 70% load. Given the time frame of the test (within
48h of the very first loading of the rope, i.e. much less than the
time required to install a system and to have it actually
operating) the resulting permanent elongations may be also
considered as a very lower bound.
4 INTERPRETATION
QS stiffness
From the fitting of creep and recovery plateaus (greatly
facilitated by the very good accuracy of load control) and the
extrapolation of last cycle plateaus, the elongations for
different levels and durations can be obtained : the standard
duration of 12h and 7 days, that would be more representative
of slowly growing events, such as loop currents, are considered
here. When normalising loads by full size rope MBS and
elongations by the length at 20%, including permanent
elongation), the quasi static stiffness for the standard levels (10
and 30%) is found, , to be about 14.5 for 12h (i.e. 17% less
than for the 1/2h duration in test), and about 12.5 for 7 days,
i.e. 12 % lower than for 12h. This decay is consistent with
earlier findings, as shown in Figure 7 .
QS PRACTICAL characteristics
For other loads, a stiffness between two levels could be
calculated in the same way, but this is not very representative
of the situation in a stationkeeping system, where line tension
will generally not oscillate between such levels : Together with
vessel offset, the tension will gradually increase (or decrease)
more or less monotonically in most cases, and over the same
period of time, from initial tension to a maximum (or
minimum) value which will be different in each line.
Then, by considering the relevant points, a practical
characteristic for the specified loading time can be obtained,
that is shown in Figure 8. This characteristic was derived from
data from the different samples: a very consistent behaviour
was found.
10
11
12
13
14
15
16
0 1 10 100 1000
loading time (hours)
K
r
S
1
0

3
0
OL1
Clarom 2000
Figure 7 . QS stiffness KrS
1030 ,
blue : data from reference [3] (corrected).
0
10
20
30
40
50
60
70
80
90
3.0 2.0 1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0
elongation (% ) above L
20
T
e
n
s
i
o
n
(
%
o
f
B
S
)
Xij 12h
Xij 7d
QS 12h
QS 7d
included Xp
fully relaxed
oliv1 Xtrap2 f
Figure 8 . QS PRACTICAL characteristics
included Xp, and fully relaxed characteristic (see text)
6 Copyright 2008 by ASME
A first observation is that, for the 12h (standard) loading
time, this characteristic is almost linear, i.e. the standard quasi
static stiffness derived from tests between 10 and 30 % is valid
over a large range, from about 7 to above 70%.
For the 7 days loading time, the slope of the characteristic
(dX/dT =1/Kr ) is higher between 10 and 30 % (as per the
lower stiffness noted above) but, above 30%, is the same as for
12h.
Taking the pretension T
0
as a start point, this characteristic
(elongation versus load) can be written as :
X(T)  X(T0) = ( T  T
0
) / Krs
for T between 10 and 30% (3)
X(T)  X(T0) = ( 30  T
0
) / Krs
+ ( T  30 ) / Krs
12h
for T above 30% (4)
where Krs
* ( u + 1.8 * u^3.6 )
where u = 1  T / 10 (5)
In a system working at low tensions in leeward lines, this
effect will have a similar effect as the catenary effect in
weighty components of the line.
Fully relaxed condition
During the test, at a given load, either creep can be
observed, or recovery from a larger elongation. There should be
a stable point in between. As discussed below, this point can be
obtained by extending the loading duration to t
, the time at
the intersection point of elongation versus (log) time of two
cycles terminating at same level, and correcting for the
permanent elongation (when applicable).
Taking t
as 4 10
7
s , i.e. about 15 months, all points
converge to a single curve, that is the non linear, but elastic
(reversible) fully relaxed characteristic, i.e. the characteristics
Xrx(T) for infinitely slow rate of loading , also shown (round
dots) on Figure 8.
Arriving at or approaching a point on the fully relaxed
characteristic in a shorter time than t
/ ) for << t
(9)
Xv = 0 for >> t
(with a transition over one or two decades around t
).
where a
c
is the creep (or recovery, then <0) per decade. This
supports the method above to determine t
. However a
c
is not
an intrinsic material constant but depends on load history, and
the spring and damper model suggested above does not fully
address the growth of Xv. Besides, a closer look at time traces
shows that Xpe is not only a function of Tmax, but depends also
on the time under load (as also shown by Figure 6), so Xp and
Xv are likely not fully separable.
The characteristics above are thus given for the Designer,
as PRACTICAL QS characteristics. Development of a true
time domain rheological model still requires further effort.
5 DYNAMIC STIFFNESS
The Dynamic stiffness is modelling the nearelastic
response of the rope to cyclic actions (both low frequency and
wave frequency) induced by the environment.
In tests (see [7] and [8]), after beddingin, the rope is
cycled around a preset mean tension. For harmonic (constant
amplitude, sinusoidal) loading, at least 100 cycles are typically
used today and, as the loadelongation graph converges to a
fairly linear relation, a stiffness can be defined. This is taken as
the mean slope over several cycles at the end of the sequence
(i.e. 500 data points, in the results below). Normalising loads
by rope MBS, the nondimensional dynamic stiffness KrD is
then obtained.
7 Copyright 2008 by ASME
In order to complement the characterisation work
performed earlier (see [3]), a number of dynamic stiffness tests
were performed, on the same subropes and full size ropes as
those above. Similar tests were also performed on another full
size rope with a different polyester fibre (see [11]). Results will
be briefly summarised, together with a discussion of the
dependence of dynamic stiffness on testing parameters.
Cycling period
A series of tests was performed with the same mean load
and amplitude, and cycling periods from 12.5 s to 500 s.
Results confirmed that the effect of this parameter on dynamic
stiffness is negligibly small (1.4% variation over the above
range). This is in accordance with earlier findings from bi
harmonic loading, and those of stochastic loading below.
Load history
During tests with harmonic loading, at a constant mean load,
the stiffness quickly increases in the first cycles, then tends to
stabilise (at least apparently), and 100 cycles minimum are
typically used today to obtain results in a practical time frame.
There is however some effect of previous load history affecting
the results. Indeed, the increase of stiffness during cycling is
quite linear with log of time, as is the variation of mean
elongation (be it creep or recovery) over the same time: From
different tests, it seems that the stabilisation of the dynamic
stiffness (toward the maximum stiffness for applied mean load)
and that of mean elongation (towards the fully relaxed
condition) are closely related.
Load range, stochastic loading
From a series of tests under harmonic loading, with
different mean load and range, the effect of both parameters
can be identified. Figure 9 shows the influence of load range,
with the dynamic stiffness normalised by the value at a 5%
range (obtained from a fit of data) . The dynamic stiffness is
clearly decreasing when range is increased, more sharply
towards higher amplitudes and lower mean loads. This may be
due, at least in part, to the increase in rope temperature
observed during such tests (see [12]).
However, the real loading is not harmonic. Series of tests
were thus performed on the two full size ropes, using 1hour
time series obtained by mooring analysis (i.e. a wide band
signal, with realistic low frequency and wave frequency
contents), and different combinations of mean load and
amplitude. Figure 10 shows an example of applied time series
and resulting loadelongation plot. It is noteworthy that the
loadelongation is very close to a single overall linear relation
(linear regression over 18000 points). A closer look to signals
clearly shows effect of neither the amplitude of individual
cycles, nor the period of underlying (low and wave frequency)
components.
Besides, from different tests it is apparent (see Figure 9)
that this stiffness is almost independent of the overall (min to
max) amplitude of the signal, and is slightly higher (within 1
to +5%) than the stiffness under harmonic loading at a 5%
0.75
0.80
0.85
0.90
0.95
1.00
1.05
1.10
0 2 4 6 8 10 12 14
Amplitude (range/2) (% of BS)
K
r
D
/
K
r
D
2
.
5
Figure 9 . Influence of load range on dynamic stiffness
harmonic loading (10 , 20, 30, 40% light to dark blue),
stochastic loading (red and pink)
1500
2000
2500
0 10 20 30 40 50 60
time, minutes
F
o
r
c
e
,
k
N
1500
1750
2000
2250
2500
3 3.1 3.2 3.3 3.4
Strain, %
F
o
r
c
e
,
k
N
Figure 10 Example of stochastic loading (1h)
a) time trace;
b) resulting loadelongation and linear regression
range. In this respect, it must be observed that in such realistic
stochastic signals, the energy, being related to the standard
deviation of the signal, is 5 to 10 times lower than in an
harmonic signal having the same range. Besides, it seems that a
better stabilisation, i.e. closer to real conditions, was achieved
by these sequences.
Thus, there is no point in taking into account an effect of
load range in analysis.
8 Copyright 2008 by ASME
Mean Load
Mean load has been identified before as the dominant
parameter affecting the dynamic stiffness, and a linear relation
is currently used. As an update of [3] (where a comparison with
the relation in [1] and those by other authors was shown),
Figure 11 shows the variation of dynamic stiffness with mean
loads, for 11 different ropes from 6 rope manufacturers,
including the results of recent tests.
Besides the overall trend for an increase of stiffness (about
linearly) with mean load, some scatter is observed that can be
attributed to several reasons. A first set is the effect of the
parameters noted above : load range (10% maximum on this
figure), load regime and effect of previous history/stabilisation.
There is also some small effect of measurement inaccuracy and
small discrepancies between the testing laboratories (three, in
addition to IFREMER), along the measurement and data
processing chain.
The principal reason is, however, the differences between
the ropes themselves (material and construction), that are
reflected in the dynamicstiffnessatendofbeddingin Krebi.
For the ropes in Figure 11, Krebi (shown by bigger marks, at
20% mean load) is in the range of 18.5 to 23, i.e. the range of a
normally stiff rope according to A15.7 in [8].
Normalising data by Krebi, for each rope, significantly
reduces the scatter, but does not lead to a single relation, due to
the other reasons quoted.
Values for Analysis.
As shown in Figure 11, all data points are lying within the
envelope formed by the design (upperbound) stiffness and the
minimum stiffness, as proposed in [8] :
KrD = 18.5 + 0.33 ML (design)
KrD
m
= 15 + 0.25 ML (lower bound),
where ML is mean load (in % of MBS).
Only a few points are marginally outside.
Towards low mean loads, results now available confirm
that stiffness is decreasing, and that the linear relation is on the
conservative side. Therefore, the minimum value proposed
earlier has been dropped. Towards higher mean loads, there is a
trend for the stiffness to increase less than predicted by the
linear relation. This was attributed to an insufficient number of
cycles in earlier test procedures, and does not appear in some
results, therefore was not considered.
The envelope given above is thus adequate for the design
of a mooring system, i.e. typically some significant time in a
project before a particular product is selected and
manufactured. This envelope covers the possible range of
stiffness of a normally stiff rope. Then the standard
measurements during rope qualifications will confirm that the
purchased product has properties within the range used for
Design. This envelope should be adjusted if stiffer ropes are
considered. Besides, for some cases (e.g. a site assessment)
where the rope is already known, a more accurate relation
could be defined. This will however require that further
measurements are made, e.g. using stochastic loading, and that
due care is taken is the derivation of engineering values, that
cannot be simply taken as the raw results of a few tests.
15
20
25
30
35
40
0 10 20 30 40 50 60
Mean load ( % MBS )
K
r
D
Krebi
Figure 11 . Dynamic stiffness as a function of mean load :
test data from [3] (blue and green), recent (red and pink), and relations in ref [8]
9 Copyright 2008 by ASME
CONCLUSION
Within the French CLAROM fibre rope project, tests have
been performed to confirm and complement the currently used
practical engineering model that was derived from earlier work.
This work focused principally on the response of ropes to slow
variations of mean load, under the effect of changing weather
conditions, usually modelled by the Quasistatic stiffness. In
addition to the standard tests described in [7], extended test
sequences were defined. From these specific sequences, and the
number of data points provided, an interpretation can be given
that overcomes the discrepancy between the time scale of tests
(one or two days), and the real world.
Tests reported in this paper were principally performed on
subropes. Some full size rope test results are also reported.
The scalability of results, an already known fact, was shown.
Due attention was given to the accuracy of load and elongation
measurement methods. Continuous time traces were obtained :
this is also an important factor for interpretation of test data.
Based on the interpretation of test results, a practical Q S
characteristic can be defined, addressing monotonic changes of
line mean load, from around the line pretension. This
characteristic is an extension of the current QuasiStatic
stiffness KrS. Two characteristics are proposed for the 12h
(standard) loading time, and for 7 days, a loading time more
appropriate for some other design conditions (e.g. the effect of
a loop current). Both are scaleable by KrS.
A first observation is that, for the 12h loading time, the
quasistatic stiffness (Krs
1030
) derived from standard tests is
valid over a large range of tensions, from about 7 to above
70%, i.e. the characteristic is linear over this range.
Another significant observation is that, for tensions below
10%, the characteristics show a clear increase of compliance
with decreasing load : this should be considered for systems
working at low tensions in leeward lines.
As background to the above characteristics, further
separation of the rope response is presented, based on the
derivation of a fully relaxed characteristic (the characteristic for
infinitely slow rate of loading), and two additional terms : a
permanent (non recoverable) elongation, and a (remaining) un
developed viscoelastic elongation, modelling the effect of
loading rate. Nevertheless, development of a true time
domain rheological model and a definitive understanding of
underlying mechanism will require further effort.
For the dynamic stiffness, the discussion of the
dependence of dynamic stiffness on testing parameters, based
on recent data, highlighted mean load as the principal
parameter under real stochastic loading. This confirmed the
adequacy of current practice in the analysis of a system, of
modelling the dynamic stiffness as a linear function of mean
load only. Besides, for a particular rope, it is important to note
that load history and other effects will always affect the test
results, thus due care is to be taken in the derivation of
engineering values that cannot be simply taken as the raw
results of a few tests. Using stochastic loading time series for
the testing appears an efficient method in this respect.
ACKNOWLEDGEMENTS
The work presented in this paper was performed within the
framework of the CLAROM group, the French Club for
research activities on offshore structures. The contributions of
project partners are gratefully acknowledged.
The authors wish also to thank Oliveira S for providing
test ropes, their colleagues that contributed to performing tests
and analysing results, and their respective companies for
permission to publish this paper.
The views expressed are those of the authors, and do not
necessarily reflect those of their respective companies.
REFERENCES
[1] Del Vecchio CJ M (1992), Light weight materials for
deep water moorings, PhD thesis University of
Reading
[2] Grosjean F, Davies P, Francois M, (2005), Synthetic
Fiber Ropes mooring: technical status and stiffness
prediction, Proc CMOO4
[3] Franois M, Davies P, (2000), Fibre rope deep water
mooring : a practical model for the analysis of polyester
mooring systems , IBP 247 00, Rio Offshore 2000
[4] Franois M, Giulivo R, (2000), Practical procedures for
fibre rope moorings, Continuous Advances in Mooring
& Anchoring, Aberdeen
[5] Davies P et al., (2002), Synthetic mooring lines for
depths to 3000 meters, OTC14246
[6] Franois M. (2005), Fiber ropes for Stationkeeping :
Engineering properties and qualification procedures,
OCEANS 2005, Washington
[7] ISO 18692:2007, Fibre ropes for offshore station
keeping  Polyester
[8] Bureau Veritas, (2007), Certification of fibre ropes for
Deep water offshore services, NI432R01
[9] Lechat C, Bunsell AR, Davies P, Piant A, (2006),
Mechanical behaviour of polyethylene terephthalate &
polyethylene naphthalate fibres under cyclic loading,
J ournal of Materials Science, Vol. 41, pp 17451756
[10] Davies P, Baizeau R, Grosjean F, Franois M, (1999),
Testing of large polyester cables for mooring line
applications, Proc ISOPE, Brest, p360
[11] Davies P, Lechat C, Bunsell A, Piant A, Franois M,
Grosjean F, Baron P, Salomon K, Bideaud C, Labb J P,
Moysan AG., (2008), Deepwater Moorings with High
Stiffness Polyester and PEN Fiber Ropes OTC 19315
[12] Davies P et al, (2008), Infuence of fibre Stiffness on
Deepwater Mooring Lines , OMAE200857147
OTC19450PP
Impact of the use of FullQTF on LNGC Moored in Shallow Water Studies
Cdric BRUN, Flvia REZENDE, Damien COACHE, J ohan MOMBAERTS (Bureau Veritas)
Copyright 2008, Offshore Technology Conference
This paper was prepared for presentation at the 2008 Offshore Technology Conference held in Houston, Texas, U.S.A., 58May2008.
This paper was selected for presentation by an OTC program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Offshore Technology Conference and are subject to correction by the author(s). The material does not necessarily reflect any position of the Offshore Technology Conference, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Offshore Technology Conference is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of OTC copyright.
Abstract
The prediction of slow drift motions for the design of a mooring system is usually made using the Newman approximation
[1], based on the assumption of a very low resonance frequency of the system and small contribution of the second order
wave fields. This hypothesis is commonly satisfied for most parts of the mooring systems in deep water. However, this is not
the case for LNG terminals moored in shallow water.
Unlike the Newman approximation, the FullQTF formulation to compute the low frequency wave loads is more accurate
but requires much longer time of computation, which presents limitations in practice when a large quantity of simulations is
needed.
Further to the work presented in [2] on the quadratic transfer function (QTF) of lowfrequency loading, a new
approximation has been developed in [3]. The F1 approximation gives comparable results to the FullQTF and presents the
interesting aspect that the loads time series can be reconstructed by means of simple summations, presenting the same
efficiency in computation time as Newman approximation.
In this paper, main parameters of mooring systems are analyzed to evaluate the impact of the choice of each method,
Newman, F1 or FullQTF. Indeed, this choice is a compromise between calculation time and accuracy of results. The
conclusions raised are underlined in the study of an LNG terminal.
2 OTC OTC19450PP
Introduction
More and more LNG carriers are built across the world to transport the liquefied gas from a location to another. These
locations sometimes propose LNG terminals in the proximity of the harbors, thus in shallow water. For those terminals, the
accurate prediction of lowfrequency wave loads is a key issue in the design of mooring systems.
The approximation proposed by Newman (1974) is generally used for the simulations of low frequency loading due to the
fact that it is based only on the diagonal terms of the QTF (Quadratic Transfer Functions). Those terms, the mean drift loads,
are contributed only by the firstorder wave field and body motions and are easily calculated once the first order solution is
obtained. In moderate and deep water depths it can be shown that the second order wave field contribution is small
comparing to the first order one. This way, the assumptions applied to the Newman approximation are satisfied and the
results obtained through this approximation can be considered adequate. However, this is not the case in shallow water,
where the contribution of the second order wave field can not be neglected.
The use of the complete QTF matrix gives more accurate estimations of the lowfrequency loads, but it requires the
solution of the second order problem and the time series reconstruction is more timedemanding. Recently, a new
approximation has been presented in [2] and [3] considering that the resonant frequencies of mooring systems are often small
(<<1). This approximation consists in developing the QTF as an expansion of difference frequency and keeping the
terms dependent on the first order quantities T
0
and another term T
1
linearly proportional to . From this expansion two
approximations can be derived: the F0 approximation, which is equivalent to the Newman approximation, and the F1
approximation.
In this paper the F0 (Newman) and F1 approximations are compared to the FullQTF considering the influence of the
water depth and mooring stiffness. The time efficiency of each method is also addressed.
Finally, the conclusions raised are highlighted in an LNG terminal example, under specific environmental conditions.
QTF Formulations
The full quadratic transfer function (QTF) of lowfrequency loads is composed of two parts: one depends on the quadratic
product of the firstorder quantities and another is contributed by the secondorder potentials:
( ) ( ) ( )
2 1 2 2 1 1 2 1
, , , T T T + = (1)
The Newman approximation is obtained by disregarding the second term of equation (1). This approximation is based on
the assumption that 0, where =
1

2
. For shallow water applications, however, the secondorder potentials
contributions can not be neglected and the Newman approximation may largely underestimate the low frequency loading.
Recently, a new approximation has been presented in [4]. Assuming <<1, the QTF is developed as an expansion:
( ) ( ) ( ) ( )
2 1 0
2 1
, , , + + = T i T T (2)
The quadratic transfer function T(
1
,
2
) is then composed of one component T
0
() depending on =(
1
+
2
)/2 and
another T
1
() linearly proportional to =
1

2
. From this expansion two approximations are derived. The F0
approximation, which is equivalent to the Newman approximation, is based only on the use of T
0
() so that is of ) 1 ( and the
F1 approximation that consider also the term T
1
() so that is of ) ( .
Considering the fact that:
( ) ( ) ( ) [ ] [ ]
2 1 / 0 1 / 0 1 / 0
, ,
2
1
, + + =
j j i i
T T T with
2
j i
+
= (3)
The F1 approximation in [3] can be rewritten as:
( ) ( ) ( ) [ ] ( ) ( ) ( ) ( ) [ ]
j j i i j i j j i i j i
T T i T T T , ,
2
1
, ,
2
1
,
1 1 0 0
+ + + = (4)
OTC OTC19450PP 3
Time series reconstructions
For an irregular sea, the wave elevation can be noted as follows:
( )
( )
=
+
=
x k t i
N
j
j
e a t
1
with
j j j
d S a = ) ( . 2 (5)
Where S() is the irregular wave spectral density and d is the sampling space of the spectrum.
The low frequency loads can now be defined in time domain as below.
FullQTF formulation
FullQTF is a two dimensions complex matrix. The Real part of the matrix is symmetrical whereas the Imaginary one is
antisymmetrical. The lowfrequency drift loads time signal using the FullQTF formulation is obtained by a double
summation:
( )
( ) ( ) ( ) ( )
[ ]
=
= =
+
N
i
N
j
x k k t i
j i j i FullQTF
i j i j i j
e a a T t F
1 1
. . , ) (
(6)
F0 approximation (Newman)
This method consists in approximating T(
i
,
j
) by a function of T(
i
,
i
) and T(
j
,
j
). Thus, the FullQTF double
summation can be transformed into the simple following formulation:
( )
( ) ( )
=
=
+
=
+
N
j
x k t i
j
N
j
x k t i
j j j
j j j j j j
e a e a T t F
1 1
0
0
. . , ) (
(7)
F1 approximation
Introducing (4) into (6), the time series reconstruction for the F1 approximation presented in [3] can be obtained as
follow:
( )
( ) ( )
=
=
+
=
+
N
j
x k t i
j
N
j
x k t i
j j j
j j j j j j
e a e a T
dt
d
t F t F
1 1
1
0 1
. . , ) ( ) (
(8)
Remarks
The FullQTF formulation is the most accurate method to describe the physical phenomenon. However this kind of
calculation is very time consuming due to the double summation on the frequencies and the calculation of the FullQTF
matrix. The F1 approximation is a new method that provides more accurate results comparing to the F0 approximation and
presents the advantage that the time series reconstruction can be described as simple summations instead of double
summations, hence faster to compute.
Numerical results
Different parameter influences are studied before computing a complete example of LNG terminal. Thus, varying water
depth, wave period and mooring stiffness, differences between the three methods are evaluated in terms of accuracy and time
efficiency.
I nfluence of water depth and peak period of sea state
In this case, numerical calculations are performed for a standard 138km
3
LNG vessel with main dimensions (Length,
Breadth and Draft) = (274m, 44.2m and 11m), respectively. The half hull mesh is composed of 2204 panels.
At a first stage, the vessel is considered fixed and the lowfrequency loads are calculated in frequency domain for different
waterdepths and for sea states with Hs=2.0m and peak periods varying from 6s to 18s.
Figures 1 to 6 present the spectral density of lowfrequency loads. For each peak period the left bar is used for F0
approximation, the middle bar for F1 approximation and the right bar for the FullQTF. Two differences of frequencies ()
are evaluated, 0.025 rd/s and 0.05 rd/s, which are directly linked to the resonance of the mooring system. It can be noticed
that the F0 approximation largely underestimates the loads in shallow water comparing to the FullQTF. For moderate water
4 OTC OTC19450PP
depth, such like 60m, the F0 approximation may give conservative results except for very long waves. In all the cases the F1
approximation presents results which are comparable to those obtained through the FullQTF.
0.00E+00
5.00E+04
1.00E+05
1.50E+05
2.00E+05
2.50E+05
3.00E+05
3.50E+05
4.00E+05
4.50E+05
Tp=6s Tp=8s Tp=10s Tp=12s Tp=14s Tp=16s Tp=18s
S
p
e
c
t
r
a
l
d
e
n
s
i
t
y
(
k
N
^
2
.
m
^
2
.
s
)
F0 Approximation F1 Approximation Full QTF
0.00E+00
2.00E+05
4.00E+05
6.00E+05
8.00E+05
1.00E+06
1.20E+06
1.40E+06
Tp=6s Tp=8s Tp=10s Tp=12s Tp=14s Tp=16s Tp=18s
S
p
e
c
t
r
a
l
d
e
n
s
i
t
y
(
k
N
^
2
.
m
^
2
.
s
)
F0 Approximation F1 Approximation Full QTF
Figure 1 LF spectral density WD 15m Tn ~251s Figure 2  LF spectral density WD 15m Tn ~126s
0.00E+00
2.00E+09
4.00E+09
6.00E+09
8.00E+09
1.00E+10
1.20E+10
1.40E+10
1.60E+10
Tp=6s Tp=8s Tp=10s Tp=12s Tp=14s Tp=16s Tp=18s
S
p
e
c
t
r
a
l
d
e
n
s
i
t
y
(
k
N
^
2
.
m
^
2
.
s
)
F0 Approximation F1 Approximation Full QTF
0.00E+00
5.00E+09
1.00E+10
1.50E+10
2.00E+10
2.50E+10
3.00E+10
3.50E+10
4.00E+10
4.50E+10
5.00E+10
Tp=6s Tp=8s Tp=10s Tp=12s Tp=14s Tp=16s Tp=18s
S
p
e
c
t
r
a
l
d
e
n
s
i
t
y
(
k
N
^
2
.
m
^
2
.
s
)
F0 Approximation F1 Approximation Full QTF
Figure 3  LF spectral density WD 30m Tn ~251s Figure 4  LF spectral density WD 30m Tn ~126s
0.00E+00
1.00E+09
2.00E+09
3.00E+09
4.00E+09
5.00E+09
6.00E+09
7.00E+09
8.00E+09
Tp=6s Tp=8s Tp=10s Tp=12s Tp=14s Tp=16s Tp=18s
S
p
e
c
t
r
a
l
d
e
n
s
i
t
y
(
k
N
^
2
.
m
^
2
.
s
)
F0 Approximation F1 Approximation Full QTF
0.00E+00
1.00E+09
2.00E+09
3.00E+09
4.00E+09
5.00E+09
6.00E+09
7.00E+09
8.00E+09
Tp=6s Tp=8s Tp=10s Tp=12s Tp=14s Tp=16s Tp=18s
S
p
e
c
t
r
a
l
d
e
n
s
i
t
y
(
k
N
^
2
.
m
^
2
.
s
)
F0 Approximation F1 Approximation Full QTF
Figure 5  LF spectral density WD 60m Tn ~251s Figure 6  LF spectral density WD 60m Tn ~126s
Regarding the two resonance periods studied from Figure 1 to Figure 6, it is shown that the differences between the three
methods increase for smaller resonance periods. To highlight this phenomenon, the mooring stiffness influence is studied in
time domain.
Mooring stiffness influence
LNG carrier motions are analyzed using different mooring system stiffnesses. To study the mooring stiffness influence, a
typical reference characteristic shown in Figure 7 is multiplied by coefficients from 0.2 to 2.0 with a step of 0.2.
OTC OTC19450PP 5
10000
5000
0
5000
10000
15000
20000
25000
0 5 10 15 20 25 30 35 40 45 50 55
Horizontal distance (m)
H
o
r
i
z
o
n
t
a
l
t
e
n
s
i
o
n
(
k
N
)
Figure 7: Mooring system characteristic
In each study case, the same environment is imposed. It is composed of waves (Jonswap spectrum with Hs=3m, Tp=10s
and =3.3). Only surge motion direction has been studied for head sea condition.
Figure 8 and Figure 9 give a comparison of the vessel motions time series for three different mooring stiffnesses, (from
the softest on the left to the stiffest on the right), comparing respectively F0/FullQTF and F1/FullQTF.
20% of reference stiffness
6
4
2
0
2
4
6
8
1000 1000 3000 5000 7000
Time (s)
V
e
s
s
e
l
p
o
s
i
t
i
o
n
(
m
)
Reference stiffness
6
4
2
0
2
4
6
0 1000 2000 3000 4000
Time (s)
V
e
s
s
e
l
p
o
s
i
t
i
o
n
(
m
)
F0 FullQTF
200%of reference stiffness
2
1.5
1
0.5
0
0.5
1
1.5
2
0 1000 2000 3000
Time (s)
V
e
s
s
e
l
p
o
s
i
t
i
o
n
(
m
)
Figure 8: Vessel position time series under F0 and FullQTF drift loads
With reference to Figure 8, it can be clearly shown that changing the global mooring stiffness accentuates the differences
between the F0 and FullQTF calculations. Indeed, when the mooring system is very soft, the two resulting signals from F0
and FullQTF are quite similar. On the contrary, when the stiffness is increasing, the time series become more and more
different.
20%of reference stiffness
8
6
4
2
0
2
4
6
8
1000 1000 3000 5000 7000
Time (s)
V
e
s
s
e
l
p
o
s
i
t
i
o
n
(
m
)
Reference stiffness
6
4
2
0
2
4
6
0 1000 2000 3000 4000
Time (s)
V
e
s
s
e
l
p
o
s
i
t
i
o
n
(
m
)
F1 FullQTF
200% of reference stiffness
2
1.5
1
0.5
0
0.5
1
1.5
2
0 1000 2000 3000
Time (s)
V
e
s
s
e
l
p
o
s
i
t
i
o
n
(
m
)
Figure 9: Vessel position time series under F1 and FullQTF drift loads
Figure 9 shows that the differences in the vessel motions time series when using the F1 approximation or FullQTF
formulation are not as consequent as when using the F0 approximation. Indeed, it seems to suggest that the F1 expression
gives suitable results even for the stiffest mooring stiffness evaluated.
In case the F0 approximation is used with a very soft mooring system, the ratio between the F0 and FullQTF time series
standard deviation of loads is about 0.8 whereas it could be less than 0.4 for a stiff mooring system (see Figure 10). In fact,
the stiffer is the mooring system, the bigger the difference between the F0 and FullQTF time series.
However the ratio between F1 and FullQTF time series standard deviation is less than 1.10 for most of the mooring
stiffnesses, except for the stiffest mooring system with a maximum of 1.25 (see Figure 10). In addition, it can be seen that the
ratio is always greater than 1.00, which seems to suggest that the F1 approximation is conservative.
6 OTC OTC19450PP
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 50 100 150 200
Stiffness rate (%)
R
a
t
i
o
o
f
S
t
d
.
D
e
v
.
F0/FullQTF F1/FullQTF
Figure 10: F0 and FullQTF time series standard deviation relative difference function of the stiffness rate
Computation efficiency in time domain reconstruction
As shown in equations (6), (7) and (8), the FullQTF reconstruction is based on a double summation on frequencies
whereas F0 and F1 are based on simple summations. Figure 11 shows an example of comparison between the methods when
irregular waves are reconstructed by 200 elementary airy waves.
0%
20%
40%
60%
80%
100%
FullQTF F0 F1
C
a
l
c
u
l
a
t
i
o
n
t
i
m
e
r
a
t
e
(
%
o
f
m
a
x
)
Figure 11: Calculation time comparison between FullQTF, F0 and F1
A solution to make FullQTF calculations faster is to reduce the number of terms of the double summation. Thus, a new
parameter called
max
is introduced, considering that the vessel is less sensitive to the difference frequency greater than
max
. Only terms of the FullQTF matrix that satisfy the following inequality are taken into account:
max
<
j i
(9)
Remark: if
max
>
N

0
, the complete drift load is calculated whereas if
max
= 0. only the mean drift load is computed.
Figure 12 shows that the computation time decreases significantly when
max
< 0.4.
0
20
40
60
80
100
120
0 0.2 0.4 0.6 0.8 1 1.2
max (rad/s)
T
i
m
e
r
a
t
e
(
%
o
f
m
a
x
)
Figure 12: Double summation time consumption in function
of
max
0
20
40
60
80
100
120
0 0.2 0.4 0.6 0.8 1 1.2
max (rad/s)
E
r
r
o
r
r
a
t
e
(
%
)
Figure 13: F(t) standard deviations error rate
Choosing
max
= 0., there is no more time dependency, meaning that the time series is a constant signal, and standard
deviation is logically 100%. In Figure 13, it is possible to see that the wave drift load time series is well recomposed for
s rad / 3 . 0
max
> , that gives about 20% reduction of the calculation time.
This conclusion is set for this particular example. The main notion is that it is possible to analyze a mooring system and
quickly determine a
max
in function of its resonance frequency.
OTC OTC19450PP 7
Study case: LNG terminal in shallow water
In this section, one practical example is used to compare the different low frequency wave loads calculation methods in
order to illustrate the results previously obtained.
The selected mooring system is a typical configuration for LNG terminals in shallow water. This system, called Soft
Yoke (SY) is composed of a fixed structure, mooring legs and the yoke (see Figure 14)[5].
Figure 14: SY for LNG terminals (Cortesy of SBM)
The mooring system is modeled by its characteristic horizontal distance/horizontal tension used as the reference one for
the stiffness influence study (see Figure 7). The neutral point is 30 meters from the yoke nose. The mooring system is
symmetrical. The LNG carrier can turn around the fixed structure and around the yoke nose. As previously presented, the
main properties of the 138km
3
LNG carrier are respectively (Length, Breadth and Draft) = (274m, 44.2m and 11m). The
water depth is 15 meters and studies are done with 3 hours time series of vessel motions and SY tensions.
In order to be able to compare correctly the different temporal signals, the same wave spectrum is imposed. The selected
spectrum is a Jonswap one defined by respectively (Hs, Tp, ) = (3m, 10s, 3.3). The wave spectrum is discretized into two
hundred elementary Airy waves.
A current load is applied in order to change the vessel azimuth with respect to the wave. Different incidences are studied,
thus different relative wave headings. Current is supposed to be constant with a velocity of 1.2m.s
1
.
Table 1: Simulation results under different environmental conditions
F0 F1 FullQTF F0 F1 FullQTF F0 F1 FullQTF F0 F1 FullQTF
Azimuth std. dev. (deg) 0.00 0.00 0.00 2.88 2.85 3.16 1.47 1.65 1.69 2.27 2.45 2.50
North std. dev. (m) 0.63 2.19 2.03 3.58 3.67 4.39 4.74 5.42 5.48 6.83 7.40 7.64
East std. dev. (m) 0.00 0.00 0.00 2.35 2.43 2.44 3.01 3.18 3.19 3.52 3.85 4.01
Std. Dev. (kN) 167.37 602.91 555.28 125.40 412.72 359.72 185.95 397.28 456.42 323.93 574.88 660.10
Relative diff. / FullQTF 0.30 1.09 0.35 1.15 0.41 0.87 0.49 0.87
SY horizontal
tension
Vessel
position
Wave: 0, current: 30 Wave: 0
Wave relative heading at mean
position (deg)
Wave: 295, current: 90 Wave: 0, current: 60
0.0 42.1 28.5 17.0
Table 1 illustrates the differences between the drift loads calculation methods varying the relative heading between the
wave and the vessel. The first point that could be underlined is that for all simulations, F0 approximation underestimates the
loads on the mooring system. Indeed, this difference is always over 50% on the standard deviation of loads and is more
important in head sea (70%).
Moreover, the F1 approximation presents comparable results to the FullQTF ones. The maximum difference observed is
15%, and F1 could underestimate the loads as well as overestimating them.
Conclusions
This paper presents a comparison between different formulations for the calculation of slow drift loads for the
applications in shallow water. It has been shown that the F0 approximation, commonly used in practice, largely
underestimates the lowfrequency loads for shallow water depths (below 60m) or when the mooring system is stiffened.
The FullQTF method, in spite of giving more accurate results, is less efficient in computation time due to the fact that the
time reconstruction of the loads is made by double summations instead of simple summations as in the case of F0
approximation. The F1 approximation is a new method to evaluate the lowfrequency loads in a more accurate way than F0
approximation. The most interesting aspect of this approximation is that the time reconstruction can be done by simple
summations, thus it presents the same calculation time efficiency as the F0 formulation.
8 OTC OTC19450PP
Acknowledgments
The authors wish to thank SBM for providing mooring system information for this study.
Nomenclature
SY = Soft Yoke
QTF = Quadratic Transfer Function
LNG = Liquefied Natural Gas
References
[1] NEWMAN J.N. (1974) Second Order, slowlyvarying forces on vessels in irregular waves, Intl. Symp. Dyn. Marine
Vehicle & Struc. In Waves, Mech. Engng. Pub., London (UK).
[2] CHEN X.B & DUAN W.Y. (2007) Formulation of lowfrequency QTF by O() approximation, 22
nd
IWWWFB,
Plitvice (Croatia).
[3] CHEN X.B. & REZENDE F. (2008)  Computations of lowfrequency wave loading, 23
rd
IWWWFB, Jeju (Korea)
[4] CHEN X.B. (1994) Approximation on the Quadratic Transfer Function of LowFrequency Loads, BOSS
[5] NACIRI M. & QUEAU J.P. (2003)  The Soft Yoke Mooring and Offloading System for LNG Offloading
Applications, OMAE, Cancun (Mexico)
Tables
Table 1: Simulation results under different environmental conditions...........................................................................................7
Figures
Figure 1 LF spectral density WD 15m Tn ~251s .......................................................................................................................4
Figure 2  LF spectral density WD 15m Tn ~126s ........................................................................................................................4
Figure 3  LF spectral density WD 30m Tn ~251s ........................................................................................................................4
Figure 4  LF spectral density WD 30m Tn ~126s ........................................................................................................................4
Figure 5  LF spectral density WD 60m Tn ~251s ........................................................................................................................4
Figure 6  LF spectral density WD 60m Tn ~126s ........................................................................................................................4
Figure 7: Mooring system characteristic............................................................................................................................................5
Figure 8: Vessel position time series under F0 and FullQTF drift loads........................................................................................5
Figure 9: Vessel position time series under F1 and FullQTF drift loads.........................................................................................5
Figure 10: F0 and FullQTF time series standard deviation relative difference function of the stiffness rate .............................6
Figure 11: Calculation time comparison between FullQTF, F0 and F1............................................................................................6
Figure 12: Double summation time consumption in function of
max
.........................................................................................6
Figure 13: F(t) standard deviations error rate....................................................................................................................................6
Figure 14: SY for LNG terminals..........................................................................................................................................................7
Latest evolution of Metocean analysis
practices and their applications
Thomas Barberon Hydrodynamic and Mooring technical department
Bureau Veritas
Guillaume Gourdet Offshore Project Engineer Bureau Veritas
1. Abstract
Lately hydrodynamic analysis performed for moored floating units have known a little
revolution. From the typical Metocean data in directional scatter diagrams format, the industry is
moving to the regular use of hindcast data. These data are presented as a list of seastates,
including swell, wind sea or secondary swell, wind and current data. The combination of these
effects allows a proper estimation of heading analysis for a turret moored vessel. Thus wave
loads effects can be assessed more accurately, in particular the occurrence of beam and
quartering sea conditions. As results of heading analysis ones get the common wind and wave
directions with regards to the ships heading which may lead to important information regarding
safety and marine operations.
Firstly, the paper details the methodology for heading analysis and the typical required site data..
Secondly, it presents the effect of such analysis of the wave loads applied onto the vessel hull
and some applications in hull structural and fatigue calculations.
The paper follows the different steps of a typical project: review and analysis of Metocean data,
heading analysis, hydrodynamic and spectral assessment. Finally, typical structural classification
finite elements computations are described with ways to adapt them for site conditions.
2. Hindcast data analysis
The first step of every project is the environmental conditions analysis. The results of these
analysis will then be used by most of the different studies done during the project.
In design reports, Metocean data may be given in two forms. The first one is the statistical form
(e.g. intensity versus return period), that eliminates information of simultaneous occurrence of.
waves, wind, and current. The second one is the form of n years hindcast data base issued by
meteorological companies.
Time series of environmental parameters are provided for 5, 10, 25 years at 3hourly (most of the
time) intervals and include the following Metocean parameters:
 Parameters corresponding to the lowest frequency spectral peak (sometime referred to
as swell),
 Parameters corresponding to the highest frequency spectral peak (sometime referred
to as windsea),
 Wind speed and direction,
 Current speed and direction.
In reference [9] is presented one solution for the partitioning of waves into swells and windsea.
Before processing into hydro/mooring analysis, the assumptions made by the companies
provinding data should be analysed.
 Wave described by an unimodal or bimodal / bidirectional spectrum (each part
representing respectively swell and wind seas),
 Spectrum used,
 Wave, wind and current heading (coming from or going to) and characteristics.
As some seastates present some spurious data, these ones have to be removed from the hindcast
data.
Based on the cleaned database, plots are used to assess the validity of the data. One of the most
usefull plot is steepness curves (see Figure 1). This reveals the partitioning between waves and
allows the comparison with steepness curves (ratio Hs/Tz).
Figure 1: Hs vs. Tz (Swell front) and Hs vs. Tz (Wind Sea front)
Next, in order to assess the validity of the partitioning, a plot Tp_swell vs. Tp_windsea (see
Figure 2) may be investigate. Following example shows typical results where peak periods of the
swell are higher than peak periods of the wind sea.
Figure 2: Tp swell vs. Tp WindSea
After checking the partitioning of the waves, correlation between environmental parameters
should be looked for. An obvious correlation is the wind sea with respect to the wind. Shape of
the plot close to median line indicates a connection between elements. The graph shown on
Figure 3 could also be split according to the seasons (Figure 4 and Figure 5).
Figure 3: Hs Wind sea vs. Wind
Speed
Figure 4: Hs Wind sea vs. Wind
Speed  Winter
Figure 5: Hs Wind sea vs. Wind
Speed  Summer
In case partitioning is well done, a weak correlation between swell intensity and wind intensity is
obtained as shown in Figure 6.
Figure 6: Hs Swell vs. Wind Speed
Relation between a parameter (intensity, period ) versus the direction allows to validate the
data with respect to geographical parameters.
Figure 7: Hs Swell vs. Swell Direction Figure 8: Hs Swell vs. Swell Direction
Figure 7 for example indicates that point is located far from the coast with a prevailing direction
from swell (in a channel for example). Figure 8 shows a sector without swell. This could be close
to a coast. It should be underlined that points in the sector 220360 are probably spurious wind
sea seastates.
Directional plots between correlated parameters allow to determine the intensity of this
connection. The sharpest the plot is, the closest the parameters are (see Error! Reference source
not found.).
1000Y Scatter Diagram
0
2
4
6
8
10
12
14
16
0 2 4 6 8 10 12 14
Tz [s]
H
s
[
m
]
Steepness 1/7
Steepness 1/10
Steepness 1/15
Steepness 1/18
Steepness 1/20
1Y
10Y
100Y
Figure 9: Wind Sea vs. Relative
Wind Sea direction
Figure 10: 1000 Year graph
For some Metocean analysis, 1000 years (or 10000 years) data are available. Plotting the couples
(Hs,Tz) for all data allows to get information on the extrapolation of the extreme values. On
Error! Reference source not found., it is clear that extreme values have been computed only
based on steepness curves.
3. Heading Analysis
Heading Analysis are usually performed to obtain the heading of the vessel with respect to global
axis and then to extract the relative heading of the vessel with respect to the environmental
parameters.
Results of these calculations are used for long term and short term computations of
hydrodynamic parameters (wave bending moment, wave shear forces, relative wave elevations,
accelerations, motions ...). Actual link between heading analysis and hydrodynamic computation
is performed by the mean of Fata, BV in house software. The heading analysis itself is performed
by the help of Ariane 3Dynamic software, which allow to combine wind, current and waves
loads.
Description of the mooring pattern is not fundamental for the simulation of each seastate but
recommended. However, as heading is computed over 3hours duration (mean heading computed
for each seastate of the hindcast data base), real stiffness of the mooring pattern could avoid
some errors.
Mandatory required data are generally the wind coefficients, wind areas, current coefficients, the
characteristics of the vessel (draft, length, beam ), QTFs (coming from hydrodynamic
software, such as BVHydrostar), Added mass, Damping, Metocean parameters
Optional data could also be required such as mooring lines description (length, composition, ),
anchors positions, pretension in lines at fairlead, water depth, position of the fairlead
Time domain simulation
All environmental parameters (swell, wind sea, current and wind) are taken into account during
the simulations. Mean heading is computed for each seastate based on low frequency time
simulation over 3hours after a transient time depending of the studied vessel.
As mean drift force are of interest, effect of the seed for wave or wind spectrum is not to be
considered.
Heading Analysis Results
Attention must be given on the fact that software could give equilibrium position that is not the
most stable. In the case of turret moored vessel (majority of mooring pattern encountered in
heading analysis), some wrong static position can be found at the beginning of the simulation.
For example, static position with all environmental parameters in line coming from stern can be
found instead of parameters coming from bow.
In some cases, after a time simulation (variable), the vessel could turn around its mooring point
to reach the normal steady position.
In the worst case, the wrong initial static position can subsists during a long time simulation
(more than 3 hours). That is the reason why it is recommended not to use only the static position
given by mooring software to perform heading analysis.
Result of the total heading analysis is a time serie of heading of the vessel relative to global axis
system (generally North, depending of the model).
Then, this time serie  one result per seastate (time step depending on the database) is extracted
to be post treated relatively to waves heading.
After having performed the heading analysis, some graphs are interesting to be plotted.
Among these, heading of the vessel relative to environmental parameter is probably the most
important.
Figure 11: Vessel Heading Relative to North Figure 12: Vessel Heading Relative to Wind
Direction
Figure 11 could be compared to the site position. For previous example, the position of the vessel
on a west coast gives parameters to check results.
In Figure 12, it is important to precise how relative heading is computed. Such graph could
permit to give information such as position of the flare, position of equipment onboard.
Figure 13 and Figure 14 represent the heading of the vessel relative to swell and windsea
direction. Conclusion is that vessel is governed by windsea and not by swell (in this case).
Figure 13: Vessel heading vs. swell direction Figure 14: Vessel heading vs. wind sea direction
Another interesting graph to be plotted is the absolute heading of the vessel in polar coordinate.
All lines connecting two following time step results should be in the vicinity of the circle and not
crossing the circle (large change of direction probably due to wrong heading result or spurious
seastate).
Figure 15: Heading Analysis with sudden change (inversion)
Probably wrong
Figure 16: Heading Analysis with smooth variation
Probably correct
For some projects, onsite records are available and should be compared with heading analysis
results. Some examples are given hereafter:
Figure 17: Comparison Heading analysis vs.
Excursion monitoring system
Figure 18: Comparison Heading Analysis vs. Excursion
monitoring system
Previous comparison Figure 17 gives correct results. It should be underlined that some data are
missing around the 24 of July for the onsite data. Moreover, onsite data are given every minute
and heading analysis is performed every 3hour (due to database time step).
Example Figure 18 shows some spurious cases due to inverted heading of the vessel.
Attention should be care on the fact that vessel could have some temporary periods with heading
forced by tug (for maintenance for example), offloading sequences...
Analysis of parameters time series
The results of the heading analysis are often used to compute motion time series. Motion
response can be computed for each seastate. Then, motion time serie over 10 (or 20, 25) years
can be analysed.
Figure 19: Roll response over several years Figure 20: Roll response over 12 days
Figure 19 shows the roll response of the vessel over several years. Analysis of the maxima can
be correlated with the season (depending of the hemisphere).
Time serie over lower time duration (1 week) give information on the continuity of the results. If
points are not in accordance with previous and following, results of the heading analysis is
probably wrong (see Figure 20).
Disambiguation of spurious points
When plotting Metocean parameters in function of relative heading, some spurious points can
occur (for example, points surrounded in red in Figure 21).
Figure 21: Wind Velocity vs. Vessel Heading
These points can have a statistical significance and correspond to exceptional events in the
distribution of the values of the Metocean parameters. They can correspond to spurious or odd
events in the hindcast data. First, the Metocean conditions must be recomputed to determine the
validity of the odd points. Afterward, one way to determine whether these points have a
statistical significance is to examine the distribution of the value of the analyzed parameter per
heading.
4. Hydrodynamic Analysis
In this paper, the hydrodynamic computation itself will not be detailed. By hydrodynamic
computations we mean the way to assess hull specific behaviour (motions, accelerations &
waveloads) of the hull using radiationdiffraction methods. The outputs of such analysis are
RAOs (Response Amplitude Operator). These RAOs are then analysed through spectral analysis
in order to get the site and hull specific waveloads. Additional information can be found in [11].
5. Spectral analysis
The structural analysis needs as input the response of the ship at a certain Return Period (RP). In
principle, the response should be obtained by cumulative Long Term (LT) statistics requiring
extrapolation of Metocean data. In lieu of such analysis, Short Term analysis is carried out using
the short term extremes seastates in each direction.
Short Term & Long term calculation
In the philosophy of a short term calculation, a single sea state is considered representative of the
Metocean condition to be encountered, and the response of the unit of this seastate is considered
representative of the extreme response. In a long term calculation, each sea state is taking into
account in the unit response.
This type of study in necessary for fatigue study and is generally considered more accurate for
extreme response.
The LT response is then obtained by cumulative long term statistics with BV software, FATA.
The summation over all spectrums is made. Given a seastate and assuming a Rayleigh
distribution of maximum, the probability associated to a response level S is equal to:
( )
( )
=
2
,
2 exp ,
P S S
P S
T H R
S
T H S p
And then the number of cycle associated to a response level S is:
( ) ( )
( )
P T
ref
P S P S
T R
T
T H S p T H S N , , =
The number of cycle, over all the seastates, associated to a given response level S is equal to:
( ) ( ) ( )
( )
=
j i P T
ref
P S P S
j Z
j i j i
T R
T
T H p T H S p S N
,
, ,
Where:

T
the average duration of a seastate (3h)
 ) , (
P S
T H S p the probability of a given stress level S within a seastate characterize
by the wave significant height Hs and the peak period Tp
 ( )
j i
P S
T H p , the probability to encounter the seastate Hsi, Tpj
 ( )
j E
P T
T R the mean period of the response for a seastate with a peak period of Tpj
6. Bureau Veritas FPSO Structural methodology
Once the final waveloads computed we can start structural loading. This first chapter gives an
overview of structural computation in BV. The next chapter will make the link with
hydrodynamic loads.
Bureau Veritas approach is based on the analysis of the stress level acting in the structure (for
yielding, buckling and fatigue strengths) and on its verification according to permissible stresses
ratios. The evaluation is carried out including:
3D FEM coarse mesh models to evaluate stresses and buckling behaviour.
3D fine meshes analysis to complete the necessary structural analysis of primary structural
members from 3D coarse mesh.
Fatigue analysis for connections details
It should be noted that definition of coarse mesh is changing now in Bureau Veritas Rules in
order to harmonize offshore Rules with Common Structural Rules for bulk carriers and double
hull tankers (IACS).
Typical FPSO loading patterns from loading sequences are used for fatigue analysis. While
yielding and buckling analysis can be performed for special loading conditions (inspection,
repair LC)
For all FPSO calculations, site conditions are applied.
At the end, the following criteria will be considered to identify elements which do not comply
with the Rules criteria for intended operation as FPSO:
Elements with stress ratio above 1.00 do not comply with the applied criteria.
Elements with stress ratio between 0.975 and 1.00 are considered as suspects areas.
Elements with ratio less than 0.975 comply with applied criteria.
Calculations are performed at both tanker and FPSO stages. Tanker phase gives an overview of
the structural integrity of the original vessel against the newest set of Rules. The conclusions of
such study can lead the predrydock inspections directly to the most critical areas. This study
may also help to determine the level of criticality of elements found critical in FPSO or tanker
phase.
As part of tanker phase, an accumulated damage is also calculated, based on the available trading
history of the ship.
On the other hand, the FPSO assessment enables to define where renewals or reinforcements will
be needed either because of ageing of the structure (corrosion, accumulated damage) either
because of its new use of FPSO (loading/unloading sequence, site conditions)
This VeriSTAR approach is generally preceded of a 2D analysis, performed in Mars BV
software. Mars enables to check longitudinal elements and transverse bulkhead against BV Rules
for both hull girder and local scantlings.
Figure 22: 2D Sections
The 3D models can be either a set of partial models (called 3holds models) to assess separately
each tank either a Complete Ship Model (CSM).
3holds models are used at first stage of project, to quickly and efficiently assess the strength of
the cargo region. They are generally linked to fix load cases defined in the Classification
Societies Rules.
However offshore engineers are used to adapt these load cases to offshore projects and their
particular wave behaviour (like spread or turret mooring, specific site conditions, etc)
Finally it is now a common practise to model the whole unit, from transom to bow. These
models generally tolerate a higher flexibility for the loads to use. If this flexibility allows a better
representation of the actual wave loads, it is also synonymous of a lot more of loading conditions
to run, increase the time of the computations and posttreatment.
Figure 23: 3holds coarse mesh model
(including deck fine mesh)
Figure 24: Complete Ship coarse mesh model
7. Hydro to Structure
Once hydrodynamics waveloads computed with the state of the art methods, hydrodynamic
engineers and structural engineers need to decide how these waveloads should be used in the
structural assessment.
Several types of analysis are performed using direct waveloads computations:
Hull structure analysis
Topsides structure analysis
These two cases include both extreme conditions for yielding and buckling check and fatigue
conditions
Sloshing loads on hull structure
Green waters effects
Slamming
These 3 analyses may also require hydroelastic methods. This means that stiffness of the
structure should be taken into account in the hydrodynamic analysis. These methods if already
applied for several ship and offshore designs are still quite exceptional studies.
Mooring and offset analysis
Motions analysis
Etc
Wave loads as described in ship Rules
On the contrary of traditional offshore engineering, naval architects are not used to apply directly
computed wave loads on hull structure. The common way to design a hull is to apply wave loads
from Rules (Classification Rules, mainly based on IACS unified requirements with regards to
wave loads).
These loads are issued from North Atlantic waves conditions, recognised as the worst area for a
ship to sail. A hull design with these loads should then be able to sail all around the world
without any restrictions.
It should however be noted that a few exceptions exist, mainly for shuttle tankers operating
continuously in harsh conditions like in North Sea, or in North Pacific.
For ships sailing in mild environment, classification societies also developed navigation
notations, which limit the areas where the ship is able to sail (tropical waters, sheltered waters,
etc). These navigation notations are simple coefficient applying to the IACS North Atlantic
waveloads.
The Rules always define a set of load cases to be applied on the hull structure. For example
Bureau Veritas Rules defined 4 basic load cases which combine the following effects of the
waves:
 Vertical bending moment
 Horizontal bending moment
 Vertical shear forces
 Torsion
 Accelerations
 Elevation of the waves as view per the hull
The difficulty of the combination is to know which loads are correlated with the other. For
example, it is quite obvious that maximum vertical accelerations will not happen at the same
time than maximum horizontal acceleration. While vertical bending moment is maximum at the
same time than the crest wave pressure on the side shell.
Wave loads in load case A
crest
& A
trough
Wave loads in load case B
Wave loads in load case C
Wave loads in load case D
Figure 25: BV Combinations of waveloads.
Extracted from the Offshore Rules, Part D, Chapter 1, Section 5
It is also a challenge to define the values of secondary wave loads when trying to maximise one
of them. For example to evaluate the wave vertical bending moment to be applied, when
maximising wave horizontal bending moment. Rules defines default coefficient for these
combinations.
Application of hydrodynamic wave loads on structural models
As for all engineering studies, it is possible to define different methods, depending of course of
what it is the purpose of the study or of the schedule of the project. But it may also depend of the
quality of available data.
BV keeping Rules load cases.
The simplest methods keep the Rules load cases for basis. These methods are quick to put in
place, using Rules Finite Element Software like VeriSTAR Hull.
They do not require complex interactions between structural and hydrodynamic analysis.
At first stage of FPSO projects, it is quite common to perform structural analysis by just applying
a single coefficient to Rules wave loads. This can even be done without hydrodynamic
computations, based on the experience of offshore engineers for wellknown areas (like West
Africa, Brazil).
These methods should be used with care and only to preproject analysis, before any better
hydrodynamic results are available.
When hydrodynamic results are known, it is then possible to tune the Rules load cases wave
loads by wave loads.
For example in the load case maximising the transversal accelerations, the site transversal
accelerations is combined with the site vertical accelerations multiply with the Rules coefficient
for this load case.
This method is known as conservative for extreme computations. However it presents limitations
mainly in 2 cases:
Fatigue issues, because of the difficulties to combine load cases results on a long term
approach.
Exceptional waves, like cyclones when Rules combination of waves loads may not be
realistic.
Direct transfer of wave loads on the structural model is then the next possible step.
Direct transfer
The most common direct transfer is used for Spectral fatigue analysis. In this method, basic data
from radiationdiffraction analysis are applied directly to the structural model: accelerations and
pressure of waves.
At the end of the structural we get stresses RAOs, exactly like RAOs define in the first part of
this paper.
It is then possible to follow the spectral computation defines in the section 5 to get probability of
each stress cycle. These stress cycles and their probabilities are then summed using Miner
Palmgren formula to get final damage of the details and so fatigue life.
The Spectral Fatigue Analysis takes into account the exact site conditions and the relative
heading of the vessel. It is known as the most uptodate fatigue methods for ship details.
Apart from the huge numbers of calculations to define stresses RAOs, this methodology is
limited by the linear assumptions done in spectral domain. Even if corrective methods exist, it is
still difficult to interpret results on side shell area, where the wave pressure discontinuity has not
a negligible effect on stress repartition.
Figure 26: Example of stress RAOs for representative heading. Each node required
hydrodynamic computation, transfer to structural model and run of FEM.
It is also possible to perform time domain analysis. This is mainly used for extreme
computations, or with specific waves.
In this case the first is to define the wave, defined by its height, length and phase that should be
computed. Then hydrodynamic software can issue wave pressure all along the hull and hull
girder accelerations to be applied on the structural FEM.
0
5
10
15
20
25
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Omega (rad/s)
S
t
r
e
s
s
R
a
n
g
e
(
M
P
a
/
m
)
180
225
270
315
360
This methodology avoids spectral assumptions which are not always easy to apply on a structural
model. The main point of this methodology is to select the wave to be run. It is always easy to
use RAOs to select the most critical wave for one effect. However it is much more difficult to be
sure that this wave will be the most critical once combining all the loads it generates.
Limit of transfer
The main point of transferring hydrodynamic results to structural is the difference of
fundamental assumptions in both methodologies. They are particularly critical for
Viscosity, which is difficult to take into account in structural FEM models
Linearity again wave height, which basic assumptions in hydrodynamics, but quite
difficult to apply on structural assessment of side shell area.
Viscosity
While hydrodynamic analysis take into account the viscosity of the liquid, viscosity is not easily
added to structural finite element models (FEM). If the viscosity can be omitted for longitudinal
waveloads, it is an important input for beam sea cases, where roll motion and acceleration are the
some of the main parameters for the computation of
Transversal acceleration,
Wave pressure on side shell in beam sea cases,
Horizontal bending moment.
Viscosity and in particular roll damping calibration is already a challenge for hydrodynamic
analysis. It becomes even more complicated when transferring to structural finite element model,
which in naval architecture does generally not include viscosity capabilities.
For a structural computation it is reasonable assumptions to neglect the viscosity (except in way
of particular appendix such as bilge keel). However in case of direct transfer of wave loads from
hydrodynamic to structural FEM the loads will not be equilibrated, which is not acceptable.
Several methods have been developed to solve this problem. Bureau Veritas spectral fatigue
methodology is applying equivalent forces on the bilge keel to model viscosity effect.
Linearity with regard to wave height
Another point is the linear assumptions of hydrodynamic computation versus the wave height.
Most of the hydrodynamic calculations are not time simulation but spectral simulations (for
computation time reasons). They consider the waveloads as a linear function of the wave height.
Applying such assumptions for a structural model would lead to wrong results in way of side
shell, where the wave pressure is not proportional to the wave height. The following picture
shows the two types of waves.
However this assumption is relatively correct for details far from the side shell.
Linear wave pressures as used in spectral
hydrodynamic analysis
Non linear pressure as needed for time domain
structural analysis
Several methods exist to compensate this assumption. In particular Bureau Veritas recommends
to compute the effect of the missing pressures on a separate model. The stress correction
obtained is then applied to the stress RAOs.
8. Conclusions
To conclude this paper the authors would like to highlight the increase and generalization of high
level technical computation for offshore projects, including fast track conversion projects. This
increase in the number of studies has transferred existing tools and methodologies from research
departments to engineering common practice. It has also highlighted points where the industry
need to improve its standard like for Metocean data, whose quality is not always at the level of
the expected following studies , or areas where research and returns of experience are still to be
gathered.
Nowadays tools and methodologies are available to analyze site data and compute accurate
waveloads for a specific hull design. And these waveloads are now used at different levels in
structural extremes and fatigue computations.
9. Acknowledgement
The authors wish to thank Bureau Veritas for permission to publish this paper. The views
expressed are those of the authors and do not necessarily reflect those of Bureau Veritas.
10. Reference
[1] C. Bran, R. Veyer Feedback on structural issues from FPSOs purpose built or converted
from oil tankers OMC2007
[2] Bureau Veritas Rules for classification of steel ships 2007
[3] C. Chauviere, J; Esteve Conversion of offshore floating facilities  How to tackle them with
an asset integrity focus OSEA 2008
[4] Bureau Veritas Rules for offshore units 2007
[5] P. Biasotto, V. Bonniol, P. Cambos Selection of trading tankers for FPSO conversion
projects OTC2005
[6] G. Gourdet Connection hulltopsides: principles, design and return of experience
DOT2008
[7] IACS Common structural rules for double oil tanker 2008
[8] Bureau Veritas NI 493 R01 E, July 2008
[9] R.Lawford, J.Brandon, T.Barberon, C.Camps, R.Jameson Directional Wave Partitioning and
its Application to the Structural Analysis of an FPSO, OMAE200857041
[10] M.Franois, C.Morandini, P.Gorf, M.Lewin Relative Wave Heading of an FPSO in Harsh
Environment OMAEFPSO040043
[11] HYDROSTAR FOR EXPERT Reference guide and Tutorial for Naval and Offshore
Hydrodynamic Application (Summary); XB C, Research Department Bureau Veritas.
[12] XiaoBo Chen Hydrodynamics in offshore and naval application, Part I, Keynotes lecture
at the 6
th
Intl Conf Hydrodynamics, Perth, Australia, 2004.
Connection hulltopsides:
principles, designs and returns of experience
Guillaume Gourdet  Bureau Veritas  Naval and Offshore Engineer
I . Abstract
In the 7 last years several findings have highlighted the importance of the system connecting the
topsides structure to the hull. This area at the border between hull design and topsides design is of
uttermost importance for the production integrity. Based on this observation, this paper will
present different existing designs of connection of the topsides to the hull, from stiff stools, to
sliding support. It will then focus on the way to assess such connections in order to insure the
integrity of the connection but also of the hull structures below the deck and of the topsides first
level structures. Interaction between the hull and the topsides will be developed. The paper
finishes on the challenges regarding inspection, maintenance and repairs of these connections. All
along the paper, feedbacks from existing units will be used to illustrate the speech, for converted
and new built units in mild and severe environment. There are approximately 170 Floating Unit
systems in operation or available worldwide. If about 45% of them are FSO with none or limited
topsides, about 55% are FPSO concerned by the question of the connection between their hulls
and their more or less complex topsides.
I I . I ntroduction
There are approximately 170 Floating Unit systems in operation or available worldwide. About
55% are FPSO. FPSO combines the refinery technologies for the topsides and the naval
architecture for the hull. One of the challenges in a FPSO project is to design, assess and validate
the structure between the topsides structure and the hull structure. The hull is under continuous
deformation due to the passing waves and these deformations should be absorbed by topsides
structures.
Since the first FPSO in the world, more than 20 years of experience have been gathered. Several
designs have been proved inefficient, expensive to build or on the contrary satisfactory choice.
And if the first FPSO were equipped with light topsides, the topsides are still becoming heavier
and heavier, and not limited to mild environment. Following the technology needs, the structural
assessment of FPSO is increasing in complexity, accuracy and quickness. Nowadays integrated
models, including hull and topsides, are a standard tool of validation of FPSO design for yielding
and spectral fatigue analysis.
This paper describes the different steps to assess the interface between topsides and hull
structure; starting by identifying the different loads to be taken into account, continuing on a
review of design of these interfaces and finishes the presentation of the different ways to assess
them.
I I I . Topsides loads and hull accelerations
Initially we can considered the hull as a rigid structure, to identified the first set of load than
apply to the interface hulltopsides. The effects of the elasticity of the hull will be described in
section IV.
I I I .1/ Topsides loads
The first aim of the interface hulltopsides is to transmit the topsides weights to the hull structure!
If topsides weights are generally well known when the project enter in the structural design
phase, their distribution on the topsides structure is known only after a first design analysis.
Moreover the loads distribution between the different support between the hull and the topsides
also depend of the design engineering of the topsides structure. To correctly assess these loads it
is needed to know:
the topsides main structure
the weight and centre of gravity of main equipements
the weight, centre of gravity and inertia of the rest of the topsides (both structure and
smaller equipments)
Depending of the stage of the project, the accuracy of the estimation of these data can lead to
different assumptions regarding the interface assessment. From a basic vertical weight to be
applied on each topsides supports to a fully coupled model. This will be detailed in section VI.
Of course the topsides loads to be transmitted to the hull are also depending of the hull onsite
motions and accelerations. These results are coming from the hydrodynamic analysis.
I I I .2/ Hull motions and accelerations from hydrodynamic analysis
Hull motions and accelerations are the first results of hydrodynamic computations. It takes into
account the unit shape, weight and inertia to get Response Amplitude Operators (RAOs) for each
of the motions and accelerations. These RAOs are specific to the unit and to the loading
conditions analysed.
Short terms accelerations can then be easily assessed for specific wave types.
However, it is recommended to perform long term calculations based on extended site condition
data. Such site data are now quite common and can be obtained either by onsite measurements
over several years, either by hindcast data.
In particular in case of units equipped with a weathervaning system (such as a turret) correlated
data of waves, wind and current are necessary to estimate the real heading of vessel with regards
to the waves. These heading calculations allow to determine the wave directions as seen by the
unit, depending of the wave heights and periods, but also of the wind and current speed and
direction. This is particularly relevant to assess the quartering and beam sea cases and estimate
how much transversal accelerations (and others transversal effects) will happen on the
weathervaning unit.
Once this heading analysis performed, long term analysis for defined return periods can be done
based on spectral computations.
These accelerations, computed at the centre of gravity of the unit can then be translated to any
point of the unit, considered as a rigid body.
Hydrodynamic analysis taking into account the real stiffness of the hull body, are now being
developed in research department, papers are mentioned in the references.
I V. Hull deformations
As said previously, the main reason of specificity of the interface hulltopsides is that the hull is
far from a rigid structure. The hull, much more than moving to follow the incoming waves, is
also continuously deformed under the wave loads and the internal pressures.
I V.1/ Hull girder deformations
The hull girder deformations can be divided in three main deformations:
Vertical bending moment
Horizontal bending moment
torsion
Those 3 deformations are always a combination of both mass distribution in the hull girder and
wave effect on the hull. Their RAOs are the next results of hydrodynamic analysis, after motions
and accelerations RAOs. Short and long term vertical and horizontal bending moment and torsion
are then computed with similar spectral computation then described for accelerations.
In ship engineering the terms static or still water analysis are used when assessing the self
weight of the unit versus the buoyancy. The buoyancy is defined by the hull shape, draft and trim.
The selfweigh is a combination of lightship distribution (steel, piping, equipments, topsides
weights) and compartment fillings. The hull girder effects due to still water loads are known as
accurately as the selfweight distribution is known using stability software, as the onboard
loadmaster (at least for vertical bending moment). At early design stage a maximum still water
vertical bending is defined called permissible. All loading patterns used during the life of the unit
should then stay below this value.
Torsion is traditionally not computed for oil tankers and considered as negligible. However
recent FSO studies lead to such extreme inspection/repair loading pattern, that torsion was no
longer negligible
The wave types of effect on the hull girder depend of:
Waves and ship relative heading
wave length
The amplitude of the effect induced is then linked also to the wave height, the hull shape and
inertia.
The 2 following sections will illustrate each effect separately. But it should be kept in mind, that
the hull always sees a combination of them.
I V.1.1/ Vertical Bending Moment
Static vertical bending moment is generally maximized homogenous loading pattern.
Wave vertical bending moment is maximized in head sea seastates, for waves about as long as
the length of the ship.
FPSO model in trough of waves
This effect is the one that creates the greater hull girder deflections. And due to the stiffness of
the hull compared to the topsides structure, the topsides are working under forced displacement.
Naval engineers say that elements longer than the tenth of the unit length are fully contributed to
the hull girder vertical bending moment. Even if class societies have generally in their Rules a
more refined way to estimate this participation, finite element models are often the only way to
determine exactly this participation.
For the interface hull topsides, it means that special care is to be paid to stiff modules, if they are
close or longer than the tenth of the hull length.
I V.1.2/ Torsion
Static torsion is generally negligible. When not, it is due to specific loading patterns, such as the
one on the following picture. This type of loading pattern is not common in shipping industry and
so, should be analyzed with the greatest care during FPSO project.
Wave torsion is maximized in quartering sea seastates. It is strongly linked to the hull shape. It
should be noted than to assess directly the wave torsion, hydrodynamic assessment should know
the centre of torsion.
Deformation under both torsion and bending moment
(pictures of container vessel is used because torsion is easier to show on opendeck vessel)
Torsion is said negligible for ship with a complete deck. (On the contrary torsion is critical for
ships like container vessel with large deck openings). However FPSO loading pattern are
generally more severe than on a ship and it has been shown on several cases than torsion should
not be disregarded.
I V.1.3/ Horizontal bending moment
In ship design, horizontal bending is the moment around the vertical axis. Static horizontal
bending moment is null, since there is no transversal loads in stillwater conditions. This remark is
also applicable in case of pure head sea condition.
On the contrary, when the waves come transversally (beam or quartering seas) the gradient of
pressure between each side of fore and aft parts of the hull can generate a horizontal bending
moment.
View frombelow of a converted unit
beamsea wave loads inducing horizontal bending moment
Horizontal bending moment should be assessed particularly when wind sea and swell directions
are not directly associated, like for Campos Basin, offshore Brazil. In such case the hull is most
of the time seeing both head and side seas.
I V.2/ Effects leading to deck stresses concentration
Effects described in the previous sections were hull girder effects, affecting the whole ship. In
this section we will detail the effects than can locally modify the deck deformations, being
generated either globally on the hull or locally.
I V.2.1/ Effects of alternate filling
Cargo tank fillings have a direct effect on the hull deformations and so, on the deck structure. In
case of alternate filling the deck and the structures below will be sensible to the global
deformations of transverse and longitudinal bulkheads.
I V.2.2/ Poisson effect
When put in tension a steel piece will of course lengthen, but it will also transversally shorten.
The same thing happens for the deck. When the ship is in hogging, the deck is in tension. So it
will also have the tendency to retract itself transversally. This effect can create nonnegligible
transverse effects in pure head sea conditions. Combined with the alternate filling between wing
and center tanks, this is one of the critical effects for the design of deck transverse web frames in
ships.
Typical section of single hull FPSO, under hogging moment and crest of waves
deck transverse webs are under transversal compression
I V.2.3/ Steel dilatation
Dilatation due the variation of temperature between the outside air, the fluid (air, sea water, oil)
below the deck and the temperature of the structure will also generate additional deformations.
This is particularly the case if the oil is warmed, or in tropical area (where the sun is harsh, but
the nights can be cold) or close to the flare tower
If the dilation is quite easy to estimate at first stage, a refine assessment is much more
complicated due to the number of heating sources and the complexity of screen effect of topsides
elements.
I V.2.4/ Effect of attached structures
Last but not least, the deck deformations are locally depending of the close by structures. The
hull girder effects are generally considered as a forced displacement, due the high stiffness of the
hull. However, it should not be forgotten than the local deformations can be significantly
influenced by the local structure such as:
longitudinal or transverse bulkhead;
Under deck secondary structure;
Under deck reinforcements for the hull equipment;
And, of course, the topsides stools
The stools arrangement and design depend of the several factors, it will also influence the type of
assessment to be performed.
Typical longitudinal and transverse deformations of ship with superstructures under bending
Discussion regarding the effects of superstructure on neutral fiber of hull section and hull girder
stresses can be found in reference [2]
V. Main interfaces hull topsides:
V.1/ Topsides supports philosophy
Topsides supporting stools arrangements and designs are almost as numerous as FPSOs. They
depend of the topsides equipments needed, of the weather severity of the site and of the hull.
Without trying to be exhaustive here are some criteria to class and assess the stools arrangements.
Number of stools
A high number of stools has the advantage to limit the loads by stools and so limiting the stresses
in the stool and the need for underdeck structure reinforcements. On the other hand that means
that the hull deformations will be much more transmit to the topsides pallet, which will have to
sustain it. In harsh environment it can appear to be difficult to design, particularly if the topsides
pallet are long compared to the unit length (around and more than one tenth of Lpp).
This picture highlights the effect of long modules combined with stiff stools: the fore and aft
stools are taking all the deformations
Length of the topside modules
The number of stool has very few effects on the stress concentration in way of first and last stool
at deck level. These stress concentrations are much affected by the length of the modules
compared with the unit length, as already mentioned in this paper.
Stool stiffness
The stool stiffness will generate the stress concentration at the connection with the deck. The
more stiff the design is the more their will be stresses between the hull and the topsides. Several
level of stiffness can be found from the complete stiff stools, which are generally large stools,
with heavy brackets in longitudinal and transversal direction in order to smooth the stress
concentration factor. These brackets should be backed by underdeck deck structure. This type of
stool allows to transmit heavy loads. Then there are some stools stiff in only one direction,
generally done with transverse or longitudinal bracket on the deck. These stool are less efficient
in case of heavy topsides. Stools stiffness can also be determined with the help of sliding joints.
In this case no stiffness will be seen by the connection in the sliding direction.
Example of four stiff stools module topsides.
Four stools, isostatic design
This arrangement aims to be quasiisostatic. If one stool is generally stiff in every direction the
three others one are sliding in one or two directions in order to avoid too much hyperstatic. These
sliding arrangements allow the deck of the hull to move and be slightly deformed without
transferring too much displacement at the topsides level. Separate like this hull and topsides is
quite satisfying from structural point of view, however the sliding joints and their aging are still a
major engineering piece of work for huge topsides. Nevertheless some of the heaviest topsides in
the world are supported using such stools design.
It should be noted that a critical design load case for sliding joint is a damaged load case with a
lot of heel. Several degrees of heel can happen very quickly on a floating unit, as a recent
example remind us. If these cases are generally not critical for the hull (damaged stability should
be checked by Class Society) the topsides supports should also be designed for such extreme load
cases.
On the left, view of stools of a FPSO, under sagging conditions.
On the right, a view of deformed 4 stools isostatic support.
No real stools, but an integrated deck.
For mild environment, it can also be found a design of integrated deck, which allows a large
flexibility in term of topsides module integration on a deck linked to the main deck of the hull by
a series of transversal supports. These transversal supports let the hull move around its
transversal axis without transferring too much displacement on the integrated deck.
Double plate
In some designs, double plates are welded on the deck. Then the topsides stools are welded on
this double plate. If from structural point of view, double plate could be easily assess, in
particular in way of the weld which should transfert most of the loads, from integrity point of
view, this design is not recommended. In fact it leads to risk of undetectable problem such as root
cracking or corrosion between double plate and deck. This 2 damage could both lead to an
unexpected breach on deck and the linked risk of explosion.
Whatever the stools arrangement their integration within the hull structure is a critical point.
V.2/ I ntegration with hull
Looking at general arrangement of a ship, several points appear to be more adapted to support
high weight, in particular the connection between transverse web frames and longitudinal
bulkheads. Due to the hull girder loads this area is generally strong enough to support heavy
loads. On the contrary this are will support high longitudinal stresses. The stools welded above
such longitudinal primary members should be designed to support theses stresses.
However it is not possible to put all the stools on such areas, and most of them are put above
typical transverse web frames. Of course the deck transverse webs should be assessed for
yielding and buckling due to the additional loads but also local calculations should be performed
for stress concentration areas.
Generally thick insert plates are put in the deck below the stools. However, great care should be
taken to the underdeck structure. This area is quite difficult to inspect and even more difficult to
repair on site due to the necessity of scaffolding. The connection between deck longitudinal
stiffeners and transverse web can be critical in fatigue due to bending. In way of topsides stools
stresses longitudinal stresses can be increased at longitudinal bracket toes increasing this risk of
low fatigue life.
Cutout on deck transverse web around deck longitudinal stiffeners are very common in shipping
industry. The deck is not loaded and such cutouts are not critical. But in way of topsides stool
the shear can be significantly increased and the cutouts may become critical in shear and fatigue
due to shear. In such cases they will need to be closed by collar plates.
It should also be noted than the structure below the deck is generally found very corroded on
older F(P)SOs. This gaseous corrosion has already leaded to onsite steel renewal. It is strongly
recommended to paint this area.
Before detailing the different ways to assess structural integrity of this integration of the stool
topsides within the hull structure in section VII; the next section will introduce typical hull
structural analysis for FPSO.
VI . Hull structural analysis
Bureau Veritas approach is based on the analysis of the stress level acting in the structure (for
yielding, buckling and fatigue strengths) and on its verification according to permissible stresses
ratios. The evaluation is carried out including:
3D coarse mesh models to evaluate stresses and buckling behaviour.
3D fine meshes analysis to complete the necessary structural analysis of primary
structural members from 3D coarse mesh.
Fatigue analysis for connections details
This VeriSTAR approach is generally preceded of a 2D analysis, performed in Mars. Mars
enables to check longitudinal elements and transverse bulkhead against BV Rules for both hull
girder and local scantlings.
Typical FPSO loading patterns from loading sequences are used for fatigue analysis. While
yielding and buckling analysis can be performed for special loading conditions (inspection, repair
LC)
For all FPSO calculations, site conditions are applied.
At the end, the following criteria will be considered to identify elements which do not comply
with the Rules criteria for intended operation as FPSO:
Elements with stress ratio above 1.00 do not comply with the applied criteria.
Elements with stress ratio between 0.975 and 1.00 are considered as suspects areas.
Elements with ratio less than 0.975 comply with applied criteria.
Calculations are performed at both tanker and FPSO stages. Tanker phase gives an overview of
the structural integrity of the original vessel against the newest set of Rules. The conclusions of
such study can lead the predrydock inspections directly to the most critical areas. This study
may also help to determine the level of criticality of elements found critical in FPSO or tanker
phase.
As part of tanker phase, an accumulated damage is also calculated, based on the available trading
history of the ship.
On the other hand, the FPSO assessment enables to define where renewals or reinforcements will
be needed either because of ageing of the structure (corrosion, accumulated damage) either
because of its new use of FPSO (loading/unloading sequence, site conditions)
VI I . Structural Assessment of the interface
The following sections will detail the different ways to assess the interface hullstructure
regarding yielding, buckling and fatigue. They are classed from the most basic to the most
complex one. Each of them can be relevant depending of the status of the project (particularly of
the knowledge of the topsides main structure and weight distribution), but depending of what is
the question to answer, from the easiest: can the hull structure support my topsides loads? to
the most difficult: is my design able to stay 20 years on site?
For each of the different assessment types, we will try to highlight the advantages and
disadvantages. Of course this list is not exhaustive; project needs and specificities on one hand
and engineer imagination and subtlety on the other hand could lead to slightly different
calculations. But the main methods are summarized here:
VI I .1/ Hull partial models and topsides loads
The hull structure is analyzed trough 3holds FEM models.
Topsides loads are distributed on deck structure by mean of
forces. The lower of the topsides can be modeled or not. For
such analysis the topsides loads are generally obtained through
basic assumptions.
This first level of analysis is particularly adapted for a quick estimation of the global yielding and
buckling capacities of the primary members under the deck:
Upper strakes of longitudinal bulkheads
Upper plates of transversal bulkheads
Deck transverse web and brackets.
For a local assessment of the connection between the stool and the deck structure, the lower part
of the stool can be modeled, and the concentrated loads from topsides analysis input on top of the
modeled stools. With such models, the local design of topsides stools support can be checked,
regarding yielding and buckling.
However these models do not take into account the topsides stiffness between stools. Most of the
time the topsides loads are also maximum topsides loads, combining gravity and hull
accelerations) which are not relevant for fatigue analysis.
Fatigue analysis based on such model will only assess the fatigue due to the hull
dynamic effects.
VI I .2/ Hull partial models and main structure of topsides
The hull structure is still analyzed trough 3holds
model. But this time the topsides are modeled up to the
first main structural level. Topsides weights and inertia
are them distributed either by mean of concentrated
loads on the model, either by an additional weight at
the centre of gravity of the topsides modules.
Based on such models, the effect of hull deformations
on the interaction hulltopsides can be analyzed, and
relevant yielding and fatigue assessment can be done.
VI I .3/ Complete models and main structure of topsides
The next level of analysis corresponds to a similar hulltopside interface modelling, but extended
on the whole ship length.
These models increase the accuracy of the hull
girder deformations, in particular in the fore and aft
part of the cargo area, where the hull scantlings and
the topsides design show discontinuities. These
areas are generally more difficult to analyze using
3holds models.
These models can also better assess the secondary hull girder effects such as horizontal bending
moment and torsion, described previously.
Nevertheless, these models are longer to build than a 3holds model, due the complexity of the
hull design in the fore and aft part.
VI I .4/ I ntegrated models
The higher level of analysis is a coupled model,
including the whole ship model, the complete
topsides main structures and weight distribution.
The main advantage of these models is that the
topsides weight distribution and inertia is directly
including in the model. The load distribution on topsides stools is then directly calculated by the
software, using an accurate modeling of hull and topsides and the wave induced effects.
These models allow to assess the effect of the interaction hull topsides even in the higher
structural nodes.
Whatever the level used the global weight of topsides and their longitudinal weight distribution is
taken into account for hull girder still water load. For 3holds, the results from stability or load
master calculations are input in the model. For complete ship model, the hull girder loads are re
calculated for each load cases.
Wave loads computed trough hydrodynamic studies, will always take into account the topsides
and superstructures mass, centre of gravity and inertia to determine on site accelerations. Wave
bending moment, wave shear and relative wave elevation computation will also use the total
weight distribution.
VI I I . Conclusions
To conclude this paper, which was focused on design, here are a few remarks regarding the
inspection/maintenance of the interface hull/tospides. Without carrying out a full risk analysis, it
appears that the interface hull/topsides is a critical area regarding both stress concentrations and
possible consequences. In case of damage and due to the proximity of the deck, the risks of gas
leaking (and so explosion) or hull weakening are quite high. Generally repairs will require hot
works.
On the other hand, in most cases stresses at the deck and topsides level can be reduced by adapted
loading and unloading sequence. It should also be noted that if the part above the deck is easy to
inspect regularly, others part such as underdeck structure are difficult to access, particularly if no
passageways is put in place, but not less critical.
Finally, behind the complexity of FPSO the connection between the hull and topsides structure, at
the border between the hull and the topsides, the marine engineers and the offshore engineers, the
shipyard and the oil industry, should not be neglected. Using the uptodate naval and offshore
tools, it is now possible to build effective coupled models, integrating both the complete hull
structural mode and the complete topsides structure. These models can then be used to assess
accurately the yielding and fatigue of these connections, under realistic wave loads.
I X. Acknowledgements
The author wishes to thank Bureau Veritas for permission to publish this paper. The views
expressed are those of the authors and do not necessarily reflect those of Bureau Veritas.
Thank also to all of my colleagues of the VeriSTAR team who participated to the various studies
which inspired and guided the writing of this paper. In particular: Cristian Bran and Nolwenn
Lorans, who put in place the fully integrated model methodology; J ose Esteve, who took the time
to read and comment this paper.
X. References:
[1] Owen F. Hughes, Ship Structural Design: A rationallybased, computeraided optimization
approach, 1988
[2] Babinet J N, Passenger Ships: Contribution of superstructures to overall bending strength,
Bulletin Technique Bureau Veritas, 4  1995
[3] Bureau Veritas, Rules for the Classification of Steel Ships, J uly 2007
[4] Bureau Veritas, Rules for Offshore Units, J une 2007
[5] Arselin E, Cambos P, Frorup U. Bureau Veritas : New FPSO rules based on return of
experience , OMAE FPSO 2004
[6] Malenica Bureau Veritas ; Hydrostructure interactions in sea seakeeping, International
Workshop on Coupled Numerical Dynamics 2007.
[7] Blandeau F., Franois M., Malenica , Chen X.B. Bureau Veritas, Linear and nonlinera
wave loads on FPSO, ISOPE 1999
OSEA 2008 1 Conversion oI OIIshore Floating Facilities
Conversion of Offshore Floating Facilities: How to tackle them with an
asset integrity focus
Christophe Chauviere, Deputy Head oI the Oil & Gas Marine Technical Department, Jose Esteve, OIIshore
Projects Manager.
ABSTRACT
These last years have seen an ever increasing demand Ior Iloating Iacilities, many oI
which have been made possible thanks to converting existing trading vessels. Most oI
those tankers had only minor issues, iI any at all, during their sailing times and it is quite
tempting to conclude that as it didn`t have problems in the past it will probably not have
problems in the Iuture as long as corrosion is kept at bay.
This is not quite true as the approach towards a trading ship and an oIIshore Iloating
Iacility (be it Ior production, storage or both) are completely diIIerent since the very
beginning, starting Irom the design. In new built units the Iormer is designed considering
that it will be Irom time to time in harbor saIety, Irequent drydocking and the captain
always has the possibility to act on speed and heading to avoid extreme weather. The
latter is designed expecting it to remain on site Ior several years and being able to
centennial conditions.
This paper will present the stateoItheart on conversion attitude and everything that can
be done as a preparation Ior eIIicient perIormance and risk management. Examples oI the
typical concerns that have been reported in converted unit will be given, as well as the
diIIerent numerical analyses that are available in order to assess the condition oI the unit
in order to avoid them.
The authors expect this paper to help push up the quality and level oI outstanding
converted units.
INTRODUCTION
F(P)SO Conversions are usually considered Ior Iast track projects as they normally
provide the lowest times to Iirst oil, ReI. 1. The main structural concern oI such units,
as the action implies, is that they are modiIied in order to Iit a diIIerent purpose other
than what it was originally designed Ior, trading tanker. It is true that both purposes are
very alike but they still have diIIerences that must be kept in mind:
A tanker is designed usually under an IACS member Rules Ior sailing ships.
These rules have been draIted considering that the each trip can be measured, at
most, in months. The maximum loads calculated take this into account. Corrosion
margins and local strength criteria used in its design have as background the Iact
that a trading tanker must go to drydock periodically Ior a Iull survey, where
renewals and repairs can be done easily.
A tanker wasn`t designed to continuously support upper deck structure throughout
its length, such as topsides in a FPSO.
A trading tanker inspection plan usually limits to the Class Surveys, predeIined
Irom Rules.
In opposition to all oI the above a converted F(P)SO intended to stay Ior several years on
site with continuous production must have a taylormade inspection plan to account Ior
the particularities oI its operation. The consequences oI Iinding areas oI concern during
OSEA 2008 2 Conversion oI OIIshore Floating Facilities
any oI the surveys must be acknowledged as repairs will be costly (direct and opportunity
costs Irom eIIects on production).
This paper will move along the structural key issues that need to be closely regarded
when doing a conversion. Examples oI the diIIerent analyses that can be carried out to
avoid the main concerns are given and the main points will be illustrated with images.
STRUCTURAL KEY ISSUES
1. Design Parameters
BeIore starting the conversion oI an oil tanker, a certain number oI design parameters
should be careIully assessed in order to evaluate the potential consequences on the
structure. The asbuilt design parameters oI the existing tankers should be checked and
possible modiIication oI these parameters due to the conversion should be assessed. In
the Iollowing chapters some oI these most important parameters are detailed together
with their inIluence. It should be noted that most oI the asbuilt parameters could be
included in the Midship Section and Trim and Stability Booklet (Loading Manual) oI the
oil tanker.
1.1 Allowable loading cases
The list oI actual allowable loading conditions oI the oil tanker is deIined in the Trim and
Stability Booklet and/or Loading Manual. Generally the asbuilt design loading
conditions are not relevant Ior the use as an OIIshore unit Ior which the loading
conditions will reIer to the loading/oIIloading sequences. It should be noted Ior example
that some existing oil tankers where not designed Ior partial Iillings conditions and/or
alternate loading conditions. The maximum cargo oil density Ior which the vessel has
been designed could also be an issue because this density will be related to the maximum
scantling draught. These points should be identiIied at the very beginning oI the
conversion project since this could induce some necessary reinIorcements oI the structure
and/or induce limitations on the allowable loading conditions Ior the FPSO, see ReI. 8.
As an example an oil tanker Ior which the Iull load homogeneous was obtained Ior a
cargo density oI 0.87 t/m3 means that the unit may not be allowed to be Iully loaded at a
density oI 0.9t/m3 because the maximum draught will certainly be exceeded. To
summarize the Iollowing non exhaustive list oI design parameters oI the existing oil
tanker should be checked and compared with the requested ones Ior the use on oIIshore
site:
 Maximum cargo density
 Scantling draught
 Allowance oI partial Iillings
 Allowance oI alternate loading cases
Figure 1 shows the design loading condition Ior newbuilt double side units. Some
converted units were not expected to see such internal Iillings at their design. II
inspection and repairs on site are an accepted possibility while maintaining production
then this means such alternate Iillings need to be addressed.
OSEA 2008 3 Conversion oI OIIshore Floating Facilities
Figure 1: Design Loading Conditions
1.2 Still water bending moments and shear Iorces
When starting the conversion oI an oil tanker, any additional equipment to be Iitted on
board is to be identiIied with their weights and locations. Each oI these added weights
will not only induce local reinIorcements oI the structure but will also modiIy the
lightship weight distribution oI the vessel. This will induce modiIications oI the
longitudinal strength assessment oI the unit, i.e. the still water bending moments
(SWBM) and shear Iorces (SWSF). For each oI the expected loading conditions,
calculation oI SWBM and SWSF should be perIormed and compared with the admissible
values which were assigned to the oil tanker (remark : most oI the existing oil tankers are
Iitted with a loading calculator on board that can be used to make such veriIication).
When listing all the loading cases to be checked due consideration is to be given to
Inspection and Maintenance cases which are generally the most critical loading cases.
See ReI. 3. The ideal is to Iind realistic loading patterns that allow inspection and
repairs on any tank while keeping the SWBM and SWSF within original admissible
values. However this is not always possible. In such cases it is still possible to use the
actual meteocean conditions at site to justiIy an increase oI SWBM and SWSF as Iar as
these site conditions are less severe than North Atlantic conditions (see also 1.3 below).
1.3 Site wave environment
One oI the key parameter that will greatly inIluence the needed modiIications and/or
reinIorcements oI the existing oil tanker is the expected wave environment oI the
oIIshore site. The existing oil tankers are generally designed Ior navigation during 20
years in North Atlantic based on deterministic wave parameters deIined by Class Rules.
However the design oI oil tanker does consider that it will always be possible Ior an oil
tanker to avoid being trapped in the most severe weather conditions Ior the structure oI
the vessel. For example it is assumed that the captain oI the vessel will avoid typhoon
conditions or will change the route oI the vessel iI being in very severe head seas or beam
seas conditions. For FPSO the 100 year maximum statistical wave on site is used Ior the
design oI the structure see ReI.8 and 9. In some parts oI the world, Ior example Ior
FPSO located in North Sea or GulI oI Mexico, the wave environment is very harsh and
will be more severe than the 20 years North Atlantic basis oI oil tankers. This means that
OSEA 2008 4 Conversion oI OIIshore Floating Facilities
quiet important reinIorcements will be needed in order to avoid any structural damage on
site. On the opposite, some FPSOs are located in very mild wave environment such as
West AIrica and this gives some additional margin Ior the design oI the FPSO that could
be used to increase some design parameters (see 1.2 above). ThereIore a hydrodynamic
analysis oI the FPSO on site conditions should always be perIormed to assess wave
parameters (motions and sea pressures) on the unit. (Remark: Ior some FPSO, the most
severe wave environment will not be the site one but could be met by the unit during
towing phase. In such case it is important to consider this towing condition Ior the
design.)
Figure 2: Hydrodynamic results and comparison with Rules values
The assessment oI the actual site conditions that the unit will see in F(P)SO service is
carried out through hydrodynamic analysis, iI possible using 3D diIIractionradiation
models hydrodynamic models oI the hull and its weight distribution. Such models, in
Irequency domain, calculate the vessel motions and the main hull girder parameters, such
as Relative Wave Elevation, Vertical Wave Bending Moment, Vertical Wave Shear
Force, Horizontal Wave Bending Moment, torsional moment. Their result is usually
given in Iorm oI transIer Iunctions or Response Amplitude Operators. Combining these
to the site wave spectra enables to obtain the shortterm response and when the
meteocean data is rich enough obtain the longterm distribution, which is very important
Ior Iatigue analysis. For turretmoored units a heading analysis is needed Iirst beIore an
appropriate longterm distribution can be calculated.
Figure 4: Max draIt and minimum draIt hydrodynamic models
OSEA 2008 5 Conversion oI OIIshore Floating Facilities
RAO Roll
Long Term Distribution
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08 1.E+09
Figure 5: TransIer Iunction Ior max draIt Roll response and long term reponse Ior
VWBM Ior max and minimum draIt
Two complementary means are used to veriIy the primary structure or hull girder with the
loading conditions selected and under the calculated environmental loads, 2D section
analysis and 3D Coarse Mesh Models, see ReI. 2. Figure 6 below illustrates a Complete
Ship Model (CSM) Ior the assessment oI the primary hull structure with the application
oI the internal loads and the external wave eIIects resulting Irom the hydrodynamic
analysis.
Figure 6: Complete Ship Finite Element Model
2. Local Strength
Even though a converted tanker has suIIicient global strength as discussed on previous
sections it may well Iail locally Irom high stresses or buckling when assessed in oIIshore
operation. Once again, the main reason is that loading patterns change and the assessed
external loads probability level is diIIerent Irom the original tanker design.
There are also additional Iactors, such as the weight, placed above the deck, oI the
topside modules. The eIIect oI such loads must not be analyzed only at the deck structure
but on the whole transversal ring, as buckling Irom vertical compression may now occur.
OSEA 2008 6 Conversion oI OIIshore Floating Facilities
Figure 7 : FEM Ior local strength assessment oI a typical transverse section.
3. Fatigue issues
3.1 General
Many existing oil tankers were designed at time oI their building without any
consideration to Iatigue issues because Iatigue calculation was not requested by Class
Rules at that time. Fatigue issues became mandatory in the Class Rules only in the end oI
the Nineties. Even though the oil tankers are since then designed with consideration to
Iatigue, the design Iatigue liIe is generally about 20 years oI navigation in North Atlantic.
Attention is drawn to some oil tankers that are designed with higher Iatigue lives but with
the reduced so called Worldwide` environment. The actual age oI the vessel and the
expected design liIe on site should be taken into account Ior evaluation oI Iatigue risks
when considering the conversion into an FPSO unit. Whereas Iatigue cracks on oil
tankers could be managed thanks to the regular drydock surveys, every Iatigue crack on
an FPSO will be very diIIicult to repair. Indeed FPSO are not expected to go in drydock
and so cracks should be repaired on site and probably aIIecting the production. In
addition to the commercial impact, repairs on site also imply a lot oI saIety measures to
be taken in order to minimize accident risk. Examples exist where this has been carried
out successIully, see ReI. 4. But there will always be some risk. The best repair is the
one that does not to be done.
OSEA 2008 7 Conversion oI OIIshore Floating Facilities
Figure 8: Iatigue crack in transverse web Irame
3.2 Actual accumulated Iatigue damage
Even though the oil tanker does not present any existing Iatigue crack and/or history oI
cracks, actual Iatigue damage oI the unit should be assessed. Indeed it could appear that
actual Iatigue damages are so close to the limits that the cracks will appear very soon on
site. ThereIore the actual Iatigue damages are to be calculated by taking into account the
navigation history oI the vessel looking at the routes that have been Iollowed by the
tanker during its liIe. This kind oI history is not always accurately available and so some
conservative hypotheses are generally needed to assess the accumulated Iatigue damages
Figure 9: example oI Iatigue studies at diIIerent hot spot location on a bracket
OSEA 2008 8 Conversion oI OIIshore Floating Facilities
3.3 Onsite Iatigue damage
The expected Iatigue damage on site conditions should be evaluated by taking into
consideration the site wave environment. Whenever possible the actual measured
thickness should be taken into account Ior this calculation (see 4.1 below). The onsite
damage will then be added to the previous service accumulated one (see 3.2 above) and
then risk oI Iatigue crack appearing during liIe on site is assessed. Depending on the
conclusion oI this assessment preventive measures could then be undertaken during
conversion to reduce the Iatigueinduce crack risk. Possible preventive actions are
reinIorcements oI the structure, postwelds improvements by using Iatigue enhancement
techniques such as grinding, TIG dressing, hammer peening.
Fatigue damage not only comes Irom the external sea states but might also have the
loading/oIIloading scheme as a source. See ReI. 2. As such it should be accounted Ior.
Figure 10: Hammer peening operation
4. Corrosion issues
4.1 General
The corrosion level oI the oil tanker is a very important aspect to be considered beIore
starting the conversion. Indeed the corrosion will reduce the strength oI the structure but
will also accelerate Iatigue propagation as described in ReI.2. ThereIore an exhaustive
campaign oI thickness measurement should be done beIore the conversion in order to
identiIy the area with high level oI corrosion. Depending on structural analysis oI the
elements under consideration necessity oI renewal oI the plates could be decided.
DiIIerent approaches Ior renewal philosophy are given in ReI.2 and 6. An estimation
oI the corrosion on site based on actual measurements is strongly recommended in order
to anticipate potential problems and take preventive actions iI needed.
OSEA 2008 9 Conversion oI OIIshore Floating Facilities
Figure 11: pitting oI bottom plating
IMPORTANCE OF SURVEY DURING CONVERSION
Generally schedule Ior conversion is very tight and so it is quiet diIIicult to perIorm
extensive surveys. However it is a very important to check that there is no existing crack
on board that could propagate on site and then create troubles Ior the production. This is
why the surveys should be prepared in advance by identiIying the critical areas oI the
structure (called hot spot` areas) that will be subject to a close inspection during
conversion. Structural calculation Ior yielding, buckling and Iatigue can allow identiIying
the critical areas oI the structure and based on these analyses, hot spot maps could be
created and delivered to surveyors prior to their inspection. Since structural calculation is
time consuming it is necessary to start the calculation as early as possible in the project.
Anticipating the calculation is a key point to have an eIIicient survey that will prevent
problems on site and maximize reIurbishment eIIorts.
NECESSITY OF A NUMERICAL MODEL
All oI the previous sections clearly lead to understand the converted unit as a 'new
design that needs validation/improvement prior to service. Any past liIe must be used to
learn Irom errors and not to layback unworried. Finite Element Models allow
understanding the structure behavior under the estimated Iuture loads and can highlight in
advance any potential issue. It can also help direct the surveys and identiIy locations that
otherwise may pass unseen being the surveyor concentrated on more immediate issues,
see ReI. 2 and 5.
A new numerical model is being discussed within IACS that will contain an accurate
description oI the structure, plates, secondary stiIIening, seams and buts.Such a model
will have a standard Iormat known as Hull Condition Monitoring, HCM. It will be able to
store the gauging results and the highlighted elements Irom the FEM. It will thus be
possible to evaluate in a rational manner the required steel weight renewal.
In addition, once on service the model can be updated with the inspection campaigns and
be used to display survey results in a very intuitive manner, helping deIine next actions.
The next Iigures show some snap shots oI a real FPSO structure.
OSEA 2008 10 Conversion oI OIIshore Floating Facilities
Fig. 12 Snapshop oI a Hold being 'Ilythrough in HCM Iormat
Fig. 13 Thickness measurements taken on a bottom transverse plate
CONCLUSION
The Iact that no signiIicant deIect was recorded while sailing doesn`t mean that nothing
will happen in the Iuture. Tools exist that can help avoid bad surprises and minimize
Iuture risks (deIined as probability oI Iailure x consequences oI Iailure). II a saIe long
oIIshore service is wanted Irom a converted F(P)SO eIIorts must be done at the
conversion time.
AKNOWLEDGEMENTS
The authors wish to thank Bureau Veritas Ior permission to publish this paper. The views
expressed are those oI the authors and do not necessarily reIlect those oI Bureau Veritas.
The authors also wish to thank the VeriSTAR team Ior providing the images oI the 3D
models.
OSEA 2008 11 Conversion oI OIIshore Floating Facilities
REFERENCES
1 'Reducing Cycle Time To First Oil For Deepwater Field Developments, Beverley F.
Ronalds. OMAE200567016
2 'Converted Fpso's. Making A Worthwhile Conversion, J. Esteve Otegui, M. Franco
Orsini. DOT 2004, New Orleans.
3 'Converted Single Hull F(P)SO Challenges Regarding Inspection Repair and
Maintenance Guillaume Gourdet, Paulo Biasotto, DOT 2006
4 'Major Steel Renewals in FPSO Ballast Tanks Without Ceasing Production, P.
Nelson, OTC 1990.
5 'FPSO Hull Maintenance Field Experience, Rodrigo Andre Hoppe, Rodrigo Klim
Gomes, Viviane C.S. Krzonkalla, OMAEFPSO 040070.
6 'Hull Structural Assessment and ReIurbishment Ior Conversion FPSOs OMAE
FPSO'040030, T. de Beer, N. van Nood
7 Bureau Veritas, 'Rules Ior the ClassiIication oI Steel Ships, July 2007
8 Bureau Veritas, 'Rules Ior OIIshore Units, June 2007
9 Arselin E, Cambos P, Frorup U. Bureau Veritas : 'New FPSO rules based on return
oI experience , OMAE FPSO 2004
1 Copyright 2008 by ASME
INFLUENCE OF FIBRE STIFFNESS ON DEEPWATER MOORING LINE RESPONSE
Peter Davies / IFREMER
29280 Brest, France, pdavies@ifremer.fr
Patrice Baron / Doris Engineering
75013 Paris, France
Karine Salomon / Saipem
13321 Marseille, France
Charles Bideaud / Technip
92973 La Dfense, France
JP Labb / Acergy
92156 Suresnes, France
Stphane Toumit / Principia
13600 La Ciotat, France
Michel Francois / Bureau Veritas
92400 La Dfense, France
Francois Grosjean / IFP
69390 Solaize, France
Tony Bunsell / ENSMP
91003 Evry, France
AG Moysan / Total
92069 La Dfense, France
ABSTRACT
Polyester Iibre ropes are now an accepted solution Ior
deepwater mooring oI production platIorms and a single high
tenacity Iibre grade is widely used. Few studies oI other Iibres
have been reported but polyesters can be produced with a range
oI properties by varying drawing parameters, and other stiIIer
Iibres are also available. This paper presents a study oI these
alternative Iibre rope solutions, perIormed within the French
Mooring Line project. First, in order to obtain the input data
necessary Ior mooring line analyses an extensive test program
was perIormed to characterize polyester, improved polyester,
PEN, LCP, aramid and HMPE Iibre ropes Irom yarns up to 800
ton break load ropes. Tests at diIIerent scales have allowed
property transIer to be quantiIied. Rope modelling has been
used in parallel to examine the inIluence oI material and
structural parameters. Then, using these data, a series oI
analyses was run by engineering contractors, which quantiIied
the beneIits oI higher stiIIness Ior diIIerent supports (semi
submersible, production platIorm and oIIloading buoy) down to
2500 meter depth. Under certain conditions the stiIIer ropes can
result in signiIicantly reduced rope diameter and weight.
Keywords: Synthetic, Stationkeeping, Fibre, Rope
INTRODUCTION
Following studies by del Vecchio and Chaplin 1,2
Bureau Veritas approved the Iirst polyester mooring system Ior
a Iloating platIorm Ior Petrobras in 1997 in 800 meters water
depth. Since then polyester ropes have been used extensively
oIIshore Brazil Ior station keeping 3, and in 2004 the Iirst
synthetic mooring lines were installed in the GulI oI Mexico
4,5. The Iibre used Ior all oI these applications has been a
high tenacity polyester, with a tensile modulus around 15 GPa.
However, there is a very large range oI polymeric Iibres
commercially available, with moduli up to 280 GPa, and even
polyester can be drawn and heat treated to obtain moduli over a
wide range. Trials have been perIormed successIully with
higher stiIIness Iibres such as HMPE on drilling rigs 6 and
ropes composed oI mixtures oI Iibres are being used Ior deep
sea handling 7. The choice oI the best rope Ior a given
mooring application is not simple, and extensive data are
necessary to validate the selection.
The paper is organized as Iollows: Iirst, a short overview
oI work perIormed over the last 10 years within the French
Mooring Line project is given. Then the diIIerent materials
which have been studied are presented, both Iibres and ropes.
Next the tests which have been perIormed are discussed and
results are given. The use oI rope modelling to complement
tests is then discussed. Finally the data which have been
generated are used in mooring line analyses Ior three case
studies, to illustrate how Iibre stiIIness aIIects mooring line
perIormance.
Proceedings of the ASME 27th International Conference on Offshore Mechanics and Arctic Engineering
OMAE2008
June 1520, 2008, Estoril, Portugal
OMAE200857147
2 Copyright 2008 by ASME
FRENCH MOORING LINE PROJECT
Since 1996 a group oI research institutes, engineering
companies, a classiIication society and an oIIshore operator
have been working together within the French Mooring Line
project to understand the behaviour oI synthetic mooring lines
and characterize their short and long term behavior. Part oI that
work has been published previously and has included stiIIness
tests on ropes and subropes 8,9 development oI rope stiIIness
models 10,11, creep and long term tests 12, modelling oI
long term behaviour 13 and a study oI tensiontorsion
interactions 14. Rope construction has also been examined
and correlations between model predictions and some oI the
experimental data, notably Ior aramid ropes, have been
presented 15,16. In a parallel project tests at sea were also
perIormed in order to measure loads insitu 17.
Throughout the project there has been an interest in
alternative materials to polyester, and some work presented
earlier on very stiII Iibres such as HMPE and aramid indicated
that, while reducing rope weight, these materials could cause
long term Iatigue problems in the chain sections oI a taut
mooring line system 18. However, since then more tests
particularly on intermediate modulus Iibres such as PEN 19
and Jectran have been perIormed, and the present paper
presents a wider view oI how stiIIness aIIects mooring line
perIormance and the role oI Iibre stiIIness and strength.
MATERIALS CONSIDERED
High tenacity polyester is the reIerence Ior oIIshore
mooring applications today, and the perIormance oI all other
materials is compared to polyester ropes. Two aspects must be
considered, the behavior oI the Iibre itselI, which provides the
material response, and the translation oI this behavior into the
rope. The eIIiciency oI this transIer is critical and both aspects
must be quantiIied and understood iI an eIIicient mooring line
is to be designed. Table 1 lists the seven materials which have
been studied in the project and the type oI specimens tested.
Polymer ReIerence Fibre Yarn Subrope
(Tons)
Rope
(Tons)
PET
Standard
855TN
1W81
25 70
30
500  800
450
PET
ModiIied
1W83
35 500
PEN 1P70 36
LCP HS 50
Aramid T1000 25 205
HMPE SK75
SK78
70
70
700, 750
PBO HM 60
Table 1. Samples tested
Two types oI PET (polyethylene terephthalate) have been
studied, standard grades Irom Diolen and PerIormance Fibers
and a grade modiIied to give higher initial stiIIness supplied by
the latter. PEN (polyethylene naphthalate), also Irom
PerIormance Fibers, was chosen to represent the intermediate
stiIIness range, Iibre initial stiIIness is around twice that oI PET
at 30 GPa compared to 15 GPa. The other Iibres are all in the
high perIormance category, LCP (liquid crystal polyester) is a
generic name Ior the Iibre more commonly reIerred to as
Jectran developed by Celanese and now marketed by Kuraray.
The aramid is a Iibre Irom Teijin similar to the Kevlar 29 grade,
two HMPE (high modulus polyethylene) grades Irom DSM
have been studied, and a small number oI tests were perIormed
on PBO (polypphenylenebenzobisoxazole), a very high
perIormance Iibre Irom Toyobo sold under the trade name
Zvlon.
MATERIAL CHARACTERIZATION
In order to characterize the material, (as opposed to the
rope, which can be considered as a structure), tests are
perIormed on either single Iibres or yarns. These are the basic
elements which will conIer to the rope all or part oI their
properties, and they provide the input data Ior subsequent
modelling.
Single filament testing
Testing oI single Iilaments oI diameters in the range Irom
10 to 30 microns is a very delicate operation. Within the project
this activity was perIormed at the Ecole des Mines in Paris,
using the special test machine shown below, Figure 1 20. This
enables quasistatic, creep and cyclic loads to be applied.
Figure 1. Single filament test machine
Yarn testing
Yarns are bundles oI tens or hundreds oI Iilaments and
yarn testing is the basic quality control method used in the rope
industry. The main concern is the gripping system and machine
suppliers provide special grips Ior this type oI test. In most
cases only yarn tenacity (strength) is measured, but Ior
modeling purposes stiIIness is also required. In order to
measure strain accurately the machine crosshead displacement
is not accurate enough and a noncontact extensometer is
essential. In this work two digital cameras and inhouse image
analysis soItware were used at IIremer to obtain stressstrain
data with the test setup shown in Figure 2a. Additional tests
were perIormed on assembled yarn samples in which tens oI
yarns are twisted together, Figure 2b.
3 Copyright 2008 by ASME
aa
bb
Figure 2. Yarn test fixtures and camera extensometer
a) yarn, b) assembled yarn
ROPE CHARACTERIZATION
The best way to characterize rope properties is to test Iull
size ropes. Within the project Iive campaigns were perIormed
on a total oI ten Iull size rope samples at the LCPC civil
engineering center in Nantes, on a 2400 ton capacity test
machine, Figure 3. The samples were spliced at both ends
overall length was around 12 meters.
Figure 3. Full size rope on test machine, LCPC Nantes
UnIortunately Iull size rope tests are expensive and there
are very Iew suitable test machines available, so in parallel with
the tests on Iull size samples over 50 subrope samples were
tested, on three test machines with up to 100 tons capacity, at
IFP and IIremer, Figure 4.
Figure 4. Subrope on test machine
The parallel construction oI most oI the ropes means that
subrope tests can provide data representative oI that Irom Iull
size ropes provided exactly the same subrope geometry is
retained. Subrope samples were 6 to 8 meters long. End
terminations were spliced loops, as shown in Figure 4.
In order to obtain input data Ior mooring line analyses the
Iollowing stiIIness values are required:
 quasistatic (QS) 1030 MBS (minimum break
strength), 1 hour test
 dynamic (typically in the ranges 2030, 3040 and
4050 oI MBS)
It is also important to quantiIy the inIluence oI mean load and
amplitude on dynamic stiIIness values 18,21. The recent ISO
document 22 provides guidelines on testing Ior polyester,
while guidelines Ior testing oI both polyester and other,
including higher stiIIness ropes are given in 23. Testing oI the
latter is however, a specialist activity requiring many
precautions to be taken iI valid, useIul data are to be obtained.
Extensometry is again a key point, and the systematic doubling
oI extensometers, using cameras and wire displacement
transducers on the same sample, to check displacement
measurements can help to detect sources oI error such as
transducer clamp movements.
One point which is oIten raised is the need to perIorm
measurements on wet ropes, and the eIIect oI testing samples
dry rather than in water. While it is clear that the environment
will be critical Ior Iatigue tests, the majority oI the work
perIormed within the project involved stiIIness measurements
lasting Ior relatively short periods. When such tests are not
perIormed in water there may be spurious heating eIIects, not
representative oI the Iinal application. In order to examine
when such eIIects might inIluence measurements an InIraRed
camera was employed during one campaign on polyester ropes.
4 Copyright 2008 by ASME
This enabled surIace temperature to be measured continuously,
and an embedded temperature probe between the outer strands
and the core strand provided additional measurements oI
temperature inside the rope. A series oI cyclic tests was run
with progressively increasing amplitude. Figure 5 shows
thermographs at the end oI cycling at three amplitudes, Figure
6 shows an example oI temperature probe results. These tests
were perIormed on a 500 ton break load twisted strand parallel
construction rope, loaded at increasing amplitudes oI +3.5,
+5, +7.5, +10 and +12.5 MBS, with 100 cycles at each
level, Ior a mean load oI 30 Fr and a loading Irequency oI 0.1
Hz.
+3.5
+7.5
+12.5
Figure 5. IR Thermographs showing rope surface
temperature for different loading amplitudes.
Temperature increase, 100 cycles at each amplitude.
10
11
12
13
14
15
16
17
18
19
20
0 200 400 600 800 1000
Time, s
T
C
7%
10%
15%
20%
25%
Figure 6. Temperature variations inside the rope
measured by temperature probe during loading at
different amplitudes.
A signiIicant temperature increase, above 1C, is only noted at
high amplitudes (above +7.5). It is interesting to note,
however, that above this amplitude there is a change in slope oI
the plot oI stiIIness versus amplitude, Figure 7. This plot is
normalized with respect to the stiIIness at the lowest amplitude.
0,85
0,9
0,95
1
0 10 20 30
Amplitude % MBS
K
r
10
15
20
25
30
T
e
m
p
e
r
a
t
u
r
e
,
C
Kr
T C
Figure 7. Influence of amplitude on stiffness.
These results suggest that it is not necessary to make dynamic
stiIIness measurements in water provided the applied amplitude
is below +7.5MBS. Above this value the results should be
treated with caution.
TEST RESULTS
In this section some examples oI test results will be
presented. In the limited space available it is not possible to
give all the data so only some examples will be shown. Figure
8 presents a comparison oI typical material properties oI the
diIIerent types oI Iibre tested in the project, based on single
Iilament and yarn tensile tests.
5 Copyright 2008 by ASME
Fibre behavior
0
1
2
3
4
0 5 10 15
Strain, %
T
e
n
a
c
i
t
y
,
N
/
t
e
x
PBO HM
PET
Pentex
Vectran
Aramid
HMPE
Figure 8. Properties of different fibres, as received.
In most oI the test procedures Ior ropes a beddingin
procedure is recommended. This involves cycling the rope to
load levels above those which will be subsequently used Ior
stiIIness testing in order to stabilize it. During this cycling, Iive
cycles up to 50 oI break load Ior example, the material can be
reoriented and the rope structure can change.
Tests at diIIerent scales with and without bedding in enable
the inIluence oI the rope structure to be evaluated. Figure 9
shows one example, Ior a modiIied polyester, which has been
tested aIter bedding in at single Iilament, yarn, assembled yarn
and subrope scales. This shows how Iilament properties are
transIerred to subropes
0
0,2
0,4
0,6
0,8
1
0 3 6 9 12 15
Strain (%)
F
o
r
c
e
(
N
/
t
e
x
)
filament
yarn
ropeyarn
subrope
Figure 9. Modified polyester, beddedin, tested at four
scales [24]
ROPE MODELLING
When material properties are available Irom either single
Iilament or yarn tests they can be used in rope modelling
soItware to predict rope properties. The most developed tool is
FRM (Fibre Rope Modeller), a commercial set oI soItware
Irom TTI 25. This is a useIul tool and proved very successIul
in simulating tension/torsion results Irom Iull scale tests
perIormed within the project 14. It can take nonlinear Iorce
strain plots Irom low level tests such as those perIormed on
Iibres, yarns or assembled yarns, and use these data together
with details oI the rope structure (dimensions, lay lengths) to
predict the behavior oI Iull size ropes. This oIIers the
possibility to reduce the amount oI testing, provided the models
have been validated, and one oI the aims oI the present project
was to develop a data base enabling the models to be checked
Ior a range oI materials.
MOORING LINE CASE STUDIES
Using data generated within the project a large number oI
mooring line analyses were perIormed over the 10 year period.
Initially the conditions oI interest were oII West AIrica at
depths oI 1000 to 1500 meters, but gradually the analyses were
extended and Ior the last series 25003000 meter depths Ior
both West AIrica and the GulI oI Mexico were examined.
Three examples oI mooring line applications will be
discussed here. Five materials will be examined, using the
properties shown in Figure 10 and Table 2. Here Kr is a
normalized stiIIness parameter, deIined as (Load/MBS)/Strain.
Kr values
0
10
20
30
40
50
60
70
80
90
100
0 10 20 30 40 50 60
Mean Load, % MBS
polyester
Pentex
Vectran
Aramid
HMPE
Figure 10. Mean load dependence of stiffness. Open
symbols: quasistatic, Closed symbols: dynamic.
Material Static Dynamic response
PET 15 18.5 0.33 (ML)
PEN 20 23 0.33 (ML)
LCP 33 30 0.61 (ML)
Aramid 47 45 0.58 (ML)
HMPE 56 59 0.54 (ML)
Table 2. Normalized stiffness values used in analyses
It should be emphasized that this is not a deIinitive set oI
data Ior these materials, but simply one set oI data generated by
tests on samples oI these materials with particular rope
6 Copyright 2008 by ASME
constructions. From these tests we know that we can obtain
these stiIIness values, but by changing rope constructions it is
undoubtedly possible to obtain other higher or lower values.
This approach was preIerred, as results such as those shown in
Figure 9 indicated that there can be very large diIIerences
between Iibre and yarn behaviour and rope perIormance.
Modelling was also used within the project, but the models do
not account Ior unloading behaviour during dynamic cycling.
The rope break strengths and weights used in the analyses
were taken Irom both measurements and rope suppliers data.
Figure 11 shows the properties used Ior the diIIerent materials.
0
500
1000
1500
2000
2500
40 60 80 100 120 140 160 180 200
Diameter, mm
B
r
e
a
k
l
o
a
d
,
t
o
n
s
HMPE, Vectran
Aramid
PET, PEN
Figure 11. Nominal rope strengths versus diameter
Key issues in dimensioning with materials other than
polyester are what saIety Iactors to use and how low a value oI
minimum tension can be allowed. Table 3 shows the saIety
Iactors and minimum tensions adopted in the Iirst analyses Ior
each material. These were subsequently varied, as will be
discussed Iurther below.
Material Minimum
tension
SaIety Iactors
Intact / Damaged
PET 2 1.925/1.375
PEN 5 2.1/1.5
LCP 5 2.1/1.5
Aramid 10 2.1/1.5
HMPE 5 2.1/1.5
Table 3. Assumptions used in analyses for each
material.
Case study 1.
The Iirst case examined is that oI a semisubmersible
drilling rig in 2500 meters water depth in the GulI oI Mexico.
The rig is moored with 8 lines, with a length oI chain at the top,
3300 meters oI synthetic rope, and 100 meters oI steel cable at
the sea bed. The analysis was perIormed using the Ariane
soItware and a quasidynamic analysis. For each analysis there
are three parameters to Iix : the material, the rope diameter and
the pretension. From the analysis, with a maximum oIIset
criterion oI 3 oI water depth, the active length oI the line and
the maximum and minimum horizontal and axial tensions are
obtained. Given the environmental conditions (wave, wind and
current including loop currents) the two lines with the highest
and lowest tensions are examined to check that the criteria oI
minimum tension and maximum tension (saIety Iactor) are
respected. DiIIerent pretensions are studied in order to
optimize the response Ior each material.
Particular attention was paid to the minimum tension, and
Ior each material three minimum tension values were
examined. Figure 12 shows an example oI results.
0
30
60
90
120
150
180
PET PEN LCP Aramid HMPE
D
i
a
m
e
t
e
r
,
m
m
2%
5%
10%
Figure 12. Influence of minimum tension on rope
dimensions, semisubmersible drilling rig GOM.
First it is clear that the high perIormance ropes oIIer very
signiIicant reductions in rope diameter. This results in even
greater weight savings. There is also a strong inIluence oI
minimum tension, simply reducing the value Irom 5 to 2
minimum tension can result in a weight saving oI around 10.
The value oI 2 was chosen Ior polyester Iollowing extensive
Iatigue tests which showed that Iatigue liIetimes were not
aIIected when minimum tension was reduced Irom 5 to 2
26. This exercise has not been perIormed Ior the other ropes,
which is why higher values were taken initially.
For this application, where long term durability is not the
main concern, these alternative materials appear very
promising, and indeed, sea trials with HMPE were very
encouraging 6. An analysis oI the Iatigue behaviour showed
that the Iatigue liIe (oI the chain section) decreased as rope
stiIIness increased, Figure 13, but even Ior the stiIIest Iibres a
liIetime oI 250 years (saIety Iactor oI 10) was achieved.
7 Copyright 2008 by ASME
0
0,5
1
PET PEN LCP Aramid HMPE
> 250 years
Figure 13. Relative fatigue life of chain, compared to
chain fatigue life in PET line
Case study 2.
The two Iollowing examples illustrate production
applications expected to remain in operation Ior 20 years. Here,
in addition to stationkeeping perIormance Iatigue behaviour
must be guaranteed.
First a production barge moored in 2500 metres oI water
oII West AIrica by 16 lines was considered. Figure 14a shows
the diameters required Ior the diIIerent materials, Figure 14b
shows potential weight gains on the synthetic part oI the lines.
a) Diameter
0
20
40
60
80
100
120
140
160
180
PET PEN LCP Aramid HMPE
D
i
a
m
e
t
e
r
,
m
m
b) Rope weight
100
77
52
66
33
0
20
40
60
80
100
120
PET PEN LCP Aramid HMPE
%
Figure 14. Influence of material on a) rope
dimensions and b) weight gain, production barge
West Africa.
Once again signiIicant reductions in rope diameter and
weight can be achieved. The limiting Iactor in this case is the
Iatigue perIormance oI the line, and in particular the high loads
introduced in chain elements due to the high line stiIIness 18.
This was one oI the main incentives Ior studying intermediate
modulus Iibres such as PEN, which allow a more modest
reduction in diameter and weight but without greatly reducing
Iatigue liIetime.
Case study 3.
A second example Ior West AIrica involves an oIIloading
buoy in 1000 meter water depth, Figure 15. The buoy is
moored by 9 (3x3) lines.
CALM
Buoy
CALM
Buoy
CALM
Buoy
Figure 15. Case study 3, offloading buoy
Here a Iull dynamic analysis was perIormed using the
Deeplines soItware Ior the diIIerent materials. Figure 16 shows
an example oI rope diameters Ior the diIIerent materials. Again
the stiIIer Iibre ropes can reduce mooring line diameters.
8 Copyright 2008 by ASME
0
20
40
60
80
100
120
140
PET PEN LCP Aramid HMPE
D
i
a
m
e
t
e
r
,
m
m
Figure 16. Rope diameters, offloading buoy.
CONCLUSION
This paper summarizes work perIormed on synthetic mooring
lines within the French Mooring Line project. The work
perIormed within the project has contributed to the
development oI standards 22, practical rope models 10,11
and classiIication society documents 23. In this paper
particular emphasis is placed on a study into how alternative
Iibres to the polyester currently used can improve mooring line
perIormance.
Extensive testing oI Iibres and yarns has enabled a large
database oI material properties to be obtained. Tests on sub
ropes and Iull size ropes then allowed stiIIness data to be
generated, in a Iorm which is suitable Ior mooring line analysis
soItware. Results Irom mooring line analyses Ior diIIerent
supports and environments have shown the beneIits oI higher
perIormance Iibre ropes Ior certain applications. However, long
term behaviour must also be characterized as many oIIshore
structures will be operational Ior 20 years or more. Current
installations are providing valuable experience Ior polyester
ropes but a signiIicant eIIort is needed to characterize the long
term behaviour oI other materials iI mooring systems are to be
improved.
ACKNOWLEDGMENTS
The technical expertise oI the many colleagues at IIremer, IFP,
ENSMP and LCPC who were involved in the work presented
in this paper is grateIully acknowledged.
The opinions expressed in this paper are those oI the authors
and do not necessarily represent those oI their companies.
NOMENCLATURE
Kr Normalized stiIIness
MBS Minimum Break Strength
PET Polyethylene Terephthalate
PEN Polyethylene Naphthalate
QS Quasistatic
Tex weight in grams per km
REFERENCES
1 Del Vecchio CJM (1992), Light weight materials Ior deep water
moorings, PhD thesis University oI Reading
2 Chaplin CR, Del Vecchio CJM, (1992) Appraisal oI lightweight
moorings Ior deepwater OTC6965.
3 De Pellegrin I, (1999), Manmade Iiber ropes in Deepwater
Mooring Applications, OTC 10907.
4 Bugg DL, Vickers DT, Dorchak CJ, (2004) Mad Dog project:
Regulatory approval process Ior the new technology oI synthetic
(polyester) moorings in the GulI oI Mexico, OTC 16089.
5 Haslum HA, Tule J, Huntley M, Jatar S, (2005) Red Hawk
polyester mooring system design and veriIication, OTC 17247.
6 Corbetta I, Sloan F, (2001) HMPE mooring line trial Ior Scarabeo
III, OTC 13272
7 Ingeberg P, Torben S, Bull S, SuccessIul use oI a Iiber rope
deployment system, Sea Technology July 2007, 2325.
8 Davies P, Baizeau R, Grosjean F, Franois M, (1999) Testing oI
large polyester cables Ior mooring line applications, Proc ISOPE,
Brest, p360.
9 Grosjean F, Davies P, Francois M, (2005) Synthetic Fiber Ropes
mooring: technical status and stiIIness prediction, Proc CMOO4.
10 Francois M, Davies P (2000) Fibre rope mooring a practical
model Ior the analysis oI polyester mooring systems, Rio Oil &
Gas, 2000.
11 Francois M, Davies P (2008) Characterization oI polyester
mooring lines, OMAE 200857136
12 Davies P, Huard G, Grosjean F, Francois M, (2000) Creep and
Relaxation oI Polyester Mooring Lines, OTC 12176.
13 Davies P, Chailleux E, Francois M, Grosjean F, Bunsell A,
(2003), Prediction oI the long term behavior oI synthetic mooring
lines OTC 15379.
14 Davies P, Bouquet Ph, Conte M, DeuII A, (2006) Tension
Torsion Behavior oI Deepwater Synthetic Mooring Lines
OTC 17872
15 Ghoreishi SR, Cartraud P, Davies P, Messager T, Analytical
modeling oI synthetic Iiber ropes subjected to axial loads. Part I:
A new continuum model Ior multilayered Iibrous structures, Int
J Solids and Structures, Vol. 44, Issue 9, 1 May 2007, 29242942
16 Ghoreishi SR, Davies P, Cartraud P, Messager T, Analytical
modeling oI synthetic Iiber ropes. Part II: A linear elastic model
Ior 1 6 Iibrous structures, Int J Solids and Structures, Volume
44, Issue 9, 1 May 2007, 29432960
17 Foulhoux L, Pennec S, Damy G, Davies P, (1999), Testing oI a
large diameter polyester rope oIIshore West AIrica, Proc ISOPE
1999, Brest.
9 Copyright 2008 by ASME
18 Davies P, Francois M, Grosjean F, Baron P, Salomon K,
Trassoudaine D, Synthetic mooring lines down to 3000 meters
depth, (2002) OTC 14246.
19 Lechat C, Bunsell AR, Davies P, Piant A, Mechanical behaviour
oI polyethylene terephthalate & polyethylene naphthalate Iibres
under cyclic loading, Journal oI Materials Science, Vol. 41,
(2006), pp17451756
20 Bunsell AR Hearle JWS, Hunter RD, (1971) An apparatus Ior
Iatigue testing oI Iibres, Jnl Physics E, 4, 86872.
21 Fernandes AC, Del Vecchio CJM, Castro GAV, Mechanical
Properties oI Polyester Mooring Cables, Int. J. OIIshore & Polar
Eng., 9, 3, Sept. 1999, pp207213.
22 ISO 18692 (2007), Fibre ropes Ior oIIshore station keeping
Polyester
23 Bureau Veritas, (2007) CertiIication oI Iibre ropes Ior deep water
oIIshore services, NI432R01.
24 Lechat C, (2007), PhD thesis, Mechanical behaviour oI polyester
Iibres and Iibre assemblies Ior oIIshore mooring lines, Ecole des
Mines de Paris, (in French).
25 BanIield SJ, Flory JF, (1995), Computer modeling oI large high
perIormance Iiber rope properties, Proc OCEANS 95, San Diego,
October.
26 BanIield SJ, Casey NF, Nataraja R, (2005) Durability oI polyester
deepwater mooring rope, OTC 17510.
DIRECTIONAL WAVE PARTITIONING AND ITS APPLICATION TO THE
STRUCTURAL ANALYSIS OF AN FPSO
Ruth Lawford
Fugro GEOS, Wallingford, UK
Jill Bradon
Fugro GEOS, Wallingford, UK
Thomas Barberon
Bureau Veritas, Paris, France
Claude Camps
Bureau Veritas, Paris, France
Richard Jameson
Hess Limited, Aberdeen, UK
ABSTRACT
A full characterisation of the individual components of a
seastate is key to enabling the response of an offshore structure
to be accurately calculated. This paper discusses the
partitioning of a time series of directional wave spectra into
windsea and swell components with distinct frequency and
direction characteristics. Once the wave data have been
partitioned, JONSWAP or PiersonMoskowitz parameters can
be fitted to each spectrum using bestfit techniques. The
result of the partitioning and fitting analyses is a time series of
wave parameters defining the wave spectrum for each
component of the sea state.
A 10year site specific time series of directional wave
spectra has been partitioned in this way and used in the analysis
of the Triton FPSO, a turret moored FPSO in the central North
Sea. The representation of the directionality and magnitude of
each environmental force acting simultaneously on the vessel,
allows the relative heading of the vessel to be determined and
the mooring and hydrodynamic analyses to be performed.
These analyses provided input to a structural analysis of the
FPSO, which resulted in an inspection plan for monitoring the
effects of the metocean conditions on the unit.
1. INTRODUCTION
There is increasing interest among offshore operators and
classification societies in performing hydrodynamic and
mooring analyses in the time domain, i.e. for several years
duration. To calculate the response of a vessel at each time step
it is necessary to define the parameters of wave spectra that
represent the wave conditions at a particular location. At any
one point in time a wave climate consists of a combination of
locallygenerated
Figure 1. Triton FPSO
wind sea, plus swells which have travelled to the site from
distant storms. The wind sea and the swells often approach the
vessel from different directions and hence it is important to be
able to separate these components before fitting analytical
spectra to the wave data.
Directional wave data for the vessel location can be in the
form of hindcast or measured data. Typically, hindcast data
provides spectral energy values for a range of wave frequencies
and directions at 6hourly time intervals. The spectral data are
read into partitioning software which identifies the spectral
peaks and then fits a JONSWAP spectrum to each peak using a
leastsquares fitting procedure. The output from the program is
a time series of Hs, Tp, direction and spectral parameters and
for each identified seastate.
a
and
b
may also be provided if
required.
1 Copyright 2008 by ASME
Proceedings of the ASME 27th International Conference on Offshore Mechanics and Arctic Engineering
OMAE2008
June 1520, 2008, Estoril, Portugal
OMAE200857341
Hindcast data such as this has a variety of applications.
Here it is applied to the case of the Triton FPSO seen in Fig. 1.
Triton is a turret moored FPSO operated by Hess Limited in the
central North Sea, see Fig. 2. The vessel is free to weather
vane around the turret in response to the wave, wind and
current forces acting on it. Depending on the magnitudes and
directions of the applied forces the vessel will take up a heading
relative to prevailing environmental conditions.
Figure 2. Location of Triton FPSO in the central North Sea:
5705N, 00053E
During late 2006 Hess decided that a cost effective,
proactive development program to monitor vessel integrity
would be advantageous. Analyses were performed by Bureau
Veritas to assess the current state of the unit, focusing on the
mooring system and the vessel structure. FPSO responses are
based on a hydrodynamic analysis, performed by combining the
environmental loads at the appropriate headings with the
response amplitude operators of the vessel. Key to this analysis
is the ability to describe analytically the wave spectra
approaching the vessel from different directions. Current and
wind effects may also be a factor. Fugro GEOS provided a ten
year hindcast metocean time series containing wave spectral
parameters for the bimodal spectrum along with wind and
current, speed and direction data.
2. PARTITIONING AND FITTING DIRECTIONAL
WAVE SPECTRA
2.1 DATA SOURCE
Hindcast directional wave spectra and wind time series
data are obtained from the Fugro Oceanor Worldwaves
database. The database is derived from the European Centre for
Medium Range Weather Forecasts (ECMWF) Wave Model
(WAM) and calibrated against satellite data. Model output may
be provided in the form of integrated wave parameters or for a
more complete set of information on the sea state a spectral
output is available. Bidlot et al [1], provide an overview of
recent changes to the ECMWF model and the ongoing
improvement of spectral information. The model currently
assimilates wave heights from the ENVISAT satellite altimeter
mission launched by the European Space Agency (ESA).
Previous to October 2003 the ERS2 altimeter was used, also an
ESA mission. In order to provide spectral output the model
utilises more comprehensive, two dimensional observations
from Synthetic Aperture Radars (SAR). Since 2000, directional
spectra from the model have been available for a range of 24
directions and 30 frequencies. Data from the wa ve model are
available on a global grid and for the purpose of this work, ten
years of spectral data from the nearest point on the model grid
were utilised.
Homogeneity changes in the model present challenges for
the provision of continuous long term hindcast data. For
example the failure of an altimeter has been shown to cause
changes in the bias and variance when compared to wave buoy
data (Bidlot et al [1]). Each satellite mission has a set length
and may be replaced from time to time with a more
sophisticated altimeter or SAR.
In order to combat these difficulties and provide a fully
integrated source of wave data, Fugro Oceanor AS perform
bespoke calibration of the model with a range of further
satellite data and, if available, insitu buoy networks. The main
source of altimeter data for the calibration has been the
Topex/Poseidon mission, which ran from 1992 to 2002 before
replacement with the Jason mission. Topex 2 was introduced to
follow ground tracks midway between Jason tracks, increasing
the amount of data from 2001. The satellite data are reorganised
from individual tracks to data contained within a 10 x 10
square areas. All data are retained and no averaging is
undertaken. The wave height and wind speed data are corrected
to remove bias relative to a large satellite/buoy matched data set
and are run through a carefully designed automatic data control.
The closest alongtrack locations for each satellite mission to
each grid point are found and time series of the calibrated
satellite wave heights are extracted. Scatter plots are then
produced for the matched wave model against satellite
significant wave heights. Any inhomogeneities occurring in the
data are investigated and corrected if necessary. This method
of calibration has regularly yielded high quality results
throughout the North Sea when compared to sources of insitu
data such as that from the Ekofisk and K13 platforms. The wind
speed is also calibrated using an analgous procedure. Global
comparison statistics for the model against satellite data are
presented in Barstow et al [2].
Calibration of this type is applied to each grid point on a
global grid both for the integrated parameters and for the
spectrum as a whole. In order to obtain a data set unique to the
Triton site, a further site specific calibration was undertaken
using data from the closest satellite tracks; Topex and Geosat
FollowOn. The resulting calibrated values of significant wave
height (Hs) from the wave model are plotted against the closest
satellite data in Fig. 3, demonstrating the high quality of these
data.
2 Copyright 2008 by ASME
Data are provided as a sequence of spectral matrices for
each time step ready for input to the partitioning process. A
concurrent time series of integrated wave parameters, wind
speed and wind direction is also required as an output from this
process
Figure 3. Calibrated model significant wave height (Hm0) against
satellite (Hs).
2.2 PARTITIONING PROCEDURE
An established partitioning algorithm detailed in work by
Hanson and Phillips [3] and Aarnes and Krogstad [4] which in
turn used ideas from Hasselmann et al [5] and Gerling [6] is
used to identify the peaks within the spectra for each time step
and assign the energy to windsea or swell components.
Partitioning may be undertaken using spectra from numerical
model hindcast data or measured wave buoy data, or to allow
comparison between the two, as described in Feld and Mork
[7]. A summary of the partitioning methodology is provided in
the following sections.
2.2.1 Peak Isolation and Partitioning
Each value or point in the spectral matrix represents a
frequencydirection combination. Before analysis begins the
number of peaks required is specified. In the North Sea two
peaks, one windsea and one swell, is usually sufficient to
capture the dominant conditions. However the same
methodology may be applied to more complex systems where
additional swell peaks are expected. The spectral matrix is
analysed using the steepest ascent method to find the local
energy maxima. This is achieved by comparing the spectral
energy at each point in the matrix, to each adjacent point and
assessing which energy belongs to which local peak. If the
number of required peaks has been correctly estimated then the
number of local peaks should be equivalent to the number of
required peaks. In some cases there may be additional local
peaks but these are likely to contain small amounts of energy.
All energy belonging to that peak will be reassigned to a more
dominant peak or discarded according to whether or not it
meets certain criteria detailed in the following sections.
2.2.2 Identify and Combine Wind Sea Peaks
After the peaks have been identified they are assigned as
either a windsea or swell component. Concurrent wind speed
and direction data are required to assess whether each peak
meets the following criterion:
c
p
=
w
U
10
cos (1)
where c
p
is the phase speed of the spectral
component.
w
is a factor applied to the wind speed to
ensure all possible wind peaks are included.
U
10
is the wind speed at 10m elevation.
?is the angle between the wind and the wind
sea (0 = ?=???/2)
Or in deep water: f
p
= g/2 . [
w
U
10
cos]
1
(2)
where f
p
is the peak wave frequency
The factor
w
is variable although is usually specified between 1
and 1.5. A typical value is 1.33.
Any partition whose peak falls within the parabolic
boundary defined in the criterion is considered to be forced by
the local wind, and all peaks within this region are combined
and defined as windsea.
2.2.3 Combine Swell Peaks
All peaks not assigned as windsea are considered to be
swell. If more than one peak is assigned to swell then the peaks
are analysed to assess whether or not they are part of the same
swell system. If either of the following criteria are met, then the
peaks are combined. The factors ? and v are variable userinput
thresholds.
Peak separation: Two peaks are combined if the spread
of either peak satisfies:
f
2
= .f
2
(3)
where f
2
is the distance between the peaks
f
2
is the spread of each individual peak
? ? ? ?is the spread function
Minimum height between peaks: Two peaks are
combined if the lowest point between them is greater
than the peak minimum factor, , times the smaller of
the two peaks:
s > . m (4)
3 Copyright 2008 by ASME
2.2.4 Energy Threshold Check
Low energy peaks are negligible in comparison to the
total energy contained within the dominant peaks and these
partitions are therefore removed. When working with measured
data this step also serves to remove spectral noise belo w a
certain energy threshold.
The threshold is calculated from the peak frequency f
m
and two userdefined parameters a and b:
e = a / (b + f
m
4
) (5)
2.2.5 Evaluation of Partitioning Method
The Hansen and Phillips partitioning method was
compared with wave spectra measured off the west coast of
New Zealand by Ewans et al [8]. The comparison showed that
in general the method performed well, although in a few cases
small low frequency swells were missed because a maximum of
two partitions had been specified. Ewans et al also present
results using a bimodal spectral fitting method proposed by
Guedes Soares [9], which gave a good fit to the measured
spectra. However, this method only partitions the data in the
frequency domain, and so cannot separate sea states coming
from different directions if they have similar peak frequencies.
2.3 FITTING PROCEDURE
2.3.1 Calculation of JONSWAP or Pierson
Moskowitz Spectra
The partitioning methodology allows identification of the
dominant peaks within the wave spectra for each time step. In
order to use this information, omni directional spectra for each
peak must be calculated by summing the amount of energy in
each partition. Then each omnidirectional spectra is
approximated using the JONSWAP spectrum, or the Pierson
Moskowitz spectrum if the calculated peak enhancement factor
is less than or equal to 1. In the North Sea case this will
result in one or two spectra for each time step according to the
number of peaks identified during the partitioning process.
The aim of the fitting procedure is to obtain values of the
JONSWAP parameters alpha ()? ? ?, gamma (), sigmaa(
a
) and
sigmab(
b
) for each peak. The enhancement widths may be
fitted along with alpha and gamma or may be fixed at the
accepted values 0.07 and 0.09.
Approximation to the JONSWAP spectra is achieved
using the Gunther method. A parabola is fitted to the highest
spectral density estimate and one point either side to identify
the peak frequency. Alpha is calculated by assuming that the
spectrum in the range 1.35f
m
to 2.00f
m
can be approximated by
the PiersonMoscowitz spectrum. Given this information it is
possible to use a least squares fitting procedure t o find the
remaining parameters, gamma, sigma a and sigmab or just
gamma ,?if?sigmaa and sigmab are fixed values. This involves
the input of initial values for the parameters and iterating until
the closest match to the function is found.
An example result for one time step is presented in Fig. 4.
The dark and light blue areas illustrate how the spectrum is
partitioned across the range of frequency and direction. The
bottom left hand plot shows the omnidirectional spectrum for
each of the two peaks, the associated Hs values and whether the
peak has been assigned as windsea or swell. The bottom right
hand plot shows the full omnidirectional spectrum of the total
seastate and how the combined fitted spectrum compares to the
original omni directional spectrum before partitioning. In this
case a good fit has been achieved and the total energy within
the sea state is conserved.
2.3.2 Large Gamma Cases
In cases where two spectra are fitted and one of these is a
small swell with a low significant wave height, this spectrum
will often have a high gamma. As the time series of parameters
is prepared for entry to engineering software it is sometimes
necessary to enforce a sensible range of values. Typically
gamma will be limited to a value of 10. However, limiting
gamma will reduce the total wave energy represented by the
calculated spectral parameters unless some adjustment is made
to conserve the energy in the sea state.
If significant wave height values are calculated from a
peak with reduced gamma then they would be reduced and in
turn extreme wave height values derived from the output time
series would be underestimated. Therefore a routine has been
developed in order to allow limitation of gamma values without
loss of energy from the sea state.
This method is undertaken for each time step where one
of the peaks has a gamma value greater than 10 and has been
refitted using a gamma equivalent to 10. Alpha must be
adjusted in order to compensate for the reduction in energy
caused by setting a limit on gamma. The key considerations are
the accuracy of the significant wave height and zerocrossing
period for the entire sea state. These may be assessed by
comparing the original time series data for the sea state as a
whole, to the new values calculated from the fitted omni
directional spectrum. Adjustments are made to the alpha value
of one of the peaks for that time step so that the correct
significant wave height for the total wave energy is obtained
and the total mean zero crossing period value is as close as
possible to the required value.
4 Copyright 2008 by ASME
Figure 4. Partitioning and fitting the spectrum.
2.3.3 Quality Control
The rigorous fitting procedure is the starting point for a
high level of quality control within the output time series of
spectral parameters. In addition for each spectral fit, a
normalised rms error and a bias are calculated so that the
goodness of fit may be assessed and unacceptable fits may be
removed. Each fit is identified in terms of the spectrum used;
JONSWAP or Pierson Moskowitz, and whether it has been
designated wind sea or swell.
The rms error and bias are available for each peak and if
the values fall outside of an acceptable defined range then
individual sea states are flagged or discarded. High error
values are most commonly found in the smallest sea states
where it is a challenge to fit spectra using standard methods.
Error values will also be higher if sigma values are fixed
rather than found using the least squares method. If treated as
free parameters then sigma values are limited a maximum of 1
and 2 respectively.
2.4 DATA FORMAT
Data are presented in time series format for each of the
required parameters. The wave parameters are provided in two
components, or more depending on the number of peaks
specified. Component one contains the parameters defining
the first spectral peak and component two the parameters
defining the second peak. Presentation of results is such that
component one corresponds to high frequency spectral peaks
and component two corresponds to low frequency spectral
peaks. For the majority of records component one will be the
windsea peak and component two will be the swell peak. In
periods of lesser wind influence there may be no windsea
peak. In this case the spectrum may contain one or two swell
peaks. In the case of two swell peaks, component one will
contain the higher frequency of the two peaks and component
two, the lower.
Further quality control must be applied to maintain
continuity of spectral peaks. This may be achieved by
applying an algorithm to track each peak as it grows and
decays through time and to ensure continuity in the
presentation of the parameters in the two components.
Table 1 lists the parameters that are available from the
analysis. Sigma values for each peak may also be provided if
they are not specified as fixed.
The procedure described in this section may be applied
to any site using any number of required peaks. For example
in areas exposed to swell waves from more than one direction
at the same time, more than one swell peak would be required
to accurately describe conditions. The approach may then be
extended to track individual swells to their original source.
5 Copyright 2008 by ASME
Column Heading Definition
Time Time stamp (GMT)
Hs 1
Significant Wave Height (m) for low
frequency spectral peak
Tp 1
Peak Wave Period (s) for low frequency
spectral peak
Alpha 1
JONSWAP Equilibrium Range Constant
for low frequency spectral peak
Gamma 1
JONSWAP Peak Enhancement Factor for
low frequency spectral peak
WavDir 1
Mean Wave Direction (T, from) for low
frequency spectral peak
Hs 2
Significant Wave Height (m) for high
frequency spectral peak
Tp 2
Peak Wave Period (s) for high frequency
spectral peak
Alpha 2
JONSWAP Equilibrium Range Constant
for high frequency spectral peak
Gamma 2
JONSWAP Peak Enhancement Factor for
high frequency spectral peak
WavDir 2
Mean Wave Direction (T, from) for high
frequency spectral peak
WindSpd
10minute Mean Wind Speed at 10m
above sea level (m/s)
WindDir
10minute Mean Wind Direction at 10m
above sea level (T, from)
Table 1. Parameters in the hindcast data set.
3. TRITON FPSO
3.1 METOCEAN REGIME
Directional spectral wave data, partitioned into windsea
and swell components and approximated to omnidirectional
spectra, allow the comprehensive study of vessel response in
the time domain. It is important that each wave energy source
is identified and represented analytically so that all cases are
included within calculation of vessel motion and mooring
response. If omnidirectional or single peak spectra alone are
used then the resulting responses could be over conservative.
Conversely, important design cases could be missed. Fitted
spectral parameters alpha and gamma (and optionally sigmaa
and sigmab) are available for each time step providing a
much more accurate description of the sea state than if a
typical or average value of gamma were used.
The North Sea is a windsea dominated environment.
The strongest winds typically come from the sectors between
NorthWest and SouthWest and conditions are dominated by
the passage of anticyclonic systems often resulting in strong
storms. However, the North Sea also experiences longer
period swell waves especially in the more exposed Northern
region. In the central and southern North Sea the fetch is more
limited in most directions and significant wave heights are
lower, although storm waves are steeper with short periods.
The Triton FPSO is located in the central area and as such
experiences both wind sea and swell effects, although
conditions are dominated by the wind. Therefore within the
hindcast data set, the sea state for each time step is represented
by up to two peaks. Figure 6 plots all wind sea and swell
peaks against the peak wave direction ( from). As expected,
the data show that the highest waves were generated by the
windsea peak. The windseas approached from all directions
with a fairly consistent magnitude apart from the NorthEast
sector. The highest significant wave heights are generated by
westerly winds. The swells wave heights were much lower,
with the largest swells coming from the north.
Figure 6: Significant wave heights with direction for 2
peaks (each dot represents a sea state).
For the Triton analysis, 10year hindcast data were
provided containing wave spectral parameters for up to two
peaks. In addition, a concurrent time series of wind speed,
wind direction, current speed and current direction were
provided, all of which may impact the heading of the vessel.
The following analyses, similar to a previous study
undertaken by Bureau Veritas [10], were undertaken for the
Triton FPSO using the hindcast data:
Calculation of vessel heading, and hence the heading
of swell and windsea relative to the vessel.
Mooring analyses including calculation of fatigue.
The 100 year response of wave bending moments,
wave shear forces, accelerations and relative wave
elevation using a long term approach.
Structural analyses.
Comparison with insitu measurements.
3.2 HEADING ANALYSIS
Heading analyses were completed using the Bureau
Veritas ARIANE3Dynamics software package.
The availability of the time series of partitioned windsea and
swell, with associated wind and current data allowed vessel
heading to be calculated for each sea state at each time step.
Further inputs required for this analysis are first order motion
6 Copyright 2008 by ASME
response amplitude operators (RAOs), quadratic transfer
functions, wind and current vessel response characteristics
(obtained from wind tunnel tests) and total mooring pattern
description in the model.
Results of the heading analysis are presented in Figure 7
below. Each dot represents one threehour seastate. The study
showed that the vessel heading is generally governed by the
windsea and by the wind directions when intensities are
important. However, the Triton FPSO can face windsea with
Hs equivalent to 4 metres from all directions. There is a weak
correlation between vessel heading and swell directions. High
intensities of these parameters can occur from almost all
incidences.
3.3 HYDRODYNAMIC ANALYSIS
Once the vessel heading, and hence the relative headings
of each environmental force, are known the hydrodynamic
responses of the vessel can be computed. The computation
used the spectral parameters fitted to the partitioned wave data
at each time step. The RAOs for the vessel at the relevant
headings were combined with the wave spectra to compute the
100year vessel response at threehourly intervals. Figure 8
shows the RAO for the vertical wave bending moment
amidships compared to a typical energy density spectrum for
the site.
The parameters of interest for structural assessment are
wave bending moments, wave shear forces, accelerations and
relative wave elevation at any point on the structure. The long
term distribution of each parameter was calculated based on
the whole hindcast database. Figure 9 represents the longterm
distribution of the vertical wave bending moment amidships
and the 100year extremes. The same approach was repeated
along the vessel in order to obtain the 100year response (Fig.
10). Checks were then made to compare the actual stress
levels against the design values in order to monitor vessel
integrity. In addition, knowledge of the actual load ranges
experienced by the hull enabled fatigue calculations to be
carried out with a higher accuracy.
The final aim of the analysis was to ensure vessel
integrity for the future. This included consideration of the
possible evolution of the weight of the vessel, for example the
addition of topsides, and the extension of the operating life of
the vessel. To meet this objective, calculations needed to be as
refined as possible to allow an accurate description of the unit
to be drawnup. The availability of hindcast data enabled the
conditions that the vessel had experienced over the past 10
years to be reproduced and analysed. It also allowed a good
estimate to be made of what the vessel will have to face in the
future. Practically, the results from the Bureau Veritas study
have yielded a plan for inspections of both the structural
elements of the FPSO, and also the mooring lines, prioritizing
areas where high stresses are expected. Further work involves
long term insitu measurement aboard the vessel which may
be crosschecked with the results obtained using metocean
data. This will allow the actual vessel responses to be
predicted and ensure that the model and the vessel are in
accordance.
Figure 7 (left). Relative headings between each
element incidence and the vessel heading. 180 means that
the element is going towards the vessel bow (head waves).
7 Copyright 2008 by ASME
0.0E+00
5.0E+07
1.0E+08
1.5E+08
2.0E+08
2.5E+08
3.0E+08
3.5E+08
4.0E+08
4.5E+08
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Wave frequency (rad/s)
VWBM (N.m/m)
0.00
0.04
0.08
0.12
0.16
0.20
0.24
0.28
0.32
0.36
S(w) (m/s)
VWBM RAO amidships 
head seas
Energy density spectrum
Hs=1m, Tp=12s, gamma=2
Figure 8.Vertical wave bending moment RAO
Longterm distribution of vertical wave bending moment amidships
0.0E+00
5.0E+08
1.0E+09
1.5E+09
2.0E+09
2.5E+09
3.0E+09
3.5E+09
4.0E+09
1.E01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08 1.E+09
Number of cycles
N.m
LT distribution  full LT distribution  intermediate LT distribution  ballast
100Y  full 100Y  intermediate 100Y  ballast
Figure 9. Longterm distribution of vertical wave bending moment amidships for different drafts and 100year extremes.
Vertical wave bending moment along the vessel
0.0E+00
5.0E+05
1.0E+06
1.5E+06
2.0E+06
2.5E+06
3.0E+06
3.5E+06
4.0E+06
0 25 50 75 100 125 150 175 200 225 250
x from aft perpendicular (m)
kN.m
Full Intermediate Ballast
Figure 10. 100year longterm response of vertical wave bending moment along the vessel for different drafts.
8 Copyright 2008 by ASME
4. CONCLUSION
A time series of wave spectral output from the ECMWF
WAM model was obtained and passed through a bespoke
calibration procedure using various satellite altimeter data
sets. Each individual spectrum from the calibrated time series
was analysed using partitioning software. The spectra were
partitioned into 1 or 2 peaks using a steepest ascent algorithm
and then fitted using the JONSWAP or PiersonMoscowitz
spectrum to give time step specific values of the defining
spectral parameters alpha and gamma. This in addition to the
usual parameters Hs, Tp and peak wave direction. Quality
control checks were applied to the parameters and the data
were presented so that parameters associated with a windsea
peak were contained within the first component of the
presentation and parameters associated with the swell peak
were contained in the second.
The data were then applied to the case of the Triton
FPSO, located in the central North Sea. The partitioned data
showed that the highest waves were caused by windsea
dominated seastates and that these may originate from a range
of direction sectors. Swell waves were highest when
originating from the North sector, as a result of the fetch
length. Heading analyses were completed for each time step
and showed that the heading was governed by the windsea
and by the wind. Swell waves and current had little effect on
the vessels heading. Headings were then combined with the
wave spectra and motion RAOs of the vessel to assess vessel
response, specifically for the 100year case. Stress levels were
compared with design values. These analyses have resulted in
a comprehensive view of vessel integrity providing an
inspection plan for monitoring the effects of the metocean
conditions on the unit.
ACKNOWLEDGMENTS
The authors would like to thank their respective
companies and in particular Hess Limited, for permission to
publish the paper. The views expressed are those of the
authors and do not necessarily reflect those of their
companies.
Thanks are due to Gunnar Mrk and Stephen Barstow at
Fugro OCEANOR AS, for advising on and undertaking the
site specific calibration of wave model data with available
satellite data for the Triton site.
REFERENCES
[1] Bidlot, J.R., Janssen, P.A.E.M., and Abdalla, S., 2005.
On the Importance of Spectral Wave Observations in the
Continued Development of Global Wave Models , Paper 207,
Ocean Wave Measurements and Analysis, Fifth International
Symposium WAVES 2005, Madrid, Spain.
[2] Barstow, S.F., Mrk, G., Lnseth, L., Mathisen, J.P.,
and Schjlberg, P., 2005. The Role of Satellite Wave Data in
the Worldwaves Project, Paper 206, Ocean Wave
Measurements and Analysis, Fifth International Symposium
WAVES 2005, Madrid, Spain.
[3] Hanson, J.L., and Phillips, O.M., 2001. Automated
Analysis of Ocean Surface Directional Wave Spectra, J.
Atmos. Oceanic Technol., 18, 277293.
[4] Aarnes, J.E., and Krogstad, H.E., 2001. Partitioning
Sequences for the Dissection of Directional Ocean Wave
Spectra: A Review, Part of work package 4 (Wp4) of the
EnviWave (EVG200100017) research programme under the
EU Energy, Environment and Sustainable Development
programme.
[5] Hasselmann, S., Hasselmann, K., and Bruning, C.,
1994. Extraction of Wave Spectra from SAR Image Spectra,
Dynamics and Modelling of Ocean Waves, G. Komen (ed.),
Cambridge University Press, Cambridge, UK, 391401.
[6] Gerling, T.W., 1992. Partitioning Sequences and
Arrays of Directional Wave Spectra into Component Wave
Systems , J. Atmos. Oceanic Technol., 9, 444458.
[7] Feld, G. , and Mrk, G. , 2004. A Comparison of
Hindcast and Measured Wave Spectra, The 8th International
Workshop on Wave Hindcasting and Forecasting, Hawaii.
[8] Ewans, K. C., BitnerGregersen, E. M. and Guedes
Soares, C., 2006. Estimation of WindSea and Swell
Components in a Bimodal Sea State. Journal of Offshore
Mechanics and Arctic Engineering., 128, 265270.
[9] Guedes Soares, C., 1984. Representation of Double
Peaked Sea Wave Spectra. Ocean Eng. 11(2) 185207.
[10] Francois, M., Morandini, C., Gorf, P. and Lewin., M,
2004. Relative Wave Heading of an FPSO in a Harsh
Environment, OMAEFPSO040043, Proc. OMAEFPSO
2004,Houston,USA.
9 Copyright 2008 by ASME
OTC 19315
Deepwater Moorings with High Stiffness Polyester and PEN Fiber Ropes
Davies P /IFREMER Brest, Lechat C, Bunsell A, Piant A /ENSMP, Franois M /BV, Grosjean F / IFP,
Baron P/Doris, Salomon K/Saipem, Bideaud C /Technip, Labb J P /Acergy, Moysan AG /Total.
Copyright 2008, Offshore Technology Conference
This paper was prepared for presentation at the 2008 Offshore Technology Conference held in Houston, Texas, U.S.A., 58May2008.
This paper was selected for presentation by an OTC program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Offshore Technology Conference and are subject to correction by the author(s). The material does not necessarily reflect any position of the Offshore Technology Conference, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Offshore Technology Conference is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of OTC copyright.
Abstract
The benefits of synthetic fiber ropes for deepwater station keeping are now well established and their use is expanding.
Nearly all current applications use a single grade of polyester fiber, but for different supports and environments this may not
be the optimal choice. Properties of polyester fibers can be modified by adjusting processing parameters and there are other
fibers available such as PEN, which offer higher stiffness. This study examines the benefits of intermediate stiffness fibres,
stiffer than standard polyester but less stiff than the high performance fibers. The results indicate that there is scope for
improving mooring line performance and reducing line weight by careful evaluation of material options.
Introduction
Polyester fiber ropes are finding increasing applications in offshore mooring systems as production moves to deeper
water. Following successful installations offshore Brazil in the late 1990s [Pellegrin 1999] the first Gulf of Mexico mooring
was for the Mad Dog spar [Bugg 2004] in 2004, which employed 1200 tons of polyester down to 1670 meters water depth.
The recently installed Independence Hub platform also used polyester moorings, in 2440 meters water depth [Paganie 2007].
Different rope constructions have been used but these mooring lines were all composed of similar high tenacity polyester
fibers. The Red Hawk spar [Haslum 2005], also installed in 2004, used a modified polyester fiber with a higher initial
stiffness to facilitate installation, and this raised the question of whether a higher fiber stiffness might be beneficial for other
supports and allow rope diameter to be reduced. Previous work within the French Mooring line project [Davies et al 2002]
studied high performance fibres such as aramids and HMPE and concluded that their very high stiffness, while allowing
much smaller rope diameter and weight, did not improve durability as it would result in high fatigue loading of the metallic
components of the mooring line. However, there is an intermediate stiffness region, shown in Figure 1, situated between the
currently used fibres, with initial tensile modulus around 1015 GPa, and the high performance fibers (>60 GPa) which has
not been explored previously for deepwater mooring applications.
Fiber behavior
0
0,2
0,4
0,6
0,8
1
1,2
1,4
1,6
1,8
2
0 5 10 15
Elongation, %
T
n
a
c
i
t
y
,
N
/
t
e
x
PEN
HMPE, Aramids, Vectran,
Polyester (PET)
INTERMEDIATE
STIFFNESS
RANGE
Figure 1. Available fiber properties and unexplored intermediate stiffness region
2 OTC 19315
The aim of the present study is to examine under what conditions intermediate stiffness fibers would be attractive for
deepwater moorings. In order to do this it was first necessary to examine how the properties of these intermediate stiffness
fibers transfer to ropes. Samples were then manufactured in order to characterize the ropes under appropriate loading
conditions, as few stiffness data were available when the study started. Based on these results mooring line analyses were
performed for different supports and environmental conditions to determine the influence of the improved rope stiffness.
Finally, a series of long term characterizations including creep and cyclic loading was performed, to check whether increased
stiffness results in deterioration of long term properties.
Materials and samples
Three fibers have been studied, a standard polyester (PET, polyethylene terephthalate) fiber, widely used today for offshore
tethers, an improved polyester, and PEN (polyethylene naphthalate), chosen to represent the intermediate stiffness range
(fibre modulus around 30 GPa). The difference between the PET and PEN polymers is the presence of a double aromatic ring
in the PEN molecule, Figure 2. Figure 3 shows the tensile response of new yarns, the PEN shows an initial modulus roughly
twice that of the standard polyester.
Ester group
Figure 2. Structures of pol yester PET (pol yethylene terephthalate) and PEN (polyethylene naphthalate) molecules
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0 2 4 6 8 10 12 14
% strain
N
/
t
e
x
Modified
polyester
Standard
polyester
PEN
Figure 3. Tensile behavior of standard and modified PET, and PEN, tests on new yarns.
Stiffness data for the standard fiber ropes are available from previous work e.g. [Franois 2000], so the present study
concentrated on testing the two improved stiffness variants. Table 1 shows the samples studied.
Scale Dimensions
diameter/test length
Nominal break load
Single Filament 2030m / 50mm 0.5 N
Yarn 110 tex 0.4 mm / 1 meter 90 N
Assembled yarn 4 mm / 1meter 5 kN
Subrope 32 mm / 8 meters 330 kN
Rope 132 mm / 12 meters 5000 kN
Table 1. Sample sizes tested
All materials were supplied by Performance Fibers, Longlaville, France, with marine finish. Rope yarns and 3strand
twisted subropes were produced by Bexco, Belgium with the same construction for both fibers. Full size modified polyester
ropes, an Integra parallel 7strand construction, were manfactured by Scanrope, Norway.
OTC 19315 3
Test procedures
Various test machines were employed to measure the properties over the range from single fiber up to full size rope. The
single filaments were tested on a special machine, Figure 4a [Bunsell 1971]. Yarns and assembled yarns were tested on a 10
kN test frame, Figure 4b. 500 and 1000 kN test frames were used to test subropes. Two digital cameras linked to an image
analysis system were used to measure strains, together with more conventional extensometry. More details of tests can be
found elsewhere [Davies 1999, Lechat 2007]. Some samples were in the new state, others were subjected to a beddingin
sequence of 5 loadunload cycles to 50% break load.
aa
cc
bb
Figure 4. Test machines
a) Single filament testing, b) yarn tests, c) subrope test machine, d) full size rope test.
Single fiber and some assembled yarn tests were performed at the Ecole des Mines de Paris (ENSMP), yarn and
assembled yarn tests were also performed at Ifremer in Brest, tests on over 20 subrope samples were carried out at IFP in
Lyon and at Ifremer, and full size ropes were tested at the LCPC civil engineering laboratory in Nantes.
Fiber to rope transfer
By testing the components of large ropes at the different steps of fabrication it is possible to evaluate the influence of the
fiber properties on the rope behavior. This is important, as if the transfer is poor there may be little to be gained by improving
fiber properties. Figure 5 shows mean results from tests at each scale, for the improved PET and PEN fibers.
a) Modi fi ed PET
0
0,2
0,4
0,6
0,8
1
0 3 6 9 12 15
Strain (%)
F
o
r
c
e
(
N
/
t
e
x
)
fi l ament
yarn
ropeyarn
subrope
b) PEN
0
0,2
0,4
0,6
0,8
1
0 2 4 6 8 10
Strai n (%)
F
o
r
c
e
(
N
/
t
e
x
)
fi l ament
yarn
ropeyarn
subrope
Figure 5. Stressstrain results, tension loading to failure after beddingin. (Note different strain scales).
4 OTC 19315
Figure 5 shows the similarities and differences between the two materials. The failure strains are significantly lower for
PEN fibers but failure stresses are similar. There is a drop in stiffness of subropes compared to rope yarns but the stiffness
transfer coefficients are similar for both materials. The behavior of the PEN fiber is more linear than that of PET after
beddingin. A permanent strain after beddingin of around 1% is noted for both materials, significantly lower than that
measured on standard polyester fibers and ropes. The similar permanent strains for the three levels of construction indicates
that this strain is mainly due to orientation of the polymer structure rather than alignment of the rope construction. Based on
low level material properties and geometrical parameters rope mechanics equations can be used to predict the quasistatic
behavior of subropes quite accurately. This will be reported elsewhere.
Stiffness properties
Mooring line analysis requires stiffness values measured under appropriate loading conditions. These are available for
standard polyester but not for modified PET nor PEN, so it was necessary to perform a complete stiffness characterization for
these materials. Tests were performed on subropes to measure these values under quasistatic and dynamic loading as
specified in the recent ISO document [ISO 2007]. All samples were bedded in by 5 cycles to 50% break load followed by a
one hour hold period before starting stiffness measurements. Figure 6 shows examples of both types of stiffness test for PEN.
(a) last 10 of 60 QS cycles
0
20
40
60
80
100
120
2 2,2 2,4 2,6 2,8 3 3,2 3,4
Strain, %
F
o
r
c
e
,
k
N
(b) Dynamic stiffness, last 10 of 500 cycles
160
165
170
175
180
185
190
4,1 4,15 4,2 4,25 4,3
Strain, %
F
o
r
c
e
,
k
N
Figure 6. Stiffness tests, a) Quasistatic, b) Dynamic high load , PEN subropes
Tests were also performed on a full size modified PET rope to examine whether these subrope values correspond to those
of full size ropes. Table 2 presents results, Kr is the normalized load/strain.
Material Quasistatic stiffness Kr
Standard PET [Franois 2000] 15
Modified PET subrope 14
Modified PET fullsize rope 14
PEN subrope 20
Table 2. Example of measured quasistatic stiffness values over the range 1030% break load
Although the stiffness of the modified PET yarn is higher initially, ie before beddingin, as shown in Figure 3, after
beddingin values are very similar to that of the standard PET material. Part of the permanent reorientation strain has been
removed during processing but working properties are very similar to those of the standard material. This is not the case for
the PEN ropes, which are significantly stiffer before and after beddingin, so this material was used subsequently to examine
how stiffness affects mooring line response. However, this example shows the importance of understanding the behavior of
these materials, as the relationship between yarn properties and useful rope stiffness can be complex. An initial factor of two
between standard PET and PEN is reduced to an increase of around 30% for quasistatic rope stiffness and around 15% for
dynamic stiffness at 45% mean load. Dynamic stiffness is strongly dependent on mean load [Fernandes 1999, Davies et al
2002].
In addition to these standard stiffness measurements some additional tests were performed on the full size rope to
examine the influence of stochastic loading. Various sequences were defined using mooring line analyses for different
scenarios and Figure 7 shows one example, for a production platform in the Gulf of Mexico.
OTC 19315 5
(a)
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
0 500 1000 1500 2000 2500 3000 3500 4000
Time, s
F
o
r
c
e
,
k
N
(b)
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
4,8 4,85 4,9 4,95 5 5,05 5,1 5,15 5,2 5,25 5,3
Strain, %
F
o
r
c
e
,
k
N
Figure 7. Stochastic loading of 500 ton modified PET rope, mean load 200 tons. a) Sequence applied, b) Measured response.
These results are discussed in more detail elsewhere [Franois, 2008] but suggest that in a stochastic sequence, the stiffness
does not depend on the amplitude of individual cycles, nor on the period of underlying (low and wave frequency)
components.
Mooring line analyses
Once the necessary property data had been obtained a number of mooring line analyses were performed to examine how the
increased stiffnesses of a rope such as PEN would affect mooring line dimensions. Different supports were considered by the
engineering companies participating in the project, for a depth of 2500 meters in two locations, West Africa and the Gulf of
Mexico. The Ariane software was used, together with fatigue analyses.
Material Quasistatic Dynamic
(function of mean load (ML))
Standard PET 15 18.5 +0.33 ML
PEN 20 23 +0.33 ML
Table 3. Values of Kr used in first mooring line anal yses.
Then in a second series of analyses, designed to look for the optimal rope properties, parametric studies of the influence
of stiffness were performed. These studies generated a large amount of data. Three examples of of results are shown below,
for a semisubmersible drilling rig and a production platform, both in the Gulf of Mexico, and for a production barge off
West Africa. For analysis, two parameters must be specified, the safety factor T
max
/Break load, and the minimum allowable
tension T
min
. The choice of values to be taken for a new material clearly has an impact on whether a particular solution is
interesting compared to a standard polyester choice. For this test case higher safety factors were taken and a 5% minimum
tension was used for PEN, rather than the 2% now usually applied for polyester, taking into account the lack of experience
with this product and the uncertainties on loadelongation properties. For the semisubmersible drilling rig with 8 lines and
and production rig (a 12 lines system), loop current and hurricane conditions were applied. Tables 4 and 5 show examples of
results.
Material
Criteria:
SF Dyn Intact
SF Dyn Damaged
T
min
applied
Sizing:
Diameter
Break load
Polyester
1.83
1.375
2%
130 mm
542 T
PEN
2
1.5
5%
144mm
601 T
Table 4. Case study 1, Semisubmersible production platform, GOM 2500 meters depth.
6 OTC 19315
In the first example, the improved behavior of the PEN is clearly occulted by more conservative safety factors and, above
all, by the criterion of minimum tension that sets the level of pretension in the mooring lines, the level of maximum tensions,
hence the size.
Material
Criteria:
SF Dyn Intact
SF Dyn Damaged
T
min
applied
Sizing:
Diameter
Break load
Polyester
1.83
1.375
2%
246 mm
1778 T
PEN
2
1.5
5%
246 mm
1778 T
Table 5. Case study 2, Semisubmersible production platform, GOM 2500 meters depth.
It is clear for both examples that with these model parameters there is little to be gained by increasing the line stiffness.
The third example involves a large production barge moored using 16 lines in 2500 meters depth off West Africa.
Here the environment is much less severe and the offset criterion is governing the design. Table 6 shows an example of
results from this case. The minimum tension criterion was also varied : Figure 8 shows how this affects the rope diameter
and weight gain.
Material
Criteria:
SF Dyn Intact
SF Dyn Damaged
T
min
applied
Sizing:
Diameter
Break load
Prtension
Polyester
1.83
1.375
2%
168mm
782 T
150 T
PEN
2
1.5
5% / 2%
152 / 144mm
648 / 590 T
160 / 120 T
Table 6. Case study 3, Production barge, West Africa 2500 meters depth.
a) Rope diameter, mm
130
135
140
145
150
155
160
165
170
PET 2% PEN 5% PEN 2%
b) Rope weight, tons
0
200
400
600
800
1000
1200
PET 2% PEN 5% PEN 2%
16%
23%
Figure 8. Influence of minimum tension criterion on PEN rope diameter and weight gain.
These results indicate that the use of a higher stiffness material can lead to significant savings, over 200 tons of material
here. This reduction may also result in associated savings on account of easier handling of a smaller diameter rope and
reduced storage requirements during installation. A further gain is to be found in the line pretensions, these are somewhat
lower in the PEN line and may allow the use of a smaller tensionning system. It should also be noted that these are all
OTC 19315 7
conservative estimates, based on higher safety factors for the higher stiffness material. In order to justify lowering these
factors it would be necessary to provide evidence that the long term behavior of the higher stiffness material is at least as
good as that of the currentlyused polyester ropes.
Long term behavior
Two aspects of the long term behavior of the two materials were considered in this study. This was not intended to
be an exhaustive study, clearly further qualification work would be essential if a PEN rope was to be considered for this type
of application, but simply to give a first indication of differences compared to the currentlyused fiber.
First, creep tests were performed. Several types of test were run; first, constant loads were applied to single
filaments and rope yarns for short periods, then subropes were subjected to 3day creeprecovery cycles, and finally longer
creep tests (hundreds of hours) were performed to measure creep rates. Figure 9 shows an example of results for filaments,
which indicate no significant difference in creep rates for the two fibers. This suggests that the intrinsic creep mechanisms are
similar at a material level.
Creep of filaments
0,0
0,1
0,2
0,3
0,4
0,5
0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7
Creep load, N/tex
C
r
e
e
p
r
a
t
e
,
%
/
d
e
c
a
d
e
PET
PEN
Figure 9. Creep rates from tests on single filaments.
Then subropes were subjected to short term creep/recovery cycles at different load levels, Figure 10, designed to produce the
parameters for a creep model developed previously (Davies 2003, Chailleux 2005). This allowed a direct comparison of the
strains to be expected for load levels of 10, 30 and 50% break load to be determined. The creep rates are again similar for the
two materials, but more importantly the overall strains and the permanent strain after unloading of the PEN ropes are much
smaller than those measured on the modified polyester. As the modified polyester has already a significantly lower
permanent strain than the standard polyester this indicates that the PEN rope may offer further significant advantages during
installation in crowded areas. Finally some longer creep tests were performed, up to 50 days, to check the creep response over
a longer period, Figure 11.
a) 72h Creeprecovery cycle
0
20
40
60
80
100
120
140
160
0 10 20 30 40 50 60 70 80
Time, h
L
o
a
d
,
k
N
b) Subrope response
0
1
2
3
4
5
6
7
0 20 40 60 80
Ti me, h
S
t
r
a
i
n
,
%
PEN
PET mod.
Figure 10. Creep/recovery cycle results for new subropes of modified PET and PEN
8 OTC 19315
a) PEN creep 18 days, 40% break load
0
1
2
3
0 2 4 6 8
log(t (s))
S
t
r
a
i
n
(
%
)
(b) strain =mx log(time, h)
0
0,02
0,04
0,06
0,08
0,1
0,12
0,14
0 20 40 60
Creep load, % break load
C
r
e
e
p
r
a
t
e
,
m
PET
PEN
Figure 11. Long term creep on subropes, a) example of plots for 40% break load, b) creep rates versus applied load
The creep of PEN is linear on a semilog plot, Figure 11a, and creep rates are plotted on Figure 11b. It is apparent from these
results and other data generated during the project that the creep rates are very similar for the two materials, and similar to
previouslypublished results for standard PET [Del Vecchio 1992, Davies 2000, Grosjean 2005 ].
The other property of interest for mooring line applications is fatigue life. It has now been clearly demonstrated by extensive
testing that the fatigue life of standard polyester tethers is more than satisfactory [Banfield 2005]. The aim of the present
study was not to generate fatigue data, but in order to obtain a first indication of whether fatigue behavior of PEN is
significantly different to that of PET fibers two types of test were performed, tensile fatigue on single filaments and yarnon
yarn abrasion. First, single filament tensile fatigue tests were performed in air at ENSMP. Figure 12 shows an example of the
results, which indicate that the intrinsic tensile fatigue behavior of the PEN is very similar to that of the polyester.
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
1,E+02 1,E+03 1,E+04 1,E+05 1,E+06 1,E+07
Lifetime (cycles)
M
a
x
i
m
u
m
f
o
r
c
e
(
N
/
t
e
x
)
Fi l aments PET
Fi l aments PEN
Figure 12. Tensile fatigue SN curves, zero minimum load, various maximum loads, 50 Hz.
Then yarnonyarn abrasion tests were performed in natural sea water at Ifremer according to the standard test procedure
[Cordage Institute 2001]. Figure 13 shows results for the two materials compared to the response of a standard yarn. The
modified PET behaves in a very similar way to the standard material, the PEN lifetimes are a little lower and close to the
recommended values [ISO, 2007]. However, it should be emphasized that whereas the marine finish on the standard
polyesters has been developed over many years, the finish for the PEN fibers was not specifically developed for this fiber.
Improved yarnonyarn abrasion lifetimes could probably be obtained for the latter if further development was carried out.
OTC 19315 9
YarnonYarn abrasion tests, 110 tex
in natural seawater
0
1
2
3
4
5
6
0 10 20 30 40 50 60
Applied load, mN/tex
L
o
g
c
y
c
l
e
s
t
o
f
a
i
l
u
r
e
PEN
Standard PET
Modified PET
ISO 18692
Figure 13. Yarn on yarn abrasion results
Qualification of higher stiffness polyester fiber ropes for mooring lines
These results suggest that the PEN material behavior is similar to that of standard PET fibers, so that additional safety
margins, as were considered in the above analyses, could be removed, subject to adequate qualification of fibre and rope. The
modified polyester fiber ropes studied here were tested and installed on the Red Hawk cell spar [Haslum 2005]. A standard
API/ABS test program was performed including full scale break tests and a range of tests on scaled (1:4) ropes including
80000 cycles over a range of 1545% break load. Those provide confidence in the modified polyester, but, as was shown
above, the stiffness properties of the worked rope are very simlar to those of the standard material. For ropes with a new fibre
such as PEN or other improved polyester fibre (see below), the process of qualification [Bureau Veritas, 2007] shall focus,
besides standard tests, on the two following aspects, as discussed in [Franois 2005]:
 careful identification of the load elongation properties, that could be made for example by tests on subropes, in addition to
standard tests on a full size rope,
 within the qualification of fibre, which is a prerequisite to rope qualification, assessment of inrope properties of the
fibres for long term endurance and capacity to withstand low minimum load under cyclic loading, by tests on small ropes,
and a comparison with results for current polyester grades (see [Banfield 2005]) .
Other material options
In the current work PEN was studied, as this is the only fiber in the intermediate range commercially available for rope
applications at present. It should be emphasized that this fiber was initially developed for tyre applications and there are
potentially a very large range of other PEN fiber properties which could be obtained by appropriate drawing and heat
treatment cycles [Wu, 2000]. A lower cost alternative to PEN might be to produce stiffer polyester fibers by drawing and heat
treatments, and again development work has shown that polyester yarns can also be produced with a wide range of
stiffnesses, up to those of PEN [Parguez, 2004].
While stiffness can be tailored to a particular application the strength of these fibers does not increase with increasing
stiffness. However, ongoing research work in J apan has shown that polyester fiber strength can also be significantly
improved [Kikutani, 2007] and this may result in an even wider range of possibilities for optimization in the future.
10 OTC 19315
Conclusions
While the polyester fiber grade used extensively today for station keeping ropes has proved very satisfactory, it is only
one of a very large family of materials which could be used for this application. In this study the influence of increasing rope
stiffness has been examined, and PEN ropes have been used to illustrate how in some locations where offset is the critical
design parameter a higher stiffness rope can provide improved performance through smaller diameter lines, weight gain and
lower pretensions. When strength is the dimensioning factor higher fiber tenacity is required compared to the standard fiber,
and the currently available intermediate stiffness fibers do not provide increased tenacity.
This paper has concentrated on the technical advantages of intermediate stiffness fiber ropes. For a particular application
the economic benefits of smaller diameter, lower weight ropes, in terms of transport, handling and installation must be offset
against the material cost penalty. The latter will depend on the manufacturing route, and in particular on whether the existing
polyester route can be modified or if a fiber based on a different molecular structure such as PEN is necessary.
Acknowledgements
This work was performed within Phase 5 of the French Mooring line project. The project, which started in 1996 and was
completed in December 2007, involved research institutes (Ifremer, IFP, ENSMP), engineering companies (Acergy, Doris,
Principia, Saipem, Technip), the Bureau Veritas and Total. It was performed within the framework of the CLAROM group,
the French Club for Research Activities on Offshore Structures.
The authors are grateful to Caroline Muller and Ren Soenen of Performance Fibers for supplying material samples. The
expertise of technical staff at Ifremer (N. Lacotte, A. Deuff, B. Forest, D. Choqueuse and L. Riou) at IFP (J. Bonnaves, P.
Hamdoun), at the ENSMP (Y. Favry, C. Teissedre) and at the LCPC (G. LeRoux, B. Philippot) is also gratefully
acknowledged.
The views expressed here are those of the authors, and do not necessarily reflect those of their respective companies.
References
Banfield SJ , Casey NF, Nataraja R, (2005) Durability of polyester deepwater mooring rope, OTC 17510.
Bugg DL, Vickers DT, Dorchak CJ , (2004) Mad Dog project: Regulatory approval process for the new technology of synthetic (polyester)
moorings in the Gulf of Mexico, OTC 16089.
Bunsell AR Hearle J WS, Hunter RD, (1971) An apparatus for fatigue testing of fibres, J nl Physics E, 4, 86872.
Bureau Veritas , (2007) Certification of fibre ropes for Deep water offshore services, NI432R01
Chailleux E, Davies P, (2005) A nonlinear viscoelastic viscoplastic model for the behaviour of polyester fibres, Mechanics of Time
Dependent Materials, 9, pp147160.
Cordage Institute standard, Test method for yarnonyarn abrasion, CI 150300, August 2001.
Davies P, Baizeau R, Grosjean F, Franois M, (1999) Testing of large polyester cables for mooring line applications, Proc ISOPE, Brest,
p360.
Davies P, Huard G, Grosjean F, Francois M, (2000) Creep and Relaxation of Polyester Mooring Lines, OTC 12176
Davies P, Francois M, Grosjean F, Baron P, Salomon K, Trassoudaine D, Synthetic mooring lines down to 3000 meters depth, (2002) OTC
14246.
Davies P, Chailleux E, Francois M, Grosjean F, Bunsell A, (2003), Prediction of the long term behavior of synthetic mooring lines OTC
15379
Del Vecchio CJ M (1992), Light weight materials for deep water moorings, PhD thesis University of Reading
De Pellegrin I, (1999), Manmade fiber ropes in Deepwater Mooring Applications, OTC 10907.
Fernandes AC, Del Vecchio CJ M, Castro GAV, Mechanical Properties of Polyester Mooring Cables, Int. J . Offshore & Polar Eng., 9, 3,
Sept. 1999, pp207213.
Francois M, Davies P (2000) Fibre rope mooring a practical model for the analysis of polyester mooring systems, Rio Oil & Gas, 2000.
Francois M, (2005) Fibre ropes for Stationkeeping : Engineering properties and qualification procedures, OCEANS 2005 MTS/IEEE 
Washington
Francois M, Davies P (2008) Characterization of polyester mooring lines, OMAE2008.
Grosjean F, Davies P, Francois M, (2005) Synthetic Fiber Ropes mooring: technical status and stiffness prediction, Proc CMOO4.
Haslum HA, Tule J , Huntley M, J atar S, (2005) Red Hawk polyester mooring system design and verification, OTC 17247.
ISO 18692 (2007) , Fibre ropes for offshore station keeping Polyester.
Kikutani T, (2007) Private communication
Lechat C, Mechanical behaviour of polyester fibres and fibre assemblies for mooring offshore platforms, PhD thesis (in French), Ecole des
Mines de Paris, 2007.
Lechat C, Bunsell AR, Davies P, Piant A, Mechanical behaviour of polyethylene terephthalate & polyethylene naphthalate fibres under
cyclic loading, J ournal of Materials Science, Vol. 41, 2006, pp17451756
Paganie D, (2007) Independence Hub breaks records through collaboration, innovation, Offshore, volume 67, issue 12
th
December,
Parguez O (2004), Private communication
Wu G, Li Q, Cuculo J A, (2000) Fiber structure and properties of poly(ethylene2,6naphthalate) obtained by highspeed melt spinning,
Polymer, 41, 81398150.
IPTC12546PP
LESSONS LEARNT FROM 12 YEARS OPERATIONS OF A HUGE FLOATING
PRODUCTION UNIT MADE OF PRESTRESSED HIGH PERFORMANCE
CONCRETE
Bertrand LANQUETIN TOTAL S.A., Heidi DENDANI TOTAL S.A., Pascal COLLET TOTAL S.A. and J ose
ESTEVE Bureau Veritas
Copyright 2008, International Petroleum Technology Conference
This paper was prepared for presentation at the International Petroleum Technology Conference held in Kuala Lumpur, Malaysia, 35 December 2008.
This paper was selected for presentation by an IPTC Programme Committee following review of information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the International Petroleum Technology Conference and are subject to correction by the author(s). The material, as presented, does not necessarily
reflect any position of the International Petroleum Technology Conference, its officers, or members. Papers presented at IPTC are subject to publication review by Sponsor Society
Committees of IPTC. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the International Petroleum Technology
Conference is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, IPTC, P.O. Box 833836, Richardson, TX 750833836, U.S.A., fax +19729529435.
Abstract
The last years have seen an impressive increase of Floating Production Storage and Offloading (FPSO) units deployed all
over the world both in number and in variety of designs. But all of them in steel.
This paper presents the lessons learned from the huge floating production unit made of prestressed high performance
concrete in operation for TOTAL E&P Congo. They touch all aspects of the life of the unit, design, construction and
maintenance on site. The paper will also present the main key issues for the floating concrete structures and some guidelines
on how to avoid them, resting on some of the available existing codes.
The unit is in production since 12 years on the NKOSSA oil field in 170 m water depth. Main characteristics: L 220 m, B 46
m, D 16 m, displacement 107,000 t including 73,000 t for the hull and 34,000 t for the topsides. This concrete unit used for
construction 27,000 m3 concrete, 2,350 t prestressed steel and 5,000 t passive steel.
This paper focuses on the following aspects specific to this floating production unit:
 Structural modeling techniques taking into account the nonlinear characteristics of concrete and incorporating a
description of passive and active steel.
 Ageing processes of the concrete units and the general typology of the defects encountered.
 Development of a tailormade inspection program.
Prestress concrete has many virtues and the fabrication process gets more and more industrialized, with better knowledge of
the parameters and how controlling them. This makes it a potential material candidate that can allow newbuilds outside
traditional shipyards, providing new opportunities.
However the paper is not intended to make a recommendation between steel and concrete because this entails many other
considerations that are not only technical; the paper remains on the technical ground and shares valuable knowledge on the
behavior of floating concrete units on the long term prospective of a field life time and on the integrity management
techniques developed to minimize the risk of production shutdown and optimize maintenance and repair costs.
Introduction
It will now be twelve years since the concrete FPU (Floating Production Unit) NKP was installed offshore Congo. During
that time it has undergone one technical stop for process maintenance as scheduled in the design, else has been on
uninterrupted service.
The unit was built in south of France in 199495 and installed one year later in the NKOSSA oil field in 170 m water depth
and some 60 km offshore the Congo Coast. To the authors knowledge it is the biggest floating production concrete unit ever
built and in service. Its main dimensions are 220 m long, 46 m wide and 16 m deep. This concrete barge used in its
construction 27,000 m3 of concrete, 2,350 t of prestressed steel and 5,000 t of passive steel, references [3], [4]. Its
displacement (weight of the equivalent volume of water occupied by the submerged body of the vessel) is 107,000 t. Two
thirds account for the hull and the remaining 34,000 t are the topsides weight.
The production facilities and living quarters for 160 people are fitted on the 10,000 m deck area which, for construction
purposes, is subdivided into six modules: accommodation and central control, utilities, electric power generation, gas
compression for reinjection, crude oil, and gas.
2 IPTC12546PP
Design production is 120,000 b/d of oil sent to the shore terminal and 1,300 metric tons/day of Liquefied Petroleum Gas sent
to an 80,000 cu.m LPG FSO.
The unit is hold in place, 70 meters away of the NKF2 platform, in a spread moored configuration by means of 12 mooring
lines. Figure 1 shows the unit in place with the fixed platform vaguely seen behind.
Figure 1: FPU NKP
Asset Integrity Management
In order to constantly analyze and monitor the condition of the units, a tailormade methodology has been developed and
implemented since 2004 for the Integrity Management of Total Floating Units currently in operation.
The aim of Floating Units Integrity Management is to ensure management and continuous follow up of Floating Units
from the safety, environmental, operational, maintenance and quality management viewpoints. It includes recommendations
on inspection, maintenance and repairs. This calls for:
1. Structural and anchoring modeling and analysis (1
st
assessment and subsequent annual reassessments).
2. Qualitative RBI implementation (Risk Based Inspection).
3. Yearly reviews of the IRM plan (Inspection, Repair and Maintenance).
4. Data management and storage (including reports).
5. Assistance for Emergency Response.
6. Gives the framework for exceptional analysis.
A detailed description of the Floating Units Integrity Management program is given in references [1] and [2].
The program is divided into four complementary, interacting modules as shown on Fig. 2:
1. Structural non linear Finite Element Analysis (FEM) model and dynamic mooring model. (ABAQUS,
HYDROSTAR and ARIANE Models).
2. IRM (Inspection Repair Maintenance): inspection plan and schedule incorporating class requirements (renewal of
certificates, repairs), and incorporating RBI (Risk Based Inspection).
3. Database (plans, results of models, inspection reports, class status, etc.), with information shared in a network
system.
4. ERS (Emergency Response Service).
IPTC12546PP 3
Figure 2: Units I ntegrity Management Modules
FEA,
Anchoring
Models &
Management
Tool
RBI Analysis
& IRM Cycle
Data
Management
& Reporting
Emergency
Response
=interaction
Periodical
(re)assessment
In alignment with the integrity management program the unit has been placed within a Classification scope. This holds an
important part as the surveys and maintenance actions required by the Classification Society are introduced and accounted for
within the system. As they are known in advance they can be arranged to minimize impact on production. The french class
society also reviews the work carried out by the Third Party Assistance company who contributes to the content and
deployment of the program. Falling within the Class scope of work are the mooring, hull and marine systems,
accommodation quarters and helideck structures and topsides connection to deck. In some cases risers and subsea equipment
are also included.
PreStressed Concrete FPU presentation
The unit is divided in 26 lateral capacities (B and T capacities) and 13 central capacities (C capacities). The lateral capacities
can be used as ballast tanks but only the 4 tanks in the corners are used on site to maintain trim and pitch. The central
capacities are void spaces.
Running through the central void spaces as a spine is the Technical gallery that connects the aft and fore pump rooms (see
dark grey color in figure 3) and provides access to the internal capacities.
The nose that can be seen on the fore end is a concrete cantilever that supports the flare tower as far away as it can be from
the accommodation quarters.
Figure 3: Capacity plan & starboard profile
4 IPTC12546PP
The shell is not a full watertight continuous skin. It is pierced in several
Molto più che documenti.
Scopri tutto ciò che Scribd ha da offrire, inclusi libri e audiolibri dei maggiori editori.
Annulla in qualsiasi momento.