response of a tension leg platform to stochastic wave forcescoebenaroya/publications/response... ·...
TRANSCRIPT
Response of a tension leg platform to stochastic wave forces
Ron Adrezina, Haym Benaroyab
aDepartment of Mechanical Engineering, University of Hartford, West Hartford, CT 06117-1599, USAbDepartment of Mechanical and Aerospace Engineering, Rutgers University, Piscataway, NJ 08855-0909, USA
Abstract
The equations of motion and the response of a Tension Leg Platform with a single tendon undergoing planar motion are presented. The hull
is represented by a rigid cylindrical body and the tendon as a nonlinear elastic beam. Tendons have often been modeled as massless springs,
which neglects contribution to the response by the wave forces on the tendons and its varying stiffness due to changes in its tension. The
structure is subjected to random wave loading. Linear wave theory is applied and the random wave height power spectrum is transformed into
a time history using Borgman's method. The surge and pitch responses for the hull, and the surge response along the tendon are presented for
two cases. Case 1 represents a tendon with a hull mass two orders of magnitude smaller than the tendon's mass. In Case 2, the hull mass is two
orders of magnitude greater than the tendon. Inclusion of tendon forces were found to signi®cantly increase the amplitude of the surge
response for Case 1 but not for Case 2. q 1998 Elsevier Science Ltd. All rights reserved.
Keywords: Tension Leg Platform; Linear wave theory; Random wave loading
1. Introduction
Of the classes of offshore structures, the Tension Leg
Platform (TLP) is particularly well suited for deepwater
operation. Unlike ®xed structures, its cost does not drama-
tically increase with water depth. The TLP is vertically
moored at each corner of the hull minimizing the heave,
pitch and roll of the platform. This small vertical motion
results in less expensive production equipment than would
be required on a semi-submersible [1, 2].
TLPs are complete oil and natural gas production facil-
ities costing $1 billion or more [3]. The supporting structure
of a TLP consists of a hull, tendons and templates as shown
in Fig. 1. The hull is a buoyant structure with a deck at its
top that supports the oil production facility and crew hous-
ing. Pontoons and columns provide suf®cient buoyancy to
maintain the deck above the waves during all sea states.
These columns are moored to the sea ¯oor through tendons,
and ®xed in place with templates. The hull's buoyancy
creates tension in the tendons.
The tallest TLP at the time of its construction, Shell Oil's
Auger TLP in the Gulf of Mexico, began production in 1994
after an investment of 6 years and $1.2 billion. The Auger
TLP with a crew of 112 has two main decks 90 m by 90 m
(300 ft by 300 ft) with a well bay at its center. Four cylind-
rical columns 22.5 m (74 ft) in diameter and pontoons
[8.5 m by 10.7 m (28 ft by 35 ft) cross-section] comprise
the hull. There are three tendons at each column. Each
tendon, also known as a tether or tension leg, was assembled
from 12 steel pipes with a 66 cm (26 in) diameter and a
3.3 cm (1.3 in) wall thickness, connected end to end, and
a total length of 884 m (2900 ft). During severe storms it
may surge 72 m (235 ft) [3, 4].
The surge, sway and yaw resonance frequencies of TLPs
are below that of the wave frequency range as de®ned by a
power spectrum such as the Pierson±Moskowitz. The
heave, pitch and roll resonance frequencies are above this
range. The resulting response is a desirable feature of TLPs.
Wind, waves and current will cause a TLP to oscillate about
an offset position rather than its vertical position. This offset
in the surge direction has a corresponding setdown, the
lowering of the TLP in the heave direction, which increases
the buoyancy forces. This results in a higher tension in the
tendon than if it was in the vertical position. Higher-order
effects due to the nonlinear nature of the waves and
nonlinear structural properties will affect the dynamic
response and may be of interest. Papers that include varying
levels of higher-order effects are discussed in a review paper
by Adrezin et al. [5].
Song and Kareem [6] analyzed a TLP as a platform±
tendon coupled structure. Here the tendon's curvature,
mass and varying tension are included in the discretized
mass, damping, stiffness matrices and the force vector,
rather than approximated as an equivalent spring. The
stochastic response to random wind and wave forces was
determined. The response of the coupled system was then
Probabilistic Engineering Mechanics 14 (1999) 3±17
0266-8920/99/$ - see front matter q 1998 Elsevier Science Ltd. All rights reserved.
PII: S0266-8920(98)00012-5
compared to a system in which the 16 tendons were modeled
as equivalent springs. For surge, sway and heave motions
the difference between results of the mean value was
between 3 and 8%. The differences for the heave, roll and
pitch motions were between 30 and 58%. These large differ-
ences were attributed to the heave, roll and pitch motions
being controlled by the elastic force in the tendons and
hence more sensitive to its characteristics. Use of an equiva-
lent spring does not include the effect of roll and pitch on the
tendon's stiffness. It was also shown that the importance of
including the nonlinear effects of the tendon increases with
greater water depth.
Kim et al. [7] studied the nonlinear response of a TLP
subjected to random waves, steady winds and currents. In
the model, the hull and tendons are coupled and large displa-
cements are allowed. The ®rst and second-order wave forces
were determined using a higher-order boundary element
method. The response was studied in the time domain utiliz-
ing a three-dimensional hybrid element method. The hull
was treated as a rigid body and the tendons were modeled
with three-dimensional beam elements connected at the top
and bottom with spring±damper elements. Responses were
analyzed for surge, heave and pitch motion, offset and
setdown. The mean tension in the tendons was found to be
about 15% higher than the pre-tension due to the large mean
displacements. For a TLP model with 415 m tendons, the
mean offset and setdown were found to be approximately 20
and 0.5 m respectively.
Wave-induced forces for a TLP in an offset position may
differ from those at the undisplaced position. Li and Kareem
[8] analyzed this effect by adding displacement-induced
feedback forces to the wave-induced loads calculated at
the TLP's undisplaced position. Applying this method to a
TLP model resulted in a surge response due to the feedback
forces that were of the same magnitude as the wave
frequency response. Drift induced by second-order wave
potential was not included.
Patel and Lynch [9] modeled the hull as a rigid body with
six degrees of freedom subject to wave forces. It was
coupled to the ®nite element model of the tendon. Differ-
ences between the quasi-static and dynamic tendon models
were minimal except for tendons of lengths of the order of
1500 m or greater. Therefore, the effect of the tendon's
dynamics on the hull's response was only signi®cant when
the tendons were long, had a large mass per unit length and
the hull's displacements were small. The bending stresses in
the tendons were also found to be small. The highest values
were found for short tendons with large outer diameters and
thin wall thicknesses.
Patel and Park [10] studied the combined axial and lateral
response of tendons. The tendon was modeled as a straight
simply supported column subject to both horizontal and
vertical motions at its top. This motion represented the
response of the hull. At the tendon's top, the lateral force
was assumed sinusoidal. Results from studies of three differ-
ent tendon lengths were presented. The analysis showed that
a greater amplitude of vibration will result from the
combined excitation than the individual axial and lateral
excitations. This was particularly signi®cant in even
numbers of the instability region of the Mathieu stability
chart. The frequency of oscillation was also found to be
dependent on the relative magnitudes of the individual exci-
tations for the combined response.
Park et al. [11] performed a reliability analysis on 4- and
6-column TLPs. Two failure criteria were considered for the
tendons at the four corners, ultimate tensile strength and
negative tension. The results indicated that negative tension,
or slacking in the tendon, may result in high impulsive
stresses at the return of positive tendon tension. This was
determined to be the governing failure criterion.
Mekha et al. [12] created three different tendon models.
The ®rst was as a massless spring to provide a constant
lateral stiffness with TLP setdown disregarded. The second
allowed for time varying axial forces in the spring with one-
third of the tendon mass lumped at the attachment points.
The third tendon model consisted of ¯exural beam elements
and allowed for the time varying axial forces. Hydrody-
namic loads on the tendons were included in the third
model only. The hull was modeled as a rigid body subject
to regular waves and analyzed at different wave depths,
wave heights and wave periods. The amplitude of the
surge response was found to be independent of the model
used. For a constant wave period it increased linearly with
wave height, and decreased linearly with increasing wave
frequency for a constant wave height. The surge mean offset
may more than double for the third model where hydrody-
namic loads on the tendons were included. The maximum
tendon forces were not affected by the hydrodynamic loads.
However, the minimum decreased by 10%. Neglecting the
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±174
Fig. 1. Schematic of a Tension Leg Platform.
inclusion of the hull setdown in the ®rst model resulted in a
maximum tendon force that was about 15% higher.
2. Problem description
In this paper, the response of a TLP with a single tendon
undergoing planar motion will be presented. The hull is
represented by a rigid cylindrical body and the tendon as a
nonlinear elastic beam. Tendons have often been modeled
as massless springs, which neglects the contribution to the
response by the wave forces on the tendon and its varying
stiffness due to changes in its tension. The TLP is subjected
to random wave loading.
2.1. Assumptions
² Tendon length is much greater than its diameter
² Tendon is inextensible and shear effects are ignored
² Planar motion
² Hull is treated as a rigid body
² Hull pitch angle is small due to excessive buoyancy
forces, therefore f is small
² The top of the hull always remains above the wave eleva-
tion
² The bottom of the hull is always submerged (otherwise
unstable)
² The hull is represented by a hollow cylinder
² Center of gravity of the hull is located at the geometrical
center of the hull
² Water level is assumed constant and horizontal across the
hull's column at a given instant of time
The assumption that the tendon length is much greater
than its diameter allows shear-deformation effects to be
disregarded because these will be considerably less than
the effects due to the bending energy. In addition, high
frequency motions are not of interest in the present study
because they are well above the range of loading. These two
assumptions allow for the use of Euler±Bernoulli beam
theory [13].
The natural frequencies for transverse vibration are
signi®cantly lower than those for longitudinal vibration.
The ratio of the transverse to longitudinal frequency is of
the order of the tendon diameter to its length. Because the
tendon length is much greater than its diameter, the long-
itudinal frequencies are well above the range of the envir-
onmental loading and will be disregarded in this study.
Therefore, the assumption that the tendon is inextensible
is valid.
The assumption of planar motion is reasonable if vortex
shedding is disregarded. Vortex shedding results in an out-
of-plane response which needs to be explored further
[14±21].
Treatment of the hull as a rigid body is a reasonable
assumption. Its stiffness results in natural frequencies well
above the range for wave, wind and current loading. The
requirement that the top of the hull always remains above
the waves is necessary to allow it to safely support oil and
gas production activities. Due to the mass of the hull and the
¯exibility of the tendons, the bottom of the hull will always
remain submerged for the system to remain stable.
The representation of the hull as a hollow cylinder with a
center of gravity located at its center (together with the
planar motion restriction) may affect the pitch response.
An actual hull is top heavy with the production facility
and crew housing at its top and buoyant pontoons and
columns at its bottom. The derived equations allow for the
center of gravity to be at any position.
The water level is assumed constant and horizontal across
the hull's column at a given instant of time. This simpli®es
the calculation of the center of buoyancy. The validity of
this assumption is dependent on the wave height, wave
frequency and the column's diameter. The free water
surface will also remain below the top of the hull.
A typical TLP has one or more tendons at each corner.
This model has a single tendon pinned to the hull and the sea
¯oor. The pitch response from a single tendon model may
differ signi®cantly from that of a multiple tendon model.
The additional tendons can restrict the pitch response.
3. Derivation of the Lagrangian
The Lagrangian for the tendon, L, is the sum of the ener-
gies:
L � Ek 2 Es 2 Ep (1)
where Ek represents the kinetic energy, Es represents the
strain energy and Ep includes all potential energy other
than due to strain. The tendon is modeled as a beam in
this paper. Due to the length of a tendon compared to its
diameter, modeling the tendon as a cable may be a reason-
able assumption. However, in order to maintain generality
and retain higher-order effects, a beam model is used which
retains the tendon's stiffness [10, 12, 22].
3.1. Strain energy
A body subject to forces may undergo rigid body motions
and deformations. The rigid body motions do not induce
stress and therefore do not produce strain. The study of a
deformable body is required to determine the strain energy
in a system. A point on the undeformed body has the coor-
dinate (X,Y) and an attached in®nitesimal line segment has
the length dS. After the body is deformed the coordinate for
the point (X,Y) is now (x,y) and the new length is ds. The
vertical and horizontal displacement of a point due to defor-
mation is represented by u and v respectively. They are
de®ned as:
u � x 2 X (2)
v � y 2 Y (3)
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±17 5
The transverse and shear strain are substantially smaller
than the axial strain; therefore only this strain will be
considered. Strain induced by torsion and temperature
changes will also be disregarded.
It is assumed that deformations are within a material's
linearly elastic range with a modulus of elasticity E. It is
also assumed that a cross-sectional plane will remain plane
after deformation; therefore a neutral axis will exist.
The axial strain is derived due to bending stresses only for
the reasons discussed above. The tendon is assumed to be in
its undeformed position when vertical. Therefore, it is desir-
able to express the orientation of an in®nitesimal segment ds
with respect to the x-axis. As shown in Fig. 2, this angle is u .
Assuming in®nitesimal strain, the strain energy may be
written as
Es �Xs�l
s�0
1
2EI�s� 2u
2s
� �2
ds (4)
I(s) is the area moment of inertia of the tendon and the
relationship 2u2s
is equal to the inverse of the radius of curva-
ture R. This is derived in Cartesian coordinates as
R�x� �1 1
2v
2x
� �2 !3=2
22v
2x2
(5)
The coordinate s is a path variable along the neutral axis.
An incremental length, ds, may be derived as follows:
ds2 � dx2 1 dv2 (6)
ds ����������������1 1 vx�x; t�2
qdx (7)
3.1.1. Constraint equation between u and v
The assumption of an inextensible tendon leads to a
constraint relationship between coordinates u and v. Recall
that u � x 2 X. From Fig. 2, we can conclude that if the
beam is vertical and straight in its undeformed position then
coordinate X � S. Furthermore, if inextensible, S � s. This
leads to the relationship
X � s (8)
which will be of use in ®nding the constraint relationship.
Assigning j as a dummy variable for x, the coordinate x may
be found as follows:
s �Zj�x
j�0
��������������������1 1
2v�j; t�2j
� �2s
dj (9)
If the slope 2v2j is assumed small, which is a reasonable
assumption based on the observed motion of a TLP, then
applying a Taylor Expansion yields:
s � x 11
2v2v
2j
����j�x
j�02
1
2
Zj�x
j�0v22v
2j2dj (10)
The boundary condition that v� 0 at the base, j � 0, will
be applied. Solving for x and substituting into Eq. (2) along
with Eq. (8) results in the following holonomic constraint
relationship:
u � 21
2v2v
2j
����j�x
11
2
Zj�x
j�0v22v
2j2dj (11)
The second term of the above equation must be solved
numerically if no further assumptions are made. However,
for a TLP, the second term will be much smaller than the
®rst. This is due to the high tension in the tendons minimiz-
ing the variation of the slope along its length. The vertical
velocity is derived from the above equation as
2u
2t� 2
1
2
2v
2t
2v
2j
����j�x
21
2v22v
2j2t
�����j�x
11
2
Zj�x
j�0
2v
2t
22v
2j21 v
23v
2j22t
( )dj (12)
3.2. Kinetic energy
For an incremental length of tendon in the perturbed state,
ds, the corresponding incremental mass, dm, is as follows:
dm � 1
4p rTp�s� D2
o 2 D2i
� �1 CArflp�s�D2
o
h ids (13)
where Do and Di are the outer and inner diameters of the
tendon. The tendon and the seawater densities are repre-
sented by rTp(s) and r¯p(s) respectively, in the perturbed
con®guration. If the tendon is assumed inextensible, this
results in a constant volume and lengths dX and ds are
equal. Therefore the tendon density is assumed constant:
rTp(s) � rTp. Water density will also be assumed constant:
r¯p(s)� r¯p. When the perturbed length ds is projected onto
the x-axis, it results in a length dx # ds (see Fig. 2). We can
see that the upper bound dx � ds occurs when the tendon is
in its undeformed (vertical) position.
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±176
Fig. 2. Incremental element of mass dm shown in its undeformed position
dS, its perturbed position ds and its projection dx.
The incremental mass in the x±y frame may be written
with a constant density and varying diameters to represent
an elliptical horizontal cross-section of the deformed
tendon. A second alternative assumes that the constant
diameters along the path variable s will be used, and the
density will vary. For instance, to account for dx , ds the
density will increase, rT(x) . rTp, where the tendon density
in the projected con®guration is represented by rT(x).
Applying the second alternative with these assumptions
and satisfying the continuity equation results in the follow-
ing relationships between the densities for the perturbed and
present Cartesian con®gurations:
rT�x� � rTp
���������������1 1 vx�x; t�2
q(14)
rfl�x� � rflp
���������������1 1 vx�x; t�2
q(15)
The second term of Eq. (13) is the added mass, where CA
is the non-dimensional added mass coef®cient that is related
to the inertia coef®cient CM by
CA � CM 2 1 (16)
For a very large tendon length to diameter ratio, as is the
case for these structures, CM approaches its theoretical limit
of 2 and CA approaches 1. This added mass is due to the
entraining of ¯uid particles by the motion of the tendon.
The area moment of inertia for the tendon's cross-section
is
I1�s� � 1
64p D4
o 2 D4i
� �(17)
and its mass moment of inertia for a thin hollow disk
dJ1 � dmI1�s�A�s�
� 1
64p rT D2
o 2 D2i
� �1 CArflD
2o
h i� D2
o 1 D2i
� � ���������������1 1 vx�x; t�2
qdx (18)
where subscript 1 denotes the tendon and 2 the hull in the
following sections.
The kinetic energy due to the tendon's linear and angular
velocities may be represented by
Ek �Zs�l
s�0
1
2dm V2
x 1 V2y
h i1Zs�l
s�0
1
2dJ1
_u 2 (19)
Substituting Eqs. (7), (13) and (18) into Eq. (19) with the
velocities in the x-direction, Vx, and the y-direction, Vy,
represented by ut(x,t) and vt(x,t) respectively, results in the
kinetic energy expressed in the present Cartesian frame,
Ek �Zx�Lx
x�0
1
8p rTp D2
o 2 D2i
� �1 CArflpD2
o
h i�
����������������1 1 vx�x; t�2
qut�x; t�2 1 vt�x; t�2h i
11
128p rT D2
o 2 D2i
� �1 CArflD
2o
h iD2
o 1 D2i
� ��
���������������1 1 vx�x; t�2
q_u �x; t�2
�dx (20)
where the angular velocity, _u �x; t�, is found from the geome-
try as follows:
_u �x; t� � vxt�x; t�1 1 vx�x; t�2
(21)
The top of the tendon, coordinate s� l, projected onto the
x-axis, is represented by Lx.
3.3. Potential energy due to gravity and buoyancy
The potential energy due to gravity and buoyancy is
expressed as
Ep �Zx�Lx
x�0Fgbdsx
�Zx�Lx
x�02
1
4rT�x�p D2
o 2 D2i
� ��l 2 s�g
�
21
4rfl�x�pD2
o�l 2 s�g�
dsx (22)
where Fgb is the force due to gravity and buoyancy of the
tendon. Eq. (22) is an approximation to the inclusion of the
membrane strain in the Lagrangian and the work due to
vertical forces represented as 2R
Fgbudx. The change in
the vertical projection of an element ds is
dsx � ds 2 dx ����������������1 1 vx�x; t�2
q2 1
� �dx (23)
Substituting Eqs. (4), (20) and (22) into Eq. (1) yields the
Lagrangian, which is then substituted into Hamilton's
Principle.
4. Coupling the tendon with the hull
In Fig. 3 the bottom beam represents the tendon with a
horizontal displacement of v(x,t). The top mass represents
the entire hull and its payload. It is modeled as a rigid body
with length LH and diameter DH and rotates about the hinge
connecting it to the tendon with angle f (t).
4.1. Equation of motion for hull pinned to tendon
The kinematics and kinetics of the hull are now
developed.
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±17 7
4.1.1. Kinematics
The displacement vector from the origin to the hull's
center of mass, ~r�t�, is written as
~r�t� � �Lx�t�1LH
2cos f�i 1 �v�Lx�t�; t�1
LH
2sin f�j (24)
The velocity and the acceleration of the hull are
~_r�t� � � _Lx�t�2LH
2_f sin f�i 1 � _v�Lx�t�; t�1
LH
2_f cos f�j
~�r�t� � � �Lx�t�2LH
2�f sin f 2
LH
2_f 2 cos f�i 1 � �v�Lx�t�; t�
1LH
2�f cos f 2
LH
2_f 2 sin f�j (25)
The moment arm, lb, between the hull's buoyancy forces
and the pivot in the inertial coordinate system can be shown
to be
lb � D2H
32LHS�v; t� 2 1 tan2f� �
1LHS�v; t�
2
" #sin f (26)
4.1.2. Kinetics
The sum of the inertia, weight and buoyancy forces of the
hull are resisted by the tendon (see Fig. 4). These are given
by
F2b � rflgVHS�v; t� (27)
F2g � Mg (28)
and the vertical and horizontal reaction forces at the hinge
(point c) are respectively Fcx and Fcy.
The relationship between the mass moment of inertia for
the hull about its center of mass and the pivot is
JH � Jcm 1 �MLH
2
� �2
(29)
where �M is the mass of the hull plus its added mass.
Performing a force balance on the hull yields the follow-
ing equations for the reaction forces and the equation of
motion of the hull in terms of generalized coordinate f :
Fcx � 2 �M� �Lx�t�2LH
2�f sin f 2
LH
2_f 2cos f�
1 rflgVHS�v; t�2 Mg (30)
Fcy � 2 �M� �v�Lx�t�; t�1LH
2�f cos f 2
LH
2_f 2 sin f� (31)
JH�f 2
1
2MgLH sin f 1 rflgVHS
� �v; t� D2H
32LHS�t� 2 1 tan2f� �
1LHS�v; t�
2
" #sin f
1 �MLH
2� �v�Lx�t�; t� cos f 2 �Lx�t� sin f� � Qf (32)
where
LHS�v; t� � d 1 h�v; t�2 Lx�t�cos f
(33)
The tension due to the hull equals Fcx, where the hull's
inertia is �M �Lx�t�, the weight of the hull is Mg and its buoy-
ancy is rflgVHS�v; t�. The volume of the submerged portion
of the hull, VHS(v,t), varies with the displacement of the hull
in time. If the hull is assumed to pivot about its attachment
to the tendon, then for a constant cross-sectional area the
submerged volume is:
VHS�v; t� � pD2
H
4
d 1 h�v; t�2 Lx�t�cos f
(34)
4.2. Equations of motion for a tendon with a pinned hull
The total work W may be expressed as
W �Zx�Lx
x�0Fcx 1 Fx
Hfl
ÿ �dsx 1 Fcyv�x; t� x�Lx
��� (35)
and 2FyHfl is added to the boundary condition at the top of
the tendon. The forces due to the wave forces will be
included in Qf¯, Qy¯, FxHfl, F
yHfl and M¯, which represent
the generalized moment on the hull about the pin between
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±178
Fig. 4. Free body diagram of the hull.Fig. 3. Model of tendon with hull.
the hull and tendon, the generalized force on the tendon, the
horizontal and vertical forces due to the ¯uid forces on the
hull, and the moment at the ends of the tendon, respectively.
Applying the extended Hamilton's Principle will result in
the equations of motion and the boundary conditions for the
tendon, the top of which corresponds to the surge and heave
response of the hull. The following equation is for the pitch
response of the hull:
Jcm�f 2
1
2MgLH sin f 1 rflgVHS
� �v; t� D2H
32LHS�t� �2 1 tan2f�1LHS�v; t�
2
" #sin f
2 �MLH
2� �v�Lx�t�; t�cos f 2 �Lx�t�sin f 1
LH
2�f � � Qffl
(36)
5. Stochastic wave forces
5.1. Wave kinematics
In the analysis presented in this paper, linear wave theory
is applied. The following are the assumptions required for
its proper application [23]:
² The wave amplitude is small compared to the wave
length and the water depth.
² The water depth is constant.
² The water is non-viscous and irrotational.
² The water is incompressible and homogenous.
² Coriolis force due to the earth's rotation is negligible.
² Surface tension is negligible.
² The sea ¯oor is smooth and impermeable.
² The sea level atmospheric pressure is uniform.
The random wave height power spectrum is transformed
into a time history using the method by Borgman [24] and
applied in the paper by Bar-Avi and Benaroya [25]. The
wave surface pro®le is given by
h�y; t� � H
2cos�ky 2 vt 1 1� (37)
The Pierson±Moskowitz spectrum for the signi®cant
wave height Hs is of the form
Sh�v� � A0
v5e2B=v4
(38)
where the constants A0 and B are de®ned as
A0 � 8:1e 2 3g2 (39)
B � 3:11
H2s
(40)
The wave elevation is approximated as
�h�y; t� ��������
A0
4BN
r XNn�1
cos kny 2 vnt 1 1n
ÿ �where the partition frequencies v n are
vn � B
ln�N=n�1 B=F4
� � 14; n � 1; 2;¼N (42)
and the relationship between the wave numbers, kn, and the
frequencies is de®ned by
v2n � gkn tanh knd (43)
F is the terminal frequency, where the wave energy spec-
trum is truncated at Sh (F). The horizontal wave velocity is
uw �XNn�1
�������A0
4BN
rvn
cosh kny
sinh kndcos kny 2 vnt 1 1n
ÿ �(44)
and the vertical wave velocity is
ww �XNn�1
�������A0
4BN
rvn
sinh kny
sinh kndsin kny 2 vnt 1 1n
ÿ �(45)
From the literature, linear wave theory appears to be an
excellent approximation for the forces on the tendon. On the
hull, however, nonlinear effects may be of importance. For
example, the variable wetting of the hull's columns may
lead to such higher-order effects as ringing [26±28].
5.2. Forces
Morison's equation is generally applied to slender
offshore structures. For a cylindrical element, the equation
for the force per unit length normal to the cylinder is given
by
Ffl � CDrD
2Vn
fl 2 VnTj j Vn
fl 2 VnT
ÿ �1 CMrp
D2
4_Vn
fl (46)
where the viscous, frictional drag coef®cient CD and the
inertia coef®cient CM are functions of the Reynolds number,
the wave characteristics, the cylinder diameter and its
roughness. The ®rst term of Eq. (46) is the drag force,
which is the force per unit length required to hold the cylin-
der in place subject to a stream of relative velocity
Vnfl 2 Vn
T
ÿ �. The second term is the force per unit length
required to hold the cylinder in place subject to a constant
free stream acceleration of _Vnfl. The acceleration of the
cylinder _VnT was already accounted for in the added mass
term of Eq. (13).
5.2.1. Fluid forces on the tendon
The generalized force per unit length on the tendon will
be determined from Eq. (46), with the following relation-
ships for the velocities and accelerations:
VnT � vt��������
1 1 v2x
p (47)
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±17 9
Vnfl � uw 2 wwvx��������
1 1 v2x
p (48)
_Vnfl � _uw 2 _wwvx��������
1 1 v2x
p (49)
resulting in the following equation:
Qyfl � CDrfl
Do
2
uw 2 vt 2 wwvx��������1 1 v2
x
p���������� uw 2 vt 2 wwvx��������
1 1 v2x
p !
1 CMrflpD2
o
4
_uw 2 _wwvx��������1 1 v2
x
p !(50)
5.2.2. Fluid forces on the hull
The displacement vector, rH, from the origin of the iner-
tial frame to any point along the centerline of the hull is as
follows:
rH � Lx�t�1 x cos f�t�� �i 1 v Lx�t�; t
ÿ �1 x sin f�t�� �
j (51)
and its velocity is
_rH � _Lx�t�2 x _f �t� sin f�t�� �i
1 _v Lx�t�; tÿ �
1 x _f �t� cos f�t�� �j (52)
which provides the value of VnT. The force per unit length on
the hull due to the wave forces in the inertial frame is
�FHfl ��Fx
Hfl
�FyHfl
( )(53)
resulting in a moment about the pivot between the hull and
the tendon of
Qffl �Zx�Lx 1 LHS cos f�t�
x�Lx
�FyHflx 2 �Fx
Hflx tan f�t�� dx (54)
and resultant forces on the tendon of
FxHfl �
Zx�Lx 1 LHS cos f�t�
x�Lx
�FxHfldx (55)
and
FyHfl �
Zx�Lx 1 LHS cos f�t�
x�Lx
�FyHfldx (56)
6. Numerical solution
The partial differential equations are discretized spatially
using the ®nite difference method. The resulting ordinary
differential equations are solved using the Runge±Kutta
method. This is implemented with Matlab [29]. The hull's
mass includes both structural and non-structural (e.g. crew
housing, storage tanks) mass. The submerged hull volume
as expressed by Eq. (34) was reduced by a factor of 4. These
adjustments to the mass and volume were made to raise the
TLP's ®rst natural frequency into the range of the Pierson±
Moskowitz spectrum for a hull mass of 4.53e6 kg (see Table
1). It is desirable in an actual design for this frequency
corresponding to surge to be below this range. Here, we
are looking to study extreme conditions. Figs 5 and 11 exhi-
bit a growing high frequency component in the response of
the surge at 282 m along the length. This results from using
a maximum time step in the Runge±Kutta method that is too
large. Reducing this parameter eliminates these high
frequencies.
7. Responses
The surge and pitch responses for the hull, and the surge
response along the tendon, are presented for two cases. Case
1 represents a tendon with a hull mass two orders of magni-
tude smaller than the tendon's mass. In Case 2, the hull mass
is two orders of magnitude greater than the tendon. The
mass moment of inertia of the hull varies with the change
in mass but all other parameters are constant, as shown in
Table 1. Figs 5±15 show the responses and the power spec-
tral densities for the two cases. For each case, a damped free
vibration response, TLP subjected to random wave forces on
the hull only, and wave forces on the hull and tendon are
presented. Fig. 10 shows the response with a signi®cant
wave height of 15 m, all other forced responses use a signif-
icant wave height of 9 m.
8. Discussion and summary
The responses exhibit both transient and steady-state
behavior. For the lower hull mass, the ®rst damped natural
frequency was found to be approximately 0.44 Hz for surge
and 0.83 Hz for pitch. For the larger hull mass, the pitch and
surge responses exhibited a frequency of 0.06 Hz, which is
within the range of the Pierson±Moskowitz spectrum. The
response of the tendon at a node 282 m from its bottom also
shows less dominant frequencies at 0.12, 0.17 and 0.22 Hz.
This was typical of the nodes away from the boundaries.
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±1710
Table 1
TLP properties
Property Value
Tendon length 470 m
Tendon outer diameter 0.6096 m
Tendon inner diameter 0.5890 m
Tendon density 7800 kg/m3
Hull length 50 m
Hull outer diameter 30 m
Hull inner diameter 29 m
Hull inertia 171 and 1.71e6 kg/m2
Hull mass 453 and 4.53e6 kg
Modulus of elasticity 204e9 N/m2
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±17 11
Fig. 6. PSD of the damped free vibration of the TLP with a hull mass of 453 kg.
Fig. 5. Damped free vibration of the TLP with a hull mass of 453 kg.
8.1.1. Case 1
Inclusion of tendon force signi®cantly increases the
amplitude of the surge response. This supports the conclu-
sions by Patel and Lynch [9]. However, because the funda-
mental frequency is well above the range of the Pierson±
Moskowitz spectrum, the amplitudes were small.
8.1.2. Case 2
Inclusion of the tendon force does not signi®cantly affect
the amplitude of the surge response as supported by the
literature [6, 12]. However, researchers have also shown
that at greater tendon lengths, the importance of including
a more sophisticated tendon model increases [6, 9]. The
amplitudes of the response for Case 2 were greater than
for Case 1 because the TLP has a fundamental frequency
which corresponds to the wave frequencies.
8.1.3. Case 2
The hull pitch response shows a non-zero equilibrium
point. This is due to the combination of insuf®cient hull
buoyancy and a buoyant tendon. The tendon's buoyancy
is greater than its weight resulting in an upward force at
the attachment point with the hull. This causes the hull to
tip in order to increase its submerged volume rather than
sinking vertically. When a third case was studied (not
presented here) with an increased hull volume, the equili-
brium con®guration became the vertical upright position
(f � 0).
The analytical model presented in this paper would be
useful to the designers of TLPs in determining the overall
dynamic response characteristics. The structural and ¯uid
parameters listed in Tables 1 and 2 may be varied to deter-
mine their effects. Additionally, deterministic wave forces,
current and wind may be applied [30]. This analysis should
be performed prior to the creation of a large ®nite element
model, scale model tank testing or other design-speci®c
analysis. The usefulness of an analysis such as this is in
specifying the design space.
Acknowledgements
This work was supported by the Of®ce of Naval Research
Grant no. N00014-94-1-0753. The ®rst author is grateful for
his ONR-sponsored Fellowship, and the second author for
additional support. Our program manager, Dr. T. Swean, is
sincerely thanked for this support as well as his interest in
our work.
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±1712
Fig. 7. Response of the TLP with a hull mass of 453 kg and waves on the hull only.
Table 2
Fluid properties
Property Value
Mean water level 500 m
Water density 1025 kg/m3
Signi®cant wave height 9 and 15 m
Drag coef®cient 1.0
Inertia coef®cient 2.0
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±17 13
Fig. 9. PSD of the TLP with a hull mass of 453 kg and waves on the tendon and the hull.
Fig. 8. Response of the TLP with a hull mass of 453 kg and waves on the tendon and the hull.
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±1714
Fig. 10. Response of the TLP with a hull mass of 453 kg and waves on the tendon and the hull, signi®cant wave height of 15.
Fig. 11. Damped free vibration of the TLP with a hull mass of 4.53e6 kg.
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±17 15
Fig. 13. Response of the TLP with a hull mass of 4.53e6 kg and waves on the hull only.
Fig. 12. PSD of the damped free vibration of the TLP with a hull mass of 4.53e6 kg.
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±1716
Fig. 15. Response of the TLP with a hull mass of 4.53e6 kg and waves on the tendon and the hull.
Fig. 14. PSD of the TLP with a hull mass of 4.53e6 kg and waves on the hull only.
References
[1] Natvig BJ, Teigen P. Review of hydrodynamic challenges in TLP
design. International Journal of Offshore and Polar Engineers
1993;3(4):241±249.
[2] Chakrabarti SK. Hydrodynamics of offshore structures. Computa-
tional Mechanics Publications, 1987.
[3] Salpukas A. Oil companies drawn to the deep. The New York Times
December 7 1994:D1,D5.
[4] Robison R. Bullwinkle's big brother. Civil Engineering 1995:44±47.
[5] Adrezin R, Bar-Avi P, Benaroya H. Dynamic response of compliant
offshore structures ± review. Journal of Aerospace Engineering
1996;October:114±131.
[6] Song X, Kareem A. Combined system analysis of tension leg plat-
forms: A parallel computational scheme. OMAE 1994;1:123±134.
[7] Kim CH, Kim MH, Liu YH, Zhao CT. Time domain simulation of
nonlinear response of a coupled TLP system in random seas. In:
Proceedings of the Fourth International Offshore and Polar Engineer-
ing Conference, vol. 1, 1994:68±77.
[8] Li Y, Kareem A. Computation of wind-induced drift forces intro-
duced by displaced position of compliant offshore platforms. Journal
of Offshore Mechanics and Arctic Engineering 1992;114:175±184.
[9] Patel MH, Lynch EJ. Coupled dynamics of tensioned buoyant plat-
forms and mooring tethers. Engineering Structures 1983;5:299±308.
[10] Patel MH, Park HI. Combined axial and lateral responses of tensioned
buoyant platform tethers. Engineering Structures 1995;17(10):687±
695.
[11] Park WS, Yun CB, Yu BK. Reliability analysis of tension leg plat-
forms by domain crossing approach. In: Proceedings of the First
International Offshore and Polar Engineering Conference, vol. 1,
August 1991:108±116.
[12] Mekha BB, Johnson CP, Roesset JM. Implications of tendon model-
ing on nonlinear response of TLP. Journal of Structural Engineering
1996;122(2):142±149.
[13] Clark SK. Dynamics of Continuous Elements. Prentice-Hall, 1972.
[14] Billah KYR. A study of vortex-induced vibration. PhD thesis, Prin-
ceton University, June 1989.
[15] Lee YG, Hong SW, Kang KJ. A numerical solution of vortex motion
behind a circular cylinder above a horizontal plane boundary. In:
Proceedings of the Fourth International Offshore and Polar Engineer-
ing Conference, vol. 3, April 1994:427±433.
[16] Moe G, Holden K, Yttervoll PO. Motion of spring supported cylinders
in sub-critical and critical water ¯ows. In: Proceedings of the Fourth
International Offshore and Polar Engineering Conference, vol. 3,
April 1994:468±475.
[17] Sibetheros IA, Miksad RW, Ventre AV, Lambrakos KF. Flow
mapping on the reversing vortex wake of a cylinder in planar harmo-
nic ¯ow. In: Proceedings of the Fourth International Offshore and
Polar Engineering Conference, vol. 3, April 1994:406±412.
[18] Sumer BM, Kozakiewicz A. Visualization of ¯ow around cylinders in
irregular waves. In: Proceedings of the Fourth International Offshore
and Polar Engineering Conference, vol. 3, April 1994:413±420.
[19] Sunahara S, Kinoshita T. Flow around circular cylinder oscillating at
low keulegan-carpenter number. In: Proceedings of the Fourth Inter-
national Offshore and Polar Engineering Conference, vol. 3, April
1994:476±483.
[20] Chung JS, Whitney AK, Lezius D, Conti RJ. Flow-induced torsional
moment and vortex suppression for a circular cylinder with cables. In:
Proceedings of the Fourth International Offshore and Polar Engineer-
ing Conference, vol. 3, April 1994:447±459.
[21] Bokaian A. Lock-in prediction of marine risers and tethers. Journal of
Sound and Vibration 1994;175(5):607±623.
[22] Patel MH, Park HI. Tensioned buoyant platform tether response to
short duration tension loss. Marine Structures 1995;8:543±553.
[23] Wilson JF. Dynamics of Offshore Structures. New York: Wiley, 1984.
[24] Borgman LE. Ocean wave simulation for engineering design. Journal
of the Waterways and Harbors Division 1969;95:557±583.
[25] Bar-Avi P, Benaroya H. Non-linear dynamics of an articulated tower
in the ocean. Journal of Sound and Vibration 1996;190(1):77±103.
[26] Kim SB, Powers EJ, Miksad RW, Fischer FJ. Quanti®cation of
nonlinear energy transfer to sum and difference frequency responses
of TLP's. In: Proceedings of the First International Offshore and Polar
Engineering Conference, vol. 1, August 1991:87±92.
[27] Natvig BJ, Vogel H. Sum-frequency excitations in TLP design. In:
Proceedings of the First International Offshore and Polar Engineering
Conference, vol. 1, August 1991:93±99.
[28] Chen XB, Molin B, Petitjean F. Numerical evaluations of the spring-
ing loads on tension leg platforms. Marine Structures 1995;8:501±
524.
[29] Metlab reference guide. The MathWorks, Inc., August 1992.
[30] Adrezin RS. The nonlinear stochastic dynamics of tension leg plat-
forms. PhD thesis, Rutgers, 1997.
R. Adrezin, H. Benaroya / Probabilistic Engineering Mechanics 14 (1999) 3±17 17