You are on page 1of 12

16TH CONGRESS OF IABSE, LUCERNE, 2000

Dynamic Fluid-Structure Interaction Analysis of Floating Platforms

Antonio PARONESSO Antonio Paronesso, born 1958,


Civil Engineer, Ph. D. received his civil engineering
degree from the Univ. of Rome
Passera & Pedretti (Italy) in 1986 and his Docteur
Lugano, Switzerland ès sciences techniques degree
from the Ecole polytechnique
fédérale de Lausanne
(Switzerland) in 1997.

Summary
Oscillations of floating platforms, to be used as temporary surfaces, are studied in the frequency
domain. For the water, a fluid governed by the Helmholtz equation is assumed. The structure
together with an irregular adjacent finite region of the fluid is discretized by applying the technique
of finite elements. For the regular part of the fluid domain extending to infinity, the boundary in
contact with the finite region is discretized and the relationship between the corresponding nodal
pressures and nodal fluxes is derived. Results for a 2-D fluid-structure interaction problem
computed using the software developed by the author are discussed.

Keywords: floating platforms; fluid-structure interaction; Helmholtz equation; frequency domain;


finite element discretization; analytical solution; anchors.

1. Introduction
Floating structures are an innovative solution to providing infrastructures in urban areas located in
proximity of bays or lakes where space is limited. As an example, wide floating temporary
platforms connected to a system of floating pontoon bridges linked to the beach can be constructed
for expo events.
Oscillations of such structures are mainly caused by surface water waves induced by the wind
action. In the evaluation of the response of such structures, a dynamic fluid-structure interaction
problem must be solved.
For platforms with length in axial direction much larger than the width in transverse direction, the
main characteristics of the problem can be grasped analysing the effects produced by regular
incident waves of angular frequency ω acting at an angle π / 2 with respect to the axis of the
platform. In these conditions the 2-D model of the problem depicted in Fig. 1a is appropriate.
Based on the substructure method, the fluid domain is separated in a finite irregular
region ( x, z ) (bounded fluid domain) adjacent to the structure ( x3 ) and in two regular unbounded
regions ( x1 , z ) and ( x2 , z ) , both representing the unbounded fluid domain.
The analysis of the system is performed in the frequency domain. The response is harmonic and can
easily be computed taking the real part of the output of the system to an input characterised by the
time dependency e i ω t .
For the modelling of the water, a fluid with small displacements, irrotational motion and negligible
viscous effects governed by the 2-D frequency domain version of the Helmholtz equation is
assumed:
16TH CONGRESS OF IABSE, LUCERNE, 2000

a) BOUNDED UNBOUNDED
FLUID DOMAIN FLUID DOMAIN
Lw Lp Lw

S.W.L. Am
w(x3,ω) u(ω) PLATFORM
Le
O3 x3

S
R
8

8
d

O
z z z

H
b
p (x,z,ω)

C
AN
x1 O1 O x O 2 x2

{W(ω)} DISCRETIZATION OF THE PLATFORM


b)
{P1(ω)} {P2(ω)}
u(ω)
µ2 µ2 µ6 µ6
µ3 µ4 µ5

8
8

µ1 {P4(ω)} µ7
µ8
DISCRETIZATION OF THE FLUID DOMAIN

Fig. 1, a)Platform-fluid system, b) discretization

∂ 2 p ( x, z , ω ) ∂ 2 p ( x, z , ω ) ω2
+ = − p ( x, z , ω ) (1)
∂x 2 ∂z 2 cw2
p ( x, z, ω ) denotes the pressure of the fluid at point ( x, z ) in excess of hydrostatics (i.e. gravity
forces have been implicitly taken into account). cw is the compression wave velocity.
For the modelling of the structure, the platform is supposed to be infinitely rigid in compression and
with a draft Le << L p . In these conditions, the motion of the structure with respect to the
configuration of hydrostatic equilibrium can be properly described by the vertical displacement
function w( x3 , ω ) and by the horizontal displacement u (ω ) (see Fig. 1a). The equilibrium equation
of the platform in vertical direction is written as
∂4w
( ρ pω 2 + k w ) w( x3 , ω ) + EI ( x3 , ω ) = p ( x3 , ω ) (2)
∂x34
ρ p and EI are, respectively, the density per square meter and the bending stiffness per unit depth of
platform. k w = ρ w gC red , with g equal to the acceleration of gravity, denotes the reduced
hydrostatic restoring force coefficient per unit depth. The reduction coefficient C red ≤ 1 equals the
ratio A flo Age om of the floating surface A flo , to the geometric surface Age om of the platform. In the
case of floating supports composed by non-continuous polystyrene blocks, A flo equals the sum of
the horizontal surfaces of all the blocks, while Age om equals the horizontal surface of the platform.
p( x3 , ω ) is the pressure applied by the fluid to the platform along the boundary µ 4 (see Fig 1b).
The 2-D interaction problem here considered consists of solving eqs. (1) and (2) together with the
following set of boundary conditions
16TH CONGRESS OF IABSE, LUCERNE, 2000

∂p ( x, z , ω )
= ρ wω 2u (ω ) on µ 3 and µ 5 (3)
∂x
∂p ( x, z , ω )
= ρ wω 2 w( x3 , ω ) on µ 4 (4)
∂z
∂p( x, z , ω ) ω 2
= p ( x, z , ω ) on µ 2 and µ 6 (5)
∂z g
∂p ( x, z , ω )
=0 on µ 8 (6)
∂z
p ∞ ( x1 , z , ω ) = p b ( x, z , ω ) on µ1 and µ 7 (7)
Eqs. (3) and (4) are compatibility conditions between the bounded fluid domain and the platform.
Eq. (5), represents the well-known linearized free surface condition. Eq. (6) express the condition of
rigid bottom. Eq. (7) is the compatibility condition between the bounded (superscript b ) and the
unbounded (superscript ∞ ) fluid domains. p ∞ ( x1 , z , ω ) in eq. (7) must verify the condition that at
infinity no gain of energy exists for the unbounded domain (radiation condition).
The problem is solved discretizing the finite fluid domain together with the structure (generalised
structure) according to the technique of the finite elements and deriving an analytical solution for
p ∞ ( x1 , z , ω ) .

2. Bounded fluid domain


In the finite element discretization of the bounded fluid domain, the pressure field p b ( x, z , ω ) ,
(superscript b for bounded domain), is approximated in terms of the vector of the nodal values
{P b (ω )} and of the shape functions [ N wb ( x, z )] (subscript w for water), that is

p b ( x, z , ω ) = [ N wb ( x, z )]{P b (ω )} (8)
Applying in the discretization the Galerkin procedure with the virtual field of pressures defined by
using the same shape functions [ N wb ( x, z )] of the real field p b ( x, z , ω ) , leads to the relationship
relating the vector of nodal fluxes {Pnb (ω )} to the vector {P b (ω )} of the nodal pressures
b
[ S PP (ω )]{P b (ω )} = {Pnb (ω )} (9)
with
ω2
b
[ S PP (ω )] = − [M b ] + [ K b ] {Pnb (ω )} = ò [ N wb ( x, z )]p nb ( x, z , ω )dµ (10)
cw2 µ

The curvilinear integral in eq. (10) is computed along the boundaries µ of the bounded domain
where the pressure p b ( x, z , ω ) ≠ 0 . pnb ( x, z , ω ) denotes the derivative of pressure with respect to
the external normal n of the boundaries µ . The fluid mass matrix [ M b ] and the fluid elastic matrix
[ K b ] equal

[ M b ] = ò [ N wb ( x, z )]T [ N wb ( x, z )]dΩ [ K b ] = ò ([ D][ N wb ( x, z )]) T [ D][ N wb ( x, z )]dΩ (11)


Ω Ω

with [ D ] diagonal matrix containing the partial derivative operators ∂ ∂x and ∂ ∂z . The integrals
16TH CONGRESS OF IABSE, LUCERNE, 2000

in eq. (11) are computed on the surface Ω of the bounded fluid domain.
Denoting with {P1b (ω )} , {P2b (ω )} and {P3b (ω )} the nodal pressures, respectively, on the boundaries
µ1 , µ 7 and µ 3 , µ 4 and µ 5 and denoting with {P4b (ω )} the pressures of the remaining nodes (see
Fig. 1b), eq. (9) can be partitioned as
é [ S11b ] [0] [0] [ S14b ]ù ì{P1b }ü ì{Pnb1}ü
ê b úï ï ï ï
ê
b
[ S 22 ] [0] [ S 24 ]ú ï{P2b }ï ï{Pnb2 }ï
ê b úí ý=í ý (12)
[ S 33b ] [ S 24 ] ï{P3b }ï ï{Pnb3 }ï
ê b úï
ëêSYM [ S 44 ]ûú î{P4b }ïþ ïî{Pnb4 }ïþ

{Pnb1 (ω )} and {Pnb2 (ω )} are the contributions of the bounded domain to the nodal fluxes along the
boundaries µ1 and µ 7 , respectively. {Pnb3 (ω )} are the nodal fluxes along the boundaries µ 3 , µ 4
and µ 5 associated to the motion of the platform while the non-vanishing elements of {Pnb4 (ω )}
correspond to the nodal fluxes on the free surfaces µ 2 and µ 6 of the fluid. Substituting eq. (8) in
(5) and the resulting equation in eq. (10) with integral evaluated along the boundaries µ 2 and µ 6 , it
is straightforward to demonstrate that {Pnb4 (ω )} can be expressed as a function of the vector of the
nodal pressures {P b (ω )} , that is

{Pnb4 (ω )} = [ S s (ω )]{P b (ω )} (13)


Partitioning the right hand side of eq. (13) accordingly to eq. (12) and introducing the result in the
same equation leads to
é [ S11bs ] [0] [0] [ S14bs ]ù ì{P1b }ü ì{Pnb1}ü
ê bs bs ú ï ï ï ï
ê [ S 22 ] [0] [ S 24 ]ú ï{P2b }ï ï{Pnb2 }ï
ê bs bs ú í ý=í ý (14)
[ S 33 ] [ S 24 ] ï{P3b }ï ï{Pnb3 }ï
ê bs ú ï
êëSYM [ S 44 ]úû î{P4b }ïþ ïî {0} ïþ
where
[ Sijus (ω )] = [ Siju (ω )] + [ Sijs (ω )] (15)
Equation (14) represents the input-output relationship of the bounded fluid domain with the effects
of surface waves taken into account.

3. Unbounded fluid domain


The starting point for the derivation of the input-output representation of the unbounded fluid
domain is the frequency domain solution

p ∞ ( x1 , z , ω ) = C0 (ω ) cosh(α 0 (ω ) z )e k0 (ω ) x1 q0 (ω ) + å C j (ω ) cos( β j (ω ) z )e
k j (ω ) x1
q j (ω ) (16)
j =1

This analytical solution, obtained by using the method of separation of variables, verifies eqs. (5)
and (6) and the radiation condition. C0 (ω ) cosh(α 0 (ω ) z ) and C j (ω ) cos( β j (ω ) z ) represent a
complete set of frequency dependent orthonormal eigenfunctions with the normalising coefficients
16TH CONGRESS OF IABSE, LUCERNE, 2000

1 1
C0 (ω ) = C j (ω ) = (17)
d sinh (2dα 0 (ω )) d sin (2dβ j (ω ))
+ +
2 4α 0 (ω ) 2 4 β j (ω )

The frequency dependent eigenvalues α 0 (ω ) and β j (ω ) are obtained as solutions of equations

ω2 ω2
α tanh(αd ) = β tan( βd ) = (18)
g g
The first term in eq. (16), with the wave number

ω2
k 0 (ω ) = −i α 0 (ω ) +
2
2
(19)
cw

where ‘i’ denotes the imaginary unit, corresponds to a surface wave of period T = 2π ω . For an
incompressible fluid ( cw → ∞ ), the wave number reduces to − iα 0 and the wavelength of the
surface wave equals λ = α 0 2π . Substituting this expression in eq. (18) and expressing the
frequency ω in terms of the period T, lead to the well known implicit relationship relating the wave
length λ to the period T of the surface wave
T 2g 2π
λ= tanh( d ) (20)
2π λ
The terms in eq. (16) under the sign å , with the wave numbers

ω2
k j (ω ) = − β j (ω ) − 2
2
(21)
cw
describe the propagation of the compression waves throughout the unbounded domain.
The modal amplitudes q0 (ω ) and q j (ω ) in eq. (16), can be determined as coefficients of the
expansion of p ∞ ( z , ω ) , ( p ∞ ( x1 , z , ω ) at x1 = 0 ), in the basis of the eigenfunctions previously
defined, that is
d d
q0 (ω ) = C0 (ω ) ò cosh(α 0 (ω ) z ) p ( z ,ω )dz

q j (ω ) = C j (ω ) ò cos( β j (ω ) z ) p ∞ ( z ,ω )dz (22)
0 0

Accordingly to eq. (16), the spatial derivative of p ∞ ( x1 , z , ω ) evaluated at x1 = 0 equals

∂p ∞ ( x1 , z , ω ) ∞
= C0 (ω )k 0 (ω ) cosh(α 0 (ω ) z )q0 (ω ) + å C j (ω )k j (ω ) cos( β j (ω ) z )q j (ω ) (23)
∂x1 j =1

To match the boundary condition of eq. (7) along the boundary µ1 , p ∞ (0, z , ω ) = p b (0, z , ω ) , the
boundary x1 = 0 of the unbounded domain is discretized accordingly to the discretization
introduced for the adjacent bounded domain (see Fig. 1b). This implies that the pressure p ∞ ( z , ω )
has to be expressed as
p ∞ ( z , ω ) = [ N w∞ ( z )]{P1∞ (ω )} (24)

with [ N w∞ ( z )] = [ N wb ( x, z )] on the boundary µ1 .


16TH CONGRESS OF IABSE, LUCERNE, 2000

Applying the principle of virtual work to eq. (23), and making use of eqs. (22) and (24), lead to the
relationship relating the vector of the nodal fluxes {Pn∞1 (ω )} to the vector of the nodal pressures
{P1∞ (ω )} along the boundary µ1 of the unbounded domain
{Pn∞1 (ω )} = [ S11∞ (ω )]{P1∞ (ω )} (25)
with
d d
[ S11∞ (ω )] = − ò [ N w∞ ( z )]T cosh(α 0 (ω ) z )dz (C02 (ω )k 0 (ω )) ò [ N w∞ ( z )] cosh(α 0 (ω ) z )dz
0 0
(26)
∞ d d
− å ò [ N ( z )] cos( β j (ω ) z )dz (C (ω )k j (ω )) ò [ N ( z )] cos( β j (ω ) z )dz

w
T 2
j

w
j =1 0 0

Analogously, applying the same procedure to the unbounded domain lying to the right of the
bounded domain leads to the relationship expressing the nodal fluxes {Pn∞2 (ω )} to the nodal
pressures {P2∞ (ω )} along the boundary µ 7 (see Fig. 1b)

{Pn∞2 (ω )} = [ S 22

(ω )]{P2∞ (ω )} (27)
For the general 3-D fluid-structure interaction problem, the derivation of the nodal pressures-nodal
fluxes relationship of the regular region of the fluid extending to infinite, cannot be derived starting
from an analytical solution. In this case, the more general scaled boundary finite-element method
[1] or, alternatively, the boundary-element procedure [2] can be applied.
In general all methods when applied in the frequency domain lead to a non rational transfer function
[ S ∞ (ω )] that is with elements which cannot be expressed as the ratio of two polynomials in the i ω
variable. This means that when the analysis must be performed, for efficiency, in the time domain
(non-linear structure and anchors), the computation of the nodal fluxes at time t implies the
evaluation of convolution integrals. To avoid convolution and then to increase the efficiency of the
analysis in the time domain, a rational approximation of [ S ∞ (ω )] can be performed [3]. The
resulting approximated rational transfer function of the unbounded domain, is then represented
(realized) with a first or a second order system of ordinary differential equations. In this manner the
unbounded domain is described using a mass matrix, a damping matrix and a stiffness matrix, as is
the case for the generalised structure. After assemblage of the corresponding matrices, the total
system can be solved using a simple time integrator.

4. Coupling of the fluid domains


Accordingly to eqs. (7), (8) and (24), the coupling of the bounded domain to the unbounded domain
along the boundaries µ1 and µ 7 is formulated in terms of nodal pressures, that is

{P1∞ (ω )} = {P1b (ω )} = {P1 (ω )} {P2∞ (ω )} = {P2b (ω )} = {P2 (ω )} (28)


As a consequence of the parallel connection between the representations of the bounded and of the
unbounded domain, (same input in terms of pressures along µ1 and µ 7 ), the total fluxes (output of
the two systems) along µ1 and µ 7 , equal

{Pnu1∞ (ω )} = {Pnu1 (ω )} + {Pn∞1 (ω )} {Pnu2∞ (ω )} = {Pnu2 (ω )} + {Pn∞2 (ω )} (29)


Expressing the fluxes in eq. (29) as described in eqs. (26) and (27) and making use of eq. (14) lead
to the input output relationship of the coupled fluid domains
16TH CONGRESS OF IABSE, LUCERNE, 2000

[ S PP (ω )]{P(ω )} = {Pn (ω )} (30)


with
é[ S11bs ] + [ S11∞ ] [0] [0] [ S14bs ]ù ì{P1}ü ì{Pnu1∞ }ü
ê ∞ bs ú ï{P }ï ï u∞ ï
[ S 22 ] + [ S 22 ] [0]
bs
[ S 24 ]ú ï ï ï{P }ï
[ S PP ] = ê {P} = í 2 ý {Pn } = í n 2b ý (31)
ê [ S 33bs ] bs ú
[ S 24 ] ï{P3 }ï ï {Pn3 } ï
ê bs ú ïî{P4 }ïþ ïî {0} ïþ
êë SYM [ S 44 ]úû

5. Discretization of the platform and coupling with the fluid domain


To derive a finite element discretization of the platform, the weak form of eq. (2) is formulated (for
the weak or variational form of a problem see Ref. [4]). The displacement field w( x3 , ω ) is
approximated in terms of the vector of the nodal displacements {W (ω )} , (vertical displacements
and rotations, see Fig. 1b), and of the shape functions [ N p ( x3 )] defined along the boundary µ 4

w( x3 , ω ) = [ N p ( x3 )]{W (ω )} (32)

To better represent the effects of the pressure p b ( x3 , ω ) acting under the platform, a mesh of the
fluid domain finer than that of the structure is introduced along the boundary µ 4 . This implies that,
between two adjacent nodes of the structural mesh, at least one additional node of the fluid mesh is
introduced.
Again, applying the Galerkin procedure with the virtual field of displacements defined by using the
same shape functions [ N p ( x3 )] of the real field w( x3 , ω ) , and introducing for p b ( x3 , ω ) the
expression of eq. (8) evaluated along the boundary µ 4 , lead to the system of linear equations
describing the vertical motion of the discretized platform
[ SWW (ω )]{W (ω )} = {FW (ω )} (33)
with {FW (ω )} vector of the interaction forces. The matrix [ SWW (ω )] equals

[ SWW (ω )] = (− ρ pω 2 + k w )[ M p ] + EI [ K p ] (34)

[ M p ] and [ K p ] are, respectively, the mass matrix and the stiffness matrix of the platform
Lp Lp

[ M p ] = ò [ N p ( x3 )] [ N p ( x3 )]dx3
T
[ K p ] = ò ([ D][ N p ( x3 )]) T [ D][ N p ( x3 )]dx3 (35)
0 0

with [D] the diagonal matrix containing the partial derivative operators ∂ 2 ∂x32 and ∂ 2 ∂z 2 .
The vector {FW(ω)} of the interaction forces can be written as
Lp

{FW {ω}} = ò [ N p ( x3 )] p b ( x, z , ω ) µ3
dx3 = [ SWP ]{P(ω )} (36)
0

where
LP

[ SWP ] = ò [ N p ( x3 ]T [ N wb ( x, z ] dx3 (37)


µ3
0
16TH CONGRESS OF IABSE, LUCERNE, 2000

As an alternative, the vertical displacement function w( x3 , ω ) can be expanded in the basis of the
eigenvectors associated to the homogeneous form of eq. (2). Following a procedure analogous to
that previously illustrated leads to the same symbolic equation (33) but with [ SWW (ω )] as a diagonal
matrix of dimension equal to the number of modes selected in the analysis and {W (ω )} the
corresponding vector of the modal amplitudes. This alternative is not presented in detail because it
is easily applicable only when the platform exhibits uniform inertia and no anchors are introduced.
The equilibrium equation of the platform in horizontal direction can be easily written as
Suu (ω )u (ω ) = Fu (ω ) (38)

with Suu (ω ) = − ρ p L pω 2 and for the horizontal interaction force Fu (ω )


Le Le

Fu (ω ) = òp ( x, z , ω ) dz − ò p b ( x, z , ω ) dz = [ S uP ]{P(ω )}
b
(39)
µ3 µ5
0 0

where
Le Le

[ S uP ] = ò [ N ( x, z )] dz − ò [ N wb ( x, z )] dz
b
w (40)
µ3 µ5
0 0

Actually, the non-vanishing terms in the matrices [ SWP ] and [SuP] are those corresponding to the
nodal pressures on the boundary µ 4 , µ 3 and µ 5 , that is, to the vector {P3 (ω )} only. More in detail,
the hypothesis Le << L p , implies negligible bending moments at the ends of the platform caused by
the distribution of pressures along the boundaries µ 3 and µ 5 . As a consequence, no coupling
between the pressures of nodes lying on such boundaries (not belonging to the boundary µ 4 ) and
the displacement vector {W (ω )} exists.
Linear anchors, with stiffness R j acting on the vertical degree of freedom W j of the platform at an
angle α j with respect to the bottom boundary µ 8 , can be easily introduced in the system, adding
the contributions R j sin (α j ) and R j cos(α j ) to the diagonal term SWWjj of [ SWW ] in eq. (33) and to
Suu in eq. (38), respectively.
The vector {Pnb3 (ω )} in eq. (31), represents the nodal fluxes along the boundaries µ 3 , µ 4 and µ 5
associated to the motion of the platform. For their computation, eq. (32) is substituted in eq. (4) and
eqs. (4) and (3) are introduced in eq. (10). The latter is then evaluated along the boundaries µ 3 , µ 4
and µ 5 . It is straightforward to demonstrate that the resulting equation can be written as

ì {0} ü
ï {0} ï
[ ì{W }ü
{Pnp } = ρ wω 2 [ S PW ]T [ S Pu ]T í ý ] with
ï
{Pnp } = í b ý
ï
(41)
î u þ ï{Pn3 }ï
ïî {0} ïþ

Decomposing the vector {Pn (ω )} in eq. (30) as


16TH CONGRESS OF IABSE, LUCERNE, 2000

ì{Pnb1∞ }ü
ï b∞ ï
ï{P }ï
{Pn (ω )} = {Pnb∞ (ω )} + {Pnp (ω )} with b∞
{Pn } = í n 2 ý (42)
ï {0} ï
ïî {0} ïþ

and combining the resulting equation with eqs. (33) and (38) both premultiplied by ρ wω 2 , lead to
the symmetric linear system of equations of the unknown nodal pressures {P(ω )} of the fluid
domain and of the nodal displacements {W (ω )} and u (ω ) of the platform, that is

é[ S PP ] − ρ wω 2 [ SWP ]T − ρ wω 2 [ S uP ]T ù ì {P} ü ì{Pnb∞ }ü


ê úï ï ï ï
ê ρ wω 2 [ SWW ] [ 0] ú í{W }ý = í {0} ý (43)
êSYM
ë ρ wω 2 S uu úû ïî u ïþ ïî 0 ïþ

6. Left hand side term


The left hand side term of eq. (43) can be evaluated as specified in Ref. [5], that is starting from the
free field solution of the fluid domain subjected to the same surface wave of the coupled system.
For perturbations propagating from the left to the right of the platform, the free field solution equals
the nodal pressures {P1 f (ω )} induced by the surface wave of amplitude Am along the boundary µ1
of the uniform fluid domain without platform (fluid channel of constant depth d)
cosh(α 0 (ω ) z j )
P1 fj (ω ) = Am ρ w g (44)
cosh(α 0 (ω )d )
Following Ref. [5], it results that the non-vanishing part of the left hand side of eq. (43) (see also
eq. 42) equals
{Pnb1∞ (ω )} = 2[ S11∞ (ω )]{P1 f (ω )} {Pnb2∞ (ω )} = {0} (45)

7. Results
1.6 To validate the software developed by the author,
dimensionless disp. at bow

pitch
1.4 2-D many comparisons with experimental data and
1.2 results computed and published by different
3-D
1.0 authors were accomplished. As an example, the
0.8 displacement at the bow of a floating platform of
0.6 heave length 200 m, width L p = 500m and draft
0.4 Le = 20m , with EI =4.92 109 kNm and water
0.2 depth d=150m, is computed selecting in the 2-D
0.0 analysis, respectively, 80 and 11 fluid elements
8 10 12 14 16 18 20 22 24 26
along the boundaries µ 4 and µ1 . The results are
wave period (sec) compared in Fig. 2 (solid lines) with those of the
3-D analysis reported in [6] (dashed lines).
Fig. 2 Vertical displacement amplitude
16TH CONGRESS OF IABSE, LUCERNE, 2000

1.2 1.0
dimensionless max. moment
Nstif=0.00289 Nstif=0.00289

dimensionless max. disp.


1.0 0.8
Nstif=0.0289 Nstif=0.0289
0.8
Nstif=0.1445 0.6 Nstif=0.1445
0.6
0.4
0.4

0.2 0.2

0.0 0.0
0.5 1.0 1.5 2.0 2.5 3.0 0.5 1.0 1.5 2.0 2.5 3.0
dimensionless wave period dimensionless wave period
Fig. 3 M d for Nd=5 (without anchors) Fig. 4 Wd for Nd=5 (without anchors)

1.2 1.0
dimensionless max. moment

Nd=5 Nd=5
1.0 dimensionless max. disp. 0.8
Nd=10 Nd=10
0.8
Nd=20 0.6 Nd=20
0.6
0.4
0.4

0.2 0.2

0.0 0.0
0.5 1.0 1.5 2.0 2.5 3.0 0.5 1.0 1.5 2.0 2.5 3.0
dimensionless wave period dimensionless wave period

Fig. 5 M d for Nstif=0.00289 (without anchors) Fig. 6 Wd for Nstif=0.00289 (without anchors)

The displacement is normalised with respect to amplitude Am of the regular incident wave and is
plotted for the heave and the pitch motions of the platform as a function of the period T of the
incident wave. Even if the width L p of the platform is larger than the length, an acceptable
agreement between the results of the two analyses exists.
To better grasp the influence of the different geometric and mechanical properties of the platform
on the structural behaviour, some dimensionless parameters are introduced. In the following, all
results will be expressed as functions of the dimensionless wave period Td , which is defined as the
ratio of the period T of the incident wave to the period Tr = 2π ρ p k w of the rigid body motion
of the platform on an elastic spring bed of constant k w . To compare the transversal stiffness of the
platform EI L4p , with the reduced hydrostatic restoring force coefficient k w , the dimensionless
parameter Nstif= EI L4p k w is introduced. For the normalisation of the depth d of the water, the ratio
Nd= d Am is used.
The behaviour of a platform without anchors, of width L p = 90m , draft Le = 1m , C red = 1 , excited
by regular waves of amplitude Am and period T , is addressed in Figs. 3 and 4 for different values of
16TH CONGRESS OF IABSE, LUCERNE, 2000

1.2 1.0
dimensionless max. moment
Nstif=0.00289 Nstif=0.00289

dimensionless max. disp.


1.0 0.8
Nstif=0.0289 Nstif=0.0289
0.8
Nstif=0.1445 0.6 Nstif=0.1445
0.6
0.4
0.4

0.2 0.2

0.0 0.0
0.5 1.0 1.5 2.0 2.5 3.0 0.5 1.0 1.5 2.0 2.5 3.0
dimensionless wave period dimensionless wave period

Fig. 7 M d for Nd=5 (with anchors) Fig. 8 Wd for Nd=5 (with anchors)
the parameter Nstif, with Nd=5, and in Figs. 5 and 6 for different ratios Nd, with Nstif=0.00289.
The analysis is performed by introducing along the boundary µ 4 12 structural finite elements with
8 fluid elements between two adjacent nodes of the platform and, along the boundary µ1 7 and 12
fluid elements for Nd=5 and Nd ≥ 10 , respectively.
In Figs. 3 and 5, the maximum M of the amplitude of the moment along the platform is normalised
with the yelding moment per unit depth M y , ( M d = M M y ), and plotted as a function of Td . For
Nstif=0.00289, M y is assumed to be equal to the maximum value M 1 registered in the analysis.
This case corresponds to a flexible platform with EI = 1.86 106 kNm. For all others values of Nstif,
the corresponding yielding moment M y is computed starting from M 1 and with hypothesis that the
area of the cross-section of the platform is the same for all Nstif. In Figs. 4 and 6, the maximum
W of the amplitude of the vertical displacement along the width of the platform is normalised with
the amplitude Am of the incident wave (Wd =W/Am). The peaks in the figures correspond to resonant
frequencies of the total system. Their amplitudes are reduced when the analysis is performed taking
into account the structural damping of the platform. The location of the peaks moves to the right
when, for a fixed value of Nd, the stiffness of the platform is increased or when, for a fixed value of
Nstif, the depth d of the water is reduced. In both cases, this can be interpreted as an increasing of
the mass of the fluid involved in the motion of the platform, which lead to a reduction of the values
of the resonant frequencies. As expected an increasing of the stiffness of the platform produces a
reduction of both the normalised moment M d and the normalised displacement Wd . The decreasing
is more important for Wd . For very large periods Td (not included in the pictures), the dynamic
interaction reduces to the simple hydrostatics with M d → 0 and Wd → Am .
To improve the performance of the platform, 5 anchors at an angle of α = π / 4 with respect to the
boundary µ 8 , spaced at a distance 30m apart, with stiffness R = 0.25 104 kN/m, are connected to
the structural nodes of the previously defined mesh. The results of the analysis are potted in Figs. 7
and 8. From the comparison of Fig. 3 with Fig. 7, it results that for a flexible platform, the
normalised moment M d is hardly reduced by the introduction of the anchors. To reduce M d in this
case, an increasing of the stiffness of the platform is necessary. From the comparison of Fig. 4 with
Fig. 8, it follows that anchors are very efficient in limiting the displacements of platforms. As
shown in the figures, their benefit reduces with the increasing of the stiffness of the platform.
16TH CONGRESS OF IABSE, LUCERNE, 2000

8. Conclusions
Oscillations of floating platforms, to be used as temporary surfaces, can be properly studied in the
frequency domain.
In the evaluation of the response of such structures, a dynamic fluid-structure interaction problem
must be solved.
For platforms with the length in axial direction much larger than the width, the main characteristics
of the problem can be grasped solving a 2-D fluid-structure interaction problem.
The structural efficiency of a floating platform is strongly related to the ratio Nstif of transversal
stiffness of the platform EI L4p to the hydrostatic restoring force coefficient k w = ρ w gC red .
To limit the bending moments along the width of the platform, Nstif must be selected adequately
large. For instance, when the width L p of the platform is increased, the bending stiffness EI per
unit depth of the platform in axial direction must be increased accordingly to L4p . Improvements of
limited benefit can be obtained reducing C red , that is, the ratio of the floating surface A flo to the
geometric surface Age om of the platform.
For a selected Nstif, the amplitude of the displacements can be controlled by introducing anchors
along the width of the platform. The efficiency of the anchors decreases with the increasing of
Nstif.

8.1 References
[1] WOLF J.P. and SONG C., Finite-Element Modelling of Unbounded Media, Wiley, Chichester,
1996.
[2] BREBBIA, C.A., TELLES, J.C.F. and WROBEL L.C., Boundary Element Techniques,
Springer, Berlin, 1984.
[3] PARONESSO A., Rational approximation and realization of generalized force-displacement
relationship of an unbounded medium, Ph.D. thesis, Laboratory of Hydraulic Constructions,
Department of Civil Engineering, Swiss Federal Institute of Technology, Lausanne, February
1997.
[4] HUGHES, T.J.R., The Finite Element Method, Prentice-Hall, Englewood Cliffs, NJ, 1987.
[5] WOLF J.P., Dynamic Soil-Structure Interaction, Prentice-Hall, Englewood Cliffs, NJ, 1985,
pp. 18.
[6] WANG S. and ERTEKIN R.C., “Hydroelastic-Response Analysis of a Box-Like Floating
Airport of Shallow Draft”, Proceedings of the Fifth (1995) International Offshore and Polar
Engineering Conference, Vol. 1, The Hague, The Netherlands, June 11-16, 1995, pp. 145-152.

You might also like