SeaFEM Theory Ma SeaFEM Theory Manual Theory ... - Compass
SeaFEM Theory Ma SeaFEM Theory Manual Theory ... - Compass
SeaFEM Theory Ma SeaFEM Theory Manual Theory ... - Compass
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
1. Table of contents<br />
1. Table of contents ....................................................................................................... 2<br />
2. Introduction ............................................................................................................... 3<br />
3. Problem statement ..................................................................................................... 4<br />
Governing equations ..................................................................................................... 4<br />
Velocity potential decomposition ................................................................................. 4<br />
Radiation condition and wave dissipation .................................................................... 5<br />
Free surface potential flow problems with currents ..................................................... 6<br />
4. Numerical formulation .............................................................................................. 8<br />
Free surface boundary condition .................................................................................. 8<br />
Stremalines scheme for the free surface boundary condition ....................................... 9<br />
SUPG stabilization scheme for the free surface boundary conditions ....................... 12<br />
Free surface kinematic boundary condition................................................................ 15<br />
Free surface boundary condition with absorption ...................................................... 19<br />
Radiation boundary condition .................................................................................... 19<br />
Body boundary condition ........................................................................................... 20<br />
Boundary condition for limit height (Hfs) .................................................................. 20<br />
Boundary condition for transom stern ........................................................................ 21<br />
5. Multi-body dynamics .............................................................................................. 22<br />
Body dynamics ........................................................................................................... 22<br />
Non-linear hydrostatics............................................................................................... 23<br />
Body links ................................................................................................................... 23<br />
7. Body matrices referred to an arbitrary generic point .............................................. 25<br />
8. Statistical description of an irregular sea ................................................................ 27<br />
Spectrum discretization .............................................................................................. 27<br />
Convergence ............................................................................................................... 28<br />
Spectral moments ....................................................................................................... 28<br />
Waves spectrum .......................................................................................................... 29<br />
9. Mooring system modelling...................................................................................... 31<br />
Catenary equations ..................................................................................................... 31<br />
Elastic catenary formulation ....................................................................................... 33<br />
Dynamic cable formulation ........................................................................................ 36<br />
10. Morison’s forces .................................................................................................. 41<br />
11. Response Amplitude Operators (RAOs) ............................................................. 43<br />
12. Fluid-Structure interaction algorithm .................................................................. 44<br />
13. References ........................................................................................................... 45<br />
14. Glossary ............................................................................................................... 47
3 Introduction<br />
2. Introduction<br />
<strong>SeaFEM</strong> deals with real problems regarding ocean waves interacting with floating<br />
structures. These sorts of problems can be simulated using potential flow theory along<br />
with Stokes perturbation approximation. In this sense, <strong>SeaFEM</strong> is a time domain solver<br />
based on the finite element method which is capable of solving multi-body seakeeping<br />
problems on unstructured meshes. Since it is based on Stokes wave theory, no remeshing<br />
or moving mesh technique are needed, which keeps computational costs and<br />
computational times low. The algorithm has been adapted to include non-linear external<br />
forces, like those used to define mooring systems.
4 Problem statement<br />
3. Problem statement<br />
Governing equations<br />
In <strong>SeaFEM</strong>, the first order diffraction-radiation problem of a set of floating bodies is<br />
considered:<br />
∇ = 0<br />
+ + = 0<br />
<br />
= <br />
− = 0 = (1)<br />
∇ ∙ n = v ∙ n <br />
φ = 0<br />
<br />
= −<br />
where ∇ = ( , ) is the gradient in the horizontal plane, Ω is the fluid domain,<br />
Γ represents the wetted surface of the ship, is the water depth, v is the local ship<br />
velocity of a point over the wetted surface, n is the normal of the ship wetted surface<br />
pointing outwards the ship, and is the free surface pressure. The domain is assumed to<br />
be infinite in the horizontal directions.<br />
Velocity potential decomposition<br />
The aim of <strong>SeaFEM</strong> is to simulate the dynamics of a set of floating bodies subjected to<br />
the action of waves. To do so, the environment is modelled as the sum of a number of<br />
airy waves. This can be expressed in terms of a velocity potential given by:<br />
cosh(| |( + ))<br />
= cos(|<br />
cosh(| |)<br />
|( + − + )<br />
<br />
(2)<br />
where are the wave amplitudes; are the wave frequencies; are the wave<br />
numbers; are the wave directions; are the wave phases. From this point on, we<br />
will refer to as the incident potential. This potential, along with the dispersion<br />
relation = | | tanh(| |) , fulfils eq. (1), and therefore is solution of the<br />
mathematical model in the absence of bodies.<br />
In order to obtain the solution to the governing equations, a velocity potential<br />
decomposition is used. Let be the solution to the governing equations. Then can be<br />
decomposed as = + and = + , where represents the velocity potential of<br />
waves diffracted and radiated by the bodies.<br />
Introducing the velocity potential decomposition into the governing equations the<br />
system of equations to be fulfilled by is obtained:<br />
∇ = 0<br />
+ + = 0<br />
<br />
=
5 Problem statement<br />
− = 0 = (3)<br />
∇ ∙ n = (v − ∇) ∙ n <br />
= 0<br />
<br />
= −<br />
Hence, the purpose is to find for a given incident potential and for a given v .To this<br />
aim, eqs. () will be solved in a finite domain by means of the finite element method.<br />
Once the velocity potential has been obtained, the pressure at any point can be<br />
calculated from:<br />
= −( + ∙ ∇ + ∇ ∙ ∇ + ∇ ∙ ∇ + ) (4)<br />
Radiation condition and wave dissipation<br />
Waves represented by φ are born at the bodies and propagate in all directions away<br />
from the bodies. These waves have to either be dissipated or to be let go out the domain<br />
so they will not come back and interact with the bodies. Because of this, a Sommerfeld<br />
radiation condition at the edge of the computational domain is introduced:<br />
∂<br />
tφ<br />
+ c∇φ<br />
⋅ n<br />
R<br />
= 0 on Γ<br />
R<br />
(5)<br />
where Γ<br />
R<br />
is the surface limiting the domain in the horizontal directions, and c is a<br />
prescribed wave velocity. eq. (5) will let waves moving at velocity c to escape out the<br />
domain. However, waves with very different velocities will not be leaving the domain.<br />
Hence, wave dissipation is also introduced into the dynamic free surface boundary<br />
condition by varying the pressure such that:<br />
P / ρ = κ(<br />
x) ∂ φ<br />
(6)<br />
a<br />
z<br />
where κ ( x) is a damping coefficient. eq. (6) increases pressure when the free surface is<br />
moving upwards, while decreases the pressure when the free surface is moving<br />
downwards. By doing this, energy is transferred from the waves to the atmosphere and<br />
waves are damped. However, the coefficient κ ( x) will be set to zero in the area nearby<br />
the bodies so that damping will have no effect on the solution of the wave structure<br />
interaction problem.<br />
Combining the dynamic and kinematic boundary conditions, introducing eq. (5), and<br />
adding eq. (6), and choosing C ' = 0 , the governing equations for φ become:<br />
φ<br />
2<br />
∇ = 0<br />
in Ω<br />
∂ φ = −g∂ φ −κ(<br />
x) ∂ ∂ φ<br />
on z = 0<br />
tt z t z<br />
∂ zφ = 0<br />
on z = − H<br />
( ψ )<br />
∇φ<br />
⋅ = −∇ ⋅<br />
nB vB n<br />
B<br />
on Γ<br />
B<br />
∂<br />
tφ<br />
+ c∇φ<br />
⋅ n<br />
R<br />
= 0<br />
on Γ<br />
R<br />
(7)
6 Problem statement<br />
η = − 1 Pa<br />
t<br />
g<br />
∂ φ − ρ g<br />
on z = 0<br />
where the free surface elevation has been decoupled from the problem of obtaining the<br />
velocity potential.<br />
Free surface potential flow problems with currents<br />
<strong>SeaFEM</strong> is also capable of solving free surface potential flow problems in the presence<br />
of water currents. Within this context, the governing equations for the first order<br />
diffraction-radiation wave problem are:<br />
∇ = 0<br />
+ · ∇ + 1 2<br />
∇ · ∇ + + = 0<br />
+ ( · ∇ ) · ∇ − = 0<br />
( + ∇) · = · <br />
= 0<br />
in Ω<br />
in z = 0<br />
in z = 0<br />
in Γ b<br />
in z = -H<br />
(8)<br />
where ∇ is the gradient in the horizontal plane, is the local body velocity of a point<br />
over the wetted surface, and is the water current. In this case, once the velocity<br />
potential is obtained the pressure field can be calculated straightforward using the<br />
Bernouilli equation:<br />
= − + · ∇ + ∇ <br />
+ (9)<br />
2<br />
As in the case without currents, the problem is solved by first decomposing the velocity<br />
potential and the free surface elevation as follows:<br />
= + <br />
= + (10)<br />
where the first terms in the right hand side account for scattered wave components and<br />
the second terms concern incident waves.<br />
By doing this, it is possible to split the governing equations into two sets of equations.<br />
The first set of equations reads as follows:<br />
∇ = 0<br />
+ · ∇ + = 0<br />
+ · ∇ − = 0<br />
= 0<br />
in Ω<br />
in z = 0<br />
in z = 0<br />
in z = -H<br />
(11)<br />
and has Airy waves transported by a uniform current as analytical solution.
7 Problem statement<br />
cosh(| |( + ))<br />
= cos(<br />
cosh(| |)<br />
( − ) − + )<br />
<br />
(12)<br />
= ( · ( · ) − + )<br />
<br />
The second set of equations concerns the scattered wave’s potential and, after<br />
neglecting second order terms, reads as follows:<br />
∇ = 0<br />
+ · ∇ + 1 2<br />
∇ · ∇ + ∇ · ∇ + + = 0<br />
+ ( · ∇ ) · ∇ + ∇ · ∇ − = 0<br />
∇ · = ( − − ∇) · <br />
= 0<br />
in Ω<br />
in z = 0<br />
in z = 0<br />
in Γ b<br />
in z = -H<br />
(13)<br />
Note that in the presence of water currents, the velocity potential and the free surface<br />
elevation problems cannot longer be decoupled. Notice also that the terms ∇ · ∇ <br />
and ∇ · ∇ account for the deviation of the incident Airy waves due to the fact that<br />
incident waves are transported by a non-uniform flow field.<br />
Within the approach used in <strong>SeaFEM</strong>, it is convenient from a numerical point of view to<br />
solve the governing equations in a frame of reference fix to the floating body rather than<br />
on the global frame of reference. For an observer moving with the floating body, the<br />
flow field around the body will be:<br />
() = − () (14)<br />
and the governing equations in the local frame of reference become:<br />
∇ = 0<br />
+ · ∇ + 1 2<br />
∇ · ∇ + ∇ · ∇ + + = 0<br />
+ ( · ∇ ) · ∇ + ∇ · ∇ − = 0<br />
∇ · = (− − ∇) · <br />
= 0<br />
in Ω<br />
in z = 0<br />
in z = 0<br />
in Γ b<br />
in z = -H<br />
(15)<br />
where the incident wave potential and incident wave elevation must be transformed to<br />
the local frame of reference.
8 Numerical formulation<br />
4. Numerical formulation<br />
This section presents the formulation based on the finite element method (FEM) to<br />
solve the system of equations presented in sections 2.1 – 2.3. This formulation has been<br />
developed to be used in conjunction with unstructured meshes. The use of unstructured<br />
meshes enhances geometry flexibility and speed ups the initial modelling time.<br />
*<br />
Let Q<br />
h<br />
be the finite element space to interpolate functions, constructed in the usual<br />
manner. From this space, we can construct the subspace Qh,<br />
φ<br />
, that incorporates the<br />
Dirichlet conditions for the potential φ . The space of test functions, denoted by Q<br />
h<br />
, is<br />
constructed as Qh,<br />
φ<br />
, but with functions vanishing on the Dirichlet boundary. The weak<br />
form of the problem can be written as follows:<br />
Q φ<br />
φ ∈ , by solving the discrete variational problem:<br />
Find [ h ] h,<br />
∇ · ∇ Ω =<br />
Ω<br />
= · Γ + · Γ +<br />
<br />
· <br />
Γ +<br />
<br />
∈ <br />
where<br />
) φ ,<br />
) Z0<br />
φ , ) )<br />
H<br />
φ and φ Z−<br />
B<br />
n<br />
R<br />
n<br />
n<br />
n<br />
<br />
· <br />
<br />
Γ<br />
<br />
∀<br />
<br />
(16)<br />
are the potential normal gradients corresponding to the<br />
Neumann boundary conditions on bodies, radiation boundary, free surface and bottom,<br />
respectively.<br />
At this point, it is useful to introduce the associated matrix form<br />
L b b b b<br />
B R Z0 Z−<br />
H<br />
φ = + + +<br />
(17)<br />
B R Z0<br />
Z H<br />
where L is the standard laplacian matrix, and b , b , b and b −<br />
are the vectors<br />
resulting of integrating the corresponding boundary condition terms. Regarding the<br />
bottom boundary for the refracted and radiated potential, it is imposed naturally in FEM<br />
H<br />
by b Z − = 0 .<br />
Free surface boundary condition<br />
Solving the velocity potential free surface boundary condition efficiently is the most<br />
important point of the problem stated since this is where a difference is made when<br />
solving the mathematical model in eqs. (7) using the FEM.<br />
The free surface conditions can be rewritten as:
9 Numerical formulation<br />
∂<br />
tη + U ⋅∇<br />
hη + ∇hφ ⋅∇hξ − φz<br />
= 0<br />
123<br />
Convective term<br />
1<br />
∂<br />
tφ + U⋅∇hφ −<br />
2<br />
∇hφ ⋅∇<br />
hφ + ∇hψ ⋅∇<br />
hφ + P / ρ + gη<br />
= 0<br />
123<br />
Convective term<br />
(18)<br />
where = ( + ∇ ) is the base flow. Linearization respect to this based flow is<br />
quite common. Linearizations most commonly used are: the Kelvin linearization, which<br />
assumes that ∇ hφ ~ ε and hence = and the double body, which assumes that<br />
DB *<br />
DB<br />
∇<br />
hφ = ∇<br />
hφ + ∇<br />
hφ<br />
, where ∇<br />
hφ<br />
~ O(1)<br />
and<br />
* DB<br />
∇<br />
hφ<br />
~ O( ε ) , being ∇ hφ the<br />
scattered velocity potential obtained when solving the double body problem. As we<br />
mentioned before, in this work no linearization is assumed since it is considered that<br />
∇<br />
hφ<br />
~ O(1)<br />
. Therefore, the convective velocity U will be different in each time step,<br />
and the numerical scheme will be adapted accordingly. Three main issues must be kept<br />
in mind:<br />
1. Convective terms are likely to introduce numerical dispersion leading to<br />
unrealistic free surfaces.<br />
2. Pressure field over the free surfaces must be imposed in order to be capable of<br />
reproducing aircushion effects.<br />
3. Scattered waves must be absorbed in the far field in order to mimic an infinite<br />
domain in the horizontal directions.<br />
The numerical schemes adopted for solving the kinematic-dynamic free surface<br />
boundary conditions are based on Adams-Bashforth-Moulton schemes, using an explicit<br />
scheme for the kinematic condition, and implicit one for the dynamic condition. Then<br />
n 1<br />
can be imposed as a Dirichlet Boundary condition. The schemes read as follows:<br />
φ +<br />
( U )<br />
n+<br />
1 n n<br />
n n n<br />
= − ∆t ⋅∇h − ∆t∇h ⋅∇<br />
h<br />
+ ∆t<br />
z<br />
η η η φ ξ φ<br />
n+<br />
1/ 2<br />
1<br />
n<br />
( U<br />
h ) 2 ( h h ) h h ( / )<br />
φ = φ − ∆t ⋅∇ φ + ∆t ∇ φ ⋅∇ φ − ∆t∇ ψ ⋅∇ φ − ∆ t P ρ + gη<br />
n+ 1 n n n n+ 1 n+<br />
1<br />
(19)<br />
It is noted that conditions of pressure acting on the free surface can be introduced in eq.<br />
(19) straightforward.<br />
Stremalines scheme for the free surface boundary condition<br />
It was pointed out before that care must be taken when solving the free surface<br />
conditions, especially when evaluating the convective terms. In this work, the<br />
convective term is obtained by differentiating along streamlines:<br />
( U η )<br />
h<br />
n n n<br />
U<br />
Lη<br />
⋅∇ = ∂<br />
n+ ( ) 1/ 2 n+ 1 n n+<br />
1<br />
U ⋅∇ φ = V + ∇ φ ∂ φ<br />
h b h L<br />
(20)<br />
where ∂<br />
L<br />
denotes the derivative along the streamline. This streamline derivatives is<br />
estimated using a two points upstream and one point downstream differential operator
10 Numerical formulation<br />
inspired by the quickest scheme [18]. Figure 1 shows the tracing of the streamline at<br />
node C. The left (L) and forward left (FL) points are the upstream points, while the right<br />
(R) point corresponds to the downstream point. The values of the scattered velocity<br />
potential φ and scattered free surface elevation η at L, FL and R points are obtained by<br />
linear interpolation between the nodes of the edges where they lie on. The stream line<br />
differential operator reads as:<br />
∂ φ ≈ δ φ = α φ + α φ + α φ + α φ<br />
n n<br />
L L R R C C L L FL FL<br />
∂ η ≈ δ η = α η + α η + α η + α η<br />
n n<br />
L L R R C C L L FL FL<br />
(21)<br />
where the stencils are:<br />
α<br />
α<br />
2 ⎛ 1 1 1<br />
∆x<br />
R<br />
R<br />
= ⎜ − CrR<br />
− −<br />
∆x R<br />
+ ∆xL 2 2 3 ∆ xR + ∆xL<br />
⎝<br />
2<br />
( 1 CrR<br />
)<br />
2 ⎛ 1 1 1 ∆x<br />
1<br />
∆x<br />
2 L<br />
2<br />
( ) ( )<br />
R<br />
C<br />
= ⎜ CrR + CrL + 1- CrR + 1-CrL<br />
∆x R<br />
+ ∆xL 2 2 3 ∆x L<br />
3 ∆x L<br />
+ ∆xFL<br />
⎝<br />
2 ⎛ 1 1 1 ∆x<br />
1 ∆x<br />
αL<br />
= − + + +<br />
∆x + ∆x ⎝ 2 2 3 ∆x 3 ∆x ( ∆ x + ∆x )<br />
2<br />
L<br />
⎜ CrL CrL CrL<br />
R L FL L R L<br />
αFL<br />
2<br />
2 ⎛ 1 ∆xL<br />
⎜<br />
x<br />
R<br />
+ xL ⎝ 3 x<br />
FL( x<br />
L<br />
+ ∆x FL<br />
)<br />
Cr = U ∆t / ∆x<br />
α<br />
2<br />
= ∆ ∆ ∆ ∆<br />
( 1-Cr L )<br />
α<br />
⎞<br />
⎟<br />
⎠<br />
2 R<br />
2<br />
( 1- ) ( 1- )<br />
⎞<br />
⎟<br />
⎠<br />
⎞<br />
⎟<br />
⎠<br />
⎞<br />
⎟<br />
⎠<br />
(22)<br />
Stream line<br />
FLx<br />
o<br />
fl1<br />
o<br />
fl2 l2<br />
r2<br />
V<br />
V<br />
o<br />
Lx<br />
o<br />
l1<br />
V<br />
V<br />
o<br />
o<br />
C<br />
o<br />
x<br />
V<br />
o<br />
x<br />
R<br />
o<br />
r1<br />
V<br />
V<br />
∆XFL<br />
∆XL<br />
φFL<br />
φL<br />
φC<br />
Figure 1: Streamline discretization.<br />
∆XR<br />
φR
11 Numerical formulation<br />
Above equations are used to integrate the free surface boundary conditions when the<br />
convective terms are dominant. However, when these terms can be neglected, the free<br />
surface condition can be simplified. For the sake of clarity, in the following, details of<br />
the integration of the free surface boundary conditions are given for this simplified case.<br />
Neglecting higher order terms, eq. (18) can be rewritten as:<br />
− = 0<br />
+ = 0<br />
(23)<br />
Or equivalently,<br />
1<br />
− = 0<br />
+ = 0<br />
(24)<br />
For efficiency and flexibility reasons, <strong>SeaFEM</strong> uses a forth order compact Padé scheme<br />
in order to solve this problem. This scheme is implicit with symmetric stencils, which<br />
provides good stability properties and requires solving the linear system in eq. (17) once<br />
per time step.<br />
Although the free surface boundary condition is usually implemented as a Dirichlet<br />
boundary condition by imposing the value of the velocity potential at the time step to be<br />
calculated, in <strong>SeaFEM</strong> it is implemented as a Neumann boundary condition that fulfils<br />
the flux boundary integral:<br />
b = M φ (25)<br />
Z0 Z<br />
Z 0<br />
0<br />
Γ z<br />
where<br />
Z<br />
M Γ 0 is the corresponding boundary mass matrix. Rather than obtaining the<br />
Z0<br />
Z0<br />
vector φ<br />
z<br />
and calculating the values of b , <strong>SeaFEM</strong> proceeds in a different manner.<br />
Let’s consider the free surface boundary condition outside the absorbing zone (where<br />
the absorbing factor equals zero, which is inside the analysis area). The forth order<br />
compact Padé scheme reads, for the second eq. in (17) as:<br />
n+ 1 n n−1<br />
φ − 2φ + φ<br />
n 1<br />
n n n<br />
= −g∂ 2<br />
zφ − g ∂<br />
zφ − ∂<br />
zφ + ∂<br />
zφ<br />
∆t<br />
12<br />
+ 1 −1<br />
( 2<br />
)<br />
(26)<br />
Introducing Taylor series expansion around time t in eq. (26) and using eq. (17), we<br />
4<br />
recover the free surface boundary condition with O( ∆ t ) . Eq. (26) is an implicit scheme<br />
which has to be solved along with the linear system given in eq (17). At first sight, it<br />
seems like an iterating procedure should be used requiring solving the linear system<br />
n 1<br />
several times per time steps. However, this can be avoided by decoupling and<br />
n 1<br />
n 1<br />
∂ . To this aim, from eq. (26) φ +<br />
z<br />
zφ +<br />
n 1<br />
is written as a function of φ + :<br />
φ +<br />
12<br />
∂ φ = −10∂ φ − ∂ φ − φ − 2φ + φ<br />
g∆t<br />
( )<br />
n+ 1 n n− 1 n+ 1 n n−1<br />
z z z<br />
2<br />
(27)
12 Numerical formulation<br />
Z0<br />
This approximation is used to evaluate ) φ<br />
n 1<br />
t +<br />
can be calculated as follows:<br />
z<br />
at<br />
n 1<br />
t + Z0<br />
, and therefore, the integral of<br />
b at<br />
⎡<br />
12<br />
⎤<br />
φ ⎢ φ φ φ φ φ<br />
g∆t<br />
⎥ (28)<br />
⎣<br />
⎦<br />
Z<br />
n+ 1<br />
0 Z<br />
n+ 1<br />
0 Z<br />
n<br />
0 Z<br />
n−1<br />
0<br />
n+ 1 n n−1<br />
Z<br />
( ) ( ) 10( ) ( ) ( 2 )<br />
0 Z<br />
b = MΓ<br />
φz = MΓ<br />
− φ 0<br />
z<br />
− φz<br />
− φ − φ +<br />
φ<br />
2<br />
Introducing eq. (28) into Eq. (17) we obtain:<br />
+ 12<br />
∆ <br />
<br />
= 12<br />
∆ (2 − ) − 10 <br />
<br />
− <br />
<br />
<br />
+ ( ) + ( ) <br />
(29)<br />
Eq. (29) imposes a strong coupling between the free surface boundary condition and the<br />
Laplace equation. This is carried out by modifying the system matrix L .<br />
n 1<br />
Once the system is solved, ∂ zφ + at the free surface is obtained using eq. (27) Then,<br />
whenever the velocity potential is solved at the present time step, the free surface<br />
elevation is computed by means of η = −(1/ g) ∂ t<br />
ϕ using the following fourth order<br />
finite difference scheme:<br />
1 ⎛ 25 4 1<br />
η = − ⎜ ϕ − 4ϕ + 3ϕ − ϕ + ϕ<br />
g∆t<br />
⎝ 12 3 4<br />
n+ 1 n+ 1 n n−1 n−2 n−3<br />
⎞<br />
⎟<br />
⎠<br />
(30)<br />
SUPG stabilization scheme for the free surface boundary conditions<br />
Alternatively, a SUPG stabilization scheme is also available in <strong>SeaFEM</strong> for the<br />
integration of the free surface boundary conditions. By using this FEM based<br />
stabilization scheme, the dynamic and kinematic free surface boundary conditions are<br />
solved as follows.<br />
P<br />
v 0<br />
(31)<br />
2 ρ<br />
∂ φ<br />
1<br />
+ 2<br />
( + ∇ hψ + ∇ hφ ) ⋅∇ hφ − ∇ hφ + + g ζ =<br />
∂t<br />
Introducing V = v + vψ + v<br />
φ<br />
P<br />
V 0<br />
(32)<br />
2 ρ<br />
∂ φ 1<br />
+ ⋅∇ 2<br />
hφ − ∇ hφ + + g ζ =<br />
∂t<br />
Weak formulation:
13 Numerical formulation<br />
∫<br />
⎛ ∂ φ 1 2 P ⎞<br />
W ⎜ + ⋅∇<br />
hφ − ∇<br />
hφ + + gζ ⎟ dσ<br />
= 0<br />
⎝ ∂t<br />
V 2 ρ ⎠<br />
(33)<br />
Ω<br />
Discrete Galerkin method:<br />
⎛ ∂φ<br />
j<br />
1 2 1<br />
⎞<br />
∀ i ∫ i ⎜ j + ⋅∇ j j j<br />
h φ<br />
j<br />
− ∇<br />
hφ j<br />
+ Pj + g ζ<br />
j ⎟dσ<br />
= 0<br />
∂t<br />
V 2 ρ<br />
Ω ⎝<br />
⎠<br />
(34)<br />
Time marching scheme:<br />
<br />
∀ − <br />
+ ( · ∇<br />
Δ<br />
) <br />
Ω<br />
+ − 1 2 ∇ + <br />
<br />
+ = 0<br />
Ω<br />
<br />
(35)<br />
SUPG Stabilization:<br />
<br />
∀ − <br />
+ (( ) · <br />
<br />
) <br />
<br />
<br />
<br />
+ − 1 2 + <br />
<br />
+ <br />
Ω <br />
+ h ( ) <br />
2|( ) <br />
| <br />
− <br />
<br />
<br />
<br />
(36)<br />
+ (( ) · <br />
) <br />
<br />
+ − 1 2 + <br />
+ = 0<br />
<br />
n+<br />
α<br />
e<br />
ke<br />
Where ( )<br />
α<br />
V = V .<br />
n+<br />
ke
14 Numerical formulation<br />
∀ <br />
<br />
<br />
+ h ( ) <br />
2|( ) | ∇ <br />
− <br />
, ,<br />
<br />
<br />
<br />
<br />
<br />
+ h ( ) <br />
2|( ) | <br />
<br />
<br />
<br />
<br />
= −Δ · <br />
<br />
<br />
<br />
+ h ( ) <br />
2|( ) | · <br />
<br />
<br />
(37)<br />
− <br />
<br />
<br />
+ h ( ) <br />
2|( ) | <br />
<br />
− 1 2 + <br />
<br />
<br />
In matrix form:<br />
r<br />
n α<br />
n+ 1 r<br />
+ n+ α n+<br />
α<br />
n+<br />
1<br />
( M + M<br />
S ) φ + ∆ t ( C + CS<br />
) αφ =<br />
r<br />
n α<br />
n<br />
r<br />
+ n+ α n+<br />
α<br />
n<br />
M + M φ − ∆ t C + C 1−α φ<br />
( S ) ( S )( )<br />
n α ⎛ 1 r ur ur<br />
+ ( M + M<br />
S ) ⎜ ( ∇hφ<br />
) − P − ∆tg<br />
⎝ 2<br />
n+ α<br />
2<br />
+<br />
n+ α n+<br />
α<br />
∆t<br />
ρ<br />
ζ<br />
⎞<br />
⎟<br />
⎠<br />
(38)<br />
Where:<br />
= ;<br />
Ω<br />
= h ( ) <br />
2|( ) | ;<br />
<br />
Ω<br />
(39)<br />
= · ∇ ;<br />
Ω
15 Numerical formulation<br />
= h ( ) <br />
2|( ) | ∇ · ∇ ;<br />
<br />
Ω<br />
Free surface kinematic boundary condition<br />
∂ ζ + ( v + ∇<br />
hφ ) ∇<br />
hζ + ∇<br />
hφ ∇<br />
hη − φz<br />
= 0<br />
(40)<br />
∂t<br />
introducing<br />
v<br />
φ<br />
= ∇<br />
hφ<br />
and U = v + vφ<br />
= v + ∇<br />
h<br />
φ<br />
∂ ζ + U ∇<br />
hζ + v<br />
φ ∇<br />
hη − φz<br />
= 0<br />
(41)<br />
∂t<br />
Weak formulation<br />
∫<br />
⎛ ∂ζ<br />
⎞<br />
W ⎜ + ⋅∇<br />
hζ +<br />
φ<br />
⋅∇hη − φz<br />
⎟dσ<br />
= 0<br />
⎝ ∂t<br />
U v ⎠<br />
(42)<br />
Ω<br />
Discrete Galerkin method:<br />
∫<br />
⎛ ∂ζ<br />
⎞<br />
⎜ ( h ) ζ<br />
j hη h<br />
φ<br />
j<br />
φz j ⎟ σ 0<br />
∂t<br />
U Ω ⎝<br />
⎠<br />
(43)<br />
i j j<br />
∀ i + ⋅∇ j + ∇ ⋅∇ j − j d =<br />
Time marching scheme:<br />
∀ <br />
− <br />
+ ( · ∇<br />
Δ<br />
) <br />
Ω<br />
<br />
(44)<br />
+ ((∇ ) · ) − <br />
<br />
<br />
= 0<br />
SUPG Stabilization:
16 Numerical formulation<br />
<br />
∀ − <br />
+ (( ) · <br />
<br />
) <br />
<br />
<br />
<br />
+ (( ) ) · <br />
<br />
Ω <br />
− <br />
<br />
<br />
<br />
+ h ( )<br />
2|( <br />
)| <br />
− <br />
<br />
<br />
<br />
(45)<br />
+ (( ) · <br />
) <br />
<br />
+ (( ) ) · <br />
<br />
− = 0<br />
<br />
n+<br />
α<br />
e<br />
ke<br />
Using ( )<br />
U = U<br />
n+<br />
ke<br />
α
17 Numerical formulation<br />
∀ + h( ) <br />
2|( ) | <br />
<br />
<br />
<br />
<br />
− + h( ) <br />
2|( ) | <br />
<br />
<br />
<br />
<br />
= −Δ · <br />
<br />
− h( ) <br />
2|( ) | · <br />
<br />
<br />
(46)<br />
− ( ) · <br />
<br />
<br />
+ h( ) <br />
2|( ) | ( ) · <br />
<br />
<br />
+ + h( ) <br />
2|( ) | <br />
<br />
<br />
<br />
<br />
In matrix form:<br />
ur<br />
n α<br />
n+ 1 ur<br />
+ n+ α n+<br />
α<br />
n+<br />
1<br />
( + S<br />
) ζ + α∆ t ( D + DS<br />
) ζ =<br />
ur<br />
n<br />
n<br />
ur<br />
+ α n+ α n+<br />
α<br />
n<br />
( + S<br />
) ζ − (1 −α ) ∆ t ( D + DS<br />
) ζ<br />
uur<br />
n α<br />
n α<br />
r<br />
+ n+ α n+<br />
α<br />
n<br />
+ ∆ t ( + S ) φz − ∆ t ⎡<br />
⎣E + E ⎤<br />
S ⎦φ<br />
+ + α<br />
(47)<br />
Where the matrices are defined as follows:
18 Numerical formulation<br />
=<br />
<br />
n+<br />
α<br />
S<br />
∑ ∫<br />
e e<br />
Ω<br />
ie<br />
je<br />
d<br />
e<br />
n+<br />
α<br />
he ( U )<br />
ie<br />
je<br />
∑<br />
h<br />
e<br />
n+<br />
α ∫<br />
e<br />
e<br />
2 ( U ) Ω<br />
ke<br />
α<br />
∑ ∫ ( Uk<br />
)<br />
e h<br />
n+ α<br />
i n+<br />
j<br />
D = ⋅∇ d<br />
e e<br />
Ω<br />
σ<br />
= ∇ dσ<br />
e<br />
n+<br />
( U )<br />
e<br />
n+<br />
( U )<br />
he<br />
⎡<br />
⎤<br />
n+ α<br />
i ke<br />
n+<br />
α j<br />
DS = ∑<br />
h ( <br />
k )<br />
e h dσ<br />
α ⎢ ∫ ∇ U ⋅∇ ⎥<br />
e<br />
e<br />
2 ⎢⎣<br />
Ω<br />
⎥⎦<br />
∑ ∫<br />
k<br />
n+<br />
α<br />
e<br />
( ( hη<br />
)<br />
k h )<br />
= ∇ ⋅∇<br />
n+<br />
α<br />
i j<br />
E d<br />
e<br />
e e<br />
Ω<br />
e<br />
n+<br />
( U )<br />
e<br />
n+<br />
( U )<br />
α<br />
α<br />
k<br />
n+<br />
α<br />
e<br />
( ( η )<br />
k )<br />
he<br />
⎡<br />
⎤<br />
n+<br />
α<br />
i j<br />
ES = ∑<br />
∇h ∇h ⋅∇h dσ<br />
α ⎢ ∫<br />
⎥<br />
e<br />
e<br />
e<br />
2 ⎢⎣<br />
Ω<br />
⎥⎦<br />
σ<br />
σ<br />
(48)
19 Numerical formulation<br />
Free surface boundary condition with absorption<br />
In order to include the wave damping effect, <strong>SeaFEM</strong> discretizes eq. (16) as follows:<br />
n+ 1 n n−1<br />
n+ 1 n−1<br />
φ − 2φ + φ<br />
n 1<br />
n+ 1 n n−1<br />
∂<br />
zφ<br />
− ∂<br />
zφ<br />
= −g∂ 2<br />
zφ − g ( ∂<br />
zφ − 2 ∂<br />
zφ + ∂zφ ) −κ(<br />
x) (49)<br />
∆t<br />
12 2∆t<br />
Eq. (49) is simply eq. (27) plus a second order finite difference in time for the absorbing<br />
n 1<br />
term. Then ∂ zφ + can be obtained as:<br />
10g∆t<br />
⎛ g∆t<br />
−6 κ( x) ⎞<br />
12<br />
∂ φ = − ∂ φ −⎜<br />
⎟∂ φ − φ − φ + φ<br />
g∆ t + 6 κ( x) ⎝ g∆ t + 6 κ( x) ⎠ g∆ t + 6 κ(<br />
x) ∆t<br />
( 2 )<br />
n+ 1 n n− 1 n+ 1 n n−1<br />
z z z<br />
2<br />
Proceeding like in the previous section, we obtain:<br />
⎛ 12 ⎞ n+ 1 ⎛ 12<br />
n n−1<br />
⎞<br />
Z0 Z0<br />
⎜L + M<br />
2<br />
Γ<br />
Γ<br />
2 ( 2 )<br />
g t 6 κ ( t<br />
⎟ φ = M ⎜ φ −<br />
φ<br />
∆ + ∆ g∆ t + 6 κ ( ∆t<br />
⎟<br />
⎝ x) ⎠ ⎝ x)<br />
⎠<br />
⎛ 10g∆t<br />
⎛ g∆t<br />
− 6 κ ( x) ⎞<br />
⎜<br />
g∆ t + 6 κ (<br />
⎜<br />
g∆ t + 6 κ (<br />
⎟<br />
⎝ x) ⎝ x) ⎠<br />
Z0 Z0<br />
B R<br />
( φ<br />
z ) ( φ<br />
z ) ⎟ ( ) ( )<br />
(50)<br />
n n− 1 n+ 1 n+<br />
1<br />
Z<br />
− MΓ<br />
0<br />
+ + b + b<br />
This implementation has several advantages over traditional. The linear system has to<br />
be solved only once per time step; the free surface numerical scheme is implicit; only<br />
information from two previous times at the free surface is required; there is no<br />
restriction about the grid structure, hence unstructured meshes can be used with no<br />
restriction; and the vertical fluid velocity at the free surface is easily computed using eq.<br />
(50).<br />
Radiation boundary condition<br />
When the fluid domain is bounded, an implementation of a radiation boundary<br />
condition is recommended to avoid artificial wave reflexions at the edges of the<br />
computational domain, where waves are supposed to go through without reflection.<br />
<strong>SeaFEM</strong> makes use of the Sommerfeld radiation condition given in eq. (5). In this<br />
sense, it is assumed that the waves scattered by the bodies hit the outlet boundary in<br />
perpendicular. Hence, the radiation condition is approximated by:<br />
⎞<br />
⎠<br />
(51)<br />
)<br />
R<br />
( φn<br />
)<br />
n+<br />
1<br />
n+ 1 φ −<br />
= −<br />
n<br />
φ<br />
c∆t<br />
(52)<br />
Then the term<br />
R<br />
b can be calculated through:
20 Numerical formulation<br />
1<br />
φ φ φ (53)<br />
∆t<br />
R<br />
n+ 1<br />
R<br />
n+<br />
1<br />
n n−1<br />
R<br />
R<br />
( b ) = MΓ<br />
( φn<br />
) = MΓ<br />
( φ −<br />
φ )<br />
R<br />
where Γ represent the outlet boundary that has been assumed to be vertical.<br />
Introducing eq. (53) into eq. (51) we get:<br />
⎛ 12 c ⎞ n+ 1 ⎛ 12<br />
n n−1<br />
⎞<br />
Z0 R<br />
Z0<br />
⎜ L + MΓ + MΓ Γ<br />
2 2 ( 2 )<br />
g t 6 κ ( t t<br />
⎟ φ = M ⎜ φ −<br />
φ<br />
∆ + ∆ ∆ g∆ t + 6 κ ( ∆t<br />
⎟<br />
⎝ x) ⎠ ⎝ x)<br />
⎠<br />
M<br />
⎛ 10g∆t<br />
⎛ g∆t<br />
− 6 κ ( x) ⎞<br />
⎜<br />
g∆ t + 6 κ (<br />
⎜<br />
g∆ t + 6 κ (<br />
⎟<br />
⎝ x) ⎝ x) ⎠<br />
Z<br />
− Γ 0<br />
+<br />
c<br />
∆t<br />
n<br />
R<br />
+ MΓ<br />
φ +<br />
B<br />
( b )<br />
n+<br />
1<br />
Z<br />
n<br />
0 Z0<br />
( φ<br />
z ) ( φ<br />
z )<br />
(54)<br />
A strong coupling between the Sommerfeld radiation condition and the Laplace<br />
equation has been defined, as it was done with the free surface in the previous section.<br />
Then the boundary condition is integrated within the system matrix, avoiding iterating<br />
among the equations.<br />
n−1<br />
⎞<br />
⎟<br />
⎠<br />
Body boundary condition<br />
The boundary condition to be imposed over the surface of the bodies is given by Eq.<br />
(11). The movements of the bodies are assumed to be small enough so that the<br />
computational domain can remind steady, as well as the normals to the bodies’ surface.<br />
B<br />
Hence, the term b is calculated by:<br />
B<br />
n<br />
B<br />
B<br />
( ) = Γ ( n )<br />
+ 1 n+<br />
1<br />
b M φ<br />
(55)<br />
Boundary condition for limit height (Hfs)<br />
The limit height boundary condition is formulated as finding a pressure field to be<br />
applied over the free surface such that the elevation of this one is limited by the location<br />
of the given surface. That is to say, the given surface act as an upper limit for the free<br />
surface elevation.<br />
The free surface boundary conditions are applied in different ways depending on<br />
whether the free surface is in contact with the surface or not. If the free surface is not in<br />
contact, the boundary conditions are applied as if there is no condition, but if there is<br />
contact, the implementation will be different in order to ensure that the free surface does<br />
not penetrate de surface and the necessary pressure to fulfil this condition is calculated.<br />
It will be said that the free surface node where the algorithm is to be applied is dry if the<br />
seal is not in contact with the free surface at that location, and wet if it is.<br />
The main challenge for an algorithm like this is to be capable of capturing when a node<br />
goes from dry to wet and vice versa, as well as estimating the pressure field on the wet<br />
nodes. For a dry node, the implementation of both, the kinetic and dynamic boundary
21 Numerical formulation<br />
condition is the same as for any other node not interacting with the surface. However,<br />
for wet nodes, the free surface boundary condition is imposed via imposing that the free<br />
surface elevation matches the surface elevation, and ensuring that there is no flow<br />
across the seal. These two conditions are represented by the following equations:<br />
+ = → = − <br />
= · ∇ + ∇ ∇ + 1 ∆ ( − )<br />
(56)<br />
The change from being a dry node to become a wet node is identified via the kinematic<br />
BC though the condition + = . On the other hand, the switch from being a wet<br />
node to become a dry one is carried out by comparing the dynamic pressure with the<br />
reference pressure plus a detachment condition. This detachment condition requires for<br />
the free surface to detach at a specific node that this node is not completely surrounded<br />
by attached nodes. Should this be the case, there would be not path for the air to move<br />
in. Therefore, if there is no connection with the air, pressure might drop below the<br />
reference pressure. The dynamic pressure on wet nodes is obtained by applying the<br />
second eq. in (18).<br />
Boundary condition for transom stern<br />
The boundary condition to be applied on the body subject to study is the usual no flux<br />
across the body surface. For this purpose, a normal flux is induced via a Newman<br />
boundary condition in order to cancel out the normal component of the incident velocity<br />
plus the velocity of the body at the specific location.<br />
However, when considering transom sterns, flow detachment happens at the lower edge<br />
of the stern. While potential flow is incapable of predicting this sort of detachment, a<br />
transpiration model will be used to enforce it. To do so, the null normal flux boundary<br />
condition is not used. On the contrary, a flux is allowed in order to enforce that the<br />
detachment edge belong to the free surface stream surface.<br />
Figure 2 explains the transpiration model idea.<br />
2 2<br />
=U 8U +0.5gh<br />
U<br />
8<br />
U<br />
h U<br />
8<br />
Figure 2: Transpiration model for transom stern Boundary condition with flow detachment.
22 Multi-body dynamics<br />
5. Multi-body dynamics<br />
Once the velocity potential has been obtained, the pressure at any point can be<br />
calculated from:<br />
P = −ρgz<br />
− ρ∂ ϕ<br />
(57)<br />
t<br />
Eq. (57) requires estimating the value of ∂ tϕ . The same fourth order finite difference<br />
scheme that has been used for the free surface elevation is used here:<br />
P<br />
ρ ⎛ 25 4 1<br />
= −ρgz<br />
− ⎜ ϕ − 4ϕ + 3ϕ − ϕ + ϕ<br />
∆t<br />
⎝ 12 3 4<br />
n + 1 n + 1 n n − 1 n − 2 n − 3<br />
⎞<br />
⎟<br />
⎠<br />
(58)<br />
Body dynamics<br />
Integrating the pressure over the bodies’ surface, the resulting forces and moments are<br />
obtained. On the other hand, the body dynamics is given by the equation of motion:<br />
+ = (59)<br />
where is the mass matrix of the multi-body system; is the hydrostatic restoring<br />
coefficient matrix of the multi-body system; F are the hydrodynamic forces induced<br />
over the bodies plus any other external forces; and X represent the vector containing the<br />
movements of the six degrees of freedom of each body. Both and are assumed to<br />
be constants.<br />
In the specific case where the bodies are fixed, only refracted waves are calculated and<br />
the linear system given in eq. (54) is to be solved just once per time step. However,<br />
when solving the body dynamics along with the wave problem requires an iterative<br />
procedure since interaction between the waves and the movements of the structure exist,<br />
giving birth to waves radiated by the bodies.<br />
In order to solve Eq.(63), <strong>SeaFEM</strong> uses an implicit Bossak-Newmark´s algorithm [17]:<br />
(1 − ) + = − (60)<br />
= + ∆(1 − ) + (61)<br />
= + ∆ + ∆<br />
2 (1 − 2) <br />
+ 2 <br />
In the above integration algorithm, α is the parameter defining the numerical damping<br />
added in the integration. This damping created a desirable stabilizing effect in the body<br />
dynamics integration. α is usually set to α =-0.1, and in <strong>SeaFEM</strong>, γ and β are calculated<br />
as γ= 0.5- α and β=0.5γ +0.025α.<br />
Within each time step, the system of eqs. (60)-(62) is solved iteratively along with eq.<br />
(54). This requires predicting the body velocities by solving eq. (61), introducing this<br />
(62)
23 Multi-body dynamics<br />
velocities into eq. (55), and solving the linear system in eq. (54), introducing the<br />
pressure forces into eq. (60), and repeating the process until convergence is reached. In<br />
order to accelerate convergence, <strong>SeaFEM</strong> uses a second order prediction method for<br />
estimation of the bodies’ velocities and positions for the next iteration.<br />
The algorithm implemented also allows considering non-linear external forces acting on<br />
the bodies such as mooring forces. In this implementation they are evaluated for every<br />
iteration of the solver and added to the right hand side of eq. (59).<br />
Non-linear hydrostatics<br />
Eq. (59) is valid for small rotations of the bodies. However, in the case of switching on<br />
the non-linear hydrostatics option, the rigid body dynamics solver of <strong>SeaFEM</strong> is<br />
extended for taking into account large motions. For this purpose, the Euler equations are<br />
used to integrate the body dynamics. Equation (59) is now rewritten as:<br />
= ∗ (63)<br />
with<br />
∗ = <br />
<br />
− <br />
∧ · (64)<br />
Where , are the 3x1 vectors containing the external forces and moments acting on<br />
the body, is the instantaneous angular velocity of the body and the inertia tensor.<br />
The Euler equations are derived in a rotating reference frame fixed to the body, and<br />
therefore, the inertia tensor is assumed to be constant.<br />
It is important to remark that, in equation (64), must be evaluated in the local<br />
reference frame.<br />
Bossak-Newmark´s algorithm shown in eqs. (60)-(62) is then applied to the integration<br />
of eq. (64) as follows:<br />
(1 − ) + = ∗ (65)<br />
= + ∆(1 − ) + (66)<br />
= + ∆ + ∆<br />
2 (1 − 2) <br />
+ 2 <br />
Body links<br />
When links exist between bodies, eq. (59) must be solved along with the algebraic<br />
system of equations<br />
<br />
() = ∑ + = 0<br />
(68)<br />
which represents the different restrictions existing between various degrees of freedom<br />
x j of two bodies.<br />
(67)
24 Multi-body dynamics<br />
Within this context, let ( ) be such that ( ) = − and ( ) =<br />
<br />
− h ( ) = () = ∑<br />
<br />
.<br />
Then, let’s formulate the following optimization problem:<br />
<strong>Ma</strong>ximize ( ) subject to: h ( ) = 0 i=1, 2, …, m. The solution to the problem<br />
can be obtained using Lagrange multipliers. Then, the above problem is reformulated<br />
as:<br />
∇() + () = − + = 0<br />
() = = 0<br />
(69)<br />
Where = and = … .
25 Body matrices referred to an arbitrary generic point<br />
7. Body matrices referred to an arbitrary generic point<br />
In this section we briefly introduce the formulation used to relate the body’s matrices<br />
(e.g. mass, damping or stiffness matrices) from two different points of reference. This is<br />
useful when body properties are known from a reference point which is not coincident<br />
with the gravity centre of the body or more generally with the point to which all output<br />
results (i.e. displacements, forces and moments, etc.) want to be referred.<br />
Let A be a matrix relating forces and momentums as a function of displacements and<br />
rotations respect to a point of reference I .<br />
⎛ FI ⎞ ⎛ xI ⎞ ⎛ AFx<br />
AF<br />
θ ⎞⎛ xI<br />
⎞<br />
⎜ ⎟ = A⎜ ⎟ = ⎜ ⎟⎜ ⎟<br />
⎝ M<br />
I ⎠ ⎝θI ⎠ ⎝ AMx<br />
AM<br />
θ ⎠⎝θI<br />
⎠<br />
(70)<br />
Let B be a matrix relating forces and momentums as a function of displacements and<br />
rotations respect to a point of reference R .<br />
⎛ FR ⎞ ⎛ xR ⎞ ⎛ BFx<br />
BFθ<br />
⎞⎛ xR<br />
⎞<br />
⎜ ⎟ = B⎜ ⎟ = ⎜ ⎟⎜ ⎟<br />
⎝ M<br />
R ⎠ ⎝θR ⎠ ⎝ BMx<br />
BMθ<br />
⎠⎝θR<br />
⎠<br />
(71)<br />
Let r be the vector r<br />
= uur IR . The following relationships hold:<br />
F = F ; M = M + F × r<br />
R I R I I<br />
x = x − θ × r;<br />
θ = θ<br />
I R R I R<br />
(72)<br />
Here, it must be noted that the vector cross product can always be written as the product<br />
of a skew-symmetric matrix and a vector in the form:<br />
× = (73)<br />
where the skew-symmetric matrix is given in terms of the vector components as<br />
follows:<br />
0 − <br />
= − 0 (74)<br />
− 0<br />
Then:<br />
F = F ; M = M + R F<br />
R I R I I<br />
x = x − Rθ ; θ = θ<br />
I R R I R<br />
(75)<br />
and using F = F ; M = M + R F we can write<br />
R I R I I
26 Body matrices referred to an arbitrary generic point<br />
F = F = A x + A<br />
θ<br />
R I Fx I Fθ<br />
I<br />
( θ )<br />
M = M + R F = A x + A θ + R A x + A<br />
R I I Mx I Mθ<br />
I Fx I Fθ<br />
I<br />
(76)<br />
Inserting x = x − Rθ ; θ = θ<br />
I R R I R<br />
F = A x − A Rθ<br />
+ A<br />
θ<br />
R Fx R Fx R Fθ<br />
R<br />
( )<br />
M = A x − A Rθ + A θ + R A x − A Rθ + A θ<br />
R Mx R Mx R Mθ<br />
R Fx R Fx R Fθ<br />
R<br />
(77)<br />
Reordering terms:<br />
F = A x + A θ − A Rθ<br />
R Fx R Fθ<br />
R Fx R<br />
M = A x + A θ + R A x − A Rθ + R ⋅ A θ − R A Rθ<br />
R Mx R Mθ<br />
R Fx R Mx R Fθ<br />
R Fx R<br />
(78)<br />
In matrix form:<br />
R R ⎛ 0 −AFx<br />
R ⎞ R<br />
⎛ F ⎞ ⎛ x ⎞ ⎛ x ⎞<br />
⎜ ⎟ = A⎜ ⎟ + ⎜ ⎟⎜ ⎟<br />
⎝ M<br />
R ⎠ ⎝θR ⎠ ⎝ R AFx − AMx R + R AF θ<br />
− R AFx<br />
R ⎠⎝θR<br />
⎠<br />
(79)<br />
Hence, matrices and are finally related by the following expression:<br />
<br />
<br />
= <br />
+ 0 − <br />
(80)<br />
− − +
27 Statistical description of an irregular sea<br />
8. Statistical description of an irregular sea<br />
Spectrum discretization<br />
Let be S( ω, α ) an energy density spectrum describing a sea state in terms of the wave<br />
frequency and direction of propagation. The discretization procedure to obtain a<br />
stationary and ergodic realization based on monochromatic waves is as follows:<br />
Let be ωmin<br />
the minimum frequency to be considered, ωm<br />
ax<br />
the maximum frequency to<br />
be considered, αmin<br />
the lower direction of propagation to be considered, α<br />
m ax<br />
the larger<br />
w<br />
direction of propagation to be considered, the number of wave frequencies, and α<br />
the number of wave directions to be considered. Then, the frequency and direction<br />
discretization sizes are given by:<br />
∆ = ( − )/ <br />
Δ = ( − )/( − 1) (81)<br />
Then, the wave elevation is given by:<br />
w<br />
<br />
α<br />
<br />
i= 1 j=<br />
1<br />
( )<br />
η = ∑∑ Aij cos kij cos( α<br />
j<br />
) x + kij sin( α<br />
j<br />
) y − Ω<br />
ijt<br />
+ δij<br />
(82)<br />
Where<br />
Ωij<br />
is the wave<br />
δ is the wave phase, t represents time, and<br />
Aij<br />
is the monochromatic wave amplitude, kij<br />
is the wave number,<br />
angular velocity, α<br />
j<br />
is the wave direction,<br />
ij<br />
x,<br />
y are the horizontal Cartesian coordinates. Each parameter is obtained as follows:<br />
Ω is a random variable with uniform distribution in [ ω ω / 2, ω ω / 2]<br />
ij<br />
i<br />
−∆ + ∆ .<br />
i<br />
= + ( − 1/2)Δ<br />
= + ( − 1)Δ<br />
Ω = tanh <br />
(83)<br />
δij<br />
is a random variable with uniform distribution in [ 0,2π ].<br />
A<br />
= 2 ∆ω∆α<br />
S ,<br />
( ω α )<br />
∑<br />
ij i j<br />
l,<br />
m<br />
1<br />
16<br />
H<br />
2<br />
S<br />
( ω α )<br />
2 ∆ω∆αS<br />
,<br />
where H = 4 m0<br />
is the significant wave height, and m0<br />
S( ω, α)<br />
dωdα<br />
S<br />
spectrum wave energy.<br />
l<br />
m<br />
∞ π<br />
0 −π<br />
(84)<br />
= ∫ ∫ is the
28 Statistical description of an irregular sea<br />
Convergence<br />
Convergence of the discretized spectrum will happen as ωmin → 0, ωmax<br />
→ ∞ , ∆ω<br />
→ 0<br />
, αmin → 0, αmax → 2π<br />
, and ∆α<br />
→ 0 .<br />
The rate of convergence with<br />
integration.<br />
∆ ω and ∆ α is that of the rectangle rule of numerical<br />
Spectral moments<br />
1. Zero order moment : m<br />
0<br />
The spectral energy of a wave spectrum is given by:<br />
∞ π<br />
1<br />
0<br />
= ∫ ∫ ω α ω α =<br />
16<br />
0 −π<br />
2<br />
m S( , ) d d HS<br />
The discrete spectrum is scaled such that the spectral moment m<br />
0<br />
is conserved.<br />
Therefore:<br />
1 1<br />
2 2<br />
∑ Aij<br />
= HS<br />
(85)<br />
i,<br />
j 2 16<br />
2. First order moment: m1<br />
m = S( ω , α ) Ω ∆ω∆ α = S( ω , α ) ω ∆ω∆ α + S( ω , α ) ε ∆ω∆ α = m + m<br />
m<br />
m<br />
*<br />
1 i j ij i j ij i j ij<br />
D<br />
1<br />
P<br />
1<br />
i, j i, j i,<br />
j<br />
D<br />
1<br />
P<br />
1<br />
∑ ∑ ∑<br />
∑<br />
= S( ω , α ) ω ∆ω∆α<br />
i,<br />
j<br />
∑<br />
= S( ω , α ) ε ∆ω∆α<br />
i,<br />
j<br />
i j ij<br />
i j ij<br />
(86)<br />
Where<br />
ε is uniform distributed between [ ω / 2, ω / 2]<br />
ij<br />
P<br />
component of the first moment, and m1<br />
ωmax<br />
→ ∞ , ω<br />
min<br />
= 0 , α<br />
min<br />
= 0, αmax 2π<br />
D<br />
−∆ ∆ , m<br />
1<br />
is a deterministic<br />
is a random component. Assuming that<br />
= , the deterministic component converges to:<br />
lim<br />
ω<br />
m m ωS( ω, α)<br />
dωdα<br />
α π<br />
D<br />
∆ →0 1 1<br />
∆ →0<br />
∞ π<br />
= = ∫ ∫ (87)<br />
0 −<br />
On the other hand, for large values of ω P<br />
, the probabilistic part m1<br />
is a random variable<br />
with normal distribution. The mean and variance of this distribution are:<br />
∆ω<br />
/ 2<br />
1<br />
µ = S( ω , α ) ∆ω∆ α ω dω<br />
= 0<br />
∆ω<br />
∑ ∫ (88)<br />
i j<br />
i, j<br />
−∆ω<br />
/ 2
29 Statistical description of an irregular sea<br />
/<br />
= , ΔΔ 1<br />
Δ <br />
= <br />
Δ<br />
, ΔΔ<br />
12 (89)<br />
,<br />
/<br />
Hence, the probabilistic component converges to a random variable with zero mean and<br />
zero variance. Therefore:<br />
,<br />
lim m = lim m + lim m = m + 0<br />
(90)<br />
*<br />
D<br />
P<br />
∆ω→0 1 ∆ω→0 1 ∆ω→0 1 1<br />
∆α →0 ∆α →0 ∆α<br />
→0<br />
Waves spectrum<br />
Several standard waves spectrum are predefined in <strong>SeaFEM</strong>:<br />
Pearson Moskowitz spectrum:<br />
This is probably the simplest idealized spectrum, obtained by assuming a fully<br />
developed sea state, generated by wind blowing steadily for a long time over a large<br />
area [19]. The resulting spectrum is [10]:<br />
() = (0.11/2)( /) .(/) (91)<br />
where is the wave period; is the significant wave height; is the mean wave<br />
period, which is obtained via = 2 / , with and the zero and first<br />
moments of the wave spectrum.<br />
Jonswap2 spectrum:<br />
The JONSWAP spectrum was established during a joint research project, the "JOint<br />
North Sea WAve Project" [20]. This is a peak-enhanced Pierson-Moskowitz spectrum<br />
given on the form [21]:<br />
() = 5<br />
32π T<br />
<br />
T · ε · e . <br />
. ()<br />
<br />
= e (. )/(√) (92)<br />
where = 2/ , = 0.07 for ≤ 6.28/ , = 0.09 for > 6.28/ , is the<br />
wave period; is the significant wave height, T is the peak wave period and is the<br />
peakedness parameter.<br />
Jonswap spectrum:<br />
An alternative definition of the JONSWAP spectrum is given by [10]:<br />
() = 155 <br />
T ω · 3.3 · e <br />
= e (. )/(√) (93)
30 Statistical description of an irregular sea<br />
where = 2/ , = 0.07 for ≤ 5.24/ , = 0.09 for > 5.24/ , is the<br />
wave period; is the significant wave height; is the mean wave period, which is<br />
obtained via = 2 / , with and the zero and first moments of the wave<br />
spectrum.
31 Mooring system modelling<br />
9. Mooring system modelling<br />
<strong>SeaFEM</strong> can handle complex mooring systems made up of various mooring lines<br />
attached to the floating structure. Each mooring line can be in turn composed of various<br />
segments each one resembling a chain, a steel cable or even a synthetic fiber. Forces<br />
resulting from the action of buoys and sinkers acting at the junctions between mooring<br />
line segments can also be considered. Hence, <strong>SeaFEM</strong> can deal with a wide variety of<br />
multi-segmented mooring line systems.<br />
Cable tensions depend on the buoyancy and lateral displacements of the floating<br />
structure, the cable weight in water, the elasticity in the cable and the geometrical layout<br />
of the mooring system. Hence, as the floating structure moves in response to unsteady<br />
environmental loadings, the mooring restraining forces change with the changing cable<br />
tension. This means that the mooring system has an effective compliance whose<br />
response is in general non-linear. Within <strong>SeaFEM</strong>, mooring inertia and damping are<br />
ignored, but the non-linear response is accounted for in the mooring system dynamics.<br />
Mooring systems within <strong>SeaFEM</strong> are solved using a quasi-static approach in the sense<br />
that at any step of the calculation, and once the floating body displacements are known,<br />
the mooring system solver calculates the tensions and the geometrical configuration of<br />
each mooring line segment assuming that each cable is in static equilibrium at that<br />
instant. The mooring loads resulting from the calculation are added to the total load<br />
acting on the floating body, and the resulting dynamic equations of motion of the<br />
structure are solved again until convergence.<br />
At each time step the implicit non-linear system of equations describing the mooring<br />
system is solved using a classical Newton-Raphson scheme.<br />
The formulation implemented within <strong>SeaFEM</strong> for dealing with catenary based mooring<br />
systems is outlined in what follows. Additionally, cables behaving as springs (both in<br />
tension and/or compression) can also be modeled.<br />
Catenary equations<br />
In a local coordinate system with its origin located at the lower cable point, the mooring<br />
equations for the catenary read as follows:<br />
= <br />
h <br />
+ + (94)<br />
= <br />
h <br />
+ + <br />
(95)<br />
where z is the vertical position, s is the catenary arc length, is the catenary weight per<br />
unit length in water, and is the horizontal component of the cable tension which is<br />
constant everywhere. If the conditions ( = 0) = 0 and ( = 0) = 0 are applied, the<br />
equations result to be:
32 Mooring system modelling<br />
= <br />
h <br />
+ − cosh () (96)<br />
= <br />
h <br />
+ − sinh ()<br />
(97)<br />
The action of any part of the line upon its neighbor is purely tangential, and the<br />
tangential direction can be written as:<br />
= <br />
= sinh <br />
+ (98)<br />
The equilibrium equations can be written as:<br />
· = · = <br />
· − · = · <br />
(99)<br />
From the horizontal component of the tension at any point of the catenary, it is possible<br />
to write the modulus of the cable tension at any point of the catenary as:<br />
=<br />
<br />
= · 1 + = · h <br />
+ (100)<br />
Hence, the vertical component of tension at any point of the catenary can be written as:<br />
= · = · h <br />
+ (101)<br />
In particular, at the upper and bottom vertex of the catenary Eq.(83) reads respectively:<br />
= · h <br />
+ (102)<br />
= · h() (103)<br />
And the vertical equilibrium equation will result in:<br />
= <br />
h <br />
+ − h () = ( = )<br />
(104)<br />
For a given length of the catenary (L) and for given horizontal and vertical distances<br />
(l,h) between the initial and end points of the catenary, it is possible to solve the<br />
equations by using a classical Newton-Raphson iterative process.<br />
If the catenary has a seabed contact in its lower point, the following condition must be<br />
added<br />
<br />
<br />
= 0<br />
<br />
(105)
33 Mooring system modelling<br />
and the catenary equations result in:<br />
= <br />
h <br />
− 1<br />
= <br />
h <br />
<br />
(106)<br />
Elastic catenary formulation<br />
The elastic catenary formulation used within <strong>SeaFEM</strong> is similar to that presented in<br />
[12].<br />
Each mooring line is analyzed in a local coordinate system that originates at the anchor.<br />
The local z-axis of this coordinate system is vertical and the local x-axis is directed<br />
horizontally from the anchor to the instantaneous position of the fairlead. When the<br />
mooring system module is called for a given structure displacement, <strong>SeaFEM</strong> first<br />
transforms each fairlead position from the global frame to this local system to determine<br />
its location relative to the anchor, x F and z F .<br />
z<br />
V F<br />
H F<br />
Fairlead<br />
V A<br />
Anchor<br />
H A<br />
Figure 3: Catenary system of reference.<br />
x<br />
In the local coordinate system, the analytical formulation is given in terms of two<br />
nonlinear equations in two unknowns—the unknowns are the horizontal and vertical<br />
components of the effective tension in the mooring line at the fairlead, H F and V F ,<br />
respectively.<br />
( , ) = <br />
<br />
+ 1 + <br />
<br />
− − <br />
<br />
( , ) = <br />
1 + <br />
<br />
<br />
<br />
+ 1 + <br />
− <br />
+ <br />
<br />
(107)<br />
− 1 + <br />
− <br />
+ 1<br />
− <br />
2 <br />
The analytical formulation of two equations in two unknowns is different when a<br />
portion of the mooring line adjacent to the anchor rests on the seabed:
34 Mooring system modelling<br />
( , ) = − <br />
+ <br />
<br />
+ 1 + <br />
<br />
+ <br />
<br />
( , ) = <br />
1 + <br />
<br />
<br />
<br />
+ <br />
2 − − <br />
<br />
+ − <br />
− <br />
− <br />
− <br />
, 0<br />
− 1 + <br />
− <br />
+ 1<br />
− <br />
2 <br />
(108)<br />
= − <br />
<br />
The last term on the right-hand side of the x F equation, which involves C B , corresponds<br />
to the stretched portion of the mooring line resting on the seabed that is affected by<br />
static friction. The seabed static friction was modeled simply as a drag force per unit<br />
length. The MAX function is needed to handle cases with and without tension at the<br />
anchor. Specifically, the resultant is zero when the anchor tension is positive; that is, the<br />
seabed friction is too weak to overcome the horizontal tension in the mooring line.<br />
Conversely, the resultant of the MAX function is nonzero when the anchor tension is<br />
zero. This happens when a section of cable lying on the seabed is long enough to ensure<br />
that the seabed friction entirely overcomes the horizontal tension in the mooring line.<br />
These equations consider that slack catenary is always tangent to the seabed at the point<br />
of touchdown.<br />
The mooring system module uses a Newton-Raphson iteration scheme to solve<br />
nonlinear for the fairlead effective tension (H F and V F ), given the line properties (L, ,<br />
EA, and C B ) and the fairlead position relative to the anchor (x F and z F ). The mooring<br />
system module uses the values of H F and V F from the previous time step as the initial<br />
guess in the next iteration of the Newton-Raphson scheme. As the model is being<br />
initialized, the following starting values, <br />
<br />
and , are used [12]:<br />
= <br />
<br />
2 <br />
= 2 (109)<br />
<br />
h( ) + <br />
where the dimensionless catenary parameter, λ 0 , depends on the initial configuration of<br />
the mooring line:<br />
1000000 = 0<br />
<br />
0.2 + ≥ <br />
=<br />
(110)<br />
<br />
3 <br />
− <br />
<br />
− 1 h<br />
<br />
<br />
Once the effective tension at the fairlead has been found, determining the horizontal and<br />
vertical components of the effective tension in the mooring line at the anchor, H A and<br />
V A , respectively, is simple. From a balance of external forces on a mooring line, one can<br />
easily verify that
35 Mooring system modelling<br />
= <br />
= − <br />
(111)<br />
When no portion of the line rests on the seabed, and<br />
= ( − , 0)<br />
= 0<br />
(112)<br />
The mooring system module solves the configuration of, and effective tensions within,<br />
the mooring line. When no portion of the mooring line rests on the seabed, the equations<br />
for the horizontal and vertical distances between the anchor and a given point on the<br />
line, x and z, and the equation for the effective tension in the line at that point, Te, are as<br />
follows:<br />
() = <br />
+ <br />
<br />
+ 1 + <br />
+ <br />
− <br />
+ 1 + <br />
<br />
+ <br />
<br />
<br />
() = <br />
1 + + <br />
<br />
<br />
<br />
− 1 + <br />
<br />
+ 1<br />
+ <br />
2 <br />
(113)<br />
() = + ( + ) <br />
where s is the unstretched arc distance along the mooring line from the anchor to the<br />
given point. The equations with seabed interaction are:<br />
0 ≤ ≤ − <br />
<br />
<br />
<br />
+ <br />
2 − 2 − <br />
<br />
<br />
+ − <br />
− <br />
, 0 − <br />
≤ ≤ <br />
() =<br />
<br />
+ <br />
( − )<br />
+ 1 + ( − ) <br />
+ <br />
<br />
<br />
<br />
<br />
<br />
+ <br />
<br />
2 − + − <br />
− <br />
, 0 ≤ ≤ <br />
<br />
0 0 ≤ ≤ <br />
<br />
() = <br />
<br />
1 + ( − <br />
)<br />
− 1 + ( − ) <br />
<br />
2<br />
≤ ≤ <br />
<br />
<br />
<br />
( + ( − ), 0)<br />
() = <br />
+ ( − ) <br />
≤ ≤ <br />
(114)<br />
The final calculation is a computation of the total load on the system from the<br />
contribution of all mooring lines. This mooring system-restoring load is found by first<br />
transforming each fairlead tension from its local mooring line coordinate system to the<br />
global frame, then summing up the tensions from all lines.
36 Mooring system modelling<br />
Dynamic cable formulation<br />
<strong>SeaFEM</strong> includes a finite element model (FEM) for solving mooring line dynamics. For<br />
this purpose, the line is divided into a series of straight segments modeled by nonlinear<br />
truss elements, with three translational degrees of freedom per node. Lagrangian<br />
formulation is used to describe the dynamics of the mooring line.<br />
Figure 4: Scheme of the FEM cable model.<br />
Applying FEM formulation to this nonlinear elastodynamics problem, the equations of<br />
the dynamics of the line can be written as follows,<br />
+ + + + = (115)<br />
Where is the vector of nodal translational degrees of freedom, the vector of external<br />
loads, M the inertia matrix of the line, the added mass matrix, is the pretension<br />
vector in the initial configuration, and , the vector of internal forces of the cable.<br />
Damping effects of mooring cable can be inserted through a Rayleigh-type damping<br />
matrix C.<br />
So, considering the standard linearization of the internal forces, above equations can be<br />
expressed as,<br />
+ + + + + + = (116)<br />
where and are the so called material stiffness matrix and geometric stiffness<br />
matrix, respectively [22]. The corresponding elemental material stiffness matrix , and<br />
geometric stiffness matrix, , can be obtained in terms of the strain of the truss<br />
element, ,
37 Mooring system modelling<br />
+ = <br />
= <br />
⊗ <br />
+ <br />
(117)<br />
Where is the volume of the element, the axial stress and the axial elastic<br />
modulus. The FEM allows calculating those matrixes as follows [22]:<br />
= <br />
⊗ <br />
= B B d<br />
= <br />
= B B d<br />
<br />
<br />
The elemental inertia matrixes are evaluated as:<br />
<br />
= N Nd<br />
<br />
And the added mass matrix is calculated as:<br />
<br />
<br />
(118)<br />
(119)<br />
<br />
= N N d<br />
<br />
(120)<br />
In the above equations, is the general matrix of shape functions [23], is the section<br />
area of the cable, and is the added mass coefficient.<br />
Finally, the damping matrix is evaluated using the Rayleigh formulation:<br />
= 1 + 2 M (121)<br />
The coefficients 1 , can be calculated as:<br />
<br />
= 1 2 1<br />
<br />
+ 2 (122)<br />
where <br />
is the damping ratio corresponding to the natural frequency of the structure.<br />
The external forces vector is evaluated by assembling the contribution of the different<br />
forces acting on the element:<br />
= + + + + <br />
<br />
(123)<br />
where is the weight, h is the buoyancy force, is the drag force related to the<br />
currents, is the force due to the seabed interaction and is the drag force due to the<br />
waves.<br />
The weight and buoyancy force are calculated as follows:
38 Mooring system modelling<br />
<br />
+ = h<br />
N d,<br />
0<br />
= + = ( − )<br />
<br />
(1 + ) <br />
(124)<br />
where <br />
is the density of the cable, <br />
is the water density, weight per meter and <br />
the strain of the considered element.<br />
On the other hand, the drag force acting on the element, is evaluated as:<br />
<br />
+ = <br />
N d,<br />
0<br />
<br />
= <br />
1 <br />
2 | <br />
| d + <br />
1 2 | <br />
| d<br />
<br />
<br />
(125)<br />
where D is the characteristic diameter of the element, and the normal and<br />
tangential drag coefficients and is the velocity relative to the element.<br />
The seabed interaction is modeled with the spring and normal and tangential damping<br />
terms given by,<br />
<br />
= N d + N d + N d<br />
<br />
<br />
<br />
<br />
<br />
(126)<br />
where the vertical forces per unit length, due to the stiffness of the seabed <br />
<br />
, are<br />
expressed as:<br />
<br />
<br />
<br />
= <br />
− ( ) <br />
, if − ( ) ≤ <br />
, if − ( ) > <br />
(127)<br />
where is the vertical coordinate of the corresponding node, is the length of the cable<br />
associated to the node, and <br />
is a coefficient of the model. is the coefficient of the<br />
seabed normal stiffness force which has the magnitude:<br />
= (128)<br />
Where is the diameter of the line, and is the ground normal stiffness per unit<br />
length. The term is then evaluated as:<br />
= + <br />
<br />
(129)<br />
The seabed damping forces are applied in the normal and tangential directions. The<br />
normal damping force is only applied when the penetrating object is travelling into the<br />
seabed. It is applied in the seabed outwards direction and has a magnitude given by:<br />
= , if < 0<br />
0 , if ≥ 0 <br />
(130)
39 Mooring system modelling<br />
and is formulated as a fraction, , of the critical damping:<br />
= 2,0 <br />
( )<br />
(131)<br />
Finally, the tangential damping force is again evaluated as a fraction, , of the critical<br />
damping, and is applied in the direction opposing the tangential component of the<br />
velocity of the penetrator:<br />
<br />
= <br />
(1321)<br />
= 2,0 <br />
( )<br />
An implicit time integration scheme based on called Bossak-Newmark method [7] is<br />
applied to solve the system. It will lead to a system of algebraic equations to be solved<br />
in iterative manner.<br />
(1 − ) , + , + ,<br />
+ , + , , <br />
= , + , , + , − + <br />
− , − ( + 1) − , + , <br />
<br />
(133)<br />
where is time step, i denotes iteration, is a parameter related with Bossak-<br />
Newmark implicit method, and and are parameters related to Newmark time<br />
integration scheme.<br />
Finally, the new position and velocity of each node can be evaluated as,<br />
Δ = + <br />
+ 2<br />
2 (1 − 2) + 2<br />
<br />
(134)<br />
= + d(1 − ) + <br />
The dynamic cable solver is integrated within <strong>SeaFEM</strong> dynamic solver. The scheme of<br />
this integration is presented in Figure 5. In order to accelerate the scheme, the mooring<br />
forces are linealised within the body dynamics group, by evaluating the stiffness matrix<br />
of the line.
40<br />
Mooring system modelling<br />
Figure 5: Scheme of the dynamic integration solver of <strong>SeaFEM</strong>.<br />
A detailed description of the FEM cable model implemented in <strong>SeaFEM</strong> can be found<br />
in [23].
41 Morison’s forces<br />
10. Morison’s forces<br />
When viscous effects may be advanced to have a significant effect on the dynamic<br />
behavior of an offshore structure, Morison's equation can be used to evaluate wave<br />
loads on slender cylindrical elements of the structure [14-16]. In <strong>SeaFEM</strong>, force<br />
corrections due to viscous effects can be also taken into account by using the Morison's<br />
equation. For this purpose, an auxiliary framework structure, associated to a body must<br />
be defined. See the <strong>SeaFEM</strong> user manual for details on how to define the auxiliary<br />
framework structure elements using the Tcl interface of <strong>SeaFEM</strong>.<br />
Based on the information provided by the user, <strong>SeaFEM</strong> evaluates Morison's forces per<br />
unit length acting on the framework structure. After integration along the different<br />
elements, the resultant forces are incorporated to the dynamic solver of the rigid body to<br />
which the idealized framework structure has been associated.<br />
It is useful to write the Morison’s equation in a vectorial formulation that automatically<br />
takes into account the actual orientation of structural elements and force components.<br />
Considering a segment of a long slender structural element submerged into water its<br />
local orientation is given by a unit vector<br />
= + + (135)<br />
being l, m, n the directional cosines and (l,j,k) the unit vectors of the global coordinate<br />
system.<br />
Similarly, the relative fluid velocity vector and the relative acceleration vector of the<br />
submerged body are given by:<br />
= v + v + v <br />
= a + a + a <br />
(136)<br />
Remark: The relative fluid velocity vector and the relative acceleration vector are<br />
evaluated based on the undisturbed wave potential equations.<br />
The force per unit length on a slender cylindrical element may be written as the sum of<br />
inertia, drag, friction and lift forces:<br />
= + + + + (137)<br />
where the inertia force is oriented along the acceleration vector component normal to<br />
the element member, and its magnitude is proportional to the acceleration component. Lift<br />
force is oriented normal to the velocity vector and normal to the axis of the element, and<br />
its magnitude is proportional the velocity squared. Drag force is proportional to the<br />
squared velocity component normal to the element and normal to the lift force, while the<br />
linear drag force is proportional to the velocity component normal to the element.<br />
Finally, friction force is aligned along the axis of the element and proportional to the
42 Morison’s forces<br />
squared velocity components tangential to the element axis. All these are satisfied if the<br />
various force components are defined as follows:<br />
= ( × × )<br />
= 1 2 | × × |( × × )<br />
= 1 2 ( × × )<br />
= 1 2 | · |( · ) · <br />
(138)<br />
= 1 2 | × |( × )<br />
where D is a linear dimension (the diameter in the case of a cylinder), S is the cross<br />
section area, C M is the added mass coefficient, C D is the non-linear drag coefficient, C V<br />
is the linear drag coefficient, C F is the friction coefficient, and C L is the lift coefficient.<br />
Remark: If the defined C M coefficient is greater than 1, then F M component is evaluated<br />
in the following alternative way:<br />
= ( − 1)( × × ) + ( × × ) (139)<br />
where is the fluid acceleration of the incident wave, and therefore, the second term in the<br />
right hand side of the above equations represents the Froude-Kriloff force.<br />
As stated above, equations (104-108) can estimate the different components of the force<br />
per unit length on a long (slender) structural element. Therefore, they can be integrated<br />
along the element axis, to obtain the additional forces an moments acting on the center<br />
of gravity of the associated body.
43 Response Amplitude Operators (RAOs)<br />
11. Response Amplitude Operators (RAOs)<br />
RAOs are transfer functions of the relation between the wave exciting forces and ship<br />
movements, used to determine the effect that a sea state will have upon the motion of a<br />
ship through the water.<br />
Calculation of Response Amplitude Operator (RAO) in <strong>SeaFEM</strong> is done by analysing<br />
the time series response of the ship, using a discretized white noise spectrum. This<br />
spectrum is defined by a number = 2 of waves of equal amplitude and periods<br />
varying between the maximum and minimum values defined by the user. The value of<br />
these se periods are selected to match the discrete Fourier transform of the output<br />
signal, give by:<br />
<br />
= · ∆∗ <br />
<br />
(140)<br />
Given and , the minimum and maximum periods of the analysis, the<br />
frequency increment is:<br />
∆ = ( − ) ⁄ = <br />
−<br />
<br />
⁄<br />
<br />
<br />
(141)<br />
The well known Fast Fourier Transform algorithms give a procedure to obtain an exact<br />
evaluation of the transfer functions defined above. This way, the time step and the total<br />
computing time are internally fixed to match the required sampling time and total<br />
sampling points. Then, the holding frequency ∆ ∗ is evaluated as<br />
∆ ∗ = (∆ , ) ⁄ <br />
And the discrete frequencies are:<br />
= · ∆ = 0, 1, 2, … , − 1<br />
The required sampling frequency defines the time step as:<br />
(142)<br />
(143)<br />
∆ =<br />
1<br />
2∆ ∗ · (144)<br />
The required number of sampling points defines the total calculation time as:<br />
= 1<br />
∆ ∗ (145)
44 Fluid-Structure interaction algorithm<br />
12. Fluid-Structure interaction algorithm<br />
<strong>SeaFEM</strong> features a fluid-structure interaction algorithm, able to perform coupled<br />
analysis with the structural solver of the Tdyn’s suite Ramseries.<br />
The fluid-structure coupling is performed by an implicit iterative algorithm. This way,<br />
the pressure field computed in every iteration of the diffraction-radiation solver is sent<br />
to the structural solver to compute the body deformation. The resulting displacements<br />
are sent back to the fluid solver, and used as boundary condition to compute the new<br />
pressure field for the following iterations. The iterative process continues until the<br />
convergence norm condition is fulfilled; the relative distance of the pressure vectors<br />
between two successive iterations is below a given value. The iterative algorithm is<br />
accelerated using the Aitken method [21].<br />
Since the diffraction-radiation and structural solvers are independent, the strategy used<br />
to communicate both solvers is based on the interchange of information at memory level<br />
by means of TCP-IP sockets. The library developed for this purpose, also takes care of<br />
the data interpolation from one mesh to another mesh.
45 References<br />
13. References<br />
[1] Oñate E. & García, J., A finite element method for fluid-structure interaction with<br />
surface waves using a finite calculus formulation, Comp. Methods Appl. Mech. and Eng.<br />
2001; 191: 635-660.<br />
[2] García, J., Valls A., & Oñate, E., ODDLS: A new unstructured mesh finite element<br />
method for the analysis of free surface flow problems, Int. J. umer. Meth. Fluids<br />
2008; 76 (9): 1297-1327.<br />
[3] Wu, G.X. & Eatock Taylor, R., Finite element analysis of two dimensional nonlinear<br />
transient water waves, Appl. Ocean Res. 1994 ; 16: 363-372.<br />
[4] Wu, G.X. & Eatock Taylor, R., Time stepping solution of the two dimensional nonlinear<br />
wave radiation problem, Ocean Eng. 1995; 22: 785-798.<br />
[5] Greaves, D.M., Borthwick, A.G.L., Wu, G.X. and Eatock Taylor, R., A moving<br />
boundary finite element method for fully nonlinear wave simulation, J. Ship Res. 1997;<br />
41: 181-194.<br />
[6] <strong>Ma</strong>, Q.W., Wu, G.X. and Eatock Taylor, R., Finite element simulation of fully<br />
nonlinear interaction between vertical cylinders and steep waves--part 1 methodology<br />
and numerical procedure, Int. J. umer. Meth. Fluids 2001; 36: 265-285.<br />
[7] <strong>Ma</strong>, Q.W., Wu, G.X. and Eatock Taylor, R., Finite element simulation of fully<br />
nonlinear interaction between vertical cylinders and steep waves--part 2 numerical<br />
results and validation, Int. J. umer. Meth. Fluids 2001; 36: 265-285.<br />
[8] Westhuis, J.H., The numerical simulation of nonlinear waves in the hydrodynamic<br />
model test basin, Ph.D. Thesis 2001, Universiteit Twente, The Netherlands.<br />
[9] P.X. Hu, G.X. Wu and Q.W. <strong>Ma</strong>, Numerical simulation of nonlinear wave radiation<br />
by a moving vertical cylinder, Ocean Engn . 2002; 29: 1733–1750.<br />
[10] Faltinsen O.M., Sea loads on ships and offshore structures, Cambridge Ocean<br />
Technology Series. 1998.<br />
[11] Clauss, G. F. , Schmittner, C., and Stutz, K., Time-domain investigation of a<br />
semisubmersible in rouge waves, 21st International Conference on Offshore Mechanics<br />
and Arctic Engineering June 23-28, 2002, Oslo, Norway. OMAE2002-28450.<br />
[12] Jonkman, J.M., Dynamic modelling and loads analysis of an offshore floating wind<br />
turbine, Technical report REL/TP-500-41958; November 2007.<br />
[13] Serván, B. and García, J., Advances in the Development of a Time-domain<br />
Unstructured Finite Element Method for the Analysis of Waves and Floating Structures<br />
Interaction. <strong>Ma</strong>rine 2011. Lisbon, Portugal.
46 References<br />
[14] Morison J. R., O’Brien M. P., Johnson J. W., and Schaaf S. A., 1950, The force<br />
exerted by surface waves on piles. Petroleum Transactions, American Institute of<br />
Mining Engineers, 189, pp. 149–154.<br />
[15] Recommended practice DNV-RP-C205. Environmental conditions and environmental<br />
loads. April 2007.<br />
[16] Det Norske Veritas. A Course in Ocean engineering. Available at:<br />
.<br />
[17] Wood W. L., Bossak M., and Zienkiewicz O. C. An alpha modication of<br />
Newmark's method. International Journal for Numerical Methods in Engineering, 1980,<br />
15(10):1562-1566.<br />
[18] Leonard, B. P. A stable and accurate convective modelling procedure based on<br />
quadratic upstream interpolation. Computer methods in applied mechanics and<br />
engineering 19 (1979) 59-98.<br />
[19] Pierson J.P. and Moskowitz L. A Proposed Spectral Form for Fully Developed<br />
Wind Seas Based on the Similarity <strong>Theory</strong> of S.A.Kitaigorodskii. Journal of<br />
Geophysical Research, 1964, Vol. 69, No.24.<br />
[20] Hasselmann K., Barnett R.C., Bouws E., Carlson H., Cartwright D.E., Enke K.,<br />
Ewing J.A., Gienapp H., Hasselmann D.E., Kruseman P., Meerburg A., Müller P.,<br />
Olbers, D.J., Richter K., Sell W., Walden H. Measurements of Wind-Wave Growth and<br />
Swell Decay during the Joint North Sea Wave Project (JONSWAP). Deutches<br />
Hydrographisches Institut, 1973, No.12.<br />
[21] Irons B. M., Tuck R.P. A Version of the Aitken Accelerator for Computer Iteration.<br />
International Journal for Numerical Methods in Engineering, 1969, 1:275-277.<br />
[22] Bathe, K.H. Finite element procedures. Prentice Hall, 1996.<br />
[23] Gutiérrez, J.E. Desarrollo de herramientas software para el análisis de<br />
aerogeneradores “offshore” sometidos a cargas acopladas de viento y oleaje. PhD<br />
dissertation, Universidad Politécnica de Cartagena (2014).
47 Glossary<br />
14. Glossary<br />
∇ = ( , ) : Horizontal gradient<br />
Ω : Fluid domain<br />
Γ : Body wet surface<br />
H : Water depth<br />
ρ : Water density<br />
v : Local body velocity<br />
n : Normal of the body wetted surface<br />
u : Water current<br />
U = − : Flow field from the point of view of the floating body<br />
: Free surface pressure<br />
: Wave potential<br />
: Incident wave potential<br />
: Difracted/radiated wave velocity potential<br />
: Free surface elevation<br />
: Incident wave free surface elevation<br />
: Free surface elevation of diffracted and radiated waves<br />
: Wave amplitudes<br />
: Wave frequencies<br />
: Wave number vectors<br />
: Wave directions<br />
: Wave phases<br />
Γ : Outlet boundary of the computational domain<br />
c : Prescribed wave velocity<br />
n : Normal of the outlet boundary surface of the domain
48 Glossary<br />
(): Wave dissipation damping coefficient