1-s2.0-S2092678221000613-main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

Contents lists available at ScienceDirect

International Journal of Naval Architecture and Ocean Engineering


journal homepage: http://www.journals.elsevier.com/
international-journal-of-naval-architecture-and-ocean-engineering/

Study of the lift effect on wind load estimation for a semi-submersible


rig using the maritime atmospheric boundary layer model
Seong Mo Yeon*, Chang Seop Kwon, Yoo Chul Kim, Kwang Soo Kim
Korea Research Institute of Ships & Ocean Engineering (KRISO), Daejeon, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: Wind loads on a semi-submersible rig were investigated using computational fluid dynamics. A maritime
Received 13 September 2021 atmospheric boundary layer model for wind profile was implemented such that the wind profile shapes
Received in revised form were retained throughout the computational domain. Wind loads on the semi-submersible rig were
11 November 2021
calculated under the maritime atmospheric boundary layer and matched well with the results from wind
Accepted 16 November 2021
tunnel within a ±20% error. Among the topside structures, five structures including deckbox were
Available online 24 November 2021
selected and analyzed on the contribution to the wind load, particularly overturning moments by
decomposing the moments into drag and lift components. Overall, moments ignoring lift components
Keywords:
Maritime atmospheric boundary layer
tended to overestimate overturning moments by approximately 20% at maximum. The majority of the lift
OpenFOAM components originated from the deckbox, which served as a lifting body owing to the accelerated
Semi-submersible rig streamlines between the waterline and the bottom of the deckbox.
Wind load © 2021 Society of Naval Architects of Korea. Production and hosting by Elsevier B.V. This is an open access
article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

1. Introduction decomposing all the complex topside structures into rudimentary


bodies such as cylinders, cuboids, and flat plates whose drag co-
Offshore floating structures are frequently designed for a spe- efficients are known. In the building block methods, the wind loads
cific site and exposed to environmental loads such as wind, wave, are calculated as a linear combination of drags of the primitive
current, and tides. In specific scenarios such as the Arctic and sub- bodies. Occasionally, databases storing the shielding effect between
Arctic regions, rain and snow may be considered important since primitive bodies can be used to improve the estimation, as in
the accumulation on the offshore structures can reduce both the WINDOS (Walree, 2015) or Eagle WIND of the American Bureau of
stability and operability of the platform and impair structural Shipping (ABS). However, it is known that these techniques tend to
integrity(Lie et al., 2017; Meløysund et al., 2007), which may be overestimate the wind load owing to the simplified calculation of
subject to wind. Since the environmental loads are mostly sto- the three-dimensional flow phenomenon and exhibit inconsistent
chastic, the environmental loads are frequently determined results between practitioners owing to the strong dependency on
through probabilistic approaches (Rahman et al., 2015; Bhatia and the experience of engineers.
Khan, 2019; Ralph, 2016). For the estimation of wind load on offshore structures, Norwe-
Owing to the importance of aerodynamics and hydrodynamics gian Petroleum Directorate (NPD) profile, which is based on the
for the afloat stability and mooring analysis of the offshore floating wind measurement in Frøya island of Norway (Andersen and
structures, the wind load is one of the main environmental loads in Løvseth, 2006), or API (API, 2000) profile, which is an approxi-
the design of offshore floating structures. Wind loads on offshore mate NPD profile, are frequently used. The NPD profile, also known
structures are traditionally estimated for a prescribed wind profile as the Frøya wind model, is a log wind profile observed in the ocean
through wind tunnel tests instead of probabilistic approaches. In (Andersen and Løvseth, 2006). Since the design wind speeds of
the initial design stage, in which it is difficult to perform model offshore structures are in the regime of fully turbulent flow and the
tests, empirical building block methods are often used by aerodynamic characteristics are plateaued against wind speed
variations, we can observe that the wind load is more affected by
the shape of the wind profile rather than the design wind speed
itself.
* Corresponding author.
Owing to the recent advancement in hardware performance and
E-mail address: seongmo.yeon@kriso.re.kr (S.M. Yeon).
Peer review under responsibility of The Society of Naval Architects of Korea. numerical methods used in Computational Fluid Dynamics (CFD),

https://doi.org/10.1016/j.ijnaoe.2021.11.002
2092-6782/© 2021 Society of Naval Architects of Korea. Production and hosting by Elsevier B.V. This is an open access article under the CC BY license (http://
creativecommons.org/licenses/by/4.0/).
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

Nomenclature k Turbulent kinetic energy


K0 Normalized moment for Mx, defined as qMrefx
n Kinematic viscosity of air L Characteristic length
M
nt Turbulence eddy viscosity M0 Normalized moment for My, defined as qrefy
u Specific dissipation rate Mx Moment around the x direction
qref Reference dynamic pressure defined as 12 rU 2ref My Moment around the y direction
r Density of air Mz Moment around the z direction
z0 Roughness length N0 Normalized moment for Mz, defined as qMrefz
Uref Reference speed at the reference height R Convergence Ratio
u* Friction velocity Re Reynolds number defined as Re ¼ UL n
zref Reference height U Wind speed
fine coarse
Er Relative Error defined as Solution Solution X0 Normalized force for Fx, defined as qFrefx
coarse
Solution F
Fx Force in the x direction Y0 Normalized force for Fy, defined as qrefy
Fy Force in the y direction Z0 Normalized force for Fz, defined as qFrefz
Fz Force in the z direction

research has been conducted on reproducing log wind profiles The NPD wind profile is also a log wind profile and has the
numerically as an alternative tool to wind tunnel tests. Richards following form:
and Hoxey (1993) was the first researchers to successfully imple- " !#
ment log wind profiles using commercial CFD software. They z
UðzÞ ¼ Uref 1 þ Clog
proved the log wind profiles could be the solutions of the Navier- zref (2)
Stokes equations and derived boundary conditions for the steady- pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
state atmospheric boundary layer. The wind profile they imple- C¼ 0:0573 1 þ 0:15Uref
mented is referred to as Horizontally Homogeneous Turbulent
Boundary Layer (HHTBL), representing a fully-developed turbulent where Uref is the reference wind speed measured at zref, which is
boundary layer profile since the wind profile is maintained the reference height, typically 10 m above still water.
throughout the computational domain without the necessity for a Equating Eqs. (1) and (2) yields analytic forms of the roughness
fetch length. Numerically, the HHTBL is advantageous in that it length and friction velocity as follows:
provides a practical method of generating wind profiles with a  
small computational domain. The SNAME OC-8 Comparative Wind 1
z0 ¼ zref exp 
Load Study (2016  2018) (Kim et al., 2018) applied Richards and C (3)
Hoxey (1993)'s approach to the NPD profile and derived the mari- u* ¼ kUref C
time atmospheric boundary layer model (MABL) and demonstrated
that the MABL is generated within 1% tolerance from the prescribed
NPD profile between participants. Recent research performed by
four participants including TechnipFMC, Exxon Mobil, Samsung 2.2. Boundary conditions for turbulence
Heavy Industries (SHI), and Korea Research Institute of Ships &
Ocean Engineering (KRISO) applied the MABL to a Floating Pro- It is known that for the NPD profile, the turbulent kinetic energy
duction Storage and Offloading (FPSO) and a semi-submersible rig k has the following form (Blocken et al., 2007; Richards and Hoxey,
and observed that the wind loads matched closely with model test 1993; Richards and Norris, 2011):
results (Yeon et al., 2019).
This study aimed to reproduce a wind profile and determine the u2*
kz ¼ pffiffiffiffiffi (4)
load factor that the building block methods missed by examining b*
the wind load distribution of the semi-submersible rig using an
open source CFD development framework, OpenFOAM (Weller Since the turbulence eddy viscosity nt satisfies the following
et al., 1998). Section 2 outlines the generation process of the relation:
MABL. Section 3 discusses the results of wind load simulation in
k
terms of drag and lift components on each topside structures. nt ¼ (5)
u
2. Modeling of the wind profile the specific dissipation rate u used in the k  u SST turbulence
model of RANS method can be derived as follows:
2.1. Boundary condition for velocity
* u
u ¼ pffiffiffiffiffi (6)
A frequently used wind profile in the design of offshore floating k b* z
structures is a log wind profile, which is expressed as (DNV, 2010):
 
u* z
UðzÞ ¼ log (1)
k z0 2.3. Sustainability condition

where k is the von Karman constant between 0.4 and 0.42, u* the Assuming a steady-state and the velocity profile as a function of
friction velocity, z0 the roughness length which is determined elevation, a high Re reduces the Navier-Stokes equations without
experimentally and z the elevation. pressure gradient to the following relation (Richards and Hoxey,1993):
2
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

vU
nt ¼ constant (7)
vz
Since the constant shear stress should be maintained to the
bottom boundary, Eq. (7) should be balanced with the wall shear
stress tw:

vU 
rnt ¼ tw ¼ ru2* (8)
vz z¼0
Thus, the turbulence eddy viscosity should satisfy the following
form to retain the wind profile.

nt ¼ ku* z (9)
In meteorology, the section in which the shear stress is constant
corresponds to the section referred to as the surface layer (Mo€lders
and Kramm, 2014). If nt in flow field satisfies Eq. (9), the flow field
can be considered to follow the MABL model. Most ships and
offshore structures operate in the surface layer because approxi-
mately 100e150 m above the water surface is generally considered
the surface layer, although it varies depending on atmospheric
conditions.

3. Results

3.1. Wind profile test

To verify that the MABL is maintained throughout the compu-


tational domain, a test was performed on an empty domain of the
cuboid shape as shown in Fig. 1. The origin of the coordinate is
located at (0m, 0m, 0m). Six boundaries were designated as inlet,
outlet, sides, bottom and top at x ¼  3.52m, x ¼ 3.52m,
y ¼ ±3.52m, z ¼ 0m, and z ¼ 3.0m, respectively. The mesh size of
the domain was 26 cells in the x and y directions and 37 cells in the
z direction. Uniform distributions were considered in the x and y
directions. In the z direction, first cell size was set to 0.004m cor-
responding to yþ z 286. The target wind speed was 51.444 m/s at
the model scale, and corresponding roughness length z0 and fric-
tion velocity u* were approximately 0.000113m and 3.655 m/s with Fig. 1. Empty fetch domain.
k ¼ 0.42, respectively. To measure the vertical distribution of wind
profiles, line probes were installed at 5 locations along the x axis as
shown in Fig. 2; inlet, x ¼  1.76m, x ¼ 0m, x ¼ 1.76m and outlet.
The k  u SST model was used for the turbulence model. line-
arUpwind and linear schemes, which are second-order accuracy
schemes, were applied to the convection and diffusion terms,
respectively. The coupling of pressure and velocity were calculated
using the semi-implicit method for pressure linked equation-
s(SIMPLE) algorithm.
Since MABL implementation is a Couette flow (Yeon et al., 2020),
the boundary condition for the top boundary should be a Dirichlet
type. Thus, velocity, the k and u profiles described in Eqs. (2), (4)
and (6) should be imposed on the boundaries including top, inlet,
outlet, and side. For the bottom boundary, wall boundary condi- Fig. 2. Position of line probes.
tions with roughness should be applied as described in Kim et al.
(2018) and Yeon et al. (2020).
The results of wind profile test are shown in Fig. 3. The field
values measured in the line probes were compared with the NPD 3.2. Wind load distribution of semi-submersible rig
profile assigned in inlet boundary as shown in Fig. 3(a), and the
relative error of the wind profiles measured using the line probes is The target hull shown in Fig. 4 was the semi-submersible rig
shown in Fig. 3(b). Up to z/zref ¼ 2(20 m in full scale) from the used for the benchmark test in the SNAME OC-8 Comparative Wind
bottom boundary, the relative error was within ±2%. From z/zref ¼ 2 Load Study (Kim et al., 2018). The volume mesh for the rig is as
and above, the field values were almost identical to the given NPD shown in Fig. 5. To better capture the wake field separated from the
profile. The calculated eddy viscosity nt followed linear behavior as rig, three concentric circular cylinders were added around the body
shown in Fig. 3(c). Thus, the MABL was well maintained throughout as volume refinement zones. The mesh was created as a
the calculation domain. hexahedral-based unstructured grid. The mesh distribution around
3
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

Fig. 4. Semi-submersible rig.

Fig. 3. Wind profiles in empty fetch domain.

the body is shown in Fig. 6. Since the body is primarily composed of


blunt bodies with sharp edges, not much attention was given to the
creation of the prism layers resolving the boundary layers as shown
in Fig. 6(b). The main dimensions of the rig are described in Table 1.
Based on the dimensions described in Table 1, the blockage ratio
was 1.0%. The heading angle used in the calculation was considered Fig. 5. Mesh in computational domain. (a) overall mesh around the body. (b) enlarged
as a total of 12 wind directions at intervals of 30 from 0 to 360 . mesh around the body.

4
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

Fig. 7. Heading convention.

Table 2
Mesh sensitivity test.

Mesh Size X0 Y0

[mil.] 0 30 0 30
Coarse 5.5 0.114 0.125 e 0.055
(10.1%) (26.3%) e (8.9%)
Medium 17.0 0.109 0.115 e 0.052
(5.1%) (16.4%) e (14.2%)
Fine 35.0 0.107 0.111 e 0.051
(3.9%) (12.7%) e (16.0%)
R e 0.252 0.376 e 0.347
% UG e 0.22 0.52 e 0.16
EFD e 0.103 0.099 e 0.061

normalized to obtain wind load coefficients to avoid the use of a


Fig. 6. Mesh around the semi-submersible rig. specific reference area or reference length as per SNAME T&R
Bulletin 5-4 (2020) as follows:

Table 1
Fx
X 0 ¼ CFx  S ¼
Principal dimensions of a semi-submersible rig.

Full scale
qref
Fy
Distance between columns [m] about 88 Y 0 ¼ CFy  S ¼
Height from waterline [m] about 124 qref
Deckbox width [m] about 75
Fz
Deckbox length [m] about 66 Z 0 ¼ CFz  S ¼
Draft [m] 53.28 qref
scale ratio 240 (10)
Mx
K 0 ¼ CMx  S  L ¼
qref
The heading angles were defined as a counterclockwise direction My
M 0 ¼ C My  S  L ¼
with 0 blowing from the stern and 180 blowing from the bow qref
according to the OCIMF convention (OCIMF, 1994) as shown in Mz
Fig. 7. For the numerical method used for the calculation, the same N 0 ¼ CMz  S  L ¼
qref
settings as those used in the wind profile test were applied.
To determine the appropriate mesh size, grid sensitivity tests where qref is the reference dynamic pressure at the reference
were performed for the wind direction angles of 0 and 30 as in
height, defined as 12 rU 2ref , S the projected area and L the charac-
Table 2. Simulations were run for 10, 000 iterations. The averaged
teristic length.
forces and moments were calculated using the last 1/3 data set and
The convergence ratio of the grid was determined using the
5
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

convergence ratio R defined as follows (Xing and Stern, 2010):

e21

e32
e21 ¼ Solutionmedium  Solutionfine (11)

e32 ¼ Solutioncoarse  Solutionmedium


The convergence ratio for 0 heading angle was 0.252, which
means a monotonic convergence. The wind load in the transversal
direction was excluded from the comparison owing to its small
magnitude. The convergence ratios for 30 heading angle were 0.376
and 0.347 in the longitudinal and transversal directions, respec-
tively, which indicated monotonic convergence. Uncertainty anal-
ysis, UG, was conducted using a Correction Factor (CF) method
following Wilson et al. (2004). Since iterative convergence was
Fig. 8. Wall time for each heading angle.
achieved and iterative uncertainty is negligible with respect to grid
errors in engineering problems (Wilson et al., 2004; Xing and Stern,
2010), the iterative uncertainty, UI, was not considered. The grid and viscous component. Since the topside structures are composed
uncertainty, UG, is defined as follows: of structures with sharp edges, most of the flow can be considered
( to be dominated by the flow caused by the separation at the edge
*
½9:6ð1  Ck Þ2 þ 1:1jdRE j j1  Ck j < 0:125 rather than the flow caused by the development of the boundary
UG ¼ *
(12) layer. Therefore, only the pressure components are considered in
½2j1  Ck j þ 1jdRE j j1  Ck j  0:125
the subsequent analysis.

where Ck is the correction factor and d*RE the one-term Richardson ! !


^ k A k þ s,n
! ! !
^k A k ¼ F p þ F s
F ¼ pI,n
Extrapolation (RE) estimates for the error, defined as 2 3
pnx
rp  1 ! 6 7 !
Ck ¼ (13) Fp ¼ 6 7
4 pny 5k A k
r pest  1
pnz (16)
e 2 32 3 2 3
d*RE ¼ p 21 (14) sxx sxy sxz nx tx
r 1 ! 6 76 7 ! 6 7 !
6
F s ¼ 4 syx syy syz 54 ny 5k A k ¼ 4 ty 7
6 7 7 6
5k A k
lnðe32 =e21 Þ szx szy szz nz tz
p¼ (15)
ln r

where r is the grid refinement ratio, which was set to 1.41, p the
order of accuracy and pest the estimate for the limiting order of 3.3. Moment distribution on the topside structures
accuracy of the first term in the RE expansion as the grid spacing
size tends to zero and asymptotic range is reached. Since pest is Generally, the building block method uses an empirical formula
frequently approximated to the theoretical order of the numerical expressed in Eq. (17).
method pth, pth was set to 2 for a second order of accurate method. It
was observed that the estimated grid uncertainties were 0.22%, 1
F¼ rCs Ch U 2ref A (17)
0.52%, and 0.16%, respectively and significantly less than relative 2
error between CFD and EFD. Since the relative error between the
where Cs is the shape coefficient, Ch the height coefficient derived
medium mesh and fine mesh was approximately 1e3% and the
from the API profile and A the projection area. Since the overturning
discrimination of the results between medium and fine meshes was
moment is considered an important design factor for the design of
negligible, the medium mesh was selected for the subsequent wind
semi-submersible rigs, roll and pitch moments related to the
load simulation. The relative errors between the medium grid and
overturning moment are calculated in the initial design. In the
the model test results (EFD), which were obtained from the
building block methods, the overturning moment is obtained by
benchmark results of KRISO for SNAME OC-8 Comparative Wind
multiplying the distance from the water surface to the centroid of
Load Study (Kim et al., 2018), were 5.1%, 16.4%, and 14.2%,
the simplified topside structure by the force obtained in Eq. (17).
respectively.
Therefore, only roll and pitch moments are used in subsequent
The simulation for the all heading angles were calculated using
analysis.
320 cores in a Xeon Phi Cluster (Intel Xeon Phi 7250 1.4 GHz), and
The moment can be divided into two components as shown in
the wall time for each heading angle is plotted in Fig. 8. An average
Eq. (18). For roll and pitch moments, it is possible to separate into
of approximately 7.3 h was required for each heading angle for
the component due to the horizontal force(drag) and vertical
10,000 iterations.
force(lift). The yaw moment is expressed only using drag
The results of the simulation were compared with the results of
components.
wind tunnel test, as shown in Figs. 9 and 10, in which the wind
tunnel test results are overlaid with 20% error bars. It is observed 2 3 2 3
ry Fz þ ðrz Fy Þ
that the CFD calculation results corresponded closely to the ! 4 Mx 5 X ! !
M ¼ My ¼ ð r  F Þ ¼ 4 rz Fx þ ðrx Fz Þ 5 (18)
experimental results within the overall error range. Mz rx Fy þ ðry Fx Þ
The wind load can be divided into pressure and viscous com-
ponents as in Eq. (16). Fig. 11 compares the pressure component Fig. 12 is a chart comparing the moment components for the roll
6
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

Fig. 9. Wind load coefficients (forces). Fig. 10. Wind load coefficients (moments).

and pitch moments due to wind direction. Wind load coefficients practitioner's experience. Although the coefficients using the
from building block method were overlaid according to ABS (2001). building block method are overestimated, the overall trend fol-
Since the target hull is composed of mostly blunt bodies with sharp lowed the CFD results. The overestimation was primarily observed
edges, the shape coefficient for each block was set to 1.0 except for where the lift component existed and the results using CFD and
flare tower which was set to 0.5 for cylindrical body. Please note building block method were almost identical where the lift
that the wind load coefficients from the building block method component hardly existed. For the roll moment shown in Fig. 12(a),
could vary since the selection of shape coefficients depends on the the moment due to lift occupied approximately 10% to the moment

7
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

Fig. 12. Moment compsrison for drag and lift components.

Fig. 11. Comparison of pressure and viscous components.

by drag at heading angles of 60 , 120 , 240 and 300 . For the pitch
moment shown in Fig. 12(b), the moment due to lift occupied
slightly larger parts than in the roll moment and exhibited
approximately 15  20% to the moment due to drag at heading
angles of 60 , 120  240 and 300 . This implied that the building
block methods can overestimate the overturning moments for the
semi-submersible rig by approximately 10e20% owing to the
moment contribution from the lift not being considered.
The distribution of moment components due to the drag and lift Fig. 13. Main topside structures.
in the wind load moments was examined by classifying the topside

8
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

Fig. 14. Roll moment distribution by topside structures. Fig. 15. Pitch moment distribution by topside structures.

structures into LQ, helideck, derrick, columns, flare tower, lower Fig. 15 is a chart showing the pitch moment distribution for each
deckbox, and upper deckbox as shown in Fig. 13. structure and generally shows a similar tendency to the roll
Fig. 14 is a chart showing the roll moment component distri- moment distribution. Fig. 15(a) shows the moment distribution due
bution for each structure. Fig. 14(a) shows the moment distribution to the lift component of each structure. For the heading angles at
due to the lift component for each structure. For the heading angles which the pitch moments were large, the deckboxes occupied
at which the pitch moments were large, the deckboxes including 60e70% of the total magnitude followed by the helideck. Fig. 15(b)
lower and upper deckboxes occupied 80e90% of the total magni- shows the moment distribution due to the drag component of each
tude followed by the helideck, and magnitudes were comparable. structure. The derrick, which had a large projected area exposed to
For the helideck, the sheared wind profile across the flat plate, the freestream in the longitudinal direction, occupied approxi-
which is for takeoff and landing of helicopters, produced the lift due mately 70e85% of the total magnitude followed by the helideck.
to the different pressure distribution on the top and bottom Figs. 14 and 15 show that most of the moment caused by the lift
surfaces. component were primarily due to the flat plates such as the
Fig. 14(b) shows the moment distribution due to the drag deckbox and helideck. The lift center, lift, and derived roll moment
component for each structure. The derrick, which had a large and pitch moment due to lift on the deckboxes are shown in Fig. 16.
projected area exposed to the freestream in the transverse direc- The lift centers were calculated using Eq. (18) and are plotted in
tion, occupied approximately 80e90% of the total magnitude fol- Fig. 16(a). For the upper deckbox, the lift centers were distributed
lowed by columns. primarily in the second (270  300 heading angles) and fourth

9
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

Fig. 16. Moment distribution on deckboxes.

quadrant (90  120 heading angles). For the lower deckbox, the lift
centers were distributed primarily in the first quadrant (210  270).
The magnitudes of the lift forces on deckboxes are plotted in
Fig. 16(b). The lower deckbox had a larger magnitude than the
upper deckbox and was clustered in the third quadrant (0  90
heading angles). The magnitude of the roll moment by lift is plotted
in Fig. 16(c). The moment distribution for both deckboxes was
primarily clustered around the 90 and 270 heading angles in which
roll moments were maximized. The magnitudes of the both deck-
boxes were comparable owing to the different lift center and lift
force distributions. The magnitudes of the pitch moment due to lift Fig. 17. Streamlines around semi-submersible at a heading angle of 90.
is plotted in Fig. 16(d). The moment distributions for both deck-
boxes were primarily clustered around the 0 and 180 heading an-
the lower deckbox.
gles in which pitch moments were maximized. The lower deckbox
had a larger magnitude than the upper deckbox.
To examine the involved physics around the deckboxes more 4. Conclusion
clearly, the streamlines passing above and below the deckboxes
were visualized as shown in Fig. 17. It was observed that since the In this study, a numerical approach using the MABL model was
field above the upper deckbox was blocked by topside structures used to identify the factors causing the inaccuracy of the wind load
such as the derrick, the streamlines exhibited three-dimensional estimation methods based on the empirical formula used in initial
behavior and then formed complex flow structures in the wake. designs.
In the lower part of the lower deckbox, the flow was accelerated By implementing the MABL using OpenFOAM, it was observed
and rectified into a two-dimensional form in a relatively narrow that the prescribed NPD profile was maintained in almost the same
space between the waterline and lower deckbox. As a result, the lift shape throughout the computational domain. The reproduced
was generated by the low pressure formed in the bottom surface of wind profile was applied to a semi-submersible rig, and showed

10
S.M. Yeon, C.S. Kwon, Y.C. Kim et al. International Journal of Naval Architecture and Ocean Engineering 14 (2022) 100419

that the calculated results matched closely with model test results environment” funded by KRISO (PES3911).
within a ±20% error.
The wind load moment acting on the rig was separated into drag
and lift components. Most of the moment was derived using the
References
drag component, but it was observed that the moment due to the
lift component occupied more than approximately 10e20% to the ABS, 2001. Rules for Building and Classing Mobile Offshore Drilling Units.
total magnitude. Therefore, if only the moment due to the drag Andersen, O.J., Løvseth, J., 2006. The Frøya database and maritime boundary layer
wind description. Mar. Struct. 173e192.
component were considered as in the building block methods, it
API, 2000. Recommended Practice for Planning, Designing and Constructing Fixed
could be expected that overestimation of 10e20% would occur Offshore PlatformseWorking Stress Design. American Petroleum Institute.
when estimating the overturning moment. Bhatia, K., Khan, F., 2019. A predictive model to estimate ice accumulation on ship
The topside structures were divided by major structures and the and offshore rig. Ocean Eng. 173, 68e76.
Blocken, B., Stathopoulos, T., Carmeliet, J., 2007. CFD Simulation of the Atmospheric
overturning moment components due to the drag and lift were Boundary Layer: Wall Function Problems. Atmospheric Environment,
compared over heading angles. Most of the overturning moments pp. 238e252.
were caused by the drag, but it was observed that the contribution DNV, 2010. DNV-RP-C205: Environmental Conditions and Environmental Loads
October 2010 - Recommended Practice. DNV.
due to the lift component was dominated by the flat plate objects Kim, J., Jang, H., Xu, W., Shen, Z., Kara, M., Yeon, S., 2018. Numerical modeling of
like deckboxes up to 70% or more of the total magnitude of the neutrally-stable and sustainable atmoshperic boundary layer for the wind load
moment due to the lift. The streamline distribution around the estimation on an offshore platform. In: ASME 2018 37th International Confer-
ence on Ocean. Offshore & Arctic Engineering, Madrid, Spain. V001T01A006.
deckboxes demonstrated that the streamlines passing through the Lie, B.B., Teigen, S.H., Hansen, E.S., Eik, K.J., 2017. Estimation of characteristic snow
narrow space between the bottom surface of the lower deckbox loads on offshore structures in the barents sea. Cold Reg. Sci. Technol. 139,
and the waterline at high speed were formed and caused the 22e35.
Meløysund, V., Lisø, K.R., Hygen, H.O., Høiseth, K.V., Leira, B., 2007. Effects of wind
deckbox to behave as a lifting body such as an airfoil. It can be exposure on roof snow loads. Build. Environ. 42, 3726e3736.
deduced that the consideration of the contribution of the lift Mo€ lders, N., Kramm, G., 2014. Lectures in Meteorology. Springer.
component to the moment estimation can be improved for objects OCIMF, 1994. Prediction of Wind and Current Loads on VLCCs. Oil Companies In-
ternational Marine Forum London.
such as deckboxes and helideck, which are frequently idealized as ~ es Re
, A., Veitch, B., 2015. Probabilistic
Rahman, M.S., Taylor, R.S., Kennedy, A., Simo
flat plates in the building block methods. analysis of local ice loads on a lifeboat measured in full-scale field trials.
In the future research, it is considered necessary to study a J. Offshore Mech. Arctic Eng. 137.
technique that can supplement the accuracy of the initial predic- Ralph, F.E., 2016. Design of Ships and Offshore Structures: a Probabilistic Approach
for Multi-Year Ice and Iceberg Impact Loads for Decision-Making with Uncer-
tion by considering the effect of the lift of the topside structures of tainty. Ph.D. thesis. Memorial University of Newfoundland.
the offshore structures. Richards, P., Hoxey, R., 1993. Appropriate boundary conditions for computational
wind engineering models using the k-e turbulence model. J. Wind Eng. Ind.
Aerod. 46, 145e153.
Declaration of competing interest Richards, P.J., Norris, S.E., 2011. Appropriate boundary conditions for computational
wind engineering models revisited. J. Wind Eng. Ind. Aerod. 257e266.
The authors declare that they have no known competing Walree, F.v., 2015. WINDOS. MARIN, Wageningen, the Netherlands [Computer
program], Version 9.0.3.
financial interests or personal relationships that could have Weller, H.G., Tabor, G., Jasak, H., Fureby, C., 1998. A tensorial approach to compu-
appeared to influence the work reported in this paper. tational continuum mechanics using object-oriented techniques. Comput. Phys.
12, 620e631.
Wilson, R., Shao, J., Stern, F., 2004. Discussion: criticisms of the “correction factor”
Acknowledgements verification method. J. Fluid Eng. 126, 704e706.
Xing, T., Stern, F., 2010. Factors of safety for richardson extrapolation. J. Fluid Eng.
The semi-submersible geometry used in this study was derived 132, 061403.
Yeon, S., Jang, H., Kim, J., Kim, J., Nam, B., Huang, Z., O'Sullivan, J., Kim, H., Hong, S.,
from Houston Offshore Engineering's Paired-Column Semi-
2019. Numerical modeling practice and verification of the wind load estimation
submersible (PC-Semi). This research was supported by a grant for FPSO and semi-submersible. In: ASME 2019 38th International Conference
from Endowment Project of “Development of original technology on Ocean. Offshore and Arctic Engineering, Glasgow, Scotland, UK.
for smart navigation of advanced ships considering low speed and V001T01A007.
Yeon, S.M., Kim, J.S., Kim, H.J., 2020. Numerical wind load estimation of offshore
scale effect” funded by KRISO (PES3860) and “Development of floating structures through sustainable maritime atmospheric boundary layer.
evaluation technology for ship's performance in extreme International Journal of Naval Architecture and Ocean Engineering 12, 819e831.

11

You might also like