Beruflich Dokumente
Kultur Dokumente
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/267605999
CITATIONS
READS
259
3 authors:
Frederick Jaouen
Arjen Koop
7 PUBLICATIONS 67 CITATIONS
22 PUBLICATIONS 94 CITATIONS
SEE PROFILE
SEE PROFILE
Guilherme Vaz
Maritime Research Institute Netherlands
88 PUBLICATIONS 444 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Proceedings of ASME 30th International Conference on Ocean, Offshore and Arctic Engineering
OMAE2011
June 19-24, 2011, Rotterdam, Netherlands
OMAE2011-49085
PREDICTING ROLL ADDED MASS AND DAMPING OF A SHIP HULL SECTION
USING CFD
Frederick Jaouen
MARIN
Offshore Department
P.O. Box 28, 6700 AA Wageningen
The Netherlands
Email: f.jaouen@marin.nl
Arjen Koop
MARIN
Offshore Department
P.O. Box 28, 6700 AA Wageningen
The Netherlands
Email: a.koop@marin.nl
INTRODUCTION
Roll motion is the most difficult response of a ship to calculate, because it contains the largest viscous effects. Studying the
damping of this roll motion is necessary for a better understanding of ship motions in waves and ship motions during manoeuvring operations. The roll damping of ships is caused by various
fluid flow phenomena.
As pointed out by Ikeda et al. [6], see also Himeno [5], the
total roll damping coefficient for a ship hull can be divided into a
number of components, i.e. skin-friction stress on the hull, eddy
damping due to pressure variations on the hull, wave damping
due to free surface waves and bilge keel damping. The change
in pressure on the hull due to the bilges and bilge keels is highly
dominated by viscous effects, especially due to generated eddies
separating from the sharp edges of the bilges or bilge keels.
Computational methods based on potential flow theory are
not capable of predicting these viscous effects. Computational
Fluid Dynamics (CFD) can play an important role in the assessment of damping coefficients, viscous effects and in the understanding of complex flow behavior around the bilges and bilge
keels of the hull. CFD may offer a cost-effective method complementary to model tests to obtain damping coefficients with
reasonable accuracy and to provide detailed flow visualizations.
Therefore, a better understanding of the possibilities and limitations of CFD for determining roll damping and the sensitivity
of CFD results for various parameters by hands-on experience is
required.
To understand the complex flow behaviour in the vicinity of
bilges and bilge keels, two-dimensional calculations using CFD
are carried out on a hull section under imposed roll motion at a
certain period and amplitude. That allows to capture roll damping estimates for the type of bilges or bilge keels considered.
Furthermore, detailed visualizations of the fluid flow around
the bilge keels and the hull shape can be assessed to obtain insight in the complex flow due to bilges. An example of a PIV
study on the flow around bilge keels is given by Di Felice et
ABSTRACT
In this paper, the flow around a forced rolling body is analyzed with MARIN in-house CFD code ReFRESCO. The objective is to assess if the code can correctly predict the added
mass and damping coefficients of a rolling vessel. After a description of code and numerical methods, the results for the flow
computed around a 2D rolling hull section are presented. Sharp
and rounded bilges are investigated for three roll amplitudes and
three roll periods. The influence of grid and time discretisation
and iterative errors are analyzed. The CFD results with ReFRESCO are compared to experiments and to results obtained
with the commercial CFD package CFX. The results shown here
indicate that ReFRESCO is capable of accurately predicting the
added mass and damping coefficients. However, it is also shown
that fine grids and time-steps are required to obtain a grid and
time-step converged solution.
DEFINITIONS
B
D
g
T
Re
a
b
a
b
Guilherme Vaz
MARIN
RD Department
P.O. Box 28, 6700 AA Wageningen
The Netherlands
Email: g.vaz@marin.nl
[m]
[m]
[m/s2]
[s]
[m2/s]
[-]
[rad]
[rad/s]
[-]
[-]
[-]
[-]
[-]
[-]
c 2011 by ASME
Copyright
0 [rad]
b (Ikeda)
b wave (DIFFRAC)
b viscous
0.085
1.04E-2
7.39E-4
9.66E-3
0.102
1.28E-2
7.39E-4
1.21E-2
0.117
1.37E-2
7.39E-4
1.30E-2
0.132
1.49E-2
7.39E-4
1.42E-2
0.143
1.84E-2
7.39E-4
1.77E-2
0.153
2.16E-2
7.39E-4
2.09E-2
0.165
1.93E-2
7.39E-4
1.86E-2
0.177
2.39E-2
7.39E-4
2.32E-2
0.204
2.75E-2
7.39E-4
2.68E-2
0.232
3.03E-2
7.39E-4
2.96E-2
EXPERIMENTAL DATA
For the study herein presented, the CFD results are compared to the experimental data of Ikeda [6]. Ikeda [6] measured
the damping coefficients for a rectangular hull of dimensions
LxBxD=0.8x0.28x0.112m. The tests were carried out with a
free-surface, which means that the results include a wave damping component. For our study, we used the Ikeda hull with sharp
bilges for a period T = 1.5s.
As described by Ikeda, the damping can be decomposed (assuming linearity) in the sum of the friction damping bF , eddy
damping bE and wave damping bW , i.e. b = bE + bW + bF .
Because we carry out the CFD computations with a fully submerged hull, the wave damping is not computed and needs to be
subtracted from the experimental data of Ikeda before comparison. Therefore, before that the CFD computations were carried
out, the added mass and wave damping coefficients of the two
hull shapes were determined by diffraction analysis, using the
MARIN in-house code DIFFRAC [12]. DIFFRAC solves the
linearized velocity potential problem using a three-dimensional
source distribution technique. The part of the hull section below the mean free surface is approximated by a number of panel
elements. The distribution of source singularities on these panels forms the velocity potential describing the fluid flow around
THEORETICAL FORMULATION
The harmonic roll motion is prescribed by moving the hulls
by a roll angle (t) defined as follows
(t) = 0 sin t,
(1)
+ 12 , for t < TF T,
2TF
2
(2)
G(t) = 2
1
, for t TF T.
The roll angle (t) is now defined by
(t) = G(t)0 sin t.
(3)
The prescribed roll angle (t) leads to a hydrodynamic reaction roll moment ME (t). The equation of motion for uncoupled
c 2011 by ASME
Copyright
-10deg
a =
X
Figure 1.
M=
(4)
where a is the added mass for roll motion, B( , 0 ) the damping moment, C( ) the restoring moment, which is equal to zero
in the absence of a free surface, and ME (t) the excitation/reaction
moment. Various forms for the damping moment exist. Here, we
consider a so-called equivalent linear form, i.e. B( , 0 ) = b 0 ,
with b the roll damping coefficient. Finally, the equation of
motion becomes:
(5)
(6)
(10)
where r is the lever arm vector, i.e. the vector from the axis of
rotation to the point of force application. S is the surface of the
object, p the pressure and the viscous stress on that surface. n
is the unit normal vector pointing into the object, i.e. out of the
computational domain and t is the tangential unit vector. Since
a double body geometry is considered and the experiments of
Ikeda et al. [6] were performed on a single body geometry, the
integral in Equation (10) is only taken over the bottom half of the
configuration shown in Figure 3.
a 00 + b 0 = ME (t).
(pr n + r t) dS,
a 00 + B( , 0 ) +C( ) = ME (t),
(8)
a =
, b =
,
(9)
B2
B2 2g
M0 sin()
M0 cos()
, b =
.
2
0
0
(V Vd ) = 0,
(11)
(Vi )
+ [ (V Vd )Vi ] = ( + t ) V + VT ij ,
t
2
(p + k)ii ,
3
( V) ,
(12)
(7)
c 2011 by ASME
Copyright
where t , k are determined by a turbulence model. In the specific case here dealt with, there is no translation velocity Vs = 0
and therefore Vd = r, with = (0, 0, 0 ) and 0 derived
from Equation 3. The advantages of using a ABF formulation
are: 1) moving-grid algorithms are not needed, which decreases
the CPU time needs; 2) the variables are defined in the earthfixed reference frame and therefore there is no need for additional
post-processing; 3) the usual Coriolis and centrifugal forces are
no longer added explicitly to the right-hand-side of the equations,
which eases the iterative convergence of the calculations. For this
ABF method, boundary conditions must be defined in the earthfixed reference frame. The same method is used at MARIN for
calculations of open-water propeller characteristics with moderate CPU time costs and good iterative convergence properties.
Table 2.
Package
Grid
Grid name
Total number
elements
ICEM-CFD
Hexpress
NUMERICAL SETUP
Geometry and computational mesh
Two rectangular hull sections are investigated: one with
sharp 90 degree corners and one with rounded corners as presented in Figure 2. The thickness of the computational domain
in the third dimension is equal to 0.1m.
Table 3.
Coarse
42k
CoarseMedium
97k
Medium
173k
MediumFine
392k
Fine
698k
Coarse
61k
Medium
199k
Fine
713k
Package
Grid
Grid name
Total number
elements
ICEM-CFD
Coarse
42k
Medium
173k
Fine
698k
Numerical settings
A forced roll motion is prescribed to the hull section, and
the start-up function is applied for TF = 4 periods. Unsteady
computations are carried out solving the momentum, continuity
and turbulence equations. As initial conditions, the domain is
full of water and the pressure and velocity components are set to
zero in all cells.
For ReFRESCO, a three-time-level discretization scheme
with a fixed time-step is used for all equations. All simulations
are twenty roll periods long. Within each time-step a RMS residual drop of 106 nominal value is attempted by carrying out several outer-loops with a maximum of 30, also here tested. For
convection terms in the momentum equation a QUICK scheme
is used, while in the turbulence equation an UPWIND scheme
is used. All gradients are calculated using Gauss theorem. All
calculations are performed with the k SST turbulence model
with the eddy viscosity ratio / equal to 1 and all other turbulent quantities automatically defined in ReFRESCO.
For the calculations with CFX the high resolution convection scheme is utilized together with the second-order backwards
time discretization scheme and the SST turbulence model using
a first order UPWIND discretization scheme. The eddy viscosity / is set equal to 1 and the turbulence intensity is chosen
equal to I=1%.
c 2011 by ASME
Copyright
Outer-loop sensitivity
In the grid and time-step sensitivity studies, 30 outer-loops
per time-step were used, which was believed to be a good compromise between accuracy and CPU time. The effect of the outerloops on the results is investigated in this section. For T = 1.5s
with 0 = 0.1rad, CFD computations are performed using ReFRESCO on the medium ICEM grid with a time-step T /800 with
30, 50, 100 and 200 outer-loops respectively. It is found that the
influence of the number of outer loops on the calculated moment
is still visible. After analysis of the moment time trace, it was
found that the added mass and damping coefficients change by
10% and 8% when using 30 iterations compared with 200 iterations. It should be noted that the level of convergence only improves from 1104 to 4105 (pressure residuals). The influence
of the outer-loops on the added mass and damping coefficients is
plotted on Figure 8. In a next phase of this work, 100 outer-loop
per time-steps should be used.
c 2011 by ASME
Copyright
T = 1.5s and 0 =
= T /800s.
D = 0.112m. Re-
c 2011 by ASME
Copyright
From these comparisons we conclude that using CFX or ReFRESCO gives results that lie within 10% from each other depending on grid resolution and time-step size used. Nevertheless,
we emphasize that a fair comparison between different codes using different numerical schemes and orders of accuracy should be
only done for the grid and time-step refined solution, and taking
into account numerical uncertainties of the results. This should
be done in a next phase of this project using a V&V verification
and validation procedure presented in [3].
Solver comparison
CFD computations have been carried out for the sharp corner using the commercial CFD package CFX, and results compared to the ReFRESCO results. For CFX the three structured grids Coarse, Medium and Fine are used with time-steps
t = T /100, T /200 and T /400. The two cases T = 1.5s with
0 = 0.1rad and T = 0.75s with 0 = 0.3rad are investigated.
It is found out that the calculated moment by CFX is not
significantly affected by the time-step. However, the grid resolution clearly changes both phase and amplitude of the calculated
moment. Figure 9 compares the moment time-traces computed
by CFX and ReFRESCO and shows significant differences in
amplitude and in phase, especially for the case T = 1.5s with
0 = 0.1rad. For the case T = 0.75s with 0 = 0.3rad the differ-
B = 0.28m, D =
0.112m. Presented are the moment solutions obtained with CFX and
ReFRESCO using the medium grid and t = T /400.
Figure 9.
c 2011 by ASME
Copyright
VALIDATION
After the sensitivity studies carried out on the grid refinement, time-step, outer-loops, grid generator and solver, the amplitude and periods of the forced roll motion were varied and the
c 2011 by ASME
Copyright
T = 1.5s and
0 = 0.1rad .
c 2011 by ASME
Copyright
the CFD calculations should be done with care. For the hull
with rounded bilge keels, where viscous effects are believed to
be limited, a very good agreement is found with the results of
DIFFRAC.
Finally, Figure 16 gives a clear overview of the differences
in the viscous damping between the sharp and rounded bilge
keels.It is observed that the damping coefficient increases both
when the periods decreases and when the amplitude increases. It
can be seen that the damping of the sharp configuration is much
higher than for the rounded configuration. This is due to the fact
that more vortices are generated by the sharp edges than by the
rounded edges, which leads to a higher eddy damping component bE . On Figure 16, a comparison between the experimental
data from Ikeda (corrected for free-surface effects) and numerical results from ReFRESCO is also shown for the damping coefficients. Note that the wave making component calculated by
potential code DIFFRAC is subtracted from the experimental results as presented in Table 1. The agreement between results is
considered very satisfactory.
(a) Results for a .
CONCLUSIONS
In this paper MARINs URANS code ReFRESCO is verified
and validated for roll damping applications. The unsteady flow
around a forced rolling hull section is computed for sections with
sharp and rounded bilges. For sharp bilges extensive sensitivity
studies have been carried out. Simulations are then computed
for 3 different periods and 3 different amplitudes on both section shapes and numerical results are compared to experimental
values.
Generally, the influence of grid resolution and time-step
size on the calculated moment is found to be relatively small.
However, a larger change is found in the analyzed added mass
and damping coefficients when refining the grid and time-step.
The added mass coefficient increases for smaller time-steps and
finer grids. The damping coefficient does not show a significant change for smaller time-steps, but a clear decrease can be
observed for finer grids. These coefficients do converge when
the grid and time-step are refined. A very fine grid (more than
700,000 cells in our case) and a time-step of t = T /800 or lower
should be used. Also, the iterative error study showed that a minimum of 100 outer-loops per time-step should be used in future
work.
For the hull with sharp bilge keels, computations are carried out with structured ICEM and unstructured hexahedral Hexpress grids. The differences in the moment time trace are very
small, but differences can be noticed on the hydrodynamic coefficients. This is believed to be due to a lower convergence level
achieved when using unstructured Hexpress grids than for the
structured ICEM grids, since the same numerical settings were
used. Therefore, we recommend to increase further the number
of outer-loop iterations when using Hexpress grids to improve
the iterative convergence.
Comparing the CFX results with those obtained with ReFRESCO it is observed that the shape of the computed moment
time-trace is slightly different in amplitude as well as in phase.
For the more difficult case T = 0.75s with 0 = 0.3rad the differences are smaller. The difference between CFX and ReFRESCO
in added mass and damping coefficient are approximately 10%
10
c 2011 by ASME
Copyright
[12]
[13]
[14]
[15]
[16]
ACKNOWLEDGMENTS
The authors acknowledge Tim Bunnik and Adri van Dijk
(MARIN) for the discussions held on added mass and damping
modelling issues. We also acknowledge Francois Chanony (ENSIETA, France) and Sebastien Martin (Ecole Centrale Nantes,
France) for their interest and help in this project.
[17]
[18]
on a rolling cylinder with bilge keel in a free surface. Masters thesis, Delft University of Technology, 2005.
J.A. Pinkster and R.H.M. Huijsmans. Wave Drift Forces in
Shallow Water. BOSS1992, 1992.
S.L. Toxopeus and G. Vaz. Calculation of Current or Manoeuvring Forces using a Viscous-Flow Solver. In Proceedings of OMAE2009, Honolulu, Hawaii, USA, June 2009.
G. Vaz, F. Jaouen, and M. Hoekstra. Free-Surface Viscous
Flow Computations. Validation of URANS Code F RE SC O.
In Proceedings of OMAE2009, Honolulu, Hawaii, USA,
June 2009.
G. Vaz, S.L. Toxopeus, and S. Holmes. Calculation of
Manoeuvring Forces on Submarines Using Two ViscousFlow Solvers. In Proceedings of OMAE2010, Shanghai,
China, June 2010.
G. Vaz, O. Waals, F. Fathi, H. Ottens, T. Le Souef,
and K. Kwong. Current Affairs - Model Tests, SemiEmpirical Predictions and CFD Computations for Current Coefficients of Semi-Submersibles. In Proceedings of
OMAE2009, Honolulu, Hawaii, USA, June 2009.
J.H. Vugts. The Hydrodynamic Forces and Ship Motions in
Waves. PhD thesis, Delft University of Technology, 1970.
Y.H. Yu and S.A. Kinnas. Roll and Heave Response of Hull
Sections of Variable Shapes in Waves. In Proceedings of the
29th International Conference on Offshore Mechanics and
Arctic Engineering, number OMAE2010-20407, 2010.
REFERENCES
[1] R. Broglia et al. Experimental and numerical analysis of
the roll decay motion for a patrol boat. ISOPE, June 2009.
[2] F. Di Felice, A. Dolcini, F. Perreira, and C. Lugni. Flow
Survey around the Bilge Keel of a Ship Model in Free Roll
Decay. In 6th International Symposium on Particle Image Velocimetry, PIV05, number 115, Pasadena, CA. USA,
2005.
[3] L. Eca, G. Vaz, and M. Hoekstra. A Verification and Validation Exercise for the Flow Over a Backward Facing Step.
In Proceedings of ECCOMAS-CFD2010, Lisbon, Portugal,
June 2010.
[4] F. Fathi, C.M. Klaij, and A. Koop. Predicting Loads On
a LNG Carrier with CFD. In Proceedings of OMAE2010,
Shanghai, China, June 2010.
[5] Y. Himeno. Prediction of Ship Roll Damping, State of the
Art. Technical Report 239, University of Michigan, 1981.
[6] Y Ikeda, Y. Himeno, and N. Tanaka. Components of Roll
Damping of Ship at Forward Speed. JSNA japan, 143,
1978.
[7] J.M.J. Journee and W.W. Massie. Offshore Hydromechanics. Delft University of Technology, 2001.
[8] B. Kacham. Inviscid and Viscous 2D Unsteady Flow
Solvers Applied to FPSO Hull Roll Motions. Masters thesis, University of Texas, 2004.
[9] A. Koop, C.M. Klaij, and G. Vaz. Predicting Wind Shielding for FPSO Tandem Offloading using CFD. In Proceedings of OMAE2010, Shanghai, China, June 2010.
[10] J.H. Na, W.C. Lee, H.S. Shin, and I.K. Park. A Design of
Bilge Keels for Harsh Environments FPSOs. In Proceedings of the Twelfth International Offshore and Polar Engineering Conference, 2002.
[11] M. Paap. Verification of CFD calculations with experiments
11
c 2011 by ASME
Copyright