You are on page 1of 5

MARSTRUCT 2011

3rd International Conference on Marine Structures


28 30 March 2011, Hamburg, Germany

Influence of Raised Invar Edges on Sloshing Impact Pressures - Numerical Investigations


Stefano Brizzolara & Diego Villa
Marine CFD Group Faculty of Engineering, University Of Genova, Italy

&
Thomas Gazzola, Nikita Tryaskin, Nicolas Moirod, Jrme De Lauzon & Louis Diebold
Bureau Veritas, Marine Division, Research Department
92571 Neuilly-Sur-Seine Cdx - France

ABSTRACT
This paper presents a numerical investigation of the influence of raised
invar edges on wet drop tests pressures.
The aim of this study is to evaluate the capabilities of the CFD software
OpenFOAM ([6]) to deal with fluid structure impact problems in the
context of sloshing inside membrane tanks. More precisely, the
objective is to evaluate OpenFOAM capabilities for water drop tests
simulations for a smooth wedge on one hand and a wedge equipped
with invar edges (like these which equip the Cargo Containment
System NO96 produced by GTT) on the other hand, both falling into
calm water without inclination. The obtained numerical results are then
compared to Wagners solution for the smooth wedge and to
experimental measurements for the wedge with edges presented in
([1]). Numerical simulations are in very good agreement with Wagners
solution and experimental results ([1]) showing the OpenFOAM
capability to deal with fluid structure impact problems in the context of
sloshing inside membrane tanks. The agreement between numerical and
experimental results confirms also that raised invar edges tend to
enhance the magnitude of sloshing pressures. This confirmation
emphasizes the importance of considering the physics of invar edge
effects in defining the design pressure to be used in assessing the
integrity of membrane LNG tanks.

KEY WORDS: LNG carrier; Membrane Containment system; Raised

Indeed, sloshing model tests are carried out using tanks with smooth
walls. However at prototype scale, the two widely-used cargo
containment systems (CCS) both have raised elements, corrugations in
the case of MarkIII and raised invar edges in the case of NO96. The
MarkIII primary membrane includes a square pattern of corrugation
cells formed by the crossing rows of larger and smaller corrugations,
both with spacings of roughly 340mm. In the case of NO96 system,
parallel rows of raised invar edges are present with spacings of roughly
500 mm and contain the weld used to join the sheets of invar that make
up the primary membrane.
In order to better understand the influence of these raised elements on
sloshing pressures, different studies have been carried out. Due to
complexity of sloshing model tests using raised edges, most of these
studies consisted in performing wedge drop tests widely used in the
industry as a means to investigate fluid impact problems. The main
conclusion of these studies was that corrugations significantly reduced
the magnitude of impact pressures by factors at least 2 when compared
to pressures measured on smooth wedge. The main explanation for
such reduction was the trapped air cushioning effect.
However, a recent study ([1]) based on drop tests and 2D sloshing tests
showed that raised elements effect on sloshing pressures is complex
and may not lead necessarily to lower pressures when compared to the
smooth cases.
The purpose of the present paper is first to check/confirm the findings
presented in ([1]) and then to evaluate OpenFOAM capability to deal
with complex fluid structure impact problems by comparing our
numerical results to the experimental results ([1]) for the wet drop tests
with & without invar edges.

Invar Edges; Liquid sloshing, OpenFOAM

1 INTRODUCTION
In order to manage the risk of failure due to sloshing, adequate
assessment of sloshing loads and structural capacities are required.
Even if the state of the art of sloshing model tests has improved a lot
these last few years - by including more physics such as reproduction
of realistic irregular 6 degrees of freedom motion for the tank,
introduction of a special ullage gas mixture in order to respect the LNG
gas/fluid ratio density in the model tank - some physical phenomena are
still not reproduced in actual sloshing model tests.

MARSTRUCT 2010 Lead Author: Stefano Brizzolara

Present work is complementary to the long validation study performed


on different CFD methods (unsteady free surface commercial RANSE
solvers and proprietary SPH method) with regards to impact load of
different ship like section shapes, performed within MARSTRUCT
activities and already published in different occasions [2].

2 WEDGE DROP TESTS SETUP


2 - 1 Equations for the impact problem
The simulation of the impact of the falling wedge in calm water is
assumed uncoupled. The body is rigid, and its position and velocity is
prescribed and imposed.

page 1/5

2 - 2 Setup of the numerical simulations


The wedges configurations are the same one as the ones described in
([1]). For sake of simplicity, only the scale 1:6 is here presented in this
section. For the 2 other scales, one should refer to ([1]).

Fig. 1: Drop Test scheme


This impact problem can be solved in several equivalent ways. It is
possible to impose a flow at a specific velocity that comes upward and
hits a static structure, or to let the structure go down in a calm flow.
The second method was chosen, because it will offer more possibilities
for structure motions in future studies.
The solver used for this study is interDyMFoam from
OpenFOAM(Errore. L'origine riferimento non stata trovata.). The
interDyMFoam is a solver for two incompressible, isothermal
immiscible fluids using a VOF (volume of fluid) phase-fraction based
interface capturing approach, in a moving mesh domain. This solver
solve the well know Navier-Stokes equations, recombined for a moving
mesh. Following are presented the equation in the punctual versions
and write for a moving control volume:

U 0

U
U U p U f

t
U dV U G dV

U U dV

p U f dV U U dV
G

(1)

(2)

Where UG represent the Velocity of the mesh. Basically a new virtual


flux is added to the equation to take in account the moving of the mesh.
That solver use a PISO loop at each time step to ensure the pressurevelocity coupling. In particular, but an addition correction to the
standard PISO algorithm has been sued to better predict all the nolinear terms of N-S equations. The time marching is performed with a
Eulerian implicit approach, that increases the stability of the solution.
The linearized system has been solved with a GAMG (Generalized
geometric-algebraic multi-grid) for the Poisson equation and with a
Smooth solver (GaussSeidel as smoother) for the N-S equation. This
solver use a VoF method to track the interface between the two fluids,
so it solves the classical VoF equation, where alpha represents the
fraction of fluid present in each cell, and all the physical quantities
representing of the fluids characteristics are weighted by this fraction.

U 0
t
1 1 2

1 1 2

Fig. 2: Wedge configuration for scale 1:6


The set-up of the numerical simulations is the following:

length of the wedge=1000 mm

height of the wedge=100 mm

width of the wedge=278 mm (however 2D simulations were


carried out)

dead rise angle =10 (to ensure validity of Wagners solution)

drop height=0.6 m
On the contrary to experiments presented in ([1]) where one side of the
wedge is smooth and the other side equipped with raised edges, two
configurations of the wedge were simulated. The first one without invar
edge (smooth case) and the second one with invar edges scaled at 1:6.
For both configurations, a symmetry plan was considered as shown in
Fig. 3. In the case of the wedge with invar edges (on both side for our
numerical calculations on the contrary of ([1]), the first invar edge is
located at 230mm away from apex of the wedge and the distance
between the two invar edges is 83mm.
The locations of the pressure sensors are chosen accordingly to ([1])
and are reminded below. The dimensions are given with respect to the
apex of the wedge.
NSensor
1
2
3
4
5
6
7
8
9

before 1st edge (mm) between edges (mm)


151.611
234.611
160.833
243.833
170.056
253.056
179.278
262.278
188.500
271.500
197.722
280.722
206.944
289.944
216.167
299.167
225.389
308.389

NSensor
10
11
12
13
14
15
16
17
18

Table 1:Pressure sensors location

(3)

Three prescribed constant velocities of the wedge equal to 2.80, 3.43 &
3.96m.s-1 were simulated corresponding to the theoretical free fall
velocities obtained for the 3 drop heights of h=0.4, 0.6 & 0.8m. Thus
variation of the wedge velocity during impact is not taken into account.

To solve the VoF equation, that is a standard fully convective equation,


the interDyFoam use the MULES method (multidimensional universal
limited explicit solver), that is a new generic solver for simply
convective equations with a bounded range of free variable values.

For the two other scales (1:20 & 1:35), the 3 drop heights were scaled
from the scale 1:6.
Thus in our numerical simulations, the flow is going from the pressure
sensor n1 to the pressure sensor n18.

MARSTRUCT 2010 Lead Author: Stefano Brizzolara

page 2/5

3 WEDGE DROP TESTS RESULTS


3 - 1 Results for the smooth wedge (without invar edge)
For sake of simplicity, only results at scale 1:6 are here presented. Our
numerical procedure displaces the mesh at V=2.80, 3.43 & 3.96 m.s-1.
The mesh used for this simulation is figured out below. The total
number of cells used to define this mesh is equal to 290,000 cells.

Fig. 3: Mesh used for the smooth wedge simulation


(290,000 cells), simulation scale 1:6

1:6, h=0.4m
Sensor 10
Sensor 11
Sensor 12
Sensor 13
Sensor 14
Sensor 15
Sensor 16
Sensor 17
Sensor 18
1:6, h=0.6m
Sensor 10
Sensor 11
Sensor 12
Sensor 13
Sensor 14
Sensor 15
Sensor 16
Sensor 17
Sensor 18
1:6, h=0.8m
Sensor 10
Sensor 11
Sensor 12
Sensor 13
Sensor 14
Sensor 15
Sensor 16
Sensor 17
Sensor 18

OpenFOAM
310608
311117
314067
315277
319110
319133
317646
318834
319689
OpenFOAM
466241
465773
470233
472968
480360
481153
478298
479665
480882
OpenFOAM
623481
623824
628800
632826
642751
643088
638552
640465
642697

Wagner
311000
311000
311000
311000
311000
311000
311000
311000
311000
Wagner
467000
467000
467000
467000
467000
467000
467000
467000
467000
Wagner
622000
622000
622000
622000
622000
622000
622000
622000
622000

(OF/Wagner)
0.999
1.000
1.010
1.014
1.026
1.026
1.021
1.025
1.028
(OF/Wagner)
0.998
0.997
1.007
1.013
1.029
1.030
1.024
1.027
1.030
(OF/Wagner)
1.002
1.003
1.011
1.017
1.033
1.034
1.027
1.030
1.033

Table 2: Comparison of the maximum peak impact pressures obtained


OpenFOAM / Wagner Comparison
by OpenFOAM with
the Wagners solution
Smooth Case - Scale=1:6 - h=0.8m Sensor 18

For the case of the wedge without invar edge at scale 1:6 and drop
drop test for wedge without invar edges
height h=0.8m, the pressure
time
histories calculated at pressure sensor
pressure time histories for the 18 pressure sensors locations
location (10-18) are presented in figure
scale 4.
1:6, 0.8m

Wagner s18
OpenFOAM s18
600000

700000
500000

p10
p11

600000

p12

400000

p13

500000

p14
300000

pressure (Pa)

p15
400000

p16
p17

300000

200000

p18

200000

100000

100000

0
0.008

0.01

0.012

0.014

0.016

0.018

0
-100000
0

0.005

0.01

0.015

0.02

0.025

time (s)

Fig. 4: Pressure time histories for bare wedge, pressure sensors 10-18
for scale 1:6 and h=0.8m
In order to validate this CFD numerical solution obtained with
OpenFOAM, this one is compared with the Wagners analytical
solution ([5], Errore. L'origine riferimento non stata trovata., &
[3]) for the smooth wedge without inclination. Comparisons of the
maximum peak impact pressures for all the pressure sensors from #10
to #18 are figured out below for scale 1:6 and the 3 dropping heights.
There is an excellent agreement between the Wagners analytical
solution and the CFD solution for the maximum peak impact pressures.
The difference for the maximum peak impact pressure between the
CFD solution (OpenFOAM) and the Wagners analytical solution does
not exceed 3% for the pressure sensors of interest #10-#18.
OpenFOAM and Wagner pressure time histories for the pressure sensor
#18 which presents the maximum error (of 3%) in comparison with
Wagners solution is figured hereafter on Fig. 5.

MARSTRUCT 2010 Lead Author: Stefano Brizzolara

Fig. 5: OpenFOAM & Wagner pressure time histories for pressure


sensor 18 for scale 1:6 and h=0.8m
The difference between the CFD solution (blue) and the Wagners
analytical solution (red) arises when the separation point reaches the
end of the wedge as figured below.

Fig. 6: Classical Wagners analytical solution is no more valid when


the separation point reaches the end of the wedge.

page 3/5

Drop Test with Invar Edges

3 - 2 Results for the wedge with invar edges

Pressure time histories for the 18 pressure sensors locations

figured out below.

scale 1:6, 0.8m

1.8e+006
1.6e+006
1.4e+006
1.2e+006

pressure (Pa)

For the sake of simplicity, only results at scale 1:6 are here presented.
Our numerical procedure displaces the mesh at V=2.80, 3.43 &
3.96m.s-1. The mesh used for this simulation is figured out below. The
total number of cells used to define this mesh is equal to 290,000 cells..

1e+006
800000

p1

p10

p2

p11

p3

p12

p4

p13

p5

p14

p6

p15

p7

p16

p8

p17

p9

p18

600000
400000
200000
0

Drop Test with Invar Edges

-200000

Pressure
time histories
0.005
0.01 for the 18 pressure
0.015 sensors locations
0.02

0.025

scale 1:6, 0.8m


time (s)

1.6e+006
1.4e+006

pressure (Pa)

1.2e+006
1e+006
800000

p1

p10

p2

p11

p3

p12

p4

p13

p5

p14

p6

p15

p7

p16

p8

p17

p9

p18

600000
400000
200000
0
-200000
0.004

0.005

0.006

0.007

0.008

0.009

0.01

0.011

time (s)

Fig. 8: Pressure time histories for all pressure sensors. Drop test
simulation of the wedge with invar edges (zoom for the right figure)

Fig. 7: Mesh used for the drop test simulation of the wedge with invar
edges (290,000 cells)
For the case of the wedge without invar edge at scale 1:6 and drop
height h=0.8m, the pressure time histories calculated at pressure sensor
location (10-18) are represented in figures 8.
Fig. 9: Comparison between raw and filtered (low-pass filter 2.5 kHz)
Drop Test with Invar Edges
time pressure histories
for pressure
sensor #3 for the smooth wedge
Pessure time histories for the 18 pressure sensors locations
Scale 1:6, 0.8m

1.2e+006

1e+006

pressure (Pa)

One can notice that the time pressure histories signals are noisy. This is
the reason why it was decided to apply a low-pass filter (2.5 kHz, 8th
order Butterworth low pass filter) to these signals. However, to ensure
that this low-pass filter does not affect the impact peak pressure of
interest, this low-pass filter is applied on the time pressure histories
obtained for the smooth wedge which was validated in the previous
section (Errore. L'origine riferimento non stata trovata.). For
instance, the difference between the raw and filtered time pressure
history for the pressure sensor #3 for scale 1:6 and h=0.6m is figured
out below.

800000

600000

p1

p10

p2

p11

p3

p12

p4

p13

p5

p14

p6

p15

p7

p16

p8

p17

p9

p18

400000

The low-pass filter has a limited influence (experiments postprocessing procedure is not available) on the impact pressure time
history for the case of the smooth wedge (without invar edge).
Therefore, this low-pass filter was also applied to time pressure
histories for all the pressure sensors for the drop test simulation of the
wedge with invar edges. These filtered time pressure histories are

MARSTRUCT 2010 Lead Author: Stefano Brizzolara

200000

0.005

0.006

0.007

0.008

0.009

0.01

time (s)

Figure 10: Filtered pressure time histories for all pressure sensors. Drop

page 4/5

test simulation of the wedge with invar edges for scale 1:6 and h=0.8m.

3 CONCLUSIONS
The aim of this study was to check/confirm the findings presented in
([1]) from one hand and to evaluate the capabilities of the CFD
software OpenFOAM to deal with fluid structure impact problems in
the context of sloshing inside membrane tanks from the other hand.
More precisely, the objective was to evaluate OpenFOAM capabilities
for drop tests simulations for a smooth wedge on one hand and a wedge
equipped with invar edges (like these which equip the Cargo
Containment System NO96 produced by GTT) on the other hand, both
falling into calm water without inclination.
Figure 11: Exp. drop test with invar edges ([1])
The Figure 11 depicts the time pressure histories obtained in ([1]) for
one case which is not specified. However, it is possible to compare
both results (Figure 10 & Figure 11) qualitatively. Doing so, one can
notice that:

the first pressure sensors (before the first edge), pressures are
similar to these ones obtained for the smooth wedge. This can
be explained by the fact that the first edge does not have any
influence on upstream pressure sensors
OpenFOAM well predicts the over peak impact pressure (in
comparison with smooth wedge) before the first edge (see
sensor #8 & #9)
OpenFOAM well predicts the under peak impact pressure (in
comparison with smooth wedge) just after the first edge (see
sensor #10 & #11)
OpenFOAM well predicts the over peak impact pressure (in
comparison with smooth wedge) before the second edge (see
sensor #17 & #18)

3 - 2 Results for the ratio (Invar Case/Wagner)


All these qualitative observations can be quantified by comparing the
ratio between the filtered maximum pressures for the wedge with invar
edges and the Wagners solution as it is done in [1].
The agreement between OpenFOAM and Exxon (experiments) curves
is satisfactory. Furthermore, precise information on the experimental
procedure and its post-processing lack to perform clearer quantitative
conclusions. Hence, we can conclude that OpenFOAM well predicts
qualitatively and quantitatively the peak impact pressures for the drop
test simulation of a wedge equipped with invar edges.
3

In section 3-1, the drop test simulation of the smooth wedge was
performed and the obtained time pressure histories were then compared
to the Wagners analytical solution (valid for this studied
configuration). The agreement between the two solutions is excellent
and thus OpenFOAM drop test simulation for smooth wedge is
validated.
In section 3-2, the drop test simulation of the wedge equipped with
invar edges was performed and the obtained time pressure histories
were analysed. As these time pressure histories signals are noisy, it was
decided to filter these signals with a low-pass filter (2.5 kHz, 8th order
Butterworth). In order to quantify the influence of this low-pass
filtering on peak impact pressures of interest, this filter was applied on
the time pressure histories obtained for the smooth wedge. It was then
shown that this low-pass filtering has a limited influence on peak
impact pressures of interest.
Then, the time pressure histories were compared qualitatively and
quantitatively with the experimental ones obtained in ([1]). The
agreement is very good considering the complexity of such simulation.
As a conclusion, it can be said that OpenFOAM can be used to simulate
drop tests simulations not only for smooth wedge but also for wedge
equipped with invar edges which represent a much more complicated
case. The next step is to perform drop tests simulations for wedges
equipped with corrugations (like these which equip the Cargo
Containment System MarkIII produced by GTT).

REFERENCES
[1]

H. He, J.F. Kuo, A. Rinehart and T.W. Yung Influence of


Raised Invar Edges on Sloshing Impact Pressures, .1st Sloshing
Dynamics Symposium, ISOPE 2009, Vol3, www.isope.org.

[2]

M. Viviani, S. Brizzolara, L. Savio, Evaluation of slamming


loads using smoothed particle hydrodynamics and Reynoldsaveraged NavierStokes methods. Journal of engineering for the
maritime environment, 2009, 223:17-31, ISSN: 1475-0902, doi:
10.1243/14750902JEME131
T. Gazzola Contribution au problme dimpact non linaire: le
problme de Wagner coupl , Ecole Centrale Paris 2007 (written
in English).
B. Molin Hydordynamique des Structures Offshore, Editions
Technip.
H. Wagner ber Stoss und Gleitvorgnge an der Oberflche
von Flssigkeite , ZAMM, Vol 12, 193-215.
OpenFOAM, v.1.7.1, www.openfoam.com/

1:6; 0.4m
2.5

1:6; 0.6m
1:6; 0.8m

1:20; 0.12m
1:20; 0.18m

1.5

[3]

1:20; 0.24m
1:35; 0.069m

1:35; 0.086m
0.5

[4]

1:35; 0.103m
Exxon

[5]

0
9

10

11

12

13

14

15

16

17

18

19

Figure 12: OpenFOAM (red) and Exxon (green) maximum impact


pressures ratios between wedge with invar edges and the smooth one

MARSTRUCT 2010 Lead Author: Stefano Brizzolara

[6]

page 5/5

You might also like