Professional Documents
Culture Documents
A Consistent Method To Design and Evaluate The Performance of Anti Roll Tanks For Ships
A Consistent Method To Design and Evaluate The Performance of Anti Roll Tanks For Ships
A Consistent Method To Design and Evaluate The Performance of Anti Roll Tanks For Ships
Schiffstechnik
To cite this article: Geert Kapsenberg & Nicolas Carette (2023) A consistent method to design
and evaluate the performance of Anti-Roll Tanks for ships, Ship Technology Research, 70:2,
117-145, DOI: 10.1080/09377255.2022.2117496
CONTACT Nicolas Carette n.carette@marin.nl Maritime Research Institute Netherlands, Wageningen, the Netherlands
This article has been republished with minor changes. These changes do not impact the academic content of the article.
© 2022 Maritime Research Institute Netherlands. Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-
nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or
built upon in any way.
118 G. KAPSENBERG AND N. CARETTE
(van Daalen et al. 2000; Souto Iglesias et al. 2004; Bun- 1.3. Nomenclature
nik and Veldman 2010; Diebold et al. 2016), but at the
price of computationally intensive calculations. Nomenclature,
general
Since quite some time, the design of an ART is
g [m/s2] Acceleration due to gravity
verified by performing bench tests. During such an L [m] Length of the ART (in the ship
experiment, a model of the tank is fitted on an oscil- longitudinal direction)
lating table, and the reaction forces are being B [m] Width of the ART (in the ship transversal
direction)
measured as a function of the motion frequency D [m] Height of the ART (in the ship vertical
and amplitude. The tank is normally positioned at a direction)
p [Pa] Fluid pressure
vertical distance from the rotation axis of the table x, y, z [m] Linear wave induced displacements of
similar to the vertical distance of the tank to the the CoG of the ship
xART, yART, zART [m] Position of the reference point of the ART
CoG of the ship. Since the sixties and seventies of relative to the CoG of the ship
the last century, the combination of vessel and ART ε [rad] Phase angle of the force or moment
produced by the tank relative to the
is also tested in a wave basin to verify the effect in motion
selected sea states. ρ [kg/m3] Specific mass of the fluid in the tank
The objective of this article is to present practical ω [rad/s] Oscillation frequency
f, u , c [rad] Angular wave induced displacements of
methods to design an anti-roll tank and to predict the ship around respectively the X, Y
the performance of the system ship + ART in a sea and Z axis.
w [rad] Tank excitation angle EGA
state. These methods include the effect of lateral Nomenclature, U-tank
accelerations, the non-linear response of the tank h [m] Height of the duct of a U-tank
n [m] Local width of the U-tank, perpendicular
and saturation of the tank. To achieve this, the to the s-coordinate
model of Stigter and the model of Verhagen and v [m/s] Fluid velocity along the s-coordinate of
the U-tank
van Wijngaarden have been extended to also predict s [m] Coordinate along the U-tank centre of
the lateral force produced by the tank. The models the duct and reservoir. (s = 0 at plane
were improved to use a more complete driving par- of symmetry, positive in negative Y-
direction)
ameter for the tank motions. The roll angle and the w1..3 [m] Geometrical dimensions of the U-tank,
local lateral acceleration have been combined to a see Figure 7.
Nomenclature, FS-tank
single parameter, the Effective Gravity Angle h0 [m] Water level at rest in a FS-tank
(EGA). The use of this parameter allows the correct h1 [m] Water level in front of the bore in an
active FS-tank
prediction of the non-linear force and moment of h2 [m] Water level behind the bore in an active
the ART for any location on the ship, using for FS-tank
v [m/s] Fluid velocity in the FS-tank
instance the results of bench tests for only one axis λ [m] Actual water level in the FS-tank
of rotation. A complication is that the coupling of
the tank force and moment to the equations of
motion of the ship needs to be changed into an
iterative procedure; this aspect is also detailed in 1.4. Reference system
the presented study. The motions of the Anti-Roll Tank fluid and the forces
produced by it are defined in a right-handed reference
1.2. Design philosophy system, Figure 1, fixed to the ART. The X-axis is posi-
tive forward, the Y-axis is positive to port-side and the
The methodology to design an Anti-Roll tank advo- Z-axis is positive upwards. The origin O is located at
cated in this article is to use the analytical models the mid-length, symmetry plane and the bottom of
for the U-tank and the Free Surface tank that are the anti-roll tank. This system is used for as well the
described in Ch. 3 and Ch. 4. A choice for the Free Surface Tank as the U-tank.
required internal damping of the tank should be The reference condition is the ‘frozen’ tank con-
based on considerations related to the operating dition. This means that the force, the moment and
conditions and the limits of the tank (Ch. 8). The the phase angle are defined zero for the infinite fre-
design of the ART should be verified by coupling quency limit. The ‘frozen’ mass of the water should
the database containing the force and moment pro- be included in the mass and inertia of the vessel.
duced by the tank to a ship motion programme
using the method described in Ch 9. If this result
is satisfactory, the design of the tank can be detailed 2. Fundamentals of anti roll tanks
using CFD (Ch 6.) or using experimental methods
2.1. Physics
(Ch. 7). In particular, the baffles can be designed
to provide the required internal damping for the The purpose of fitting an anti-roll tank into a ship is
U-tank. The procedure is illustrated by an example to increase the roll damping, and thereby reducing
ART design for a yacht, Ch. 10. the roll motion, rather than modifying the stability
SHIP TECHNOLOGY RESEARCH 119
Figure 1. Anti-roll tank reference systems for Free Surface tank (left) and U-tank (right).
of the ship to shift its resonance frequency. The ART and this, combined with the width of the tank, deter-
is essentially a separate mass-damper-spring system mines the natural period of the tank. The height of
added to the ship. The damping that the system gen- the bore is quite significant for low amplitudes of
erates is proportional to the product of the moment it oscillation and does only slowly increase for increas-
generates and the roll velocity of the ship. Therefore, ing amplitudes. A first limit of the tank is reached
the efficiency of the system is at maximum when the when the rotation angle is such that one side of the
maximum moment is in-phase with the roll velocity, tank is dry; a second limit is reached when the pile-
which means that it has a 90 deg phase shift with the up of water created during the impact of the bore
roll motion. at the side interacts with the top of the tank. These
There are two basic designs for an ART, the U- phenomena ‘de-tune’ the tank such that the damping
tank and the Free-Surface tank. To produce the effect decreases quite quickly. Internal damping cre-
required moment, both ART designs rely on res- ated with bulkheads and baffles can reduce these
onant water motions. Both tanks will reduce the effects, but they will strongly reduce the height of
roll around the resonance frequency but increase it the bore. However, such structures can be necessary
elsewhere (see Figure 4). The physics of both tanks to reduce either the noise level of the tank or the
are fundamentally non-linear. The response, risk of damage due to the impact of the bore against
moment divided by excitation amplitude, is largest the side of the tank.
for small values of the excitation (Figure 2). The The principle of the U-tank is based on resonance
response of the tank can be reduced by applying of oscillating water columns in two reservoirs on both
internal damping, e.g. water or air flow restrictions. sides of the ship connected by a duct. The duct is
The amount of internal damping is a critical aspect connected to the lowest point of the reservoirs, giving
of the design of the tank as it determines the upper the tank a U-shape. Ventilation at the top of the
limit of the operating range. reservoirs is necessary, this is done with either vents
The Free Surface tank is usually a rectangular to the outside or a second duct connecting the top
reservoir that is filled with a relatively small amount of the reservoirs. In contrast to a free surface tank,
of water. Its working principle relies on a travelling a U-tank can be actively controlled by temporarily
wave bore created by the tank oscillations. The vel- closing the water or air duct to increase the period
ocity of the bore is a function of the water depth of its peak response. A control valve in the air duct
is usually chosen for practical reasons. The valve
can also be used to switch the tank completely off.
The limits of this type of tank are reached when
one reservoir is empty at some stage and air flows
into the connecting duct, or when the fluid hits the
top of the reservoir.
A big difference between the U-tank and the FS-
tank is the frequency range in which the tank is
effective. The passive U-tank has a very narrow
response curve; it is only effective in a small fre-
quency range. An example of typical responses of
the two tanks is given in Figure 3. The roll response
of a vessel is in general quite narrow, but the
Figure 2. Moment (at resonance) produced by a U-type ART response of a passive U-tank is often narrower
compared to the same for a FS-tank. The moment has been
normalized by the maximum moment at a 1 deg excitation still. A second difference is that a free surface tank
angle. The different types of tanks show an identical degree has a much larger free surface effect (reduction of
of non-linearity. metacentric height) than a U-tank. Finally, a
120 G. KAPSENBERG AND N. CARETTE
practical difference is that the U-tank is silent, and This approach properly adds the effect of the lat-
the FS-tank is noisy. eral acceleration due to the sway motion and the
The fact that ART’s work very well at low exci- induced effect due to the roll and yaw motions.
tation levels make them very useful to increase the Combining these motion components in this single
operability of a vessel with very stringent roll criteria, parameter is essential to correctly account for the
but due to their decreasing performance with non-linear response of the ART. The validity of
increasing motions, they are not suitable to reduce this EGA concept was demonstrated by using the
the extreme values of the roll motion. In practice results of van den Bosch and Vugts (1966a) who
this means that an ART should always be used measured the moment produced by FS-tanks when
together with bilge keels. Bilge keels are increasingly they were oscillated around a rotation point that
more effective at larger amplitudes, the limit is only varied in height relative to the bottom of the tank.
reached when they surface. A second property of Using the results measured for the rotation point
the ART is, that it responds to the ship motions at the bottom of the tank, zART/B = 0, the exper-
only, independent of the sailing speed and the imental results at different zART/B values could be
environmental conditions. It is thus one of the few reproduced accurately, Figure 5. An even more
devices that works as well at zero as at forward direct validation was carried out by Carette (2015);
speed Figure 4. he carried out pure roll bench tests (rotation point
at the bottom of the tank) and pure sway bench
2.2. The driving parameter tests with exactly the same EGA value and compared
the moment and phase angle produced by the tank.
Originally the roll motion was used as the driving par- The resulting time traces of lateral force and roll
ameter for the fluid motions in the tank, and thereby moment compared spot-on, Figure 6, as demon-
the tank reaction forces. It was soon realized that the strated by Carette (2015).
vertical position of the tank in the ship was also an The concept of the EGA to determine the ART reac-
important parameter; bench tests were carried out tion forces is used in this publication from now on.
using the actual position of the ART relative to the
CoG of the ship. Although mathematically the roll
Figure 6. Results of experiments with a FS-tank on a moving The parameters in this equation are defined in the
platform. Pure roll experiments (green line) are compared to nomenclature, Ch. 1.3. Since the duct has a constant
pure sway experiments (blue line). The comparison is based cross-section and assuming incompressibility of the
on an identical EGA values (EGA = 1 deg) for both motions. fluid, the velocity does not change in magnitude in
Results from Carette (2015). the duct, so:
∂v
=0 [s−1 ] (3)
∂s
3. Analytical model for the U-tank
Expressing the changes in the water level of the reser-
3.1. Model for the lateral force and the roll voirs as a change in the angle ψ, Figure 7, and consid-
moment of the U-tank ering continuity of mass we have:
A mathematical model for the moment around the ∂v w2 w3
longitudinal axis produced by a U-tank has been = c̈ [m/s2 ] (4)
∂t 2n
developed by Stigter (1966). His model assumes
that the dynamics of the tank are governed by the The double dot used for c̈ indicates the second time
dynamics in the duct connecting the two reservoirs. derivative of parameter ψ. Parameter w3 is defined as:
This is the case when the cross-section of the duct is w3 = w + w2 [m] (5)
considerably smaller than that of the reservoirs, and
thus the velocities and accelerations in the fluid The spatial pressure gradient is governed by the effects
considerably higher. In elaborating the equations of gravity, friction of the fluid along the wall of the
of motion for the fluid in the duct, Stigter only con-
sidered the lateral acceleration in the duct resulting
from the roll angle. He also assumed that the fric-
tion in the duct is proportional to the velocity,
which makes his model fully linear with respect to
the amplitude of the excitation. It is straightforward
to adapt this to a friction in the duct that is pro-
portional to the square of the velocity and to use
a linearized value in the solution of the equations
of motion.
The original model, as developed by Stigter (1966),
was adapted on the following aspects:
w2 w3 1 ∂p w2 w3 ċ
c̈ = − −k + g cos 11
2n r ∂s 2n2
− rf̈ sin 12 (6)
+ g w3 w
inertia of the mass of the fluid moving from the PS
=0 (7)
to the SB reservoir. This mass mMOV is determined by:
with:
mMOV = rLw2 w3 c [kg] (10)
sr
ds w 1 2y − h
E1 = = + 1 + ln 1 + [−]
0 n 2h a h + 2w1 /a
sr
ds w + 2h 1 2y − h
E2 = = + [m−1 ]
0 n
2 2h2 a2 (y + w1 /a)(h + 2w1 /a)
sr
′ w + w1 aR′
F2 = r sin 12 ds = (R + y − h) + (2y − h) [m2 ]
0 2 4
(8)
The solution of the equations of motion, equation 7, This mass moves over a total distance (double ampli-
results in: tude) of w3 ; this results in an inertia force of:
Figure 9. Roll moment produced by a U-tank with different internal arrangements resulting in different internal damping coeffi-
cient. Left: Roll moment for EGA = 1 deg. Right: Roll moment for EGA = 5 deg.
Figure 10. Validation of the model for the prediction of the moment and phase angle of a U-tank. Calculations using a damping
factor k = 1.4; comparison to the experimental results (Conf-1) by Gunsing et al. (2014).
124 G. KAPSENBERG AND N. CARETTE
Table 1. Main dimensions CRS U-tank. 4. Analytical model for the free surface tank
Parameter Abbr. Value units
Length tank (X-direction) L 6.53 m
4.1. Model for the roll moment of the FS-tank
Width tank (Y-direction) B 14.72 m
Height tank (Z-direction) H 3.68 m An analytical model for the moment of an oscillating
Length duct w 7.82 m FS tank was developed by Verhagen and Wijngaarden
Height duct h 0.989 m
Width reservoir bottom w1 3.45 m
(1965). They acknowledged that the physics of the
Width reservoir at water level w2 3.45 m fluid in the tank are fundamentally different for fre-
Water level y 1.84 m quencies around the resonance compared to frequen-
cies outside this range. At resonance, a shallow water
bore travels up and down the tank; this physical
significant amount of damping. These results were phenomenon does not occur at frequencies well
obtained by the CRS.1 Figure 12 away from resonance.
Using equation (15) in a non-linear time domain The main parameters used to describe the physics
solver for equation (7) gave also very good results in a FS-tank are illustrated in Figure 14. The resonance
using the same damping coefficient λ as in the linear frequency of the tank is determined by the velocity of
solution as presented above Figure 13. the shallow water bore (16). Using the first order
Figure 11. Validation of the model for the prediction of the moment and phase angle of a U-tank. Calculations using a damping
factor k = 8; comparison to the experimental results (Conf-3) by Gunsing et al. (2014).
Figure 12. Result of the model for the prediction of the lateral force (top) and moment (bottom), and corresponding phase angle,
of a U-tank. Calculations were carried out with a damping coefficient λ = 0.9. The experimental results were obtained by the CRS.
SHIP TECHNOLOGY RESEARCH 125
Figure 13. Result of the model in time domain for the prediction of the roll moment (top) and lateral force (bottom) of a U-tank.
The experimental results were obtained by the CRS.
approximation c = gh0 for the velocity together with The associated phase angle is defined as:
the width B of the tank, the resonance frequency of the
FS tank can easily be determined (17): 1Mx = −12p + a for v , v0
(20)
1Mx = −12p −a for v . v0
gh1 (h1 + h2 ) With α defined by:
c= ≈ gh0 [m/s] (16)
2h2 ⎧ ⎫
⎨ p2 B (v − v )2 ⎬
0
a = 2 arcsin
p ⎩ 24 g wa ⎭
v0 = gh0 [rad/s] (17)
B
p2 B (v − v0 )2
− arcsin (21)
The theory used here to solve the shallow water 96 g wa − 3p2 B(v − v0 )2
bore is based on similarity with a moving piston in a The bore height is determined by:
gas cylinder that creates a pressure wave. This theory
uses a perturbation parameter 1:
h2 − h1 BV 2
= 4A1 1 − (22)
h0 6Ac0
B wa
1= [−] (18)
p h0 Equation (22) uses parameters A and V that are
defined in equation (23). The resonance frequency
Although no strict limits were given in the publi- v0 is defined in equation (17).
cation, a practical limit of ε < 0.5 is recommended.
2 p2 h20
If 1 = (2/p) 0.8, the heel angle of the tank is A= 1+ 2
such that, in a static condition, the fluid is in a triangu- 3p B (23)
lar shape just covering the full width of the tank. In v − v0
V=
dynamic conditions this occurs at slightly lower values 1
of ε. Outside the region of resonance, the fluid motions can
Verhagen and van Wijngaarden derived the be described by classical shallow water theory based on
equation for the moment amplitude close to resonance the continuity (24) and the momentum equations
as: (25):
∂l ∂l ∂v
LB3 4 4 2h0 wa h0 v2 +v +l =0 (24)
Mx,a = rg 1+ ∂t ∂y ∂y
12 p 3B g
2
p B(v − v0 )
2 (19) ∂v ∂v ∂l
× 1− + v + g + g w sin vt = 0 (25)
32g wa ∂t ∂y ∂y
[Nm] Linearizing the equations result in the following
126 G. KAPSENBERG AND N. CARETTE
4.2. Lateral force prediction for the FS-tank 4.3. Effect of internal damping in the FS-tank
The shallow water theory used to predict the moment Internal damping can be generated in different ways in
for frequencies outside resonance can also be used to a free surface tank. Experiments were carried out in
predict the lateral force. The wave elevation in the which partial bulkheads were placed in the tank per-
tank resulting from shallow water theory is pendicular to the velocity of the bore. These bulkheads
or ‘gates’ were fitted on both sides on a distance ± B/4
l = h0 from the centre line. Two sets of gates were used: the
wa Bv0 1 + (h0 v2 /g) pvy small gates reduced the effective cross-section by 25%
− sin (vt) sin (27)
pv cos (pv/2v0 ) Bv0 and the large ones by 50%. The results of the exper-
[m] iments show that (Figure 17):
If we consider the wave elevation at the outer ends of . The velocity of the bore, and hence the resonance
the tank, y = + B2 , and if we integrate the pressure, the frequency of the tank, is not affected to an
Figure 15. Left: Result of the shallow water equations for a wide frequency range showing multiple resonance peaks. Right: Com-
posite prediction of the moment produced by a FS tank; result of shallow water equations (red) and the model for resonance
frequencies (blue line), the spline function (black) connects the two parts.
SHIP TECHNOLOGY RESEARCH 127
Figure 16. FS-tank with width ‘B’ and length of bore ‘s’. Figure 18. FS-tank at the time instant of maximum moment.
important degree. This is concluded from the fre- When the height of the bore is now reduced by a
quency for which the phase angle is −90 deg of factor h representing the effect of the gates, the
the three configurations, Figure 17. new height can be estimated by multiplying the
. The amplitude of the roll moment produced by the result of (31) by factor (1-η):
tank reduces significantly as a function of the size of
the gates. This effect is modelled as a reduction of h∗2 − h∗1 BV 2
= (1 − h)4A1 1 − (33)
the height of the bore. h0 6Ac0
. Although not shown, a gate in the centreline of the
tank has the largest influence on the response, gates Since the magnitude of the lateral force was
have a decreasing effect as they are moved further coupled to the roll moment, (30), it is straightfor-
away from the centreline. These last can be helpful ward to adapt this estimate also for the effect of
in reducing the magnitude of the impact against the restrictions in the tank.
side walls Figure 18.
Figure 17. Roll moment and phase angle for a Free Surface tank. The figure compares experimental results for a smooth tank (No
gates) and two sizes restrictions (Small gates and Large gates reducing the cross-section by respectively 25 and 50%).
128 G. KAPSENBERG AND N. CARETTE
The limits of this tank were exceeded on several formulated as a sort of inside-out seakeeping problem;
occasions during the experiments, either when the the walls of the container are described by panels and
combination of a low water level and the excitation the fluid is inside. There is no incoming wave, and
angle EGA result in a partly dry tank (ε > 0.8), or at thus also no diffracted wave, only the wave created by
a high water level when the bore impacted partly moving the walls. Since the waves cannot leave the
against the top of the tank. These limits of the tank domain, no energy can be dissipated. The internal damp-
are further discussed in Ch. 8. ing is, therefore, zero, so the response at resonance is
Results of calculations and experiments for a low infinite. In order to model internal damping, the original
water level are shown in Figure 19. The larger excitation work proposed to use a damping coefficient on the body
amplitudes EGA of 5 and 10 deg result in values for the boundary condition:
perturbation parameter that are larger than the
∂w
specified range of applicability. The roll moment and = i1B kw + vn (34)
∂n
its phase angle are quite well predicted for the two
lower EGA values of 0.5 and 1.0 deg. The lateral force In which:
is however not very accurate for all amplitudes. w [m2/s] Potential of the flow
Figures 20 and 21 show results of calculations and n [m] Normal
ε [-] Damping factor
experiments for a water level of 1.91 m. These results κ [m-1] Wave number
are all in the range of validity of the ε parameter, the vn [m/s] Normal velocity of the surface
predicted and experimental results for the lateral
force and the roll moment, and their phase angles, This approach appeared not to give satisfactory results
agree now much better. These two figures also demon- in comparison to experimental data; therefore, an
strate the effect of adding rather large restrictions in alternative dissipation term 1FS was added to the
the tank. The resonance frequency does not change, boundary condition of the free surface:
but the peak moment amplitude is reduced, especially
∂w
for the lower excitation amplitudes. = (k + i1FS )w (35)
∂z
Parameter k is the wave number and 1FS is the dis-
5. Boundary element models for FS tanks sipation term that needs to be chosen. This method is
A practical Boundary Element Method (BEM) has been however not able to model the shallow water bore at
developed by Malenica et al. (2003). The problem is resonance in the tank; this phenomenon is not part
Figure 19. Lateral force (top) and roll moment (bottom) produced by the FS-tank with water level h0 = 0.382 m. 25% obstructions
in the tank. The predictions (full lines) are compared to results of bench tests (symbols). The perturbation parameter ε = 0.37, 0.52,
0.90, 1.64 for the cases EGA = 0.5, 1, 3, 10 deg respectively. No damping applied in the calculations.
SHIP TECHNOLOGY RESEARCH 129
of the solution of the linear BVP. The effect of the bore (2014). An early study with a free surface tank was car-
is essential for the phase difference between motion ried out by van Daalen et al. (2000) using a Volume of
and produced moment, and hence for the effect of Fluid method (VoF). He showed that good predictions
the ART on the roll motion. The consequence is that could be obtained with a relatively coarse grid and with-
the method is more suited for tanks with a high out solving the boundary layer. Also Bunnik and Veld-
filling rate, such as might occur for an LNG tank, man (2010) showed good results for this VoF method
than for an ART with a low filling rate. coupled to a ship motion programme. They also used
This BEM method was also applied by Bunnik and the results of Molin et al. (2008) as a validation case
Veldman (2010), they compared results of calculations and showed much better results than those obtained
to experimental results by Molin et al. (2008). They with the BEM that was mentioned before, Figure 23.
considered a very large tank on a barge. The tank Solutions with a RANSE solver were presented by
had a high filling rate, so shallow water effects were Delaunay (2012), Thanyamanta and Molyneux (2012)
absent. Bunnik and Veldman showed that they could and Kerkvliet et al. (2014) for a U-tank. It appeared
reproduce experimental results well by tuning the to be important to closely monitor the conservation
value of the dissipation term 1FS , Figure 22. However, of mass as the moment produced by the tank is directly
they also concluded that they could not accurately pre- proportional to it. Mass losses often occur in initial
dict the 2nd resonance frequency of the barge – tank steps due to the mismatch of the grid with the free sur-
combination; this 2nd resonance frequency is the face. The U-tank considered was the one for which
sloshing mode of the tank. experimental results were published by Field and Mar-
tin (1975). New experiments were done by Gunsing
et al. (2014); these researchers repeated the tests by
6. Computational fluid dynamics models
Field and Martin and then added various baffles in
An anti-roll tank is an ideal object for modellisation in the duct. Figure 24 shows results for the tank without
CFD. The domain is limited, and the boundary con- baffles. There are some differences in the various CFD
ditions are well defined. As a consequence, only a rela- predictions; the results as presented by Kerkvliet et al.
tively small computational effort is required, and very (2014) compare quite well to the experimental results,
good results can be obtained. Nevertheless, it is essential Figure 24. It is of paramount importance to accurately
to carry out a systematic grid study so that the results can model the geometry of the edges of the duct and the size
be analysed to determine the numerical accuracy of the and type of vents in a U-tank as this drives the internal
calculation using the method of Eça and Hoekstra damping, hence the response.
Figure 20. Lateral force (top) and roll moment (bottom) produced by the FS-tank with water level h0 = 1.91 m. The predictions
(full lines) are compared to results of bench tests (symbols). No obstructions in the tank. The perturbation parameter ε = 0.16, 0.23,
0.40, 0.73 for the cases EGA = 0.5, 1, 3, 10 deg respectively. No damping applied in the calculations.
130 G. KAPSENBERG AND N. CARETTE
Figure 21. Lateral force (top) and roll moment (bottom) produced by the FS-tank with water level h0 = 1.91 m. The predictions
(full lines) are compared to results of bench tests (symbols). 50% obstructions in the tank. The perturbation parameter ε = 0.16,
0.23, 0.40, 0.73 for the cases EGA = 0.5, 1, 3, 10 deg respectively. Damping k = 0.40 applied in the calculations.
Results of CFD calculations compared to results of complicated to model accurately due to air entrap-
experiments for a Free-Surface tank have been pre- ment, water jets and high velocities.
sented by Kerkvliet et al. (2016). The tank considered
has a complicated shape with inward sloping ends and
baffles. The results of the CFD calculations agree clo- 7. Experimental methods for ART
sely to the experimental results as shown in Figure 25.
Free surface tanks are more complicated to model in Bench tests are used to obtain a broad and accurate
CFD than U-tanks due to the travelling bore. The determination of forces and moments produced by
bore has an almost vertical front that is difficult to cap- an ART. They are typically used to investigate the
ture by a compressive free surface scheme. The physics effect of geometrical variations, flow obstructions or
of the bore impacts against the sides is also extreme motions.
Figure 22. Roll motion of a barge with a large FS-tank. The Figure 23. Roll motion of a barge with a large FS-tank. The
plot shows results of calculations using a BEM for the motions plot shows results of calculations using the VoF method Com-
in the tank with different values of the damping coefficient, Flow (different grids) coupled to the ship motion programme
Bunnik and Veldman (2010). These results are compared to aNySIM, Bunnik and Veldman (2010). These results are com-
results of experiments by Molin et al. (2008). pared to results of experiments by Molin et al. (2008).
SHIP TECHNOLOGY RESEARCH 131
Figure 24. Experimental results for the U-type ART used by Field and Martin (1975) and Gunsing et al. (2014). Results of CFD
calculations of various authors are added.
Since the response of the tank is based on the EGA accurately measure the mass, position of the centre
parameter, roll oscillations around one axis of rotation of gravity and the inertia matrix of the model for
suffice; an example of an oscillating table is given in later correction. The correction for the dry mass and
Figure 26. To be able to transfer the response of the inertia is then easily performed using accurate motion
ART to any location in the ship, it is important to measurements during all other tests. This method is
measure not only the moment response, but also the preferred to a one-on-one subtraction between results
lateral force. It is recommended to measure the for an empty and a filled tank. The latter is much more
response of the tank directly underneath it with a prone to errors and inaccuracies, in addition to the
six-component force frame rather than measure the cost of doubling the number of tests. The calibration
torque in the axis of rotation. This approach also tests typically consist of a series of static inclination
ensures that the measured data only needs to be cor- tests to derive the mass and the position of the CoG,
rected for the mass and inertia of the tank model, and super-resonance oscillations to derive the inertia.
this correction is linear and changes only the in- Those tests can also be used to derive the accuracy of
phase part of the response. the six-component force frame using calibrated
The main advantage of experiments over a calcu- weights. After this calibration procedure, but before
lation method like CFD is that a broad range of con- dynamic tests, it is useful to also perform inclination
ditions can quickly be covered at a reasonable cost. tests with water. This allows verification of the water
Most of the cost of experiments is in manufacturing mass and free surface effect, which should be perfectly
the model and in the test setup. The test procedure known beforehand. Such tests give an indication of the
can be automated to generate a large range of ampli- accuracy of the model and filling height.
tudes and frequencies of motions, as well as to analyse A test campaign should mostly contain regular
the results. It is important to include sufficient waiting oscillation tests if the purpose is to derive the global
time between different tests. It can take more than ten response of the tank. The motions should cover a
oscillation periods for a tank with little internal broad range of amplitudes and frequencies because
damping to have a calm water surface in the tank the response of the tank cannot be extrapolated.
after a test. Very small amplitude test are the most critical ones
Each test should be preceded by careful calibration due to high accuracy needed regarding motions, iner-
of the setup. Dry oscillation tests are necessary to tia correction and forces in combination with very
Figure 25. CFD results for a FS-type ART with baffles; Prediction of the moment and phase angle by CFD calculations (red lines and
symbols) compared to experimental results from a moving platform (blue lines and symbols). Results from Kerkvliet et al. (2016).
132 G. KAPSENBERG AND N. CARETTE
Figure 27. Frequency sweep roll motion at 0.5 (blue line) and 5.0 deg (orange dotted line) maximum amplitude and correspond-
ing response moment.
upper corner of the tank wall and bottom when If the water level is high relative to the height of
√
1 = 2/p = 0.8. Results of bench tests with oscil- the tank, the bore can easily hit the top of the tank.
lation amplitudes that exceed this limit are shown in Figure 33 shows results of experiments with a FS-
Figure 31. This figure clearly shows that not too tank with a 50% filling height (h0/D = 0.5). The
much can be expected from the theoretical model. figure shows four different excitation angles, observ-
If the results of these experiments are compared to ing these shows that for EGA = 1.0 deg the bore just
those of the theoretical model, it appears that the error hits the top of the tank resulting in a small jet. For
in the predicted moment is not so bad; the main pro- larger angles there is massive pile-up of water
blem is in the phase angle. The phase angle at the natu- against the top of the tank. Considering the theor-
ral frequency of 0.33 rad/s changes from −90 deg for etical height of the bore, Figure 34, it appears that
low values of ε, to −45 deg for the highest value of ε the maximum height of the bore when it impacts
during the experiments, and this is not captured by the side of the tank is a factor 1.4 higher than
the model. This means that the tank will no longer the theoretical maximum value when the bore is
perform as expected and will produce a lower amount half-way the tank. This means that, when we design
of damping than predicted Figure 32. this tank to operate at a maximum EGA = 5 deg,
the tank needs to be a factor 2.7 times higher
Figure 28. Effect of the size of the venting openings on the moment produced by an U-tank. The legend and Table 3 indicate the
size of the opening in the reservoirs.
134 G. KAPSENBERG AND N. CARETTE
Table 3. Area ratio (cross sectional area venting pipes / same out for EGA angles larger than 3.5 deg, therefore the
for the reservoir) of the venting pipes, amplitude of water operating range is very limited.
motions in the reservoirs (Zw-a) and air velocities (Vair)
through the venting pipes of a U-tank.
Large Medium Small
9. Analysis of motions of a ship with an anti-
Area ratio [-] 0.030 0.008 0.002
Zw-a (wa = 3deg) [m] 1.50 0.65 0.15 roll tank
Vair (wa = 3deg) [m/s] 36 54 36
9.1. Frequency domain solutions
The classical method of including the effect of an ART
than the fluid level to avoid the bore hitting the in the equations of motion of a ship is to derive a
top. The theoretical moment and phase angle are damping and a restoring coefficient B44 and C44,
compared to the experimental results in Figure 35. from ART calculations or bench tests, and to add
This result shows that the consequences for these these coefficients to the equation for the roll motion
parameters are not very large. The maximum of the ship. This means that data should be available
moment reduces a little, the peak shifts to a for the ART at the position of the rotation axis similar
lower frequency but the change in resonance fre- to the height of the ART above the CoG of the ship. If
quency is not very large. this is done, the effect of the local lateral acceleration
The design value for the height of the tank depends due to a roll arm is included, but not the effect the
therefore on the design condition. A minimum height local lateral acceleration due to other modes of
appears to be a value of three times the design water motion. Using just one set of B44 and C44 coefficients
level at rest. This value was already mentioned by means that the ART reaction forces are linearized
van den Bosch and Vugts (1966b). for the roll amplitude, this underestimates the ART
effect at low amplitudes and over-estimates the effect
8.3. Safety aspects of anti-roll tanks at high amplitudes. The effect on the ship motions
of the lateral force produced by the tank is ignored
Both types of Anti-Roll tanks have limits above altogether in this approach.
which they do not work as intended. Essentially The method presented here solves these problems
this means that they will have hardly any effect by:
on extreme values. A ship designed with an ART
should therefore also be equipped with bilge keels. . using the EGA as the driving parameter for the tank
These devices increase in efficiency (roll damping) . performing calculations for a range of amplitudes
with increasing roll amplitude, opposite to ART’s. (and frequencies) of the roll amplitude
ART’s and bilge keels are therefore the perfect com- . adding both the sway force and the roll moment
bination for a ship to provide roll damping at the generated by the ART to the right-hand side
full range of amplitudes. (RHS) of the equations of motion.
Figure 29. U-tank with rounded corners between duct and reservoir, large venting openings (3%), duct height 0.989 m. Oscillation
with EGA = 3 deg, v = 0.7 rad/s. The picture shows the moment of maximum fluid motion (water level indicated by a yellow
line); the fluid level is really close to the top of the reservoir on SB and is well below the top of the opening of the duct at PS.
required so that no roll motion is started for a sym- This consequence is however irrelevant: the final
metric ship in head seas. The Newton scheme takes answer is a zero roll RAO. The condition with a zero
care of the proper combination of the six motions or very small value of the roll RAO is checked in the
including their phase angles to calculate the EGA at (fa /za ) . 11 box in the diagram of Figure 37; it is a
the location of the ART and to have the proper condition to stop the iteration procedure.
response from the ART. Since normally the result of For cases in quartering wave directions, the pro-
the EoM is a motion amplitude per unit wave ampli- cedure changes the wave amplitude until the value of
tude: xa i /za , the wave amplitude is used to determine the roll RAO multiplied with this wave amplitude is
the dimensional values of the motions (in particular equal (within margin 12 ) to the required linearization
the roll motion) so that the EGA in [deg] can be deter- angle fla .
mined. This dimensional value is necessary to interp- A comparison of the resulting roll motion using
olate the proper values from the ART database. this procedure to the traditional method based on
For some conditions this procedure results in pro- just using a roll moment damping and restoring
blems. The calculated roll motion of a symmetric ship coefficient is shown in Figure 38. Using the EGA
in head seas will be zero, so the necessary wave ampli- parameter for the coupling results in lower roll
tude to match the input roll amplitude will be infinite. angles for low frequencies and higher roll angles
Figure 30. U-tank with rounded corners between duct and reservoir, large venting openings (5.9%), duct height 0.989
m. Comparisons of predicted and measured lateral force and roll moment.
136 G. KAPSENBERG AND N. CARETTE
Figure 31. Stills from a bench test with a Free Surface tank with a low filling rate and a relatively large motion amplitude. Filling
level h0/B = 0.02, 2 sets of small obstructions, oscillation frequency 0.40 rad/s (maximum moment). Top: amplitude wa = 3 deg,
Perturbation parameter ε = 0.90. Bottom: amplitude wa = 10 deg, ε = 1.64.
for higher frequencies. This is the effect of the lat- et al. (2015). As mentioned before, an ART is a good
eral acceleration in the EGA, its contribution candidate for CFD modelling due to its closed domain
increases for higher frequencies. with well-defined boundary conditions. Most non-lin-
ear seakeeping models use an explicit solver with lar-
ger time steps than the CFD ART model; this
facilitates the coupling. It is however a computation-
9.2. Time domain solutions
ally intensive solution as long simulations are necess-
Fully or weakly non-linear seakeeping codes are ary to gather statistical data.
designed to include non-linear external forces. These
codes are suitable to investigate the performance of
an ART. Using a time domain solver to model the 9.2.1. Impulse response functions
ship motions including the effect of the ART allows A much faster approach is to use the frequency
to better investigate the change of roll and lateral domain response of the tank to build so-called impulse
acceleration distributions, which is not possible with response functions. The damping operator of the tank
frequency domain tools. This can be of importance satisfies the zero-bounding and smoothness con-
to further investigate the effect of the tank on the oper- ditions for such an approach. The impulse response
ability of the ship, and to verify criteria. Most criteria function has thus often a nice decaying shape as
use RMS values, this implies some linearity of the shown in Figure 39.
response. An ART can strongly modify the response The relation between the in- and out-of-phase part
spectrum and change the distribution of peak values. of the tank response is less adapted for impulse
Time domain solutions allow a first investigation of response functions than with normal wave damping.
such changes. In the case of the wave potential, the added mass has
To the knowledge of these authors, there exists no the same offset at zero and infinite frequency. This
publication of analytical time-domain solution for an offset is thus corrected with a mean value. In case of
ART including non-linear effects. Therefore, the an ART, the in-phase part, the restoring term, does
most common approach is to use a coupling with a not have the same offset. The restoring term at zero
CFD model of the ART, as shown by Cercos-Pita frequency corresponds to the free surface effect, but
SHIP TECHNOLOGY RESEARCH 137
Figure 32. Moment and phase angle of a FS-tank for excessive values of the perturbation parameter. Filling level of the tank h0/B
= 0.02, 2 sets of small obstructions.
Figure 33. Photos of the FS-tank bench tests. Filling height h0/D = 0.5, resonance frequency. The images show the moment the
bore fully hits the outer end of the tank. From top to bottom: EGA = 0.5, 1, 3 and 10 deg.
138 G. KAPSENBERG AND N. CARETTE
Figure 35. Experimental (symbols) and theoretical (lines) moment and phase angle of a FS-tank with a high filling rate, h0/D = 0.5.
The bore hits the top of the tank for EGA = 3 and 10 deg at the resonance frequency.
SHIP TECHNOLOGY RESEARCH 139
Figure 36. Design diagram for a FS-tank. Required water depth (blue lines) and maximum excitation angle EGA (red lines) as a
function of the width of the tank for three values of the resonance period T.
directly under the tank, and the tank model and main approaches. Either the water in the tank is com-
attachments should be made as light and stiff as poss- pensated for to maintain the natural roll period of the
ible to avoid excessive inertia force correction. If pre- ship independent of the water mass in the tank, or the
sent, ballast weights should never be placed on the loading conditions with tank filled and empty must
balance. Measuring the forces when the tank is match the real-life conditions. The first approach is
empty must be done accurately to be able to correct mostly applicable during design or validation tests. It
the water forces measurements, this is not possible offers the advantage that resonant conditions remain
when ballast is placed on the balance during some the same for the empty and the filled tests. The
tests. other approach is relevant for final seakeeping tests.
The loading condition of the model must be cali- For seakeeping tests, irregular seas are preferred as
brated carefully, and consistently. There are two being closer to real-life conditions. Including an ART
during such tests should not affect the test pro-
gramme: the tank must not be evaluated on its own,
the focus lies on the performance of the vessel. Such
tests are meant to assess that the ship meets the appli-
cable criteria, be it with or without the tank.
If the purpose of the test is also to investigate the
performance of the tank, it is important to choose
sufficient (encounter) wave periods. Tests should not
only be done in waves at the resonant frequency, but
in sea states that are relevant for the sailing region of
the vessel. Tests should then be repeated accurately
with the tank filled and empty.
The accuracy of model experiments can suffer from
scale effects. The response of FS-tanks is driven by the
dynamics of the bore. The speed of the bore is related
to shallow water wave theory, which scales according
to Froude’s law. Flow obstruction effects are believed
to be dominated by pressure drag and flow separation
at the gates. These effects do not suffer from noticeable
scale effects at the Reynolds numbers of interest. The
air flow in a FS-tank is not a driving factor in the
Figure 37. Procedure to include the non-linear response of
the ART in the equations of motion of a ship. This procedure tank’s response.
is repeated for a series of values of the roll amplitude – and Scale effects in U-tanks can be more complicated.
for all frequencies, speeds and headings. The driving factor for the tank response is the water
140 G. KAPSENBERG AND N. CARETTE
Figure 38. Roll motion and phase angle of a yacht equipped with a U-tank. The plot compares results from a calculation using a
coupling with just a damping and restoring coefficient (dashed lines identified as BC44) and using a coupling through the EGA
parameter (full lines). Results are given for three values of the linearization angle.
Figure 39. Impulse response function based on ART damping for various amplitudes.
motion in the reservoirs which is driven by gravity and coefficient for the vents should scale without issues
tank rotation and accelerations. These parameters fol- as long as the Reynolds number is not changing too
low Froude scaling. The flow around obstructions in much. The air flow through the vents should stay
the ducts is, similar to the flow around a flat plate, well below the speed of sound.
dominated by flow separation. The flow at duct exits We believe that scale effects are negligible as
and entries have a fairly constant discharge coefficients long as the dimensions of the tank are not too
over a broad range of Reynolds numbers, covering small. We build model tanks with a width of 1
model and full-scale flow regimes. The air flow in metre at minimum. Our belief is supported by the
and out of the reservoirs is important for the response results of validation experiments in comparison to
of the tank. The geometry of the venting openings analytical models and CFD calculations. These latter
must correspond to full scale. The discharge have shown, Kerkvliet et al. 2014, 2016, that
Figure 40. Input and reconstructed restoring coefficient using impulse response functions.
SHIP TECHNOLOGY RESEARCH 141
Table 4. Main dimensions of a motor yacht. calculated using the 3D panel programme Precal-R.
Parameter Abbr. Value units This programme has been developed by the CRS.
Length between perpendiculars Lpp 70.00 m The damping of the bilge keels is quadratic in relation
Beam B 10.94 m
Draft T 2.88 m to the local flow velocity due to the roll motion. From
Displacement Δ 1167 ton this non-linear damping an equivalent linear damping
Metacentric height GM 1.28 m
Roll gyradius kxx/B 0.35 –
has been determined to approximate the damping for
Roll resonance period Tw 7.5 s a series of roll amplitudes. The result is shown in as an
Bilge keel length LBGK 40.00 m amplitude dependent roll RAO, Figure 42. The roll
Bilge keel height HBGK 0.40 m
damping for small amplitudes is very low, resulting
in a very high peak of the roll motion; the roll angle
although vortices are slightly different at model and / wave slope parameter has a value of 19; an indication
full scale, this did not show noticeable effects on of a very low ratio of the roll damping / critical roll
the first order response of the tank. damping (0.4%). The right plot in Figure 42 is more
meaningful for a designer, it shows the roll amplitude
at resonance as a function of the wave amplitude. Sup-
10. Design procedure: ship with U-tank and
pose that the criterion for the roll amplitude is 5 deg, it
free surface tank
results in a very low operability in this condition
An ART design study has been carried out for a motor (beam seas, zero speed).
yacht. The hull form of the yacht is taken from the This design is ideally suited for an ART. A U-type
FDS series, Kapsenberg et al. (2015). The main hull ART is designed with a volume of 2% of the displace-
form parameters are listed in Table 4, a small-scale ment, so containing 23 m3 water. The size of the tank
body plan is given in Figure 41. The yacht has bilge is normally a parameter derived from the amount of
keels, but initially no other roll damping devices. damping required to meet the roll criterion as found
The motions at zero speed in beam waves have been in the design procedure; in this example the size is
Figure 42. Roll motion of the motor yacht in beam seas, Vs = 0. Left: Roll RAO for different roll linearization amplitudes. Right:
Relation between roll angle at resonance and wave amplitude for the roll natural frequency.
142 G. KAPSENBERG AND N. CARETTE
Figure 43. Amplitude of the water motions in the U-type ART Figure 45. Bore height in the FS-type ART (defined in Table 6)
(defined in Table 5) as a function of the oscillation frequency as a function of the oscillation frequency and the excitation
and the internal damping. Linearization angle EGA = 10 deg. angle EGA.
SHIP TECHNOLOGY RESEARCH 143
Figure 47. Moment and phase angle produced by the designed U-tank (full lines) versus the same produced by the FS-tank
(dashed lines). The moment is given for three values of the excitation angle EGA.
Figure 48. Performance of the yacht in irregular seas of different peak periods. The plot shows the significant double amplitude
roll angle (SDA) of the yacht without ART as dashed lines, the performance of the yacht with a U-tank as full lines (left), the per-
formance of the yacht with a FS-tank as full lines (right).
144 G. KAPSENBERG AND N. CARETTE
The roll motion of the yacht equipped with the FS- necessary to achieve the same performance. This has
tank is better than the one with the U-tank. This is as side effect that the ‘side lobes’ in the roll response
demonstrated by the RAO’s, Figure 46, and by the sig- will be brought further apart. This effect may be ben-
nificant double amplitude (SDA) value of the roll eficial for operability, however at the cost of a heavier
motion in a sea state, Figure 48. tank.
Vent size for U-tank should not be underestimated.
These can produce a significant internal damping.
11. Conclusions
Very small vents may completely annihilate the tank
The classical models for respectively U-type and Free- response as air velocity may become excessively
Surface type Anti-Roll tanks by respectively Stigter high. This may also introduce noise issues.
(1966) and Verhagen and Wijngaarden (1965) have
been improved and extended. The improvement con-
cerns the calculation of the resonance frequency for Notes
the U-tank and the extension concerns the calculation 1. The Cooperative Research Ships (CRS) is an inter-
of the lateral force produced by the ART. national organization that develops knowledge for
An important step has been made in the coupling of ships. Website: www.crships.org.
the ship motions to the ART and the effect of the force 2. More information: www.crships.org.
and moment produced by the ART on the equations of
motion (EoM) of the ship. Rather than using just the
Acknowledgement
roll angle for the force/moment produced by the
ART, a combination of local lateral acceleration and The main part of this work has been performed in the Coop-
roll angle, the Effective Gravity Angle (EGA) is used erative Research Ships (CRS) organization in the Roll
Reduction (I and II) working groups. The CRS is a group
as the driving parameter. It has been shown that the of 26 companies carrying out hydrodynamic research.2 Per-
use of this parameter properly covers the effect of mission to publish these results is gratefully acknowledged.
the height of the rotation axis relative to the bottom
of the tank.
Information is given on the limit of the use of the Disclosure statement
ART. For the U-tank, the limit is determined by the No potential conflict of interest was reported by the
height of the reservoirs and the water level in the reser- author(s).
voir relative to the opening of the duct. For the FS-
tank one limit is given by the combination of width
of the tank, water level and excitation angle. If this References
limit is exceeded, part of the tank is dry, thus reducing Bosch J, Vugts JH. 1966a. Roll damping by free surface
the effectiveness of the tank. A second limit is given by tanks, Neth. Ship Research Centre TNO, Report 83 S,
the bore not only impacting the side of the tank, but Apr 1966.
the tank top as well. Exceeding this limit has an Bosch J, Vugts JH. 1966b. On roll damping by free surface
effect on the resonance frequency of the tank. tanks, Trans. RINA, 1966.
Bunnik T, Veldman AEP. 2010. Modelling the effect of
An example of a design exercise for a 70 m motor sloshing on ship motions, Proc. 29th Int Conf on
yacht has been given. A U-tank and a FS-tank have Ocean, Offshore and Arctic Eng (OMAE), 6-11 Jun
been designed, both tanks contain the same volume 2010, Shanghai, China.
of water. It appears that the FS-tank is more effective Carette N. 2015. A study of the response to sway motions of
for this case, the tank produces a moment that is free surface Anti-Roll Tanks, Proc. World Mar. Techn.
Conf (WMTC), 2015, Providence, Rhode Island, US.
roughly 50% larger than the moment produced by
Carette N. 2016. Fast time domain evaluation of anti-roll
the U-tank. An important reason is the amount of tank and ship coupling using non-linear retardation
fluid that moves from side to side in the tank, this is functions, Proc. 15th Int. Ship Stability Workshop
roughly a factor 2 more for the FS-Tank. An important (STAB), 13-15 June 2016, Stockholm, Sweden.
part of the fluid (46%) in the U-tank is in the duct; this Cercos-Pita JL, Bulian G, Pérez-Rojas L, Francescutto A.
volume does not contribute to the roll damping. This 2015. Coupled simulation of nonlinear ship motions
and Free Surface Tanks, 12th Int. Conf. On the Stability
consideration, however, does not include other design of Ships and Ocean Veh., 14–19 Jun 2015, Glasgow, UK.
aspects such as noise, or active control of the U-tank. Daalen E, Luth HR, Kleefsman KMT, Veldman AEP. 2000.
Internal damping of an ART is an important par- Anti roll tank simulations with a volume of fluid (VOF)
ameter in the design of a tank. It should be minimum based Navier-Stokes solver, Proc. 23rd Symp on Naval
to ensure that the tank has the largest response poss- Hydrodynamics, 17-22 Sep 2000, Val de Reuil, France.
Delaunay J. 2012. Numerical simulation of motion stabiliz-
ible. This allows the use of a smaller tank to effectively
ation by U-tube Anti-Roll Tanks using CFD, Master’s
reduce the roll motions. This is however at the cost of thesis, Un. de Bretagne Occ.
tank height: high water motions require more head- Diebold L, Derbanne Q, Malenica Š. 2016. Etude numerique
room. If this is not available, a longer tank will be des cuves anti-roulis (Numerical study of anti-roll tanks),
SHIP TECHNOLOGY RESEARCH 145
15èmes Journées de l’Hydro-dynamique, 22 - 24 Nov 2016, Kerkvliet M, Gunsing M, Carette N. 2014. Analysis of U-type
Brest. anti-roll tanks using URANS, Sensitivity and Validation,
Eça L, Hoekstra M. 2014. A procedure for the estimation of Proc. 33th Int. Conf. on Ocean, Offshore and Arctic
the numerical uncertainty of CFD calculations based on Eng. (OMAE), 8-13 Jun 2014, San Francisco, USA.
grid refinement studies. J. of Comp. Physics. 262:104– Malenica S, Zalar M, Chen XB. 2003. Dynamic coupling of sea-
130. keeping and sloshing, Proc. 13th Int. Offshore and Polar Eng.
Field SB, Martin JP. 1975. Comparative Effects of U-Tube Conf. (ISOPE), 25 - 30 May 2003, Honolulu, US.
and Free Surface Type Passive Roll Stabilisation Molin B, Remy F, Ledoux A, Ruiz N. 2008. Effect of roof
Systems, Trans, RINA, pp. 73-92, Apr 1975. impacts on coupling between wave response and sloshing
Frahm H. 1911. Results of trials of the anti-rolling tanks at in tanks of LNG carriers, Proc. 27th Int. Conf. on Ocean,
sea. Trans. Inst. Naval Arch. 53:1911. Offshore and Arctic Eng. (OMAE), 15–20 Jun 2008,
Froude W. 1861. On the rolling of ships, Inst of Nav. Arch., Estoril, Portugal.
March, 1861. Souto Iglesias A, Perez Rojas L, Zamora Rodriguez R. 2004.
Gunsing M, Carette N, Kapsenberg GK. 2014. Experimental Simulation of anti-roll tanks and sloshing type problems
data on the systematic variation of the internal damping with smoothed particle hydrodynamics. Ocean Eng.
inside an U-shape Anti Roll Tank, 33th Int Conf on 31:1169–1192.
Ocean, Offshore and Arctic Eng (OMAE), 8-13 Jun Stigter C. 1966. The performance of U-tanks as passive anti-
2014, San Francisco, USA. rolling device, Neth. Ship Research Centre, Report 81S,
Journee JMJ. 1997. Liquid cargo and its effect on ship Feb 1966.
motions, Proc. 6th Int. Conf.on Stability of Ships and Thanyamanta W, Molyneux D. 2012. Prediction of stabiliz-
Ocean Struct. (STAB-97), 22-27 Sep 1997, Varna, ing moments and effects of U-tube Anti-Roll
Bulgaria. Tank geometry using CFD, Proc. 31st Int. Conf. on
Kapsenberg GK, Aalbers AB, Koops A, Blok JJ. 2015. Fast Ocean, Offshore and Arctic Eng (OMAE), OMAE2012-
displacement ships, The MARIN Systematic Series, 83104.
MARIN. Verhagen JHG, Wijngaarden L. 1965. Non-linear oscil-
Kerkvliet M, Carette N, Straten O. 2016. Analysis of free sur- lations of fluid in a container. J. Fluid Mech. 22(4):
face anti-roll tank using URANS. Verification and vali- 737–751.
dation, Proc. PRADS2016, 4-8 Sep 2016, Copenhagen, Watts P. 1883. On a method of reducing the rolling of ships
Denmark. at sea. Trans. Royal Inst. of Naval Arch. 12:165–190.