Professional Documents
Culture Documents
CFD Analysis of Ship-To-Ship Hydrodynamic Interaction
CFD Analysis of Ship-To-Ship Hydrodynamic Interaction
CFD Analysis of Ship-To-Ship Hydrodynamic Interaction
https://doi.org/10.1007/s11804-018-0010-z
RESEARCH ARTICLE
Abstract
A numerical study of ship-to-ship interaction forces is performed using a commercial CFD code, and the results are compared
with experimental data and with the results of a panel method analysis. Two ship models have been used in the interaction forces
analysis: a tug and a tanker, advancing parallel to each other with different lateral distances and two different values of the fluid
depth. Computations are carried out with four different flow models: inviscid and viscous flow with the free surface modeled as a
rigid wall and inviscid and viscous flow with the deformable free surface. A fair agreement was obtained with available
experimental data and results obtained by panel method. The influence of viscosity in the computations is found to be compar-
atively weak, while the wavemaking effects may be important, at small magnitude of the horizontal clearance.
Keywords Hydrodynamic interaction . Computational fluid dynamics (CFD) . Flow model . Comparative computations .
Ship-to-ship interaction . Shallow water
1
Centre for Marine Technology and Ocean Engineering (CENTEC),
There exists a large group of recent contributions dedicated
Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco mainly to experiments with scaled models although some of
Pais, 1049-001 Lisbon, Portugal them are complemented with CFD modeling. One of the most
Journal of Marine Science and Application
possible presence of tank walls. It was found that the code is Agreement in the passing by maneuver for two identical
able to predict reliably interaction forces in encounter Wigley hulls at a low Froude number was perfect. The
maneuvers with certain underestimation of peak values. The RANS computations with the KVLCC2 forms were carried
latter inspired a suggestion to introduce appropriate empiric out for the Froude number values 0.037 and 0.055. Zhang and
corrections. Zou (2011) have studied the interaction of two ships (KCS and
Sutulo et al. (2012) presented a validation of the potential Osaka types) in encountering and overtaking in a shallow
interaction code against experimental results that were obtain- canal with the code FLUENT using the layering method of
ed in the towing tank for a tug operating near a large vessel. dynamic meshing. The sway force and yaw moment coeffi-
The comparisons of the results corresponding only to parallel cients were computed for Fn = 0.04, 0.08 and for various
motion and equal speeds showed good agreement with exper- values of the canal width, relative depth, and lateral distance
iments for the sway force and yaw moment in most cases. to the bank.
Large discrepancies were, however, noticed at very small dis- Sadat-Hosseini et al. (2011) performed CFD computations
tances between the ship sidewalls which were attributed to the of the interaction between two tanker models (KVLCC2 and
influence of viscosity and/or wavemaking effects neglected by Aframax) advancing with the same speed in shallow water.
the double-body potential code. The in-house RANS code CFDShip-Iowa adapted to
Recently, Kadri and Weihs (2015) have revisited the multi-object configurations was used, and results of the com-
slender-body theory applying it to the problem interaction putations were compared to the towing tank measurements.
of two bodies of revolution in parallel motion in unbound- The overall number of grid points exceeded 8 × 106, and the
ed fluid. The authors have compared their numerical re- Froude number varied from 0.046 to 0.06, i.e., was very small
sults with classic experimental data obtained by Newton but the free-surface effects, though not very significant, were
(1960). The authors focused on obtaining higher-order captured by the level-set method. The agreement between
solutions and demonstrated that they provide better agree- measured and computed values was rather large and in most
ment with experiments. The slender-body technique may cases varied from 30% to 200% but it must be also noted that
result in very fast algorithms but is not able to properly measured absolute values of the loads were very small for 3–
account for the specifics of shapes of interacting bodies 4 m models, due to very small velocities.
that may become important in very close proximity. Much better agreement was reached by Simonsen et al.
(2011) who applied the commercial code Star CCM+ for study-
1.3 Application of Field Computational Methods ing interaction between a tanker ship and a tug in deep water
using an unstructured grid in coarse (1.8 × 106 cells), medium,
Lately, a number of publications were dedicated to application and fine (8.9 × 106 cells) variants. The isotropic k − ε/k − ω
of various CFD codes implementing the Reynolds-averaged SST turbulence model with wall functions was used.
Navier–Stokes (RANS) model. Such codes cannot be applied Satisfactory and even good agreement with the tank data was
directly in maneuvering simulations but they are very impor- demonstrated for parallel motion and also for the tug’s relative
tant for acquiring better insight into the details of the flow and heading 20°, bow inwards. It must be noted that in some cases,
can serve as a substitute for physical experimentation. much better agreement was reached with the coarse grid than
Benedict et al. (2011) presented a comprehensive study with more refined ones, which probably is related to the fact
emphasizing that “the implementation into simulators requires that monotone convergence over the grid refinement was not
a holistic model valid for the entire range of parameter com- quite certain.
binations”. To handle the most general cases with the CFD, Zou and Larsson (2012) studied interaction between an
the so-called overset method using overlapping grids was de- Aframax tanker and a KVLCC2 type ship using the finite-
veloped on the basis of the finite volume and level-set tech- volume RANS solver XCHAP. Only the case of parallel
niques. However, most of the results refer to the parallel mo- centerplanes and equal velocities was considered. The
tion where they were compared with the experimental data wavemaking was neglected, and the double-body model was
showing rather good overall (over the range of instantaneous used. Not only the hulls but also the rudders and propellers
stagger in the overtaking) agreement though very large dis- were modeled with the overall number of cells reaching 5.5 ×
agreement could be present at some specific relative positions. 106. The computed results were validated against available
The crossing motion case was studied only numerically using independent computational and experimental data. While the
the “transformed space”, i.e., reversing the flow around the agreement with the CFD results obtained at the Iowa
larger vessel. University is always good, this is much less so when it goes
Yang et al. (2011) studied Wigley and KVLCC2 hulls by about comparison with the tank data: the discrepancies in the
means of the commercial code FLUENT, and the RANS re- sway force and yaw moment reached 100% and even more.
sults were compared with those obtained with the potential Nikushchenko et al. (2012) have presented data about interac-
flow method developed by Xiang and Faltinsen (2010). tion forces and moments in parallel steady motion computed
Journal of Marine Science and Application
for two Wigley hulls using three commercial codes: ANSYS and 2091 panels on the lock walls. Agreement with the
Fluent (data presented for the surge force only), NUMECA experimental data was demonstrated to vary from reasonable
FINE/Marine, and Star CCM+. The agreement between re- to good, especially at smaller time steps. Refinement of the grid
sults obtained with different codes varied from quite good to in RANS computations improved agreement for the surge force
relatively large (30%–50%) discrepancies depending on the but worsened it for the sway force. The potential flow model in
component and on relative position of the hulls. Some devel- most cases heavily underpredicted the magnitudes of the sway
opment of these studies for the KCS container ship in over- force and of the yaw moment.
taking maneuver at relatively high (up to 0.26) values of the Yuan and Incecik (2016) described results of application of
Froude number and using different turbulence models was the in-house potential flow method MHydro. Unlike many
presented by Nikushchenko and Zubova (2015). other potential interaction codes, this code accounts for
Fonfach et al. (2011) used commercial CFD codes to model wavemaking effects applying the linear steady flow
the flow between the same tug and tanker forms as studied in free-surface condition which is quite adequate for the
(Sutulo et al. 2012) using four different flow models with ship-bank interaction but much less so for the overtaking ma-
viscous or inviscid fluid and with or without wavemaking. neuver with substantial difference of velocities of the
In that study, viscous flow was modeled for the full-scale interacting ship. Comparisons with tank data, two CFD codes
Reynolds numbers. The obtained data showed that, likely, and the double-body potential code ROPES did not show any
account for the free-surface effects is more important than that definite patterns. For instance, while the code MHydro pre-
of viscosity. However, it must be emphasized that in general, dicts the sway interaction force better than other codes, it fails
the results of the cited reference are harder to assess and to predict the yaw moment.
discuss as computations were performed using two different A very comprehensive attempt to investigate capabilities of
commercial codes with different grids and different turbulence various interaction prediction methods was undertaken by Van
models so that effects of those factors were mixed with the Hoydonck et al. (2015) who compared tank results for a
effect of the flow model per se. In addition, viscous KVLCC2 model with those obtained with the CFD codes
computations were performed for the full scale that made Re FRESCO and ISIS-CFD and with the potential code
comparisons with the tank experimental data less ROPES mentioned in the previous subsection. The Froude
meaningful. In many cases, iterative convergence was not number was 0.055, and the Reynolds number was 1.5 × 106
achieved. As result, the conclusions, presented there, even if (model scale). In some ReFRESCO computations, not only
they happened to be correct, must be viewed as uncertain the hull but also the rudder and the propeller (on the body
let alone any quantitative assessments. force level) were modeled. The overall number of cells varied
In the paper by Sian et al. (2016) interaction of an LNG from 2.9 to 34.7Mcells. Even for the finest grid, no interaction
carrier with a ship of Series 60 hull was studied in a towing convergence was achieved with oscillating deviations
tank and with the commercial CFD code Fluent V15 for the reaching 6.6% in sway interaction force, and for the yaw mo-
case of parallel overtaking motion in a shallow canal was con- ment, these deviations were even much larger than the average
sidered. The SST k − ωturbulence model was used, and all re- value. Also, practically, no convergence over the grid density
ported computations were performed with 2Mcells. The exper- was reached, especially for the surge force. Most of computa-
imental and numerical results showed satisfactory agreement tions with the code ISIS-CFD were carried out for 16M cells.
with some under-prediction (from slight to moderate) of the The grid convergence studies carried out for four grids with
sway force and some over-prediction for the yaw moment the number of cells varying from 5 to 29.7 M showed better
while data for the surge force were only presented for few trends than in the case of ReFRESCO but uncertainty at 16M
cases. In the contribution (Toxopeous and Bhawsinka 2016), cells is not quite excluded. Comparisons between experimen-
interaction effects on a full-bodied ship entering a lock were tal responses and those obtained with various codes look rath-
predicted using the in-house CFD code ReFRESCO, validated er disappointing in the sense that no one of the codes was able
with the scaled-model experimental data and compared with to predict reliably all components, and the divergences are
predictions obtained with the in-house double-body potential often significant reaching 50%, 100%, and even 200%.
flow code ROPES. In the CFD modeling, a combined sliding Also, it must be noted that the observed influence of the pro-
and deforming grid approach was used but the free-surface peller was substantially stronger than could be predicted with
effects were neglected as the maximum Froude number 0.026 the potential flow method (Sutulo and Guedes Soares 2012).
(or 0.091 when based on the water depth) was considered small Shevchuk et al. (2016) have performed a thorough numer-
enough for application of the rigid-wall condition. The SST k ical study of the flow between the ship hull and the bottom of a
− ωturbulence model was applied and wall functions were used shallow canal with an in-house CFD code based on the
for modeling the flow close to the hull surface. Two grids with OpenFoam software. In particular, it was found that at very
1.6 and 4.6 Mcells were applied. The panel method small underkeel gaps the boundary layers on the hull and on
computations were performed using 1650 panels on the hull the bottom are fused which leads to an additional pressure
A.D. Wnęk et al.: CFD Analysis of Ship-to-Ship Hydrodynamic Interaction
drop and intensifies the squat. Besides the more or less usual Table 1 Main characteristics of the ship
RANS modeling, a hybrid (RANS+LES) closure was applied Characteristics Tug (ship 1) Tanker (ship 2)
with a much finer grid (13 instead of 4Mcells) and then
self-sustained motions of the ship in heave and pitch caused Full scale Model Full scale Model
by the unsteady flow separation were captured. However, that
Length (o.a.)/m 29.5 1.18 189.5 7.58
does not change substantially the time averaged values of the
Length (p.p.)/m 25.6 1.024 186.2 7.44
sinkage and trim.
Breadth/m 11.0 0.44 31.6 1.26
Rattanasiri et al. (2014) applied the commercial code
Draft/m 4.6 0.184 10.3 0.41
ANSYS CFX to a less common problem of hydrodynamic
interaction between several autonomous underwater vehicle Displacement/m3 649.0 0.0416 49,197.0 3.15
operating in tight formations. Numerical results were com-
pared to experimental ones, and special attention was paid to
larger tanker model (ship 2 or tanker) at the scale 1:25. The
effects of alteration of resistance and to possible gains that can
main characteristics of the vessels in two scales are presented
be achieved on optimized configurations.The presented re-
in Table 1 and their hull forms are outlined in Fig. 1.
view of the published results rather clearly demonstrates that
Considered were two values of the water depth given in
so far, in spite of considerable efforts, it has not been possible
Table 2, and corresponding to the conditions of the tank tests
to obtain definite and crisp estimates of possible discrepancies
is described in (Sutulo et al. 2012), where also all necessary
between predictions of hydrodynamic interaction loads ob-
details about the experimental setup are given. Two ship
tained with various methods, and the performed comparative
lengths are presented in Table 1, the length overall (o.a.),
studies are not sufficient and must be continued.
and the length between perpendiculars (p.p.).
As result, it was decided to revisit one of the scenarios
The axes Cxyz were linked to each ship with the x-axis
considered earlier and perform thorough comparative compu-
pointing forward, y-axis—to the starboard, and the z-axis—
tations with the recognized commercial code Star CCM+.
downwards. The origin was defined as the intersection of the
Although the scenario covers the cases studied in (Sutulo
centerplane, midship plane, and waterplane. The both
et al. 2012) and (Fonfach et al. 2011), the present study is
centerplanes were remaining parallel in the present study,
quite independent and original and is aiming at obtaining
and the position of the tug relative to the tanker was described
more credible data on relative importance of accounting for
by location of the origin of the tug’s frame in the body axes of
wavemaking and viscous effects in interaction load prediction.
the tanker, i.e., by its longitudinal stagger ξand the signed
Special attention was paid to appropriate choice of computa-
lateral distance ηbetween the centerplanes which is always
tional domains and grids, to selection of a single turbulence
negative for a tug located at the port side from the tanker.
model, and to convergence studies. At each interaction con-
The following nondimensional position parameters will be
figuration, CFD computations were performed for four possi- 2 / L2, 2 / B2, where the subscript 2
used further:
ble combinations formed by accounting or non-accounting for
means the ship number associated with the tanker. It is clear
viscosity and wavemaking. In addition, experimental and pan-
that ξ′ = 0 when the midship planes of the both vessels coin-
el method data are also shown for comparisons. To avoid any
cide, and ξ′ = 1 means that the midship of the tug coincides
ambiguities, it should be taken into account that through the
with the fore perpendicular of the tanker (Fig. 2).
present paper, the term “Free Surface case” means the case
Due to specifics of the interaction configuration character-
when the actual kinematic and dynamic boundary conditions
ized by a large difference in the dimensions of the interacting
on the free surface are met and the wavemaking effects are
ships, the interaction forces of surge X, sway Y, and the
modeled. This is opposed to the “Rigid Wall” case when the
non-penetration (and free-slip when the viscosity is present)
condition is applied on the free surface of the fluid and all
wavemaking is neglected. The main purpose of the study is
to try to demonstrate relative importance of purely inertial
forces (provided by the double-body potential theory) and of
influence of viscosity and wavemaking in problems of hydro-
dynamic interactions.
Table 2 Depth of the ship behind it. The depth was defined equal to H1 or H2 (depending
Absolute depth in Depth [m] relative to draft of: on the case) in model scale from the undisturbed free surface.
full scale/m In the case of the deformable free surface handled with the two
Tug T1 Tanker T2 phase flow model, the air layer extended 0.1L2 up from the
still water-free surface. The width of the domain was always
H1 = 12.25 2.66 1.19
8 m which is not arbitrary but is the actual width of the towing
H2 = 15.57 3.38 1.51
tank used in the tests described in (Sutulo et al. 2012). Also,
exactly as it was in the tank tests, the transverse location of the
interaction yaw moment N, were only computed for the small- smaller ship was in the central longitudinal plane of the do-
er ship (tug). Each interaction force was obtained as the dif- main or tank while the larger model was shifted in the lateral
ference between its total values obtained in presence and in direction to set the required distance |η| between the
absence of the interacting large ship. All the forces and mo- centerplanes of the ship hulls.
ments were non-dimensionalized as: X ′ = 2X/(ρAV2), Y ′ = 2Y/
(ρAV2), N ′ = 2N/(ρALV2), where ρ is the water density, A ¼
p ffiffiffiffiffiffiffiffiffiffiffiffi 3.2 Governing Equations, Turbulence Model
3
∇ 1 ∇ 2 is the reference area, ∇ is the ship displacement, and L
and Boundary Conditions
is the length between perpendiculars. The subscripts 1 and 2
here and further correspond to the tug and tanker respectively.
The standard Reynolds-averaged Navier–Stokes equations
Computations were carried out for two different speed
were applied in the whole computational domain and used
values as shown in Table 3, where standard definitions for
for the both flow phases:
the Froude number based on the ship length and water depth
pffiffiffiffiffiffi pffiffiffiffiffiffiffi !
H are used, i.e., Fn ¼ V= gL and FnH ¼ V= gH respec- ∂ u u
∂ui i j 1 ∂p ∂τ ij
tively with g being the magnitude of the gravity acceleration. þ ¼ Fi− þ ð1Þ
No results for the loads in heave, roll, and pitch are pre- ∂t ∂x j ρ ∂xi ∂x j
sented in this paper as, first, they are less important in inter-
action scenarios, and, second, the authors did not possess any where t is the current time; Fi are components of the volumet-
experimental data for them. ric force which is only here represented by the vertical gravity
force; ui are the components of the averaged fluid velocity; xi
are the Cartesian coordinates; ρ is the water or air density; p is
the averaged pressure; and τ ij are the turbulent stresses
3 Problem Formulation disappearing when the inviscid fluid model is considered
and the momentum equation becomes the Euler equation.
3.1 Computational Domain Obviously, this equation must be complemented with the
continuity equation
The finite computational domain has the shape of a parallele-
piped (Fig. 3) whose transverse section and location of the ∂u j
hulls corresponded to the actual tank layout and had longitu- ¼0 ð2Þ
∂x j
dinal extensions: 2.5L2 to the front of the larger ship and 4L2
and with the closing equations representing the turbulence
model.
Although the review presented in the Introduction indicates
that the SST k − ω turbulence model was preferred by most
researchers, no solid explanations of this choice for interaction
problems were presented. That is why, an auxiliary compara-
tive study performed for the smallest water depth H1 and lat-
eral distance ξ′ = 1.36 with account for the deformable free
surface with the following turbulence models:
4 2.06 0.412 0.13 0.048 0.188 0.167 5.3 × 107 4.2 × 105
7
6 3.09 0.618 0.195 0.072 0.282 0.25 7.9 × 10 6.3 × 105
The Turbulence suppression model allows the user to mim- & Air–water separation surface: the non-penetration condi-
ic the effect of transition from the laminar to turbulent flow if tion for the rigid-wall model (RW) or the velocity and
the location of transition is known priori while the γ − Reθ pressure continuity otherwise;
model is determining that location automatically and therefore & Hull surface: no-slip condition (zero velocity) for the vis-
is used more often. In all variants, the SST model requires a cous fluid model and non-penetration condition for the
finer mesh with an increased number of mesh cells around the inviscid fluid.
surface to capture the viscous sublayer effects which result in a
significantly larger computation time. It was found that while Of course, in computations with the inviscid fluid, the tur-
the force/moment coefficients obtained with various turbu- bulent stresses τ ij vanishes, and the RANS (1) reduce to the
lence models varied insignificantly, the speed of iterative con- Euler equations, and part of the boundary conditions are sim-
vergence turned out rather different (Fig. 4) with somewhat plified accordingly.
better behavior for the k − ε model which pre-determined the
final choice as the required computations are very time- 3.3 Computational Grid
consuming.
Hence, the main set of computations was performed using The final meshes adopted for the present study were generated
the k − ε turbulence model described by: after a number of comparative computations, and they were of
the same type as adopted in Wnęk and Guedes Soares (2015).
∂ðρkÞ ∂ ρu j k ∂ μ ∂k
þ ¼ μþ t þ 2μt S ij S ij −ρε ð3Þ Although in general, inviscid flow computations allow much
∂t ∂x j ∂x j σk ∂x j
and
∂ðρεÞ ∂ ρu j ε ∂ μt ∂ε
þ ¼ μþ
∂t ∂x j ∂x j σε ∂x j ð4Þ
ε ε2
þ2C 1ε μt S ij S ij −C 2ε ρ
k k
where k is a turbulent kinematic energy; ε is a turbulent dissi-
pation rate; μ is the fluid viscosity; μt is the dynamic turbulent
viscosity, Sij is a component of a fluid element deformation
rate, with i, j = 1, 2, 3, and σk, σε, C1ε, C2ε are the model
constants, as presented by Versteeg and Malalasekra (2007).
As the considered problem was steady, the reversed flow
was modeled with the following boundary conditions:
Fig. 4 Surge force X convergence plots for viscous flow with free
surface; η' = − 1.36, H1 = 0.49 m; the k - ε model (top); the SST model
without transition (upper middle), the SST model with turbulence
suppression (lower middle), the SST model with γ - Reθ transition
Fig. 3 Computational domain: the horizontal plane view (bottom)
Journal of Marine Science and Application
X′ Y′ N′
Fig. 7 Residuals for viscous flow (rigid wall): (1) continuity equation, (2) Fig. 9 Surge force X (top), sway force Y (middle), yaw moment N
x-momentum, (3) y-momentum, (4) z-momentum, (5) turbulent kinetic (bottom) convergence for viscous flow with free surface, η’ = − 1.36,
energy, and (6) turbulent dissipation rate H2 = 0.623 m
Journal of Marine Science and Application
Fig. 10 Plots of interaction forces for smaller depth H1: surge force (top),
sway force (middle), and yaw moment (bottom) Fig. 11 Plots of interaction forces for larger depth H2: surge force (top),
sway force (middle), yaw moment (bottom)
responses could be identical as in both cases, the fluid was
perfect and the wavemaking was absent. The observed differ- expected that the most realistic flow model accounting for
ence can be partly explained by a much finer discretization in both viscosity and the wavemaking, this does not happen in
the field method but also by accounting for the tank sidewalls all the cases: often the perfect fluid model provides better
while in the panel method computations, the fluid was assumed agreement although the differences caused by viscosity are
unbounded in the horizontal plane. On the other hand, finite not so large. But rather certain is that neglecting the
distances of the inlet and outlet boundaries do not exactly cor- wavemaking effects makes the agreement worse. In particular,
respond to the absence of such boundaries in the panel method. relatively large magnitudes of the surge force at close dis-
Accounting for the viscosity and/or wave making always tances are caused by the flow blockage mainly caused by
improved the agreement with experimental data, especially at wave making although also to some extent—by viscosity.
closer lateral distances but separate influence of these two Of course, neither the panel method, nor the IF+RW field
factors looks more complicated. Although it could be method are able to capture this effect.
A.D. Wnęk et al.: CFD Analysis of Ship-to-Ship Hydrodynamic Interaction
Fig. 12 Values of the dimensionless surge force: top— smaller depth H1, Fig. 14 Values of the dimensionless yaw moment: top— smaller depth
bottom— larger depth H2 H1, bottom— larger depth H2
4.1.2 Sway Force methods neglecting wavemaking are not able to capture the
drop of the suction force and even its transition to repulsion.
Agreement between the panel method and the CFD computa- Obviously, ever increasing suction at unlimited approach is
tion without viscosity and wavemaking is again good. Some theoretically correct for the pure potential flow without
difference at the smallest lateral distance is certainly caused by free-surface effects but it has always been clear that in real
a relatively small number of panels. For relative lateral dis- fluid, the flow will be sooner or later blocked, the suction will
tances greater than 1.4 agreement is good for all data although disappear, and the zone of the higher pressure in the fore part
account for the wave making and viscosity gives better agree- of the hull will result in repulsion. This should happen even in
ment with the experiment. At smaller distances, however, the the absence of the free surface and wavemaking effects but the
results presented here showed decisive role of the
wavemaking. Although sometimes the full RANS model
VF + FS gave better agreement than the model IF+FS, in other
cases, the observed agreement was even worse. However, it
must be taken into account that at smaller lateral clearances,
the uncertainty of the RANS computations is higher as long as
the grid convergence was not achieved. Apparently, viscous
blocking without free surface should also come into the effect
but it will likely happen at much smaller gaps than explored
here. There is also little doubt that a panel method with ac-
count for the actual free surface boundary condition would
also provide fair prediction of the sway force inversion effect.
and the IF+RW model. Also, the trend shown in the latter case
is very different from that provided by all remaining data. The
initial guess that some occasional confusion happened, was
not confirmed, and the same phenomenon was present in ab-
solutely independent computations by Fonfach et al. (2011).
So far, it has not been possible to obtain absolutely clear ex-
planation of such anomalous behavior but, apparently, this
interaction configuration was especially sensitive to
free-surface effects, which are expected to be the most pro-
nounced in the gap between the vessels, although such effects
exist over all domain. A better agreement observed for the
panel method should be rather viewed as occasional, caused
by some error cancelation effects. This hypothesis is some-
what confirmed by the fact that for the larger depth such
anomalies are much less pronounced.
Furthermore, despite a rough overall agreement between
various flow models, account for the viscosity is not always
Fig. 15 Dynamic pressure contours on the hull of the tug model for ξ' =
beneficial from the view of closeness to empiric data and 0.014 (top) and ξ' = 0.62 (bottom) at viscous free surface flow, η’ = 1.36
although in most cases it increases the absolute value of the (pressure values in Pascals)
yaw moment, sometimes the effect is reversed. Besides, in-
creased uncertainty caused by insufficient grid convergence at
small lateral gaps mentioned above, it can be supposed that midship of the tug model was observed at inviscid flow for
this is the effect of inherent uncertainties of the RANS tech- lateral separation η′ = 1.47. It can be noticed that, as should be
nique dependent on specific turbulence models whose effec- expected, the velocity near the hull is higher in the inviscid
tiveness and adequacy have not been exhaustively explained. flow. Distribution of shear stresses in the viscous flow is pre-
Summing up, it can be stated that the disagreement be- sented in Fig. 17.
tween the most physically adequate VF + FS computational Figure 18 presents the velocity distribution around the tug
model and the experimental data is substantially higher than at the midship section of the hull for the larger depth H2 and
presented by Simonsen et al. (2011) who noticed the largest for the depth Froude number FnH = 0.25. In this case, the
discrepancies for the sway force. However, those computa- longitudinal stagger is also larger ξ′ = 0.62, and the tug is lo-
tions were performed for an apparently simpler deep water cated closer to the bow of the tanker; the velocity near the hull
case and for the sinkage, trim, and heel of the tug taken from of the tanker starts to increase. In this case, larger magnitudes
the experiment while these data were not available to the au- of the velocity are more visible also for a higher lateral dis-
thors of the present contribution. tance η′ = 1.47.
Fig. 17 Shear stress distribution on the tug for smaller depth H1, ξ' =
0.014 at viscous free surface flow, η' = 1.36
Fig. 18 Velocity distribution around tug model for larger depth H2, ξ' =
0.62. Inviscid flow (left) and viscous flow (right), three lateral Fig. 20 Velocity distribution and velocity vectors for smaller depth H1
separations: η' = 1.36 (top), 1.47 (middle), 1.97 (bottom) (top) and larger depth H2 (bottom) at viscous flow, η' = 1.36
Journal of Marine Science and Application
Fig. 23 Wave profile between the ship models and close to the tug for
smaller depth H2, ξ' = 0.62, η' = 1.36 (solid), 1.47 (dashed), and 1.97
(dotted) and inviscid flow with free surface
5 Conclusions
Fig. 21 Free-surface elevation for smaller depth H1 (top) and larger depth A CFD study of interaction forces using various flow models
H2 (bottom) in viscous flow, η' = 1.36 for a small tug ship moving parallel to a larger vessel in shal-
low water has been carried out, and the resulting estimates of
the forces and moment were also compared with the model
at smaller water depth. At the same time, the largest descent of test data and results obtained with the panel code developed
the water level was observed at a larger depth, and in both earlier. The overall impression is that the ship-to-ship interac-
cases, this happened at the intermediate lateral clearance. The tion problem is more challenging than most other problems
latter is probably related to some local resonance phenomena. related to the flow around a single ship hull, and the difficul-
In general, it can be stated that the interaction flow is of ties grow as the lateral clearance decreases. A number of au-
rather complex character, and the role of nonlinear effects thors discovered difficulties in reaching iterative and grid con-
seems to be significant. This explains, at least qualitatively, vergence using even commonly acknowledged and reliable
certain observed “irregularities” in the influence of the viscos- commercial codes. The present study has generally confirmed
ity and wavemaking effects on the estimated values of the those findings indicating that the problem still needs rather
interaction forces and moments. At the same time, it is now extensive and thorough investigation. The authors have made
clear that detailed investigation of these phenomena requires an attempt to study such factors as the fluid viscosity and
much more extensive computations with simpler geometric gravitational waves which may or may not be in effect in
forms. At the same time, the data already obtained can be various hydrodynamic models of the ship-to-ship interaction.
helpful in estimating possible uncertainties related to adapta- Unfortunately, besides a more or less obvious conclusion
tion of this or that flow model and in selecting degree of about importance of the free-surface effects, it turned out that
simplification appropriate in various applications. no definite general recommendations about application of dif-
ferent flow models could be worked out yet. Meanwhile, the
following specific conclusions can be drawn from analysis of
the results:
especially at smaller water depth where anomalous results Benedict K, Köpnick W, Baldauf M, Gronarz A, Friedhoff B, de Mello
Petey F (2011) Improved simulation model and methods for deter-
of the field method were registered.
mination of the required passing distance between ships in restricted
3. The inviscid rigid wall flow models substantially fairways. In: Pettersen B et al (ed) 2nd International Conference on
underpredict the magnitude of the surge interaction force Ship Manoeuvring in Shallow and Confined Water: Ship to Ship
for all lateral distances. While accounting for viscosity Interaction, May 18–20, 2011, Trondheim, Norway, p 29–50
somewhat improves the agreement, rather the free- Chetvertakov AM, Lebedeva MP, Nikushchenko DV (2011) Numerical
investigation of bank influence on ship motion. In: Pettersen B et al
surface effects are here decisive for reliable prediction. (ed) 2nd International Conference on Ship Manoeuvring in Shallow
4. Inversion of the sign of the interaction sway force hap- and Confined Water: Ship to Ship Interaction, May 18–20, 2011,
pening, however, at horizontal clearances, which, howev- Trondheim, Norway, p 83–88
er, are too small to be of practical significance, is only Duffy JT, Renilson MR (2011) The importance of the form of time do-
main forces and moments on berthed ship interaction. In: Pettersen
correctly captured by the “true free surface” flow models. B et al (ed) 2nd International Conference on Ship Manoeuvring in
5. The viscous flow model with the rigid wall-free surface is Shallow and Confined Water: Ship to Ship Interaction, May 18–20,
of the least interest in studies of surface ships interactions 2011, Trondheim, Norway, p 107–116
as being very heavy computationally fails to capture many Faltinsen OM (2011) Modelling of manoeuvring with attention to ship-
ship interaction and wind waves. In: Pettersen B et al (ed) 2nd
effects and does only lead to insignificant accuracy im-
International Conference on Ship Manoeuvring in Shallow and
provements compared to perfect fluid methods. Confined Water: Ship to Ship Interaction, May 18–20, 2011,
6. Account for the free-surface effects is always beneficial Trondheim, Norway, p 1–11
for improvement of the interaction force predictions at Fonfach JMA, Sutulo S, Guedes Soares C (2011) Numerical study of
least when the exact free-surface boundary conditions ship-to-ship interaction forces on the basis of various flow models.
In: Pettersen B et al (ed) 2nd International Conference on Ship
are fulfilled. As the flow in the gap between the sidewalls Manoeuvring in Shallow and Confined Water: Ship to Ship
may be of highly nonlinear character, it is unclear how Interaction, May 18–20, 2011, Trondheim, Norway, p 137–146
adequate a flow model with the linearized free-surface Gronarz A (2011) A new approach in modelling the interaction forces. In:
condition will be and additional studies are required. Pettersen B et al (ed) 2nd International Conference on Ship
Manoeuvring in Shallow and Confined Water: Ship to Ship
7. Account for viscosity in the free-surface flows in most Interaction, May 18–20, 2011, Trondheim, Norway, p 165–173
cases further improves predictions but likely this not very Hess JL, Smith AMO (1964) Calculation of nonlifting potential flow
certain improvement can be sacrificed in favor of possibly about arbitrary three-dimensional bodies. J Ship Res 8:22–44
much faster calculations in the perfect fluid. This conclu- Ibaragi H, Nakiri Y, Furukawa Y (2012) Experimental study of bank
sion may, however, become not correct in the case of non- suction. International Conference on Marine Simulation and Ship
Maneuverability (MARSIM 2012), April 23–27, Singapore, paper
parallel centerplanes not considered here. FP8–2-1410, 9p
8. Poor iterative convergence with the free-surface flow Kadri U, Weihs D (2015) Higher order hydrodynamic interaction be-
models can probably reflect possible macroscopic un- tween two slender bodies in potential flow. J Mar Sci Technol 20:
steadiness of the real flow in narrow gaps, and this effect 249–256. https://doi.org/10.1007/s00773-014-0275-0
Lataire E, Vantorre M, Vanderbranke J (2011) Ship to ship interaction
deserves special attention in the future.
forces during lightering operations. In: Pettersen B et al (ed) 2nd
International Conference on Ship Manoeuvring in Shallow and
In general, in spite of difficulties revealed in the present Confined Water: Ship to Ship Interaction, May 18–20, 2011,
study, reliable CFD codes can serve as valuable tool for ex- Trondheim, Norway, p 211–221
ploring intricacies of the hydrodynamic interaction between Maxwell GR (2011) Interaction—a practical viewpoint. In: Pettersen B
et al (ed) 2nd International Conference on Ship Manoeuvring in
surface ships in close proximity. Shallow and Confined Water: Ship to Ship Interaction, May 18–
20, 2011, Trondheim, Norway, p 233–237
Funding Information The study was performed within the project McArthur PJ (2011) New thinking in ship generated hydrodynamic
PTDC/EMSTRA/5628/2014 “Maneuvering and moored ships in ports— fields: introducing concept for predicting bank suction and rejection.
physical and numerical modeling,” funded by the Portuguese In: Pettersen B et al (ed) 2nd International Conference on Ship
Foundation for Science and Technology (FCT). The first has been fi- Manoeuvring in Shallow and Confined Water: Ship to Ship
nanced by FCT under contract number SFRH/BD/67070/2009. Interaction, May 18–20, 2011, Trondheim, Norway, p 239–251
Newton RN (1960) Some notes on interaction effects between ships close
aboard in deep water. In: Proceedings of first symposium on ship
maneuverability, DTMB Report 1461, Washington, D.C., p 1–24
References
Nikushchenko DV, Zubova AA (2015) Hydrodynamic interaction phe-
nomena investigations during the ship overtaking maneuver for ma-
Arslan T, Pettersen J, Visscher J, Muthanna C, Andersson HI rine related simulators with the use of CFD methods. International
(2011) A comparative study of PIV experiments and numer- Conference on Ship Manoeuvrability and Maritime Simulation
ical simulations of flow fields around two interacting ships. (MARSIM 2015), Newcastle University, United Kingdom, 8–11
In: Pettersen B et al (ed) 2nd International Conference on September 2015, paper 3–4-3, 13p
Ship Manoeuvring in Shallow and Confined Water: Ship to Nikushchenko DV, Zubova AA, Chetvertakov AM (2012) Investigation
Ship Interaction, May 18–20, 2011, Trondheim, Norway, p of ship-to-ship interaction phenomena on the example of Wigley
31–37 body with the use of CFD methods. International Conference on
Journal of Marine Science and Application
Marine Simulation and Ship Maneuverability (MARSIM 2012), MASHCON: International Conference on Ship Manoeuvring in
April 23–27, Singapore, paper FP5–3-1120, 6p Shallow and Confined Water, Hamburg, Germany, 23–25
Pinkster JA (2016) Progress on real-time prediction of ship-ship-shore May 2016, p 305–314
interactions based on potential flow. Proceedings 4th MASHCON: Tuck EO, Newman JN (1974) Hydrodynamic Interactions between
International Conference on Ship Manoeuvring in Shallow and Ships. Proc. 10th Symposium on Naval Hydrodynamics,
Confined Water, Hamburg, Germany, 23–25 May 2016, p 157–167 Cambridge, Mass., USA, p 35–69
Rattanasiri P, Wilson PA, Phillips AB (2014) Numerical investigation of a Van der Molen W, Moes J, Swiegers PB, Vantorre M (2011) Calculation
fleet of towed AUVs. Ocean Eng 80C:25–35. https://doi.org/10. of forces on moored ships due to passing ships in shallow water. In:
1016/j.oceaneng.2014.02.001 Pettersen B et al (ed) 2nd International Conference on Ship
Sadat-Hosseini H, Wu P-C, Toda Y, Carrica P, Stern F (2011) URANS Manoeuvring in Shallow and Confined Water: Ship to Ship
studies of ship-ship interactions in shallow water. In: Pettersen B Interaction, May 18–20, 2011, Trondheim, Norway, p 369–374
et al (ed) 2nd International Conference on Ship Manoeuvring in Van Hoydonck W, Toxopeus S, Eloot K., Bhawsinka K, Queutey P,
Shallow and Confined Water: Ship to Ship Interaction, May 18– Visonneau M (2015) Bank effects for KVLCC2, World Maritime
20, 2011, Trondheim, Norway, p 299–308 Technology Conference (WMTC). Providence, RI, USA,
Sano M, Yasukawa H, Hata H (2012) Experimental study on ship oper- November 3–7, Paper No. 132, p 1–21
ation in close proximity to bank channel. International Conference Van Hoydonck W, Delefortrie G, Peeters P, Mostaert F (2016) Evaluation
on Marine Simulation and Ship Maneuverability (MARSIM 2012), of ROPES for passing ships including a comparison with CFD re-
April 23–27, Singapore, paper FP8–3-1350, 10p sults. Version 4.0. FHR Reports, 15_011. Flanders Hydraulics
Shevchuk I, Böttner C-U, Kornev N (2016) Numerical analysis of the Research: Antwerp, Belgium, 42 p
flow in the gap between the ship hull and the fairway bottom in Versteeg H, Malalasekra W (2007) An introduction to computational
extremely shallow water, Proceedings 4th MASHCON: fluid dynamics: the finite volume method. Pearson Education Ltd,
International Conference on Ship Manoeuvring in Shallow and Harlow
Confined Water, Hamburg, Germany, 23–25 May 2016, p 37–42 Von Graefe A, Shigunov V, el Moctar O (2015) Rankine source method
Sian AY, Maimun A, Ahmed Y, Rahimuddin (2016) Simultaneous ship- for ship-ship interaction problems. J Offshore Mech Arct Eng 137.
to-ship interaction and bank effects on a vessel in restricted water. https://doi.org/10.1115/1.4029316
Proceedings 4th MASHCON: International Conference on Ship
Wnęk AD, Guedes Soares C (2015) CFD assessment of the wind loads on
Manoeuvring in Shallow and Confined Water, Hamburg,
an LNG carrier and floating platform models. Ocean Eng 97:30–36.
Germany, 23–25 May 2016, p 197–207
https://doi.org/10.1016/j.oceaneng.2015.01.004
Simonsen CB, Nielsen CK, Otzen JF, Agdrup K (2011) CFD based pre-
Xiang X, Faltinsen O M (2010) Maneuvering of two interacting ships in
diction of ship-ship interaction forces on a tug beside a tanker. In:
calm water. Proceedings, 11th International Symposium on Practical
Pettersen B et al (ed) 2nd International Conference on Ship
Design of Ships and Other Floating Structures (PRADS 2010),
Manoeuvring in Shallow and Confined Water: Ship to Ship
September 19–24, Rio de Janeiro, Brazil, p 161–171
Interaction, May 18–20, 2011, Trondheim, Norway, p 329–338
Skejić R, Breivik M, Berg TE (2011) Investigating ship maneuvers Xiang X, Skejić R, Faltinsen OM, Berg TE (2011) Hydrodynamic inter-
around a floating structure under the influence of a uniform current action loads between two ships during lightering operations in calm
in deep and calm water. In: Pettersen B et al (ed) 2nd International water. In: Pettersen B et al (ed) 2nd International Conference on Ship
Conference on Ship Manoeuvring in Shallow and Confined Water: Manoeuvring in Shallow and Confined Water: Ship to Ship
Ship to Ship Interaction, May 18–20, 2011, Trondheim, Norway, p Interaction, May 18–20, 2011, Trondheim, Norway, p 383–392
339–350 Xu Y, Sun Y (2012) Study on ship-ship hydrodynamic interaction by
STAR-CCM+ User Guide (2014) Version 9.02.005, CD-adapco ANN optimization. International Conference on Marine
Sutulo S, Guedes Soares C (2008) Simulation of the hydrodynamic in- Simulation and Ship Maneuverability (MARSIM 2012), April 23–
teraction forces in close-proximity manoeuvring. Proceedings of the 27, Singapore, paper FP4–3-0920, 10p
27th Annual International Conference on Offshore Mechanics and Yang H, Wu BS, Miao QM, Xiang X, Berg TE, Kuang XF (2011) Study
Arctic Engineering (OMAE 2008), Estoril. Portugal, 15–19 of the effects of unsteady ship to ship interaction by CFD method.
June 2008, Paper OMAE2008–57938 In: Pettersen B et al (ed) 2nd International Conference on Ship
Sutulo S, Guedes Soares C (2009) Simulation of close-proximity maneu- Manoeuvring in Shallow and Confined Water: Ship to Ship
vers using an online 3D potential flow method. Proceedings of Interaction, May 18–20, 2011, Trondheim, Norway, p 393–398
International Conference on Marine Simulation and Ship Yasukawa H, Yoshida S (2011) Hydrodynamic interaction of two thin
Manoeuvrability MARSIM 2009, Panama City, Panama, 17–20 ships with rudder in close proximity. In: Pettersen B et al (ed) 2nd
August 2009, pp. M-9-1–M-9-10 International Conference on Ship Manoeuvring in Shallow and
Sutulo S, Guedes Soares C (2012) Hydrodynamic interaction forces on Confined Water: Ship to Ship Interaction, May 18–20, 2011,
ship hulls equipped with propulsors. Proceedings of the 31th Annual Trondheim, Norway, p 399–406
International Conference on Offshore Mechanics and Arctic Yuan Z-M, Incecik A (2016) Investigation of ship-bank, ship-bottom and
Engineering (OMAE 2012), Rio de Janeiro, Brasil, 6–11 ship-ship interactions by using potential flow method. Proceedings
June 2012, Paper OMAE2012–84181 4th MASHCON: International Conference on Ship Manoeuvring in
Sutulo S, Guedes Soares C (2016) Parametric study of a modified panel Shallow and Confined Water, Hamburg, Germany, 23–25
method in application to the ship-to-ship hydrodynamic interaction. May 2016, p 83–92
Proceedings 4th MASHCON: International Conference on Ship Yuan Z-M, Incecik A, Dai S, Alexander D, Ji C-Y, Zhang X (2015)
Manoeuvring in Shallow and Confined Water, Hamburg, Hydrodynamic interaction between two ships travelling or station-
Germany, 23–25 May 2016, p 177–185 ary in shallow waters. Ocean Eng 108:620–635. https://doi.org/10.
Sutulo S, Guedes Soares C, Otzen J (2012) Validation of potential-flow 1016/j.oceaneng.2015.08.058
estimation of interaction forces acting upon ship hulls in parallel Zhang C-X, Zou Z-J (2011) Numerical investigation on ship-ship hydro-
motion. J Ship Research 56(3):129–145. https://doi.org/10.5957/ dynamic interaction in restricted waters. In: Pettersen B et al (ed)
JOSR.56.3.100031 2nd International Conference on Ship Manoeuvring in Shallow and
Toxopeous SL, Bhawsinka K (2016) Calculation of hydrpdynamic inter- Confined Water: Ship to Ship Interaction, May 18–20, 2011,
action forces on a ship entering a lock using CFD. Proceedings 4th Trondheim, Norway, p 407–412
A.D. Wnęk et al.: CFD Analysis of Ship-to-Ship Hydrodynamic Interaction
Zhou X Q, Sutulo S, Guedes Soares C (2010) Computation of ship-to- Zou L, Larsson L (2012) Investigation of ship-to-ship interaction during a
ship interaction forces by a 3D potential flow panel method in finite lightering operation in shallow water using a RANS solver.
water depth. Proceedings of the 29th Annual International International Conference on Marine Simulation and Ship
Conference on Offshore Mechanics and Arctic Engineering Maneuverability (MARSIM 2012), April 23–27, Singapore, paper
(OMAE 2010), Shanghai, People’s Republic of China, 6–11 FP6–3-0920
June 2010, Paper OMAE2010–20497