Engineering Structures 210 (2020) 110343

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Launching of ships from horizontal berth by tipping tables – CFD simulation T

of wave generation
⁎
Ivo Senjanovića, , Josip Katavića, Vuko Vukčevića, Nikola Vladimira, Hrvoje Jasaka,b
a
University of Zagreb, Faculty of Mechanical Engineering and Naval Architecture, Ivana Lučića 5, Zagreb, Croatia
b
Wikki Ltd, 459 Southbank House, SE1 7SJ London, United Kingdom

A R T I C LE I N FO A B S T R A C T

Keywords: The paper considers the side launching of ships from a horizontal berth by tipping tables. The launching process
Launching is split into the six phases and a system of nonlinear differential equations of motion is established. The system is
Horizontal berth solved using the finite difference method. The numerical calculations of the side launching of a large tanker are
Tipping table performed and the results are compared against the model test results. The generation of a high surface wave is
Numerical simulation
simulated with CFD as a 2D problem. The numerical simulation captures the physics of the side launching with
Model test
CFD
high accuracy. The advantages of building the floating structures on a horizontal berth are pointed out.

1. Introduction sideways. For the longitudinal launching of ships a single pad is suffi-
cient, while for the launching of semi–submersible platforms two pads
Launching of ships and floating structures by tipping table principle are necessary. Platforms may be launched sideways, too.
is an idea which makes their building on a horizontal berth possible. In order to determine the optimal values of the basic parameters of
This enables an improvement of the production process by introducing the launching system, an extensive theoretical and experimental in-
a high degree of mechanisation and automation, successive block vestigation has been undertaken. This includes the launching theory,
manufacturing and assembling of structural elements, and fitting–out of model testing and structural analysis of the turning pads and launching
structures on the way from the workshop to the quay [1–3]. structures, [5].
The launching system consists of a set of concrete circular slipways The launching theory comprises the definition of the forces acting
built–in into the quay and of the steel turning pads, Fig. 1. The sliding on the dynamic system of the structure and the pad in each launching
surfaces of the pads are extended to the sliding lines of the building phase, i.e. structural weight and inertia, buoyancy, resistance and water
berth. The sliding lines and arches are covered by the polymer coating inertia, friction forces and slipway reaction; the establishment of
in order to reduce friction [4]. equilibrium equations of forces and moments, the derivation of the
A structure assembled on the packing is pushed onto the braked motion equations by reducing the number of setting equations, speci-
tipping table by hydraulic gripper jacks. When the moment of structure fying the transition conditions from one to another phase, and solving
weight overcomes the friction moment of the sliding arches, the tension nonlinear differential equations of motion by the finite difference
forces remove the triggers. The structure then starts to rotate with pads method, [7,8]. The same advanced solution is achieved by transforming
until the displacement excites first its slide along the pads and then its the system of nonlinear differential equations of the second order into
erection in cradles, and finally the floating phase. Thus, the launching extended system of equations of the first order. Then the Runge-Kutta
system has three degrees of freedom, i.e. pad rotation, φ , structure ro- method is applied, [9].
tation, ϑ, and relative structure slide, s , Fig. 2. The combinations of the Special attention is paid to the returning motion of the pad, which is
displacements in the air and in the water define different launching controlled by reducing the buoyancy of the pad, [10].
phases which are shown in Fig. 3, [5,6]. The procedure for the calculation of longitudinal and side launching
Depending on the main characteristics of the launching system it is of ships and floating structures was computer programmed. The non-
possible to launch different structures regarding their shape and size, as linear differential equations of motion are solved by the finite differ-
well as structure blocks, such as grips, docks, platforms etc.; smaller ence method utilizing an iterative algorithm. A large number of nu-
structures being launched longitudinally and larger structures launched merical simulations has been performed for launching of pontoons,

⁎
Corresponding author.
E-mail address: [email protected] (I. Senjanović).

https://doi.org/10.1016/j.engstruct.2020.110343
Received 7 June 2019; Received in revised form 20 December 2019; Accepted 3 February 2020
Available online 21 February 2020
0141-0296/ © 2020 Elsevier Ltd. All rights reserved.
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Fig. 1. The launching system.

Fig. 2. Displacements of the ship and the pad.

ships and a semisubmersible platforms, [5,8]. In this paper an illustrative example related to the side launching of
In order to validate the developed launching theory and the com- a chemical tanker is presented. A correlation analysis of the calculated
puter program, model tests have been performed in Brodarski Institute – and measured results is summarised, [6,12]. A special attention is paid
Marine Research & Special Technologies in Zagreb. A model of the to the generation of a surface wave during the side launching as a
launching system, consisted of only two slipways, was built in a 1:75 criterion for the application of the current launching system in a ship-
scale. A TV camera and a coordinate net were used to record the mo- yard with a limited water basin due to a possible ashore washing. Wave
tions of the launched models. The side launching of three pontoons and generation is simulated by direct CFD simulation as a 2D problem.
two ships was tested as well as the longitudinal launching of a semi-
submersible platform, [5].
The side launching of large ships with the trim angle is question- 2. Two-phase flow model for the CFD analysis
able, since at erection, the ship is supported only by two cradles due to
yaw. As a result, a high stress concentration in the ship structure occurs. This section is dealing with the mathematical model for in-
Furthermore, it is necessary to check the depth of the water and the compressible, viscous, two-phase flow, which is the basis for the CFD
height of the generated waves. framework. The detailed model can be found in [15], while a brief
To investigate the above facts, the model tests of the side launching summary is presented here.
of large tanker were performed in the Maritime Research Institute
Netherlands (MARIN). An existing wooden model in scale 1:68 was
2.1. Incompressible two-phase flow model
adapted for that purpose, [11,12].
In addition, the correlation analysis was performed between the
As both fluids are assumed incompressible, the continuity constraint
calculated and the measured results, which encouraged the building of
reads
the launching system in Shipyard Brodosplit. Five out of the total 13
designed circular slipways of 22 m were completed. In 1993 the first ∇ ·u = 0, (1)
structure, a 4450 DWT barge, was built on the horizontal berth and side
launched by two turning pads, [13,14]. where u is the velocity field. The momentum equation has the following
form

2
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Fig. 3. Launching phases.

3
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

∂u ∂ 1 3. Overset mesh approach

+ ∇ ·(uu) − ∇ ·(νe ∇u) = − ⎡ ∇pd ⎤,
∂t ∂t ⎢
⎣ρ ⎥
⎦ (2)
The Overset mesh approach refers to the use of multiple dis-
where the decomposition of pressure into the hydrostatic and the dy- connected meshes to discretise the flow domain. The component me-
namic component is used ( pd = p + ρg ·x ) . ρ is the piece–wise constant shes, which can be any size, type, or shape, need only overlap each
density field, while νe is the effective kinematic viscosity. other to cover the solution domain completely [23]. Furthermore, a
The interface between the two fluids is considered infinitesimally component mesh resolving one geometric feature may intersect another
thin and the jump interface discontinuities are treated with the Ghost geometric feature [24].
Fluid Method (GFM). In each fluid, density is assumed constant, i.e. This approach has much of its advantages in the computation of
ρ = ρw in water and ρ = ρa in air. Following the notation used by the multi–body and moving body problems as well as in the optimisation
GFM authors [16,17], the jump of density across the free surface can be studies [24].
written as Candidate hydrodynamic applications include ships or submarines
[ρ] = ρw − ρa , (3) with rotating propellers, the launch of torpedo or mini-submarines from
a parent ship, moving appendages and other control surfaces, ship
To facilitate the implementation of the GFM in arbitrary polyhedral motion relative to the sea surface and coastline, and seakeeping simu-
finite volume framework, effective kinematic viscosity is assumed lations with multiple ships in close proximity [25].
continuous across the free surface using the volume fraction α from the Overset mesh generation is conceptually split into the off–body or
VOF approach: background meshes and a number of near-body meshes which resolve
νe = ανe, w + (1 − α ) νe, a, the geometry and the boundary layers. Structured hexahedral meshes
(4)
are often used for their accuracy. However, the overset technique is
where νe, w is the effective kinematic viscosity in water and νe, a is the routinely applied using the hybrid unstructured meshes for the highly
effective kinematic viscosity in air, allowing the use of general eddy- automated meshing of complex configurations [23].
viscosity turbulence models, [18]. Eq. (4) states that the effective ki- A CFD solution on the system of meshes requires the coupling of the
nematic viscosity does not have a discontinuity across the free surface, solution between meshes in the overlapped regions. This is typically
although it might have a steep gradient if the smearing of the interface performed by identifying appropriate inter–mesh boundary locations in
in the VOF approach is confined to a narrow region. one mesh and obtaining the value to be applied by interpolating the
The kinematic boundary condition [19,20] defines a continuous solution from meshes that overlap the region [25].
velocity field across the free surface The Domain Connectivity Information (DCI) consists of the locations
that are to be excluded from the computation, the location of the in-
[u] = u w + ua = 0. (5)
ter–mesh boundary locations, and the corresponding interpolation
Eq. (5) states that the velocity field infinitesimally close to the free sources. This domain connectivity information is computed by a code
surface in the water must be the same as the velocity field in- typically called an overset fringe assembly code [25]. The overset fringe
finitesimally close to the free surface in the air. assembly algorithm used in this work is described in Section 3.2.
Furthermore, neglecting the surface tension effects as in [19,20]
yields a continuous pressure field 3.1. Overset mesh cell types
[p] = 0. (6)
According to their role in the solution process of the governing
Taking into account Eqs. (3), (5) and (6), it is possible to arrive at equations, there are three base overset mesh cell types (also called
two jump conditions for the dynamic pressure [15] overset types): acceptor, live and hole cells.
Basic and special overset types are presented in the form of a dia-
[pd ] = −[ρ] g ·x , (7)
gram in Fig. 4. Arrow points from derived to a base type. The diagram
should be read from the bottom to the top. For example, both master
⎡ 1 ∇p ⎤ = 0. and extended donor are donors, while the donor is a live cell and the
⎢ d⎥
⎣ρ ⎦ (8) live cell is an overset mesh cell.
The jump conditions presented in Eqs. (6) and (7) require a re- The live cells are used to discretise the governing equations. Field
presentation of the interface between the two fluids. In present work, values in those cells are obtained by solving a system of linear equa-
the capturing of the free surface between is achieved with the VOF tions.
method [21]. The VOF method is based on the indicator function α , A donor cell is a special type of the live cell. It is used for overset
which represents the volume fraction interpolation, i.e. donor is the interpolation source for providing the
value applied to the appropriate acceptor cell.
Vw
α= , Field values at acceptor cells are obtained exclusively by inter-
V (9)
polation from the donor cells. The field value at an acceptor cell is
where Vw is the volume of water inside a control volume V . VOF method therefore not affected by the values of its neighbouring cells, but field
is conservative because α represents a physical, conserved property, values at acceptor's neighbouring cells are affected by the field value in
bounded between 0 and 1. Following Rusche [22], transport equation the acceptor cell. After the acceptor candidates are identified, the donor
for α reads: search procedure is employed to find the appropriate master donor and
extended donors for all acceptors.
∂α
+ ∇ ·(uα ) + ∇ ·(ur α (1 − α )) = 0, The extended donors are neighbours of the master donor, and to-
∂t (10)
gether with the acceptor and the master donor, they form an inter-
where the last term serves to prevent excessive smearing of the free polation stencil. Master donor is a cell closest to the acceptor cell. The
surface based on the compressive velocity field ur [22]. The term is interpolation stencil depends on the chosen overset interpolation
active only in the smeared interface region due to the α (1 − α ) non- scheme.
linear pre-factor. Each acceptor has a unique master donor, while one donor can be
After the α is advected, the location of the interface given by α = 0.5 paired with multiple different acceptors. Pairing one donor with a
is used in the discretisation of the momentum equation, Eq. (2), obeying number of acceptors may lead to an unphysical overall solution, which
the dynamic pressure jump conditions given by Eqs. (7) and (8). occurs when the fringe layer of the overset meshes does not match in

4
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Fig. 4. Overset mesh cell types.

Table 1 form unsuitable donor/acceptor pairs are converted to holes in the next
Algorithm parameters. iteration. Hole cells which are the result of fringe assembly procedure
Parameter Data type Default value
are called fringe holes.
Also, it is possible that holes or other acceptors surround a specific
minLocalSuit scalar 0 acceptor cell completely, so this cell may become a hole as well in order
specifiedIterationsNumber label 4 to lower possibly time–consuming interpolation. For that purpose, the
additionalIterations Switch true
filtering procedure is employed. The filtering procedure is the final step
orphanSuitability scalar −1
of the fringe assembly procedure.
The field value in a hole cell does not depend on the field values in
neighbouring cells, nor does the hole cell value affect the field values at
the neighbouring cells. Field values at hole cells are user–prescribed,
i.e. are defined via overset boundary conditions.

3.2. Fringe assembly algorithm

The algorithm uses a fringe quality gradient to choose optimal

donor/acceptor assembly. Four parameters control the algorithm and
all of them have default values. The parameters together with their data
types and default values are presented in Table 1. The scalar data type is
equivalent to standard double C++, label to int and Switch to bool.
Among the parameters, the user should choose the Donor Suitability
Function (DSF) function which calculates the suitability of the donor/
acceptor pair indicating their similarity regarding user–prescribed
Fig. 5. Iterative procedure diagram.
property, usually cell volume.
The presence of orphans is undesirable because if donor and ac-
resolution. Therefore, in the overlapping zone, cells should be of ceptor, which form a single donor/acceptor pair, are far apart, an un-
comparable size in both meshes. Also, the coarser of the two meshes physical solution can be obtained. The orphan suitability is therefore
determines the level of the discretisation (interpolation) error. not calculated in a standard manner using DSF, rather it is user pre-
An orphan is an acceptor cell for which donor is found, but its cell scribed. By prescribing negative suitability to an orphan, average suit-
centre does not fall within acceptor’s bounding box. The orphan cell is ability is being degraded. That reduces the possibility that fringe as-
also called invalid donor/acceptor pair. The presence of an orphan cells sembly with an orphan present is being chosen for the final overset
generally indicates that there is insufficient overlap between meshes or assembly. By default, orphan suitability is −100% which means that
that the mesh resolutions in the overlapping region do not match well. the chosen donor is not suitable at all.
If the overset body meshes overlap each other, it is necessary to The first step in fringe assembly procedure is to get an initial guess
deactivate all regions on these meshes which are outside of the com- for holes and acceptors. The neighbourhood search starts from hole
putational domain, i.e. covered by other bodies or lying outside the neighbouring cells and optionally from user–specified patches and
background mesh [24]. This process is called hole cutting, and deacti- holes.
vated cells are called cut holes. In addition, since the parts of each mesh After an initial guess is obtained, the algorithm loops through the
may be covered by another overset mesh, the holes may be created in received donor/acceptor pairs, checks if an orphan is present and if it is,
these meshes as well [24]. prescribes user–defined suitability. Otherwise, it calculates donor/ac-
During the iterative procedure, acceptors that together with donors ceptor suitability using DSF.

5
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Fig. 6. The water basin with ship model on the tipping table.

Table 2 Pairs for which suitability is lower than minimum local suitability
The ship particulars in launching condition. are categorised as unsuitable while pairs for which suitability is higher
Length between perpendiculars Lpp = 310 m than minimum local suitability are categorised as suitable. After passing
Breadth B = 53.96 m through all donor/acceptor pairs, the average suitability is calculated.
Draft fore Tf = 0.45 m The algorithm tracks iteration history, i.e. it stores information that
Draft aft Ta = 5.91 m is needed to reconstruct fringe layer assembly from a specific iteration
Displacement D = 40, 000t and to calculate the slope using the linear regression.
Centre of gravity above base ¯ = 18.3 m
KG
After a user-defined number of iterations is performed and if user
Centre of buoyancy from station 0 x b = 131.4 m
Metacentric height ¯ = 56.52 m prescribed that additional iterations should be done, algorithm per-
GM
Longitudinal radius of gyration kyy = 87 m forms a linear regression to calculate the suitability slope coefficient,
Transverse radius of gyration k xx = 22.68 m i.e. quality or suitability gradient. Here, average suitability is the pre-
dicted variable (on y axis) and the iteration number is the predictor
variable ( x axis).
It is assumed that if the gradient is positive, fringe assembly with
greater average suitability may be found and an additional iteration

6
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Fig. 7. The arrangement of force transducers.

will be made. Data stored during the first iteration is then being deleted. filtering procedure: it is possible that holes or other acceptors surround
In Fig. 5, as an example, the iterative procedure is demonstrated a certain acceptor cell completely, so this cell becomes a hole as well to
where the average suitability against the iteration number diagram is lower the expense of the overset interpolation.
shown. After four iterations, the suitability gradient is calculated. Due For more details on the fringe assembly algorithm and overset mesh
to a positive suitability gradient, the data from iteration 1 is then de- approach as implemented in foam-extend, the reader is referred to
leted. One additional iteration is then made, i.e. iteration 5. After the Katavić [26].
iteration 5, due to a negative suitability gradient, the iterative proce-
dure is terminated. For final overset fringe assembly, the iteration with 4. Model test and numerical simulation of a chemical tanker
the highest average suitability value is chosen, i.e. iteration 4. launching
The algorithm checks two conditions on each iteration: (1) If there is
any unsuitable donor/acceptor pair and (2) If a desired number of The water basin of the Shipyard “Brodosplit” – Split, Croatia, is
iterations is made (i.e. global criterion is satisfied). If those two criteria shown in Fig. 6, indicating the tipping table, layout of the water basin
are not reached, it finds a new set of acceptors. The new acceptors are with the water depth. The proposed launching system consists of 13
immediate neighbours of unsuitable acceptors which are eligible (are circular slipways, having radius of 18 m, and distributed equidistantly
not holes nor acceptors from current iteration). Acceptors from un- each 22 m.
suitable donor/acceptor pairs are converted to holes. Otherwise, the The side launching of large ships with trim from the horizontal
iterative procedure is finished. berth by tipping table is problematic because the ship at erection is
When the iterative process is finished, the algorithm identifies an supported only by a couple of cradles due to yaw. Thus, the cradle
iteration with the highest average suitability and performs the pairs reactions cause high stress concentration on the ship structure, which

7
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

coordinates of the ship's aft and fore peaks, x , y , z , were recorded too.
The tests were performed varying the ship displacement and position of
the centre of gravity in the longitudinal and vertical direction. The
recorded ship motion is shown in Fig. 8 at the fore and aft pads. Some
differences due to yaw motion may be noticed.
Model test and numerical calculation results performed by two pads
are elaborated in details in [11,12] respectively. Here, the most im-
portant results are presented and compared in the time domain.
The inclination of the ship and the pad are shown in Fig. 9. The
calculated values for the ship and the pad are related to the cross-sec-
tion at the longitudinal position of the centre of gravity. The mea-
surement was performed separately for the ship cross-section at the
centre of gravity, and the fore and aft pads. The agreement between the
calculated and the measured results is relatively good up to the phase 5,
when their discrepancies start to increase. The measured inclinations of
the ship and the pad are bounded by the calculated values. Thus, the
calculation results are on the conservative side.
In [12] additional information on correlation analysis between the
calculated and measeured results can be found, i.e. ship yaw and pitch,
ship transverse velocity, berth normal reaction and buoyancy as func-
tion of time, etc.

Fig. 8. Sketch of the motion of the ship and the pad. 5. CFD simulation of sideways ship launching

has to be checked. Also, a large ship launched in such a way in shallow In this section, the numerical simulation setup and the numerical
water may touch the bottom, while a high generated wave in a limited results are presented in detail. The goal of the simulation is to validate
water basin may cause other issues. the Naval Hydro Pack [27–35], combined with the new overset fringe
In order to investigate the above facts, the model tests of the side algorithm, on the sideways launching case. The primary validation item
launching of a 260 000 DWT tanker were performed in the Maritime is the generated wave amplitude during the sideways launching where
Research Institute Netherlands (MARIN), [11]. An existing wooden model the numerical results are compared with the experimental data [11].
of scale 1:68 was adapted for that purpose. The ship particulars in According to the experimental data [11], the highest observed wave
launching conditions are presented in Table 2. had a maximum amplitude of 2.2 m, while the lowest maximum am-
The ship model on the tipping table is shown in Fig. 6. Two circular plitude was 1 m. Between those values, a wide scatter of maxima was
slipways were used. The main particulars of slipways and turning pads observed. Measures should be taken for a generated wave of approxi-
with force transducers are shown in Fig. 7, where full-scale dimensions mately 2 m [11]. Due to the diffraction effects, the generated wave can
are indicated. The port side bilge is supported by two ship–bound reach as much as twice the value observed at the wave probe [11].
cradles. All slide surfaces are equipped with stainless steel polished Therefore, as a relevant physical quantity for testing and validation, the
strips. first generated wave amplitude is considered.
Transverse and vertical forces, Fy and Fx , were measured at both
ends of the pads by strain gauge force transducers, Fig. 7. The

Fig. 9. Inclination diagram: ψ -ship angle, φ -pad

angle.

8
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Fig. 10. The arrangement of two

overset meshes and free surface posi-
tion at initial condition.

Fig. 11. Dimensions of the

backgroundMesh region.

5.2. Finite volume overset meshes

The meshes are block-structured and are generated using the

Pointwise software [23]. The backgroundMesh consists of 52 724 cells,
while the shipMesh mesh consists of 21 174 cells. All cells are hex-
ahedral. The overset mesh system thus consists of 73 898 cells. To
capture the wave properly, the backgroundMesh mesh is made with 20
cells per expected wave height. Also, to ensure a good quality overlap
assembly, the shipMesh is generated in a way that the bottom and side
blocks consist of 40 layers of cells, while upper blocks, which are not
relevant for wave elevation, consist of 18 layers of cells. The combined
meshes at the initial condition of side launching are presented in
Fig. 13.
The donor region for the shipMesh is the backgroundMesh, and vice
versa. For overset mesh assembly the adaptiveOverlap algorithm pre-
Fig. 12. Dimensions of the shipMesh region. sented in Section 3.2 with default parameters is employed. The
boundaries that determine holes are presented in Table 4.
5.1. Computational domain
5.3. Boundary conditions
The main data of the water basin with bottom topography and the
All computational boundaries are marked in Fig. 14. Prescribed
ship model for the test case No. 5874 from [11] are shown in Fig. 6.
boundary conditions for velocity, phase fraction and dynamic pressure
According to the coordinate system shown in Fig. 10, the deepest point
fields are given in Table 3. In addition to the boundary conditions given
of the Water Basin is at −17 m, while the shallowest point is at −9 m.
in Table 3, overset boundary conditions for every field are presented.
A 2D overset mesh is created according to the water basin (see
The overset boundary conditions are used for overset interpolation, and
Fig. 6). The computational domain is spatially discretised with two
to prescribe hole cell values for certain fields.
overset meshes. The first mesh is named backgroundMesh and the
For a general variable ϕ , following keywords are used for certain
second one is named shipMesh. The arrangement of the overset meshes
boundary conditions:
and the free surface position at the initial condition are presented in
Fig. 10.
Dimensions of the backgroundMesh mesh are presented in Fig. 11. • Zero gradient (zeroGradient): n·∇ϕ = 0
The overall mesh length is 396 m, while the overall mesh height is • Fixed value (fixedValue): ϕ = ϕ . b

63 m. Initial free surface position is set to (0, 0) in accordance with

The other boundary conditions are explained hereafter:
x − y coordinate system (Fig. 10).
The shipMesh mesh dimensions are shown in Fig. 12. The overall
mesh length is 64 m, while the overall mesh height is 41.05 m. • Inlet outlet (inletOutlet): the same as zeroGradient, but it switches
to fixedValue if the velocity vector next to the boundary is directed
towards the inside of the domain (backward flow),
• Pressure inlet outlet velocity (pressureInletOutletVelocity): the same

Fig. 13. The combined meshes at the

initial condition of the side launching
simulation.

9
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Table 3
Prescribed boundary conditions.
Patch ID Patch name u α pd

1 coastHorizonal fixedValue zeroGradient zeroGradient

2 coastVertical fixedValue zeroGradient zeroGradient
3 left inletOutlet inletOutlet zeroGradient
4 right fixedValue zeroGradient zeroGradient
5 seabed fixedValue zeroGradient zeroGradient
6 top pressureInletOutletVelocity inletOutlet fixedValue
7 boat movingWallVelocity zeroGradient zeroGradient
8 boatOuter emptyOverset emptyOverset emptyOverset

Table 4
Hole–defining boundaries.
Boundary ID Name

1 coastHorizonal
2 coastVertical
4 left
5 right
6 seabed
7 top
8 boat

as zeroGradient, but in the case of the backward flow assigns a ve-

locity based on the flux in the patch-normal direction,
• Moving wall velocity (movingWallVelocity): sets the velocity to the
desired value for moving walls when employed in moving mesh
cases,
• Empty overset (emptyOverset): boundary condition used for overset Fig. 15. Horizontal displacement (experimental data).
interpolation and to prescribe hole cell values for certain fields.

5.4. Prescribed motions

The numerical simulations are performed with both the theoretical

and the experimental motions prescribed, and wave elevations are
compared with the model test result. The experimental data is extracted
from diagrams given in [11]. The translational motions are measured
for aft and fore of the ship. As 2D computations are performed, the
motions are reduced to the centre of gravity of the ship (CoG) using
inverse distance weighting (IDW).
The extracted and interpolated motions are shown in Figs. 15 and
16. The extracted motions are denoted with the solid line (fore motion)
and the long–dashed line (aft motion), while the interpolated, i.e.
prescribed motions are denoted with the short–dashed line.
The roll angle is measured only for the first 28 s and is extrapolated
to capture the first wave at the wave probe location. The roll angle is
extrapolated in such a way that the ship returns to a neutral position
and stays therein, Fig. 17. Fig. 16. Vertical displacement (experimental data).
In Figs. 18 and 19 prescribed theoretical motions are shown. The
theoretical motions are obtained by solving equations of motion

Fig. 14. The computational domain with boundary IDs.

10
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

5.5. Simulation properties

The density of the water is ρw = 998 kg/m3, while the kinematic

viscosity is νw = 1.05·10−6 m2/s . The density of the air is ρa = 1 kg/m3,
while the kinematic viscosity is 1.48·10−5 m2/s . The surface tension
coefficient is set to zero because surface tension effects are considered
negligible for large-scale flows [37]. No turbulence modelling has been
used in this simulation.
At the initial condition the water in the basin is calm and waves are
generated exclusively due to ship's motion in water and diffraction ef-
fects of the walls and the bottom.

5.6. Results and discussion

In Fig. 20 the ship and the water are shown, where the darker
shading denotes deeper locations.
As reported in [11], it was observed that the reflected wave flooded
Fig. 17. Roll angle (experimental data). the quay from which the ship was launched. In [11], the time of the
event is not reported, while in CFD simulation it occurs at t ≥ 69.5 s
(see Fig. 20).
The surface elevation at the probe is given in Fig. 21, where CFD –
experimental motions stands for calculated surface elevation using CFD
with prescribed experimentally measured motions, EFD stands for the
experimentally measured surface elevation, and CFD – theoretical mo-
tions stands for the calculated surface elevation using CFD with pre-
scribed theoretical motions.
The maximum measured wave amplitude is 1.997 m, and it is re-
gistered at t = 36.04 s , which corresponds to the wave amplitude from
experimental test case No. 5874. The maximum calculated wave am-
plitude (with CFD using the prescribed theoretical motions) is 2.004 m
at t = 37.22 . The relative discrepancy for the wave amplitude is
0.36%. The phase difference of 1.17 s between the experimental and the
numerical data is larger, with relative discrepancy of 3.16%. The re-
lative discrepancy is calculated as the absolute difference between the
measured value and the value obtained by CFD divided by the value of
the measured value.
Fig. 18. Prescribed horizontal and vertical displacement.
The second wave amplitude is underestimated which is expected
because the prescribed roll motion is simply set to zero for t > 28 s (see
Fig. 19).
The difference between the CFD results with prescribed experi-
mental motions and the CFD results with prescribed theoretical motions
is noticeable. It occurs because the theoretical motions overestimate the
vertical displacement of the ship, i.e. according to the theoretical mo-
tions, the ship penetrates deeper into the water than it is the case in
experiments.
When the ship penetrates the water, the mean free surface level is
raised for about 0.5 m due to the 2D effects. The disturbance propagates
after the first wave amplitude is captured.
This indicates that the 2D simulation is an appropriate tool only for
capturing the first wave elevation. Also, the maximum measured yaw
angle for the test case No. 5874 is 6.5°, which indicates that the ship
performs 3D motion in reality. 3D simulation is not performed, because
not all water basin dimensions were available, which would also affect
Fig. 19. Prescribed roll motion. the results due to 3D effects.
The numerical calculations were carried out on a desktop computer
whose characteristics are presented in Table 5. Both of the calculations
presented in [7]. Nonlinear differential equations of motion are solved
lasted for 7 h, therefore less than 30 CPU hours were spent per calcu-
by the finite difference method using the software named LATUP
lation.
(LAunching by TUrning Pads) [36].
Theoretical motions are available for the first 5 launching phases,
6. Conclusion
i.e. until the final phase, where the ship freely floats performing com-
plex motion. To capture the first wave elevation, motions are extra-
The extensive numerical and experimental investigation of long-
polated.
itudinal launching of the platform described in [7] and side launching
of the tanker show that this idea is feasible also in the case of large
floating units. The correlation between the calculated and the measured
results is acceptable from an engineering point of view, especially in the

11
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Fig. 20. Time–series of the ship launching simulation.

12
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Fig. 20. (continued)

first part of launching. side launching of a ship. Furthermore, comparing simulations with
Side launching of a ship is more complicated than the longitudinal prescribed experimental and theoretical motions shows significant
launching since trim causes yawing during the ship's sliding along the sensitivity of the wave amplitude to the imposed motions.
pads. Thus, in phases 4 and 5 the ship is supported only by two cradles, The advantage of launching the ships and offshore structures by
respectively. In order to reduce the load concentration on the ship tipping table arrangement is the possible introduction of higher level of
structure it is necessary to trim the ship close to an even keel. mechanisation and automation in the ship–building process, simulta-
The main goal of the CFD launching simulation was to compare neous building of several units employing only one launching system,
experimental and numerical results for the first wave amplitude and and cost reduction in comparison with the classical method of ship–-
phase. The relative error between the measured and the calculated building. Sideways ship launching is also performed when a ship is built
wave amplitudes is 0.36%, while the relative error between their phases in a yard on a river or canal either because the hull is especially long or
is 3.16%. It is concluded that the numerical simulation captures side the channel it is being launched into is narrow.
launching physics with high accuracy, thus providing the proof of the
validity of 2D CFD approach. More detailed analysis of the wave field
would require a 3D CFD simulation. Declaration of Competing Interest
It is concluded that a 2D simulation is an appropriate tool only for
capturing the first wave elevation, due to 3D effects which occur during The authors declared that there is no conflict of interest.

13
I. Senjanović, et al. Engineering Structures 210 (2020) 110343

Appl Math 1980;6(1):19–26.

[10] Anzulović I. A dynamic model of turning slipway's returning motion. Brodogradnja
1997;45(1):44–50.
[11] Feikema G, Luisman H, Pinkster J. Report No. 010001-1-BT: Launching tests for a
260 000 DWT ship. Maritime Research Institute Netherlands (MARIN); 1989.
[12] Senjanović I. Numerical and experimental investigation of side launching of tanker
by tipping table principle. Brodogradnja 1998;46(3):264–70.
[13] Abdulaj R, Ljubetić M. The effect of eccentricity on kinematics of side launching
with rotative standing ways (in Croatian). Proceedings of the 11th symposium
theory and practice of ship-building in memoriam Professor Leopold Sorta,
Dubrovnik. 1994.
[14] Cagalj A. Building a barge on a horizontal berth (in Croatian). Brodogradnja
1995;43(3):58–62.
[15] Vukčević V, Jasak H, Gatin I. Implementation of the ghost fluid method for free
surface flows in polyhedral finite volume framework. Comput Fluids
2017;153:1–19.
[16] Desjardins O, Moureau V, Pitsch H. An accurate conservative level set/ghost fluid
method for simulating turbulent atomization. J Comput Phys
2008;227(18):8395–416.
[17] Huang J, Carrica P, Stern F. Coupled ghost fluid/two-phase level set method for
curvilinear body-fitted grids. Int J Numer Meth Fluids 2007;55(9):867–97.
[18] Wilcox D. Turbulence modeling for CFD vol. 2. La Canada (CA): DCW industries;
Fig. 21. Comparison of the surface elevation. 1998.
[19] Batchelor CK. An introduction to fluid dynamics. Cambridge University Press; 1967.
[20] Ito K, Li Z. Interface conditions for stokes equations with a discontinuous viscosity
Table 5 and surface sources. Appl Math Lett 2006;19(3):229–34.
Hardware specifications. [21] Ubbink O, Issa R. A method for capturing sharp fluid interfaces on arbitrary meshes.
J Comput Phys 1999;153(1):26–50.
CPU [email protected] GHz [22] Rusche H. Computational fluid dynamics of dispersed two-phase flows at high
phase fractions PhD thesis Imperial College London (University of London); 2003
Number of cores 4 [23] Wyman N. Why you should be using overset grids; 2019. http://www.pointwise.
RAM 16 GB DDR4 com/theconnector/2014-September/Why-You-Should-Be-Using-Overset-Grids.htm
[Online, accessed 5 June 2019].
[24] Hadzic H. Development and application of finite volume method for the compu-
tation of flows around moving bodies on unstructured, overlapping grids.
Acknowledgements Technische Universität Hamburg; 2006.
[25] Celeritas Simulation Technology, LLC. What is overset?; 2019. http://
This investigation received funding within the International co- celeritassimtech.com [Online, accessed 5 June 2019].
[26] Katavić J. Development of automatic fringe assembly algorithm for overset meshes
operation project Global Core Research Centre for Ships and Offshore
in foam-extend software Master's thesis Croatia: Faculty of Mechanical Engineering
Plants (GCRC SOP Grant No. 2011-0030669), established by the South and Naval Architecture, University of Zagreb; 2018
Korean Government (MSIP) through the National Research Foundation [27] Jasak H. CFD analysis in subsea and marine technology. IOP conference series:
materials science and engineering 2017;vol. 276. page 012009.
of South Korea (NRF). The authors would like to express their gratitude
[28] Jasak H, Vukčević V, Gatin I, Lalović I. Cfd validation and grid sensitivity studies of
to Dr. Guilherme Vaz for sharing the original report of the experimental full scale ship self propulsion. Int J Nav Archit Ocean Eng 2019;11(1):33–43.
campaign performed in MARIN, Netherlands. [29] Gatin I, Cvijetić G, Vukčević V, Jasak H, Malenica Š. Harmonic balance method for
nonlinear and viscous free surface flows. Ocean Eng 2018;157:164–79.
[30] Gatin I, Vukčević V, Jasak H, Rusche H. Enhanced coupling of solid body motion
References and fluid flow in finite volume framework. Ocean Eng 2017;143:295–304.
[31] Gatin I, Vukčević V, Jasak H. A framework for efficient irregular wave simulations
[1] Vukman Š. Tipping launching ways in modern shipbuilding technology (in using higher order spectral method coupled with viscous two phase model. J Ocean
Croatian). Brodogradnja 1981;29(5):317–22. Eng Sci 2017;2(4):253–67.
[2] Vukman Š. Launching by means of tipping table arrangement (in Croatian). [32] Gatin I, Vukčević V, Jasak H, Seo J, Rhee S. CFD verification and validation of green
Brodogradnja 1982;30(2):59–68. sea loads. Ocean Eng 2018;148:500–15.
[3] Vukman Š. Construction of ships on horizontal level with launching over tipping [33] Jasak H, Gatin I, Vukčević V. Monolithic coupling of the pressure and rigid body
table. Brodogradnja 1982;30(2):135–41. motion equations in computational marine hydrodynamics. J Mar Sci Appl
[4] Vukman Š, Krstulović A, Papić J. Method of launching by means of polymeric 2017;16(4):375–81.
materials instead of lubricants. Brodogradnja 1986;34(2):99–102. [34] Vukčević V, Jasak H, Malenica Š. Decomposition model for naval hydrodynamic
[5] Senjanović I, Ljubetić M, Vukman Š. Investigation of launching of ships and offshore applications, part I: Computational method. Ocean Eng 2016;121:37–46.
structures from horizontal berth by tipping table principle. Int Shipbuild Prog [35] Vukčević V, Jasak H, Malenica Š. Decomposition model for naval hydrodynamic
1986;33(388):218–30. applications, part II: Verification and validation. Ocean Eng 2016;121:76–88.
[6] Senjanović I. Launching of ships and floating structures from horizontal berth by [36] Senjanović I. Building of ships and floating structures on horizontal berth and their
tipping table. J Ship Prod 1998;14(4):265–76. launching by tipping table principle. The proceedings of the international offshore
[7] Senjanović I. Launching theory of ships from horizontal berth by tipping table ar- and polar engineering conference. International Society of Offshore and Polar
rangement. Brodogradnja 1998;46(4):359–69. Engineers; 1999.
[8] Senjanović I, Grubišić I. Side launching of ships from horizontal berth by tipping [37] Carrica PM, Wilson RV, Noack RW, Stern F. Ship motions using single-phase level
table arrangement. Brodogradnja 1999;47(2):135–43. set with dynamic overset grids. Comput Fluids 2007;36(9):1415–33.
[9] Dormand JR, Prince PJ. A family of embedded Runge-Kutta formulae. J Comput

14