You are on page 1of 11

Alexandria Engineering Journal (2019) 58, 849–859

H O S T E D BY
Alexandria University

Alexandria Engineering Journal


www.elsevier.com/locate/aej
www.sciencedirect.com

ORIGINAL ARTICLE

Numerical study on the hydrodynamic drag force of


a container ship model
Ahmed G. Elkafas a,*, Mohamed M. Elgohary a, Akram E. Zeid b

a
Department of Naval Architecture and Marine Engineering, Faculty of Engineering, Alexandria University, 21544
Alexandria, Egypt
b
Department of Marine Engineering Technology, Faculty of Maritime Transport and Technology, Arab Academy for Science,
Technology & Maritime Transport, 21937 Alexandria, Egypt

Received 6 May 2019; revised 7 July 2019; accepted 28 July 2019


Available online 16 August 2019

KEYWORDS Abstract In recent years, importance has been recognized increasingly for the reduction of fuel
Computational fluid dynam- consumption of ships in a seaway to reduce green-house gas emissions from shipping. From a ship
ics; design viewpoint, it is of crucial importance to establish reliable prediction methods for ship’s resis-
ANSYS-CFX; tance and propulsive power. The required power for the propulsion unit depends on the ship resis-
Ship resistance; tance and speed. There are three solutions for the prediction of ship resistance as follow analytical
Holtrop method; methods, model tests in tanks and Computational Fluid Dynamics (CFD). The rapid developments
Container ship; in computers and computational methods increased the opportunities of the CFD simulation to be
Hydrodynamic drag used in the ship design process. The present paper aims at simulating ship resistance using CFD
simulations method which is conducted using ANSYS-CFX software package. As a case study,
Container ship scale model is investigated. The results show the ship resistance which calculated
at various ship speeds and Froude number. Predicted results for resistance components at various
Froude numbers were compared against Resistance results computed by using Holtop method. It is
shown that the simulation results agree fairly well with the results computed from Holtrop method,
and that ANSYS-CFX code can predict ship resistance.
Ó 2019 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria
University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

1. Introduction reduce green-house gas emissions from shipping. It is esti-


mated that almost 90% of global trade is mobilized by ship-
In recent years, importance has been recognized increasingly ping. In the process of carrying such an immense amount of
for the reduction of fuel consumption of ships in a seaway to goods, ships produce roughly 3% of global CO2 emissions,
14–15% NOX emissions and 16% of SOX emissions [1].
Increasing environmental concerns and adoption of different
* Corresponding author.
emission related regulations have motivated both shipbuilders
E-mail addresses: es-ahmed.gamal1217@alexu.edu.eg, marineengi- and owners to opt for more efficient and environment-friendly
neer36@gmail.com (A.G. Elkafas).
vessels. From a ship design viewpoint, it is of crucial impor-
Peer review under responsibility of Faculty of Engineering, Alexandria
tance to establish reliable prediction methods for ship’s resis-
University.
https://doi.org/10.1016/j.aej.2019.07.004
1110-0168 Ó 2019 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
850 A.G. Elkafas et al.

tance and propulsive power for a range of operational speeds. CFX package which has implemented a RANS solver model.
With the development of power-driven vessels in the nine- In this paper, CFD simulations of the hull would be conducted
teenth century, Evaluation of ship hydrodynamic parameters with the different velocities of flows. Predicted results for resis-
and components of ship resistance began to be important. tance components at various Froude numbers were compared
The resistance of a ship at a selected speed and displacement against resistance results computed by using Holtop method.
is the fluid force acting on the ship in the opposite direction The study presented herein aims to assess the deviation of
of its forward motion [2]. It can be defined as the force the CFD simulation results when compared with the results
required to move the ship at a particular speed [3]. There are of Holtrop Method. The detail information about the geome-
three solutions for the prediction of ship resistance as follows try of the model, boundary layer, boundary domain, meshing
analytical methods, model tests in tanks and Computational process, study conditions, testing installations would be pre-
Fluid Dynamics (CFD). sented on the following sections of the paper. Generally, it is
Towing tank model testing has been since early times, the shown that the simulation results agree fairly well with the
most reliable method for power predictions, because of its high results computed from Holtrop method, and that ANSYS-
cost and demand for a fixed and given geometry it can be used CFX code can predict ship resistance
only in the basic design stage when the design parameters
related to the vessel’s hull geometry are fixed and serves as a
final validation and benchmark to be used afterward in the 2. Theoretical background
conclusion of the shipbuilding process during sea trials [4].
On the other hand, the last decade has seen the exponential 2.1. Ship resistance and power
growth of computational fluid dynamic solvers that solve Rey-
nolds Averaged Navier Stokes equations over the hull form in Ship Resistance is the total force that opposes the forward
finite volume approaches [5]. Although originally the computa- motion of the ship at a corresponding speed in calm water.
tional cost was penalizing its application in early design stages, Alternately, the force required to tow a ship in calm water at
the advances in computing hardware and software allowed the a constant speed. In order to achieve a forward motion, ves-
integration of Computational Fluid Dynamics (CFD) in the sel’s thrust must overcome the total resistance. The total resis-
early ship hull form design and optimization [6]. For the last tance of ship consists of air and hydrodynamic resistances.
few decades computer technology has been exponentially Hydrodynamic resistance is affected by the wetted surface area
evolving in an unrestrained manner bringing massive Central of the ship hull. It can be divided into two main components
Processing Unit (CPU) power for acceptable price to the reg- according to two approaches. It is composed of either the fric-
ular high-end users. Computational Fluid Dynamics (CFD) tional and residual resistances or the viscous and wave resis-
is greatly dependent on CPU power since the basis in solving tances [2]. Particularly, frictional resistance component plays
a fluid flow of any kind is found in Navier-Stokes (NS) equa- an important role as it takes the largest portion of the total
tions with extension. These require high computational effort ship resistance for the majority merchant ships. For example,
to acquire a satisfactory solution. So that Computational skin friction can account for up to 90% of the total resistance,
Fluid Dynamic (CFD) approaches for studying the influence for a slow-speed ship [12]. The total resistance of a vessel can
of hydrodynamic forces on ships are increasingly used in the be calculated by Eq. (1) [2].
marine field. CFD application is an easy and less time consum- RT ¼ RF þ RR Or RT ¼ RV þ RW ð1Þ
ing application [7]. The accuracy of using CFD simulation is
proven to be high and most naval architects use this method where RT, RF, RR, RV and RW are the total resistance, friction
instead of towing tank experiment method which is tedious resistance, residual resistance, viscous and wave making resis-
and time consuming [2]. CFD application has advanced in tance respectively. All above components of resistance are cal-
recent years and become one of the most important methods culated using the generic form [2]:
used in ship building industries [8]. CFD methods can analyze 1
flow problems in resistance estimation. CFD techniques give R ¼ qCAV2 ð2Þ
2
practical results with less effort in cost and time. Viscous flow
gives more accurate results of drag than potential flow [9,10]. where
Also, empirical or statistical methods are considered suited.
The most prominent of these is the approximate resistance and C – The resistance coefficient
power prediction method by Holtrop and Mennen together q – The density of the medium
with its revision [11]. Although this methodology provides suf- A – Wetted Surface area
ficient accuracy, the statistical sample of the hull forms on V – The vessel speed
which it is based dates back to the 1970s and 1980s. Such hulls,
although roughly similar, have some distinct deviations from For the current ship model, the total resistance in deep
modern commercial vessels. The Holtrop and Mennen method water can be calculated using Holtrop-Mennen method. The
is currently considered as one of the most accurate and effi- Holtrop method [11] is currently considered as one of the effi-
cient methods for the estimation of the resistance and propul- cient methods for the estimation of the resistance and propul-
sion power requirements of conventional monohull vessels at sion power requirements of conventional vessels. It is an
the initial stages of design. empirical method consisting of equations for the various resis-
In this paper, example of CFD simulation of ship resistance tance components that derive from the statistical analysis and
components is presented for Container ship model on calm regression of a database with a large number of model test
water. The software used for computations was ANSYS- results. The model developed by Holtrop is a numerical
Numerical study on the hydrodynamic drag force 851

description of the ship’s resistance, subdivided into compo- equations of mass, momentum can be written as follows in
nents of different origin. Each component was expressed as a Eqs. (8) and (9).
function of the speed and hull form parameters. The applica- @q
tion of Holltop’s method for the deep water resistance consist þ r  ðqUÞ ¼ 0 ð8Þ
@t
many ship types, one of them is specified for our case as dis-
cussed in [11]. The Holtrop-Mennen Method has acquired @ðqUÞ
widespread recognition. Holtrop and Mennen tried to include þ r  ðqU  UÞ ¼ rp þ r  s þ SM ð9Þ
@t
physical aspects in their formulas, but used the experimental
data for determining the coefficients. A summary of their where the stress tensor, s is related to the strain rate as
method which is based on test results from 334 models of tan- follow:
kers, cargo ships, trawlers, ferries, etc., is given below. 2
s ¼ lðrU þ ðrUÞT  dr  UÞ ð10Þ
The resistance is split into viscous and wave resistance. For 3
the viscous resistance, the standard formula is used as pre-
The Third governing equation of CFD is the total energy
sented in Eq. (3).
equation which can be presented in the following form.
CV ¼ ð1 þ kÞCF ð3Þ @ðqhhot Þ @q
 þ r  ðqUhhot Þ ¼ rðkrTÞ þ rðU  sÞ þ U  SM þ SE
where CF is obtained from the ITTC-57 formula. The form @t @t
factor k is determined from a formula obtained statistically ð11Þ
as follow. where htot is the total enthalby, related to the static enthalpy h
 
B T L L3 (T, p) by:
k¼f ; ; ; ; CP ; c ð4Þ
L L LR r 1
htot ¼ h þ U2 ð12Þ
where c is a coefficient dependent on the shape of the after 2
body and LR is the length of the after body. If LR or S are The term rðU  sÞ represents the work due to viscous stres-
unknown, they may be obtained from other statistically ses and is called the viscous work term. The term U  SM rep-
derived formulas. The appendage resistance is considered as resents the work due to external momentum sources and is
a correction to the form factor. For the wave resistance, Hol- currently neglected.
trop and Mennen use a theoretical expression attributed to Generally, the Navier-stokes equations describe both lami-
Havelock (1913), obtained by replacing the hull by two pres- nar and turbulent flows without the need for additional infor-
sure disturbances separated by the wave making length of mation. However, turbulent flows at realistic Reynolds
the hull. The original expression is elaborated in the following numbers span a large range of turbulent length and time scales.
equation. In general, turbulence models seek to modify the original
RW unsteady Navier-stokes equations by the introduction of aver-
¼ C1 C2 C3 emFn þ m2 cosðkFn2 Þ
d
ð5Þ aged and fluctuating quantities to produce the Reynolds Aver-
W
aged Navier-Stokes (RANS) equations. Turbulence models
where W is the weight of the ship and C1, C2, C3, m, and m2 are based on RANS equations are known as statistical Turbulence
coefficients, which are functions of the form parameters of the models due to the statistical averaging procedure employed to
hull. Different coefficients are used for Fn = 0.40 and obtain the equations [14]. The Reynolds averaged equations
Fn = 0.55. In the intermediate range, the residuary resistance given below in Eqs. (13) and (14).
is obtained by an interpolation formula between the two limits.
Holtrop and Mennen also suggest a formula for the roughness @q @  
þ qUj ¼ 0 ð13Þ
allowanceDCF and compute the total resistance as expressed in @t @xj
Eq. (6).
@ðqUÞ @   @p @
1 RW þ qUi Uj ¼  þ ðsij  qui uj Þ þ SM ð14Þ
RT ¼ qV2 S½CF ð1 þ kÞ þ DCF  þ W ð6Þ @t @xj @xj @xj
2 W
where s is the molecular stress tensor (including both normal
After determining the resistances then the corresponding and shear components of the stress).
effective power (PE) which help vessels moving through in The need of turbulence models for RANS simulation is to
water with a determined speed can be calculated by Eq. (7). determine the Reynolds stresses. This process can be con-
PE ¼ RT  V ð7Þ ducted by three categories of RANS turbulence.

 Linear eddy viscosity


2.2. Computational fluid dynamics theory
This is a turbulence models which using Reynolds stress
The proposed CFD model was developed based on the that obtained from the Reynolds averaging from Navier stokes
Reynolds-averaged Navier-Stokes (RANS) method for three equations. This involves the relation between Reynolds stresses
dimensional unsteady viscous incompressible flow using and mean strain. This models has disadvantages in flow situa-
ANSYS-CFX software package. The averaged continuity tions, such as, over predict turbulence energy levels in stagnant
and momentum equations for incompressible flows may be regions, Misinterpretation of normal stresses and does not
given as in the following two equations [13]. The instantaneous reproduce the asymmetry in the velocity profile
852 A.G. Elkafas et al.

 Nonlinear eddy viscosity models


Table 1 Vessel and model particulars.
This turbulence model is more accurate than linear eddy Symbol Full-scale Scaled vessel Unit
viscosity model. It is due to the fact the turbulence is a highly vessel model (1:100)
nonlinear phenomena Length between 247 2.47 m
perpendiculars (LPP)
 Reynolds stress model (RSM) Breadth (B) 32 0.32 m
Draught (D) 12 0.12 m
This turbulence model which is also referred as the second Displacement 64,000 ton 63.4 kg –
Metacentric height (GM) – 8.75 mm
moment closure model is most complete turbulence model. In
Vertical centre of – 153 mm
this turbulence model, the eddy viscosity is removed and Rey- gravity (KG)
nolds stress are directly computed. LCG length (from aft – 1160 mm
Referring to ANSYS-CFX, there are three type of turbu- perpendicular)
lence model, which are k-epsilon, k-omega, and shear stress
transport (SST) [15].

K-e Model Module of ANSYS Workbench version 19. Fig. 1 shows the
bare hull of model built in Rhinoceros 5.
The k-e model consists of k, kinetic energy and e, balance of
dissipation along with the complete RANS equation. The k-e 3.2. Fluid domain geometry
model works well away from the wall around the boundary
layer edge and for fully turbulent flows in the high Reynolds
The boundary domain is the area where flows (Air and Water)
regime.
will influence the hull of the vessel. The boundaries of the fluid
domain are designed to be placed with a sufficient distance
K-x Model
from the area of investigation. This is to ensure the accuracy
of the solution. The fluid domain for container ship model’s
The k-x model consists of k, kinetic energy and x, specific
investigation was built based on the International Towing
dissipation turbulent frequency. Works well within the low
Tank Conference (ITTC)’s recommendation in order to pre-
Reynolds regime and should have been applied for the
vent flow reflections [18]. The inlet and the exterior boundary
transitional Reynolds for this simulation. The model predicts
such as top, bottom and side of wall whereby the flow is undis-
separation early and requires a mesh inflation layer near the
turbed usually occurred, they are required to be located
wall.
around one or two total length of the object of investigation
which in this case is the length between perpendiculars. As
SST Model
the outlet of the fluid domain is normally where the fluid is
unsteady, it is required to be placed around three to five
The shear stress transport SST model, which is applied in
perpendicular length away from the ship to prevent the
this simulation, gives high accuracy modelling of the boundary
interference or reflection of the flow, where half of the body
layer and is a combination of both k-e and k-x. The SST
is modelled to decrease the computational domain size and
model gives accurate predictions of the onset and the amount
time. The ship axis is located along the x-axis with the bow
of flow separation under opposing pressure gradients, where
located at x = LBP and the stern at x = 0. The still water level
turbulence is present. By applying both previous models, it
lies at z = 0. The dimensions of the computational domain
covers both regions of the boundary layer, close to the wall
satisfy the well-known ITTC procedure. Detailed information
and far away from the wall close to the boundary layer limit
about the principles of computational domain dimensions’
and applies the Bradshaw relation for good separation predic-
selection strategy can be found in [19] and [20]. Also a detailed
tion [16].
description of the boundary conditions is given in [21]. The
The shear stress transport (SST) model was used for this
general view of the computational domain and the boundary
simulation to gain the longitudinal forces, resistance forces,
conditions are shown in Fig. 2. Table 2 shows the selected
acting on a container ship model, as it gives the best results
dimension for the fluid domain which represented in
for maritime engineering applications.
ANSYS-CFX.

3. Methodology and setup 3.3. Mesh

3.1. Model details


The meshing of the fluid domain and the model were con-
ducted using ANSYS Meshing 19.0 with the CFD as the phy-
The case study for this paper is selected to be a Container ship sic preference and the CFX as the solver. The size function for
model. The vessel is a twin-propeller container ship [17]. A the mesh is set as curvature, the mesh will not change the shape
1:100 scale model of this vessel was built in software. The gen- of hull or resemblance the tendency of hull’s shape. While the
eral parameters of the model are listed in Table 1. minimum size of the mesh is set to be 1 mm such that the mesh
The 3D model of the vessel is processed using Rhinoceros is allowed to capture the curvature of the model [22]. Table 3
5.0 and MAXSURF Modeler and Imported to Geometry presents the finalized mesh criteria.
Numerical study on the hydrodynamic drag force 853

Fig. 1 Geometry of container ship model.

Fig. 2 Boundary Domain Conditions.

Table 2 Size of the fluid domain geometry.


Fluid domain entities B.C ITTC Selected Distance (m)
Rectangular Inlet 1–2 LPP 1 2.5
Outlet 3–5 LPP 3 7.5
Top & Bottom wall 1–2 LPP 1 2.5
Side wall 1–2 LPP 1 2.5

In order to ensure that the mesh represents the actual body


Table 3 Mesh size and specification.
for analysis, refinement were made on the body of the fluid
Criteria Value domain, face of Hull. The refinements were as below:
Nodes 239,385
Body sizing (m) 0.3 3.3.1. Body sizing
Face sizing (m) 0.008 The body sizing is applied with unstructured tetrahedral ele-
Inflation layer 5 ment size of 0.3m. The aim of body sizing applied is such that
Growth rate 1.2
the mesh size for both side of interface for the Box are same.
Maximum thickness boundary layer (m) 0.05
This action allows the optimal interpolation between fluid
domains [14].

Fig. 3 Meshing of hull and fluid domain.


854 A.G. Elkafas et al.

3.3.1.1. Face sizing. Further refinement is conducted on the  The analysis is conducted based on steady state approached
face of the Hull. The element size applied on the face of the as RANS based CFD stimulation method is used.
Hull was 0.008 m. The refinement is conducted so that the flow
behaviour around the Hull can be modelled with a higher mesh
resolution which hence increase the accuracy of the results. 3.4.2. Domain characteristic
Fig. 3 below shows Meshing and inflation layer around hull. Table 4 presents the domain characteristics that being applied
to the flow.
3.4. CFX pre solver
3.4.3. Boundary condition
The completed mesh in the previous module is then imported Table 5 presents the type of boundary conditions that being
to the CFX Pre Solver of ANSYS CFX to define its analysis applied to the fluid domains and Fig. 4 shows the boundary
method, fluid flow characteristic and its boundary conditions. condition.
The details of the setup are as below:
3.4.4. Output control
3.4.1. Analysis setting
The main interested output of the analysis is the hydrodynamic
performance of the Hull in term of drag force. In order to
 The turbulence model that is selected for the analysis is monitor the progress of the CFX-Solver, the hydrodynamic
Shear Stress Transport (SST) Model. characteristic is expressed in the following expression [for-
ce_x()@Hull].

3.4.5. Solver control


Table 4 Domain characteristics.
Solver control is used to increase the efficiency of the computer
Criteria Value
resources and control the quality of the CFX solution; ensur-
Fluid type Water
Fluid density (kg/m3) 997
Temperature (°C) 25 Table 6 Solver control inputs.
Kinematic viscosity (m2/s) 8.9E7
Convergence control
Morphology Continuous fluid
Minimum iteration 200
Buoyancy model Buoyant
Maximum iteration 1200
Gravity Z component (m/s2) 9.81
Timescale control Auto
Domain motion Stationary
Reference pressure (atm) 1 Convergence criteria
Turbulence model SST Residual type RMS
Turbulent wall functions Standard Residual target 0.00001

Table 5 Boundary condition setting specification.


Boundary type Settings
Inlet Inlet Velocity Range = 1–2 [m/s]
Turbulence model: SST model
Outlet Opening Mass and momentum: Entrainment Static Pressure: 0 [Pa]
Top, Bottom and Side wall Walls Mass and momentum: Free slip wall
Symmetry Symmetry –
Hull Wall No Slip wall

Fig. 4 Boundary Condition on the Fluid Domain.


Numerical study on the hydrodynamic drag force 855

Fig. 5 Graph showing the convergence of the solution.

ing the result will converged to ensure the accuracy of the ing. As the residuals decrease further, the monitor values
results. The solver input and the convergence criteria were change less and less between iterations. Once the monitor point
defined in Table 6. values have ‘‘flattened out”, so the solution is assumed to be
converged.
3.5. CFX solver
4. Independent study and results
The defined CFX-Pre files are imported into the CFX-Solver
for analysis. The run mode of the CFX solver was set to be 4.1. Grid independence study
the Intel MPI Local Parallel with 2 partitions core and the
computations are made on 4 CPU with 2.50 GHz, on windows In order to reach the results with high accuracy in CFD simu-
Win10 system. Explanation of the numerical method can be lation, the mesh must divide elements as much as possible.
found in [23]. During the simulation progress, inspections on However, the number of nodes of the simulated model directly
the solution convergence is conducted and the results show a depends on three main factors (Technology, Time and Cost)
converge trend. Figs. 5 and 6 show CFX Solver when simulate [24]. In this paper, Richardson’s extrapolation method for grid
the performance and reach the balance condition at results. convergence could be a proper choice for estimation of the
The residual monitors in Fig. 5 demonstrate monotonic con- mesh error. To have a clear view of the method, the following
vergence, indicating a well-posed problem and a tightly con- section have an illustration about calculation of the drag coef-
verged solution. Fig. 6 shows the change in the monitor ficient, which is based on the Richardson’s extrapolation
point(drag) values vs. iteration number. After approximately method [25] as shown in Fig. 7.
100 iterations, the drag monitor point is within just a few per- The grid convergence study was conducted based on the
cent of its final value. However, the drag value is still far from ITTC, uncertainty analysis recommendation [4]. The conver-
its final value, so stopping the analysis here could be mislead- gence study was made based on three varying mesh resolution
856 A.G. Elkafas et al.

Fig. 6 Change of monitor point (drag) values vs. iteration number.

Fig. 7 Richardson Extrapolation method.


Numerical study on the hydrodynamic drag force 857

which were categorized into coarse, medium and fine mesh.


Table 7 The coarse, medium and fine mesh details.
The mesh were varied by the modification of the face sizing
Detail Coarse Medium Fine while keeping the body sizing with a constant element size.
Body sizing (m) 0.3 0.3 0.3 The inflation layer was kept constant throughout the analysis
Face sizing (m) 0.008 0.006 0.0046 as the mesh resolution was based on the standard wall
Number of nodes 248,863 344,331 503,421 calculation. It should be noted that there is a constant
Number of elements 781,248 1,045,750 1,496,670 refinement ratio exists between the nodes count on every mesh
Drag force (N) 5.13448 3 2.334 that being conducted. Table 7 shows the details of the meshes
Drag coefficient 0.010107517 0.005905671 0.004594612 used for the convergence studies and drag force calculated at
each case at 1 m/s.
Based on formulas at the used equation section, outcomes
have been calculated and presented at Table 8.
Figs. 8 and 9 show the Richardson Extrapolation conver-
Table 8 Results of meshing error estimation.
gence studies and the relative error estimation that conducted
Outcome Equation Value on the Hull to determine the mesh performance. It is expected
Change in solution (e32) e32 = C(h3)  C(h2) 0.004201 that the simulations which have the higher number of nodes
Changes in solution (e21) e21 = C(h2)  C(h1) 0.001311 will be more accurate. The relative error estimation study is
Convergence ratio Ri = e21/e32 0.312019 conducted on the convergence result to determine is the error
Order of convergence q = ln(e32/e21)/ln 3.461470 between the results vary with number of nodes.
(ri) The drag coefficient was converged based on the graphs
Error of the finest grid d = e21/rpi  1 0.000594 which prove that the mesh is converged with different fineness
Richardson extrapolated CR = C1 – d 0.004000
of mesh. However, the fine mesh is still chosen for analysis as it
solution
Relative error estimate (er) er = Ci  CR/CR
provides a higher accuracy to the stimulation which hence min-
imizes the error of investigation.

4.2. Resistance results

0.012
After installation of simulation for the hull and boundary
Drag Coefficient domain in CFX-Pre, then simulations will run and export out-
0.01
CR
put data under tested conditions. It is the fact that this simu-
0.008
lation only tests a half of the ship hull. Therefore, in order
Drag Coefficient

to obtain the full resistance of the whole hull, the generated


0.006 result are doubled. Simulations presented in this study were
performed for 8 selected test conditions for different Froude
0.004 Number (Fn) and also different ship speeds as presented in
Table 9 which show the selected test conditions. The resistance
0.002 components resulted from CFD simulation are divided to skin
friction resistance and pressure resistance component. The
0 simulation results show that the skin friction resistance
200 250 300 350 400 450 500 550
account up to 90% of the total resistance as shown in Fig. 10.
Number of Nodes(×1000)
In order to examine the accuracy of resistance results of the
Fig. 8 Number of nodes against Drag Coefficient. simulated ship model by using ANSYS-CFX solver, the simu-
lation results for the tested Froude numbers are thoroughly
validated by comparing directly with the resistance results
from the Holtrop prediction method [11] in MAXSURF soft-
ware. The comparison determines the accuracy of the pro-
10
gramming and the computational implementation of the
conceptual model in which it examines the mathematical
errors. CFD simulation and Holtrop method are investigated
similar test condition and Froude numbers to calculate the dif-
Relave Error

ference between both resistance components results.


1 Fig. 11 shows the comparison of skin friction resistance
100000 1000000 results from CFD simulations to Holtrop skin friction resis-
tance results. The comparison between two results is very good
for the whole simulated conditions. Within the simulated range
of Froude Number, the maximum error is 5.15% at Fn = 0.4.
The model results from CFD simulation indicate that the com-
0.1
Number of Nodes putations predicted 3.125% lower skin friction resistance at
Fn = 0.24 and 0.28, 4.2% lower skin friction resistance at
Fig. 9 Drag’s Relative Error Estimation against the number of Fn = 0.35 compared to the results from Holtrop. Generally,
nodes. it is shown that the CFD simulation results agree fairly well
858 A.G. Elkafas et al.

Table 9 The selected test conditions (Speed-Froude Number).


Speed (m/s) 1 1.2 1.4 1.6 1.7 1.8 1.9 2
Fn ¼ pVffiffiffiffiffiffi 0.20315 0.24378 0.28441 0.32504 0.345355 0.36567 0.385985 0.4063
gL

18 with the results computed from Holtrop method, and that


Total Resistance ANSYS-CFX code can predict ship resistance.
16
Fig. 12 shows the comparison of pressure resistance
Skin Fricon Resistance
14 results from CFD simulations to Holtrop form component
Pressure Resistance results. Within the simulated range of Froude Number, the
12
Resistance (N)

model results from CFD simulation indicate that the computa-


10 tions predicted greater results than those from Holtrop
8
method.

6
5. Conclusion and recommendation
4

2 In this paper, a Reynolds averaged Navier Stokes (RANS)


method is presented to predict the resistance components
0
0.15 0.2 0.25 0.3 0.35 0.4 0.45
results for a container ship model. The software used for com-
Froude Number (Fn) putations was ANSYS-CFX package which has implemented a
RANS solver model. The turbulence model that is selected for
Fig. 10 Resistance components values calculated by CFD the analysis is Shear Stress Transport (SST) Model which gives
simulation Versus Froude Number. high accuracy modelling of the boundary layer and is a combi-
nation of both k-e and k-x. To predict the resistance of ship
with different running attitudes conveniently, a plenty of
16 numerical simulations of ship advancing at different speeds
CFD Simulaon
14 and Froude numbers are carried out. After installation of sim-
Skin Fricon Resistance (N)

Holtrop Method ulation for the hull and boundary domain in CFX-Pre, then
12
simulations will run and export output data under tested con-
10 ditions. The resistance components resulted from CFD simula-
tion are divided to skin friction resistance and pressure
8
resistance component. The simulation results show that the
6 skin friction resistance account up to 90% of the total
4 resistance.
Accuracy of the computations is evaluated by comparing
2
with the results from Holtrop Method in MAXSURF soft-
0 ware. First, computed skin friction resistance results from
0.15 0.2 0.25 0.3 0.35 0.4 0.45 CFD simulation are compared with Holtrop skin friction resis-
Froude Number(Fn) tance results. The comparison between two Results is very
good for the whole simulated conditions. Within the simulated
Fig. 11 Comparison of Skin friction resistance between CFD
range of Froude Number, the maximum error is 5.15% at
simulation and Holtrop Method.
Fn = 0.4. Then, the comparison of pressure resistance results
from CFD simulations to Holtrop form component results is
1.6
presented. Within the simulated range of Froude Number,
1.4 the model results from CFD simulation indicate that the com-
CFD pressure component putations predicted greater than results from Holtrop method.
1.2
Generally, it is shown that the degree of agreement with Hol-
Resistance(N)

1 Holtrop Form component trop prediction method is satisfactory for resistance compo-
0.8 nents. The present simulation seems to indicate the
0.6 availability of ANSYS-CFX as a tool for the prediction of ship
resistance and its application to the development of ship
0.4
design.
0.2 A grid convergence study has been conducted to determine
0 the accuracy of the mesh and hence increase the accuracy
0.15 0.2 0.25 0.3 0.35 0.4 0.45 of the predicted result. The grid convergence study done and
Froude Number (Fn) the analysis conducted with Low Reynolds Wall Treatment,
the stimulation result seem to be acceptable and further
Fig. 12 Comparison between CFD simulation’s pressure com- investigation can be conducted.
ponent and Holtrop’s form component.
Numerical study on the hydrodynamic drag force 859

In general, the CFD simulations by computer software, based panel method, Ocean Eng. 34 (2007) 1892–1900, https://
which would give us the results with high accuracy, time- doi.org/10.1016/j.oceaneng.2006.06.010.
saving and presented in visual images must be highly used. [10] K.A. Belibassakis, T.P. Gerostathis, K.V. Kostas, C.G. Politis,
So that more and more CFD simulations would be applied P.D. Kaklis, A.I. Ginnis, C. Feurer, A BEM-isogeometric
method for the ship wave-resistance problem, Ocean Eng. 60
on the whole fields simply because of its efficiency and
(2013) 53–67, https://doi.org/10.1016/J.
reliability. OCEANENG.2012.12.030.
The recommendations for a better future of this work can [11] J. Holtrop, A statistical re-analysis of resistance and probulsion
be concluded, the verification of the accuracy of CFD simula- data (1984) 272–276.
tion results based on the results from Holtrop method. It is not [12] H. Lackenby, The thirty-fourth Thomas Lowe gray lecture:
really objectivity when evaluates the results of a software by resistance of ships, with special reference to skin friction and hull
another one’s. However, due to limitations of research which surface condition, Proc. Inst. Mech. Eng. 176 (1962) 981–1014,
have led to not carry out simulations by the experimental https://doi.org/10.1243/PIME_PROC_1962_176_077_02.
model in towing tank in order to get data for validation pur- [13] J.H. Ferziger, M. Peric, Computational methods for fluid
poses. Therefore, one of the future recommendations, the dynamics/J.H. Ferziger, M. Peric., 2002. doi: 10.1007/978-3-
642-56026-2.
Results of CFD simulation must be compared with Experi-
[14] J.D. Anderson, Computational Fluid Dynamics the Basics With
mental data generated from Towing tank or using another Applications, first ed., McGraw-Hill Education, 1995.
software in order to be more validated. <https://soaneemrana.org/onewebmedia/
For further investigation, it is necessary to solve the prob- COMPUTATIONAL FLUID DYNAMICS BY JHON D
lem of nodes number limitation. The higher use of nodes num- ANDERSON2 JR.pdf>.
ber will increase the accuracy of result even though the [15] F.R. Menter, Two-equation eddy-viscosity turbulence models
simulation time will be longer. for engineering applications, AIAA J. 32 (1994) 1598–1605,
https://doi.org/10.2514/3.12149.
[16] F. Menter, M. Kuntz, R.B. Langtry, Ten years of industrial
References
experience with the SST turbulence model, Heat Mass Transf. 4
(2003).
[1] S. Sherbaz, W. Duan, Ship trim optimization: assessment of
[17] Y. Wang, S. Chai, H.D. Nguyen, Modelling of a surface vessel
influence of trim on resistance of MOERI container ship, Sci.
from free running test using low cost sensors, in: 2017 3rd Int.
World J. 2014 (2014) 1–6, https://doi.org/10.1155/2014/603695.
Conf. Control. Autom. Robot., 2017, pp. 299–303. doi: 10.1109/
[2] A.F. Molland, S. Turnock, D. Hudson, in: Ship Resistance and
ICCAR.2017.7942707.
Propulsion. Practical Estimation of Ship Propulsive Power,
[18] International Towing Tank Conference (ITTC), Practical
Cambridge University Press, New York, USA, 2011, https://doi.
guidelines for ship CFD applications, in: Proc. 26th ITTC,
org/10.1017/CBO9780511974113.
2011. <http://ittc.sname.org/CD 2011/pdf Procedures 2011/7.5-
[3] L. Larsson, H.C. Raven, J.R. Paulling, Ship Resistance and
03-02-03.pdf>.
Flow, Society of Naval Architects and Marine Engineers, Jersey
[19] T. Tezdogan, Y.K. Demirel, P. Kellett, M. Khorasanchi, A.
City, N.J., 2010, http://app.knovel.com/hotlink/toc/id:
Incecik, O. Turan, Full-scale unsteady RANS CFD simulations
kpPNASSRF2/principles-of-naval.
of ship behaviour and performance in head seas due to slow
[4] H. Jasak, V. Vukčević, I. Gatin, I. Lalović, CFD validation and
steaming, Ocean Eng. 97 (2015) 186–206, https://doi.org/
grid sensitivity studies of full scale ship self propulsion, Int. J.
10.1016/j.oceaneng.2015.01.011.
Nav. Archit. Ocean Eng. 11 (2019) 33–43, https://doi.org/
[20] Y. Hakan Ozdemir, A. Dogrul, B. Barlas, T. Cosgun, A
10.1016/j.ijnaoe.2017.12.004.
numerical application to predict the resistance and wave
[5] H. Schneekluth, V. Bertram, Ship propulsion, in: Sh. Des. Effic.
pattern of kriso container ship, Brodogradnja 67 (2016) 47–65.
Econ., second ed., Butterworth-Heinemann, Oxford, 1998, pp.
doi: 10.21278/brod67204.
180–205. doi: 10.1016/B978-075064133-3/50006-2.
[21] Y. Ozdemir, B. Barlas, T. Yilmaz, S. Bayraktar, Numerical and
[6] International Towing Tank Conference (ITTC), The specialist
experimental study of turbulent free surface flow for a fast ship
committee on computational fluid dynamics-Final report and
model, Brodogradnja 65 (2014) 39–54.
recommendations to the 26th ITTC, in: Proc. 26th ITTC, 2011,
[22] Henk Kaarle Versteeg, W. Malalasekera, An introduction to
pp. 337–377.
computational fluid dynamics – the finite volume method,
[7] A.G. Elkafas, N.R. Ammar, M.M. Elgohary, A. Zeid,
second, Pearson Education Limited, 1995. doi: 10.1017/
Prediction of shallow water resistance for a new ship model
cbo9780511761348.011.
using CFD simulation: case study container barge, J. Sh. Prod.
[23] ANSYS, CFX-solver theory GUIDE, (n.d.). <https://www.
Des. 35 (2) (2019) 198–206, https://doi.org/10.5957/
sharcnet.ca/Software/Ansys/17.0/en-us/help/cfx_thry/cfx_thry.
jspd.11170051.
html>.
[8] H. Abdelkhalek, D.F. Han, L.T. Gao, Q. Wang, Numerical
[24] J. Tu, G. Yeoh, C. Liu, in: Computational Fluid Dynamics: A
estimation of ship resistance using CFD with different turblence
Practical Approach, third ed., Butterworth-Heinemann, 2019,
model, in: Adv. Res. Struct. Mater. Eng. Inf. Technol. III, Trans
https://doi.org/10.1016/C2015-0-06135-4.
Tech Publications Ltd, 2014, pp. 209–213. doi: 10.4028/
[25] D. ZINGG, Viscous airfoil computations using Richardson
www.scientific.net/AMR.1021.209.
extrapolation, in: 10th Comput. Fluid Dyn. Conf., American
[9] M.S. Tarafder, K. Suzuki, Computation of wave-making
Institute of Aeronautics and Astronautics, 1991. doi: 10.2514/
resistance of a catamaran in deep water using a potential-
6.1991-1559.

You might also like