You are on page 1of 126

V

Hydrodynamics

Turbulent Flow Computations For Slender Ships with Free Surface


Yavuz Hakan zdemir*, Bar Barlas**, Seyfettin Bayraktar*
*

**

Dept of Naval Architecture and Marine Engineering, Yildiz Technical University, Istanbul, Turkey, Istanbul

Technical University, Turkey, yozdemir@yildiz.edu.tr, barlas@itu.edu.tr, sbay@yildiz.edu.tr

Abstract
In this study, the unsteady turbulent flow around slender ship hulls is investigated. The components of
the flow are computed by solving the RANS (Reynolds Averaged Navier-Stokes) equations. In the
turbulent flow calculations, the Reynolds stresses in the momentum equations are modelled in
accordance with the Boussinesq's hypothesis of turbulence viscosity. The relationship between the
turbulence viscosity and the velocities are obtained through the SST (shear stress transport) k-
turbulence model. In the numerical scheme, the velocities, pressure and the turbulent viscosity values
are calculated by using the hexagonal FVM (finite volume method) discretization of the spatial
domain. The validation studies are carried out by comparing the results obtained for a slender hull
form, the Wigley hull with corresponding experimental results available from literature. The resistance
and wave elevation are compared with available experimental results. The comparisons are found to be
satisfactory for turbulent flow regimes.

Keywords:RANSE, Turbulent flows, slender ships, SST k- turbulence model.


1. Introduction
For more than a century, scientists carried out many studies to model the flow around ship hulls. Most
of the research was deal with the inviscid flows. This is due to the fact that the Navier-Stokes
equations are complex, compared with the potential theory used for inviscid flows. Because of the
compelxity of the problem both experimental and numerical techniques have been the only practical
way to obtain information on flow around ship hulls. With the recent developments in computers, both
in hardware and software, it has become possible to obtain meaningful numerical results on flow
around ship hulls.
From the open literature it can be reported that ship hydrodynamics computations based on NavierStokes (N-S) solvers were initiated in the 1980s, and since then a number of useful codes have been
developed (Sato et al., 1999). In the recent years, computational fluid dynamics (CFD) techniques
have been incorporated into optimization procedures for hull configuration. In this way CFD
simulation plays an important role in ship design, performance analysis and form optimization, etc. A
flow-simulation method was developed to predict the performance of a sailing boat in unsteady motion
on a free surface by Akimoto and Miyata (2002). The sailing conditions of the boat are virtually
279

realized by combining the simulations of water-flow and the motion of the boat. Skytt (2004) has
given the detailed information about how to get a computer-aided design (CAD) to CFD meshes for
ship geometries. Gorski (2002) and Parolini and Quarteroni (2004) reported recent innovative aspects
of the numerical models used in CFD studies. Successful design and optimization of marine structures
requires information about the flow problem on both integral and field quantity levels.
In this work, the finite volume solution method for the RANS equations is applied to the simulation of
flow field around a Wigley hull. The turbulence model used is the well known SST (shear stress
transport) k- two equation turbulence model. The numerical algorithm is divided into three stages: in
the first stage the velocity components, in the second stage the pressure and in the third stage the
turbulence quantities are calculated. The main goal of this numerical study is to show the capability a
general-purpose CFD code of Star CCM+ for design, analysis and feasibility of such a simulation for
shipping industry.

2. Mathematical Formulation
The governing equations are the RANS equations and the continuity equation for mean velocity of the
unsteady, three-dimensional incompressible flow. The continuity equation and momentum equations
in Cartesian coordinates can be given as;
Ui

(1)

xi

for the continuity,

P
xi

xj

Uj

Ui
xj

xi

ui u j

(2)

for the momentum equations.


The well known SST (shear stress transport) k- model has been used to simulate the turbulent flows.
The turbulence kinetic energy, k, and the specific dissipation rate, , are obtained from the following
transport equations:

xi

kui

xj

k
xj

Gk Yk

Sk

(3)

and

xi

ui

xj

xj

280

(4)

G represents the generation of

and

represent the effective diffusivity of k and

respectively, Yk and Y represent the dissipation of k and

due to turbulence. All of the above

terms are calculated as described below. Sk and S are user-defined source terms.
The effective diffusivities for the k

model are given by

(5)

where

and

viscosity,

are the turbulent Prandtl numbers for

, is computed by combining

and

and

, respectively. The turbulent

as follows:

(6)

The use of a k- formulation in the inner parts of the boundary layer makes the model directly usable
all the way down to the wall through the viscous sublayer hence the SST k- model can be used as a
Low-Re turbulence model without any extra damping functions. The SST formulation also switches to
a k- behaviour in the free-stream and thereby avoids the common k- problem that the model is too
sensitive to the inlet free-stream turbulence properties. The SST k- model has good behaviour in
adverse pressure gradients and separating flows like ship boundary layers.
Calculations are made in hexahedral computational domain for the hull of the Wigley hull symmetric
to its centerline. The general view of the computational domain with Wigley hull and the notations of
boundary conditions is depicted in Fig. 1.

Fig. 1. The general view of the domain and boundary conditions.

2.1 Boundary Conditions


To perform turbulent flow computations with the use of a k- formulation, at least one grid point is
required very close to the wall, especially to resolve the viscous sublayer. At the upstream boundary,

281

the uniform flow condition is used, i.e. U=1, V=W=0, P is hydrostatic.At the downstream boundary,
zero derivative condition in x- direction is used, for P, pressure is hydrostatic. At the symmetry plane
boundaries zero derivative condition in the normal directions are utilized.

3. Numerical Method
The governing equations described above are discretized using a node based finite volume method, the
advection terms are discretized using a first-order upwind interpolation scheme. The governing
equations are solved successively. Because the equations are nonlinear, several iterations must be
performed before a converged solution is obtained. First, the initial values of U, V, W, I, t/ , and P
are defined. Then U, V, and W are solved from the momentum equations. The pressure field is solved
by using the well known SIMPLE algorithm . The turbulence quantities are then solved for k and

Since the computations involve certain approximations, an iterative procedure is needed. The solution
is considered converged when the normalized residuals of all the variables is lower than 10 -5. The
computations is made on a 4 CPU workstation with 3.4GHz, on windows XP system.
The Courant-Friedrichs-Lewy (CFL) number in the main flow direction, i.e. U t/ x should be less
than 1 for better results. In our computations the CFL number for each node on the Wigley hull is in
the range between 0.3 and 0.6. The hexahedral finite volume discretization of the spatial domain (Fig.
2) is performed with GAMBIT preprocessor package which allows the user to generate the
computational grids. Flow simulations are made by using Star CCM+ code.

Fig. 2. A sketch of the hexahedral finite volume discretization of the spatial domain.

4. Numerical Results And Discussion


Wigley hull is a mathematically definable ship hull form which is frequently used as a benchmark for
flow computations around ships. As the main purpose of this study is to show the capability of the
general-purpose CFD solver of Star CCM+ code to use at shipping industry, the CFD calculations are
282

compared with experimental data of Sarda (1986) The mathematical formulation of Wigley hull is
given below.
(7)
The hull length is L=6.096m, the breadth is B=L/5, and the draft is T=L/16. The outer boundary is
taken at 1.2xL from the centerline. The upstream starts 2xL from the begining, and the downstream
ends 3xL from the aft end of Wigley hull. A grid improvement study was performed using three
geometrically similar grids. The unsteady flow around Wigley hull is computed at Fn=0.20 using a
coarse grid of 145x32x25 grid system, a medium grid of 150x50x30 grid system, and a fine grid of
180x70x40 grid system (see Table 1). Because of the SST k- turbulence model, some results are
obtained after five days of computer running. The geometry together with a visualization of the free
surface is given in Fig. 3.

Fig. 3. Wigley hull and free surface visualization (Fn=0.20).

The upstream flow velocity is taken as 1.546 m/s which gives a Fn based on ship length of 0.2 and a
Rn of 9.062x106. The time step t is chosen to be 0.15 s.
Table 1: Grid resolution and y+ values
Coarse grid

Medium grid

Fine grid

Number of Nodes

116,000

225,000

504,000

Average y+

26

18

11

Fig. 4 presents the wave elevation contours. Fig. 5 illustrate the computed wave profile along the ship
hull. The comparison with the experimental measures is very good. The results indicate that grid
dependence at the aft end of the ship is small, hence the computed wave profile at the bow of the ship
is less than expected with the coarse grid used. The computed drag for three grid cases is compared in
Table 2.

283

Fig. 4. Predicted wave elevation contours (Fn=0.20).

0.010

Wave elevation y/L

0.008

Fine grid (500000)


Medium grid (225000)
Coarse grid (116000)
Experiment

0.006
0.004
0.002
0.000
-0.002

0.0

0.1

0.2

0.3

0.4

-0.004

0.5

0.6

0.7

0.8

0.9

1.0

x/L

Fig. 5. Predicted wave profile along the hull for Fn=0.20.

As can be seen, results obtained from fine grids are in good agreement with experimental towing test
data. It should be noted here that these CFD results are obtained from fine grids, 504000 hexahedral
mesh elements and all the results including the experiment are calculated for only half-ship hull.
Table 2: Computed drag for different grid sizes for Fn=0.20

Total Resistance (RT) [N]

Experiment

Coarse grid

Medium grid

Fine grid

12.45

11.07

11.89

11.95

Conclusions
In this work,the finite volume solution method for the RANS equations is applied to the simulation of
flow field around a Wigley hull. The turbulence model used is the SST k-

two equation turbulence

model. The main goal of this numerical study is to show the capability a general-purpose CFD code of
284

Star CCM+ for design, analysis and feasibility of such a simulation for shipping industry. This work
highlighted the importance that CFD analysis is achieving in the ship resistance calculation process
and the SST k- two equation turbulence model is in good agreement with experimental towing test
data. For the future work, a robust and accurate semi-empirical model for predicting free surface flow
around merchant ship hulls can be developed by using Star CCM+ code.

References
Akimoto, H., Miyata, H., 2002, Finite-volume Simulation Methods to Predict the Performance of a Sailing Boat,
J. Mar. Sci.Tech.7, 31-42.
Gorski, J.J., 2002, PresentState of Numerical Ship Hydrodynamics and Validation Experiments, J. Offshore
Mechanics and Arctic Engineering, vol.1254, Issue 2, pp.74-80.
Parolini, N., Quarteroni, A., 2004, Numerical Simulation for Yacth Design, Proceedings of the 6th Conference
on Informatics and Mathematics, HERCMA 2003, vol.1, pp.38-44.
Sarda,O.P., 1986, Turbulent flows past ship hullsan experimental and computational study. Ph.D. thesis.
University of Iowa, Iowa City
Sato, Y., Miyata, H., Sato, T., 1999, CFD Simulation of 3-Dimensional Motion of a Ship in Waves: Application
to an Advanced Ship in Regular Heading Waves, J. Mar. Sci. Tech.4: 108-116.
Skytt, V., 2005, From CAD to CFD Meshes for Ship Geometries, Progress in Industrial Mathematics at ECMI
2004, Springer, 1 Edition.

285

287

288

= +

+ + =

=
+

= ( )
+ =

+ =

+ =

= ( )


=
+

+ + + =

289


+ +

+ =

+ =

( ) =

290

+
= =

=

( )


+ + =

+
= +

291

=
+ + +

( )

= =

292

293

294

295

A Desingularized Method for Ship-Wave Computation in Time Domain


Dario Bruzzone, Cristina Gironi
University of Genoa, Italy, bruzzone@dinav.unige.it

Abstract
This paper presents a numerical application of a fully desingularized method for the calculation of wave
induced motions of a marine vehicle in deep water.
In the present approach, the problem is solved in terms of velocity potential in the time domain. The
relevant Green function is obtained distributing desingularized Rankine sources inside the body and
above the calm water surfaces. The amount of the desingularizing length is determined in relation to the
dimensions of flat panels approximating the relevant surfaces. The boundary value problem is linearized
by assuming small motions and small wave amplitudes. The body boundary conditions are satisfied on the
mean body surface while the free-surface conditions are linearized on the calm water level and are
numerically solved and updated at each time step referring to an Eulerian time-stepping integration
scheme.
The results regarding two dimensional and three dimensional cases are presented. They include radiation
forces, added mass, damping, diffraction forces for 2D ship sections and for the S-175 hull with forward
speed. The computed results are compared with experimental data and other numerical solutions.

Keywords: Time Domain, Ship Motions , Rankine Source, Desingularized Surface.


1. Introduction
A very important criterion to evaluate the quality of a ship hull is its performance in a seaway. Prediction
of ship motions by numerical methods represents a relevant subject of the naval hydrodynamic research,
since an accurate sea keeping prediction corresponds to a good hull design realization. There are several
methodologies able to find a numerical solution for solving problems related to ship motions and they can
be characterized by a different range of complexity.
An important classification of these studies may be given by the option of analyzing the interactions
between hull and fluid in terms of linear, weakly non linear or fully non linear effects. The possibility to
consider non linearity increases the complexity of the calculations and the computational effort, but the
results obtained are more accurate then the linear ones when the wave steepness is increasing.

297

A linear analysis may be considered as a first step for subsequent non linear computations and, in
addition, it can be used to evaluate in a reasonable computer time the influence on the ship behaviour
when hull form or other important parameters are varied. The primary limitation that can be found in this
approach is that the analysis may give excessively approximated or even wrong responses for the cases in
which the amplitude of motions and the wave steepness are not small.
A 3D Rankine source panel method to evaluate ship motions for mono and multi-hull in the frequency
domain (Bruzzone, 2003) has been chosen by the authors as a reference for following developments
including the study here reported.
The introduction of non linearity directly in the frequency domain results difficult and limited. However,
introducing the so called hybrid or blended methods, based on the impulse functions and on the relation
between frequency and time domain through the Fourier transform, some results from frequency domain
can be exploited.
An example of a blended approach can be found in Bruzzone et al. (2010), where radiation and diffraction
forces are calculated in the frequency domain, while non linear Froude-Krylov and hydrostatic forces are
evaluated in the time domain. This solution permits to consider some sources of non linearity without
increasing very significantly the computational time.
To move toward a fully non linear approach, a linear panel method completely based in the time domain
has been developed as a first step and is presented in this paper. This choice is related to the future
development in course of realization: the initial linear analysis permits to put the base for the introduction
of non linearity without a significant growth of computational time. In addition, to achieve a further
reduction, a desingularized approach has been adopted. In particular, desingularized sources are
distributed above the panels of the calm water surface and slightly inside the body surface. Using this idea
the singularity problem and the integration over the single panel surfaces can be overcome obtaining an
easier procedure.
After a brief theoretical outline, results dealing with two-dimensional sections and a three dimensional
application for the S-175 hull are here presented.

2. Outline of the numerical method


An inertial right handed orthogonal coordinate system (X,Y,Z) advancing at the body constant speed U0 is
chosen; the XY plane is coincident with the calm water surface, the Z-axis is positive upwards and the Xaxis is chosen as the symmetry axis of the waterplane, positive astern. The motions of the body are
described by the vector

t k=1, , 6, defining the instantaneous position of a hull fixed reference

system (x,y,z) with respect to the system (X,Y,Z).

298

The fluid domain is bounded by the free surface S F and by the hull surface S H , the depth is considered
infinite. The fluid is assumed to be inviscid, incompressible and the flow irrotational.
With the previous hypotheses, a velocity potential

may be assumed, that satisfy the following

conditions:
a) The Laplace equation into the fluid domain:
2

(2)

b) The non-penetration boundary condition applied on the hull surface SH:


n VB n

(3)

where V B represents the velocity of a point on the hull surface and n

nx , ny , nz is the normal

vector pointing toward the fluid;


c) A kinematic and a dynamic condition on the free surface S F :

t
1
2

where z

(4)

p
t

1
U 02
2

(5)

x, y; t is the free surface elevation and g is the gravity acceleration.


can be

To simplify the general problem linear assumptions are exploited and the total velocity potential
expressed as the sum of two principal contributions: a steady base flow

, representing the potential of

a ship advancing with a certain velocity U 0 in calm water, and a perturbation unsteady potential
this unsteady potential can be further decomposed into an incident wave potential
potential

and six radiation potentials

0,

US

a diffraction

, related to the six degree of freedom of the ship.

As expressed in (6), this decomposition permits, after linearization, to separate the singular potentials as
the outcome of separate problems, therefore the hydrodynamic problem may be divided into a diffraction
problem, implying

and

7,

and six radiation problems for each

(6)
So, with the previous decomposition, a set of BVP has been stated for each of the motions ( j
and for wave diffraction ( j

7 ), assuming

1,...,6 )

as a known reference potential of lower order. The free

surface boundary conditions are as follows:

299

on z

(7)

on z

(8)

and the body boundary conditions are:


j

en j

mj

on SH

(9)

on SH

(10)

where n is the generalized normal vector and m j represents the components of the vector expressed by:
mj

mj

for

j 1,2,3

for

(11)

4,5,6

For the numerical solution, both the body and the free surface are geometrically described using flat
quadrilateral panels: NSH represents the number of panels of the body surface, NSF the number of free
surface panels, for a total of N T

N SH

N SF .

For the body surface, the discretization depends on the complexity of the geometry shapes, therefore the
panel dimensions are decreased for the zones characterized by a complex shapes as, for example, the bow.
For the free surface, the grid dimension depends on the distance of the panel center from the body. In the
zone closed to the hull, the area of the free surface panels is not too different from the area of the adjacent
body panels, in order to maintain a geometrical continuity that gives stability to the numerical solution of
the problem. In the zone of the free surface away from the body, the grid spacing will be increased to
create a sort of numerical beach in order to absorb the outgoing waves.
Whereas in the frequency domain method of Bruzzone (2003) flat panels have been considered upon the
relevant surface as described in Hess and Smith (1964), in this study desingularized sources are used for
the free and for the body surface, in order to reduce the computational amount and to simplify the
influence matrix.
Every panel is identified by two reference points: the panel center xC , that can be assumed as the
collocation point, and the source point xS , placed exactly above the center of the panel at a distance Dd
given by the square root of the local mesh size or panel diagonal.
The velocity potential at any collocation point xC on the boundary of the fluid domain can be expressed
as:

300

xCi

N SH
j 1

xS j G xCi ; xS j

N SF
j 1

xS j G xCi ; xS j

(13)

where the function G xC ; xS is the relevant Rankine source Green functions for the 2D and 3D case.
From expression (13) and from its derivatives the boundary conditions (7-10) may be discretized and a
system of linear equations can be obtained and solved for the unknown source strengths; the resulting
influence matrix Aij is inverted using an LU factorization. Once the source strengths are found, the
potential and its derivatives can be found for every panel of the fluid domain.
In particular, the velocity potential computed on the free surface at a certain time step allow to update the
kinematic and dynamic boundary conditions (7) and (8); the time stepping is obtained using the AdamsBashforth third-order linear method. If T is the period of the oscillation the time-step size
ensure the numerical stability of the method may be assumed as T 100

sufficient to

T 200 , as suggested in

Bandyk (2009).
Once the current value of potential is obtained, the pressure on the hull surface can be calculated using the
Bernullis equation, in which time-derivative of the potential term is directly solved by incremental ratio
and the once related to the forward velocity of the body is considered.
The forces and moments acting on the hull are given by the integration of pressure on the body.

Applications
The fully desingularized approach previously described has been tested on both two dimensional and
three dimensional cases. A half circular section and a rectangular section (B/T=28) have been considered
for the 2D examples. The S-175 hull has been used as a 3D case. The numerical results must be validated
by comparison with adequate and reliable experimental results. It is not always possible to find them in
the open literature so, in order to test the accuracy of a method, in some cases numerical data obtained
from different approaches may be also considered.
For the 2D problem Vugts (1968) offers experimental results for various sections while the 3D outcomes
have been compared with the results from three other numerical approaches, in particular: a strip theory, a
3D Rankine source panel method in the frequency domain (Bruzzone 2003) and the semi-desingularized
3D results found in Zhang et al. (2010).
Figg.1-2 show the comparison between numerical and experimental data regarding added mass and
damping coefficients for the circular section, in sway and heave motions respectively. In Fig.3 the
horizontal and vertical diffraction forces has been reported. The comparison between two different
rectangular sections shown in figg.4-5 points out the influence of the B/T value, in particular for added

301

mass coefficient that assume high values for the low adimensional frequencies tested. The results seem to
exhibit good agreement with the experimental data and thus indirectly confirm also the choice of the
desingularization parameters.
Dealing with the 3D example, figg.68 present the added mass and damping coefficients computed for the
S175 hull in heave and pitch motions. In every condition a forward velocity is considered (Fn=0.275).
Comparisons show a good correspondence between the different methodologies since the coefficients
seem to agree adequately. The results obtained with the fully desingularized approach always stay
between the frequency domain method and the strip theory results, even if in the case of the B55
coefficient the current approach seems to be underestimated with respect to the strip theory values. The
cross-coefficients present a higher spreading and this may be found also in other similar applications
reported in literature. The 3D results obtained from the frequency domain code show limited differences
for added mass coefficients while the damping terms turns out to be more underestimated with respect to
the other approach: this behavior is a critical issue noted also in some previous applications of the
frequency code.
Fig.9 presents the ship responses for heave and pitch motions in head sea. The results has been compared
with experimental data available from Fonseca and Guedes Soares (2005) and with the numerical results
obtained from the frequency domain method. The RAO values seem to well agree for both cases, except
around the heave peak frequency, where the results from the 3D frequency domain method appear to be
higher, probably due to the previous described damping underestimation. The fully desingularized
approach gives a good approximation respect to the experimental data.

Conclusions
A linear method in the time domain for evaluating ship motions in waves has been presented. A fully
desingularization has been adopted in order to reduce computing time.
The method has been applied to two dimensional and three dimensional cases in regular waves including
the well known S175 containership. The obtained numerical results evidence a satisfactory correlation
when compared with experiments. Added mass and damping for the S175 hull, where experimental results
are not available, exhibit a behavior similar to other numerical methods. Further steps will regard the
inclusion of non linear effects and the transition toward a fully non linear approach.

References
Bandyk P., (2009). A body-exact strip theory approach to ship motions computations, PhD Thesis, Department of Naval
Architecture and Marine Engineering, University of Michigan, Ann Arbor, USA.

302

Bruzzone, D., (2003). Application of a Rankine source method to the evaluation of motions of high speed marine vehicles,
Proceeding of the 8th International Marine Design Conference, Athens, Vol. II, pp. 69-79.
Bruzzone D., Grasso A., Zotti I., (2008). Nonlinear seakeeping analysis of catamarans with central bulb, Proc. Of the 6th
conference of high-speed performance marine vehicles, Naples, pp. 47-61.
Bruzzone D., Grasso A., (2009). Weakly nonlinear analysis of wave induced motions and load in head seas, Proceedings of
NAV09.
Bruzzone D., Gironi C., Grasso A., (2010). Nonlinear effects on motions and loads using an iterative time-frequency solver,
Proceeding of ITTC Workshop on Seakeeping, Seoul, Vol. I, pp. 17-27.
Fonseca N., Guedes Soares C., (2005). Comparison between experimental and numerical results of the nonlinear
vertical ship motions and loads on a containership in regular waves. Int. Shipbuilding Prog., 52, n. 1, pp. 57-89.
Hess J.L., Smith A.M.O., (1964). Calculation of non-lifting potential flow about arbitrary three-dimensional bodies. Journal of
Ship Research, Vol. 8, N.2, pp. 22-44.
Vugts J.H., (1968). The hydrodynamic coefficients for swaying, heaving and rolling cylinders in a free surface. International
Shipbuilding Progress 15, 251276.
Zhang X., Bandyk P., Beck R., (2010). Nonlinear Time-domain simulations of radiation and diffraction forces, Journal of Ship
Research, Vol. 54, N.2, June 2010, pp. 79-94.

303

Fig. 1. 2D added mass and damping coefficients in swaying for a circular section.

Fig. 2. 2D added mass and damping coefficients in heaving for a circular section.

Fig. 3. 2D horizontal and vertical diffraction forces for a circular section.

304

Fig. 4. 2D added mass and damping coefficients in swaying for a circular and a box section (B/T=28).

Fig. 5. 2D added mass and damping coefficients in heaving for a circular and a box section (B/T=28).

Fig. 6. S 175 added mass and damping coefficients of heave due to heave for Fn=0.275

305

Fig. 7. S 175 added mass and damping coefficients of pitch due to pitch for Fn=0.275

Fig. 8. S 175 added mass and damping coefficients pitch due to heave for Fn=0.275

Fig. 9. S 175 OdR for heave and pitch motions for Fn=0.25

306

Application of Classical Blade Element Momentum Theory and a Boundary


Element Method to Cavitating Marine Current Turbines
D. Usar, . Bal
stanbul Technical University, Turkey, usar@itu.edu.tr, sbal@itu.edu.tr

Abstract
In this study, both the classical blade element momentum (BEM) theory and a boundary element method
are applied to optimize a cavitating marine current turbine (MCT). The classical BEM theory refers to the
determination of optimal blade shape for any MCT by combining the equations of momentum theory and
blade element theory. These equations are basically developed for wind turbines which are very similar to
MCTs except cavity characteristics on blades. In addition, the tip speed ratios on MCTs are lower than
those of wind turbines due to the strength and cavitation requirements. On the other hand, a boundary
element method for cavitating 2-D hydrofoil sections is applied along the radius of blades to calculate the
section lift and cavity drag coefficients. The pressure is set to be equal to constant vaporize pressure of
water on the cavity surface and normal dipole and source panels are used on the cavity surfaces. The
cavity length and volume is solved in terms of cavitation number which can be computed for each
hydrofoil section along the blades of MCT. The application of the model is illustrated in a simple design
example and the results are discussed.
Keywords: Marine Current Turbine, Blade or Boundary Element Momentum Method, Cavitation, Hydrofoil

1. Introduction
In recent years, electricity energy generation from renewable sources such as wind, sun, marine and tidal
currents etc. has increased very rapidly than before. Their potential is very high to meet the worlds future
energy demand safely, cleanly and economically. The marine current turbines (MCT) that exploit
underwater currents for power generation are one of the renewable sources in oceans and seas. It is noted
in (Fraenkel, 2002) that MCTs are technically feasible and current resource is large enough to have the
potential to make a major contribution towards meeting the future energy demand. They provide also with
the regular and predictable energy as mentioned in (Charlier, 2003) and (Blunden and Bahaj, 2007). It is
very well-known that the MCT is basically very similar to the wind turbine in terms of its working
mechanism and its principle (Khan et.al, 2009). One of the main differences between MCT and wind
turbine is the cavitation which is an avoidable physical phenomenon. As is also very well-known that

307

cavitation is defined as the formation of vapor regions inside of a flowing liquid due to a decrease in the
local pressure. It appears in most of the blades of MCTs and should be included into the calculations.
In the past, a methodology for the hydrodynamic characteristics of MCT was presented in (Batten et.al,
2006). The turbine was modeled using the blade element momentum (BEM) theory. It was noted that
suitable section performance data, which also included the cavitation characteristics, were required for the
detailed design of the MCT blades. This investigation demonstrated how blade pitch angle or changes in
camber altered the stall performance and delayed the possibility of cavitation for MCTs. However, levels
of acceptance of cavitation are currently not yet clear (Batten et.al, 2006). On the other hand, in (Batten
et.al, 2008), the BEM theory was further developed for the design of MCTs. The theory included routines
for interpolation of 2-D section data and extrapolation for stall delay. Experimental verifications of
numerical predictions for hydrodynamic performance of MCTs were also presented in (Bahaj et.al,
2007a). It was indicated that two developed codes demonstrated similar trends in the results and provided
a satisfactory representation of experimental turbine performance (Bahaj et.al, 2007a). In addition, some
measurements from cavitation tunnel and a towing tank were given in (Bahaj et.al, 2007b). The results of
this experimental investigation provided an insight into the effect on performance of changes in the tip
immersion of the rotor and possible areas of cavitation inception. An experimental investigation on
cavitation, noise and slipstream characteristics of MCTs was conducted in (Wang et.al, 2007). The section
(foil) characteristics are also very important to design the blades of MCTs. Some measurements and
predictions of forces, pressures and cavitation characteristics on 2-D sections suitabe for MCTs were
given in (Molland et.al, 2004). An experimentally validated and developed BEM method was presented in
(Batten et.al, 2007) and some results (in terms of power, thrust and cavitation) were given for a model
turbine. Wake studies, on the other hand, for a model MCT could also be found in (Myers and Bahaj,
2007).
In the present study, however, the flow pattern and flow characteristics around a marine current turbine
(including cavity characteristics) are modeled by a BEM method. The blades of the marine current turbine
are divided into strips along the radius for the BEM method. The blade element analysis uses two
dimensional hydrofoil (section) characteristics and the coupling between momentum analysis and blade
element analysis allows the performance at a given tip speed ratio and a given radius to be calculated. In
order to compute the cavitating characteristics of flow around two dimensional sections, a potential based
panel method is used. This panel method discretizes with panels the exact surface of blade sections. The
method is inherently non-linear with either the thickness or the angle of attack that are very important to

308

model the proper cavity surface on the surface of blade section, since it makes no assumption about the
magnitude of these quantities. The panels are located on the exact cavity surface of which the shape is
determined iteratively, until the kinematic and dynamic boundary conditions are both satisfied on that
surface. In the first iteration, the cavity panels are located on the wetted foil surface. Cavity shapes are
predicted from applying the dynamic cavity condition as the last iteration of the panel method. A cavity
closure condition is also applied to require the sum of the sources to be equal to zero. It should be
mentioned that it is determined the corresponding cavitation number and cavity shape for a given cavity
length which can be extended up to a given point. After computing the flow characteristics (including
cavity) of foil sections of blades, a span-wise integration produces the total torque, thrust and power. The
method is also very suitable for blade optimization. The results of blade element momentum method with
the corresponding panel method to include the cavity characteristics are compared with those of
experiments and other numerical methods given in literature. The strip-independence study is also
performed to get the converged results.
2. BEM Method
The hydrodynamic performance of a MCT, similar to wind turbine (Burton et.al, 2001), (Manwell et.al,
2002) and (Hansen, 2008), can be modeled by BEM method (Batten et.al, 2008). The BEM theory couples
the momentum theory with the blade element theory for which local events take place at the actual blades.
The stream tube introduced in the 1-D momentum theory is discretized into N annular elements of height
dr. It is assumed that there is no flow across the elements. The momentum theory is used to derive the
axial and circumferential inflow factors (a and a', respectively), with the introduction of a tip loss factor
(F), including the effect of finite number of rotor blades. The momentum solution for the differential
thrust and torque can be given as,

dT

4 r V 2 a (1 a) F dr

(1)

dQ

4 r 3V a (1 a) F dr

(2)

The blade element theory, on the other hand, is used to model the section drag and torque by dividing the
rotor blade into a number of strips along its span-wise direction. By the combination of these theories, at
each blade radius (strip), the rotor thrust loadings and power loadings are determined by matching the
fluid momentum changes to blade forces based on lift and drag coefficients (CL and CD) at the angles of
attack of the blade sections. The integration of the loadings across the blade allows the derivation of
torque, drag, and power coefficients for the blade. Since the details of BEM theory can be found in
(Hansen, 2008) and (Batten et.al, 2008), it is not repeated here.

309

3. Section Cavity Model and Coupling with BEM Theory


Consider a cavitating 2-D hydrofoil (section) surface subject to a uniform inflow, as shown in Fig. 1, for
section cavity model.

Fig. 1. A cavitating 2-D hydrofoil

The x axis is positive in the direction of uniform inflow and the z axis is positive upwards. S c, Sh and Sw
are the cavity, hydrofoil and wake surfaces, respectively. The fluid is assumed to be inviscid,
incompressible, and irrotational. Then the flow field can be expressed in terms of the total velocity
potential, , or the perturbation potential, , as follows:

( x, z)

( x, z) Vx

(3)

where, V is the velocity of incoming flow. The perturbation potential , and the total potential, , should
satisfy Laplaces equation in the fluid domain:
2

(4)

In addition, the following boundary conditions should be satisfied by .


i. The kinematic boundary condition: The flow should be tangent to the surface of the hydrofoil and cavity

V n

(5)

where n is the unit normal vector to the hydrofoil or cavity surface directed into the fluid domain.
ii. The dynamic boundary condition on the cavity surface: The pressure is constant and equal to pv (pv :
vapor pressure) on the cavity surface. By applying Bernoullis equation, the total velocity on the cavity, q c
, can be given as (Kinnas and Fine 1993)

qc

V (1

(6)

310

where is the cavitation number,

p
1

pv
2

(7)

V2

iii. The Kutta condition: The velocity at the trailing edge of the hydrofoil is finite,

finite

(8)

iv. Cavity closure condition: The cavity closes at its trailing edge. The complicated physical phenomena
occurring at the cavity trailing edge are ignored and the region is represented by a termination model.
Please refer to (Kinnas and Fine 1993) for details. The cavity detachment point is assumed to be known
and in this study the leading edge of the hydrofoil is chosen as the cavity detachment point.
By applying Greens third identity to the governing equation (4) in the fluid domain, the following integral
equation for the perturbation potential on the hydrofoil surface and the cavity surface can be written,

SH

ln r
n

ln r dS

W
SW

ln r
dS
n

(9)

where, SH and SW are the boundaries of the hydrofoil (including the cavity surface) and the wake surface,
respectively. r is the distance between source point and field point.

is the potential jump across the

wake surface, and n+ is the unit vector normal to the wake surface pointing upwards. After applying the
kinematic boundary condition (5) to equation (9), the following integral equation can be written for the
hydrofoil part, including cavity surface Sc,

SH

ln r
n

(V n ) ln r dS

W
SW

ln r
dS
n

(10)

Here, the hydrofoil and the cavity boundaries are discretized into straight panels with piecewise constant
source and dipole distributions. The discretized integral equation provides a matrix of equations with
respect to the unknown potential values and can be solved by any matrix solver. This is called a boundary
element method (panel method) for cavitating hydrofoils. Please refer to (Kinnas and Fine, 1993) and (Bal
and Kinnas, 2002) for details.
Then the coupling between BEM method and the cavity model which is solved by a boundary element
method is carried out by the following methodology. First, the cavitation numbers at different blade angles
are calculated by equation (7) for each section (strip) and then the above cavity model is applied to get the

311

lift value, drag value and cavity shape. The computed lift coefficients and drag coefficients (viscous drag
(from (Abbott and Doenhoff, 1959) + cavity drag) are now the input values for BEM method. After
applying BEM method (integrating the 2-D sectional characteristics), the thrust and power values versus
current speed can be found.

4. A Numerical Example
A three bladed MCT is chosen to compare the results of present methodology with those of given in
(Batten et.al, 2006). The chord radius ratio (c/R) is changing linearly from 0.125 at r=0.2R (at the hub) to
0.05 at r=R (at the tip). The twist along the span (radius) of blade is defined as

a tan(

r
)
4R

2 .

The blade sections (hydrofoils) are selected as NACA 0012 profiles. The chord/radius ratio and incidence
angles are shown in Fig. 2 and Fig.3, respectively.

Fig. 2. Chord/radius ratios as given in (Batten et.al, 2006).

Fig. 3. Incidence angles as given in (Batten et.al, 2006)

312

The lift coefficient and drag coefficient (viscous drag + cavity drag (viscous drag values are taken from
(Abbott and Doenhoff, 1959))) values of each strip versus angle of attack and computed cavitation
numbers are given in Fig. 4 and Fig. 5, respectively.

Fig. 4. Lift and drag coefficients of cavitating NACA 0012 sections

Fig. 5. Computed cavitation numbers of NACA0012 section

Moreover, the cavity shapes for each strip are shown in Fig. 6. The BEM method can now be applied
easily using Figs. 4 and 5.

313

Fig. 6. Cavity shapes near tip of blade.

In Fig. 7, the comparison of calculated power coefficient and thrust coefficient values with and without
blade cavitation versus tip speed ratio is presented.

Fig. 7. Power and thrust coefficients with and without cavitation

Note that although the differences between the power coefficient values without cavitation and with
cavitation are small for higher tip speed ratios (higher than 5), the power coefficient with cavitation is
lower than that of without cavitation for small tip speed ratios (smaller than 5). The possible reason for
this is that the cavitation induces higher drag values for small tip speed ratios. Note also that thrust

314

coefficient values with cavitation are higher than those of without cavitation, since the cavitation surface
behaves like an added camber effect (cavity increases the lift coefficient of the section).
In Fig. 8 the calculated power coefficient values are compared with those of (Batten et.al, 2006).
The possible reason for the differences between the results of (Batten et.al, 2006) and those of present
methodology is that NACA 63-212 sections are used in the work of (Batten et.al, 2006). On the other hand
NACA 0012 sections are applied in the present calculations.

Fig. 8. Comparison of coefficient of thrust

Conclusions
In this study, the classical BEM theory and a boundary element method for cavitating blade sections are
coupled to analyze a MCT. The followings are found:
1. The power coefficient, CP, has been reduced under the working condition with cavity, especially for tip
speed ratios considered to be relatively low at which the cavitation inception is more likely to occur. In
addition the difference between the power coefficient with and without cavitation is very small for higher
tip speed ratios.
2. The thrust coefficient, CT, has not been influenced significantly from cavity characteristics. CT values
are slightly increased due to the additional camber effect of cavitation (additional chamber means an extra
lift and thrust value).

315

3. There are differences between the power coefficient values of (Batten et.al, 2006) and the present
results since NACA 0012 sections are used instead of NACA 63-212 which are applied in (Batten et.al,
2006). This problem will be tackled in near future.

References
Abbott IH, von Doenhoff AE, (1959), Theory of Wing Sections, Dover Publ., NY, USA.
Bahaj AS, Batten WMJ, McCann G, (2007a), Experimental verifications of numerical predictions for the
hydrodynamic performance of horizontal axis Marine Current Turbines, Renewable Energy, Vol. 32, pp: 24792490.
Bahaj AS, Molland AF, Chaplin JR, Batten WMJ, (2007b), Power and thrust measurements of Marine Current
Turbines under various hydrodynamic flow conditions in a cavitation tunnel and a towing tank, Renewable Energy,
Vol. 32, pp: 407-426.
Bal S, Kinnas SA, (2002), A BEM for the prediction of free surface effect on cavitating hydrofoils, Computational
Mechanics, Vol. 28, pp: 260-274.
Batten WMJ, Bahaj AS, Molland AF, Chaplin JR, (2006), Hydrodynamics of Marine Current Turbines,
Renewable Energy, Vol. 31, pp: 249-256.
Batten WMJ, Bahaj AS, Molland AF, Chaplin JR, (2007), Experimentally Validated Numerical Method for the
Hydrodynamic Design of Horizontal Axis Tidal Turbines, Ocean Engineering, Vol. 34, pp: 1013-1020.
Batten WMJ, Bahaj AS, Molland AF, Chaplin JR, (2008), The prediction of Hydrodynamic performance of Marine
Current Turbines, Renewable Energy, Vol. 33, pp: 1085-1096.
Blunden LS, Bahaj AS, (2007), Tidal energy resource assessment for tidal stream generators, Proc. Inst. Mech.
Eng., Part A, Journal of Power and Energy, Vol. 221, No:2, pp: 137146.
Burton T, Sharpe D, Jenkins N, Bossanyi E, (2001), Wind Energy Handbook, John Wiley and Sons Inc., NY, USA.
Charlier RH, (2003), A sleeper awakes: tidal current power, Renewable and Sustainable Energy Reviews, Vol.
7, Issue 6, pp: 515-529.
Fraenkel PL, (2002), Power from marine currents, Proc. Inst. Mech. Eng., Part A, Journal of Power and Energy,
Vol. 216, Special Issue, pp: 114.
Hansen, MOL, (2008), Aerodynamics of Wind Turbines, 2nd edition, Earthscan, London, UK.
Khan MJ, Bhuyan G, Iqbal MT, Quaicoe JE, (2009), Hydrokinetic Energy Conversion Systems and Assessment of
Horizontal and Vertical Axis Turbines for River and Tidal Applications: A Technology Status Review, Applied
Energy, Vol. 86, pp: 1823-1835.
Kinnas SA, Fine NE, (1993), A Numerical Nonlinear Analysis of the Flow around Two- and Three-Dimensional
Partially Cavitating Hydrofoils, Journal of Fluid Mechanics, Vol. 254, pp: 151-181.
Manwell JF, McGowan JG, Rogers AL, (2002), Wind Energy Explained Theory, Design and Application, John
Wiley and Sons Inc,NY, USA.
Molland AF, Bahaj AS, Chaplin JR, Batten WMJ, (2004), Measurements and predictions of forces, pressures and
cavitation on 2-D sections suitable for marine current turbines, Proc. Inst. Mech. Eng., Part M, Journal of
Engineeing for Marine Environment, Vol. 218, No: 2, pp:127138.
Myers L, Bahaj AS, (2007), Wake Studies of a 1/30th Scale Horizontal Axis Marine Current Turbine, Ocean
Engineering, Vol. 34, pp: 758-762.

316

Comparative Seakeeping Performance Analysis of a Warship


Tahsin Tezdoan, Ali Ihsan Aldogan
Faculty of Naval Architecture and Ocean Engineering., Istanbul Technical University, Istanbul, Turkey,
tezdogan@itu.edu.tr

Abstract
A successfully designed ship is expected to fulfill her mission in almost all weather and sea states without
compromising her safety. This is particularly important for a warship and crew onboard to be able to
perform their complex tasks in good physical and mental state.
This paper presents comparative seakeeping performance analysis of a warship operating in Turkish
waters, which include Mediterranean Sea, Aegean Sea and Black Sea, for varying sea states, wave
headings and ship speeds.
The comparative analysis was conducted by using a commercial seakeeping package (ShipmoPC), which
is a strip theory based software, for the 6 degrees of freedom motion responses as well as the vertical
accelerations, and added wave resistance. The effect of active fins on the roll motion responses was also
explored. The analysis results were compared with the NATO Standardization Agreement (STANAG
4154) criteria. The results were presented in standard graphical format and polar diagrams, and discussed
in details in the paper.

Keywords: Seakeeping, Strip Theory, Ship Motions, Warships


1. Introduction
The overall performance of a ship depends on the seakeeping performance in specified sea areas where the
vessel operates. She is supposed to perform her duties even in severe sea conditions. Therefore, prediction
of ship motions and seakeeping performance are very important for a ship in the preliminary design stage.
In this paper, some seakeeping analyses were conducted for a warship operating in Turkish waters for
varying sea states, wave headings and ship speeds. The results of analysis were presented in graphical
format.
The factors which lead to restrict performance of a warship in a seaway may be listed as follows:
Severe ship motions due to waves
Motion induced interruptions
High accelerations
Slamming
Deck wetness
Propeller emergence

317

In order to assess of seakeeping performance of a warship in a specified sea environment, these inputs
must be identified completely:
Type of a warship and her missions
Principal dimensions and hull geometry
Mass distribution
Coordinates of critical points for the vessel such as helipad, bridge deck, combat operations center
Sea areas where the ship operates and sea states
Seakeeping criteria which is determined in accordance with her type, her mission, and her
armament
After the determination of foregoing inputs, seakeeping analysis of a warship was carried out by the aid of
a commercial seakeeping package, ShipmoPC, which is a 2-D strip theory based software (BMT, 2001).
These calculations include ship responses in regular and irregular seas, added resistance due to waves,
vertical acceleration, and slamming. In addition, the effect of active fins on the roll motion responses was
also explored.
1.1. Basic properties of the sample warship
The sample warship studied in this paper is a landing ship, which is a form of amphibious warship
designed to support amphibious operations. These amphibious assault ships transport and launch
amphibious craft and vehicles with their crews and embarked personnel (Web 1).
Principal dimensions of the vessel are given in table 1. Ship geometry is divided into 20 stations and 0.
station is regarded as aft perpendicular. At the same time, mass properties related to these stations are
entered into the software. Besides, bilge keel, skeg, rudder, shaft brackets and active roll fins are modeled
in the software.
Table 1. Principal dimensions of the warship

LBP

k yy

208 m.

52 m.

7 m.

CB
25,430 ton

0.540

2. Ship Responses in Regular Seas


Ship responses in regular seas are calculated in order to obtain ship responses in irregular seas using the
linear superposition principle.
Ship motions in regular seas can be predicted experimentally, but this may not be appropriate in
preliminary design stage because data of the ship may be changed frequently. Therefore, it is really
expensive and laborious to conduct experiments for every changing situation.

318

It may be noted that warships with slender geometry are very suitable for 2-D linear strip theory
application. ShipmoPC provides motion predictions using a frequency domain strip theory of Salvesen
et al. (1970). For lateral plane motions, appendage and viscous forces are highly important, so their
effects are computed using Schmitkes method (1978).
In ShipmoPC, two dimensional sectional hydrodynamic coefficients are determined using either Lewis
form method (1929) or boundary element method (Sclavounos and Lee, 1985). In this paper, the
boundary element method is chosen to compute sectional hydrodynamic properties.
Heave and pitch responses of the warship for varying headings are presented in figure 1 and 2. In these
graphics, wave frequency is given in the apsis (rad/sec), whereas the ordinate represents response
amplitude operator (RAO). These values are computed for 22 knot ship speed.
Linear motion amplitudes are non-dimensionalised by dividing by the wave amplitude ( a) for
translation motions (surge, sway, heave), and by dividing by the wave slope amplitude (k a) for angular
motions (roll, pitch, yaw).
As seen in fig. 1 and 2, heave amplitudes reach maximum values in beam seas; on the other hand the
highest pitch amplitudes are occurred in head waves.

Fig. 1. Heave RAO for 22 knots in regular waves

319

Fig. 2. Pitch RAO for 22 knots in regular waves

2.1. Added wave resistance in regular waves


The resistance of a ship in a seaway is known to be greater than the ship resistance in calm water. The
difference between these two values is called the added resistance. Added resistance due to the waves is
predicted not only experimentally but also analytically from the ship motions using the strip theory. The
added resistance prediction in ShipmoPC is executed using the near-field method given by Faltinsen et al.
(1980).
The added wave resistance of the warship for varying speeds is shown in figure 3. Horizontal axis of the
graphic is non-dimensional encounter frequency coefficient, and the vertical axis is non-dimensional
added resistance coefficient. These coefficients are derived as given in equation 1 and 2:
Non-dimensional encounter frequency coefficient: e=e(L/g)0.5
Non-dimensional added resistance coefficient:

RAW
g a2 ( B 2 / L)

AW

(1)
(2)

The maximum added resistance is to be expected in head waves, so the added resistance prediction in
figure 3 is computed for regular head waves. It is obvious from figure 3 that added resistance increases
with increasing ship speed.

320

Fig. 3. Added resistance curves of the warship for varying ship speeds in regular head waves

3. Definition of the Seaway


The regular waves are seldom found in nature and hence the RAOs are of little consequence on their own.
The natural seaway in which a ship operates can only be described by means of a statistical model. The
spectrum or spectral density function is the primary device used for representing the seaway and the
oscillatory response of the vessel to the seaway. The wave characteristics of an area must be known in
terms of the distribution of wave energy with respect to frequency and direction, as well as the severity of
seas as indicated by the wave height probability distributions. The wave energy distribution within various
wave height bands can be represented through the use of a wave spectral family, Sarioz and Narli (2005).
The most used mathematical sea spectrum model is two-parameter ITTC spectrum (ITTC, 1978).

A
5

exp

(3)

where A and B constants are defined by

A 173

2
H1/3
and B
T

691
T14

(4)

where T1 is mean wave period and modal wave period equals Tm=1.2958T1.
In this paper, two-parameter ITTC spectrum is used to model Turkish waters including Mediterranean
Sea, Aegean Sea and Black Sea. Significant wave heights and modal wave periods to represent Turkish
waters are given in table 2.

321

Table 2. Significant wave heights and modal periods for varying sea states for Turkish waters (Tezdoan, 2011)
Sea State

Significant Wave
Height (m)

1
2
3
4
5
6

0.05
0.30
0.88
1.88
3.25
5.00

Modal Wave Period (sec)


Black sea

Mediterranean

3.53
4.14
5.41
7.28
9.1
10.19

4.42
5.00
6.25
8.15
10.16
11.74

Aegean sea
3.63
4.01
4.86
6.25
7.96
9.81

4. Prediction of Responses in a Seaway


Ship responses in a seaway are obtained by superposition of the transfer functions with the wave spectral
family (Sarioz and Narli, 2005).
Szz=SRAO2

(5)

According to Lloyd, for ship design purposes the most common practise is to use short crested sea with
90 spreading angle (1989), so all calculations in this paper are carried out accordingly.

Fig. 4. Comparison of RMS heave amplitudes in Turkish waters with 22 knot ship speed (sea state: 6)

Fig. 5. Comparison of RMS pitch amplitudes in Turkish waters with 22 knot ship speed (sea state: 6)

322

RMS heave and pitch displacements in irregular seas are shown in figure 4 and 5, respectively. In these
calculations, significant wave height and modal period values are chosen at sea state 6 given in table 2.
According to these graphics, blue curves represent Mediterranean Sea, red curves represent Black Sea, and
green ones present Aegean Sea. As shown in figure 4 and 5, heave and pitch amplitudes reach maximum
values in Mediterranean among the other seas.
4.1. Rolling analysis
Rolling has a remarkable importance on human comfort and safety of the cargo. Hence, it should be
predicted with enough accuracy. Despite the great impact of rolling on ship operations, it is the most
difficult motion to predict because of the viscous effects. According to McTaggart, ShipmoPC uses the
Schmitkes method to include viscous effects in lateral motions (1997).
The polar diagram showing RMS roll amplitudes of the warship for different ship speeds in Mediterranean
(sea state: 6) is given in figure 6. It can be said that the most severe roll amplitude is predicted for zero
speed in beam seas (approx. 4). Roll motion is decreasing with the increasing ship speed between 0-22
knots speed interval.

Fig. 6. Polar diagram shows RMS roll amplitudes for different ship speeds (Mediterranean, sea state: 6)

4.2. Effect of the active roll stabiliser fin on roll motion


Active roll stabiliser fins are usually mounted on rotatable stocks at the turn of the bilge near the middle of
the ship. The angle of incidence of the fins is continually adjusted by a control system which is sensitive
to the rolling motion of the ship. The fins develop lift forces which exert roll moments about the centre of

323

gravity of the ships. There roll moments are arranged to oppose the moment applied by the waves and the
roll motion is reduced, Lloyd (1989).
The warship has active roll fin besides bilge keel. Some properties of the active fin are given in table 3.
In this part, performance of the active roll fin is assessed. For this purpose, rms roll amplitudes are
calculated considering the effect of roll fin firstly, and then the same analysis is carried out without roll fin
in the same sea conditions. Analysis of roll motion is performed in Mediterranean (sea state: 6) for 22
knots ship speed. In all circumstances, the contribution of the bilge keel to roll motion is included to the
calculations. The comparative graphic is shown in figure 7.
Table 3. Properties of the active roll fin (Tezdoan, 2011)
Properties
Roll acceleration gain
Roll velocity gain
Control system natural frequency
Control system damping ratio

Values
4.130 sec2
4.070 sec
0.492 rad/sec
0.076 (-)

Fig. 7. The effect of active roll fin on roll amplitudes for 22 knots ship speed (Mediterranean, sea state: 6)

It is stated that the active roll fin reduces maximum roll amplitudes by approx. 44% according to figure 7.
The vessel should purpose to minimize roll amplitudes to carry on her tasks safely.
4.3. Vertical acceleration analysis
The amplitude of vertical acceleration ( z )a at any point along the ship length is given by (Bhattacharyya,
1978):

( z )2a

( z )2a

xb2 ( )a2

2( z )a ( )a xb cos

(6)

where ( z )a is the amplitude of heaving acceleration at the CG, and ( ) a is the amplitude of pitching
acceleration at the CG.

is the phase angle.

324

Figure 8 shows the effect of changing severity of the sea on vertical acceleration. The calculations are done in
head seas for different vessel speeds.
The analyses are conducted at a point on the bridge deck. The ordinate of the figure 8 is given as significant
vertical acceleration, whereas the x-axis represents varying sea states. As seen in figure 8, the significant
vertical acceleration values are increasing as the severity of the sea goes up. As expected, vertical acceleration
increases with ascending ship speed.
Figure 9 illustrates the effect of changing longitudinal location on vertical acceleration in sea state 6 in
Mediterranean at 16 and 22 knots. The calculations are done on the centerline of the warship at the same height
as the vertical center of gravity (VCG). It may be noted that there is a strong dependence on longitudinal
location, and RMS vertical acceleration is 3.75 times greater at forward perpendicular than at Station 8 (at 22
knots).

Fig. 8. Effect of severity of the sea on vertical acceleration for varying ship speeds in head seas (Black Sea)

Fig. 9. Effect of longitudinal location on vertical acceleration in head seas (Mediterranean, sea state: 6)

325

5. Seakeeping Criteria
Sarioz and Narli point out that in order to assess the effect of seakeeping performance on the mission
capability of the vessel the mission requirements need to be translated into seakeeping performance
requirements (Sarioz and Narli, 2005).
Criterions for seakeeping performance are different for all vessels with respect to their types, missions and
armament. Also, most of criterions can vary depending on the location and region. Every mission has its
own special limit value that makes seakeeping criteria a complex issue.
Some limit values for a warship which has transit and patrol missions are given in table 3. These values
may be appropriate for the warship discussed in this paper.
Table 3. Seakeeping criteria: transit and patrol mission (NATO, 2000)
Parameter

Limit Value

Roll angle

4.0 RMS deg

Pitch angle

1.5 RMS deg

Vertical acceleration

0.2 RMS g

Deck wetness index

30 per hour

Bottom slamming index

20 per hour

Helicopter take off (roll)

3.0 RMS deg

Helicopter take off (pitch)

1.0 RMS deg

Conclusions
Examining analyses results, it is appeared that the highest values of ship motions, added resistance, and
vertical acceleration are occurred in Mediterranean Sea, and then it is followed by Black Sea and Aegean
Sea.
The motions in regular seas are computed in order to calculate ship responses in irregular seaways using
the linear superposition principle. The challenging part of predicting ship responses in irregular seas is to
model real sea waves adequately. To do this, some mathematical sea spectrums are derived and
consequently they simplify the calculations.
In this paper, two-parameter ITTC spectrum has been used to model Turkish waters. All the analyses have
been done according to this.
The effect of active roll fin has been evaluated and it may be noted that it reduces maximum roll
amplitudes by approx. 44%.

326

In the next part, vertical accelerations have been computed and the effect of longitudinal position on
vertical acceleration has been assessed. It may be concluded that the influence of longitudinal location is
very significant and RMS vertical acceleration is 3.75 times greater at forward perpendicular than at
Station 8.
Finally, seakeeping criteria is explained briefly. Seakeeping performance of a warship enormously
depends on the chosen limit values. Sample seakeeping criterions taken from STANAG 4154 have been
given in the paper.

Acknowledgements
This paper has been dedicated in the memory of Prof. Dr. Ali Ihsan ALDOAN, who passed away on 11th
of February, 2011.
I wish to thank to Prof. Dr. Mehmet ATLAR who gave me his support at Newcastle University. I am
really grateful to him for his assistance. I would like to also thank to my thesis advisor, Prof. Dr. Metin
TAYLAN, who shared all his experience and knowledge with me.

References
Bhattacharyya, R., (1978). Dynamics of Marine Vehicles. McCormick, M. E. (Ed.), Wiley, New York.
British Maritime Technology (BMT), (2001). ShipmoPC Version 3 User Manual. Revision 10, BMT Fleet
Technology Limited, Canada, pp. 107.
Faltinsen, O. M., Minsaas, K. J., Liapis, N. and Skjordal, S. O., (1980). Prediction of resistance and propulsion of a
ship in seaway. 13th Symposium on Naval Hydrodynamics, Tokyo, 505-529.
ITTC Seakeeping Committee Report, (1978). 15th International Towing Tank Conference, The Hague, 1, 55-114.
Lewis, F. M., (1929). The inertia of water surrounding a vibrating ship. Transactions of SNAME 37, 1-20.
Lloyd, A. R. J. M., (1989). Seakeeping: Ship Behaviour in Rough Weather. Ellis Horwood, Chichester, UK.
McTaggart, K. A., (1997). Shipmo7: An Updated Strip Theory Program for Predicting Ship Motions and Sea Loads
in Waves. Defence Research Establishment Atlantic, Technical Memorandum 96/243.
NATO, (2000). Common Procedures for Seakeeping in the Ship Design Process. Standardization Agreement
(STANAG 4154).
Salvesen N., Tuck E. O. and Faltinsen O., (1970). Ship motions and sea loads. Transactions of SNAME 78, 250-287.
Sarioz, K. and Narli, E., (2005). Effect of criteria on seakeeping performance assessment.Ocean Eng.32, 1161-1173.
Schmitke, R. T., (1978). Ship sway, roll, and yaw motions in oblique seas. Transactions of SNAME 86, 26-46.
Sclavounos, P. D. and Lee, C., (1985). Topics on boundary element solutions of wave radiation-diffraction problems.
4th International Conference on Numerical Ship Hydrodynamics, Washington.
Tezdoan, T., (2011). Investigation of ship motions and application to ships (in Turkish) (master thesis). Istanbul
Technical University Graduate School of Science Engineering and Technology.
Web 1, http://en.wikipedia.org/wiki/Dock_landing_ship

327

A Comparative Study on Ship Motions: Theory vs. Model Experiments


Tahsin Tezdoan, Metin Taylan
Dept. of Naval Architetcure and Marine Engineering, Istanbul Technical University, Istanbul, Turkey,
tezdogan@itu.edu.tr, taylan@itu.edu.tr

Abstract
The best way of validating theory on ship motions is to compare the outcome with that of the experimental
analysis. In this work, a commercial seakeeping package, which is based on the strip theory, has been
utilized. The software provides motion predictions and sea loads using 2-D linear strip theory of Salvesen
et al. (1970).
This paper presents the comparative study of theoretical ship motions with experimental work. Detailed
analyses were performed for a series 60 ship having block coefficient of 0.70 and a cargo ship in regular
head waves using the software, and then the results were compared with the experimental data. For a
series 60 ship form, experimental results are available in the literature, whereas for the cargo ship, the
model experiments were conducted at ITU Ata Nutku Ship Model Testing Laboratory.
The results were presented in graphical form, and discussed in details. Comparing the output of the
software with the experimental results for both vessels, it may be said that there is a considerable
correlation between them.
Keywords: Seakeeping, Strip Theory, Ship Motions, Experiment

1. Introduction
In this paper, comparison of theoretical ship motions with experiment is expressed. Two different ships
were taken into the consideration for this purpose. Transfer functions of ship responses in regular seas
were obtained by the aid of the software for varying ship speeds, headings and wave frequencies. The
results were compared with model experiments for both ship types.
First of all, transfer functions for series 60 form with in regular head waves at zero speed are computed
and compared with experimental data found in literature. In the following section, motions of a cargo ship
with transom stern is examined. These results are compared with experiments data obtained in model ship
laboratory of ITU.

329

2. Comparison of Response Amplitude Operators


Under the assumptions that the responses are linear and harmonic, the equations of motion for a ship
advancing at constant forward speed with arbitrary heading in regular sinusoidal waves can be written in
the following form:
6
k 1

M jk

Ajk

B jk

C jk

Fj ei et , j 1, 2,3,..., 6

(1)

where Mjk are the components of the generalized mass matrix, Ajk and Bjk are the added mass and damping
coefficients, Cjk are the hydrostatic restoring coefficients, and Fj are the complex amplitudes of the
exciting force and moment. j=1,2,3,,6 refer to the surge, sway, heave, roll, pitch, and yaw motion,
respectively. The dots stand for time derivatives (Sarioz et al, 2000).
2.1 Series 60 CB=0.70 form
The general characteristics of sample Series 60 ship studied in this paper is given in table 1. The body plan
of the ship form is illustrated in figure 1.
Table 1. Principal dimensions of the series 60 CB=0.70 form
LBP

LCB

CB

121.920

17.417

6.857

+0.5 LBP % (to fore)

0.700

Fig. 1. Body plan of the series 60 CB=0.70 form (Sarioz et al, 2000)

Heave and pitch RAOs for series 60 CB=0.70 form are calculated by using the software, and the outputs of
the software are compared with experimental results, extracted from reports of Gerritsma and Beukelman
(1966). The experiments were conducted for different model speeds corresponding to Fn=0.15, 0.20, 0.25,
and 0.30 in regular head waves. Heave and pitch amplitudes and phases in head seas were measured in
regular waves with two different wave height to wave length ratio, namely 1/50 and 1/40 (Sarioz et al.,
2000)

330

All these comparisons in head sea conditions are shown between figures 2 and 9 for the Series 60 hull
having CB=0.70.
As given between figure 2 and figure 9, it may be said that the outputs of the software are quite
compatible with the experimental results even in high Froude numbers. These results may indicate that the
software is successful at application of linear strip theory to series 60 forms in regular head seas.

Fig. 2. Comparison of heave RAO at Fn=0.15

Fig. 3. Comparison of pitch RAO at Fn=0.15

Fig. 4. Comparison of heave RAO at Fn=0.20

Fig. 5. Comparison of pitch RAO at Fn=0.20

Fig. 6. Comparison of heave RAO at Fn=0.25

Fig. 7. Comparison of pitch RAO at Fn=0.25

331

Fig. 8. Comparison of heave RAO at Fn=0.30

Fig. 9. Comparison of pitch RAO at Fn=0.30

2.2 Cargo ship form with transom stern


Some seakeeping experiments have been conducted in ITU Ata Nutku Ship Model Testing Laboratory for
a cargo ship with transom stern for varying model speeds corresponding to Fn=0.0, 0.089, 0.1425, and
0.2138 in regular head waves. Heave and pitch RAOs for the ship and added resistance due to waves are
computed in consequence of the experiments. The analytical analyses for the same conditions are made
by the aid of the software. All analyses values are illustrated in the same graphic in order to see the
difference and make a clear comparison.
The principal properties of the ship form are given in table 2. Strip theory is actually very suitable for
slender ship forms. The cargo ship has a fuller hull form (CB=0.80), so it may not be appropriate for linear
strip theory application. Besides she has a transom stern which makes end effects correction mandatory.
This is a correction to the hydrodynamic coefficients for the effects at the aftermost station (BMT,
2001). Classic strip linear theory does not include these corrections, since it assumes that ship sectional
area varies gradually along the ship length (McTaggart, 1997). However, this is not applicable to a
transom stern ship. End effect correction is applied to the cargo ship int.
Due to these reasons, the correlation between experimental results and the outputs of the software has a
great importance to detect reliability of the software.
Table 1. Principal dimensions of the cargo ship (Tezdogan, 2011)

LBP

B/T

CB

Vs

84.94 m.

2.44

0.800

12 kn.

5906 t.

Heave and pitch RAOs for the cargo ship at zero ship speed are given in Figure 10. Heave and pitch RAOs
for the cargo ship at Fn=0.089, 0.1425, and Fn=0.2138 are given in Figure 11, Figure 12 and Figure 13,
respectively. Blue points on the graphics symbolize experimental results, and green lines present outputs

332

of the software. According to these graphics, it may be noted that there is a strong concordance between
experiments and the software.

Fig. 10. Comparison of a) heave RAO, b) pitch RAO for the cargo ship (Fn=0.00, head waves)

Fig. 11. Comparison of a) heave RAO, b) pitch RAO for the cargo ship (Fn=0.089, head waves)

Fig. 12. Comparison of a) heave RAO, b) pitch RAO for the cargo ship (Fn=0.1425, head waves)

Fig. 13. Comparison of a) heave RAO, b) pitch RAO for the cargo ship (Fn=0.2138, head waves)

333

Fig. 13. Comparison added resistance for


the cargo ship for varying Froude numbers
in regular head waves
a)Fn=0.089, b)Fn=0.1425, c)Fn=0.2138

The added resistance prediction in the software is performed using the near-field method given by
Faltinsen et al. (1980). The added wave resistances of the cargo ship for varying Froude numbers are
shown in figure 13. Horizontal axis of the graphic is wave length to ship length ratio, and the vertical axis
is non-dimensional added resistance coefficient. This coefficient is given in equation 2:
Non-dimensional added resistance coefficient:

AW

RAW
g a2 ( B 2 / L)

(2)

As seen in figure 13, there is a good agreement between added resistance analyses and experimental
results for the cargo ship model in regular head seas. Peaks in the software curves (green lines) are higher
than the experimental data.

Conclusions
A comparative study of ship motions between the theory and experiment has been presented in this study.
Basic ship motions, derived responses such as vertical and lateral accelerations, added resistance, motion
induced interruption, slamming, propeller emergence, deck wetness, and sea loads, as well are computed
for ships in regular and irregular seas in the software for varying ship speeds, headings and wave
frequencies. However, only heave and pitch regular wave responses, i.e. the transfer functions or response
amplitude operators (RAO) have been calculated and compared to experimental results within the scope of

334

this paper. In addition, prediction of added resistance due to waves in regular head seas have been made
by using the software and compared to the experimental data conducted in in ITU Ata Nutku Laboratory.
Seakeeping analyses have been performed for two different ship types: a series with 60 CB=0.70 form and
a cargo ship which has a fuller form. All the results given in a comparative form were shown in details in
the paper. The intrinsic approach of the software provides fair agreement with experimental results with
respect to all graphics. Finally, reliability of the software to predict the ship motions was tried to present in
the paper.

Acknowledgements
The authors wish to thank Prof. Kadir Sarioz, who shared views and some documents and also Assoc.
Prof. Emin Korkut for his contribution in the experiments.

References
British Maritime Technology (BMT), 2001. ShipmoPC Version 3 User Manual. Revision 10, BMT Fleet Technology
Limited, Canada, pp. 107.
Faltinsen, O. M., Minsaas, K. J., Liapis, N. and Skjordal, S. O., 1980. Prediction of resistance and propulsion of a
ship in seaway. 13th Symposium on Naval Hydrodynamics, Tokyo, 505-529.
Gerritsma, J. ve Beukelman, W., 1966. Comparison of calculated and measured heaving and pitching motions of a
Series 60 ship model in regular longitudinal waves. Laboratorium voor Scheepsbouwkunde, Technische Hogeschool
Delft, Report No 139.
McTaggart, K. A., 1997. Shipmo7: An Updated Strip Theory Program for Predicting Ship Motions and Sea Loads in
Waves. Defence Research Establishment Atlantic, Technical Memorandum 96/243.
Salvesen N., Tuck E. O. and Faltinsen O., 1970. Ship motions and sea loads. Transactions of SNAME 78, 250-287.
Sarioz, K., Kukner, A. and Narli, E., 2000. Validation of a strip theory based ship motion prediction program. ITU
Faculty of Naval Architecture and Ocean Engineering Department of Ocean Engineering, 31-37.
Tezdogan, T., 2011. Investigation of ship motions and application to ships (master thesis, in Turkish)., Istanbul
Technical University Graduate School of Science Engineering and Technology.

335

Effect of Encounter Angle on Parametric Roll Motion in Regular Waves


Emre PEMAN, Metin TAYLAN
stanbul Technical University, Turkey, pesman@itu.edu.tr, taylan@itu.edu.tr

Abstract
This paper examines the influence of encounter angle on parametric roll motion of ships in regular
longitudinal and quartering waves. Parametric roll is known as a dangerous phenomenon typically occurs
in longitudinal waves where wave length is approximately equal to the ship length at an encounter
frequency twice and equal that of the roll natural frequency. A nonlinear roll motion model was set up and
solved by a numerical method in time domain. Maximum roll amplitudes of various sample ships were
calculated with respect to ship speed and encounter angle by using a singledegreeoffreedom model
incorporating heave and pitch effects by means of restoring moment variations. On the other hand,
restoring moment variations in waves with respect to time and instantaneous roll angle was modeled
analytically between the righting moment curves for the wave crest and wave trough conditions. Effect of
encounter angle was also introduced to the model by using a simple function in restoring moment term. It
was shown that the effect of encounter angle strongly depends on encounter frequency and ship hull form.
Results for several ship forms are presented in terms of polar diagrams which may guide ship operators
and designers in order to avoid the dangerous zones where parametric resonance is likely to occur.
Keywords: Parametric resonance, following and quartering waves, encounter angle.

1. Introduction
The phenomenon of parametrically excited roll motion has been known for a long time (Kempf, 1938).
Parametric roll motion have been studied by a number of researchers including Graff, Heckscher (1941),
Kerwin (1955), Paulling and Rosenberg (1959). The first experimental observation of parametric roll was
done by Paulling et al (1972) in SanFrancisco Bay. Although its theoretical existence has been known for
a long time, parametric roll attracted a great deal of interest in recent years because of the incidents that
resulted in loss of lives and money.

In October, 1998, a postPanamax, C11 class containership

encountered extreme weather and sustained extensive loss and damage to deck stowed containers (France
et al, 2003). These casualties urged designers, researchers and regulatory authorities to initiate further
research and investigations. In further studies, researchers such as Spyrou (2000), Neves, Rodrigues
(2006) and Bulian et al (2004) focused on nonlinear aspects and effect of changing frequency range on

337

parametric roll motion. In another aspect, some researchers focused on probabilistic properties of
parametric roll (Shin et al, 2004; Belenky, 2004; Hashimoto et al, 2006; Bulian et al, 2006). Besides
researchers, regulatory and classification societies have also been studying parametric roll. Current IMO
intact stability code has been mainly formed with the static and quasistatic stability rules. The need to
develop stability criteria with the latest achievements in the research of dynamic intact stability physical
mechanisms has been indicated. The IMO working group proposed the framework for the development of
the new criteria on the meetings in 2007 (SLF 50/WP.2, 2007). ABS presented a guide to predict
parametric roll resonance in preliminary design (Shin et al., 2004). The state of the art in methodology
development and regulations for assessment of ship intact stability can be found in Francescutto (2007).
This paper examines the influence of encounter angle on parametric roll motion of ships in regular
longitudinal and quartering waves. For this purpose, various hull forms having different geometries were
investigated for numerous wave conditions. In the present study, parametrically excited roll motion is
modeled as a single degree of freedom system incorporating heave and pitch effects by means of restoring
moment variations. Bulian (2006) approximated GZ surface with polynomial coefficients and Fourier
series. The model presented in this study may be considered as simplified version of the earlier work.
Restoring moment variations in waves with respect to time and instantaneous roll angle was modeled
analytically between the righting moment curves for the wave crest and wave trough conditions. Simply,
the polynomial coefficients of wave crest and wave trough GZ curves and sinusoidal function were
utilized unlike the previous approach. Effect of encounter angle was also introduced to the model by using
a simple function in restoring moment term. Model was set up and solved by DormandPrince Method in
time domain (Dormand & Prince, 1980).

2. Mathematical Model
In general, the equation of roll motion in regular longitudinal waves can be written as follows:

(I xx + I xx ) + B(, ) + GZ ( , t ) = 0

(1)

where:

(I xx + I xx )

: Moment of inertia,

: Roll angle,

B (, )

: Damping function,

GZ ( , t )

: Restoring function.

Eq. (1) may be rewritten as;

338

+ b(, ) +

0 2

GM 0

GZ ( , t ) = 0

(2)

In the above equation, GZ ( , t ) may be approximated by the following expression neglecting surge and
FroudeKrylov forces. In this study, the restoring moment variation is modeled by using only wave crest
and wave trough restoring moment curves.
N

GZ ( , t ) = (m2n1 + k 2n1 cos(e t ) )

2 n 1

(3)

n =1

The coefficients m and k in Eq. (3) are obtained from righting lever curves in wave crest and wave
trough conditions.

m2 n1 =
k 2 n 1 =

c2 n 1,trough + c2 n 1,crest

(4)

c2 n 1,trough c2 n1,crest

(5)

( )

b , = 2 + + 3

(6)

In Eq. (4) and (5), c2n1,crest and c2n1,trough show the coefficients of polynomials fitted to restoring lever
curves in wave trough and wave crest conditions. When the ship is in upright position, she is usually
symmetrically loaded, so the even terms of the polynomial approximation of the righting lever curve
disappear. In this work, seventh degree polynomials are utilized for developing the restoring lever
surfaces. The nonlinear damping term is used and is given in Eq. (6). Coefficients of the damping function
were addressed by Ikeda, Himeno and Tanaka (1978).
Substitution of Eq. (6) and Eq. (3) in Eq. (2) leads to the following differential equation:

+ 2 + + 3 + 0 2

2 n 1
N

(m 2 n 1 + k 2 n 1 cos( e t ) )

n =1

=0
(7)
GM 0

339

Restoring moment term of Eq. (7) was developed to take into account encounter angle by using a simple
function in front of k2n1 coefficient. Restoring moment values in various encounter angles were calculated
by a standard stability program. Results showed that Cos() function is applicable to identify variation of
restoring moment value related to encounter angle (Fig. 1).

Fig. 1. Variation of GZ values related to encounter angle

By adding excitation term and function of encounter angle, the model was generated as follows;

+ 2 + + 3 +

0 2
GM0

(m2n1 + Cos( )k 2n1 cos(et))


n=1

2n 1

H 2

= 0 sin( ) cos(e t )

(8)

3. Sample Ships
In this paper, 8 different ship forms were used. Ships were named as C1, RR1, RR2, F1, D2, D1F, D1W
and D1T. C1 is a PostPanamax C11class containership (France et al., 2003). RR1 is a RoRo ship whose
experimental tests were carried out at the towing tank of DINMA (Bullian, 2006) and INSEAN
Laboratory (Bullian, 2006). RR2 is also a RoRo ship smaller than RR1. Experimental tests of RR2, F1
(frigate) and D2 (destroyer) were carried out at the towing tank of DINMA (Bullian, 2006). D1F, D1W
and D1T are destroyers whose underbodies are similar in size and characteristics as DDG51 but above
water shapes are different. D1F has 10 flare, D1W has a parallel body and D1T has 10 tumblehome
abovewater forms (McCue et al., 2007). Main characteristics and forms of sample ships are given in
Table 1 and Figure 2, respectively.

340

Table 1. Main characteristic of sample ships (meters).


C1
D1F

Type
Container
Destroyer Flared

LBP
262.00
154.00

B
40.00
18.80

T
12.360
5.500

KG
17.550
8.200

D1W

Destroyer Wallsided

154.00

18.80

5.500

8.200

D1T
D2
F1
RR1
RR2

Destroyer Tumblehome
Destroyer
Frigate
RoRo
RoRo

154.00
126.60
120.00
132.22
52.55

18.80
13.65
14.25
19.00
10.00

5.500
3.985
4.060
5.875
2.100

8.200
5.800
6.557
8.660
4.558

C1

RR1

RR2

F1

D2

D1F

D1W

D1T

Fig. 2. Forms of sample ships.

4. Results
The analyses were carried out for 1/30 wave slope for 8 sample ship forms. Roll amplitudes are given in
Fig. 3 as polar diagrams with respect to ship speed and encounter angle.

C1

D2

341

RR1

D1F

D1W

RR2

F1

D1T

Fig. 3. Roll amplitudes with respect to ship speed and encounter angle (H/L: 1/30).

It is observed that risky conditions regarding ships mainly depend upon their geometry and natural
frequency. As it can be seen from Fig. 3; C1, RR1 and F1 forms have large amplitudes; more than 30 in
head waves. These ships encounter parametric resonance in head waves at low speeds for encounter
angles between 0 and 60. C1, RR1 and F1 ship forms also experience parametric resonance in following
waves, but roll amplitudes which occur in higher speeds for encounter angles between 120 and 175, do
not exceed 20. However, an adverse behavior is observed for ship forms, D2, D1F and D1W. These
ships experience parametric resonance in quartering waves between the encounter angles of 100 and
150. Besides, roll amplitudes in head waves do not exceed 15. On the other hand, large amplitude
parametric roll motion is observed in both head waves and quartering waves for RR2 and D1T ship forms.
By using Mathieu equation, stability of the trivial solution, in other words, minimum and maximum

342

critical ship speeds can be determined practically (Mathieu, 1868; Hayashi, 1964; Nayfeh & Mook, 1979).
Stability of the trivial solution of Eq. (8) can be found as follows:
Let, f(t) be a small perturbation of the upright condition 0 . Substituting f(t) in Eq. (8) leads to the
following differential equation governing the evolution of the perturbation:

f + 2f + [ 1 + 1 cos( )] f = 0

(9)

In Eq. (9), since f(t) is supposed to be small, only the linear terms are kept. Eq. (9) can be recognized as a
form of the Mathieu equation. By transforming Eq. (9) into the standard form of the Mathieu equation and
utilizing known characteristics of Mathieu equation, stability of the upright condition in other words
minimum and maximum critical ship speeds are obtained. Following transformation is used to obtain the
standard Mathieu equation.

(10)

After substituting of Eq. (10) into Eq. (9) and dividing both side of equation by , the following equation
is obtained:
d2 f

+ 2

( )]

df
+ 1 + 1 cos f = 0
d

(11)

; 1 = 12 ; 1 = 12
e

(12)

( ))

f + 2 f + p + q cos f = 0

(13)

The following transformation is used to get rid of damping coefficient.

() ()

f = x .e
d2x
d

(14)

( )]

+ p + q cos x = 0 (15)

p = 1

);

q = 1

(16)

InceStrutt diagram is plotted in Fig. 4 by utilizing Floquet method (Floquet, 1883). Risky speeds of first
parametric resonance region (e=2 o) and second parametric resonance region (e=o) are determined by
using InceStrutt diagram and plotted in Fig. 3 for longitudinal waves and nondamped condition. Results
show that parametric roll motion occurs in two regions which are being encounter frequency twice of the
natural frequency (e=2 o) and equal to the natural frequency (e=o). Results also reveal that parametric
roll motion generally occurs in head waves in the neighborhood of e=2 o and occurs in quartering

343

waves in the neighborhood of e=o. Fig. 3 clearly shows that results of nonlinear model and Mathieu
equation agree with each other. In addition, results of nonlinear model are bounded by the results of
Mathieu equation. Without the necessity of solving nonlinear model, critic ship speeds with respect to
encounter angle can be determined by obtaining risky speeds in longitudinal waves by using InceStrutt
diagram as indicated in Eq. (17) and Eq. (18).

Fig. 4. InceStrutt diagram

V critic , head V critic ,longitudin al

V critic , following

V critic ,longitudin al
Cos ( )

0 0 < 60 0 head

(17)

waves

170 0 < 110 0 following waves

(18)

Conclusions
Parametric roll motion is a phenomenon that has to be considered in the preliminary design stage. In the
preliminary design stage, several precautions may be taken to avoid large roll amplitudes caused by
parametric excitation. These precautions are form optimization, increasing damping capability (bilge
keels, etc.), limiting sea state that ship serviced in and choosing service speed out of the critical ship speed
range. In the present study, potential risk regions of parametric roll motion were determined based on the
relationship between ship speed and encounter angle. It is concluded that parametric roll motion occurs in
two regions which are being encounter frequency twice of the natural frequency
(e=2 o) and equal to
the natural frequency (e=o). It is observed that parametric roll motion generally occurs in head waves in
the neighborhood of e=2 o and occurs in quartering waves in the neighborhood of e=o. The most
practical outcome of this study is capability of estimating critical ship speeds by using Eq. (17), Eq. (18)
and critical ship speeds in longitudinal waves which are determined by utilizing InceStrutt diagram.
Finally, using the procedure outlined in this paper may help designers and operators to avoid large roll
amplitudes caused by parametric excitation.

344

References
Belenky, V.L., (2004). On Risk Evaluation at Extreme Seas, Proc. of the 7th Int. Stability Workshop, 188202,
Shanghai, China.
Bullian G., (2006). Development of Analytical Nonlinear Models for Parametric Roll and Hydrostatic Restoring
Variations in Regular and Irregular Waves. University of Trieste Ph. D. Thesis, Trieste.
Bullian,G., Francescutto, A., Lugni, C., (2004). On the nonlinear modeling of parametric rolling in regular and
irregular waves, International Shipbuilding Progress, 51, 205220.
Bullian,G., Francescutto, A., Lugni, C., (2006). Theoretical, Numerical and Experimental Study on the Problem of
Ergodicity and Practical Ergodicity with an Application to Parametric Roll in Longitudianl Long Crested Irregular
Sea, Ocean Engineering, 33: 10071043.
Dormand, J. R.; Prince, P. J., (1980). A family of embedded RungeKutta formulae, Journal of Computational and
Applied Mathematics, 6 (1), 1926
Floquet, G., (1883). Sur les quations diffrentielles linaires coefficients priodiques, Ann. cole Norm. Sup., 12,
4788
France, W.N., Levaduo, M., Treakle, T.W., Paulling, J.R.,Michel, R.K. and Moore, C., (2003). An Investigation of
HeadSea Parametric Rolling and its influence on Container Lashing Systems. Marine Techn,. 40(1), 119.
Francescutto, A., (2007). Intact Stability of Ships Recent Developments and Trends, Proc. of 10th International
Symposium on Practical Design of Ships and Other Floating Structures PRADS07, 1, 487496, Houston.
Graff, W. and E. Heckscher, (1941). Widerstand und Stabilitt Versuche mit Drei Fischdampfer Modellen, Werft
Reederei Hafen, 22: 115120
Hashimoto, H., Umeda N., and Matsuda A., (2006). Experimental and Numerical Study on Parametric Roll of a Post
Panamax Container Ship in Irregual Wave, Proc. of STAB06 9th Int. Conf. on Stability of Ships and Ocean
Vechicles, 181190, Rio de Janeiro, Brazil.
Hayashi, C., (1964). Nonlinear Oscillations in Physical Systems, McGraw Hill, New York.
Ikeda, Y., Himeno, Y. and Tanaka, N., (1978). A Prediction Method for Ship Roll Damping, Report No. 00405 of
Department of Naval Architecture, University of Osaka Prefecture.
IMO SLF 50/4/12, (2007). Review of the IS Code
Kempf, G., (1938). Die Stabilitt Beanspruchung der Schiffe Durch Wellen und Schwingungen, Werft Reederei
Hafen, 19, 200202.
Kerwin, J.E., (1955). Not on Rolling in Longitudinal Waves, Int. Shipbuilding Progress, 2(16), 597614.
Mathieu, E., (1868). Mmoire sur Le Mouvement Vibratoire dune Membrane de forme Elliptique, Journal des
Mathmatiques Pures et Appliques, 137203.
McCue, L.S., Campbell, B.L. and Belknap, W.F., (2007). On the Parametric Resonance of Tumblehome Hullforms
in a Longitudinal Seaway, American Society of Naval Engineers Journal, 3, 3544.
Nayfeh, A.H., Mook, D.T., (1979). Nonlinear Oscillations, John Wiley & Sons, Inc.
Neves, M.A.S. and Rodriguez, C.A., (2006). Influence of nonlinearities on the limits of stability of ships rolling in
head seas, Ocean Engineering, 34, 16181630.
Paulling, J.R. and R.M. Rosenberg, (1959). On unstable ship motions resulting from nonlinear coupling, Journal of
Ship Research, 3, 3646
Paulling, J. R., S. Kastner, and S. Schaffran, (1972). Experimental Studies of Capsizing of Intact Ships in Heavy
Seas. U.S. Coast Guard Technical Report (also IMO Doc. STAB/7, 1973).
Shin,Y.S., Belenky,V.L., Paulling,J.R., Weems,K.M. and Lin,W.M., (2004). Assesment of Parametric Roll
Resonance in the Design of Container Carriers, ABS, 4162.
Spyrou, K.J., (2000). Designing against parametric instability in following seas, Ocean Engineering, 27, 625 653.

345

In this study, the transverse stability of ships in beam seas is analyzed by using both numeric safe
basin concept and Lyapunovs direct method. In the first part, the ultimate stability of ships is
determined by Lyapunovs direct method. In the second part, safe basin of nonlinear rolling motion is
obtained numerically. The nonlinear rolling motion equation is integrated for each initial condition by
using stiff ODE solver till to roll angle exceeds capsizing criterion or to the end of simulation time. If
the roll angle exceeds capsizing criterion, the initial condition is named as unsafe or capsizing initial
condition else it is named as safe initial condition. From the comparisons of analytically and
numerically obtained safe basins, it is observed that they are coherent with each other. It is concluded
that both the numeric safe basin concept and Lyapunovs direct method are useful to analyze the
stability of ships in beam seas.
Ultimate stability, rolling motion, safe basin


Stability against capsizing in beam seas is always popular subject of the researchers due to its major
importance on designing a ship in heavy seas. Different techniques and methods to determine the
stability against capsizing in beam seas are used by researchers in last fifty years. For example,
conditions of stability against capsizing are determined by scaled model tests (Wright and Marshfield,
1980, Grochowalski, 1989 and Cotton and Spyrou, 2001), the approximate solution are obtained by
multiple scale method (Nayfeh and Khdeir, 1986), harmonic balance method (Senjanovic, 1996) and
Bogoulibov Mitropolsky asymptotic method (Cardo, Francescutto and Nabergoj, 1981). The
conditions of stability against capsize of the ships are determined by using the concept of numeric safe
basin (Thompson 1989, Thompson and Soliman, 1991, Ucer and Odabasi, 2008 and Long, Lee and
Kim, 2010), Melnikov Method (Falzarano, 1990), Lyapunov Exponents (McCue, 2004) and Lyapunov
Direct Method (Odabasi, 1978, Ozkan, 1981 and CaldeiraSariava 1986).
A model of ship capsizing based on the application of exciting moment to a rolling ship in beam
irregular seas, due to a group of high waves, applied to the ship at some time instant is developed by
Blocki (1994). In that study, it is shown that the ship performed linear roll motions until the excitation
moment occurred. Probability density function for the initial conditions is then well established,

347

assuming that the two variables are independent one of another, and have Gaussian distributions. Since the
occurrence of a group of high waves, the ship motions are described by a nonlinear SDF equation of
motion. Despite irregular waves, a group of high waves are treated as a regular wave of known circular
frequency, equal to the average frequency, with a determinate excitation.
In this paper, the concepts of numeric and analytic safe basins are used to determine the capsizing
behaviors of BSRA Trawlers in heavy and moderate seas. In the first part, the ultimately bounded safe
basins of roll motion are obtained by using Lyapunov Direct Method (La Salle and Lefschetz, 1961)
for different excitation amplitudes. In the second part, the safe basins of rolling motion are determined
numerically and compared with the analytically obtained safe basins. It is concluded that both numeric
and analytic safe basins are coherent with each other.

In beam seas, rolling motion of the ship has a greater influence on ship stability rather than the other
modes of ship motion. Due to the difficulty of accurately determining the complete hydrodynamic
forces, a rolling model which de couples the six degrees of freedom is generally assumed. In most of
the literature, only roll and sway are considered for the purpose of ship stability analysis (Jiang, 1995).
The two degree of freedom roll and sway model can be reduced to a 1DOF rolling model if a virtual
roll centre is introduced (Hutchison 1991, Jiang 1995 and Balcer 2004).
In this paper, while the rolling motion of the ship is being modeled, interactions between rolling and
other modes of motion are ignored by defining virtual roll centre and the ship is considered to have a
rigid body and seawater is ideal and incompressible. Under these assumptions, rolling motion of a ship
is written as in Eq. (1).

()

I + B + M R () = E(t ) + M sw

(1)

where
is rolling angle with respect to calm sea surface (rad)
is roll angular velocity (rad/s)
I , Virtual moment of inertia (I=Jxx+m44) corresponds to a virtual (physical) axis of rotation, located at

the virtual ship mass centre (the mass centre of the ship along with the added mass in sway), as
discussed by Balcer (2004). The virtual axis of rotation lies commonly below the ship centre of
gravity. Consequently, the added moment of inertia m44 should correspond to the virtual axis of
rotation. Virtual moment of Inertia can be easily obtained from free roll tests.
B( ) is the nonlinear damping moment. It is approximated as follows:
Be=BF+BE+BL+BW+BBK
where Be is equivalent linear damping, BF is friction damping, BE is eddy damping, BL is lift damping
and BBK is bilge keel damping coefficient. These coefficients are determined by semiempirical
348

formulas given by Himeno (1981), BW is wave damping coefficient determined by SHIPMO program
(Beck and Troesch, 1990). Although these coefficients are seemingly linear, their values may vary
with the roll amplitude and the wave frequency (Himeno, 1981) and also the interactions among these
damping components are ignored. For brevity the effect of appendages except for rudder and bilge
keels, is not considered.
MR() is righting moment in calm water and can be represented as follows:
MR()= GZ()
where is the buoyancy force and GZ is the righting arm as a function of the roll angle. GZ can be
approximated by odd polynomials and best fit can be obtained by a seventh order odd polynomial
(Pawlowski, 1987). Therefore, seventh order polynomial is used to fit GZ curve:
GZ() = GM ( + c3 3 + c5 5 + c7 7 )
where is the angle of roll and GM is the initial metacetric height.
E(t) , roll exciting moment is the hydrodynamic moment due to a regular wave. This moment is
calculated according to the linear theory, in which the ship is in an upright position and can be
approximated as follows (Senjanovic et al., 2000).
E(t) = K0 [( k12 m44)2 + ( N44)2]1/2 cost
where is the reduction coefficient for the effective wave slope, k1 stands for the initial metacentric
height GM, K=2/w is the wave number, 0=1/2 hw is the wave amplitude, and is the wave circular
frequency. Assuming that N44 is negligible in relation to k12 m44 and k1=02 I , the above
equation takes the form E(t)=E0 cost, where E 0 = (h w w ) 02 I 2 m 44 is the amplitude of
wave excitation. The wave slope ( hw / w) is taken smaller than 12 due to the real harmonic waves
cannot be arbitrarily steep and high.
Msw, is the steady wind moment defined as follows:
Msw=a Vw2 A Z w /2
where a is the air density, Vw is the velocity of the wind, A is the lateral area including erections and
rigging exposed to the wind (m2), Z is the vertical distance between the centre of wind pressure and the
centre of water pressure: in practice it is taken as the vertical distance between the centre of exposed
area and a point at half draught (m) and w is the wind pressure coefficient; according to full scale
experiments about 1.3 (Wendel, 1967).
In the light of these assumptions and dividing both left and right hand sides of Eq. (1) with virtual
mass moment of inertia, the following equation of nonlinear rolling is obtained:

+ b( ) + g() = E 0 cos t + M sw
I
I

()

where b = B e I = b e and g() =

(2)

4
GZ() = c 2i 1 2i 1
I
I i =1

349

The nonlinear stochastic rolling motion equation, Eq. (2) is written by assuming x1= and x2= as
follows:
x 1 = x 2
x 2 = b e x 2 g(x 1 ) +

(3a)
E0
M
cos t + sw
I
I

(3b)

In Figure 1, M is a closed but otherwise quite arbitrary set of points in n space. For any positive
number r, Mr denotes the set of all points whose distance from M is less than r. Mrc denotes the set of
points outside Mr. If M is the set defined by x R , then Mr is the set R< x <R+r and MrC is the set

x >R+r (Lasalle and Lefschetz, 1961).

Defined sets in n space

In this study, it assumed that the fundamental system is in the form of x = X(x , t ) t 0 and the
following lemmas and theorems (Lasalle and Lefschetz, 1961) are satisfied:
Lemma 1:
V(x,t) is a scaler function with continuous first partials for all x and all t0, and M is a closed set in n

(x , t ) 0 for all x in Mc and if V(x1,t1)<V(x2,t2) for all t2t10, all x1 in M and all x2 in
space. If V
MrC, then each solution of the fundamental system which at some time t00 is in M can never
thereafter leave Mr.
Lemma 2:
(x , t ) < 0 for all t0 and all x in Mc,
If, in addition to the conditions of Lemma 1, V(x,t)0 and V
then each solution of the fundamental system is defined in the future s ultimately inside Mr. (That is,
if x(t) is a solution of the fundamental system defined in the future, then there exists a T such that x(t)
is in Mr, for all tT.
350

Theorem 1:
If in addition to the conditions of Lemma 2 the set M is bounded and V(x,t) uniformly for t>0 as
x then the fundamental system is ultimately bounded.
Theorem 2:
Let V(x) be a scalar function which for all x has continuous first partial derivatives with the property
(x ) for all x outside some closed and bounded set M, then the
that V(x) as x . If V
fundamental system is ultimately bounded.

In order to determine safe basins of nonlinear stochastic rolling motion, Eq. (3a) and Eq. (3b) are
written as Eq. (4a) and (4b) due to the fact that E 0 cos(t ) / I E 0 I (Caldeira Sarieva, 1986 and
Ozkan, 1981).
x 1 = x 2

(4a)

x 2 = b e x 2 g (x 1 ) +

E 0 + M sw
I

(4b)

The dynamic system is assumed to be consisted of two parts. For the positive values of roll angular
velocity (x2), the Lyaponov function shown in Eq. (5a) is used whereas for the negative values Eq.
(5b) is used. Both Lyapunov functions are valid till to the absolute value of the angle of vanishing
stability.
V (x 1 , x 2 ) =

x 22
E + M sw

x 1 F(x 1 )
+ G (x 1 ) c 0
2
I

(5a)

V (x 1 , x 2 ) =

x 22
+ G (x 1 ) d F(x 1 )
2

(5b)

where c and d are positive constant, G(x1) = g(x1) dx1 and F(x1)= be dx1
If constants c and d are taken 2 and roll angle dependency of the damping moment f(x1) is ignored,
Lyapunov functions and their derivatives respect to time become as follows:
x 2 0, V (x 1 , x 2 ) =

x 22
E + M sw

+ G (x 1 ) 2 0
b e x1
2
I

(6a)

x 2 < 0, V (x 1 , x 2 ) =

x 22
+ G (x 1 ) 2 b e x 1
2

(6b)

(x , x ) = 2b E 0 + M sw x b x 2
x 2 0, V
1
2
e
2
e 2
I

(7a)

(x , x ) = 2b + E 0 + M sw x b x 2
x 2 < 0, V
1
2
e
2
e 2
I

(7b)

351

Analytic safe basins are determined by using Lyapunov functions and their derivatives with the
constraints of Lemma 12 and Theorem 12. In figure 1, variation of analytical safe basins of BSRA
trawler (Pattullo and Thomson, 1967) due to F=(E0+Ewind / I) is shown for the ratio of equivalent linear
damping coefficient (Be) to the total mass of inertia (I) is equal to 0.03365. The ratio (Be/I) is
determined by using SHIPMO.BM (Troesch and Beck, 2011) and semiempirical formulas (Himeno,
1981).
2
E + E wind

0, V , =
+ G () 2 0
0.03365
2
I

(8a)

2
< 0, V , =
+ G () 0.0673
2

(8b)

( )

( )

Derivatives of these Lyapunov functions are shown in Eq. (9a) and Eq. (9b) respectively.

( )

, = 0.0673 E 0 + E wind
0, V
I

0.03365 2

, = 0.0673 + E 0 + E wind 0.03365 2


< 0, V
I

( )

(9a)
(9b)

From these equations, it can be concluded that the ratio F=(E0+Msw/I) cannot be greater than 0.0673. In
other words, there is not any closed and bounded set M above 0.0673.

Ultimate bounded regions of rolling motion due to F = (E0+Msw) / I


The stability of nonlinear dynamic systems is very sensitive to the variations of initial conditions. For
example, the small increment of initial roll angular velocity can change the stability of the ship (Ucer
and Helvacolu, 2008). Therefore, thousands of initial conditions should be examined to determine
the stability of the dynamic system (ship). In the numeric safe basin method (Thompson, 1989), the
safe and unsafe initial conditions are represented by white and black points respectively and the effects
of thousands of initial conditions on the stability of the dynamic system (ship) can be shown by using
just one graphic.

352

Safe basin is a set of safe initial conditions defined in the roll angle and roll angular velocity phase
space. In Figure 3, analytic and numeric safe basins of BSRA trawler (GM=0.39m and L/B=5.8) are
presented for F=0.04 and F=0.05 respectively. As can be seen from Figure 3 analytically and
numerically obtained safe basins are coherent with each other.

Comparison of analytic and numeric safe basin (F=0.04 left, F=0.05 right)

The aim of this study is to determine safe basins of nonlinear rolling motion both analytically and
numerically. From the results of this study, the following conclusions are attained:

The analytical safe basins of nonlinear rolling motion are coherent with numerically obtained safe
basins but beside this coherency analytically obtained safe basins are more conservative.

When the sum of magnitude of wind and wave excitation force is greater than 0.0673, there is not
any closed and bounded set of safe initial conditions for sample BSRA trawler.

Both increment of wave and wind force decrease the size of safe basin. In other words it increases
the capsizing probability of the trawler.
Both numeric and analytic method can be used to determine the safe basins of rolling motion and to
develop an intact stability criterion.

Balcer, L., 2004. Location of ship rolling axis. Polish Maritime Research 11(1), 37.
Beck, R.F., Troesch, A.W., 1990. Students documentation and users manual for the computer program
SHIPMO.BM., Department of Naval Architecture and Marine Engineering, University of Michigan, Ann Arbor.
CaldeiraSariava, F., The boundedness of solutions of a Leinard Equation arising in the theory of ship rolling,
IMA Journal of Applied Mathematics, Vol 36, 126139, 1986.
Cotton, B. ve Spyrou, K. J., 2001. An experimental study of nonlinear behavior in roll and capsize. International
Shipbuilding Progress, 481, 518.

353

Falzarano, J. M., 1999. Predicting complicated dynamics leading to vessel capsize. PhD Thesis, University of
Michigan, Ann Arbor.
Grochowalski, S., 1989. Investigation into the Physics of Ship Capsizing by Combined Captive and Free running
Model Tests. Trans. SNAME, 97, 169212.
Himeno, Y., 1981. Prediction of ship roll dampingstate of the art. The University of Michigan, College of
Engineering, Report No. 239.
Hutchison, B.L., 1991. The transverse plane motions of ships. SNAME Marine Technology, 282, 5572.
Hsieh, S.R. et al., A nonlinear probabilistic method for predicting vessel capsizing in random beam seas,
Proceedings of Royal Society London, Vol A446, 117, 1994.
Jiang, C., 1995. Highly Nonlinear Rolling Motion Leading to Capsize. PhD. Thesis, University of Michigan,
Ann Arbor.
Jiang, C. et al., 1996. Highly nonlinear rolling motion of biased ships in random beam seas. Journal of Ship
Research, Vol 40, No 2, 125135.
Lasalle J., Lefschetz, S., 1961. Stability by Liapunovs Direct Method with applications. Academic Press, USA.
Long, Z.Z., Lee, S.K., Kim, J.Y., 2010. Estimation of survival probability for a ship in beam seas using the safe
basin. Ocean Engineering, 37, 418424.
Odabasi, A. Y., 1976. Ultimate stability of ships, Trans. RINA, Vol 118, 237262, 1976.
Odabasi, A. Y., 1978. Conceptual understanding of the stability theory of ships. Schiffstechnik, 25, 118.
Ozkan, I. R., Total practical Stability of ships, Ocean Engineering, Vol 8, 551598, 1981.
Pattullo, A., Thomson, K., 1965. BSRA trawler series (Part 1). Trans. RINA, 107, 215241.
Rainey, R.C.T., Thompson J.M.T., 1991. The transient capsize diagrama new method of quantifying stability
analysis. Journal of Ship Research, 351, 5892.
Soliman, M. S., Thompson J.M.T., 1991. Transient and steady state analysis of capsize phenomena. Applied
Ocean Research, 132, 8292.
Thompson, J.M.T., 1989. Loss of engineering integrity due to the erosion of absolute and transient basin
boundaries. Proceedings of IUTAM Symposium on the Dynamics of Marine Vehicles and Structures in Waves,
313320.
Ucer, E., Odabasi, A.Y., 2008. Significance of roll damping on weather criteria. Trans. RINA International
Journal of Maritime Engineering, 150A1, 18.
er, E., Helvacioglu, I.H., 2008. Methods of safe basins and integrity curves for the nonlinear systems.
Symposium of modern methods in science, Osmangazi University, Eskiehir, pp. 489500, 1517 October.
Wendel, K., 1967. Safety from capsizing in Traung, J.O. Eds., Fishing Boats of the World, USA, pp. 496504.
Wright, J.H.G., Marshfield, W.B, 1980. Ship roll response and capsize behavior in beam seas. Trans. RINA, 122,
129149.

354

Experimental Validation of Viscous Free Surface Flow Computation around


Fast NPL Catamarans at Large Drift Angles
Max Haase, Stefan Winkler, Robert Bronsart, Nikolai Kornev
University of Rostock, Germany, mhaase@utas.edu.au, mail@winkst.de, robert.bronsart@unirostock.de, nikolai.kornev@uni-rostock.de

Abstract
Simulation of ship manoeuvres is one of the most important problems of ship hydrodynamics. The
reliability of manoeuvring prediction depends substantially on the accuracy of the hydrodynamic
models. At present the existing hydrodynamic models are based mostly on approximations of
empirical data obtained for a special series of ships with variation of main geometric parameters.
However, the development of more general tools like CFD (Computational Fluid Dynamics) is
becoming very important, because lately the design of non-standard ships became common business of
the European shipbuilding industry. The focus of the present paper is the validation of viscous flow
computations using RANS-equations (Reynolds Averaged Navier Stokes) performed for catamarans
with NPL hull forms. Steady single-phase computations utilizing a double body concept and unsteady
flow simulations considering a free surface have been conducted at moderate Froude numbers and
large drift angles. To validate these predictions, numerous captive model tests using a twin hull
configuration of NPL hull forms have been undertaken. The computational results have been
experimentally proven being capable of reliably predicting the hydrodynamic forces and moments on
the hull for a wide range of drift angles.
Keywords: Hydrodynamics, captive model tests, RANSE simulation, static drift, twin-hull ships

1. Introduction
One of the major tasks in the manoeuvring theory is to determine the hydrodynamic force and moment
coefficients of ships in oblique flow. Usually a variety of captive and free sailing model tests have to
be conducted to prove the ships manoeuvring abilities. It is known, that model test are time and cost
consuming, and the geometry cannot easily be changed, once the model has been manufactured.
Several empirical approximations are available to estimate the side force and yaw moment coefficients
as a function of geometric ship parameters in dependence on the effective yaw angle. These
approaches developed originally for monohull ships can be extended to catamarans using a correction
factor depending on the distance between two demi-hulls (Mastushkin, 1976). Haase et al. (2010)
showed good agreements between free sailing manoeuvres and predictions using empirical
approximation methods. However, the application of empirical approximations is restricted to hull
forms used in serial measurements. Forces on ships with specific geometry and appendages cannot be
355

predicted with desirable accuracy using existing approximations. The development of more general
tools like CFD is becoming very important, because nowadays design of non-standard ships is getting
usual business of the European shipbuilding industry. The focus of the present paper is the validation
of an universal numerical tool based on RANS-equations. The test calculations are performed for a
fast catamaran with a NPL hull form at large drift angles. Numerical simulations are compared with
measurements done at the Seoul National University Towing Tank (SNUTT).
The work has been done within the joint research project AGAPAS (Autonomously acting Rescue
Robot for Persons in Distress at Sea) aiming at the development of a novel rescue boat using a
catamaran configuration with the total length of L = 4.5 m. The boat has to operate in heavy seas under
conditions of strong wave induced flow motion with velocities comparable with boat speed. Large
effective drift angles are expected to be the common case in operation. Development of the
methodology for estimation of forces under these conditions was the motivation for the present study.
2. Literature Review
2.1 Experiments
Several experiments have been made to investigate force and moment coefficients of ships in steady
oblique motion. Sharma and Zimmermann (1982) investigated the dependency of the lateral force and
yaw moment on the drift angle and the yaw rate for all four quadrants of the horizontal plane, i.e.
= 0 to 180 . They conducted captive Computerised Planar Motion Carriage model tests with a
tanker. As a result they presented the lateral force and yaw moment decomposed in three components
utilizing the potential theory, cross ow drag concept and the wing theory.
Kobayashi and Asai (1987) developed a mathematical model for the prediction of the manoeuvring of
ships at low speeds. The simulated hydrodynamic lateral force and yaw moment at large drift angles,
i.e. = 0 to 180 , have shown a good agreement with measurements for a VLCC tanker model.
Karasuno et al. (1988) proposed a mathematical model for describing the hydrodynamic lateral force
and yaw moment within a wide range of drift angles, i.e. = 0 to 180 . The mathematical
representation for the external hydrodynamic forces and moments acting on the bare hull consists of
three major components. A linear lift force, dominant at small drift angles, a cross flow drag force,
dominant at very large drift angles and a non-linear lift force, dominant at moderate drift angles. The
results for the lateral force and the yaw moment were verified with oblique towing tests of two
monohull fishing vessels at large drift angles. The presented results of the mathematical model showed
an adequate agreement with the model tests.
Umeda and Yamakoshi (1989) conducted circular motion tests with a stern trawler manoeuvring at low
speeds with large drift angles. The developed mathematical model was valid for both, longitudinal
symmetric and non-symmetric ships. They investigated the cross ow and lift forces affected by stall.

356

The derivatives based on measurements at 0 , 90 and 180 showed a good agreement with the
theoretical derivatives derived from slender body theory.
Since the 1980s the scientific attention has more and more shifted towards the motion and wave loads,
e.g. slamming, whipping and springing, and the resistance of high-speed displacement catamarans.
Only a few scientists conducted research on manoeuvrability of fast displacement catamarans, e.g.
Ishiguro et al. (1993), Voulon and Wesselnik (1995), Dand et al. (1999). In these investigations the
manoeuvring problem has been treated as that of 3DOF.
Kaplan (1994) investigated the manoeuvrability and stability of surface effect ships (SES) and
catamarans vessels. He drew attention towards the influence of roll motion and suggested to use a
6DOF mathematical model. Unfortunately, his study mainly focused on small to moderate drift angles.
Among others, Sutulo and Guedes Soares (2005) stated that the manoeuvring motion of displacement
catamarans at high service speeds, generally described only in the horizontal plane (3DOF), depends
on the vertical degrees of freedom, namely sinkage, trim and heel. Based on Computerized Planar
Motion Carriage captive model tests, a relatively complete 6DOF mathematical model for the
manoeuvring of a river-going catamaran equipped with steered waterjets was developed. Standard
manoeuvres were simulated with a 3DOF and the 6DOF mathematical model. The results were
compared with the available full-scale data (Guedes Soares et al. 1999) and showed a clear trend that
the 6DOF model becomes more adequate to use at higher service speeds. Most comprehensive review
of hydrodynamics and the controllability theory of catamarans is presented by Dubrovsky and
Lyakhovitsky (2001).
2.2 Literature Review CFD
First viscous CFD calculations of manoeuvring forces have been done by the end of the 1990's.
Ohmori (1998) developed a finite-volume method to compute viscous flow around a ship in
manoeuvring motion. Both steady conditions such as static drift and steady turning and unsteady
PMM motion were simulated. A block-structured grid was used to discritise the domain, the ReynoldsAverages Navier-Stokes Equation was solved in a transient manner. Two ships were computed for drift
angles of < 10 : fist having U-shaped, the second having V-shaped frames in the aft section.
Convincing results were obtained for the yaw moment. While the V-shaped model shows acceptable
agreement for the transverse force, the U-shaped type shows a significant discrepancy between
experiment and computation.
Alessandrini and Delhommeau (1998) presented a method to compute the viscous flow around a
Series 60 model in steady drift and circular motion. A k- turbulence model and circular block-

357

structured grids have been utilised. Free surface simulation at Fn = 0.32 and drift angle of 5 showed
very good agreement with experimental data.
Maki and Wilson (2008) performed unsteady simulation of the naval combatant DTMB 5415 in
oblique flow motion and presented results in the SIMMAN workshop (Stern and Agdrup, 2008). The
commercial code FLUENT with polyhedral-hexaedral, unstructured grids and the k- turbulence
model was used. The free surface computation were carried out at Fn = 0.28 and drift angle of =
10.Simulations were successfully verified using experiments from the IIHR (Iowa Institute of
Hydraulic Research).
The same ship has been simulated by Miller (2008) using a piso-solver of the CFD code CFDShipIowa. A very fine overset grid with 4.5 106 points for the bare hull and 8.7 106 points for the
appended ship had been used. The turbulence was modelled using Menter's blended k- / k- model.
The simulation had been validated with model experiments from FORCE and IIHR at different model
scales for Fn = 0.28 and the drift angle = 10 . While the bare hull computations reached relative
errors compared to the experiments of up to 10 %, the relative error for the appended ship was
between 13% and 24%. The discrepancy was ascribed to very complex flow field around the
appendages.
Also Sakamoto et al. (2008) calculated the bare hull of DTMB 5415 at Fn = 0,28 and at drift angle of
= 10 and = 20 with the unsteady multi-phase solver of CFDShip-Iowa code. An isotropic
blended k- / k- turbulence model, single-phase level-set free surface modelling and dynamic overset
grid technology had been used. The static drift forces and moments overestimate the measurement data
provided by FORCE, IIHR and INSEAN up to 30%. The discrepancy was explained by complexity of
the flow at large drift angles and the uncertainty of the experimental data.
An interesting study combining numerical simulations and captive model tests had been performed by
Hua-ming et al. (2008) for the Tanker KVLCC2. The experiments were conducted at the towing tank
at SJTU (Shanghai Jiao Tong University) and at NMRI (National Maritime Research Institute). The
simulation had been done using FLUENT with a block-structured grid and SST k- (shear stress
transport) turbulence model. They obtained satisfactory agreement for Fn = 0.06 and Fn = 0.14 for
drift angles up to = 12 . The increase of the Froude number results in better agreement between
numeric and experiments. Also a scattering of experimental data between results obtained on the two
different model test basins was documented.

358

3 Experiments
The model tests have been conducted at the towing tank of the Seoul National University, Republic of
Korea. The particulars of the tank can be taken from table 1.
During static drift experiments the model was travelling, at a constant speed, through the tank in an
oblique flow due to a given drift angle . The towing device can be seen in figure Figure 11. The static
drift angle could be changed from 0 to 180 in increments of one degree. The measurements from
the static runs were used to determine the hydrodynamic derivatives depending on the lateral velocity,
i.e. Xvv, Yv, Yvvv, Nv and Nvvv.

Figure 1: Static drift towing device.

Prior testing the model was ballasted to a given displacement and to zero trim. The forces and
moments were obtained by using a three-component loadcell, located at midships between the ships
deck and the towing device.
All forces and moments have been made dimensionless by density, squared velocity, length and draft.

Table 1: Main particulars of the SNU towing tank.


Parameters
Length

[m]

110

Breadth

[m]

Depth

[m]

3,5

[m/s]

3,5

Max. towing speed

359

Table 2: Main particulars of the tested NPL hull form.


Parameter
Design length between perpendiculars

Lpp

[m]

1,6

Design breadth

[m] 0,15

Design draft

[m] 0,10

Block coefficient

CB

0,40

Prismatic coefficient

CP

0,67

Midship coefficient

CM

0,56

Wetted surface area

Sw

[m] 0,34

LCB

[m] 0,69

Longitudinal center of buoyancy

The NPL 4a with a representative length of 1,6 m was selected for investigations. Main particulars and
a lines plan can be seen in table 2 and figure 2.

Figure 2: Lines plan NPL 4a.

A series of 23 static drift tests with carriage speeds of U = 1.189 m/s and U = 1.585 m/s have been
conducted at drift angles between -20 < < 45 , this correlates with a Froude number of Fn = 0,3
and Fn = 0,4. Table 3 shows the matrix of tests presented in this paper.
Figures (3-5) show hydrodynamic force and moment coefficients for all test cases. Considering the
longitudinal force coefficient, one would expect a graph which is symmetrical with respect to the
ordinate axis. Surprisingly this was not the case, because a positive roll moment appeared at negative
drift angles. That is why only the values corresponding to positive drift angles should be considered
for validation purposes. For the lateral force coefficient a symmetrical graph with respect to the origin
was obtained. After a non-linear growth at moderate drift angles, the increase becomes almost linear
at > 10 . As seen, the lateral force coefficient is independent of the Froude number at drift angles

360

Figure 4: Measured longitudinal force coefficient


versus the drift angle for Fn = 0.3 and Fn = 0.4.

Figure 5: Measured lateral force coefficient versus the


drift angle for Fn = 0.3 and Fn = 0.4.

Figure 3: Experimentally determined yaw moment coefficient versus the drift angle for Fn = 0,3 and 0.4.

less than 20 . At large drift angles the increase of the Froude number results in the decrease of the
lateral force. Similar results were observed for the yaw moment coefficient (Fig.5): a non-linear raise
is followed by almost linear increase of the moment coefficient. Again, values for > 20 are
becoming smaller with increasing Froude number.
Table 3: Test matrix for experimental investigations.
Test

Velocity
[Fn]

Drift Angle [deg]

Static
Drift

0,3

0, 2, 4, 6, 10, 15, 20, 30, 45

0,4

0, 2, 4, 6, 10, 15, 20, 30, 45

4 CFD Simulation of Double Body Flow


Since no significant surface deformation is expected it is advantageous to model the ship flow using
the double body concept according which the submerged body is mirrored with respect to the
unperturbed water surface. The calculation is then performed for the double body in an unbounded
fluid. Generally the double body concept is valid for small Froude numbers Fn < 0.1 0.15. However
this simplification can be quite acceptable for slender body at moderate Froude numbers around 0.3.

361

Calculations were carried out using solver simpleFoam within the open source code OpenFOAM
1.7.1. The solver utilizes pressure correction SIMPLE algorithm (Semi-Implicit Method for PressureLinked Equations) (see Peric and Ferzinger (2001)). The SST k- turbulence model was used.

Figure 5: Grid structure utilised for double body


computation.

Figure 7: Grid structure for free surface flow


computation.

Figure 6: Grid structure utilised for double body


computation.

A grid (figure 6) has been created using OpenFOAM tools, such as blockMesh, snappyHexMesh, or
snapEdge. A coarse resolution in the far field, combined with fine resolution close to the hull has
shown good results using only about 100,000 cells.
Generally, as can be seen in figure 8-10, the experiments and numerical computations agree quite well.
The resistance force agrees quite well for drift angles up to 20 . However, the non-linear character of
the resistance increase observed in measurements was not reproduced numerically at very large drift
angles. The relative discrepancy between experiment and computation is between 5% and 10% for all
cases and forces excepting the longitudinal force at drift angles larger than 30 . Also it is obvious,
that the results for the transverse force agree with measurements better than these for the yaw moment
coefficient. While experimental values for the yaw moment coefficients are generally larger than the
computational one, experimental side force coefficients are slightly smaller compared to the
computation. An exception is the transverse force at a drift angle of 30 . The computations at = 30
362

has been conducted for two velocities corresponding to Froude numbers Fn = 0.3 and Fn = 0.4. There
has been no noticeable difference between the two velocities regarding force and moment coefficients.
Computational results for the drift angle of = 45 diviate significantly from the experiment due to,
perhaps, increasing gravitational (Froude number) effects.

Figu
Fi
re 8: Comparison between experiments and double
gure
9:
Comparison
between
experiments
and
double
body computation for longitudinal force coefficient of
NPL catamaran at different drift angles, V = 1.19 m/s. body computation for lateral force coefficient of NPL
catamaran at different drift angles , V = 1.19 m/s.

Figure 10: Comparison between experiments and double body computation for yaw moment coefficient
of NPL catamaran at different drift angles, V = 1.19 m/s.

CFD Simulation of Free Surface Flow

To prove the effect of the Froude number on manoeuvring forces the calculations have also been
carried out with modelling the free surface. The computations have been performed with a multi-phase
solver interFoam of the OpenFOAM in unsteady mode. For a grid considering the free surface, also
the emerged part of the ship and the domain needs to be modelled. To resolve the free surface
elevation a vertical refinement around the expected free surface has been done. The commercial tool
HexPress (Kleinsorge and Bronsart 2011) has been used to generate the grid consisting of 500,000
cells, as shown in figure 7.
363

Figure 11: Comparison between experiments and free


surface flow computation for the longitudinal force
coefficient of NPL catamaran at different drift angles
and Froude numbers.

Figure 12: Comparison between experiments and free


surface flow computation for lateral force coefficient of
NPL catamaran at different drift angles and Froude
numbers.

Figure 13: Comparison between experiments and free surface flow computation for the yaw moment
coefficient at different drift angles and Froude numbers.

In figure 11 it can be seen, that longitudinal force is well estimated using the free surface computation for
moderate drift angles. At large drift angles the longitudinal force seems to be overestimated. Surprisingly,
the results for the high Froude number of 0.4 agree with measurements better than these for Fn = 0.3
although the strong free surface effects are more pronounced at large Froude numbers and therefore the
modelling errors are more probable.
The discrepancy with measurement for the transverse force given in figure 12 is acceptable at moderate
drift angles and not satisfactory for the large ones. Even the tendency of the force decrease with growing
Froude number is not reproduced. For the yaw moment (figure 13), the agreement between simulations and
experiments is quite good at = 15 at both Froude numbers and for = 30 at the slower speed, while the
results for = 20 and for = 30 at Fn = 0.4 show a significant deviation from experimental ones.
However, for drift angles = 15 the computations considering free surface effects better correlate with
measurements then double body flow simulation.

364

Figure 14 shows the free surface deformation caused by the catamaran. The bow wave is well pronounced
and its appearance agrees well with observations done during the model tests. Figure 15 demonstrates a
strong vortex shedding from the bow section as well as the ventilation of the fore ship section.

Figure 14: Bow wave and free surface elevation


at = 15 and Fn = 0.4.

Figure 15: Vortex structure and ventilation at bow


sections of each demi-hull at = 30 and Fn = 0,4.

Conclusion

PMM model tests of a NPL 4a catamaran with separation ratio of s/L = 0.15 have been undertaken at
Fn = 0.3 and Fn = 0.4. Selected results have been used to validate single-phase and multi-phase CFD
calculations. Numerical simulations of ships in oblique motion show good agreement with
theexperiments for drift angles up to = 20 , which can be considered as a large one for conventional
ships. While computations using the double body concept deliver quite reliable results at moderate
costs, relatively high computational resources are needed for free surface computations. Consideration
of free surface effects results in the improvement of numerical results at the drift angle = 15
especially for the longitudinal force and yaw moment. A large discrepancy at large drift angles is due
to significant free surface deformations which is known to be not properly modelled in CFD. In the
future work, the investigations presented in this paper will be continued for the cases of large drift
angles and high Froude numbers. Special attention will be paid to study of grid dependency and
influence of turbulence models.

Acknowledgements
This research has been performed within the framework of the interdisciplinary project AGAPAS,
funded by the German Federal Ministry of Economics and Technology (BMWi). The authors express
their special thank to Prof. Dr. Eng. Key-Pyo Rhee from the Ship Controllability Laboratory (SCL) at
Seoul National University (SNU) in the Republic of Korea for his outstanding support. Also all staff
members of the Seoul National University Towing Tank (SNUTT) are thanked for the successful
accomplishments of the experimental model tests.

365

References
Alessandrini, B. and Delhommeau, G. 1998. Viscous Free Surface Flow Past a Ship in Drift and in Rotating Motion
Twenty-Second Symposium on Naval Hydrodynamics, Washington, D.C., August 914,
Bronsart, R. and Kleinsorge, L. 2011. CFD Meshing Tools and Their Integration into the Ship CAD Process, Proc. of
RINA ICCAS, Trieste, Italy
Dand, I. W., Dinham-Peren, T. A. and King, L. 1999. Hydrodynamic aspects of a fast catamaran operating in shallow
water, in Proc. International Conference Hydordynamics of High-Speed Craft, RINA, London, UK, p. 17.
Dubrovsky, V. and Lyakhovitsky, A. 2001. Multi-Hull Ships, Backbone Publishing Company, Fair Lawn, United States
of America.
Guedes Soares, C., Sutulo, S., Francisco, R., Santos, F. and Moreira, L. 1999. Full scale measurements of the
manoeuvring capabilities of a catamaran, in Proc. Hydrodynamics of High-Speed Craft, RINA, London, UK, p. 17.
Haase, M., Bronsart, R., Kornev, N. and Nikolakis, D. 2010. Simulation of the dynamics of an autonomously acting
small catamaran for search and rescue process, in Proc. IFAC CAMS, Rostock, Germany
Hua-ming, W., Xi-min, T., Zao-jian, Z. and Bao-shan, W.2008. Experimental and Numerical Researches on the
Viscous Hydrodynamic Forces Acting on a KVLCC2 Model in Oblique Motion, in Proc of SIMMAN workshop, p.
F69, Copenhagen, Denmark
Ishiguro, T., Uchida, K., Manabe, T. and Michida, R. A. 1993. Study on the manoeuvrability of the super slender twin
hull, in Proc. 2nd Int. Conference on Fast Sea Transportation FAST93, Yokohama, Japan, Vol. 1, pp. 283294.
Kaplan, P. 1994. Maneuverability and stability of surface effect ships SES and catamaran vessels, in Trans. RINA,
pp. 1736.
Karasuno, K., Yoneta, K. and Januma, S. 1988. A new mathematical model of hydrodynamic force and moment acting
on a hull in manoeuvring motion at slow speed and oblique direction in Japanese, Journal of the Kansai Society of
Naval Architects, Japan, Vol. 209, pp. 111122.
Kobayashi, E. and Asai, S. 1987. A simulation study on ship manoeuvrability at low speeds, in Proc. International
Conference on Ship Manoeuvrability - Prediction and Achievement, London, UK
Maki, K. J. and Wilson W. M. 2008 Steady Drift force Calculation on the Naval Destroyer Hull 5415, in Proc. of
SIMMAN workshop, p. F34, Copenhagen, Denmark
Mastushkin, Y. M. 1976. Calculating Controllability of Twin-Hull Ships. Shipbuilding no 6,
Miller, R. W. 2008. PMM Calculations for the Bare and Appended DTMB 5415 Using RANS Solver CFDShip-Iowa,
in Proc of SIMMAN workshop, p. F40, Copenhagen, Denmark
Molland, A., Wellicome, J. and Couser, P. 1994. Resistance experiments on a systematic series of high-speed
displacement catamaran forms: Variation of length-displacement ratio and breadth-draught ratio, Technical Report,
Ship Science Report 71, Southampton, England
Ohmori, T. 1998. Finite-volume simulation of flows about a ship in maneuvering motion, in Journal of Marine Science
and Technology, pp 3:82-93
Peric, M. and Ferzinger, J. H. 2001. Computational Fuild Dynamics, Springer Verlag, 2001
Umeda, N. and Yamakoshi, Y. 1989 Hydrodynamic forces acting on a longitudinally non-symmetric ship under
manoeuvring at low speed in Japanese, Journal of the Kansai Society of Naval Architects, Japan, Vol. 211, pp.
127137.
Sakamoto, N., Carrica, P. M. and Stern, F. 2008. URANS Simulation of Static and Dynamic Maneuverng for
Surface Combatant, in Proc of SIMMAN workshop, p. F49, Copenhagen, Denmark
Sharma, S. and Zimmermann, B. 1982. Schrgschlepp- und Drehversuche in vier Quadranten - Teil 2 in German,
Schriftenreihe Schiffbau TUHH, Vol. 423
Stern, F. and Agdrup, K. 2008 SIMMAN - Workshop on Verication and Validation of Ship Manoeuvring
Simulation Methods, Copenhagen, Denmark
Sutulo, S. and Guedes Soares, C. 2005. Simulation of a fast catamarans manoeuvring motion based on a 6DOF
regression model, in Proc. International Conference on Fast Sea Transportation FAST 2005, St Petersburg,
Russia
Voulon, S. and Wesselnik, A. F. 1995. Manoeuvrability of waterjet-propulsed passenger ferries, in Proc. 3rd
International Conference on Fast Sea Transportation FAST95, pp. 11311156, Lbeck, Germany

366

367

As the exact evaluation of vertical relative motions at the bow is of utmost importance for deck
wetness predictions, detailed measurements at theoretical cross sections of no. 18, 19 and 20 have
been implemented at wide range of sea states and speeds of advance.

2. Ship / Model Dimensions and Bow Form Geometry


The models used in the series are listed in Table 1. In selection of ship models for compiling the
test series, variety in hull geometry was intentionally sought. Both U and V bow forms were
included, with or without bulb.
able 1 Vessel features (lengths in meters)
No
1
2
3
4
5
6
7
8
9

Vessel Type
Motor Yacht
70 m Motor Yacht
Container Ship
Container Ship (S175)
19000-20000 tdw Bulk Carrier
Trawler
Patrol Vessel
Frigate
13900 tdw Bulk Carrier

Abbrev.
MY829
MY335
CS
S175
BC640
TR
PV
FR
BC724

Lpp
62.10
66.00
195.40
175.00
159.00
72.60
40.00
122.00
133.00

Scale
1:16.5
1:17.8
1:55.0
1:50.0
1:38.8
1:16.5
1:12.0
1:27.0
1:35.8

Model Lpp
3.764
3.712
3.553
3.500
4.098
4.400
3.333
4.518
3.712

The entrance angle of waterline as well as knuckle angle with reference to water surface are
given in Table 2:
able 2 Bow features of the vessels (angles in degrees)
No

Abbrev.

1
2
3
4
5
6
7
8
9

MY829
MY335
CS
S175
BC640
TR
PV
FR
BC724

WL angle of
enterance
17.5
14.2
14.8
8.5
37.7
16.6
13.6
11.2
26.6

Knuckle angle

Bow form

39.2
42.5
52.1
67.5
61.1
48.4
45.0
41.0
65.4

bulbous
no bulb
bulbous
bulbous
no bulb
bulbous
no bulb
Bulbous
Bulbous

The board flare angles at the bow of considered vessels measured with reference to the center
plane are systematized in Table 3 and Fig. 1.
able 3 Frame definitions
heoretical Frame
20
19
18

V - type Frames
46 - 66
47 - 65
52 - 64

368

U - type Frames
51 71
47 64
58 74

Fig. 1 Sectional flare angles

The local geometry of considered cross sections are consequently given in Table 3:
Table 3 Local geometric features
Type of Ships
MY829
MY335
CS
S175
BC640
TR
PV
FR
BC724

Frame
Type
V
V
U
U
U
U
V
V
U

C p 20th TF

C p 19th TF

C p 18th TF

C b 20-18th TF

0.695
0.000
0.673
0.498
0.000
0.558
0.000
0.265
0.512

0.820
0.520
0.848
0.751
0.685
0.730
0.495
0.663
0.796

0.620
0.635
0.667
0.681
0.858
0.671
0.501
0.585
0.747

0.3795
0.2894
0.3204
0.4247
0.4185
0.4590
0.1301
0.4674
0.3520

Water surface distortion at the bow strongly depends on the local form. The relationship between
local geometry and the statical swell-up has been profoundly investigated in models and
regression dependence drawn, further used for proper evaluation of the effective freeboard.

3. Test Program
All models have been tested fully loaded, heading with three different speeds (see Table 4) in
irregular head seas of three different intensities. Significant wave heights have been chosen so
that deck wetness process to be firmly manifested. For uniformity, wave heights have been taken
as percentage of geometrical free board (GFB) at Frame 19, as follows:
- Low rate of deck wetness expected at Hs = 50 % GFB;
- Intensive deck wetness expected at Hs = 65 % GFB,
- Extreme deck wetness expected at Hs = 80 % GFB.
Pierson-Moskovitz 2-D wave spectrum has been assumed.

369

able 4 The test program envisaged measurements of heave and pitch motions, relative vertical
motions at three bow cross sections, effective freeboard, as well as visual observations of deck
wetness process development, in order to get statistics of exceeding.
Table 4. Vessel speeds in full scale
Vessel type
Hs1
Hs2
Hs3

Vs, [kn]
Fn
Vs, [kn]
Fn
Vs, [kn]
Fn

MY
829

MY
335

0.000
0.000
10.00
0.208
12.50
0.260

0.000
0.000
10.00
0.202
14.00
0.282

BC
640
FULL SCALE SHIP SPEEDS
14.89
0.000
9.00
0.175
0.000
0.117
18.09
15.50
10.00
0.212
0.193
0.130
21.28
22.20
12.50
0.250
0.275
0.163
CS

S 175

TR

PV

FR

BC
724

6.785
0.131
8.321
0.160
9.715
0.187

11.92
0.309
14.46
0.375
17.04
0.442

16.86
0.250
20.46
0.304
24.01
0.357

0.00
0.00
11.0
0.17
13.7
0.16

4. Theoretical Approach
The common probability approach for deck wetness assessment is suitable for compiling general
seakeeping prediction procedures, but is not sensitive to the whole group of bow geometry
parameters influencing the process most, as shown in (Kondrikov, 1989). Analysis of a number of
published data about deck wetness shows considerable differences between numerically predicted
number of green water chancing on deck and values obtained by model or full scale tests (Kishev
and Sirakov, 1987), (Rakitin et al., 1993).
Theoretical procedure applied in this study for deck wetness estimation is based on a modified
probabilistic approach described in details in (ishev and Rakitin, 1997) and (Kishev et al.,
1997). It combines probability of exceeding of effective freeboard with the duration time of
exceeding by introducing a threshold period:
Exp{-*/ }= Nobs / Np

(1)

where Nobs is the number of observed freeboard exceeding; Np is the number of predicted
exceeding.
To increase the accuracy of prediction, effective freeboard is evaluated directly by model tests or
using approximations of systematic model test data. Among all observed evidences of water
chancing on deck, only these with high intensity have been considered.

5. Comments on Results
5.1 Static Swell-up at the Bow
The static swell-up raised around the bow at speed strongly depends on bow geometry. It reduces

370

actual freeboard thus bringing more intensive deck wetness. This phenomenon was thoroughly
investigated by calm water towing of models at various speeds. The water surface elevation along
the board was measured by wire probes and simultaneously filmed, data being compared also
with existing data base.
Sample illustrations of the static swell-up distribution along the bow length in case of V-shaped
or U-shaped bow cross sections are given in Figure 2 and Figure 3.

Fig. 2 V-Shape Bow Frames

Fig. 3 U-Shape Bow Frames

Sample results for freeboard alteration by the static swell-up, including the influence of the
running trim and sinkage, are shown in Figs. 4 and 5 for the case of tested patrol vessel (PV) and
trawler (TR).

Fig. 4 V-Shape Bow Frames

Fig. 5 U-Shape Bow Frames

5.2 Relative Vertical Motions and Deck Level Exceeding


As already mentioned, special attention has been paid to relative motion assessment. Theoretical
predictions have been drawn using linear strip theory and linear spectral theory. Generally good

371

agreement has been found between measured and calculated values, as illustrated in Fig. 6,
showing significant amplitudes of relative vertical motions at 19-th cross section for the trawler
form considered.

Fig. 6 Comparison of predicted and measured significant VRM Amplitudes

For all ship models considered, comparative analysis has been made of theoretically predicted
frequency of deck line exceeding and that observed by model tests or full scale, separating
counts of real water chancing on deck from the total number of exceeding. Sample illustration
of the results is given in Fig. 7.

Fig. 7 Number of exceeding per hour (Containership)


Further, time durations of exceeding at every count have been directly estimated from the time
series of relative motions, extracting the effective freeboard at location. It has been shown, that
statistical distribution of time intervals of water presence on board follows Rayleigh law and is
statistically related to the relative motion amplitudes, as shown on Figs. 8 and 9 for the case of the

372

frigate.
For each ship model, the critical time parameter * (threshold period) has been evaluated on the
basis of experimental observations and calculations. For two of the vessels, comparison has been
made with full scale observations as well. Then, regression analysis has been made of the results
as a function of sensitive geometry parameters. A sample relationship between the threshold
period and the ship length is shown in Fig. 10.

Fig. 8 Duration of Exceeding

Fig. 9 Free Board Exceeding Amplitudes

Fig. 10 Threshold period related to ship length

Similar relations have been drawn for threshold period as a function of local form parameters.
More statistics is however needed for detailed validation and generalization, though reverse self
assessment of the rate of deck wetness occurrence for the ships in consideration shows principal
correctness of the approach.

373

6. Conclusions
Based on analysis of collected experimental data, following conclusions about the influence of
bow forms on spray generation and deck wetness could be made:
- The sharp V-like bow forms characteristic for naval ships (like the frigate and patrol vessel
tested) inflict significant number of deck line exceeding even at low sea states and it increase
with speed and sea intensity. However, real chancing of water on deck is rare even at high
speeds, intensive spray generation is only observed.
- At forms typical for high-speed motor yachts, with large flare angles, the number of exceeding
is considerably less, deck wetness is rare and spray generation is low even at high waves.
- For ships with typical U-shaped bow forms, the number of deck wetness occurrences grows
evenly with the frequency of exceeding. Deck wetness intensity is moderate at operational
speeds and waves of high expectance, but sharply increases in extreme waves.
- The threshold period approach reveals more reliable picture of the deck wetness process and
makes predictions compatible to operational routines. Collection of more data is necessary for
drawing complete relationship of the process to local bow geometry.

References
Kishev R., Sirakov A., 1987, Criteria for Good Seakeeping - a Look from the Bridge, IMAEM 1987.
ishev R., Rakitin V., 1997, Bringing Deck Wetness Predictions in Conformity with Operational Practice - IMAM
1997, Istanbul, Turkey.
Kishev R., Rakitin V., Chalakov V., Maron A., 2005, Experimental Verification Of An Advanced Deck Wetness
Prediction Method - IMAM 05, Lisbon, Portugal.
Kondrikov D., 1989, Slamming Evaluation by Limiting Acceleration of Impact - Annual CMRDI, Leningrad.
Rakitin V., Kishev R. et. al., 1993, Full Scale Observations on Board two 19000 tdw Bulkcarrier Sister Ship IMAM93, Varna.
Rakitin V., Nachev R., 1995, Seakeeping Model Tests of 14000TDW Bulkcarrier - BSHC KP0724 Report, Varna.
Rakitin V., Kishev R., 2005, Development of An Advanced Deck Wetness Prediction Method -BSHC B051105 Report,
Varna.
ITTC 23rd-Recommended Procedures, 2002,Section 7.5-02-02-02; Section 7.5-02-03-01,2; Section 7.5-02-03-02,2.

374

Numerical Simulation of Flow past a Cavity with Vortex Generators


Serkan Trkmen*, Dimitrios Mylonas*, Nurten Vardar**
*

Department of Naval Arch. and Marine Eng., University of Strathclyde, UK, serkan.turkmen@strath,ac,uk,

dimitrios.mylonas@strath.ac.uk
**

Department of Naval Arch and Marine Eng., Yldz Technical University, TR, vardar@yildiz.edu.tr

Abstract
The current paper investigates the effect of vortex generators placed in front of a cavity-like opening
representing a structure such as a bow thruster and examines through parameterisation the effect of the
dimensions of the VG on the flow. Initial stages of the study are presented in this paper, focusing on
incompressible flow over a 2D section of a cavity opening, in view of later on performing 3D
simulations extended to a fore section of a ship. A CFD approach based on the Navier-Stokes
equations is used in the study. Typical results are presented in the form of streamlines and vorticity
contours. The results reveal complex 3D like vertical structures when the vortex generators are
introduced that influence the flow past the cavity.

Keywords: Vortex Generator, cavity, bow thruster, vorticity, laminar flow


1. Introduction
The flow past a ship is a complex and challenging problem because of its nature and characteristics. In
the bow region, breaking waves are generated which influence the free-surface effects and the flow
regime further down the ship. In the stern region, surface tension effects become more important and
the turbulent boundary layer represents a challenge for any design, as the propeller has to operate in a
non-uniform flow behind the ship.
On one hand, improving the wake into the propeller will reduce cavitation, and increase the ships
efficiency; on the other, lengthening the range of laminar regime along the ships boundary layer will
give more streamlining effect thus benefiting the resistance.
In addition, the presence of open-type cavities such as bow thrusters or sea chest, and the appendages
behind the propeller increase the complexity of the problem. The existence of a cavity usually
generates unsteady velocities and pressure fluctuations that can extend further downstream of the hull.
High Reynolds Number incurs flow separation on the surface and oscillations in the shear layer of a
cavity.
Pressure and velocity oscillations due to eddies generated in the shear layer of bow thrusters and sea
chests induce noise and vibration at low speeds, separation in the tunnel, resistance increase to forward

375

motion and cavitation. Current solutions include proper fairing aft of the tunnel apertures and
installation of a protective grillage.
There are a number of devices that can be fitted to the hull of a ship to improve the flow. They are
usually retrofitted or they can be included in the initial design. One of these devices is the vortex
generator (VG). The use of VGs as a way to control the flow is a well-known solution, widely used in
the aircraft and automotive industries mostly for external flows. VGs prevent the flow separation on
the leeward side of an airplanes wing, the vortices introduced to the boundary layer are attaching to
the wing surface. By influencing the boundary layer of the flow, the target of achieving drag reduction
can be obtained.
The application of vortex generators on ships and marine-like bodies is not new and has been studied
over the past decades, focusing mainly on aspects of after body flow realignment and separation. Their
purpose in ship hydrodynamics is to improve the propulsion efficiency of the hull, by increasing the
average flow velocity in the propeller disc, reducing the boundary layer thickness in the stern and
improving the uniformity of the wake flow.
Experimental studies appeared through work from Matheson (1980) on a cargo ship and a bulk carrier,
both at model scale, to look at the effects of VGs on flow field, vibration and manoeuvrability. Both
cases reported improvements in wake flow uniformity and velocity field, helping vibration and
steering issues.
Oledal (1997) carried out an experimental study of VGs for hydraulic-related inlet flows applications
such as waterjets or other propulsive systems. The focus was mainly on cavitation by effects induced
by three different VGs designs through tank observations. A wise choice in the design could lead to
advantages at the inlet flow of the systems.
Brandner and Walker (2003) and Brandner et al. (2010) also investigated the hydrodynamic
performance of VGs through experimental measurements and analysis initially in a cavitation tunnel
then with further application to a waterjet inlet. It was found that the influence of cavitation from the
VGs was not significant at moderate cavitation numbers for the first study and the VGs were found to
have little effect on pressures in the water-jet; though cavitation occurrence on the VGs was found to
present potential performance limitations at low cavitation numbers for the second study.
Johnannsen (2006) reported on the advantages of vortex generators in reducing propeller induced hull
pressure pulses by improving the wake flow into the propeller. It was found that reductions greater
than 50% in all the first four harmonics, both at model scale and full scale could be achieved.
Schmde (2008) looked at numerical RANS-based computations to assess the effectiveness of VGs on
a container ship, both at full and model scale. Computations using several turbulence models showed
that VGs reduce the boundary layer thickness and the areas of separated flows in the propeller disc
region. Carlton (2009) showed by using a system of vortex generators that the flow around the after
body of a container ship could be modified to resolve propeller cavitation induced vibrations.

376

Dymarski and Kraskowski (2010) stepped up from experimental approach and performed numerical
simulations using CFD to look at VGs as part of the propulsive efficiency and improvement of wake
flow of the stern of merchant ships (car-carrier and bulk carrier). It was found that improvement of the
wake flow was obtained for both vessels, but also that resistance largely increased for the car carrier
case. The same authors continued their work (2011) by looking at the optimisation of the configuration
of VGs developed in their previous study. The results showed positive outcome in forming improved
wake flow by vortex generators, but also that there were no gains in propulsion efficiency achieved.
Heinke and Hellwig-Rieck (2011) looked at the scale effects (Reynolds number) of retrofitting
technologies, including VGs fins, on the flow around the appendages and into the propeller of a
container ship. The study highlighted the importance of full-scale simulations when using such
devices, to predict wake field and propeller flow, which can be compromised at model-scale.
Although the presence of VGs can cause some additional drag, they have the potential to control the
flow separation hence reducing the total resistance and pressures on the surface of the hull. Therefore,
as can be seen, the benefits of VGs on ships are numerous.
The scope of this new research initiative is the application of vortex generators in the vicinity of bow
thruster tunnels (laminar region) and sea chest cavities (turbulent region), in order to investigate their
influence on the flow. The aim of the research is to find out if VGs are beneficial to the bow thruster
tunnels, in the same way as protective grids and fairing/scallops of the thruster are. Though only early
stages of the research are presented here, future developments will include influence on acoustic noise
and vibration, resistance and flow characteristics.
The current paper investigates the effect of vortex generators placed in front of a cavity-like opening
representing a structure such as a bow thruster and examines through parameterisation the effect of the
dimensions of the VG on the flow. Initial stages of the study are presented in this paper, focusing on
flow over a 2D section of a thruster opening, in view of later on performing 3D simulations extended
to a fore section of a ship. A CFD approach based on the Navier-Stokes equations is used in the study.
Typical results are presented in the form of pathlines and vorticity contours. The results reveal
complex 3D like vertical structures when the vortex generators are introduced.

2. Numerical Simulations
2.1. Overview of the Simulations
To validate the calculation, the study of Pereira and Sousa (1994, 1995) was used. They performed an
experimental and numerical investigation of the unsteady, incompressible, laminar flow past a
rectangular cavity. The computational domain considered in the current work is similar to that of the
water tunnel experiments performed in their study. To reduce the computational time, a 2D approach
is used. Future developments will involve a more complex 3D computational analysis. Fig. 1 shows
the geometrical configuration of the domain.

377

The cavity has a length L, a height h and the domain a total height of H from the bottom of the cavity
to the top of the domain. The aspect ratio of the cavity is L/h = 2 and the Reynolds number based on
the cavitys height is Reh = 3360. The Reynolds number based on the total height of the domain is ReH
= 23800. the ratio of cavity to total domain height is h/H = 0.142 which is small enough to neglect the
effects of the top wall of the domain. A Cartesian coordinate system is used to define the problem,
with x defining the longitudinal direction and y the vertical direction. The free stream velocity is
defined as U0.

Fig. 1. : Computation domain used in validity simulations

The second stage of the numerical simulations involves the introduction of a vortex generator to the
domain, located in the top left corner of the cavity. The main parameterisation study involved
modifying the VG height and observing its influence on the flow pathlines and vorticity. A semicircular shape for the design of the VGs is used. The height of the VG is linked to the cavity height h
in a manner that the different VGs investigated have a radius of h/4, h/5, h/8, h/10 and h/20
respectively.
2.2. Governing Equations
The flows considered in the current study are assumed to be 2D, fully incompressible with constant
viscosity. Calculation of the flows for the validation case are performed in both steady and unsteady
conditions in the laminar regime. Calculations for the VGs are carried out in unsteady conditions.
A finite volume method approach is used based on the Reynolds Averaged Navier-Stokes equations
(RANS) solved in the domain. The governing equations can be expressed, assuming incompressible
flow, as follows:
For Continuity:

(1)

378

For Momentum:
(2)
Where is the characteristic velocity vector, F is any external body forces (e.g. gravity) applied to
the system, p is the pressure, is the kinematic viscosity and is the fluid density. A pressure Poisson
equation is used to couple the changes in the velocity field with the changes in the pressure field and
satisfying the continuity equation.
2.3. Numerical Model
The computation domain extended to 6h in the longitudinal direction from each side of the cavity and
in the vertical direction. The total height of the domain was set as 7h to reduce the influence of the top
wall in the numerical computations. For the validation study, a coarse, medium and fine mesh were
developed, with a size of 9 400, 27 000 and 78 000 cells respectively. A very fine mesh of 300 000
cells was also created to investigate any potential differences arising in the flow due to increase in
clustered cells. For the unsteady calculations, a Courant number of 0.1 was set to keep a small timestep size. For the VGs calculations, only the coarse mesh runs are presented (9 400 cells). All grids
were generated with clustering of nodes near walls and in the shear layer region. These clustered
nodes account for greater gradients in velocity and pressure expected in these regions.
The inlet of the domain was treated as a velocity input, where are the outflow was defined as a
pressure outlet. The initial pressure is set to its free stream value. The cavity and bottom walls were
treated as non-slip surfaces, whereas the top wall was assumed to have a slip condition as it was found
that it did not influence the flow in the domain.
boundary layer is developed on the fore bottom wall and at the entrance of the cavity. The entire flow
field above the cavity is initialized with the in-flow condition, and the flow field within the cavity is
initialized to zero. A second-order upwind scheme for the convective terms is used along with a
standard pressure interpolation scheme using an implicit SIMPLE pressure-velocity coupling for the
discretization of the problem. Fig. 2 shows the coarse (a) and fine (b) meshes for the normal case and
the mesh used for the VG simulations.

(a)

(b)

Fig. 2. coarse (a) and fine (b) mesh for the validity runs

379

3. Results and Discussion


Fig. 3 shows the mean pathlines corresponding to the simulations for the coarse and fine mesh
compared to the ones obtained from the experiments of Pereira & Sousa (1994). Two large
recirculation eddies can be observed in all simulations. The size and position of the eddies inside the
cavity are in good agreement with the equivalent from the mean flow streamlines of the experiments.
The CFD computations depict well the elongated contour of the main recirculation eddy in the upper
upstream region of the cavity.

Fig.3. Mean flow streamlines (left: exp., middle: coarse, right: fine mesh)

Fig.4 shows the mean pathlines for the very fine mesh in steady (left) and unsteady (right) laminar
flow. We can observe a much more complex pattern of large and small eddies inside the cavity. The
steady calculation exhibits the main eddy in good agreement with the experiments, and the second
larger eddy is correctly represented in the lower part of the cavity whereas near the top edge of the
opening there seem to be another vortex developing. For the unsteady run the we observe small eddies
appearing near the opposite corners of the cavity and at the bottom between the two main eddies. The
main vortices are not properly represented compared to the experiments.

Fig.4. Mean flow streamlines for the very fine mesh; left: steady, right: unsteady

Fig. 5 depicts the instantaneous contours of vorticity at three different times for the cavity without and
with the vortex generators, for each of the different parameterised structures. The simulations were
performed for the unsteady laminar conditions. The unsteady nature of the flow is illustrated, showing the
separation that can occur when the flow is passing the VGs and the edge of the cavity in the shear-layer.

380

t= 3.5 sec

t= 5.25 sec

t= 7.0 sec

No VG

VG h/4

VG h/5

VG h/8

VG h/10

VG h/20

Fig.5. instantaneous contours of vorticity of flow without and with the Vortex Generators.

381

The figure clearly shows the free shear layer oscillations above the cavity and in the downstream
boundary layer. There is a noticeable link between the recirculating flow and the shear layer for all the
VGs. The oscillations are propagated downstream of the cavity for the largest VGs, shear stresses
cause flow instability. For the smallest VGs, the flow inside the cavity is the closest to the original
case, the largest vortex is accurately represented and does not seem to be influence by the presence of
the VGs.

4. Conclusions
In this paper, the initial results of the effects of vortex generators on the laminar, incompressible flow
past open cavities are numerically investigated. Results are compared to experimental studies of
Pereira and Sousa (1994, 1995). The simulations are performed using a 2D CFD approach based on
the solution of the Navier-Stokes equations, in view of extending them to a 3D case. The flow past the
vortex generators exhibit strong three-dimensional nature with complex vertical structures in the shear
layer.

References
Brandner, P.A. and Walker, G. J., 2003, Hydrodynamic Performance of a Vortex Generator, Exp. Therm. Flu.
Sci., Vol. 27, pp. 573-582.
Carlton, J.S., 2009, Ship hydrodynamic propulsion: some contemporary issues of propulsive efficiency,
cavitation and erosion, Lloyds Register Technology Day Proceedings, February 2009.
Brander, P.A., Dawson, E.C. and Walker, G.J., 2010, An Experimental Investigation into the Influence of RampMounted Vortex Generators on the Performance of a Flush Waterjet Inlet, JSR, Vol. 54, No. 3, pp. 209-223.
Dymarski, P. and Kraskowski, M., 2010, Using the vortex generators for improving the wake flow of large ships
preliminary results of CFD analyses, 18th International Conference on Hydrodynamics in Ship Design, Safety
and Operation, Gdansk, Poland.
Dymarski, P. and Kraskowski, M., 2011, Numerical and Experimental Investigation of the Possibility of
Forming the Wake Flow of Large Ships by Using the Vortex Generators, Second International Symposium on
Marine Propulsors, Hamburg, Germany.
Heinke, H-J. and Hellwig-Rieck, K., 2011, Investigation of Scale Effects on Ships with a Wake Equalizing Duct
or with Vortex Generator Fins, Second International Symposium on Marine Propulsors, Hamburg, Germany.
Johannsen, C., 2006, HSVA Prediction confirmed: vortex generator fins reduced the vibration excitation level in
full scale, HSVA NewsWave, the Hamburg Ship Model Basin Newsletter.
Matheson, N., 1980, Experimental Studies to Modify the Velocity in the Wake of Ships Hulls, seventh
Australian Hydraulics and Fluid Mechanics Conference, Brisbane, Australia.
Oledal, M., 1997, Applications of vortex generators is ship propulsion system design, Transactions of the Built
Environment, Vol. 24.
Pereira, J.C.F., Sousa, J.M.M., 1994, Influence of Impingement Edge Geometry on Cavity Flow Oscillations,
AIAA Journal, Vol. 38, No. 8, pp. 1737-1740.
Pereira, J.C.F., Sousa, J.M.M., 1995, Experimental and Numerical Investigations of Flow Oscillations in a
Rectangular Cavity, ASME J. of Flu. Eng., Vol. 117, pp. 68-74.
Schmde, D., 2008, RANS computations for wake improving vortex generators, Proceedings of 11th Numerical
Towing Tank Symposium (NuTTS), Brest, France.

382

Prevention of Parametric Rolling Onboard Fishing Vessels


Cihan Emre Uzunoglu, Srgio Ribeiro e Silva, Jos Luis Mantari, Carlos Guedes Soares
Centre for Marine Technology and Engineering (CENTEC), IST, TU Lisbon, Lisboa, Portugal,
ceuzunoglu@gmail.com, guedess@mar.ist.utl.pt

Abstract
The proneness of a fishing vessel to the occurrence of parametric rolling when sailing in longitudinal
waves is discussed. The susceptibility is addressed based on recently proposed design criteria. The
application is taken further by the introduction of different wave length to ships length ratios. The
vessel is then studied for the severity of the roll motion by means of extended numerical simulations.
These preliminary numerical results are relevant to the development of an assessment tool of a fishing
vessels susceptibility to parametric rolling, which could be used in the preliminary design stage.
Keywords: Parametric rolling, fishing vessel, stability variations in longitudinal waves, GZ curves,
design criteria

1. Introduction
The operational environments of fishing vessels impose considerable difficulties to the crew onboard.
If these conditions are made even more severe by further dynamic instabilities, serious problems that
endanger safety onboard may arise. These consequences underline the need for certain design criteria
with the aim of preventing parametric rolling onboard fishing vessels.
The problem of parametric rolling in fishing vessels was noted by Neves et al (1999) and further
studied by Bulian and Francescutto, (2006). Parametric rolling is a phenomenon that has been studied
more intensively in the recent past and an overview of the existing approaches can be found in Turk et
al (2010).
To avoid the consequence of parametric roll, design measures need to be taken and their effect on the
ship design can be ascertained with existing codes. This can be complemented by decision support
systems that provide advice to the ship master (Rodrigues et al, 2011).
An initial indication of the susceptibility to parametric rolling can be obtained from the variations of
dynamic transverse intact stability of the fishing vessels under longitudinal waves in transit conditions
(Mantari et al 2011a).
This paper considers one of the fishing vessels whose dynamic behavior during fishing operations was
studied by Tello et al. (2011) and (Mantari et al 2011b) aims to study a fishing vessel against the
susceptibility and severity criteria. Two wavelengths to ships length ratios and three loading
383

conditions were studied in order to evaluate reactions of vessel in head seas, under different scenarios.
Firstly, the vessel is checked against the susceptibility criteria similar to the one developed by the
American Bureau of Shipping (ABS, 2008). Then, further examination is carried out to obtain the
severity of the motion by means of numerical simulations.

2. The Parametric Rolling Prediction Model


Ribeiro e Silva and Guedes Soares (2000) have demonstrated that both linear and nonlinear theories
could be used to predict parametric rolling. The linear model is in form of a Mathieu equation and
utilizes the stability variations that were evaluated from the linearised righting arm curves. However
this model has proved to be limited in estimating the ships roll resonance magnitudes since deck
submergence effect could not be included in the calculations. Therefore, a nonlinear model having
three degrees of freedom (DoF) was proposed by the same authors as a replacement. While the new
model has obtained better results, the implicit 3 DoF limitation did not allow calculations for irregular
waves.
To simulate unidirectional long crested irregular wave reactions of the vessel in the time domain, a
more recent model with five degrees of freedom was proposed by Ribeiro e Silva et al. (2005). The
model has been extended to six degrees of freedom to include the surge motion using a semi-empirical
formulation. The code has lately been enhanced to use different models for roll damping estimation to
give refined values of maximum roll amplitudes (Ribeiro e Silva and Guedes Soares, 2008).
As in this study the vessel will first be checked against the susceptibility criteria, the method of
calculation regarding the susceptibility is introduced next.
2.1. Susceptibility Criteria for Parametric Rolling in Regular Waves
The equations that describe the nonlinear dynamics of ship motions, which can be used to study
parametric rolling, are given by;
M 44

A44

in which (M 44

B44

B44

C44

,t

F4 cos

(1)

A44 ) are the added inertia terms (structural + added mass) in respect to the x axis;

is the damping term considering the viscous and wave-making components; C44

, t is

the restoring term, including the wave induced time dependent variations and F4 is the roll exciting
moment.
In the particular case of longitudinal waves, if the starboard and port symmetry of the hull is
considered, it is clear that there will be no external excitation and the condition F4

0 will hold true.

Considering the time-dependent roll restoring moment, due to the instantaneous righting arm, equation
1 may be rewritten as follows:
384

M 44

A44

B44

.g.GZ ( 3 , 4 , 5 ,

a
w

) 0

(2)

g is the ships weight, and GZ ( 3 ,

If the restoring term is further detailed,

instantaneous righting arm lever. The wave amplitude is

a
w

, 5,

a
w

is the wavelength and

) is the
G

is the

relative position between waves and the ships hull.


The righting arm GZ in the equation number (2) can be used to represent the GM values using
available methodologies. This value of GM will change due to different loading conditions and due to
waves and may lead to parametric resonance. In order to check if this will be the case, the roll
equation should first be expressed in form of a Mathieu equation. Then the Ince-Strutt diagram may be
utilized to represent the results. A set of assumptions are required for this process. If the fluctuation of
the wave position along the vessels length is assumed sinusoidal:
(3)
where:
is the mean value of the metacentric height and
is the amplitude of GM change in waves.
Using commertial software for stability calculations, the maximum and minimum GM values included
in equation 3 may be determined for a number of wavelengths and positions. As a result, the following
formulation may be used for the estimation of roll motion:

d2 4
dt 2

d 4
dt

2
( mean )

2
( amp )

cos( et )

(4)

The terms that have been introduced in this equation are as follows:
= linear
44 ( mean )

44 ( amp )

roll damping coefficient

gGM ( mean )
M 44

A44

gGM ( amp )
M 44

A44

mean value of roll frequency in longitudinal waves;

= amplitude

of roll frequency variation in longitudinal waves;

Equation 4 can be transformed into the standard form of the Mathieu equation by the introduction of
non-dimensional coefficients:
e

m
e

m
e

Furthermore, the damping term may be eliminated through the following formulation:

385

a
e

( )e(

( )

(5)

After these calculations, the roll motion in form a Mathieu equation is obtained:

d2
d 2

( p q cos )

(6)

Solution of Mathieu equation may be found in many references and depends on the values of p and q
and its solution can be periodic, increasing or decreasing in nature. The value of p depends on the
square ratio of natural frequency and the excitation frequency and q dictates the amplitude of the
motion:
2
m

(7)

2
a

(8)

On the bounded and unbounded periodic solutions of the Mathieu equation, Shin et al. (2004) stated
that here is a threshold roll damping value for each pair of these Mathieu parameters p and q.
Although this solution is difficult to express in terms of elementary functions, Hayashi (1953) has
suggested the following threshold value:
qk1 1 k32

THayashi

(9)

In this formulation, the coefficients k1 and k3 are calculated utilizing the following formulae:
k1

k3

(10)

1 0.1875q 2

q 2 16

q4

352q 2 1024 p
16q

(11)

The resulting susceptibility criteria are utilized to check the vessels vulnerability to parametric
rolling. It imposes two conditions: a frequency condition formulated in terms of the Mathieu
parameters p and q, and a damping threshold condition.
2.1.1. Frequency Condition
Depending on the p parameters that lead to instabilities of the of the Mathieu equation, the vessel
will be found susceptible to parametric rolling if the parameter is between the limit values of the given
inequality:

LB

RB

LB

1
4

q
2

RB

1
4

q
2

(12)

q2
8

q3 q 4
32 384

(13)
(14)

386

2.1.2. Damping Threshold Condition


Based on the calculations for containerships Shin et al. (2004) have concluded that the damping was
underestimated in Hayashis formulations. They have found another solution by setting the damping
equal to the threshold according to Hayashi (equation 9) and then increasing it until the solution
became bounded. The same procedure can be repeated for fishing vessels and this paper presents the
resulting damping threshold:
( ABS )

qk1k2 1 k32

(15)

In this formulation, a new coefficient, k 2 , is introduced apart from the k1 and k3 coefficients
that have been presented before:

k2

1.002 p

0.16 q

0.759

(16)

2.2. Severity Criteria for Parametric Rolling


Once the vessel is found to be vulnerable to parametric rolling, the severity of the motion should be
determined by using an appropriate numerical model. The mathematical code that has been developed
by Ribeiro e Silva and Guedes Soares, (2008) has been used to evaluate the severity of the roll motion
in this work. The approach utilizes quasi-static hydrodynamic coefficients of added mass and damping
for heave and pitch motions. In calculation of the restoring forces, the submerged body of the vessel is
taken into account and the instantaneous restoring forces as well as the Froude-Krylov forces are
evaluated on each time step. The program utilizes an interface which has been developed by Uzunoglu
(2011) for ease of operation and iteratively estimating the roll damping values.

3. Results
3.1. The Fishing Vessel
The fishing vessel proposed is a Portuguese fishing vessel with the length between perpendiculars of
20 meters. The beam of the vessel is 7.4 meters while the depth is 5.5 meters. The draught of the
vessel is 3.03 meters at design conditions. The vessel has a rounded hull form and lacks a chine.
3.2. The Critical Loading Condition
Previously it was shown that vessels can be very sensitive to the change of wave profiles and large
variations of roll restoring energy as far as partial or total stability failures are concerned (Mantari, et
al 2011a). Taking this into account, this paper utilizes the most critical condition defined as where the
maximum variations occur in the metacentric height. There are three loading conditions that have been
examined in this work. The first condition, designated C1, is the port departure condition with cargo
holds empty (0%) and the consumables full at 100%. The second condition is the C2 condition, the
387

Figure 1 Ince-Strutt diagram showing linear and higher order instability zones

fishing ground departure, with full cargo (100%) and 35 to 50 percent of consumables. The port arrival
condition is the last condition at 100% cargo and 10-20% consumables. The maximum variation of
GM and their respective loading conditions are presented in table1. For this particular fishing vessel,
the critical condition turns out to be the port arrival condition; although the fishing ground departure
condition is also very similar.
Table 1 Maximum variation of GM at different loading conditions (m)
Condition

GM Variation (m)

Port Departure (C1)

0.096

Fishing ground departure (C2)

0.180

Port Arrival (C3)

0.185

3.3. Results of the Susceptibility Criteria


The fishing vessel was studied for the most critical ratios of wave length to ships length which were
suggested by Bulian and Francescutto (2006), to be 1 and 1.6. This differs from the ABS criteria
which suggest a ratio of unity. After the calculation of the limit values that are formed using the p and
q pairs, the results are presented in an Ince-Strutt diagram. The zones inside the boundaries are the
instable zones, therefore if the vessel ends up being in that zone for a wavelength to ships length
ratio, it may be defined as prone to parametric rolling due to the frequency condition. Figure 1
presents such instability zones for the ratios of 1 and 1.6. From this figure, it can be seen that the
vessel is far from the danger zone for the wavelength to ships length ratio of 1 but the ratio 1.6 places
the vessel in danger as it is clearly inside the instability zone far from the borders. The wave steepness
in both cases is 1/20.
In the case that the vessel fails the frequency criteria, the procedure is taken one step forward and it is
checked against the damping criteria. Therefore the damping criteria were considered for the
wavelength to ships length ratio of 1.6. Table 2 presents the complete results for the vessel and shows
that it also fails the suggested damping criteria as the damping value is lower than the threshold value.

388

Table 2 The results of the ABS susceptibility criteria for the fishing vessel

Frequency criteria

Damping criteria

/Lpp

LB

RB

Fail/Pass

k1

k2

k3

Threshold

(ABS)

Fail/Pass

1.0

0.23

0.27

0.25

0.05

Pass

1.0

0.94

-4.77

N/A

0.047

N/A

1.6

0.23

0.27

0.17

0.04

Fail

1.0

1.03

0.37

0.045

0.018

Fail

3.4. Results of the Severity Criteria


The results presented in here are the numerical simulations in critical loading condition and the
wavelength to ships length ratio that the vessel failed to pass the above mentioned criteria. As a
reminder, the wavelength to ships length ratio was 1.6 while the vessel is in port arrival loading
condition. The vessel has failed both the frequency condition and the damping condition in this case.
Apart from the vessel being prone to parametric roll, the severity of the motion is an important factor
in parametric rolling especially for fishing vessels. Due to nature of the onboard equipment, relatively
smaller roll angles that normally would not lead to dire consequences in other types of vessels may
lead serious injuries to the crew in fishing vessels apart from damaging the vessel itself.

The numerical codeFigure


used 1utilizes
close-fit
method
and therefore
the modeling
ResultsFranks
of the parametric
rolling
simulations
for the fishing
vessel of the vessel
can be quickly done by defining around 20 points per section. After the definition of the geometry and

389

introduction of several structural and geometrical properties of the vessel, results describing the
severity of the motion are obtained for a given frequency and speed.
Figure 1 shows these results at 8 knots. The heave and pitch motions are presented to show that the
period of roll motion is two times as long as the period of heave and pitch, demonstrating the
eminence of parametric rolling. The amplitude of the roll motion reaches 30 degrees, indicating severe
danger for the crew if such a scenario is encountered. It develops to its maximum under 2 minutes,
which may not allow enough time to react and take necessary measures.

4. Conclusions
The availability of dependable criteria that will allow early assessment of a vessels proneness to
parametric rolling is important for safety measures. While such criteria have been studied for container
vessels, fishing vessels still need to be more studied. However, due to the nature of their operation,
large roll angles onboard fishing vessels may lead to serious damage to the vessel or injuries to the
crew. For that reason, this paper has applied such criteria to fishing vessels with the addition of
different wave length to ships length ratios. The severity criteria were introduced by means of
numerical simulations. The two different wavelength to ships length ratios that were focused on have
been 1.0 and 1.6 and the results have shown the ratio of 1.6 to be critical as opposed to the ratio of
unity suggested by the ABS criteria for container vessels. This information outlines the importance of
considering different criteria depending on the hull form. The findings of the susceptibility criteria
have been compatible with the numerical code outcome that provided the severity criteria for the
fishing vessel. While both checks indicate a dangerous scenario for the vessel, if would be important
to confirm these results experimentally, in order to validate these results.
5. Acknowledgment
This work has been performed within the project SADEP-Decision support system for the safety of
fishing vessels subjected to waves. The project has been financed by the Foundation for Science and
Technology (Fundao para a Cincia e a Tecnologia), from the Portuguese Ministry of Science and
Technology, under contract PTDC/EME-MFE/75233/2006.

6. References
ABS (American Bureau of Shipping), 2008, Guide for the Assessment of Parametric Roll Resonance in the
Design of Container Carriers, ABS Publications
Bulian, G., Francescutto, A., 2006, Safety and operability of fishing vessels in beam and longitudinal waves
International Journal of Small Craft Technology, London, UK: RINA
Hayashi, Ch., 1953, Forced oscillations in nonlinear systems, Nippon Printing and Publishing Company,
Osaka

390

Mantari, J.L., Ribeiro e Silva, S., Guedes Soares, C., (2011a), Loss of fishing vessels intact stability in
longitudinal waves International Journal of Small Craft Technology, (RINA Transactions Part B) Vol. 153, pp.
23-37.
Mantari, J. L.; Ribeiro e Silva, S., and Guedes Soares, C. (2011b) Intact Stability of Fishing Vessel under Action
of Fishing Gear and Wind, Ocean Engineering, (accepted for publication)
Neves, M. A. S., Perez, N. A., Valerio, L., (1999), Stability of Small Fishing Vessels In Longitudinal Waves,
Ocean Engineering, 26(12):1389-1419.
Ribeiro e Silva, S., Santos, T. and Guedes Soares, C., (2005), Parametrically Excited Roll in Regular and
Irregular Head Seas, International Shipbuilding Progress, 52(1):29-56
Ribeiro e Silva, S.; Guedes Soares, C., (2000), Time Domain Simulation of Parametrically Excited Roll in Head
Seas. Proceedings of the 7th International Conference on Stability of Ships and Ocean Vehicles (STAB'2000);
Renilson, M. (ed) Launceston, Tasmania, Australia. Pages 652-664.
Rodrigues, J. M.; Perera, L. P., and Guedes Soares, C. (2011), Decision support system for the safe operation of
fishing vessels in waves, Proceedings 1st Int. Conf. on Maritime Engineering and Technology, Lisbon (in press)
Shin, Y.S., Belenky, V.L., Paulling, J.R., Weems, K.M., Lin, W.M., (2004), Criteria for parametric roll of large
containerships in longitudinal seas SNAME, Maritime Technology Conference and Expo in Washington, DC,
30 September1 October
Tello, M.; Ribeiro e Silva, S., and Guedes Soares, C. (2011), Seakeeping Performance of Fishing Vessels in
Irregular Waves. Ocean Engineering. 38:763-773.
Turk, A.; Prpic-Oric, J.; Ribeiro e Silva, S., and Guedes Soares, C. (2010), Methods for Estimating Parametric
Rolling, Advanced Ship Design for Pollution Prevention. Guedes Soares, C. and Parunov, J. (Eds.), London,
U.K.: Taylor & Francis Group; pp. 43-54.
Uzunoglu, E., 2011, Numerical and Experimental Study of Parametric rolling of a Container Ship in Waves,
M.Sc. Thesis, Technical University of Lisbon, Instituto Superior Tecnico.

391

CFD Analysis of Supersonic Combustion


A. C. Benim*, K. Kuppa*, P. Wollny*, B. Pfeiffelmann*, A. Al-Halbouni**, H. Rahms**,
A. Giese***
*

Duesseldorf University of Applied Sciences, Germany, alicemal.benim@fh-duesseldorf.de,


kalyan.kuppa@fh-duesseldorf.de, patrick.wollny@fh-duesseldorf.de, bjoern.pfeiffelmann@fh-duesseldorf.de

**

Clyde Bergemann Brinkmann GmbH, Germany, ahmad.al-halbouni@clydebergemann.de,


hendrik.rahms@clydebergemann.de

***

Gaswrme-Institut, Germany, a.giese@gwi-essen.de

Abstract
Steady-state, two-dimensional problems in the area of supersonic combustion are computationally
investigated. Predictions are performed for non-combustion and combusting cases. Different turbulence
and combustion modeling approaches are applied. Results are compared with the measurements found in
literature.

Keywords: CFD, Supersonic, Combustion, Turbulence Modelling, Combustion Modelling


1. Introduction
Supersonic combustion is a vital area of concern not only in the field of Aerodynamics (Segal, 2009), but
also in specialized areas in industry such as Vacuum metallurgy (Mahoney et al, 2008). This paper aims
an initial validation study of CFD based modelling of supersonic combustion, as a part of a sponsored
German national research project, which aims in the development of an industrial burner operating in the
supersonic regime. Numerous investigations in this field have already occurred (Baurle et al, 1998;
Kumaran and Babu, 2009). However, the challenges of a truly satisfying combustion modelling are yet to
be met. The views for the modelling philosophy are bifurcated, as there are investigations supporting the
turbulence influence on the time-averaged reaction rates in the high-speed flows (Baurle et al, 1998;
Kumaran and Babu, 2009), on the other hand there are investigations indicating the importance of
turbulence-chemistry interaction (Engman, 2008; Villasenor et al, 1992; Baurle et al, 1998) In this work, a
detailed validation study of different turbulence and combustion models is conducted and is applied to a
number of test cases from literature and validated by comparisons with the experimental results.

393

2. Modelling
2.1. Formulation and Solution methodology
The computations are implemented in the general-purpose finite volume code Fluent (Ansys Fluent 13.0,
2010). All equations are to be understood as time averaged equations (Durbin and Reif, 2003), adopting a
Reynolds averaging (Durbin and Reif, 2003) for the density and a Favre averaging (Durbin and Reif,
2003) for the remaining dependent variables. In the momentum equations, the gravity is neglected. In the
energy equation, the pressure work and viscous dissipation terms are neglected. Radiation is neglected.
2.1.1. Turbulence Modelling
The turbulence is modelled by the turbulent viscosity approach. The SST model (including
incompressibility corrections) (Ansys Fluent 13.0, 2010; Menter, 1994), the two equation standard k(Ansys Fluent 13.0, 2010; Launder and Spalding, 1974), RNG k- (Ansys Fluent 13.0, 2010; Yakhot and
Orszag, 1986) and the realizable k- model (Ansys Fluent 13.0, 2010; Shih et al, 1992). The one equation
Spalart-Allmaras (S-A) model (Ansys Fluent 13.0, 2010; Spalart et al, 1992) is also tested. Although some
authors suggest the use of modified values (Baurle, 2004), or variable (Xiao et al, 2007) turbulent Prandtl
and Schmidt numbers in supersonic flows, the standard, constant values (Ansys Fluent 13.0, 2010) are
used. For treating the turbulent flow near the wall in wall-bounded flows, the following two approaches
are adopted: Firstly, the near-wall flow is sufficiently resolved, and no-wall functions are applied. . In the
second approach the near-wall layer is not resolved, and non-equilibrium wall-functions are used that are
sensitized to pressure gradient (Kim and Choudhury, 1995). An exception here is the S-A model (Ansys
Fluent 13.0, 2010), since the model was originally designed as a low-Reynolds number model.
2.1.2. Combustion Modelling
In the present computations, hydrogen is used as the fuel. A single-step global reaction scheme for the
oxidation of hydrogen is considered.
Three different approaches are applied in calculating the time-averaged conversion rates of the species,
which are necessary as the source/sink terms of the averaged species transport equations.
Model 1 (M1): The reaction-rates are considered to be purely kinetic based and computed using
Arrhenius kinetics, by directly using the mean temperatures and mean concentrations (Kuo, 2005).
Turbulence influence on the time-averaged reaction rate is completely neglected. This approach is quite
frequently used in modelling supersonic combustion (Baurle et al, 1998; Kumaran and Babu, 2009;
Engman, 2008), assuming that the mixing is too fast at the supersonic speeds for the reactions to be

394

mixing controlled and the time scales are too short to assume equilibrium. The kinetic data for evaluating
the Arrhenius equation, is obtained from Hsu and Jemcov (Hsu and Jemcov, 2000).
Model 2 (M2): The reaction-rates are assumed to be purely controlled by turbulent-mixing, i.e. the
Time-averaged conversion rates are assumed to be controlled purely by the rate of fine-scale mixing, as it
is modelled using the Eddy Dissipation Model (EDM) (Magnussen and Hjertager, 1976).
Model 3 (M3): The time-averaged reaction rates given by the previously mentioned two models are
computed and the minimum of the two is assumed to be the net time-averaged reaction rate, thus,
considering that the reaction may be either mixing or kinetics controlled.
2.1.3. Boundary Conditions
At inlets, the static pressure and stagnation pressure, stagnation temperature and species mass fractions are
specified. For the cases with measured inlet distributions, the profiles are considered in prescribing the
inlet boundary conditions. The turbulence intensity and hydraulic diameters are specified at the inlets. At
outlets, the static pressure is prescribed, which however, wasnt effective for supersonic flow at the exit.
In some of the considered cases, the flames were rather free-jet type and blowing into quite large
domains. In such cases, pressure boundaries were prescribed for segregating the solution domain from the
environment. In defining the position of such boundaries, several trials were performed for ensuring that
they were placed sufficiently far away from the jet/flame and do not practically affect the solution. No-slip
conditions were applied at walls, in combination with the wall-functions approach for the near-wall
turbulence modelling, if the grid wasnt fine enough to sufficiently resolve the near-wall layer, as already
discussed above. Either isothermal or adiabatic conditions were applied for the energy equation at walls.
Zero-gradient boundary conditions applied for the species transport equations at the walls.
2.2. Numerical Modelling
Grid independence studies were performed, the details of which are not included in the paper. The
presented results are grid independent. For the momentum and continuity equations, pressure based
(segregated and coupled) and density based solvers are alternatively used (Ansys Fluent 13.0, 2010). The
performance of the coupled solver is observed to be comparable to that of the density based one, whereas
the segregated schemes observed to be less satisfactory, in general. For the discretization of the convective
terms, a second-order upwind scheme (Ansys Fluent 13.0, 2010) is used for all variables except the
turbulence quantities. For the latter, a first-order upwind scheme is used. For convergence, the residuals
are required to be at least two orders of magnitude smaller than the default settings.

395

3. Results
Identical results were obtained from the Computations with and without wall-functions, confirming the
adequacy of the wall-functions method in the present class of flows. Thus, this point will not be elaborated
further. Similar results delivered by the different k-

(standard, RNG, realizable) models. However, the

realizable k- model showed better predictions. Thus, here, only the results of the realizable k- model
will be presented as the k- prediction (k-eps), along with that of the SST and S-A models. The numerical
domain contained about 50,000 cells, and a grid independent solution was obtained.
3.1. Cases (Pure mixing)
Firstly the test cases here are considered without combustion, in order to validate the capability of the
applied procedures in predicting the Flow field at supersonic/transonic velocities and pure-mixing,
independent from the uncertainties of combustion modelling.
3.1.1. Case WOC-1 (Cutler et al, 2006):
This test case is based on the measurements of Cutler et al. (Cutler et al, 2006). A coaxial-jet assembly is
blowing supersonically a mixture consisting of 95% He and 5% O2 (mole-fractions) through the central jet
and air through the coaxial jet into the environmental air. The both streams have Mach numbers of 1.8 at
the jet exit. The experimentally set total pressures and total temperatures for the axial and coaxial jet were
(628 kPa, 306 K) and (580 kPa, 300 K) respectively (Cutler et al, 2006).
Fig. 1 shows the velocity magnitude and Mach number distributions in the near-field of the jet exit as
predicted by the SST model. Although the jets have the same Ma at the inlet, they have considerably
different velocities due to the different material properties, and thus sound speed. Mixing of the axial and
coaxial streams and the widening of the jet by the ambient air entrainment can be observed.
Fig. 2 shows the radial total temperature profile at x=101mm. Fig. 3 shows the radial distributions of
the Helium molar concentrations at several axial positions. A fair agreement of the predictions with the
measurement is observed. However, the accuracy of the predictions decreased with increasing distance to
the jet exit. The realizable k- model showed a rather better performance than the SST model.

(a)

(b)

Fig. 1. Predicted (SST) detail contours of (a) velocity magnitude, (b) Mach number (WOC-1)

396

Fig. 2. Total temperature profiles at x=101mm (WOC-1)

(a)

(b)

(c)

Fig. 3. He mole fraction profiles at (a) x=28mm, (b) x=81mm, (c) x=151 (WOC-1)

3.1.2. Case WOC-2 (Burrows and Kurlov, 1973):


This test case is based on the measurements of Burrows and Kurkov (Burrows and Kurlov, 1973). An inert
gas mixture (67% N2 and 33%H2O) is flowing supersonically through the main channel above the
hydrogen slit, the Mach number in this case is 2.44. The mixture exits the hydrogen channel at pure sonic
velocity. The inlet stagnation-temperatures are 300K and 1150 K for the hydrogen and inert gas (a profile
was considered for the main channel). The walls have a constant temperature of 298K throughout.
Fig. 4 shows the static pressure contour-plots as predicted by the SST model. An expansion wave
structure can be observed. Through the Profiles of the Mach number and the total temperature shown in
Fig. 5, the predicted and measured traversal profiles at the exit of the test section are compared. An over
prediction by all models is observed, especially in the near-wall region. A slightly better performance is
showed by the SST and k- models in comparison to the Spalart-Allmaras model (Fig. 5). Fig. 6 presents
the profiles of H2 and H2O mole fractions. The SST and k- show very similar predictions as it was also
the case for the Mach number and the total temperature, Fig. 5. Although their quantitative agreement with
the experiments leaves more to be desired, they may still be considered to show a fair qualitative

397

Fig. 4. Predicted contours of static pressure (SST) (WOC-2)

(a)

(b)

Fig. 5. Profiles of (a) Mach number, (b) Total Temperature (WOC-2)

agreement with the experiments. A rather too intensive mixing is predicted by the Spalart-Allmaras model
which shows a comparably less satisfactory performance.
3.2. Cases With Combustion
3.2.1. Case WIC-1 (Cheng et al, 1994):
These measurements are obtained from the experiments of Cheng et al. (Cheng et al, 1994), the co-axial
burner geometry and arrangement are principally very similar to that of the experiments of Cutler et al.
(Cutler et al, 2006) (WOC-1). Pure H2 is used as the fuel. The oxidizer was vitiated air with the molar
composition: 20.1% O2, 54.4%N2, 25.5% H2O. The stagnation pressure and temperature of the oxidizer
and fuel flow were given as (778 kPa, 1750 K) and (213 kPa, 657 K), respectively. The Mach numbers at
the nozzle exit for the co-axial oxidizer and the central fuel jet were 2.0 and 1.0, respectively. Fig. 7 shows
the radial profiles of temperature and O2 mole fraction at x/D=43. Again a fair qualitative agreement
between the measurements and predictions is observed, while there are quantitative discrepancies. M2 and
M3 perform quite similar, where M1 provides a slightly different and a somewhat better prediction.

398

(a)

(b)

Fig. 6. Mole fraction profiles (a) H2, (b) H2O (WOC-2)

3.2.2. Case WIC-2 (Burrows and Kurlov, 1973):


These measurements are obtained from Burrows and Kurkov (Burrows and Kurlov, 1973), who also
performed combustion measurements on the same test rig. The test conditions remain quite similar to case

(a)

(b)

Fig. 7. Profiles at x/D=43 (a) T, (b) O2 mole fraction (WIC-1)

WOC-2, with the difference: vitiated air is now flowing through the main channel with the molar
composition of 20.3% O2, 43.8%N2 and 35.9% H2O and the static temperature of 1250 K. The SST model
is used and different combustion models: M1, M2, M3 are used for the computations.
Fig. 8 shows the mole-fractions at the exit. The radial extension of the reaction zone is under-predicted
by all models, while a fair qualitative agreement with the experiments can be observed. For the mole
fraction, M1 seems to perform slightly better, where M2 and M3 deliver again practically identical results.
Fig. 9 shows the traversal profiles of the Mach number and the total temperature at channel exit. The

399

predictions show a reasonable qualitative agreement with the experiments. Quantitatively, they overpredict the measurements.A very similar Mach number distribution is predicted by the combustion
models. For the total temperature practically identical results are predicted by M2 and M3. M1 predicts a
higher peak value, this is due to the higher static temperature predicted by the model, which on the use of

(a)

(b)

(c)

Fig. 8. Mole fraction profiles (a) H2, (b) O2, (b) H2O (WIC-2)

a much more sophisticated mechanism with a number of inter-mediate reactions, would predict lower
temperatures accommodating for dissociation effects.

(a)

(b)

Fig. 9. Profiles of (a) Mach number, (b) Total Temperature (WIC-2)

4. Conclusions
The numerical simulations of the combusting and non-combusting cases in a supersonic regime are
presented. The two-equation turbulence models show similar overall performance, which is better than

400

that of the single-equation one, for the cases considered. In combustion modeling, M2 and M3 approaches
lead to similar results. The M1 approach predicts not much different but is generally slightly better.

Acknowledgements
We thank Prof. T. S. Cheng, Prof. A. D. Cutler and Prof. R. W. Pitz for making their measured data
available, Prof. V. Babu and Dr. Gerlinger for the technical discussions. We gratefully acknowledge the
financial support by the German Federal Ministry of Economy of Technology for this research project.

References
Ansys Fluent 13.0, 2010. Users Guide, Ansys Germany, Darmstadt, Germany.
Baurle, R. A. Mathur, T., Gruber, M. R., Jackson,K.R., 1998. A numerical and experimental investigation of a
scramjet combustor for hypersonic missile applications, AIAA-1988-3121.
Baurle, R. A., 2004. Modeling of high speed reacting flows: established practices and future challenges,
AIAA 2004-0267.
Burrows, M. C., Kurkov, A. P., 1973. Analytical and experimental study of supersonic combustion of hydrogen
in a vitiated air stream, NASA Technical Memor., NASA TM X 2828.
Cheng, T. S., Wehrmeyer, J. A., Pitz, R. W., Jarret Jr.,O., Northam, G.B., 1994. Raman measurement of mix- ing
and finite-rate chemistry in a supersonic Hydrogen-Air diffusion flame, Combustion and Flame, 99, pp. 157-173.
Cutler, A. D., Diskin, G. S., Drummond, J. P., White, J. A., 2006. Supersonic coaxial jet experiment of
computational fluid dynamics code validation, AIAA Journal, 44, pp. 585-592.
Durbin, P. A. and Reif, B. A. P., 2003. Statistical Theory and Modelling for Turbulent Flows, Wiley, New York.
Engman, E., 2008. Numerical Simulation of Scramjet Combustion, M.Sc. Thesis, Lule University of
Technology, Sweden.
Hsu, K. and Jemcov, A., 2000. Numerical investigation of detonation in premixed hydrogen-air mixture-assessment
of simplified chemical mechanism, AIAA Paper 2000-2478.
Kim, S. E., Choudhury, D., 1995. A near wall treatment using wall functions sensitized to pressure gradient, In:
Separated and Complex Flows ASME FED, 217.
Kumaran, K., Babu, V., 2009. A comparison of the numerical predictions of the supersonic combustion of
Hydrogen using the S-A and SST k- models, Prog. Comput. Fluid Dyn., 9, pp.475-489.
Kuo, K. K., 2005. Principles of Combustion, 2nd Ed., Wiley & Sons, New York.
Launder, B. E., Spalding, D. B., 1974. The numerical computation of turbulent flows, Comput. Meths. Appl. Mech.
Eng.,3,pp.269-289.
Magnussen, B. F., Hjertager, B. H., 1976. On mathematical models of turbulent combustion with special emphasis
on soot formation and combustion, Proc.16th Symp.(Int.. The Comb. Inst., Pittsburgh, PA, pp. 719-729.
Mahoney, W. J., Deneys, A. C., Vardian, G. T., 2008. High Mach number sub-atmospheric, coherent jets and their
application in vacuum metallurgy, Proc. 3rd Int. Conf. on Process Development in Iron and Steelmaking, 8-11 June
2008, Lule, Sweden, 1, pp. 367-376.

401

Menter, F. R., 1994. Two equation eddy viscosity turbulence models for engineering applications, AIAA Journal,
32, pp. 1598-1695.
Segal, C., 2009. The Scramjet Engine:Processes and Characteristics, Cambridge Univ. Press.
Shih, T. -H., Liou, W. W., Shabbir, A., Yang, Z., Zhu, J., 1995. A new k- eddy-viscosity model for high Reynolds
number turbulent flows, Computers and Fluids, 24, pp. 227-238.
Spalart, P., Allmaras, S., 1992. A one-equation turbulence model doe aerodynamic flows, Technical Report AIAA92-0439.
Villasenor, R., J.-Y. Chen., R. W. Pitz., 1992. Interaction between chemical reaction and turbulence in supersonic nonpremixed H2-Air combustion, AIAA Journal, 30, pp. 2552-2554.
Xiao, X., Hassan, H.A., Baurle, R. A., 2007. Modelling scramjet flows with variable turbulent Prandtl and Schhmidt
numbers, AIAA Journal, 45, pp. 1415-1423.
Yakhot, V., Orszag, S.A., 1986. Renormalization group analysis of turbulence: I. basic theory, Journal of
Scientific Computing, 1, pp. 1-51.

402