Proceedings of

Proceedings of ICOE 2009

International
Conference
in Ocean Engineering,
Sub domain based numerical
modeling ofICOE
ship2009
hull propeller interaction
IIT Madras, Chennai, India.
1-5 Feb. 2009

SUB DOMAIN BASED NUMERICAL MODELING OF SHIP HULL PROPELLER

INTERACTION
G.Dhinesh1, V. Anantha Subramaian2 and K. Murali3

Abstract: Viscous flow around the self-propelled hull is investigated at model

scale with RANSE (Reynolds Averaged Navier stokes Equations) solver using
realizable k- turbulence model by coupling the flow around the model hull with
the flow generated by the rotating propeller in time domain. The challenges
involved in coupling of flow past ship hull with the propeller flow, have been
brought out and addressed. The technique used is based on a sliding model method
where the numerical tank is divided into two regions which are modeled separately.
They are the propeller domain and the tank domain respectively. The latter includes
the bulk of the liquid, the hull and the boundary regions of the domain. The grid in
the propeller region rotates with the propeller. The grid in the tank remains
stationary. The two grids slide past each other at a cylindrical interface. In the tank
region the standard conservation equations for mass and momentum are solved. In
the sliding mesh model motions of the propeller region is accounted for by a grid
motion of the propeller domain, and the flow variables are interpolated across a
sliding interface. Comparisons were made between the simulated results and
corresponding experimental results for the hull-propeller interaction effects. It is
shown that the flow field in the self-propelling condition is well reproduced in the
simulation.
Keywords: hull-propeller interaction; sliding mesh method; turbulence modelling.
INTRODUCTION
The hydrodynamics of the ship hull and ship hull-propeller interaction effects for design are
assessed through numerical simulations and / or model tests. It is also well known that in
model scale tests, the inequality of Reynolds number in model and prototype necessitates
extrapolation of results to the prototype by decoupling the viscous component of resistance.
Numerical modeling can conceptually be done on prototype scale, but practical computational
limitations constrain the numerical solution to model scale and this is fairly common today.
Results based on RANSE simulations in model scale are fairly common in current literature.
RANSE modeling requires appropriate choice of turbulent models and interface tracking or
1
Research scholar, Department of Ocean Engineering, Indian Institute of Technology Madras, Chennai 600
036, India, Email: dhineshg80@gmail.com
2
Professor, Department of Ocean Engineering, Indian Institute of Technology Madras, Chennai 600 036,
India, Email: subru@iitm.ac.in
3
Associate Professor, Department of Ocean Engineering, Indian Institute of Technology Madras, Chennai
600 036, India, Email: murali@iitm.ac.in

Proceedings of ICOE 2009

Sub domain based numerical modeling of ship hull propeller interaction

capturing methods, Ferziger and Peric (1999).

The flow in the propeller region is heterogeneous and highly unsteady in nature. The
interactive components of wake fraction and thrust deduction fraction are generally
determined from towing tank model experiments, however such tests give only limited
information. It is known from recent literature that the theoretical estimation of the propulsive
performance of a ship is done by combining various computational methods of potential and
viscous flow. However, in these methods the viscous effect is only dealt with approximately.
Following the progress of Computational fluid dynamics (CFD) technology, numerical
methods based on the Navier-Stokes equations have been applied in the recent past, and some
attempts have been made to solve the self-propelling problem of ships. Stern et al. (1988)
presented a RANSE method for propeller hull interaction in which the propeller effects are
quantified by an equivalent body-force distribution. Tahara et al. (2005) coupled RANS
solver with a propeller-performance program based on infinitely-bladed propeller theory.
MODEL EXPERIMENTS
Model experiments form an important part of the present study. Self propulsion tests were
carried out using the scaled model (1:66).The propeller used was right handed type. The main
particulars of the prototype and corresponding values for model of the vessel chosen for the
study are given in Table 1. The particulars of the propeller are given in Table 2.
Table 1. Particulars of the oil carrier.
Particulars

Prototype

Model

Length overall (LOA), m

Beam (B), m
Depth (D), m
Draft (T), m

332.0
58.0
31.20
20.80

5.03
0.878
0.473
0.315

Design speed (V), m/s

7.87

0.968

Froude number (Fn)

0.14

0.14

Table 2. Particulars of the propeller

Model
Particulars
Prototype
Number of blades, z
4
4
Diameter, D, (m)
9.8
0.15
Pitch, P

6.48

0.1

Disk Area Ratio, AE/AO

0.6

0.6

The experiments were carried out in a towing tank of dimensions 82m 3.2m 2.8m at IIT
Madras. The objective was to compare measurements to validate CFD based numerical

Proceedings of ICOE 2009

Sub domain based numerical modeling of ship hull propeller interaction

predictions. A servo motor (Unimotor, Cussons Technology, U.K.) with a feedback control
drive unit for precise setting of RPM (Rotations per minute), R-31 thrust torque
dynamometers capable of measuring range 100N thrust and 4 N-m torque, (Kempf and
Remmers German design and supplied by Cussons Technology U.K.) were assembled in the
model for the self propulsion tests. A data logger (HP Benchlink) was used to acquire the
thrust developed by the propeller. The physical model was fitted out with stern tube with the
propeller in position at frame as per the design drawing. As in typical model tests, the model
is held in braked condition during the acceleration phase. Using the control drive unit the
propeller was given an initial RPM such that it operates in the right handed direction. When
the carriage reaches the steady speed the brake is released now the model is free and is self
propelled. Depending on the behavior of the model the RPM is adjusted to make the ship
model move with same speed as that of the towing carriage. Thus the RPM and the thrust
generated by the propeller at design speed are obtained. Typical sampling interval of 10
milliseconds was used for acquiring data from thrust torque dynamometer. The thrust
measurements from the experiments have been used for comparison with the numerical
predictions.
THE COMPUTATIONAL MODEL AND APPROACH
The numerical studies are based on model scale in order to directly compare with the results
of model tests. This paper describes a new approach namely, the numerical simulation of a
rotating propeller considering the free surface and turbulence model in predicting the hull
propeller interaction effects. The technique used is based on a sliding model method where
the numerical tank is divided into two regions which are modeled separately. They are the
propeller domain and the tank domains respectively, see figure 1.

1(a) Rectangular domain

around the hull

1(b) Closer view of

1(c) Cylindrical
propeller domain along domain encompassing
with hull and rudder inside
the propeller
the rectangular domain

Fig. 1. Numerical domain for simulating self propulsion condition

The latter includes the bulk of the liquid, the hull and the boundary regions of the domain.
The grid in the propeller region rotates with the propeller. The grid in the tank remains
stationary. The two grids slide past each other at a cylindrical interface. In the tank region the
standard conservation equations for mass and momentum are solved. In the sliding mesh
model motions of the propeller region is accounted for by a grid motion of the propeller
domain, and the flow variables are interpolated across a sliding interface. In this unsteady
problem all interaction effects can be captured with complete fidelity. The numerical studies

Proceedings of ICOE 2009

Sub domain based numerical modeling of ship hull propeller interaction

were conducted on model scale and the results are therefore directly comparable with physical
model studies. The computations were carried out for a model speed corresponding to Fn =
0.14, and Rn = 5.2x106. The computations have been performed using the general purpose
RANSE solver STAR CCM+.
GOVERNING EQUATIONS FLUID FLOW
In the rotating propeller region a modified set of balance equations is solved as given below.
The integral form of the generic conservation equation for the total control volume
bounded by control surface is given by,
dt
dV + (v vb ).ndS = .ndS + q ,V dV
(1)
d











unsteady term

convective term

diffusive term

source term

where t is time, denotes the density of the fluid, v is the fluid velocity vector whose
Cartesian components are ui, vb is the velocity of the CV surface (propeller rotation), and n is
the unit vector normal to the CV face and directed outwards. is the diffusion coefficient,
q ,V is the volumetric source of the conserved scalar quantity.
For incompressible flows, the corresponding equations for mass and momentum conservation
is given by
dt
dV + (v vb ).ndS =0
(2)
d V
S

dt
vdV + v(v vb ).ndS = (T pI ).ndS + bdV
(3)
d V
S
S
V
In the above equations, T is the viscous stress tensor, p is the pressure, I is the unit tensor, and
b represents the body force vector. In Cartesian coordinates the components of the viscous
stress tensor are defined as
u u
2
ij = [ i + j ] ij .v
(4)
x j xi
3
where being the molecular viscosity of the fluid and ij is the Kronecker delta (1, when i=j,
otherwise 0).At the sliding interface a conservative interpolation is used for both mass and
momentum, using a set of fictitious control volumes.
Methods for computation of free surface flows can be grouped into two broad categories,
interface tracking and interface capturing. Interface tracking method adapts the free surface as
the solution progresses, whereas interface capturing method treats the continuum as a mixture
of two species (air and water) In this work, interface capturing method has been used which
is more suitable to analyze complex wave phenomenon and ship geometry. This method
assumes that the molecular level mixing is absent and two species share the same velocity,
pressure and temperature field. In addition to the conservation equation for mass and
momentum, a transport equation for void fraction of the liquid phase c is solved:
dt
cdV + c(v vb ).ndS =0
(5)
d V
S

Proceedings of ICOE 2009

Sub domain based numerical modeling of ship hull propeller interaction

Here the void fraction of the liquid phase c is defined as

V
c= l
(6)
V
where Vl denotes the volume occupied by the liquid phase in a CV with the volume V. The
computational domain extends over the water and air phase; void fraction c is equal to 1 for
CVs filled with water, 0 for CVs filled by air, and between 0 and 1 if one CV is partly filled
with air and water. As it is assumed both fluids share the same velocity and pressure if
0< c <1, both fluids are treated as a single effective fluid whose properties vary in space
according to the volume fraction of each phase,
= 1c + 2 (1 c), = 1c + 2 (1 c)
(7)
where subscript 1 denotes water and 2 denotes air.

TURBULENCE MODELLING
Reynolds Averaged Navier-Stokes equations have the same form as conservation equation for
laminar flow. The only difference being the variables is now replaced by averaged ones by its
average and a fluctuation over time:
ui = ui + ui
(8)
i = i + i
(9)
where the over bar denotes the averaged quantity and the prime denotes the fluctuating part of
a variable.
Reynolds averaging results in an additional term in the momentum equations,
ui u j = (ui + ui )(u j + u j ) = ui u j uiu j

(10)

The last term in the equation uiu j is the Reynolds stress term and it brings more
unknowns than the number of equations available and therefore has to be modelled to close
the system of equations.
The present study uses two equation Realizable k turbulence model. One of the weakness
of the standard k model lies with the equation for the dissipation rate ( ). The wellknown round-jet anomaly (prediction of the spreading rate for axisymmetric jets is poor) is
considered to be mainly due to the modelled dissipation equation. The realizable k model
proposed by Shih et al. (1995) to address these deficiencies of standard k models by
adopting the following:

A new eddy-viscosity formula involving a variable C originally proposed by

Reynold.

A new model equation for dissipation ( ) based on the dynamic equation of the
mean-square vorticity fluctuation.

The term "realizable'' means that the model satisfies certain mathematical constraints on the
normal stresses, consistent with the physics of turbulent flows. This model has been
extensively validated for a wide range of flows including rotating homogeneous shear flows,
free flows including jets and mixing layers, channel and boundary layer flows, and separated
flows. The performance of the model has been found to be substantially better than that of the
standard k model for all the above cases. Especially noteworthy is the fact that the

Proceedings of ICOE 2009

Sub domain based numerical modeling of ship hull propeller interaction

realizable k model resolves the round-jet anomaly; i.e., it predicts the spreading rate for
axisymmetric jets as well as that for planar jets. In the present study the model is tested for
the prediction of drag on surface ship and the thrust on rotating propeller.
The starting point for generating a mesh in STAR CCM+ is an imported surface description
of some kind or a translated surface mesh. In this study hexahedral cells have been used.
Additionally, prism layers have been be included next to wall boundaries around the hull and
the propeller blades. The minimum and maximum cell sizes are 0.012m and 0.25m. Volume
sources using such shapes as rectangles (bricks) have been included to increase the mesh
density locally around the appendages and the free surface region. Along the free surface a
refined region between z = -0.1m and z = 0.15m is defined. Figures 2 and 3 show the mesh
generated on the hull and the propeller

Fig. 2. Mesh generated on the hull surface

Fig. 3. Mesh generated on the propeller

In our approach for simulating the self propulsion condition the domain mesh is composed by a
rectangle and cylinder domain. This simplifies the complexity involved in meshing the region
around the propeller as it can be meshed separately in a small cylindrical domain, which is
imported into the rectangular domain. A fixed external rectangle represents the towing tank, on its
forward surface, inlet condition is imposed, on the aft surface outlet condition is imposed, wall
with slip is used on the lateral surfaces of the domain, wall with no slip boundary condition is
imposed on the surface of the hull and the propeller, while around the cylinder interface condition
is imposed as shown in fig. 4. The number of cells for simulating the towing condition is 0.5
million cells, whereas for that of self propulsion test is 0.72million cells in the rectangular domain
and 0.25million cells in the cylindrical domain totalling it to 0.97 million cells.

Proceedings of ICOE 2009

Sub domain based numerical modeling of ship hull propeller interaction

Flow inlet
Hull (wall
with no slip)
B

A
C

D
Flow outlet

A
B

Wall with no slip on the

shaft and propeller and
interface on
all other boundaries of
the cylindrical domain

Fig. 4. Boundary conditions used for numerical simulation of Self propulsion.

A smaller cylinder containing the propeller is set inside the external rectangular one. At each time
step the internal cylinder is rotated of a small amount, and the propeller with it. Computational
variables are interpolated at the sliding interface of the common surfaces of the internal and
external domain. The mesh inside each block uses hexahedral cells. The Rigid Body Motion
model is used for unsteady simulations in which rigid mesh motions are specified. The mesh
motions typically involve sliding meshes and repeating interfaces. Flux continuity will be
calculated across the interface.

RESULTS
Transient interaction between a ship and its propeller has been modelled using a CFD tool.
The focus of the modelling has been to avoid using different computational domain and
model for the ship and the propeller. Hence a challenging task of coupling the flow around

Proceedings of ICOE 2009

Sub domain based numerical modeling of ship hull propeller interaction

the propeller has been taken up using multiple mesh zones. The results of the investigations
are discussed below. The experimental hydrodynamic investigation consisted of resistance
and propulsion experiments in the towing tank. The analysis of the model test data revealed
that the experimental results were in good agreement with the numerical results. The results
are given in table 3.
Table 3. Validation of the self propulsion scheme
Model speed
Propeller thrust, Hull drag (towed),
t
(m/s)
T (N)
Rt (N)
Experiment
0.968
20.1
14.4
0.283
Numerical
0.968
19.73
14.45
0.276
Figures 5and 6 indicate, thrust on the propeller and the drag on the hull in self propulsion
condition. Figures 5 and 6 represent the simulation in achieving the self propulsion point.
Were the initial RPM of the propeller is set as 400 and the solution is allowed to proceed after
1000 iterations the propeller thrust is 2 N and the augmented drag on the hull is around 19 N,
since there is a mismatch in the values the propeller RPM is varied until the propeller thrust
equals the drag on the hull. For 900 RPM the thrust on the propeller is 19.73N and the
augmented hull drag 19.8N, thus the self propulsion point is achieved.
45
40
Propeller thrust(N)

35
30
25

900RPM

20
15

600RPM

10

400RPM

5
0
0

500

1000

1500

2000

2500

Iteration

Fig. 5. Convergence history of the thrust on the propeller (self propulsion condition)

Proceedings of ICOE 2009

Sub domain based numerical modeling of ship hull propeller interaction

50

Augmented hull drag(N)

45
40
35
30
25
20
15
10
5
0
0

500

1000

1500

2000

2500

Iteration

Fig. 6. Convergence history of the augmented drag on the hull (self propulsion
condition)
Figure 7 depicts the nominal wake and the effective wake at the propeller plane for the oil carrier
considered in this study. While a streamlined flow past the propeller plane is observed in the case
of nominal wake the figure for effective wake shows the effect of the propeller and acceleration
the flow due to propeller action. It is to be noted that the contour plot of velocity indicates that the
liquid sucked upstream of the propeller.

Fig. 7. Nominal wake and effective wake at 0.1D upstream of the propeller plane
Figure 8 shows the qualitative comparison of the axial velocity contours with that of
experimental values for a similar vessel at a downstream side of propeller plane.

Proceedings of ICOE 2009

Sub domain based numerical modeling of ship hull propeller interaction

Ref: Karl et al (2005)

Fig. 8. Comparison of axial velocity contours downstream side of propeller

CONCLUSIONS
Application of the sliding mesh model to numerical simulation for the oil carrier model has
shown promising results. The interaction between the hull and the propeller is taken into
account. The proposed procedure gives the effective inflow to the propeller can be obtained
from simulations which are very important for propeller designers.

REFERENCES
Ferziger, J. H., and Peric, M. 2002. Computational Methods for Fluid Dynamics, 423pp.
Karl, Y., Chao., and Carrica, P. 2005. Numeric propulsion for the KCS container ship,
Proceedings of CFD workshop Tokyo, 483-489.
Stern, F., Kim, H.,T., Patel, V.,C. 1988. A viscous-flow approach to the computation of
propeller-hull interaction. Journal of Ship Research, 32, 246-262
Shih, T., H., Liou, W., W., Shabbir, A., Tang, Z.,and Zhu, J., 1995. A New k-epsilon Eddy
Viscosity Model for High Reynolds Number Turbulent Flows, Computers and Fluids, 24,
227-238.
Tahara, Y., Wilson, R., and Carrica, P., 2005. Comparison of free-surface capturing and
tracking approaches to modern container ship and prognosis for extension to selfpropulsion simulator, Proceedings of CFD workshop Tokyo, 548-555.