You are on page 1of 12

Applied Ocean Research 33 (2011) 309–320

Contents lists available at ScienceDirect

Applied Ocean Research


journal homepage: www.elsevier.com/locate/apor

Computations of self-propulsion free to sink and trim and of motions in head


waves of the KRISO Container Ship (KCS) model
Pablo M. Carrica a,∗ , Huiping Fu b , Frederick Stern a
a
IIHR-Hydroscience and Engineering, The University of Iowa, USA
b
School of NAOCE, Shanghai Jiao Tong University, China

a r t i c l e i n f o a b s t r a c t

Article history: Two computations of the KCS model with motions are presented. Self-propulsion in model scale free to
Received 8 March 2011 sink and trim are studied with the rotating discretized propeller from the Hamburg Model Basin (HSVA)
Received in revised form 27 May 2011 at Fr = 0.26. This case is particularly complex to simulate due to the close proximity of the propeller
Accepted 7 July 2011
to the rudder. The second case involves pitch and heave in regular head waves. Computations were
Available online 30 July 2011
performed with CFDShip-Iowa version 4.5, a RANS/DES CFD code designed for ship hydrodynamics. The
self-propulsion computations were carried out following the procedure described in Carrica et al. [1],
Keywords:
in which a speed controller is used to find the propeller rotational speed that results in the specified
Self-propulsion
Seakeeping
ship velocity. The rate of revolutions n, sinkage, trim, thrust and torque coefficients KT , KQ and resistance
Sinkage and trim coefficient CT(SP) are thus obtained. Comparisons between CFD and EFD show that the rate of revolutions
KCS Container Ship n, thrust and torque coefficients KT and KQ have higher prediction accuracies than sinkage and trim.
Computational fluid dynamics For the simulation of pitch and heave in head waves, the geometry includes KCS hull and rudder under
Dynamic overset grids three conditions with two Froude numbers and three wave length and amplitude combinations. 0th and
1st harmonic amplitudes and 1st harmonic phase are computed for total resistance coefficient CT , heave
motion z and pitch angle . Comparisons between CFD and EFD show that pitch and heave are much better
predicted than the resistance. In both cases comparisons with simulations by other authors presented at
the G2010 CFD Workshop [2] using different CFD methodologies are included.
© 2011 Elsevier Ltd. All rights reserved.

1. Introduction Because CFD solvers are based on more realistic physics, the results
are typically better and require no empirical input. This comes,
Except for a handful of applications, such as resistance, drift and however, with a considerable increase in cost in computational
steady turn, most ship tests require motions. CFD codes able to time and problem setup complexity.
simulate seakeeping and maneuvers need thus capability to pre- In this paper we use CFD to analyze two problems with mov-
dict ship motions and model propellers. Conventional research ing bodies for the same geometry, the KRISO Container Ship (KCS)
methods of ship maneuvering and sea-keeping are based on Sys- developed by the Korean Maritime and Ocean Engineering Research
tems Based Simulation (SBS) and potential flow codes. Both of Institute (MOERI, formerly KRISO) appended with rudder and the
these methods are generally flexible, but they also have limitations. HSVA propeller version. The first problem is the prediction of the
The SBS method is based on integral quantities, i.e. the equa- self-propulsion point free to sink and trim in calm water, with
tions of motion (EOM) are solved based on large sets of force data the gridded propeller rotating at a speed defined by an autopilot
obtained from empirical, experimental or numerical methods. Rel- controller. The second problem is pitch and heave in head waves
ative forces and moments are treated separately, and interaction without propeller.
among hull and appendages is hard to include. Not being a physics- Computations of the KCS model fixed at even keel without rud-
based approach, no information on the flow field is obtained. As der with rotating propeller have been for the first time presented
for ship motions in waves, sea-keeping codes based on poten- by Lübke [3] imposing the rotational speed, and by Carrica et al. [1]
tial theory also have some obvious limitations. They are inviscid computing the self-propulsion point. In the G2010 CFD Workshop
and frequently linearized. Forces due to waves are approximate. [2] the same case with discretized rotating propeller was computed
by 5 groups using commercial software: Greve et al. [4] using CFX,
Bugalski and Hoffman [5] using StarCCM+, Wu et al. [6], Jin et al. [7]
∗ Corresponding author at: 300 South Riverside Dr, Iowa City, IA 52242, USA. and Lee and Rhee [8] using Fluent. In all cases the rotating propeller
Tel.: +1 319 335 6381; fax: +1+319 335 5238. was modeled with sliding grids, with the exception of Carrica et al.
E-mail address: pablo-carrica@uiowa.edu (P.M. Carrica). [1] who use dynamic overset grids.

0141-1187/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apor.2011.07.003
310 P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320

Fig. 1. KCS geometry for pitch and heave in head waves (top) and self-propulsion in calm water (bottom).

One feature of the HSVA KCS model is that the propeller is very iteratively. The overset connectivity is obtained at run time using
close to the rudder, making it difficult to rotate the blades without the code Suggar [15]. Waves are introduced as initial and bound-
conflicting with the rudder grids. This case was also part of G2010, ary conditions, with regular and irregular waves of any direction
but no computations were presented with sliding grids, likely due implemented, including JONSWAP, Bretschneider, Hurricane and
to the complexity of creating a cylindrical grid that can handle the Pierson-Moskowitz spectra. Passive and feedback controllers are
rotating propeller in close proximity to the rudder, and one used implemented for speed, heading, course and roll control. Details of
overset grids, which can naturally handle objects rotating respect mathematical formulation, numerical methods, and implementa-
to each other in close proximity. Two contributions used lifting tion of these features in CFDShip-Iowa version 4 can be seen in the
surface methods as body-force models for the propeller [9,10]. cited references [1,16–19] and the literature therein.
Simulations of KCS pitching and heaving in head waves have
been performed first by Simonsen et al. [11], who obtained good 2.1. Controller
results for motions but poor agreement with the measured forces.
This case was included in G2010 and additional results were con- A PI speed controller is used to act on the propeller RPS to
tributed by El Moctar et al. [12], Manzke and Rung [13] and Akimoto achieve the target speed. The error is defined as the difference
et al. [14], these last authors neglecting the rudder. between the instantaneous ship speed and target speed
In this paper results of self-propulsion and sinkage and trim
for the first problem and pitch, heave and resistance for the sec- eU = Uship − Utarget (1)
ond problem are presented, and compared to data from G2010. In and the instantaneous RPS is
addition to integral results (forces, propeller rotational speed and  t
torque and thrust coefficient) and time histories of resistance and n = P eU + I eU dt (2)
motions required by G2010, the paper includes analysis of the flow 0
fields for these two problems.
where P and I are the proportional and integral constants of the
controller. For details refer to Carrica et al. [1].
2. Mathematical and numerical modeling
3. Geometry and conditions
CFDShip-Iowa version 4.5 solves the RANS/DES equations using
a blended k − ␧/k – ω model for turbulence modeling. The free sur- The KCS is a conceptual modern containership design with a
face is modeled using a single phase level-set approach, in which bulbous bow, conceived to provide data for both exploration of flow
the air/water interface is the zero level set of the level-set distance physics and for CFD validation. Data for pitch, heave, and added
function. The domain is discretized using multiblock/overset struc- resistance are available from Force Technology in Denmark [20], as
tured grids. The overset capabilities are fully dynamic, which allows well as self-propulsion tests, in both cases using the model from
the code to simulate large-amplitude motions. Numerical methods HSVA.
include finite differences discretization, with second- to fourth- Unless otherwise specified all variables are non-
order upwind (for DES) discretization of the convection terms and dimensionalized using a reference velocity U0 taken to be the ship
second-order centered scheme for the viscous terms. The tempo- service speed and a length scale taken to be the length between
ral terms are discretized using a second order backwards Euler perpendiculars LPP . Then two dimensionless numbers define the
scheme. Incompressibility is enforced by a strong pressure/velocity simulation conditions: the  Reynolds number Re = U0 LPP / and the
coupling, achieved using Pressure Implicit Split Operator (PISO) or Froude number Fr = U0 / gLPP , where  is the kinematic viscosity
projection algorithms. The equations of motion are solved only on of water and g is the acceleration of gravity. With these defini-
the water side, while quantities are extended into the air to sat- tions the dimensionless rotational speed, forces and torque are
isfy the proper boundary conditions at the interface. Regular waves n* = n LPP /U0 , F ∗ = F/(U02 LPP
2 ) and M ∗ = M/(U 2 L3 ), respectively,
0 PP
are implemented through initial and boundary conditions. 6DoF where  is the water density. The asterisk superscript to identify
capabilities are enabled in a hierarchy of bodies. The ship (par- the dimensionless variables will be omitted from now on.
ent object) and appendages move together with respect to fixed The KCS geometry is shown in Fig. 1, and details of the dimen-
grids, while propellers, rudders and other control surfaces follow sions and geometrical properties are listed in Table 1. The ship’s
the ship but add motion relative to it (children objects). The fluid reference system follows the customary CFD standard, with the ori-
flow equations are solved in an Earth-fixed inertial reference sys- gin located on the centerline at the bow where it meets the design
tem, while the rigid-body equations are solved in the ship system, water line. The x-axis lies in the center plane and water plane at
so forces and moments are projected appropriately to perform the rest, being the longitudinal axis pointing to stern. The y-axis points
integration of the rigid-body equations of motion, which are solved from port to starboard, and z-axis points upwards. The Earth system
P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320 311

Table 1 Table 3
Geometrical properties of KCS. Conditions for Case 2.

Main particulars Full scale Condition no. C1 C2 C3

Length between perpendiculars LPP (m) 230.0 Speed (m/s) 1.701 1.701 2.151
Length of waterline LWL (m) 232.5 Froude number (Fr) 0.26 0.26 0.33
Maximum beam of waterline BWL (m) 32.2 Reynolds number (Re) 6.52 × 106 6.52 × 106 8.27 × 106
Depth D (m) 19.0 Wave length /LPP 1.15 2.00 1.33
Draft T (m) 10.8 Wave height Hs (mm) 84 146 97
Displacement  (m3 ) 52030
Wetted area w/o rudder SW (m2 ) 9424
Wetted surface area of rudder SR (m2 ) 115.0
Block coefficient (CB) /(LPP BWL T) 0.6505
Midship section coefficient (CM) 0.9849
LCB (%LPP ), fwd+ −1.48
Vertical center of gravity (from keel) KG (m) 7.28
Metacentric height GM (m) 0.60
Moment of inertia Kxx /B 0.40
Moment of inertia Kyy /LPP , Kzz /LPP 0.25
Propeller center, long. location (from FP) x/LPP 0.9825
Propeller center, vert. location (below WL) −z/LPP 0.02913
Service speed (knots) U 24
Froude number Fr 0.26
Scale factor 52.667

is coincident with the ship system at rest at the initial condition,


and is denoted by (Xe , Ye , Ze ).
Two cases are studied. Case 1 is a self-propulsion test at model
point in calm water but free to sink and trim with the rudder
and propeller appended to the hull, corresponding to Case 2.3b in
G2010. Case 2 includes 3 conditions of pitch and heave motions in
head waves with the geometry including only the rudder, corre-
sponding to Case 2.4 in G2010.
Following the experimental conditions for Case 1, self-
propulsion of KCS, the SVP1193 propeller from HSVA is used. The Fig. 2. Mesh on the wall boundaries for the KCS self-propulsion computation.

propeller is a 5 bladed, variable pitch design with diameter 150 mm.


The propeller’s principal particulars are given in Table 2. In this case The background grid is a box that extends from −1 to 2 in x, from −1
the pitch to diameter ratio is 1. The rotational speed of the pro- to 1 in y and from −1 to about 0.2 in z, all in non-dimensional coor-
peller is determined by a PI controller that modifies the propeller dinates. Since Case 2 is symmetric respect to the ship’s centerplane
rotational speed n in revolutions per second (RPS) to achieve the due to the absence of the propeller, only the starboard side of the
target speed. Details of these controllers and how they are used to domain is simulated, that is 0 ≤ y ≤ 1. Refinements blocks are added
achieve the self-propulsion point are provided in Carrica et al. [1]. near the propeller and rudder regions and on the wake to resolve
In this case a small gap between the shaft hub at the hull and the features and vortical structures in those regions, and to allow for
propeller hub is used to allow grids to rotate and still obtain a valid proper overlapping in regions where propeller blades and rudder
overset interpolation. are very close to each other. Appendages are added by use of overset
For Case 2 in head waves three conditions are studied, listed in grids. These grids have a collar surface grid that conforms to the hull,
Table 3. For all conditions ak = 0.0525. Heave and pitch motions are and other surface grids that conform to the appendage itself. From
evaluated at midships (Lpp /2) and the phase is defined such that these surface grids volume grids are generated using hyperbolic
the wave crest is at midships at t = 0. Notice that Condition C3 has solvers provided by the commercial grid generation code Gridgen.
a higher speed than Case 1 and Conditions C1 and C2. Fig. 2 shows the mesh at the hull, propeller and rudder on the wall
Both cases were computed using a k − ␧/k – ω based DES turbu- boundary for Case 1. The grid is comprised of 26 base blocks with
lence model with integration to the wall (no wall functions were a total of 23.2 million points, split for parallel processing in 160
used). domains. This includes 4.4 million points for the hull grids, 9.35
million points for the propeller (see Fig. 3), 0.75 million points for
4. Grids the rudder and 8.7 million points for background and refinements
on the free surface. Basic grids and decomposition information for
Since the code supports dynamic overset grids, the general strat- Case 1 are shown in Table 4.
egy is to generate a background grid which extends to the far field The grid used for Case 2 adds a superstructure to the bow area
and is appropriate to define the boundary conditions on that region. to avoid wet deck condition under heavy pitching, mimicking the

Table 2
Propeller principal particulars.

Diameter (mm) D 150.0 Cordl. at r/R = 0.75 (mm) C0.75 43.6081

Pitch at r/R = 0.7 (mm) P0.7 150.0 Thickness at r/R = 0.75 (mm) T0.75 1.8375
Pitch at r/R = 0.75 (mm) P0.75 148.2095 Pitch ratio P0.7 /D 1.0
Mean pitch (mm) Pmean 150.0722 Mean pitch ratio Pmean /D 1.00048
Area ratio Ae /Ao 0.7 Skew 12.65898◦
Hub ratio dh /D 0.16667 No. of blades z 5
Direction of rotation Right-handed
312 P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320

Table 4
Original grids and decomposition information for Case 1.

Grid Points Processors Moves with ship Rotates with


propeller

Boundary layer starboard 1 241 × 55 × 89 8 Yes No


Boundary layer starboard 2 241 × 58 × 53 5 Yes No
Stern starboard 61 × 48 × 51 1 Yes No
Hub starboard 71 × 41 × 51 1 Yes No
Boundary layer port 1 241 × 55 × 89 8 Yes No
Boundary layer port 2 241 × 58 × 53 5 Yes No
Stern port 61 × 48 × 51 1 Yes No
Hub port 71 × 41 × 51 1 Yes No
Propeller hub 101 × 32 × 91 2 Yes Yes
Propeller blades 5 × 79 × 51 × 74 10 Yes Yes
Propeller tips 5 × 51 × 41 × 71 5 Yes Yes
Rudder starboard 71 × 51 × 81 2 Yes No
Rudder port 71 × 51 × 81 2 Yes No
Refinement propeller 211 × 161 × 201 48 Yes Yes
Refinement rudder 57 × 55 × 48 1 Yes No
Refinement wake 101 × 113 × 101 8 No No
Refinement 241 × 189 × 101 32 No No
Background 226 × 141 × 91 20 No No

Total 23,182,004 160

experimental setup (see Fig. 1). This results in a total of 8.0 million the ship but also rotates at the rotational speed imposed by the
grid points, with 2.2 corresponding to the hull grids, 0.3 million for controller.
the flared superstructure, 0.3 million for the rudder and 3.47 million
for refinements and 1.1 million for the background, for a total of 10
5. Results and discussion
base blocks split into 55 processors. Basic grids and decomposition
information for Case 2 are shown in Table 5.
Results are presented for Cases 1 and 2 on a single grid and time
In this grid system we can identify three grid levels. The back-
step, chosen based on experience from previous calculations. Due
ground and refinement grids are not subject to ship motions, but
to the high computational cost verification and validation (V&V)
follow the ship by surging with it (only for Case 1, they move at con-
studies are not performed. The cost of refining and coarsening is
stant speed with the ship for Case 2). Since the background grid is
specially demanding on overset approaches, as discussed in Car-
refined about the free surface, this assures that a reasonable refine-
rica et al. [17]. The only CFD V&V analysis for self-propulsion with
ment is available everywhere independently of the motions that
discretized propeller known to the authors has been performed by
the ship undergoes. All the other grids form the ship object and
Bugalski and Hoffman [5] for the unappended KCS model ignoring
move with it following the surge, pitching and heaving motions
the free surface. V&V studies of the pitch and heave problem are
computed for the object. The propeller is a child object and follows
more common, see for instance the analysis for the surface com-
batant model DTMB 5415 performed by Carrica et al. [17]. In these
two V&V studies the grids are coarser than those used in this study,
giving some confidence that the grid size is appropriate for these
types of computations.

5.1. Case 1: KCS self-propulsion at model point

For this case the solution is computed on the Earth reference sys-
tem, so all the grids actually advance forward as the computation
progresses, but the refinement grids for the wake and free sur-
face and the background grid are restrained from pitch and heave
motions.

Table 5
Original grids and decomposition information for Case 2.

Grid Points Processors Moves with


ship

Boundary layer 1 241 × 55 × 89 8 Yes


Boundary layer 2 241 × 58 × 53 5 Yes
Flared superstructure 95 × 60 × 52 2 Yes
Stern 61 × 48 × 51 1 Yes
Hub 71 × 41 × 51 1 Yes
Rudder 71 × 51 × 81 2 Yes
Refinement fairing/rudder 101 × 57 × 101 4 Yes
Refinement wake 101 × 57 × 101 4 No
Refinement 241 × 95 × 101 16 No
Background 226 × 71 × 107 12 No

Total 8,000,250 55
Fig. 3. Details of the propeller and rudder grids.
P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320 313

the self-propulsion point is reached in an oscillatory fashion, show-


ing reflected waves from the exit boundary. These oscillations have
an amplitude of <0.1 RPS, and are thus of the same order of mag-
nitude of the experimental errors. Results of the computation are
condensed in Table 6, where the resistance coefficient is based on
the wetted surface area (S/Lpp 2 ) = 0.1803 including the rudder in
static orientation in calm water. The comparison error is defined as
E%D = (S − D)/D × 100, where D is the EFD value and S is the simu-
lation value. The table includes also results from Kim et al. [9], and
Kim and Li [10], both computed with advanced propeller models.
From Table 6, comparisons between CFD and EFD show that the
rate of revolution n, thrust and torque coefficients KT and KQ have
higher prediction accuracies than sinkage and trim. The error in
the prediction of the resistance is small, increasing to 5–6% for the
thrust and torque coefficients, mostly due to the overprediction of
the propeller rotational speed. These errors are comparable to those
for other self-propulsion computations with resolved rotating pro-
pellers [19]. Notice that the experiment by Otzen and Simonsen
[20] exhibits an error of 2.0%, and that the predicted self-propulsion
rotational speed is off by 1.6%. The predicted values for sinkage
and trim are around 15%, consistent also with values from other
simulations.
Two animations of the ship advancing under self-propulsion
condition are shown in Electronic Annexes I and II, where isosur-
faces of Q = 5000 are colored with axial velocity. Fig. 5 shows a rear
view of the free surface and isosurfaces of Q = 5000, the second
invariant of the rate of strain tensor. The propeller blade tip vortices
are clearly identified as they interact with the rudder. A very strong
hub vortex is also observed. Notice that the propeller creates a flow
field strong enough to break the symmetry of the free surface.
Fig. 4. Time histories of sinkage, trim and rate of revolutions for KCS self-propelled. Fig. 6 shows a close-up of the flow field near the propeller and
rudder with isosurfaces of Q = 5000 colored with axial velocity,
Case 1 was run using a total of 184 processors, 160 for the flow while the solid surfaces are colored with pressure. The animated
computation and 24 for the overset connectivity calculation run- version of this figure is presented in Electronic Annex I. The inter-
ning 3 Suggar instances. The wall clock time for this case was about action with the rudder causes the tip vortices to climb higher on the
three weeks, for a total of about 93,000 CPU hours on a Cray XT5 port side of the rudder, while they flow slightly down on the star-
computer. Once the controller reaches the self-propulsion point, board side, see also Fig. 5. The hub vortex combines with horseshoe
the rotational speed n, thrust and torque coefficients KT , KQ and vortices created at the root of each blade to produce a rich set of
resistance coefficient CT(SP) are obtained. structures leaving the propeller axis and interacting with the rud-
Time histories of the rate of revolutions n, sinkage and trim dur- der. A strong horseshoe vortex forms at the root of the rudder, as
ing the self-propulsion simulation are shown in Fig. 4. Notice that expected, but it strongly interacts with the propeller flow. The port

Fig. 5. Free surface and vortical structures as isosurfaces of Q = 5000 for Case 1.
314 P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320

Table 6
Self-propulsion results for KCS.

CT ×103  × 102 (m) (◦ ) KT 10 KQ n

EFD 5.14 −1.0 −0.155 0.252 0.408 14.15


CFD (E%D) 5.01 (−2.5%) 1.183 (18.3%) −0.172 (11%) 0.236 (−6.3%) 0.387 (−5.1%) 14.44 (2%)
Kim et al. [9] 5.16 (0.4%) −0.922 (−7.8%) −0.124 (−20%) 0.226 (−10.3%) 0.368 (−9.8%) 15.02 (6.1%)
Kim and Li [10] 4.72 (−8.2%) Imposed Imposed 0.267 (6.0%) 0.415 (1.7%) 13.17 (−3.2%)

Fig. 6. Close-up view of the propeller and rudder with isosurfaces of Q = 5000 colored with axial velocity for Case 1.

Fig. 7. Boundary layer and propeller wake represented by slices colored with axial velocity limited to U = 0.9 (boundary layer) and U = 1.5 (propeller wake) for Case 1.
P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320 315

Fig. 8. Free surface and vortical structures for Case 2, C1.

side of the vortex is largely unaffected by the propeller presence, see ary conditions as described in Carrica et al. [17]. Case 2 required 55
Electronic Annex I. On the other hand, the starboard side sees signif- processors for the flow solver and 1 for the overset solver, running
icant instability and growth as the net flow upstream of the rudder in 7 nodes with 8 cores each of a Cray XT5.
near the root moves from port to starboard due to the propeller As described in Table 3, three conditions C1, C2 and C3 with three
rotation, causing a small angle of attack that ultimately results in wave lengths and two velocities were studied with the ship free to
a vortex of considerable strength, see also Electronic Annex II. See
also in Fig. 6 the presence of two small bilge vortices approaching
the propeller from below the shaft hub, which are of note but prob-
ably do not affect the propeller flow much since they act on the root
of the propeller where the loads are small.
The boundary layer represented by slices colored with axial
velocity limited to U = 1.5 for Case 1 is shown in Fig. 7. The boundary
layer remains attached and uneventful until it reaches the hull con-
traction near the stern, where it thickens dramatically, resulting on
a wake factor at the propeller plane of about 0.7 [1]. A cross section
at 80% chord of the rudder shows clearly the high momentum wake
of the propeller as it interacts with the rudder, and the deviation up
on the port side and down on the starboard side. It is also evident
the presence of the already described strong horseshow vortex that
exhibits low momentum at its core. Downstream a low-momentum
wake is observed near the free surface, but the upward momentum
created by the interaction of the propeller with the port side of
the rudder keeps climbing, forming a wake composed of two low-
speed regions separated by a finger of the high-speed wake of the
propeller.

5.2. Case 2: KCS pitching and heaving in head waves

As discussed in Section 4, a symmetry boundary condition has


been used in the center plane to save on computational costs.
When used in conjunction to a DES turbulence approach, this is an
approximation since the instantaneous large-scale structures that
are being resolved with DES will not be symmetric, only the aver-
aged flow field will satisfy symmetry. Though the consequences
in the wake field can be significant, the symmetry approximation
should have a small effect on the predicted forces and motions,
which will be compared with data. Waves were imposed as bound- Fig. 9. Time histories of CT , z and , Case 2, C1.
316 P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320

pitch and heave. Time histories of the total resistance coefficient


CT , heave motion z and pitch angle  are obtained from the compu-
tation. The data from Otzen and Simonsen [20], available from the
G2010 website [2], was reprocessed to obtain 0th and 1st harmonic
amplitudes and 1st harmonic phase using only the periodic part of
the data after dropping the initial transient.
Irvine et al. [21] suggest that the maximum response of the ship
will occur when fe = fn and /Lpp = 1.33, where fe is the encounter
frequency and fn the natural frequency of the pitching and heaving
system in wet condition. For the present model the natural frequen-
cies in heave and pitch are close to fn ≈ 0.9 Hz. Conditions C1 and
C3 have frequencies of encounter of 0.896 and 0.8885, respectively,
both very close to the natural frequency of the system. The wave-
length in condition C2 is very large (/Lpp = 2.0) and thus exhibits a
very linear behavior with amplitudes and phases as expected from
a long wave: the heave amplitude the same as the waves and in
phase with them, and the pitch amplitude given by the slope of the
wave with a 90◦ lead in phase. Condition C3 also has the maximum
response wavelength /Lpp = 1.33 and is thus closer to resonance.
As an example of the type of flow generated by the ship pitch-
ing and heaving, free surface and vortical structures obtained as
isosurfaces of Q = 100 are shown in Fig. 8 for condition C1. An ani-
mated version of Fig. 8 is presented in Electronic Annex III. Notice
the bilge vortices coming off the stern, modulated by the motion
of the ship. Time histories of CT , z and  for conditions C1–C3 are
shown in Figs. 9–11. Results of the 0th harmonic amplitude of CT , z
and  are given in Table 7, with the resistance coefficient based on
the same wetted surface area used for Case 1. Table 7 shows also
results presented by Simonsen et al. [11] for condition 1 and by El
Moctar et al. [12], Manzke and Rung [13] and Akimoto et al. [14] for
Fig. 10. Time histories of CT , z and  for Case 2, C2. all three conditions at the G2010 CFD Workshop [2].
The motions observed are mostly linear, but the forces are not
with the exception of condition C2. The first harmonic amplitude
and phases for pitch and heave are in excellent agreement, with the
heave amplitude a bit under predicted for conditions C1 and C3. The
resistance coefficient is completely off respect to the experiments.
The strong non-linear behavior observed in the experimental forces
in conditions C1 and C3 are somewhat captured by CFD, but the
amplitude of the fluctuations is grossly underestimated while the
phase is also off. The reasons for these discrepancies are not clear,
and need further investigation. Results from Simonsen et al. [11],
El Moctar et al. [12], Manzke and Rung [13] and Akimoto et al. [14]
are also in strong disagreement with resistance data for this same
experiment, casting doubts on the validity of the experimental data.
Quantitative comparisons between CFD and EFD show that the
0th harmonic amplitude of the resistance coefficient is generally
well predicted in this work, with under prediction of the forces
ranging from 1% to 7% for all conditions, see Table 7. The predictions
by Simonsen et al. [11] are also good (underpredicted by 5.6%), but
other authors had differences with EFD closer to ±25%. The first
harmonic amplitudes for conditions C1 and C3 are strongly under-
predicted, in agreement with calculations of other authors, ranging
from 15% to 30% of the reported experimental value. Even condi-
tion C2, which is mostly linear, shows differences on first harmonic
amplitude ranging from −18.7% to 24.8% for all authors. These dis-
crepancies between several independent CFD studies and EFD cast a
shadow of doubt on the reliability of the experimental data, as it was
discussed in G2010, since CFD is usually within 3–10% of EFD for
resistance for this type of computations. The first harmonic phase
of the resistance is better predicted with errors within 8% of the
experiments, similar to the errors reported by other researchers.
Notice that the phase difference is non-dimensionalized using 180◦
as reference phase.
The heave and pitch responses are much closer to those reported
Fig. 11. Time histories of CT , z and , Case 2, C3.
by EFD, see Figs. 9–11. The worst results occur for condition C1,
where errors in heave for the 0th and 1st harmonic amplitudes
P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320 317

Fig. 12. Free surface colored with elevation for the four quarter periods for Case 2, condition C1. Top left: t/T = 0, top right: t/T = 0.25, lower left: t/T = 0.5, lower right: t/T = 0.75.

Fig. 13. Free surface colored with elevation for the four quarter periods for Case 2, condition C2. Top left: t/T = 0, top right: t/T = 0.25, lower left: t/T = 0.5, lower right: t/T = 0.75.
318 P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320

Table 7
Amplitudes and phases of CT , z and  for Case 2.

CT × 103 z (mm)  (◦ )

C1
0th, 1st, 1st ␸EFD (D) 7.159, 20.42, 82 −4.89, 40.04, 4.7 −0.058, 2.11, 34
0th, 1st, 1st ␸CFD (E%D) 6.971, 6.03, 67.0 (−2.6, −70.4, −8.3) −6.37, 33.97, −7.5 (30.0, −15.1, −6.8) −0.077, 2.21, 30 (32.7, 4.7, −2.2)
Simonsen et al. [11] 6.676, 3.61, 77.9 (−5.6, −82.33, −2.3) −5.714, 43.54, −11.35 (16.9, 8.7, −8.9) −0.028, 2.39, 23 (−51.7, 11.7, −6.1)
El Moctar et al. [12] Not reported −5.869, 38.82, −19.8 (20.0, −3.1, −8.4) 0.023, 2.165, 14.1 (−139.7, 2.6, −11.1)
Manzke and Rung [13] 8.628, 3.153, 117.85 (20.3, −84.56, 19.9) −6.001, 37.93, −22.29 (22.7, −5.3, −15.0) 0.002, 1.984, 22.34 (−103.4, −6.0, −6.5)
Akimoto et al. [14] 5.627, 0.292, 87.2 (−21.4, −98.6, 2.9) −6.244, 33.61, −29.6 (27.7, −16.1, −19.1) 0.045, 2.11, 19.6 (−177.6, 0.0, 8.0)
C2
0th, 1st, 1st ␸EFD (D) 6.033, 32.07, 70.8 −6.52, 65.8, 9.0 −0.13, 3.12, 61.0
0th, 1st, 1st ␸CFD (E%D) 5.599, 25.2, 82.3 (−7.2, −21.4, 6.4) −6.49, 62.4, 2.0 (−0.4, −5.2, −3.9) −0.21, 2.99, 70.0 (61.5, −4.2, 5.0)
El Moctar et al. [12] Not reported −7.074, 65.1, 0.76 (8.5, −1.1, −4.6) 0.08, 3.14, 65.1 (−161.5, 0.6, 2.3)
Manzke and Rung [13] 7.841, 26.06, 72.8 (30.0, −18.7, 1.1) −6.348, 69.98, −2.0 (−2.6, 6.3, −6.1) 0.025, 3.56, 60.5 (−119.2, 14.1, −0.2)
Akimoto et al. [14] 4.168, 24.13, 84.3 (−30.9, −24.8, 7.5) −7.228, 91.6, −8.5 (10.9, 39.2, −9.7) 0.035, 3.12, 71.6 (−126.9, 0.0, 5.9)
C3
0th, 1st, 1st ␸EFD (D) 8.716, 35.4, 99.8 −11.4, 58.1, 21.0 0.075, 2.43, 45.0
0th, 1st, 1st ␸CFD (E%D) 8.189, 5.09, 86.4 (−6.0, −85.6, −7.4) −11.8, 51.0, −14.0 (3.5, −12.2, −19.4) 0.04, 2.59, 21.0 (−46.6, 6.6, −13.3)
El Moctar et al. [12] Not reported −11.28, 56.0, −24.62 (−1.1, −3.6, −25.3) 0.117, 2.57, 17.7 (56.0, 5.8, −15.2)
Manzke and Rung [13] 10.533, 5.041, −224 (28.8, −86.8, 20.1) −15.21, 61.23, −20.9 (33.4, 5.4, −23.3) −0.202, 2.458, 33.8 (−369.3, 1.2, −6.2)
Akimoto et al. [14] 4.07, 14.05, −77.3 (−53.3, −60.3, −98.4) −6.6, 56.3, −17.2 (−42.1, −3.1, −21.2) 0.021, 2.37, 33.1 (−72.0, −2.5, −6.6)

and 1st harmonic phase are 30%, −15.1% and −6.8%, respectively, other two conditions. First harmonic amplitude and phase were
with other authors showing also large discrepancies in 0th har- also very well predicted by other authors for all conditions.
monic amplitude. Errors in condition C2 are within 5% of error for Notice in the data for pitch and heave a significant low frequency
all variables, increasing somewhat for condition C3. This trend is content, which is not reproduced by CFD. This could be due to wave
also observed in the results of other authors. As expected, condition reflections or direct low frequency content in the carrying wave.
C3 exhibits the largest response to heave in dimensionless terms. This generates inaccuracies in the calculation of the 0th and 1st
The 0th harmonic of the pitch angle is very small, reaching harmonics of the experimental data.
−0.13◦ for condition C2, and thus large relative errors are observed Free surface colored with elevation for the four quarter periods
when compared with EFD. The absolute errors are small, since CFD for Case 2, conditions C1–C3 are shown in Figs. 12–14. Though all
also predicts small 0th harmonic amplitude. The pitch 1st harmonic conditions share the same relatively large wave slope ak = 0.0525,
amplitude is within 6% of the data for all conditions, and the first C1 and C2 produce motions of moderate violence. Condition C3
harmonic phase is −13◦ for condition C3 and much better for the results in stronger motions with some water on deck (see Fig. 14

Fig. 14. Free surface colored with elevation for the four quarter periods for Case 2, condition C3. Top left: t/T = 0, top right: t/T = 0.25, lower left: t/T = 0.5, lower right: t/T = 0.75.
P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320 319

Fig. 15. Boundary layer represented by slices colored with axial velocity limited to U = 0.9 for the four quarter periods for Case 2, condition C3. Top left: t/T = 0, top right:
t/T = 0.25, lower left: t/T = 0.5, lower right: t/T = 0.75.

at t/T = 0.5). In all cases the waves are diffracted significantly, with in waves. Notice also the modulation of the tip vortices generated
condition C3 showing particularly large modulation of the wave in by the shaft hub and their interaction with the rudder.
the near field of the ship.
Fig. 15 shows the boundary layer represented by x = constant 6. Conclusions and future work
slices colored with axial velocity limited to U = 0.9 for the four quar-
ter periods for Case 2, condition C3. As the wave passes through High-performance computations of self-propulsion and pitch
midships (t/T = 0), the bow is starting to dip into the water (see and heave in waves for KCS appended with rudder in model scale
Fig. 14), and the bilge vortices that originate near the bow when it are presented. The self-propulsion results, computed with a dis-
comes out of the water are transported downstream. As the wave cretized propeller, are in very good agreement with the data for
trough travels downstream bilge vortices are generated along the forces, sinkage and trim, and consistent in accuracy with other
hull up to x/L ∼ = 0.8, where the hull shape is faired to accommo- simulations of self-propulsion using discretized propellers. Pitch
date the rudder and shaft hub. In this region the boundary layer and heave motions agree well with data for all three conditions
becomes significantly thicker and the modulation by the vortices evaluated, as mostly does also the 0th harmonic of the forces. The
generated upstream by the action of the wave and ship motions predicted amplitude and 1st harmonic amplitude and phase of the
is dramatic. At the propeller plane, immediately upstream of the resistance forces are considerably off, in agreement with reported
rudder, the boundary layer is considerable thinnest at t/T = 0.25, CFD results by other researchers.
and thickest at t/T = 0.75, with presumably very large differences Future work in this area includes study of the problem of self-
in nominal wake. These differences can cause important unsteady propulsion in waves and the effect of the modulated boundary layer
propeller loads and reduce the propeller efficiency when operating on propeller efficiency.
320 P.M. Carrica et al. / Applied Ocean Research 33 (2011) 309–320

Acknowledgements [8] Lee JH, Rhee SH. Flexible CFD meshing strategy for prediction of ship resis-
tance and propulsion performance. CFD Workshop Gothenburg, Gothenburg,
Sweden; 2010.
This research was sponsored by the US Office of Naval Research [9] Kim J, Park IR, Kim KS, Van SH. Feasibility study on numerical towing tank
under the administration of Dr. Patrick Purtell. Discussions and help application to predictions of resistance and self-propulsion performances for a
by H. Hosseini are appreciated. Computations were performed at ship. CFD Workshop Gothenburg, Gothenburg, Sweden; 2010.
[10] Kim K, Li DQ. Estimation of numerical uncertainty of SHIPFLOW in self-
the Artic Region Supercomputer Center and at the Navy DoD Super- propulsion simulation of KCS. CFD Workshop Gothenburg, Gothenburg,
computing Resource Center Cray XT5 machines. Sweden; 2010.
[11] Simonsen C, Otzen J, Stern F. EFD and CFD for KCS heaving and pitching in
regular head waves. In: Proc. 27th Symp. Naval Hydrodynamics, Seoul, Korea;
Appendix A. Supplementary data 2008.
[12] El Moctar B, Kaufmann J, Ley J, Oberhagemann J, Shigunov V, Zorn T. Prediction
Supplementary data associated with this article can be found, in of ship resistance and ship motions using RANSE. CFD Workshop Gothenburg,
Gothenburg, Sweden; 2010.
the online version, at doi:10.1016/j.apor.2011.07.003.
[13] Manzke M, Rung T. Resistance prediction and seakeeping analysis with FreSCo+.
CFD Workshop Gothenburg, Gothenburg, Sweden; 2010.
References [14] Akimoto H, Omori T, Saito H, Masiur R. Numerical simulation of KCS container
carrier in head wave conditions. CFD Workshop Gothenburg, Gothenburg,
Sweden; 2010.
[1] Carrica PM, Castro AM, Stern F. Self-propulsion computations using speed con-
[15] Noack R. SUGGAR: a general capability for moving body overset grid assembly.
troller and discretized propeller with dynamic overset grids. J Mar Sci Technol
AIAA paper 2005-5117. 17th AIAA CFD Conf., Toronto, Canada; 2005.
2010;15:316–30.
[16] Carrica PM, Wilson RV, Stern F. An unsteady single phase level set
[2] G2010, A workshop on CFD in ship hydrodynamics, Gothenburg, Sweden, 2010.
method for viscous free surface flows. Int J Num Methods Fluids 2007;53:
http://www.gothenburg2010.org.
229–56.
[3] Lübke LO. Numerical simulation of the flow around the propelled KCS. CFD
[17] Carrica PM, Wilson RW, Noack R, Stern F. Ship motions using single-phase level
Workshop Tokyo 2005, Tokyo, Japan; 2005.
set with dynamic overset grids. Comput Fluids 2007;36:1415–33.
[4] Greve M, Abdel-Maksoud M, Eder S, De Causmaecker. Steady viscous flow cal-
[18] Carrica PM, Paik K, Hosseini H, Stern F. URANS analysis of a broaching event in
culation around the KCS model with and without propeller under consideration
irregular quartering seas. J Mar Sci Technol 2008;13:395–407.
of the free-surface. CFD Workshop Gothenburg, Gothenburg, Sweden; 2010.
[19] Carrica PM, Huang J, Noack R, Kaushik D, Smith B, Stern F. Large-scale DES
[5] Bugalski T, Hoffman P. Numerical simulation of the interaction between ship
computations of the forward speed diffraction and pitch and heave problems
hull and rotating propeller. CFD Workshop Gothenburg, Gothenburg, Sweden;
for a surface combatant. Comput Fluids 2010;39:1095–111 [2010G].
2010.
[20] Otzen J, Simonsen C. Uncertainty assessment for KCS resistance and propulsion
[6] Wu Q, Feng XM, Yu H, Wang JB, Cai RQ, Chen XL. Prediction of ship resistance
tests in waves. Force Technology report ONRIII187-01; 2008.
and propulsion performance using multi-block structured grids. CFD Workshop
[21] Irvine Jr M, Longo J, Stern F. Pitch and heave tests and uncertainty assess-
Gothenburg, Gothenburg, Sweden;, 2010.
ment for a surface combatant in regular waves. J Ship Res 2008;52:
[7] Jin W, Gao Q, Vassalos D. The prediction of KCS resistance and self-propulsion
146–51.
by RANSE. CFD Workshop Gothenburg, Gothenburg, Sweden; 2010.

You might also like