Experiments in Fluids 31 (2001) 567±578 Ó Springer-Verlag 2001

Measurement of flows around modern commercial ship models

W. J. Kim, S. H. Van, D. H. Kim

567
Abstract To document the details of ¯ow characteristics benchmark database for CFD validation for resistance and
around modern commercial ships, global force, wave propulsion of a ship (ITTC 1999; see also Longo and Stern
pattern, and local mean velocity components were mea- 1996; Stern et al. 1998). For the cargo-container ship, Series
sured in the towing tank. Three modern commercial hull 60 (Fry and Kim 1985; Toda et al. 1990, 1992; Longo et al.
models of a container ship (KRISO container ship = KCS) 1993; Suzuki et al. 1997) and Hamburg Test Case (Bertram
and of two very large crude-oil carriers (VLCCs) with the et al. 1994; Gietz and Kux 1995) are given. DTMB model
same forebody and slightly different afterbody (KVLCC 5415 is recommended for a combatant model (Fry and Kim
and KVLCC2) having bow and stern bulbs were selected 1985; Ratcliffe 1998; Olivieri and Penna 1999; Longo and
for the test. Uncertainty analysis was performed for the Stern 1999). For the full-form tanker, HSVA/Dyne tanker
measured data using the procedure recommended by the models (Knaack 1992; Denker et al. 1992; Dyne 1995) and
ITTC. Obtained experimental data will provide a good Ryuko-Maru (Ogiwara 1994; Suzuki et al. 1998) are given.
opportunity to explore integrated ¯ow phenomena around Previously, two workshops (Larsson et al. 1991; Kodama
practical hull forms of today. Those can be also used as the 1994) were arranged for the computational analysis of ¯ow
validation data for the computational ¯uid dynamics around a ship, and HSVA/Dyne tanker models and a Series
(CFD) code of both inviscid and viscous ¯ow calculations. 60 model were chosen for the test cases. However, those
data are often partial and not enough to understand the
1 complicated ¯ow phenomena. The hull forms used in those
Introduction experiments are old-fashioned and quite different from the
As most cargo worldwide today is transported via ship, it modern hull forms of ships today.
is very important to design the ship hull forms such that The aim of the present study is to provide a set of
they operate economically. To propel a ship, its engine has experimental data for the modern commercial hull forms
to provide enough power to overcome the hydrodynamic with bulbous bow and stern bulb. The experimental data
drag due to viscosity and wave generation. It is necessary presented in this paper encompass resistance and pro-
to understand the complicated ¯ow characteristics to de- pulsive ef®ciency, wave elevation along the hull surface,
sign the hull forms with lower drag and higher propulsive wave pattern along longitudinal and transverse cuts, and
ef®ciency. For better understanding of the ¯ow around a mean velocity components in boundary layer and wake.
modern commercial ship, it is of primary importance to Measurements were performed in the towing tank of Korea
produce reliable experiment data of practical hull forms. Research Institute of Ships and Ocean Engineering (here-
The experiment data describing the local ¯ow details are after, KRISO).
also invaluable in the ®eld of computational ¯uid dy- It is desirable to select the actual hull forms built by a
namics (CFD) for the validation of the developed physical shipyard and subjected to sea-trial; however, they can
and numerical modeling. hardly be disclosed. Instead, the hull forms for the con-
There have been some experimental data for the ¯ows tainer ship and the VLCC (very large crude-oil carrier)
around ship models. The International Towing Tank were designed by KRISO. In the case of the container ship,
Conference (hereafter, ITTC) summarized available it was designed to carry 3,600 containers, while the VLCC
was supposed to transport 300,000 tons of crude oil. For
the VLCC, two hull forms with the same forebody and
Received: 26 September 2000/Accepted: 22 May 2001 slightly different afterbody were designed to identify the
stern ¯ow change due to hull form variation. For briefness,
W. J. Kim (&), S. H. Van, D. H. Kim the 3600TEU container ship was named KCS and the two
Korea Research Institute of Ships and Ocean Engineering 300 KVLCCs were named KVLCC and KVLCC2, respec-
KORDI, Jang-Dong 171, Taejon, 305±343, Korea
e-mail: wjkim@kriso.re.kr tively. Their main particulars and hull geometries are very
close to those of the commercial ships today. For KCS with
The present work was based on the project ``Improvement of moderate speed and low block coef®cient, wave generation
Resistance Performance of Commercial Ships,'' supported by the on free surface was of the primary interest, since the big
Ministry of Trade and Industry of Korea. The authors thank the
technicians of the KRISO towing tank for their valuable contri- bulbous bow was supposed to suppress bow wave gener-
bution in manufacturing models and helping in experiments. The ation and the ¯at stern overhang and the transom stern
hull information and experiment data discussed in this paper are produced very complicated stern waves. On the other
available at http://hyperteam.krio.re.kr. hand, for KVLCC and KVLCC2 with low speed and large
 block coef®cient, viscous boundary layer ¯ow in the stern Table 1. Principal particulars of KCS
region was focused upon, since strong bilge vortices were
Designation Prototype TT model
entering into the propeller plane.
Recently the Gothenburg 2000 Workshop on CFD in Scale ratio 1.0 1/31.6
ship hydrodynamics chose three test ships (Larsson et al. Speed (m/s) 12.3467 2.1964
2000), one was the DTMB model 5415 combatant model Froude number (Fn) 0.26 0.26
mentioned earlier and the others were the KCS and Reynolds number (Re) 2.4 ´ 109 1.4 ´ 107
KVLCC2 of the present paper. International collaborative Length (m) 230.0 7.2786
Breadth (m) 32.2 1.0190
works were performed to provide the well-prepared data Depth (m) 19.0 0.6013
set to the ship hydrodynamics community for the CFD Draft (m) 10.8 0.3418
validation. The Ship Research Institute of Japan measured Wetted surface area (m2) 9,498.0 9.5121
568
stern ¯ow ®eld and surface pressure distribution with and Displacement (m3) 52,030.0 1.6490
without propeller rotating behind the KCS model (Fujis- Block coef®cient (CB) 0.6505 0.6505
awa et al. 2000; Tsukada et al. 2000). For the KVLCC2
model, wind tunnel experiments were carried out for Table 2. Principal particulars of KVLCC
turbulence quantities (Lee et al. 1998; Van et al. 1999). In
the present paper, the experimental procedures and the Designation Prototype TT model
measured data for KCS, KVLCC, and KVLCC2 procured in
Scale ratio 1.0 1/58.0
the KRISO towing tank are described. The hull forms and
Speed (m/s) 7.9739 1.047
measured data presented in this paper are publicly avail- Froude number (Fn) 0.142 0.142
able in the Internet (http://hyperteam.kriso.re.kr or http:// Reynolds number (Re) 2.1 ´ 109 4.6 ´ 106
www.iihr.uiowa.edu/gothenburg2000/). Length (m) 320.0 5.5172
In the following, the hull forms and experimental Breadth (m) 58.0 1.0
apparatus used to document global force, wave pattern, Depth (m) 30.0 0.5172
and local velocity ®eld are given along with the uncertainty Draft (m) 20.8 0.3586
Wetted surface area (m2) 27,320.0 8.1213
analysis results based on the procedure recommended by
Displacement (m3) 312,737 1.6029
ITTC (ITTC 1999; Coleman and Steele 1999). Measured Block coef®cient (CB) 0.8101 0.8101
resistances, wave pro®les along the hull surfaces, wave
patterns obtained from transverse, and longitudinal cut
methods are discussed, followed by the mean velocity Table 3. Principal particulars of KVLCC2
®elds around the stern region. For the VLCC models with
stern frame line modi®cation, the propeller plane wakes Designation Prototype TT model
are also analyzed with the viewpoint of propulsive ef®- Scale ratio 1.0 1/58.0
ciency as well as resistance, since the effect of stern hull Speed (m/s) 7.9739 1.047
form variation on the speed of a ship should be considered Froude number (Fn) 0.142 0.142
as an outcome of both resistance and propulsion charac- Reynolds number (Re) 2.1 ´ 109 4.6 ´ 106
teristics. As proven in the Gothenburg 2000 Workshop, Length (m) 320.0 5.5172
authors expect the present data to be useful for the un- Breadth (m) 58.0 1.0
Depth (m) 30.0 0.5172
derstanding of the ¯ows around modern commercial hull Draft (m) 20.8 0.3586
forms and also for the validation of CFD tools in ship Wetted surface area (m2) 27,194.0 8.0838
hydrodynamics. Displacement (m3) 312,621 1.6023
Block coef®cient (CB) 0.8098 0.8098
2
Experimental apparatus
afterbody geometries (A1 and A2) with the same side
2.1 pro®le were designed and used for the global-force and
Model ships local-¯ow measurement.
In the present experiment, three test hull forms were se- The two VLCC hull forms, i.e., F1 + A1 and F1 + A2,
lected, i.e., a 3600-TEU container ship and two 300 were named KVLCC and KVLCC2, respectively. The ®rst
KVLCCs, recently designed by KRISO. The test ships are of VLCC stern hull form (A1) has barge-type stern frame
very similar shapes to commercial ships today. Tables 1, 2 lines with a ®ne stern bulb, while the second stern hull
and 3 provide the principal particulars of the test ships, form (A2) has more U-shaped stern frame lines than the
whose body plans and side pro®les are also given in Figs. 1 former. Recently, CFD tools are being used in the initial
and 2. design stage in order to evaluate the resistance perfor-
Initially, three hull forms were designed and resistance/ mance and wake distribution of hull forms in shipyards. It
self-propulsion tests were performed for the container ship is often asked if the computational tools can tell the ¯ow
(Kang et al. 1998). Of the three container hull forms, difference from the frame line changes the designer actu-
F1 + A2 (i.e., the ®rst forebody and the second afterbody) ally applies during the hull form optimization. To ful®ll
was selected for local ¯ow measurement and named KCS, the aforementioned request, the experimental data for CFD
meaning KRISO container ship. For the VLCC, two hull validation should be for comparative purposes like the
forms with the same forebody (F1) and slightly different ¯ow measurements around two slightly different hull
 Fig. 1. Body plan and side pro®le of KCS

569
Fig. 2. Body plans and side pro®les of
KVLCC and KVLCC2 (Soild lines: KVLCC,
Dotted lines: KVLCC2)

forms. The hull form changes of VLCC sterns (A1 and A2)
correspond to those shipyards usually apply, which en-
sures the usefulness of the present data in the evaluation of
CFD tools for wake prediction.
For the towing tank test, a model of the KCS was made
of wood with a scale ratio of 1/31.6, while the two models
of KVLCC and KVLCC2 were quite small with a scale ratio
of 1/58 in order to keep the width of the models accom-
modated in the measurement frame. Studs with trapezoi-
dal heads as turbulence stimulators were nailed at St. 19
and at the middle of the bow bulb at 10-mm girthwise
intervals, so as to make sure that the ¯ow became fully
turbulent. In shipbuilding engineering, St. 0 and St. 20
mean the after perpendicular (A.P.) at the stern and the
forward perpendicular (F.P.) at the bow of a ship, re-
spectively. Stations and waterlines were marked densely to
locate the probes in the proper positions and to make it
easy to read wave pro®les along the hull surface. The
model ship was ®xed at the towing carriage by using two
clamping devices located around St. 6 and St. 14. All the
local ¯ow measurements in the towing tank were carried
out in the ®xed condition, where neither trim nor sinkage
was allowed, in order to prevent the ship model from Fig. 3. Schematics of towing tank measurement frame
breaking the probes. Besides, the computational condition
is easier in the ®xed condition, since the posture of ship
does not change during the run. the measurement rail to position the probes accurately and
ef®ciently.
2.2
Towing tank 3
The measurement of global force, wave pattern, and Measurement and uncertainty analysis
mean velocity components were carried out in the tow- In the towing tank test, a model ship was towed at the
ing tank of KRISO. The size of the towing tank is speed of Froude's similarity law such that the speed ratio
200 m ´ 16 m ´ 7 m (length ´ width ´ depth) and the of the model and the prototype is the square root of the
towing carriage can run up to 6 m/s. Blockage coef®cients, scale ratio. Since the design speed (VS) of the KCS was
de®ned as the ratio of the sectional area of the model and 12.247 m/s (i.e., 24.0 knots), the model ship was towed at
the towing tank, were less than 0.35% for all test models, the speed (V
pM ) of 2.196 m/s.pThe
 resulting

 Froude number
allowing the blockage effect to be ignored. …Fn ˆ VS = …gLpp †S ˆ VM = …gLpp †M † was 0.26. The cor-
To keep the side of model ship spacious, the ship model responding Reynolds number (Re ˆ VM (Lpp)M/m)
was moved into the starboard direction by 25  30 cm of the KCS model was 1.4´107. Here, (LPP)S and (LPP)M are
from the center of the measurement frame, so that the the length between perpendiculars of the prototype and
probe-holding device connected to the traversing mecha- the model ship, respectively. g is the gravitational accel-
nism could be accommodated, as shown in Fig. 3. The eration and m is the kinematic viscosity. For KVLCC and
traversing mechanism driven by three servomotors with a KVLCC2, the Froude number was 0.142 and the speeds of
precision of 0.02 mm for each direction was installed at the prototype and the model were 7.974 m/s (i.e., 15.5
 knots) and 1.047 m/s, respectively. The Reynolds number single run. The wave height gauge holder, driven by an
of both KVLCC and KVLCC2 models was 4.6 ´ 106. electric motor, could slide along a truss attached at the
sidewall of towing tank. An optical switch at 3 m ahead of
3.1 F.P. provided the triggering signal to identify the relative
Global-force measurement location between the model ship and the wave height
The total resistance (drag) coef®cients were measured gauges. Data were obtained at the rate of 50  100 Hz
using the resistance dynamometer (R25, Kempf and using a portable analog-digital converter (DaqBook, IO-
Remmers, Hamburg, Germany). When a ship is running, Tech, Cleveland, Ohio) for 50 s. The duration of sampling
its posture is changed. The trim and sinkage are usually was long enough to include the re¯ected waves from the
allowed in the towing tank resistance test to match the real sidewall of the towing tank. The servo-needle-type wave
situation (i.e., free condition). However, in the present height gauges attached to the cantilever truss for the wave
570 study, trim and sinkage were not allowed in the local ¯ow measurement along longitudinal cuts are shown in Fig. 4.
measurement (i.e., ®xed condition). It is not easy to Next, waves around the bow and stern region were
measure drag in the ®xed condition since the ®xing device measured by the transverse-cut method using the gauges
can easily disturb the axial force balance. Thus, in the attached to the traversing mechanism moving with the
resistance measurement, negative trim and sinkage was towing carriage. In the case of the KCS having a large angle
initially given to make the ®xed running condition. of bow ¯are, a gauge was inclined by 19° to access close to
Initially, the weight distribution for the model ship was the hull surface. To measure bow and stern waves, two or
adjusted to have the negative amount of trim and sinkage three wave gauges were attached to the traversing mecha-
of the free condition to obtain the resistance in zero trim nism and moved after gathering 100 samples in 5 s at each
and sinkage. After several iterations, the resistance position. It was found that the measured wave height
coef®cient could be procured in zero trim and sinkage behind the stern was hardly stable, especially for the KCS.
condition. The data given here for the bow and stern region other than
The measured total resistance coef®cients were divided the longitudinal cuts are simply averaged at each position.
into frictional resistance coef®cients, based on the 1957 On the other hand, to measure the wave elevation along
ITTC formula (SNAME 1967), and residuary resistance the hull surface, a telescope, a digital video recorder, and a
coef®cients representing the sum of wave-making drag still camera were used. It was very dif®cult to read the
and form drag. The measured residuary resistance coef®- wave elevation near the stern area, since waves along the
cient of KCS at design speed was 0.725 ´ 10)3, while ¯at overhang near the design waterline were not clearly
those of KVLCC and KVLCC2 were 0.638 ´ 10)3 and seen. To reduce human error, three persons used three
0.660 ´ 10)3, respectively. The frictional resistance different means to read the elevation and the average value
coef®cient of KCS from the 1957 ITTC formula was was taken as the wave elevation along the hull surface.
2.832 ´ 10)3, while those of KVLCC and KVLCC2 were
both 3.450 ´ 10)3. 3.3
The self-propulsion test with rotating propeller was also Local mean velocity measurements
carried out for KVLCC and KVLCC2 to compare the For the measurement of the three-dimensional mean ve-
propulsive ef®ciency of the two hull forms. The stern hull locity ®eld around stern region, a 5-hole Pitot-tube rake
form change usually modi®es the wake distribution at the was utilized. The rake included ®ve 5-hole Pitot tubes with
propeller plane. Thus, the propeller in¯ow condition be- 4 cm intervals to speed up the measurement as shown in
comes different. A model propeller formerly used for the Fig. 5. An adapter was designed to connect the rake to the
similar VLCC was chosen for the assessment of the ®nal three-dimensional traversing mechanism installed at the
propulsive power of the two VLCCs. To achieve the self-
propelled condition in the towing tank, the towing rod
provided the difference of frictional resistance coef®cients
between the prototype and the model ship, while the re-
maining drag was overcome by the thrust from the pro-
peller. The thrust and torque dynamometer (NR330,
Kempf and Remmers) was utilized. The measured pro-
pulsive ef®ciency and the resulting power for the proto-
type will be discussed in Sect. 4.2.2 with the local ¯ow
measurement results.

3.2
Wave pattern measurements
The servo-needle-type wave height gauges (SHT3±30, West
Japan Fluids Engineering Laboratory Co., Nagasaki, Japan)
were used to document the full spectrum of generated
waves. First of all, the longitudinal cut method was utilized
to measure the generated waves along 36  39 lateral
positions. Three gauges were tied up in one unit, so that
three lines of wave height data could be obtained in a Fig. 4. Photograph of wave measurement along longitudinal cuts
 been performed (Van and Park 1993) and was not repeated
in the present study. The uncertainty interval with 95%
con®dence level of resistance measurement was about
1.0% of the total resistance, including both bias and pre-
cision errors.
In the case of wave height measurement, two different
methods were used. When obtaining the hull surface
pro®le, the major source of error stems from the human
eye, because wave heights were read from the photo, video,
or by telescope during running. The bias error for the
wave elevation along the hull surface was within 1.0 mm,
571
except for F.P., where the wave itself tended to move up
and down by a few millimeters during the test. For the
wave measurement along the longitudinal cuts, the on-
coming waves (time-varying quantity) were measured.
Fig. 5. Pitot-tube rake used to measure the velocities in the
Thus, the precision error was not taken into account. The
towing tank
bias error came from the nonlinearity and resolution of
servo-needle-type gauge and ampli®er. The location of
towing carriage. Before and after the local ¯ow measure- wave probes (initial setting and movement), and the phase
ment, the nulling procedure without the model ship was due to the triggering signal were also taken into consid-
completed to con®rm and adjust the Pitot-tube rake set- eration. The resulting bias error was within 0.5 mm for the
ting. The 5-hole Pitot tube had a semi-spherical tip and the wave height along longitudinal cuts.
outer diameter of the tube was 8 mm. Before the mea- For local mean velocity measurement, both bias and
surement, the Pitot tubes were calibrated in the range of precision errors were considered. As sources of bias error,
‹40° for both pitch and yaw angles simultaneously. The the resolution and nonlinearity of pressure transducers,
calibration device provided ®rst the pitch (vertical) angle demodulators, A/D converters, the location of the probes,
and then the yaw (horizontal) angle, which resulted in the carriage speed, and the water temperature and density
pitch-yaw chart calibration (Kim et al. 1997). A total of 121 were considered. The bias error was found to be less than
angle variations, i.e., 11 angles both for pitch and yaw, 0.1% of the model speed. The precision error was esti-
provided three correlation coef®cients for velocity mag- mated from the measured data series for each probe lo-
nitude, pitch, and yaw angle, respectively. cation. The time history of the measured data showed that
Each tube was connected directly to a Validyne pressure the pressure value near the hull surface and behind the
transducer (DP-15) and carrier demodulator (CD-10). An hull body was quite unstable, however, it became very
analog±digital signal converter (NEFF) on the towing stable in the outside region. The precision error was much
carriage provided pressure data to compute three-dimen- larger than the bias error. The resulting error with 95%
sional velocity components. A pressure signal from each con®dence interval was ‹0.8% of the model speed in the
hole of the tube was obtained for 8 s at a rate of 50 Hz and near-wall region and ‹0.3% in the outside region. How-
then averaged. Calculated correlation coef®cients were ever, it became ‹1.2% at a very few points just behind the
used to determine the magnitude and angle of the local stern bulb and transom.
¯ow, utilizing the already obtained calibration chart. Bi-
cubic spline interpolation was used to ®nd the correlation 4
coef®cients of the corresponding calibration angles. Results
It was possible to obtain data at 3  6 positions during In the following, the measured wave patterns and local
one carriage run, depending on the model ship speed. mean velocity components are given. The Cartesian co-
However, local ¯ow angles were sometimes outside of the ordinates (X,Y,Z) are used for displaying the data, where X
calibration ranges of ‹40° of pitch and yaw angles, espe- denotes the downstream direction, Y starboard, and Z the
cially below the stern cap and behind the transom. If the upward direction. The origin of the coordinates is located
local ¯ow angle was outside of the calibration range, the at midship and on a calm free surface. All the coordinates
measured data was discarded and velocity components are nondimensionalized by the length between perpen-
could not be determined at those positions. This happened diculars ((Lpp)M) of the model ships. U, V, and W denote
when the axial velocity component was relatively small but the mean velocity components in the X, Y, and Z direc-
the transverse ¯ow angle was large. Measurements were tions, non-dimensionalized by the towing carriage speed
carried out at ®ve planes (St. 2, 1, 0.35, )0.5767, )2) for the (VM).
KCS and at six planes (St. 3, 2, 1, 0.35, )0.4525, )2) for
both KVLCC and KVLCC2. 4.1
Wave patterns
3.4 When a ship advances in calm water, it generates waves on
Experimental uncertainties the disturbed air±water interface (free surface). The ship
Uncertainty analysis was performed for the measured data has to supply energy continuously to the generated wave
following the procedure recommended by ITTC (1999). pattern following the ship. Thus, the ship should overcome
For the resistance test, uncertainty analysis had already the drag induced by the wave generation on the free sur-
 face. The so-called wave resistance (drag) related to this wave elevation near the stern is much ¯atter than in the
phenomenon is important in ship design. bow region. The thick boundary layer existing around the
Modern commercial ships have a bulbous bow to re- stern region suppresses the pressure recovery and re-
duce wave generation around the bow region. The waves strains the wave rising. Moreover, modern container ships
generated by the bow bulb under the free surface interact like the KCS usually have the ¯at overhang pushing down
with the waves generated near the design waterline. It is the water surface in the stern region.
believed that wave resistance can be reduced if the bow The global view of the wave pattern of the KCS is shown
bulb is properly located. The generated waves can be di- in Fig. 7. It should be noted that only the port side was
vided into transverse and divergent wave components measured and re¯ected on the starboard side for plotting.
depending on the wave-propagating direction. The trans- It is observed that a wave exists inside a certain angle
verse wave components have the crest line normal to the known as the Kelvin angle from linear potential wave
572
X-direction with the phase velocity similar to the ship theory. The bow wave having the maximum height on the
speed, while the divergent wave components are located at hull surface gradually disappears, followed by a trough and
the edge of the so-called Kelvin triangle with the shorter crest lines of divergent wave components with decreasing
wavelength and slower phase velocity (SNAME 1967). amplitude along the edge of the wave region. Transverse
However, in reality, it is not easy to clearly identify the two wave components are seen inside the divergent wave
components of the waves generated, since the waves are patterns.
seen in superposed form. An interesting phenomenon in this wave pattern around
There are some efforts to ®gure out the generated wave the KCS occurred behind the transom stern. When the
patterns around a hull form by using CFD techniques. The ship was at rest, the transom stern was above the calm
wave measurements for the KCS and the VLCCs provide water surface, since the design waterline was located below
severe test cases for wave resistance prediction, since they the transom. However, as the ship advanced and reached
contain various aspects of the free-surface phenomena steady state at the design speed, the generated wave rose
around modern commercial ships. Viscous or inviscid over the transom stern. If the speed of the ship is high, the
¯ow solvers with the capability of handling the nonlinear wave surface would be parallel to the hull surface under
free-surface condition can be validated against the present the transom stern. Thus, the transom stern could appear
data to prove the accurate prediction of free-surface ¯ow out of water region, resulting in the so-called ``dry tran-
around an actual hull form. som''. When the speed of the ship is very low, wave gen-
eration could be ignored. However, if the ship advances at
4.1.1
The KCS case
Before the global wave pattern is explored, the wave pro®le
along the hull surface of the KCS is given in Fig. 6. The
Froude number of the experiment was 0.26, thus the cor-
responding length of transverse wave propagating with the
ship speed is 2pFn2, i.e., 0.425 for this case. About two
principal wavelengths are expected along the hull surface.
The crests are located at X ˆ )0.475 and 0.05, while the
troughs are seen at X ˆ )0.15 and 0.25. The ®rst crest is
over 1% of the ship length, while the ®rst trough is )0.5%.
It is observed that a small wave from the forebody
shoulder exists at X ˆ )0.4 just after the ®rst crest. The Fig. 6. Measured wave elevation along KCS (Fn ˆ 0.26)

Fig. 7. Measured wave pattern around

KCS (Fn ˆ 0.26)
the moderate speed as in the present KCS case, waves short waves with the relatively small wave heights. It is
rising over the transom stern and ¯ow reversal behind the observed that the bow wave is dominant and small short
transom make an unsteady and complicated wave pattern, waves follow. For the VLCCs, the contour levels of mea-
as shown in Fig. 7. It is observed that the wave surface sured wave height in Fig. 9 are a quarter of those in Fig. 7
looks like unsteady bubbling up behind the transom. In of the KCS case.
Fig. 7 the hull line is the intersection of the real free sur-The wave resistance of this low-speed ship is usually
face, where the effect of the transom stern is clearly negligible; however, the big bow wave associated with
identi®ed. It will be quite dif®cult to apply numerical wave-breaking phenomena of real ships sometimes draw
modeling to the unstable wave pattern behind the transom attention. However, for the present VLCC hull forms in the
as described above. However, the apparent wave pattern model scale, wave breaking was not observed. Transverse
generated by the real container ships today is very similarwave components are hardly identi®ed. Short divergent
573
to the present results. wave components exist along the edge of the Kelvin tri-
angle. The present VLCCs have the design waterline lo-
4.1.2 cated above the transom, although there is apparently no
The KVLCC and KVLCC2 cases transom effect. But the local ¯ow measurement revealed
Most VLCC hull forms have a large block coef®cient over that the ¯ow angle seemed to undergo an abrupt change
0.8 with a full bow shape and long parallel middle body. due to the transom with ¯ow reversal.
The wave pattern generated by VLCC hull forms with a low
Froude number can be characterized by a big bow wave 4.2
followed by much smaller waves. Wave elevations along Local mean velocity fields
the hull surfaces of KVLCC and KVLCC2 are compared in Most of drag occurs due to viscosity for modern com-
Fig. 8. The discrepancy in the forebody should be con- mercial ships with moderate or low speed. Viscous drag is
sidered as a measurement uncertainty since the forebodies dependent on velocity pro®le and pressure distribution
of the two VLCC hull forms are the same. There is a small inside the turbulent boundary layer. To understand the
difference around X ˆ 0.3 due to the slight difference in relation between hull form and viscous drag, boundary
waterline angle of the two ship models. Wave patterns are layer ¯ows should be explored. Furthermore, the velocity
compared in Fig. 9. It should be noted that the difference distribution in the stern region is very important to design
of wave pattern between the two VLCC models is within the proper propeller with high ef®ciency, since the pro-
the uncertainty level of the present measurement. The pulsor of a modern commercial ship operates inside the
Froude number of the VLCCs was 0.142, resulting in the wake behind stern bulb.
A thin boundary layer is formed in the forebody of the
ship. Flow angles are dependent on the shape of the bul-
bous bow and shoulder in the forebody. The boundary
layer becomes thicker along the hull surface and parallel in
the midship region. Flows are swept upwards from the
bottom in the stern bilge region, where bilge vortices are
formed. For ®ne slender ships like the KCS, the weak bilge
vortices are found at the propeller plane. However, if the
sectional area change is large at the stern, as in the case of
VLCCs, strong bilge vortices can be located inside the
propeller plane. The location and strength of these bilge
Fig. 8. Measured wave elevation along KVLCC and KVLCC2 vortices play the key role in determining the nominal wake
(Fn ˆ 0.142) distribution and the resulting propulsive ef®ciency.

Fig. 9. Measured wave pattern around

KVLCC and KVLCC2 (Fn ˆ 0.142)
 As mentioned earlier, the two VLCC hull forms have the prevent the abrupt change after stern boss. The measure-
same forebody and slightly different afterbodies. The dif- ment was carried out across the center plane to con®rm
ference in wake distribution of KVLCC and KVLCC2 will the symmetry of ¯ow. Axial velocity contours and trans-
provide valuable information to explore the in¯uence of verse vectors show good symmetry with respect to the
stern frame line change on wake distribution. Viscous ¯ow Y ˆ 0 plane. The notable fact in the axial velocity contours
solvers can be applied in the initial hull form design to at X ˆ 0.4 is a thick boundary layer on the concave sur-
predict the turbulent boundary layer and the wake around face. It is likely that streamlines from bilge area converge
a ship. However, before practical application of such onto the concave surface after the midship, resulting in a
computational tools, they should be validated with the thickening of the boundary layer thereafter. It is also
reliable data of realistic hull forms. The present experi- observed that the boundary layer is very thin along the
mental data will provide good test cases to ascertain convex keel region, since low-momentum ¯uids are moved
574
whether physical and numerical modeling can predict towards the concave side.
boundary layer and wake ¯ows around modern hull forms. The axial velocity contours at the propeller plane
(X ˆ 0.4825) are of round shape, which is favorable to the
4.2.1 propeller ef®ciency. The stern bulb clearly affected the
The KCS case wake distribution at the propeller plane. A very weak
The mean velocity components around the stern region of vortex near the upper corner of the propeller hub is ob-
the KCS model are shown in Fig. 10. Five stern stations (St. served, however, the strength is too small to identify from
2, 1, 0.35, )0.5767, )2) of X ˆ 0.4, 0.45, 0.4825, 0.5288, and the ®gure. It is likely that the transverse ¯ow at the pro-
0.6 were chosen for the stern ¯ow measurement around a peller plane is not rolled up enough to form a discrete
modern container ship. A dummy hub was attached to vortex in the KCS case due to the ®ne stern shape. The

Fig. 10a±e. Measured mean velocity ®elds

around KCS. a X ˆ 0.4 (St. 2), b X ˆ 0.45
(St. 1), c X ˆ 0.4825 (St. 0.35), d X ˆ
0.5288 (St. )0.5767), e X ˆ 0.6 (St. )2)
shaft centerline of the model container ship was for both KVLCC and KVLCC2. It is seen that a thicker
located at Z ˆ )0.02913, where the propeller with the boundary layer than in the previous KCS case is formed on
radius RP ˆ 0.03435 (i.e., 7.9 m for the prototype and the concave surface with low-momentum ¯uids fed from
25 cm for the model ship) would be located. The propeller the bilge region. It is likely that the trace of the full mid-
hub is marked as a semi-circle in Fig. 10c. The transverse ship section still remains for both ships. There is not a big
velocity vectors at X ˆ 0.5288 in the wake region illustrate difference before X ˆ 0.45 between the two VLCC hull
the vortex more clearly than in the propeller plane. The forms.
next station at X ˆ 0.6 shows the wake ®eld diffusing A distortion of axial velocity contours is found just
slowly. above the stern bulb at X ˆ 0.45. The boundary layer
Velocity pro®les at the propeller plane are measured pro®les of KVLCC2 show more distortion than those of
separately using a nominal wake-measuring device using KVLCC, as expected from the stronger turn of the bilge
575
the same Pitot tubes. Details of nominal wake distribution shape. At X ˆ 0.4825, the distortion of axial velocity
are given in Fig. 11. The measured propeller plane wake is contours is clearly observed near the stern bulb (Z ˆ )0.04,
shown in local propeller coordinates (YP, ZP), where the Y ˆ )0.015) of both ship models. The strong distortion of
inner and outer circles denote 30% and 110% of the pro- axial velocity contours implies the formation of bilge
peller radius (RP), respectively. The axial velocity contours vortices. Both ships show hook-like axial velocity contours
show the near-wake ®eld behind the stern bulb. Most and strong transverse vortical ¯ows are seen. The distor-
transverse ¯ow vectors are directed upward and inward, as tion of velocity contours is severer in KVLCC2, implying
expected from hull tangents in the stern region. Downward that stronger bilge vortices are formed. The formation of
¯ow vectors are observed above the hub along the plane of longitudinal vortices are well identi®ed in the axial vor-
symmetry and the transverse velocity vectors shows weak ticity contours at X ˆ 0.4825 given in Fig. 13. A vortex
bilge vortices. core is more clearly seen in KVLCC2, while KVLCC has a
narrower vortex region. Two contra-rotating bilge vortices
4.2.2 are formed since the hull form is symmetric with respect to
The KVLCC and KVLCC2 cases Y ˆ 0 plane. The relationship between the stern hull form
For the VLCCs, the stern boss cap of the semi-sphere was and the strength and location of bilge vortices is very
attached instead of a dummy hub, since the hub made it important in hull form design, since it will determine the
dif®cult to measure the velocity around the shaft center- in¯ow to the propeller blades.
line. The propeller plane (X ˆ 0.4825, St. 0.35) was located There are kinks in the axial velocity contours at
just behind the stern cap. The shaft centerline of the Y ˆ )0.04 of the X ˆ 0.4825 plane of KVLCC. However, it
VLCCs was located at Z ˆ )0.04688. The measured mean is believed that these arise from possible errors in data
velocity components around the stern region of KVLCC acquisition or the location of probes during the experi-
and KVLCC2 are compared in Fig. 12. Mean velocity ®elds ment, and thus should be ignored. It should be noted that
were measured at six stern stations (St. 3, 2, 1, 0.35, the measurement points with local ¯ow angle larger than
)0.4525, )2) of X ˆ 0.35, 0.4, 0.45, 0.4825, 0.5226, and 0.6 ‹40° are discarded since they are outside the calibration
range. This occurred below the stern cap and behind the
transom.
It is noteworthy that low-momentum ¯uids accumulate
near the free surface. A thick boundary layer is already
observed at X ˆ 0.45 and X ˆ 0.4825 near the free surface.
A weak vortical structure in the axial vorticity contours in
Fig. 13 is found. This thick boundary layer near the water
plane was also reported in the wind tunnel measurement
for KVLCC2 (Van et al. 1999); thus, it is certain that these
low-momentum ¯uids do not stem from the free-surface
effect. Moreover, the Froude number of the present cases
is very low (Fn ˆ 0.142). It is believed that the thick
boundary layer near the water plane is due to ¯ow con-
vergence along the upper edge of hollow stern frame lines.
The near wake regions of X ˆ 0.5226 and X ˆ 0.6
planes show wake pattern. The axial velocity component in
the centerline is recovered up to 0.7 at X ˆ 0.6. As the ¯ow
goes downstream, the shear layer of wake ¯ow is expanded
and the center of longitudinal vortex moves downward.
This can be explained with the entrainment of inviscid
¯ow into the wake region. The velocity outside the longi-
tudinal vortex region increases, while the retarded ¯ow in
the core of wake is moved downward. Thick boundary
layers near the water plane still exist and diffuse slowly.
Fig. 11. Measured wake distribution of KCS at the propeller The axial velocity at the propeller plane is extracted and
plane compared in Fig. 14. KVLCC has more vertical contour
 576

Fig. 12a±f. Measured mean velocity ®elds around KVLCC and KVLCC2. a X ˆ 0.35 (St. 3), b X ˆ 0.4 (St. 2), c X ˆ 0.45 (St. 1), d X ˆ 0.4825 (St. 0.35), e X ˆ 0.5226 (St. )0.4525),
f X ˆ 0.6 (St. -2)
 Fig. 13. Axial vorticity contours around
KVLCC and KVLCC2 at X ˆ 0.4825 577

frame lines. This distribution is more favorable and pref-

erable for higher propulsive ef®ciency, although strong
bilge vortices can increase the form drag of a ship.
Resistance and self-propulsion test results con®rmed
the above observations. The same ship models used for
local measurement were used for the self-propulsion test
with a stock propeller, which was previously used for other
VLCCs with similar wake distribution. The diameter of the
propeller is 0.0308125 (i.e., 9.86 m for the real ship and
0.17 m for the model ship). Full-scale prediction was
performed following the ITTC method (ITTC 1999). The
results are compared at the design speed (15.5 knots) in
Table 4. The resistance of KVLCC2 is slightly greater than
that of KVLCC due to the generation of stronger bilge
vortices. However, the quasi-propulsive ef®ciency of
KVLCC2 is higher due to the favorable wake distribution.
The effective wake of KVLCC2 is larger than that of
KVLCC and the hull ef®ciency is higher for KVLCC2.
Consequently, slightly less power is required for KVLCC2.
As noted earlier, the difference in stern hull forms of
KVLCC and KVLCC2 is as much as shipyards usually
apply for hull form optimization. The global features of
the stern ¯ows are similar, however, the changes of ¯ow
Fig. 14. Propeller plane wake distribution of KVLCC and are notable at the propeller plane. It will be very inter-
KVLCC2 esting to see whether the numerical modeling can tell the
difference in the propeller plane wake of KVLCC and
shapes than KVLCC2. The depth of hook is clearer at KVLCC2.
U ˆ 0.4 and 0.35 contours of KVLCC2. This difference in
nominal wake distribution will contribute a notable 5
change of the propulsive ef®ciency. It should be men- Concluding remarks
tioned that velocities at some locations below stern cap in In order to provide a benchmark database for CFD vali-
the propeller plane could not be determined since the local dation in ship hydrodynamics, global force, wave pattern,
¯ow angle was out of calibration range. Thus, the velocity and local mean velocity components around a 3,600-TEU
components at the lower part of 0.3 Rp and 0.4 Rp position container ship model (KCS) and two 300K VLCC models
is extrapolated. However, authors believe that the extrap- (KVLCC and KVLCC2) were measured in the towing tank.
olation does not affect the results much. The shape of axial These hull forms are very similar to commercial ships of
velocity contours of KVLCC2 in Fig. 14 shows clearer hook today. Focus was directed on wave generation on the free
shape and becomes closer to a concentric circle than surface for the KCS, while for the KVLCC and KVLCC2,
KVLCC. Moreover, the nominal wake fraction at the pro- the formation of bilge vortices and the wake difference
peller plane of KVLCC2 is larger due to the U-shaped stern between two similar tanker hull forms were emphasized.

Table 4. Resistance and propulsive coef®cients of KVLCC and KVLCC2 (Vs = 15.5 knots)

Ship Cr (X1000) EHP (PS) DHP wM wS t gH gR gO gD

KVLCC 0.638 18,223 25,467 0.407 0.297 0.197 1.142 1.005 0.624 0.716
KVLCC2 0.660 18,438 25,384 0.443 0.305 0.190 1.165 1.005 0.620 0.726
 For KCS with moderate Froude number (Fn ˆ 0.26), Knaack T (1992) Investigation of structure of Reynolds tensor
wave patterns composed of transverse and divergent wave ®elds in a three-dimensional ¯ow. (in German) Report 527,
components were clearly seen. However, a complicated University of Hamburg. Institute of Shipbuilding (IfS), Uni-
versity of Hamburg, Hamburg, Germany
unstable transom wave ®eld was observed, although the Kodama Y (ed) (1994) Proceedings of CFD workshop, Tokyo,
transom stern was located initially above the design wa- Japan
terline. For KVLCC and KVLCC2 with low Froude number Larsson L; Patel VC; Dyne G (1991) Ship viscous ¯ow. In: Larsson
(Fn ˆ 0.142) and full bow hull form, a big bow wave was L, Patel VC, Dyne G (ed) Proceedings of SSPA-CTH-IIHR
observed followed by small short waves. Workshop, 1990, Gothenburg, Sweden
Measured velocity components revealed the boundary Larsson L; Stern F; Bertram V (2000) Gothenburg 2000: a work-
layer structure developing over the stern and evolving into a shop on numerical ship hydrodynamics. Gothenburg, Sweden;
(see also http://www.iihr.uiowa.edu/gothenburg2000/V)
complicated three-dimensional wake ¯ow. A thick bound-
578 Lee J; Lee SJ; Van SH (1998) Wind tunnel test on a double-deck-
ary layer was formed along the concave surface in the stern shaped ship model. In: Kim H, Lee SH, Lee SJ (eds) Proceedings
region. A weak vortex was observed with a thick boundary of the 3rd International Conference on Hydrodynamics
layer near the free surface. For the KCS with a ®ne stern (ICHD), 1998, Seoul, Korea. pp 815±820
shape, the formation of bilge vortices was not so prominent. Longo J; Stern F (1996) Technical note: evaluation of surface-ship
However, in the case of the VLCCs, strong bilge vortices resistance and propulsion model-scale database for CFD vali-
were formed and distorted the axial velocity pro®les in the dation. J Ship Res 40: 112±116
Longo J; Stern F (1999) Resistance, sinkage and trim, wave
propeller plane. The difference in wake distribution and its pro®le, and nominal wake and uncertainty assessment for
relation to propulsive ef®ciency was also explored from the DTMB model 5512. In: Proceedings of the 25th ATTC, 1999,
viewpoint of stern hull form modi®cation. Iowa City, USA
The experimental data presented in the paper were Longo J; Stern F; Toda Y (1993) Mean-¯ow measurements in the
recommended as a benchmark database by the Interna- boundary layer and wake and wave ®eld of a Series 60 CB=0.60
tional Towing Tank Conference and used as test cases in ship model ± Part 2: scale effects on near-®eld wave patterns
the Gothenburg 2000 CFD Workshop in Ship Hydrody- and comparisons with inviscid theory. J Ship Res 37: 16±24
Ogiwara S (1994) Stern ¯ow measurements for the tanker 'Ryuko-
namics. The authors believe that the present experimental Maru' in model-scale, intermediate-scale, and full-scale ships.
data can be used in evaluating physical and numerical In: Kodama Y (eds) Proceedings of CFD Workshop, vol. 1,
modeling to predict wave generation on the free surface 1994, Tokyo, Japan. pp 341±349
and three-dimensional boundary layer and wake around Olivieri A; Penna R (1999) Uncertainty assessment in wave ele-
modern commercial ships with bow and stern bulb. vation measurements. In: ISOPE, 1999, Brest, France
Ratcliffe T (1998) http://www50.dt.navy.mil/5415/
References Society of Naval Architect and Marine Engineers (SNAME)
Bertram V; Chao K-Y; Lammers G; Laudan J (1994) Experimental (1967) Principles of naval architecture. Society of Naval Ar-
validation data of free-surface ¯ows for cargo vessels. In: chitect and Marine Engineers (SNAME), Jersey City, NJ
Kodama Y (ed) Proceedings of CFD Workshop, 1994, Tokyo, Stern F; Longo J; Maksoud M; Suzuki T (1998) Evaluation of
Japan. pp 311±320 surface-ship resistance and propulsion model-scale database
Coleman HW; Steele WG (1999) Experimentation and uncer- for CFD validation. In: Proceedings of the 1st Symposium on
tainty analysis for engineers. Wiley-Interscience, New York Marine Applications of Computational Fluid Dynamics, 1998,
Denker J; Knaack T; Kux J (1992) Experimental and numerical McLean, USA
investigations of HSVA-tanker 2 ¯ow ®eld. Report 521, Uni. Suzuki H; Suzuki T; Miyazaki S (1997) Turbulence measurements
Institute of Shipbuilding (IfS), Hamburg, Germany in stern ¯ow ®eld of a ship model ± series 60, CB=0.60. (in
Dyne G (1995) An experimental investigation of the tanker model Japanese) J Kansai Soc Nav Archit 227: 29±40
`Dyne' in a towing tank. Report CHA/NAVR/-95/0036. Chal- Suzuki H; Miyazaki S; Suzuki T; Matsumura K (1998) Turbulence
mers University of Technology, Sweden measurements in stern ¯ow ®eld of two ship models. In: Hi-
Fry DJ; Kim YH (1985) Bow ¯ow ®eld fo surface ships. In: Pro- meno Y (ed) Proceedings of the 3rd Osaka Colloquium on
ceedings of the 15th ONR Symposium on Naval Hydrody- Advanced CFD Applications to Ship Flow and Hull Form De-
namics, 1985, Hamburg, Germany sign, 1998, Osaka, Japan. pp 229±240
Fujisawa J; Ukon Y; Kume K; Takeshi H (2000) Local velocity ®eld Toda Y; Stern F; Tanaka I; Patel VC (1990) Mean-¯ow measure-
measurements around the KCS Model (SRI M.S.No.631) in the ments in the boundary layer and wake of a series 60 CB=0.60
SRI 400 m towing tank. Ship Performance Division Report No. model ship with and without propeller. J Ship Res 34(4): 225±252
00±003±02. The Ship Research Institute of Japan, Mitaka, Japan Toda Y; Stern F; Longo J (1992) Mean-¯ow measurements in the
Gietz U; Kux J (1995) Flow investigations on the Hamburg test boundary layer and wake and wave ®eld of a series 60 CB=0.60
case model in the wind tunnel. (in German) Report 550, Uni. ship model ± Part 1: Froude numbers 0.16 and 0.316. J Ship Res
Hamburg. Institute of Shipbuilding (Ifs), University of Ham- 36: 360±377
burg, Hamburg, Germany Tsukada Y; Hori T; Ukon Y; Kume K; Takeshi H (2000) Surface
ITTC (1999) Quality manual. In: Yang SI (ed) Proceedings of the pressure measurements on the KCS model (SRI M.S.No.631) in
22nd International Towing Tank Conference (ITTC), 1999, the SRI 400 m towing tank. Ship Performance Division Report
Seoul, Korea/Shanghai, PR China No. 00±004±01. The Ship Research Institute of Japan, Mitaka,
Kang K; et al. (1998) The in¯uence of the stern frame shape for a Japan
high-speed container ship on the powering performance. In: Van SH; Park YJ (1993) Uncertainty analysis for the experiments
Oosterveld MCW, Tan SG (eds) Proceedings of the 7th Inter- in KRISO towing tank. (in Korean) J Ships Ocean Eng 16: 60±70
national Symposium on Practical Design of Ships and Mobile Van SH; Kim WJ; Kim HR; Lee SJ (1999) Wind tunnel test on ¯ow
Units (PRADS), 1998, The Hague, The Netherlands. pp 691±698 characteristics of KRISO 300 K VLCC double model. In: Naito
Kim WJ; et al. (1997) Study on calibration method of ®ve-hole S, Choi HS (eds) Proceedings of the 4th Japan-Korea Joint
Pitot probes for wake measurements. (in Korean) J Soc Nav Workshop on Marine & Ship Hydrodynamics (JAKOM), 1999,
Archit Korea 34: 11±19 Fukuoka, Japan. pp 157±164