Ocean Engineering 108 (2015) 336–349

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

An experimental study on flow separation control of hydrofoils

with leading-edge tubercles at low Reynolds number
Zhaoyu Wei a, T.H. New a,n, Y.D. Cui b
a
School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore
b
Temasek Laboratories, National University of Singapore, Engineering Drive 1, Singapore 117411, Singapore

art ic l e i nf o a b s t r a c t

Article history: Hydrodynamic characteristics of hydrofoils with leading-edge tubercles were experimentally investi-
Received 6 January 2015 gated in a water tunnel at a Reynolds number of Re¼1.4  104. Particle image velocimetry measurements
Accepted 4 August 2015 and particle-streak visualizations reveal that the tubercles improve flow separation behaviour. In
Available online 29 August 2015
particular, hydrofoils with larger wave amplitudes and smaller wavelengths tend to perform significantly
Keywords: better in flow separation control. Cross-stream flow measurements indicate that streamwise counter-
Hydrofoil rotating vortex pairs are generated over the tubercles and mitigate flow separation. Analysis confirms
Passive flow control that the tubercles function as vortex generators, due to their comparable heights relative to the
Leading-edge tubercles boundary layer thickness. The vortex pairs meander and interact with adjacent flows, causing the flow
Flow visualization
separation behaviour to be occasionally unstable, thus leading to variable flow separation region sizes.
Particle-image velocimetry
This suggests that measures may have to be taken to ensure the stability of the counter-rotating vortex
NACA634-021
pairs for more persistent and predictable improvements.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction their control surfaces. Lower Reynolds number flows have thicker
boundary layers and more prone to flow separations, thus often
1.1. Research background leading to poor performance.

In the past decades, drawing inspirations from nature for solutions

to engineering problems have seen increased interests from a number
of new research areas. Lately, a novel passive flow control technique
derived from the morphology of humpback whales has attracted
increasing attention of scientists from different areas (Fish et al.,
2011; Bolzon et al., 2015). The humpback whale is capable of perform-
ing complex underwater manoeuvres to catch prey, including sharp,
high-speed, banked turns, as well as loops and rolls. This has been
attributed to their pectoral flippers, where several tubercles are
distributed along the leading-edge (Fish and Battle, 1995; Fish and
Lauder, 2006). Since whales use pectoral flippers as hydrodynamic
control surfaces (with a Reynolds number of about Re¼ 106), questions
are being raised as to whether these leading-edge tubercles could be
applied on marine engineering devices. In fact, potential marine
applications of leading-edge tubercles have been reviewed by Fish
et al. (2011). Some of these applications include rudders (Weber et al.,
2010), dive planes, stabilizers, surfboard skegs and propellers. Another
potential application of these tubercles is on unmanned underwater
vehicle (UUV) control surfaces. These vehicles often cruise at low
speeds and tend to have lower Reynolds number flows associated with
n
Corresponding author. Tel.: þ 65 6790 4443; fax: þ 65 6792 4062.
E-mail address: dthnew@ntu.edu.sg (T.H. New).
1.2. Literature review
The idealized humpback whale flipper model with leading-
edge tubercles was tested by Miklosovic et al. (2004) in a wind
tunnel at Re¼5  105. They found that the wing with leading-edge
tubercles increases the maximum lift by 6% and delays the stall
angle-of-attack by about 40%, as compared to one without tuber-
cles. Murray et al. (2005) showed that the three-dimensional
flipper models could increase maximum lift values by 9% and 4%
for sweep angles of 151 and 301 respectively, as compared to one
without leading-edge tubercles. Johari et al. (2007) investigated
the effects of tubercle wavelength and amplitude on the flow
separation behaviour of full-span NACA634-021 models in a water
tunnel at Re¼ 1.83  105. It was found that wings with tubercles
reduce the maximum lift coefficient. However, the wings stall
more gently and the lift coefficient could be raised by as much as
50% in the post-stall regime.
Flow visualizations using surface tufts and dye showed that the
flow behind tubercle peaks remains attached to part of the wing
surface in the post-stall regime of the baseline hydrofoil (Johari et
al., 2007; Custodio, 2007). Similar trends for the lift and drag of a
full-span NACA 0020 aerofoil with leading-edge tubercles were
also observed by Miklosovic and Murray (2007) at Re¼2.74  105.

http://dx.doi.org/10.1016/j.oceaneng.2015.08.004
0029-8018/& 2015 Elsevier Ltd. All rights reserved.
 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
In addition, using dye visualization technique, Custodio (2007)
showed that each tubercle behaves like the leading-edge of a small
delta wing and a counter-rotating vortex pair is created behind
each tubercle. The analogy of a tubercle as a small delta wing was
later confirmed by Stanway (2008). This study also measured the
lift and drag of the whale flipper model with and without
tubercles at Re¼4.4  104 to 1.2  105, where it was found that
there is no significant improvement on the maximum lift except
for the highest Reynolds number. In addition, the flow field around
the flipper model with tubercles was measured by particle image
velocimetry (PIV) technique in that study, where delays in flow
separation and formations of streamwise vortices were shown.
Furthermore, force characteristics of the whale flipper were tested
under dynamic conditions (i.e., as flapping foils).
Later, Ozen and Rockwell (2010) applied leading-edge tuber-
cles onto a flapping wing and PIV results showed that the
tubercles could attenuate the positive and negative spanwise
flows along the wing surface, as well as demonstrating the onset
and developments of the large-scale concentrations of positive
and negative streamwise vortices at inboard locations. Goruney
and Rockwell (2008) have also implemented them on a delta
wing in an attempt to look into their flow effects on the flow
structures and topology along the wing surface. Hansen et al.
(2011) investigated the influence of leading-edge tubercles on the
aerodynamic characteristics of two NACA aerofoils. They sug-
gested that the tubercles behave in a manner similar to conven-
tional vortex generators and that the optimum tubercle
amplitude and wavelength bear strong resemblances to the
optimum vortex generator spacing and height. Zhang et al.
(2014) investigated the effects of leading-edge tubercles on the
aerodynamic performance of a full-span NACA634-021 aerofoil in
a wind tunnel for a wide range of angles-of-attack. Particle image
velocimetry results confirmed the formation of streamwise
counter-rotating vortex pairs, though it should be highlighted
that only one measurement location was considered. Further-
more, Delgado et al. (2014) tested a unity aspect-ratio wing with
a sinusoidal leading-edge at Re ¼1.4  105 using stereoscopic PIV
to aid their numerical modelling.
More recent experimental studies include that by Custodio
et al. (2015). They conducted a systematic testing of the hydro-
dynamic forces on finite-span wings with leading-edge tubercles
at Reynolds numbers up to Re¼4.5  105, as well as making use of
three different planforms and a wide range of tubercle geometries.
Besides the above experimental works, Van Nierop et al. (2008)
developed a mathematical model based on lifting-line theory to
investigate the stall characteristics of semi-span wings with
tubercles. The control mechanism has been attributed to modifi-
cations to the pressure gradient on the suction side of the wing.
Pedro and Kobayashi (2008) used detached-eddy simulation (DES)
to simulate the flow field over a flipper model at Re¼5.0  105 and
they noticed that the tubercles could reduce the flow separation
near the tip of the flipper. Similar simulation work was also
conducted by Weber et al. (2011) at Reynolds number between
5.05  105 and 5.2  105. Câmara and Sousa (2013) used DES
Table 1
Configurations and experimental conditions for hydrofoils with leading-edge tubercles.
Hydrofoil label A (mm) λ (mm) A/λ U0 (m/s) Re
NACA634-021 baseline – – – 0.19 1.4  104
A9λ18.75 9 18.75 0.48 0.19 1.4  104
A9λ37.5 9 37.5 0.24 0.19 1.4  104
A1.875λ18.75 1.875 18.75 0.1 0.19 1.4  104
A1.875λ37.5 1.875 37.5 0.05 0.19 1.4  104
337
technique and evaluated the aerodynamic characteristics of wings
with leading-edge tubercles at Re¼1.6  105, while the case with
Reynolds number up to 3  106 was also considered by Dropkin
et al. (2012).
These above CFD simulations demonstrate that they can
provide information on the surface pressure distributions and 3D
flow structures which are not readily available from experimental
and theoretical investigations. Numerical simulations were con-
ducted by Rostamzadeh et al. (2014) to investigate the formation
mechanism of streamwise vortices and surface pressure distribu-
tions were also captured in wind tunnel testing. They found that a
skew-induced mechanism accounts for the generation of stream-
wise vortices and the development of these vortices is accompa-
nied by flow separation in delta-shaped regions close to the
trailing-edge. More interestingly, they made use of critical-point
theory to depict and explain the flow topology above the wing
surface. Skillen et al. (2015) used large-eddy simulation (LES) and
studied infinite-span wings with leading-edge undulations at
Re ¼1.2  105 and the results showed that a secondary flow toward
the regions behind the troughs are driven by the spanwise
pressure gradient. The stall delay properties behind the peak were
explained by the entrainment of higher-momentum fluid and the
re-energization of the boundary layer.
1.3. Current research motivation
It should be noted that although much work has been done on
the implementation of leading-edge tubercles on wings, the flow
separation behaviour still has not been entirely clarified. Further-
more, most of the previous studies on either aerodynamic or
hydrodynamic applications were performed at relatively high
Reynolds numbers. It should be mentioned that in an earlier
study by the authors (New et al., 2014b), a full-span NACA634-
021 wing with leading-edge tubercles has been investigated in a
water tunnel at Re ¼1.4  104, where PIV and particle-streak
visualization techniques were applied to reveal some preliminary
understanding on the flow separation behaviour. In this work
however, the hydrodynamic characteristics of the hydrofoils with
leading-edge tubercles are considered at a similar Reynolds
number but with a wider variety of tubercle amplitudes and
wavelengths. The aim is to reveal the fundamental flow beha-
viour of tubercles and to relate them to the resulting flow
separation behaviour. It should also be kept in mind that the
flow-fields associated with these tubercles will be highly three-
dimensional. Hence, to avoid the limitation of two-dimensional
PIV and particle-streak photography techniques, three laser sheet
positions along the tubercle trough, peak and a location mid-way
between them were included during the streamwise experi-
ments. For cross-stream experiments, the laser sheet planes were
aligned at four different chordwise locations downstream of the
leading-edge to elucidate the nature of the streamwise counter-
rotating vortex pairs.
α Measurement planes
Streamwise Cross-stream
01, 101 151, 201 Mid-span –
01, 101 151, 201 Trough (z/s¼ 0.50) Mid (z/s ¼ 0.48) Peak (z/s ¼ 0.46) x/c¼ 0.12, 0.25, 0.38, 0.52
01, 101 151, 201 Trough (z/s ¼ 0.50) Mid (z/s ¼ 0.46) Peak (z/s ¼0.44) –
01, 101 151, 201 Trough (z/s¼ 0.50) Mid (z/s ¼ 0.48) Peak (z/s ¼ 0.46) –
01, 101 151, 201 Trough (z/s ¼ 0.50) Mid (z/s ¼ 0.46) Peak (z/s ¼0.44) –
338 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
2. Experimental details
2.1. Design of hydrofoils
The baseline hydrofoil in the present study was based on a
NACA634-021 profile, which has been studied by other researchers
(Johari et al., 2007; Zhang et al., 2014) at higher Reynolds numbers
due to its close resemblance to the humpback whale pectoral flipper.
Two sets of aluminium hydrofoils with tubercles were considered
(see Table 1). The mean chord and span for all hydrofoils are
c¼75 mm and s¼ 300 mm, respectively. For the first set, the wave-
length is λ ¼0.25c, and amplitudes are A¼0.025c and 0.12c. For the
second set, the wavelength is λ ¼0.5c, with amplitudes remaining the
same with the first set. These amplitudes and wavelengths fall within
the range of values associated with humpback whale flippers (Fish
and Battle, 1995; Johari et al., 2007). The design methodology for the
hydrofoils was similar to that used by Lohry et al. (2012). A Cartesian
coordinate system was first defined with its origin located at the
lower end of the mean leading-edge (see Fig. 1). x is on the lower end
face pointing towards the trailing-edge, z was on the leading-edge
and pointing upwards, and y is determined by the right-hand law.
The trailing-edges of hydrofoils are located at xTE ¼ cand the
leading-edge with n sinusoidal waves can be described as


2nπ z
xLE ¼ A sin ; 0 rz r s: ð1Þ
s
where A is the amplitude of the sinusoidally-shaped tubercles and
z is the coordinate along the hydrofoil span.
To create the hydrofoils with tubercles, cross sections of the
baseline hydrofoil was modified by a non-linear shearing trans-
formation which maintained the leading-edge radius of the base-
line hydrofoil, position of maximum thickness, profile behind
maximum thickness point and continuity of the cross section
profile over the maximum thickness point. The nonlinear shear
transformation equations used are
( dynamic characteristics of the hydrofoils, both particle-streak
x þ 0:5xLE ð1 þ cosðπ x=0:3cÞÞ; 0 r x o 0:3c
x1 ¼ ; ð2Þ
x; x Z 0:3c
where x is the abscissa of baseline profile and x1 is abscissa of the
modified cross section. The derivatives of Eq. (2) can be expressed
as
(
dx1 1  5xLE π sinðπ x=0:3cÞ=3c; 0 r x o 0:3c
¼ : ð3Þ
dx 1; x Z 0:3c
Note that Eq. (3) maintains the lead-edge radius, as well as the
smoothing of the two sections over maximum thickness point and
boundary conditions. The dimension definitions, as well as the
Trough-plane and Peak-plane cross sections obtained along the
peak and valley of the sinusoidally-shaped leading-edge are
shown in Fig. 1.
2.2. Water tunnel facility
The experiments were conducted in a low-speed recirculating
water tunnel, where the test section size measured 450 mm (W) 
Fig. 1. (a) 2D plan view and (b) 3D view of a hydrofoil with tubercles, (c) Trough- and Peak-plane cross sections obtained by the non-linear transformation method.
600 mm (H)  1100 mm (L) (see Fig. 2). The two sides and bottom
walls were made of tempered glass, which allowed excellent
optical access. In addition, there was another 40 mm (W) 
250 mm (H) tempered glass window at the back of water tunnel
for cross-stream measurements/visualizations. The free-stream
velocity was regulated by an axial pump with a frequency-
inverter speed controller. Before entering into the 4-to-1 ratio
contraction section and test-section, flow was conditioned by
several honeycomb structures and fine-mesh screens to ensure
low free-stream turbulence intensity. In the present study, the
free-stream turbulence intensity was determined to be approxi-
mately 1.1% at U0 ¼0.19 m/s from time-averaged PIV measure-
ments taken for the region-of-interest. In addition, the uncertainty
in free-stream Reynolds number has also been ascertained to be
approximately 71.2%. All experiments were conducted at
U0 ¼ 0.19 m/s and the corresponding Reynolds number was
Re¼ U0c/ν ¼1.4  104, where c is the mean hydrofoil chord and ν
is water kinematic viscosity at working conditions. Note that lift
and drag force measurements were not performed here, as the
present focus would be on the flow separation behaviour.
To mount the hydrofoils for testing, two flat transparent acrylic
plates were installed beneath free-stream water surface and the
hydrofoils were installed between these two flat plates. Note that
the upper plate was covered by black adhesive papers to reduce
laser reflection and to provide a black background for greater
contrast against the illuminated particles. In order to control the
angles-of-attack precisely, a computer-controlled high-torque
stepper motor (MDrives34ac Hybrid) was mounted rigidly to
the upper end-plate, with its shaft coupled to the test hydrofoil
(see Fig. 2(a)). The angle-of-attack, α, was positive in the clockwise
direction from the top-view.
2.3. Flow visualization and particle image velocimetry procedures
To evaluate the effects of leading-edge tubercles on the hydro-
photography and PIV techniques were applied. They were used
to show the development of the vortices and flow separation
regions in both x–y and y–z planes. Fig. 2 shows the schematic
diagram of particle-streak photography experimental setup in
both cross-stream and streamwise experiments.
For particle-streak photography experiment, 20 μm polyamide
particles were uniformly seeded in water. A 2 W, 532 nm
continuous-wave diode-pumped solid-state laser in conjunction with
beam-steering and sheet-forming optics was used to create a 2 mm
thick laser sheet to illuminate these particles. Using relatively long
exposure timing, the streaks of these particles were captured by a
Canon 550D Digital Single-Lens-Reflex (DSLR) camera with a Canon
EF 50 mm f/1.4 USM lens. For streamwise visualization (x–y plane),
the laser sheet was aligned perpendicular to the hydrofoil span. The
camera was positioned beneath the test section (Fig. 2(b)). For cross-
stream visualizations, the laser sheet was aligned perpendicular to
free-stream direction, and the camera with a 100 mm f2.8 lens was
located behindthe small window at the back of the water tunnel. For
streamwise visualization, the laser sheet was aligned at three different
spanwise positions named as Trough-plane, Mid-plane and Peak-
 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
Fig. 2. Schematics of the experimental setup and apparatus for (a) cross-stream and (b) streamwise particle-streak photography/PIV experiments.
plane. Trough-plane is the spanwise location that coincides with the
middle of all troughs present on the leading-edge of hydrofoils, which
also corresponds to the mid-spans of all hydrofoils. The Peak-plane is
the spanwise location coinciding with the peak point closest to the
Trough-plane in the lower half parts of the hydrofoils. The Mid-plane
is at the inflection point of the same tubercle between the Trough-
plane and Peak-plane. For cross-stream experiments, the laser sheet
positions were referenced from the quarter-chord location.
For particle-streak photography experiments, the DSLR camera
was connected to a computer and controlled remotely by a
proprietary software, which provided full user control over the
camera shutter, aperture and ISO settings. Images were taken
using an exposure time of Δt¼ 1/25 s. When the images were
taken at this exposure time, they could indicate the overall flow
separation bubbles and instantaneous flow-fields. It should be
noted that 30 s videos were also taken alongside the images to
assess the unstable nature of flow-fields as well, which had proved
to be highly useful in shedding light upon the flow structures
associated with different highly dynamic flow fields (New and Tsai,
2007; New and Tsovolos, 2011, 2012; Shi and New, 2013).
Particle image velocimetry was used to measure the velocity and
vorticity fields. The setup for PIV experiment was almost the same
with particle-streak photography experiment in Fig. 2. The continuous
wave laser was replaced by a Litron double-pulsed Nd:YAG laser with
a maximum energy of 200 mJ/pulse. The instantaneous particle
images were captured by a Dantec Dynamics FlowSense 2M/E
double-frame CCD camera attached to a Nikon f/1.4D 50 mm lens,
which was synchronized with the double-pulsed Nd:YAG laser by a
Dantec Dynamics PIV Timer Box connected to a workstation equipped
with timer and image-grabber boards. Both streamwise and cross-
stream PIV experiments were carried out to interpret the flow
dynamics. Similar measurement procedures can be found in a wide
variety of different flow scenarios previously (Soria et al., 2003; New
and Tay, 2006; New and Tsovolos, 2009; Shi et al., 2014; New et al.,
2014a, 2014b, take for instance). For streamwise cases, the three
measurement planes corresponded to the Trough-plane, Mid-plane
and Peak-plane in particle-streak photography experiments. A total of
615 image-pairs (i.e., instantaneous velocity fields) were captured for
each angle-of-attack to ensure satisfactory convergence in the mean
flow field. The physical measurement window was approximately
120 mm  90 mm, mapped out by the 1600  1200 pixel CCD
camera.
The time interval between two frames was 1450 ms, which was
set smaller than the time required for the particles to translate 20%
of the interrogation window (i.e. 32  32 pixel). During post-
processing procedures, 2-step refinement multi-grid cross-corre-
lation technique with initial and final interrogation window sizes
of 128 pixel2 and 32 pixel2 were applied, where the square inter-
rogation areas overlapped 50% in both horizontal and vertical
339
directions, resulting in 99  74 vectors in each velocity map. The
vorticity component ωz was evaluated by a central-differencing
scheme on the PIV measured velocity maps. Based on the PIV
measurement procedures, velocity map resolution was approxi-
mately 1.25 mm/vector. In addition, moving-average validation
was applied to validate vectors by comparing each vector with
the average of other vectors in a defined neighbourhood. Erro-
neous vectors would be replaced with vectors interpolated from
surrounding ones. According to the above post-process techniques,
the uncertainty levels of measured velocity components were
limited to within 7 1% (Keane and Adrian, 1992). Considering
the flow conditions, experimental setup (camera, laser and stepper
motor) and image processing, the total experimental uncertainty
in the velocity measurements and vorticity calculation were
estimated to be approximately 3.6% and 4%, according to the
principles described by Moffat (1988).
For cross-stream PIV cases, the laser sheet was aligned normal
to the free-stream direction, and a 105 mm f2 lens was used
instead. Time interval between the two frames was chosen such
that most of particles would remain illuminated by the laser sheet
within this time interval. As such, a time-interval of 10,000 μs was
used and a total of 1000 image pairs were captured for each case.
The physical measurement window was approximately
130 mm  100 mm. The raw datasets were processed by using
the same procedure as in streamwise experiment, resulting in
99  74 vectors in each velocity map. The vorticity component, ωx,
was also evaluated from the PIV velocity maps.
3. Results and discussion
Particle-streak flow visualization images highlight the charac-
teristic flow features for hydrofoils with leading-edge tubercles
and provide a first-hand appreciation of the flow fields. Since
these images only show the flow patterns in relatively short time
durations, it is difficult to shed light on the mean near-field flow
structures. Thus, mean flow field from PIV data is used to reveal
the overall flow separation. The rationale behind these reported
results is to illustrate the important flow features and physics
associated with tubercles, as well as to serve as comparisons
between different tubercle configurations.
3.1. Flow control behaviour for hydrofoil A9λ18.75
In this section, to reveal the fundamental mechanism and flow
separation behaviour of hydrofoils with leading-edge tubercles,
the A9λ18.75 hydrofoil with a relatively larger tubercle amplitude
and smaller wavelength is selected for a closer inspection. As its
tubercle amplitude-to-wavelength ratio of 0.48 is relatively large,
340 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
Fig. 3. Instantaneous streamwise particle-streak photography flow fields along the upper surface at α ¼01, 101, 151 and 201 for (a) baseline NACA634-021 hydrofoil along its
mid-span, (b) A9λ18.75 hydrofoil along its Trough-plane, (c) Mid-plane, and (d) Peak-plane.
very pronounced streamwise counter-rotating vortex pairs (CVPs)
would be produced in the presence of a free-stream. Particle-
streak photography images and mean vorticity field with stream-
lines from PIV results are presented in Figs. 3–6. These figures
include the streamwise and cross-stream results at four angles-of-
attack of α ¼01, 101, 151 and 201. The results for the baseline
hydrofoil are also included as a benchmark for comparison
purposes.
For the baseline hydrofoil (Fig. 3(a)), adverse pressure gradient
due to the relatively thick profile exerts a significant influence at
the present low Reynolds number. Obvious flow separation can be
detected even at α ¼ 01, where the separation point is located at
approximately half-chord position. As the angle-of-attack
increases to α ¼101, the flow has already separated from the
leading-edge. For angles-of-attack of α ¼ 151 and 201, very evident
recirculating flows could be observed. It was described by Zhang
et al. (2014) that at Re¼2.0  105, the flow remains attached to
most of the baseline NACA634-021 wing surface at α ¼191 (see
Fig. 3(b) in Zhang et al. (2014)). It could be discerned that at such a
low Reynolds number, such a hydrofoil will stall more easily than
those at significantly higher Reynolds numbers (Johari et al., 2007;
Zhang et al., 2014).
Note that the images in Fig. 3 only show the instantaneous flow
separation behaviour. To ascertain that these observations continue to
hold true in a more persistent manner, streamlines and vorticity fields
derived from mean PIV data are presented in Fig. 4 to inspect the
resultant flow separation behaviour. For the baseline hydrofoil, it can
be observed that the PIV results agree well with particle-streak
photography images (see Fig. 4(a)). At α ¼01, the flow separates from
the hydrofoil surface at about half-chord location, and forms a small
flow recirculating bubble. At α ¼ 101, a significant separated shear
layer and flow recirculation bubble are generated. As the angle-of-
attack increases to α ¼ 151 and 201, large flow recirculation regions
can be observed. These results are similar to those of other NACA
hydrofoils, such as the NACA0010 described by New et al. (2014a,
2014b).
When the leading-edge tubercles are introduced, the relatively
stable two-dimensional flows give way to unstable three-
dimensional flow behaviour. As shown in Fig. 3(b), along the
Trough-plane of the A9λ18.75 hydrofoil, the flow always separates
from the leading-edge for the four angles-of-attack. It has to be
mentioned that the separation regions are smaller than that of the
baseline hydrofoil. In the particle-streak photography video, it can
be seen that the flow tends to reattach to the hydrofoil surface
once separated flow is produced. In addition, no obvious recircu-
lating flow can be observed from the video (i.e. not shown here for
the sake of brevity). In Fig. 4(b), strong separated shear layers
could be noticed at α ¼01 and 201. Closer inspection of the
streamlines shows that no flow separation bubble is generated.
As soon as the flow separates from the leading-edge, it reattaches
to the hydrofoil surface. In addition, there is a critical point
observed in the mean streamlines in Fig. 4(b) for α ¼151. As
described by Rostamzadeh et al. (2014), it is in fact a saddle point
showing flow separation in a delta-shaped region near the trailing
edge in three-dimensional cases.
Along the Mid-plane of the A9λ18.75 hydrofoil shown in Fig. 3(c),
it is noticed that the mildly separated flow remains mostly attached to
the hydrofoil surface at α ¼01, and the flow attaches to the entire
hydrofoil surface at α ¼101. Even when the angle-of-attack increases
to as high as α ¼201, the flow remains attached to the surface till
about 1/3 c from the leading-edge. It can be observed from the video
that no flow recirculation bubble is formed. It should be mentioned
that for this plane at α ¼151 and 201, the flow separation becomes
more unstable as well. Some images show the flow remaining
attached to most of the hydrofoil surface, while other images show
the flow detaching from the surface from about 1/3 c from the
leading-edge. Video playback also shows that the flow only reattaches
to the hydrofoil surface intermittently. During the experiment, more
than 20 images were taken for each case and the one in Fig. 3(c) at
α ¼15 and 201 is a typical representation of the observed flow
separation behaviour. It can be seen from the mean streamlines and
vorticity results (Fig. 4(c)) that the level of flow separation control
 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349 341

Fig. 4. Mean streamlines and vorticity fields ðω z Þ of (a) baseline hydrofoil along its mid-span, as well as A9λ18.75 hydrofoil along its (b) Trough-plane, (c) Mid-plane and
(d) Peak-plane. For each laser sheet position, angles-of-attack of α¼ 01, 101, 151 and 201 are considered.

Fig. 5. Cross-stream particle-streak photography images along the upper surface of A9λ18.75 hydrofoil at α ¼01, 101, 151 and 201along (a) x/c¼ 0.12, (b) x/c¼ 0.25 and (c) x/
c ¼0.38 cross-stream locations.

appears to be more promising. At α ¼01, the flow separates from the
hydrofoil surface at about half-chord location and forms a very weak
shear layer and small flow recirculation bubble. Similarly for α ¼101,
151 and 201, only weak shear layers are generated and the flow
reattaches to the whole hydrofoil surface.
It is interesting to note that flow separation control along Peak-
plane appears to be better. The delay in flow separation along
Peak-plane has already been reported by Johari et al. (2007) and
Zhang et al. (2014). It can be noticed from Fig. 3(d) that the flow
remains mostly attached to the entire hydrofoil surface at α ¼01,
101 and 151. As the angle-of-attack increases to α ¼201, the flow
would still be attached to the hydrofoil surface till about 1/3c
location from the leading-edge. From the corresponding mean
streamlines and vorticity field (Fig. 4(d)), no obvious separated
shear layer is detected and the flow attaches to the whole
hydrofoil surface for all four angles-of-attack. This phenomenon
is quite different from that observed by Zhang et al. (2014) along
the Peak-plane at a higher Reynolds number of Re¼2.0  105,
342 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
Fig. 6. Mean cross-stream vorticity fields ðω x Þ of A9λ18.75 hydrofoil along (a) x/c ¼0.12, (b) x/c ¼0.25, (c) x/c¼ 0.38 and (d) x/c ¼ 0.52 cross-stream locations. Four angles-of-
attack of α ¼01, 101, 151 and 201 are considered.
where separation shear layers could be observed. Interestingly,
there is a critical point lying along the hydrofoil surface at α ¼201,
indicating the presence of strong three-dimensional flows.
As to how exactly leading-edge tubercles alter flow separation
characteristics, some researchers have attributed it to the genera-
tion of streamwise CVPs (Miklosovic et al., 2004; Custodio, 2007;
Stanway, 2008; Hansen et al., 2011; Zhang et al., 2014 and
Rostamzadeh et al., 2014, take for instance). However, the forma-
tion mechanism of streamwise CVPs and their role in flow
separation control remain to be fully revealed. As mentioned
previously, each tubercle behaves like a small delta-wing
(Custodio, 2007 and Stanway, 2008) and thus a CVP is generated
over each tubercle in the presence of a free-stream. This is also the
purpose of presenting cross-stream particle-streak photography
images and mean PIV vorticity field in Figs. 5 and 6. To zoom into
these vortices, only the results for a few tubercles are shown.
However, the readers should keep in mind that the flow behaviour
for other tubercles is generally similar.
In Fig. 5(a), the CVPs can be clearly discerned at α ¼ 101, 151 and
201 when the laser sheet was located at x/c ¼0.12 position. As the
laser sheet moved downstream to the quarter-chord position (i.e.
x/c¼ 0.25), the vortex patterns reveal that no CVPs can be observed
at α ¼01. However, it can be detected that the flow streaks
converge along the Peak-plane but diverge along the Trough-
plane. As the angle-of-attack increases to α ¼101, very regular
CVPs can be observed. Compared to the CVPs for α ¼ 101 at x/
c¼ 0.12 (see Fig. 5(a)), the size of these CVPs increases and they
move away from each other, which agree well with the expected
behaviour of CVPs in general.
At higher angles-of-attack of α ¼151 and 201, some vortex cores
of the CVPs cannot be discerned clearlyand is believed to be a
result of vortex interactions between adjacent CVPs. As the CVPs
evolve, they tend to move away from each other. When two
adjacent vortices near the Trough-plane move closer to each other
such that they interact intensely, enhanced viscous dissipations
will render them incoherent rapidly. In Fig. 5(c), as the laser sheet
was moved to x/c ¼0.38 position, only weak CVPs can be observed
at α ¼ 01. In contrast, no CVPs can be observed at α ¼101, 151 and
201. However, evidence of spanwise flow still exists and the
particle streaks show that they diverge and converge along
Trough-plane and Peak-plane, respectively.
To further ascertain these observations in Fig. 5, attention should
be paid towards the mean vorticity derived from PIV data. In Fig. 6,
mean vorticity results agree well with the particle-streak photogra-
phy images. At x/c¼0.12 and 0.25 locations, these CVPs can be
observed regularly, especially at α ¼01, 101 and 151. Compared to the
results in Fig. 5, it is found that these CVPs do not show up very well
through the particle-streak photography technique at α ¼ 01. In Fig. 6
(c), at α ¼01 and 101, regular CVPs can still be clearly seen. However,
at α ¼151 and 201, some of the CVPs have already dissipated. To
confirm how far these CVPs could travel along the hydrofoil surface,
the laser sheet was also positioned at x/c¼0.52 location (see Fig. 6
(d)). Results show that the CVPs could not be discerned clearly at
α ¼01 and 101. For larger angles-of-attack of α ¼151 and 201, no
coherent vortex structures can be found.
As mentioned earlier, a streamwise CVP is created over each
tubercle. It is also found that this CVP share similar characteristics
with vortices formed by delta-wings and conventional vortex
generators (VGs). They grow in size but decrease in intensity,
which coincides with the results in Figs. 5 and 6. Fig. 7 depicts a
sketch for the formation and evolution of these streamwise CVPs.
As a result of spanwise flow arising from the variation in the
tubercle leading-edge sweep, streamwise CVPs are generated
along the leading-edge close to the peak for each tubercle (see
Fig. 7(a)). From the end-view sketch (see Fig. 7(b)), the vortex on
the left side rotates clockwise, while the one on the right side
rotates counter-clock wise. Due to the tubercle geometry and
characteristics of conventional CVPs, each vortex rotates in a way
that causes them to migrate towards the Trough-planes (Fig. 5(b)).
These CVPs drive flows to converge along the Peak- plane while
 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
Fig. 7. Sketches of vortex structures produced by leading-edge tubercles. (a) Three-dimensional view, (b) end-view.
Fig. 8. Unstable flow separation caused by interactions between counter-rotating vortex pairs along the Mid-plane of A9λ18.75 hydrofoil at α ¼151. Both (a) and (b) flow
instances are captured using a short exposure time of 1/25 s at two different time points.
causing them to diverge along the Trough-plane, which agrees
well with the behaviour described in Fig. 5. The two adjacent
vortices close to the Trough-plane will combine to create low
pressure and high pressure gradient at the troughs of tubercles.
Furthermore, as the cross section along Trough-plane is relatively
blunt, the flow always separates from the leading-edge at the
troughs (Custodio, 2007).
It is also clear from the results presented above that the CVPs
interact with each other as soon as they approach closer to one
another. When they interact, unstable flow separation behaviour
can be observed. This phenomenon is more evident at larger
angles-of-attack. For the present A9λ18.75 hydrofoil, the flow
separation becomes unstable at α ¼151 and 201. Fig. 8 shows
two particle-streak photography images taken at two different
time points for the A9λ18.75 hydrofoil along the Mid-plane at
α ¼151, where the unstable flow separation behaviour can be seen.
In Fig. 8(a), the flow attaches to most of the hydrofoil surface. In
contrast, the flow only attaches to a small part of the hydrofoil
surface as shown in Fig. 8(b) and leads to a sudden flow
separation.
To further ascertain this unstable behaviour, instantaneous PIV
vorticity fields for the baseline and A9λ18.75 hydrofoils along the
Mid-plane at α ¼ 151 are also presented in Figs. 9 and 10,
respectively. For the baseline hydrofoil, the flow separation is
relatively stable with a series of strong separation vortices being
produced. From the image-sequence, they also lead to relatively
consistent flow separation region size (Fig. 9). In contrast, highly
unstable flow separations are observed for the A9λ18.75 hydrofoil
along the Mid-plane, as shown in Fig. 10, where the separation
region varies over time. In Fig. 10(a and b), strong separated
vortices can be noticed with significant separation region sizes.
On the other hand, Fig. 10(e and f) show that as soon as the flow
separates from the leading-edge, it reattaches back to the hydrofoil
surface at approximately half-chord location. The entire vorticity
field sequence also shows that the separation region becomes
smaller as the separated vortices become weaker.
So far, it is known that the formation of CVPs accounts for flow
separation control in hydrofoil with leading-edge tubercles. As
343
these CVPs interact with one another and dissipate, unstable flow
separation will be produced. Despite the mean vorticity results
depicting the CVPs as stable vortical entities, it is expected that
their instantaneous behaviour are not as stable as these in the
mean data. The size and relative position of each vortex core in the
CVPs also vary over time. Figs. 11 and 12 show the distribution of
these instantaneous CVPs over the tubercles at α ¼101 and 151. As
the laser sheet was moved to a position farther behind at x/
c¼ 0.25, the vortices become far more complex and some of the
CVPs become far too weak or incoherent to be observed. Hence,
only the results for the two laser sheet positions at x/c¼ 0.12 and x/
c¼ 0.25 locations are presented. In Fig. 11, the laser sheet position
is closer to the leading-edge, thus the CVPs can be clearly observed
over each tubercle. The vortices vary both in size and position, and
overlapping of vortices can be clearly seen. As indicated in Figs. 11
and 12, typical vortex pairs from two adjacent CVPs along the
Trough-plane are circled and marked with numbers. For Pair 1, the
positive vortex is below the negative vortex, while the positive
vortex is over the negative vortex for Pair 2. For Pairs 3 and 4,
interaction of the two adjacent negative and positive vortices
becomes very pronounced. For the laser sheet position at x/c ¼0.25
(Fig. 12), the size and relative position of the vortices change more
significantly. In Pairs 1 and 6, the positive vortex is above the
negative vortex, while the negative vortex is just above the
positive vortex for Pairs 2 and 3. For Pairs 4 and 5, interaction of
the positive and negative vortices along the Trough-plane can also
be clearly observed.
3.2. Effect of amplitude and wavelength for tubercles
The effects of amplitude and wavelength for the leading-edge
tubercles on flow separation will be presented and discussed in
this section. These results can also be applied to understand the
flow separation and vortex structures for high Reynolds number
cases, such as the study by Johari et al. (2007). The Reynolds
number used by Johari et al. (2007) was Re¼ 1.83  105 and
substantially larger than Re¼1.4  104 in this work. However, the
large-scale vortical structures are fundamentally similar. This
344 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349

Fig. 9. Time-sequenced vorticity fields (ωz) obtained for the baseline hydrofoil at α ¼151, where the first image was arbitrarily set to t¼ 0.0 s and subsequent images are at
t¼ 3.3 s apart. (a) t ¼ 0s (b) t ¼3.3s (c) t ¼ 6.7s (d) t ¼ 10s (e) t ¼ 13.3s (f) t ¼ 16.7s (g) t ¼20s (h) t ¼ 23.3s.

Fig. 10. Time-sequenced vorticity fields (ωz) obtained for the A9λ18.75 hydrofoil at α ¼151 along the Mid-plane, where the first image was arbitrarily set to t¼ 0.0 s and
subsequent images are at t¼ 3.3 s apart. (a) t ¼ 0s (b) t ¼3.3s (c) t ¼ 6.7s (d) t ¼ 10s (e) t ¼ 13.3s (f) t ¼ 16.7s (g) t ¼20s (h) t ¼ 23.3s.

Fig. 11. Instantaneous cross-stream vorticity fields (ωx) of A9λ18.75 hydrofoil along x/c ¼ 0.12 cross-stream location at angles-of-attack of (a) α ¼101 and (b) α¼ 151. (I), (II) and
(III) represent three different arbitrary time points.

phenomenon was also described by Erm (2003), where it was
found that both flow patterns are very similar for Reynolds
numbers differing by over an order of magnitude.
Johari et al. (2007) suggested that as the tubercle amplitude
varies, maximum lift coefficient, stall characteristics and drag
coefficient will vary correspondingly as well. In contrast, varying
the wavelength has little effects. In recent research, it was
noticed that for a given amplitude, there is a limit to how much
hydrodynamic performance improvement that can be gained
through wavelength variations. This means that the amplitude-
to-wavelength ratio, A/λ, is also an important parameter (Hansen
et al., 2011). In fact, as the ratio of A/λ varies, the leading-edge
sweep angle is altered as well, which in turn modifies the
strength of CVPs. The streamwise PIV mean streamlines with
vorticity field for the A1.875λ18.75 hydrofoil is shown in Fig. 13.
This hydrofoil has the same wavelength as the previous
 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349 345

Fig. 12. Instantaneous cross-stream vorticity fields (ωx) of A9λ18.75 hydrofoil along x/c ¼0.25 cross-stream location at angles-of-attack of (a) α ¼ 101 and (b) α ¼ 151. (I), (II)
and (III) represent three different arbitrary time points.

Fig. 13. Mean flow streamlines and vorticity fields ðω z Þ of A1.875λ18.75 hydrofoil along its (a) Trough-plane, (b) Mid-plane and (c) Peak-plane. For each laser sheet position,
angles-of-attack of α¼ 01, 101, 151 and 201 are considered.

A18.75λ18.75 hydrofoil, while the amplitude becomes much
smaller. Similarly, three laser sheet positions and four angles-
of-attack are considered. In Fig. 13, although the flow separation
control is worse than A9λ18.75 hydrofoil, it remains better than
the baseline hydrofoil. Along the Mid-plane at α ¼01, the flow
separation point is at about half-chord location, similar to the
baseline case. Along the Trough-plane, the separation point
moves slightly upstream at α ¼ 01, while it moves downstream
for the Peak-plane at α ¼01. The flow separation region size
decreases as the laser sheet location moves from the Trough-
plane to the Peak-plane.
Along the Peak-plane, no flow recirculation bubble is detected
at α ¼101 and 151, and a moderately-sized recirculation bubble
close to the trailing-edge can be observed at α ¼ 201. Similar
phenomenon can also be seen along the Trough-plane at α ¼101
and 151. Along the Mid-plane, the flow separation behaviour
becomes rather irregular. At α ¼101, the separation point is at
about 1/3c from the leading-edge with a resulting large recircula-
tion region. As the angle-of-attack increases to α ¼ 151, the flow
surprisingly remains attached to most of the hydrofoil surface,
with only a very small separation bubble produced. Interestingly, a
large flow recirculation bubble can be observed for α ¼201, formed
346 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
Fig. 14. Mean flow streamlines and vorticity fields ðω z Þ of A9λ37.5 hydrofoil along its (a) Trough-plane, (b) Mid-plane and (c) Peak-plane. For each laser sheet position,
angles-of-attack of α¼ 01, 101, 151 and 201 are considered.
after the flow attaching to the hydrofoil surface till approximately
1/3c location.
Fig. 14 shows the mean PIV streamlines with vorticity field for
A9λ37.5 hydrofoil at three laser sheet positions and four angles-of-
attack. The A9λ37.5 hydrofoil has the same tubercle amplitude
with its A9λ18.75 counterpart, even though the wavelength is
doubled. Compared to the results in Figs. 4 and 13, flow separation
control for A9λ37.5 is better than A1.875λ18.75, but worse off than
A9λ18.75. Similar to the results of A9λ18.75 in Fig. 4, very strong
separated shear layers are generated along Trough-plane at the
four angles-of-attack (see Fig. 14(a)). Close inspection of the
streamlines shows that the flow in fact remains attached to the
hydrofoil surface with no recirculation bubble. In addition, a node
of detachment (ND) can be identified, which is particularly evident
at α ¼201. This ND marks the location where limiting streamlines
converge in three-dimensional flow field, which agree well with
the phenomenon revealed by Rostamzadeh et al. (2014). The flows
along the Peak-plane converge at ND, and they seem to be
emanating from the hydrofoil surface and then diverging in the
two-dimensional streamline figures along the Trough-plane. The
flow even moves upstream, as shown in Fig. 14(a) at α ¼ 201. Along
the Mid-plane, there is only a small flow recirculation region at
α ¼201. The flow attaches to the whole hydrofoil surface at α ¼ 01,
101 and 151. Compared with the results for A9λ18.75 hydrofoil (see
Fig. 4(d)), the flow behaviour is more unstable and irregular along
the Peak-plane. At α ¼01, there is no flow separation. As the angle-
of-attack increase to α ¼ 101, the flow starts to separate from the
hydrofoil surface at about 1/3c from the trailing-edge, where a
small flow recirculation region is formed. As for α ¼151 and 201,
the regular flow separation region could not be observed, the
modified flow topology (critical points) reveals very unstable and
strong turbulent three-dimensional flow behaviours.
Fig. 15 shows the mean PIV streamlines with vorticity field for
the A1.875λ37.5 hydrofoil. Undoubtedly, the tubercles have lost
their effectiveness in terms of flow separation control. For all the
four angles-of-attack and three laser sheet positions, the flow
separation behaviour is almost the same as that of baseline
hydrofoil (see Fig. 4(a)). This once again confirms that
amplitude-to-wavelength ratio, A/λ, plays an important role in
flow separation control. With reference to the results presented,
flow separation improvement has a strong dependence on A/λ,
showing an order of A9λ18.75, A9λ37.5, A1.875λ18.75 and
A1.875λ37.5, in terms of worsening flow separation control
effectiveness. Note that this also coincides with decreasing order
of A/λ.
It was mentioned by Simons (2002) that low Reynolds number
boundary layer flows and separation bubbles are surprisingly
complicated and difficult to predict. Hence, in the present work
where a significantly lower Reynolds number of Re¼ 1.4  104 is
considered, the boundary layer flows and separation bubbles will
be more complicated. The results show that the flow separation
could be significantly suppressed in both pre-stall and post-stall
regimes for A9λ18.75, A9λ37.5 and A1.875λ18.75 hydrofoils, espe-
cially along the Mid-plane and Peak-plane. However, whether the
lift and drag characteristics could be improved needs to be
confirmed in future experiments.
 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
Fig. 15. Mean flow streamlines and vorticity fields ðω z Þ of A1.875λ37.5 hydrofoil along its (a) Trough-plane, (b) Mid-plane and (c) Peak-plane. For each laser sheet position,
angles-of-attack of α¼ 01, 101, 151 and 201 are considered.
This work highlights the formation of CVPs and effect of
tubercles on the near-field hydrodynamic characteristics for a
given hydrofoil profile (NACA634-021) at very low Reynolds
number. From earlier research (Johari et al., 2007; Zhang et al.,
2014), it is clear that the tubercles generate CVPs. It has also been
argued by Van Nierop et al. (2008) that it is implausible for these
tubercles to act as vortex generators, since their wavelength and
amplitude are much larger than the boundary layer thickness, but
instead they alter the pressure distribution on the hydrofoil sur-
face, increasing the extent of flow attachment.
Despite the above view point, results in this work clearly show
that leading-edge tubercles create very regular CVPs and evolution
of these CVPs greatly affects the flow separation behaviour. As
shown in Fig. 7 that just like a small delta-wing, a CVP is created
over each tubercle, and there are also several analogies between
these tubercles and vortex generators (Lin, 2002; Hansen et al.,
2011). The CVPs produced by VGs were found to increase in size
but decrease in intensity along the streamwise direction as shown
Figs. 5 and 6 (Godard and Stanislas, 2006). In fact, both VGs and
performance enhancement with tubercles can be understood
through detailed flow patterns in the boundary layer (Zverkov
and Zanin, 2008). As described by Lin (2002), the device angle has
been found to be an important performance parameter for a vortex
generator, and this parameter has also been suggested to describe
the leading-edge tubercles by Hansen et al. (2011). In this case, the
device angle is also the tubercle sweep angle, governed by the
ratio of A/λ. It is expected that there exists an optimal amplitude-
to-wavelength ratio, and the best hydrofoil performance
347
characteristics can possibly be achieved when this is coupled with
the best amplitude. In this study, based on the performance of
flow separation control, the best tubercle configuration can be
found in A9λ18.75 hydrofoil.
It can be seen from the above results that the vortex generating
mechanism for tubercles would become more prominent at larger
angles-of-attack. Hansen et al. (2011) suggested that the effective
device height to boundary layer ratio, heff/δ, is an important
parameter to describe the tubercles. The value of heff increases
for larger angle-of-attack, and the tubercles will perform more like
vortex generators. This parameter could also be used as a reference
to determine the optimum tubercle amplitude for a given hydro-
foil profile. For each angle-of-attack, heff represents the height of
tubercles seen by free-stream, it is 2Asinα or (cPeak-plane - cTrough-
plane) sinα. δ is the boundary layer thickness obtained at a distance
of A from the leading-edge of baseline hydrofoil in the mean PIV
velocity field.
The procedures for obtaining the boundary layer thickness
will be briefly explained here. The cross section profile of baseline
hydrofoil was firstly fitted in the mean velocity field. Then, a
point at the chordwise location of x ¼ A on the baseline hydrofoil
surface was marked. A line passing this point and normal to the
hydrofoil surface was further drawn in the mean velocity field.
Finally, the entire velocity values along this line were exported,
and the boundary layer thickness corresponds to the distance
from the hydrofoil surface to a location on the line where velocity
is 99% of the free-stream velocity. The boundary layer thickness
for hydrofoils with A ¼ 9 mm, was determined as 4.17 mm,
348 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
Fig. 16. Dependence of heff/δ on angle-of-attack α for hydrofoils with leading-edge
tubercles. Results for two different wave amplitude are included.
5.96 mm and 7.2 mm for α ¼101, 151 and 201, respectively. Note
that these values are smaller than the tubercle amplitude. For
hydrofoils with amplitude of A ¼ 1.875 mm, the thickness of
boundary layer was estimated to be 1.976 mm, 1.973 mm and
2.52 mm, for α ¼ 101, 151 and 201, respectively. Obviously, these
values are close to the amplitude. Fig. 16 shows the dependence
of parameter heff/δ upon the angle-of-attack α for the hydrofoils
with tubercles considered in this work. It is evidently seen that
for the two wave amplitudes, the parameter of heff/δ increases
first and then decreases after reaching the maximum value at
approximately α ¼151. For A ¼ 0.12c, the value is around 0.7. As to
A ¼0.025c, heff/δ has a value around 0.5, which means that
compared with large amplitude tubercles, small amplitude
tubercles works more like vortex generator with normal heff/δ
range of 0.1–0.5 (Lin, 2002).
4. Conclusions
In this paper, flow separation control for hydrofoils with
leading-edge tubercles was experimentally conducted under a
very low Reynolds number of Re ¼1.4  104. Two tubercle ampli-
tudes and two wavelengths are considered to investigate their
effects on the flow separation behaviour. On the other hand,
results are also presented to shed more light on the flow separa-
tion control behaviour associated with the application of tubercles
in hydrofoils. The main findings obtained in the present research
are summarized as follows:
Under very low Reynolds number, the flow can separate rather
easily from the baseline hydrofoil surface. The tubercles behave
like small delta-wings, where a CVP is formed over each tubercle.
The CVPs significantly modify the boundary layer flow through
chordwise and spanwise interactions. For the four hydrofoils with
tubercles, performance of overall flow separation control depends
on the parameter of A/λ. For larger A/λ, flow separation control is
better, while the tubercles almost lose their flow separation
control capabilities for the smallest A/λ. For the A9λ18.75 hydro-
foil, the CVPs are significantly unstable with strong interactions
with their adjacent neighbours along the Trough-plane. When
they dissipate, very unstable flow separation can be observed.
Furthermore, the flow separation region varies with time, which is
particularly noticeable at angles-of-attack of α ¼151 and 201. The
CVPs are observed to be responsible for flow separation mitiga-
tions and the tubercles are deduced to take on the role of vortex
generators.
Acknowledgments
The authors gratefully acknowledge the support of the present
study under Singapore MINDEF Defense Innovative Research
Programme (Project Agreement number: 9012101770), as well as
G. C. Koh for his assistance in the experiments and data analysis.
References
Bolzon, M.D., Kelso, R.M., Arjomandi, M., 2015. Tubercles and their applications.
Journal of Aerospace Engineering, 04015013.
Custodio, D., 2007. The Effect of Humpback Whale-like Leading-edge Protuberances
on Hydrofoil Performance. Worcester Polytechnic Inst, Worcester, MA, M.S.
Thesis.
Câmara, J.F.D., Sousa, J.M.M., 2013. Numerical study on the use of a sinusoidal
leading edge for passive stall control at low Reynolds number. In: 51st AIAA
Aerospace Sciences Meeting, AIAA Paper 2013-0062.
Custodio, D., Henoch, C.W., Johari, H., 2015. Aerodynamic characteristics of finite
span wings with leading-edge protuberances. AIAA J. 53 (7), 1878–1893.
Dropkin, A., Custodio, D., Henoch, C.W., Johari, H., 2012. Computation of flowfield
around an airfoil with leading-edge protuberances. J. Aircr. 49 (5), 1345–1355.
Delgado, H.E.C., Esmaeili, A., Sousa, J.M.M., 2014. Stereo PIV measurements of low-
aspect-ratio Low-Reynolds-number wings with sinusoidal leading edges for
improved computational modeling. In: Proceedings of the 52nd Aerospace
Sciences Meeting, AIAA, National Harbor, MD.
Erm, L.P., 2003. Measurement of flow-induced pressures on the surface of a model
in a flow visualization water tunnel. Exp. Fluids 35, 533–540.
Fish, F.E., Battle, J.M., 1995. Hydrodynamic design of the humpback whale flipper. J.
Morphol. 225, 51–60.
Fish, F.E., Lauder, G.V., 2006. Passive and active flow control by swimming fishes
and mammals. Annu. Rev. Fluid Mech. 38, 193–224.
Fish, F.E., Weber, P.W., Murray, M.M., Howle, L.E., 2011. Marine applications
biomimetic humpback whale flipper. Mar. Technol. Soc. J. 45 (4), 198–207.
Godard, G., Stanislas, M., 2006. Control of a decelerating boundary layer. Part 1:
Optimisation of passive vortex generators. Aerosp. Sci. Technol. 10, 181–191.
Goruney, T., Rockwell, D., 2008. Flow past a delta wing with a sinusoidal leading-
edge: near-surface topology and flow structure. Exp. Fluids 47 (2), 321–331.
Hansen, K.L., Kelso, R.M., Dally, B.B., 2011. Performance variations of leading-edge
tubercles for distinct aerofoil profiles. AIAA J. 49 (1), 185–194.
Johari, H., Henoch, C., Custodio, D., Levshin, A., 2007. Effects of leading-edge
protuberances on aerofoil performance. AIAA J. 45 (11), 2634–2642.
Keane, R.D., Adrian, R.J., 1992. Theory of Cross-correlation Analysis of PIV Images.
49. Applied Scientific Research, pp. 191–215.
Lohry, M., Clifton, D., Martinelli, L., 2012. Characterization and design of tubercle
leading-edge. In: Seventh International Conference on Computational Fluid
Dynamics, ICCFD. , Princeton, NJ.
Lin, J.C., 2002. Review of research on low-profile vortex generators to control
boundary-layer separation. Prog. Aerosp. Sci. 38, 389–420.
Moffat, R.J., 1988. Describing the uncertainties in experimental results. Exp. Therm.
Fluid Sci. 1 (1), 3–17.
Miklosovic, D.S., Murray, M.M., Howle, L.E., Fish, F.E., 2004. Leading-edge tubercles
delay stall on humpback whale (megaptera novaeangliae) flippers. Phys. Fluids
16 (5), 39–42.
Murray, M.M., Fish, F.E., Howle, L.E., Miklosovic, D.S., 2005. Effects of leading edge
tubercles on a representative whale flipper model at various sweep angles. In:
Proceedings of the Fourteenth International Symposium on Unmanned Unteth-
ered Submersible Technology (UUST), Autonomous Undersea Systems Inst., Lee,
NH, Aug.
Miklosovic, D.S., Murray, M.M., 2007. Experimental evaluation of sinusoidal
leading-edges. J. Aircr. 44 (4), 1404–1407.
New, T.H., Chan, Y.X., Koh, G.C., Chung, H.M., Shi, S.X., 2014a. Effects of corrugated
aerofoil surface features on flow separation control. AIAA J. 52 (1), 206–211.
New, T.H., Tay, W.L., 2006. Effects of cross-stream radial injections on a round jet. J.
Turbul. 7 (57), 1–20.
New, T.H., Tsai, H.M., 2007. Experimental investigations on indeterminate-origin V-
and A-notched jets, AIAA J. 4 (45), 828–839.
New, T.H., Tsovolos, D., 2009. Influence of nozzle sharpness on the flow fields of V-
notched nozzle jets. Phys. Fluids 21, 1–8.
New, T.H., Tsovolos, D., 2011. On the vortical structures and behaviour of inclined
elliptic jets. Eur. J. Mech. B. Fluids 30 (4), 437–450.
New, T.H., Tsovolos, D., 2012. On the characteristics of minor-plane inclined elliptic
jets. Exp. Therm. Fluid Sci. 38, 94–106.
New, T.H., Wei, Z.Y., Goh, G.C., Cui, Y.D., 2014b. On the role and effectiveness of
streamwise vortices in leading-edge modified wings at low Reynolds number.
In: Proceedings of the 16th International Symposium on Flow Visualization.
Okinawa, Japan.
Ozen, C.A., Rockwell, D., 2010. Control of vortical structures on a flapping wing via a
sinusoidal leading-edge. Phys. Fluids 22 (2), 021701.
Pedro, H.T.C., Kobayashi, M.H., 2008. Numerical Study of Stall Delay on Humpback
Whale Flippers. AIAA Paper No 2008-0584.
Rostamzadeh, N., Hansen, K.L., Kelso, R.M., Dally, B.B., 2014. The formation
mechanism and impact of streamwise vortices on NACA 0021 airfoil’ s
 Z. Wei et al. / Ocean Engineering 108 (2015) 336–349
performance with undulating leading edge modification. Phys. Fluids 26,
107101.
Shi, Shengxian, New, T.H., 2013. Some observations in the vortex-turning behaviour
of noncircular inclined jets. Exp. Fluids 54 (11), 1614.
Shi, Shengxian, New, T.H., Liu, Yingzheng, 2014. Flapping dynamics of a low aspect-
ratio energy-harvesting membrane immersed in a square cylinder wake. Exp.
Therm. Fluid Sci. 46, 151–161.
Simons, M., 2002. Model Aircraft Aerodynamics. Special Interest Model Books,
Dorset, England, UK p. 112.
Skillen, A., Revell, A., Pinelli, A., Piomelli, U., Favier, J., 2015. Flow over a wing with
leading-edge undulations. AIAA J. 53 (2), 464–472.
Soria, J., New, T.H., Lim, T.T., Parker, K., 2003. Multigrid CCDPIV measurements of
accelerated flow past an airfoil at angle of attack of 301. Exp. Therm. Fluid Sci.
27, 667–676.
Stanway, M.J., 2008. Hydrodynamic Effects of Leading-edge Tubercles on Control
Surfaces and in Flapping Foil Propulsion. Massachusetts Inst. of Technology,
Cambridge, MA, M.S. Thesis.
349
Van Nierop, E.A., Alben, S., Brenner, M.P., 2008. How bumps on whale flippers delay
stall: an aerodynamic model. Phys. Rev. Lett. 100 (5), 054502.
Weber, P.W., Howle, L.E., Murray, M.M., 2010. Lift, drag and cavitation onset on
rudders with leading-edge tubercles. Mar. Technol. 47 (1), 27–36.
Weber, P.W., Howle, L.E., Murray, M.M., Miklosovic, D.S., 2011. Computational
evaluation of the performance of lifting surfaces with leading-edge protuber-
ances. J. Aircr. 48 (2), 591–600.
Zhang, M.M., Wang, G.F., Xu, J.Z., 2014. Experimental study of flow separation
control on a low Re aerofoil using leading-edge protuberance method. Exp.
Fluids 55, 1710.
Zverkov, I., Zanin, B., 2008. Disturbances growth in boundary layers on classical and
wavy surface wings. AIAA J. 46 (12), 3149–3158.