The Effects of Leading-Edge Tubercles On Dynamic Stall
The Effects of Leading-Edge Tubercles On Dynamic Stall
The Effects of Leading-Edge Tubercles On Dynamic Stall
893 A5-1
Published by Cambridge University Press
doi:10.1017/jfm.2020.216
region, appeared to affect the convective behaviour of the DSV. The results suggest
that the leading-edge tubercles observed on Humpback whale flippers act as passive
flow-control mechanisms to control or delay dynamic stall.
Key words: vortex dynamics
1. Introduction
Dynamic stall is a fundamental unsteady flow event that occurs when the apparent
angle of attack (AoA) of a lifting surface is changed dynamically. Usually this is
due to a wing being either pitched or oscillated, though it can also be caused by
changes in the approach flow direction. This dynamic change in the AoA results in
the formation of a strong vortex, i.e. the dynamic stall vortex (DSV), near the leading
the suction surface of the airfoil. More specifically, that the tubercles of the humpback
whale are passive dynamic stall control mechanisms that enable the high degree of
manoeuvrability displayed by the humpback whale. If true, these structures could be
used to delay dynamic stall to higher angles. This study investigates the effect of
idealized leading-edge tubercles on an airfoil undergoing constant pitch rate motion
to study the dynamics of the formation and convection of the DSV.
2. Background
Dynamic stall effects on helicopter rotors, and to a lesser extent on turbomachinery,
were the common driving force behind early experiments on dynamic stall (Carr
et al. 1977; McAlister, Carr & McCroskey 1978; McCroskey 1982). During forward
helicopter flight the effective AoA of a rotor blade rapidly changes as it advances into
and away from the free-stream flow during rotation. Carr et al. (1977) focused on a
number of airfoil sections, but went into great detail on their results for the NACA
0012. That airfoil has since become one of the most commonly studied, but has one
of the most complicated dynamic stall behaviour patterns of any of the airfoils used
Effects of leading-edge tubercles on dynamic stall 893 A5-3
but again no difference was found in the stall behaviour (Carr et al. 1977). Factors
identified by Carr et al. (1977) as affecting dynamic stall behaviour, in order of
maximum effect were: airfoil shape, pitching frequency/rate, pitching amplitude and
Reynolds number. McCrosky’s (1982) review of unsteady flow behaviours extended
the list of quantities affecting dynamic stall to include Mach number. The effect of
the separation bubble on vortex formation was studied for the NACA 0012 airfoil by
McAlister et al. (1978). A boundary layer trip was used to transition the separation
bubble at a range of oscillating amplitudes and frequencies, but this had little to
no effect on dynamic stall. McAlister et al. (1978) theorized that the DSV strength
was directly related to the total circulation around the airfoil at the time of vortex
formation. Also of note here is that the DSV formation was delayed with increased
pitching frequency (McAlister et al. 1978).
Shih, Lourenco & Krothapalli (1995), Oshima & Ramaprian (1997) and Pruski
& Bowersox (2013) all studied the dynamic stall phenomenon using particle image
velocimetry to obtain quantitative measurements of the flow field. Results from these
particle image velocimetry experiments showed that the DSV was not formed as a
result of a Kelvin–Helmholtz instability (Shih et al. 1995). Secondary vorticity near
893 A5-4 J. T. Hrynuk and D. G. Bohl
mid-chord formed and may (Shih et al. 1995; Oshima & Ramaprian 1997) or may
not (Pruski & Bowersox 2013) affect the DSV. Oshima & Ramaprian (1997) defined
this vortex as shear-layer vortex (SLV) and it was shown that it could reach up to
35 % of circulation found in the DSV (Shih et al. 1995).
Gendrich (1999) studied the formation of the DSV with highly resolved experimental
data using molecular tagging velocimetry (MTV). The MTV technique can be
effectively viewed as a molecular counterpart to particle image velocimetry. This
experimental technique was also used in the current study and is explained in detail
in later sections. Gendrich (1999) studied a NACA 0012 airfoil pitched at constant
dimensionless pitch rates of Ω ∗ = 0.1–0.4, where Ω ∗ = α̇c/2U∞ , at a chord Reynolds
number of 12 000 in a water tunnel. The spatial resolution in this experiment was
0.003c which allowed for detailed analysis of the leading-edge dynamics to be
conducted. Low-Reynolds-number experiments were selected for ease of comparison
to the numerical model of Choudhuri, Knight & Visbal (1994). Focus was placed
on the leading-edge region to generate high-resolution data of the DSV formation
process (i.e. up to the angle at which the DSV began to convect away from the
airfoil surface). Extensive work was also done to show that the flow was highly
repeatable and thus a candidate for phase averaging to airfoil pitch angle. Gendrich
(1999) detailed specific formation angles and a highly resolved vortex structure, but
did not measure the flow downstream of the DSV formation region.
Gendrich (1999) found that at low incidence angles the boundary layer remained
attached to the airfoil surface but with increasing boundary layer thickness as a
function of the AoA. A thin region of vorticity, with sign opposite to that of the
boundary layer, formed underneath the primary vorticity along the airfoil surface
in the recirculating region. Gendrich (1999) defined this flow condition as detached
but not separated, where detached flow was defined as when ‘a thin region of
opposite-sign vorticity is present under the primary shear layer’. Separated flow was
defined as when the flow no longer followed the contours of the airfoil shape. When
the flow detached, a ‘bulge’ formed near the leading edge which was associated
with a growth in the reversed flow region. This reversed flow region induced the
boundary layer flow back towards the airfoil surface cutting off the reversed flow
region downstream. The bulge region near the leading edge is commonly referred
to as the separation bubble, which sometimes is said to ‘burst’ forming the DSV
(Carr et al. 1977). Around the time that the separation bubble formed, the reversed
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
flow shear layer began to expand up into the flow away from the airfoil surface.
The pressure gradients inside this bubble generated another thin layer of vorticity
forming a three-layer vortical structure that had previously only been observed in the
computational model of Choudhuri et al. (1994). The expansion of this third region
of vorticity away from the airfoil pushes up the overlaying reversed flow region which
in turn develops the full DSV along with secondary and tertiary vorticity regions. The
results of Gendrich (1999) indicated that the source of the vorticity within the DSV
was derived from the original airfoil boundary layer. Further this source continuously
fed with vorticity from the leading edge during formation. A secondary vortex
structure, originating from the first reversed flow region, was formed by vorticity
from the reversed shear layer. A third vortex formed from the leading-edge vorticity
interacting with the secondary vortex. The tertiary region of vorticity, which was near
the surface underneath the secondary vortex, did not form a vortex-like structure but
was the driving factor that pushed the reversed shear flow away from the surface
forming the secondary vortex. Gendrich (1999) suggested that the formation of this
tertiary region of reversed flow was the initiator of the secondary vortex which in
turn pinched off the large DSV.
Effects of leading-edge tubercles on dynamic stall 893 A5-5
Numerical simulations and the experimental data of Gendrich (1999) were used
to investigate the impact of pitch rate on DSV formation. Increased pitch rate
resulted in a more compact leading-edge separation process that occurred closer
to the leading edge at high pitch rates (Ω ∗ > 0.2; Gendrich 1999). These results
showed the formation of the thin reversed flow region near the leading edge can
occur in two different ways: via the progression of the reversed flow from the
trailing edge, or independently near the leading edge. The independent formation
method was only observed for high pitch rates of Ω ∗ > 0.3. The transition between
these two stall mechanisms was evaluated computationally and observed to switch
at around Ω ∗ = 0.2. Vorticity was significantly higher for higher pitch rate cases
and the vortex structures that formed were more compact and nearer the leading
edge. Phase-averaged experimental results at high pitch rates proved to be an issue
for observing the formation mechanisms due to decreased repeatability. Increases
in small-scale structures and three-dimensional effects appeared to play a role in
repeatability at higher pitch rates (Gendrich 1999).
The frame of reference of the data presented may also have a significant effect
on the results of the data (Gendrich 1999). A common first method of rotating
the airfoil into a rotating frame of reference is the transformation of rotating the
coordinate system and velocity vectors to align with the airfoil. With this simple
coordinate transformation, the vector magnitudes and angles remain the same with
respect to the airfoil surface, but the airfoil surface retains a velocity from the
pitching motion. This unwanted velocity of the airfoil surface can be eliminated with
the addition of a solid-body rotation to the coordinate transformation developed by
Gendrich (1999):
velocity (U∞ ), dimensionless pitch rate and spatial locations of the velocity vectors.
Note that here x and y are dimensionless (x/c, y/c) and are left in the form shown
in Gendrich (1999).
More recently, work by Visbal & Garmann (2017) computationally evaluated the
dynamic stall of a NACA 0012 wing at low Reynolds number and at a lower pitch
rate than used by Gendrich (1999). They observed a laminar separation bubble and
determined that the breakdown of that flow feature preceded the onset of DSV
formation. Visbal & Garmann (2017) also observed the SLV features noted by Shih
et al. (1995), Oshima & Ramaprian (1997) and Pruski & Bowersox (2013).
One dynamic stall control study of particular relevance to the current work is one
that used small vortex generators (Mai et al. 2008). Small circular vortex generators
(6 mm in diameter, 0.54–1.1 mm in height) were applied to the leading-edge area
of an airfoil. The optimal location was found to be around 0.17c on the pressure
side of the airfoil, not on the suction side where one might naturally assume. The
vortex generators were placed at 0.05c spacing along the span and tested for both light
and deep dynamic stall. Mai et al. (2008) observed that lift hysteresis was reduced
significantly during the reattachment phase of the motion. Reductions in lift hysteresis
893 A5-6 J. T. Hrynuk and D. G. Bohl
of up to 39 % were observed when the vortex generators were applied. Both drag and
moment coefficients saw improvements of approximately 25 % during the downswing
phase of the motion. The results were relatively independent of Mach number and
pitching condition. These devices appeared to improve reattachment by energizing the
boundary layer of the airfoil in a way similar to small tubercles on the humpback
whale flipper. These results on the control of dynamic stall mimic some of the findings
for humpback whale tubercles, where small tubercles were suggested to act as vortex
generators (Borg 2012).
The analysis of a humpback whale flipper as an aerodynamic/hydrodynamic shape
was first performed by Fish & Battle (1995), where they discussed the tubercle
structure on the leading edge of the flipper and its aerodynamic/hydrodynamic impact.
The authors analysed a flipper recovered from a beached whale in coastal New Jersey.
The flipper was roughly elliptic in shape with a slight backward sweep, had a span
of 2.5 m and a wing area of approximately 1.02 m2 . The flipper also had a mean
aerodynamic chord of 0.82 m, a thickness-to-chord ratio of 0.2–0.26 and an aspect
ratio of 6.1, making it strikingly similar in size and shape to modern aircraft wings.
The flipper dissected by Fish & Battle (1995) had a total of 11 tubercles on the
leading edge, with tubercle height decreasing along the span. Cross-sectional slices
of the flipper were compared with known aerodynamic airfoils and Fish & Battle
(1995) found that the flipper closely resembled a NACA 634 -021 airfoil at mid-span.
The function of the tubercles was assumed to be a passive stall control mechanism
similar to vortex generators (strakes) on full-scale aircraft.
Humpback whales are known to hunt using a bubble-net technique in which they
release bubbles while circling prey from below thereby corralling their prey. When
the bubble net becomes small enough the whale turns into the middle of the net and
surfaces with its mouth open capturing the prey. Bubble-net diameters vary based
on specific prey animals from 1.5 up to 50 m. The minimum turning radius of a
‘typical’ humpback whale was estimated to be 7.4 m at a 90◦ bank angle using basic
aerodynamic lift and drag assumptions. The authors explained the difference between
their calculated minimum radius and field-observed minimum (∼1.5 m) in two
separate ways: first that the humpback whale must be using its flippers as more than
just static wings and overall manoeuvrability would be higher for an actual whale,
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
case which can be evaluated aerodynamically. They recorded a rapid downward flipper
stroke that rotated the flippers 90◦ in approximately 0.8 s. Given the conditions noted
here and a few reasonable assumptions, this motion is easily shown to be in the
range of the unsteady aerodynamic conditions encompassing dynamic stall.
The actuation motion observed by Segre et al. (2017), combined with the mean
chord length of the humpback whale from Fish & Battle (1995) of 0.82 m, and an
assumption of 3 m s−1 forward velocity of the whale, leads to a non-dimensional pitch
rate of Ω ∗ = 0.21. This result places the motion significantly above the threshold of
Ω ∗ > 0.05 with a peak AoA greater than the static stall angle, which defines the
presence of a dynamic stall event. Segre et al. (2017) assumed a static lift maximum
of 120 kN for this particular humpback.
One can easily perform a ‘back-of-an-envelope’ calculation to show the potential
of dynamic, rather than static, lift to generate the large forces observed in humpback
whales. The results of the dynamic stall study by Gendrich (1999) showed that the
unsteady force on an infinite span airfoil was nominally three times the static lift
maximum. If instead of using the Cl,max of the static case one conservatively assumes a
more realistic 2Cl,max for the dynamic case, the whale could be generating an unsteady
893 A5-8 J. T. Hrynuk and D. G. Bohl
peak lunging force roughly equivalent to its own body weight. With this back-of-an-
envelope result in mind, the evaluation of the effect of tubercles on dynamic stall
becomes even more interesting.
The current work studies the effects of tubercles on a dynamically pitching airfoil
at pitch rates in the deep dynamic stall range. Suction-side planar flow field data
were collected to compare a tubercled airfoil to a baseline straight airfoil. Information
derived from that dataset includes vortex formation angle, vortex convective path data
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
and circulation. Vortex convective speed and the dynamic force generated by the
vortex were inferred from these datasets. In particular, the circulation and convection
characteristics of the DSV were used to infer the effects of the tubercles on the
dynamic stall process.
3. Experimental methods
The airfoils created for the current study were designed with comparison to prior
studies in mind. Tubercles applied to the airfoil had an amplitude of 0.04c similar
to the tubercles used in Miklosovic et al. (2004) and the 0.05c ‘medium’ tubercles
from Johari et al. (2007). The span of the model was 36.8 cm (14.5 in) based on the
width of a tunnel originally to be used for the study. This resulted in an aspect ratio
of 3.1 for the current study. False walls were placed on both ends of the airfoil to
limit end effects. An odd number of tubercles (13) was selected so that the middle
tubercle would align to the centreline of the water tunnel (figure 2). The wavelength
of the tubercles was approximately 0.24c which closely matches that of (Johari et al.
2007) while maintaining spanwise symmetry for mounted model.
Effects of leading-edge tubercles on dynamic stall 893 A5-9
Trough (4)
Near trough (3)
Near peak (2)
Peak (1)
Gendrich (1999) also drove other key experimental and model parameters. The mean
chord length (c = 12 cm), free-stream velocity (U∞ = 10 cm s−1 ), Reynolds number
(Rec = 12 000) and non-dimensional pitch rates (Ω ∗ = 0.1, 0.2, 0.4) were selected to
match those in Gendrich (1999).
The models were three-dimensionally printed with a precision of 0.1 mm in
three pieces using a Viper Si2r SLA system with DSM 11120 epoxy resin. The
models were assembled, sanded and painted a matte black finish to minimize surface
reflections. The planar flow measurements were recorded for the modified airfoil at
the planes shown in figure 3. The baseline case was measured at the midspan location.
The peak and trough planes will be primarily discussed in this paper, along with the
baseline NACA 0012 measurement plane.
The current experiments utilize a constant pitch-up and hold motion. The wing was
pitched up from 0◦ to 55◦ at a constant pitch rate and held at the maximum angle
for several convective times to allow the DSV to convect downstream. The dynamic
pitching motion was controlled by a Galil DMC 40-20 servocontroller with a 4000
count per revolution servo motor. Encoder motion data from the experiments were
analysed for sources of error in the pitch angle of the airfoil. The data acquisition
893 A5-10 J. T. Hrynuk and D. G. Bohl
(a) 60 Control
(b) 16
Encoder
50 14
12
40 10
å (deg.)
30 8
6
20 4
10 2
0
0
-1 0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
t (s) t (s)
F IGURE 4. Control–response curves for (a) full pitching motion and (b) accelerating
region.
Pitch rate Ideal pitch Measured pitch Control error Mean range STD Error
rate (deg. s−1 ) rate (deg. s−1 ) (deg.) (deg.) (deg.) (deg.)
0.1 9.8 9.67 0.088 0.09 0.027 ±0.10
0.2 19.6 19.0 0.281 0.08 0.029 ±0.10
0.4 39.2 39.04 0.372 0.18 0.057 ±0.13
TABLE 1. Motion control errors.
frequency of the system was limited to 5 Hz due to laser and camera limitations.
The previous work of Gendrich (1999) has shown that the results are repeatable from
run to run and therefore the temporal resolution was increased by phase-averaging the
results from multiple experimental runs in the following way. The start of the pitching
motion of the airfoil was controlled with respect to the laser and camera operation. For
the Ω ∗ = 0.1 experiments, a series of 10 different start time delays, with respect to
the start of motion, were used. The data from the different runs were phase-averaged
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
with respect to the airfoil angle to provide an effective data rate of 50 Hz. For the
higher pitch rates (Ω ∗ = 0.2, 0.4) five start time offsets were used and phase-averaged
for an effective data rate of 25 Hz. The resulting angular resolution of the data was
approximately 0.19◦ for Ω ∗ = 0.1, 0.73◦ for Ω ∗ = 0.2 and 1.47◦ for Ω ∗ = 0.4. The
lower angular resolution for the higher pitch rates was a combination of the reduced
data rate and the significant increases in angular velocity of the airfoils. The timing
presented here is normalized time (t∗ = tU∞ /c) with the start of the pitching motion
defining t∗ = 0.
The average control error, measured by comparing encoder position and the
command position, varied as a function of pitch rate. Larger control errors were
observed during the accelerative phase, as expected due to the effects of starting
motion. The average control error for the entire pitching motion was small (see
table 1). Figure 4 shows the control–response curves for the different pitch rates
tested (Ω = 0.1, 0.2, 0.4) with figure 4(b) showing the accelerative region. Oscillations
in the motion were observed that worsened with increased pitch rate, but typically
dampened out before α = 10◦ . Past work (e.g. Gendrich 1999) showed the flow
dynamics was insensitive to these small deviations in the controller motion.
Effects of leading-edge tubercles on dynamic stall 893 A5-11
F IGURE 5. (a) Undelayed MTV image t = t0 . (b) Delayed MTV image t = t0 + 3.5 ms.
(c) Resulting displacement vector field.
Two images were taken in rapid sequence where the first image was taken immediately
after laser firing (figure 5a) and the second image at some specified time later (i.e. the
delay time, 1t) (figure 5b). Images were captured using a PCO Dicam Pro CCD
camera with a 1280 × 1024 pixel resolution. The camera was oriented so that the
longer axis of the image (1280 pixels) was aligned with the flow direction. An 80 mm
Nikon lens was used to capture all the MTV images. The delay time for all MTV
image pairs in this work was 1t = 3.5 ms to limit the maximum displacement between
images to less than 10 pixels as recommended by Gendrich & Koochesfahani (1996).
Laser intersection angles were nominally 90◦ to reduce error in the measurements also
as per Gendrich & Koochesfahani (1996).
Image pairs were converted into displacement fields using the direct correlation
method described in Gendrich & Koochesfahani (1996) in the following way.
Each grid intersection was identified in the undelayed images and a small region
(21 × 21 pixels) was defined around each intersection in the undelayed images. These
subregions were then moved individually around the delayed image to determine the
correlation field. The location of the maximum correlation for each subregion was
taken as the displacement for that particular intersection. The velocity was found
893 A5-12 J. T. Hrynuk and D. G. Bohl
by dividing the displacement for a particular intersection by 1t. This provides a
Lagrangian measurement of the velocity which is then taken to exist at the midpoint
location between the undelayed and delayed intersections (figure 5c).
Two fields of view were used to increase the spatial extent of the measurements
in this work. The two fields of view were overlapped to allow for alignment of the
full velocity field. Locations of the laser intersections were shifted within each field
of view for each of the 25 repeated trials to increase the effective spatial resolution of
the data. The final spatial resolution was approximately 1 mm (0.008c) with a total
field of view of approximately 17 cm × 14 cm.
The MTV data are typically irregularly spaced. The irregularly spaced data results
were mapped onto a regularly spaced grid using the polynomial fitting technique
described by Cohn & Koochesfahani (2000). Vorticity was calculated based on this
regular grid mapping of the data, using a second-order finite difference calculation.
Circulation was calculated by integrating the vorticity field over a specified spatial
area. The spatial area for the integration was allowed to vary with time based on the
instantaneous measurement of the vortex core radius. This was done to accommodate
the increasing size of the DSV as circulation was added from the leading edge.
The error in the velocity measurements is directly related to the error in the
correlation process (i.e. the error in the determination of the displacement). A series
of 100 undelayed images, with over 55 000 vectors, were correlated against an
averaged image of all the undelayed images for that case. This, in effect, determined
the error level in the displacement based on all system uncertainties including camera
noise, tagging line orientation, tagging line width, etc. The vectors at each grid
intersection were averaged to determine the average error in displacement at each of
approximately 550 grid intersections. The 95 % error level was calculated using this
method to be 0.05 pixels. This pixel error, when converted to an error in velocity,
based on the image magnification and delay time, was found to be 0.24 cm s−1 . It
is noted that the error in the velocity measurement assumes that the delay time was
known exactly. The jitter in the timing of the system was several orders of magnitude
lower than the delay times used and therefore not significant. Error in the vorticity
calculations was estimated to be 2.14 s−1 based on the analysis described in Bohl &
Koochesfahani (2009).
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
4. Results
4.1. Baseline NACA 0012
Results for the baseline NACA 0012 airfoil were consistent with observations in
previous studies by Carr et al. (1977), Gendrich (1999) and others. The case shown
in figure 6 is that of Ω ∗ = 0.1, which was the lowest reduced frequency and the
best temporally resolved case investigated. For ease of viewing, the vector maps
in figures 6–8 show only every eighth vector of the datasets. The origin of the
data presented is set such that (x, y) = (0, 0) is located at the pitching axis and
corresponded to the 14 chord location of the airfoil.
The flow remained attached near the leading edge at α = 15◦ , which is higher
than the static stall angle. The velocity field showed separation and reverse flow was
present near the trailing edge. This region grew in thickness and advanced towards the
leading edge as the AoA, α, increased. Trailing edge vortices also formed through the
pitching motion. By 20◦ , a distinct region of vorticity along the airfoil surface was
observed near the leading edge as the DSV began to form. When the airfoil reached
25◦ the DSV had fully formed into a coherent vortex, though the DSV appeared
Effects of leading-edge tubercles on dynamic stall 893 A5-13
0.6
0.4
0.2
y/c
0
ø
-0.2 30.0
15° 20° 25° 22.5
-0.4
15.0
7.5
0.6 -7.5
0.4 -15.0
-22.5
0.2 -30.0
y/c
0
-0.2
°
-0.4 30 35° 40°
-0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.4 -0.2 0 0.2 0.4 0.6 0.8
x/c x/c x/c
0.6
0.4
0.2
y/c
0
ø
-0.2 30.0
-0.4 15° 20° 25° 22.5
15.0
7.5
0.6 -7.5
-15.0
0.4
-22.5
0.2 -30.0
y/c
0
-0.2
-0.4 30° 35° 40°
-0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.4 -0.2 0 0.2 0.4 0.6 0.8
x/c x/c x/c
F IGURE 7. Tubercled airfoil peak plane dynamic stall flow field.
0.6
0.4
0.2
y/c
0
ø
-0.2 30.0
22.5
-0.4 15° 20° 25°
15.0
7.5
0.6 -7.5
-15.0
0.4 -22.5
0.2 -30.0
y/c
0
-0.2
-0.4 30° 35° 40°
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
-0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.4 -0.2 0 0.2 0.4 0.6 0.8
x/c x/c x/c
F IGURE 8. Tubercled airfoil trough plane dynamic stall flow field.
behaviour extended to high angles where the feeding vorticity for the DSV left the
leading edge more in line with the free-stream flow direction when compared to the
baseline.
Comparing the baseline (figure 6) and peak of the tubercled airfoil (figure 7),
other significant differences existed between the two, the most notable of which
were formation angle, vortex strength and vortex shape. At α = 25◦ , the DSV was
clearly observed on the baseline airfoil while only initial indications of a DSV and
secondary recirculating region below it were seen on the peak plane. The strong
reverse flow region that formed beneath the DSV did not separate or ‘erupt’ from
the surface (figures 6 and 7 at 35◦ ) for the peak plane. Further this reverse flow
region did not develop a secondary vortex for the peak plane. Feeding vorticity from
the leading edge was noticeably discontinuous from the leading edge to the vortex
Effects of leading-edge tubercles on dynamic stall 893 A5-15
region (figure 7; α = 30◦ ). The data recorded in this study were planar and thus
out-of-plane motion, which is known to be present for the static case with tubercles
(Watts & Fish 2001), may have been the source of apparent discontinuities observed
in vorticity. Compared to the baseline, the DSV formed later in the motion, further
downstream along the airfoil chord, had a more elliptic shape and appeared to be
more diffuse.
Flow measurements over the trough of the tubercle showed stark differences from
the baseline and the peak plane of the tubercled airfoil. Figure 8 shows the trough
plane flow field with the baseline airfoil shape again outlined in grey for reference
and the tubercled airfoil trough in black. Key differences observed for this data plane
were: an earlier formation of the DSV and a lack of secondary vortex formation from
the reverse flow under the DSV, the DSV vortex formation location and its convective
path. First, the leading-edge vorticity at this plane began to pinch off and roll up into
a vortex at much lower angle than the baseline. The DSV, not observed until α ≈ 20◦
for the baseline, was apparent by α = 15◦ in the trough of the tubercled airfoil. In
addition to early formation angle, the DSV formed close to the leading edge and
rolled up into a notably strong vortex in this region. However, that compact vortex
structure was short-lived as by α = 30◦ the DSV had convected to mid-chord and
diffused significantly (figure 8). At α = 30◦ , the centre of rotation of the DSV was
in approximately the same location as the DSV on the peak plane (figures 7 and 8).
Compared to the baseline, the DSV formed on the trough plane was more diffuse
and feeding vorticity from the leading edge was at a steeper angle with respect to the
airfoil surface. In this case, vorticity shed from the leading edge during the pitching
process was almost perpendicular to the airfoil. On the baseline airfoil this feeding
vorticity was more aligned with the free-stream flow and was almost in line with the
free stream on the peak plane.
were used to determine the vortex formation angle in this work. First, the application
of streamlines to the flow field was used to identify the first frame that showed
the spiraling flow behaviour typically associated with a vortex. Second, the presence
of secondary vorticity is known to be the driver for pushing the DSV up from the
airfoil surface (Gendrich 1999). The angles at which secondary vorticity and spiraling
streamlines were first observed are shown in table 2. One major consideration to
evaluating the results of each of these methods is their heavy dependence on the
temporal and spatial resolution of the data. In essence, the values presented in table 2
are reflective of the angles at which a DSV can first be observed at the data resolution
from these experiments. The actual formation angle will occur between this angle
and the prior measurement angle. These results will be likely to lag experiments with
higher temporal resolution. Spatial resolution will also affect the observation time of
the DSV formation as finer spatial resolution allows for a more precise picture of
the velocity and vorticity fields to be observed. The formation angle of the baseline
results was found to lag those of the experiments done by Gendrich (1999), but more
closely match the computational results in the same study. The goal of that study,
which had both higher spatial and temporal resolution, was to investigate the early
893 A5-16 J. T. Hrynuk and D. G. Bohl
formation of the DSV. The focus of the current study was more global and so spatial
resolution near the airfoil surface was sacrificed, somewhat limiting the observation
of vortex structures until they were sufficiently large. However, the current results
provide trends in the formation angle between cases.
The results show that the DSV formed earliest in the trough of the tubercled airfoil
and then progressively increased towards the peak plane. The baseline results showed
a formation angle nominally the same as for the near-trough plane and significantly
earlier than for the near-peak and peak planes. These results are consistent with what
was observed in the full flow field results (figures 6–8). The methods for determining
formation angle agreed well, with the only discrepancy occurring for the baseline.
One factor likely causing these methods to converge for the tubercled airfoil is the
eruption-like behaviour of DSV formation on the modified wing versus the roll-up
behaviour of the baseline DSV. While the DSV on the trough plane formed earlier
than the baseline, between 0.7◦ and 2.5◦ earlier depending on method, the DSV on
the peak plane formed significantly later in the pitch motion, 5.4◦ –7.1◦ later than the
baseline. Using the data in table 2, the mean formation angle of the DSV on the
tubercled airfoil was α = 21.7◦ using the streamlines method and α = 21.4◦ using
secondary vorticity. A linear average of the formation angle is unlikely to reflect the
intricacies of the spanwise flow on the tubercled airfoil. Instead it suggests that taken
as a whole, the DSV formation is delayed on the tubercled airfoil with the peak and
trough planes bounding the formation behaviour.
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
Equation (4.1) calculates the average angle between the position vector from the
calculation point and velocity vector of all measurement points in a user-defined
region. The value of Γ1 is limited by ±1 (θM = 90◦ ) with the sign of Γ1 describing
the direction of rotation and a magnitude of 1 indicating a true circular motion of
the velocity field around the analysis location. Here Γ1 was chosen for tracking the
vortices because the centre of rotation was found to be well defined even for the
diffuse DSV. The spatial tracks of the DSV for the baseline and tubercled airfoils
Effects of leading-edge tubercles on dynamic stall 893 A5-17
are shown in figure 9, with near-peak and near-trough plane data excluded. Velocity
field data from these intermediary planes are not shown in this paper because of
their similarity to the peak and trough plane data. All data in this section are shown
in the airfoil frame of reference as the location of the DSV relative to the airfoil
is important when determining relational effects. Recall that the origin location was
taken at the pitching axis, the 14 chord location.
Figure 9 shows that the baseline DSV formed forward of the 14 chord pitching
axis and moved towards the trailing edge before convecting away from the airfoil
surface. The baseline DSV moved towards the trailing edge up to xAF /c ≈ 0.4, roughly
65 % from the leading edge, before convecting away from the airfoil. At later times
the baseline DSV convected upstream towards the leading edge reaching xAF /c ≈ 0.3
before again moving towards the trailing edge and drifting out of the field of view.
In the trough plane of the tubercled airfoil, the DSV also formed near the leading
edge, just slightly ahead of the baseline DSV formation location. However, the trough
DSV lingered near the leading edge before moving quickly towards the trailing edge
to align with and convect along the same path as the peak plane DSV, which formed
behind the pitching axis. On the tubercled airfoil the DSV moved away from the
airfoil surface around xAF /c ≈ 0.3 and did not exhibit the upstream motion exhibited
by the baseline DSV.
Tracking the DSV as a function of time, t∗ = tU∞ /c, allows for comparison of the
location of the DSV across tubercled airfoil planes, answering the question of where
the DSVs are at the same time in the pitching motion. Normalized time was used
here instead of pitch angle as it allows tracking of the DSV after the pitching motion
stopped. Figure 10 shows the xAF /c and yAF /c location of the DSVs for the different
planes. Baseline and trough plane DSVs were formed at similar times while the peak
plane DSV formed at a much later time. The data showed that there was spatial
convergence of the DSV location of the peak and trough planes at around t∗ = 2.75,
after which the DSV on the tubercled airfoil was nominally aligned across the span.
This was notable given the differences in the flow conditions at the different locations
on the tubercle geometry. The baseline DSV exhibited the greatest spatial variation
throughout its convection where the baseline moved up and downstream in the flow
more than the tubercled airfoil DSV.
The vortex tracking data also allowed for the calculation of DSV convective velocity.
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
Convective velocity was calculated by fitting the vortex location in time for both xAF
and yAF directions and then differentiating the fit. Twenty neighbouring data points
(10 before and 10 after the current location) were used in the fit for each measurement
time, generating the data in figure 11. All cases showed fluctuation in the convective
velocities indicating that the formation and convection process was dynamic. Consider
first the baseline case. After formation, the DSV moved away from the airfoil. Local
peaks in Vconv,AF indicate that this process was dynamic as the vortex accelerated and
decelerated during this process. The maximum acceleration began to occur at t∗ = 3.0
after which the convective speed reached a maximum of Vconv,AF ≈ 0.6U∞ . The DSV
continued to move away from the surface, though the rate decreased. The convective
speed in the x direction, Uconv,AF , highlighted the forward/backward motion of the DSV
relative to the pitching location. The convective speeds in this direction were smaller
and oscillated about zero. The convective speed away from the airfoil (Vconv,AF ) was
always positive but significantly less than the free-stream speed, of the order of 30 %
of U∞ . Speed in the chordwise direction (Uconv,AF ) fluctuated about zero.
A peak in the convective velocity of the baseline DSV occurred around t∗ = 3.75
and decelerated after that time. The convective velocity exceeded 60 % of that of the
893 A5-18 J. T. Hrynuk and D. G. Bohl
1.0
Baseline
Peak
0.8 Trough
0.6
yAF/c
0.4
0.2
0
-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
xAF/c
0.8
Baseline
Peak
0.6 Trough
0.4
xAF/c
0.2
0
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
0.8
0.6
yAF/c
0.4
0.2
0
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
*
t = tU∞/c
Ucon√AF/U∞
Vcon√AF/U∞
0.4 0.4
0.2 0.2
0 0
-0.2 -0.2
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
t * = tU∞/c t * = tU∞/c
free stream for the baseline DSV at this point. At that time on the tubercled airfoil
the opposite effect was noticed, with a local minimum in convective speed. While the
speed of the DSVs on the tubercled airfoil tended to oscillate more, their convective
speed remained lower than that for the baseline. These results also show that after the
alignment motion of the trough DSV to match the location with the peak DSV, the
trough DSV led the motion with a slight phase shift between the peak and trough.
Despite the unique motion of the tubercled airfoil DSV, the mean convective velocity
of the tubercled airfoil DSV was lower than that of the baseline.
4.5. Circulation
Results so far have shown that, compared to a baseline airfoil, a tubercled airfoil
generated DSVs in different locations, with different shapes and different convective
behaviour. Changes in the dynamic lift and drag on the airfoil models were of interest
to the current study, but lift and drag were not directly measured. Prior studies have
shown that the dynamic lift generated in the dynamic stall process is caused by the
strength of the low-pressure DSV and its proximity to the suction side of the airfoil
(Carr et al. 1977; McCroskey 1982). Fundamental differences in the flow were the
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
primary focus of the current study, but inferences on the dynamic lift can be made
from quantifying the circulation of the DSV. By comparing the circulation observed in
Gao, Wei & Hrynuk (2018) with the dynamic lift measured in Gendrich (1999) there
is a strong correlation between dynamic lift and circulation of the DSV. This shows
that the strength of the vortex, as quantified through the circulation, correlates with
the dynamic lift generated on the airfoil.
Circulation is usually calculated from data either by summing the vorticity over the
area of the vortex or by performing a line integral around the edge of the vortex. In
both cases, calculating the circulation of a vortex is difficult to do as one must identify
the outer edge of the vortex, which is hard to define. In this work the circulation
was calculated in the same manner for all cases so that comparisons could be made.
Circulation presented here is the circulation of only the DSV and was calculated as
the sum of DSV vorticity within a varying radius in the following way. At each time,
the radius of the vortex was calculated using the radius of gyration of the distributed
vorticity in the manner described in Bohl & Koochesfahani (2009). The circulation
was then obtained by summing the vorticity over an area using 1.75rcore . This factor
was used as it was found to define a circular area that was just slightly larger than the
893 A5-20 J. T. Hrynuk and D. G. Bohl
5
d˝ */d t * = 2.14
Baseline
Peak
d˝ */dt * = 1.88
4 Trough
3
d˝ */dt * = 0.96
˝/cU
0
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
t * = tU∞/c
observed vorticity field of the DSV. The vorticity of the secondary vortex, which was
of opposite sign compared to the DSV, was not included in the calculation to isolate
the properties of the DSV.
Comparing the baseline and tubercled airfoil DSV circulations shows a significantly
higher circulation for the tubercled airfoil at all planes compared to the baseline
(figure 12). While the baseline and trough DSVs formed at similar times, the trough
DSV exhibited higher circulation levels than the baseline, in part due to a plateau
in circulation for the baseline around t∗ = 2. This initial plateau in circulation was
not observed on the trough although there was a decreased rate of growth around
t∗ = 2 for that plane. Peak plane DSV circulation showed a similar trend despite
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
a later formation time. A plateau in circulation occurred for the tubercled airfoil
data around t∗ = 2.75, but was more pronounced for trough data. Circulation for the
peak plane of the tubercled airfoil grew almost linearly up to a peak normalized
circulation near 3.8 with other tubercled airfoil data having maximum circulation
around 3.25. The second plateau in circulation occurred later for the baseline and
maximum circulation was around 2.5. While the circulation appears to be increasing
at the last data point, the data shown include only data where the entire DSV was
inside the measurement field of view. All three cases shown exhibit regions of linear
growth in the circulation with time. The slopes of these regions were computed using
a linear fit and are also shown in figure 12. The growth rate in the circulation was
similar for all measurement planes on the modified airfoil and nominally double the
baseline growth rate. Further, while there were minor differences, the growth rate of
the circulation was nominally doubled for the modified airfoil versus the baseline
airfoil in these regions.
The aerodynamic forces on the airfoils were not measured directly in the current
work; however, the impact of the tubercles on the airfoil can be inferred. Gao
et al. (2018) showed that applying the Kutta–Joukowski lift theorem resulted in lift
Effects of leading-edge tubercles on dynamic stall 893 A5-21
(a) (b)
Upwash
Downwash
Downwash
Static flow behaviour Approximate DSV
(å < åstall) core location
F IGURE 13. (a) Schematic view of flow generated by tubercles for the static case and
(b) schematic of the DSV core axis line.
coefficients that closely resemble prior lift measurements of dynamic stall. The authors
calculated the total circulation through the summation of both positive (trailing edge)
and negative (DSV) vorticity, but the excess of negative vorticity in the DSV drove
the total calculated lift. While this method likely does not accurately calculate lift
after the DSV convects away, the correlation between their calculation and prior lift
measurements suggests that increasing negative circulation while limiting the positive
circulation of the trailing-edge vortex would generate higher dynamic lift.
In the current study, consider first the flow conditions late in the pitch, t∗ ≈ 4.
The DSV for the modified airfoil was located closer to the airfoil surface and had a
measured circulation 33 % higher than that of the baseline DSV. These two factors
indicate that the tubercled airfoil likely generated more dynamic lift than the baseline
airfoil at these late times. For earlier times while the baseline DSV was slightly
closer to the airfoil, the tubercled airfoil had higher circulation at lower pitch angle
and significant increases in circulation over the baseline at high angles. These factors,
along with the weaker secondary vortex for the tubercled airfoil, strongly imply that
the tubercled airfoil would have a higher lift over the majority of the pitching time
compared to a baseline straight airfoil.
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
DSV core alignment can also be observed in the convective velocity oscillations in
figure 11. It is important to note that the DSV never truly became aligned as the
process was dynamic with spanwise oscillations observed in the relative location of
the DSV (figure 10a). The nominal alignment of the DSV was likely due to the
decreased induced velocity from the streamwise vortices and increased strength of
the DSV as the pitch angle was increased. The spatial alignment of the DSV is a
feature that is characteristic of the dynamics of the tubercled airfoil not observed for
finite-span wings.
Circulation was also likely increased for the tubercled wing due to the pinning of
the DSV to the tubercled wing for a longer period of time compared to the baseline.
This allowed more time for the feeding vorticity from the leading edge to build up
within the tubercled wing DSV. It is likely that the DSV also interacted with the
streamwise vorticity formed by the tubercles, increasing the total circulation. The
difference in circulation between the trough and peak planes adds further evidence
to the streamwise vortex interaction theory, but a full three-dimensional flow field
would be needed to identify the streamwise vortices as the source of the added DSV
circulation.
Effects of leading-edge tubercles on dynamic stall 893 A5-23
(a) (b) 50
0.4 Baseline 45
Peak
40
Trough
35
åformation
Ø* 30
0.2
25
0.1 20
15
10
-0.25 -0.15 -0.05 0 0.05 0.15 0.25 0.1 0.2 0.4
xAF/c Ø*
F IGURE 14. Pitch rate effects on formation angle and streamwise position.
Maximum circulation for the baseline airfoil increased by 16 % and 20 % for the
Ω ∗ = 0.2 and Ω ∗ = 0.4 cases, respectively, when compared to the Ω ∗ = 0.1 case.
While the maximum circulation on the tubercled airfoil did increase slightly when
the pitch rate was doubled to Ω ∗ = 0.2, it actually decreased slightly when the pitch
rate was doubled again to Ω ∗ = 0.4. Despite minimal changes in performance as a
function of pitch rate, the tubercled airfoil still outperformed the baseline one at all
pitch rates. These results suggest that the features of the tubercled airfoil maximize
the peak circulation independent of pitch rate. At the highest pitch rate, the tubercled
airfoil DSV circulation reached a plateau early in the convection, meaning circulation
was high for most of the time that the DSV was near the airfoil.
Some changes in the DSV convective path occurred as a function of pitch rate.
Figure 16 shows the track of the DSV at the three pitch rates. At the higher pitch
rates, the large-scale difference in path between the baseline and tubercled case
became less pronounced. However, the oscillatory motion of the DSV for the tubercled
airfoil showed an increasing frequency as the pitch rate was increased. In general,
the path of the DSVs on the baseline and tubercled airfoils became more similar as
the pitch rate increased.
893 A5-24 J. T. Hrynuk and D. G. Bohl
4.0
Baseline
3.5 Peak
3.0 Trough
2.5
˝/U∞c
2.0
1.5
1.0
0.5 Ø * = 0.1 Ø * = 0.2 Ø * = 0.4
0
1.5 2.5 3.5 4.5 5.5 1.5 2.0 2.5 3.0 3.5 4.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
t * = tU∞/c t * = tU∞/c t * = tU∞/c
1.0
Baseline
0.8 Peak
Trough
0.6
yAF/c
0.4
0.2
Ø * = 0.1 Ø * = 0.2 Ø * = 0.4
0
-0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6
xAF/c xAF/c xAF/c
F IGURE 16. Dynamic stall vortex path for Ω ∗ = 0.1, 0.2 and 0.4.
0.8
Baseline Ø * = 0.1 Ø * = 0.2 Ø * = 0.4
0.6 Peak
Trough
Vcon√AF/U∞
0.4
0.2
0
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
-0.2
1.5 2.5 3.5 4.5 5.5 1.5 2.0 2.5 3.0 3.5 4.0 1.0 1.5 2.0 2.5 3.0
t * = tU∞/c t * = tU∞/c t * = tU∞/c
Comparing the convective speeds using the VconvAF metric (figure 17), the trends in
convective speed remained much the same despite the convergence of the convective
paths. The velocity spike associated with the convergence of the tubercled airfoil
DSVs was less pronounced with increased pitch rate, although that feature might
be too short to be easily observed with the reduced time resolution of the higher
pitch rates. The convective speeds for the different planes on the tubercled airfoil
remained slightly out of phase after the DSV aligned, again highlighting the dynamic
nature of the DSV convection process. The average convective speed of the tubercled
DSV appeared to increase slightly as a function of pitch rate, but overall trends were
similar at all pitch rates tested.
Effects of leading-edge tubercles on dynamic stall 893 A5-25
0
-0.2 0 0.2 0.4 -0.2 0 0.2 0.4 0.6
x/c x/c
0.4
y/c ø
0.2 30.0
22.5
15.0
7.5
0.6 -7.5
t * = 2.5 t * = 2.8 t * = 3.2 -15.0
-22.5
0.4 -30.0
y/c
0.2
0
-0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6 0.8
x/c x/c x/c
behaviour. One flow feature that appeared to be most affected by the addition of
tubercles and changes in pitch rate was the SLV, which formed near the trailing
edge and in some cases interacted with the DSV. The strength of the SLV was
observed to be a function of pitch rate for both the baseline and tubercled airfoils
(figure 18). While it is difficult to make a direct comparison between pitch rates,
figure 18 compares the SLVs at times when the baseline SLV was at x/c ≈ 0.4 and
the corresponding time step on the peak plane of the tubercled airfoil. The SLV
on the tubercled airfoil was more diffuse and at many times appeared more like a
separated shear layer than the compact isolated SLV that formed on the baseline
airfoil.
A comparison of the convective velocities of the DSV shows some hints of the
impact that the SLV has on DSV convective behaviours. Looking at the baseline
airfoil convective velocity (figure 17) for Ω ∗ = 0.2, the DSV accelerated beginning at
around t∗ = 1.75 up to the peak, then decelerated rapidly until t∗ = 3.75. During the
acceleration period of the DSV, the SLV was moving rapidly towards the trailing edge
(figure 19). A reversal in the direction of the SLV occurred around t∗ = 2.8, directly
corresponding to the peak convective speed of the DSV. After the peak convective
893 A5-26 J. T. Hrynuk and D. G. Bohl
0.4
y/c
0.2
0.4
y/c
0.2
0
-0.2 0 0.2 0.4 -0.2 0 0.2 0.4 -0.2 0 0.2 0.4 -0.2 0 0.2 0.4
x/c x/c x/c x/c
speed of the DSV occurred, the SLV reversed direction and orbited the DSV as the
DSV continued convecting downstream.
When comparing the SLV interaction with the tubercled airfoil, results suggest
that the SLV, or general lack thereof, caused a lower peak convective speed of the
DSV and introduced oscillations into the convection behaviour. Figure 20 shows
the formation and convection of the SLV on the peak plane of the tubercled airfoil.
Compared to the baseline airfoil (figure 19), the tubercled airfoil suppressed the
formation of the SLV and a much weaker SLV convected upstream.
While the tubercled airfoil SLVs were notably weaker compared to the baseline,
higher pitch rate caused the formation of stronger SLVs (figure 18). Thorough
comparison of flow field and tracking data has shown that each major acceleration
and deceleration in the convective velocity of the DSV correlated to interaction and
motion of the SLVs, independent of the leading-edge condition. This feature of the
flow was not easily observed for the Ω ∗ = 0.1 pitch rate case because the tubercled
airfoil inhibited the formation of the SLVs. However, the feature observed for the
baseline DSV at Ω ∗ = 0.1 is observable for tubercled airfoil planes at Ω ∗ = 0.2.
Figure 21 shows the peak and trough planes on the tubercled airfoils with a focus on
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
the DSV acceleration between t∗ = 2.5 and t∗ = 3 (figure 17). At around t∗ = 2.5 there
were signs of the SLV near the trailing edge which began to orbit the DSV as time
progressed. The weak SLV near the trailing edge was stretched as it passed near the
surface of the airfoil before disappearing near the end of the acceleration observed
in figure 17. It also becomes apparent at t∗ = 2.7 that the SLV on the trough plane
had advanced further forward compared to the peak plane, which aligns well with
the tracking data showing the trough DSV accelerating before the peak DSV.
Taken as a whole, these results indicate that the dynamics of the DSV is affected
by the induced velocity of the SLV (and vice versa) and the presence or absence of
the SLV is critical in the dynamic stall physics. For the tubercled airfoil the SLV was
weakened at the lower reduced frequency. The differences between the airfoils were
most pronounced at this low pitch rate. As the pitch rate increased, the SLV became
stronger and more pronounced, and the differences in the convective dynamics of the
DSV became less noticeable (Shih et al. 1995). The SLV as a flow feature has had
limited focus in prior works. Shear-layer vortices were observed to form at a range of
Reynolds numbers from Re = 5000–25 000 (Shih et al. 1995) to Re = 0.54 × 106 –1.5 ×
106 (Oshima & Ramaprian 1997), which suggests that their formation is independent
Effects of leading-edge tubercles on dynamic stall 893 A5-27
0
Peak t * = 2.7 Trough t * = 2.7
0.2
0
Peak t * = 3.1 Trough t * = 3.1
0.2
0
-0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6
x/c x/c
F IGURE 21. Interaction between SLV and DSV on the tubercled airfoil at Ω ∗ = 0.2.
of Reynolds number. However, the specific effects of airfoil shape, Reynolds number,
turbulence, etc., on SLVs are still largely unknown.
5. Conclusions
The effect of leading-edge tubercles on airfoil aerodynamics, previously only studied
under fixed AoA conditions, has been studied experimentally under deep dynamic stall
conditions. A NACA 0012 airfoil was modified with tubercles having an amplitude
of 4 % of the chord and a wavelength of 24 % of the chord and was compared to
a baseline straight-leading-edge airfoil at Rec = 12 000 using MTV. The velocity and
vorticity fields were quantified at constant pitch rates (Ω ∗ = 0.1, 0.2, 0.4) with a
motion of α = 0–50◦ . Tracking of the DSV was performed by locating the peaks in
the Γ1 fields, rather than the location of the peak vorticity to account for the diffuse
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press
Declaration of interests
The authors report no conflict of interest.
REFERENCES
B OHL , D. G. & K OOCHESFAHANI , M. M. 2009 MTV measurements of the vortical field in the
wake of an airfoil oscillating at high reduced frequency. J. Fluid Mech. 620, 63–88.
B ORG , J. 2012 The effect of leading edge serrations on dynamic stall. Masters thesis, University of
Southampton, Faculty of Engineering and the Environment.
B RANDON , J. M. 1991 Dynamic stall effects and applications to high performance aircraft. NASA
Langley Tech. Rep. 10.5555/887303.
C AI , C., L IU , S., Z UO , Z., M AEDA , T., K AMADA , Y., L I , Q. & S ATO , R. 2019 Experimental
and theoretical investigations on the effect of a single leading-edge protuberance on airfoil
performance. Phys. Fluids 31, 027103.
Effects of leading-edge tubercles on dynamic stall 893 A5-29
C AI , C., Z UO , Z., M AEDA , T., K AMADA , Y., L I , Q., S HIMAMOTO , K. & L IU , S. 2017 Periodic
and aperiodic flow patterns around an airfoil with leading-edge protuberances. Phys. Fluids
29, 115110.
C ARR , L. W., M C A LISTER , K. W. & M C C ROSKEY, W. J. 1977 Analysis of the development of
dynamic stall based on oscillating airfoil experiments. NASA-TN-D-8382, A-6674.
C HEN , J. H., L I , S. S. & N GUYEN , V. T. 2012 The effect of leading edge protuberances on the
performance of small aspect ratio foils. In 15th International Symposium on Flow Visualization,
June 25–28. Belarus National Academy of Sciences, Luikov Heat & Mass Transfer Institute,
Minsk, Belarus.
C HOUDHURI , P. G., K NIGHT, D. D. & V ISBAL , M. R. 1994 Two-dimensional unsteady leading-edge
separation on a pitching airfoil. AIAA J. 32 (4), 673–681.
C OHN , R. K. & K OOCHESFAHANI , M. M. 2000 The accuracy of remapping irregularly spaced
velocity data onto a regular grid and the computation of vorticity. Exp. Fluids 29, Supplemental,
S61–S69.
C ORSINI , A., D ELIBRA , G. & S HEARD , A. G. 2013 On the role of leading-edge bumps in the
control of stall onset in axial fan blades. Trans. ASME J. Fluids Engng 135, 081104.
C OTON , F. N., G ALBRAITH , R. A. M. & G REEN , R. B. 2001 The effect of wing planform shape
on dynamic stall. Aeronaut. J. 105 (1045), 151–159.
C USTODIO , D. 2008 The effect of humpback whale-like leading edge protuberances on hydrofoil
performance. PhD Dissertation, WPI.
D ICKINSON , M. H., L EHMANN , F.-O. & S ANE , S. P. 1999 Wing rotation and the aerodynamic
basis of insect flight. Science 284, 1954–1960.
E KATERINARIS , J. A. 2002 Numerical investigations of dynamic stall active control for incompressible
and compressible flows. J. Aircraft 39 (1), 71–78.
E LLINGTON , C. P., B ERG , C. VAN DEN, W ILLMOTT, A. P. & T HOMAS , A. L. R. 1996 Leading-edge
vortices in insect flight. Nature 384, 626–630.
FAVIER , J., P INELLI , A. & P IOMELLI , U. 2012 Control of the separated flow around an airfoil using
wavy leading edge inspired by humpback whale flippers. C. R. Méc. 340, 107–114.
F ISH , F. E. & BATTLE , J. M. 1995 Hydrodynamic design of the Humpback Whale flipper. J. Morphol.
225, 51–60.
G AO , H., W EI , M. & H RYNUK , J. T. 2018 Data-driven ROM for the prediction of dynamic stall. In
AIAA Aviation Forum, doi:10.2514/6.2018-3094.
G ENDRICH , C. P. 1999 Dynamic stall of rapidly pitching airfoils: MTV experiments and Navier–
Stokes simulations. PhD thesis, Michigan State University.
G ENDRICH , C. P., B OHL , D. G. & K OOCHESFAHANI , M. M. 1997 Whole-field measurements of
unsteady separation in a vortex ring/wall interaction. In 28th AIAA Fluid Dynamics Conference,
https://doi.org/10.1017/jfm.2020.216 Published online by Cambridge University Press