Vortex Dynamics around Pitching Plates

Ryan T. Jantzen,1, a) Kunihiko Taira,1, b) Kenneth O. Granlund,2, c) and Michael V.

Ol2, d)
1)
Department of Mechanical Engineering and Florida Center for Advanced
Aero-Propulsion, Florida State University, Tallahassee, Florida 32310,
USA
2)
Aerospace Systems Directorate, U.S. Air Force Research Laboratory,
Wright-Patterson Air Force Base, Ohio 45433, USA
(Dated: 29 April 2014)
Vortex dynamics of wakes generated by rectangular aspect-ratio 2 and 4 and two-
dimensional pitching flat plates in free stream are examined with direct numerical
simulation and water tunnel experiments. Evolution of wake vortices comprised of
tip, leading-edge and trailing-edge vortices is compared with force history for a range
of pitch rates. The plate pivots about its leading edge with reduced frequency from
π/8 to π/48, which corresponds to pitching over 1 to 6 chord lengths of travel.
Computations have reasonable agreement with experiments, despite large differences
in Reynolds number. Computations show that the tip effects are confined initially
near the wing tips, but begin to strongly affect the leading-edge vortex as the motion
of the plate proceeds, with concomitant effects on lift and drag history. Scaling
relations based on reduced frequency are shown to collapse aerodynamic force history
for the various pitch rates.

PACS numbers: Vortex dynamics (fluid flow), 47.32.C-, Separated flows, 47.32.Ef-,
Wakes, laminar, 47.15.Tr-

a)
Electronic mail: [email protected]
b)
Electronic mail: [email protected]
c)
Electronic mail: [email protected]
d)
Electronic mail: [email protected]

1
I. INTRODUCTION

The humble rigid flat plate remains a useful abstraction for the study of separated flow
with unsteady boundary conditions, for applications of aircraft undergoing agile maneu-
vers and operating in harsh environments such as gusts. Various types of small to large
unmanned air vehicles with fixed, rotating, or flapping wings1,2 oftentimes take inspiration
from biological flyers to take advantage of vortical effects for enhanced lift. In gusty en-
vironments, the effective free stream velocity and the angle of attack can change within a
short period of time3 , which with a buoyancy correction is the inverse problem to that of
maneuver in steady freestream4 .
Flow separation and the resulting vortex formation are inherently nonlinear phenomena,
with nontrivial deviations from classical unsteady theories5,6 that motivate an aim for de-
tailed understanding of the physics to predict the flow field and the aerodynamic forces7 .
Study of the vortex dynamics around plates undergoing unsteady motions has been identi-
fied as an international research task by the NATO AVT-202 Group and the AIAA Fluid
Dynamics Technical Committee’s Low Reynolds Number Discussion Group. The identified
research task encompasses a wide variety of motions, including pitching8,9 , rotation10,11 ,
various acceleration profiles12,13 , and the combination thereof14 .
There have been extensive experimental studies on the pivot point location for purely
pitching plates. Two-dimensional flat plates were studied using direct force measurements
and qualitative dye flow visualization by Granlund et al.15 to compare the flow field evolution
and aerodynamic forces for pitching maneuvers over a range of reduced frequencies and
pivot point locations. They were able to correlate lift and drag coefficients as functions
of both angle of attack and pivot point location for reduced frequencies greater than 0.1.
For finite-aspect ratio wings, Yu and Bernal16 studied the effect of pivot point location and
reduced frequency on the flow structure and aerodynamic forces for an AR = 4 pitching
flat plate using direct force measurements and two-dimensional PIV velocity measurements.
Granlund et al.17 performed an extensive parametric study on the influence of aspect ratio,
pivot-point location, and reduced frequency for rotational and translational accelerating flat
plates from Re = 14 to 10,000. For all Reynolds numbers considered, both non-circulatory
and circulatory loading resulting from the acceleration and deceleration of the plate were
found to be highly dependent on the pivot-point location; for instance, a more forward pivot
point produces a higher peak lift due to an induced camber effect5 .
On the numerical side, Taira and Colonius7 used direct numerical simulations to analyze
the three-dimensional separated flow over low-aspect-ratio flat plates in translation. They
observed that the tip vortices from the low-aspect-ratio flat plates help stabilize the separated
flow for impulsively started plates. The stability of the wake dynamics was also characterized
over a wide range of aspect ratios and angles of attack. High-fidelity simulations were
performed to investigate the effect of reduced frequency and Reynolds number on the flow
structure and unsteady loading for pitching flat plates of AR = 2 and spanwise periodic
configuration by Visbal18 and Garmann and Visbal19 , respectively. They found that for all
pitch rates considered, there is a significant increase in the maximum lift achieved compared
to plates at static angles of attack. Visbal18 noted that the increase in lift for finite-aspect-
ratio plates can be attributed to the three-dimensional dynamic stall process resulting from
the formation of a LEV that evolves into an arch-type vortex. He showed the flow fields to
be qualitative similar over Re = 1,000 to 40,000. In a companion study to Granlund et al.17 ,
Jantzen et al.20 reported on the three-dimensional wake dynamics and unsteady forces for

2
both pitching and accelerating low-aspect-ratio plates. It was found that reduced tip effects
for higher-aspect-ratio plates influence the formation and evolution of the LEV.
The current investigation examines the vortex dynamics around plates with aspect ratio 2
and 4 (and 2D) pitching about the leading edge, using both numerical (immersed boundary
method) and experimental (water tunnel) approaches. We consider such canonical pitch-up
motion that can potentially be encountered by miniature aircraft in gusty conditions to
provide a better understanding of how the flow evolves around the wing and how unsteady
aerodynamic forces are generated. The pitch rates are selected to bracket the range of time
scales identified with the vortex formation time21,22 . The present work particularly highlights
the finite-aspect-ratio effects (tip effects) for pitching wings which has not been examined in
detail with computations. By performing parameter studies with varied pitching frequency
and plate aspect ratio, we aim to provide insights towards designing new vehicles that are
able to withstand the evolving three-dimensional flow structure on the maneuvering wings
and also utilize the unsteady forces generated during these motions efficiently for improved
performance and stability.
The present paper is organized in the following manner. In Section II, we present the
problem setup and methodology. Section III discusses the formulation and dynamics of the
vortices around the pitching plate at various pitch rates and aspect ratios, to elucidate the
influence of the leading-edge and tip vortices. Section IV analyzes the generation and growth
of the leading-edge vortices during the pitching motion. We provide discussions on the un-
steady aerodynamic forces experienced by the wing in Section V. A non-dimensional scaling
that incorporates pitch rates is provided to collapse the force history during pitching. The
numerical and experimental results are compared in Section VI to illustrate the differences
due to Reynolds number effects. Despite the large difference in Reynolds numbers, the flow
fields and force histories are found to be in qualitative agreement. Concluding remarks are
offered in Section VII to summarize the findings from this study.

II. PROBLEM SETUP

The present investigation considers flat-plate wings with rectangular planform of var-
ious aspect ratios undergoing a pitching maneuver about the leading edge in a constant
freestream. The spatial coordinates are defined with x, y, and z representing the stream-
wise, vertical, and spanwise directions, respectively. Length scales are non-dimensionalized
by the chord length c, and the velocity vector is non-dimensionalized by the freestream ve-
locity value U∞ . Temporal variable t is the non-dimensional convective time, normalized
by the freestream velocity and the chord length. The Reynolds number is Re = U∞ c/ν,
where ν is the kinematic viscosity. The forces on the flat plate
(Fx , Fy ) are reported as
1 1
2 2


non-dimensional lift and drag, defined by CL = Fy / 2 ρU∞ A and CD = Fx / 2 ρU∞ A ,
respectively, where A is the planform area of the plate and ρ is the density of the fluid.
The flat plate pitches from zero-incidence to a post-stall angle of attack αmax in a
smoothed linear ramp. Smoothing of the start and end of the pitch ramp is through a
form of relationship proposed by Eldredge23 :
 
Ω◦ cosh [a (t − t1 )] αmax
α(t) = log + , (1)
2a cosh [a (t − t2 )] 2

where Ω0 is the nominal pitch rate given by Ω0 = αmax /tp , where αmax is the maximum angle

3
of attack at the end of the pitching motion (taken in all cases as 45 deg), and tp = t2 − t1
is the pitching interval. Here, t1 = 0 and t2 is determined by the reduced frequency (K ≡
Ω0 c/2U∞ = αmax c/2U∞ tp ) of the pitching motion, t2 = t1 + αmax c/(2U∞ K). The parameter
a is a smoothing parameter chosen to regularize the sharp jump in α̈. The time history of
the pitch incidence angle α is illustrated in Figure 1 for different rates. This motion profile
has been selected to be the test case for a collective study by the NATO AVT-202 working
group. Throughout this paper, we refer to the different pitch-rate cases as CN , where N
indicates the convective time units (chord lengths) over which the plate pitches.

α [de g] 45

30

C1
15 C2
C4
C6
0
0 1 2 4 6
t
FIG. 1: The smoothed linear ramp used for the pitching maneuver from α = 0◦ to 45◦ .

The smoothing parameter is chosen to be a = 21, 16, 11, and 4, for the cases of C1, C2,
C4, and C6, respectively. These values are selected to match work by Visbal18 and Yilmaz
and Rockwell24 . In numerical simulations, we consider rectangular wings of aspect ratios 2
and 4 and the two-dimensional case for the freestream Reynolds number Re = 300, which is
a value chosen to elude issues of turbulence and to highlight the large-scale wake structures
generated by the unsteady motion of the wing. The reduced frequency K = π/8, π/16,
π/32, and π/48 correspond to the cases of C1, C2, C4, and C6, respectively.

A. Computational setup

Three-dimensional incompressible flow over the flat-plate wing is numerically solved with
the immersed boundary projection method25 , which creates a plate in the domain with a
set of Lagrangian points where appropriate boundary forces are introduced to enforce the
no-slip condition on the wing surface. This method has been used to simulate a wide variety
of flows and has been well-validated7,25,26 . We use a computational domain with a typical
size of (x, y, z) ∈ [−4, 6] × [−5, 5] × [−5, 5]. The plate is positioned in the computational
domain with its midspan point on the leading edge at the origin, as shown in Figure 2. The
inlet and side boundary conditions are set to constant uniform flow with U∞ = 1 and the
outlet boundary condition uses a convective boundary condition of ∂u/∂t + U∞ ∂u/∂x = 0
to allow the wake vortices to freely exit the computational domain without disturbing the
near-field solution. A simulation is first performed to determine the steady-state flow over
the plate at zero degree angle of attack. This steady state is then used as the initial condition
for all simulations in which the wing undergoes the pitching motion prescribed by Eq. (1).

4
The thickness of the plate in the simulation is modeled to be infinitely thin with a discrete
delta function.

U1
y
z x
z ⌦

x z
FIG. 2: The computational setup showing the xz, xy, and zy grid planes with every fourth
gridline and the AR = 2 pitching plate shown.

To ensure numerical simulations are performed with sufficient spatial grid resolution, we
perform a pitching plate study but with varying grid sizes. Shown in Figure 3 gives the
lift and drag histories with varied grid resolutions from a case where the AR = 2 plate is
pitching over one chord length of travel (C1) for Re = 300 with grid sizes listed in Table
I. The results in Figure 3 and Table I show that the medium size grid provides sufficient
resolution to achieve convergence. This is also illustrated by the flowfield images shown on
the right side of Figure 3. For the three cases, we essentially see no observable differences
in the vortical structures. For all cases discussed below, the resolution used in this study is
based on the medium resolution. The computational domain size is increased with added
points in the spanwise direction for AR = 4 cases. The CFL number is limited to 0.5 in all
cases.
Resolution Coarse Medium Fine
nx × ny × nz 141 × 91 × 116 170 × 110 × 140 212 × 137 × 175
max |CL −CL,fine |
max |CL,fine |
6.9 % 3.0% –
max |CD −CD,fine |
max |CD,fine |
11.2 % 5.7% –

TABLE I: Number of grid points used for the grid resolution study.

5
 6
Fine
5 Medium
Coarse

Fine
4
CL
3

Medium
0
0 0.5 1 1.5 2
t
6

4
CD

2 Coarse
1

0
0 0.5 1 1.5 2 t=1 t=2
t

FIG. 3: Grid resolution study showing the lift (top left) and drag (bottom left) force
coefficients along with snapshots of the instantaneous Q-criterion (Q = 3) and vorticity
magnitude (kωk = 3) showing the grid dependence on the three-dimensional flow structure
for the AR = 2 flat plate.

B. Experimental Setup

Direct force measurement and fluorescent dye flow visualization for pitching plates at
Re = 20, 000 were conducted in the Horizontal Free-surface Water Tunnel at the U.S.
Air Force Research Laboratory at Wright–Patterson Air Force Base27 . The tunnel has 4:1
contraction ratio and 0.46 m wide by 0.61 m high test section, speed range of 3 to 45 cm/s,
and u-component turbulence intensity of 0.4% at 15 cm/s. The tunnel is fitted with a three-
degree of freedom electric rig, consisting of a triplet of H2W linear motors, driven by AMC
DigiFlex servo-drives controlled by a Galil DMC 4040 4-channel card, with user-selected
proportional/integral/derivative (PID) constants for each channel. The model motion of
pitch and plunge are controlled via two motors mounted vertically on a plate above the
tunnel test section, shown in Figure 4 (left). For the present study, AR = 2 and 4 flat plates
are used with a chord length of 117mm and 75mm, respectively. The thickness of the plate
is 1.59mm (1/16 in).
Flow visualization is limited to qualitative inferences from dye illuminated by planar
laser fluorescence. Rhodamine 6G dissolved in water is injected at the leading and trailing-
edge 3/4 semispan location by a positive-displacement pump, connecting to a set of 0.5mm
internal-diameter rigid lines glued to the surface of the plate, in an approach similar to that
of Ol et al.27 The dye is illuminated by an Nd:YLF 527nm pulsed laser sheet of ≈ 2mm
thickness at 50Hz, and images are recorded with a PCO DiMax high-speed camera through

6
FIG. 4: The experimental setup in the U.S. Air Force Research Laboratory Water Tunnel.
Water tunnel with motion mechanism (left) and flat-plate model (right).

a Nikon PC-E 45mm Micro lens. A Tiffen orange #21 filter is used to remove the incident
and reflected laser light, leaving only the fluorescence.
Force data are recorded from an ATI Nano-17 IP68 6-component integral load cell. The
load cell is visible in Figure 4 (right) at the junction of the triangular mounts. Load cell
strain gage electrical signals are A/D converted in an ATI NetBox interface and recorded
using a Java application, and are filtered in three steps. The first is a low-pass filter in
the ATI NetBox at f = 34 Hz, to avoid introducing noise not correlated with motion force
data as well as structural eigenfrequency. These are on the order of 50Hz. The second
step uses a moving-average of 11 points to smooth the data and to numerically stabilize the
third filtering operation, which is the fourth order Chebychev II low-pass filter with −20dB
attenuation of the stopband. The cutoff frequency is five times the motion frequency, were
the ramp motion to have been a 1/4-sine wave. In the present study, the force measurements
have an uncertainty level of < 2% of the steady state values (of those after the pitch-up
motion ends). The uncertainty in experimental measurements were obtained from loadcell
quoted uncertainty and the temporal average of 10 repetitions of each experimental case.

III. VORTEX DYNAMICS

Here, we focus on the evolution of the vortical structures around the wing and the effects
that the pitch rate and aspect ratio have on the vortex dynamics. The three-dimensional
vortical structures for the AR = 2 and 4 flat plates are shown in Figure 5 for the pitching
case occurring over one chord of travel (C1), visualized by the iso-surface of the Q-criterion
(Q = 3) in blue and the iso-surface of the magnitude of vorticity (kωk2 = 3) in gray. The
Q-criterion shows the vortex cores and the vorticity norm highlights the vortex sheets. The
values for the iso-surfaces were chosen based on previous studies of low Reynolds number
simulations7 .
As the plate begins to pitch and the incidence angle increases to α = 11◦ , the vortex
sheet over its top surface begins to roll up, initiating the formation of the LEV. As the
angle further increases from α = 23◦ to 38◦ , tip vortices arise from the pressure difference
between the top and bottom surfaces, creating a vortex loop with the starting vortex that
has detached from the trailing edge of the plate, the two tip vortices and the LEV. The

7
 K = ⇡/8 t=0 t = 0.25 t = 0.5 t = 0.75 t=1
(C1) (↵ = 0 ) (↵ = 11 ) (↵ = 23 ) (↵ = 38 ) (↵ = 45 )

AR = 4

AR = 2

FIG. 5: Instantaneous Q-criterion iso-surfaces (Q = 3, blue) and magnitude of vorticity

(kωk = 3, gray) showing the three-dimensional flow structure for the AR = 4 (top) and 2
(bottom) plates for the C1 case at Re = 300. The flow field is visualized for a reference
frame fixed with the plate.

growth of the LEV is uniform across most the span of the plate, while the legs of the LEV
stay pinned to the corners at the leading edge. By the time that the plate has reached the
maximum angle of α = 45◦ , the trailing-edge (starting) vortex has advected downstream.
During the pitching motion, the tip vortices remain pinned to the leading-edge corners of the
plate. The LEV has continued to grow, remaining fairly uniform while still being attached to
the plate. In the present discussion, the LEV is considered detached18 when the legs of the
LEV are no longer attached to the corners of the leading edge. We note that the definition
of the detached LEV does not depend significantly on the values of the iso-surfaces chosen
to visualize them. Throughout the entire motion for this pitch rate, the flow field evolution
is very similar between AR = 2 and 4 at each respective snapshot of time. As will be seen
subsequently, aspect-ratio independence of the flow field is attenuated with decreasing pitch
rate.
The evolution of the flow structure for all four of the pitch rates considered is given in
Figures 6 and 7 for the AR = 2 and 4 plates, respectively. To extend upon the discussion
about the fastest pitching rate, C1, we consider the flow structure later in time after the
plate has completed its pitching maneuver (t > 1). By t = 2, the tip vortices separate
from the trailing edge of the plate while staying attached to the corners of the leading edge.
During this process, a second vortex loop, created by the quick angular deceleration of the
plate, sheds from the trailing edge which wraps around the initially generated tip vortices.
By t = 3, the LEV begins to lift off at the centerline of the plate and resembles an arch-type
vortex that has been previously observed numerically by Visbal18 and experimentally by
Yilmaz and Rockwell24 for AR = 2 pitching plates at moderate Reynolds numbers.
After t = 3, we start to observe differences in the flow structure between the AR = 2 and
4 plates, especially in the evolution of the LEV. For the AR = 2 plate, the LEV remains
relatively close to the surface of the plate long after (t > 8) it has detached from the corners
of the leading edge. This delayed advection is due to the presence of the stronger influence
from the tip vortices on the mid-span region for the AR = 2 plate, similar to what is observed
by Taira and Colonius7 . At t = 5, we begin to notice that the tip vortices begin to roll in
towards the midspan and by t = 8, the counter rotating tip vortices have become very close
to one another, creating a significant downwash which essentially pulls the LEV towards the
surface of the plate, slowing down the advection of the LEV.

8
 AR = 2 t=1 t=2 t=4 t=6 t=8

K = ⇡/8
(C1)

↵˙ = 0

K = ⇡/16
(C2)

↵˙ > 0

K = ⇡/32
(C4)

K = ⇡/48
(C6)

FIG. 6: Instantaneous Q-criterion iso-surfaces (Q = 3, blue) and vorticity magnitude

iso-surfaces (kωk = 3, gray) showing the three-dimensional flow structure for the AR = 2
flat plate at various times during the four different pitching motions at Re = 300. The flow
field is visualized for a reference frame fixed with the wing. The images to the left of the
dotted line correspond to when the plate is still in motion (α̇ > 0), and the images to the
right correspond to when the plate has reached the maximum α and is no longer moving
(α̇ = 0, α = 45◦ ).

In contrast, for the AR = 4 plate, the LEV lifts away from the suction-side surface and
detaches from the corners at the leading edge. By t = 8, the LEV develops into a large
horseshoe vortex as it travels away from the plate due to the vortex being stretched along
the streamwise and plate normal directions. Another difference that is present in the AR = 4
case that is not observed for the AR = 2 plate are the smaller structures that appear to wrap
around the head of the horseshoe type vortex at later times (t > 5). These structures result
from the continued roll up of the trailing-edge vortex and the large spanwise variation of the
flow structure. The trailing-edge vortex (TEV) rolls up faster along the midspan compared
to the AR = 2 case, while at the tips of the plate there is a delay in this roll up due to the
downwash induced by the tip vortices and the legs of the LEV on the surface of the plate.

9
 AR = 4 t=1 t=2 t=4 t=6 t=8

K = ⇡/8
(C1)

↵˙ = 0

K = ⇡/16
(C2)

↵˙ > 0

K = ⇡/32
(C4)

K = ⇡/48
(C6)

FIG. 7: Instantaneous Q-criterion iso-surfaces (Q = 3, blue) and vorticity magnitude

iso-surfaces (kωk = 3, gray) showing the three-dimensional flow structure for the AR = 4
flat plate at various times during the four different pitching motions at Re = 300. The flow
field is visualized for a reference frame fixed with the plate. The images to the left of the
dotted line correspond to when the plate is still in motion (α̇ > 0), and the images to the
right correspond to when the plate has reached the maximum α and is no longer moving
(α̇ = 0, α = 45◦ ).

We next consider the influence of various pitch rates (C1, C2, C4, and C6) for the
AR = 2 and 4 plates as shown in Figures 6 and 7, respectively. The times necessary for
LEV formation and detachment are seen to depend strongly on pitch rate. For the C1 case,
we previously observed that the pinch-off of the LEV occurs around t = 4, which is well
after the pitching motion has stopped. It can be seen that for the C2 case, the formation
of the LEV also occurs well after the end of the pitching motion and it detaches between
t = 4 and 5. For the two slowest pitching rates, C4 and C6, the LEV forms by the end
of the motion (t = 4 and 6, respectively) and detaches roughly around one convective time
unit afterwards for the AR = 2 plate. In the case of AR = 4 plate, this detachment occurs
around one convective time unit after the end of the pitching motion. For the AR = 2

10
 AR = 2 AR = 4 2D
4 4 4
C1 C1 C1
C2 C2 C2
3 3 3
C4 C4 C4
C6 C6 C6
2 2 2
x

x
1 1 1

0 0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
t t t
2 2 2

1 1 1

0 0 0
y

y
−1 −1 −1

−2 −2 −2
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
t t t
1 1 1
0.2
0.2
0.2

0.5 0.5 0.5

00 0

0
y

0 0
y

y
−0.5 -0.2
0.20.20.75 −0.5 −0.5
1 1.25
0.75
0.75 1 1.25
1.25
−1 −1 −1
0 1 2 3 0 1 2 3 0 1 2 3
x x x

FIG. 8: Comparison of the temporal evolution of the spatial location of the LEV for the
four pitch rates in the x-direction (top row) and y-direction (middle row) and the spatial
location of the LEV with respect to the plate (bottom row) for the AR = 2 (left column),
4 (center column) and 2D (right column) flat plates. The solid line represents the first
LEV and the dashed line represents the formation of a second LEV. For the time period
considered in this study, no additional LEVs are observed for the AR = 2 cases.

plate, at the furthest convective time considered in the present study, the LEV is located at
roughly the same location for all pitch rates considered. This is not the case for the AR = 4
wing, since the tip vortices have a much weaker influence on the detachment of the LEV.

IV. LEADING-EDGE VORTEX TRACKING

In order to better understand the vortex dynamics associated with the present pitching
wing simulations, the vortex identification method of Graftieaux et al.28 is employed to track
the formation and evolution of the LEV. The vortex identification function γ1 captures the
center of a vortex when its value takes the maximum value (theoretically max(γ1 ) = 1). In
the current study, γ1 is calculated on the midspan plane. We select a threshold of γ1 ≥ 0.9
to identify a vortex and track its center where the maximum spatial value of γ1 is attained.
The trajectory of the LEV centroid on the midspan plane for all pitch rates considered
for the AR = 2, 4, and 2D plates are given in Figure 8. We note that the calculation of the
LEV centroid begins at different times for the four different pitch rates, due to the time it

11
takes for the LEV to initially form (with γ1 ≥ 0.9). For the AR = 2 plate, the LEV centroid
travels from the leading edge in roughly the same direction for all four pitch rates, until
it advects one chord length downstream. For the two fastest pitch rates, C1 and C2, the
LEV actually moves upstream before continuing to travel downstream (for t < 8). The C4
case for AR = 2 shows some slight upstream motion at the end of the trajectory but is not
as district as the lower pitch rate cases. As observed earlier, at t = 8 the position of the
LEV is in roughly the same position for all pitch rates. For the AR = 4 plate, the center
of the LEV travels the same path for the four pitch rates for roughly one chord length, and
then begins to vary. The LEV advects along the same path but with different velocity for
different pitch rates as evident from the x-t diagram. The dashed lines in Figure 8 for the
AR = 4 wing represent a second LEV that has developed, which was not observed for the
AR = 2 wing for the shown time frame. The emergence of the second LEV is obstructed
due to the initial LEV remaining over the wing surface for a longer period of time. For the
2D plate, the LEV centroid appears to initially take a similar path towards the trailing edge
as observed for the AR = 4 wing (although somewhat higher at x = 1). Once the LEV
reaches the trailing edge of the 2D wing, it advects downstream along similar paths for the
four pitch rates considered. The difference between the trajectories of the LEV between the
AR = 4 and 2D plates is caused by the lack of tip effects for the two-dimensional wing.
The strong tip vortices that develop while the wing pitches induce large downward velocity
that pushes the LEV towards the centerline of the wing and restricts it from immediately
lifting off. When comparing the two finite aspect ratios, the LEV remains closer to the wing
surface for the AR = 2 plate, which again is due to the tip vortices inducing a relatively
stronger downward velocity on the LEV along the midspan plane. For the AR = 4 plate,
the tip vortices concurrently pull fluid into the region between the arch-shaped LEV and
the tip vortices that in turn pushes the LEV away from the plate once the LEV reaches
a certain size. On the other hand, the tip vortices for the AR = 2 plate cover a majority
of the wing surface, resulting in significant downward induced velocity which restricts any
significant upward motion of the LEV.
Evolution of vortex strength (spanwise circulation Γz ) along the midspan is tracked using
the second vortex identification method of Graftieaux et al.28 , γ2 , to define the boundary
of the LEV. Graftieaux et al.28 defines regions where |γ2 | > 2/π to be locally dominated by
rotation and therefore represents the core of a vortex. Cutoff values for |γ2 | are normally
between 0.6 and 0.75 in the literature29,30 . Since we are interested in the circulation of the
LEV, we seek for regions where γ2 < −0.75, and then numerically integrate the spanwise
vorticity inside the vortex core boundary to approximate the midspan circulation of the
LEV.
Results from the Γz calculations at the midspan plane are given in Figure 9 for the
AR = 2, 4, and 2D plates. Each of the solid lines illustrates the growth of the first LEV
created by the different pitching motion. The dashed lines represent the development of
the second LEV for the AR = 4 and 2D plates. For the fast pitching C1 case, the LEV
circulation increases almost linearly during the pitching motion for the AR = 2 plate until
it levels off around t = 1.5. By t = 4, the circulation increases again and then levels off
for the AR = 2 case. For the AR = 4 plate, the circulation of the first LEV continues to
increase until t = 3. Afterwards, there is a decrease in its strength. This decrease is due to
the detachment of the LEV from the plate (the tail of the first LEV turns into the second
LEV) and diffusive vorticity flux escaping through the γ2 boundary. As can be seen from
the dashed lines on the right side of Figure 9, a second LEV begins to develop around this

12
time. We observe a similar behavior for the C2 cases for both aspect ratios, but offset by
roughly one convective unit.
For the 2D plate, the circulation of the first LEV increases linearly for the C1 case during
the pitching interval and then increases again until t = 3. At this point, the LEV centroid
begins to travel away from the surface of the plate, which was observed in Figure 8. While
we expect for the 2D plate that the circulation should monotonically increase until it levels
off to a constant value, there is a slight reduction in Γz for the C1 and C2 cases that occurs
at the end of the motion (at t = 1 and 2, respectively). This is due to how the boundary
of the γ2 cutoff contour is generated. While the first LEV is attached to the plate, the γ2
contour includes part of the shear layer near the leading edge, which provides added strength
to the calculated LEV circulation. The observed decease is actually an artifact of added
circulation from the high value of vorticity in the vicinity of the leading edge during the
pitching motion no longer being captured by the γ2 contour.
A much slower development of the LEV circulation is observed for all aspect ratios be-
tween t = 0 and 2 for the C4 and C6 pitching cases. The circulation begins to increase at
nearly the same rate for the AR = 2 and 4 plates beyond t = 2, and then somewhat levels
off for the AR = 2 case around t = 4.5. We notice for the AR = 4 plate, the calculated
circulation begins to decrease after it reaches its maximum value around t = 5 due to the
pinch off of the first LEV. The tail of the LEV structure is no longer included in the first
γ2 contour which results in the apparent decrease in Γz . The second LEV forms with the
excluded vorticity and leads to the growth of the second LEV, shown in Figure 9. For both
the AR = 4 and 2D plates, the midspan circulation of the second LEV begins to increase at
a similar time for the four pitch rates, but the rate at which it increases is greater for the
2D plate.

AR = 2 AR = 4 2D
5 5 5
C1
4 C2 4 4
C4
3 C6 3 3
Γz

Γz
Γz

2 2 2

1 1 1

0 0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
t t t

FIG. 9: Comparison of the temporal evolution of the midspan circulation for the four pitch
rates due to the LEV for the AR = 2 (left), AR = 4 (center), and 2D (right) flat plate.

The main difference between the flows around AR = 2 and 4 plates is how the LEV
circulation grows. For plate of AR = 2, the LEV is influenced significantly by tip effects,
keeping the roll up of the attenuating streamwise convection and detachment of the LEV to
take place directly above the wing. For the AR = 4 plate, the reduced tip effect allows for
the LEV to lift up and deform more freely, as shown by Figures 6 and 7. This in turn allows
for the increased circulation to be accumulated by the LEV. The first LEV grows large to
critical threshold, allowing for the second LEV to start forming earlier than the case with
AR = 2.

13
V. AERODYNAMIC FORCES

Figure 10 shows the temporal evolution of the aerodynamic lift and drag coefficients for
the AR = 2, 4, and 2D plates undergoing the four pitching motions considered in this study.
For all pitch rates and aspect ratios considered, there is a spike in CL centered around t = 0
due to non-circulatory6 effects from the angular acceleration of the wing. The amplitude
of the peak is related to the smoothing value a in Eq. (1), where larger values of a result
in greater peak values. As the plate continues to pitch, CL begins to increase to a second
maximum value for all pitch rates considered. This second increase in CL is attributed to
suction from the LEV.
As the pitch rate is decreased, we notice a substantial reduction in the slope of the lift
curve between the first three pitch rates (C1, C2, and C4), but a more gradual reduction
between the two slowest cases, C4 and C6. By the end of the pitching motion, there is a
sharp reduction in lift due to the angular deceleration. After the plate has completed its
motion, the CL curves for AR = 2 collapse. For AR = 4, the CL curves for the C1 and
C2 cases collapse t ≥ 4 and then gradually begin to increase again after t = 5 due to the
development of a second LEV. We also observe a similar collapse between the C4 and C6
curves by t = 6.5, although there is no gradual secondary increase in lift due to further LEV
development.

AR = 2 AR = 4 2D
7 7 7
C1 6 6
6
C2
5 C4 5 5

4 C6 4 4
CL

CL
CL

3 3 3

2 2 2

1 1 1

0 0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
t t t
7 7 7

6 6 6

5 5 5

4 4 4
CD

CD

CD

3 3 3

2 2 2

1 1 1

0 0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
t t t

FIG. 10: Aerodynamic force coefficients for various pitch rates at Re = 300 for AR = 2,
AR = 4, and 2D plates.

An interesting observation made between the three aspect ratios is the time at which the
maximum CL is attained for the C4 and C6 cases. For the AR = 2 plate, the maximum
CL is attained at t = 3.5 and t = 5 for the C4 and C6 cases, respectively. For both the
AR = 4 and 2D plates, the maximum lift is achieved at about half a convective time earlier
for C4 case (at t = 3) and nearly two convective units for the C6 case (at t = 4). After

14
this maximum is achieved, there is a decrease in CL for both the AR = 4 and 2D plates,
signifying that the LEV lifts off earlier in time when compared to the AR = 2 plate for the
slower pitching rates. We note that the maximum lift being achieved before the end of the
pitching motion is due to the formation and detachment of the LEV, which provides the
enhanced lift. This lift off of the LEV for the case of AR = 4 is evident in Figures 7 and 8.
In comparison, we have observed that the LEV for the case of AR = 2 stays relatively close
to the plate throughout the time shown in the figures.
The second row of Figure 10 presents the CD histories for the AR = 2, 4 , and 2D plates.
The initial drag value is the steady-state value at zero incidence. As the pitch ramp begins,
drag increases due to the increase in the projected area seen by the flow. At the end of the
motion, there is a negative spike in CD due to the deceleration of the motion. For the C1
cases for the three aspect ratios, this rapid deceleration actually results in a slight thrust.
The CD curves for all pitch rates for the AR = 2 plate collapse to one another after the
respective pitching motion is complete. This is also observed for the AR = 4 plate except
for the C1 and C2 pitching cases after t = 6 due to the slight increase in drag resulting from
the development of the second LEV. For the 2D wing, there is no collapse in the CL or CD
curves due to the subsequent development of additional LEV’s.

AR = 2 AR = 4 2D
3 3 3
C1
2.5 C2 2.5 2.5

2
C4 2 2
C6
C L∗

C L∗

C L∗
1.5 1.5 1.5

1 1 1

0.5 0.5 0.5

0 0 0
0 0.25 0.5 0.75 1 1.25 0 0.25 0.5 0.75 1 1.25 0 0.25 0.5 0.75 1 1.25
t ∗
t ∗
t∗
3 3 3

2.5 2.5 2.5

2 2 2
CD

CD

CD
∗

1.5 1.5 1.5

1 1 1

0.5 0.5 0.5

0 0 0
0 0.25 0.5 0.75 1 1.25 0 0.25 0.5 0.75 1 1.25 0 0.25 0.5 0.75 1 1.25
t∗ t∗ t∗

FIG. 11: Scaled lift and drag coefficients over scaled time for various pitch rates at
Re = 300 for the AR = 2, AR = 4, and 2D plates.

Parameter study of various acceleration rates suggests opportunity for non-dimensional

scaling. During the pitching motion, the total streamwise travel distance and therefore
velocity varies along chord of the plate. At the leading edge, where the pivot point is
located, the plate travels at a speed of U∞ . As the plate is pitching, the velocity of the plate
increases along the chord of the wing from the minimum value of U∞ at the pivot point to
maxt (||U∞ || + ||uTE ||) at the trailing edge. For the present scaling analysis, we include the
trailing-edge speed into the characteristic velocity used in the non-dimensionalization of the

15
lift and drag coefficients. The scaled lift and drag coefficients are given by
Fx Fy
CL∗ = 1 2
∗
CD = 1 2
, (2)
2
ρUchar A 2
ρUchar A

where Uchar = U∞ + uTE, max , and uTE, max = maxt ||uTE || = cα̇ = 2cK is the maximum
velocity of the trailing edge during the pitching motion. The maximum velocity achieved
by the trailing edge occurs during the constant angular velocity (α̇) portion of the pitching
motion. We also scale the time variable with t∗ = t/tp , where tp = αmax /2K is the pitching
interval previously defined for Eq. (1). Figure 11 presents the scaled lift (top) and drag
(bottom) coefficients (CL∗ and CD ∗
, respectively) over the scaled convective time, t∗ . Here,
we observe a good collapse in both scaled lift and drag curves for all pitch rates and aspect
ratios considered. This collapse of data confirms that reduced frequency K is the driving
variable that determines the unsteady force. The definition of the characteristic velocity
used above has resemblance to how Milano and Gharib31 chose the characteristic leading-
edge velocity for flapping plates. We have also considered the use of maxt ||U∞ + uTE ||
as the characteristic velocity for the non-dimensionalization of the forces. We found that
this choice of characteristic velocity does not collapse the force histories well for the various
acceleration rates. This suggest that the generation of the hydrodynamic forces on the plate
is due to the additive effects from the freestream interacting with the leading edge and
the trailing edge undergoing a rotational motion. Therefore, we use the previously defined
characteristic value based on the sum of the norms, U∞ + uTE,max .
Alternatively, the forces can be collapsed by considering the lift-to-drag ratio (L/D) since
such scaling does not require the use of characteristic velocity in the definition. For each
aspect ratio, we present L/D in Figure 12 with the time variable scaled in the same manner,
t∗ = t/tp . At the beginning of the motion, we observe large peaks in the L/D curves due
to the initial angular acceleration. After that initial peak, we observe a collapse in the L/D
curves for the four acceleration rates. For the three aspect ratios considered, the collapse
of the L/D curves is qualitatively similar. As the plate increases its angle of attack from
α = 0◦ to 45◦ , the L/D curves decrease resulting from the increased pressure drag due to
the highly separated wake.

AR = 2 AR = 4 2D
25 25 25
C1
C2
20 C4 20 20
C6
15 15 15
L/D

L/D
L/D

10 10 10

5 5 5

0 0 0
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
t∗ t ∗
t∗

FIG. 12: Lift-to-drag ratios over scaled time for various pitch rates at Re = 300 for the
AR = 2, AR = 4, and 2D plates.

16
 AR = 2 AR = 4 2D
7 7 7

6 6 6

5 C1 5 5

4 4 4
CL

CL

CL
3 3 3
C6
2 2 2

1 1 1

0 0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
t t t
7 7 7

6 6 6

5 5 5

4 4 4
CD

CD

CD
3 3 3

2 2 2

1 1 1

0 0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
t t t

FIG. 13: Comparison between the experimental (Re = 20, 000, dashed) and present
simulation (Re = 300, solid) force coefficients for the C1 and C6 pitch rates with AR = 2,
4, and 2D plates.

VI. EFFECTS OF REYNOLDS NUMBER

We examine the aerodynamic force coefficients for the AR = 2, 4, and 2D pitching plates
from the simulation at Re = 300 and the experiments at Re = 20,000 in Figure 13. The
first row compares the CL curves for AR = 2, AR = 4, and 2D plates pitching from 0
to 45◦ about the leading edge for the C1 and C6 cases. We note that the 2D simulations
are performed in only in the xy-plane, whereas the 2D experiments are conducted with a
flat plate that extends wall to wall within the water tunnel (≈ 1mm tip gap). For the C1
cases, a large non-circulatory spike in lift is present centered around t = 0 resulting from
the fast acceleration of the plate. This non-circulatory spike is less evident for the C6 cases
due to the reduced a value which smooths the motion profile (Eq. (1)), and due to lower
acceleration at lower pitch rate.
As the plate continues to increase its angle of attack, CL increases due to the development
of a leading-edge (dynamic stall) vortex over the plate surface. After the pitching motion is
complete (α = αmax = 45◦ has been attained), we observe a reduction in CL over time for
all plates. It is interesting to note the qualitative comparison between the long-term force
histories for the AR = 4 C1 case, where an increase in CL is observed for experimental and
computational results around t = 7 due to the development of a second LEV. We observe
similar comparisons between the CD curves, with the main difference being the initial drag
value before the plate begins to pitch, which is expected due to increased viscous effects at
Re = 300. Note that the lift-to-drag ratio is close to unity, signifying that the resultant
aerodynamic force is primarily normal to the plates surface. Although there are two orders
of magnitude of difference in Reynolds number between simulations and experiments, we

17
observe reasonable agreement in the forces for all aspect ratios considered. Differences in the
aerodynamic forces between the two results should be expected due to the noticeable viscous
losses present in the low Reynolds number simulations. The enhanced diffusion of vorticity
can account for the reduced lift slopes and maximum force values. While the magnitude
of the force coefficients deviate somewhat between the two, we observe similar behavior of
when maxima in CL and CD are achieved, and also similar characteristics even after the
pitching motion ends.
Figure 14 presents a comparison between the flow field around the AR = 4 plate for
the C1 case. The experimental fluorescent dye flow visualization (left column) was obtained
along the 3/4, 7/8, and tip spanwise locations and is compared to the magnitude of vorticity
from the simulation at the same locations (right column). The vorticity magnitude contours
(3 ≤ kωk ≤ 15) were chosen to match the experimental dye injection photos. The results
are presented in a plate-fixed reference frame at α = 23◦ , 32◦ , and 45◦ corresponding to
t = 0.5, 0.75, and 1, respectively. As the angle of attack reaches 23◦ , we observe similar
features between the experiments and computations, namely the shedding of the starting
vortex from the trailing-edge, the development of the tip vortices, and the initial roll up of
the vortex sheet from the leading edge forming the LEV. The obvious difference between
the two flow fields is the absence of the Kelvin–Helmholtz instability in the shear layer at
the trailing edge at the Reynolds number of the computation.

45

U1
↵ = 23
30
α

15

0
0 0.25 0.5 0.75 1
U1 t
↵ = 38

3/4 7/8 tip

midspan

U1 ↵ = 45

FIG. 14: Comparison between experimental fluorescent dye flow visualizations at

Re = 20, 000 (left) and computed contours of vorticity magnitude (3 ≤ kωk ≤ 15) slices
from the present simulations at Re = 300 (right), taken along the 3/4, 7/8, and tip span
locations for the AR = 4 plate pitching from 0 to 45◦ over one chord of travel (C1).

Clear differences between the development of the LEV between the experiment and com-
putation are observed as the plate continues to increase its angle of attack. The streamwise

18
elongation of the LEV core for the low Reynolds number is observed for DNS, whereas for
the LEV is more compact towards the leading edge of the plate for higher Reynolds number
experiment. We believe this to be the reason between the difference in the CL curves given in
Figure 13, since the compact LEV has a stronger suction associated with it compared to the
elongated LEV, which creates a more profound suction on the top surface of the wing, thus
leading to a higher CL . Although the differences in Reynolds number between the two results
is of two orders of magnitude and there is a lack of any instabilities in the present numerical
results, there is a good agreement in the qualitative behavior of the wake vortices which
leads us to believe that the present simulations are valid for the vortex dynamics around
low-aspect-ratio pitching plates, especially in the developmental stage of vortex formation.

VII. CONCLUDING REMARKS

Three-dimensional direct numerical simulations via the immersed boundary projection

method have been performed to examine the vortex dynamics for pitching flat plates of
aspect ratio 2, 4 and 2D, and compared with experiment in a water tunnel, at a range of
pitch rates that span across the vortex formation time. While the Reynolds number is 300
in the computation and 20,000 in the experiments, the results were in good agreement in
terms of the wake structures and lift and drag histories. The differences between the two
studies were the appearance of Kelvin–Helmholtz instabilities and the compactness observed
for the LEV core found in the experiments, which are both attributed to Reynolds number
effects.
During the early phase of pitching, the tip vortices do not affect the flow field near
the midspan significantly. However, as the tip vortices grow, tip effects influence the wake
dynamics in a global manner. While the AR = 2 plate kept the LEV to remain in the vicinity
of the wing, the reduced tip effects from the AR = 4 plate allowed the LEV to detach earlier
and led to the subsequent formation of the second LEV. The first LEV detachment caused
the tip vortices to deform and interact strongly with the LEV structure that resembles the
arch-type vortex. The detaching LEV creates a peak in the lift force before the end of motion
for the slower pitching cases. For both the AR = 2 and 4 plates, lift and drag histories for the
examined pitch rates evinced good collapse to a common curve, if coefficients are normalized
by the running speed of the trailing edge, instead of the free-stream relative speed.

ACKNOWLEDGMENTS

RJ and KT were supported by the 2012 USAF Air Vehicles Directorate Summer Re-
search and Development Program and the 2013 ASEE Summer Faculty Fellowship Program
during their stays at the Wright–Patterson Air Force Base. RJ also acknowledges the Aero-
Propulsion, Mechatronics, and Energy Fellowship from the Florida State University.

REFERENCES
1
T. J. Mueller, ed., Fixed and flapping wing aerodynamics for micro air vehicle applications
(AIAA, 2001).
2
D. J. Pines and F. Bohorquez, “Challenges facing future micro-air-vehicle development,”
J. Aircraft 43, 290–305 (2006).

19
3
S. Watkins, J. Milbank, B. J. Loxton, and W. H. Melbourne, “Atmospheric winds and
their implications for microair vehicles,” AIAA J. 44, 2591–2600 (2006).
4
K. Granlund, M. Ol, B. Monnier, and D. Williams, “Airfoil longitudinal gust response
in separated vs. attached flows,” (42nd AIAA Fluid Dynamics Conference and Exhibit
(AIAA2012-2695), 2012).
5
G. Leishman, Principles of helicopter aerodynamics, 2nd ed. (Cambridge Univ. Press,
2006).
6
T. Theodorsen, “General theory of aerodynamic instability and the mechanism of flutter,”
Tech. Rep. 496 (NACA, 1935).
7
K. Taira and T. Colonius, “Three-dimensional flows around low-aspect-ratio flat-plate
wings at low Reynolds numbers,” J. Fluid Mech. 623, 187–207 (2009).
8
J. H. J. Buchholz and A. J. Smits, “On the evolution of the wake structure produced by
a low aspect ratio pitching panel,” J. Fluid Mech. 546, 433–443 (2006).
9
M. A. Green, C. W. Rowley, and A. J. Smits, “The unsteady three-dimensional wake
produced by a trapezoidal panel.” J. Fluid Mech. 685, 117–145 (2011).
10
D. J. Garmann, M. R. Visbal, and P. D. Orkwis, “Three-dimensional flow structure and
aerodynamic loading on a revolving wing.” Phys. Fluids 25, 034101–27 (2013).
11
A. Jones and H. Babinsky, “Unsteady lift generation on rotating wings at low Reynolds
numbers.” J. Aircaft 47, 1013–1021 (2010).
12
K. K. Chen, T. Colonius, and K. Taira, “The leading-edge vortex and quasisteady vortex
shedding on an accelerating plate,” Phys. Fluids 22, 033601–11 (2010).
13
C. W. P. Ford, R. Stevens, and H. Babinsky, “Flexible leading edge flap on an impulsively
started flat plate at low Reynolds numbers.” (50th AIAA Aerospace Sciences Meeting
(AIAA2012-2840), 2012).
14
R. Stevens, C. W. P. Ford, and H. Babinsky, “Experimental studies of an accelerating,
pitching, flat plate at low Reynolds numbers.” (51st AIAA Aerospace Sciences Meeting
(2013-0677), 2013).
15
K. O. Granlund, M. V. Ol, and L. P. Bernal, “Unsteady pitching flat plates,” J. Fluid
Mech. 733, R5.1–13 (2013).
16
H. Yu and L. P. Bernal, “Effect of pivot point on aerodynamic force and vortical structure
of pitching flat plate wings,” (51st AIAA Aerospace Sciences Meeting (AIAA 2013-0792),
2013).
17
K. Granlund, M. Ol, K. Taira, and R. Jantzen, “Parameter studies on rotational and trans-
lational accelerations of flat plates,” (51st AIAA Aerospace Sciences Meeting (AIAA2013-
0068), 2013).
18
M. R. Visbal, “Flow structure and unsteady loading over a pitching and perching low-
aspect-ratio wing,” (42nd AIAA Fluid Dynamics Conference and Exhibit (AIAA 2012-
3279), 2012).
19
D. J. Garmann and M. R. Visbal, “Numerical investigation of transitional flow over a
rapidly pitching plate,” Phys. Fluids 23, 094106 (2011).
20
R. Jantzen, K. Taira, K. Granlund, and M. Ol, “On the influence of pitching and ac-
celeration on vortex dynamics around a low-aspect-ratio rectangular wing,” (51st AIAA
Aerospace Sciences Meeting (AIAA2013-0833), 2013).
21
M. Gharib, E. Rambod, and K. Shariff, “A universal time scale for vortex ring formation,”
J. Fluid Mech. 360, 121–140 (1998).
22
A. C. DeVoria and M. J. Ringuette, “Vortex formation and saturation for low-aspect-ratio
rotating flat-plate fins,” Exp. Fluids 52, 441–462 (2012).

20
23
J. D. Eldredge, C. Wang, and M. V. Ol, “A computational study of a canonical pitch-up,
pitch-down wing maneuver,” (39th AIAA Fluid Dynamics Conference (AIAA2009-3687),
2009).
24
T. O. Yilmaz and D. Rockwell, “Flow structure on finite-span wings due to pitch-up
motion,” J. Fluid Mech. , 518–545 (2012).
25
K. Taira and T. Colonius, “The immersed boundary method: a projection approach,” J.
Comput. Phys. 225, 2118–2137 (2007).
26
T. Colonius and K. Taira, “A fast immersed boundary method using a nullspace approach
and multi-domain far-field boundary conditions,” Comput. Methods Appl. Mech. Engrg.
197, 2131–2146 (2008).
27
M. V. Ol, L. Bernal, C.-K. Kang, and W. Shyy, “Shallow and deep stall for flapping low
Reynolds number airfoils,” Exp. Fluids 46, 883–901 (2009).
28
L. Graftieaux, M. Michard, and N. Grosjean, “Combining PIV, POD and vortex identi-
fication algorithms for the study of unsteady turbulent swirling flows,” Meas. Sci. Tech.
12, 1422–1429 (2001).
29
A. R. Jones and H. Babinsky, “Reynolds number effects on leading edge vortex development
on a waving wing,” Exp. Fluids 51, 197–210 (2011).
30
Y. S. Baik, L. P. Bernal, K. Granlund, and M. V. Ol, “Unsteady force generation and
vortex dynamics of pitching and plunging aerofoils,” J. Fluid Mech. 709, 37–68 (2012).
31
M. Milano and M. Gharib, “Uncovering the physics of flapping flat plates with artificial
evolution,” J. Fluid Mech. 534, 403–409 (2005).

21