JID: HFF

ARTICLE IN PRESS [m5G;January 23, 2017;21:24]

International Journal of Heat and Fluid Flow 0 0 0 (2017) 1–14

Contents lists available at ScienceDirect

International Journal of Heat and Fluid Flow

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

Computational study of a pitching bio-inspired corrugated airfoil

T.J. Flint a, M.C. Jermy a, T.H. New b, W.H. Ho c,∗
a
Department of Mechanical Engineering, University of Canterbury, New Zealand
b
School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore
c
Department of Mechanical and Industrial Engineering, University of South Africa, South Africa

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

Article history: The phenomenon of insect flight has been of scientific interest for many years and is more recently in-
Received 21 February 2016 spiring modern engineering devices such as Micro Aerial Vehicles (MAVs). Insect flight is characterized
Revised 22 November 2016
by unsteady fluid dynamics at low Reynolds numbers. The importance of viscous effects to the successful
Accepted 28 December 2016
flapping flight of insects has been identified and with the current state of computing and Computational
Available online xxx
Fluid Dynamics (CFD) these effects can now be studied in detail. The present work attempts to simplify
Keywords: this complex phenomenon by considering symmetric oscillating rotational motion of a wing (pitching).
CFD What is of interest in this study is how the shape of a corrugated idealized insect wing affects the per-
PIV formance and flow characteristics around the pitching wing. Two dimensional CFD on an oscillating wing
Oscillating airfoil has been performed and reported. Measurements were taken to ensure the accuracy of the computational
Pitching airfoil solution and the results validated against experimental PIV results. A range of frequencies and rotational
Bio-inspired airfoil
amplitudes have been investigated. Lift and drag coefficients have been analyzed for all cases to quantify
Biomimicry
the effects of unsteady flow features on the performance of the oscillating wing. It was found that the
wing shape used in this study resulted in the viscous features formed on the top of the wing exhibiting
high sensitivity to the oscillating conditions and these influenced the performance of the wing. The flow
features formed on the bottom of the wing remained similar throughout the cases tested. In the pitch-
ing regime this wing profile did not perform as well as published results for smooth airfoils in terms of
thrust and propulsion efficiency. However this may be due to reduced frequency effects becoming impor-
tant at our high pitching amplitude which need to be investigated further. There may be other oscillatory
regimes that more accurately represent flapping flight in which the corrugated foil outperforms a smooth
counterpart but these are yet to be investigated. Further research in this area may help answer the ques-
tion as to how evolutionarily significant other benefits of a corrugated wing, such as being light and
strong, are compared to its aerodynamic properties, the present results seem to favor the former.
© 2016 Elsevier Inc. All rights reserved.

1. Introduction to 104 (Kesel, 20 0 0; Rüppell, 1989) meaning that viscous effects

Adapting solutions engineered by nature to improve techno-
logically advanced products is a familiar and increasingly popular
practice by engineers and researchers to come up with implemen-
tations that are simple and robust (Nawroth et al., 2012). Dragon-
flies are incredibly agile fliers, able to dip and dart in all directions.
It was shown that the lift coefficients required for the flapping
flight of dragonflies far exceeded the steady state values measured
from non-flapping dragonfly wings (Norberg 1975; Wakeling and
Ellington 1997a,c). This suggested that unsteady aerodynamic ef-
fects, among others, were significant to the generation of lift. Drag-
onfly wings operate at Reynolds numbers ranging from about 102
∗ studies of corrugated bio-inspired wing sections in steady flow
Corresponding author.
E-mail addresses: howh@unisa.ac.za, weihua0202@gmail.com (W.H. Ho).
such as the formation and subsequent shedding of vortices become
important to the flow structures and forces acting on the wings.
Low Reynolds number flows are often complicated since separa-
tion, transition and subsequent reattachment occur within a short
distance and are dominated by large scale vortex motions (Lin and
Pauley, 1996).
Unlike smooth airfoils seen on aeroplanes and birds, dragon-
flies (and most flying insects) have corrugated wings (Kesel, 20 0 0;
Newman, et al., 1977; Rees, 1975). The corrugations of the wings
provide strength benefits with minimal weight penalties (Rees,
1975; Lian et al., 2014). Aerodynamic and biological studies have
demonstrated that such superior flight may be due not only to the
flapping motion but the cross-sectional profiles of the wings also.
The latter being supported by recent experimental and numerical

http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
0142-727X/© 2016 Elsevier Inc. All rights reserved.

Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
JID: HFF
ARTICLE IN PRESS [m5G;January 23, 2017;21:24]

2 T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14

thors list a number of theories that suggest the source of lift gen-
Notation eration including the effects of leading-edge vortices (LEV), cap-
turing of wake and pitching-up rotation. Dickson and Gotz (1993)
A amplitude of oscillation [˚] first showed that lift was enhanced by the presence of LEVs. Fur-
C courant number [dimensionless] (defined in Eq. (3)) ther studies by Liu and Kawaguchi (1998) and van den Berg and
Cd sectional drag coefficient [dimensionless] (defined Ellington (1997) confirmed the status of the leading-edge vortex.
in Eq. (7)) Although there has been much previous work detailing the flow
Cip power coefficient [dimensionless] (defined in field near to oscillating airfoils (Lu et al., 2013; Zhou et al., 2013;
Eq. (9)) Panda and Zaman, 1992; Lee, 2011), these have been confined to
Cl sectional lift coefficient [dimensionless] (defined in smooth conventional NACA type airfoils or flat plates. There are
Eq. (4)) few studies of flow characteristics about oscillating corrugated bio-
Cp pressure coefficient [dimensionless] (defined in inspired airfoils (Saharon and Luttges, 1987; Okamoto and Azuma,
Eq. (6)) 2005) and to date the authors are unaware of any other work that
CT thrust coefficient [dimensionless] (−1∗ Cd ) makes use of PIV or computational methods (CFD) to visualise the
c airfoil chord length [m] flow. While work has been performed for curved deforming wings
l relative difference in lift coefficient [%] (defined in (Molki and Sattari, 2013) reviews of Micro Aerial Vehicle (MAV)
Eq. (5)) development have neglected to consider the effect that corruga-
l mean relative difference in lift coefficient [%] tions have on the flow field (Pines and Bohorquez, 2006). There
Fd drag force per meter span [N/m] has been a significant research effort to understand the unsteady
Fl lift force per meter span [N/m] effect around moving wings for MAV applications but none includ-
f frequency of oscillation [Hz] ing corrugations or bio-inspired cross-sections (Shyy et al., 2008).
h wake width [m] The present study intends to expand the scope of knowledge in
k reduced frequency [dimensionless] (defined in this field by studying the aerodynamics of a pitching corrugated
Eq. (2)) airfoil which is an idealised geometry inspired by dragonfly wings.
M pitching moment per unit span [N] The prevalence of corrugated wings on dragonflies and other in-
η propulsion efficiency [dimensionless] (defined in sects indicates some form of evolutionary advantage, it is not cer-
Eq. (10)) tain whether this is solely structural or if unsteady aerodynam-
P pressure [Pa] ics are improved also. Either way, such a wing is well suited for
P∞ far-field pressure [Pa] technology such as MAVs (Pines and Bohorquez, 2006) as proven
ρ fluid density [kg/m3 ] by nature’s own MAVs, insects. Our aim is to provide insight into
Re Reynolds number [dimensionless] (defined in the characteristics of viscous flows forming around oscillating air-
Eq. (1)) foil and how corrugations similar to those seen on dragonfly wings
St Strouhal number [dimensionless] (defined in affect their unsteady aerodynamic properties. The geometry is sim-
Eq. (8)) ilar to the one used in previous, steady, investigations (Levy and
T period [s] Seifert, 2009; New et al., 2014).
t time [s]
t simulation time-step [s] 2. Methodology
θ˙ wing rotation rate [rad/s]
U free stream velocity [m/s] 2.1. CFD methodology
ν kinematic viscosity [m2 /s]
x distance along the chord from the leading edge of A rigid two-dimensional airfoil was modeled using ANSYS Flu-
the airfoil [m] ent version 14.5 to simulate the flow-field and calculate the lift
x i height of the first inflation layer on airfoil surface and drag on a bio-inspired corrugated airfoil undergoing two di-
[m] mensional oscillations subject to a constant horizontal flow at a
y+ distance from the wall normalised by the viscous chord Reynolds number (Re) of 14,0 0 0 as defined by:
length scale [dimensionless] Uc
Re = (1)
ν
where U, c, and ν are the free stream velocity, airfoil chord length,
(Hu and Tamai, 2008; Murphy and Hu, 2010; Levy and Seifert,
and the fluid kinematic viscosity respectively. The range of oscilla-
2009; New et al., 2014). The evidence shows that the corrugated
tions used in the calculations had reduced frequencies, k (Eq. (2)),
wing sections delay stall and results in smaller separation regions
between 1.24 – 4.96 and angular amplitude between 5° - 20°
at high angle of attack compared to smooth counterparts resulting
in superior lift to drag characteristics in some regimes. It has been π fc
k= (2)
suggested that the corrugations provoke an early transition to tur- U
bulent flow resulting in the reattachment of flow and thus reduced where f is the frequency of oscillation. The center of rotation of
separation (Newman et al., 1977). Murphy and Hu (2010) con- the airfoil was located at quarter-chord position in the middle of
ducted PIV experiments and concluded that the corrugations help the plate thickness. The geometry used is shown in Fig. 1(a) and
to overcome adverse pressure gradient and discourage large-scale corresponds to the “Corrugated B” section used in the experiments
separation and airfoil stall by first tripping the boundary layer to of New et al. (2014). Locations A-F are also defined here to aid dis-
promote transition from laminar to turbulent and then “trapping” cussion of the results. This particular wing geometry is described
the unsteady vortex structures by pulling high-speed flow into in detail by Levy and Seifert (2009), where full details of the ge-
the near wall region. Levy and Seifert (2009) also reached similar ometry can be found there. The only modifications made to the
conclusions when studying a different profiled corrugated airfoil. geometry is an increased thickness which was a requirement for
Flapping wing research has focused on furthering understand- the experimental portion of the study. This general idealised shape
ing of the unsteady aerodynamic mechanisms that result from was first used by Newman et al. (1977) for the wings of model
wing movements (Viieru et al., 2006). In the same paper, the au- gliders that were flown to assess the performance of such a wing.

Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
JID: HFF
ARTICLE IN PRESS
T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14
a b
A BC D E
2/75 c
c/4
c
Fig. 1. Geometry of airfoil and mesh with (a) airfoil dimension showing location of pivot point and locations A-F to aid discussion of the results and (b) overall dimensions
in entire flow domain. Flow from left to right.
Newman created this shape from the section of an Aeschna Inter-
rupta dragonfly specimen. The wide use of this geometry in the
literature made it an attractive choice for this study.
The flow domain consists of an inner smaller circular domain
containing the airfoil which rotates inside a stationary outer rect-
angular domain. The rotation of the inner domain was achieved
with the sliding mesh method in ANSYS Fluent. The sliding mesh
method eliminates the need for re-meshing or mesh deformation
during the rotation and in the process reduces computational ex-
penses. Fig. 1(b) shows the main dimensions associated with the
overall flow domain.
The boundary conditions used consist of a uniform velocity in-
let on the upstream side of the domain, a zero pressure outlet
downstream, symmetry conditions on the top and bottom of the
domain, a matching interface between the inner and outer fluid
domains and a no slip wall condition on the airfoil surface. The
matching interface allows the interpolation and transfer of data be-
tween cells on either side of a non-conformal interface. By allow-
ing data transfer between separate inner and outer fluid zones, the
inner zone can be allowed to rotate inside the outer zone while
maintaining fluid flow across the boundary.
A quad based mesh was used with inflation layers on the air-
foil and a zone of finer cells directly behind the airfoil as seen in
Fig. 1. The mesh size was equal on both sides of the matching in-
terface. A lower cell growth rate was used on the inner domain as
to allow a finer mesh in the vicinity of the airfoil. Inflation layers
were used around the airfoil to capture boundary layer effects. The
first inflation layer had a y+ ranging from 0.028 to 4.7 calculated
locally based on the smallest cell dimension.
A coupled pressure based solver was used to solve the flow
around the airfoil. Unsteady flow structures around the wing were
of interest to this study so a turbulence model needed to be se-
lected that could resolve large turbulent structures in time. The
Scale-Adaptive Simulation (SAS) turbulence modelling method was
selected because it can provide detail of unsteady flow structures
well. The SAS method uses a modified Unsteady Reynolds Aver-
aged Navier-Stokes (URANS) approach which modifies how well it
resolves fluctuating quantities based on the local flow conditions
(Menter, 2012; Menter and Egorov, 2010). This provides well re-
solved fluctuating quantities in areas where the flow is most un-
steady while reverting to a k-w SST URANS approach in steady re-
Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
[m5G;January 23, 2017;21:24]
3
7c
F 3c
6c
14/3 c
gions. The PRESTO! Scheme (Patankar, 1980) was used for the pres-
sure discretization as it performs well in swirling flows. A second
order upwind discretization scheme was used for all other vari-
ables. Gradients were evaluated using the least squares cell based
method. Higher order term relaxation, double precision solver and
a bounded second order implicit transient formulation was used
for accuracy and stability. The root mean squared (rms) residual
convergence criterion for each time step was set to 10−5 . The in-
coming turbulence intensity was set to 1.1% as recorded in experi-
ment (New et al., 2014). The time dependent solution was initial-
ized from a steady state solution run as pseudo transient with the
SST turbulence model and settings equivalent to that used in the
time dependent simulation. A user defined function was used to
set the oscillatory rotational motion of the internal fluid domain
and hence the airfoil. The time dependent solution was solved
with a Courant number (C) of one relative to the height of the first
inflation layer as defined by:
U t
C= (3)
x i
where t is the simulation time-step and xi is the height of the
first inflation layer. The fluid was assumed incompressible as rele-
vant to water tunnel PIV experimental results and low Mach num-
ber flight in air.
2.2. Mesh convergence
Because of the complex nature of this problem, a mesh conver-
gence study was performed in multiple stages and the effects of
the courant number were tested also. Three different studies were
performed, they tested how the mesh resolution, courant number,
and local mesh refinement effected the solution of the simulation.
By reaching independence with each study before moving on to
the next study the final simulation will be independent to all
the metrics tested. The studies were performed with the case of
k = 4.96 and A = 20◦ . This was the largest oscillating frequency
and amplitude used in this study hence providing the most vig-
orous motion to be encountered. A mesh convergence study of
this case was presumed to be applicable to all the other scenarios.
The metric used to compare meshes was the unsteady section lift
JID: HFF
4 T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14
a b
35
30
Sectional Lift Coefficient, Cl
25
20
15
10
5
0 Elements
-5 6.7
-10
-15
-20
-25
-30
-35
Fig. 2. Initial global mesh convergence in (a) graphical and (b) numerical form for C = 2.5.
coefficient, Cl (Eq. (4)), generated by the airfoil throughout the
oscillations.
Fl
Cl =
1
2
ρU 2 c
where L is the lift force per meter span on the wing and ρ is the
fluid density. Relative differences in unsteady lift data between
meshes are presented as a percentage of the maximum magnitude
observed during a typical cycle and will be denoted by l .
C (t ) − Cl (2) (t )
l (t ) =    l (1 )     × 100%
max Cl (1) (t ); 0 ≤ t ≤ T ∪ Cl (2) (t ); 0 ≤ t ≤ T
where the subscripts (1) and (2) denote the meshes to be com-
pared, t is time, and T is the period of oscillation of the wing.
The root mean square (rms) of the error between meshes, rms(l ),
was chosen as the metric used to compare meshes. This gives
an idea of how well two lift coefficient signals compare over one
oscillation cycle.
2.2.1. Global mesh resolution
A number of meshes were refined while keeping the first infla-
tion layer height and t constant. The first inflation layer height
was selected to be the smallest possible for convergence of the
steady initialization case. These cases were all run with a Courant
number of 2.5.
Looking at Fig. 2(a) visually, the lift curves collapse onto a com-
mon curve in the medium and fine case with the exception of a
small region at t/T = 7.1. Fig. 2(b) indicates that the rms error be-
tween the coarse mesh and the medium mesh is double that of
the rms error between the medium mesh and the fine mesh which
implies that the solutions are converging. This problem must be
run for a long period of time (∼8T) to ensure that the simulation
reaches a steady state and from a computational standpoint the
finest mesh was prohibitively expensive with the resources that
were available. With a good visual comparison the medium mesh
was deemed sufficient.
2.2.2. Courant number
Because an implicit scheme was used, the requirement of hav-
ing C ≤ 1 for a stable time dependent solution was not strict and
a Courant number of 2.5 could be used. The dependence on C was
next investigated using the Medium mesh as the baseline case.
One simulation was run with the medium mesh using a decreased
t to bring C down to one (‘Small dt’) and a second case was run
with the same t used for the Medium mesh but with a larger
first inflation layer height to bring C down to one (‘Large Infla-
tion’). The results can be seen in Fig. 3. With the only difference
Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
ARTICLE IN PRESS [m5G;January 23, 2017;21:24]
Mesh Number of [%]
6.9 7.1 7.3 7.5 7.7 Coarse 38317 -
Medium 52245 14
Fine 72584 7
t/T
Coarse Medium Fine
being t, the difference between the ‘Medium’ and the ‘Small dt’
cases indicates that, while the solution converged with C = 2.5 it
is in fact dependent on the Courant number. Looking at the two
(4) cases run with C = 1, they appear to qualitatively agree well and
have an rms error of less than 10%. The ‘Large Inflation’ case was
chosen to continue because it gave a similar solution to the ‘Small
dt’ case while saving computational effort. It is not expected that
decreasing the Courant number further below 1 will have any sig-
nificant effect on the solution.
(5)
2.2.3. Local mesh refinement
After preliminary comparison of the simulated results with the
experimental results it was determined that more resolution was
required in the region behind the airfoil to better resolve the com-
plex vortex pattern shed from the trailing edge of the airfoil. A
further spatial mesh convergence was conducted by refining a cir-
cular zone of cells behind the airfoil as can be seen in the image
of the final mesh shown in Fig. 1. As seen in Fig. 4, the ‘Large In-
flation Refined’ mesh showed a relatively large difference from the
original ‘Large Inflation’ mesh indicating that improving resolution
behind the airfoil was important to the accuracy of the simulation.
The ‘Small dt’ case was then modified to have a similar region of
refinement behind the airfoil providing a comparison case while
retaining C = 1. The ‘Small dt Refined’ case agreed well with the
‘Large Inflation Refined’ case. The ‘Large Inflation Refined’ case was
used for the final simulations.
2.3. Validation against experimental data
The computational data was validated against a PIV experiment.
Before presenting the validation data, a brief mention of the ex-
perimental setup and methodology will be presented here. Further
details can be found in New et al. (2014). An airfoil with geometry
similar to that shown in Fig. 1(a) but with a physical chord length
of 75 mm was studied in a re-circulating water tunnel using par-
ticle streak photography and two-dimensional particle-image ve-
locimetry (PIV) techniques. The Reynolds number (based on chord
length) used in the experiments was the same as that used in
the numerical simulations here (Re = 14,0 0 0). The test airfoil has
a span of 300 mm and thus an aspect ratio of four, which was suf-
ficiently large to minimize any significant 3D spanwise flow effects.
During the experiments, it was located vertically between two hor-
izontal flat plates with rounded leading-edges fully-immersed in
the open-channel type water-tunnel, with an MDrive 34AC Hy-
brid high-torque stepper motor system driving its sinusoidal os-
cillations. The use of the flat-plates was to mitigate end-effects, as
JID: HFF
ARTICLE IN PRESS [m5G;January 23, 2017;21:24]

T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14 5

a b
35
30

Sectional Lift Coefficient, Cl

25
20
15
10
5 Mesh Number of [%]
0 Elements
-5 6.7 6.9 7.1 7.3 7.5 7.7 Medium 52245 -
-10 Small dt 52245 14
-15 Large 18937 7
Inflation
-20
-25
-30
-35
t/T
Medium Large Inflation Small dt
Fig. 3. Courant number dependence, ‘Large Inflation’ and ‘Small dt’ have C = 1 while ‘Medium’ has C = 2.5 in (a) graphical and (b) numerical form.

a b
35
30
Sectional Lift Coefficient, Cl

25
20
15
10
5 Mesh Number of [%]
Elements
0 Large Inflation 18937 -
-5 6.7 6.9 7.1 7.3 7.5 7.7 Large Inflation 49723 19
-10 Refined
-15 Small dt 135134 8
-20 Refined
-25
-30
-35
t/T
Large Inflation Large Inflation Refined Small dt Refined
Fig. 4. Further spatial mesh convergence in (a) graphical and (b) numerical form for C = 1.

well as to ensure a more uniform free-stream distribution along tex being shed from the corrugations and travelling along the wing
the entire span of the airfoil. with moderate agreement. This vortex looks irregular and occa-
For the PIV experiments, 20 μm diameter, 1.03 gm/cm3 den- sionally broken up in the PIV data. The traveling vortex is relatively
sity polyamide seeding particles dispersed uniformly in the wa- weak and the experimental data may not have sufficient resolution
ter tunnel were illuminated by a double-pulsed Nd: YAG laser (lo- or sensitivity to pick up the vortex completely, however its loca-
cated above the “top surface” of the airfoil resulting in an op- tion agrees well with the CFD results. The vortices shed from the
tical shadow “beneath” the airfoil and hence no data) and their trailing edge of the airfoil (TEV) are present in both the CFD and
reflected light was captured by a 1600px × 1200px double-frame experimental data with the CFD data capturing the thin stretched
CCD camera at 15 Hz. Image-pairs were captured for about two section of the vortex while the experimental data tended to only
complete oscillation cycles before they were transferred to a work- capture the main vortex region. A thin stretched vortex region is
station for post-processing. Multi-pass, multi-grid cross-correlation physical and expected as the vortex is shed from the trailing edge
based interrogations of the image-pairs were performed, with a giving an indication of the resolution and sensitivity limitation of
50% overlap in both directions and a final interrogation window the experimental data. The positions of TEVs are well matched
size of 32px × 32px. Global and local vector validations were per- by the CFD results, slight differences in locations or orientations
formed to remove spurious vectors, before they were replaced by are attributed to small time differences between when each image
substitute vectors determined via a 3-point by 3-point neighbour- was captured and small errors in the location of the airfoil in the
hood scheme to arrive at the final velocity vector maps. experiment.
Vorticity contour plots from the CFD simulations and the exper-
imental PIV measurements for one period of the k = 2.48, A = 10◦
case are shown in Fig. 5. The experimental results do not show 3. Results and discussion
any flow features below the airfoil because of the optical shadow
due to the experimental set up. The region of negative vorticity Before presenting and discussing the results, some important
around the leading edge and airfoil corrugations compares very locations, A-F, on the airfoil were defined in Fig. 1(a) to aid subse-
well throughout the pitching cycle. Both datasets capture the vor- quent discussion, these will be referred to throughout this section.

Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
JID: HFF
ARTICLE IN PRESS
6 T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14
Fig. 5. Comparison with experimental data. Plots show contours of vorticity for the
scenario k = 2.48, A = 10◦ . CFD for one full oscillation period. Experimental PIV
data is shown in the top half of each frame with CFD results below, the scale used
for both is the same.
A few prominent vortices are shed from the wing and travel
down its surface during each oscillation, modifying surface pres-
sures as they go. Three of these traveling vortices have been fo-
cused on in this analysis to aid comparison of vortex dynamics
and surface pressure variations between cases. They were chosen
because they are present in most cases and they have a noticeable
effect on the wing surface pressures as they travel downstream.
Fig. 6. Contours of vorticity (a) formation of LEV R and vortex under curved section of the wing S on bottom surface of wing; (b) formation of vortex behind corrugations,
D, on top surface of wing Q.
Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
[m5G;January 23, 2017;21:24]
These vortices are labeled Q, R, and S and are shown in Fig. 6. This
is a couple of snapshots of the vorticity contour for the k = 1.24,
A = 10◦ case. There is a large vortex shedding off the peak, at D,
on the top surface of the wing, Q; the Leading Edge Vortex (LEV)
formed and shed off the leading edge on the bottom surface of the
wing, R; and the vortex forming under curved section of the wing,
at F, on the bottom surface of the wing, S. These are the largest
vortices that form near the surface of the wing and often remain
quite close to the wing as they travel downstream. The formation
and shedding of these vortices have the most notable effect on the
pressure distributions on the top and bottom surface of the airfoil
and will be referred to in future discussion.
3.1. Interpretation of time dependent pressure contours
Qualitative results regarding the formation, traveling and shed-
ding of vortices during the oscillating motion of the wing through-
out the range of motion and across all scenarios are difficult to
show without copious snapshots at consecutive time steps. The re-
sults have been presented here as time dependent pressure con-
tours (Fig. 9), where a combination of contours and vortex transit
lines illustrate the movement of the prominent vortices being con-
vected downstream throughout the oscillation cycle. To become fa-
miliar with the interpretation of the time dependent pressure con-
tours, snapshots of results of the k = 1.24, A = 10◦ scenario are
presented in Fig. 7 and are compared to the time dependent pres-
sure contour of the same scenario (Fig. 8).
Each snapshot (Fig. 7) shows colour contours of pressure with
a black outline of the vorticity contours superimposed for refer-
ence. The vorticity contours are intended only to help the reader
visualise the structure and location of vortices traveling down the
wing. The effect that these vortices have on the surface pressure
as they travel along the wing can be realised in the time depen-
dent pressure contour (Fig. 8). The time dependent pressure con-
tour presents the evolution of the surface pressure distribution on
the wing with time. The three traveling vortices identified earlier
have been labelled on the snapshots to allow their position to be
easily tracked with time (in Fig. 7) and compared to their corre-
sponding transit lines on the time dependent pressure contour plot
(in Fig. 8). The transit lines in Fig. 8 are intended only as visual
guides to highlight the streaks of lower pressure formed by the
traveling vortices as they move down the wing, they have not been
mathematically derived.
Pressure coefficient, Cp , is plotted in all of these results as de-
fined as:
P − P∞
Cp = (6)
1
2
ρU 2
were P is the pressure at the point of interest and P∞ is the pres-
sure in the far field.
JID: HFF
ARTICLE IN PRESS
T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14
a b
e f
i j
m n
q r
Fig. 7. Snapshots of the k = 1.24, A = 10◦ case during one oscillation. Colour contours of pressure are shown with an outline of the vorticity contour superimposed (black
lines). The irregular outlines in the far field are due to near zero vorticity magnitude.
Note that the pressure scale for the snapshots in Fig. 7 has been
truncated to C p = ±1.4 to aid visualisation. As an example of in-
terpretation of the time dependent pressure plots, the R vortex
can be clearly seen in both figures. The R vortex is formed from
a region of separation below the leading edge of the wing. The
flow separation causes a decrease in pressure and lasts from about
t/T = 0.30 to t/T = 0.65. The separation below the leading edge can
be seen in the snapshots at times t/T = 0.30 – 0.65 and in the
time dependent pressure contour as the large dark blue area at
the bottom centre of the middle plot (corresponding to the pres-
sures on the lower surface of the wing) between times around
t/T = 0.30 – 0.65. The low pressure contour extends to location
x/c = 0.2 or point C on the wing which corresponds to the size of
the separation region as seen in the snapshots. From this region of
separation emerges the vortex, R, at about t/T = 0.65 and it sub-
sequently travels along the wing. The shedding and motion of R
down the wing is clearly seen in the snapshots and can be seen in
the time dependent pressure contours as a light blue low pressure
streak projecting upwards and towards the right from the large
Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
[m5G;January 23, 2017;21:24]
7
c d
g h
k l
o p
s t
blue area caused by the separation region. R subsequently passes
the trailing edge of the wing at about t/T = 0.30 and no longer
affects the surface pressure distribution of the wing significantly.
One must remember that the motion of the wing is cyclic and one
full cycle has been shown meaning that the snapshot at t/T = 0.95
precedes t/T = 0.00 and that any vortex path in the time dependent
pressure plots that reaches the right hand edge of the plot should
continue on the left hand edge.
The labels Q1 and Q2 have been used for this example to dis-
tinguish on the pressure snapshots (Fig. 7) the two separate Q vor-
tices that are shed during each oscillation.
Q1∗ has been introduced because the Q1 vortex does not pass
the trailing edge of the wing by the time the Q1 for the next cycle
(Q1∗ ) is formed at t/T = 0.75.
One weakness of the time dependent pressure plot as a visual-
isation tool is that it is difficult to observe events happening near
the trailing edge of the wing which may be under the influence
of flow features that are beyond the trailing edge and don’t di-
rectly affect the surface pressure distribution. For example in the
JID: HFF
ARTICLE IN PRESS [m5G;January 23, 2017;21:24]

8 T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14

Q1 Q1
Q2

(a) Top Surface

(b) Bottom Surface

(c) Wing Position

Fig. 8. Pressure distributions on the surface of the wing throughout one full oscillation for the k = 1.24, A = 10◦ case showing the location of the Q, R, and S vortices. A
reference line, U, indicates the free-stream velocity.

k = 1.24, A = 10◦ case shown in Fig. 7 the vortex Q1 reaches the
trailing edge of the wing but does not immediately enter the wake.
Instead the shedding of a trailing edge vortex of the opposite sense
during the clockwise rotation of the wing prevents Q1 from mov-
ing past the trailing edge until it is subsequently absorbed by Q2 at
t/T = 0.80 and can continue past the trailing edge when the clock-
wise rotation slows. It is very difficult to observe this in the time
dependent pressure plot. This feature is not included in the path
lines that have been superimposed onto the plots resulting in the
discrepancy between when Q1 and Q2 leave the trailing edge as
observed in the snapshots (t/T = 0.95) and when their path lines
reach x/c = 1 (t/T ≈ 0.60 and t/T ≈ 0.85 respectively) on the time
dependent pressure plots. This effect is exaggerated by the reduc-
tion in pressure at the trailing edge of the wing due to the cyclic
shedding of trailing edge vortices.
3.2. Time dependent pressure plots
Now Fig. 9 shows the time dependent pressure plots of all
the scenarios studied in this investigation. One of the immedi-
ate observations the reader can make is that the vortex R always
remains beneath the bottom surface of the airfoil except in the
k = 4.96, A = 20◦ scenario where it shows up on the top surface of
the airfoil. This has implications on the lift and drag experienced
by the airfoil and will be discussed later. Another observation that
can be made is that the time dependent pressure profile on the top
surface of the airfoil shows significant variation between the differ-
ent cases. Whereas the profile of the bottom surface shows simi-
lar features for all the cases apart from relative magnitudes and
relative vortex transit time. One can also observe a change in the
overall appearance of the time dependent pressure contours as k
is increased. The pressure scales for these figures were determined
by the maximum absolute value of surface pressure occurring in
each case hence the images cannot be directly compared. How-
ever, the darkest blue and red areas indicate where on the wing
the largest pressures are occurring and give insight into what flow
features are dominating the pressure on the surface of the wing. It
can be concluded that the k = 1.24 cases are dominated by leading
edge vortices, the k = 2.48 cases are dominated by vortices travel-
ing along the wing, and the k = 4.96 cases are dominated by iner-
tial effects of the rotation of the wing. This section of the results
concentrates on the diverse vortex shedding mechanisms occurring
on the top surface of the wing.
For k = 1.24, the vortices that are formed are relatively weak
due to the low rotation rate of the wing. This can be seen in the
k = 1.24, A = 5◦ case in which the vortex traveling along the top
surface of the wing does not make it to the trailing edge but is
instead absorbed into the region of vorticity formed by the bound-
ary layer and is no longer distinguishable as an individual vortex.
At the higher amplitudes vortices are more readily formed. On the
top surface two vortices are shed from the corrugated region ev-
ery oscillation which is unique to the k = 1.24 case. Steady vortex

Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
JID: HFF
ARTICLE IN PRESS [m5G;January 23, 2017;21:24]

T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14 9

a b c

d e f

g h i

Fig. 9. Pressure distributions on the surface of the wing throughout one full oscillation. (a) k = 1.24, A = 5◦ (b) k = 1.24, A = 10◦ ; (c) k = 1.24, A = 20◦ ; (d) k = 2.48,
A = 5◦ ; (e) k = 2.48, A = 10◦ ; (f) k = 2.48, A = 20◦ ; (g) k = 4.96, A = 5◦ ; (h) k = 4.96, A = 10◦ ; (i) k = 4.96, A = 20◦ .

Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
JID: HFF
ARTICLE IN PRESS [m5G;January 23, 2017;21:24]

10 T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14

a b c

Fig. 10. Vorticity contours showing vortex transfer around the leading edge of the wing for the case of k = 4.96, A = 20◦ at three consecutive times. (a) Negative vortex
(blue) is formed just in front of the first corrugation on the top of the wing; (b) The vortex travels around the leading towards the bottom surface of the wing; (c) Vortex
is now on the bottom side of the wing and is being absorbed by the larger positive vortex (Red) forming under the leading edge of the wing. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)

shedding mechanisms prevail in this case which cause vortices to
be shed at a high enough frequency that two can be formed in
the time that vortex shedding from the corrugations is favourable.
This effect is not apparent in the pressure plot of the 5° case as
the second vortex is too weak to be distinguished. A similar but
unique, steady, mechanism causes a vortex to be shed from the
leading edge due to the increased angle of attack in the A = 20◦
case which subsequently combines with the typical Q vortex be-
fore traveling down the wing.
For k = 2.48, the results for A = 5◦ and A = 10◦ are very simi-
lar with the regular formation of vortices convecting downstream
whilst remaining attached to the airfoil. However with the higher
frequency of oscillations, the vortices no longer reach the trailing
edge during one cycle. For A = 20◦ , the Q vortex no longer remains
attached to the airfoil during its travel downstream. A large lead-
ing edge vortex formed due to the high angle of attack is thrown
up as it interacts with the corrugations towards the front of the
wing. The Q vortex is also thrown towards the leading edge by
the high rotation rate and becomes entrained to combine with the
shed leading edge vortex. The resulting vortex detaches from the
surface of the wing showing little influence on the surface pres-
sure distribution.
For k = 4.96, the most striking feature is at A = 20◦ , where the
vortex R (which normally forms on the bottom surface of the air-
foil) appears on the top surface of the airfoil. The leading edge vor-
tex forms on the top surface (t/T = 0.9) and instead of convecting
downstream along the top surface, it is thrown forwards by the
high rotation rate of the wing and transfers to the bottom sur-
face around the leading edge. It then merges with vortex R and
convects downstream. The presence of the R vortex transit line on
the top surface represents the leading edge vortex that will trans-
fer and form the R vortex. Vorticity contours of snapshots at three
different times are given in Fig. 10 and show the transfer of the
vortex from the top to the bottom surface. This phenomenon has
not been previously reported and could be the cause of a reduction
in lift produced by the k = 4.96, A = 20◦ case.
3.3. Lift and drag coefficients
The detailed flow features presented in Section 3.2 have impor-
tant implications on the lift and drag characteristics and hence per-
formance of the oscillating airfoil. Before presenting periodic re-
sults, the mean lift and drag coefficients will first be discussed,
these results are shown in Fig. 11. It can be immediately observed
that with increased frequency, there is a drop in the mean drag
coefficients for all amplitudes. Sectional lift and drag coefficients
were used for this comparison, the sectional drag coefficient is de-
fined as:
2Fd
Cd = (7)
ρU 2 c
where Fd is the drag force per meter span on the wing.
The trends in mean drag coefficient are very different for the
different frequencies. At the two lower frequencies, the drag values

a b
0.4 1
Mean Drag Coefficient, Cd

0.2 0.8
Mean Lift Coefficient, Cl

0 0.6
,

0 5 0 15 20
-0.2 0.4
-0.4 0.2
-0.6 0
0 5 10 15 20
-0.8 -0.2
-1 -0.4
-1.2 -0.6
Amplitude- (deg) Amplitude (deg)

k=1.24 k=2.48 k=4.96

Fig. 11. Mean (a) Drag and (b) Lift Coefficients.

Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
JID: HFF
ARTICLE IN PRESS [m5G;January 23, 2017;21:24]

T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14 11

10 30

20

Pitch Angle (deg)

5

Coefficient
10

0 0
0 0.2 0.4 0.6 0.8 1
-10
-5
-20

-10 -30
t/T

Lift Drag Pitch Angle

Fig. 12. Lift and Drag Coefficients for k = 2.48, A = 20◦ , plotted with pitch angle (−1∗ AOA).

are always positive with an increasing trend for k = 1.24 and re-
maining more or less stable for k = 2.48. At the highest frequency
of k = 4.96 however, the mean drag coefficient decreases into neg-
ative values with increased oscillation amplitude. This implies that
there is forward momentum being generated and there is thrust
being produced.
For lift on the other hand, there are trends of increasing mean
lift coefficient with increasing oscillation frequencies and ampli-
tude with the exception of k = 4.96, A = 20◦ case. This may be a
result of the vortex transferring of from the top to the bottom sur-
face as discussed in Section 3.2. Symmetric oscillations were tested
in this study and it is expected that a symmetric airfoils would
exhibit zero mean lift. The mean lift coefficients observed in this
study are hence due to asymmetry of the wing profile and vortex
interaction effects. It is apparent that the asymmetry of the wing
is the dominating effect for most cases until cases of high rotation
where the vortex dynamics around the wing change significantly
(i.e. k = 4.96, A = 20◦ ).
The lift and drag coefficients for k = 2.48, A = 20◦ are shown in
Fig. 12. It can be observed that the lift coefficient is not a smooth
sinusoidal curve but has a kink near the position of maximum lift.
While almost unnoticeable in the low amplitude cases, this feature
is common to all cases tested though its location varies widely. An-
other observation is the lack of symmetry for both the lift and drag
curves. This is more clearly indicated by the lift and drag hysteresis
curves shown in Fig. 13. For k = 1.24, 2.48, and 4.96 (A = 5◦ and
10°), it is can be observed that the pitch-down (clockwise rotation)
action generates positive lift but also covers a larger area than the
pitch-up cycle. This indicates that the overall lift is positive as pre-
vious discussed. However at k = 4.96, A = 20◦ , the lower section
covers a marginally larger area indicating overall negative lift.
The difference between the upper and lower sections of the
drag coefficients show positive drag for k = 1.24, 2.48, and 4.96
with A = 5 and negative drag for k = 4.96, A = 10◦ and 20° indicat-
ing thrust being produced. The thrust produced during the pitch-
up action was almost always observed to be lower than that of
the pitch-down action. This is likely geometric with the exception
at k = 4.96, A = 20◦ possibly caused by enhanced vortex dynam-
ics. Another interesting observation is in the cross-over point of
the drag hysteresis. For the lower frequencies, the cross-over point
is to the left of the y-axis whereas for k = 4.96, A = 20◦ , it is on
the right. This indicates that the relative magnitude of drag at an
angle-of-attack of 0° between the pitch-up and pitch-down cycle is
reversed. The implication or explanation of this is not yet certain.
3.4. Comparison with a smooth airfoil
The dynamics of smooth pitching airfoils have been studied ex-
tensively both experimentally (Koochesfahani, 1989) and numeri-
cally (Ramamurti and Sandberg, 2001; Xiao and Liao, 2009). The
thrust produced by a pitching airfoil is related to the Strouhal
number as defined by
fh
St = (8)
U
where h is the width of the wake which in this case is approxi-
mated as the displacement of the trailing edge between the maxi-
mum and minimum angular positions. When plotted against St the
mean thrust coefficient, CT (defined as −1 ∗ Cd ), for various reduced
frequencies, k, and angular amplitudes, A, should collapse onto one
smooth curve (Triantafyllou et al., 1991). The present results are
plotted with some experimental and numerical results for smooth
airfoils in Fig. 14. A power coefficient, Cip , has been defined as
Mθ˙
Cip = (9)
1
2
ρU 3 c
where M is the pitching moment per meter span and θ˙ is the rota-
tion rate. This can be combined with the thrust coefficient to give
a propulsion efficiency, η, as
CT
η= (10)
Cip
The Strouhal number for the other experimental and numerical
cases were calculated using:
3k sin(A )
St = (11)
2π
Under the assumption that the center of rotation lies at ¼ c
from the leading edge of the airfoil in each case. The values k, CT ,
Cip , and η have been estimated from their figures. Note that the
Strouhal number is linearly related to the reduced frequency by a
constant proportional to the wake thickness, 2 sin(A).
The experimental and numerical studies shown in Fig. 14 used
a NACA 0012 airfoil at Reynolds numbers of 12,0 0 0 (Koochesfahani,
1989; Ramamurti and Sandberg, 2001) and 10,000 (Xiao and Liao,
2009), which compare well to the present Reynolds number of
14,0 0 0 though these results should be independent of Reynolds
number (Triantafyllou et al., 1991). There is a moderate amount
of spread in the thrust coefficient data between past experimental
and numerical results. Ramamurti and Sandberg (2001) attributed
this to the possibility of three dimensional effects and the way that
the experimental thrust coefficients were calculated in the experi-
ments of Koochesfahni (1989). This does not, however, explain the
difference between the numerical datasets. Nevertheless these re-
sults provide context to the results that have been obtained in the
present study.
The thrust coefficients for the corrugated foil are consistently
below that reported for the smooth airfoil (Fig. 14(a)). The pres-
ence of separated flow regions caused by the corrugated geome-
try and the multiple sites where vortex shedding can occur are

Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
JID: HFF
ARTICLE IN PRESS
12 T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14
Fig. 13. Lift and Drag hysteresis for the different scenarios plotted against pitch angle (−1∗ AOA). The arrow head indicates the direction of wing oscillation.
likely the culprit of decreased thrust compared to the smooth case.
While the 5˚ and 10˚ amplitude cases collapse onto a common line
the 20˚ cases appear out of place. This may be due to a combi-
nation of the low reduced frequencies tested and the high angle
of attack at the extremes of the wings pitching cycle. This corru-
gated profile begins to stall between 15˚ and 20˚ angle of attack
in a static configuration (New, et al., 2014). Perhaps the airfoil in
this study was oscillating slow enough to allow large scale flow
separation to begin to form resulting in a large increase in drag
and a hence a decrease in thrust coefficient compared to the lower
amplitude cases. All of the experimental and numerical results
that have been shown in Fig. 14 remained below the critical an-
gle of attack for the NACA 0012 airfoil during the entire oscillation
cycle.
The power coefficients for the present results collapse very
well and do not differ greatly from that of the smooth airfoil
(Fig. 14(b)). If power requirements and oscillation Strouhal num-
ber are most important to a given application, for high agility a
high oscillation frequency might be desired for example, there are
not obvious disadvantages to either wing profile. However, when
the thrust coefficient is also considered and propulsion efficiency
is investigated (Fig. 14(c)) the smooth airfoil offers superior per-
formance for all Strouhal numbers studied in this regime. All of
the data from smooth airfoils shown here were produced with re-
duced frequencies from 0–13 whereas the present study considers
Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
[m5G;January 23, 2017;21:24]
k in the range 1–5. While the Strouhal number is the primary pa-
rameter that determines the wake structure of oscillating airfoils
there is evidence for heaving motions that the reduced frequency
is also required to characterise the flow structure in some cases
(Triantafyllou et al., 20 0 0; Wang, 20 0 0), perhaps the same is true
for pure pitching motions. This may also justify the discrepancy
seen here though more research is needed.
There is evidence to show that Strouhal numbers of 0.2–0.4
provide an optimum in propulsion efficiency for flapping motions
(Triantafyllou et al., 1991). Many organisms that use flapping
motions for propulsion such as birds, fish, and insects have been
observed to operate in this optimum range for cruising flight
where efficiency is important (Taylor et al., 2003). Data for cruis-
ing flight of dragonflies is scarce, estimates of St from the flight
data of Wakeling and Ellington (1997b) and Rüppell (1989) for
forward flight range from 0.32 to 1.62. It is unknown if dragon-
flies also operate in the quoted optimal St range or if most of
their flight involves manoeuvring at high St rather than efficient
cruising. The present results include one case at a high St which
produces high thrust though no data for smooth wings is available
for comparison.
Further research would be needed to determine if there was
in fact a periodic pitching, heaving, flapping, or a combination
of these in which a corrugated airfoil outperforms conventional
airfoils. In the case that this regime cannot be found it might
JID: HFF
ARTICLE IN PRESS [m5G;January 23, 2017;21:24]

T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14 13

a b

Fig. 14. (a) Thrust coefficient, (b) power coefficient, and (c) propulsion efficiency ploted against Strouhal number comparing the present results with the experimental results
of (Koochesfahni 1989) and the numerical results of (Ramamurti and Sandberg 2001) and (Xiao and Liao 2009).

suggest that the other advantages of having corrugated wings such
as strength to weight ratio and good high angle of attack perfor-
mance may evolutionarily outweigh any oscillatory performance
loses compared to a smooth wing.
4. Conclusion
The sinusoidal oscillation of a bio-inspired corrugated airfoil in
the presence of upstream flow at Re = 14 0 0 0 has been simulated
using CFD codes and validated against experimental PIV measure-
ments. The calculations show previously unreported intricacies of
the near field flow structure with good spatial and time resolu-
tions as well as lift and drag calculations of such corrugated airfoil.
The complex set of results have been presented as time dependent
pressure contours. This allows comparison of the dynamics of vor-
tices traveling along the surface of the wing at different operating
conditions without showing copious snapshots in time. Traveling
vortices appear to be the most influential to the surface pressure
field in the k = 2.48 regime with large amplitudes acting to disturb
the vortex dynamics.

Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
JID: HFF
ARTICLE IN PRESS
14 T.J. Flint et al. / International Journal of Heat and Fluid Flow 000 (2017) 1–14
It has been shown that whilst the instantaneous pressure on
the top of the airfoil differs from case to case, the pressure along
the bottom surface is largely similar for all cases. At low oscillat-
ing frequencies, k = 1.24, vortices that are shed and travel down
the wing are relatively weak and at the lowest amplitude tested,
5˚, vortices were dissipated before reaching the trailing edge of the
wing. This indicates that in order for the oscillation to have signif-
icant effect on the flow field around the wing, it has to be above
k = 1.24 and A = 5◦ (St > 0.05), below which it will behave similar
to a static airfoil.
Symmetrical oscillation of the chosen wing profile produced a
positive mean lift for most of the cases considered. This is likely
due to the geometry of the wing including asymmetric corru-
gations and a cambered tail. Increasing the oscillation frequency
and amplitude tended to increase the lift generated. At the high-
est tested frequency and amplitude k = 4.96, A = 20◦ , the leading
edge vortex formed on the top surface did not convect downstream
and instead reversed its direction of travel, transferring itself to the
bottom surface around the leading edge. This has implications in
terms of reducing the mean lift generated below zero. Conversely,
the violent rotation experienced at k = 4.96, A = 20◦ produced a
large mean thrust as inertial effects begin to dominate. A mean
thrust was observed in only the higher frequencies and amplitudes
beyond k = 4.96 and A = 10◦ (St > 0.4 )below this, no thrust was
generated.
Comparison with published data for smooth airfoils reveals
that the corrugated foil has a lower propulsive efficiency in the
regimes that have been studied. However, the difference in re-
duced frequency between the present results and published results
for smooth airfoils may be important considering the high angles
of attack that have been simulated here though this effect is not
known. These results may indicate that other benefits of having
corrugated wings, such as structural benefits, may be more evolu-
tionarily favourable for the dragonfly than oscillatory pitching per-
formance. There may be other oscillatory regimes, ones that better
resemble flapping flight, in which corrugated foils provide perfor-
mance benefits over a smooth one, further research is necessary
on this topic.
Finally the location of the cross-over point of the drag hys-
teresis indicates that at the highest frequency and amplitude (k =
4.96, A = 20◦ ), the relative values of drag at angle-of-attack of 0˚
are reversed compared to the lower frequencies. The implications
and explanation of this is not yet known.
References
Dickinson, M.H., Gotz, K.G., 1993. Unsteady aerodynamic performance of model
wings at low Reynolds numbers. J. Exp. Biol. 174, 45–64.
Hu, H., Tamai, M., 2008. Bioinspired corrugated airfoil at low Reynolds numbers. J.
Aircraft 45 (6), 2068–2077.
Kesel, A.B., 20 0 0. Aerodynamic characteristics of dragonfly wing sections compared
with technical aerofoils. J. Exp. Biol. 203, 3125–3135.
Koochesfahani, M.M., 1989. Vortical patterns in the wake of an oscillating airfoil.
AIAA J. 27 (9), 1200–1205.
Lee, T., 2011. Flow past two in-tandem airfoils undergoing sinusoidal oscillations.
Exp. Fluids 51 (6), 1605–1621.
Levy, D.-E., Seifert, A., 2009. Simplified dragonfly airfoil aerodynamics at Reynolds
numbers below 80 0 0. Phys. Fluids 21.
Please cite this article as: T.J. Flint et al., Computational study of a pitching bio-inspired corrugated airfoil, International Journal of Heat
and Fluid Flow (2017), http://dx.doi.org/10.1016/j.ijheatfluidflow.2016.12.009
[m5G;January 23, 2017;21:24]
Lian, Y., Broering, T., Hord, K., Prater, R., 2014. The characterization of tandem and
corrugated wings. Prog. Aerosp. Sci. 65, 41–69.
Liu, H., Kawachi, K., 1998. A numerical study of insect flight. J. Comput. Phys. 146,
124–156.
Lin, J.C.M., Pauley, L.L., 1996. Low-Reynolds-number separation on an airfoil. AIAA J.
34 (8), 1570–1577.
Lu, K., Xie, Y.H., Zhang, D., Lan, J.B., 2013. Numerical investigations into the asym-
metric effects on the aerodynamic response of a pitching airfoil. J. Fluids Struct.
39, 76–86.
Menter, F.R., 2012. In: Best practice: scale-resolving simulations in ANSYS CFD, AN-
SYS Germany GmbH, Version 1.02.
Menter, F.R., Egorov, Y., 2010. The scale-adaptive simulation method for unsteady
turbulent flow predictions. Part 1: theory and model description. Flow Turbul.
Combust. 85 (1), 113–138.
Molki, M., Sattari, N., 2013. Vortex generation in low-speed flow over an oscillating
and deforming arc airfoil. J. Fluids Eng. 135.
Murphy, J., Hu, H., 2010. An experimental study of a bio-inspired corrugated airfoil
for micro air vehicle applications. Exp. Fluids 49 (2), 531–546.
Nawroth, J.C., Lee, H., Feinberg, A.W., Ripplinger, C.M., McCain, M.L., Grosberg, A.,
Dabiri, J.O., Parker, K.K., 2012. A tissue-engineered jellyfish with biomimetic
propulsion. Nat. Biotechnol. 30, 792–797.
New, T.H., Chan, Y.X., Koh, G.C., Hoang, M.C., Shi, S., 2014. Effects of corrugated aero-
foil surface features on flow-separation control. AIAA J. 52 (1), 206–211.
Newman, B.G., Savage, S.B., Schouella, D., 1977. Model tests on a wing section of an
Aeschna dragonfly. In: Scale Effects in Animal Locomotion, pp. 445–477.
Norberg, R.Å., 1975. Hovering flight of the dragonfly Aeschna Juncea L. kinematics
and aerodynamics. In: Swimming and Flying in Nature, pp. 763–781.
Okamoto, M., Azuma, A., 2005. Experimental study on aerodynamic characteristics
of unsteady wings at low Reynolds number. AIAA J. 43 (12), 2526–2536.
Panda, J., Zaman, K.B.M.Q., 1992. Experimental investigation of the flowfield of an
oscillating airfoil. 10th AIAA Applied Aerodynamics Conference.
Patankar, S.V., 1980. Numerical heat transfer and fluid flow. Hemisphere., Washing-
ton, DC.
Pines, D.J., Bohorquez, F., 2006. Challenges facing future micro-air-vehicle develop-
ment. J. Aircraft 43 (2), 290–305.
Ramamurti, R., Sandberg, W., 2001. Simulation of flow about flapping airfoils using
finite element incompressible flow solver. AIAA J. 39 (2), 253–260.
Rees, C.J.C., 1975. Form and function in corrugated insect wings. Nature 256,
200–203.
Rüppell, G., 1989. Kinematic analysis of symmetrical flight manoeuvres of Odonata.
J. Exp. Biol. 144, 13–42.
Saharon, D., Luttges, M., 1987. Three-dimensional flow produced by a pitching–
plunging model dragonfly wing. AIAA 25th Aerospace Sciences Meeting.
Shyy, W., Lian, Y., Tang, J., Liu, H., Trizila, P., Stanford, B., Bernal, L., Cesnik, C., Fried-
mann, P., Ifju, P., 2008. Computational aerodynamics of low Reynolds number
plunging, pitching and flexible wings for MAV applications. Acta Mech. Sin. 24,
351–373.
Taylor, G.K., Nudds, R.L., Thomas, A.L.R., 2003. Flying and swimming animals cruise
at a Strouhal number tuned for high power efficiency. Nature, 425, 707–711.
Triantafyllou, M.S., Triantafyllou, G.S., Gopalkrishnan, R., 1991. Wake mechanics for
thrust generation in oscillating foils. Phys. Fluids A 3 (12), 2835–2837.
Triantafyllou, M.S., Triantafyllou, G.S., Yue, D.K.P., 20 0 0. Hydrodynamics of fishlike
swimming. Annu. Rev. Fluid Mech. 32, 33–53.
van den Berg, C., Ellington, C.P., 1997. The three-dimensional leading-edge vortex of
a ’hovering’ model Hawkmoth. Phil. Trans. R. Soc. Lond. B, 352, 329–340.
Viieru, D., Tang, J., Lian, Y., Liu, H., Shyy, W., 2006. Flapping and flexible wing aero-
dynamics of low Reynolds number flight vehicles. 44th AIAA Aerospace Sciences
Meeting and Exhibit.
Wakeling, J.M., Ellington, C.P., 1997a. Dragonfly flight. I. Gliding flight and steady-
state aerodynamic forces. J. Exp. Biol. 200, 543–556.
Wakeling, J.M., Ellington, C.P., 1997b. Dragonfly flight. II. Velocities, accelerations and
kinematics of flapping flight. J. Exp. Biol. 200, 557–582.
Wakeling, J.M., Ellington, C.P., 1997c. Dragonfly flight. III. Lift and power require-
ments. J. Exp. Biol., 200, 583–600.
Wang, Z.J., 20 0 0. Vortex shedding and frequency selection in flapping flight. J. Fluid
Mech. 410, 323–341.
Xiao, Q., Liao, W., 2009. Numerical study of asymmetric effect on a pitching foil. Int.
J. Mod. Phys. C 20 (10), 1663–1680.
Zhou, Z., Li, C., Nie, J.B., Chen, Y., 2013. Effect of oscillation frequency on wind tur-
bine airfoil dynamic stall. In: IOP Conference Series: Materials Science and En-
gineering, 52 (5).