Professional Documents
Culture Documents
and the assessment of the mesh motion solvers, the physical and
numerical modelling are described. The incompressible Navier-Stokes equations
are rewritten in the rotating reference frame in order to identify dimensionless
numbers related to the wing motion. The most important number is the Rossby
number, which represents the wing stroke path curvature.
First a two-dimensional study is performed to investigate the eects of dier-
ent wing kinematic models, with increasing complexity, on hovering ight perfor-
mance. The results show that the sawtooth amplitude has a small eect on the
mean lift but the mean drag is aected signicantly. The second model simplica-
tion, the trapezoidal angle of attack, caused the leading-edge vortex to separate
during the translational phase. This led to an increase in mean drag during each
half-stroke. The extra bump in angle of attack as used by the fruit y model is
not aecting the mean lift to a large extent. The other realistic kinematic feature
is the deviation, which is found to have only a marginal eect on the mean lift and
mean drag in this two-dimensional study. However, the eective angle of attack is
altered such that the deviation leads to levelling of the force distribution.
Additionally, a numerical model for two-dimensional ow was used to investi-
gate the eect of foil kinematics on the vortex dynamics around an ellipsoid foil
subject to prescribed apping motion over a range of dimensionless wavelengths,
dimensionless amplitudes, angle of attack amplitudes, and stroke plane angles.
Both plunging and rotating motions are prescribed by simple harmonic functions
which are useful for exploring the parametric space despite the model simplicity.
Optimal propulsion using apping foil exists for each variable which implies that
aerodynamics might select a range of preferable operating condition. The condi-
tions that give optimal propulsion lie in the synchronisation region in which the
ow is periodic.
Furthermore, dierent results relevant to three-dimensional apping wing aero-
v
dynamics, are described. First, the ow around a dynamically scaled model wing
is solved for dierent angles of attack in order to study the force development
and vortex dynamics at small and large mid-stroke angle of attack. Secondly, the
Rossby number is varied at dierent Reynolds numbers. A varying Rossby number
represents a variation in stroke path curvature and thus angular acceleration. It is
shown that a low Rossby number is benecial for the stability of the leading-edge
vortex, leading to an increase in lift and eciency. Thirdly, the three-dimensional
wing kinematics is varied by changing the shape in angle of attack and by applying
a deviation, which may cause a gure-of-eight pattern. As in two-dimensional
studies, the deviation may inuence the force distribution to a large extent, by
changing the eective angle of attack. Additionally, the three-dimensional ow is
compared with the two-dimensional studies performed on apping forward ight.
Finally, a preliminary investigation is performed to show the eect of wing
exing. Therefore, a pre-dened exing deformation is applied to a plunging airfoil
in two-dimensional forward ight and to a three-dimensional apping wing in
hovering ight. Concerning the exible airfoil in forward ight, a similar behaviour
was observed as for a rigid plunging airfoil, subjected to additional rotation.
The present simulations have led to important insight to understand the inu-
ence of wing kinematics and deformation on the aerodynamic performance. These
results may be important to design and optimise Micro Air Vehicles.
Contents
Summary iii
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Physics of apping ight . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Experimental and numerical methods . . . . . . . . . . . . . . . . 7
1.4 Objectives and approach . . . . . . . . . . . . . . . . . . . . . . . . 9
2 Finite volume discretisation 13
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 The Navier-Stokes equations . . . . . . . . . . . . . . . . . . . . . . 16
2.2.1 Constitutive relations . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Incompressible laminar ow simplications . . . . . . . . . 18
2.2.3 Dimensionless numbers . . . . . . . . . . . . . . . . . . . . 18
2.3 Spatial and temporal discretisation . . . . . . . . . . . . . . . . . . 19
2.4 Measures of cell quality . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Discretisation of an incompressible momentum equation . . . . . . 21
2.5.1 Face interpolation schemes . . . . . . . . . . . . . . . . . . 23
2.5.2 Convection term . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.3 Diusion term . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5.4 Temporal term . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7 Solution of the Navier-Stokes equations . . . . . . . . . . . . . . . 27
2.7.1 Pressure equation and Pressure-Velocity coupling . . . . . . 28
2.7.2 Procedure for solving the Navier-Stokes equations . . . . . 29
2.7.3 Arbitrary Lagrangian Eulerian approach . . . . . . . . . . . 30
viii Contents
2.8 Swept volume calculation . . . . . . . . . . . . . . . . . . . . . . . 32
2.9 Numerical ow solvers . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.10 Code validation and verication . . . . . . . . . . . . . . . . . . . . 34
2.10.1 2D vortex decay and convection . . . . . . . . . . . . . . . . 35
2.10.2 Validation using cylinder ows . . . . . . . . . . . . . . . . 39
2.11 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3 Mesh deformation techniques for apping ight 49
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.2 Dierent mesh deformation techniques . . . . . . . . . . . . . . . . 52
3.2.1 Laplace equation with variable diusivity . . . . . . . . . . 53
3.2.2 Solid body rotation stress equation . . . . . . . . . . . . . . 54
3.2.3 Radial basis function interpolation . . . . . . . . . . . . . . 55
3.3 Mesh quality measures . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4 Comparison of mesh motion solvers . . . . . . . . . . . . . . . . . . 59
3.4.1 Translation and rotation of a two-dimensional block . . . . 59
3.4.2 Flapping of a three-dimensional wing . . . . . . . . . . . . . 62
3.4.3 Flexing of a two-dimensional block . . . . . . . . . . . . . . 64
3.5 Improving computational eciency . . . . . . . . . . . . . . . . . . 66
3.5.1 Boundary point coarsening and smoothing . . . . . . . . . . 67
3.5.2 Iterative techniques and parallel implementations . . . . . . 69
3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4 Physical and numerical modelling of apping foils and wings 73
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.2 Governing equations for apping wings . . . . . . . . . . . . . . . . 75
4.3 Wing shape and kinematic modelling . . . . . . . . . . . . . . . . . 78
4.3.1 Wing shape and planform selection . . . . . . . . . . . . . . 79
4.3.2 Kinematic modelling . . . . . . . . . . . . . . . . . . . . . . 80
4.3.3 Modelling of active wing exing . . . . . . . . . . . . . . . . 82
4.3.4 Numerical implementation of the wing kinematics . . . . . 83
4.4 Dynamical scaling of apping wings . . . . . . . . . . . . . . . . . 86
4.5 Computational domain and boundary conditions . . . . . . . . . . 88
4.6 Denition of force and performance coecients . . . . . . . . . . . 89
4.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5 Inuence of wing kinematics in two-dimensional hovering ight 93
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.1.1 Similarity and discrepancy between two- and
three-dimensional ows . . . . . . . . . . . . . . . . . . . . 94
5.1.2 Inuence of kinematic modelling . . . . . . . . . . . . . . . 94
5.2 Numerical simulation methods . . . . . . . . . . . . . . . . . . . . 96
5.2.1 Flow solver and governing equations . . . . . . . . . . . . . 96
5.2.2 Mesh generation and boundary conditions . . . . . . . . . . 97
Contents ix
5.2.3 Validation using harmonic wing kinematics . . . . . . . . . 99
5.3 Modelling insect wing kinematics . . . . . . . . . . . . . . . . . . . 99
5.3.1 Insect wing selection and model parameters . . . . . . . . . 100
5.3.2 Dynamical scaling of the wing model . . . . . . . . . . . . . 101
5.3.3 Force and performance indicators . . . . . . . . . . . . . . . 102
5.3.4 Dierent wing kinematic models . . . . . . . . . . . . . . . 103
5.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.4.1 Overall model comparison . . . . . . . . . . . . . . . . . . . 105
5.4.2 Kinematic features investigation . . . . . . . . . . . . . . . 109
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6 Vortex wake interactions of a two-dimensional apping foil 121
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.2 Flapping foil parametrisation . . . . . . . . . . . . . . . . . . . . . 122
6.3 Force coecients and performance . . . . . . . . . . . . . . . . . . 124
6.4 Numerical model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.5 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.5.1 Inuence of dimensionless wavelength . . . . . . . . . . . . 126
6.5.2 Inuence of dimensionless amplitude . . . . . . . . . . . . . 127
6.5.3 Inuence of angle of attack amplitude . . . . . . . . . . . . 129
6.5.4 Inuence of stroke plane angle . . . . . . . . . . . . . . . . 130
6.5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
7 Vortical structures in three-dimensional apping ight 135
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.2 Three-dimensional apping wing simulations . . . . . . . . . . . . 137
7.2.1 Modelling and parameter selection . . . . . . . . . . . . . . 138
7.2.2 Simulation strategy and test matrix selection . . . . . . . . 141
7.3 Flow solver accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.4 Vortex identication methods for ow visualisation . . . . . . . . . 145
7.5 Flapping wings at low Reynolds numbers . . . . . . . . . . . . . . 148
7.5.1 The angle of attack in apping ight . . . . . . . . . . . . . 149
7.5.2 Inuence of apping stroke curvature . . . . . . . . . . . . . 149
7.5.3 Inuence of Reynolds number . . . . . . . . . . . . . . . . . 152
7.5.4 Inuence of trapezoidal angle of attack . . . . . . . . . . . 158
7.5.5 Inuence of deviation . . . . . . . . . . . . . . . . . . . . . 160
7.6 Flapping wings in forward ight . . . . . . . . . . . . . . . . . . . . 163
7.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
8 Inuence of wing deformation by exing 171
8.1 Airfoil exing in two-dimensional forward apping ight . . . . . . 171
8.2 Wing exing in three-dimensional hovering ight . . . . . . . . . . 175
8.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
x Contents
9 Conclusions and recommendations 179
9.1 Overall conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
9.2 Conclusions on mesh motion techniques . . . . . . . . . . . . . . . 180
9.3 Conclusions on hovering apping ight . . . . . . . . . . . . . . . . 181
9.3.1 Two-dimensional hovering . . . . . . . . . . . . . . . . . . . 181
9.3.2 Three-dimensional hovering . . . . . . . . . . . . . . . . . . 182
9.4 Conclusions on forward apping ight . . . . . . . . . . . . . . . . 184
9.4.1 Two-dimensional forward apping . . . . . . . . . . . . . . 184
9.4.2 Three-dimensional forward apping . . . . . . . . . . . . . 184
9.5 Preliminary conclusions on wing exing . . . . . . . . . . . . . . . 185
9.6 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
A Grid generation for apping wings 189
A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
A.2 BlockMesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
A.3 Gambit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
A.4 GridPro
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
A.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
B Flow solver settings 195
B.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
B.2 Fluent
C
L
= 1.540
x
c
[]
-3 -2 -1 0 1 2 3
(a) (b)
Figure 1.3 Forces and vortices in apping wing aerodynamics. (a) Two-dimensional il-
lustration of the wing kinematics and the resulting force vector generated by the apping airfoil at
Re = O(100) (Bos et al., 2008). (b) Three-dimensional leading-edge vortex generated by a apping
wing at Re = O(1000).
Leading-edge vortex
The potential benet of vortices attached to the wing was already discussed
by Maxworthy (1979) and Dickinson & Gotz (1993). It was Ellington et al. (1996)
who identied the presence of a leading-edge vortex (LEV) generated on top of the
apping wing, increasing the lift force to values much higher than predicted by con-
ventional wing theory. The stability of the helical three-dimensional leading-edge
vortex is still not yet fully understood and appears to heavily depend on the wing
kinematics and Reynolds number. It appears that the leading-edge vortex is more
stable around a three-dimensional apping wing compared to two-dimensional
apping foil situations.
In apping foil aerodynamics the vortices are shed and form either a periodic
or chaotic wake pattern, depending on the kinematics, notably advance ratio and
dimensionless apping amplitude (Thaweewat et al., 2009, Lentink et al., 2008).
The origin of the leading-edge vortex is the roll-up of shear layers, present in
highly viscous ows, which is the case at low Reynolds numbers. It is thought that
the kinematics in two- and three-dimensional apping inuences the shear layer
direction and ow accelerations, which will undoubtedly inuence the development
of the leading-edge vortex (Lentink & Dickinson, 2009b).
In order to understand the physics of apping wing aerodynamics, it is im-
portant to obtain insight in how an insect moves its wing. Figure 1.3(a) shows
a two-dimensional illustration of the wing kinematics of a fruit y, operating at
Re = 110, from (Bos et al., 2008).
Inuence of insect wing kinematics on forces
The relevance of experiments and ow simulations of insect ight has been found
to depend on how reliably true insect wing kinematics are reproduced. Wang
1.2 Physics of apping ight 5
et al. (2004) and Sane & Dickinson (2001) showed that the kinematic modelling
signicantly inuences the mean force coecients and its distribution. Addition-
ally, Hover et al. (2004) showed that the angle of attack inuences the apping
foil propulsion eciency to a large extent. This illustrates the appreciable eects
which details of the wing kinematics, like parameter values and stroke patterns,
may have on ight performance. It further emphasises the need to critically assess
the inuence of kinematic model simplications.
In literature, dierent kinematic models have been employed to investigate the
aerodynamic features of insect ight. For example, Wang (2000a,b) and Lentink
& Gerritsma (2003) numerically investigated pure harmonic translational motion
with respectively small and large amplitudes. Wang (2000a,b) varied apping am-
plitude and frequency and showed that at a certain parameter selection the lift
is clearly enhanced. Lewin & Haj-Hariri (2003) performed a similar numerical
study for heaving airfoils. Besides lift enhancement at certain reduced frequen-
cies, they found periodic and aperiodic ow solutions which are strongly related
to the aerodynamic eciency. Lentink & Gerritsma (2003) varied airfoil shape
with amplitude and frequency xed at values representative to real fruit ies.
They concluded that the airfoil geometry choice is of minor inuence, but large
amplitudes lead to an increase of lift by a factor of 5 compared to static forces
generated by translating airfoils. It was also shown that wing stroke models with
only translational motion could not provide for realistic results, such that includ-
ing rotation is essential. In addition to the harmonic models with pure translation
(Dickinson & Gotz, 1993), rotational parameters were investigated by Dickinson
(1994). They varied rotational parameters and showed that axis-of-rotation, rota-
tion speed and angle of attack during translation are of great importance for the
force development during each stroke. Harmonic wing kinematics, including wing
rotation, where used by Pedro et al. (2003) and Guglielmini & Blondeaux (2004)
in their numerical models to solve for forward ight. Both studies emphasised the
importance of angle of attack to inuence the propulsive eciency. Slightly more
complex fruit y kinematic models were used by Dickinson et al. (1999) and Sane
& Dickinson (2001) with their Roboy. Based on observation of true insect ight,
the wing maintains a constant velocity and angle of attack during most of the
stroke, with a relatively strong linear and angular acceleration during stroke re-
versal. This results in the typical sawtooth displacement and trapezoidal angle
of attack pattern of the Roboy kinematic model. Using these models, the eect
of amplitude, deviation, angle of attack and the timing of the latter were explored.
The present thesis deals with dierent kinematic models from literature, both
the pure harmonic and the Roboy model, in order to investigate their inuence
on the aerodynamic performance (Bos et al., 2008). Furthermore, the results were
compared with more realistic fruit y kinematics obtained from the observation of
free ying fruit ies. Instead of performing a parameter study within the scope of
one kinematic model, the objective of the present study is to compare the eect
6 Introduction
(a) (b) (c)
Figure 1.4 Vortex wakes generated by cylinders and apping wings. (a) Von K arm an vortex
street behind a stationary cylinder at Re = 150. (b) Periodic vortex wake behind a plunging airfoil at
Re = 110, one single and one vortex pair is generated each plunging period. (c) Chaotic vortex wake
behind a plunging airfoil at Re = 110, depending on the kinematics a chaotic wake pattern may occur
with unpredictable forces as the result.
of the available models as a whole. This leads to better insights into the conse-
quences of simplications in kinematic modelling, which is of great importance to
both experiments and numerical simulations. Also, it may reveal the importance
of certain specic features of the stroke pattern, in relation to aerodynamic per-
formance.
The similarity between two- and three-dimensional ows
To limit both the parametric space as well as the computational eort, many stud-
ies have been performed as two-dimensional simulations (Thaweewat et al., 2009,
Bos et al., 2008, Wang et al., 2004, Lewin & Haj-Hariri, 2003). One of the major
(and partially unresolved) issues in modelling of insect ight and apping wing
propulsion, is the possibly restrictive applicability of two-dimensional results to
true insect ight. Additional important aspects are unsteady ow mechanisms,
wing exibility (uid structure interaction) and Reynolds number eects. In a
recent paper Wang et al. (2004) compared three-dimensional Roboy results with
two-dimensional numerical results. Both Dong et al. (2005) and Blondeaux et
al. (2005b) concluded that two-dimensional studies over predict forces and per-
formances since the energy-loss, which is present in three dimensions, is not re-
solved. Dong et al. (2005), Blondeaux et al. (2005b) numerically investigated the
wake structure behind nite-span wings at low Reynolds numbers. They observed
three-dimensional vortical structures around apping wings with low aspect ratio,
as was mentioned by Lighthill (1969).
Notwithstanding the possible discrepancy between two- and three-dimensional
ow, two-dimensional analysis has often been applied to obtain insight into the
aerodynamic eects of wing kinematics and geometry. Wang et al. (2004) con-
rmed that the similarities between two- and three-dimensional approaches are
sucient to warrant that a reasonable approximation of insect ight can be ob-
tained using a two-dimensional approach. First, in case of advanced and symmet-
ric rotation the forces were found to be similar in the two-dimensional simulations
compared to the three-dimensional experiments. Secondly it was observed that
in both simulations and experiments the leading-edge vortex did not completely
separate for amplitude-to-chord ratios between 3-5 (Dickinson & Gotz, 1993, Dick-
1.3 Experimental and numerical methods 7
inson, 1994). The current research deals with amplitudes that are in this range.
In view of the excessive computational expense required to perform accurate
three-dimensional simulations, and with the above justication, the rst part of
this thesis makes extensive use of two-dimensional simulations. In the second
part, various three-dimensional simulations were performed using limited varia-
tions wing kinematics.
1.3 Experimental and numerical methods
In literature, dierent methods were used to solve and visualise the ow around
apping insect wings, from realistic fruit y measurements to three-dimensional
simulations using a representative model wing. In this section, dierent methods
will be briey addressed, from experimental methods to computational uid dy-
namics simulations.
Experimental investigations and quasi steady theory
Several experimental studies considered the ight performance of insects, and re-
vealed the complex nature of insect ight aerodynamics. The ow induced by the
motion of insect wings is highly unsteady and vortical, as visualised by Weish-
Fogh & Jensen (1956) using tethered locusts and by Willmott et al. (1997) using
a hawkmoth (Manduca Sexta), see gure 1.2. More recently, Srygley & Thomas
(2002) and Thomas et al. (2004) performed free ight and tethered experimental
visualisations using butteries and dragonies to show the complicated vortical
structures. This unsteady and vortical ow behaviour is a consequence of the
high relative frequencies, amplitudes and the very low Reynolds number involved
(Re < 1000 for a large number of insects and Re 110 for the fruit y, Drosophila
Melanogaster, in particular).
Ellington (1984) indicated that the lift in insect ight is signicantly higher
than expected on the basis of quasi-steady aerodynamics, hence revealing that
important unsteady and vortical ow phenomena play a major role in insect ight.
In several studies (Dickinson & Gotz, 1993, Dickinson, 1994, Dickinson et al.,
1999) it was conrmed that important aspects, like delayed stall and wake capture
enhance the lift force beyond values predicted by quasi-steady theory. Ellington
et al. (1996) discovered that these lift increasing mechanisms are amplied by
the generation of a leading-edge vortex (LEV). It was shown that this leading-
edge vortex arises during the translational part of the wing motion rather than
during the rotational ip between up and down stroke. The lift increasing eect of
the leading-edge vortex strongly depends on the kinematics of the apping wing
(Dickinson et al., 1999, Wang, 2000b, Sane & Dickinson, 2001, 2002, Lentink &
Dickinson, 2009a,b).
In order to understand insect ight performance Dickinson et al. (1999) and
Wang (2000b) applied the quasi-steady theory to compare with unsteady forces.
The quasi-steady approach was revised by Sane & Dickinson (2002) to include ro-
8 Introduction
tational eects but even then the results require further improvement. According
to Sane & Dickinson (2001) the mean lift is well predicted by quasi-steady theory,
but the mean drag is underestimated. This conrms the restricted applicability of
the quasi-steady theory due to lack of unsteady mechanisms like rotational lift and
wake capture. Several experimental studies have been performed with the aim of
characterising the unsteady aerodynamics of insect ight. Dickinson et al. (1999)
investigated the ow around a apping Roboy model which moves in oil to meet
the same ow conditions as the real fruit y encounters (reproduction of Reynolds
number in particular).
Numerical simulations
Notwithstanding important advances in experimental techniques for non-intrusive
ow eld analysis, Particle Image Velocimetry in particular (Bomphrey et al., 2006,
Poelma et al., 2006), it remains dicult to capture all the relevant details of the
ow using only experimental techniques. An appealing approach, therefore, is to
supplement experiments with numerical ow simulations. A number of numerical
studies on full three-dimensional congurations have been reported, in relation to
specic insect geometries (moth: Liu & Kawachi (1998), fruit y: Ramamurti &
Sandberg (2002), Sun & Tang (2002), dragony: Young & Lai (2008), Isogai et al.
(2004)).
To perform numerical simulations around moving objects, such as apping
wings, one can use either immersed boundary methods (Peskin, 2002, Mittal &
Iaccarino, 2005), deforming mesh techniques (Boer de et al., 2007, Jasak, 2009),
see gure 1.5, or even complete re-meshing (Young & Lai, 2008, Zuo et al.,
2007). Although, the computational eort involved in three-dimensional stud-
ies is presently still extremely demanding, an integrated computational study was
performed by Aono et al. (2008) who developed a code to incorporate two wings
and a body using overset mesh techniques. In an immersed boundary method, the
moving boundary is projected on a xed Cartesian background grid, which is not
allowed to deform. Besides interpolation issues, the conservation of mass and mo-
mentum in current immersed boundary methods is not obvious, even not for xed
boundaries (Mittal & Iaccarino, 2005). Nevertheless, when two wings touch, as in
the manoeuvre clap-and-ing (Weish-Fogh & Jensen, 1956), one will undoubtedly
need methods like overset, immersed boundary or re-meshing techniques.
Together with the unavailability of an accurate ow solver with parallel sup-
port, it was chosen to assess and improve existing mesh motion techniques. The
commonly used mesh motion techniques result in high quality meshes as long as
the rotation of the moving boundaries is limited. In order to cope with high ro-
tation rates, mesh motion based on radial basis function (RBF) interpolation is
implemented in this thesis and improved in terms of accuracy and eciency. This
modern mesh motion technique is incorporated in OpenFOAM
1
, which is an
open-source framework to solve the Navier-Stokes equations on three-dimensional
1
OpenFOAM
software.
1.4 Objectives and approach 9
(a) (b)
Figure 1.5 Dierent mesh motion solvers. Two illustrations of mesh motion solutions, (a) shows
a Laplacian mesh motion, while (b) shows the mesh motion obtained by using radial basis function
interpolation.
unstructured grids of polyhedral cells with full parallel support. This code is thor-
oughly tested and used for apping foil and wing simulations.
Arbitrary Lagrangian-Eulerian formulation
The governing equations to solve the ow are generally discretised using the Eule-
rian description, where the uid is allowed to ow through the xed mesh. This is
in contrast to the Lagrangian formulation, where the mesh is xed to the uid or
material. If the material or uid deforms, the mesh deforms with it. This method
is commonly used to discretise the governing equations encountered in structure
mechanics. However, when the ow domain moves or deforms in time due to
a moving boundary, a xed mesh becomes inconvenient, because it requires the
explicit tracking of the domain boundary. Therefore, the Arbitrary Lagrangian
Eulerian (ALE) formulation is used to discretise the ow equations on moving and
deforming meshes (Donea, 1982). This method incorporates and combines both
Lagrangian and Eulerian frameworks. The Lagrangian contribution allows the
mesh to move and deform according to the boundary motion, whereas the Eule-
rian part takes care of the uid ow through the mesh. At the time of writing, the
ALE method has become the standard implementation in most popular codes to
solve for the ow around moving boundaries while the mesh deforms accordingly.
1.4 Objectives and approach
Flapping ight aerodynamics is governed by many parameters, like advance ratio,
wing kinematics, Reynolds number, etc. In order to perform accurate numerical
simulations it is important to use an ecient code which is capable to solve for
10 Introduction
various conditions, using realistic wing kinematics. For large wing translations and
rotations the numerical grid needs to deform accordingly to maintain high accuracy
of the ow solver. Therefore, the overall goal of this research is to develop a reliable
mesh deformation technique, in terms of accuracy and eciency, to solve the ow
around apping wings. This method is used to study the complex vortical patterns
to identify optimal strategies in apping foil and wing aerodynamics. In order to
satisfy this aim, the following objectives are dened:
1. improve current mesh motion techniques and implementation, using an ac-
curate and ecient framework,
2. validate and verify the numerical solver with the implemented and improved
mesh motion technique,
3. solve for the ow around two-dimensional apping foils to study the wing-
wake interaction as well as the inuence of wing kinematics,
4. solve for the ow around three-dimensional apping wings to assess the im-
portance of parameters like apping amplitude, frequency or Reynolds num-
ber,
5. study the three-dimensional structure of vortical patterns, especially the
leading-edge vortex.
Approach and outline
In order to solve for the ow around apping foils and wings, an improved mesh
motion technique, based on radial basis function interpolation, is implemented in
the open-source framework OpenFOAM
and the
non-commercial open-source code OpenFOAM
is a general-
purpose CFD solver, which has been an authority in the eld of computational
uid dynamics for decades. Two major drawbacks of a commercial solver are the
unavailability of the source code and the potential lack of sucient support from
the company or the user community in code development. The used open-source
solver, OpenFOAM
and OpenFOAM
t
+
(u) = 0. (2.1)
Here, [kg/m
3
] is the density and u [m/s] the ow velocity vector. The nabla
operator is dened in three dimensions as
=
.
x
+
.
y
+
.
z
.
2.2 The Navier-Stokes equations 17
Secondly, for momentum conservation the following expression can be derived
(neglecting gravity and additional body forces):
(u)
t
+
(uu) =
, (2.2)
where [N/m
2
] is the surface stress tensor, necessary in viscous uid ow. For
compressible ow calculations also the energy conservation equation is specied:
(e)
t
+
(eu) =
(u)
q +Q, (2.3)
where e [J/kg] is the total specic energy (including kinetic and potential energy),
q [W/s] is the heat ux vector and Q [Jm
3
/kg] equals the nett energy generation.
The full set of equations describing unsteady, compressible viscous ows, are called
the Navier-Stokes equations. The Navier-Stokes equations are non-linear, which
makes them dicult to solve; only for very simplied problems there exists an
analytical solution.
2.2.1 Constitutive relations
In order to close the system of equations (2.1), (2.2) and (2.3), constitutive relations
are needed. For a Newtonian uid, the stress tensor, , which is dened for a
Newtonian uid as
=
_
p +
2
3
u
_
I +
_
u +u
T
_
.
Here I represents the identity tensor, p [N/m
2
] is the pressure and [Ns/m
2
] is the
dynamic viscosity. To close the energy equation, the equation of state is specied,
such as the perfect gas law:
p = RT,
in which T [K] is the temperature and R [J/(molK)] the specic gas constant.
The constitutive relation for the total specic energy yields as follows:
e = e(p, T).
Additionally, the heat conduction is described using Fouriers law:
q = T,
with [W/(mK)] the heat conduction transport coecient.
The governing equations (2.1), (2.2) and (2.3) in combination with additional tur-
bulence modelling can be used in a wide variety of engineering problems. Without
other restrictions, these equations are used for high and low speed ows, turbu-
lence research, multi-phase ows and a lot of other applications. However, these
equations can be dicult to solve and simplications can be made if applicable to
the concerning problem.
18 Finite volume discretisation
2.2.2 Incompressible laminar ow simplications
The current research deals with the ow around apping wings at insect scale,
which is considered to be incompressible (Lentink, 2003, Bos et al., 2008) and
laminar (Williamson, 1995). Therefore, the incompressible laminar Navier-Stokes
equations are solved for Reynolds numbers ranging from Re = 100 to 1000.
A ow can be assumed to be incompressible, when the velocity is lower than 0.3
times the speed of sound (Lentink, 2003, Bos et al., 2008)) and thermal expansion
eects can be neglected. The incompressible Navier-Stokes equations are:
u = 0, (2.4)
u
t
+
(uu) =
p
+
2
u, (2.5)
with = / [m
2
/s] being the kinematic viscosity. For an incompressible ow,
this system of equations is closed such that there is no need to use the energy
equation and additional turbulence modelling.
2.2.3 Dimensionless numbers
In general, the relative relevance of the dierent terms in equations (2.4) and (2.5)
is revealed by making those equations dimensionless. Therefore, the main vari-
ables, u, t, x, p and are scaled with their reference values, as follows:
u
=
u
U
ref
, t
= t f
ref
, x
=
x
L
, p
=
p
ref
U
2
ref
,
ref
. (2.6)
The star (*) is used to indicate the dimensionless variables. In case of incom-
pressible ow, the density is constant, such that
= 1. When substituting
equation (2.6) into equations (2.4) and (2.5) the following non-dimensional form
of the incompressible continuity and momentum equations is obtained:
= 0, (2.7)
and
St
u
t
+
(u
) = p
+
1
Re
2
u
. (2.8)
In these equations, two main dimensionless numbers are identied as relevant
parameters, the Strouhal (St) and Reynolds number (Re):
St =
f
ref
L
ref
U
ref
=
T
conv
T
motion
, (2.9)
Re =
U
ref
L
ref
=
T
visc
T
conv
. (2.10)
These dimensionless numbers represent order estimates for time-scale ratios in the
ow. In (2.9) and (2.10), these relevant time-scales are, respectively, the time-scale
2.3 Spatial and temporal discretisation 19
for convective transport, T
conv
, viscous transport, T
visc
and the relevant time-scale
of the body motion, T
motion
. In order for the dimensionless numbers to have a
proper physical meaning, the reference values need to be chosen appropriately.
It was seen that uid ow is governed by non-linear partial dierential equa-
tions, which can only be solved analytically for extremely simplied model prob-
lems. The full Navier-Stokes equations, combined with the constitutive relations,
are applicable to all kind of ows, where the computational costs strongly depend
on the desired resolution and solution methods. When the ow is considered lam-
inar and incompressible, the governing equations are signicantly simplied, such
that the costs for solving may be reduced. However, these simplications need to
be justied by the concerned uid ow problem. Concerning apping wing physics
at lower Reynolds numbers (100 Re 1000) the ow inherently is incompress-
ible and laminar. Therefore, solving the unsteady incompressible laminar ow can
be seen as performing a Direct Numerical Simulation, at suciently low Reynolds
numbers.
2.3 Spatial and temporal discretisation
This section deals with the spatial and temporal discretisation of the governing
mathematical equations. Since time can be interpreted as a parabolic coordi-
nate (Patankar & Spalding, 1972), it is sucient to specify the initial time-step,
which is used to march linearly in time, starting at the initial solution. The
time-step may vary, dependent on the maximal Courant number, which will be
explained in section 2.5. Space, however, needs to be discretised throughout the
entire computational domain. The nite volume approach needs a domain sub-
division into a nite number of convex polyhedral control volumes without overlap,
completely lling the domain. OpenFOAM
N
(a) Cell non-orthogonality.
P N f
i
S
f
f
m
d
(b) Cell skewness.
Figure 2.3 Quality measures using cell non-orthogonality and cell skewness. Two-
dimensional representation of cell non-orthogonality (a) and cell skewness (b) as a measure for the
nite volume cell quality. Cell non-orthogonality is dened as the angle between the face normal vector
S
f
and the direction vector between two cell centres P and N. The cell skewness is dened by the
vectors m and d. From (Jasak, 1996).
error of the diusion term. The second quality criterion is the cell skewness, see
gure 2.3(b). When the line connecting the two neighbouring face centres does
not coincide with the connecting face centre, the cell is skewed. The degree of
skewness is dened by:
=
|m|
|d|
,
where m and d are dened in gure 2.3(b). Assessing cell skewness is important,
since the interpolation from cell centre to face centre strongly depends on this
quality criterion as will later be seen in this chapter.
2.5 Discretisation of an incompressible momen-
tum equation
This section deals with the temporal and spatial nite volume discretisation of the
incompressible momentum equation, which forms the basis for the incompressible
Navier-Stokes equations. This partial dierential equation has the following form:
u
t
+
(uu)
(u) =
p
. (2.13)
Here, [kg/m
3
] is the reference density, u [m/s] the transport velocity and =
/ [m
2
/s] is the kinematic viscosity. This expression contains a temporal, con-
22 Finite volume discretisation
vection and a diusion term, given by:
u
t
: temporal term,
(uu)dV
_
V
CV
(u)dV =
_
V
CV
p
dV. (2.14)
This equation is solved in both CFD codes used, Fluent
and OpenFOAM
. In
the remainder of this section, the dierent terms of equation (2.14) are elaborated
in more detail.
Before dealing with the discretisation of the dierent terms of equation (2.14)
it is important to discuss the evaluation of the volume, surface, divergence and
the gradient integrals, necessary to understand the evaluations of the convection
and diusion terms. For this, the scalar variable is used, which may represent
the dierent velocity components. When substituting equation (2.11) into the
volume integral, a second-order approximation is obtained, such that the result is
a multiplication of the scalar value multiplied by the cell volume.
_
V
P
(x)dV =
_
V
P
(
P
+ (x x
P
)()
P
)dV
=
P
_
V
P
dV + (
_
V
P
(x x
P
)dV ) (
P
)
P
V
P
.
Similar, for the surface integral the following yields:
_
f
dS a = S
f
a
f
, (2.15)
where S
f
is the face surface area. The divergence and gradient terms are evaluated
using Gauss theorem (Panton, 2005, Anderson Jr., 1991), which denes a relation
between the volume and the surface integrals. Using Gauss theorem, the volume
integral of the divergence of a vector a can be written as the sum of all faces, like:
_
V
P
adV =
_
S
CV
dS
a,
=
f
_
S
f
dS
a,
f
dS
f
a
f
. (2.16)
2.5 Discretisation of an incompressible momentum equation 23
Here, C
V
is the control volume with surface normal vectors S
f
and a
f
is a vector
interpolated to the cell faces using a second-order linear interpolation method.
Discretisation of the gradient integral of a scalar variable can be written,
using Gauss theorem, as:
_
V
P
dV =
_
S
CV
dS,
=
f
_
S
f
dS,
f
dS
f
.
f
.
2.5.1 Face interpolation schemes
Similar to the divergence term, the ux
f
is interpolated from the cell centre to
the face centres. A linear interpolation is performed using the following expression:
f
= f
x
P
+ (1 f
x
)
N
,
which is illustrated in gure 2.5. The linear interpolation factor f
x
is dened as the
ratio of two distances, f
x
= |fD|/|PD|. So, both divergence and gradient volume
integrals can be reduced to a summation of the corresponding vector or scalar
variable over the cell faces. The standard face interpolation scheme is obtained by
central dierencing.
In OpenFOAM
(
r
)
0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
2
2.5
3
(a) SuperBee (Roe)
r
(
r
)
0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
2
2.5
3
(b) Koren
r
(
r
)
0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
2
2.5
3
(c) Van Leer
Figure 2.4 Dierent ux splitting limiters. Flux splitting schemes are used to limit the gradient
of the solution in order to avoid spurious wiggles. The ux splitting scheme are a function of r, which
represents the ratio of successive gradients on the mesh. Dierent ux limiters are employed, (a)
SuperBee (Roe, 1986), (b) Koren (Koren, 1993) and (c) Van Leer (Van Leer, 1979, Sweby, 1984).
method (midpoint, least squares):
_
V
P
(u)dV =
f
S
f
(u)
f
=
f
S
f
(u)
f
f
=
f
F
f
.
Here, F is the mass ux, given by F = S
f
(u)
f
. The scalar variable needs to
be interpolated using a second-order interpolation method in combination with a
ux limiter, e.g. linear, Gamma, Van Leer.
2.5.3 Diusion term
The volume integral of the diusion term from equation (2.13) is discretised and
approximated using linearization as
_
V
P
()dV =
f
S
f
()
f
=
f
(S )
f
.
Here, the terms (S )
f
and
f
need to be approximated using a proper method.
The face viscosity
f
is obtained by interpolation from cell centre to faces. The
other term (S )
f
is obtained on a non-orthogonal mesh by the following ex-
pression:
(S )
f
= |m|
N
P
|d|
+k ()
f
.
2.5 Discretisation of an incompressible momentum equation 25
ts
Flow direction
U P D
D
f
P d f m N
S
f
N
k
Figure 2.5 Variation of the ux . The
value of at the face f is determined as a
function of upstream and downstream values.
Figure 2.6 Cell non-orthogonality
treatment. Illustration of the cell non-
orthogonality correction which is used
on meshes with large skewness and non-
orthogonality.
Here d is the vector between two adjacent cell centres and m is parallel to d with
magnitude of the surface normal vector S
f
. The decomposition of S
f
is shown
in gure 2.6 and derived in Jasak (1996), Juretic (2004) such that the following
relation holds:
S
f
= m+k,
where k is orthogonal to the surface normal vector S.
2.5.4 Temporal term
Since the unsteady Navier-Stokes equations are solved, a proper discretisation of
the temporal scheme is necessary. The time derivative represents the temporal
rate of change of which needs to be discretised using new and old time values.
This time dierence is dened using prescribed time-step size t such that:
n+1
=
n
+ t,
where
n
and
n+1
are the scalar variable at the old and new time instances,
respectively. Two implicit time discretisation methods are considered, one rst-
order and and one second-order scheme. The rst-order discretisation is simply
the temporal dierence:
t
=
n+1
n
t
,
and the second-order discretisation, see (Wesseling, 2001, Hirsch, 1988), is given
by:
t
=
3
2
n+1
n
+
1
2
n1
t
.
This implicit scheme is referred to as the second-order backward dierencing
scheme, where
n1
is the old-old value of . Consequently, the corresponding
26 Finite volume discretisation
volume integrals obey the following relations:
_
C
V
t
dV =
n+1
n
t
V
P
,
_
C
V
t
dV =
3
2
n+1
2
n
+
1
2
n1
t
V
P
. (2.17)
Note, that these relations are only valid on xed meshes and constant time-steps.
According to (Wesseling, 2001, Hirsch, 1988), the explicit rst-order time integra-
tion method may be unstable if the Courant number is larger than 1, where the
Courant number is dened as
Co =
u t
x
.
Implicit methods are in general more stable, compared to (semi-) explicit methods,
such that in the current research the implicit rst- and second-order backward
scheme have been used. While the implicit methods are bounded and stable, a
pre-dened maximal Courant number Co
max
is used to vary the corresponding
time-step during the simulation. In that case, the coecients
3
2
, 2 and
1
2
in (2.17)
should be elaborated to incorporate the ratio of the old and current time-steps.
In section 2.7.3 this will be discussed in more detail.
2.6 Boundary conditions
In order to solve the discretised governing equations, boundary conditions need
to be dened at the boundaries of the computational domain. There are four
boundary conditions (Hirsch, 1988, Wesseling, 2001), which are used to close the
system, namely:
1. zero-gradient boundary condition, dening the solution gradient to be zero.
This condition is known as a Neumann-type condition, /n = a,
2. xed-value boundary condition, dening a specied value of the solution.
This is a Dirichlet-type condition, = b,
3. symmetry boundary condition, treats the conservation variables as if the
boundary was a mirror plane. This condition denes that the component
of the solution gradient normal to this plane should be xed to zero. The
parallel components are extrapolated from the interior cells,
4. moving-wall-velocity boundary condition is used on a moving boundary to
keep the ux zero, using the Arbitrary Lagrangian Eulerian approach.
For external ow simulations, a distinction is made between the outer and the
inner boundaries, the latter corresponds to the moving wing or body. To minimise
the eects of the outer boundaries it is desirable to specify a symmetry boundary
2.7 Solution of the Navier-Stokes equations 27
condition (3) at those xed boundaries, unless a free-stream is specied. In case of
forward apping ight, two domain boundaries are dened as inow and outow,
respectively. At the inow boundary the velocity is dened as xed-value (2) and
the pressure as zero-gradient (1). On the other hand, at the outow boundary,
the pressure has to be xed-value and the velocity zero-gradient (Hirsch, 1988,
Wesseling, 2001). On a stationary wall the no-slip condition needs to be guaran-
teed, therefore a xed-value (u = 0) is specied for the velocity in combination
with a zero-gradient for the pressure. If the boundary of the wall moves, than
the proper boundary condition is the moving-wall-velocity (4) which introduces
an extra velocity in order to maintain the no-slip condition and ensures a zero ux
through the moving boundary.
2.7 Solution of the Navier-Stokes equations
Previously, the dierent terms to discretise the general momentum equation (2.13),
were described. This section briey deals with the discretisation of the Navier-
Stokes equations and the solution procedure. The incompressible laminar Navier-
Stokes equations were given by (2.4) and (2.5):
u = 0,
u
t
+
(uu) =
p
+
2
u.
There are two items, requiring special attention, namely the non-linear term
present in the momentum equation and the pressure-velocity coupling (Ferziger &
Peric, 2002). The non-linear term in these governing equations,
(uu), can be
solved either by using a solver for non-linear systems or by Newton linearization.
Previously, it was seen that the convection term can be written as:
_
V
P
(uu)dV =
f
S
f
(u)
f
(u)
f
=
f
F(u)
f
= a
p
u
p
+
N
a
N
u
N
,
where a
p
, a
N
and F are still depending on u. a
p
and a
N
represent the diagonal
and o-diagonal terms of the sparse system of equations, respectively. A complete
derivation can be found in (Jasak, 1996). Since F should satisfy the continuity
equation (2.4), both equations (2.4) and (2.5) should be solved together as if it
was a coupled system. In order to avoid the use of expensive solvers for non-linear
systems, this convection term is linearised such that existing velocity elds will be
used to calculate the matrix coecients a
p
and a
N
.
28 Finite volume discretisation
2.7.1 Pressure equation and Pressure-Velocity coupling
Since the pressure depends on the velocity and vice-versa, a special treatment of
this inter-equation coupling is needed. In order to derive the pressure equation, a
semi-discrete formulation of the momentum equation is written as:
a
p
u
p
= H(u) p. (2.18)
This equation is derived from the integral form of the momentum equation using
the previously described discretisation methods and divided by the volume. Fol-
lowing Rhie & Chow (1983) the pressure gradient in equation (2.18) is not yet
discretised. The H(u) term contains two parts, a convection and a source con-
tribution. The convection part includes the matrix coecients for all neighbours
multiplied by their corresponding velocities. The source contribution consists of
all source terms, except the pressure term, including the transient term. Therefore
H(u) can be written as follows:
H(u) =
N
a
N
u
N
+
u
0
t
.
Additionally, the discretised continuity equation (2.4) is given by:
u =
f
S
f
u
f
= 0, (2.19)
Now equation (2.18) is rewritten to nd an expression for u
p
:
u
p
=
H(u)
a
p
1
a
p
p. (2.20)
The velocities on the face of a nite volume cell can be expressed as the interpolated
value on the face of equation (2.20):
u
f
=
_
H(u)
a
p
_
f
_
1
a
p
_
f
(p)
f
. (2.21)
This equation will be used to determine the face uxes. If equation (2.21) is
substituted into equation (2.19), the following pressure equation can be obtained:
_
1
a
p
p
_
=
_
H(u)
a
p
_
. (2.22)
The Laplacian operator is discretised using existing methods, which are previously
explained. Combining equations (2.18) to (2.22), the nal form of the discretised
Navier-Stokes equations can be written as:
a
p
u
p
= H(u)
f
S
f
(p)
f
, (2.23)
2.7 Solution of the Navier-Stokes equations 29
and
f
S
f
__
1
a
p
_
(p)
f
_
=
f
S
f
_
H(u)
a
p
_
f
, (2.24)
additionally, the face ux is calculated using:
F = S
f
u = S
f
_
_
H(u)
a
p
_
f
_
1
a
p
_
(p)
f
_
. (2.25)
When equation (2.22) is satised, the face uxes are guaranteed to be conser-
vative (Ferziger & Peric, 2002). For the discretised form of the Navier-Stokes
equations (2.23) and (2.24) it can be observed that both equations are coupled
through the pressure and velocity, which requires special attention. Since a simul-
taneous approach would be too computationally demanding, this system is solved
in a segregated manner, which means that these equations are solved in sequence.
The inner-equation coupling is established using either PISO (Issa, 1986) or SIM-
PLE (Patankar & Spalding, 1972) based algorithms. Both ow solvers, Fluent
and OpenFOAM
used the PISO scheme for transient ows and the SIMPLE
scheme for steady ows. Since the PISO scheme was used, a brief description is
given. The PISO algorithm consists of the following steps:
1. Momentum predictor stage. The momentum equation (2.23) is solved
using the pressure gradient, known from the previous time-step, since the
actual pressure gradient is not yet calculated. Furthermore, equation (2.23)
provides an approximation of the new velocity eld.
2. Pressure solution stage. Using the predicted velocity, from the previ-
ous stage, the H(u) term can be constructed such that the pressure equa-
tion (2.22) can be formed. Using this pressure equation a better approxima-
tion of the new pressure eld can be obtained.
3. Explicit velocity correction stage. The last equation in this sequence
is (2.25), which determines the conservative uxes, which are consistent with
the new pressure eld. Since the approximated pressure eld, from stage 1,
is replaced by a better pressure eld, from stage 2, the velocity eld has to be
corrected accordingly. This is performed using (2.20) in an explicit fashion.
For more detailed information, please consult (Jasak, 1996, Ferziger & Peric,
2002, Juretic, 2004).
2.7.2 Procedure for solving the Navier-Stokes equations
After dealing with the discretisation of the Navier-Stokes equations in combination
with the PISO algorithm, it has become possible to describe the solution procedure
to obtain the solution of the Navier-Stokes equations. In unsteady simulations all
other inter-equation couplings, besides the pressure-velocity equations, are lagged,
such that they are included in the PISO loop. For incompressible unsteady ow
30 Finite volume discretisation
with additional turbulence modelling, the solution sequence can be summarised as
follows:
1. Initialisation of all elds, including pressure and velocity, using the initial
condition,
2. Start the simulation to obtain the velocity and pressure values at the new
time-step,
3. Create and solve the momentum predictor equations using the obtained face
uxes,
4. Iterate through the PISO loop until the pre-dened tolerance of the pressure-
velocity system is reached. The pressure and velocity elds are obtained for
the current time-step, as well as a new set of conservative uxes,
5. Using the new conservative uxes, all remaining equations of the system are
solved. If turbulence modelling is included, update the turbulent viscosity
at this stage,
6. Unless the nal time is reached, go back to step 2.
This procedure results in solution elds for all solved variables, like pressure, veloc-
ity and possible turbulence variables. Since the present thesis deals with deforming
mesh problems, special attention is necessary to describe the modications to the
discretised equations dealing with the Arbitrary Lagrangian Eulerian approach,
commonly used to satisfy conservation on deforming meshes.
2.7.3 Arbitrary Lagrangian Eulerian approach
The governing equations to solve the ow are generally discretised using the Eule-
rian description, where the uid is allowed to ow through the xed mesh (Ferziger
& Peric, 2002). This is in contrast to the Lagrangian formulation, where the mesh
is xed to the uid or material. If the material or uid deforms, the mesh deforms
with it. This method is commonly used to discretise the governing equations
encountered in structural mechanics. However, when the ow domain moves or
deforms in time due to a moving boundary, a xed mesh becomes inconvenient,
because it requires the explicit tracking of the domain boundary. Therefore, the
Arbitrary Lagrangian Eulerian (ALE) formulation is used to discretise the ow
equations on moving and deforming meshes (Donea, 1982). This method incorpo-
rates and combines both Lagrangian and Eulerian frameworks. The Lagrangian
contribution allows the mesh to move and deform according to the boundary mo-
tion, whereas the Eulerian part takes care of the uid ow through the mesh. At
the time of writing, the ALE method has become the standard implementation in
most popular CFD codes to solve for the ow around moving boundaries while the
2.7 Solution of the Navier-Stokes equations 31
mesh deforms accordingly. In general, the momentum equation (2.14) for a scalar
eld can be derived on a moving mesh as
t
_
V
CV
dV +
_
S
CV
n(uu
s
)dS
_
S
CV
ndS =
_
V
CV
S
()dV, (2.26)
where V
CV
is the arbitrary volume and u
s
the velocity of the moving surface.
The relationship between the rate of change of the volume V
CV
and the veloc-
ity u
s
of the boundary surface S is dened by the so-called Space Conservation
Law (SCL) (Ferziger & Peric, 2002) or Geometric Conservation Law (Lesoinne &
Farhat, 1996):
t
_
V
CV
dV
_
S
CV
n u
s
dS = 0.
Using the current nite volume discretisation, the computational domain is split
into a nite number of polyhedral cells with varying shape and volume, since the
mesh is deforming. The cells do not overlap and completely ll the domain (Jasak,
2009). In time, the temporal dimension is marched using a variable time-step,
corresponding to a maximal Courant number, using either an implicit rst-order
Euler method or a second-order backward scheme (Tukovic & Jasak, 2007). For
eciency, the second-order accurate three time levels backward scheme was used
throughout this research. If equation (2.26) is discretised in space and time the
following relation is obtained for a constant time-step:
3
n+1
P
n+1
P
V
n+1
P
4
n
P
n
P
V
n
P
+
n1
P
n1
P
V
n1
P
2t
+
f
( m
n+1
f
n+1
f
V
n+1
f
)
n+1
f
=
f
(
)
n+1
f
S
n+1
f
n
n+1
f
()
n+1
f
+s
n+1
V
n+1
P
,
where the subscript P denotes the cell values and f represents the values at the
face centres. The superscripts n+1, n and n1 are, respectively, the new, old and
old-old values. The mass ux through the face is given by m
f
= n
f
u
f
S
f
and
the cell face volume change by
V
f
= n
f
u
s
f
S
f
, where u
s
represents the cell face
velocity. The uid mass ux m is obtained as part of the solution, satisfying mass
conservation. Furthermore, it is important to determine the volume face ux
such that it satises the Space Conservation Law. The temporal discretisation
scheme should be similar to the one used in the momentum equation, otherwise
inconsistency could introduce numerical errors. It is very important to determine
the volume face ux in a consistent way such that it equals the swept volume
calculation, see section 2.8.
Note, that the previous equation is derived for a constant time-step, i.e. the
maximal Courant number varies during the simulation. The current research uses
32 Finite volume discretisation
x
y
z
P
f
(a)
t
n
V
t
n+1
f
(b)
Figure 2.7 Finite volume cell decomposition to calculate swept volumes. The nite volume
cell decomposition is used to form a tetrahedral mesh used for point-based mesh motion solvers and to
calculate the swept volumes. (a) shows the decomposition of a polyhedral cell into tetrahedral volumes
and faces. The swept volume V of a decomposed face is shown in (b).
a dened maximal Courant number leading to a varying time-step. Therefore, the
following relation is derived for a non-constant time-step:
_
1 +
t
n+1
t
n+1
+ t
n
_
n+1
P
n+1
P
V
n+1
P
_
1 +
t
n+1
t
n
_
n
P
n
P
V
n
P
+
_
1 +
(t
n+1
)
2
t
n
(t
n+1
+ t
n
)
_
n1
P
n1
P
V
n1
P
+
f
( m
n+1
f
n+1
f
V
n+1
f
)
n+1
f
=
f
(
)
n+1
f
S
n+1
f
n
n+1
f
()
n+1
f
+s
n+1
V
n+1
P
,
where the new and old time-steps are respectively given by t
n+1
= t
n+1
t
n
and
t
n
= t
n
t
n1
.
2.8 Swept volume calculation
The swept volume is dened as the volume swept by a face of a polyhedral cell
between two subsequent time-steps, t
n
and t
n+1
. This calculation is necessary in
order to satisfy the Space Conservation Law, described in the previous section.
If a polyhedral face is swept from one time-step to the next, it may occur that
the volume becomes warped, such that a volume calculation will not be trivial.
Therefore, the polyhedral cells and faces are decomposed into tetrahedral cells, see
gure 2.7(a). The polyhedral face is decomposed into triangles, using its centroid,
2.9 Numerical ow solvers 33
which is illustrated in gure 2.7(b). The swept volume of a polyhedral face f
is equal to the sum of the swept volumes of the dierent decomposed triangles,
which need to be accurately calculated. As is illustrated in gure 2.7(b), the swept
volume of such a triangle is similar to a prism with a triangle-shaped bottom area.
Since this prism may be warped, the volume is calculated as the sum of three
tetrahedron volumes. One tetrahedron is shown in the gure, but the remaining
volume of the prism contains two more tetrahedrons. These two additional tetra-
hedrons can be constructed in two dierent unique ways, using either one of the
two diagonals in the right side face of the swept volume shown in gure 2.7(b),
see (Zuijlen van, 2006). Therefore, the total swept volume using the two dierent
unique tetrahedron decompositions is:
V
1
=
1
6
(V
p
1
+V
p
2
+V
p
3
),
and
V
2
=
1
6
(V
p
1
+V
p
4
+V
p
5
),
where the same base tetrahedron V
p
1
is used and V
p
2
, V
p
3
, V
p
4
and V
p
5
corre-
spond to dierent tetrahedron volumes. To obtain the total swept volume of the
decomposed triangle V
triangle
, the average of both V
1
and V
2
is taken as
V
triangle
=
1
2
(V
1
+V
2
).
This swept calculation is successfully validated on test cases using unsteady ow
with mesh motion and proved to be suciently accurate. Therefore, this method
is implemented in OpenFOAM
.
2.9 Numerical ow solvers
Section 2.7 described the solution procedure to solve the Navier-Stokes equations,
necessary for uid ow. This section deals with a brief elaboration of the computer
software, i.e. the CFD codes, used throughout the present research. Two dierent
CFD codes are used, one commercial (Fluent
).
Fluent
.
The pressure-velocity coupling in incompressible ow simulations was obtained
using the iterative PISO scheme (Ferziger & Peric, 2002). The accuracy was set
to double-precision and the initial conditions were chosen to be uniform. The
boundary condition on the body was set to no-slip. The convergence criterion for
34 Finite volume discretisation
the iterative method was satised with mass and momentum residuals dropping
O(10
5
) in magnitude.
The other code used in this research, OpenFOAM
, is a general object-oriented
toolbox, written in C++, which is used to solve partial dierential equations, e.g.
the Navier-Stokes equations, on a nite volume mesh containing polyhedral cells,
making this code very versatile. One of the main assets of the code is that the
user writes the code in an intuitive way himself, without the need to dig deep in
the underlying code.
All terms are discretised using standard second-order central dierencing, ex-
cept for the convection term. Section 2.10 will show that the best method to
discretise the convection term, for our low Reynolds number problems, turned out
to be the linear scheme with the Van Leer limiter (Van Leer, 1979). Concerning
the temporal discretisation scheme, the implicit second-order backward scheme is
used, in combination with a variable time-step corresponding a maximal Courant
number (Wesseling, 2001, Ferziger & Peric, 2002).
Additionally, the iterative solvers and their corresponding convergence criteria
need to be specied. The convergence criterion is based on the residual, which is
derived from the complete system of equations, which is written as
Ax = b,
such that the residual Res is dened by:
Res = b Ax
The pressure equation is solved using a pre-conditioned conjugate Gradient (PCG)
iterative solver, while the pressure-velocity coupling equation employs its asym-
metric counterpart pre-conditioned bi-stab conjugate Gradient (PBiCG) solver.
The pre-conditioning method varies from incomplete Choleski to incomplete LU
decomposition (Wesseling, 2001, Jasak et al., 2007). Appendix B summarises the
used discretisation schemes and iterative solvers combined with the convergence
criteria for both ow solvers Fluent
and OpenFOAM
.
2.10 Code validation and verication
This section deals with the validation and verication of the CFD solvers that
were used throughout the current research. The commercial ow solver Fluent
has already been tested earlier specically for low Reynolds number ows, relevant
to apping insect ight in (Lentink & Gerritsma, 2003, Zuo et al., 2007, Bos et
al., 2008, Thaweewat et al., 2009). OpenFOAM
= t
m
f(), (2.27)
where V
= 2e
1
2
e
2
2
, (2.28)
where is a function of the similarity parameter :
=
2
_
(2)
.
Vortex decay
As a rst test problem, a Taylor vortex is considered, which decays in a two-
dimensional squared domain with dimensions (5 x 5). This squared domain is
discretised with a Cartesian nite volume grid of 100 x 100 mesh cells as is shown
in gure 2.9, which shows the vorticity of the vortex at t = 0. To solve the
incompressible Navier-Stokes equations, a standard OpenFOAM
solver is used,
icoFoam, without mesh motion, which is not necessary for this problem.
The temporal term is discretised using a second-order backward scheme. Fur-
thermore, all other terms, except for the convection term, are discretised using
a second-order linear interpolation. To study the eect of the face interpolation
of the convection term, the following schemes are used: Gamma (Jasak et al.,
1999), Koren limiter (Koren, 1993), SuperBee limiter (Roe, 1986), Van Leer lim-
iter (Van Leer, 1979) and standard linear interpolation. The mesh size was xed
to (100 x 100) and the time-step is varying to meet a maximal Courant number of
Co
max
= 1.0, which are both considered to be ne enough. The Reynolds number
2.10 Code validation and verication 37
Figure 2.9 Initial solution of a Taylor vortex on an Cartesian mesh. Starting from this initial
Taylor vortex solution the ow diusion is solved, without the presence of a convection free-stream.
The velocity proles and total energy are monitored to identify the accuracy of the ow solver.
is xed to Re = 100 by setting the kinematic viscosity to = 0.01 and the velocity
vector to u = (1.0, 0.0, 0.0). The ow solver solves for 40 seconds such that the
temporal eect of the dierent face interpolation schemes on the shape and mag-
nitude of the velocity proles can be compared. Any excess or lack in numerical
diusion may become visible.
Results
Figure 2.10(a) and 2.10(b) show the initial and nal velocity variations in X- and
Y -direction for the dierent face interpolation schemes. Besides the evolution of
the velocity prole the total energy is shown in gure 2.10(c), which is a measure
for the diusion. The total energy is calculated as
E
tot
=
N
i=1
0.5|u
i
|
2
,
where i represents the cell index, N the total number of cells and u
i
the velocity
in cell i.
The rst important observation, from gure 2.10, is that the total energy is
increasing for the SuperBee ux limiter. This scheme clearly introduced a large
amount of negative numerical diusivity which causes the Taylor vortex to grow,
which is not physical. The velocity prole as well as the total energy of both
Gamma and linear interpolation schemes are similar, without a signicant amount
of diusion. One major drawback of these methods is that the vortex looses sym-
metry after 5 s, gure 2.10(b). The Koren and Van Leer limiters are slightly more
38 Finite volume discretisation
U velocity [m/s]
Y
c
o
o
r
d
i
n
a
t
e
[
m
]
Initial solution
Gamma
Koren
SuperBee
van Leer
Linear
-1 -0.6 -0.2 0.2 0.6 1
3
3.5
4
4.5
5
5.5
6
(a)
X coordinate [m]
V
v
e
l
o
c
i
t
y
[
m
/
s
]
Initial solution
Gamma
Koren
SuperBee
van Leer
Linear
3 3.5 4 4.5 5 5.5 6
-1
-0.6
-0.2
0.2
0.6
1
(b)
Time [t]
T
o
t
a
l
e
n
e
r
g
y
[
m
2
/
s
2
]
Gamma
Koren
SuperBee
van Leer
Linear
0 2 4 6 8 10 12 14 16
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
1.12
(c)
Figure 2.10 Velocity and total energy variations of a decaying Taylor vortex. (a) shows
u(y), at t=20 s, for dierent face interpolation schemes, while (b) provides v(x). (c) shows the
decaying total energy due to numerical diusion.
diusive, see gure 2.10(c), but provide similar results. The Van Leer limiter is
called shape preserving and provides good results for both vortex decay and con-
vection as will be seen in the next section.
Vortex convection
The second validation case concerns a Taylor vortex, which is convected through
a channel with dimensions (20 x 5), shown in gure 2.11. Similar to the vortex
decay problem, the discretisation of the face interpolation is varied and all other
discretisation terms are xed to second-order linear interpolation. The Cartesian
mesh resolution was set to (400 x 100) and the max Courant number was equal to
Co
max
= 1.0. The ow solver solved the governing equations for 20 s, such that
the vortex was convected through the entire domain. The Reynolds number was
xed to Re = 100 by setting a kinematic viscosity to = 0.01 and an inlet velocity
2.10 Code validation and verication 39
Figure 2.11 Initial solution of a Taylor vortex on an Cartesian mesh, representing a
channel. This Taylor vortex solution is used as the initial solution for the convection validation case.
At suciently large grid resolution and time-step size, the eect of dierent face interpolation schemes
is compared.
to u = (1.0, 0.0, 0.0).
Results
First of all, gure 2.12(c) shows that the total energy in the entire domain is de-
creasing with time for all schemes. Extra diusion is clearly visible after t=14 s
when the vortex approaches the outlet boundary, smearing the vortex. Since
convection induces physical diusion, depending on the convection velocity, the
SuperBee scheme possibly still has negative numerical diusion, as was seen in
the results for vortex decay. Furthermore, the total energy does not provide more
information about the accuracy of the scheme, since the integrated energy is close
for all schemes, except for the SuperBee scheme. Additionally, gure 2.12(a)
and 2.12(b) show, respectively, the velocity in X- and Y -direction through the
vortex core at t = 10 s. Besides the possibly negative numerical diusion in the
SuperBee scheme (Juntasaro & Marquis, 2004), another observation can be made.
The second-order linear and Gamma schemes lead to an overshoot of the velocity
in Y -direction. This eect increases with time and therefore these two methods
are not appropriate to study the vortical wake patterns in insect ight. Again, the
Koren and Van Leer limiters are very close, without overshoots and with a proper
symmetry preservation of the vortex. When looking in real detail to these results,
the Van Leer limiter slightly outperforms the Koren limiter. The Van Leer limiter
leads to smoother, and more symmetrical vortices (Juntasaro & Marquis, 2004,
Kuzmin & Turek, 2004), such that this scheme was used throughout this study.
2.10.2 Validation using cylinder ows
To validate the accuracy of the ow solver for unsteady and vortical ow, two
cylinder example problems are dened at suciently low Reynolds numbers. At
Reynolds numbers less than Re = 47 the ow exhibits steady behaviour (see
Williamson, 1998), which is not relevant for unsteady insect ight, so a Reynolds
number Re > 47 needs to be chosen. On the other hand, when taking a Reynolds
40 Finite volume discretisation
U velocity [m/s]
Y
c
o
o
r
d
i
n
a
t
e
[
m
]
Gamma
Koren
SuperBee
van Leer
Linear
0.4 0.6 0.8 1 1.2 1.4 1.6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
(a)
X coordinate [m]
V
v
e
l
o
c
i
t
y
[
m
/
s
]
Gamma
Koren
SuperBee
van Leer
Linear
12 13 14 15 16 17 18
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
(b)
Time [t]
T
o
t
a
l
e
n
e
r
g
y
[
m
2
/
s
2
]
Gamma
Koren
SuperBee
van Leer
Linear
0 2 4 6 8 10 12 14 16 18 20
50
50.1
50.2
50.3
50.4
50.5
50.6
50.7
50.8
50.9
51
(c)
Figure 2.12 Velocity proles of a convected Taylor vortex. The velocity in X-direction u(y)
is shown in (a) at time t=10 s. (b) illustrates the velocity prole in Y -direction as a function of the
X-coordinate, v(x). (c) shows the total energy for the convected vortex.
number larger than about Re = 185, the ow becomes turbulent and additional
turbulence modelling becomes necessary. Since the main objective of this research
is to solve for unsteady, vortical ow around apping wings, the characteristics of
that kind of ow needs to be present in the validation cases. The apping wing
simulations are performed in the laminar ow regime Re = O(100), with periodic
force histories. Therefore, in the range 100 Re 200, two validation cases
were selected, one concerns the ow around a stationary cylinder at Re = 150 and
the other involves a transversely oscillating cylinder at Re = 185 (Guilmineau &
Queutey, 2002).
The main parameter selected for comparison is the time-averaged drag coe-
cient, which is well-documented in literature. Figure 2.14 shows the computational
domain used for both validation cases, the boundaries are located at 10D before,
above and below the cylinder, where D is the cylinder diameter. The outow
2.10 Code validation and verication 41
Reynolds number [-]
S
t
r
o
u
h
a
l
n
u
m
b
e
r
[
-
]
40 60 80 100 120 140 160 180 200
0.12
0.13
0.14
0.15
0.16
0.17
0.18
0.19
0.2
(a)
Reynolds number [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
10
1
10
2
10
3
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
(b)
Figure 2.13 Relationships between Reynolds number, Strouhal number and drag coef-
cient. (a) shows the relation between Strouhal number, which is related to the vortex shedding
frequency, and the Reynolds number (Williamson, 1998). The dot () shows a Strouhal number of
St = 0.183 for Re = 150 at which the stationary cylinder validation case is performed. (b) relates the
viscous drag coecient (), pressure drag coecient () and total drag coecient () to the Reynolds
number (Henderson, 1995).
Figure 2.14 Computational grid around a cylinder. This grid, with sizes 25k, 50k and 100k is
used to validate the accuracy of the ow solver. The ow is from left to right and the inlet boundary
is located 10D upstream, the outlet 40D downstream and the upper and lower boundaries are located
10D from the cylinder surface, where D is the cylinder diameter.
boundary is located at a distance of 40D. Previous studies, e.g. Lentink & Ger-
ritsma (2003), Bos et al. (2008), showed that boundary eects are minimal at this
domain size.
Flow around a stationary circular cylinder
The rst case, dealing with the ow around a static cylinder at Re = 150 is
inherently laminar and unsteady, resulting in a periodic vortex wake. Henderson
(1995) performed a spectral element numerical study which is used as the baseline
reference for this case. Figure 2.13(a) shows the relation between the Strouhal
and Reynolds number, according to (Williamson, 1998). A Strouhal number of
St = 0.183 is obtained for a stationary cylinder at Re = 150. Additionally,
gure 2.13(b) shows the results from an extensive study performed by Henderson
(1995) to identify a relation between drag coecient and Reynolds number. The
42 Finite volume discretisation
Figure 2.15 Vorticity visualisation of the Von Karman vortex street. The ow around a
stationary cylinder shows a periodic vortex street, of which the frequency depends on the Reynolds
number. Vorticity = u is used to identify the vortical structures, which are clearly visible at a
Reynolds number of Re = 150.
resulting drag coecient at Re = 150 is found to be C
D
= 1.333. The drag and
lift coecient are respectively dened as:
C
D
=
D
1
2
U
2
ref
, C
L
=
L
1
2
U
2
ref
. (2.29)
In order to investigate the temporal and spatial convergence of the solution, the
grid size is varied from 25k, 50k and 100k. The time-step is systematically
decreased according to a maximal Courant number corresponding to Co
max
=
2.0, 1.0, 0.5 and 0.25.
Results
To illustrate the ow behaviour, gure 2.15 shows the instantaneous vorticity
( = u) contours, which reveals the presence of the Von K arm an vortex street
behind the stationary cylinder. In practical applications such a vortex street ex-
ists behind struts in water, for example. The alternating vortex shedding pattern
leads to periodic force variations which can be visualised using C
L
-C
D
limit cycles.
The results, obtained for dierent mesh resolutions (25k, 50k and 100k) are shown
in gure 2.16. These limit cycles are determined by taking the periodic part of
the force histories as shown in gure 2.19. From these periodic forces, the time-
averaged drag coecient is determined and compared with literature in table 2.1.
From this table, it can be observed that the drag coecient of the coarsest case,
25k and Co
max
= 2.0, has the largest dierence with literature, 5.63%. When the
grid is rened and the time-step decreased, it is seen that the solution decreases
asymptotically. The drag on the nest grid with the smallest time-step is about
2.7% larger compared to the value obtained by Henderson (1995). In addition,
the calculated time-averaged Strouhal number (shedding frequency) is shown in
table 2.2. From that table, it can be seen that the Strouhal number matches the
value from Henderson (1995) even more closely than the time-averaged drag coef-
cient. The dierences of the mean Strouhal number with literature ranges from
2.10 Code validation and verication 43
Drag coecient [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
Grid Size=25k
Grid Size=50k
Grid Size=100k
1.34 1.36 1.38 1.4 1.42 1.44 1.46
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
(a) Co
max
= 2.0
Drag coecient [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
Grid Size=25k
Grid Size=50k
Grid Size=100k
1.34 1.36 1.38 1.4 1.42 1.44 1.46
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
(b) Co
max
= 1.0
Figure 2.16 Lift-Drag limit cycles of the ow around a stationary cylinder. The periodicity
of the ow around a stationary circular cylinder at Re = 150 is illustrated by a C
L
-C
D
limit cycle.
Additionally, the grid convergence of the solution can be observed for Co
max
= 2.0, (a), and Co
max
=
1.0 (b).
Drag coecient, C
D
Mesh size Co
max
2.0 1.0 0.5 0.25 Richardson
25k 1.408 1.393 1.385 1.381 1.379 (+3.5%)
50k 1.392 1.381 1.376 1.373 1.372 (+2.9%)
100k 1.385 1.377 1.372 1.370 1.369 (+2.7%)
Henderson (1995) 1.333
Table 2.1 Drag comparison for the ow around a stationary cylinder. Comparison of
the time-averaged drag coecient for dierent grids, 25k, 50k and 100k, and dierent time-steps
corresponding to a maximal Courant number, Co
max
= 2.0, 1.0, 0.5 and 0.25.
3.28% to 1.64%, which is considered suciently small.
From table 2.1 and 2.2, it is clear that the ow solver produces results which
are suciently close to the values obtained from literature. Besides a comparison
solely with literature it is important to investigate the convergence of the solution
with increasing grid resolution and decreasing time-step size. If the ow solver is
developed in a numerically consistent way, the ow solution should converge to
an asymptotically value with increasing spatial and temporal resolution. In order
to illustrate if the solution converges, gure 2.17 shows the time-averaged drag
coecient for increasing mesh resolution (for each time-step) and for decreasing
time-step (for each mesh). As can be seen, the solution decreases asymptotically,
which should be the case. The last value in these two gures, is the extrapolated
values, using Richardsons extrapolation (Ferziger & Peric, 2002). The extrapo-
44 Finite volume discretisation
Spatial resolution
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
Co
max
= 2.0
Co
max
= 1.0
Co
max
= 0.5
Co
max
= 0.25
1 2 4
1.36
1.37
1.38
1.39
1.4
1.41
(a)
Temporal resolution
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
Grid Size=25k
Grid Size=50k
Grid Size=100k
1 2 4 8
1.36
1.37
1.38
1.39
1.4
1.41
(b)
Figure 2.17 Time-averaged drag coecient as a function of spatial and temporal resolution
for the stationary cylinder. A ow solver is numerically consistent if the ow solution converges
with increasing grid resolution and decreasing time-step size. (a) shows the drag coecient with
increasing grid renement level for dierent time-steps. (b) shows the drag coecient with decreasing
time-step size for the dierent grid sizes. The Richardson extrapolated values are plotted at .
lated value is obtained using the following expression:
extrap
=
fine
+
fine
coarse
2
p
1
, (2.30)
where
extrap
is the extrapolated value,
fine
and
coarse
are the two most accurate
solutions available. Theoretically, the order of the scheme p can be obtained
using (2.30) and should be 2 for a second-order discretisation scheme. This is true
for uniform Cartesian meshes (Ferziger & Peric, 2002) which is not the case for
the cylinder simulations. However, from table 2.1 and 2.2 it can be deduced that
for both second-order spatial and temporal schemes, the value of p lies between
1.5 and 2.
Strouhal number, St
Mesh size Co
max
2.0 1.0 0.5 0.25
25k 0.186 (+1.64%) 0.187 0.188 0.188 (+2.73%)
50k 0.187 (+2.19%) 0.188 0.188 0.189 (+3.28%)
100k 0.188 (+2.73%) 0.188 0.188 0.189 (+3.28%)
Henderson (1995) 0.183
Table 2.2 Strouhal number comparison for the ow around a stationary cylinder. Com-
parison of the time-averaged Strouhal number for dierent grids, 25k, 50k and 100k, and dierent
time-steps corresponding to a maximal Courant number, Co
max
= 2.0, 1.0, 0.5 and 0.25.
2.10 Code validation and verication 45
Spatial resolution
L
i
f
t
c
o
e
c
i
e
n
t
a
m
p
l
i
t
u
d
e
[
-
] Co
max
= 2.0
Co
max
= 1.0
Co
max
= 0.5
Co
max
= 0.25
1 2 4
0.54
0.56
0.58
0.6
0.62
0.64
(a)
Temporal resolution
L
i
f
t
c
o
e
c
i
e
n
t
a
m
p
l
i
t
u
d
e
[
-
]
Grid Size=25k
Grid Size=50k
Grid Size=100k
1 2 4 8
0.54
0.56
0.58
0.6
0.62
0.64
(b)
Figure 2.18 Time-averaged lift convergence with grid resolution and time-step size for the
stationary cylinder. (a) shows the average lift coecient amplitude with increasing grid renement
level for dierent time-steps. (b) shows the time-averaged lift coecient amplitude with decreasing
time-step for the dierent grid sizes.
Time [s]
F
o
r
c
e
c
o
e
c
i
e
n
t
[
-
]
Drag
Lift
Co
max
= 2.0
Co
max
= 1.0
Co
max
= 0.5
Co
max
= 0.25
0 50 100 150
-0.5
0
0.5
1
1.5
Figure 2.19 Forces around a stationary cylinder, 25k. Lift and drag coecients for a stationary
circular cylinder case at Re = 150, the grid size is 25k and the time-step was varied according to a
maximal Courant number of Co
max
= 2.0, 1.0, 0.5 and 0.25.
Additionally to the temporal and spatial convergence of the drag coecient,
gure 2.18 shows the convergence of the lift coecient amplitude in order to prove
the consistency of the ow solver. Concerning the extrapolated values of the drag
coecient, from table 2.1, it can be determined that the dierences varies from
2.7% to 3.5% compared to literature with decreasing mesh resolution. For all
meshes, the dierences in drag coecient compared to the extrapolated values,
are smaller than 1.0% for maximal Courant numbers Co
max
= 1.0, 0.5 and 0.25.
Therefore, it seems sucient to consider the mesh of 50k and Co
max
= 1.0 to result
in an accurate solution, these settings were used throughout the present research.
Flow around a transversely oscillating circular cylinder
The last validation case concerns the ow around a transversely oscillating cylinder
46 Finite volume discretisation
at Re = 185 using a numerical study, performed by Guilmineau & Queutey (2002).
The oscillating direction is perpendicular to the free-stream direction. The cylinder
motion is dened as
y(t) = A
e
sin(2f
e
t), (2.31)
where the amplitude is set to A
e
= 0.2D, with D the cylinder diameter. The
frequency was set to f
e
= 0.154, corresponding to 0.8 times the natural shedding
frequency of a stationary cylinder at a Reynolds number Re = 185. An amplitude
of 0.2D is relatively small for insect aerodynamics, which employs amplitudes of
several chord lengths, but sucient to investigate the moving wing capabilities
of the numerical model. Previously conducted simulations on stationary cylinder
ow showed that a mesh of 50k provides a suciently accurate solution, therefore
that mesh is also used for this test case. Although, a time-step corresponding to
Co
max
= 1.0 was found to be sucient, the following values are used to show that
the ow solver converges to an asymptotic solution for an oscillating cylinder as
well, Co
max
= 2.0, 1.0, 0.5 and 0.25 were considered.
Results
To assess the accuracy of the ow solver applied to apping wings, the results of
an oscillating cylinder case were compared with literature. Guilmineau & Queutey
(2002) found a drag coecient of C
D
= 1.2. Figure 2.20(a) shows the limit cy-
cle results, using the present ow solver, with second-order temporal and spatial
discretisation. From this gure it is obvious that the ow solution converges with
decreasing time-step. This statement is conrmed if the time-averaged drag coef-
cient is plotted in gure 2.20(b), for the 50k grid. The value for all time-steps
was within 2% compared to the extrapolated value, which is considered to be
suciently accurate.
2.11 Conclusions
This chapter has presented the nite volume discretisation of the incompressible
laminar Navier-Stokes equations. The discretisation concerns arbitrary polyhedral
meshes, such that this method can easily be applied to a wide variety of problems
with complex geometries. To obtain accurate and ecient results, the mesh qual-
ity should be high in terms of non-orthogonality and skewness, both mesh quality
measures. The dierent terms of the governing equation were discretised using
second-order schemes and dierent ux splitting methods were described concern-
ing the face interpolation. In order to solve the ow on a computational grid, four
dierent types of boundary conditions were specied, xed-value (Dirichlet), zero-
gradient (Neumann), symmetry and moving-wall-velocity. Using those boundary
conditions, the discretised Navier-Stokes equations can be solved using a PISO
pressure velocity coupling in combination with an Arbitrary Lagrangian Eulerian
(ALE) approach if dynamic meshes are used.
2.11 Conclusions 47
Drag coecient [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
Co = 2.0
Co = 1.0
Co = 0.5
Co = 0.25
1.2 1.25 1.3 1.35
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
(a)
Temporal resolution
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
1 2 4 8
1.255
1.26
1.265
1.27
1.275
1.28
(b)
Figure 2.20 Forces for the ow around an oscillating cylinder. (a) shows the C
D
-C
L
limit
cycles for dierent time-steps on a 50k mesh. (b) shows the time-averaged drag coecient with
decreasing time-step for the 50k mesh. The Richardson extrapolated values are plotted at .
In principle, the described discretisation method and solution procedure is ap-
plicable to dierent commercial and non-commercial CFD solvers. The present
research used the commercial CFD solver Fluent
community.
These dierent mesh motion methods are described in section 3.2. In order to
assess the mesh quality, section 3.3 discusses two dierent criteria, based on non-
orthogonality and skewness, described in chapter 2. The resulting mesh quality
of the dierent mesh motion methods is studied using a two-dimensional test case
of a block which translates and rotates. The mesh quality is investigated using
a visualisation and histograms of the skewness and non-orthogonality criterion.
This is the subject of section 3.4. In addition to the simplied moving block, two
more relevant test cases were considered, one using a three-dimensional apping
wing and the other involves a two-dimensional exing airfoil. Since the radial ba-
sis function mesh motion method is computationally expensive, section 3.5 deals
with two techniques to increase its eciency. Finally, the conclusions are drawn
in section 3.6.
3.2 Dierent mesh deformation techniques
When a moving mesh problem is considered, the shape of the computational do-
main is varying in time. Therefore, a distinction can be made between the motion
of the boundary points and the motion of the internal (uid) points. The displace-
ment of the boundary points can be considered to be given, either it is externally
dened, i.e. a prescribed rigid body motion, or it is part of the solution, which is
3.2 Dierent mesh deformation techniques 53
the case in uid-structure interaction problems. According to the given boundary
point motion, the internal points need to be moved in order to maintain mesh qual-
ity and validity. The internal point motion inuences the solution only through
the discretisation errors (Ferziger & Peric, 2002), provided that the ALE formula-
tion is correctly implemented. The internal point motion can be calculated using
dierent methods, as will be shown in the next section.
3.2.1 Laplace equation with variable diusivity
One can think of a deforming computational domain as if it was a solid body under-
going internal stresses given by the Piola-Kirchho stress-strain equation (Baruh,
1999). That equation is non-linear and thus expensive to solve using existing
numerical techniques. Therefore, other type of equations were used, namely the
Laplace equation and the solid body rotation stress (SBR Stress) equation, which
is a variant of the linear stress equation (Dwight, 2004). A mesh motion method
based on one of these equations is computationally cheap since the resulting matrix
system is sparse, such that existing iterative solvers can be used eciently.
When the mesh motion is governed by the Laplace equation, the given bound-
ary point motion may be arbitrary and non-uniform. The nature of the Laplace
equation is that the point displacements will be largest close to the moving bound-
ary and small at large distance. Ideally, a user input is not desired, since it
decreases the robustness of the method. However, this method needs the speci-
cation of a variable diusivity. This leads to the following denition of the Laplace
equation:
(x) = 0,
where x is the displacement eld and the diusion coecient, which decreases
with the radius r from the deforming boundary as follows:
(r) =
1
r
m
. (3.1)
The resulting mesh quality strongly depends on the chosen (r) function, which
depends on the distance from the moving boundary. This variable diusion coe-
cient can be chosen such that a region next to the deforming or moving boundary
closely moves with the boundary. The resulting mesh contains less cell quality
deterioration next to the boundary. The current research uses a (r) function like
equation (3.1). In addition to the freedom of choosing a diusion function, it is also
possible to dene (r) for every internal mesh cell for all time-steps independently.
This, however, appears to be very problem dependent and thus optimisation of
(r) seems not cost eective. To maintain robustness, in the current work we use a
quadratically, m = 2, decreasing diusion coecient, which was found to provide
ecient and a smooth mesh motion (Jasak & Tukovic, 2004). Additionally, one
could also have used an exponentially decreasing diusion coecient or a diusion
coecient related to the mesh deformation energy.
54 Mesh deformation techniques for apping ight
3.2.2 Solid body rotation stress equation
The second method to deform the mesh is based on the linear elasticity equation
and is called the solid body rotation stress (SBR Stress) equation (Dwight, 2004).
The equation of linear elasticity, valid for small displacements, may be written as
= f , (3.2)
where is the stress tensor and f the acting force vector. The stress tensor is
given in terms of the strain, which is given by the following constitutive relation:
= tr ()I + 2, (3.3)
where tr is the trace and and are Lame constants (Baruh, 1999), which are a
property of the elastic material. The constants can be related to Youngs modulus,
E, as
=
E
(1 +)(1 2)
, =
E
2(1 +)
,
where is Poissons ratio, meaning the material contraction ratio as it stretches.
The following equation:
=
1
2
(x +x
T
), (3.4)
denes the relative change in length, where x is the position of an internal mesh
point, which is treated as if it was a linear solid. Although equation (3.4) does
not allow for rotations, there is nothing against changing this strain equation such
that rigid body rotations are allowed. In Dwight (2004) an extra term was added
to obtain the following strain relation:
=
1
2
(x +x
T
+x
T
x). (3.5)
Combining equations (3.5), (3.3) and (3.2), together with = E and = E, the
following solid body rotation stress equation is obtained:
(x) +((x x
T
)) tr(x) = 0, (3.6)
where is a similar diusion coecient as in equation (3.1). Equation (3.6) allows
for rigid body motion and is still linear and therefore the computational costs are
of the same order as the costs necessary to solve the Laplace equation.
Solving the Laplace or the SBR Stress equation leads to a sparse system of equa-
tions, such that standard iterative techniques can be used, like the pre-conditioned
Conjugate Gradient (PCG) method. However, it is also possible to explicitly de-
ne the point motion using interpolation techniques, like the transnite interpo-
lation (Wang & Przekwas, 1994) usually applied to the points of multi-blocks.
In section 3.4 it is shown that both previously described methods maintain high
mesh quality for problems with limited boundary rotation. In order to deal with
large rotations, a newly implemented mesh motion solver is based on radial basis
function interpolation, such that it can be used for apping wing simulations.
3.2 Dierent mesh deformation techniques 55
3.2.3 Radial basis function interpolation
In the current work we use radial basis function interpolation to nd the dis-
placements of the internal uid points for given boundary displacements. The
interpolation function s(x) describing the displacement of all computational mesh
points, is approximated by a sum of basis functions:
s(x) =
N
b
j=1
j
(||x x
b
j
||) +q(x), (3.7)
where the known boundary value displacements are given by x
b
j
= [x
b
j
, y
b
j
, z
b
j
],
q is a polynomial, N
b
is the number of boundary points and is a given basis
function as a function of the Euclidean distance ||x||. The minimal degree of
polynomial q depends on the choice of the basis function (Boer de et al., 2007).
A unique interpolant is given if the basis function is a conditionally positive denite
function. If the basis functions are conditionally positive denite of order m 2, a
linear polynomial can be used (Beckert & Wendland, 2001). We only applied basis
functions that satisfy this criterion. A consequence of using a linear polynomial is
that rigid body translations are exactly recovered. The polynomial q is dened by
the coecients
j
which can be dened by evaluating the interpolation function
s(x) in the known boundary points:
s(x
b
j
) = x
b
j
.
Here x
b
j
contains the known discrete values of the boundary point displacements.
Together with the additional requirements:
N
b
j=1
j
p(x
b
j
) = 0,
which holds for all polynomials p with a degree less or equal than that of polynomial
q, the
j
values can be determined (Boer de et al., 2007).
The values for the coecients
j
and the linear polynomial can be obtained by
solving the system:
_
x
b
0
_
=
_
bb
Q
b
Q
T
b
0
_ _
_
, (3.8)
where is containing all coecients
j
, the four coecients of the linear poly-
nomial q,
bb
an n
b
n
b
matrix contains the evaluation of the basis function
b
i
b
j
= (||x
b
i
x
b
j
||) and can be seen as a connectivity matrix connecting all
boundary points with all internal uid points. Q
b
is an (n
b
(d + 1)) matrix
with row j given by [ 1 x
b
j
]. In general, (3.8) leads to a dense matrix system,
which is dicult to solve using standard iterative techniques. Therefore, it needs
to be solved directly, by doing a LU decomposition. The possibilities of solving
the system in a more ecient way are discussed in section 3.5.
56 Mesh deformation techniques for apping ight
When the coecients in and are obtained they are used to calculate the
values for the displacements of all internal uid points x
in
j
using the evaluation
function (3.7),
x
in
j
= s(x
in
j
). (3.9)
The result of (3.9) is transferred to the mesh motion solver to update all internal
points accordingly. This interpolation function is equal to the displacement of the
moving boundary or zero at the outer boundaries. Every internal mesh point is
moved based on its calculated displacement, such that no mesh connectivity is
necessary. The size of the system of this method (3.8) is ((N
b
+ 4) x (N
b
+ 4)),
which is considerably smaller than other techniques using the mesh connectivity.
The mesh connectivity techniques encounter systems of the order (N
in
x N
in
),
with N
in
the total number of mesh points, which is a dimension higher than the
total number of boundary points. Solving the system (3.8) gives the values of the
necessary coecients and , which are then used for step two, the evaluation
using equation (3.9).
In contrast to the Laplace and SBR Stress methods, no partial dierential
equation needs to be solved and the evaluation of all internal boundary points is
straightforward to implement in parallel, since no mesh connectivity is needed.
Concerning robustness, this method is not using a variable diusion coecient
which has to be tuned by the user. Instead, the proper radial basis function needs
to be chosen to satisfy the need for robustness.
Radial basis functions with compact support
From literature, various radial basis functions are available, which are suitable for
data interpolation. Two types of radial basis functions can be distinguished: func-
tions with compact and functions with global support. Functions with compact
support have the following property:
(x/r) =
_
f(x/r) 0 x r,
0 x > r,
where f(x/r) 0 is scaled with a support radius r. When a support radius is used,
only the internal mesh points inside a circle (two-dimensional problem) or a sphere
(three-dimensional problem) with radius r around a centre x
j
are inuenced by
the movement of the boundary points. When choosing r, it must be noted that
larger values for the support radius lead to more accurate mesh motion. On the
other hand, a very large support radius leads to a dense matrix system, while a
low support radius results in a sparse system which can be solved eciently using
common iterative techniques.
In table 3.1 various radial basis functions with compact support are shown
using the scaled variable = x/r. The rst four are based on polynomials (Wend-
land, 1996). These polynomials are chosen in such a way that they have the
lowest degree of all polynomials that create a C
n
continuous basis function with
n {0, 2, 4, 6}. The last four are a series of functions based on the thin plate spline
3.2 Dierent mesh deformation techniques 57
Ref. nr. RBF Name f()
1 CP C
0
(1 )
2
2 CP C
2
(1 )
4
(4 + 1)
3 CP C
4
(1 )
6
(
35
3
2
+ 6 + 1)
4 CP C
6
(1 )
8
(32
3
+ 25
2
+ 8 + 1)
5 CTPS C
0
(1 )
5
6 CTPS C
1
1 +
80
3
2
40
3
+ 15
4
8
3
5
+ 20
2
log()
7 CTPS C
2
a
1 30
2
10
3
+ 45
4
6
5
60
3
log()
8 CTPS C
2
b
1 20
2
+ 80
3
45
4
16
5
+ 60
4
log()
Table 3.1 Radial basis functions with compact support. Radial basis functions with compact
support are non-zero within the range of the support radius r. Note that = x/r. Taken from
Wendland (1996).
which creates C
n
continuous basis functions with n {0, 1, 2} (Wendland, 1996).
There are two possible CTPS C
2
continuous functions which are distinguished by
subscript a and b.
Radial basis functions with global support
In contrast to the functions with compact support, functions with global support
are not equal to zero outside a certain radius, but cover the whole interpolation
space. Radial basis functions with global support generally lead to dense matrix
systems, which can be improved by multiplication with a smoothing function, as
will be discussed in section 3.5.
Table 3.2 shows six radial basis functions with global support which are com-
monly used in e.g. neural networks, computer graphics (Carr et al., 2003) and
for data transfer in uid-structure interaction computations (Smith et al., 2000,
Boer de et al., 2007).
Ref. nr. RBF Name Abbrev. f(x)
9 Thin plate spline TPS x
2
log(x)
10 Multiquadratic Bi-harmonics MQB
a
2
+x
2
11 Inverse Multiquadratic Bi-harmonics IMQB
_
1
a
2
+x
2
12 Quadric Bi-harmonics QB 1 +x
2
13 Inverse Quadric Bi-harmonics IQB
1
1+x
2
14 Gaussian Gauss e
x
2
Table 3.2 Radial basis functions with global support. Radial basis functions with global
support cover the whole interpolation space, i.e. the computational domain.
58 Mesh deformation techniques for apping ight
The MQB and IMQB methods use a shape parameter a, which controls the
shape of the radial basis function. A large value of a gives a at sheetlike function,
whereas a small value of a gives a narrow cone-like function. The value of a is
typically chosen in the range 10
5
10
3
. More information about RBFs and
their error and convergence properties can be found in (Buhmann, 2000, Wend-
land, 1999, 1998). Boer de et al. (2007) compared the resulting mesh quality using
all radial basis functions from table 3.1 and 3.2. The best results were obtained
using the continuous polynomial C
2
, the functions based on a continuous thin
plate spline, C
1
, C
2
a
and C
2
b
and nally the globally supported thin plate spline.
In section 3.4 these six radial basis functions are applied to our test problem and
the best one is used throughout the current thesis. First, the mesh quality mea-
sures, skewness and non-orthogonality, are discussed to compare the mesh quality
for the dierent mesh motion solvers.
Absolute and relative radial basis function interpolation
In principle there are two dierent ways to implement this RBF mesh motion
method, the absolute and the relative implementation. The absolute method per-
forms a direct solve of the system (3.8) only once at the beginning of the simulation.
The coecient arrays and are calculated and used to calculate the internal
point displacements at all time-steps. This method is very ecient since the di-
rect matrix solve, which is more expensive than the evaluation, is only performed
initially. On the other hand, the mesh quality is limited since the coecients are
not dened with respect to the previous time-step. Therefore, the relative method
is used when very large boundary displacements occur, like a 180
rotation.In this
method the inverse is calculated at every time-step and the motion is dened with
respect to the previous time-step. For reasonably small boundary displacements,
it is much cheaper to use the absolute method. When using the relative imple-
mentation it is important to use dierent techniques to decrease the number of
boundary points, resulting in lower computation costs (see section 3.5).
3.3 Mesh quality measures
In section 2.4 the skewness and non-orthogonality denition were introduced. In
order to compare the quality of the dierent meshes after mesh motion it is impor-
tant to elaborate on how to interpret those two mesh quality measures (Knupp,
2003). These mesh quality measures are based on the cell properties such as size,
orientation, shape and skewness. The skewness and non-orthogonality are written
to scalar elds such that they can be used for post-processing.
The test cases used to compare the mesh quality of the dierent mesh motion
solvers, contain a Cartesian grid around a square box, leading to optimal initial
mesh quality, this is shown in section 3.4. It is important that the ideal mesh
motion solver maintains high quality in terms of skewness and non-orthogonality
after mesh deformation.
3.4 Comparison of mesh motion solvers 59
In section 2.4 it is shown that mesh skewness should be within 0 and 1, and
that the mesh non-orthogonality, which is an angle, should be within 0
and 90
.
In both cases, a lower value means a higher mesh quality. Therefore, desirable
mesh quality bounds are:
0.0 f
skewness
1.0
0
f
nonortho
90
(3.10)
When assessing the mesh quality it is important to analyse the maximal and aver-
age values. The maximal value provides an indication if the numerical simulation
will be stable and converge at all. If the worst cell quality is too low, the simulation
will diverge within a couple of iterations. On the other hand, the average value
of the mesh quality measure will provide an indication of the average quality of
the mesh. The higher the average quality of the mesh, the more stable, accurate
and ecient the computation will be. In the next section, both the average and
minimal value of the skewness and non-orthogonality mesh quality measures are
used to compare the mesh motion solvers.
3.4 Comparison of mesh motion solvers
In section 3.2, three dierent kinds of mesh motion solvers are described, based on
solving the Laplace equation, solving the solid body rotation (SBR Stress) equa-
tion and based on interpolation using radial basis functions (RBF). This section
introduces with three numerical test cases which were performed to investigate the
dierences in mesh quality obtained with the dierent mesh motion solvers. Addi-
tionally, the eect of dierent radial basis functions is discussed. The rst simple
case is a two-dimensional block which performs a combination of translation and
rotation. The initial computational mesh is shown in gure 3.1. The domain size
of the test problem is limited to 25D x 25D and the size of the moving block
is 5D x 1D, the grid spacing corresponds to 1D in order to obtain a Cartesian
grid as can be seen in the gure. Mesh motion simulations are performed using
the dierent mesh motion solvers and a variation of the radial basis function. Af-
ter this simple model problem, section 3.4.2 deals with the mesh motion around
a three-dimensional apping wing, followed by an example of a two-dimensional
exible moving boundary in section 3.4.3.
3.4.1 Translation and rotation of a two-dimensional block
The rst test case considers a combined motion of translation and rotation to com-
pare the mesh motion solvers under these conditions. The two-dimensional block
is initially centred and translates 2.5D in both X-and Y -direction. In addition,
the block is rotated around its translating centre with 57.3
, and the
mesh remains valid. This method is very robust but computationally expensive as
will be shown in section 3.5.
Table 3.3 shows a quantitative comparison of the resulting mesh quality ob-
tained with dierent mesh motion solvers, including dierent RBFs. The max-
imal and averaged values of both skewness and non-orthogonality are compared.
The mesh quality calculated by the Laplace method is low, since the maximal
skewness and maximal non-orthogonality are large, respectively f
s
max
= 0.95 and
f
o
max
= 72.1. The results of the Laplace and SBR Stress mesh motion solver are
very similar, within 6.1%. Concerning the RBF method, four RBFs with compact
support are used (CP C
2
, CTPS C
1
, CTPS C
2
a
and CTPS C
2
b
) and one with global
support, the thin plate spline TPS. In order to neglect the eect of the support
62 Mesh deformation techniques for apping ight
(a) RBF absolute (b) RBF relative
Figure 3.3 Cell non-orthogonality of relative and absolute RBF mesh motion solvers.
The cell non-orthogonality is shown for the relative and absolute versions of the RBF mesh motion
solver. (a) shows the absolute and (b) the relative implementation, lower (blue) is better. The non-
orthogonality is visualised for a two-dimensional block with a combined motion of translation and
rotation. The boundary points translate a distance of 2.5D in both X- and Y -direction. The rotation
is 57.3
sin(2ft),
where, A
t
and A
d
c
2 c
f
) sin(2ft)
d
f
,
where A
f
is the exing amplitude vector, d
c
is the direction vector in-plane of the
exing surface, c
f
the length of the exing surface and d
f
represents the direction
vector of the exing. In this model problem of a moving two-dimensional block, the
amplitudes of both translation and rotation were xed to A
t
= (2.5, 2.5, 0.0) and
A
= (0.0, 0.0, 1.0), respectively. The exing was dened such that the main ex-
ing direction is perpendicular to the exible boundary surface, d
f
= (0.0, 1.0, 0.0)
with a exing amplitude of A
f
= 0.5 which is about 10% of the exible boundary
length, c
f
= 5.0. Figure 3.7 shows the resulting mesh deformation at t = 0.25T
and t = 0.75T where T = 1/f is the motion period. It can be seen that the whole
mesh is deformed by the RBF mesh motion solver, like was the case with a rigid
airfoil. Still, some high non-orthogonality can be observed within a region of 12
block lengths, which is mainly caused by the xed points on the outer boundary
and the small computational domain. A larger domain will undoubtedly lead to
high quality meshes when using RBF mesh motion.
66 Mesh deformation techniques for apping ight
(a) (b)
Figure 3.7 Deforming mesh around a exible block using RBF mesh motion. A exible
rectangular block which is translating, rotating and exing using the RBF mesh motion solver with
the globally supported thin plate spline (TPS). The exing is dened by simple harmonic functions.
(a) shows the mesh deformation at t = 0.25T and (b) at t = 0.75T.
3.5 Improving computational eciency
When using direct methods, the computational costs of the new mesh motion solver
based on radial basis function interpolation, increases fast with an increased num-
ber of boundary and internal points. The method consists of two computationally
expensive steps:
1. Solving the system of equations (3.8) for given boundary points x
b
i
and cor-
responding displacements x
b
i
to nd the coecient vectors and . The
upper diagonal block matrix
bb
is in general a dense symmetric matrix of
size (N
b
x N
b
). Therefore, standard direct solvers require O(N
3
b
) operations.
2. Evaluation of the radial basis function summation, equation (3.7), at all N
in
(of order O(N
in
)) internal mesh points, using the given boundary points x
b
i
and the in step 1 computed coecient vectors and . This evaluation
leads to a computational cost of order O(N
b
N
in
).
For large two- and three-dimensional meshes both the system solve and the eval-
uation procedures may become very computationally expensive, especially when
direct methods are used. In order to illustrate the increasing costs with increasing
number of mesh points, a two- and three-dimensional Cartesian grid is considered
with a uniform mesh distribution. The computational domain is square shaped
with an equal number of points on all edges of the boundary, equal to N
b
. Table 3.4
shows the total number of internal and boundary points as a function of N
b
, for
both the two- and three-dimensional example. Additionally, the total number of
operations for a direct solve and the RBF evaluation is given in the table, which
3.5 Improving computational eciency 67
Internal points All boundary points Direct solve RBF evaluation
2D N
2
b
4N
b
64N
3
b
4N
3
b
3D N
3
b
6N
2
b
216N
6
b
6N
5
b
Table 3.4 Computational costs for solving the system and evaluate the RBFs. Computa-
tional costs for solving the system and evaluation the RBFs on all internal mesh points. Illustration of
a two-dimensional and three-dimensional uniform square shaped Cartesian mesh with an equal number
of boundary nodes N
b
in x- and y-direction.
shows that the computational costs for both a direct solve and evaluation scales
with N
3
b
for the two-dimensional case. Concerning the three-dimensional case, the
computational costs for the direct system solve are a factor N
b
larger compared
to the costs for the RBF evaluation. So when a complex three-dimensional case,
like a apping wing, is considered, special treatment is necessary concerning the
system solve, e.g. by reducing the number of boundary nodes or using advanced
direct solver techniques.
3.5.1 Boundary point coarsening and smoothing
From table 3.4 it is seen that the total computation costs will decrease if a con-
straint is put on the number of mesh points. This can be achieved in two dierent
ways. First, a major part of the computational cost is spent at solving the system,
step 1, which is a factor N
b
more expensive than the evaluation. It seems rea-
sonable to reduce the number of moving boundary points by performing a sound
selection procedure. Especially when the body displacement follows a rigid body
motion, not all boundary points are necessary. Therefore, a coarsening technique
was incorporated which selects a boundary point for every points, where is
the coarsening factor. More advanced coarsening techniques, based on greedy
algorithms are applied by Rendall & Allen (2008a).
Secondly, the eciency of the RBF method is improved, by the notion that
all outer boundary points are xed in general. Therefore, the outer boundary
points can be neglected. This is achieved by specifying a smoothing function
such that the RBF contribution reduces to zero at the outer boundary, which is
dened (Jakobsson & Amoignon, 2007) as
( x) =
_
_
_
1, x 0,
1 x
2
(3 2 x), 0 x 1,
0, x 1,
(3.11)
where, x is given by:
x =
||x
i
R
inner
||
R
outer
R
inner
, (3.12)
x
i
represents the coordinate of the ith inner mesh point, to evaluate this smooth-
ing function at that particular location in space. R
inner
and R
outer
are two radii,
68 Mesh deformation techniques for apping ight
between those radii this smoothing function is decreasing from 1.0 to 0.0. When
the RBF evaluation function (3.9) is multiplied by ( x) the corrected RBF eval-
uation is obtained by:
x
in
j
= s(x
in
j
)
( x). (3.13)
Within R
inner
the contribution of the RBF evaluation remains unaltered, but out-
side R
outer
the value of the RBF becomes zero, such that all xed outer boundary
points may be neglected. In principle, the inner radius is chosen to be multiple
boundary lengths (wing chords) and the outer radius is chosen as the distance
from the moving boundary to the outer boundary. Therefore, the system to be
solved, only contains the control points on the moving boundary, selected by the
coarsening function.
In addition, it is interesting to address the computing times concerning the
dierent mesh motion solvers. Figure 3.8(a) shows the computing times with in-
creasing grid resolution for the Laplace, SBR Stress and RBF mesh motion solvers.
Concerning the RBF mesh motion solvers, the three described variants are used,
the absolute implementation, relative implementation and the relative method in
combination with the previously dealt coarsening and smoothing techniques. It
is clear that the mesh motion solvers based on solving a partial dierential equa-
tion are very fast, since standard iterative techniques can be used for these sparse
systems. If the absolute and relative RBF methods are compared, it is observed
that these methods require very large computing times, at least an order of mag-
nitude more. On the other hand, if the coarsening and smoothing techniques are
applied the computing times are of similar order compared with the fast Laplace
mesh motion. The non-linear behaviour of the nal curve is caused by the choice
of the coarsening ratio to select the moving boundary points. While keeping the
coarsening ratio xed, the mesh resolution is increased, the number of boundary
points used in the system solving is still growing non-linearly. For high resolution
meshes, the order of computing times can be decreased further by increasing the
coarsening ratio.
Figure 3.8(b) shows the computing times, per time-step, concerning a three-
dimensional apping wing simulation (6 apping periods). These three-dimension-
al simulations are performed for grid sizes of 100k, 200k, 400k, 800k and 1600k
cells and needed a total computing time of about 8, 32, 48, 103, 190 hours, respec-
tively. All simulations were performed on four CPU cores of an AMD Opteron
280 cluster. Figure 3.8(b) shows super-linear curves for both solving the ow
equations and the RBF mesh motion. Validation showed that a mesh resolution
of 800k provided accurate results for apping wing aerodynamics. Additionally,
gure 3.8(b) shows that the computing time used for RBF mesh motion is less
than 10% of the computing time for the complete time-step. This is considered to
be very ecient, it must be noted that mesh coarsening and a smoothing function
are applied for acceleration.
3.5 Improving computational eciency 69
Grid spacing
C
o
m
p
u
t
i
n
g
t
i
m
e
[
s
]
10
0
10
1
10
2
10
0
10
1
10
2
10
3
10
4
(a) Two-dimensional
Grid spacing
C
o
m
p
u
t
i
n
g
t
i
m
e
[
s
]
Mesh motion
Flow equation
Total timestep
0 1 2 4 8 16
0
10
20
30
40
50
60
70
(b) Three-dimensional
Figure 3.8 Computing times for dierent mesh motion solvers. (a) shows a comparison
of computing times for the mesh motion solvers based on the Laplace equation (), the SBR Stress
equation (), RBF absolute method (), RBF relative method () and RBF relative method in combi-
nation with coarsening and smoothing techniques (). The computing times for the Laplace equation
() and the SBR Stress equation () are nearly identical. (b) shows the computing times for one
time-step of a three-dimensional apping wing simulation. The times are subdivided by solving the
RBF mesh motion and the ow equations. Meshes are used from 100k to 1600k.
3.5.2 Iterative techniques and parallel implementations
As was shown in the previous section, the two dierent phases of the radial ba-
sis function mesh motion, solving (I) and evaluation (II) can be very expensive.
However, it is not necessary to solve (I) and (II) exactly, since the internal point
motion can be arbitrary, as long as the resulting mesh is of suciently high quality.
Dierent pre-conditioning techniques are described in literature (Boer de et al.,
2007) to approximate the system of equations (3.8), leading to a system which is
more ecient to solve using existing iterative solver techniques.
Furthermore, a mesh motion problem leads in general to an ill-conditioned
system, which is dicult to solve directly and iteratively. The condition number in
case of the model problem of a moving square, section 3.4, was about O(10
10
). The
high condition numbers, which occur in this type of problems, are caused by the
boundary point locations. In general moving mesh applications, the combination of
cell clustering on the moving surface and the large distance to the outer boundary,
causes large dierences between points in
bb
. Despite the pre-conditioning, the
ill-conditioned system cannot be eciently solved using iterative techniques. A
possible better choice would be to use parallel direct techniques, available in the
linear algebra packages SuperLU and ScaLAPACK, a parallel version of LAPACK.
A dierent way to improve the eciency of the computation has already been
applied in section 3.4, which is about reducing the number of boundary points
by applying coarsening and smoothing techniques, with a signicant gain in com-
puting time. Concerning the coarsening, complex greedy algorithms (Rendall &
70 Mesh deformation techniques for apping ight
Allen, 2008a) may be applied to select the necessary boundary points such that
the eciency is increased even further. Another method to increase the eciency
to solve the system (3.8) is to decrease the condition number by only taking the
boundary points with a low mutual distance (Boer de et al., 2007), in addition,
re-ordering can be applied for further enhancement.
Finally, the speed of the evaluation (II) can be increased by various fast evalu-
ation algorithms (Boer de et al., 2007). Furthermore, the evaluation can be easily
implemented in parallel since it only involves a matrix-vector multiplication. The
major diculty, concerning a parallel implementation, is that all processor parti-
tions need to know which control point belongs to itself and which to the other
partitions. Then every processor only performs the evaluation of the internal
points, of that particular partition, using all control points and corresponding co-
ecients,
i
and
i
, which are distributed over all partitions. Currently, this is
being implemented in OpenFOAM
.
3.6 Conclusions
In this chapter two dierent mesh motion techniques were described which are
commonly used within the code of OpenFOAM
u = 0, (4.1)
u
t
+
(uu) =
1
p +
2
u, (4.2)
76 Physical and numerical modelling of apping foils and wings
These equations are derived by analysis of the forces on an innitely small uid el-
ement in an inertial reference frame. Using Computational Fluid Dynamics (CFD)
techniques, these equations are solved on a discretised computational domain in
combination with appropriate initial and boundary conditions. When ight under
hovering conditions is considered, the initial velocity eld is zero as well as the
boundary conditions at the outer domain boundary. At the boundary of the ap-
ping wing, a no-slip condition (White, 1991) needs to be specied, which means
that the velocity, relative to the wing, has to be zero, in all directions. This is
accomplished by dening the mathematical velocity on the moving boundary to
be equal to the actual wing motion, which moves according to a specic kinematic
model, derived from realistic insect data (e.g. Fry et al., 2003).
Momentum analysis in rotating reference frame
A dierent numerical approach to solve this problem is to transform the governing
equations and boundary conditions from the inertial reference frame (XY Z) to
a rotating reference frame (xyz), which is xed to the apping wing and moves
accordingly. In the present study, the reference frame approach is only used to
identify certain dimensionless numbers that are related to the rotation of the three-
dimensional wing, i.e. the actual ow computations are made with respect to the
inertial reference frame.
The rotating reference frame is attached with its origin to the joint around
which the wing rotates. The resulting boundary condition on the apping wing
will be such that the eective velocity will be zero, since the reference frame
moves with the boundary. The corresponding velocity transformation, to make
the velocity at the moving boundary equal to zero, is dened as (Ginsberg, 1998,
Baruh, 1999)
u
XY Z
= u
xyz
+ (u
trans
+
wing
r),
where u
XY Z
corresponds to the velocity in the inertial reference frame, whereas
u
xyz
represents the velocity in the local rotating reference frame, r is the distance
from a rotating point to the origin. u
trans
is the translating velocity of the reference
frame itself and can be used to specify the translation of the insect body. In the
present study, the translation velocity of the rotating reference frame is assumed
to be zero, which the case in hovering ight, approximately.
wing
is the angular
velocity of the rotating reference frame, i.e. representing the apping motion of
the wing.
The boundary condition needs to take care of the rotation of the reference
frame, resulting in the following expression:
u
wing
= u
WING
(u
trans
+
wing
r),
where u
WING
is the ow velocity at the wing in the inertial reference frame. This
relation results in a no-slip condition in the rotating reference frame. Additionally,
it is interesting to study the accelerations (Lentink, 2008) with respect to the
inertial frame (XY Z). The accelerations in the inertial and rotating reference
4.2 Governing equations for apping wings 77
X
Y
Z
x
y
z
O
Figure 4.1 Illustration of the rotational reference frame. The rotational reference frame xyz is
moving with the wing and obtained by rotating the inertial reference frame XY Z by three orientation
angles.
frames are related using the following (Ginsberg, 1998, Baruh, 1999):
a
XY Z
= a
xyz
+ (a
ang
+a
cen
+a
cor
).
The angular, a
ang
[m/s
2
], centripetal, a
cen
[m/s
2
], and Coriolis a
cor
[m/s
2
] accel-
erations are respectively dened as
a
ang
=
r,
a
cen
= (r),
a
cor
= 2u
xyz
.
The main parameter in these three dierent accelerations is the angular velocity
[rad/s], which strongly depends on the wing kinematics. u
xyz
[m/s] is the
velocity in the rotating frame. In order to explore the dierent acceleration terms,
needs to be related to the apping motion. This will be elaborated in detail
in section 4.3. Now the expressions for velocity and accelerations in the rotating
reference frame are substituted into the Navier-Stokes equations (4.1) and (4.2)
such that the following transformed Navier-Stokes equations are obtained (for the
sake of simplicity, the subscripts are dropped):
Du
Dt
+ (
r) + ((r)) + (2u) =
1
p +
2
u. (4.3)
This transformed equation describes the momentum balance for a uid particle
close to the wing (in the boundary layer) in the rotating reference frame. With this
approach it is possible to derive dimensionless numbers representing the dierent
acceleration terms, in addition to the already described Reynolds and Strouhal
numbers. These new dimensionless numbers will become available if the dierent
78 Physical and numerical modelling of apping foils and wings
terms in equation (4.3) are scaled as
u
=
u
U
ref
, t
=
U
ref
c
t,
= c ,
, r
=
r
R
, p
=
p
ref
U
2
ref
,
leading to (dropping the stars for simplicity):
U
ref
f c
Du
Dt
+
Rc
U
2
ref
(
r)+
2
Rc
U
2
ref
((r))+
c
U
ref
(2u) = p+
U
ref
c
2
u,
where U
ref
[m/s] is the reference velocity, f [1/s] the apping frequency, which
is dened as f =
U
ref
c
, where c the average chord length. L [m] is the reference
length, [ rad/s] and
[rad/s] are the average rotational velocity and acceler-
ation, respectively. Furthermore,
ref
[kg/m
3
] is the reference density (constant
in incompressible ows). R [m] is a radius length. In addition to the Reynolds
and Strouhal numbers, other dimensional numbers can be identied, related to
the rotation of the reference frame, which is still attached to the wing, rewriting
gives:
1
St
Du
Dt
+
1
C
ang
(
r) +
1
C
cen
((r)) +
1
Ro
(2u) = p+
1
Re
2
u,
where the additional dimensionless number are respectively dened as
C
ang
=
U
2
ref
R c
, (4.4)
C
cen
=
U
2
ref
2
R c
, (4.5)
Ro =
U
ref
c
. (4.6)
By the rotation amplitude , these dimensionless numbers strongly depend on the
wing kinematics, especially the velocity and acceleration due to the wing rotation.
Therefore, the eect of dierent kinematics and ow features may be related to
these dimensionless numbers.
4.3 Wing shape and kinematic modelling
When investigating the three-dimensional aerodynamics around apping wings,
one may try to simulate the geometry and conditions of specic species, in order
to fully understand them, or to opt for a more generic approach. Previous inves-
tigations of specic insect species have been reported for e.g. a fruit y (Sane &
Dickinson, 2001, Birch & Dickinson, 2003), hawkmoth (Liu & Kawachi, 1998) or
4.3 Wing shape and kinematic modelling 79
a dragony (Isogai et al., 2004). The present study follows the second approach
and considers a three-dimensional ellipsoidal model wing (Bos et al., 2008); a sim-
ilar ellipsoid was used by Wang et al. (2004) for their two-dimensional research.
Section 4.3.1 briey describes the wing shape and morphology, while section 4.3.2
deals with the kinematic modelling. The general kinematic modelling consists
of a translation and a rotation component. In addition, the kinematic model
was appended with an active wing exing component, which is described in sec-
tion 4.3.3. The numerical implementation of the wing kinematics is the subject of
section 4.3.4.
4.3.1 Wing shape and planform selection
The general model wing is described by an ellipsoid in the three-dimensional wing
reference frame:
_
x
a
_
2
+
_
y
b
_
2
+
_
z
c
_
2
= 1, (4.7)
where a, b and c are the semi-axes of the ellipsoidal wing. In order to obtain a wing
which has a single wing span, b
s
= 2.0, a maximal chord length c
max
= 1.0 and a
thickness of 10% of the chord, the semi-axes are chosen as a = 0.05, b = 0.5 and
c = 1.0 (for the two-dimensional airfoil, a similar elliptical cross-section is applied
for z = 0). Previous studies show that specic insect features, like the corrugated
wing planform (Luo & Sun, 2005) are of minor inuence on the resulting uid
behaviour. The three-dimensional elliptical planform is shown in gure 4.2 in
comparison to a more realistic representation of the fruit y wing planform.
Since the planform is analytically given by equation (4.7), the average chord
length can be obtained by integration of the chord distribution c(r) along the wing
span from root to tip:
c =
1
R
_
R
0
c(r)dr, (4.8)
where R [m] is the radius of the wing tip and c(r) [m] the chord distribution along
the wing span. Assuming that the chord c(r) is represented by y(z) the following
relation is used to calculate the average chord:
c(r) =
b
2
_
1
r
2
c
2
_
, (4.9)
which is obtained by rewriting equation (4.7). Evaluation of equation (4.8) and
(4.9) leads to an average chord length of c = /4 for c = 1.0, both dening the
planform. Figure 4.2 shows the wing planform and corresponding parameters for
a fruit y wing and a ellipsoidal model wing.
Radius of gyration
In order to dene a sound framework of comparison for dierent three-dimensional
and two-dimensional simulations, it is necessary to dene all reference parameters,
80 Physical and numerical modelling of apping foils and wings
O
z
y
R
g
S
r
R
g
R
root
b
s
R
tip
Axis of rotation
c(r)
dr
(a) Fruit y shape
O
z
y
R
g
S
r
R
g
R
root
b
s
R
tip
Axis of rotation
c(r)
dr
(b) Ellipsoidal shape
Figure 4.2 Model wing geometry and planform. Schematic illustration of the geometry and
planform of a fruit y (a) and ellipsoidal model wing (b). The planform is dened by the chord
variation c(r), the single wing span b
s
and the planform surface area S. The radius of gyration
R
g
is used to dene a sound framework for comparison between mutual three-dimensional and two-
dimensional simulations. In three-dimensional simulations the wing revolves around the origin O, of
which the location is varied to study the eect of the angular accelerations.
introduced in section 4.2, at a representative cross-sectional area of the wing. As
the local velocity of each cross-section varies during apping, the spanwise refer-
ence location is chosen to be at the radius of gyration. According to Ellington
(1984), this is the location where the resulting lift acts. Besides the average chord
length c and the single wing span b
s
the radius of gyration is another impor-
tant geometric parameter, especially when the comparison of dierent kinematic
models is concerned (Bos et al., 2008). The radius of gyration, R
g
is dened as
the weighted second moment of inertia (Luo & Sun, 2005, Lentink, 2008) and is
calculated as
R
g
=
1
S
_
R
0
r
2
c(r)dr. (4.10)
Here S is the wing planform, r the spanwise coordinate, R the distance from the
rotation origin to the wing tip and c(r) represents the chord distribution along
the wing. Additionally, Luo & Sun (2005) compared the ow induced by apping
wings with dierent aspect ratios and found that the radius of gyration provided
a reliable framework for force comparison when the apping velocity is varied.
4.3.2 Kinematic modelling
Besides the numerical interest in the development and improvement of mesh mo-
tion techniques, described in chapter 3, the purpose of the present research is also
to investigate the three-dimensional ow around apping wings at low Reynolds
numbers, the scale at which insects operate. Previous two-dimensional studies
4.3 Wing shape and kinematic modelling 81
X
Y
Z
(t)
(t)
(t)
O
R
root
R
tip
start downstroke
start upstroke
mid-stroke
Stroke plane
Figure 4.3 Schematic illustration of the governing apping angles. Flapping wing motion
is governed by three angles, (t) corresponds to the stroke variation, (t) to the geometrical angle of
attack and deviation (t), which is measured in a plane normal to the stroke plane.
showed the tight relationship between the aerodynamic forces or performance and
the kinematic model (Wang et al., 2004, Bos et al., 2008, Thaweewat et al., 2009).
Since an accurate three-dimensional numerical method has been developed to solve
for the ow around apping wings, the eect of dierent kinematic models on the
aerodynamic performance can be evaluated. Besides the use of an idealisation of
the insect wing planform, simplied kinematics was used, like harmonic motion,
to study dierent parameters independently, e.g. the centre of rotation (which is
equivalent to the Rossby number Ro) or the angle of attack amplitude.
The kinematic wing motion is dened by the variation of three independent
attitude angles, see gure 4.3. In this three-dimensional model the three degrees
of freedom of the wing motion are dened as the apping angle, (t), in the mean
stroke plane, the angle of attack, (t), with respect to the horizontal plane and
the deviation (t), which is measured in a plane normal to the stroke plane, as is
shown in gure 4.3. The deviation may be used to create a gure-of-eight pattern
which is present in realistic fruit y kinematics (Fry et al., 2003).
In literature, dierent kinematic models have been studied, from purely har-
monic motion to complex realistic fruit y kinematics. Using two-dimensional
numerical techniques, Bos et al. (2008) showed that the apping wing perfor-
mance may be inuenced by the specic features of a particular kinematic model.
Figure 4.4 shows the least and the most complex of the kinematics models, which
are considered. Figure 4.4 illustrates (a) the harmonic model (Wang et al., 2004)
compared to (b) the realistic fruit y kinematics (Fry et al., 2003). Figure 4.4(b)
shows that the realistic fruit y kinematics is characterised by an asymmetry in
apping angle, and angle of attack. Furthermore, the angle of attack shows a dip
(at t=0.1T), lowering the eective angle of attack. In addition, the shape of the
fruit y angle of attack clearly shows a plateau of constant value, which is ex-
82 Physical and numerical modelling of apping foils and wings
ploited by Sane & Dickinson (2001, 2002), Lehmann et al. (2005) using a Roboy.
Additionally, the realistic fruit y kinematics in characterised by the presence of
deviation which may result in a gure-of-eight pattern (Shyy et al., 2008b)
In order to illustrate the variation of the motion angles, the denition of the
harmonic model is described accordingly. The apping angle, (t), is described by
a cosine function. The geometric angle of attack, (t), is dened by a sine function
with respect to 90
cos(2ft),
(t) =
2
A
sin(2ft), (4.11)
(t) = A
sin(2ft).
Here, A
0
in Ellington, 1984). f is the apping frequency and A
(t) = A
(2f) sin(2ft),
(t) = A
(t) = A
(2f) cos(2ft).
In general, this harmonic model is crude but fairly reasonable representation of
the apping motion of a fruit y. Also, simplied (harmonic) kinematics may
be interesting for Micro Air Vehicles (MAV) implementation. As will be shown in
chapter 5 (Bos et al., 2008), the apping wing performance may be signicantly in-
uenced by modications of the basic kinematics, such as the previously described
plateau in angle of attack, modelled by a trapezoidal shape and the presence of
deviation (Birch & Dickinson, 2003, Fry et al., 2003, Lehmann et al., 2005, Lu
& Shen, 2008, Bos et al., 2008). The trapezoidal shaped angle of attack is im-
plemented using a piece-wise continuous function and the deviation by dening a
non-zero harmonically variation of (t).
4.3.3 Modelling of active wing exing
In addition to the rigid body motion as shown in gure 4.3 it is possible to dene
an extra displacement concerning exing of the wing. Since a full uid structure
interaction (FSI) simulation is too expensive and beyond the scope of the current
research, a exing displacement of the wing surface is dened. The exing dis-
placement is dened with respect to the initial wing position and can be written
4.3 Wing shape and kinematic modelling 83
t/T [-]
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-80
-60
-40
-20
0
20
40
60
80
100
(a) Harmonic kinematics
t/T [-]
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-80
-60
-40
-20
0
20
40
60
80
100
(b) Fruit y kinematics
Figure 4.4 Comparison of the harmonic and fruit y kinematic models. (a) shows the
least complex kinematic model, representing pure harmonic variations of the apping angle (t) (),
angle of attack (t) () and deviation (t) (). The variation of realistic fruit y kinematics (Fry
et al., 2003) is shown in (b), which is the most complex kinematics available. The realistic fruit y
kinematics is characterised by an asymmetric variation of apping angle, an extra bump and a degree
of trapezoidal in angle of attack and the presence of deviation.
as
x(t) = A
f
cos
_
2x
0
f
_
sin(2ft),
where the rst cosine function denes the wing shape and the sine function rep-
resents the time-variation. A
f
is the exing amplitude vector and x
0
the location
of the initial boundary points. The cosine shaped wing exing is dened by
f
,
which corresponds to the cosine ratio, i.e.
f
= 0.5 means that the shape is like a
half cosine function. Figure 4.5 shows a plunging airfoil incorporating exing for
f
= 0.25 (a) and
f
= 0.5 (b).
4.3.4 Numerical implementation of the wing kinematics
In previous sections the physical kinematic modelling was described. The current
section deals with the numerical implementation of this particular wing kinematics.
In general, the wing kinematics can be decomposed into a translation, a rotation
and a exing (deformation) component. The translational component is not used
in the current three-dimensional simulations, which is limited to hovering and for-
ward ow conditions with stationary position of the rotation origin. However, in
the two-dimensional simulations, the apping motion is dened by a translation in
combination with one rotation angle, the angle of attack. In addition to the trans-
lation and rotation, a limited number of two- and three-dimensional simulations
using a pre-dened wing exing have been performed.
In general, the wing kinematics is calculated beforehand and applied to the
numerical ow solver. At every time-step the location of the boundary points is
determined leading to three distinct displacement arrays, due to translation, rota-
84 Physical and numerical modelling of apping foils and wings
(a) Quarter cosine shape,
f
= 0.25 (b) Half cosine shape,
f
= 0.5
Figure 4.5 Illustration of a exing two-dimensional airfoil. The two-dimensional exing airfoil
is modelled by a time-varying cosine shape. This airfoil shape is either a quarter cosine,
f
= 0.25
(a) or a half cosine,
f
= 0.5 (b). Both gures (a) and (b) show the upstroke (left) and downstroke
(right).
tion and exing.
Translation
The displacement array of the boundary points, which is due to translation is
obtained from
x(t) = A
t
sin(2f
t
t),
where x(t) = (x(t), y(t), z(t)). A
t
and f
t
are respectively the translation ampli-
tude and frequency vectors. The sine function in this denition is used when the
wing needs to move according to an ordinary rigid body motion (Ginsberg, 1998).
When a apping motion is desired, a cosine function is used to dene the motion
in the inertial reference frame.
Rotation
The second boundary point displacement array, due to rotation, is calculated at
subsequent (old and new) time-steps with respect to the initial mesh:
x
old
= R
old
x
0
, (4.13)
and
x
new
= R
new
x
0
, (4.14)
such that:
x
rot
= x
new
x
old
.
Here x
rot
is the boundary displacement due to rotation. R
old
and R
new
are the
rotation transformation matrices at respectively the old and new time instances.
4.3 Wing shape and kinematic modelling 85
The initial boundary points are given by x
0
. These rotation transformation ma-
trices consist of three dierent components, due to a rotation around the X-, Y -
and Z-axis. According to (Ginsberg, 1998, Baruh, 1999), these three matrices are
dened as
R
X
(t) =
_
_
1 0 0
0 cos((t)) sin((t))
0 sin((t)) cos((t))
_
_
, (4.15)
R
Y
(t) =
_
_
cos((t)) 0 sin((t))
0 1 0
sin((t)) 0 cos((t))
_
_
, (4.16)
and
R
Z
(t) =
_
_
cos((t)) sin((t)) 0
sin((t)) cos((t)) 0
0 0 1
_
_
. (4.17)
Here, the rotation around the X-, Y - or Z-axis correspond to the deviation angle,
(t), apping angle, (t), and the angle of attack, (t), respectively. Using the
motion convention, shown in gure 4.3, it is clear that a hovering wing aps in the
X-Z plane, accelerating the uid from top to bottom. After a general sequence of
matrix multiplications, the total rotation matrix is obtained by combining equa-
tions (4.15) to (4.17) as
R
rot
(t) = R
X
(t) R
Y
(t) R
Z
(t), (4.18)
which is substituted into equation (4.13) and (4.14) to nd the displacement of
the boundary points at the old and new times, due to rotation:
x
rot
= [R
rot
(t
new
) R
rot
(t
old
)](x
0
r
0
), (4.19)
where r
0
is the direction vector of the initial rotation origin, which is varied sys-
tematically in chapter 7.
Flexing
In addition to the translation and rotation, the apping wing is able to perform a
exing motion as well. This wing exing is dened as
x
ex
(t) = A
f
cos(2x
0
) sin(2f
f
t),
where A
f
is the exing amplitude vector, f
f
the exing frequency and x
0
repre-
sents the initial boundary points of the apping wing at t = 0. This denition
permits exing in each orientation of a three-dimensional wing, i.e. spanwise and
chordwise.
When combining the previous results for the displacements due to translation,
rotation and exing, the following is obtained:
x
b
= x
trans
+ x
rot
+ x
ex
.
86 Physical and numerical modelling of apping foils and wings
This total displacement eld for the boundary points can be applied to the mesh
motion solver, available in most commercial and non-commercial CFD codes.
Numerical initialisation
If a apping wing is considered, it is desirable to start the numerical simulation at
maximal apping angle, in order to minimise the initial acceleration of the mesh.
Therefore, it was chosen to use a rigid body motion until the maximal apping
angle was reached, in general this occurs at T/4, where T is the apping period.
This approach has two major advantages. First, the mesh deformation is symmet-
ric, yielding high mesh quality at the extreme wing positions. Secondly, the initial
velocity is as small as possible, yielding good conditions for numerical convergence.
4.4 Dynamical scaling of apping wings
In section 4.2 dierent dimensionless numbers, concerning rotational motion, were
identied in order to scale the Navier-Stokes equations, governing uid ow. To de-
ne these rotational dimensional numbers, C
ang
= U
2
ref
/
R c, C
cen
= U
2
ref
/
2
R c
and Ro = U
ref
/ c, it is important to identify the average rotation amplitude,
and its time-derivative
for the particular kinematics implemented:
=
1
T
_
T
0
|
|dt =
1
T
_
T
0
| A
(2f) sin(2ft)|dt = 4A
f, (4.20)
=
1
T
_
T
0
|
|dt =
1
T
_
T
0
| A
(2f)
2
cos(2ft)|dt = 8f
2
A
= 2f, (4.21)
where the apping velocity
was taken from equation (4.12). As shown, evaluation
gives = 4A
f and
= 2f. Besides the angular velocity , angular accelera-
tion
and average chord length c it is necessary to dene an appropriate reference
velocity. As previously described and in accordance with (Bos et al., 2008, Lentink,
2008, Luo & Sun, 2005), the reference cross-section of the three-dimensional wing
is positioned at the radius of gyration, R
g
, and the reference velocity is calculated
at that particular location. This time-averaged velocity follows from:
U
ref
=
1
T
_
T
0
|u
R
g
|dt =
1
T
_
T
0
_
u
2
(t) +v
2
(t) +w
2
(t)dt, (4.22)
where |u
R
g
(t)| is the absolute velocity at R
g
which can be decomposed into three
components u(t), v(t) and w(t) in respectively X-, Y - and Z-direction. Using
equation (4.22) it is straightforward to nd U
ref
by multiplying the expression for
(4.20) by R
g
:
U
ref
= 4A
fR
g
. (4.23)
These relations hold if the wing is the only driving force behind the resulting ow
velocity, but in case of forward ight conditions, there is an additional free-stream
4.4 Dynamical scaling of apping wings 87
90
tan
1
(2Stsin)
2A
sin
Figure 4.6 Schematic illustration of the kinematic parameters in forward ight.
velocity, U
. Lentink & Gerritsma (2003) showed that the following relation can
be used as a good approximation for both hovering (U
+ (4A
fR
g
)
2
. (4.24)
In forward ight another important parameter is the advance ratio (Shyy et al.,
2008b), J, which describes the forward speed with respect to the apping velocity
at a certain radius, R:
J =
U
4A
fR
g
=
4A
,
where the reduced frequency (Ellington, 1984),
, is given by
= U
/fc, with f
the apping frequency. According to (Shyy et al., 2008b, Thaweewat et al., 2009),
=
A
R
g
c
,
which is a measure for the dimensionless translation of the selected cross-sectional
area. In order to create an appropriate framework for comparison, the dimension-
less amplitude at R
g
, A
=
4A
fR
g
c
.
The area that is swept by the revolving wing (Usherwood & Ellington, 2002),
A
swept
, is obtained by subtracting the area swept by the wing tip from the swept
88 Physical and numerical modelling of apping foils and wings
area by the wing root:
A
swept
= 2A
(R
2
tip
R
2
root
),
= 2A
(R
tip
R
root
) (R
tip
+R
root
),
= 2A
b
s
(R
tip
+R
root
),
where R
tip
and R
root
are the radii at respectively the wing tip and root. The
distance from R
tip
to R
root
is identied as the single wing span b
s
. From (Lentink,
2008), it is deduced that keeping the swept area constant is similar to maintaining
a constant Froude eciency (Stepniewski & Keys, 1984, Lentink, 2008). Fur-
thermore, the Rossby number, Ro, which is inversely proportional to the Coriolis
acceleration, needs to be obtained by rewriting equation (4.6) as
Ro =
U
ref
c
=
R
g
c
. (4.25)
When the reference cross-sectional area is located at the radius of gyration (Elling-
ton, 1984), the Rossby number is given by Ro = R
g
/ c, the radius of gyration
divided by the average chord length. For a translating wing, the value for Ro is
innite; for a rotating wing, Ro is nite. If Ro is varied, the eect of dierent
rotation origins can be investigated, from nearly translating to strongly revolving.
However, the determination of Ro would be easier if dened as Ro = R
tip
/ c,
where R
tip
is the wing tip radius, since those values are readily available from lit-
erature, in general Ro = 3.0 for insect and sh (Lentink, 2008), generating thrust
by moving the uid. If the wing planform is complex, like in real insects, it may be
dicult to obtain the radius of gyration. Additionally, for R
root
= 0, the Rossby
number is equivalent to the single wing aspect ratio AR
s
= R
tip
/ c, a geometric
wing characteristic.
4.5 Computational domain and boundary
conditions
In this section the general approach for mesh generation is described, that has been
applied for the two- and three-dimensional apping wing simulations. Chapter 5, 6
and 7, which are dealing with respectively two-dimensional hovering, forward and
three-dimensional hovering ight, explain the specic mesh generation, domain
size and boundary conditions for these simulations in more detail.
When performing a Computational Fluid Dynamics (CFD) study, a compu-
tational domain is necessary to contain the mesh on which the governing partial
dierential equations are solved. It is necessary to use a suciently large compu-
tational domain to minimise disturbances, with appropriate boundary conditions
in order to obtain large convergence rates and the correct solution. When gener-
ating a grid around a two-dimensional thin and ellipse-shaped airfoil, it is ecient
4.6 Denition of force and performance coecients 89
X
Y
left
wing
right
(a) Two-dimensional
Z X
Y
wing
left
right
bottom
top
front
back
(b) Three-dimensional
Figure 4.7 Computational domain and boundary conditions. The two-dimensional domain
uses the O-type topology (a), while the three-dimensional apping wing simulations uses the boxed
topology (b).
to use an O-type mesh, which is shown in gure 4.7(a). Using conformal map-
pings (Bos et al., 2008, Thaweewat et al., 2009), a high quality mesh is generated
between the ellipse-shaped wing boundary,
wing
, and the cylinder-shaped outer
boundaries,
left
and
right
. The outer boundary is split into
left
and
right
in
order to be able to specify inow and outow boundary conditions, which are
necessary (Wesseling, 2001) when forward ight conditions are simulated. When
simulating hovering ight, the apping wing still induces a signicant amount of
downwash, leading to a small inow and outow at the boundaries of the compu-
tational domain. The meshes for the two-dimensional simulations were generated
using Gambit software, see appendix A.
Generation of a three-dimensional structured mesh around a wing is not an easy
task, which can be very cumbersome using manual procedures used in programs
like Gambit or Gridgen
. Using GridPro
Figure 4.8 Forces on a general three-dimensional apping wing. The forces are dened in the
three-dimensional inertial reference frame. Depending on the type of motion, hovering or forward ight,
two- or three-dimensional, the lift and drag are constructed from the forces in X-, Y - or Z-direction.
is important to properly dene force and performance coecients. In this section,
the general determination of the force coecients is explained in combination with
performance characteristics.
Forces
In general three dimensions, a two-dimensional derivation is trivial, the denitions
of the forces, F
X
, F
Y
and F
Z
are shown in gure 4.8. The centre of the axes co-
incides with the origin of rotation of the three-dimensional wing. The total force
vector is integrated over the wing surface and contains a pressure and a viscous
contribution. These forces are calculated using the following expression:
F
tot
=
_
S
pdS
_
S
u
n
dS,
where F
tot
[N] is the total force vector, S [m] the wing surface and dS [m] represents
an innitely small surface area element, p [N/m
2
] is the pressure and [Ns/m
2
]
the dynamic viscosity. The term u/n is the gradient of the velocity vector with
respect to the normal vector to the wall, i.e. together with this forms the wall
shear stress.
Two dierent force denitions can be dened, depending on the type of three-
dimensional motion, hovering or forward apping ight. In hovering conditions,
when the main apping direction is around the Y -axis, as shown in gure 4.3, the
lift force F
lift
is dened in the vertical direction and equal to F
Y
. The drag force
F
drag
, however, is dened in opposite direction of the apping wing motion. If
three-dimensional apping in hovering ight without deviation is considered, the
following lift and drag variations are found:
F
lift
(t) = F
Y
(t),
F
drag
(t) = F
X
(t) sin((t)) F
Z
(t) cos((t)),
where (t) is the apping angle. If the motion includes a deviation velocity, such
that the wing is not moving in the horizontal plane, the drag force derivation
is more elaborate but similar. Besides the lift and drag, there is a force in the
4.6 Denition of force and performance coecients 91
direction of the spanwise coordinate. This spanwise force is dominated by the
viscous wall shear stress and therefore small compared to the lift and drag forces.
It is justied to neglect this spanwise force, also because it has little relevance to
performance.
In forward ight conditions, the same inertial reference frame is used as shown
in gure 4.3. The main apping direction is still around the Y -axis, but the
direction of the uniform ow is from top to bottom in direction of the negative
Y -axis. The complete system needs to be rotated around the Z-axis in order to get
a horizontal orientation of the free-stream. In the case of forward apping ight
the lift force is dened in the positive X-direction and the drag in the negative
Y -direction, opposite to the free-stream velocity:
F
lift
(t) = F
X
(t),
F
drag
(t) = F
Y
(t).
Commonly the forces are made dimensionless using the dynamic pressure based
on the average velocity. With the strong variation in velocity, however, it is deemed
more appropriate to scale the forces with the mean dynamic pressure itself (Bos
et al., 2008). Hence, the forces are dened as:
C
D
=
F
drag
q S
, C
L
=
F
lift
q S
,
where C
D
and C
L
are the drag and lift coecients and S the wing surface. The
mean dynamic pressure q is dened as:
q = 1/2U
2
ref
= 1/2
1
T
_
T
0
_
(U
+U
ap
(t))
2
+ (U
dev
(t))
2
dt,
where the integration is evaluated over one apping cycle with period T [s]. The
reference velocity contains the free-stream velocity U
1
x
y
1
x
y
0.4 0.42
0.56
0.58
Figure 5.1 O-type mesh topology with
boundary conditions on
1
,
2
and
3
.
Figure 5.2 Body conformal moving
mesh around a 2% ellipsoid airfoil.
Navier-Stokes equations (2.1) and (2.2) are used. Additional solver settings can
be found in (Bos et al., 2008, appendix B).
5.2.2 Mesh generation and boundary conditions
In order to compute the ow around the moving airfoils, an O-type computational
domain is used, which is shown schematically in gure 5.1. The computational
domain is divided into two parts:
1
and
2
for the inner and outer mesh respec-
tively. The body surface
1
is located in the centre of the computational domain.
It has the reference length L which corresponds to the wing chord length. The
outer boundary
3
is located at 25L such that the inuence of the far eld bound-
ary condition is negligible (Lentink & Gerritsma, 2003). At the body surface a
no-slip boundary condition is applied. Since the moving wing simulations concern
hovering insect ight, such that a free-stream is absent, a symmetry boundary
condition was applied at
3
for numerical reasons. The inuence of this symmetry
condition has been investigated and found to be suciently small.
For the wing, which is modelled as an ellipse of 2% thickness, generation of a
high quality mesh is not as straightforward as for a cylinder. The geometric surface
gradient is high, especially at the leading and trailing edges. This complicates the
creation of a high quality mesh, i.e. high cell orthogonality. In order to create
this body conformal mesh (see gure 5.2) a conformal mapping was applied (see
Wang, 2000b). The intermediate interface
2
divides the mesh into two separate
elds, corresponding respectively to the inner conformal mesh (
1
) and the outer
mesh (
2
). The complete inner mesh moves according to the wing kinematics,
while re-meshing takes place in the outer eld
2
. Since re-meshing occurs at
a distance of 25 to 30 body lengths away from the wing, the ow around the
wing is not aected by the mesh regeneration. The described computational setup
was thoroughly validated using the ow around stationary and moving circular
98 Inuence of wing kinematics in two-dimensional hovering ight
t +dt
t
ref
ref
+d
y
x
Figure 5.3 Relative cell displacement in rotation.
cylinders (Bos et al., 2008, appendix C).
The airfoil simulations were performed on a mesh of 50000 cells with 2000
time-steps within one motion period. At this mesh, the size of the rst cell at the
wing surface varies between 2% and 50% of the wing thickness at the leading-edge
and in the middle of the prole respectively. The grid resolution near the wing, up
to 1 chord length, was 8800 (176x50) cells such that the leading and trailing edge
vortices where captured with at least 1000 cells. One run, simulating 18 apping
periods needed approximately 10 days on one serial AMD Athlon 2500+ CPU.
In order to minimise the interpolation errors from one time-step to the next
it is important to analyse the inuence of the relative cell displacements. There-
fore, the motion of a reference cell was investigated, which is illustrated for the
rotational motion in gure 5.3. From the relative displacements in rotational and
translational direction follow the constraints for the size of the time-step in or-
der to keep the interpolation errors within limits. The relative displacements in
rotational and translational direction are dened as
r
=
ref
and
y
=
y
y
ref
=
2f
e
A
e
Nt
y
ref
.
Here corresponds to the angular displacement of the reference cell, while
ref
is the original radial length of this cell. The linear displacement of this cell is y
and y
ref
is the original length of this cell. Furthermore, f
e
, A
e
and N correspond
respectively to, the frequency, amplitude and number of cells on the surface.
In (Bos et al., 2008, appendix C), it was shown that a relative displacement of
10% in both rotational and translational direction leads to accurate results with
dierences in drag coecients remaining below 5%. The computational eorts are
acceptable: 2000 time-steps within one excitation period. Additionally, Bos et al.
(2008) (appendix D) investigated the mesh and time-step independence for the
nominal solver settings using harmonic wing kinematics for hovering ight.
5.3 Modelling insect wing kinematics 99
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
-0.5
0
0.5
1
1.5
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
-0.5
0
0.5
1
1.5
(b)
Figure 5.4 Comparison of force coecients between the present simulations and Wang
et al. (2004). Comparison of lift (a) and drag (b) coecients using harmonic wing kinematics with
A = 2.8, Re = 75 for the present study () and obtained by Wang et al. (2004) ().
5.2.3 Validation using harmonic wing kinematics
The main numerical parameters, a mesh size of 50000 cells and 2000 time-steps
within one excitation period, are used to validate our results with those obtained
by Wang et al. (2004) for similar but not entirely identical conditions. A two-
dimensional case was selected, with a moving wing according to harmonic kinemat-
ics. The amplitude was 2.8 times the chord length, which corresponds to Re = 75.
Figure 5.4 shows the lift and drag coecients for validation purposes. Our forces
are normalised with the maximum of the quasi-steady force, just as in (Wang et
al., 2004). In similarity to (Wang et al., 2004) the drag in gure 5.4(a) is dened
to be positive in the direction opposite to the horizontal motion.
Generally, our force distribution looks similar for both cases. Only just after
stroke reversal our computation nds a larger lift and drag which is probably
the result of dierent numerical dissipation properties of both codes. The mean
lift and drag coecients are 0.84, 1.47 for our simulation, compared to 0.82, 1.44
obtained by Wang et al. (2004), which is a dierence of only 2% in lift and drag and
therefore, the computations were considered to be suciently accurate. Moreover,
within the context of comparing results of dierent stroke patterns, the present
numerical method is proved to be accurate.
Further details of the validation and verication studies can be found in (Bos
et al., 2008, appendix C and D).
5.3 Modelling insect wing kinematics
In order to derive the two-dimensional kinematic models the three-dimensional
degrees of freedom need to be converted to their two-dimensional counter-parts.
A common procedure is to dene an equivalent two-dimensional geometry, while
100 Inuence of wing kinematics in two-dimensional hovering ight
Figure 5.5 Illustration of the main motion directions. (t) corresponds to the stroke variation,
(t) to the geometrical angle of attack and (t) to the deviation from the horizontal stroke plane.
From (Sane & Dickinson, 2001).
maintaining the characteristic aspects of the wing motion. This two-dimensional
set-up is derived in section 5.3.1 in terms of wing selection and model parameters.
The dynamical scaling and the force denitions are described respectively in 5.3.2
and 5.3.3.
5.3.1 Insect wing selection and model parameters
The computational approach is applied to investigate the inuence of dierent
kinematic wing motion models on the aerodynamic performance. The dierent
kinematic models are illustrated using the Roboy experimental set-up, shown
in gure 5.5 (see Sane & Dickinson, 2001, Dickinson et al., 1999). In this three-
dimensional model the three degrees of freedom of the wing motion are dened
as the angular displacement in the mean stroke plane, the angle of attack ,
with respect to the horizontal plane and the deviation from the horizontal plane
, as is shown in gure 5.6. The deviation causes a gure-of-eight pattern which
is present in real fruit y kinematics (see Fry et al., 2003). The two-dimensional
airfoil shape is chosen to be a 2% thick ellipsoid. Lentink & Gerritsma (2003) found
this airfoil an acceptable choice to model insect wings at low Reynolds numbers,
Re = O(100). The two-dimensional projection is to be dened at a representative
spanwise location such that the motion is conned to an arc around the wing root.
Birch & Dickinson (2003) found strongest vorticity at a spanwise location of 0.65R
from the wing root, where R is the wing span. Therefore, Wang et al. (2004) used
this distance to derive their two-dimensional model.
In the present study, a dierent argument for the selection of the projection
5.3 Modelling insect wing kinematics 101
x
ac
x
cg
c
F
y
F
r
F
x
M
=
fR
g
c
_
1
0
_
(
t
)
2
+ (
t
)
2
(5.3)
and
St =
fc
U
=
c
R
g
1
_
1
0
_
(
t
)
2
+ (
t
)
2
. (5.4)
Here f = 1/T is the frequency, and the three-dimensional kinematic an-
gles for the displacement and deviation. From (5.3) and (5.4) it can be observed
that the Reynolds number Re depends solely on the frequency f for a given dis-
placement (t) and deviation (t). The Strouhal number St is not to be varied
independently. We xed the Reynolds number to Re = 110.
5.3.3 Force and performance indicators
The denition of the drag and lift forces is shown in gure 5.6. The lift is equal
to the vertical force F
y
, while the drag is taken equal to the horizontal force
F
x
, dened positive in the positive x-direction. Commonly the forces are made
dimensionless using the dynamic pressure based on the average velocity. With the
strong variation in velocity, however, it is deemed more appropriate to scale the
forces with the mean dynamic pressure itself. Hence, the forces are dened as
C
D
=
F
x
q c
, C
L
=
F
y
q c
,
where C
D
and C
L
are the drag and lift coecients. The mean dynamic pressure
q is dened as
q = 1/2U
2
= 1/2
1
T
_
T
0
_
_
x
t
_
2
+
_
y
t
_
2
_
dt,
where the integration is evaluated over one apping cycle. The force coecients
are the major parameters used to assess the inuence of the dierent wing motion
models. In addition, the ratio between time-averaged lift coecient, C
L
, and time-
averaged drag coecient, C
D
, is used to characterise performance. These force
averages are obtained by integration of C
L
and C
D
. The lift is averaged over the
complete period, while for the drag the averages are per half stroke. The average
lift-to-drag ratio, C
L
/C
Dave
is chosen as an indicator of aerodynamic performance,
also known as the glide number in aerospace engineering. Since the average lift
coecients of the dierent kinematic models are matched, the lift-to-drag ratio is
corrected for any dierences in lift. Therefore, a high lift-to-drag ratio eectively
means low drag at equal lift.
5.3 Modelling insect wing kinematics 103
5.3.4 Dierent wing kinematic models
Since the main purpose of this study is to investigate the inuence of wing kinemat-
ics on the aerodynamic performance during hovering fruit y ight, four dierent
kinematic models, with dierent degree of complexity, have been analysed. Two
of these models, the pure harmonic motion and the Roboy experimental kine-
matics have appeared in literature. The third model represents the actual fruit y
kinematics as observed in experiments and the last one was a modication of the
latter, chosen to investigate the eect of symmetry in the wing motion.
In order to facilitate the comparison the model parameters are chosen based
on matching the mean quasi-steady lift coecient (Bos et al., 2008, appendix
A). Although according to Sane & Dickinson (2001) the mean drag is strongly
inuenced by the unsteady ow physics, which are not fully present in the quasi-
steady theory, the mean lift coecient is predicted well using this theory. Using
quasi-steady theory, dierent kinematic models were constructed, such that their
quasi-steady lift coecients are matched within 1%. For the symmetric models
this force is equal to the resultant force. In view of the limitations of the quasi-
steady theory, the dierence between predicted and simulated values is expected
to exceed this 1% tolerance. However, in section 5.4 it is shown that the computed
mean lift coecient of the numerical simulations are reasonably well matched for
all models, which provides an a-posteriori justication of our choices for the model
parameters.
The characteristic shapes of each model are described. Subsequently they are
used to investigate the inuence of the models on the force histories and the per-
formance in section 5.4. Analysing those aspects leads to a better understanding
of how the fruit y may benet from kinematic features which are absent in the
simpler models, and reveals the relevance of including these aspects in theoretical
models.
The rst of the four models is described by pure sine and cosine functions and
will therefore be referred to as the harmonic model (see Wang et al., 2004). The
displacement, angle of attack and deviation, are shown in gure 5.7(a). The second
model takes the wing kinematics as used in the Roboy model (Dickinson et al.,
1999). In gure 5.7(b) it is shown that the ip from down to upstroke is postponed
to the end of the translational phase which results in the sawtooth shape of the
displacement. Large accelerations at stroke reversal are the result. The deviation
is zero, just as in the harmonic model. The third model, shown in gure 5.7(c),
is derived from measurements on real fruit ies (Fry et al., 2003) and is therefore
considered as the most realistic fruit y kinematic model. This model does include
the deviation which results in a gure-of-eight pattern. Neither the displacement,
angle of attack nor deviation is symmetric during the apping period.
In order to investigate the fact that the observed fruit y kinematics lacks an
exact symmetry in the wing stroke pattern, a symmetrical model was constructed,
referred to as the symmetric fruit y model, displayed in gure 5.7(d). Within this
model the motion is identical for the downstroke and upstroke. Like the realistic
104 Inuence of wing kinematics in two-dimensional hovering ight
t/T [-]
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-80
-60
-40
-20
0
20
40
60
80
100
(a)
t/T [-]
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-80
-60
-40
-20
0
20
40
60
80
100
(b)
t/T [-]
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-80
-60
-40
-20
0
20
40
60
80
100
(c)
t/T [-]
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-80
-60
-40
-20
0
20
40
60
80
100
(d)
Figure 5.7 Kinematic angles of the dierent kinematic models. (a) Harmonic model. (b)
Roboy model. (c) fruit y model. (d) simplied fruit y model. : displacement angle , : angle
of attack , : deviation angle .
fruit y model this symmetric model includes a time-dependent deviation such
that the observer sees a gure-of-eight pattern of the wing. Neither of those
last two realistic kinematic models can be described by using simple analytical
functions without losing signicant detail.
When comparing the motion parameters, , and for each model it becomes
possible to identify certain important dierences. The Roboy initially has a
larger gradient in time of the angle of attack compared to the harmonic case, see
gure 5.7(a) and (b). During translation from about t = 0.1T to t = 0.4T the
angle of attack attens at a value of almost 40
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
0
1
2
3
4
5
(a)
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
0
1
2
3
4
5
(b)
Figure 5.8 Lift coecient histories of the baseline kinematic models. : harmonic model,
: Roboy model, : realistic fruit y model, : simplied fruit y model.
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-8
-6
-4
-2
0
2
4
6
8
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-8
-6
-4
-2
0
2
4
6
8
(b)
Figure 5.9 Drag coecient histories of the baseline kinematic models. : harmonic model;
: Roboy model, : realistic fruit y model, : simplied fruit y model.
kinematic model C
L
C
Ddown
C
Dup
C
L
/C
Dave
harmonic 1.483 3.7% 1.848 1.839 0.805 29%
Roboy 1.417 8.0% 2.466 2.448 0.577 49%
realistic fruit y 1.540 baseline 1.387 1.335 1.132 baseline
simplied fruit y 1.454 5.6% 1.012 1.596 1.115 1.5%
Table 5.1 Time-averaged force coecients using the complete baseline models.
5.4 Results and Discussion 107
C
L
= 1.483
x
c
[]
-3 -2 -1 0 1 2 3
(a)
C
L
= 1.417
x
c
[]
-3 -2 -1 0 1 2 3
(b)
C
L
= 1.54
x
c
[]
-3 -2 -1 0 1 2 3
(c)
C
L
= 1.454
x
c
[]
-3 -2 -1 0 1 2 3
(d)
Figure 5.10 Force vectors during each half-stroke. (a) harmonic model, (b) Roboy model,
(c) realistic fruit y model, (d) symmetric fruit y model.
108 Inuence of wing kinematics in two-dimensional hovering ight
(a) harmonic model (b) realistic fruit y model
Figure 5.11 Vorticity contours around a apping airfoil. Vorticity contours are shown for
t = 0.1T (blue: clock-wise, corresponding to negative vorticity values).
compared to the fruit y models. This is also illustrated in gure 5.8 and 5.9 (lift
and drag histories) and gure 5.10 (force vectors). Figure 5.11 shows the vorticity
contours of the realistic fruit y model compared with the harmonic model. It can
be seen in gure 5.7(a) that the eective angle of attack is higher in the harmonic
case, compared to the realistic fruit y model, gure 5.7(c). Therefore, the mean
drag contribution of the leading-edge vortices (LEV) is higher. The decrease in
eective angle of attack in the realistic fruit y model is also enlarged by the
presence of the bump. This drag increasing eect is even larger in case of the
Roboy model due to the trapezoidal angle of attack. The sawtooth shaped
Roboy displacement could possibly play an important role as is discussed in the
next section. The dierent kinematic patterns are also illustrated in gure 5.10,
which shows the resultant force vectors during a full stroke for those baseline
kinematic models.
Furthermore, the mean drag coecient of the simplied fruit y is not sym-
metric, i.e. the drag during the upstroke is about 57% higher than during the
downstroke, which is attributed to the complex vortex dynamics. Nevertheless,
the average value during a complete stroke matches the mean drag coecient
obtained with the realistic fruit y model.
When comparing the lift-to-drag ratios in table 5.1, it can be observed that
within the model assumptions, the fruit y models perform better than the less
complex models. Compared to the harmonic model, the realistic fruit y model
results in a signicant decrease in drag of 29% at comparable lift. The dierence
with the Roboy model is even larger, 49%. These performance increases are the
result of the lower drag coecients in both fruit y models due to certain benecial
5.4 Results and Discussion 109
kinematic features. The current results provide insight into the eects of certain
specic kinematic features. However, one has to be cautious when extrapolating
these results to real ying ies since in reality not every apping period displays
exactly the same kinematic prole. Next, the individual inuences of the dierent
interesting kinematic shapes are studied.
5.4.2 Kinematic features investigation
Inuence of sawtooth displacement used by the Roboy
The sawtooth shaped displacement of the Roboy is investigated in isolation
to assess its inuence on the force histories and the aerodynamic performance.
Therefore, the purely harmonic model was appended with the Roboy displace-
ment and the results were compared with the ones obtained using the original
harmonic model. Figure 5.12(a) shows the force vectors acting on the wing during
the up and downstroke. In addition, the force histories during one full stroke are
shown in gure 5.13. From gure 5.13 it is observed that compared to the har-
monic model the global force histories look similar. Two force peaks are observed
close to t = 0.1T and t = 0.4T, respectively, which are repeated since the motion
is symmetric. The lift peaks are almost equal but the drag peaks are signicantly
larger for the sawtooth case, see gure 5.13(b). This also explains the larger
mean drag compared to the harmonic model which can be read from table 5.2.
In gures 5.14(a) and (b) the vorticity contours are plotted at t = 0.1T for the
harmonic model and the one with the appended sawtooth shaped displacement.
From gure 5.14 it can be seen that the LEV is stronger for the sawtooth case
which explains the higher drag peak. The stronger LEV at the beginning of
the downstroke in the sawtooth case is most likely caused by the higher velocity
gradient. This leads to a larger shear layer to form a stronger vortex. On the other
hand, at the end of the half-stroke the wing decelerates faster in the sawtooth
case which results in a lower strength in the LEV. Since the wing orientation is
almost vertical, at t = 0.1T, the drag peak is larger than the lift peak.
The larger mean drag is reected in the integrated values in table 5.2. Due
to this larger drag during each stroke, the sawtooth shaped displacement leads
to a lower lift-to-drag ratio, which shows a decrease of 24.3% with respect to the
harmonic case.
Inuence of trapezoidal angle of attack used by the Roboy
In combination with the sawtooth displacement, the Roboy uses a trapezoidal
shape for the angle of attack. In order to determine the eect of this shape the
harmonic model is extended by this trapezoidal angle of attack. The results are
compared with those obtained with the original harmonic model, see gure 5.12(a)
for the force vectors. The lift and drag coecients are plotted in gure 5.15. A
surprising and unexpected observation is the asymmetry in the periodic force dis-
tribution for the trapezoidal angle of attack notwithstanding the symmetry of
the kinematics. This leads to the non-zero mean horizontal force along a complete
110 Inuence of wing kinematics in two-dimensional hovering ight
C
L
= 1.366
x
c
[]
-3 -2 -1 0 1 2 3
(a)
C
L
= 1.351
x
c
[]
-3 -2 -1 0 1 2 3
(b)
C
L
= 1.483
x
c
[]
-3 -2 -1 0 1 2 3
(c)
C
L
= 1.323
x
c
[]
-3 -2 -1 0 1 2 3
(d)
Figure 5.12 Force vectors during each half-stroke. (a) harmonic model with sawtooth , (b)
harmonic model with trapezoidal , (c) harmonic model with extra bump , (d) harmonic model
with deviation .
5.4 Results and Discussion 111
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-6
-4
-2
0
2
4
6
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-6
-4
-2
0
2
4
6
(b)
Figure 5.13 Lift and drag coecients. Lift (a) and drag (b) histories to investigate the inuence
of the sawtooth displacement compared to the harmonic model. : harmonic ,,; : harmonic ,
and Roboy .
(a) harmonic model (b) harmonic model with sawtooth displacement
Figure 5.14 Vorticity contours around a apping airfoil. Vorticity contours are shown for
t=0.1T (blue: clock-wise, corresponding to negative vorticity values)
112 Inuence of wing kinematics in two-dimensional hovering ight
kinematic model C
L
C
Ddown
C
Dup
C
L
/C
Dave
harm. , and 1.483 (baseline) 1.848 1.839 0.804 (baseline)
harm. , + Roboy 1.366 (7.9%) 2.240 2.250 0.608 (24.3 %)
harm. , + Roboy 1.351 (8.9%) 2.302 2.733 0.537 (33.3 %)
harm. , + fruit y. 1.483 (0.0%) 1.221 1.969 0.930 (+15.6 %)
harm. , + fruit y. 1.323 (10.8%) 1.807 1.776 0.738 (8.2 %)
Table 5.2 Time-averaged force coecients to investigate the inuence of kinematic shapes.
Each characteristic shape is varied with respect to the harmonic motion model.
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-6
-4
-2
0
2
4
6
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-6
-4
-2
0
2
4
6
(b)
Figure 5.15 Lift and drag coecients. Lift and drag histories to study the inuence of the
trapezoidal angle of attack compared to harmonic model. : harmonic ,,; : harmonic , and
Roboy .
5.4 Results and Discussion 113
(a) harmonic model (b) harmonic model with trapezoidal
Figure 5.16 Vorticity contours around a apping airfoil. Vorticity contours are shown for
t=0.6T (blue: clock-wise, corresponding to negative vorticity values)
stroke cycle. Although this model is symmetric, the force distributions are not,
since the complex vortex dynamics are non-linear and asymmetric.
From gure 5.15 it is clear that at the beginning of a stroke the lift peak of the
trapezoidal case is larger. Using gure 5.16 this is illustrated at the beginning of
the upstroke using vorticity contours. The LEV is larger in case of the trapezoidal
angle of attack. This can be explained as follows. In the trapezoidal case the
wing reaches the maximum angle of attack earlier in the stroke, see gure 5.12(b).
Therefore, the angle of attack is larger at the early start of a stroke compared to
the harmonic model. Since large angle of attacks cause high velocity gradients
over the leading-edge, larger vortices occur in the beginning of a stroke.
Another interesting result is the low second peak in the lift, at the end of each
stroke, compared to the harmonic model. Taking a closer look at gure 5.17(b),
one observes stronger and more pronounced vortices in the wake of the trape-
zoidal case. This could indicate a larger amount of vortex shedding during the
period when the angle of attack is nearly constant. This results therefore in a
lower second peak since the LEV has decreased in size and strength. Altogether,
the mean lift is slightly decreased whereas the mean drag is increased. This leads
to a signicant performance decrease of 33.3% due to the trapezoidal angle of
attack variation, see table 5.2.
Inuence of extra bump in angle of attack used by the fruit y
The fruit y models are subject to an extra bump in angle of attack. To make
comparison plausible the symmetric bump variation used in the simplied fruit
y model is used to compare results with the harmonic model. Figure 5.12(c)
114 Inuence of wing kinematics in two-dimensional hovering ight
(a) harmonic model (b) harmonic model with trapezoidal
Figure 5.17 Vorticity contours around a apping airfoil. Vorticity contours are shown for
t=0.4T (blue: clock-wise, corresponding to negative vorticity values)
shows the force vectors during up and down stroke. In gure 5.18 the lift and
drag forces are shown for the harmonic model with and without the symmetric
bump in angle of attack. From table 5.2 it is seen that using this feature the
mean lift does not change signicantly. However, the drag during the downstroke
is very much aected. A decrease of at least 30% in mean drag is found, compared
to the harmonic case. It is also noted that this case results in asymmetric force
distributions as was the case when using the trapezoidal angle of attack. On the
other hand the drag is slightly increased during the upstroke such that the mean
lift-to-drag ratio is still increased with more than 15.6%. From gure 5.18 it is
observed that the extra bump generates an extra lift peak at the beginning of the
downstroke. The change in angle of attack due to the extra bump is shown when
gure 5.19(a) and (b) are compared. The decrease in eective angle of attack as
a result of the bump is considerable compared to the harmonic case. The same
was found for the Roboy case. Therefore, for the case with the bump in angle
of attack, the LEV provides nearly complete lift since the wing orientation is ap-
proximately horizontal. This is also the main reason for the lower drag during the
downstroke.
Figure 5.20 shows the vorticity at the beginning of the upstroke at the time
of the bump. The LEV is larger compared to the case with the extra bump in
angle of attack. This causes the loss in lift just after stroke reversal in case of the
bump angle of attack compared to the harmonic model.
Inuence of wing deviation used by the fruit y
The last important characteristic of the kinematics is the deviation, present in the
5.4 Results and Discussion 115
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-6
-4
-2
0
2
4
6
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-6
-4
-2
0
2
4
6
(b)
Figure 5.18 Lift and drag coecients. Lift (a) and drag (b) histories to study the inuence of
the extra bump in angle of attack. : harmonic ,,; : harmonic , and fruit y .
(a) harmonic model (b) harmonic model with extra bump in
Figure 5.19 Vorticity contours around a apping airfoil. Vorticity contours are shown for
t=0.1T (blue: clock-wise, corresponding to negative vorticity values)
116 Inuence of wing kinematics in two-dimensional hovering ight
(a) harmonic model (b) harmonic model with extra bump in
Figure 5.20 Vorticity contours around a apping airfoil. Vorticity contours are shown for
t=0.6T (blue: clock-wise, corresponding to negative vorticity values)
realistic and simplied fruit y model. This deviation causes a gure-of-eight
pattern, described by the wing tip instead of wing motion solely in the stroke
plane. Since deviation could introduce a large velocity component perpendicular
to the stroke plane, the eective angle of attack is highly aected. This motion
perpendicular to the stroke plane is illustrated in gure 5.12(d) which also shows
the force vectors.
Figure 5.21 shows the force coecients during one apping period with de-
viation added to the harmonic model. The mean lift and drag are not strongly
inuenced by the deviation, see table 5.2. The mean lift is decreased by 10.8% and
the mean drag is almost not aected by the presence of deviation, about 2%4%
dierence in both strokes. It is also revealed that the force distributions remain
symmetric.
The large inuence of the deviation on the variation of the lift force is observed
at the start (t = 0.1T and t = 0.6T) and end (t = 0.4T and t = 0.9T) of each
stroke. Just after stroke reversal a lift peak occurs, which is higher compared to
the harmonic case. However, at the end of each stroke the harmonic lift peak was
decreased by the deviation. It appears that the force distribution is levelled or
balanced by the deviation.
The ow dynamic mechanism for this is shown in the vorticity visualisations
of gure 5.22 which shows the vorticity at the beginning of the stroke. Compared
to the harmonic model, the deviation causes a slightly stronger LEV at t = 0.1T.
The inuence of the deviation is relatively large since the deviation increases the
eective angle of attack considerably just after stroke reversal. At the end of a
stroke the wings move up again which leads to a decrease in eective angle of
5.4 Results and Discussion 117
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-6
-4
-2
0
2
4
6
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-6
-4
-2
0
2
4
6
(b)
Figure 5.21 Lift and drag coecients. Lift (a) and drag (b) histories to study the inuence of
the deviation compared to harmonic model. : harmonic ,,; : harmonic , and fruit y .
(a) harmonic model (b) harmonic model with deviation
Figure 5.22 Vorticity contours around a apping airfoil. Vorticity contours are shown for
t=0.1T (blue: clock-wise, corresponding to negative vorticity values).
118 Inuence of wing kinematics in two-dimensional hovering ight
(a) harmonic model (b) harmonic model with deviation
Figure 5.23 Vorticity contours around a apping airfoil. Vorticity contours are shown for
t=0.6T (blue: clock-wise, corresponding to negative vorticity values).
attack. Figure 5.23(a) and (b) show LEVs of comparable strength for both cases.
Summarising, the deviation is levelling the force distributions while the mean
lift and drag are almost unaected. This leads to the suggestion that a fruit y
may use the deviation to level the wing loading over a apping cycle. Three-
dimensional studies are needed to investigate to what extent this eect is also
present in real insect ight.
5.5 Conclusions
The eect of wing motion kinematics on the aerodynamic characteristics of hov-
ering insect ight was investigated by means of two-dimensional numerical ow
simulations. The results of the present two-dimensional study may provide useful
insights in the understanding of real three-dimensional insect ight (Wang et al.,
2004).
Four dierent kinematic models, with dierent complexity, have been anal-
ysed using two-dimensional time-dependent Navier-Stokes simulations. Two of
these models, pure harmonic motion and Roboy experimental kinematics have
appeared in literature. The third model represents the actual fruit y kinematics
as observed in experiments and the last one is a modication of the latter, chosen
to investigate the eect of symmetry. The most prominent aspects of the Roboy
kinematic model are the sawtooth displacement and the trapezoidal angle of
attack. The fruit y models are characterised by a bump in angle of attack and
the presence of deviation. To facilitate the comparison all models are dynami-
cally scaled at Re = 110 and constructed such that their mean quasi-steady lift
5.5 Conclusions 119
coecient was matched.
It was found that the realistic fruit y wing kinematics result in signicantly
lower drag at similar lift compared with the simplied wing kinematic models
used in literature. The trend that the fruit y kinematics increases aerodynamic
performance agrees well with the predictions of the quasi-steady theory, but the
numerical ow simulations provide a more complete quantitative analysis of the
ow behaviour. To investigate which aspects of the kinematic shapes are the most
important, they were compared to the harmonic model.
First, an overall comparison of the complete kinematic models was given. It
was shown that the dierence in performance in terms of mean lift-to-drag ratio
between the dierent kinematic models was signicant. The mean aerodynamic
drag at equal lift of the fruit y models is about 49% lower compared to the Roboy
model and about 29% lower with respect to the harmonic model. Therefore, the
eect of the characteristic features has been studied. Hereto the harmonic model
was extended by respectively the sawtooth displacement, trapezoidal angle of
attack, extra bump in angle of attack and the presence of deviation. The actual
vortex dynamics, as well as the resulting lift and drag histories were studied.
The results showed that the sawtooth amplitude used in the Roboy model
has a small eect on the mean lift but the mean drag is aected signicantly. Due
to the high acceleration during stroke reversal of the sawtooth shaped amplitude,
the mean drag at comparable lift is increased by 24.3%. The second model sim-
plication used by the Roboy, the trapezoidal angle of attack, caused the LEV
to separate during the translational phase. This led to an increase in mean drag
during each half-stroke. Also in this case large accelerations at stroke reversal lead
to a decrease in lift-to-drag ratio of 33.3%.
The extra bump in angle of attack, as used by the fruit y model, is not
aecting the mean lift to a large extent. During the beginning of the up and
downstroke the bump decreases the angle of attack such that the wing orientation
is almost horizontal. This leads to a signicant decrease in drag which improves
aerodynamic performance in the sense of lift-to-drag ratio by 15.6%. The other
realistic kinematic feature is the deviation, which is found to have only a marginal
eect on the mean lift and mean drag. However, the eective angle of attack is
altered such that the deviation leads to levelling of the force distribution.
The results from the present study show that special features of insect ight
have an appreciable eect on the accuracy of performance models of insect ight.
In particular they indicate that kinematic features, found in fruit y kinematics,
like the extra bump in angle of attack and deviation, may lead to drag reduction
or force levelling compared to harmonic kinematics.
CHAPTER 6
Vortex wake interactions of a
two-dimensional forward apping
foil
AIAA paper 2009-791.
A two-dimensional numerical investigation is performed to study the vortical ow
around a apping foil that models an animal wing, n, or tail in forward motion.
The vortex dynamics and performance are studied to determine the inuence of
foil kinematics. The baseline kinematic model is prescribed by harmonic functions
which can be characterised by four variables, the dimensionless wavelength, the
dimensionless apping amplitude, the amplitude of geometric angle of attack, and
the stroke plane angle. The foil motion kinematics has a strong inuence on the
vortex dynamics, in particular on the vortex-wake pattern behind the foil which
can be either periodic or aperiodic. Both symmetric and asymmetric solutions
are found. Evidence was found that the attachment of a leading-edge vortex
(LEV) is not signicantly advantageous for the force enhancement during the
full stroke. Plots of eciency versus the independent variable show that, for
symmetric kinematics, the largest eciency is achieved at an intermediate value
of each variable within the parameter range considered, where periodic ow occurs.
122 Vortex wake interactions of a two-dimensional apping foil
6.1 Introduction
The ow around a apping wing, n, or tail is highly unsteady and governed by the
dynamics of the generated vortices (Weish-Fogh & Jensen, 1956, Dickinson et al.,
2000). An experimental study (Lentink et al., 2008) showed that these shed vor-
tices interact with each other and organise themselves, similarly to an oscillating
cylinder as described by Williamson & Roshko (1988), into specic wake patterns
depending on the foil kinematics. The wake pattern can be either periodic or
aperiodic and directly determines the periodicity of the aerodynamic forces acting
on the foil. Periodic ow is the result of a match between the driving frequency
and the natural shedding frequency which is referred to synchronisation of the
ow (Williamson & Roshko, 1988, Lentink & Gerritsma, 2003). The wake will be
aperiodic if synchronisation of the vortex-wake does not occur. The synchronisa-
tion band organisation for the apping foil may be very complex due to the large
extent and high dimension of the parametric space. In contrast to the cylinder,
vortex shedding from a apping foil displays a variation of the natural shedding
frequency as a function of angle of attack (Katz, 1981, Dickinson & Gotz, 1993).
A numerical study by Lentink & Gerritsma (2003) showed that symmetric foil
kinematics can result in either a symmetric or an asymmetric wake. In the case
of an asymmetric wake, the initial condition determines the orientation of the
wake and hence the orientation of the time-averaged lift over a complete apping
period. Several studies have shown that the wing benets from the attachment of
the LEV because of the low pressure core of the LEV acting on the wing during
the full stroke (Lentink & Dickinson, 2009b,a, Ellington et al., 1996, Dickinson,
1994). However, the propulsive performance of plunging foil kinematics without a
pitching motion is poor (Lentink & Gerritsma, 2003). Therefore, it was suggested
that foil rotation is an important source for production of thrust to increase the
aerodynamic performance.
In the present research, we studied the vortex structure generated in the wake of
an ellipsoid foil undergoing apping motion, plunging and pitching, at a Reynolds
number of Re = 150 which corresponds to the ight of a small insect, e.g. a fruit
y. Here only the near wake of the foil is studied. The motivation for this is
that performance of a apping foil is inuenced mainly by near wake dynamics.
The objective of the present study is to investigate the inuence of dierent foil
kinematics on the vortex-wake structure, force coecients, and performance.
6.2 Flapping foil parametrisation
The baseline kinematic model is based on harmonic motion, such as used by Wang
(2000b,a). The ow around a apping foil and the foil kinematics can be char-
acterised by dimensionless parameters. The method used to make the govern-
ing equations dimensionless is the same as used by Lentink & Gerritsma (2003)
and Lentink et al. (2008). This approach enables us to perform a systematic inves-
6.2 Flapping foil parametrisation 123
tigation of the inuence of dierent foil kinematic parameters on the vortex-wake
pattern. The main important parameters are the frequency f [1/s] of both the
translation and rotation, which are coupled with a phase shift of 90
, the ampli-
tude of translation A [m], the amplitude of the sinusoidal foil rotation A
),
the forward velocity of the foil U
=
U
fc
.
The dimensionless amplitude A
=
A
c
.
The Strouhal number St is based on the stroke amplitude A, and is hence equal
to the ratio of the dimensionless amplitude A
:
St =
fA
U
=
A
.
It corresponds to the maximum induced angle of attack A
ind
at mid-stroke due
to the translation of the apping motion of the foil. The mean velocity U [m/s] is
obtained by averaging the velocity components over one apping period:
U =
1
T
_
T
0
_
(U
+U
ap
X
)
2
+ (U
ap
Y
)
2
dt .
Here T [s] is the period, U
ap
X
[m/s], and U
ap
Y
[m/s] the velocity of the foil kine-
matics in X and Y directions respectively. The time-averaged Reynolds number
Re becomes
Re =
Uc
,
where [m
2
/s] is the kinematic viscosity and changed for every computation to
match the Reynolds number. For the basic model the dimensionless wavelength
,
the dimensionless amplitude A
,
and the stroke plane angle are chosen as independent variables. The Reynolds
number is kept constant at Re = 150.
In order to study the inuence of the kinematics, each parameter is varied from
the baseline model, dened by
= 6.8, A
= 1.5, A
= 15
, and = 90
. The
dimensionless wavelength was varied from
3.0 with a 0.5 increment. The amplitude of angle of attack varies from
0
45
with a 15
90
(tan
1
)2Stsin()
2A
sin()
(a)
y
x
Y
X
2A
U
(b)
Figure 6.1 Schematic illustration of the foil kinematics in forward ight. (a) Illustration
of the foil parameters in forward ight: the dimensionless wavelength
= 15
and 45
in combination
with 15
90
with a 15
, (6.2)
P
req
=
1
T
_
T
0
d U
ap
dt
1
T
_
T
0
M
ap
dt . (6.3)
Here D [N] represents thrust, U
e
which leads to
positive drag. Thrusting modes are found for 12
5.7 the LEVs are shed before stroke reversal. The amount of LEVs
6.5 Results and discussion 127
pattern C
L
C
Ldownstroke
C
Lupstroke
C
D
24.0 no vortices 0.002 0.324 -0.320 0.218 -
20.0 no vortices 0.002 0.481 -0.478 0.185 -
12.0 2P+2S -0.001 0.925 -0.927 -0.028 4.71
10.0 2P+2S -0.003 1.095 -1.100 -0.092 10.66
7.9 2P+2S 0.009 1.495 -1.440 -0.186 13.05
6.8 2P+2S 0.005 1.633 -1.624 -0.252 13.47
6.3 2P+2S -0.004 1.704 -1.712 -0.289 13.48
6.0 2P+S -0.071 1.577 -1.719 -0.287 13.00
5.7 2P+S -0.103 1.580 -1.786 -0.302 12.70
5.3 2P 0.004 1.874 -1.833 -0.342 12.16
4.5 P+S 0.241 2.254 -1.772 -0.418 11.35
4.0 P+S 0.502 2.937 -1.934 -0.568 11.43
3.0 Aperiodic -0.494 2.034 -3.021 -0.617 9.06
Table 6.1 Inuence of dimensional wavelength. The numerical results are shown for 13 dierent
values for the dimensionless wavelength, A
= 1.5, A
= 15
, and = 90
.
and TEVs shed from the foil is decreasing with the dimensionless wavelength be-
cause the vortices have less time to develop and shed, see gure 6.3. Therefore,
the LEVs stay attached to the foil relatively longer at lower dimensionless wave-
lengths. The LEVs increase in size and strength due to increasing eective angle
of attack. As a result of this, the foil produces higher lift and thrust during each
half-stroke for decreasing dimensionless wavelength. A further decrease in dimen-
sionless wavelength results in stronger vortex-wake interactions which lead to an
aperiodic wake at
= 24, A
eff
= 7
(b) 2P+2S,
= 6.8, A
eff
= 39
(c) 2P+S,
= 6.0, A
eff
= 43
(d) 2P,
= 5.3, A
eff
= 46
(e) P+S,
= 4.5, A
eff
= 49
(f ) Aperiodic,
= 3, A
eff
= 53
Figure 6.3 Vorticity contours for decreasing wavelength. Vorticity contours of various wake
patterns for decreasing dimensionless wavelength
= 1.5, A
= 15
, and = 90
.
A
pattern C
L
C
Ldownstroke
C
Lupstroke
C
D
0.5 2S 0.002 0.376 -0.373 0.198 -
1.0 2P+2S 0.002 1.151 -1.147 -0.109 12.70
1.5 2P+2S 0.005 1.633 -1.624 -0.252 13.47
2.0 Aperiodic -0.038 1.561 -1.638 -0.217 9.94
2.5 Aperiodic -0.086 1.598 -1.770 -0.274 8.71
3.0 Aperiodic -0.064 1.830 -1.959 -0.302 7.13
Table 6.2 Numerical results of the kinematics for six dierent dimensionless amplitudes.
= 6.8,
A
= 15
, and = 90
.
6.5 Results and discussion 129
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
k = 24, No vortices
k = 6.8, 2P+2S
k = 6.0, 2P+S
k = 5.3, 2P
k = 4.5, P+S
k = 3.0, Aperiodic
0 0.2 0.4 0.6 0.8 1
-8
-6
-4
-2
0
2
4
6
8
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
k = 24, No vortices
k = 6.8, 2P+2S
k = 6.0, 2P+S
k = 5.3, 2P
k = 4.5, P+S
k = 3.0, Aperiodic
0 0.2 0.4 0.6 0.8 1
-2
-1.5
-1
-0.5
0
0.5
1
1.5
(b)
Figure 6.4 Force coecients to study the inuence of wavelength. Lift (a) and drag (b)
histories of dierent wake patterns for six dimensionless wavelengths. A
= 1.5, A
= 15
, and
= 90
.
A
pattern C
L
C
Ldownstroke
C
Lupstroke
C
D
0 2P+S 0.216 2.282 -1.850 0.050 -
15 2P+2S 0.005 1.633 -1.624 -0.252 13.47
30 2P 0.001 1.097 -1.094 -0.348 28.18
45 2P 0.004 0.588 -0.579 -0.089 14.58
Table 6.3 Numerical results of the kinematics for four dierent angle amplitudes.
= 6.8, A
= 1.5,
and = 90
.
form a LEV which leads to lift enhancement and thrust. For high dimensionless
amplitude A
2.0, vortices with a diameter larger than chord length are formed.
Some of these vortices are hit by the foil during stroke reversal. Strong foil-vortex
interactions lead to an aperiodic wake pattern causing aperiodic force coecients.
6.5.3 Inuence of angle of attack amplitude
Table 6.3 shows numerical results for dierent angle of attack amplitudes. The
plunging kinematic model, A
= 0
pattern C
L
C
Ldownstroke
C
Lupstroke
C
D
90 15 2P+2S 0.005 1.633 -1.624 -0.252 13.47
75 15 2P+S 0.518 2.124 -1.087 -0.103 5.42
60 15 2P 1.092 3.053 -0.868 0.303 -
45 15 Aperiodic 1.282 3.183 -0.619 0.521 -
30 15 Aperiodic 1.632 3.082 0.182 2.402 -
15 15 Aperiodic 1.160 2.264 0.056 2.464 -
90 45 2P 0.004 0.588 -0.579 -0.089 14.58
75 45 2P+S 0.600 1.262 -0.062 -0.002 0.36
60 45 3P+S 0.988 1.892 0.084 0.242 -
45 45 P+3S 1.344 2.624 0.063 0.626 -
30 45 P+2S 1.500 2.844 0.156 1.120 -
15 45 Aperiodic 1.768 3.240 0.296 2.045 -
Table 6.4 Inuence of the stroke plane angle. Numerical results of the kinematics for six dierent
stroke plane angles in combination with two dierent angle amplitudes.
= 6.8 and A
= 1.5
pared to other cases. This is because the thrust component of the resulting aero-
dynamic force is high compared to the normal component, due to the foil rotation.
At A
= 45
e
= 9
= 15
and A
= 45
.
A similar trend is found for both angle amplitudes that the lift coecient
is increasing for decreasing stroke plane angle during the downstroke until the
ow becomes aperiodic. Also the negative lift in the upstroke is decreasing in
magnitude. This is because during the downstroke of asymmetric kinematics the
foil undergoes a greater relative velocity. Therefore, the averaged lift is mainly
generated during the downstroke. The dierence in relative velocity between up
and downstroke also aects the drag contribution in a similar way.
6.5.5 Discussion
In the symmetric kinematics, non-zero average lift exists only as a result of an
asymmetry in wake pattern. The orientation of the mean lift depends on initial
6.5 Results and discussion 131
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
Aperiodic P+S 2P 2P+S 2P+2SNo vortices
C
L
(a) Mean lift coecient of symmetric kinematics.
Present study
Diptera
No vortices No vortices
Aperiodic
P+S
2P 2P+S
2P+2S
0 5 10 15 20 25
0
0.5
1
1.5
2
2.5
3
3.5
4
(b) Vortex-wake synchronisation A
dia-
gram.
Figure 6.5 Inuence of the kinematics on the vortex wake pattern and force generation.
(a) The mean lift coecient over a complete period of symmetric kinematics as a function of wake
pattern. (b) Vortex-wake synchronisation A
= 15
and = 90
and
as:
St =
A
>
1
2
tan(A
geo
+A
stall
) , (6.4)
which is illustrated by the dashed line. Besides, it is thought that the results could
also shed light on the Micro Air Vehicle (MAV) design. When the wing operates
outside the synchronisation region, the wake and consequently the forces become
aperiodic which will inuence the stability and controllability of the MAVs. Fig-
ure 6.6 shows plots of eciency versus the independent motion parameters. In
symmetric kinematics there is an optimal value for each variable, see gure 6.6(a),
(b) and, (c). The peak eciency of 28.18% could conrm that the wing rotation
plays an important role in the unsteady aerodynamic force production.
132 Vortex wake interactions of a two-dimensional apping foil
A
p
e
r
i
o
d
i
c
P
+
S
2
P
2
P
+
S
2
P
+
2
S
2 4 6 8 10 12 14
0
5
10
15
(a)
A
A
p
e
r
i
o
d
i
c
2
P
+
2
S
0 0.5 1 1.5 2 2.5 3 3.5 4
0
5
10
15
(b)
A
2
P
2
P
+
2
S
0 10 20 30 40 50
0
10
20
30
(c)
= 15
= 45
2P
2
P
+
S
2P+2S
60 65 70 75 80 85 90 95 100
0
5
10
15
(d)
Figure 6.6 Inuence of apping kinematics on the eciency. Eciency as a function of
the independence variables, (a) dimensionless wavelength, (b) dimensionless amplitude, (c) angle
amplitude and (d) stroke plane angle.
6.6 Conclusions
A numerical model for two-dimensional ow was used to investigate the eect of
foil kinematics on the vortex dynamics around an ellipsoid foil subjected to pre-
scribed apping motion over a range of dimensionless wavelengths, dimensionless
amplitudes, angle of attack amplitudes, and stroke plane angles at the Reynolds
number of 150. Both plunging and rotating motions are prescribed by simple har-
monic functions which are useful for exploring the parametric space despite the
model simplicity.
The resulting wake patterns behind the foil are categorised using the concept
of Williamson & Roshko (1988). Although such an attempt at classifying vortex
patterns can lead to confusion due to the shedding, tearing, or merging of tiny
vortices, it is suitable for straightening out the shedding vortices in our simulations.
The results are in satisfactory agreement with the comparable experiments.
6.6 Conclusions 133
Optimal propulsion using apping foil exists for each variable, which implies
that aerodynamics might select a range of preferable operating conditions. The
conditions that give optimal propulsion lie in the synchronisation region in which
the ow is periodic. Since the computational costs are high and the parameters
cannot be varied continuously, the synchronisation band was not investigated com-
pletely. However, the present study is benecial for understanding the inuence
of wing kinematics on the performance characteristics.
CHAPTER 7
Vortical structures in
three-dimensional apping ight
Submitted to J. Fluid Mech. (January 2010).
Results are obtained by performing numerical simulations of the three-dimensional
ow around a apping wing. A parameter study is performed to investigate the
performance in apping ight and to get insight into the vortex dynamics and force
generation. Dierent aspects, relevant for three-dimensional apping wing aero-
dynamics, have been studied, namely the angle of attack, the Rossby number, the
Reynolds number and the stroke kinematics. First, the ow around a dynamically
scaled model wing is solved for dierent angles of attack in order to study the force
development and vortex dynamics at small and large mid-stroke angles of attack.
Secondly, the Rossby number is varied at dierent Reynolds numbers. A varying
Rossby number represents a variation in the radius of the stroke path and thus
the magnitude of the angular acceleration. Thirdly, the three-dimensional wing
kinematics is varied by changing the shape in angle of attack and by applying a
deviation, which may result in a gure-of-eight, a gure-of-O or a gure-of-U
pattern. Finally, the three-dimensional ow is compared with the two-dimensional
studies performed on apping forward ight.
7.1 Introduction
To understand the aerodynamic performance of apping wings at low Reynolds
numbers, relevant for insect ight, it is important to obtain insight into the vortex
136 Vortical structures in three-dimensional apping ight
dynamics and its inuence on force development. The most important feature
in apping wing aerodynamics has been established to be the generation of a
stable leading-edge vortex (LEV) on top of the wing, which is responsible for the
unexpectedly large force augmentation in hovering insect ight (Maxworthy, 1979,
Ellington et al., 1996, Dickinson et al., 1999, Srygley & Thomas, 2002, Lentink
& Dickinson, 2009b). In order to gain insight into the three-dimensional ow
eld induced by the apping wings, several two-dimensional studies have been
performed (Dickinson & Gotz, 1993, Dickinson, 1994, Wang, 2005, Bos et al., 2008).
It was shown that the leading-edge vortex generated by a two-dimensional moving
foil is shed after several travelled chord lengths, while a three-dimensional LEV
remains stably attached to a three-dimensional revolving (Usherwood & Ellington,
2002) or apping (Dickinson et al., 1999, Lehmann, 2004, Birch et al., 2004) model
wing, which rotates around its base. Those results indicate that three-dimensional
ow eects are essential for the LEV stability. Previously conducted research
addressed a possible analogy between the LEV on apping wings and the LEV
generated by swept and delta wings (Ellington et al., 1996, Van Den Berg &
Ellington, 1997). The spiral leading-edge vortex generated by a translating swept
or delta wing is stabilised by the induced spanwise ow, which could suggest
that a spanwise ow may play an important role concerning the LEV stability in
insect ight (Ellington et al., 1996, Van Den Berg & Ellington, 1997). Lentink
& Dickinson (2009b) discussed that the stability of the LEV growth specically
might be increased by the spanwise ow through the LEV core, driven by either
the dynamic pressure gradient on the wings surface, the centrifugal acceleration of
the boundary layer or the induced velocity eld of the spiral vortex lines (Ellington
et al., 1996). Additionally, the LEV stability may be strengthened by a reduction
of the eective angle of attack as a result of the tip vortex generation (Birch &
Dickinson, 2001, Shyy et al., 2008b). However, Birch & Dickinson (2001) showed
no signicant eect of the spanwise ow on the LEV strength and stability, using
plates at dierent spanwise locations to block the spanwise ow, but they did not
completely explain the LEV stability in their experiments.
In order to investigate the vortex dynamics and the stability of the leading-edge
vortex in particular, an accurate simulation method is developed to perform a CFD
simulation of a three-dimensional apping wing. Based on the discussion about
LEV stabilisation due to wing revolving (Usherwood & Ellington, 2002, Birch et
al., 2004, Lentink & Dickinson, 2009b,a, Bos et al., 2010b), a three-dimensional
wing was modelled which was able to ap around a base of which the location
can be varied. By varying the location of the centre of rotation, the inuence of
the revolving strength (Rossby number) and the eect of the tip vortices can be
studied. In addition, the kinematics is varied from simple harmonics by adding a
deviation and trapezoidal shaped angle of attack. Recent two-dimensional sim-
ulations (Bos et al., 2008) suggested that the wing kinematics may also have a
large inuence on the apping performance in three-dimensional hovering. Addi-
tionally, (Lentink, 2008) showed interesting results concerning the stability of the
three-dimensional leading-edge vortex depending on the Rossby number (equiv-
7.2 Three-dimensional apping wing simulations 137
F
X
F
Y
F
Z
(t)
(t)
(t)
O
R
root
R
tip
start downstroke
start upstroke
mid-stroke
Stroke plane
F
drag
F
span
F
normal
Figure 7.1 Illustration of the wing motion and force denitions. Illustration of the wing
motion and force denitions. (t) corresponds to the stroke variation, (t) to the geometrical angle of
attack and (t) to the deviation from the horizontal stroke plane.
alent to the stroke path curvature) and the Reynolds number. Therefore, the
Rossby number is systematically varied for dierent Reynolds numbers and mid-
stroke angles of attack. In agreement with (Bos et al., 2008), the kinematic model
is extended with a trapezoidal shaped angle of attack and a non-zero deviation
is applied. Dierent deviation patterns are investigated, following the shape of
gure-of-O, gure-of-U and gure-of-eight.
The apping wing modelling is described in section 7.2, which also addresses
wing geometry, kinematic models and the simulation strategy. In order to show
that the CFD method is accurate and ecient, section 7.3 briey discusses the
validation and verication of the ow solver. The vortical ow needs to be visu-
alised in such a way that the resulting vortices are clearly visible. Dierent vortex
identication methods are described in 7.4. Furthermore, the results are discussed
in 7.5 and 7.6, while the conclusions are summarised in 7.7.
7.2 Three-dimensional apping wing simulations
In order to study the vortex dynamics and stability of the leading-edge vortex, the
ow is solved using Computational Fluid Dynamics (CFD), of which the details
are described in chapter 2. The three-dimensional apping wing is modelled in
order to provide a framework for comparison, which is still representative for true
insect ight, this is the subject of section 7.2.1. In view of limiting computing
resources, a selection of geometric and kinematic parameters is made to systemi-
cally investigate the ow phenomena of our interest. Additionally, the simulation
strategy is discussed in section 7.2.2.
138 Vortical structures in three-dimensional apping ight
7.2.1 Modelling and parameter selection
In general, most investigations concerning apping wing aerodynamics make use
of the modelling convention as previously described by Sane & Dickinson (2002)
and Dickson & Dickinson (2004) as applied in the experiments with a dynamically
scaled robotic fruit y wing. The current research uses a model wing with an
ellipsoidal shaped planform with 10% thickness, since Lentink & Gerritsma (2003)
showed that airfoil shape was of minor inuence on the forces and the ow eld. In
addition, Luo & Sun (2005) showed that the airfoil corrugation, present in dragon-
y wings, did not inuence the force development signicantly. The length scales
of the corrugation are orders of magnitude smaller compared to the length scale
of the separated ow region or the leading-edge vortex, such that signicant eect
of corrugation on the ow can be neglected.
Planform selection
The single wing span is xed to b
s
= 2.0 and the chord at mid-span is c = 1.0.
The hinge around which the wing is able to ap is xed to a distance of 0.5 from
the wing root, such that the wing tip radius becomes R
tip
= 2.5, while the rota-
tional distance of the Roboy was xed to 0.7 (Sane & Dickinson, 2002, Poelma
et al., 2006). Since the wing planform is chosen to be ellipsoidal, the wing sur-
face is dened by S = ab, where a = 0.5 and b = 1.0 are the semi-minor and
semi-major axes, respectively, such that S = /2. The average chord length of
this ellipsoidal planform is found to be c = S/b
s
= /4. So, the three geomet-
ric parameters important for the apping wing simulations are dened: b
s
, S and c.
Kinematic models
The apping wing motion is prescribed by three dierent motion angles dening
the deviation angle (t), apping angle (t) and the angle of attack (t). The de-
viation angle is the angle with respect to the horizontal stroke plane, as described
in chapter 4 and (Bos et al., 2008). During the stroke, the deviation is varied
harmonically with an amplitude between A
= 0 and A
= 20
. A combined devi-
ation and apping angle variation leads to a wing tip pattern. Depending on the
variation of the deviation angle, gure 7.2 shows the resulting gure-of-O, 7.2(a),
gure-of-eight, 7.2(b), or gure-of-U, 7.2(c).
Realistic fruit y kinematics (Fry et al., 2003, Lentink & Dickinson, 2009b,
Bos et al., 2008) resembles an harmonically varying deviation angle, a sawtooth
shaped apping angle and a trapezoidal shaped angle of attack (Sane & Dick-
inson, 2002, Dickson & Dickinson, 2004) with an incidental bump, shortly after
stroke reversal (Bos et al., 2008). A two-dimensional investigation (Bos et al.,
2008) showed that the eect of the trapezoidal angle of attack was most promi-
nent.
From the discussion in section 7.1 it can be concluded that there is need for a
detailed three-dimensional numerical study to investigate the eects of the Rossby
number, Reynolds number, angle of attack and stroke kinematics, i.e. trapezoidal
7.2 Three-dimensional apping wing simulations 139
Flapping angle, [
]
D
e
v
i
a
t
i
o
n
a
n
g
l
e
,
]
[
-
]
0 20 40 60 80 100 120 140 160 180
-80
-60
-40
-20
0
20
40
60
80
upstroke
downstroke
(a) gure-of-O
Flapping angle, [
]
D
e
v
i
a
t
i
o
n
a
n
g
l
e
,
]
[
-
]
0 20 40 60 80 100 120 140 160 180
-80
-60
-40
-20
0
20
40
60
80
upstroke
downstroke
(b) gure-of-eight
Flapping angle, [
]
D
e
v
i
a
t
i
o
n
a
n
g
l
e
,
]
[
-
]
0 20 40 60 80 100 120 140 160 180
-80
-60
-40
-20
0
20
40
60
80
upstroke
downstroke
(c) gure-of-U
Figure 7.2 Dierent wing tip patterns. Dierent wing tip patterns as a result of the variation
in deviation with a combined apping motion. (a) gure-of-O. (b) gure-of-eight. (c) gure-of-U.
shape and deviation. The current study varied the Rossby number from Ro = 3.2,
which is relevant for vortex induced propulsion in nature (Lentink & Dickinson,
2009a,b), to a nearly translating wing, Ro = 130. In addition to the variation of
Reynolds number from Re = 100, 500 and 1000, the (geometric) angle of attack is
varied from = 15
to = 90
with increments of = 15
R
g
, the Reynolds number at the radius of gyration Re
R
g
,
and the area swept by the wing A
swept
. Using R
g
=
_
1
S
_
R
0
r
2
c(r)dr, the radius
of gyration is determined from the rotation origin to the tip r = 0 to R
tip
. For
the baseline case, where R
tip
= 2.5 the radius of gyration becomes R
g
= 1.58.
Using (4.23), the average Reynolds number, based on the radius of gyration is
dened as:
Re
R
g
=
4A
fR
g
c
, (7.1)
where A
R
g
= A
R
g
/c. On the other hand, maximal values, occurring at the
wing tip are still varying like Re
R
, U
R
and A
R
= A
R
g
2.2, which is of similar order as used in the two-dimensional analysis in (Bos
et al., 2008).
In order to investigate the eect of three-dimensional wing kinematics in hov-
ering ight, which is the main subject of the present thesis, the hovering wing
kinematics is substituted into the expressions for the angular and centripetal co-
ecients, C
ang
and C
cen
, and the Rossby number Ro, equation (4.6) to nd the
following:
C
ang
=
2
R
g
c
= A
R
g
, (7.2)
C
cen
=
R
tip
c
= AR
s
, (7.3)
Ro =
R
tip
c
= AR
s
. (7.4)
Here, AR
s
is the single wing aspect ratio. It remains clear that both the centripetal
C
cen
, and the Rossby number Ro, are dened by the wing geometry, whereas the
angular acceleration number C
ang
, depends on the wing kinematics.
Force and performance denitions
In order to determine the eect of the dierent motion and geometric parameters
on the forces and performance, proper denitions are necessary. Since the wing
rotates with a rotating reference frame, two dierent force denitions are possible,
7.2 Three-dimensional apping wing simulations 141
in the inertial and the rotating reference frame, which is shown in gure 7.1 in re-
lation to the motion angles. Because the present research concerns hovering ight,
the lift force is by denition vertical and thus equal to F
Y
. On the other hand,
the drag force is opposite to the motion direction, derived by a decomposition of
F
X
, F
Y
and F
Z
in the rotating reference frame. Therefore, the three-dimensional
lift and drag are given by:
F
lift
= F
Y
, (7.5)
and
F
drag
= F
X
sin() F
Z
cos(). (7.6)
The force in spanwise direction is not used throughout our analysis, since that
force is small compared to the lift and drag. As discussed previously (Bos et al.,
2008) in chapter 4, the force coecients, C
L
and C
D
, are obtained by division
using the average dynamic pressure, q = 0.5U
2
ref
. Two performance indicators
are used, the lift-to-drag ratio, also known as the glide factor, C
L
/C
D
and the
power factor, C
3/2
L
/C
D
(see Ruijgrok, 1994).
7.2.2 Simulation strategy and test matrix selection
The following variables are varied throughout the current research, the Rossby
number Ro (due to varying rotation origin), the angle of attack amplitude , the
Reynolds number Re and the wing kinematic model. In order to investigate the in-
uence of the wing kinematics, the shape of the angle of attack variation, reected
by T
rot
and the deviation amplitude, A
to = 90
with increments of = 15
. This
provides insight in the force development as a function of angle of attack for a
fully revolving, i.e. apping, and translating wing. Additionally, the Reynolds
number is varied from Re = 100, Re = 500 and Re = 1000, including a variation
in angle of attack, to study its eects on the behaviour of the leading-edge vortex.
142 Vortical structures in three-dimensional apping ight
Rotation origin, R
tip
2.5 3.0 4.0 5.0 6.0 7.0 12.0 102.0
A
]
90 + +
75 + +
60 + + + + + + + +
45 + + + + + + + +
30 + +
15 + +
Table 7.1 Simulation matrix: wing stroke curvature origin, angle of attack and Reynolds
number. Simulation matrix to vary the rotation origin of the apping wing. This matrix is used to
study the inuence of the stroke curvature on the structure of the leading-edge vortex and corresponding
forces. The angle of attack at mid-stroke is varied from 90
to 15
_
A
sin(2ft) 0 t < T
rot
,
A
T
rot
t <
1
2
T T
rot
,
A
cos(2ft)
1
2
T T
rot
t < T
rot
+
1
2
T,
A
T
rot
+
1
2
T t < T T
rot
,
A
cos(2ft) T T
rot
t < T.
(7.7)
Here T
rot
is the rotation duration, such that T
rot
= 0.25 recovers a fully harmonic
angle of attack variation. For dierent values of T
rot
, the angle of attack is plotted
in gure 7.3. Note that the geometric angle of attack is given by
geom
=
2
.
Table 7.2 shows dierent deviation amplitudes A
0 5 10 15 20
T
r
o
t
0.25 + + + + +
0.20 +
0.15 +
0.10 + + + + +
Table 7.2 Simulation matrix: kinematic modelling. Simulation matrix to vary the rotation
duration and the deviation of the apping stroke. The rotation duration is varied from T
rot
= 0.25
to T
rot
= 0.10 in order to get a trapezoidal shaped angle of attack. The deviation is varied by the
deviation amplitude A
and 60
, which is
shown to result in maximal lift coecients. The Reynolds number is xed to Re = 100, the grid
resolution to 800k and the time-step was chosen corresponding to Co
max
= 2.0.
t/T [-]
[
-
]
T
rot
= 0.10
T
rot
= 0.15
T
rot
= 0.20
T
rot
= 0.25
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
50
60
70
80
90
100
110
120
130
Figure 7.3 Angle of attack variation with a trapezoidal shape. To investigate the inuence
of a trapezoidal shape in angle of attack, the amount of this shape is systematically varied by T
rot
.
7.3 Flow solver accuracy
In order to test the accuracy of the used ow solver, concerning highly unsteady
and vortical ows, numerical comparisons are performed. A verication is per-
formed by varying the grid resolution (grid independence study) and the time-
step size, by decreasing the maximum Courant number. The meshes for these
three-dimensional simulations are constructed with GridPro
using a structured
approach. Grid renement is uniform and the cells are clustered close to the ap-
ping wing boundary. More detailed information on grid generation can be found
in appendix A.
In order to show that the numerical solution is grid and time-step indepen-
dent, a verication study is performed using the ow around a three-dimensionally
apping wing. The kinematics is according to the simple harmonic model. The
apping angle amplitude was xed to A
= 63
= 45
= 63
, A
= 45
and A
= 0
.
grid resolution and maximal Courant number. The smaller Co
max
, the smaller
the time-step. The grid resolution was varied from 100k to 1600k cells and the
Co
max
from 2.0 to 0.5. The spatial grid independence study was performed for a
maximal Courant number of Co
max
= 1.0 and the temporal convergence for two
grid sizes of 400k and 800k cells.
In order to assess the accuracy of the ow solver, the drag and lift coecients
are plotted in gure 7.4 for meshes from 100k to 1600k cells and Co
max
= 1.0. The
corresponding limit cycles are shown in gure 7.5. As can be clearly seen from
the gures, the ow is periodic and the force coecients (lift and drag) appear
to be close for the grid resolutions considered. In order to assess spatial and
temporal convergence both time-averaged lift and drag coecients are plotted with
increasing spatial and temporal resolution in gure 7.6. Figure 7.6(a) and 7.6(b)
shows a converging time-averaged lift and drag coecients with increasing grid
resolution, the value at is determined by Richardson extrapolation (Ferziger &
Peric, 2002), see chapter 2. Temporal convergence is shown in gure 7.6(c), where
the time-averaged lift and drag coecients are plotted with decreasing time-step.
In order to justify the choice for grid and temporal resolution for the three-
dimensional apping wing simulations, table 7.4 and 7.5 show the spatial and
temporal errors in average lift and drag with the Richardson extrapolated values.
Table 7.4 shows that even the dierences in lift and drag for 100k mesh cells and
Co
max
= 1.0 are reasonably small, i.e. less than 4%. This may be explained by
the fact that the forces are mainly dependent on the near wake, with several chord
lengths from the wing. Apparently, all grid resolutions considered, from 100k to
1600k, are suciently ne to capture the near wake, on which the forces depend.
However, in order to visualise the vortices in the far wake (Bos et al., 2008) at least
400k, but preferably 800k mesh cells are desired. Table 7.5 shows the errors in lift
and drag with respect to the Richardson extrapolated values for both 400k and
800k with decreasing time-step (Co
max
= 2.0, 1.0 and 0.5). Again, the dierences
are small, even for the largest time-step, corresponding to Co
max
= 2.0, the error
is less than 0.2%.
Summarising, all generated grids provide suciently accurate force coecients,
but to capture the far wake vortex dynamics at least 800k cells are required.
7.4 Vortex identication methods for ow visualisation 145
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
100k
200k
400k
800k
1600k
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
2.5
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
100k
200k
400k
800k
1600k
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-3
-2
-1
0
1
2
3
(b)
Figure 7.4 Three-dimensional verication: force coecients. Lift and drag coecients for the
verication cases with varying grid size, 100k 1600k. The time-step is taken such that Co
max
= 1.0.
The kinematics of the three-dimensional wing is simple harmonics with A
= 63
, A
= 45
and
A
= 0
.
N
lift
[%]
drag
[%]
100k -3.66 -3.19
200k -2.52 -2.32
400k -1.27 -1.23
800k -0.49 -0.50
1600k -0.12 -0.13
Table 7.4 Error values of lift and drag coecients for varying grid sizes. Error values of
lift and drag coecients [%] with respect to the extrapolated values for varying grid sizes, ranging
from 100k to 1600k. The time-step was determined by a max Courant number of Co
max
= 1.0. The
kinematics of the three-dimensional wing is simple harmonics with A
= 63
, A
= 45
and A
= 0
.
Furthermore, at Co
max
= 2.0, the temporal errors are suciently small. Therefore,
a grid resolution of 800k in combination with Co
max
= 2.0 was used for all three-
dimensional apping wing simulations, described in this chapter.
7.4 Vortex identication methods for ow
visualisation
In order to study the vortex dynamics and the stability of the leading-edge vortex
in particular, a proper vortex identication criterion is essential. Dierent well-
known techniques to detect and visualise vortices are based on the velocity gradient
tensor, which requires a complete velocity eld. Two dierent vortex identication
criteria are discussed, namely the magnitude of vorticity || (Lu & Shen, 2008)
146 Vortical structures in three-dimensional apping ight
Drag coecient [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
-3 -2 -1 0 1 2 3
-0.5
0
0.5
1
1.5
2
2.5
3
Figure 7.5 Three-dimensional verication: limit cycle. The limit cycles, constructed from the
lift and drag coecients clearly shows periodic behaviour for the verication cases. The grid size was
xed to 800k. The time-step is taken such that Co
max
= 1.0. The kinematics of the three-dimensional
wing is simple harmonics with A
= 0 rad, A
= 0.785 rad.
Co
max
lift
[%] 400k
drag
[%] 400k
lift
[%] 800k
drag
[%] 800k
2.0 0.082 0.183 0.067 0.159
1.0 0.053 0.097 0.043 0.087
0.5 0.013 0.024 0.011 0.022
Table 7.5 Error values of lift and drag coecients for varying temporal resolution. Error
values of lift and drag coecients [%] with respect to the extrapolated values for varying temporal
resolutions, corresponding to Co
max
= 2.0 to Co
max
= 0.5. Shown are the errors for two grid sizes,
400k and 800k. The kinematics of the three-dimensional wing is simple harmonics with A
= 63
,
A
= 45
and A
= 0
.
and the Q criterion (Hunt et al., 1988).
The rst criterion is based on the magnitude of vorticity ||, where = u.
If || reaches a user-dened threshold, that region is identied as a vortex. More
specically, within this particular region, there is a concentration of vorticity. Since
shear layers and curved streamlines also are a source of vorticity, this criterion may
lead to undesired contours of e.g. shear layers. Especially in three-dimensional
ows this may become a diculty. In two-dimensional ow, however, || is the
common vortex visualisation method (Bos et al., 2008).
The Q criterion (Hunt et al., 1988) is the second invariant of the local velocity
gradient tensor u. For Q > 0 the region is identied as a vortex. This second
invariant of u is written as
Q =
1
2
_
||
2
|S|
2
_
,
where the rate of strain tensor S is given by S =
1
2
(u +u
T
) and the vorticity
tensor by =
1
2
(u u
T
). Hence, a positive value of Q > 0 is a measure
for any excess of rotation rate (in terms of vorticity) with respect to the strain
7.4 Vortex identication methods for ow visualisation 147
Spatial resolution [-]
A
v
e
r
a
g
e
l
i
f
t
c
o
e
c
i
e
n
t
[
-
]
12 4 8 16
1.18
1.19
1.2
1.21
1.22
1.23
1.24
(a)
Spatial resolution [-]
A
v
e
r
a
g
e
d
r
a
g
c
o
e
c
i
e
n
t
[
-
]
12 4 8 16
1.97
1.98
1.99
2
2.01
2.02
2.03
2.04
2.05
(b)
Temporal resolution [-]
A
v
e
r
a
g
e
d
r
a
g
c
o
e
c
i
e
n
t
[
-
]
400k
800k
12 4 8 16
2.015
2.02
2.025
2.03
2.035
(c)
Temporal resolution [-]
A
v
e
r
a
g
e
l
i
f
t
c
o
e
c
i
e
n
t
[
-
]
400k
800k
12 4 8 16
1.21
1.212
1.214
1.216
1.218
1.22
1.222
1.224
1.226
1.228
1.23
(d)
Figure 7.6 Spatial and temporal convergence. (a) and (b) are showing the average lift and drag
coecients for increasing spatial resolution and constant time-step, corresponding to Co
max
= 1.0.
The nal value, at , is obtained using Richardson extrapolation. The temporal convergence is
illustrated in (c) and (d), showing the average drag and lift coecients for decreasing time-step at
two specic grid sizes, 400k and 800k.
rate. Therefore, a region where Q > 0 indicates a clear swirling ow (as shown
by Chakraborty et al., 2005). It must be noted that Jeong & Hussain (1995)
found that Q > 0 is not a sucient condition to have a pressure minimum in the
vortex core of that specic region. In most cases, however, a pressure minimum
does occur. By neglecting these unsteady and viscous eects from the governing
Navier-Stokes equations the following relation can be obtained for the symmetric
tensor
2
+S
2
:
2
+S
2
=
1
(p),
where is the uid density and p the pressure. In order to identify which vortex
criterion should be used, gure 7.7 shows iso-surfaces around a apping wing.
The wing aps around a distance of 0.5 from the wing root and the apping
angles are varying harmonically. The iso-surface is visualised at t = 0.25T, during
the downstroke. At t = 0.25T the leading-edge vortex is formed on the wings
148 Vortical structures in three-dimensional apping ight
(a) || (b) Q
Figure 7.7 Comparison of the near wake ow eld using dierent vortex identication
criteria. The spiralling leading-edge vortex is visualised using contour plots of the magnitude of
vorticity, || and Q. The arrow shows the apping direction of a downstroke and the ow is visualised
at mid-stroke. (a) shows the contour of the vorticity magnitude, || = 5.0 and (b) the Q = 2.0 value.
The colours represent values of helicity, h = (u )/(|u| ||), within the range of 1.0 h 1.0.
upper surface and rolls up into a tip vortex, the vortices from the previous stroke
should be visible as well. The leading-edge vortex, rolling up into a tip vortex, is
identied using a carefully chosen threshold of the vortex identication criteria,
using the values || = 5.0 and Q = 2.0, normalised by their maximal values. The
colours show the helicity which is dened as h = (u )/(|u| ||), within the range
of 1.0 h 1.0. A positive helicity, h > 0, means that the direction vector of
vorticity ( = u) is aligned with the local ow velocity.
In gure 7.7(a) it can be observed that || shows not only the vortical struc-
tures, but also the shear layers near the wing and between the vortices. This leads
to a thicker iso-surface, such that detail of the vortical structures is lost. The Q cri-
terion shows more detail, in gure 7.7(b), a smooth leading-edge vortex is shown,
rolling up into a tip vortex. Some of the previously shed vortices are still present.
Since the Q criterion oers sucient and adequate information about the local
ow eld, e.g. a rotation dominated region is identied by Q < 0, this criterion is
used throughout the remainder of the present research, for all three-dimensional
simulations.
7.5 Flapping wings at low Reynolds numbers
In order to provide insight into the vortex dynamics (for a purely harmonic ap-
ping motion) and its inuence on the variation of forces, dierent geometric and
kinematic parameters are systematically varied. First, the inuence of the angle
of attack on the forces is briey discussed in section 7.5.1. In that section, the in-
uence of an increasing mid-stroke angle of attack is briey discussed on the force
development. In order to investigate the eect of the Rossby number, the radius
7.5 Flapping wings at low Reynolds numbers 149
of curvature is subsequently varied in section 7.5.2, this is performed for dierent
Reynolds number as well. The inuence of the Reynolds number on the forces
and leading-edge vortex stability is assessed in 7.5.3. Additionally, the kinematic
model is varied by considering a trapezoidal shape adaptation and the addition
of deviation in section 7.5.4 and 7.5.5, respectively. In addition to the variety of
hovering ight simulations, section 7.6 deals with forward apping ight with sim-
ilar conditions as the two-dimensional simulations performed by Bos et al. (2008),
Thaweewat et al. (2009).
7.5.1 The angle of attack in apping ight
Previous studies showed that the angle of attack variation during the stroke in-
uences the forces considerably. This was conrmed by a recent two-dimensional
investigation (Bos et al., 2008), concerning hovering ight at fruit y conditions.
It was already mentioned that the angle of attack at mid-stroke is varied from
= 15
to = 90
. Note that = 90
. Furthermore, it can be
seen in gure 7.8 that the maximal force values occur at approximately halfway
through the down and upstroke, t = 0.25T and t = 0.75T, where T is the apping
period.
Furthermore, it must be noted that the lift is nearly always non-negative,
which means that during the hovering conditions, lift is being generated during
the complete stroke. The drag for = 15
and = 60
) and minimal ( = 15
geom
C
L
C
Ddownstroke
C
Dupstroke
C
L
/C
Dave
90 0.000 3.546 3.543 0.000
75 0.703 3.325 3.329 0.178
60 1.127 2.750 2.739 0.436
45 1.224 2.034 2.028 0.667
30 0.977 1.339 1.333 0.722
15 0.526 0.963 0.957 0.397
Table 7.6 Force coecients for Re = 100 and Ro = 3.2. Time-averaged lift C
L
, drag C
D
and
lift-to-drag C
L
/C
Dave
are shown as a function of the mid-stroke geometrical angle of attack for given
Re = 100 and Ro = 3.2, so a apping wing with small stroke curvature.
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
= 15
= 30
= 45
= 60
= 75
= 90
c
i
e
n
t
[
-
]
= 15
= 30
= 45
= 60
= 75
= 90
,
: 30
, : 45
, : 60
, : 75
and : 90
.
7.5 Flapping wings at low Reynolds numbers 151
Ro C
L
C
Ddownstroke
C
Dupstroke
C
L
/C
Dave
3.2 1.224 (baseline) 2.034 2.028 0.603 (baseline)
3.8 1.175 (3.9%) 1.983 2.981 0.593 (1.6%)
5.1 1.105 (9.7%) 1.933 1.935 0.572 (5.2%)
6.4 1.058 (13.6%) 1.916 1.915 0.552 (8.4%)
7.6 1.023 (16.4%) 1.908 1.904 0.537 (11.0%)
8.9 0.997 (18.5%) 1.903 1.899 0.525 (13.0%)
15.3 0.943 (22.9%) 1.892 1.886 0.500 (17.2%)
130 0.922 (24.7%) 1.883 1.877 0.490 (18.7%)
Table 7.7 Time-averaged force coecients for varying Rossby numbers at Re = 100. The
variation of average lift (C
L
), drag (C
D
), lift-to-drag ratio (C
L
/C
D
) are shown for Rossby numbers
from Ro = 3.2 to Ro = 130. The mid-stroke angle of attack was xed to = 45
. It is
clear that both lift and drag are decreasing with increasing Rossby number, i.e.
decreasing curvature of the stroke path. At Ro = 130 the wing nearly performs
a two-dimensional motion leading to a decrease in lift of 24.7% in comparison to
the baseline case with Ro = 3.2. The decrease in drag is small, such that the
decrease in lift-to-drag ratio is still signicant, 18.7%. Figure 7.9 shows the force
histories concerning the nearly translating wing, i.e. Ro = 130. When compared
with gure 7.8 (which applies to the revolving wing Ro = 3.2) it is seen that
both lift and drag variations are signicantly lower when the Rossby number is
large. This is due to the loss in energy by the tip vortices which was also studied
by (Blondeaux et al., 2005a). Figure 7.10 shows the variation of the lift and
drag coecients during the apping cycle and the eect of Rossby number as it
increases. It is clear that the major loss in lift for high Rossby numbers, occurs at
mid-stroke t=0.25T and t=0.75T. The loss in drag, just after stroke reversal and
during mid-stroke are of similar magnitude.
Figure 7.11 shows the iso-surfaces of Q = 1.0 at t = 0.25T, which is at the
midst of the downstroke. It can be clearly observed that the fully apping wing
shows a pronounced leading-edge vortex which spirals towards the tip to form a
tip vortex. However, the translating wing shows a leading-edge vortex which stays
symmetric with respect to the wing centre plane, feeding two wing tip vortices, at
the root and the tip. Additionally, gure 7.13 shows the streamlines to illustrate
the leading-edge and tip vortices in more detail.
An additional observation, while comparing gures 7.8 and 7.9 is that the lift
drops signicantly (75% at least) during mid-stroke (t = 0.2T) for an angle of
attack of = 15
and = 45
c
i
e
n
t
[
-
]
= 15
= 30
= 45
= 60
= 75
= 90
c
i
e
n
t
[
-
]
= 15
= 30
= 45
= 60
= 75
= 90
, : 30
, : 45
, : 60
, : 75
and : 90
.
lift, compared to = 45
R
g
, average
Reynolds number Re
R
g
and swept area A
swept
are comparable.
Summarising, it can be stated that a apping wing motion is of crucial impor-
tance for lift generation at a small penalty of drag, compared to wing translation.
Additionally, the leading-edge vortex is important for the gain in lift. This leading-
edge vortex is larger and more stable at angles of attack larger than about 30
.
At smaller angles of attack, it was shown for both apping and translating wings
at = 15
c
i
e
n
t
[
-
]
Ro = 3.2
Ro = 3.8
Ro = 6.4
Ro = 8.9
Ro = 130
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
0
0.5
1
1.5
2
2.5
3
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
Ro = 3.2
Ro = 3.8
Ro = 6.4
Ro = 8.9
Ro = 130
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-4
-3
-2
-1
0
1
2
3
4
(b)
Figure 7.10 Variation of force coecients for a apping wing with varying Rossby num-
bers. Lift (a) and drag (b) coecients induced by a apping wing around varying Rossby numbers,
Ro = 3.2 130. The amplitude of the angle of attack variation was xed such that at mid-stroke
= 45
. The average Reynolds number remained xed at Re = 100. The Rossby number is varied
from : 3.2, : 3.8, : 6.4, : 8.9, : 130.
(a) Ro = 3.2
(b) Ro = 130
Figure 7.11 Vortex visualisation of the near wake for dierent Rossby numbers. Iso-
surfaces of Q = 1.0 are shown for Rossby numbers Ro = 3.2 and Ro = 130. A time-frame is shown at
mid-stroke, t = 0.25T. The average Reynolds number was xed to Re = 100. Colours indicate helicity.
154 Vortical structures in three-dimensional apping ight
(a) = 45
(b) = 15
Figure 7.12 Vortex visualisation of the near wake for a translating wing at low and high
angle of attack. Iso-surfaces of Q = 1.0 are shown for = 15
and = 45
c
i
e
n
t
[
-
]
Re=100, Ro=3.2
Re=100, Ro=130
Re=500, Ro=3.2
Re=500, Ro=130
Re=1000, Ro=3.2
Re=1000, Ro=130
0.5 1 1.5 2 2.5 3 3.5 4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
(a)
Glide ratio [-]
P
o
w
e
r
f
a
c
t
o
r
[
-
]
Re=100, Ro=3.2
Re=100, Ro=130
Re=500, Ro=3.2
Re=500, Ro=130
Re=1000, Ro=3.2
Re=1000, Ro=130
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.2
0.4
0.6
0.8
1
(b)
Figure 7.14 Force and performance polars. (a) shows the force polars as a function of angle
of attack amplitude. (b) illustrates the power factor C
L
3/2
/C
D
as a function of the lift-to-drag ratio
C
L
/C
D
. In both (a) and (b), the results are shown for three dierent Reynolds numbers, Re = 100,
500 and 1000, and two Rossby numbers, Ro = 3.2 and Ro = 130. The angle of attack varies from 90
to 15
the
dierences in drag become signicant. However, considering a apping motion
(Ro = 3.2) with respect to translating (Ro = 130), an average dierence in drag
of 7.5% is obtained. While lift is enhanced signicantly, combined with a small
drag penalty, there is still a large gain in lift-to-drag, which is benecial in terms
of performance.
Another observation from gure 7.14(a) is that at large mid-stroke angles of
attack, the time-averaged lift and drag coecients show marginal variations with
Reynolds number, for a translating (Ro = 130) wing. On the other hand, the
Reynolds number has a larger eect on the lift and drag, while apping. This may
be explained by considering that the leading-edge vortex was found to be more
stable on a translating wing, compared to apping. Looking at gure 7.14(a),
the variations in lift and drag with Reynolds number are larger for lower Rossby
numbers. So, the structure of the leading-edge vortex strongly depends on the
Reynolds number in cases of large angular accelerations.
Figure 7.15 shows the iso-surface of Q = 1.0 to visualise the leading-edge vortex
on a apping wing (Ro = 3.2) during the downstroke for both Reynolds numbers
Re = 100 and Re = 1000. As was already discussed, the leading-edge vortex be-
comes slightly unstable with increasing Reynolds numbers, which is visualised by
irregularities in the iso-surfaces. In addition, the streamlines for the corresponding
156 Vortical structures in three-dimensional apping ight
(a) Re = 100
(b) Re = 1000
Figure 7.15 Vortex visualisation of the near wake for dierent Reynolds numbers, Ro =
3.2. Iso-surface of Q = 1.0 are shown for two Reynolds numbers, Re = 100 and Re = 1000, for a
apping wing Ro = 3.2. A time-frame is shown at mid-stroke, t = 0.25T. Colours indicate helicity.
comparison are shown in gure 7.16. Besides the irregularities, the leading-edge
vortex clearly detaches at a smaller distance from the wing root for Re = 1000.
However, although the leading-edge vortex may be less stable, the lift increasing
eects of the leading-edge vortex are larger for higher Reynolds numbers. Addi-
tionally, for higher Reynolds numbers, the leading-edge vortex may burst as was
discussed by Lentink & Dickinson (2009b), without a signicant loss in lift. In or-
der to illustrate the irregular motion at larger Reynolds numbers, i.e. Re = 1000,
gure 7.17 shows the vortical motion just before stroke reversal from down to
upstroke.
It was noted that the leading-edge vortex may play a more important role for
apping motion, compared to translation. Figure 7.18 shows the Q iso-surfaces
for dierent Reynolds numbers Re = 100 and Re = 1000, for a translating wing.
In case of the apping wing Ro = 3.2, gure 7.15 shows some irregularities of the
leading-edge vortex, with increasing Reynolds numbers. For a translating wing
Ro = 130, these irregularities are less pronounced. While the leading-edge vortex
detaches at a smaller distance from the wing tip, on a apping wing for increasing
Reynolds numbers, this is not the case for a translating wing.
Therefore, it seems plausible that the generation of a leading-edge vortex is
important for both apping Ro = 3.2 and translating Ro = 130 ight. A Reynolds
number increase, leads to larger lift enhancement, but also to irregularities such
that the ow at low Ro is more sensitive to changes in Reynolds number.
7.5 Flapping wings at low Reynolds numbers 157
(a) Re = 100 (b) Re = 1000
Figure 7.16 Streamline visualisation of the near wake for dierent Reynolds numbers,
Ro = 3.2. Streamlines are shown for two Reynolds numbers, Re = 100 and Re = 1000, for a apping
wing Ro = 3.2. A time-frame is shown at mid-stroke, t = 0.25T. Colours indicate helicity.
(a) Re = 100
(b) Re = 1000
Figure 7.17 Vortex visualisation of the near wake for dierent Reynolds numbers, Ro =
3.2. Iso-surface of Q = 1.0 are shown for two Reynolds numbers, Re = 100 and Re = 1000, for a
apping wing Ro = 3.2. A time-frame is shown at the end of the downstroke, t = 0.5T. Colours
indicate helicity.
158 Vortical structures in three-dimensional apping ight
(a) Re = 100 (b) Re = 1000
Figure 7.18 Vortex visualisation of the near wake for dierent Reynolds numbers, Ro =
130. Iso-surface of Q = 1.0 are shown for two Reynolds numbers, Re = 100 and Re = 1000, for a
translating wing Ro = 130. A time-frame is shown at mid-stroke, t = 0.25T. Colours indicate helicity.
7.5.4 Inuence of trapezoidal angle of attack
In order to study the inuence of the trapezoidal angle of attack and later com-
pared to an earlier two-dimensional study (Bos et al., 2008), the shape of the
angle of attack is varied. Various experimental and numerical studies have been
conducted, using a trapezoidal shaped angle of attack variation. As previously
described, the trapezoidal shape is dened by a rotation timing T
rot
, which is
varied from 0.25 to 0.1, the rst representing a harmonic variation, whereas the
latter corresponds to a strong rotation during stroke reversal. Figure 7.3 in sec-
tion 7.2.2 shows the angle of attack as a function of the rotation duration T
rot
.
While varying the rotation duration, the Reynolds number and mid-stroke angle
of attack remained xed.
Table 7.8 shows the time-average lift, drag and lift-to-drag values. The most
important observation is that with decreasing rotation duration, i.e. increasing
angular acceleration during stroke reversal, a gain in average lift is obtained of
10.8%. The average drag decreases with a similar amount, leading to a signicant
increase in lift-to-drag ratio of 21.9%. Furthermore, it can be seen that the average
drag coecient, generated in both up- and downstroke, are within 1.0%, so drag
generation is symmetric during a complete stroke.
The time variation of both lift and drag is shown in gure 7.19 for varying
rotation duration during a full apping stroke. At low rotation duration T
rot
=
0.10, the angular acceleration just after stroke reversal is large, which leads to an
increase in lift, accompanied by a decrease in drag, which may be caused by a fast
decrease in eective angle of attack. Since the wing rotates relatively quickly after
reversal, it reaches its mid-stroke angle of attack early in the stroke, compared
7.5 Flapping wings at low Reynolds numbers 159
T
rot
C
L
C
Ddownstroke
C
Dupstroke
C
L
/C
Dave
0.25 1.127 (baseline) 2.750 2.740 0.411 (baseline)
0.20 1.177 (+4.5%) 2.629 2.625 0.448 (+9.2%)
0.15 1.215 (+7.8%) 2.539 2.536 0.479 (+16.6%)
0.10 1.248 (+10.8%) 2.496 2.493 0.500 (+21.9%)
Table 7.8 Time-averaged force coecients for varying trapezoidal shape of angle of
attack. The time-averaged lift C
L
, drag C
D
and lift-to-drag ratio C
L
/C
D
are shown for a rotation
duration, varying from T
rot
= 0.25 to T
rot
= 0.10. This corresponds to respectively harmonic variation
to a trapezoidal shape. The average Reynolds number was maintained at Re = 100.
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
T
rot
= 0.10
T
rot
= 0.15
T
rot
= 0.20
T
rot
= 0.25
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
0
0.5
1
1.5
2
2.5
3
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
T
rot
= 0.10
T
rot
= 0.15
T
rot
= 0.20
T
rot
= 0.25
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-5
-4
-3
-2
-1
0
1
2
3
4
5
(b)
Figure 7.19 Force coecients for a apping wing with trapezoidal angle of attack. Lift
(b) and drag (c) coecients induced by a apping wing, such that Ro = 3.2 and Re = 100. The
rotation duration is varied from T
rot
= 0.25 to T
rot
= 0.10, using an angle of attack amplitude of
A
= 60
.
with the harmonic case T
rot
= 0.25. This causes a long period of lift enhancement,
which leads to the integrated gain in lift of 10.8%. A similar, but opposite, eect
applies to the drag coecient.
At every time instance, the lift increase and drag decrease are only marginal,
but overall the results are signicant, 21.9% lift-to-drag enhancement. Therefore,
it can be stated that a trapezoidal shape in angle of attack considerably increases
performance in three-dimensional hovering apping ight. In contrast to this, two-
dimensional studies (Bos et al., 2008) showed opposite eects, due to a premature
vortex shedding of the leading-edge vortex during the long period of high angle
of attack. The explanation for this discrepancy is that in the three-dimensional
simulations the leading-edge vortex was found to remain more stable than in two-
dimensional investigations. Therefore, it can be concluded that the leading-edge
vortex stability should be studied with a three-dimensional approach.
160 Vortical structures in three-dimensional apping ight
7.5.5 Inuence of deviation
The deviation amplitude A
is varied from 0
to 20
. In addition,
three dierent stroke patterns are studied, by varying the deviation frequency,
resulting in patterns that can be characterised as gure-of-O, gure-of-U and
gure-of-eight, which are shown in gure 7.2. Considering the gure-of-eight
patterns, two dierent deviation directions are studied, corresponding to a vari-
ation of A
from 0
to 20
and from 0
to 20
= 0
to A
= 20
, according
to the gure-of-O wing tip pattern. Following gure 7.2, the wing moves consec-
utive down and up during the downstroke and up and down during the upstroke.
Since a downward motion increases the eective angle of attack, which is therefore
subjected to an increase, decrease, decrease and another increase during the four
consecutive half-strokes. Table 7.9 shows the time-averaged force coecients for
a wing following this gure-of-O motion, while the mid-stroke angle of attack is
xed to = 45
C
L
C
Ddownstroke
C
Dupstroke
C
L
/C
Dave
0 1.224 (baseline) 2.030 2.030 0.603 (baseline)
5 1.204 (-1.6%) 2.067 1.934 0.602 (-0.15%)
10 1.180 (-3.6%) 2.072 1.850 0.601 (-0.23%)
15 1.151 (-5.9%) 2.056 1.767 0.602 (-0.12%)
20 1.104 (-9.8%) 1.996 1.676 0.601 (-0.28%)
Table 7.9 Time-averaged force coecients for varying deviation using a gure-of-O
pattern. The time-averaged lift C
L
, drag C
D
and lift-to-drag ratio C
L
/C
D
are shown for a deviation
amplitude, varying from A
= 0
to 20
c
i
e
n
t
[
-
]
20
A
= 0
A
= 5
A
= 10
A
= 15
A
= 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
0
0.5
1
1.5
2
2.5
3
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
A
= 0
A
= 5
A
= 10
A
= 15
A
= 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-3
-2
-1
0
1
2
3
4
(b)
Figure 7.20 Force coecients for a apping wing with varying deviation using gure-of-
O. Lift (a) and drag (b) coecients induced by a apping wing, such that Ro = 3.2 and Re = 100.
The deviation amplitude is varied from A
= 0
to A
= 30
= 45
.
the gure-of-eight pattern may cause signicant changes in forces. Two types of
gure-of-eight patterns are used, from A
= 0
to 20
= 0
to 20
= 20
with
the baseline case A
= 0
= 20
the lift
even shows a clear minimum at t = 0.2T, which was also present in cases without
deviation but at small angles of attack, e.g. = 15
C
L
C
Ddownstroke
C
Dupstroke
C
L
/C
Dave
0 1.224 (baseline) 2.030 2.030 0.603 (baseline)
5 1.187 (-3.0%) 2.003 1.994 0.593 (-1.7%)
10 1.167 (-4.6%) 2.012 2.001 0.580 (-3.8%)
15 1.137 (-7.1%) 2.014 2.002 0.565 (-6.3%)
20 1.108 (-9.4%) 2.021 2.007 0.548 (-9.0%)
Table 7.10 Time-averaged force coecients for varying deviation using a gure-of-U
pattern. The time-averaged lift C
L
, drag C
D
and lift-to-drag ratio C
L
/C
D
are shown for a deviation
amplitude, varying from A
= 0
to 20
C
L
C
Ddownstroke
C
Dupstroke
C
L
/C
Dave
0 1.224 (baseline) 2.030 2.030 0.603 (baseline)
5 1.107 (-9.5%) 1.859 1.852 0.596 (-1.2%)
10 0.934 (-23.7%) 1.621 1.616 0.576 (-4.5%)
15 0.747 (-39.0%) 1.370 1.365 0.545 (-9.5%)
20 0.575 (-53.0%) 1.129 1.125 0.510 (-15.5%)
-5 1.220 (-0.3%) 2.067 2.064 0.600 (-0.5%)
-10 1.205 (-1.5%) 2.057 2.047 0.586 (-2.8%)
-15 1.151 (-6.0%) 2.056 2.040 0.560 (-7.1%)
-20 1.070 (-12.6%) 2.017 2.003 0.530 (-12.0%)
Table 7.11 Time-averaged force coecients for varying deviation using a gure-of-eight
pattern. The time-averaged lift C
L
, drag C
D
and lift-to-drag ratio C
L
/C
D
are shown for a deviation
amplitude, varying from A
= 0
to 20
and A
= 0
to 20
to 45
,
while maintaining a Reynolds number of Re = 150. The three-dimensional wing
motion is harmonic and scaled such that the dimensionless amplitude A
R
g
, based
on the radius of gyration R
g
is comparable between dierent cases. Additionally,
the average Reynolds number Re
R
g
is matched.
In table 7.12, the time-averaged force coecients are shown for three dier-
ent apping situations: a two-dimensional plunging airfoil, a three-dimensional
translating wing (Ro = 130) and a three-dimensional apping wing (revolving,
Ro = 3.2). For both apping and translating wings, the force coecients are com-
164 Vortical structures in three-dimensional apping ight
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
A
= 0
A
= 5
A
= 10
A
= 15
A
= 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
0
0.5
1
1.5
2
2.5
3
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
A
= 0
A
= 5
A
= 10
A
= 15
A
= 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-3
-2
-1
0
1
2
3
(b)
Figure 7.21 Force coecients for a apping wing with varying deviation using type 1 of
gure-of-eight. Lift (a) and drag (b) coecients induced by a apping wing, such that Ro = 3.2
and Re = 100. The deviation amplitude is varied from A
= 0
to A
= 20
= 45
.
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
A
= 0
A
= 5
A
= 10
A
= 15
A
= 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
A
= 0
A
= 5
A
= 10
A
= 15
A
= 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-4
-3
-2
-1
0
1
2
3
4
(b)
Figure 7.22 Force coecients for a apping wing with varying deviation using type 2 of
gure-of-eight. Lift (a) and drag (b) coecients induced by a apping wing, such that Ro = 3.2
and Re = 100. The deviation amplitude is varied from A
= 0
to A
= 20
, using an angle of
attack amplitude of A
= 45
.
7.6 Flapping wings in forward ight 165
A
C
Lave
C
D
C
L
/C
Dave
2
D
0 2.376 (baseline) 0.0676 35.148
15 1.969 (baseline) -0.401 -4.920
30 1.403 (baseline) -0.580 -2.419
45 0.834 (baseline) -0.429 -1.947
R
o
=
1
3
0
0 1.710 (-28.0%) 0.028 61.571
15 1.414 (-28.2%) -0.231 -6.119
30 1.016 (-27.6%) -0.316 -3.211
45 0.644 (-22.8%) -0.175 -3.673
R
o
=
3
.
2 0 1.904 (-20.9%) 0.068 27.964
15 1.544 (-21.6%) -0.270 -5.719
30 1.061 (-24.4%) -0.366 -2.897
45 0.604 (-27.6%) -0.214 -2.824
Table 7.12 Time-averaged force coecients in forward ight. Two- and three-dimensional
time-averaged force coecients for a apping wing in forward ight at Re = 150,
= 6.3. The
mid-stroke angle of attack is varied from 0
to 45
= 30
c
i
e
n
t
[
-
]
= 0
= 15
= 30
= 45
c
i
e
n
t
[
-
]
= 0
= 15
= 30
= 45
= 6.3 and
mid-stroke angle of attack of A
= 45
.
two counter-rotating tip vortices. This vortex ring is shed and convected into the
wake. The apping wing (gure 7.24(b)), however, generates a more complex ow
eld, induced by a spiralling leading-edge and tip vortex, even though the angle
of attack is zero. When the angle of attack is increased, gure 7.25 shows for
A
= 30
= 0
= 30
. However,
the performance, in terms of maximal lift-to-drag ratio was found to be maximal
at = 30
. At = 15
the
leading-edge vortex is not strong enough, causing a sudden decrease in both lift
and drag during each half-stroke.
Secondly, the eects are studied of a varying stroke curvature, reected by the
Rossby number. Both time-averaged lift and drag decrease signicantly with in-
creasing Rossby number. At Ro = 130, a nearly translating wing, the lift decreases
with 24.7%, compared to the apping wing with Ro = 3.2. The major decrease
in lift and drag occurs during mid-stroke, between t = 0.25T and t = 0.5T. Flow
visualisations showed that the leading-edge vortex is signicantly reduced in size
and strength for the translating with at Ro = 130, compared to the apping wing.
In addition, the leading-edge vortex rolls-up to form two tip vortices instead of
one in case of the apping wings. This causes both lift and drag to be signicantly
lower at Ro = 130.
To study the eect of three dierent Reynolds numbers (Re = 100, Re = 500
and Re = 1000), the force polars are constructed which also shows the relation with
the angle of attack and Rossby numbers. It was seen that with increasing Reynolds
number, both time-averaged lift and drag increases, but the dierences become
smaller at Re = 1000. The eect of a changing Reynolds number is negligible
for both lift and drag at high mid-stroke angles of attack translation. This is
caused by the larger importance of the leading-edge vortex for a apping wing at
Ro = 3.2. For both apping, Ro = 3.2, and translation Ro = 130, irregularities in
the leading-edge vortex occur, when the Reynolds number increases. For higher
Reynolds numbers the vortex separates earlier, but lift and drag still increases.
When the Reynolds number increases even further, it probably reaches a limit,
beyond which the leading-edge vortex bursts, causing a loss in performance.
Besides the inuence of the angle of attack, Rossby and Reynolds numbers, the
7.7 Conclusions 169
eects of the wing kinematics has been investigated. First, the shape of the angle
of attack variation is varied along various trapezoidal shapes. The trapezoidal
shape causes a fast pitch-up motion at the beginning of each up- and downstroke.
Therefore, the mid-stroke angle of attack is reached earlier and maintained for a
longer period, compared to cases with a simple harmonic angle of attack variation.
This leads to a signicant increase in time-averaged lift of 10.8%. The lift increase
is accompanied by a decrease in drag of a similar amount, such that the lift-to-
drag ratio shows a signicant gain of 21.9%. Two-dimensional studies showed
an opposite result which was caused by the fact that the leading-edge vortex
separated, while translating at a constant angle of attack. Three-dimensional
eects, however, lead to rm and stably attacked leading-edge vortex.
The second eect of kinematic modelling is reected by the presence of de-
viation. Three dierent wing tip patterns may be caused by deviation, such as
the gure-of-O, gure-of-U and the gure-of-eight. The gure-of-O wing tip
pattern leads to an asymmetric variation in eective angle of attack, leading to
dierences in drag for the up- and downstroke. Nevertheless, the eect of this
pattern on the time-averaged forces coecients is negligible. On the other hand,
the gure-of-U pattern does inuence the forces. The eective angle of attack is
signicantly decreased during mid-stroke. The time-averaged lift coecient is de-
creased by 9.4%. While drag is maintaining nearly constant, the lift-to-drag ratio
is decreased by about 9%. The gure-of-eight aects the forces most, however,
it depends on the starting direction. Two types are considered, the rst starting
the downstroke with an downward deviation motion, whereas the type 2 starts
upwards. The eective angle of attack is being inuenced in such a way that the
type 1 gure-of-eight decreases both lift and drag considerably. Lift decreases up
to 53%. Since drag decreases fast as well, the loss in lift-to-drag ratio is limited to
15.5%. The type 2 gure-of-eight motion, which starts with an upward motion
at the beginning of the stroke, the drag is maintained constant. However, lift
decreases 12.6% such that the lift-to-drag ratio decreases with about 12.0%.
In addition to the eects of Reynolds number, Rossby number and wing kine-
matics in hovering ight, a preliminary study is performed to compare two- and
three-dimensional forward apping ight. Both two- and three-dimensional simu-
lations are dynamically scaled using the radius of gyration to justify comparison.
For a dimensionless wavelength of k = 6.3 the lift and drag forces are compared
for dierent rotation amplitudes at a Reynolds number of Re = 150. For the
translating wing (Ro = 130), the three-dimensional simulations result in 28% less
lift compared to the two-dimensional case. This dierence is mainly caused by
the loss in lift due to the generation of a tip vortex which is only present in the
three-dimensional simulations. However, an increase in Rossby number resulted
in a signicant gain in lift. In combination with a higher thrust this observation
leads to the conclusion that a stable leading-edge vortex (induced by the revolving
motion) plays an important role in three-dimensional apping aerodynamics.
CHAPTER 8
Inuence of wing deformation by
exing
A preliminary study is performed to investigate the eects of wing exing in ap-
ping wing aerodynamics. The eects of a cosine-shaped wing exing is analysed
in two-dimensional forward apping ight. For three-dimensional hovering ight,
the eects of wing exing in spanwise and chordwise directions are discussed. In
two-dimensional forward apping ight, the exing of the airfoil shows similar
eects as if wing rotation was applied, increasing its eciency. Furthermore, in
three-dimensional hovering, the exing reduces the strength of the leading-edge
vortex, compared to a rigid wing. This leads to an overall decrease in lift and
drag, this inuence is larger for chordwise compared to spanwise exing.
Section 8.1 deals with the inuence of exing for the two-dimensional plunging
airfoil in forward ight. Its inuence on the ow induced by a three-dimensional
wing in hovering ight is subject of study in section 8.2. The conclusions are
drawn in section 8.3.
8.1 Airfoil exing in two-dimensional forward
apping ight
In chapter 6, the inuence on the forces and vortex patterns of a variation in
dimensionless wavelength, apping amplitude, angle of attack and stroke plane
angle was investigated in two-dimensional forward apping ight. These results
are extended by using preliminary simulations of deformation airfoils, subjected to
a pre-dened exing. The boundary displacements due to the exing are dened
172 Inuence of wing deformation by exing
with respect to the initial boundary shape and varying in time such that the
maximal exing occurs at mid-stroke. The resulting airfoil shape is either a quarter
or half-cosine shape, corresponding to
f
= 0.25 or
f
= 0.5, respectively, as is
shown in gure 8.1. Besides the two dierent deformation mode shapes, the exing
amplitude is varied from A
f
= 0.1 to A
f
= 0.4, which is similar to maximal 40%
of the chord length. This exing was imposed on a baseline apping motion
(see chapter 6) with a dimensionless wavelength of
= 0
f
= 0.0
f
= 0.25
f
= 0.5
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
Figure 8.1 Flexing displacements to represent wing deformation. The exing displacements
of a plunging airfoil are shown for dierent shape ratios. A shape ratio
f
= 0.25 corresponds to a
quarter-cosine and
f
= 0.5 to a half-cosine shape.
8.1 Airfoil exing in two-dimensional forward apping ight 173
A
f
C
Lave
C
Ldownstroke
C
Lupstroke
C
D
0 1.956 (baseline) 1.758 -2.154 0.058
f
=
0
.
2
5
0.1 1.683 (-14.0%) 1.880 -1.485 -0.044
0.2 1.510 (-22.8%) 1.393 -1.628 -0.109
0.4 1.492 (-23.7%) 1.492 -1.492 -0.263
0.8 0.988 (-49.5%) 0.988 -0.987 -0.229
f
=
0
.
5
0.1 1.867 (-4.6%) 1.865 -1.868 -0.088
0.2 1.810 (-7.5%) 1.810 -1.810 -0.215
0.4 1.732 (-11.5%) 1.632 -1.628 -0.465
0.8 1.021 (-47.8%) 1.021 -1.021 -0.332
Table 8.1 Time-averaged force coecients for a deforming plunging airfoil. The time-
averaged lift and drag are shown for a deforming two-dimensional airfoil, for two exing shapes,
f
=
0.25 and
f
= 0.5. The Reynolds number was xed to Re = 150, the dimensionless wavelength was
k = 6.8 and the rotation amplitude remained at A
= 0.
the force asymmetry is strongly related to the wake dynamics. Therefore, airfoil
exing using a quarter-cosine shape ( = 0.25) may induce a similar eect on the
wake dynamics as airfoil rotation. However, when applying the half-cosine airfoil
exing, the lift coecient is symmetric over a complete stroke. While the quarter-
cosine increases the airfoil camber, the half-cosine shape redirects the trailing-edge
towards the uniform free-stream. This redirection of the trailing-edge compensates
for the destabilising eect of the airfoil rotation, such that a symmetric force is
the result of a symmetric wake pattern, which is benecial in terms of eciency.
Figure 8.2 shows the drag coecients for both exing shapes. It is clear that the
thrust (negative drag) is more prominent for the half-cosine ( = 0.5) exing mode.
The main regions of thrust enhancement are during mid-stroke, from t = 0.1T to
t = 0.4T and t = 0.6T to t = 0.9T. In order to obtain a better understanding of
the ow physics, gure 8.3 shows vorticity contours for both exing shapes with
an amplitude of A
f
= 0.2 at t = 0.25T, which is midway during the downstroke.
Comparing gure 8.3(a) with 8.3(b), it is apparent that the leading-edge vortex is
of similar size and strength for both exing shapes. However, the trailing edge of
the half-cosine shaped airfoil appears to enhance the speed of the vortex shedding.
Therefore, at similar time instance, the shed vortices are farther away downstream
for the half-cosine ( = 0.5) exing mode, increasing its thrust.
Summarising, this preliminary two-dimensional investigation, to understand
the eect of wing exing, has led to some interesting results. The applied wing
exing inuences the ow in a similar way as was obtained by applying airfoil
rotation, indicating a similar mechanism in terms of the eects of angle of attack.
With the introduction of exing, the drag became negative to generate thrust in
forward ight, and the lift decreased signicantly. The lift and drag development
strongly depends on the shape of the wing exing. The half-cosine exing shape
174 Inuence of wing deformation by exing
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
A
flex
= 0.0
A
flex
= 0.1
A
flex
= 0.2
A
flex
= 0.4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-2
-1.5
-1
-0.5
0
0.5
(a) = 0.25
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
A
flex
= 0.0
A
flex
= 0.1
A
flex
= 0.2
A
flex
= 0.4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-2
-1.5
-1
-0.5
0
0.5
(b) = 0.5
Figure 8.2 Comparison of drag coecients for a exing airfoil. The drag coecients are shown
for the quarter-cosine (a) and half-cosine (b) shaped airfoil exing modes. The exing amplitude A
flex
is varied from 0.0 to 0.4, the latter corresponds to 40% of the chord length.
(a) = 0.25 (b) = 0.5
Figure 8.3 Vorticity contours. Vorticity contours around a plunging exing airfoil without ro-
tation at Re = 150 are shown during mid-stroke at t = 0.25T. The apping amplitude was xed to
A
= 0
.
results in less lift than obtained using the quarter-cosine shape. Additionally, the
half-cosine exing mode also generates the largest thrust. This behaviour was
related to the fact that the half-cosine shape caused the generated vortices to
convect faster, since the trailing-edge became aligned with the ow.
8.2 Wing exing in three-dimensional hovering ight 175
8.2 Wing exing in three-dimensional hovering
ight
In addition to the investigation of exing eects on an airfoil in forward apping
ight, it was chosen to investigate a three-dimensional wing as well. Hovering
conditions are assumed, since it was extensively studied in the previous chapter.
Besides the inuence of the Rossby number, Reynolds number and the kinematic
modelling, applied under rigid wing conditions, it may be interesting to study the
additional eects of wing deformation. The deformation of the wing is dened by a
user-dened exing function, leading to a realistic bending of the three-dimensional
wing (Combes & Daniel, 2003a,b, Shyy et al., 2008a).
In order to investigate the eects of three-dimensional wing exing on the
forces and performance, the exing is dened according to the model described
in section 4.3.3. The deformation model is hence similar to the one used for
the two-dimensional airfoil in the previous section. Additionally, the exing was
independently applied to the spanwise and the chordwise directions. The three-
dimensional wing is deformed such that its shape prescribes a quarter-cosine func-
tion, which is maximal at the wing tip or the trailing edge, respectively. This
cosine shaped deformation varies harmonically in time during the stroke such that
largest change in deformation occurs during stroke reversal, which is considered
to be realistic. The spanwise and chordwise directions are investigated separately.
The exing amplitude A
f
is varied from A
f
= 0.0 to A
f
= 0.4 in spanwise di-
rection, while in chordwise direction the amplitude was varied from A
f
= 0.0 to
A
f
= 0.2, due to limitations of the mesh motion solver. An amplitude of A
f
= 0.2
corresponds to a maximal deection at mid-stroke of 20% of the representative
exing length (wing span or chord, respectively). Figure 8.1 shows an illustration
of the exing model.
A
f
C
L
C
Ddownstroke
C
Dupstroke
C
L
/C
Dave
S
p
a
n
0 1.234 (baseline) 2.034 2.028 0.608 (baseline)
0.1 1.212 (-1.74%) 2.017 2.011 0.602 (-0.9%)
0.2 1.213 (-1.73%) 2.029 2.023 0.599 (-1.5%)
0.3 1.206 (-2.3%) 2.034 2.028 0.594 (-2.3%)
0.4 1.190 (-3.6%) 2.031 1.992 0.592 (-2.6%)
C
h
o
r
d
0.1 1.137 (-7.9%) 2.129 2.126 0.534 (-12.1%)
0.2 1.036 (-16.0%) 2.267 2.266 0.457 (-24.8%)
Table 8.2 Time-averaged force coecients for a deforming three-dimensional wing. Time-
averaged force coecients of a three-dimensional exing wing. The spanwise exing amplitude is varied
from A
f
= 0.0 to A
f
= 0.4, while the chordwise exing amplitude varies from A
f
= 0.0 to A
f
= 0.2.
.
176 Inuence of wing deformation by exing
t/T [-]
L
i
f
t
c
o
e
c
i
e
n
t
[
-
]
A
f
= 0.0
A
f
= 0.2
A
f
= 0.4
A
f
= 0.1
A
f
= 0.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
2.5
(a)
t/T [-]
D
r
a
g
c
o
e
c
i
e
n
t
[
-
]
A
f
= 0.0
A
f
= 0.2
A
f
= 0.4
A
f
= 0.1
A
f
= 0.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-3
-2
-1
0
1
2
3
(b)
Figure 8.4 Force coecients for a spanwise and chordwise deforming apping wing. Lift
(a) and drag (b) coecients induced by a apping wing, which deforms in spanwise (A
f
= 0.0 0.4)
and chordwise (A
f
= 0.0 0.2) direction. The latter are shown by and . The angle of attack at
mid-stroke remained at = 45
.
Table 8.2 shows the time-averaged force coecients for dierent exing amplitudes,
using spanwise or chordwise deformation. The most important result is that both
time-averaged lift and drag are only marginally inuenced by the spanwise exing,
so where the wing tip bends. The maximal dierence with the non-deforming case
(-3.6%) occurs for 40% wing tip deection, which is a signicant deformation.
The time-averaged drag remains nearly unaected as well, such that the maximal
dierence in lift-to-drag ratio is only -2.6%.
The chordwise deformation, on the other hand, inuences the time-averaged
forces to a large extent. As is easily seen, the time-averaged lift decrease with
7.9% for only a small (A
f
= 0.1) chord deformation and 16% for A
f
= 0.2. The
drag increases for the chordwise deforming wing, with respect to the case without
exing, such that the maximal dierence in lift-to-drag ratio is -24.8%.
The force variation, during a full apping stroke, is shown in gure 8.4. It
is clear that the largest dierence in force coecients occurs during mid-stroke,
lowering the maximal lift coecient. It seems that the leading-edge vortex is ex-
pelled by the wing deformation, with a maximal eect at A
f
= 0.2 of chordwise
exing. When the lift decreases, the drag increases during mid-stroke. Addition-
ally, gure 8.5 shows the streamlines to visualise the leading-edge vortex during
mid-stroke (t = 0.25T) for both spanwise and chordwise exing. For comparison,
also the case using no exing is shown. Compared to the case without exing the
leading-edge vortex detaches earlier (less close to the wing tip) for both exing
cases. The spanwise exing has less inuence than the chordwise exing mode
shape. The inuence of spanwise exing is minor, which was also reected in the
8.3 Conclusions 177
time-averaged force coecients, which were only marginally lower. However, the
chordwise exing induced the leading-edge vortex to burst almost at the middle
of the wing, leading to signicantly less lift, see table 8.2.
8.3 Conclusions
In this chapter, the result have been described of a preliminary investigation to
understand the eects of wing exing. Therefore, a pre-dened exing deformation
has been applied to a plunging airfoil in two-dimensional forward ight and to a
three-dimensional apping wing in hovering ight. First of all, the two-dimensional
exing results in a similar eect on the ow and resulting forces as rotation of the
airfoil. With increasing exing amplitude, the larger eective angle of attack leads
to the generation of negative drag or thrust. In addition, the inuence of the exing
shape (quarter-cosine or half-cosine) of the airfoil was studied. A quarter-cosine
shaped exing results in signicantly higher thrust at less lift.
Secondly, a exing deformation has been applied to a three-dimensional ap-
ping wing in hovering ight. Two exing directions have been considered, span-
wise and chordwise exing. While keeping the exing amplitude constant, the
chordwise exing aects the ow the most. For a chordwise exing amplitude of
A
f
= 0.2 the time-averaged lift and drag coecients are signicantly decreased
and increased, respectively, such that the lift-to-drag decreases with 24.8%. The
spanwise exing does not show a comparable impact, only -2.6%.
178 Inuence of wing deformation by exing
(a) No exing
(b) Spanwise exing, A
f
= 0.2 (c) Chordwise exing, A
f
= 0.2
Figure 8.5 Streamlines on a exing wing in hovering ight. Iso-surface of Q = 1.0 are shown
for a three-dimensional apping wing in hovering ight at Re = 100 and mid-stroke angle of attack of
A
= 45
at t = 0.25T. (a) shows the wing without exing, while (b) shows the wing using spanwise
exing and (c) using chordwise exing, both with A
f
= 0.2. Colours indicate helicity.
CHAPTER 9
Conclusions and recommendations
In the present study we investigate mesh motion techniques to be able to per-
form parametric variation of the ow around apping foils and wings. Dierent
existing mesh motion methods have been compared using cell quality measures.
To improve the mesh quality a mesh motion technique has been implemented,
based on the interpolation of radial basis functions on the mesh interior. Using
that technique, it has become possible to eciently solve the ow around apping
foils and wings at low Reynolds numbers, for ow conditions corresponding to the
scale of ying insects. The improved and implemented mesh motion technique,
is used to solve the ow around apping wings in hovering and forward apping
ight. Additionally, a preliminary study of the inuence of wing exing has been
conducted as well.
The overall conclusions of this research are given in section 9.1. Secondly, the
conclusions on the assessment of dierent mesh motion techniques are drawn in
section 9.2. Section 9.3 presents the conclusions of a apping wing under hovering
conditions, for two-dimensional airfoil as well as for the three-dimensional wing.
Section 9.4 shows the results for forward apping ight, while section 9.5 describes
the conclusions of the preliminary study to understand the eects of wing exing.
Additional recommendation are made in Section 9.6.
9.1 Overall conclusions
The following overall conclusions are drawn:
1. The ow around a apping wing, at the scale relevant to insect ight, can
be solved accurately using advanced mesh motion techniques in existing ow
180 Conclusions and recommendations
solvers.
2. Radial basis function mesh motion leads to improved mesh quality, compared
to methods based on solving the Laplace or a modied stress equation. How-
ever, mesh motion based on radial basis function interpolation is much more
computationally demanding. Dierent ways to improve its eciency are
discussed.
3. The wing kinematic pattern has a large inuence on the forces in two-
dimensional hovering and forward ight.
4. In forward ight, thrust generation is the result of the forward tilting of the
wing.
5. In two- and three-dimensional apping ight, the leading-edge vortex is a
very important lift enhancement mechanism.
6. The Rossby number, which describes the wing stroke curvature, has a large
inuence on the time-averaged lift force. A low Rossby number signicantly
increases lift as the result of an induced spiralling leading-edge vortex.
7. Wing exing may play an important role in insect ight to modulate the
forces and to generate thrust.
9.2 Conclusions on mesh motion techniques
Two existing mesh motion techniques have been described which are commonly
used in existing CFD codes. The rst method solves the Laplace equation with
a variable diusion coecient, which is used to control the nal mesh quality.
Secondly, a modied stress equation was used as the basis for mesh motion. Ad-
ditionally, a mesh motion solver is implemented, which uses the interpolation of
radial basis functions (RBF). The following conclusions are drawn:
1. For both existing mesh motion solvers, based on the Laplace and a modied
stress equation, the mesh quality is not sucient for apping wing cases,
where the rotation angles are large. However, these methods are very ecient
since existing iterative solvers can be used to solve a sparse system.
2. The mesh motion solver based on the radial basis function (RBF) interpola-
tion does not need any information about the mesh connectivity and can be
applied to arbitrary unstructured meshes containing polyhedral cells. Addi-
tionally, dierent radial basis functions, concerning the RBF mesh motion,
are compared. The RBF mesh quality provides superior mesh quality over
the Laplace and modied stress equation mesh motion. Especially, when
using the thin plate spline (TPS) or the continuous polynomial C
2
as radial
9.3 Conclusions on hovering apping ight 181
basis functions, the mesh quality is high in terms of low skewness and non-
orthogonality. The TPS has global support, whereas the C
2
basis function
has compact support.
3. Since the RBF mesh motion technique encounters a dense system of equa-
tions, dierent methods are implemented to increase its eciency. First of
all, a subset of the moving boundary points was selected, because not all
points are necessary if the body performs a rigid body motion. Therefore, a
coarsening algorithm selects the desired control points. Secondly, a smooth-
ing function is used to decrease the RBF contribution to zero at the outer
domain boundaries.
9.3 Conclusions on hovering apping ight
9.3.1 Two-dimensional hovering
The eects of wing motion kinematics on the aerodynamic characteristics of hov-
ering insect ight have been investigated by means of two-dimensional numerical
ow simulations. The results of the present two-dimensional study has provided
useful insights, which may be relevant also for the understanding of realistic three-
dimensional insect ight.
Four dierent kinematic models, with dierent complexity, have been analysed.
Two of these models, pure harmonic motion and the Roboy experimental kine-
matics have been used extensively in literature. The most prominent aspects of
the Roboy kinematic model are the sawtooth displacement and the trapezoidal
angle of attack. The third model represents the actual fruit y kinematics as ob-
served in experiments and the last one was a modication of the latter, chosen
to investigate the eect of symmetry. The fruit y models are characterised by a
bump in angle of attack and the presence of deviation. To facilitate the compar-
ison these models were dynamically scaled at Re = 110 and constructed such that
their mean quasi-steady lift coecient was matched. The following conclusions
are drawn about the two-dimensional hovering simulations:
1. It was found that the realistic fruit y wing kinematics result in signicantly
lower drag at similar lift compared with the simplied wing kinematic models
used in literature. The trend that the fruit y kinematics increases aerody-
namic performance agrees with the predictions of the quasi-steady theory,
but the numerical ow simulations provide a more complete quantitative
analysis of the ow behaviour.
2. It was shown that the dierence in performance in terms of mean lift-to-drag
ratio between the dierent kinematic models was signicant. The mean
aerodynamic drag at equal lift of the fruit y models is about 49% lower
compared to the Roboy model and about 29% lower with respect to the
harmonic model.
182 Conclusions and recommendations
3. The sawtooth amplitude used in the Roboy model has a small eect on
the mean lift but the mean drag is aected signicantly. Due to the high
acceleration during stroke reversal of the sawtooth shaped amplitude, the
mean drag at comparable lift is increased by 24%.
4. The second model simplication used by the Roboy, the trapezoidal angle
of attack, caused the LEV to separate during the translational phase. This
led to an increase in mean drag during each half-stroke. Also in this case
large accelerations at stroke reversal lead to a decrease in lift-to-drag ratio
of 33%.
5. The extra bump in angle of attack as used by the fruit y model is not
aecting the mean lift to a large extent. During the beginning of the up
and downstroke the bump decreases the angle of attack such that the wing
orientation is almost horizontal. This leads to a signicant decrease in drag
which improves aerodynamic performance in the sense of lift-to-drag ratio
by 15.6%.
6. The other realistic kinematic feature is the deviation, which is found to have
only a marginal eect on the mean lift and mean drag. However, the eective
angle of attack is altered such that the deviation leads to levelling of the force
distribution over the apping cycle.
9.3.2 Three-dimensional hovering
A numerical model has been developed for solving the ow around a three-dimen-
sional wing with complex wing kinematics. Dierent aspects, relevant to three-
dimensional apping wing aerodynamics, have been studied. First, the ow around
a dynamically scaled model wing is solved for dierent angles of attack in order to
study the force development and vortex dynamics at small and large mid-stroke
angle of attack. Secondly, the Rossby number is varied at dierent Reynolds num-
bers. A varying Rossby number represents a variation in stroke path curvature
and thus angular acceleration. Thirdly, the three-dimensional wing kinematics is
varied by changing the shape in angle of attack and by applying a deviation, which
may cause a gure-of-eight pattern. Finally, the three-dimensional ow is com-
pared with the two-dimensional studies performed on apping forward ight. The
following conclusions are drawn with respect to the three-dimensional simulations,
concerning hovering ight:
1. The eect of a variation in angle of attack is such that the maximal lift
occurs at a mid-stroke angle of = 45
.
The apping motion induces a leading-edge vortex which caused a peak in
lift halfway during each up- and downstroke. This leading-edge vortex is
9.3 Conclusions on hovering apping ight 183
strong at suciently high angles of attack > 30
. At = 15
the leading-
edge vortex is less strong, causing a sudden decrease in both lift and drag
during each half-stroke.
2. Both time-averaged lift and drag decrease signicantly with increasing Ross-
by number. At Ro = 130, a nearly translating wing, the lift decreases
with 24.7%, compared to the apping wing with Ro = 3.2. The major
decrease in lift and drag occurs during mid-stroke, between t = 0.25T and t =
0.5T. Flow visualisations showed that the leading-edge vortex is signicantly
reduced in size and strength for the translating wing at Ro = 130, compared
to the apping wing. In addition, the leading-edge vortex rolls-up to form
two tip vortices instead of one in case of the apping wings. This causes
both lift and drag to be signicantly lower at Ro = 130.
3. It was seen that for increasing Reynolds number, both time-averaged lift
and drag increases, but the dierences become smaller. The eect of a
changing Reynolds number is negligible for both lift and drag at high mid-
stroke angles of attack, when the wing is nearly translation at Re = 130. This
is caused by the larger importance of the leading-edge vortex for a apping
wing at Ro = 3.2. For both apping (Ro = 3.2) and translation (Ro = 130),
irregularities in the leading-edge vortex occur, when the Reynolds number
increases. For higher Reynolds numbers the vortex separates earlier, but lift
and drag still increases. When the Reynolds number increases even further,
a limit is reached, beyond which the leading-edge vortex bursts, causing a
loss in performance.
4. The trapezoidal shape in angle of attack causes a fast pitch-up motion at
the beginning of each up- and downstroke. Therefore, the mid-stroke angle
of attack is reached earlier and maintained for a longer period, compared
to cases with a simple harmonic angle of attack variation. This leads to
a signicant increase in time-averaged lift of 10.8%. The lift increase is
accompanied by a decrease in drag of a similar amount, such that the lift-to-
drag ratio shows a signicant gain of 21.9%. Two-dimensional studies showed
an opposite result which was caused by the fact that the leading-edge vortex
separated, while translating at a constant angle of attack. Three-dimensional
eects, however, lead to a rm and stably attached leading-edge vortex.
5. Three dierent wing tip patterns, caused by deviation, have been studied,
such as the gure-of-O, gure-of-U and the gure-of-eight. The gure-
of-O wing tip pattern leads to an asymmetric variation in eective angle of
attack, leading to dierences in drag for the up- and downstroke. Nonethe-
less, the eect of this pattern on the time-averaged forces coecients is
negligible. On the other hand, the gure-of-U pattern does inuence the
forces in a non-benecial way. The eective angle of attack is signicantly
decreased during mid-stroke. The time-averaged lift coecient is decreased
184 Conclusions and recommendations
by 9.4%. While drag is maintaining nearly constant, the lift-to-drag ratio is
decreased by about 9%. The gure-of-eight aects the forces most, how-
ever, it depends on the starting direction. Two types are considered, the
rst starting the downstroke with an downward deviation motion, whereas
the type 2 starts upwards. The eective angle of attack is being inuenced
in such a way that the type 1 gure-of-eight decreases both lift and drag
considerably. Lift decreases up to 53%. Since drag decreases fast as well,
the loss in lift-to-drag ratio is limited to 15.5%. The type 2 gure-of-eight
motion, which starts with an upward motion at the beginning of the stroke,
the drag is maintained constant. However, lift decreases 12.6% such that the
lift-to-drag ratio decreases with about 12.0%.
9.4 Conclusions on forward apping ight
9.4.1 Two-dimensional forward apping
A numerical model for two-dimensional ow has been used to investigate the ef-
fect of motion kinematics on the vortex dynamics around an ellipsoid foil subjected
to prescribed apping motion over a range of dimensionless wavelengths, dimen-
sionless amplitudes, angle of attack amplitudes, and stroke plane angles at the
Reynolds number of 150. Both plunging and rotating motions are prescribed by
simple harmonic functions which are useful for exploring the parametric space
despite the model simplicity. Concerning the two-dimensional forward apping
simulations, the following overall conclusions are drawn:
1. The resulting wake patterns behind the foil are categorised. Although such
an attempt at classifying the observed vortex patterns can lead to a degree
of uncertainty in determining the exact wake pattern due to the shedding,
tearing, or merging of big and small vortices.
2. Optimal propulsion using apping foil exists for each variable which implies
that aerodynamics might select a range of preferable operating condition.
The conditions that give optimal propulsion lie in the synchronisation region
in which the ow is periodic.
9.4.2 Three-dimensional forward apping
In addition to the investigation of the eects of Reynolds number, Rossby number
and wing kinematics in hovering ight, a preliminary study is performed to com-
pare two- and three-dimensional forward ight. Both two- and three-dimensional
simulations are dynamically scaled using the radius of gyration to justify compar-
ison. The three-dimensional simulations, concerning forward ight are performed
for a dimensionless wavelength of k = 6.3 and two Rossby numbers, Ro = 130
and Ro = 3.2 at a Reynolds number of Re = 150. The following conclusions are
drawn:
9.5 Preliminary conclusions on wing exing 185
1. For the case without rotation the three-dimensional simulations result in
14.5% less lift compared to its two-dimensional counterpart. This dierence
is mainly caused by the loss in lift due to the generation of a tip vortex which
is only present in the three-dimensional simulations.
2. An increase in rotation angle resulted in signicant higher lift, up to 70.5%,
which is the opposite as in the two-dimensional plunging wing. Together
with the higher thrust, this leads to the conclusion that a stable leading-edge
vortex plays a more important role in three-dimensional apping compared
to two-dimensional plunging, in the situation that the rotation angle is non-
zero.
3. For a dimensionless wavelength of k = 6.3 the lift and drag forces are com-
pared for dierent rotation amplitudes at a Reynolds number of Re = 150.
For the translating wing (Ro = 130), the three-dimensional simulations re-
sult in 28% less lift compared to the two-dimensional case. This dierence is
mainly caused by the loss in lift due to the generation of a tip vortex which
is only present in the three-dimensional simulations. However, an increase in
Rossby number resulted in a signicant gain in lift. In combination with a
higher thrust this observation leads to the conclusion that a stable leading-
edge vortex (induced by the revolving motion) plays an important role in
three-dimensional apping aerodynamics.
9.5 Preliminary conclusions on wing exing
A exing deformation has been applied to a plunging airfoil in two-dimensional
forward ight and to a three-dimensional apping wing during hovering ight.
Concerning the exible airfoil in forward ight, a comparison is made with a
plunging airfoils, with additional rotation. The following conclusion are drawn:
1. The two-dimensional exing has a comparable eect on the ow and forces as
rotation of the airfoil. With increasing exing amplitude, the larger eective
angle of attack leads to the generation of negative drag or thrust. Besides,
the exing shape of the airfoil is important. A quarter-cosine shaped exing
results in signicantly higher thrust, while lift decreases.
2. The exing deformation was also applied to a three-dimensional apping
wing in hovering ight. Two exing directions were considered, spanwise and
chordwise exing. While keeping the exing amplitude constant, the chord-
wise exing aects the ow signicantly. For a chordwise exing amplitude
of A
f
= 0.2 the time-averaged lift and drag coecients are signicantly de-
creased and increased, respectively, such that the lift-to-drag decreases with
24.8%. The spanwise exing does not show a comparable inuence, only
-2.6%.
186 Conclusions and recommendations
9.6 Recommendations
The current thesis describes a comparison of dierent mesh deformation methods
and development of an improved method, based on radial basis function interpo-
lation. Additionally, these methods are used to solve the ow around two- and
three-dimensional apping airfoils and wings under hovering and forward ight
conditions. The following recommendation can be made for further research.
The eciency of the radial basis function mesh motion can be further im-
proved by methods which selects only the necessary boundary points. The
size of the matrix, which need to be solved, can be signicantly reduced.
Additionally, the implementation of parallel iterative solver techniques may
increase its eciency even further.
The current implementation of the radial basis function mesh motion in
OpenFOAM
, and is ca-
pable of generating high quality meshes around complex geometries. Nevertheless,
using Gambit for structured meshes can be dicult since dening an appropriate
grading can be cumbersome. The third grid generation tool is the commercial
GridPro
package. GridPro
and Fluent
. These three
tools are described in sections A.2, A.2 and A.2, respectively.
A.2 BlockMesh
The open-source mesh generator, supplied with OpenFOAM
, is blockMesh. It is
an easy to use and robust utility, applicable for simplied cases. The computational
domain is specied by points, lines and blocks, see gure A.1. The connections be-
tween the points (p1p4) are dened by lines, which can be curved. Additionally,
the domain consists of blocks (b1 b8), specied by the lines, accordingly. The
relation between the points, lines and blocks need to be specied in a dictionary
le, within the OpenFOAM
and OpenFOAM
using stationary and plunging cylinder ows. The next grid generator, GridPro
,
is capable to generate the multiple blocks automatically, which will improve the
mesh quality considerably, especially for three-dimensional cases.
A.4 GridPro
GridPro
b
X
Y
Figure A.4 GridPro
topology used to
generate the (block-) structured mesh.
The red blocks indicate extra topology, ac-
cording to which GridPro
. The
grid is generated by GridPro
according to
the user-dened mesh topology, such that the
mesh quality is optimised.
represent structured grid blocks. An illustration of the resulting grid is shown in
gure A.5. Besides the high quality meshes, GridPro
to generate
the three-dimensional meshes around a apping wing, as illustrated in gure A.6.
A.5 Conclusions
Three dierent mesh generators have been described, blockMesh, Gambit and
GridPro
, is easy
to use for simplied problems. In the current thesis, blockMesh is used to gen-
erate the two-dimensional meshes, which were used to validate the ow solvers.
The second grid generator, Gambit, has more capabilities and can be used to
generated meshes around complex geometries. This mesh generator is used for
two-dimensional ows around stationary and plunging cylinders. Using the third
grid generator, GridPro
is used for
the three-dimensional simulations around apping wings.
A.5 Conclusions 193
(a) 3D GridPro
mesh
(b) 3D GridPro
,
whereas section B.3 describes the settings used in OpenFOAM
.
B.2 Fluent
solver settings
Fluent
is a nite volume based CFD solver. This section deals with the dif-
ferent ow solver settings, that are necessary to reproduce the results from this
thesis. Fluent
solver settings
Since Fluent
(Open
Field Operation And Manipulation) is a C++ toolbox for the customisation and
extension of numerical solvers for continuum mechanics problems, including Com-
putational Fluid Dynamics (CFD). It is produced by OpenCFD
Ltd. and is
freely available and open source, licensed under the GNU General Public Licence.
Since the source-code is fully accessible, dierent ow solvers and utilities are
developed. For example, the unsteady Navier-Stokes equations were solved, using
mesh motion techniques, with icoDyMFoam or icoDyMFoamRBF, the latter using ra-
dial basis function interpolation for mesh deformation. Utilities like setTaylorVortices
or meshQuality write respectively the velocities for a Taylor vortex initial con-
dition or the quality of the mesh (skewness and non-orthogonality). One of the
main strengths of OpenFOAM
en de open-bron code
OpenFOAM