Numerical simulations of ﬂapping foil
and wing aerodynamics
Mesh deformation using radial basis functions
Copyright c 2009 by F.M. Bos
All rights reserved. No part of this material protected by this copyright notice
may be reproduced or utilized in any form or by any means, electronic or mechan
ical, including photocopying, recording or by any other information storage and
retrieval system, without written permission from the copyright owner.
Printed by Ipskamp Drukkers B.V. in The Netherlands
ISBN: 9789090251738
An electronic version of this thesis is available at http://repository.tudelft.nl
Numerical simulations of ﬂapping foil
and wing aerodynamics
Mesh deformation using radial basis functions
Proefschrift
ter verkrijging van de graad van doctor
aan de Technische Universiteit Delft,
op gezag van de Rector Magniﬁcus Prof. ir. K.C.A.M. Luyben,
voorzitter van het College voor Promoties,
in het openbaar te verdedigen op woensdag 24 februari 2010 om 10:00 uur
door
Frank Martijn BOS
ingenieur luchtvaart en ruimtevaart
geboren te Naaldwijk.
Dit proefschrift is goedgekeurd door de promotor:
Prof. dr. ir. drs. H. Bijl
Copromotor:
Dr. ir. B.W. van Oudheusden
Samenstelling promotiecommissie:
Rector Magniﬁcus, voorzitter
Prof. dr. ir. drs. H. Bijl Technische Universiteit Delft, promotor
Dr. ir. B.W. van Oudheusden Technische Universiteit Delft, copromotor
Prof. dr. ir. P.G. Bakker Technische Universiteit Delft
Prof. dr. ir. B. Koren Universiteit Leiden
Centrum Wiskunde & Informatica
Prof. dr. H. Jasak Zagreb University
Prof. dr. FO. Lehmann University of Ulm
Prof. dr. W. Shyy The University of Michigan
This research was supported by The Netherlands
Organisation for Scientiﬁc Research (NWO),
NWOALW grant 814.02.019.
Voor mijn ouders
Summary
Both biological and engineering scientists have always been intrigued by the ﬂight
of insects and birds. For a long time, the aerodynamic mechanism behind ﬂap
ping insect ﬂight was a complete mystery, until several decades ago. Experiments
showed the presence of a vortex on top of the ﬂapping wings, generates forces
larger than obtained by using conventional aircraft aerodynamics. Flapping wings
produce both lifting and propulsive forces such that it becomes possible for insects
and smaller bird species, e.g. hummingbirds, to stay aloft and hover, but also to
perform extreme manoeuvres. Because of this versatility, insects and smaller birds
are an inspiration for the development of ﬂapping wing Micro Air Vehicles, small
manmade ﬂyer’s to use in exploration and surveillance.
Several ﬂow visualisation experiments and numerical simulations have been
performed to improve the understanding of ﬂapping wing aerodynamics in order
to design and optimise Micro Air Vehicles. However, the eﬀects of wing kinemat
ics on the ﬂow and forces is still not fully understood. We performed two and
threedimensional numerical simulations in order to systematically vary relevant
parameters, related to the wing motion and ﬂow physics. In order to capture the
boundary layer and the near wake, it is important to maintain a high mesh quality
near the moving wing, especially at large rotations. Therefore, an accurate mesh
motion technique is necessary, which is able to cope with large mesh deformations.
In order to incorporate a ﬂapping wing in our numerical model, diﬀerent mesh mo
tion techniques are compared and improved. The overall goal of this part of the
research is to develop a reliable mesh deformation technique, in terms of accuracy
and eﬃciency, to solve the ﬂow around ﬂapping wings.
The ﬂow around ﬂapping wings, at the scale relevant to insect ﬂight, is highly
unsteady and vortical, described by the unsteady incompressible NavierStokes
equations. Diﬀerent dimensionless numbers are discussed, characterising the ﬂow,
i.e. Strouhal and Reynolds numbers. Since the ﬂow at the considered Reynolds
number, Re = O(100), is laminar, there is no need for additional turbulence
modelling, such that our simulations, assuming laminar ﬂow, may be treated as a
iv Summary
Direct Numerical Simulation (DNS).
In order to solve the unsteady incompressible NavierStokes equations, the com
mercial software Fluent
and the opensource code OpenFOAM
have been used
extensively. Diﬀerent mesh motion solvers are compared. Two existing methods
are assessed, solving the Laplace and a modiﬁed stress equation. Both methods
are very eﬃcient by using iterative solver techniques. However, these mesh motion
solvers are not able to maintain high mesh quality at large rotation angles, which
occur in insect ﬂight. Therefore, a new mesh motion solver is implemented, which
is based on the interpolation of radial basis functions.
This mesh motion solver is a point based method, which means that the dis
placement of all individual internal mesh points are evaluated, based on a given
boundary displacement, and updated accordingly. No mesh connectivity informa
tion is necessary, so that it can be applied to unstructured polyhedral meshes.
To increase its eﬃciency, a coarsening is applied to the set of moving boundary
points, such that only selected control points are used. This decreases the size of
the system of equations and associated computational eﬀort considerably.
After the discussion of the governing equations, ﬁnite volume discretisation in
OpenFOAM
and the assessment of the mesh motion solvers, the physical and
numerical modelling are described. The incompressible NavierStokes 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 twodimensional study is performed to investigate the eﬀects of diﬀer
ent wing kinematic models, with increasing complexity, on hovering ﬂight perfor
mance. The results show that the ‘sawtooth’ amplitude has a small eﬀect on the
mean lift but the mean drag is aﬀected signiﬁcantly. The second model simpliﬁca
tion, the ‘trapezoidal’ angle of attack, caused the leadingedge vortex to separate
during the translational phase. This led to an increase in mean drag during each
halfstroke. The extra ‘bump’ in angle of attack as used by the fruit ﬂy model is
not aﬀecting the mean lift to a large extent. The other realistic kinematic feature
is the deviation, which is found to have only a marginal eﬀect on the mean lift and
mean drag in this twodimensional study. However, the eﬀective angle of attack is
altered such that the deviation leads to levelling of the force distribution.
Additionally, a numerical model for twodimensional ﬂow was used to investi
gate the eﬀect of foil kinematics on the vortex dynamics around an ellipsoid foil
subject to prescribed ﬂapping 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 ﬂapping 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
ﬂow is periodic.
Furthermore, diﬀerent results relevant to threedimensional ﬂapping wing aero
v
dynamics, are described. First, the ﬂow around a dynamically scaled model wing
is solved for diﬀerent angles of attack in order to study the force development
and vortex dynamics at small and large midstroke angle of attack. Secondly, the
Rossby number is varied at diﬀerent 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 beneﬁcial for the stability of the leadingedge
vortex, leading to an increase in lift and eﬃciency. Thirdly, the threedimensional
wing kinematics is varied by changing the shape in angle of attack and by applying
a deviation, which may cause a ‘ﬁgureofeight’ pattern. As in twodimensional
studies, the deviation may inﬂuence the force distribution to a large extent, by
changing the eﬀective angle of attack. Additionally, the threedimensional ﬂow is
compared with the twodimensional studies performed on ﬂapping forward ﬂight.
Finally, a preliminary investigation is performed to show the eﬀect of wing
ﬂexing. Therefore, a predeﬁned ﬂexing deformation is applied to a plunging airfoil
in twodimensional forward ﬂight and to a threedimensional ﬂapping wing in
hovering ﬂight. Concerning the ﬂexible airfoil in forward ﬂight, 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 inﬂu
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 ﬂapping ﬂight . . . . . . . . . . . . . . . . . . . . . . . . 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 NavierStokes equations . . . . . . . . . . . . . . . . . . . . . . 16
2.2.1 Constitutive relations . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Incompressible laminar ﬂow simpliﬁcations . . . . . . . . . 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 Diﬀusion term . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5.4 Temporal term . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7 Solution of the NavierStokes equations . . . . . . . . . . . . . . . 27
2.7.1 Pressure equation and PressureVelocity coupling . . . . . . 28
2.7.2 Procedure for solving the NavierStokes equations . . . . . 29
2.7.3 Arbitrary Lagrangian Eulerian approach . . . . . . . . . . . 30
viii Contents
2.8 Swept volume calculation . . . . . . . . . . . . . . . . . . . . . . . 32
2.9 Numerical ﬂow solvers . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.10 Code validation and veriﬁcation . . . . . . . . . . . . . . . . . . . . 34
2.10.1 2D vortex decay and convection . . . . . . . . . . . . . . . . 35
2.10.2 Validation using cylinder ﬂows . . . . . . . . . . . . . . . . 39
2.11 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3 Mesh deformation techniques for ﬂapping ﬂight 49
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.2 Diﬀerent mesh deformation techniques . . . . . . . . . . . . . . . . 52
3.2.1 Laplace equation with variable diﬀusivity . . . . . . . . . . 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 twodimensional block . . . . 59
3.4.2 Flapping of a threedimensional wing . . . . . . . . . . . . . 62
3.4.3 Flexing of a twodimensional block . . . . . . . . . . . . . . 64
3.5 Improving computational eﬃciency . . . . . . . . . . . . . . . . . . 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 ﬂapping foils and wings 73
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.2 Governing equations for ﬂapping 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 ﬂexing . . . . . . . . . . . . . . . . 82
4.3.4 Numerical implementation of the wing kinematics . . . . . 83
4.4 Dynamical scaling of ﬂapping wings . . . . . . . . . . . . . . . . . 86
4.5 Computational domain and boundary conditions . . . . . . . . . . 88
4.6 Deﬁnition of force and performance coeﬃcients . . . . . . . . . . . 89
4.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5 Inﬂuence of wing kinematics in twodimensional hovering ﬂight 93
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.1.1 Similarity and discrepancy between two and
threedimensional ﬂows . . . . . . . . . . . . . . . . . . . . 94
5.1.2 Inﬂuence 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 Diﬀerent 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 twodimensional ﬂapping foil 121
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.2 Flapping foil parametrisation . . . . . . . . . . . . . . . . . . . . . 122
6.3 Force coeﬃcients and performance . . . . . . . . . . . . . . . . . . 124
6.4 Numerical model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.5 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.5.1 Inﬂuence of dimensionless wavelength . . . . . . . . . . . . 126
6.5.2 Inﬂuence of dimensionless amplitude . . . . . . . . . . . . . 127
6.5.3 Inﬂuence of angle of attack amplitude . . . . . . . . . . . . 129
6.5.4 Inﬂuence of stroke plane angle . . . . . . . . . . . . . . . . 130
6.5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
7 Vortical structures in threedimensional ﬂapping ﬂight 135
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.2 Threedimensional ﬂapping 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 identiﬁcation methods for ﬂow visualisation . . . . . . . . . 145
7.5 Flapping wings at low Reynolds numbers . . . . . . . . . . . . . . 148
7.5.1 The angle of attack in ﬂapping ﬂight . . . . . . . . . . . . . 149
7.5.2 Inﬂuence of ﬂapping stroke curvature . . . . . . . . . . . . . 149
7.5.3 Inﬂuence of Reynolds number . . . . . . . . . . . . . . . . . 152
7.5.4 Inﬂuence of ‘trapezoidal’ angle of attack . . . . . . . . . . . 158
7.5.5 Inﬂuence of deviation . . . . . . . . . . . . . . . . . . . . . 160
7.6 Flapping wings in forward ﬂight . . . . . . . . . . . . . . . . . . . . 163
7.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
8 Inﬂuence of wing deformation by ﬂexing 171
8.1 Airfoil ﬂexing in twodimensional forward ﬂapping ﬂight . . . . . . 171
8.2 Wing ﬂexing in threedimensional hovering ﬂight . . . . . . . . . . 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 ﬂapping ﬂight . . . . . . . . . . . . . . . . 181
9.3.1 Twodimensional hovering . . . . . . . . . . . . . . . . . . . 181
9.3.2 Threedimensional hovering . . . . . . . . . . . . . . . . . . 182
9.4 Conclusions on forward ﬂapping ﬂight . . . . . . . . . . . . . . . . 184
9.4.1 Twodimensional forward ﬂapping . . . . . . . . . . . . . . 184
9.4.2 Threedimensional forward ﬂapping . . . . . . . . . . . . . 184
9.5 Preliminary conclusions on wing ﬂexing . . . . . . . . . . . . . . . 185
9.6 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
A Grid generation for ﬂapping 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
solver settings . . . . . . . . . . . . . . . . . . . . . . . . . 195
B.3 OpenFOAM
solver settings . . . . . . . . . . . . . . . . . . . . . 197
Bibliography 201
Samenvatting 213
Acknowledgements 217
Curriculum Vitae 219
CHAPTER 1
Introduction
1.1 Motivation
The year of writing, 2009, is known as the year of Charles Darwin (18091882),
since he was born 200 years ago. At the age of 50, he published his world famous
Origin of Species. That book describes the natural selection, inspired by his scien
tiﬁc observations during a voyage (18311836) around the world with his ship, the
Beagle. At the Gal´ apagos Archipelago, Darwin discovered slightly diﬀerent bird
species living on the diﬀerent islands, whereas, he only knew one species on the
mainland of SouthAmerica. Apparently, the mainland bird species had travelled
to the islands and on every diﬀerent island it adapted to the diﬀerences in environ
mental circumstances. This process has become known as natural selection, which
is also applicable to the early era of ﬂight, millions of years ago. Birds are ancient
descendants of feathered dinosaurs (Templin, 2000) which developed the skill of
ﬂight in order to migrate over large distances and to catch prey. Long before the
origin of dinosaurs and birds, insects adapted to leave the ground to take oﬀ into
the thin air. Birds and insects are both ﬂapping their wings at diﬀerent length
scales, leading to a diﬀerent ﬂow behaviour. The larger the animal, the lower the
need for ﬂapping wings, e.g. the Andean Condor only ﬂaps when it looses height
in the thermal winds, whereas a small insect, a fruit ﬂy (WeishFogh & Jensen,
1956) ﬂaps with about 200 times per second.
Flapping wings produce both lifting and propulsive forces, such that it becomes
possible for insects and even smaller bird species, e.g. hummingbirds, to stay aloft
and hover, but also to perform extreme manoeuvres. Because of this versatility,
insects and smaller birds are a major inspiration of study to develop Micro Air
Vehicles (MAV), tiny manmade ﬂyer’s to use in exploration and surveillance, see
2 Introduction
ﬁgure 1.1.
To optimise the ﬂight performance of MAV’s it is important to get a thorough
understanding of the complex ﬂow generated by its wings, especially at smaller
length scales (< 5 cm). The ﬂapping wings induce complicated vortical struc
tures which inﬂuence the forces and performance characteristics in hovering and
forward ﬂight. In order to study this kind of ﬂows, researchers performed ﬂow
visualisations (WeishFogh & Jensen, 1956, Srygley & Thomas, 2002), detailed
experiments (Ellington et al., 1996, Sane & Dickinson, 2002, Poelma et al., 2006,
Lentink & Dickinson, 2009b) and numerical simulations (Wang et al., 2004, Sun
& Tang, 2002, Bos et al., 2008, Thaweewat et al., 2009). One limitation of doing
experiments is that they can be very expensive, in view of the need of precision
equipment and wind tunnel facilities. Secondly, the construction of models needs
to be very precise, which can be very costly as well. Additionally, when performing
wind tunnel experiments it is not straightforward to extract the force data, either
directly or indirectly, from the ﬂow visualisation obtained by Particle Image Ve
locimetry (PIV) (Poelma et al., 2006). Even when the most advanced techniques
are used, e.g. Digital Particle Image Velocimetry (DPIV), the ﬂow ﬁeld can not
be visualised in much detail, especially due to the reﬂections and shadows of the
moving wings. On the other hand, when performing numerical simulations, using
Computational Fluid Dynamics (CFD), the forces and ﬂow visualisations are a
direct result of the computations. Since it is interesting to solve for the forces
acting on a ﬂapping wing in combination with the vortical structures within the
near wake, performing CFD provides a suitable framework.
The present study deals with the development and improvement of computa
tional techniques to solve the ﬂow around ﬂapping wings at low Reynolds num
bers, O(100 − 1000). Section 1.2 brieﬂy provides background information on the
ﬂow physics concerned, while section 1.3 deals with the diﬀerent approaches for
analysing the ﬂow, experiments and numerical methods. Finally, section 1.4 de
scribes the objectives and approach of the present study as well as the outline of
this thesis.
1.2 Physics of ﬂapping ﬂight
In order to illustrate the necessity and diﬃculties with solving and visualising the
ﬂow around ﬂapping foils and wings, diﬀerent aspects of ﬂapping wing aerody
namics are discussed. The vortex dynamics, the leadingedge vortex in particular,
is brieﬂy discussed, as well as the inﬂuence of the wing kinematic modelling in
two and threedimensional problems.
Vortex generation in ﬂapping wing aerodynamics
Vortex generation in nature is fairly common in ﬂows induced by aeroplanes, birds,
insects, but also by boats and trees. Large aeroplanes generate wingtip vortices,
see ﬁgure 1.2, which can cause damage to a following aeroplane which encounters
1.2 Physics of ﬂapping ﬂight 3
(a) Wasp. (b) Entomopter. (c) Delﬂy.
Figure 1.1 Diﬀerent ﬂapping wing Micro Air Vehicle concepts. At lower Reynolds numbers,
ﬂapping MAV concepts can be used for hovering and low speed forward ﬂight, which is especially
interesting for intelligence and exploration. (a) Flying insect scale robotic model, which is able to
perform a tethered takeoﬀ (Wood, 2008). (b) The U.S. patented Entomopter has four ﬂapping wings
powered by chemicallyfuelled propulsion system (Michelson, 2008). (c) The Delﬂy Micro is camera
equipped and is able to hover (designed and developed at Delft University of Technology).
(a) (b) (c)
Figure 1.2 Vortex induced force generation. (a) Wingtip vortex causes big disturbances in the
wake, limiting the time between two successive aeroplane approaches. (b) Vortex generation in insect
ﬂight. A water strider generates vortices with its long legs to create the necessary propulsion (Hu et al.,
2003). (c) Willmott et al. (1997) performed smoke visualisation of the vortical ﬂow patterns induced
by a hawkmoth. It was observed that the leadingedge vortex was stabilised by the radial ﬂow moving
out towards the wing tip. Additionally, alternating vortex rings were seen in the wake, generated by
successive up and downstrokes.
this vortex. Another undesired eﬀect of vortex generation is ﬂow induced vibra
tion of cables, bridges or struts in water. On the other hand, vortex generation
provides possibilities to generate forces, which is used by birds, ﬁsh and insects,
e.g. ﬁgure 1.2 shows induced propulsive vortices generated by a water strider.
It is a common story that ﬂies could not ﬂy according to conventional aircraft
theory as developed by Lanchester (1907) and Prandtl (19141918). Prandtl did
develop a relation between the tip vortices, circulation and lift generation, but
this was not suﬃcient to explain the high lift generation of insects. This mystery
persisted until the discovery of the unsteady vortical ﬂow ﬁeld, ﬁgure 1.2, and
especially the generation of the leadingedge vortex.
4 Introduction
¯
C
L
= 1.540
x
c
[−]
3 2 1 0 1 2 3
(a) (b)
Figure 1.3 Forces and vortices in ﬂapping wing aerodynamics. (a) Twodimensional il
lustration of the wing kinematics and the resulting force vector generated by the ﬂapping airfoil at
Re = O(100) (Bos et al., 2008). (b) Threedimensional leadingedge vortex generated by a ﬂapping
wing at Re = O(1000).
Leadingedge vortex
The potential beneﬁt of vortices attached to the wing was already discussed
by Maxworthy (1979) and Dickinson & G¨otz (1993). It was Ellington et al. (1996)
who identiﬁed the presence of a leadingedge vortex (LEV) generated on top of the
ﬂapping wing, increasing the lift force to values much higher than predicted by con
ventional wing theory. The stability of the helical threedimensional leadingedge
vortex is still not yet fully understood and appears to heavily depend on the wing
kinematics and Reynolds number. It appears that the leadingedge vortex is more
stable around a threedimensional ﬂapping wing compared to twodimensional
ﬂapping foil situations.
In ﬂapping foil aerodynamics the vortices are shed and form either a periodic
or chaotic wake pattern, depending on the kinematics, notably advance ratio and
dimensionless ﬂapping amplitude (Thaweewat et al., 2009, Lentink et al., 2008).
The origin of the leadingedge vortex is the rollup of shear layers, present in
highly viscous ﬂows, which is the case at low Reynolds numbers. It is thought that
the kinematics in two and threedimensional ﬂapping inﬂuences the shear layer
direction and ﬂow accelerations, which will undoubtedly inﬂuence the development
of the leadingedge vortex (Lentink & Dickinson, 2009b).
In order to understand the physics of ﬂapping wing aerodynamics, it is im
portant to obtain insight in how an insect moves its wing. Figure 1.3(a) shows
a twodimensional illustration of the wing kinematics of a fruit ﬂy, operating at
Re = 110, from (Bos et al., 2008).
Inﬂuence of insect wing kinematics on forces
The relevance of experiments and ﬂow simulations of insect ﬂight has been found
to depend on how reliably true insect wing kinematics are reproduced. Wang
1.2 Physics of ﬂapping ﬂight 5
et al. (2004) and Sane & Dickinson (2001) showed that the kinematic modelling
signiﬁcantly inﬂuences the mean force coeﬃcients and its distribution. Addition
ally, Hover et al. (2004) showed that the angle of attack inﬂuences the ﬂapping
foil propulsion eﬃciency to a large extent. This illustrates the appreciable eﬀects
which details of the wing kinematics, like parameter values and stroke patterns,
may have on ﬂight performance. It further emphasises the need to critically assess
the inﬂuence of kinematic model simpliﬁcations.
In literature, diﬀerent kinematic models have been employed to investigate the
aerodynamic features of insect ﬂight. 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 ﬂapping am
plitude and frequency and showed that at a certain parameter selection the lift
is clearly enhanced. Lewin & HajHariri (2003) performed a similar numerical
study for heaving airfoils. Besides lift enhancement at certain reduced frequen
cies, they found periodic and aperiodic ﬂow solutions which are strongly related
to the aerodynamic eﬃciency. Lentink & Gerritsma (2003) varied airfoil shape
with amplitude and frequency ﬁxed at values representative to real fruit ﬂies.
They concluded that the airfoil geometry choice is of minor inﬂuence, 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 & G¨otz, 1993), rotational parameters were investigated by Dickinson
(1994). They varied rotational parameters and showed that axisofrotation, 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 ﬂight. Both studies emphasised the
importance of angle of attack to inﬂuence the propulsive eﬃciency. Slightly more
complex fruit ﬂy kinematic models were used by Dickinson et al. (1999) and Sane
& Dickinson (2001) with their Roboﬂy. Based on observation of true insect ﬂight,
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 Roboﬂy kinematic model. Using these models, the eﬀect
of amplitude, deviation, angle of attack and the timing of the latter were explored.
The present thesis deals with diﬀerent kinematic models from literature, both
the pure harmonic and the Roboﬂy model, in order to investigate their inﬂuence
on the aerodynamic performance (Bos et al., 2008). Furthermore, the results were
compared with more realistic fruit ﬂy kinematics obtained from the observation of
free ﬂying fruit ﬂies. Instead of performing a parameter study within the scope of
one kinematic model, the objective of the present study is to compare the eﬀect
6 Introduction
(a) (b) (c)
Figure 1.4 Vortex wakes generated by cylinders and ﬂapping 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 simpliﬁcations in kinematic modelling, which is of great importance to
both experiments and numerical simulations. Also, it may reveal the importance
of certain speciﬁc features of the stroke pattern, in relation to aerodynamic per
formance.
The similarity between two and threedimensional ﬂows
To limit both the parametric space as well as the computational eﬀort, many stud
ies have been performed as twodimensional simulations (Thaweewat et al., 2009,
Bos et al., 2008, Wang et al., 2004, Lewin & HajHariri, 2003). One of the major
(and partially unresolved) issues in modelling of insect ﬂight and ﬂapping wing
propulsion, is the possibly restrictive applicability of twodimensional results to
true insect ﬂight. Additional important aspects are unsteady ﬂow mechanisms,
wing ﬂexibility (ﬂuid structure interaction) and Reynolds number eﬀects. In a
recent paper Wang et al. (2004) compared threedimensional Roboﬂy results with
twodimensional numerical results. Both Dong et al. (2005) and Blondeaux et
al. (2005b) concluded that twodimensional studies over predict forces and per
formances since the energyloss, which is present in three dimensions, is not re
solved. Dong et al. (2005), Blondeaux et al. (2005b) numerically investigated the
wake structure behind ﬁnitespan wings at low Reynolds numbers. They observed
threedimensional vortical structures around ﬂapping wings with low aspect ratio,
as was mentioned by Lighthill (1969).
Notwithstanding the possible discrepancy between two and threedimensional
ﬂow, twodimensional analysis has often been applied to obtain insight into the
aerodynamic eﬀects of wing kinematics and geometry. Wang et al. (2004) con
ﬁrmed that the similarities between two and threedimensional approaches are
suﬃcient to warrant that a reasonable approximation of insect ﬂight can be ob
tained using a twodimensional approach. First, in case of advanced and symmet
ric rotation the forces were found to be similar in the twodimensional simulations
compared to the threedimensional experiments. Secondly it was observed that
in both simulations and experiments the leadingedge vortex did not completely
separate for amplitudetochord ratios between 35 (Dickinson & G¨otz, 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
threedimensional simulations, and with the above justiﬁcation, the ﬁrst part of
this thesis makes extensive use of twodimensional simulations. In the second
part, various threedimensional simulations were performed using limited varia
tions wing kinematics.
1.3 Experimental and numerical methods
In literature, diﬀerent methods were used to solve and visualise the ﬂow around
ﬂapping insect wings, from realistic fruit ﬂy measurements to threedimensional
simulations using a representative model wing. In this section, diﬀerent methods
will be brieﬂy addressed, from experimental methods to computational ﬂuid dy
namics simulations.
Experimental investigations and quasi steady theory
Several experimental studies considered the ﬂight performance of insects, and re
vealed the complex nature of insect ﬂight aerodynamics. The ﬂow 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 ﬁgure 1.2. More recently, Srygley & Thomas
(2002) and Thomas et al. (2004) performed free ﬂight and tethered experimental
visualisations using butterﬂies and dragonﬂies to show the complicated vortical
structures. This unsteady and vortical ﬂow 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 ﬂy, Drosophila
Melanogaster, in particular).
Ellington (1984) indicated that the lift in insect ﬂight is signiﬁcantly higher
than expected on the basis of quasisteady aerodynamics, hence revealing that
important unsteady and vortical ﬂow phenomena play a major role in insect ﬂight.
In several studies (Dickinson & G¨otz, 1993, Dickinson, 1994, Dickinson et al.,
1999) it was conﬁrmed that important aspects, like delayed stall and wake capture
enhance the lift force beyond values predicted by quasisteady theory. Ellington
et al. (1996) discovered that these lift increasing mechanisms are ampliﬁed by
the generation of a leadingedge vortex (LEV). It was shown that this leading
edge vortex arises during the translational part of the wing motion rather than
during the rotational ﬂip between up and down stroke. The lift increasing eﬀect of
the leadingedge vortex strongly depends on the kinematics of the ﬂapping wing
(Dickinson et al., 1999, Wang, 2000b, Sane & Dickinson, 2001, 2002, Lentink &
Dickinson, 2009a,b).
In order to understand insect ﬂight performance Dickinson et al. (1999) and
Wang (2000b) applied the quasisteady theory to compare with unsteady forces.
The quasisteady approach was revised by Sane & Dickinson (2002) to include ro
8 Introduction
tational eﬀects but even then the results require further improvement. According
to Sane & Dickinson (2001) the mean lift is well predicted by quasisteady theory,
but the mean drag is underestimated. This conﬁrms the restricted applicability of
the quasisteady 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 ﬂight. Dickinson et al. (1999)
investigated the ﬂow around a ﬂapping Roboﬂy model which moves in oil to meet
the same ﬂow conditions as the real fruit ﬂy encounters (reproduction of Reynolds
number in particular).
Numerical simulations
Notwithstanding important advances in experimental techniques for nonintrusive
ﬂow ﬁeld analysis, Particle Image Velocimetry in particular (Bomphrey et al., 2006,
Poelma et al., 2006), it remains diﬃcult to capture all the relevant details of the
ﬂow using only experimental techniques. An appealing approach, therefore, is to
supplement experiments with numerical ﬂow simulations. A number of numerical
studies on full threedimensional conﬁgurations have been reported, in relation to
speciﬁc insect geometries (moth: Liu & Kawachi (1998), fruit ﬂy: Ramamurti &
Sandberg (2002), Sun & Tang (2002), dragonﬂy: Young & Lai (2008), Isogai et al.
(2004)).
To perform numerical simulations around moving objects, such as ﬂapping
wings, one can use either immersed boundary methods (Peskin, 2002, Mittal &
Iaccarino, 2005), deforming mesh techniques (Boer de et al., 2007, Jasak, 2009),
see ﬁgure 1.5, or even complete remeshing (Young & Lai, 2008, Zuo et al.,
2007). Although, the computational eﬀort involved in threedimensional 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 ﬁxed 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 ﬁxed
boundaries (Mittal & Iaccarino, 2005). Nevertheless, when two wings touch, as in
the manoeuvre clapandﬂing (WeishFogh & Jensen, 1956), one will undoubtedly
need methods like overset, immersed boundary or remeshing techniques.
Together with the unavailability of an accurate ﬂow 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 eﬃciency. This
modern mesh motion technique is incorporated in OpenFOAM
1
, which is an
opensource framework to solve the NavierStokes equations on threedimensional
1
OpenFOAM
is a registered trade mark of OpenCFD
Limited, the producer of the
OpenFOAM
software.
1.4 Objectives and approach 9
(a) (b)
Figure 1.5 Diﬀerent 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 ﬂapping foil and wing simulations.
Arbitrary LagrangianEulerian formulation
The governing equations to solve the ﬂow are generally discretised using the Eule
rian description, where the ﬂuid is allowed to ﬂow through the ﬁxed mesh. This is
in contrast to the Lagrangian formulation, where the mesh is ﬁxed to the ﬂuid or
material. If the material or ﬂuid deforms, the mesh deforms with it. This method
is commonly used to discretise the governing equations encountered in structure
mechanics. However, when the ﬂow domain moves or deforms in time due to
a moving boundary, a ﬁxed 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 ﬂow 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 ﬂuid ﬂow through the mesh. At the time of writing, the
ALE method has become the standard implementation in most popular codes to
solve for the ﬂow around moving boundaries while the mesh deforms accordingly.
1.4 Objectives and approach
Flapping ﬂight 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 eﬃcient 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 ﬂow solver. Therefore, the overall goal of this research is to develop a reliable
mesh deformation technique, in terms of accuracy and eﬃciency, to solve the ﬂow
around ﬂapping wings. This method is used to study the complex vortical patterns
to identify optimal strategies in ﬂapping foil and wing aerodynamics. In order to
satisfy this aim, the following objectives are deﬁned:
1. improve current mesh motion techniques and implementation, using an ac
curate and eﬃcient framework,
2. validate and verify the numerical solver with the implemented and improved
mesh motion technique,
3. solve for the ﬂow around twodimensional ﬂapping foils to study the wing
wake interaction as well as the inﬂuence of wing kinematics,
4. solve for the ﬂow around threedimensional ﬂapping wings to assess the im
portance of parameters like ﬂapping amplitude, frequency or Reynolds num
ber,
5. study the threedimensional structure of vortical patterns, especially the
leadingedge vortex.
Approach and outline
In order to solve for the ﬂow around ﬂapping foils and wings, an improved mesh
motion technique, based on radial basis function interpolation, is implemented in
the opensource framework OpenFOAM
. The mesh motion technique is used
by an incompressible unsteady CFD solver to solve for the ﬂow around a three
dimensional ﬂapping wing on dense meshes in parallel.
To meet the objectives the current thesis is structured in the following chapters. In
order to solve the governing equations for ﬂuid ﬂow, a ﬁnite volume discretisation is
used, which is the subject of chapter 2. That chapter deals with the discretisation
of the diﬀerent terms as well as a deﬁnition of mesh quality. Furthermore, the solu
tion procedure is described together with a brief discussion about the opensource
framework OpenFOAM
, which is thoroughly validated and veriﬁed. Within the
code, diﬀerent mesh deformation techniques are incorporated. These mesh defor
mation techniques are described and assessed with respect to accuracy in chapter 3.
In addition to the already implemented mesh deformation techniques, a method
based on radial basis function interpolation is discussed. This mesh deformation
technique is implemented and used for ﬂapping foil and wing aerodynamics. Before
proceeding to the numerical results of the ﬂow around ﬂapping foils and wings, it
is important to discuss the numerical modelling for ﬂapping ﬂight in chapter 4.
Chapters 5 and 6 deal with the numerical investigations of twodimensional ﬂow
around a ﬂapping foil in hovering and forward ﬂight conditions, respectively. It is
1.4 Objectives and approach 11
found that the kinematic modelling has a large inﬂuence on forces, performance
and wake patterns. These twodimensional results provide good insight what to
expect of the threedimensional ﬂow around a ﬂapping wing, which is discussed
in chapter 7. Additionally, chapter 8 presents the preliminary results of a ﬂexing
wing in two and threedimensional ﬂapping ﬂight. Complex vortical structures
induced by a model ﬂapping wing can be accurately solved and analysed. It will
be seen that accurately solving the ﬂow around a ﬂapping wing is not an easy
task when the wing performs complex rotational motion. The conclusions and
recommendations can be found in chapter 9.
CHAPTER 2
Finite volume discretisation
A secondorder ﬁnite volume discretisation of the incompressible NavierStokes
equations on arbitrary polyhedral meshes is described. In addition to the mesh
quality measures nonorthogonality and skewness, the boundary conditions and the
solution procedure are presented. This ﬁnite volume approach is applicable to gen
eral commercial and noncommercial CFD codes. The commercial code Fluent
and the opensource code OpenFOAM
have been used for the simulations de
scribed in this thesis. Fluent
was already tested by Bos et al. (2008), such that
this chapter deals with the validation of OpenFOAM
using problems relevant for
low Reynolds number insect ﬂight. For test problems involving vortex decay and
convection, it was found that the Van Leer ﬂux limiter provides the most accurate
results, since the ﬂow is dominated by the convection of vortices. Furthermore,
the ﬂow around stationary and transversely oscillating cylinders showed that the
code of OpenFOAM solves the ﬂow in detail. Spatial and temporal convergence
was proved as well.
2.1 Introduction
Important aspects concerning numerical simulations, are being described. A nu
merical simulation needs to be performed in an accurate and concise way. There
fore, diﬀerent properties of a CFD simulation, like stability and convergence, are
addressed.
Important aspects of numerical simulations
Before describing the used methods in detail, four important aspects of numerical
14 Finite volume discretisation
(a) Structured (b) Unstructured
Figure 2.1 Diﬀerent mesh generation methods. Meshes can be generated in an structured way
(a) and using unstructured methods (b).
simulations are discussed, the governing equations, the discretisation method, the
numerical grid, and the solution method to solve the system.
In order to solve the incompressible NavierStokes equations, a suitable method
to discretise these equations needs to be chosen. In the ﬁeld of CFD, three methods
are commonly used, the ﬁnite diﬀerence, ﬁnite element and ﬁnite volume method.
Wesseling (2001) and Ferziger & Peric (2002) described these methods in more
detail. Traditionally, ﬁnite element methods are used for structural problems
whereas numerical simulations related to ﬂuid ﬂow are mostly solved with ﬁnite
volume methods, as is the case in the current thesis. When using the ﬁnite volume
method, the interpolation from cell centres to cell faces and how to approximate
the surface and volume integrals, needs to be described.
The third aspect, concerning a CFD simulation, is the generation of a numerical
grid, a division of the computational domain in a ﬁnite amount of cells. There are
three types of grids, structured, blockstructured and unstructured grids (Ferziger
& Peric, 2002). Figure 2.1 shows an example of a structured and an unstructured
grid. When using (block) structured grids, the cell ordering is fairly straight
forward such that the ﬂow solver uses this fact to solve the system in a more
eﬃcient way. A drawback of a (block) structured grid is that it is more diﬃcult
to create around complex geometries (commonly encountered in engineering prob
lems). This is the more important asset of unstructured grids. Besides the type
of grid, the cell shape can be varied from tetrahedral (three corners in two dimen
sions), hexahedral (four corners) to polyhedral (arbitrary number of corners) cells.
However, for less complex geometries, a structured grid is favourable in terms of
accuracy and eﬃciency of the ﬂow solver. Besides the spatial discretisation, the
time is discretised as well, which is necessary to perform unsteady simulations.
2.1 Introduction 15
Finally, the fourth aspect of a CFD simulation is the iterative solver. The dis
crete system of equations needs to be solved up to a certain convergence criterion.
Depending on the governing equations, discretisation method and the choice of
grid, the system of discretised equations can be easy or diﬃcult to solve, limiting
the iterative solver. When an appropriate iterative solver is used, a convergence
criterion needs to be applied for the inner (within the linear system) and outer
iterations (to couple the nonlinear parts and perform nonorthogonal corrections).
Properties of numerical solution methods
In order to solve the governing equations in a satisfactory manner, it is important
to discuss diﬀerent properties of the numerical solution method, consistency, con
vergence, stability, conservation and boundedness, from (Ferziger & Peric, 2002).
Since it is often not possible to ﬁnd a numerical method which outperforms on
all aspects, the choice of numerical method is usually a tradeoﬀ. The following
properties are relevant concerning numerical simulations, especially when using a
ﬁnite volume approach, applied in general commercial and noncommercial CFD
solvers.
The ﬁrst important property is consistency. The discretisation should become
exact when the mesh resolution tends to inﬁnity, i.e. when the cell size approaches
zero. The diﬀerence between the discretised and the exact solution is called the
truncation error. In order to check the consistency of the complete numerical
scheme, a grid and timestep convergence study has to be performed using in
creasing grid resolution and decreasing timestep. The second important property
of a numerical scheme is convergence. The solution of the discretised system of
equations should tend to the exact solution of the governing diﬀerential equations
as the mesh spacing tends to zero. Convergence is diﬃcult to prove theoretically
for real engineering applications, so commonly the empirical approach is followed,
where the same computation is repeated on subsequently reﬁned meshes. When
the solution converges to a gridindependent solution, the solution process is said
to be converged. However, it may happen that the exact solution is not approx
imated with decreasing timestep, when the method is not stable. Therefore,
stability is the third important aspect. When performing the iterations of the nu
merical process it should be the case that the numerical errors are not ampliﬁed.
In that case the solution process is called stable. For general engineering prob
lems the stability of the numerical process is strongly dependent on the timestep;
it should be suﬃciently small, depending on the temporal discretisation scheme.
The fourth property of a numerical scheme is conservation. Considering a steady
problem, without sources or sinks, the mass ﬂux of a conserved quantity through
a speciﬁed system should be zero. Since the governing equations in ﬁnite volume
formulation are conservative, this property should be respected by the discretised
equations. One of the advantages of the ﬁnite volume approach is that conserva
tion is guaranteed for every small control volume and therefore, for the complete
computational domain as a whole. Finally, the last aspect is boundedness. Certain
variables in the governing equations contain physical bounds, like concentration or
16 Finite volume discretisation
density and all other nonnegative variables. When the numerical process respects
these physical bounds, the method is called bounded.
In this thesis, two diﬀerent CFD codes, the commercial ﬂow solver Fluent
and the
noncommercial opensource code OpenFOAM
are used. Fluent
is a general
purpose CFD solver, which has been an authority in the ﬁeld of computational
ﬂuid dynamics for decades. Two major drawbacks of a commercial solver are the
unavailability of the source code and the potential lack of suﬃcient support from
the company or the user community in code development. The used opensource
solver, OpenFOAM
, provides the source code and there is a big user community,
providing support for code development.
The remainder of this chapter deals with a description of the governing equa
tions of ﬂuid ﬂow in section 2.2. To solve the governing equations, the spatial
and temporal discretisation methods are described in section 2.3, followed by a
discussion about the cell quality measures, like skewness and nonorthogonality
in 2.4. In section 2.5 a general transport equation will be discretised to show how
to deal with the diﬀerent terms, like diﬀusion and convection. Additionally, a brief
discussion about the treatment of boundary conditions is provided in section 2.6.
When the discretised transport equation and corresponding boundary conditions
are fully explained, the solution procedure to solve the incompressible Navier
Stokes equations will be dealt in section 2.7. These numerical solution procedures
are present in the used ﬂow solvers, Fluent
and OpenFOAM
. Section 2.9 pro
vides a brief description of the background and usage of both CFD codes. In order
to validate and verify the CFD codes for our problem, some small test problems are
deﬁned in order to test the inﬂuence of diﬀerent numerical settings, like discreti
sation schemes, grid resolution and timestep size. The validation and veriﬁcation
discussion is the subject of section 2.10. Finally, the major conclusions of this
chapter are summarised in section 2.11.
2.2 The NavierStokes equations
The governing equations for viscous ﬂuid ﬂow are a coupled set of nonlinear par
tial diﬀerential equations (Anderson Jr., 1991, Panton, 2005). These equations
are derived from conservation of mass, momentum and energy within an inﬁnitesi
mally small spatial control volume. For mass conservation, the following continuity
equation is obtained:
∂ρ
∂t
+∇
•
(ρu) = 0. (2.1)
Here, ρ [kg/m
3
] is the density and u [m/s] the ﬂow velocity vector. The nabla ∇
operator is deﬁned in three dimensions as
∇=
∂.
∂x
+
∂.
∂y
+
∂.
∂z
.
2.2 The NavierStokes equations 17
Secondly, for momentum conservation the following expression can be derived
(neglecting gravity and additional body forces):
∂(ρu)
∂t
+∇
•
(ρuu) = ∇
•
σ, (2.2)
where σ [N/m
2
] is the surface stress tensor, necessary in viscous ﬂuid ﬂow. For
compressible ﬂow calculations also the energy conservation equation is speciﬁed:
∂(ρe)
∂t
+∇
•
(ρeu) = ∇
•
(σu) −∇
•
q +ρQ, (2.3)
where e [J/kg] is the total speciﬁc energy (including kinetic and potential energy),
q [W/s] is the heat ﬂux vector and Q [J·m
3
/kg] equals the nett energy generation.
The full set of equations describing unsteady, compressible viscous ﬂows, are called
the NavierStokes equations. The NavierStokes equations are nonlinear, which
makes them diﬃcult to solve; only for very simpliﬁed 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 ﬂuid, the stress tensor, σ, which is deﬁned for a
Newtonian ﬂuid as
σ = −
_
p +
2
3
µ∇
•
u
_
I +µ
_
∇u +∇u
T
_
.
Here I represents the identity tensor, p [N/m
2
] is the pressure and µ [Ns/m
2
] is the
dynamic viscosity. To close the energy equation, the equation of state is speciﬁed,
such as the perfect gas law:
p = ρRT,
in which T [K] is the temperature and R [J/(mol·K)] the speciﬁc gas constant.
The constitutive relation for the total speciﬁc energy yields as follows:
e = e(p, T).
Additionally, the heat conduction is described using Fourier’s law:
q = λ∇T,
with λ [W/(m·K)] the heat conduction transport coeﬃcient.
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 ﬂows, turbu
lence research, multiphase ﬂows and a lot of other applications. However, these
equations can be diﬃcult to solve and simpliﬁcations can be made if applicable to
the concerning problem.
18 Finite volume discretisation
2.2.2 Incompressible laminar ﬂow simpliﬁcations
The current research deals with the ﬂow around ﬂapping wings at insect scale,
which is considered to be incompressible (Lentink, 2003, Bos et al., 2008) and
laminar (Williamson, 1995). Therefore, the incompressible laminar NavierStokes
equations are solved for Reynolds numbers ranging from Re = 100 to 1000.
A ﬂow 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
eﬀects can be neglected. The incompressible NavierStokes equations are:
∇
•
u = 0, (2.4)
∂u
∂t
+∇
•
(uu) = −
∇p
ρ
+ν∇
2
u, (2.5)
with ν = µ/ρ [m
2
/s] being the kinematic viscosity. For an incompressible ﬂow,
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 diﬀerent 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 ρ are scaled with their reference values, as follows:
u
∗
=
u
U
ref
, t
∗
= t · f
ref
, x
∗
=
x
L
, p
∗
=
p
ρ
ref
· U
2
ref
, ρ
∗
=
ρ
ρ
ref
. (2.6)
The star (*) is used to indicate the dimensionless variables. In case of incom
pressible ﬂow, the density is constant, such that ρ
∗
= 1. When substituting
equation (2.6) into equations (2.4) and (2.5) the following nondimensional form
of the incompressible continuity and momentum equations is obtained:
∇
•
u
∗
= 0, (2.7)
and
St
∂u
∗
∂t
+∇
•
(u
∗
u
∗
) = −∇p
∗
+
1
Re
∇
2
u
∗
. (2.8)
In these equations, two main dimensionless numbers are identiﬁed 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
ν
=
T
visc
T
conv
. (2.10)
These dimensionless numbers represent order estimates for timescale ratios in the
ﬂow. In (2.9) and (2.10), these relevant timescales are, respectively, the timescale
2.3 Spatial and temporal discretisation 19
for convective transport, T
conv
, viscous transport, T
visc
and the relevant timescale
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 ﬂuid ﬂow is governed by nonlinear partial diﬀerential equa
tions, which can only be solved analytically for extremely simpliﬁed model prob
lems. The full NavierStokes equations, combined with the constitutive relations,
are applicable to all kind of ﬂows, where the computational costs strongly depend
on the desired resolution and solution methods. When the ﬂow is considered lam
inar and incompressible, the governing equations are signiﬁcantly simpliﬁed, such
that the costs for solving may be reduced. However, these simpliﬁcations need to
be justiﬁed by the concerned ﬂuid ﬂow problem. Concerning ﬂapping wing physics
at lower Reynolds numbers (100 ≤ Re ≤ 1000) the ﬂow inherently is incompress
ible and laminar. Therefore, solving the unsteady incompressible laminar ﬂow can
be seen as performing a Direct Numerical Simulation, at suﬃciently 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 suﬃcient to specify the initial timestep,
which is used to march linearly in time, starting at the initial solution. The
timestep 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 ﬁnite volume approach needs a domain sub
division into a ﬁnite number of convex polyhedral control volumes without overlap,
completely ﬁlling the domain. OpenFOAM
uses a collocated variable arrange
ment (Ferziger & Peric, 2002), which means that every control volume centre is
used to store the values of all variables, like pressure and velocity. Figure 2.2 shows
an arbitrary polyhedral control volume V
P
with centre P and neighbouring centre
N. The computational point x
P
is located at the centroid of the computational
cells, which is found from the following relation (Jasak, 1996):
_
V
P
(x −x
P
)dV = 0.
Every two cells, i.e. with centres P and N from ﬁgure 2.2, share an internal
face whose geometric centre is denoted by f and has an outward pointed normal
vector S
f
. Faces which are not shared are boundary faces, consequently. Derived
boundary ﬁelds, like surface normal gradients or face ﬂuxes, are deﬁned in the face
centre.
After the domain is discretised into a set of control volumes, face surfaces
and points, the governing equations need to be approximated over these cells.
Discretisation is performed assuming a linear variation of scalar variable φ across
20 Finite volume discretisation
x
y
z
V
P
r
P
P
f
S
f N
d
f
Figure 2.2 Discretisation of the computational domain using ﬁnite volume cells. An
arbitrary polyhedral control volume is constructed around a centre P and with volume V
P
. The vector
from the cell centre to the neighbouring cell centre N is d
f
. The faces of cell P are directed with the
unit normal vector S
f
and may have an arbitrary number of corners. From (Jasak, 1996).
a cell. This scalar variable φ can be seen as pressure or a velocity component.
Using a Taylor series approximation, the following expression is obtained:
φ(x) = φ
P
+ (x −x
P
) · (∇φ)
P
+O((x −x
P
)
2
), (2.11)
where O((x − x
P
)
2
) represents the secondorder truncation error. For the tem
poral variation of this scalar variable φ a similar expression can be found:
∂φ(t)
∂t
=
φ(t + ∆t) −φ(t)
∆t
+O(∆t). (2.12)
With this linear temporal behaviour of φ the truncation error is secondorder
O(∆t
2
), similar to the spatial truncation O(∆x
2
). Both truncation errors can be
expanded using a full Taylor series expansion, which is not within the scope of the
present thesis, but can be found in (Wesseling, 2001, Ferziger & Peric, 2002, Jasak,
1996). Since this discretisation approach is able to cope with arbitrary polyhedral
cell volumes, this method can be used for complex unstructured threedimensional
meshes, including local mesh reﬁnement.
2.4 Measures of cell quality
Since the accuracy of the numerical solution heavily depends on the interpolation
from cell to face centre, one can imagine that the cell quality is very important.
We will brieﬂy describe the cell quality based on nonorthogonality and skewness,
which will both be used to assess the performance of mesh motion solvers in
chapter 3.
First of all, cell nonorthogonality is deﬁned in ﬁgure 2.3(a) by the angle α
N
between the face normal vector S
f
and the line connecting the two cell centres,
d. This angle needs to be as small as possible in order to minimise the truncation
2.5 Discretisation of an incompressible momentum equation 21
P N
f d
S
f
α
N
(a) Cell nonorthogonality.
P N f
i
S
f
f
m
d
(b) Cell skewness.
Figure 2.3 Quality measures using cell nonorthogonality and cell skewness. Two
dimensional representation of cell nonorthogonality (a) and cell skewness (b) as a measure for the
ﬁnite volume cell quality. Cell nonorthogonality is deﬁned 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 deﬁned by the
vectors m and d. From (Jasak, 1996).
error of the diﬀusion term. The second quality criterion is the cell skewness, see
ﬁgure 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 deﬁned by:
ψ =
m
d
,
where m and d are deﬁned in ﬁgure 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 ﬁnite volume discretisation of the
incompressible momentum equation, which forms the basis for the incompressible
NavierStokes equations. This partial diﬀerential equation has the following form:
∂u
∂t
+∇
•
(uu) −∇
•
(ν∇u) =
∇p
ρ
. (2.13)
Here, ρ [kg/m
3
] is the reference density, u [m/s] the transport velocity and ν =
µ/ρ [m
2
/s] is the kinematic viscosity. This expression contains a temporal, con
22 Finite volume discretisation
vection and a diﬀusion term, given by:
∂u
∂t
: temporal term,
∇
•
(uu) : convection term,
∇
•
(ν∇u) : diﬀusion term.
Using the ﬁnite volume approach, the integral form of the incompressible Navier
Stokes equations is obtained by integrating over a control volume, C
V
:
_
V
CV
∂u
∂t
dV +
_
V
CV
∇
•
(uu)dV −
_
V
CV
∇
•
(ν∇u)dV =
_
V
CV
∇p
ρ
dV. (2.14)
This equation is solved in both CFD codes used, Fluent
and OpenFOAM
. In
the remainder of this section, the diﬀerent terms of equation (2.14) are elaborated
in more detail.
Before dealing with the discretisation of the diﬀerent 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 diﬀusion terms. For this, the scalar variable φ is used, which may represent
the diﬀerent velocity components. When substituting equation (2.11) into the
volume integral, a secondorder approximation is obtained, such that the result is
a multiplication of the scalar value multiplied by the cell volume.
_
V
P
φ(x)dV =
_
V
P
(φ
P
+ (x −x
P
)(∇φ)
P
)dV
= φ
P
_
V
P
dV + (
_
V
P
(x −x
P
)dV ) · (∇φ
P
)
≈ φ
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 deﬁnes 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 secondorder linear interpolation method.
Discretisation of the gradient integral of a scalar variable φ can be written,
using Gauss’ theorem, as:
_
V
P
∇φdV =
_
S
CV
dSφ,
=
f
_
S
f
dSφ,
≈
f
dS
f
.φ
f
.
2.5.1 Face interpolation schemes
Similar to the divergence term, the ﬂux φ
f
is interpolated from the cell centre to
the face centres. A linear interpolation is performed using the following expression:
φ
f
= f
x
φ
P
+ (1 −f
x
)φ
N
,
which is illustrated in ﬁgure 2.5. The linear interpolation factor f
x
is deﬁned 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 diﬀerencing.
In OpenFOAM
, Jasak et al. (1999) applied extra face interpolation schemes,
like upwind blending using a gamma coeﬃcient and diﬀerencing using a ﬂux split
ting limiter such as SuperBee (Roe, 1986), the Koren limiter (Koren, 1993), or
Van Leer (Van Leer, 1979, Sweby, 1984). The SuperBee, Koren and Van Leer lim
iter are shown using the Sweby diagram in ﬁgure 2.4 (Sweby, 1984). The purpose
of ﬂux limiters is to limit the gradient of the solution in order to avoid spuri
ous oscillations and to improve the stability of the scheme. Section 2.10 shows a
comparison of the results using diﬀerent ﬂux limiters on the solution of a model
problem of vortex decay and convection.
2.5.2 Convection term
When the volume integral of the convection term from equation (2.13) is consid
ered, the following relation can be derived using equation (2.16) and a linearization
24 Finite volume discretisation
r
φ
(
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
φ
(
r
)
0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
2
2.5
3
(b) Koren
r
φ
(
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 Diﬀerent ﬂux splitting limiters. Flux splitting schemes are used to limit the gradient
of the solution in order to avoid spurious wiggles. The ﬂux splitting scheme are a function of r, which
represents the ratio of successive gradients on the mesh. Diﬀerent ﬂux 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
∇
•
(uφ)dV =
f
S
f
· (uφ)
f
=
f
S
f
· (u)
f
φ
f
=
f
Fφ
f
.
Here, F is the mass ﬂux, given by F = S
f
· (u)
f
. The scalar variable φ needs to
be interpolated using a secondorder interpolation method in combination with a
ﬂux limiter, e.g. linear, Gamma, Van Leer.
2.5.3 Diﬀusion term
The volume integral of the diﬀusion term from equation (2.13) is discretised and
approximated using linearization as
_
V
P
∇
•
(ν∇φ)dV =
f
S
f
· (ν∇φ)
f
=
f
ν
f
(S · ∇φ)
f
.
Here, the terms (S· ∇φ)
f
and ν
f
need to be approximated using a proper method.
The face viscosity ν
f
is obtained by interpolation from cell centre to faces. The
other term (S · ∇φ)
f
is obtained on a nonorthogonal mesh by the following ex
pression:
(S · ∇φ)
f
= m
φ
N
−φ
P
d
+k · (∇φ)
f
.
2.5 Discretisation of an incompressible momentum equation 25
ts
Flow direction
U P D
φ
U
φ
P
φ
f
φ
D
f
P d f m N
S
f
α
N
k
Figure 2.5 Variation of the ﬂux φ. The
value of φ at the face f is determined as a
function of upstream and downstream values.
Figure 2.6 Cell nonorthogonality
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 ﬁgure 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 NavierStokes equations are solved, a proper discretisation of
the temporal scheme is necessary. The time derivative represents the temporal
rate of change of φ which needs to be discretised using new and old time values.
This time diﬀerence is deﬁned using prescribed timestep size ∆t such that:
φ
n+1
= φ
n
+ ∆t,
where φ
n
and φ
n+1
are the scalar variable φ at the old and new time instances,
respectively. Two implicit time discretisation methods are considered, one ﬁrst
order and and one secondorder scheme. The ﬁrstorder discretisation is simply
the temporal diﬀerence:
∂φ
∂t
=
φ
n+1
−φ
n
∆t
,
and the secondorder discretisation, see (Wesseling, 2001, Hirsch, 1988), is given
by:
∂φ
∂t
=
3
2
φ
n+1
−φ
n
+
1
2
φ
n−1
∆t
.
This implicit scheme is referred to as the secondorder backward diﬀerencing
scheme, where φ
n−1
is the oldold value of φ. Consequently, the corresponding
26 Finite volume discretisation
volume integrals obey the following relations:
_
C
V
∂φ
∂t
dV =
φ
n+1
−φ
n
∆t
V
P
,
_
C
V
∂φ
∂t
dV =
3
2
φ
n+1
−2φ
n
+
1
2
φ
n−1
∆t
V
P
. (2.17)
Note, that these relations are only valid on ﬁxed meshes and constant timesteps.
According to (Wesseling, 2001, Hirsch, 1988), the explicit ﬁrstorder time integra
tion method may be unstable if the Courant number is larger than 1, where the
Courant number is deﬁned 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 ﬁrst and secondorder backward
scheme have been used. While the implicit methods are bounded and stable, a
predeﬁned maximal Courant number Co
max
is used to vary the corresponding
timestep during the simulation. In that case, the coeﬃcients
3
2
, 2 and
1
2
in (2.17)
should be elaborated to incorporate the ratio of the old and current timesteps.
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 deﬁned 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. zerogradient boundary condition, deﬁning the solution gradient to be zero.
This condition is known as a Neumanntype condition, ∂φ/∂n = a,
2. ﬁxedvalue boundary condition, deﬁning a speciﬁed value of the solution.
This is a Dirichlettype condition, φ = b,
3. symmetry boundary condition, treats the conservation variables as if the
boundary was a mirror plane. This condition deﬁnes that the component
of the solution gradient normal to this plane should be ﬁxed to zero. The
parallel components are extrapolated from the interior cells,
4. movingwallvelocity boundary condition is used on a moving boundary to
keep the ﬂux zero, using the Arbitrary Lagrangian Eulerian approach.
For external ﬂow simulations, a distinction is made between the outer and the
inner boundaries, the latter corresponds to the moving wing or body. To minimise
the eﬀects of the outer boundaries it is desirable to specify a symmetry boundary
2.7 Solution of the NavierStokes equations 27
condition (3) at those ﬁxed boundaries, unless a freestream is speciﬁed. In case of
forward ﬂapping ﬂight, two domain boundaries are deﬁned as inﬂow and outﬂow,
respectively. At the inﬂow boundary the velocity is deﬁned as ﬁxedvalue (2) and
the pressure as zerogradient (1). On the other hand, at the outﬂow boundary,
the pressure has to be ﬁxedvalue and the velocity zerogradient (Hirsch, 1988,
Wesseling, 2001). On a stationary wall the noslip condition needs to be guaran
teed, therefore a ﬁxedvalue (u = 0) is speciﬁed for the velocity in combination
with a zerogradient for the pressure. If the boundary of the wall moves, than
the proper boundary condition is the movingwallvelocity (4) which introduces
an extra velocity in order to maintain the noslip condition and ensures a zero ﬂux
through the moving boundary.
2.7 Solution of the NavierStokes equations
Previously, the diﬀerent terms to discretise the general momentum equation (2.13),
were described. This section brieﬂy 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
ρ
+ν∇
2
u.
There are two items, requiring special attention, namely the nonlinear term
present in the momentum equation and the pressurevelocity coupling (Ferziger &
Peric, 2002). The nonlinear term in these governing equations, ∇
•
(uu), can be
solved either by using a solver for nonlinear 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 oﬀdiagonal 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 nonlinear
systems, this convection term is linearised such that existing velocity ﬁelds will be
used to calculate the matrix coeﬃcients a
p
and a
N
.
28 Finite volume discretisation
2.7.1 Pressure equation and PressureVelocity coupling
Since the pressure depends on the velocity and viceversa, a special treatment of
this interequation coupling is needed. In order to derive the pressure equation, a
semidiscrete 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 coeﬃcients 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 ﬁnd an expression for u
p
:
u
p
=
H(u)
a
p
−
1
a
p
∇p. (2.20)
The velocities on the face of a ﬁnite 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 ﬂuxes. 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 ﬁnal form of the discretised
NavierStokes equations can be written as:
a
p
u
p
= H(u) −
f
S
f
(p)
f
, (2.23)
2.7 Solution of the NavierStokes equations 29
and
f
S
f
•
__
1
a
p
_
(∇p)
f
_
=
f
S
f
•
_
H(u)
a
p
_
f
, (2.24)
additionally, the face ﬂux 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 satisﬁed, the face ﬂuxes are guaranteed to be conser
vative (Ferziger & Peric, 2002). For the discretised form of the NavierStokes
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 innerequation coupling is established using either PISO (Issa, 1986) or SIM
PLE (Patankar & Spalding, 1972) based algorithms. Both ﬂow solvers, Fluent
and OpenFOAM
used the PISO scheme for transient ﬂows and the SIMPLE
scheme for steady ﬂows. 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 timestep, since the
actual pressure gradient is not yet calculated. Furthermore, equation (2.23)
provides an approximation of the new velocity ﬁeld.
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 ﬁeld can be obtained.
3. Explicit velocity correction stage. The last equation in this sequence
is (2.25), which determines the conservative ﬂuxes, which are consistent with
the new pressure ﬁeld. Since the approximated pressure ﬁeld, from stage 1,
is replaced by a better pressure ﬁeld, from stage 2, the velocity ﬁeld 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 NavierStokes equations
After dealing with the discretisation of the NavierStokes equations in combination
with the PISO algorithm, it has become possible to describe the solution procedure
to obtain the solution of the NavierStokes equations. In unsteady simulations all
other interequation couplings, besides the pressurevelocity equations, are lagged,
such that they are included in the PISO loop. For incompressible unsteady ﬂow
30 Finite volume discretisation
with additional turbulence modelling, the solution sequence can be summarised as
follows:
1. Initialisation of all ﬁelds, including pressure and velocity, using the initial
condition,
2. Start the simulation to obtain the velocity and pressure values at the new
timestep,
3. Create and solve the momentum predictor equations using the obtained face
ﬂuxes,
4. Iterate through the PISO loop until the predeﬁned tolerance of the pressure
velocity system is reached. The pressure and velocity ﬁelds are obtained for
the current timestep, as well as a new set of conservative ﬂuxes,
5. Using the new conservative ﬂuxes, all remaining equations of the system are
solved. If turbulence modelling is included, update the turbulent viscosity
at this stage,
6. Unless the ﬁnal time is reached, go back to step 2.
This procedure results in solution ﬁelds 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 modiﬁcations 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 ﬂow are generally discretised using the Eule
rian description, where the ﬂuid is allowed to ﬂow through the ﬁxed mesh (Ferziger
& Peric, 2002). This is in contrast to the Lagrangian formulation, where the mesh
is ﬁxed to the ﬂuid or material. If the material or ﬂuid deforms, the mesh deforms
with it. This method is commonly used to discretise the governing equations
encountered in structural mechanics. However, when the ﬂow domain moves or
deforms in time due to a moving boundary, a ﬁxed 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 ﬂow
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 ﬂuid ﬂow 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 ﬂow around moving boundaries while the
2.7 Solution of the NavierStokes equations 31
mesh deforms accordingly. In general, the momentum equation (2.14) for a scalar
ﬁeld φ can be derived on a moving mesh as
∂
∂t
_
V
CV
ρφdV +
_
S
CV
ρn·(u−u
s
)φdS−
_
S
CV
ρΓ
φ
n·∇φdS =
_
V
CV
S
φ
(φ)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 deﬁned by the socalled 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 ﬁnite volume discretisation, the computational domain is split
into a ﬁnite number of polyhedral cells with varying shape and volume, since the
mesh is deforming. The cells do not overlap and completely ﬁll the domain (Jasak,
2009). In time, the temporal dimension is marched using a variable timestep,
corresponding to a maximal Courant number, using either an implicit ﬁrstorder
Euler method or a secondorder backward scheme (Tukovi´c & Jasak, 2007). For
eﬃciency, the secondorder 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 timestep:
3ρ
n+1
P
φ
n+1
P
V
n+1
P
−4ρ
n
P
φ
n
P
V
n
P
+ρ
n−1
P
φ
n−1
P
V
n−1
P
2∆t
+
f
( ˙ m
n+1
f
−ρ
n+1
f
˙
V
n+1
f
)φ
n+1
f
=
f
(ρΓ
φ
)
n+1
f
S
n+1
f
n
n+1
f
•
(∇φ)
n+1
f
+s
n+1
φ
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 n−1 are, respectively, the new, old and
oldold values. The mass ﬂux 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 ﬂuid mass ﬂux ˙ m is obtained as part of the solution, satisfying mass
conservation. Furthermore, it is important to determine the volume face ﬂux
such that it satisﬁes 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 ﬂux 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 timestep, 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 ﬁnite volume
cell decomposition is used to form a tetrahedral mesh used for pointbased 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 deﬁned maximal Courant number leading to a varying timestep. Therefore, the
following relation is derived for a nonconstant timestep:
_
1 +
∆t
n+1
∆t
n+1
+ ∆t
n
_
ρ
n+1
P
φ
n+1
P
V
n+1
P
−
_
1 +
∆t
n+1
∆t
n
_
ρ
n
P
φ
n
P
V
n
P
+
_
1 +
(∆t
n+1
)
2
∆t
n
(∆t
n+1
+ ∆t
n
)
_
ρ
n−1
P
φ
n−1
P
V
n−1
P
+
f
( ˙ m
n+1
f
−ρ
n+1
f
˙
V
n+1
f
)φ
n+1
f
=
f
(ρΓ
φ
)
n+1
f
S
n+1
f
n
n+1
f
•
(∇φ)
n+1
f
+s
n+1
φ
V
n+1
P
,
where the new and old timesteps are respectively given by ∆t
n+1
= t
n+1
−t
n
and
∆t
n
= t
n
−t
n−1
.
2.8 Swept volume calculation
The swept volume is deﬁned as the volume swept by a face of a polyhedral cell
between two subsequent timesteps, 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 timestep 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
ﬁgure 2.7(a). The polyhedral face is decomposed into triangles, using its centroid,
2.9 Numerical ﬂow solvers 33
which is illustrated in ﬁgure 2.7(b). The swept volume of a polyhedral face f
is equal to the sum of the swept volumes of the diﬀerent decomposed triangles,
which need to be accurately calculated. As is illustrated in ﬁgure 2.7(b), the swept
volume of such a triangle is similar to a prism with a triangleshaped 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 ﬁgure, but the remaining
volume of the prism contains two more tetrahedrons. These two additional tetra
hedrons can be constructed in two diﬀerent unique ways, using either one of the
two diagonals in the right side face of the swept volume shown in ﬁgure 2.7(b),
see (Zuijlen van, 2006). Therefore, the total swept volume using the two diﬀerent
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 diﬀerent 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 ﬂow
with mesh motion and proved to be suﬃciently accurate. Therefore, this method
is implemented in OpenFOAM
.
2.9 Numerical ﬂow solvers
Section 2.7 described the solution procedure to solve the NavierStokes equations,
necessary for ﬂuid ﬂow. This section deals with a brief elaboration of the computer
software, i.e. the CFD codes, used throughout the present research. Two diﬀerent
CFD codes are used, one commercial (Fluent
) and one opensource package
(OpenFOAM
).
Fluent
is a wellknown, easy to use and proven CFD solver, which exploits
the ﬁnite volume approach. For completeness, the main settings that we used,
are brieﬂy discussed. The spatial discretisation was secondorder upwind and
the time discretisation was ﬁrstorder implicit Euler (Hirsch, 1988), which is the
only method for which the dynamic mesh module is implemented by Fluent
.
The pressurevelocity coupling in incompressible ﬂow simulations was obtained
using the iterative PISO scheme (Ferziger & Peric, 2002). The accuracy was set
to doubleprecision and the initial conditions were chosen to be uniform. The
boundary condition on the body was set to noslip. The convergence criterion for
34 Finite volume discretisation
the iterative method was satisﬁed with mass and momentum residuals dropping
O(10
−5
) in magnitude.
The other code used in this research, OpenFOAM
, is a general objectoriented
toolbox, written in C++, which is used to solve partial diﬀerential equations, e.g.
the NavierStokes equations, on a ﬁnite 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 secondorder central diﬀerencing, 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 secondorder backward scheme is
used, in combination with a variable timestep corresponding a maximal Courant
number (Wesseling, 2001, Ferziger & Peric, 2002).
Additionally, the iterative solvers and their corresponding convergence criteria
need to be speciﬁed. 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 deﬁned by:
Res = b −Ax
The pressure equation is solved using a preconditioned conjugate Gradient (PCG)
iterative solver, while the pressurevelocity coupling equation employs its asym
metric counterpart preconditioned bistab conjugate Gradient (PBiCG) solver.
The preconditioning 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 ﬂow solvers Fluent
and OpenFOAM
.
2.10 Code validation and veriﬁcation
This section deals with the validation and veriﬁcation of the CFD solvers that
were used throughout the current research. The commercial ﬂow solver Fluent
has already been tested earlier speciﬁcally for low Reynolds number ﬂows, relevant
to ﬂapping insect ﬂight in (Lentink & Gerritsma, 2003, Zuo et al., 2007, Bos et
al., 2008, Thaweewat et al., 2009). OpenFOAM
, however, has been tested exten
sively for ﬂuidstructure interaction (Tukovi´c & Jasak, 2007), mesh motion (Jasak,
2009) and Large Eddy Simulations (Jasak, 1996, Juretic, 2004, Jasak et al., 2007),
but not for low Reynolds number ﬂows. Therefore, this section only deals with
the validation and veriﬁcation of the OpenFOAM
ﬂow solver for test problems
relevant for low Reynolds number ﬂapping insect ﬂight.
2.10 Code validation and veriﬁcation 35
The opensource toolbox OpenFOAM
provides a framework of ﬁnite volume
based functions in order to build a speciﬁc application for solving partial dif
ferential equations. Within the context of the present research, an application
is developed, which solves for the unsteady, incompressible ﬂow using deforming
meshes. Besides the ﬂuid and ﬂow properties it is necessary to specify diﬀerent
interpolation schemes for the diﬀerent terms of the governing equations. Since this
ﬂow solver was used for low Reynolds number ﬂows with vortex convection it is
necessary to choose the correct face interpolation scheme for the convection term.
Therefore, the face interpolation for diﬀusion and source terms will be kept ﬁxed
at secondorder linear (central) interpolation. Concerning the convection term, on
the other hand, the most accurate scheme needs to be determined from linear dif
ferencing, gamma diﬀerencing, SuperBee splitting, Koren splitting and Van Leer
splitting, as was brieﬂy discussed in section 2.5.
In order to assess the accuracy of these interpolation schemes, two test cases
are investigated, the decay and convection of a Taylor vortex, described in sec
tion 2.10.1. After the selection of the proper ﬂow solver settings, the accuracy of
the code is assessed using unsteady ﬂows around static and plunging cylinders in
section 2.10.2.
2.10.1 2D vortex decay and convection
To validate the code, two vortex cases are considered, one concerning a decaying
vortex, the other deals with the vortex convection. This problem is relevant for
low Reynolds number ﬂapping ﬂight, since these ﬂows are dominated by unsteady
vortical structures. First a proper vortex deﬁnition is described, followed by the
two vortex simulations to discuss the results.
Vortex deﬁnition
When a vortex is used for code validation it is important that a well conﬁned
deﬁnition is used, i.e. the vortex should have ﬁnite and monotonic velocity and
vorticity proﬁles. The following deﬁnition is used from (Panton, 2005), but ana
logue to (Zhou & Wei, 2003):
V
θ
= t
−m
f(η), (2.27)
where V
θ
represents the radial velocity, m is a compactness coeﬃcient and η a sim
ilarity parameter. Figure 2.8 shows the velocity proﬁle for diﬀerent values of m to
show the eﬀect on the compactness of the vortex. In this ﬁgure some interesting
characteristics can be observed. With decreasing m, the curve proﬁle becomes
steeper until the asymptotic behaviour is lost at negative m. Conversely, the com
pactness of the vortex is increased with increasing m, until m = 1.5 after which a
region with counterrotation appears. When m is chosen to be 0.5 the wellknown
LambOseen vortex is the result, which is not of the desired compactness, although
maximal angular momentum (Panton, 2005) is obtained. For validation purposes
a vortex corresponding to m = 1.5, a Taylor vortex (Panton, 2005, Zhou & Wei,
36 Finite volume discretisation
Figure 2.8 Diﬀerent velocity proﬁles of a similar vortex deﬁnition. The vortex deﬁnition,
from (Panton, 2005), provides diﬀerent vortex proﬁles for varying m, ranging from −0.5 to 2 in steps
of 0.5. This m values determines the compactness of the vortex, where m = 1.5 provides the most
compact vortex.
2003), provides a better representation of vortical ﬂows like in low Reynolds num
ber vortex shedding problems (Panton, 2005). When the compactness coeﬃcient
m is ﬁxed to 1.5 the velocity proﬁle of this Taylor vortex can be derived (Panton,
2005) and is given by:
V
θ
= 2e
1
2
λe
−2λ
2
, (2.28)
where λ is a function of the similarity parameter η:
λ =
η
2
_
(2)
.
Vortex decay
As a ﬁrst 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 ﬁnite volume grid of 100 x 100 mesh cells as is shown
in ﬁgure 2.9, which shows the vorticity of the vortex at t = 0. To solve the
incompressible NavierStokes 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 secondorder backward scheme. Fur
thermore, all other terms, except for the convection term, are discretised using
a secondorder linear interpolation. To study the eﬀect 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 ﬁxed
to (100 x 100) and the timestep is varying to meet a maximal Courant number of
Co
max
= 1.0, which are both considered to be ﬁne enough. The Reynolds number
2.10 Code validation and veriﬁcation 37
Figure 2.9 Initial solution of a Taylor vortex on an Cartesian mesh. Starting from this initial
Taylor vortex solution the ﬂow diﬀusion is solved, without the presence of a convection freestream.
The velocity proﬁles and total energy are monitored to identify the accuracy of the ﬂow solver.
is ﬁxed to Re = 100 by setting the kinematic viscosity to ν = 0.01 and the velocity
vector to u = (1.0, 0.0, 0.0). The ﬂow solver solves for 40 seconds such that the
temporal eﬀect of the diﬀerent face interpolation schemes on the shape and mag
nitude of the velocity proﬁles can be compared. Any excess or lack in numerical
diﬀusion may become visible.
Results
Figure 2.10(a) and 2.10(b) show the initial and ﬁnal velocity variations in X and
Y direction for the diﬀerent face interpolation schemes. Besides the evolution of
the velocity proﬁle the total energy is shown in ﬁgure 2.10(c), which is a measure
for the diﬀusion. The total energy is calculated as
E
tot
=
N
i=1
0.5u
i

2
,
where i represents the cell index, N the total number of cells and u
i
the velocity
in cell i.
The ﬁrst important observation, from ﬁgure 2.10, is that the total energy is
increasing for the SuperBee ﬂux limiter. This scheme clearly introduced a large
amount of negative numerical diﬀusivity which causes the Taylor vortex to grow,
which is not physical. The velocity proﬁle as well as the total energy of both
Gamma and linear interpolation schemes are similar, without a signiﬁcant amount
of diﬀusion. One major drawback of these methods is that the vortex looses sym
metry after 5 s, ﬁgure 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 diﬀerent face interpolation schemes, while (b) provides v(x). (c) shows the
decaying total energy due to numerical diﬀusion.
diﬀusive, see ﬁgure 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 ﬁgure 2.11. Similar to the vortex
decay problem, the discretisation of the face interpolation is varied and all other
discretisation terms are ﬁxed to secondorder 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 ﬂow solver solved the governing equations for 20 s, such that
the vortex was convected through the entire domain. The Reynolds number was
ﬁxed to Re = 100 by setting a kinematic viscosity to ν = 0.01 and an inlet velocity
2.10 Code validation and veriﬁcation 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 suﬃciently large grid resolution and timestep size, the eﬀect of diﬀerent face interpolation schemes
is compared.
to u = (1.0, 0.0, 0.0).
Results
First of all, ﬁgure 2.12(c) shows that the total energy in the entire domain is de
creasing with time for all schemes. Extra diﬀusion is clearly visible after t=14 s
when the vortex approaches the outlet boundary, smearing the vortex. Since
convection induces physical diﬀusion, depending on the convection velocity, the
SuperBee scheme possibly still has negative numerical diﬀusion, 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, ﬁgure 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 diﬀusion in the
SuperBee scheme (Juntasaro & Marquis, 2004), another observation can be made.
The secondorder linear and Gamma schemes lead to an overshoot of the velocity
in Y direction. This eﬀect increases with time and therefore these two methods
are not appropriate to study the vortical wake patterns in insect ﬂight. 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 ﬂows
To validate the accuracy of the ﬂow solver for unsteady and vortical ﬂow, two
cylinder example problems are deﬁned at suﬃciently low Reynolds numbers. At
Reynolds numbers less than Re = 47 the ﬂow exhibits steady behaviour (see
Williamson, 1998), which is not relevant for unsteady insect ﬂight, 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 proﬁles of a convected Taylor vortex. The velocity in Xdirection u(y)
is shown in (a) at time t=10 s. (b) illustrates the velocity proﬁle in Y direction as a function of the
Xcoordinate, v(x). (c) shows the total energy for the convected vortex.
number larger than about Re = 185, the ﬂow becomes turbulent and additional
turbulence modelling becomes necessary. Since the main objective of this research
is to solve for unsteady, vortical ﬂow around ﬂapping wings, the characteristics of
that kind of ﬂow needs to be present in the validation cases. The ﬂapping wing
simulations are performed in the laminar ﬂow regime Re = O(100), with periodic
force histories. Therefore, in the range 100 ≤ Re ≤ 200, two validation cases
were selected, one concerns the ﬂow 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 timeaveraged drag coeﬃ
cient, which is welldocumented 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 outﬂow
2.10 Code validation and veriﬁcation 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
ﬃ
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
ﬁcient. (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 coeﬃcient (◦), pressure drag coeﬃcient (•) and total drag coeﬃcient (△) 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 ﬂow solver. The ﬂow 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 eﬀects are minimal at this
domain size.
Flow around a stationary circular cylinder
The ﬁrst case, dealing with the ﬂow 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,
ﬁgure 2.13(b) shows the results from an extensive study performed by Henderson
(1995) to identify a relation between drag coeﬃcient and Reynolds number. The
42 Finite volume discretisation
Figure 2.15 Vorticity visualisation of the Von K´arm´an vortex street. The ﬂow around a
stationary cylinder shows a periodic vortex street, of which the frequency depends on the Reynolds
number. Vorticity ω = ∇×u is used to identify the vortical structures, which are clearly visible at a
Reynolds number of Re = 150.
resulting drag coeﬃcient at Re = 150 is found to be C
D
= 1.333. The drag and
lift coeﬃcient are respectively deﬁned as:
C
D
=
D
1
2
ρU
2
ref
, C
L
=
L
1
2
ρU
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 timestep 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 ﬂow behaviour, ﬁgure 2.15 shows the instantaneous vorticity
(ω = ∇×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 diﬀerent mesh resolutions (25k, 50k and 100k) are shown
in ﬁgure 2.16. These limit cycles are determined by taking the periodic part of
the force histories as shown in ﬁgure 2.19. From these periodic forces, the time
averaged drag coeﬃcient is determined and compared with literature in table 2.1.
From this table, it can be observed that the drag coeﬃcient of the coarsest case,
25k and Co
max
= 2.0, has the largest diﬀerence with literature, 5.63%. When the
grid is reﬁned and the timestep decreased, it is seen that the solution decreases
asymptotically. The drag on the ﬁnest grid with the smallest timestep is about
2.7% larger compared to the value obtained by Henderson (1995). In addition,
the calculated timeaveraged 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 timeaveraged drag coef
ﬁcient. The diﬀerences of the mean Strouhal number with literature ranges from
2.10 Code validation and veriﬁcation 43
Drag coeﬃcient []
L
i
f
t
c
o
e
ﬃ
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 coeﬃcient []
L
i
f
t
c
o
e
ﬃ
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 LiftDrag limit cycles of the ﬂow around a stationary cylinder. The periodicity
of the ﬂow 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 coeﬃcient, 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 ﬂow around a stationary cylinder. Comparison of
the timeaveraged drag coeﬃcient for diﬀerent grids, 25k, 50k and 100k, and diﬀerent timesteps
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 suﬃciently small.
From table 2.1 and 2.2, it is clear that the ﬂow solver produces results which
are suﬃciently 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 timestep size. If the ﬂow solver is
developed in a numerically consistent way, the ﬂow solution should converge to
an asymptotically value with increasing spatial and temporal resolution. In order
to illustrate if the solution converges, ﬁgure 2.17 shows the timeaveraged drag
coeﬃcient for increasing mesh resolution (for each timestep) and for decreasing
timestep (for each mesh). As can be seen, the solution decreases asymptotically,
which should be the case. The last value in these two ﬁgures, is the extrapolated
values, using Richardson’s extrapolation (Ferziger & Peric, 2002). The extrapo
44 Finite volume discretisation
Spatial resolution
D
r
a
g
c
o
e
ﬃ
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
ﬃ
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 Timeaveraged drag coeﬃcient as a function of spatial and temporal resolution
for the stationary cylinder. A ﬂow solver is numerically consistent if the ﬂow solution converges
with increasing grid resolution and decreasing timestep size. (a) shows the drag coeﬃcient with
increasing grid reﬁnement level for diﬀerent timesteps. (b) shows the drag coeﬃcient with decreasing
timestep size for the diﬀerent grid sizes. The Richardson extrapolated values are plotted at ∞.
lated value is obtained using the following expression:
φ
extrap
= φ
fine
+
φ
fine
−φ
coarse
2
p
−1
, (2.30)
where φ
extrap
is the extrapolated value, φ
fine
and φ
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 secondorder 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 secondorder 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 ﬂow around a stationary cylinder. Com
parison of the timeaveraged Strouhal number for diﬀerent grids, 25k, 50k and 100k, and diﬀerent
timesteps corresponding to a maximal Courant number, Co
max
= 2.0, 1.0, 0.5 and 0.25.
2.10 Code validation and veriﬁcation 45
Spatial resolution
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 Timeaveraged lift convergence with grid resolution and timestep size for the
stationary cylinder. (a) shows the average lift coeﬃcient amplitude with increasing grid reﬁnement
level for diﬀerent timesteps. (b) shows the timeaveraged lift coeﬃcient amplitude with decreasing
timestep for the diﬀerent grid sizes.
Time [s]
F
o
r
c
e
c
o
e
ﬃ
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 coeﬃcients for a stationary
circular cylinder case at Re = 150, the grid size is 25k and the timestep 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 coeﬃcient,
ﬁgure 2.18 shows the convergence of the lift coeﬃcient amplitude in order to prove
the consistency of the ﬂow solver. Concerning the extrapolated values of the drag
coeﬃcient, from table 2.1, it can be determined that the diﬀerences varies from
2.7% to 3.5% compared to literature with decreasing mesh resolution. For all
meshes, the diﬀerences in drag coeﬃcient 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 suﬃcient 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 ﬂow 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 freestream direction. The cylinder
motion is deﬁned as
y(t) = −A
e
sin(2πf
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 suﬃcient to investigate the moving wing capabilities
of the numerical model. Previously conducted simulations on stationary cylinder
ﬂow showed that a mesh of 50k provides a suﬃciently accurate solution, therefore
that mesh is also used for this test case. Although, a timestep corresponding to
Co
max
= 1.0 was found to be suﬃcient, the following values are used to show that
the ﬂow 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 ﬂow solver applied to ﬂapping wings, the results of
an oscillating cylinder case were compared with literature. Guilmineau & Queutey
(2002) found a drag coeﬃcient of C
D
= 1.2. Figure 2.20(a) shows the limit cy
cle results, using the present ﬂow solver, with secondorder temporal and spatial
discretisation. From this ﬁgure it is obvious that the ﬂow solution converges with
decreasing timestep. This statement is conﬁrmed if the timeaveraged drag coef
ﬁcient is plotted in ﬁgure 2.20(b), for the 50k grid. The value for all timesteps
was within 2% compared to the extrapolated value, which is considered to be
suﬃciently accurate.
2.11 Conclusions
This chapter has presented the ﬁnite volume discretisation of the incompressible
laminar NavierStokes 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 eﬃcient results, the mesh qual
ity should be high in terms of nonorthogonality and skewness, both mesh quality
measures. The diﬀerent terms of the governing equation were discretised using
secondorder schemes and diﬀerent ﬂux splitting methods were described concern
ing the face interpolation. In order to solve the ﬂow on a computational grid, four
diﬀerent types of boundary conditions were speciﬁed, ﬁxedvalue (Dirichlet), zero
gradient (Neumann), symmetry and movingwallvelocity. Using those boundary
conditions, the discretised NavierStokes 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 coeﬃcient []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 ﬂow around an oscillating cylinder. (a) shows the C
D
C
L
limit
cycles for diﬀerent timesteps on a 50k mesh. (b) shows the timeaveraged drag coeﬃcient with
decreasing timestep for the 50k mesh. The Richardson extrapolated values are plotted at ∞.
In principle, the described discretisation method and solution procedure is ap
plicable to diﬀerent commercial and noncommercial CFD solvers. The present
research used the commercial CFD solver Fluent
and the opensource CFD code
OpenFOAM
, both were brieﬂy described. Fluent
has already been tested for
various ﬂows in literature, OpenFOAM
too, but not for low Reynolds ﬂows, rel
evant for ﬂapping insect ﬂight. Therefore, this chapter presented a validation of
OpenFOAM
for a number of relevant test cases. Vortex decay and convection
were used to study the inﬂuence of the face interpolation scheme with diﬀerent
ﬂux limiters. It was found that the Van Leer ﬂux limiter provides the most accu
rate results, concerning vortex decay and convection. In addition, stationary and
transversely oscillating cylinder ﬂows were used to successfully prove spatial and
temporal convergence. It is concluded that the opensource solver OpenFOAM
provides an accurate and eﬃcient framework to investigate the ﬂow around ﬂap
ping wings at low Reynolds numbers.
CHAPTER 3
Mesh deformation techniques for
ﬂapping ﬂight
Submitted to Comput. Meth. Appl. Mech. Engrg. (January 2010).
In order to use mesh deformation techniques to investigate ﬂapping wing aerody
namics it is necessary to maintain a high mesh quality for relevant wing kinematics.
Diﬀerent mesh deformation techniques are compared in order to identify their ap
plicability for cases with ﬂapping wings. The main diﬃculty is to maintain high
mesh quality when the wing exhibits large translations and rotations. In addition
to existing mesh motion methods, based on solving the Laplace and solid body
rotation stress equations, a mesh deformation routine based on the interpolation
of radial basis functions is introduced.
The radial basis function method can be used with diﬀerent basis functions
with global or compact support. A globally supported basis function results in
the highest average mesh quality, but is computationally more expensive, com
pared to a function with compact support. A test case, concerning a moving
twodimensional block, is used to show that the radial basis function method pro
vides superior mesh quality compared to the Laplace mesh motion solver. The
mesh quality, based on skewness and nonorthogonality, is found to be highest
when the thin plate spline is used as a basis function.
Additionally, it is shown that this method can be used for mesh deformation
for threedimensional ﬂapping wings and can handle ﬂexing boundaries. In order
to increase the eﬃciency of this method, two techniques are applied, based on
boundary coarsening and smoothing of the radial basis function.
50 Mesh deformation techniques for ﬂapping ﬂight
3.1 Introduction
In order to solve the governing equations, using the discretisation methods from
chapter 2, a computational mesh is necessary. Using prescribed initial and bound
ary conditions, the equations are iteratively solved on the computational domain.
The boundaries of the computational domain can be either ﬁxed or deforming. In
engineering, there are numerous computational ﬂuid dynamics (CFD) problems
in which the ﬂow solution involves geometrically deforming boundaries. Exam
ples of such interaction problems are ﬂuidstructureinteraction cases like blood
ﬂow through arteries or deforming ﬂags. A simpliﬁed type of interaction is that
which concerns a oneway coupling, where the ﬂow is being inﬂuenced by a chang
ing boundary shape. This may be caused by imposed external eﬀects, like the
rigid body motion, e.g. a ﬂapping wing, or a prescribed body deformation, if this
happens to be known beforehand.
If the shape of the domain boundary is timevarying, it is important that the
internal mesh preserves its validity (no negative cell volumes) and quality (cell
orthogonality and skewness). In order to deal with moving objects, it is possible
to solve this mathematical case by changing the boundary conditions as if the
boundary was deforming or to deform the complete mesh. The ﬁrst method is
called the immersed boundary method (Peskin, 2002), which deﬁnes a moving
boundary on a stationary Cartesian background mesh. The disadvantages of the
immersed boundary method are the diﬃculty to capture the boundary layer and
to meet the requirements for mass and momentum conservation. An appropriate
implementation of this method is not trivial. The second method to deal with a
deforming boundary, the one used in the current thesis, is the use of a mesh motion
solver which moves the internal mesh points. Current CFD solvers incorporate
diﬀerent mesh motion techniques in order to change the location of the internal
mesh points according to the varying domain shape. Preservation of high mesh
quality is necessary to solve the ﬂow in an accurate and eﬃcient way. When using
a mesh motion solver, the computational mesh points are moved in order to keep
track of the changing location of boundary points. In order to assess the quality
of a mesh motion solver, three diﬀerent aspects need to be formulated, quality,
eﬃciency and robustness.
The quality of the resulting mesh is deﬁned by the nonorthogonality and skew
ness of the ﬁnite volume cells. Eﬃciency is a measure of the used computation
time to calculate the displacements of the mesh points at the new timestep. Addi
tionally, robustness is used to identify if a method is userfriendly. A robust mesh
motion solver is deﬁned such that it needs little to no userinput. However, current
mesh motion techniques are not fully suitable to cope with the mesh deformation
around an object which moves with a large change in rotation. Therefore, existing
methods will be compared and an improved mesh motion solver is explored and
incorporated.
In literature, several mesh deformation methods have been presented using dif
ferent approaches to calculate the motion of the computational mesh points. For
3.1 Introduction 51
structured meshes, there are eﬃcient techniques available to deform the mesh, for
example Transﬁnite Interpolation (Wang & Przekwas, 1994). They interpolated
the displacements of the boundary points along grid lines through the entire com
putational mesh to ﬁnd the displacements of all interior mesh points. Using an
additional mapping (Wang, 2000b) the mesh quality can be improved signiﬁcantly
if the boundary is subjected to signiﬁcant rotation and deformation. These meth
ods are perfectly suitable for structured but unsuitable for unstructured grids.
Since, unstructured meshes are used for complex geometries, possibly in combina
tion with mesh reﬁnement, the focus is put on mesh deformation techniques which
can be applied to unstructured meshes containing arbitrary polyhedral cells, used
in the ﬁnite volume code of OpenFOAM
(Weller et al., 1998, Jasak et al., 2004).
The most popular mesh deformation method, applicable to both structured
and unstructured meshes, is called the spring analogy (Batina, 1990) where the
pointtopoint connection of every two neighbouring mesh points is represented
by a linear spring. However, this method proved to lack robustness, especially on
arbitrarily unstructured meshes, as was observed by (Blom, 2000), high resulting
mesh quality was only achieved by specifying a problem speciﬁc spring stiﬀness.
Additionally, Farhat et al. (1998), Degand & Farhat (2002) proposed a method to
incorporate torsional springs to improve the robustness of this method.
Other mesh deformation techniques involve solving a partial diﬀerential equa
tion on the complete ﬁeld of internal mesh displacements for given boundary point
displacements. Concerning the governing partial diﬀerential equations, the Laplace
and biharmonic operators (L¨ohner & Yang, 1996, Helenbrook, 2003) are often used
in combination with a constant or variable distancebased diﬀusion coeﬃcient to
improve the mesh quality. Another choice of equations is made by Johnson &
Tezduyar (1994) who used the pseudosolid equation, which assumes static equi
librium for small deformations of a linear elastic solid (the mesh is treated as if it
was a solid). The latter method is often used in the Arbitrary Lagrangian Eule
rian formulation of ﬁnite element codes. Dwight (2004) modiﬁed this method to
incorporate rigid body rotation, which signiﬁcantly improves the mesh quality for
meshes subjected to large boundary translations and rotations.
Since these methods solve a partial diﬀerential equation on the complete ﬁeld
of internal mesh points the existing iterative solvers can be used, already available
in existing CFD codes (Jasak & Tukovi´c, 2004, Jasak, 2009). Therefore, the par
allel implementation of these methods is fairly straightforward. Depending on the
method, a variable diﬀusivity ﬁeld needs to be deﬁned, which acts as a stiﬀness
of the system of equations, this inﬂuences the eﬃciency. One major drawback
of these methods is that they all fail in maintaining high mesh quality when the
boundary points move with high rotation angles. Therefore, a new mesh defor
mation method was developed and incorporated, based on the use of radial basis
function (RBF) interpolation to obtain the mesh point displacements. In (Boer de
et al., 2007, Bos et al., 2010a) it was shown that radial basis function interpolation
could improve mesh quality considerably.
Radial basis functions (RBF) are commonly used in literature to interpolate
52 Mesh deformation techniques for ﬂapping ﬂight
scattered data, because of their good approximation properties which is discussed
by Buhmann (2000). The application of radial basis functions is very wide, they
have been used in computer graphics, geophysics, error estimation, but also in cou
pled simulations as in ﬂuidstructureinteraction. (Boer de et al., 2007) used radial
basis function interpolation to couple two nonmatching meshes at the interface of
a ﬂuidstructure interaction computation. An RBF interpolation function is used
to transfer the known boundary point displacements to the ﬂuid boundary mesh.
Since the application of RBF’s to interpolate from and to the boundary mesh was
very accurate and eﬃcient, the idea was born to interpolate the boundary mesh
to all computational mesh points. A preliminary study was performed by Boer de
et al. (2007). Previously, radial basis functions were only applied to mesh motion
concerning the boundaries in multiblock meshes (Potsdam & Guruswamy, 2001).
They noted that applying this method to all mesh points would be too compu
tationally expensive. Since mesh deformation using RBF interpolation results in
high quality meshes even with large body rotation angles, two techniques are im
plemented to improve its eﬃciency. Only recent studies (Jakobsson & Amoignon,
2007, Rendall & Allen, 2008c,b) have been carried out to improve the eﬃciency of
mesh motion based on radial basis function interpolation.
In this chapter two existing mesh deformation techniques are compared, based on
the Laplace equation with variable diﬀusivity and a modiﬁed pseudosolid equa
tion, with radial basis function mesh motion. Both Laplace and pseudosolid
mesh motion techniques are commonly used within the OpenFOAM
community.
These diﬀerent mesh motion methods are described in section 3.2. In order to
assess the mesh quality, section 3.3 discusses two diﬀerent criteria, based on non
orthogonality and skewness, described in chapter 2. The resulting mesh quality
of the diﬀerent mesh motion methods is studied using a twodimensional test case
of a block which translates and rotates. The mesh quality is investigated using
a visualisation and histograms of the skewness and nonorthogonality criterion.
This is the subject of section 3.4. In addition to the simpliﬁed moving block, two
more relevant test cases were considered, one using a threedimensional ﬂapping
wing and the other involves a twodimensional ﬂexing airfoil. Since the radial ba
sis function mesh motion method is computationally expensive, section 3.5 deals
with two techniques to increase its eﬃciency. Finally, the conclusions are drawn
in section 3.6.
3.2 Diﬀerent 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 (ﬂuid) points. The displace
ment of the boundary points can be considered to be given, either it is externally
deﬁned, i.e. a prescribed rigid body motion, or it is part of the solution, which is
3.2 Diﬀerent mesh deformation techniques 53
the case in ﬂuidstructure 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 inﬂuences 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
diﬀerent methods, as will be shown in the next section.
3.2.1 Laplace equation with variable diﬀusivity
One can think of a deforming computational domain as if it was a solid body under
going internal stresses given by the PiolaKirchhoﬀ stressstrain equation (Baruh,
1999). That equation is nonlinear 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 eﬃciently.
When the mesh motion is governed by the Laplace equation, the given bound
ary point motion may be arbitrary and nonuniform. 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 speciﬁ
cation of a variable diﬀusivity. This leads to the following deﬁnition of the Laplace
equation:
∇
•
(γ∇x) = 0,
where x is the displacement ﬁeld and γ the diﬀusion coeﬃcient, which decreases
with the radius r from the deforming boundary as follows:
γ(r) =
1
r
m
. (3.1)
The resulting mesh quality strongly depends on the chosen γ(r) function, which
depends on the distance from the moving boundary. This variable diﬀusion coeﬃ
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 γ(r) function like
equation (3.1). In addition to the freedom of choosing a diﬀusion function, it is also
possible to deﬁne γ(r) for every internal mesh cell for all timesteps independently.
This, however, appears to be very problem dependent and thus optimisation of
γ(r) seems not cost eﬀective. To maintain robustness, in the current work we use a
quadratically, m = 2, decreasing diﬀusion coeﬃcient, which was found to provide
eﬃcient and a smooth mesh motion (Jasak & Tukovi´c, 2004). Additionally, one
could also have used an exponentially decreasing diﬀusion coeﬃcient or a diﬀusion
coeﬃcient related to the mesh deformation energy.
54 Mesh deformation techniques for ﬂapping ﬂight
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
∇
•
σ = f , (3.2)
where σ is the stress tensor and f the acting force vector. The stress tensor σ is
given in terms of the strain, which is given by the following constitutive relation:
σ = λtr (ǫ)I + 2µǫ, (3.3)
where tr is the trace and λ and µ are Lam´e constants (Baruh, 1999), which are a
property of the elastic material. The constants can be related to Young’s modulus,
E, as
λ =
νE
(1 +ν)(1 −2ν)
, µ =
E
2(1 +ν)
,
where ν is Poisson’s ratio, meaning the material contraction ratio as it stretches.
The following equation:
ǫ =
1
2
(∇x +∇x
T
), (3.4)
deﬁnes 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:
ǫ =
1
2
(∇x +∇x
T
+∇x
T
· ∇x). (3.5)
Combining equations (3.5), (3.3) and (3.2), together with λ = −E and µ = E, the
following solid body rotation stress equation is obtained:
∇
•
(γ∇x) +∇(γ(∇x −∇x
T
)) −λtr(∇x) = 0, (3.6)
where γ is a similar diﬀusion coeﬃcient 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 preconditioned
Conjugate Gradient (PCG) method. However, it is also possible to explicitly de
ﬁne the point motion using interpolation techniques, like the transﬁnite interpo
lation (Wang & Przekwas, 1994) usually applied to the points of multiblocks.
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 ﬂapping wing simulations.
3.2 Diﬀerent mesh deformation techniques 55
3.2.3 Radial basis function interpolation
In the current work we use radial basis function interpolation to ﬁnd the dis
placements of the internal ﬂuid 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
γ
j
φ(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 φ 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 φ (Boer de et al., 2007).
A unique interpolant is given if the basis function is a conditionally positive deﬁnite
function. If the basis functions are conditionally positive deﬁnite 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 deﬁned by
the coeﬃcients γ
j
which can be deﬁned 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
γ
j
p(x
b
j
) = 0,
which holds for all polynomials p with a degree less or equal than that of polynomial
q, the γ
j
values can be determined (Boer de et al., 2007).
The values for the coeﬃcients γ
j
and the linear polynomial can be obtained by
solving the system:
_
∆x
b
0
_
=
_
Φ
bb
Q
b
Q
T
b
0
_ _
γ
β
_
, (3.8)
where γ is containing all coeﬃcients γ
j
, β the four coeﬃcients of the linear poly
nomial q, Φ
bb
an n
b
× n
b
matrix contains the evaluation of the basis function
φ
b
i
b
j
= φ(x
b
i
− x
b
j
) and can be seen as a connectivity matrix connecting all
boundary points with all internal ﬂuid 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 diﬃcult 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 eﬃcient way are discussed in section 3.5.
56 Mesh deformation techniques for ﬂapping ﬂight
When the coeﬃcients in γ and β are obtained they are used to calculate the
values for the displacements of all internal ﬂuid 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 coeﬃcients γ and β, which are then used for step two, the evaluation
using equation (3.9).
In contrast to the Laplace and SBR Stress methods, no partial diﬀerential
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 diﬀusion coeﬃcient
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:
φ(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 (twodimensional problem) or a sphere
(threedimensional problem) with radius r around a centre x
j
are inﬂuenced 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 eﬃciently using
common iterative techniques.
In table 3.1 various radial basis functions with compact support are shown
using the scaled variable ξ = x/r. The ﬁrst 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 Diﬀerent mesh deformation techniques 57
Ref. nr. RBF Name f(ξ)
1 CP C
0
(1 −ξ)
2
2 CP C
2
(1 −ξ)
4
(4ξ + 1)
3 CP C
4
(1 −ξ)
6
(
35
3
ξ
2
+ 6ξ + 1)
4 CP C
6
(1 −ξ)
8
(32ξ
3
+ 25ξ
2
+ 8ξ + 1)
5 CTPS C
0
(1 −ξ)
5
6 CTPS C
1
1 +
80
3
ξ
2
−40ξ
3
+ 15ξ
4
−
8
3
ξ
5
+ 20ξ
2
log(ξ)
7 CTPS C
2
a
1 −30ξ
2
−10ξ
3
+ 45ξ
4
−6ξ
5
−60ξ
3
log(ξ)
8 CTPS C
2
b
1 −20ξ
2
+ 80ξ
3
−45ξ
4
−16ξ
5
+ 60ξ
4
log(ξ)
Table 3.1 Radial basis functions with compact support. Radial basis functions with compact
support are nonzero within the range of the support radius r. Note that ξ = 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 ﬂuidstructure 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 Biharmonics MQB
√
a
2
+x
2
11 Inverse Multiquadratic Biharmonics IMQB
_
1
a
2
+x
2
12 Quadric Biharmonics QB 1 +x
2
13 Inverse Quadric Biharmonics 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 ﬂapping ﬂight
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 ﬂat sheetlike function,
whereas a small value of a gives a narrow conelike function. The value of a is
typically chosen in the range 10
−5
− 10
−3
. More information about RBF’s 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 ﬁnally 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 nonorthogonality, are discussed to compare the mesh quality
for the diﬀerent mesh motion solvers.
Absolute and relative radial basis function interpolation
In principle there are two diﬀerent 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 coeﬃcient arrays γ and β are calculated and used to calculate the internal
point displacements at all timesteps. This method is very eﬃcient 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 coeﬃcients are
not deﬁned with respect to the previous timestep. 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 timestep and the motion is deﬁned with
respect to the previous timestep. 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 diﬀerent 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 nonorthogonality deﬁnition were introduced. In
order to compare the quality of the diﬀerent 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 nonorthogonality are written
to scalar ﬁelds such that they can be used for postprocessing.
The test cases used to compare the mesh quality of the diﬀerent 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 nonorthogonality
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 nonorthogonality, 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
non−ortho
≤ 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 eﬃcient the computation will be. In the next section, both the average and
minimal value of the skewness and nonorthogonality mesh quality measures are
used to compare the mesh motion solvers.
3.4 Comparison of mesh motion solvers
In section 3.2, three diﬀerent 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
diﬀerences in mesh quality obtained with the diﬀerent mesh motion solvers. Addi
tionally, the eﬀect of diﬀerent radial basis functions is discussed. The ﬁrst simple
case is a twodimensional block which performs a combination of translation and
rotation. The initial computational mesh is shown in ﬁgure 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 ﬁgure. Mesh motion simulations are performed using
the diﬀerent 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 threedimensional ﬂapping wing, followed by an example of a twodimensional
ﬂexible moving boundary in section 3.4.3.
3.4.1 Translation and rotation of a twodimensional block
The ﬁrst test case considers a combined motion of translation and rotation to com
pare the mesh motion solvers under these conditions. The twodimensional block
is initially centred and translates 2.5D in both Xand Y direction. In addition,
the block is rotated around its translating centre with 57.3
◦
(1.0 rad). The outer
boundary points are kept ﬁxed, such that the inﬂuence of the moving boundary
points on all internal points could be studied independently. The resulting mesh
60 Mesh deformation techniques for ﬂapping ﬂight
Figure 3.1 Initial mesh around a moving block. The initial hexahedral mesh with optimal
quality around a moving block. The domain size is (25D x 25D) around a block with size (5D x 1D),
at every unit spacing, a grid point is places such that an optimal hexahedral mesh is obtained.
quality, after the combined translation and rotation, is assessed using the ﬁelds of
skewness and nonorthogonality. Using those ﬁelds, the maximal and average val
ues are obtained as well as a complete visualisation of those mesh quality measures,
combined with mesh quality histograms.
Before proceeding to the results, some special settings, applicable to this test
case, need to be described. First, when using the Laplace and SBR Stress mesh
motion solver it is necessary to specify the diﬀusivity coeﬃcient which decreases
proportionally to the distance from the moving boundary points. Following Jasak
& Tukovi´c (2004) and Jasak (2009), a quadratically decreasing diﬀusivity coef
ﬁcient, i.e. decreasing from the moving boundary, was chosen. Secondly, the
boundary conditions need to be set for the motion solver. In both cases, Laplace
and SBR Stress motion solvers, the boundary conditions on all outer boundary
points are set to Dirichlet type with value 0, so the outer boundary points are kept
ﬁxed. Finally, the newly implemented RBF mesh motion solver is used with ﬁve
diﬀerent functions, CP C
2
, CTPS C
1
, CTPS C
2
a
, CTPS C
2
b
and TPS, based on
the assessment performed by Boer de et al. (2007). Figure 3.2 shows the cell non
orthogonality at maximal mesh deformation for the Laplace and the SBR Stress
motion solvers. As seen in the ﬁgure, with these standard methods the mesh qual
ity near the moving boundary is low, especially near the leading and trailing edges.
The mesh motion method which solves the Laplace equation with a quadratically
decreasing diﬀusion coeﬃcient, is simply not robust enough to obtain high mesh
quality when the boundary rotates. As can be seen in ﬁgure 3.2(a), the mesh
deformation is largest near the boundary, which is not desirable. In cases where
large rotation angles occur, it is best to apply a mesh motion solver, which leads
to the motion of all internal mesh points, coping with the boundary deformation.
The gain in mesh quality by using the SBR Stress method is marginal. From
3.4 Comparison of mesh motion solvers 61
(a) Laplace
(b) SBR Stress
Figure 3.2 Cell nonorthogonality of Laplace and SBR Stress mesh motion solvers. The
cell nonorthogonality is compared of the Laplace (a) and the SBR Stress (b) motion solvers, lower
(blue) is better. The nonorthogonality is visualised for a twodimensional block with a combined
motion of translation and rotation. The boundary points translate over a distance of 2.5D in both X
and Y direction and rotate 57.3
◦
(1.0 rad) around the translating centre.
ﬁgure 3.2(b) it can be seen that the cells with a high nonorthogonality occur at a
distance from the moving boundary. Still, the mesh quality near the body surface,
especially near the leading and trailing edges, needs improvement. Improvement
can be obtained by specifying a constraint to an inner mesh region, such that it
moves according to the body motion.
Figure 3.3(a) shows the resulted mesh obtained using RBF interpolation using
a thin plate spline function. When comparing the cell nonorthogonality with
ﬁgure 3.2(a) and 3.2(b) it may be seen that most of the cell deformation, with
the RBF mesh motion, occurs in the outer regions of the mesh and all cells are
dealing with the boundary displacement. Additionally, ﬁgure 3.3(b) shows the
resulting mesh, which is obtained using the relative implementation of the RBF
mesh motion method. The rotation of the boundary is very large, 180
◦
, 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 diﬀerent mesh motion solvers, including diﬀerent RBF’s. The max
imal and averaged values of both skewness and nonorthogonality are compared.
The mesh quality calculated by the Laplace method is low, since the maximal
skewness and maximal nonorthogonality 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 RBF’s 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 eﬀect of the support
62 Mesh deformation techniques for ﬂapping ﬂight
(a) RBF absolute (b) RBF relative
Figure 3.3 Cell nonorthogonality of relative and absolute RBF mesh motion solvers.
The cell nonorthogonality 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 twodimensional 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
◦
(1.0 rad) for (a) and 180
◦
(3.14 rad) for (b).
radius, it is set to r = 75 which is about 3 times the domain size such that the
results in table 3.3 are independent of the support radius for r > 75. The RBF
method provides high mesh quality, maximal and averaged, for both C
2
and TPS
compared to the other functions. The C
2
and TPS RBF’s result in an maximal
skewness of respectively −32% and −45% compared to the Laplace motion solver,
while the diﬀerence of the other RBF’s is only about 10%. Similar results are
shown in the table for the average skewness and maximal orthogonality. Con
cerning the average orthogonality, the C
2
RBF is outperformed by the TPS RBF,
which is the only mesh motion solver resulting in a lower value compared to the
Laplace method. Overall, the basis function thin plate spline provides the high
est mesh quality for both skewness and nonorthogonality, such that this globally
supported function was used for the current investigations.
Finally, in addition to the nonorthogonality visualisations, ﬁgure 3.4 and 3.5
show the histograms of cell nonorthogonality and skewness for the Laplace and
RBF mesh motion solver, which are considered to result in the worst and best
mesh quality, respectively. For both nonorthogonality and skewness, the RBF
mesh motion results in a smooth proﬁle, which emphasises the fact that all internal
cells are coping with the mesh motion.
3.4.2 Flapping of a threedimensional wing
It was shown that high mesh quality was obtained for a simpliﬁed twodimensional
test case, by using radial basis function interpolation. Especially, the globally sup
3.4 Comparison of mesh motion solvers 63
Method (f
s
)
max
(f
s
)
ave
(f
o
)
max
(f
o
)
ave
Laplace 0.95 (baseline) 0.09 () 72.1 () 20.1 ()
SBR Stress 0.96 (+0.7%) 0.08 (4%) 75.1 (+4%) 21.3 (+6%)
CP C
2
0.65 (32%) 0.065 (28%) 59.6 (17%) 23.4 (+16%)
CTPS C
1
0.86 (10%) 0.105 (+17%) 74.5 (+3%) 31.9 (+59%)
CTPS C
2
a
0.81 (15%) 0.103 (+14%) 73.8 (+2%) 31.4 (+56%)
CTPS C
2
b
0.88 (7%) 0.11 (+22%) 76.0 (+5%) 32.8 (+63%)
TPS 0.52 (45%) 0.051 (41%) 52.5 (27%) 19.0 (6%)
Table 3.3 Comparison of mesh quality for diﬀerent mesh motion solvers. The mean and
maximal values of the skewness f
s
and nonorthogonality f
o
are compared at maximal displacement
and rotation of the twodimensional rectangular block. Results are shown for the Laplace, SBR Stress
and RBF mesh motion solver, the latter using diﬀerent RBF’s.
Nonorthogonality
N
u
m
b
e
r
o
f
c
e
l
l
s
0 10 20 30 40 50 60 70 80 90
0
10
20
30
40
50
60
70
80
90
(a) Laplace
Nonorthogonality
N
u
m
b
e
r
o
f
c
e
l
l
s
0 10 20 30 40 50 60 70 80 90
0
10
20
30
40
50
60
70
80
90
(b) RBF
Figure 3.4 Cell nonorthogonality histograms for Laplace and RBF mesh motion solvers.
Mesh quality histograms show the variation in nonorthogonality for Laplace (a) and RBF mesh motion
solvers (b) using the TPS.
ported thin plate spline (TPS) provided high quality and robust mesh deformation.
To show the threedimensional capabilities of the RBF mesh motion solver, the
mesh around a ﬂapping wing is shown in ﬁgure 3.6 at midstroke. Figure 3.6(a)
shows the nonorthogonality during the downstroke, while ﬁgure 3.6(b) presents
the mesh quality halfway of the upstroke. The TPS was used as radial basis
function, without any user input, since it has global support. From the ﬁgure it is
clear that a large part of the near wake is deformed in order to deal with the three
dimensional wing motion. Concerning the RBF mesh motion solver, the cells close
to the wing take a larger part of the deformation compared to the Laplace method.
This was already illustrated for the moving twodimensional block in ﬁgures 3.4
and 3.5, showing the histograms of the nonorthogonality and skewness, respec
64 Mesh deformation techniques for ﬂapping ﬂight
Skewness
N
u
m
b
e
r
o
f
c
e
l
l
s
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
50
100
150
200
250
300
350
400
450
500
(a) Laplace
Skewness
N
u
m
b
e
r
o
f
c
e
l
l
s
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
50
100
150
200
250
300
350
400
450
500
(b) RBF
Figure 3.5 Cell skewness histograms for Laplace and RBF mesh motion solvers. Mesh
quality histograms show the variation in skewness for Laplace (a) and RBF mesh motion solvers (b)
using the TPS.
tively. The Laplace method results in a small number of cells with large values
for both nonorthogonality and skewness, while the RBF mesh motion technique
leads a smoother decline of both quality measures in the histograms.
When comparing ﬁgure 3.6(a) and 3.6(b) it is seen that the mesh during both
the upstroke and downstroke is symmetric. This is caused by the fact that the
radial basis function interpolation determines the new internal mesh points with
respect to the initial mesh, such that the initial mesh is recovered after every
ﬂapping period.
As discussed in subsection 3.2.3, the radial basis function mesh motion con
tains a direct system solve and an evaluation to determine the displacement of all
internal mesh points. Therefore, the mesh deformation for threedimensional cases
may become very expensive, such that it is necessary to implement techniques to
improve its eﬃciency. These techniques are described in section 3.5. But ﬁrst, the
RBF mesh motion solver is tested by employing the ﬂexing of a twodimensional
moving boundary.
3.4.3 Flexing of a twodimensional block
Within the context of the present research, performing a ﬂuidstructureinteraction
simulation is too computationally demanding for a full threedimensional ﬂapping
wing. Therefore, the eﬀects of wing ﬂexibility was incorporated by deﬁning the
wing ﬂexing using harmonic functions, to mimic realistic insect wing deforma
tion (Shyy et al., 2008a). In order to show that the RBF mesh motion method
is able to deal with a ﬂexing boundary, the model problem of a moving block is
used. The motion of the twodimensional block can be decomposed into transla
tion, rotation and ﬂexing, all deﬁned with respect to the initial conﬁguration. The
3.4 Comparison of mesh motion solvers 65
(a) t = 0.25T (b) t = 0.75T
Figure 3.6 Cell nonorthogonality for a threedimensional wing. Cell nonorthogonality of
a mesh around a threedimensional ﬂapping model wing, obtained with the RBF mesh motion solver
with the globally supported thin plate spline (TPS), lower (blue) is better. (a) shows the mesh quality
at t = 0.25T and (b) at t = 0.75T, respectively during the downstroke and upstroke.
translation and rotation are deﬁned by:
x
t
= A
t
· sin(2πft),
α = A
α
· sin(2πft),
where, A
t
and A
α
represent the translation and rotation amplitude vectors, f the
frequency and t the time. The ﬂexing of the boundary is deﬁned by combining
two harmonic functions like:
x
f
= A
f
· cos(2π
x
•
d
c
2 · c
f
) · sin(2πft)
•
d
f
,
where A
f
is the ﬂexing amplitude vector, d
c
is the direction vector inplane of the
ﬂexing surface, c
f
the length of the ﬂexing surface and d
f
represents the direction
vector of the ﬂexing. In this model problem of a moving twodimensional block, the
amplitudes of both translation and rotation were ﬁxed to A
t
= (2.5, 2.5, 0.0) and
A
α
= (0.0, 0.0, 1.0), respectively. The ﬂexing was deﬁned such that the main ﬂex
ing direction is perpendicular to the ﬂexible boundary surface, d
f
= (0.0, 1.0, 0.0)
with a ﬂexing amplitude of A
f
= 0.5 which is about 10% of the ﬂexible 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 nonorthogonality can be observed within a region of 1−2
block lengths, which is mainly caused by the ﬁxed 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 ﬂapping ﬂight
(a) (b)
Figure 3.7 Deforming mesh around a ﬂexible block using RBF mesh motion. A ﬂexible
rectangular block which is translating, rotating and ﬂexing using the RBF mesh motion solver with
the globally supported thin plate spline (TPS). The ﬂexing is deﬁned by simple harmonic functions.
(a) shows the mesh deformation at t = 0.25T and (b) at t = 0.75T.
3.5 Improving computational eﬃciency
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 ﬁnd the coeﬃcient vectors γ and β. The
upper diagonal block matrix Φ
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 coeﬃcient vectors γ and β. This evaluation
leads to a computational cost of order O(N
b
N
in
).
For large two and threedimensional 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 threedimensional 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 threedimensional 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 eﬃciency 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 RBF’s. Computa
tional costs for solving the system and evaluation the RBF’s on all internal mesh points. Illustration of
a twodimensional and threedimensional uniform square shaped Cartesian mesh with an equal number
of boundary nodes N
b
in x and ydirection.
shows that the computational costs for both a direct solve and evaluation scales
with N
3
b
for the twodimensional case. Concerning the threedimensional 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 threedimensional case,
like a ﬂapping 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 diﬀerent
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 ξ points, where ξ is
the coarsening factor. More advanced coarsening techniques, based on greedy
algorithms are applied by Rendall & Allen (2008a).
Secondly, the eﬃciency of the RBF method is improved, by the notion that
all outer boundary points are ﬁxed 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
deﬁned (Jakobsson & Amoignon, 2007) as
ψ(¯ 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 i−th 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 ﬂapping ﬂight
between those radii this smoothing function is decreasing from 1.0 to 0.0. When
the RBF evaluation function (3.9) is multiplied by φ(¯ x) the corrected RBF eval
uation is obtained by:
∆¯ x
in
j
= s(x
in
j
)
•
ψ(¯ 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 ﬁxed 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
diﬀerent 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 diﬀerential 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 nonlinear behaviour of the ﬁnal curve is caused by the choice
of the coarsening ratio to select the moving boundary points. While keeping the
coarsening ratio ﬁxed, the mesh resolution is increased, the number of boundary
points used in the system solving is still growing nonlinearly. 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 timestep, concerning a three
dimensional ﬂapping wing simulation (6 ﬂapping periods). These threedimension
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 superlinear curves for both solving the ﬂow
equations and the RBF mesh motion. Validation showed that a mesh resolution
of 800k provided accurate results for ﬂapping wing aerodynamics. Additionally,
ﬁgure 3.8(b) shows that the computing time used for RBF mesh motion is less
than 10% of the computing time for the complete timestep. This is considered to
be very eﬃcient, it must be noted that mesh coarsening and a smoothing function
are applied for acceleration.
3.5 Improving computational eﬃciency 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) Twodimensional
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) Threedimensional
Figure 3.8 Computing times for diﬀerent 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
timestep of a threedimensional ﬂapping wing simulation. The times are subdivided by solving the
RBF mesh motion and the ﬂow 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 diﬀerent 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 suﬃciently high quality.
Diﬀerent preconditioning 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 eﬃcient to solve using existing iterative solver techniques.
Furthermore, a mesh motion problem leads in general to an illconditioned
system, which is diﬃcult 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 diﬀerences between points in Φ
bb
. Despite the preconditioning, the
illconditioned system cannot be eﬃciently 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 diﬀerent way to improve the eﬃciency 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 signiﬁcant gain in com
puting time. Concerning the coarsening, complex greedy algorithms (Rendall &
70 Mesh deformation techniques for ﬂapping ﬂight
Allen, 2008a) may be applied to select the necessary boundary points such that
the eﬃciency is increased even further. Another method to increase the eﬃciency
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,
reordering 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 matrixvector multiplication. The
major diﬃculty, 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
eﬃcients, α
i
and β
i
, which are distributed over all partitions. Currently, this is
being implemented in OpenFOAM
.
3.6 Conclusions
In this chapter two diﬀerent mesh motion techniques were described which are
commonly used within the code of OpenFOAM
, both based on solving a partial
diﬀerential equation. The ﬁrst method solves the Laplace equation with a variable
diﬀusion coeﬃcient, which is used to control the ﬁnal mesh quality. Secondly, the
linear stress equation was modiﬁed to include rigid body rotations in order to cope
with the severe mesh deformation present in ﬂapping wing simulations. As with
the Laplace equation, the solid body rotation stress (SBR Stress) mesh motion
uses the diﬀusivity, acting as a stiﬀness, to inﬂuence the quality of the mesh. The
diﬀusivity, in both cases, is deﬁned to decrease quadratically with the distance
from the moving boundary.
Besides solving a partial diﬀerential equation the motion of mesh points can
be deﬁned using interpolation techniques. A new mesh motion solver is incor
porated in OpenFOAM
, which uses the interpolation of radial basis functions
(RBF). For given boundary point displacements the internal mesh displacements
are obtained by solving a system of equations to obtain an array of interpolation
coeﬃcients. Using those coeﬃcients, the internal point displacements are obtained
by evaluating the radial basis functions.
This new mesh motion technique does not need any information about the mesh
connectivity and can be applied to arbitrary unstructured meshes containing poly
hedral cells, the way OpenFOAM
deals with the ﬁnite volume implementation.
The three mesh motion solvers are tested using a case of a twodimensional rect
angular block which moves through a Cartesian mesh. The cell nonorthogonality
and skewness are compared. Additionally, diﬀerent radial basis functions, con
cerning the RBF mesh motion, are compared. The RBF mesh quality provides
superior mesh quality over the Laplace and SBR Stress mesh motion solvers. Es
pecially when using the thin plate spline (TPS) or the continuous polynomial C
2
3.6 Conclusions 71
as radial basis functions, the mesh quality is high in terms of low skewness and
nonorthogonality. The TPS has global support, whereas the C
2
basis function
has compact support. The RBF mesh motion was successfully tested on simple
test problems and for a threedimensional ﬂapping wing with the possibility to
incorporate a ﬂexing moving boundary.
Since the RBF mesh motion technique encounters a dense system of equations,
diﬀerent methods are implemented to increase its eﬃciency. First of all, a subset
of the moving boundary points was selected, because not all points are necessary
when the body performs a rigid body motion. So a coarsening algorithm selects
those control points. Secondly, a smoothing function is used to decrease the RBF
contribution to zero at the outer domain boundaries. Therefore, it is justiﬁed to
neglect the outer (ﬁxed) boundary points, which reduces the system of equations
considerably. After this elaborate discussion it is concluded that, concerning the
threedimensional wing simulations, the globally supported TPS should be used
in combination with the coarsening and smoothing techniques to increase the ef
ﬁciency of the RBF mesh motion method.
CHAPTER 4
Physical and numerical modelling
of ﬂapping foils and wings
In this chapter, the physical and numerical modelling of ﬂapping wing and foils is
described. The relevant dimensionless numbers (Strouhal, Reynolds and Rossby
numbers) are identiﬁed after writing the NavierStokes equations in a rotating
reference frame. To systematically study the aerodynamics around ﬂapping wings,
a model planform and kinematic model is deﬁned. The kinematic model, which
describes the wing motion, consists of a rigid body motion appended by a ﬂexing,
representing occasional wing deformation. Both geometry and wing kinematics are
dynamically scaled in order to design a sound framework for comparison, using
the radius of gyration. Additionally, the force coeﬃcients are used in conjuncture
with the lifttodrag ratio to assess the ﬂapping wing performance.
4.1 Introduction
This chapter deals with the physical and numerical modelling of ﬂapping wings, in
hovering as well as in forward ﬂight conditions. Animals that ﬂy or swim, which are
equivalent from the ﬂuiddynamic perspective (Triantafyllou et al., 1993, Taylor et
al., 2003) at certain scales, undergo signiﬁcant interactions with the environmental
ﬂuid in which they move. Therefore, for the mathematical analysis of swimming
or ﬂying it is important to formulate the governing equations and accompanying
boundary conditions in an appropriate form. These equations are used to deduce
the dimensionless numbers relevant for insect ﬂight.
74 Physical and numerical modelling of ﬂapping foils and wings
The Reynolds number
In order to improve our understanding of biological ﬂows, like the ﬂows around
ﬂapping wings or ﬁns, it is of importance to make extensive use of dimensionless
numbers, like the Reynolds (Re) and Strouhal (St) number in particular (e.g. Pan
ton, 2005, White, 1991). The Reynolds number is deﬁned as the ratio between
the inertial and viscous forces present in a ﬂuid. It is a property of the ﬂow, such
that it identiﬁes what kind of propulsive mechanism applies to the ﬂapping wing.
For example, when a ﬂapping wing operates at a very low Reynolds number, i.e.
Re = O(100), the forces in the ﬂow are dominated by the viscous term, compared
to the inertial component, such that viscous phenomena, like shear layers and
vortex generation, will be more pronounced. Additionally, the Reynolds number
determines whether the ﬂow behaves turbulent or laminar, deﬁning implicitly the
complexity of the mathematical model needed to solve the problem. A diﬀerent
approach to dimensionless numbers is to deﬁne them as the ratio of time or length
scales, instead of the ratio between two distinct forces (Tennekes & Lumley, 1972,
Lentink & Gerritsma, 2003, Bos et al., 2008). In that case the Reynolds number
can be deﬁned as the ratio between the convection time over the diﬀusion time,
see chapter 2, which is easier to understand and to apply to a ﬂow problem, such
as the Von K´ arm´ an vortex street behind a bluﬀ body.
The Strouhal number
Besides the Reynolds number, the Strouhal number plays an important role in
ﬂapping wing aerodynamics as well. The typical deﬁnition of the Strouhal num
ber is the ﬂapping frequency times ﬂapping amplitude divided by a reference ﬂow
velocity. The ﬁrst use of the Strouhal number was in the context of the natural
vortex shedding behind a stationary cylinder in a uniform ﬂow. Williamson (1988)
found a universal relation between the Reynolds and Strouhal number based on the
observed vortexshedding frequency in the laminar ﬂow regime. For ﬂapping wings
or oscillating bodies, the Strouhal number can be deﬁned based on the imposed
oscillation frequency and amplitude. For moving bodies and especially ﬂapping
wings, the Strouhal number can be very useful. For example, in forward ﬂapping
ﬂight, the Strouhal number is proportional to the maximum value of the induced
angle of attack, provided that the wing ﬂaps in a stroke plane perpendicular to the
forward velocity (Lentink & Gerritsma, 2003, Taylor et al., 2003, Thaweewat et al.,
2009). This deﬁnition of the Strouhal number, based on the ﬂapping amplitude, is
closely related to the advance ratio, as deﬁned by Ellington (1984), J = U/4Φ
0
fR,
which is the ratio between the forward and ﬂapping distance, travelled by the wing.
Here Φ
0
is half of the total ﬂapping angular amplitude, f the ﬂapping frequency
and R represents the distance from root to tip of the wing, i.e. the single wing
span. Another deﬁnition was introduced by Dickinson (1994) and Wang (2000b)
using the average chord length as reference, St
c
= f c/U. This expression is very
similar to the reduced frequency (Shyy et al., 2008b), which is commonly used
to relate the two velocities due to either ﬂapping and forward ﬂight. In general
engineering applications, the Strouhal number is commonly used to characterise
4.2 Governing equations for ﬂapping wings 75
the vortex shedding, whereas the reduced frequency number is used in ﬂapping
wing problems. Additionally, the reciprocal of the Strouhal number is known as
the dimensionless wavelength λ
∗
= U/f c which is often used in studies concerning
forward ﬂapping ﬂight as it seems reasonably intuitive in that it corresponds to
the distance travelled over one ﬂapping period, relative to the mean chord.
Equations and other assumptions
As previously discussed, it is appropriate to use dimensionless numbers to study
the eﬀect of ﬂapping characteristics on the aerodynamic performance. In order
to perform a sound and valid comparison it is important to maintain constant
dimensionless numbers while kinematic parameters or ﬂow properties are varied
to study their inﬂuence. In accordance with Lentink & Dickinson (2009a), spe
ciﬁc attention is given to the appropriate deﬁnition of the governing dimensionless
numbers to investigate the ﬂow around ﬂapping wings.
In view of simplicity, the present study deals with a model wing which is a
simpliﬁed representation of a ﬂying insect wing operating at Reynolds numbers,
Re = 100, 500, and 1000, corresponding to the operating conditions of fruit ﬂies,
house ﬂies and bumblebees, respectively. Furthermore, only one ﬂapping wing is
considered, under hovering as well as forward ﬂight conditions, which allows that
the induced vortical ﬂow can be studied in more detail. This implies that no
interaction between two wings or with the body are included. Nevertheless, the
considered ﬂapping kinematics that result in large rotation rates put the current
numerical techniques to a signiﬁcant challenge.
In section 4.2, the governing equations are formulated for forward and hovering
ﬂapping ﬂight. Secondly, the model wing selection and the deﬁnition of the kine
matic model parameters are described in section 4.3, followed by the dynamical
scaling of ﬂapping ﬂight in section 4.4. As a prelude to the numerical solution
of the governing equations, the mesh generation in combination with the bound
ary conditions is brieﬂy dealt with in 4.5. Section 4.6 describes the force and
performance deﬁnitions, followed by the conclusions of this chapter in section 4.7.
4.2 Governing equations for ﬂapping wings
Concerning ﬂapping ﬂight in nature, like the operation of insects and ﬁsh, the
ﬂow can be considered to be incompressible since the Mach number (a measure
for compressibility) is typically Ma = U/a = O(10
−3
) (Brodsky, 1994), where
U [m/s] is the reference velocity and a [m/s] the speed of sound. In section 2.2 the
incompressible NavierStokes equations were deﬁned by equation (2.4) and (2.5),
and restated here:
∇
•
u = 0, (4.1)
∂u
∂t
+∇
•
(uu) = −
1
ρ
∇p +ν∇
2
u, (4.2)
76 Physical and numerical modelling of ﬂapping foils and wings
These equations are derived by analysis of the forces on an inﬁnitely small ﬂuid 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 ﬂight under
hovering conditions is considered, the initial velocity ﬁeld is zero as well as the
boundary conditions at the outer domain boundary. At the boundary of the ﬂap
ping wing, a noslip condition (White, 1991) needs to be speciﬁed, which means
that the velocity, relative to the wing, has to be zero, in all directions. This is
accomplished by deﬁning the mathematical velocity on the moving boundary to
be equal to the actual wing motion, which moves according to a speciﬁc kinematic
model, derived from realistic insect data (e.g. Fry et al., 2003).
Momentum analysis in rotating reference frame
A diﬀerent 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 ﬁxed to the ﬂapping 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 ﬂow 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 ﬂapping wing
will be such that the eﬀective 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 deﬁned as (Ginsberg, 1998,
Baruh, 1999)
u
XY Z
= u
xyz
+ (u
trans
+Ω
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 ﬂight, approximately. Ω
wing
is the angular
velocity of the rotating reference frame, i.e. representing the ﬂapping 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
+Ω
wing
×r),
where u
WING
is the ﬂow velocity at the wing in the inertial reference frame. This
relation results in a noslip 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 ﬂapping 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 deﬁned as
a
ang
=
˙
Ω×r,
a
cen
= Ω×(Ω×r),
a
cor
= 2Ω×u
xyz
.
The main parameter in these three diﬀerent accelerations is the angular velocity
Ω [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 diﬀerent acceleration terms,
Ω needs to be related to the ﬂapping 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 NavierStokes equations (4.1) and (4.2)
such that the following transformed NavierStokes equations are obtained (for the
sake of simplicity, the subscripts are dropped):
Du
Dt
+ (
˙
Ω×r) + (Ω×(Ω×r)) + (2Ω×u) = −
1
ρ
∇p +ν∇
2
u. (4.3)
This transformed equation describes the momentum balance for a ﬂuid 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 diﬀerent
acceleration terms, in addition to the already described Reynolds and Strouhal
numbers. These new dimensionless numbers will become available if the diﬀerent
78 Physical and numerical modelling of ﬂapping foils and wings
terms in equation (4.3) are scaled as
u
∗
=
u
U
ref
, t
∗
=
U
ref
c
t, ∇
∗
= c · ∇,
˙
Ω
∗
=
˙
Ω
˙
Ω
, Ω
∗
=
Ω
Ω
, r
∗
=
r
R
, p
∗
=
p
ρ
ref
· U
2
ref
,
leading to (dropping the stars for simplicity):
U
ref
f · c
·
Du
Dt
+
˙
ΩRc
U
2
ref
·(
˙
Ω×r)+
Ω
2
Rc
U
2
ref
·(Ω×(Ω×r))+
Ωc
U
ref
·(2Ω×u) = −∇p+
ν
U
ref
c
·∇
2
u,
where U
ref
[m/s] is the reference velocity, f [1/s] the ﬂapping frequency, which
is deﬁned as f =
U
ref
c
, where c the average chord length. L [m] is the reference
length, Ω [ rad/s] and
˙
Ω [rad/s] are the average rotational velocity and acceler
ation, respectively. Furthermore, ρ
ref
[kg/m
3
] is the reference density (constant
in incompressible ﬂows). R [m] is a radius length. In addition to the Reynolds
and Strouhal numbers, other dimensional numbers can be identiﬁed, related to
the rotation of the reference frame, which is still attached to the wing, rewriting
gives:
1
St
·
Du
Dt
+
1
C
ang
· (
˙
Ω×r) +
1
C
cen
· (Ω×(Ω×r)) +
1
Ro
· (2Ω×u) = −∇p+
1
Re
· ∇
2
u,
where the additional dimensionless number are respectively deﬁned as
C
ang
=
U
2
ref
˙
ΩR c
, (4.4)
C
cen
=
U
2
ref
Ω
2
R c
, (4.5)
Ro =
U
ref
Ω c
. (4.6)
By the rotation amplitude Ω, these dimensionless numbers strongly depend on the
wing kinematics, especially the velocity and acceleration due to the wing rotation.
Therefore, the eﬀect of diﬀerent kinematics and ﬂow features may be related to
these dimensionless numbers.
4.3 Wing shape and kinematic modelling
When investigating the threedimensional aerodynamics around ﬂapping wings,
one may try to simulate the geometry and conditions of speciﬁc species, in order
to fully understand them, or to opt for a more generic approach. Previous inves
tigations of speciﬁc insect species have been reported for e.g. a fruit ﬂy (Sane &
Dickinson, 2001, Birch & Dickinson, 2003), hawkmoth (Liu & Kawachi, 1998) or
4.3 Wing shape and kinematic modelling 79
a dragonﬂy (Isogai et al., 2004). The present study follows the second approach
and considers a threedimensional ellipsoidal model wing (Bos et al., 2008); a sim
ilar ellipsoid was used by Wang et al. (2004) for their twodimensional research.
Section 4.3.1 brieﬂy 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 ﬂexing 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 threedimensional wing
reference frame:
_
x
a
_
2
+
_
y
b
_
2
+
_
z
c
_
2
= 1, (4.7)
where a, b and c are the semiaxes 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 semiaxes are chosen as a = 0.05, b = 0.5 and
c = 1.0 (for the twodimensional airfoil, a similar elliptical crosssection is applied
for z = 0). Previous studies show that speciﬁc insect features, like the corrugated
wing planform (Luo & Sun, 2005) are of minor inﬂuence on the resulting ﬂuid
behaviour. The threedimensional elliptical planform is shown in ﬁgure 4.2 in
comparison to a more realistic representation of the fruit ﬂy 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 = π/4 for c = 1.0, both deﬁning the
planform. Figure 4.2 shows the wing planform and corresponding parameters for
a fruit ﬂy wing and a ellipsoidal model wing.
Radius of gyration
In order to deﬁne a sound framework of comparison for diﬀerent threedimensional
and twodimensional simulations, it is necessary to deﬁne all reference parameters,
80 Physical and numerical modelling of ﬂapping 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 ﬂy 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 ﬂy (a) and ellipsoidal model wing (b). The planform is deﬁned 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 deﬁne a sound framework for comparison between mutual threedimensional and two
dimensional simulations. In threedimensional simulations the wing revolves around the origin O, of
which the location is varied to study the eﬀect of the angular accelerations.
introduced in section 4.2, at a representative crosssectional area of the wing. As
the local velocity of each crosssection varies during ﬂapping, 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 diﬀerent kinematic
models is concerned (Bos et al., 2008). The radius of gyration, R
g
is deﬁned 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 ﬂow induced by ﬂapping
wings with diﬀerent aspect ratios and found that the radius of gyration provided
a reliable framework for force comparison when the ﬂapping 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 threedimensional ﬂow around ﬂapping wings at low Reynolds
numbers, the scale at which insects operate. Previous twodimensional studies
4.3 Wing shape and kinematic modelling 81
X
Y
Z
θ(t)
φ(t)
α(t)
O
R
root
R
tip
start downstroke
start upstroke
midstroke
Stroke plane
Figure 4.3 Schematic illustration of the governing ﬂapping angles. Flapping wing motion
is governed by three angles, φ(t) corresponds to the stroke variation, α(t) to the geometrical angle of
attack and deviation θ(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 threedimensional numerical method has been developed to solve
for the ﬂow around ﬂapping wings, the eﬀect of diﬀerent kinematic models on the
aerodynamic performance can be evaluated. Besides the use of an idealisation of
the insect wing planform, simpliﬁed kinematics was used, like harmonic motion,
to study diﬀerent 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 deﬁned by the variation of three independent
attitude angles, see ﬁgure 4.3. In this threedimensional model the three degrees
of freedom of the wing motion are deﬁned as the ﬂapping angle, φ(t), in the mean
stroke plane, the angle of attack, α(t), with respect to the horizontal plane and
the deviation θ(t), which is measured in a plane normal to the stroke plane, as is
shown in ﬁgure 4.3. The deviation may be used to create a ‘ﬁgureofeight’ pattern
which is present in realistic fruit ﬂy kinematics (Fry et al., 2003).
In literature, diﬀerent kinematic models have been studied, from purely har
monic motion to complex realistic fruit ﬂy kinematics. Using twodimensional
numerical techniques, Bos et al. (2008) showed that the ﬂapping wing perfor
mance may be inﬂuenced by the speciﬁc 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 ﬂy kinematics (Fry et al., 2003). Figure 4.4(b)
shows that the realistic fruit ﬂy kinematics is characterised by an asymmetry in
ﬂapping angle, and angle of attack. Furthermore, the angle of attack shows a dip
(at t=0.1T), lowering the eﬀective angle of attack. In addition, the shape of the
fruit ﬂy angle of attack clearly shows a plateau of constant value, which is ex
82 Physical and numerical modelling of ﬂapping foils and wings
ploited by Sane & Dickinson (2001, 2002), Lehmann et al. (2005) using a Roboﬂy.
Additionally, the realistic fruit ﬂy kinematics in characterised by the presence of
deviation which may result in a ‘ﬁgureofeight’ pattern (Shyy et al., 2008b)
In order to illustrate the variation of the motion angles, the deﬁnition of the
harmonic model is described accordingly. The ﬂapping angle, φ(t), is described by
a cosine function. The geometric angle of attack, α(t), is deﬁned by a sine function
with respect to 90
◦
and the deviation angle, θ(t), is given by a pure sine:
φ(t) = A
φ
· cos(2πft),
α(t) =
π
2
−A
α
· sin(2πft), (4.11)
θ(t) = A
θ
· sin(2πft).
Here, A
φ
is the ﬂapping amplitude, which is deﬁned from stroke reversal to mid
stroke. Remind that this amplitude is half times the value used in literature (e.g.
Φ
0
in Ellington, 1984). f is the ﬂapping frequency and A
α
represents the amplitude
of the angle of attack with respect to π/2, which is the initial position under
hovering conditions. A
θ
is the amplitude of the deviation angle which causes
the socalled ‘ﬁgureofeight’ pattern and is sinusoidal shaped (Fry et al., 2003).
The corresponding angular velocities are found by taking the timederivatives of
equation (4.11).
˙
φ(t) = −A
φ
· (2πf) · sin(2πft),
˙ α(t) = −A
α
· (2πf) · cos(2πft), (4.12)
˙
θ(t) = A
θ
· (2πf) · cos(2πft).
In general, this harmonic model is crude but fairly reasonable representation of
the ﬂapping motion of a fruit ﬂy. Also, simpliﬁed (harmonic) kinematics may
be interesting for Micro Air Vehicles (MAV) implementation. As will be shown in
chapter 5 (Bos et al., 2008), the ﬂapping wing performance may be signiﬁcantly in
ﬂuenced by modiﬁcations 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 piecewise continuous function and the deviation by deﬁning a
nonzero harmonically variation of θ(t).
4.3.3 Modelling of active wing ﬂexing
In addition to the rigid body motion as shown in ﬁgure 4.3 it is possible to deﬁne
an extra displacement concerning ﬂexing of the wing. Since a full ﬂuid structure
interaction (FSI) simulation is too expensive and beyond the scope of the current
research, a ﬂexing displacement of the wing surface is deﬁned. The ﬂexing dis
placement is deﬁned with respect to the initial wing position and can be written
4.3 Wing shape and kinematic modelling 83
t/T []
φ
,
α
,
θ
[
◦
]
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 []
φ
,
α
,
θ
[
◦
]
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 ﬂy kinematics
Figure 4.4 Comparison of the harmonic and fruit ﬂy kinematic models. (a) shows the
least complex kinematic model, representing pure harmonic variations of the ﬂapping angle φ(t) (•),
angle of attack α(t) (◦) and deviation θ(t) (). The variation of realistic fruit ﬂy kinematics (Fry
et al., 2003) is shown in (b), which is the most complex kinematics available. The realistic fruit ﬂy
kinematics is characterised by an asymmetric variation of ﬂapping angle, an extra ‘bump’ and a degree
of ‘trapezoidal’ in angle of attack and the presence of deviation.
as
x(t) = A
f
· cos
_
2πx
0
ǫ
f
_
· sin(2πft),
where the ﬁrst cosine function deﬁnes the wing shape and the sine function rep
resents the timevariation. A
f
is the ﬂexing amplitude vector and x
0
the location
of the initial boundary points. The cosine shaped wing ﬂexing is deﬁned by ǫ
f
,
which corresponds to the cosine ratio, i.e. ǫ
f
= 0.5 means that the shape is like a
half cosine function. Figure 4.5 shows a plunging airfoil incorporating ﬂexing for
ǫ
f
= 0.25 (a) and ǫ
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 ﬂexing (deformation) component. The translational component is not used
in the current threedimensional simulations, which is limited to hovering and for
ward ﬂow conditions with stationary position of the rotation origin. However, in
the twodimensional simulations, the ﬂapping motion is deﬁned 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 threedimensional simulations
using a predeﬁned wing ﬂexing have been performed.
In general, the wing kinematics is calculated beforehand and applied to the
numerical ﬂow solver. At every timestep the location of the boundary points is
determined leading to three distinct displacement arrays, due to translation, rota
84 Physical and numerical modelling of ﬂapping foils and wings
(a) Quarter cosine shape,ǫ
f
= 0.25 (b) Half cosine shape, ǫ
f
= 0.5
Figure 4.5 Illustration of a ﬂexing twodimensional airfoil. The twodimensional ﬂexing airfoil
is modelled by a timevarying cosine shape. This airfoil shape is either a quarter cosine, ǫ
f
= 0.25
(a) or a half cosine, ǫ
f
= 0.5 (b). Both ﬁgures (a) and (b) show the upstroke (left) and downstroke
(right).
tion and ﬂexing.
Translation
The displacement array of the boundary points, which is due to translation is
obtained from
x(t) = A
t
· sin(2πf
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 deﬁnition is used when the
wing needs to move according to an ordinary rigid body motion (Ginsberg, 1998).
When a ﬂapping motion is desired, a cosine function is used to deﬁne the motion
in the inertial reference frame.
Rotation
The second boundary point displacement array, due to rotation, is calculated at
subsequent (old and new) timesteps 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 diﬀerent components, due to a rotation around the X, Y 
and Zaxis. According to (Ginsberg, 1998, Baruh, 1999), these three matrices are
deﬁned as
R
X
(t) =
_
_
1 0 0
0 cos(θ(t)) −sin(θ(t))
0 sin(θ(t)) cos(θ(t))
_
_
, (4.15)
R
Y
(t) =
_
_
cos(φ(t)) 0 −sin(φ(t))
0 1 0
sin(φ(t)) 0 cos(φ(t))
_
_
, (4.16)
and
R
Z
(t) =
_
_
cos(α(t)) −sin(α(t)) 0
sin(α(t)) cos(α(t)) 0
0 0 1
_
_
. (4.17)
Here, the rotation around the X, Y  or Zaxis correspond to the deviation angle,
θ(t), ﬂapping angle, φ(t), and the angle of attack, α(t), respectively. Using the
motion convention, shown in ﬁgure 4.3, it is clear that a hovering wing ﬂaps in the
XZ plane, accelerating the ﬂuid 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 ﬁnd 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 ﬂapping wing is able to perform a
ﬂexing motion as well. This wing ﬂexing is deﬁned as
x
ﬂex
(t) = A
f
· cos(2πx
0
) · sin(2πf
f
t),
where A
f
is the ﬂexing amplitude vector, f
f
the ﬂexing frequency and x
0
repre
sents the initial boundary points of the ﬂapping wing at t = 0. This deﬁnition
permits ﬂexing in each orientation of a threedimensional wing, i.e. spanwise and
chordwise.
When combining the previous results for the displacements due to translation,
rotation and ﬂexing, the following is obtained:
∆x
b
= ∆x
trans
+ ∆x
rot
+ ∆x
ﬂex
.
86 Physical and numerical modelling of ﬂapping foils and wings
This total displacement ﬁeld for the boundary points can be applied to the mesh
motion solver, available in most commercial and noncommercial CFD codes.
Numerical initialisation
If a ﬂapping wing is considered, it is desirable to start the numerical simulation at
maximal ﬂapping angle, in order to minimise the initial acceleration of the mesh.
Therefore, it was chosen to use a rigid body motion until the maximal ﬂapping
angle was reached, in general this occurs at T/4, where T is the ﬂapping 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 ﬂapping wings
In section 4.2 diﬀerent dimensionless numbers, concerning rotational motion, were
identiﬁed in order to scale the NavierStokes equations, governing ﬂuid ﬂow. To de
ﬁne these rotational dimensional numbers, C
ang
= U
2
ref
/
˙
ΩR c, C
cen
= U
2
ref
/Ω
2
R c
and Ro = U
ref
/Ω c, it is important to identify the average rotation amplitude, Ω
and its timederivative
˙
Ω for the particular kinematics implemented:
Ω =
1
T
_
T
0

˙
φdt =
1
T
_
T
0
 −A
φ
· (2πf) · sin(2πft)dt = 4A
φ
f, (4.20)
˙
Ω =
1
T
_
T
0

¨
φdt =
1
T
_
T
0
 −A
φ
· (2πf)
2
· cos(2πft)dt = 8πf
2
A
φ
= 2πfΩ, (4.21)
where the ﬂapping velocity
˙
φ was taken from equation (4.12). As shown, evaluation
gives Ω = 4A
φ
f and
˙
Ω = 2πfΩ. Besides the angular velocity Ω, angular accelera
tion
˙
Ω and average chord length c it is necessary to deﬁne an appropriate reference
velocity. As previously described and in accordance with (Bos et al., 2008, Lentink,
2008, Luo & Sun, 2005), the reference crosssection of the threedimensional wing
is positioned at the radius of gyration, R
g
, and the reference velocity is calculated
at that particular location. This timeaveraged 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 Zdirection. Using
equation (4.22) it is straightforward to ﬁnd U
ref
by multiplying the expression for
Ω (4.20) by R
g
:
U
ref
= 4A
φ
fR
g
. (4.23)
These relations hold if the wing is the only driving force behind the resulting ﬂow
velocity, but in case of forward ﬂight conditions, there is an additional freestream
4.4 Dynamical scaling of ﬂapping wings 87
λ
∗
90
◦
−β
A
α
tan
−1
(2Stsinβ)
2A
∗
sinβ
Figure 4.6 Schematic illustration of the kinematic parameters in forward ﬂight.
velocity, U
∞
. Lentink & Gerritsma (2003) showed that the following relation can
be used as a good approximation for both hovering (U
∞
= 0) and forward ﬂight
conditions:
U
ref
= U
R
g
≈
_
U
2
∞
+ (4A
φ
fR
g
)
2
. (4.24)
In forward ﬂight another important parameter is the advance ratio (Shyy et al.,
2008b), J, which describes the forward speed with respect to the ﬂapping velocity
at a certain radius, R:
J =
U
∞
4A
φ
fR
g
=
λ
∗
4A
∗
,
where the reduced frequency (Ellington, 1984), λ
∗
, is given by λ
∗
= U
∞
/fc, with f
the ﬂapping frequency. According to (Shyy et al., 2008b, Thaweewat et al., 2009),
λ
∗
is also known as the dimensionless wavelength, which is illustrated in ﬁgure 4.6.
The dimensionless amplitude at the radius of gyration R
g
is deﬁned as
A
∗
=
A
φ
R
g
c
,
which is a measure for the dimensionless translation of the selected crosssectional
area. In order to create an appropriate framework for comparison, the dimension
less amplitude at R
g
, A
∗
is kept constant for all relevant simulations. Additionally,
a constant A
∗
leads to similar wingwake interactions (Birch & Dickinson, 2003)
for the representing simulations. For completeness and consistency, two other im
portant parameters are kept constant as well, the Reynolds number at R
g
, which
is an implicit result of keeping R
g
constant and the area swept by the wing, A
swept
.
The Reynolds number for hovering conditions and using R
g
can be written:
Re =
U
ref
c
ν
=
4A
φ
fR
g
c
ν
.
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 ﬂapping foils and wings
area by the wing root:
A
swept
= 2A
φ
· (R
2
tip
−R
2
root
),
= 2A
φ
· (R
tip
−R
root
) · (R
tip
+R
root
),
= 2A
φ
· 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 identiﬁed 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 eﬃciency (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
Ω c
=
R
g
c
. (4.25)
When the reference crosssectional 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
inﬁnite; for a rotating wing, Ro is ﬁnite. If Ro is varied, the eﬀect of diﬀerent
rotation origins can be investigated, from nearly translating to strongly revolving.
However, the determination of Ro would be easier if deﬁned 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 ﬁsh (Lentink, 2008), generating thrust
by moving the ﬂuid. If the wing planform is complex, like in real insects, it may be
diﬃcult 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 threedimensional ﬂapping wing simulations. Chapter 5, 6
and 7, which are dealing with respectively twodimensional hovering, forward and
threedimensional hovering ﬂight, explain the speciﬁc 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
diﬀerential equations are solved. It is necessary to use a suﬃciently 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 twodimensional thin and ellipseshaped airfoil, it is eﬃcient
4.6 Deﬁnition of force and performance coeﬃcients 89
X
Y
Γ
left
Γ
wing
Γ
right
(a) Twodimensional
Z X
Y
Γ
wing
Γ
left
Γ
right
Γ
bottom
Γ
top
Γ
front
Γ
back
(b) Threedimensional
Figure 4.7 Computational domain and boundary conditions. The twodimensional domain
uses the Otype topology (a), while the threedimensional ﬂapping wing simulations uses the boxed
topology (b).
to use an Otype mesh, which is shown in ﬁgure 4.7(a). Using conformal map
pings (Bos et al., 2008, Thaweewat et al., 2009), a high quality mesh is generated
between the ellipseshaped wing boundary, Γ
wing
, and the cylindershaped outer
boundaries, Γ
left
and Γ
right
. The outer boundary is split into Γ
left
and Γ
right
in
order to be able to specify inﬂow and outﬂow boundary conditions, which are
necessary (Wesseling, 2001) when forward ﬂight conditions are simulated. When
simulating hovering ﬂight, the ﬂapping wing still induces a signiﬁcant amount of
downwash, leading to a small inﬂow and outﬂow at the boundaries of the compu
tational domain. The meshes for the twodimensional simulations were generated
using Gambit software, see appendix A.
Generation of a threedimensional 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
. Therefore, an automated topology mesh generator is
used, GridPro
. Using GridPro
, it was possible to generate a high quality mesh
around a thin ellipsoidal wing in a threedimensional boxshaped computational
domain, which is shown in ﬁgure 4.7(b). Depending on the ﬂapping conﬁgurations,
the outer boundaries, Γ
left
, Γ
right
, Γ
bottom
, Γ
top
, Γ
front
, Γ
back
, are set to inﬂow,
outﬂow or symmetry planes.
4.6 Deﬁnition of force and performance
coeﬃcients
Besides ﬂow ﬁeld analysis, using advanced visualisation techniques, the resulting
forces acting on the airfoils and wings are of primary importance to assess aero
dynamic performance. In order to make a sound comparison of forces and perfor
mance for mutual two and threedimensional simulations (chapters 5, 6 and 7) it
90 Physical and numerical modelling of ﬂapping foils and wings
PSfrag
X
Y
Z
R
g
F
X
F
Y
F
Z
˙
φ
Figure 4.8 Forces on a general threedimensional ﬂapping wing. The forces are deﬁned in the
threedimensional inertial reference frame. Depending on the type of motion, hovering or forward ﬂight,
two or threedimensional, the lift and drag are constructed from the forces in X, Y  or Zdirection.
is important to properly deﬁne force and performance coeﬃcients. In this section,
the general determination of the force coeﬃcients is explained in combination with
performance characteristics.
Forces
In general three dimensions, a twodimensional derivation is trivial, the deﬁnitions
of the forces, F
X
, F
Y
and F
Z
are shown in ﬁgure 4.8. The centre of the axes co
incides with the origin of rotation of the threedimensional 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
µ
∂u
∂n
dS,
where F
tot
[N] is the total force vector, S [m] the wing surface and dS [m] represents
an inﬁnitely small surface area element, p [N/m
2
] is the pressure and µ [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 µ this forms the wall
shear stress.
Two diﬀerent force deﬁnitions can be deﬁned, depending on the type of three
dimensional motion, hovering or forward ﬂapping ﬂight. In hovering conditions,
when the main ﬂapping direction is around the Y axis, as shown in ﬁgure 4.3, the
lift force F
lift
is deﬁned in the vertical direction and equal to F
Y
. The drag force
F
drag
, however, is deﬁned in opposite direction of the ﬂapping wing motion. If
threedimensional ﬂapping in hovering ﬂight 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(φ(t)) −F
Z
(t) · cos(φ(t)),
where φ(t) is the ﬂapping 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 Deﬁnition of force and performance coeﬃcients 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 justiﬁed to neglect this spanwise force, also because it has little relevance to
performance.
In forward ﬂight conditions, the same inertial reference frame is used as shown
in ﬁgure 4.3. The main ﬂapping direction is still around the Y axis, but the
direction of the uniform ﬂow is from top to bottom in direction of the negative
Y axis. The complete system needs to be rotated around the Zaxis in order to get
a horizontal orientation of the freestream. In the case of forward ﬂapping ﬂight
the lift force is deﬁned in the positive Xdirection and the drag in the negative
Y direction, opposite to the freestream 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 deﬁned as:
C
D
=
F
drag
q · S
, C
L
=
F
lift
q · S
,
where C
D
and C
L
are the drag and lift coeﬃcients and S the wing surface. The
mean dynamic pressure q is deﬁned as:
q = 1/2ρU
2
ref
= 1/2ρ ·
1
T
_
T
0
_
(U
∞
+U
ﬂap
(t))
2
+ (U
dev
(t))
2
dt,
where the integration is evaluated over one ﬂapping cycle with period T [s]. The
reference velocity contains the freestream velocity U
∞
, which is nonzero in for
ward ﬂapping ﬂight, the ﬂapping velocity U
ﬂap
and the deviation velocity U
dev
,
perpendicular to the horizontal plane (in hovering ﬂight).
Performance
The force coeﬃcients are the major parameters used to assess the inﬂuence of the
diﬀerent wing motion models. In addition, the ratio between timeaveraged lift
coeﬃcient C
L
and timeaveraged drag coeﬃcient C
D
is used to characterise per
formance. These force averages are obtained by integration of C
L
and C
D
. The
lift is averaged over the complete ﬂapping period, while for the drag the absolute
values are used for averaging. The drag is opposed to the ﬂapping motion, such
that the sign ﬂips at stroke reversal.
The average lifttodrag ratio, C
L
/C
Dave
is chosen as an indicator of aero
dynamic performance, also known as the glide number in aerospace engineering.
When the average lift coeﬃcients of the diﬀerent kinematic models are matched,
92 Physical and numerical modelling of ﬂapping foils and wings
the lifttodrag ratio is corrected for any diﬀerences in lift. Therefore, a high liftto
drag ratio eﬀectively means low drag at equal lift. Additionally, the power factor,
C
L
3/2
/C
D
, (see Ruijgrok, 1994), has been used to assess the required power for a
certain amount of lift (Wang, 2008).
4.7 Conclusions
This chapter dealt with the physical and numerical modelling of threedimensional
ﬂapping wings and twodimensional ﬂapping airfoils. Two important dimensionless
numbers were identiﬁed, the Reynolds and Strouhal number. These numbers are
common in general ﬂuid ﬂow, but for ﬂapping ﬂight the deﬁnition has been slightly
changed. In order to analyse the ﬂow around ﬂapping wing, the governing Navier
Stokes equations are written in a rotating reference frame. This leads to an extra
important dimensionless number, related to the wing rotation, namely the Rossby
number. The Rossby number is a way to describe the radius of curvature and thus
the angular accelerations in dimensionless terms.
In order to systematically study the aerodynamics around ﬂapping wings at
the scale of insects, a model wing planform has been deﬁned with an ellipsoidal
planform. Using that planform, the radius of gyration can be easily obtained,
which is used to dynamically scale the wing kinematics. Diﬀerent ﬂapping wing
kinematic models are brieﬂy addressed, from realistic fruit ﬂy kinematics to a
fully harmonic model. In addition to the rigid body rotations, a deformation of
the wing is deﬁned in order to study the eﬀects of wing ﬂexing. This ﬂexing motion
is deﬁned with respect to the initial wing position and has a timevarying cosine
shape.
Most importantly, to design a sound framework for comparison it is necessary
to dynamically scale the wing kinematics for all numerical simulations. This is
achieved by scaling the motion parameters such that the dimensionless amplitude,
the average Reynolds number and the area swept by the wing result in comparable
values. The radius of gyration is used as a reference crosssection for both two
dimensional ﬂapping foil as well as threedimensional ﬂapping wing simulations.
Analysis of the vortical ﬂow around the wings and foils is primarily performed
by plotting the force coeﬃcient. The lift is deﬁned in vertical direction, while
the drag is opposed to the ﬂapping velocity. The corresponding coeﬃcients are
obtained by averaging the dynamic pressure, which has proved to be proper refer
ence. In addition to the forces, the lifttodrag ratio is used to assess the ﬂapping
wing performance.
CHAPTER 5
A 2D investigation of the inﬂuence
of wing kinematics in hovering
ﬂight
J. Fluid Mech. (2008), vol. 594, pp. 341368.
The inﬂuence of diﬀerent wing kinematic models on the aerodynamic performance
of a hovering insect is investigated by means of twodimensional timedependent
NavierStokes simulations. For this, simpliﬁed models are compared with averaged
representations of the hovering fruit ﬂy wing kinematics. With increasing com
plexity, a harmonic model, a Roboﬂy model and two more realistic fruit ﬂy models
are considered, all dynamically scaled to Re = 110. To facilitate the comparison,
the parameters of the models were selected such that their mean quasisteady lift
coeﬃcient were matched. Details of the vortex dynamics, as well as the resulting
lift and drag forces were studied. The simulation results reveal that the fruit ﬂy
wing kinematics result in forces that diﬀer signiﬁcantly from those resulting from
the simpliﬁed wing kinematic models. In addition, light is shed on the eﬀect of
diﬀerent speciﬁc characteristic features of the insect wing motion. The angle of
attack variation used by fruit ﬂies increases aerodynamic performance, whereas
the deviation is most likely used for levelling the forces over the cycle.
94 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
5.1 Introduction
In order to investigate the full ﬂow around a threedimensional ﬂapping wing,
twodimensional simulations are performed to get insight in the complicated ﬂow
structures. This chapter deals with the evolution of the forces and the wake
originated by a ﬂapping foil in hovering conditions.
5.1.1 Similarity and discrepancy between two and
threedimensional ﬂows
In a recent paper Wang et al. (2004) compared threedimensional Roboﬂy results
with twodimensional numerical results. This showed that twodimensional simu
lations are useful to obtain a better understanding of the ﬂow features, which can
then be investigated more thoroughly in three dimensions.
Both Dong et al. (2005) and Blondeaux et al. (2005b) concluded that twodi
mensional studies overpredict forces and performances since the energyloss, which
is present in three dimensions, is resolved. Dong et al. (2005) and Blondeaux et
al. (2005b) numerically investigated the wake structure behind ﬁnitespan wings
at low Reynolds numbers. They observed that the ﬂapping wings with low aspect
ratio generates threedimensional vortical structures as was mentioned by Lighthill
(1969).
Notwithstanding the possible discrepancy between twodimensional and three
dimensional ﬂow, twodimensional analysis has often been applied to obtain insight
into the aerodynamic eﬀects of choices in kinematics, airfoil crosssection, Reynolds
numbers, etc. Wang et al. (2004) conﬁrmed that the similarities between two and
threedimensional approaches are suﬃcient to warrant that a reasonable approxi
mation of insect ﬂight can be obtained using a twodimensional approach. First, in
case of advanced and symmetric rotation the forces were found to be similar in the
twodimensional simulations compared to the threedimensional experiments. Sec
ondly it was observed that in both simulations and experiments the leadingedge
vortex did not fully separate for amplitudetochord ratios between 35 (Dickinson
& G¨otz, 1993, Dickinson, 1994), a similar amplitude range was used in the present
research.
In view of the excessive computational expense required for accurate three
dimensional simulations, and with the above justiﬁcation, the present study was
restricted to twodimensional simulations. In a twodimensional simulation our
mesh resolution can be higher compared to a threedimensional simulation, in
view of the limitation of computational resources.
5.1.2 Inﬂuence of kinematic modelling
The relevance of (experimental or numerical) simulations of insect ﬂight has been
found to depend on how reliable true insect wing kinematics are reproduced. Wang
et al. (2004) and Sane & Dickinson (2001) showed that the kinematic modelling
5.1 Introduction 95
signiﬁcantly inﬂuences the mean force coeﬃcients and its distribution. Addition
ally, Hover et al. (2004) showed that modelling the angle of attack inﬂuences the
ﬂapping foil propulsion eﬃciency to a large extent. This illustrates the appreciable
eﬀects which details of the wing kinematics, like parameter values and stroke pat
terns, may have on ﬂight performance. It further emphasises the need to critically
assess the inﬂuence of kinematic model simpliﬁcations.
In literature, diﬀerent kinematic models have been employed to investigate the
aerodynamic features of insect ﬂight. 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 ﬂapping am
plitude and frequency and showed that at a certain parameter selection the lift is
clearly enhanced. Lewin & HajHariri (2003) performed a similar numerical study
for heaving airfoils. Besides lift enhancement at certain reduced frequencies, they
found periodic and aperiodic ﬂow solutions which are strongly related to the aero
dynamic eﬃciency. Lentink & Gerritsma (2003) varied airfoil shape with amplitude
and frequency ﬁxed at values representative to real fruit ﬂies. They concluded that
the airfoil choice is of minor inﬂuence, 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 including rotation is essential. In addition
to the harmonic models with pure translation (Dickinson & G¨otz, 1993), rota
tional parameters were investigated by Dickinson (1994). They varied rotational
parameters and showed that axisofrotation, rotation speed and angle of attack
during translation are of great importance of the force development during each
stroke. Harmonic wing kinematics, including wing rotation, were used by Pedro
et al. (2003) and Guglielmini & Blondeaux (2004) in their numerical models to
solve for forward ﬂight. Both studies emphasised the importance of angle of at
tack modelling to inﬂuence the propulsive eﬃciency. Slightly more complex fruit
ﬂy kinematic models were used by Dickinson et al. (1999) and Sane & Dickinson
(2001) with their Roboﬂy. Based on observation of true insect ﬂight, 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 reversal. This
results in the typical ‘sawtooth’ displacement and ‘trapezoidal’ angle of attack
pattern of the Roboﬂy kinematic model. Using these models, the eﬀect of ampli
tude, deviation, angle of attack and the timing of the latter were explored.
In the present study, diﬀerent models from literature were considered, both the
pure harmonic and the Roboﬂy model, in order to investigate their inﬂuence on the
aerodynamics. Furthermore, the results were compared with more realistic fruit
ﬂy kinematics obtained from the observation of free ﬂying fruit ﬂies (Fry et al.,
2003). Instead of performing a parameter study within the scope of one kinematic
model, the objective of the present study is to compare the eﬀect of the available
models as a whole. This leads to better insights in the consequences of simpliﬁ
cations in kinematic modelling, which is of great importance to both experiments
and numerical simulations. Also, it can reveal the importance of certain speciﬁc
96 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
features of the stroke pattern, in relation to aerodynamic performance.
This study considers four diﬀerent wing kinematic models with varying degree
of complexity. These models are implemented in a generalpurpose Computational
Fluid Dynamics (CFD) code, which solves the NavierStokes equations under the
assumption of incompressible ﬂow. In brief, the ﬁrst model describes the wing
motion using basic harmonics as derived by Wang (2000a). The second model
contains the kinematics implemented by Dickinson et al. (1999) for their Roboﬂy
at UC Berkeley (presently CalTech). The third model is a representation of the real
kinematics used by a hovering fruit ﬂy (Drosophila Melanogaster), based on data
measured by Fry et al. (2003). Finally, the fourth model is a slightly simpliﬁed
version of the latter, observed fruit ﬂy model. All these kinematic models are
dynamically scaled at a Reynolds number of Re = 110 which corresponds to
the ﬂight conditions of the fruit ﬂy. In addition, these kinematic models are
constructed such that their mean quasisteady lift coeﬃcients are comparable such
that our performance comparison is justiﬁed. This basis of comparison is veriﬁed
aposteriori from the force results of the actual simulations.
The outline of this chapter is as follows. In section 5.2 the numerical simulation
methods are described. In addition, the actual modelling of the insect parameters
is discussed in 5.3. The results of the numerical simulations obtained with the
diﬀerent kinematic models are treated in section 5.4 and concluding remarks are
given in 5.5.
5.2 Numerical simulation methods
The diﬀerent kinematic models are implemented in the commercial ﬂow solver
Fluent
, which solves the governing incompressible NavierStokes equations on
a twodimensional computational mesh. The resulting model has been validated
using stationary and moving circular cylinders and veriﬁed using harmonically
moving wings.
5.2.1 Flow solver and governing equations
To simulate the ﬂow around moving wings with predeﬁned motions the commer
cial CFD solver Fluent
was used. The twodimensional timedependent Navier
Stokes equations are solved using the ﬁnite volume method, assuming incompress
ible ﬂow which is justiﬁed since the Mach number of ﬂapping insect ﬂight is typi
cally O(10
−3
) (see Brodsky, 1994). The mass and momentum equations are solved
in a ﬁxed inertial reference frame incorporating a moving mesh following the Ar
bitrary Lagrangian Eulerian (ALE) formulation (Ferziger & Peric, 2002).
At the considered Reynolds number, Re = O(100), the ﬂow is assumed to
be laminar. Henderson (1995) and Williamson (1995) showed that for circular
cylinders, the transition from laminar to turbulent ﬂow occurs at Re = 180 ± 5,
which supports this assumption. Therefore, the transient incompressible laminar
5.2 Numerical simulation methods 97
Γ
3
Γ
2
Γ
1
x
y
Ω
2
Ω
1
x
y
0.4 0.42
0.56
0.58
Figure 5.1 Otype mesh topology with
boundary conditions on Γ
1
,Γ
2
and Γ
3
.
Figure 5.2 Body conformal moving
mesh around a 2% ellipsoid airfoil.
NavierStokes 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 ﬂow around the moving airfoils, an Otype computational
domain is used, which is shown schematically in ﬁgure 5.1. The computational
domain is divided into two parts: Ω
1
and Ω
2
for the inner and outer mesh respec
tively. The body surface Γ
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 Γ
3
is located at 25L such that the inﬂuence of the far ﬁeld bound
ary condition is negligible (Lentink & Gerritsma, 2003). At the body surface a
noslip boundary condition is applied. Since the moving wing simulations concern
hovering insect ﬂight, such that a freestream is absent, a symmetry boundary
condition was applied at Γ
3
for numerical reasons. The inﬂuence of this symmetry
condition has been investigated and found to be suﬃciently 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 ﬁgure 5.2) a conformal mapping was applied (see
Wang, 2000b). The intermediate interface Γ
2
divides the mesh into two separate
ﬁelds, corresponding respectively to the inner conformal mesh (Ω
1
) and the outer
mesh (Ω
2
). The complete inner mesh moves according to the wing kinematics,
while remeshing takes place in the outer ﬁeld Ω
2
. Since remeshing occurs at
a distance of 25 to 30 body lengths away from the wing, the ﬂow around the
wing is not aﬀected by the mesh regeneration. The described computational setup
was thoroughly validated using the ﬂow around stationary and moving circular
98 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
t +dt
t
α
ref
α
ref
+dα
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
timesteps within one motion period. At this mesh, the size of the ﬁrst cell at the
wing surface varies between 2% and 50% of the wing thickness at the leadingedge
and in the middle of the proﬁle 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 ﬂapping
periods needed approximately 10 days on one serial AMD Athlon 2500+ CPU.
In order to minimise the interpolation errors from one timestep to the next
it is important to analyse the inﬂuence of the relative cell displacements. There
fore, the motion of a reference cell was investigated, which is illustrated for the
rotational motion in ﬁgure 5.3. From the relative displacements in rotational and
translational direction follow the constraints for the size of the timestep in or
der to keep the interpolation errors within limits. The relative displacements in
rotational and translational direction are deﬁned as
ǫ
r
=
△α
α
ref
and
ǫ
y
=
△y
y
ref
=
2f
e
A
e
N△t
y
ref
.
Here α corresponds to the angular displacement of the reference cell, while α
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
diﬀerences in drag coeﬃcients remaining below 5%. The computational eﬀorts are
acceptable: 2000 timesteps within one excitation period. Additionally, Bos et al.
(2008) (appendix D) investigated the mesh and timestep independence for the
nominal solver settings using harmonic wing kinematics for hovering ﬂight.
5.3 Modelling insect wing kinematics 99
t/T []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 coeﬃcients between the present simulations and Wang
et al. (2004). Comparison of lift (a) and drag (b) coeﬃcients 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 timesteps
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 coeﬃcients for validation purposes. Our forces
are normalised with the maximum of the quasisteady force, just as in (Wang et
al., 2004). In similarity to (Wang et al., 2004) the drag in ﬁgure 5.4(a) is deﬁned
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 ﬁnds a larger lift and drag which is probably
the result of diﬀerent numerical dissipation properties of both codes. The mean
lift and drag coeﬃcients are 0.84, 1.47 for our simulation, compared to 0.82, 1.44
obtained by Wang et al. (2004), which is a diﬀerence of only 2% in lift and drag and
therefore, the computations were considered to be suﬃciently accurate. Moreover,
within the context of comparing results of diﬀerent stroke patterns, the present
numerical method is proved to be accurate.
Further details of the validation and veriﬁcation studies can be found in (Bos
et al., 2008, appendix C and D).
5.3 Modelling insect wing kinematics
In order to derive the twodimensional kinematic models the threedimensional
degrees of freedom need to be converted to their twodimensional counterparts.
A common procedure is to deﬁne an equivalent twodimensional geometry, while
100 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
Figure 5.5 Illustration of the main motion directions. φ(t) corresponds to the stroke variation,
α(t) to the geometrical angle of attack and θ(t) to the deviation from the horizontal stroke plane.
From (Sane & Dickinson, 2001).
maintaining the characteristic aspects of the wing motion. This twodimensional
setup is derived in section 5.3.1 in terms of wing selection and model parameters.
The dynamical scaling and the force deﬁnitions 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 inﬂuence of diﬀerent
kinematic wing motion models on the aerodynamic performance. The diﬀerent
kinematic models are illustrated using the Roboﬂy experimental setup, shown
in ﬁgure 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 deﬁned
as the angular displacement φ in the mean stroke plane, the angle of attack α,
with respect to the horizontal plane and the deviation from the horizontal plane
θ, as is shown in ﬁgure 5.6. The deviation causes a ‘ﬁgureofeight’ pattern which
is present in real fruit ﬂy kinematics (see Fry et al., 2003). The twodimensional
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 twodimensional projection is to be deﬁned at a representative
spanwise location such that the motion is conﬁned 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 twodimensional model.
In the present study, a diﬀerent 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
α
Figure 5.6 Force deﬁnition on the twodimensional airfoil.
location (Lentink & Gerritsma, 2003) was used. Considering that the local velocity
of each crosssection varies during ﬂapping, the spanwise location was selected to
be at the radius of gyration where the mean lift acts (Ellington, 1984).
In view of providing completeness on the threedimensional setup, the used
values are for the wing surface S=0.0167 m
2
the wing tip radius R=0.254 m,
the location of centre of gravity x
cg
=0.0882 m, the location of the wing base
x
base
=0.0667 m and the moment of inertia I
cg
=40.42 · 10
−4
m
4
. For the radius
of gyration the following value was obtained R
g
= 0.6396 · R. When comparing
this distance to the value used by (Wang et al., 2004) the current crosssection
is less than 2% closer to the wing root. Apparently, the mean lift acts nearly at
the location where the vorticity is maximal. Another important parameter to be
deﬁned is the reference length, L
ref
, based on the mean chord length. A deﬁnition
of the mean chord length based on the moment of inertia around the wing root was
proposed. This leads to a value for the mean chord length of c = 0.082 m. Finally,
the conversion from threedimensional angles to nondimensional displacements is
given by:
x =
φ · R
g
c
, y =
θ · R
g
c
, (5.1)
where R
g
is the radius of gyration. Both the displacement x and the deviation
y have been made dimensionless with the mean chord c. The centre of rotation
is deﬁned in the aerodynamic centre which lies at the quarter chord point of the
mean chord.
5.3.2 Dynamical scaling of the wing model
Since the ﬂapping of the wings induces highly unsteady ﬂow the relevant ﬂow and
motion parameters have to be scaled dynamically. The period of the motion is
used to average the relevant ﬂow velocity (Lentink & Gerritsma, 2003):
U =
1
T
_
T
0
_
u
2
+v
2
dt. (5.2)
102 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
Here T [s] is the period , u represents the nondimensional velocity in the stroke
plane and v the nondimensional deviation velocity. Both are given by u = ∂x/∂t
and v = ∂y/∂t, where t = t/T is the dimensionless time.
Substituting equation (5.1) into (5.2) and evaluating, the following relations
for the Reynolds and Strouhal numbers were derived:
Re =
Uc
ν
=
fR
g
c
ν
·
_
1
0
_
(
∂φ
∂t
)
2
+ (
∂θ
∂t
)
2
(5.3)
and
St =
fc
U
=
c
R
g
·
1
_
1
0
_
(
∂φ
∂t
)
2
+ (
∂θ
∂t
)
2
. (5.4)
Here f = 1/T is the frequency, φ and θ the threedimensional 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 φ(t) and deviation θ(t). The Strouhal number St is not to be varied
independently. We ﬁxed the Reynolds number to Re = 110.
5.3.3 Force and performance indicators
The deﬁnition of the drag and lift forces is shown in ﬁgure 5.6. The lift is equal
to the vertical force F
y
, while the drag is taken equal to the horizontal force
F
x
, deﬁned positive in the positive xdirection. 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 deﬁned as
C
D
=
F
x
q · c
, C
L
=
F
y
q · c
,
where C
D
and C
L
are the drag and lift coeﬃcients. The mean dynamic pressure
q is deﬁned as
q = 1/2ρU
2
= 1/2ρ ·
1
T
_
T
0
_
_
∂x
∂t
_
2
+
_
∂y
∂t
_
2
_
dt,
where the integration is evaluated over one ﬂapping cycle. The force coeﬃcients
are the major parameters used to assess the inﬂuence of the diﬀerent wing motion
models. In addition, the ratio between timeaveraged lift coeﬃcient, C
L
, and time
averaged drag coeﬃcient, 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
lifttodrag 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
coeﬃcients of the diﬀerent kinematic models are matched, the lifttodrag ratio is
corrected for any diﬀerences in lift. Therefore, a high lifttodrag ratio eﬀectively
means low drag at equal lift.
5.3 Modelling insect wing kinematics 103
5.3.4 Diﬀerent wing kinematic models
Since the main purpose of this study is to investigate the inﬂuence of wing kinemat
ics on the aerodynamic performance during hovering fruit ﬂy ﬂight, four diﬀerent
kinematic models, with diﬀerent degree of complexity, have been analysed. Two
of these models, the pure harmonic motion and the Roboﬂy experimental kine
matics have appeared in literature. The third model represents the actual fruit ﬂy
kinematics as observed in experiments and the last one was a modiﬁcation of the
latter, chosen to investigate the eﬀect of symmetry in the wing motion.
In order to facilitate the comparison the model parameters are chosen based
on matching the mean quasisteady lift coeﬃcient (Bos et al., 2008, appendix
A). Although according to Sane & Dickinson (2001) the mean drag is strongly
inﬂuenced by the unsteady ﬂow physics, which are not fully present in the quasi
steady theory, the mean lift coeﬃcient is predicted well using this theory. Using
quasisteady theory, diﬀerent kinematic models were constructed, such that their
quasisteady lift coeﬃcients 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 diﬀerence 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 coeﬃcient of the numerical simulations are reasonably well matched for
all models, which provides an aposteriori justiﬁcation of our choices for the model
parameters.
The characteristic shapes of each model are described. Subsequently they are
used to investigate the inﬂuence 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 ﬂy may beneﬁt from kinematic features which are absent in the
simpler models, and reveals the relevance of including these aspects in theoretical
models.
The ﬁrst 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 ﬁgure 5.7(a). The second
model takes the wing kinematics as used in the Roboﬂy model (Dickinson et al.,
1999). In ﬁgure 5.7(b) it is shown that the ﬂip 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 ﬁgure 5.7(c),
is derived from measurements on real fruit ﬂies (Fry et al., 2003) and is therefore
considered as the most realistic fruit ﬂy kinematic model. This model does include
the deviation which results in a ‘ﬁgureofeight’ pattern. Neither the displacement,
angle of attack nor deviation is symmetric during the ﬂapping period.
In order to investigate the fact that the observed fruit ﬂy kinematics lacks an
exact symmetry in the wing stroke pattern, a symmetrical model was constructed,
referred to as the symmetric fruit ﬂy model, displayed in ﬁgure 5.7(d). Within this
model the motion is identical for the downstroke and upstroke. Like the realistic
104 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
t/T []
φ
,
α
,
θ
[
◦
]
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 []
φ
,
α
,
θ
[
◦
]
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 []
φ
,
α
,
θ
[
◦
]
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 []
φ
,
α
,
θ
[
◦
]
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 diﬀerent kinematic models. (a) Harmonic model. (b)
Roboﬂy model. (c) fruit ﬂy model. (d) simpliﬁed fruit ﬂy model. •: displacement angle φ, ◦: angle
of attack α, : deviation angle θ.
fruit ﬂy model this symmetric model includes a timedependent deviation such
that the observer sees a ‘ﬁgureofeight’ pattern of the wing. Neither of those
last two realistic kinematic models can be described by using simple analytical
functions without losing signiﬁcant detail.
When comparing the motion parameters, φ, α and θ for each model it becomes
possible to identify certain important diﬀerences. The Roboﬂy initially has a
larger gradient in time of the angle of attack compared to the harmonic case, see
ﬁgure 5.7(a) and (b). During translation from about t = 0.1T to t = 0.4T the
angle of attack ﬂattens at a value of almost 40
◦
. This ‘trapezoidal’ shape of α is
characteristic for the Roboﬂy and may be inﬂuencing the performance. Although
the Roboﬂy model clearly shows similarities with the fruit ﬂy models the latter
has some typical additional features. The most obvious peculiarity of the realistic
fruit ﬂy models is the extra ‘bump’ in angle of attack just after stroke reversal,
5.4 Results and Discussion 105
compared to the Roboﬂy (ﬁgure 5.7(b) and (c)). It follows the same high angular
velocity, but instead of ﬂattening α, the fruit ﬂy wing α descends to the ‘bump’.
After the ‘bump’ the angle of attack more or less matches the plateau found in
Roboﬂy but starts to increase earlier. During stroke reversal the gradient of α
matched the harmonic model closer than the Roboﬂy with its high gradients.
The harmonic and Roboﬂy models lack deviation, so no ‘ﬁgureofeight’ is
present. The deviation of the fruit ﬂy model is asymmetric during the complete
cycle, but also during each half stroke (ﬁgure 5.7(c)). This is likely to inﬂuence
the performance since the eﬀective angle of attack is altered due to deviation.
It is also observed that the deviation is negative for a certain period during the
upstroke. Therefore, the deviation of the realistic fruit ﬂy is averaged to derive the
simpliﬁed fruit ﬂy model, see ﬁgure 5.7(d). This last model is used to investigate
the inﬂuence of deviation on the force histories and performance.
5.4 Results and Discussion
In the previous section it was observed that the most interesting aspects of the
Roboﬂy kinematic model are the ‘sawtooth’ displacement and the ‘trapezoidal’
angle of attack. This implies that strong translational and rotational accelerations
occur at stroke reversal. The more realistic fruit ﬂy models are characterised
by a ‘bump’ in angle of attack and the presence of deviation. Results of two
comparative studies were presented. The ﬁrst is an overall comparison of the
complete kinematic models, which is described in section 5.4.1. In the second
study the eﬀect of the characteristic features identiﬁed above, are considered in
more in detail. In order to assess the eﬀect of these kinematic features in isolation,
the comparison is made using the simplest model, the harmonic model, as baseline.
Hereto this baseline model is subsequently modiﬁed by adding respectively the
‘sawtooth’ displacement,‘trapezoidal’ angle of attack, extra ‘bump’ in angle of
attack and the presence of deviation. The results of this comparison, in terms of
actual vortex dynamics, as well as the resulting lift and drag histories are studied
in 5.4.2.
5.4.1 Overall model comparison
In table 5.1 the mean force coeﬃcients are given for the four complete models,
the harmonic model, the Roboﬂy model, the realistic fruit ﬂy model and the
simpliﬁed fruit ﬂy model. The mean drag, for each halfstroke, and lift coeﬃcients
are given, as well as the average lifttodrag ratio, which characterises aerodynamic
performance.
The diﬀerences of the obtained mean lift coeﬃcient are signiﬁcantly smaller
than the diﬀerences in lifttodrag ratios. Therefore, the conclusions on the per
formance comparison are considered to be signiﬁcant.
The mean drag for the harmonic and Roboﬂy models is substantially higher
106 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
t/T []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 coeﬃcient histories of the baseline kinematic models. •: harmonic model,
◦: Roboﬂy model, : realistic fruit ﬂy model, ▽: simpliﬁed fruit ﬂy model.
t/T []
D
r
a
g
c
o
e
ﬃ
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
ﬃ
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 coeﬃcient histories of the baseline kinematic models. •: harmonic model;
◦: Roboﬂy model, : realistic fruit ﬂy model, ▽: simpliﬁed fruit ﬂy 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%
Roboﬂy 1.417 −8.0% 2.466 2.448 0.577 −49%
realistic fruit ﬂy 1.540 baseline 1.387 1.335 1.132 baseline
simpliﬁed fruit ﬂy 1.454 −5.6% 1.012 1.596 1.115 −1.5%
Table 5.1 Timeaveraged force coeﬃcients 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 halfstroke. (a) harmonic model, (b) Roboﬂy model,
(c) realistic fruit ﬂy model, (d) symmetric fruit ﬂy model.
108 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
(a) harmonic model (b) realistic fruit ﬂy model
Figure 5.11 Vorticity contours around a ﬂapping airfoil. Vorticity contours are shown for
t = 0.1T (blue: clockwise, corresponding to negative vorticity values).
compared to the fruit ﬂy models. This is also illustrated in ﬁgure 5.8 and 5.9 (lift
and drag histories) and ﬁgure 5.10 (force vectors). Figure 5.11 shows the vorticity
contours of the realistic fruit ﬂy model compared with the harmonic model. It can
be seen in ﬁgure 5.7(a) that the eﬀective angle of attack is higher in the harmonic
case, compared to the realistic fruit ﬂy model, ﬁgure 5.7(c). Therefore, the mean
drag contribution of the leadingedge vortices (LEV) is higher. The decrease in
eﬀective angle of attack in the realistic fruit ﬂy model is also enlarged by the
presence of the ‘bump’. This drag increasing eﬀect is even larger in case of the
Roboﬂy model due to the ‘trapezoidal’ angle of attack. The ‘sawtooth’ shaped
Roboﬂy displacement could possibly play an important role as is discussed in the
next section. The diﬀerent kinematic patterns are also illustrated in ﬁgure 5.10,
which shows the resultant force vectors during a full stroke for those baseline
kinematic models.
Furthermore, the mean drag coeﬃcient of the simpliﬁed fruit ﬂy 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 coeﬃcient
obtained with the realistic fruit ﬂy model.
When comparing the lifttodrag ratios in table 5.1, it can be observed that
within the model assumptions, the fruit ﬂy models perform better than the less
complex models. Compared to the harmonic model, the realistic fruit ﬂy model
results in a signiﬁcant decrease in drag of 29% at comparable lift. The diﬀerence
with the Roboﬂy model is even larger, 49%. These performance increases are the
result of the lower drag coeﬃcients in both fruit ﬂy models due to certain beneﬁcial
5.4 Results and Discussion 109
kinematic features. The current results provide insight into the eﬀects of certain
speciﬁc kinematic features. However, one has to be cautious when extrapolating
these results to real ﬂying ﬂies since in reality not every ﬂapping period displays
exactly the same kinematic proﬁle. Next, the individual inﬂuences of the diﬀerent
interesting kinematic shapes are studied.
5.4.2 Kinematic features investigation
Inﬂuence of ‘sawtooth’ displacement used by the Roboﬂy
The ‘sawtooth’ shaped displacement of the Roboﬂy is investigated in isolation
to assess its inﬂuence on the force histories and the aerodynamic performance.
Therefore, the purely harmonic model was appended with the Roboﬂy 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 ﬁgure 5.13. From ﬁgure 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 signiﬁcantly
larger for the ‘sawtooth’ case, see ﬁgure 5.13(b). This also explains the larger
mean drag compared to the harmonic model which can be read from table 5.2.
In ﬁgures 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 ﬁgure 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 halfstroke 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 reﬂected in the integrated values in table 5.2. Due
to this larger drag during each stroke, the ‘sawtooth’ shaped displacement leads
to a lower lifttodrag ratio, which shows a decrease of 24.3% with respect to the
harmonic case.
Inﬂuence of ‘trapezoidal’ angle of attack used by the Roboﬂy
In combination with the ‘sawtooth’ displacement, the Roboﬂy uses a ‘trapezoidal’
shape for the angle of attack. In order to determine the eﬀect 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 ﬁgure 5.12(a)
for the force vectors. The lift and drag coeﬃcients are plotted in ﬁgure 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 nonzero mean horizontal force along a complete
110 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
¯
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 halfstroke. (a) harmonic model with ‘sawtooth’ φ, (b)
harmonic model with ‘trapezoidal’ α, (c) harmonic model with ‘extra bump’ α, (d) harmonic model
with ‘deviation’ θ.
5.4 Results and Discussion 111
t/T []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 coeﬃcients. Lift (a) and drag (b) histories to investigate the inﬂuence
of the ‘sawtooth’ displacement compared to the harmonic model. •: harmonic φ,α,θ; ◦: harmonic α,
θ and Roboﬂy φ.
(a) harmonic model (b) harmonic model with ‘sawtooth’ displacement
Figure 5.14 Vorticity contours around a ﬂapping airfoil. Vorticity contours are shown for
t=0.1T (blue: clockwise, corresponding to negative vorticity values)
112 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
kinematic model C
L
C
Ddown
−C
Dup
C
L
/C
Dave
harm. φ, α and θ 1.483 (baseline) 1.848 1.839 0.804 (baseline)
harm. α, θ + Roboﬂy φ 1.366 (−7.9%) 2.240 2.250 0.608 (−24.3 %)
harm. φ, θ + Roboﬂy α 1.351 (−8.9%) 2.302 2.733 0.537 (−33.3 %)
harm. φ, θ + fruit ﬂy. α 1.483 (0.0%) 1.221 1.969 0.930 (+15.6 %)
harm. φ, α + fruit ﬂy. θ 1.323 (−10.8%) 1.807 1.776 0.738 (−8.2 %)
Table 5.2 Timeaveraged force coeﬃcients to investigate the inﬂuence of kinematic shapes.
Each characteristic shape is varied with respect to the harmonic motion model.
t/T []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 coeﬃcients. Lift and drag histories to study the inﬂuence of the
‘trapezoidal’ angle of attack compared to harmonic model. •: harmonic φ,α,θ; ◦: harmonic α, φ and
Roboﬂy α.
5.4 Results and Discussion 113
(a) harmonic model (b) harmonic model with ‘trapezoidal’ α
Figure 5.16 Vorticity contours around a ﬂapping airfoil. Vorticity contours are shown for
t=0.6T (blue: clockwise, corresponding to negative vorticity values)
stroke cycle. Although this model is symmetric, the force distributions are not,
since the complex vortex dynamics are nonlinear and asymmetric.
From ﬁgure 5.15 it is clear that at the beginning of a stroke the lift peak of the
‘trapezoidal’ case is larger. Using ﬁgure 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 ﬁgure 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 leadingedge, 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 ﬁgure 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 signiﬁcant performance decrease of 33.3% due to the ‘trapezoidal’ angle of
attack variation, see table 5.2.
Inﬂuence of extra ‘bump’ in angle of attack used by the fruit ﬂy
The fruit ﬂy models are subject to an extra ‘bump’ in angle of attack. To make
comparison plausible the symmetric ‘bump’ variation used in the simpliﬁed fruit
ﬂy model is used to compare results with the harmonic model. Figure 5.12(c)
114 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
(a) harmonic model (b) harmonic model with ‘trapezoidal’ α
Figure 5.17 Vorticity contours around a ﬂapping airfoil. Vorticity contours are shown for
t=0.4T (blue: clockwise, corresponding to negative vorticity values)
shows the force vectors during up and down stroke. In ﬁgure 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 signiﬁcantly. However, the drag during the downstroke
is very much aﬀected. 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
lifttodrag ratio is still increased with more than 15.6%. From ﬁgure 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
ﬁgure 5.19(a) and (b) are compared. The decrease in eﬀective angle of attack as
a result of the ‘bump’ is considerable compared to the harmonic case. The same
was found for the Roboﬂy 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.
Inﬂuence of wing deviation used by the fruit ﬂy
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
ﬃ
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
ﬃ
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 coeﬃcients. Lift (a) and drag (b) histories to study the inﬂuence of
the extra ‘bump’ in angle of attack. •: harmonic φ,α,θ; ◦: harmonic α, φ and fruit ﬂy α.
(a) harmonic model (b) harmonic model with ‘extra bump’ in α
Figure 5.19 Vorticity contours around a ﬂapping airfoil. Vorticity contours are shown for
t=0.1T (blue: clockwise, corresponding to negative vorticity values)
116 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
(a) harmonic model (b) harmonic model with ‘extra bump’ in α
Figure 5.20 Vorticity contours around a ﬂapping airfoil. Vorticity contours are shown for
t=0.6T (blue: clockwise, corresponding to negative vorticity values)
realistic and simpliﬁed fruit ﬂy model. This deviation causes a ‘ﬁgureofeight’
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 eﬀective angle of attack is highly aﬀected. This motion
perpendicular to the stroke plane is illustrated in ﬁgure 5.12(d) which also shows
the force vectors.
Figure 5.21 shows the force coeﬃcients during one ﬂapping period with de
viation added to the harmonic model. The mean lift and drag are not strongly
inﬂuenced by the deviation, see table 5.2. The mean lift is decreased by 10.8% and
the mean drag is almost not aﬀected by the presence of deviation, about 2%−4%
diﬀerence in both strokes. It is also revealed that the force distributions remain
symmetric.
The large inﬂuence 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 ﬂow dynamic mechanism for this is shown in the vorticity visualisations
of ﬁgure 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 inﬂuence of the deviation is relatively large since the deviation increases the
eﬀective 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 eﬀective angle of
5.4 Results and Discussion 117
t/T []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 coeﬃcients. Lift (a) and drag (b) histories to study the inﬂuence of
the deviation compared to harmonic model. •: harmonic φ,α,θ; ◦: harmonic α, φ and fruit ﬂy θ.
(a) harmonic model (b) harmonic model with deviation
Figure 5.22 Vorticity contours around a ﬂapping airfoil. Vorticity contours are shown for
t=0.1T (blue: clockwise, corresponding to negative vorticity values).
118 Inﬂuence of wing kinematics in twodimensional hovering ﬂight
(a) harmonic model (b) harmonic model with deviation
Figure 5.23 Vorticity contours around a ﬂapping airfoil. Vorticity contours are shown for
t=0.6T (blue: clockwise, corresponding to negative vorticity values).
attack. Figure 5.23(a) and (b) show LEV’s of comparable strength for both cases.
Summarising, the deviation is levelling the force distributions while the mean
lift and drag are almost unaﬀected. This leads to the suggestion that a fruit ﬂy
may use the deviation to level the wing loading over a ﬂapping cycle. Three
dimensional studies are needed to investigate to what extent this eﬀect is also
present in real insect ﬂight.
5.5 Conclusions
The eﬀect of wing motion kinematics on the aerodynamic characteristics of hov
ering insect ﬂight was investigated by means of twodimensional numerical ﬂow
simulations. The results of the present twodimensional study may provide useful
insights in the understanding of real threedimensional insect ﬂight (Wang et al.,
2004).
Four diﬀerent kinematic models, with diﬀerent complexity, have been anal
ysed using twodimensional timedependent NavierStokes simulations. Two of
these models, pure harmonic motion and Roboﬂy experimental kinematics have
appeared in literature. The third model represents the actual fruit ﬂy kinematics
as observed in experiments and the last one is a modiﬁcation of the latter, chosen
to investigate the eﬀect of symmetry. The most prominent aspects of the Roboﬂy
kinematic model are the ‘sawtooth’ displacement and the ‘trapezoidal’ angle of
attack. The fruit ﬂy 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 quasisteady lift
5.5 Conclusions 119
coeﬃcient was matched.
It was found that the realistic fruit ﬂy wing kinematics result in signiﬁcantly
lower drag at similar lift compared with the simpliﬁed wing kinematic models
used in literature. The trend that the fruit ﬂy kinematics increases aerodynamic
performance agrees well with the predictions of the quasisteady theory, but the
numerical ﬂow simulations provide a more complete quantitative analysis of the
ﬂow 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 diﬀerence in performance in terms of mean lifttodrag ratio
between the diﬀerent kinematic models was signiﬁcant. The mean aerodynamic
drag at equal lift of the fruit ﬂy models is about 49% lower compared to the Roboﬂy
model and about 29% lower with respect to the harmonic model. Therefore, the
eﬀect 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 Roboﬂy model
has a small eﬀect on the mean lift but the mean drag is aﬀected signiﬁcantly. 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
pliﬁcation used by the Roboﬂy, the ‘trapezoidal’ angle of attack, caused the LEV
to separate during the translational phase. This led to an increase in mean drag
during each halfstroke. Also in this case large accelerations at stroke reversal lead
to a decrease in lifttodrag ratio of 33.3%.
The extra ‘bump’ in angle of attack, as used by the fruit ﬂy model, is not
aﬀecting 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 signiﬁcant decrease in drag which improves
aerodynamic performance in the sense of lifttodrag ratio by 15.6%. The other
realistic kinematic feature is the deviation, which is found to have only a marginal
eﬀect on the mean lift and mean drag. However, the eﬀective 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 ﬂight
have an appreciable eﬀect on the accuracy of performance models of insect ﬂight.
In particular they indicate that kinematic features, found in fruit ﬂy 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
twodimensional forward ﬂapping
foil
AIAA paper 2009791.
A twodimensional numerical investigation is performed to study the vortical ﬂow
around a ﬂapping foil that models an animal wing, ﬁn, or tail in forward motion.
The vortex dynamics and performance are studied to determine the inﬂuence of
foil kinematics. The baseline kinematic model is prescribed by harmonic functions
which can be characterised by four variables, the dimensionless wavelength, the
dimensionless ﬂapping amplitude, the amplitude of geometric angle of attack, and
the stroke plane angle. The foil motion kinematics has a strong inﬂuence on the
vortex dynamics, in particular on the vortexwake 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 leadingedge vortex
(LEV) is not signiﬁcantly advantageous for the force enhancement during the
full stroke. Plots of eﬃciency versus the independent variable show that, for
symmetric kinematics, the largest eﬃciency is achieved at an intermediate value
of each variable within the parameter range considered, where periodic ﬂow occurs.
122 Vortex wake interactions of a twodimensional ﬂapping foil
6.1 Introduction
The ﬂow around a ﬂapping wing, ﬁn, or tail is highly unsteady and governed by the
dynamics of the generated vortices (WeishFogh & 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 speciﬁc 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 ﬂow is the result of a match between the driving frequency
and the natural shedding frequency which is referred to synchronisation of the
ﬂow (Williamson & Roshko, 1988, Lentink & Gerritsma, 2003). The wake will be
aperiodic if synchronisation of the vortexwake does not occur. The synchronisa
tion band organisation for the ﬂapping 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 ﬂapping foil displays a variation of the natural shedding
frequency as a function of angle of attack (Katz, 1981, Dickinson & G¨otz, 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 timeaveraged lift over a complete ﬂapping
period. Several studies have shown that the wing beneﬁts 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 ﬂapping motion, plunging and pitching, at a Reynolds
number of Re = 150 which corresponds to the ﬂight of a small insect, e.g. a fruit
ﬂy. Here only the near wake of the foil is studied. The motivation for this is
that performance of a ﬂapping foil is inﬂuenced mainly by near wake dynamics.
The objective of the present study is to investigate the inﬂuence of diﬀerent foil
kinematics on the vortexwake structure, force coeﬃcients, and performance.
6.2 Flapping foil parametrisation
The baseline kinematic model is based on harmonic motion, such as used by Wang
(2000b,a). The ﬂow around a ﬂapping 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 inﬂuence of diﬀerent foil kinematic parameters on the vortexwake
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
α
(
◦
),
the forward velocity of the foil U
∞
[m/s], the chord length of the foil c [m] and
the stroke plane angle β (
◦
). The deﬁnition of the dimensionless parameters is
schematically illustrated in ﬁgure 6.1 and described in more detail below. The di
mensionless wavelength λ
∗
represents the number of chord lengths travelled during
one ﬂapping period:
λ
∗
=
U
∞
fc
.
The dimensionless amplitude A
∗
represents the ratio of amplitude of the foil trans
lation and the chord length of the foil:
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
∗
and the dimensionless wavelength
λ
∗
:
St =
fA
U
∞
=
A
∗
λ
∗
.
It corresponds to the maximum induced angle of attack A
α
ind
at midstroke due
to the translation of the ﬂapping motion of the foil. The mean velocity U [m/s] is
obtained by averaging the velocity components over one ﬂapping period:
U =
1
T
_
T
0
_
(U
∞
+U
ﬂap
X
)
2
+ (U
ﬂap
Y
)
2
dt .
Here T [s] is the period, U
ﬂap
X
[m/s], and U
ﬂap
Y
[m/s] the velocity of the foil kine
matics in X and Y directions respectively. The timeaveraged Reynolds number
Re becomes
Re =
Uc
ν
,
where ν [m
2
/s] is the kinematic viscosity and changed for every computation to
match the Reynolds number. For the basic model the dimensionless wavelength λ
∗
,
the dimensionless amplitude A
∗
, the amplitude of geometric angle of attack A
α
,
and the stroke plane angle β are chosen as independent variables. The Reynolds
number is kept constant at Re = 150.
In order to study the inﬂuence of the kinematics, each parameter is varied from
the baseline model, deﬁned by λ
∗
= 6.8, A
∗
= 1.5, A
α
= 15
◦
, and β = 90
◦
. The
dimensionless wavelength was varied from λ
∗
= 24, 20, 12, 10, 7.9, 6.8, 6.3, 6.0,
5.7, 5.3, 4.5, 4.0 to 3.0. The dimensionless amplitude is varied within the range of
0.5 ≤ A
∗
≤ 3.0 with a 0.5 increment. The amplitude of angle of attack varies from
0
◦
≤ A
α
≤ 45
◦
with a 15
◦
increment. The inﬂuence of the stroke plane angle is
124 Vortex wake interactions of a twodimensional ﬂapping foil
λ
∗
90
◦
−β
A
α
(tan
−1
)2Stsin(β)
2A
∗
· sin(β)
(a)
y
x
Y
X
β
β
2A
U
∞
(b)
Figure 6.1 Schematic illustration of the foil kinematics in forward ﬂight. (a) Illustration
of the foil parameters in forward ﬂight: the dimensionless wavelength λ
∗
, the dimensionless amplitude
A
∗
, the Strouhal number St, the angle amplitude A
α
, and the stroke plane angle β. In this frame of
reference the observer is ﬁxed relative to the undisturbed air. The ﬂight direction is from right to left.
(b) The twodimensional relation between two inertial coordinate systems. The downstroke phase is
ﬁlled by dark blue and the upstroke by light blue.
investigated for two diﬀerent angle amplitudes A
α
= 15
◦
and 45
◦
in combination
with 15
◦
≤ β ≤ 90
◦
with a 15
◦
increment. In the cases that the stroke plane angle
diﬀers from 90
◦
, the resulting ﬂapping motion is asymmetric.
6.3 Force coeﬃcients and performance
In this study two inertial coordinate systems are used, see ﬁgure 6.1. The XY 
plane has the Xaxis in the direction of the freestream velocity and the Y axis in
vertical direction. The xyplane is tilted at an angle β, the stroke plane angle. The
lift force L is the component of the total aerodynamic force perpendicular to the
forward velocity of the foil and is positive when it is in the positive Y direction.
The drag force D is the component of the total aerodynamic force parallel to
the forward velocity of the foil and is positive when directed in the positive X
direction. In the present study, the force and moment coeﬃcients are scaled using
the average dynamic pressure q [N/m
2
]:
q =
1
2
ρU
2
.
Using q, the force and moment coeﬃcients are deﬁned as:
C
D
=
D
qc
, C
L
=
L
qc
, C
M
=
M
qc
2
.
Projecting the lift coeﬃcient C
L
and drag coeﬃcient C
D
onto the y and xaxes
we obtain the foil lift coeﬃcient C
l
and the foil drag coeﬃcient C
d
respectively.
C
l
= C
L
· sinβ +C
D
· cosβ ,
6.4 Numerical model 125
C
d
= −C
L
· cosβ +C
D
· sinβ .
Note that a negative drag coeﬃcient C
D
means thrust which is necessary in forward
locomotion whereas the foil drag coeﬃcient C
d
indicates the ﬂuid force that the
animal must overcome for translational motion of its wing, ﬁn, or tail and is
relevant to the required power of locomotion.
The comparative assessment of the aerodynamic performance of the diﬀerent
kinematic models is based on the mechanical eﬃciency of the foil motion. The
eﬃciency η [%] is the ratio between the eﬀective propulsive power P
eﬀ
[Nm/s]
and the required power P
req
[Nm/s] which are given in (6.1), (6.2), and (6.3)
respectively:
η =
P
eﬀ
P
req
· 100% , (6.1)
where
P
eﬀ
= −D · U
∞
, (6.2)
P
req
= −
1
T
_
T
0
d · U
ﬂap
dt −
1
T
_
T
0
M · ω
ﬂap
dt . (6.3)
Here −D [N] represents thrust, U
∞
[m/s] the freestream velocity, T [s] the ﬂapping
period, d [N] the foil drag, U
ﬂap
[m/s] the translational velocity of the foil in the
stroke plane, M [N·m] the moment about the centre of rotation and ω
ﬂap
[rad/s]
represents rotational velocity of the foil. Note that we have neglected inertial cost
of mechanical work done by the foil.
6.4 Numerical model
In the present thesis, a 2% thick ellipsoid shape with unit chord length represents
the foil. In order to obtain a good quality mesh, elliptical coordinates (µ, θ) are
used following Bos et al. (2008). The constant µ and θ correspond to confocal el
lipses and hyperbolas respectively. These elliptical coordinates can be transformed
to Cartesian coordinates via a conformal mapping:
x +iy = cosh(µ +iθ) .
The result of this conformal mapping can be seen in ﬁgure 6.2. The inner Otype
mesh of 50000 cells is surrounded by a ring of tetrahedral cells. The inner mesh
is able to move, whereas the outer mesh is remeshed every timestep. The radius
of the inner computational domain is chosen to be 25 chord lengths, such that
the inﬂuence of far ﬁeld boundary condition can be neglected. A uniform grid in
(µ, θ) is concentrated around the leading and trailing edges. This type of grid is
suitable for the problem since the vorticity is strongest near the edge of the foil.
Those twodimensional simulations are performed on mesh resolutions of about
50000 cells, more information on this mapping can be found in (Bos et al., 2008,
Wang, 2000a).
126 Vortex wake interactions of a twodimensional ﬂapping foil
X
Y 0
0
0
0
0 0 0 0
10
10
20
20
30
30
3
3
2
2
1
1
(a) (b)
Figure 6.2 The body conformal moving mesh around a 2% ellipsoid foil. (a) The Otype
body conformal mesh with a grid size of 50000 cells is moving within a ring of tetrahedral cells. (b)
The closeup of the mesh at the foil surface shows that the grid is concentrated around the leading and
trailing edge.
6.5 Results and discussion
The simulations start with the ﬂuid at rest in which the initial velocity vector is
zero. The resulting wake patterns have been classiﬁed using a symbolic code of
letters and numbers developed by Williamson & Roshko (1988) that describes the
combination of pairs (P) and single (S) vortices shed during each ﬂapping cycle.
The moment when a LEV is shed from the wing is deﬁned as the moment when
its core passes the trailing edge. The averaged aerodynamic force coeﬃcients in
table 6.1, 6.2, 6.3, and 6.4 are obtained using three simulation periods. Note that
the eﬃciency is only calculated when the drag is negative, i.e. thrusting mode.
6.5.1 Inﬂuence of dimensionless wavelength
The wake pattern and vortex behaviour are studied as a function of the dimension
less wavelength in the range of 24 ≥ λ
∗
≥ 3. The numerical results are provided
in table 6.1 by decreasing dimensionless wavelength which is equivalent to an in
crease in ﬂapping frequency at a constant ﬂight velocity. Our results are similar to
the experimental results found by Lentink et al. (2008) using a soapﬁlm tunnel.
At high dimensionless wavelengths λ
∗
= 24 and 20 the numerical results give no
strong vortices shedding from the foil in relation to the foil oscillation. Thus the
lift and drag are a function of the eﬀective angle of attack A
α
eﬀ
which leads to
positive drag. Thrusting modes are found for 12 ≥ λ
∗
because of generated LEV’s
which pull the foil toward in forward direction. For dimensionless wavelengths
12 ≥ λ
∗
≥ 5.7 the LEV’s are shed before stroke reversal. The amount of LEV’s
6.5 Results and discussion 127
λ
∗
pattern C
L
C
Ldownstroke
C
Lupstroke
C
D
η
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 Inﬂuence of dimensional wavelength. The numerical results are shown for 13 diﬀerent
values for the dimensionless wavelength, A
∗
= 1.5, A
α
= 15
◦
, and β = 90
◦
.
and TEV’s shed from the foil is decreasing with the dimensionless wavelength be
cause the vortices have less time to develop and shed, see ﬁgure 6.3. Therefore,
the LEV’s stay attached to the foil relatively longer at lower dimensionless wave
lengths. The LEV’s increase in size and strength due to increasing eﬀective angle
of attack. As a result of this, the foil produces higher lift and thrust during each
halfstroke for decreasing dimensionless wavelength. A further decrease in dimen
sionless wavelength results in stronger vortexwake interactions which lead to an
aperiodic wake at λ
∗
= 3, so that the forces of this case are varying with relative
small changes from period to period. The asymmetry in the lift coeﬃcient is a
result of the asymmetry in wake pattern. It is observed for cases when vortices are
formed on the foil that the lift changes its direction before stroke reversal. This
means that the foil cannot produce lift enhancement just before the end of each
halfstroke whether the LEV is shed before or after stroke reversal, see ﬁgure 6.4.
Therefore the foil does not fully beneﬁt from the attachment of LEV’s.
6.5.2 Inﬂuence of dimensionless amplitude
Six dimensionless amplitudes are chosen to investigate the inﬂuence of this pa
rameter. In table 6.2 the numerical results are given for the six cases. At low
dimensionless amplitude A
∗
= 0.5, no vortices are formed on the foil due to the
low eﬀective angle of attack. Nevertheless, shear layers from the foil organise
themselves into a 2S pattern. As a result, the force distributions have a sinusoidal
shape because the foil cannot produce force enhancement. For medium dimen
sionless amplitude A
∗
= 1.0 and 1.5 the eﬀective angle of attack is high enough to
128 Vortex wake interactions of a twodimensional ﬂapping foil
(a) No vortices, λ
∗
= 24, A
α
eff
= 7
◦
(b) 2P+2S, λ
∗
= 6.8, A
α
eff
= 39
◦
(c) 2P+S, λ
∗
= 6.0, A
α
eff
= 43
◦
(d) 2P, λ
∗
= 5.3, A
α
eff
= 46
◦
(e) P+S, λ
∗
= 4.5, A
α
eff
= 49
◦
(f ) Aperiodic, λ
∗
= 3, A
α
eff
= 53
◦
Figure 6.3 Vorticity contours for decreasing wavelength. Vorticity contours of various wake
patterns for decreasing dimensionless wavelength λ
∗
. The ﬂow is from left to right. All images are
taken at t = 0.35T. A
∗
= 1.5, A
α
= 15
◦
, and β = 90
◦
.
A
∗
pattern C
L
C
Ldownstroke
C
Lupstroke
C
D
η
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 diﬀerent dimensionless amplitudes. λ
∗
= 6.8,
A
α
= 15
◦
, and β = 90
◦
.
6.5 Results and discussion 129
t/T []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 coeﬃcients to study the inﬂuence of wavelength. Lift (a) and drag (b)
histories of diﬀerent wake patterns for six dimensionless wavelengths. A
∗
= 1.5, A
α
= 15
◦
, and
β = 90
◦
.
A
α
pattern C
L
C
Ldownstroke
C
Lupstroke
C
D
η
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 diﬀerent angle amplitudes. λ
∗
= 6.8, A
∗
= 1.5,
and β = 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 foilvortex
interactions lead to an aperiodic wake pattern causing aperiodic force coeﬃcients.
6.5.3 Inﬂuence of angle of attack amplitude
Table 6.3 shows numerical results for diﬀerent angle of attack amplitudes. The
plunging kinematic model, A
α
= 0
◦
, results in an asymmetric 2P+S pattern. The
LEV in the upstroke is weaker than those generated in the downstroke, which gives
positive mean lift over a period. No thrust is generated for this setting. Once the
foil is allowed to rotate, nonzero angle of attack amplitude, the eﬀective angle of
attack is lower. This results in decreasing lift in each halfstroke for increasing
angle amplitude. However, the reverse trend is found for thrust. The foil rotation
leads to thrust generation due to the frontal surface area for the pressure diﬀerence
acting toward in forward direction (Lentink & Gerritsma, 2003).
A peak performance of 28.18% is obtained which is considerably larger com
130 Vortex wake interactions of a twodimensional ﬂapping foil
β A
α
pattern C
L
C
Ldownstroke
C
Lupstroke
C
D
η
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 Inﬂuence of the stroke plane angle. Numerical results of the kinematics for six diﬀerent
stroke plane angles in combination with two diﬀerent angle amplitudes. λ
∗
= 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
α
= 45
◦
, no signiﬁcant vortices are formed on the foil because the eﬀective
angle of attack is low A
α
eﬀ
= 9
◦
at midstroke. However, shear layers which are
generated by the foil, form themselves into a 2P pattern. At this low eﬀective
angle of attack the foil produces lower lift and thrust.
6.5.4 Inﬂuence of stroke plane angle
The stroke plane angle causes an asymmetry in the kinematics. Here the results
for the inﬂuence of the stroke plane angle with two diﬀerent angle amplitudes
A
α
= 15
◦
and A
α
= 45
◦
are shown in table 6.4. From the baseline kinematics the
stroke plane angle is tilted backward by 15
◦
.
A similar trend is found for both angle amplitudes that the lift coeﬃcient
is increasing for decreasing stroke plane angle during the downstroke until the
ﬂow 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 diﬀerence in relative velocity between up
and downstroke also aﬀects the drag contribution in a similar way.
6.5.5 Discussion
In the symmetric kinematics, nonzero 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 coeﬃcient of symmetric kinematics.
λ
∗
A
∗
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) Vortexwake synchronisation A
∗
− λ
∗
dia
gram.
Figure 6.5 Inﬂuence of the kinematics on the vortex wake pattern and force generation.
(a) The mean lift coeﬃcient over a complete period of symmetric kinematics as a function of wake
pattern. (b) Vortexwake synchronisation A
∗
−λ
∗
diagram of our sinusoidal ﬂapping wing. The angle
amplitude and stroke plane angle are kept constant at A
α
= 15
◦
and β = 90
◦
. The dash line represents
our theoretical estimate of the boundary governed by equation (6.4). We have added the operating
conditions of insects belonging to the order Diptera.
conditions. The results are summarised in ﬁgure 6.5(a), where the timeaveraged
lift coeﬃcient is plotted against wake pattern. The grey bands indicate a sym
metric wake pattern in which the nearly zero mean lift is obtained. The vortex
synchronisation diagram for all models is shown in ﬁgure 6.5(b). There is an im
portant limitation in forward ﬂapping locomotion. To begin with, the (absolute)
eﬀective angle of attack should be high enough to form a LEV in order to generate
force enhancement. This approximately restricts the values of St, i.e. the ratio of
A
∗
and λ
∗
as:
St =
A
∗
λ
∗
>
1
2π
· tan(A
α
geo
+A
α
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 inﬂuence the stability and controllability of the MAV’s. Fig
ure 6.6 shows plots of eﬃciency versus the independent motion parameters. In
symmetric kinematics there is an optimal value for each variable, see ﬁgure 6.6(a),
(b) and, (c). The peak eﬃciency of 28.18% could conﬁrm that the wing rotation
plays an important role in the unsteady aerodynamic force production.
132 Vortex wake interactions of a twodimensional ﬂapping foil
λ
∗
η
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
∗
η
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
α
η
2
P
2
P
+
2
S
0 10 20 30 40 50
0
10
20
30
(c)
β
η
A
α
= 15
◦
A
α
= 45
◦
2P
2
P
+
S
2P+2S
60 65 70 75 80 85 90 95 100
0
5
10
15
(d)
Figure 6.6 Inﬂuence of ﬂapping kinematics on the eﬃciency. Eﬃciency 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 twodimensional ﬂow was used to investigate the eﬀect of
foil kinematics on the vortex dynamics around an ellipsoid foil subjected to pre
scribed ﬂapping 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 ﬂapping 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 ﬂow 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 beneﬁcial for understanding the inﬂuence
of wing kinematics on the performance characteristics.
CHAPTER 7
Vortical structures in
threedimensional ﬂapping ﬂight
Submitted to J. Fluid Mech. (January 2010).
Results are obtained by performing numerical simulations of the threedimensional
ﬂow around a ﬂapping wing. A parameter study is performed to investigate the
performance in ﬂapping ﬂight and to get insight into the vortex dynamics and force
generation. Diﬀerent aspects, relevant for threedimensional ﬂapping wing aero
dynamics, have been studied, namely the angle of attack, the Rossby number, the
Reynolds number and the stroke kinematics. First, the ﬂow around a dynamically
scaled model wing is solved for diﬀerent angles of attack in order to study the force
development and vortex dynamics at small and large midstroke angles of attack.
Secondly, the Rossby number is varied at diﬀerent 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 threedimensional wing
kinematics is varied by changing the shape in angle of attack and by applying a
deviation, which may result in a ‘ﬁgureofeight’, a ‘ﬁgureofO’ or a ‘ﬁgureofU’
pattern. Finally, the threedimensional ﬂow is compared with the twodimensional
studies performed on ﬂapping forward ﬂight.
7.1 Introduction
To understand the aerodynamic performance of ﬂapping wings at low Reynolds
numbers, relevant for insect ﬂight, it is important to obtain insight into the vortex
136 Vortical structures in threedimensional ﬂapping ﬂight
dynamics and its inﬂuence on force development. The most important feature
in ﬂapping wing aerodynamics has been established to be the generation of a
stable leadingedge vortex (LEV) on top of the wing, which is responsible for the
unexpectedly large force augmentation in hovering insect ﬂight (Maxworthy, 1979,
Ellington et al., 1996, Dickinson et al., 1999, Srygley & Thomas, 2002, Lentink
& Dickinson, 2009b). In order to gain insight into the threedimensional ﬂow
ﬁeld induced by the ﬂapping wings, several twodimensional studies have been
performed (Dickinson & G¨otz, 1993, Dickinson, 1994, Wang, 2005, Bos et al., 2008).
It was shown that the leadingedge vortex generated by a twodimensional moving
foil is shed after several travelled chord lengths, while a threedimensional LEV
remains stably attached to a threedimensional revolving (Usherwood & Ellington,
2002) or ﬂapping (Dickinson et al., 1999, Lehmann, 2004, Birch et al., 2004) model
wing, which rotates around its base. Those results indicate that threedimensional
ﬂow eﬀects are essential for the LEV stability. Previously conducted research
addressed a possible analogy between the LEV on ﬂapping wings and the LEV
generated by swept and delta wings (Ellington et al., 1996, Van Den Berg &
Ellington, 1997). The spiral leadingedge vortex generated by a translating swept
or delta wing is stabilised by the induced spanwise ﬂow, which could suggest
that a spanwise ﬂow may play an important role concerning the LEV stability in
insect ﬂight (Ellington et al., 1996, Van Den Berg & Ellington, 1997). Lentink
& Dickinson (2009b) discussed that the stability of the LEV growth speciﬁcally
might be increased by the spanwise ﬂow through the LEV core, driven by either
the dynamic pressure gradient on the wing’s surface, the centrifugal acceleration of
the boundary layer or the induced velocity ﬁeld of the spiral vortex lines (Ellington
et al., 1996). Additionally, the LEV stability may be strengthened by a reduction
of the eﬀective angle of attack as a result of the tip vortex generation (Birch &
Dickinson, 2001, Shyy et al., 2008b). However, Birch & Dickinson (2001) showed
no signiﬁcant eﬀect of the spanwise ﬂow on the LEV strength and stability, using
plates at diﬀerent spanwise locations to block the spanwise ﬂow, but they did not
completely explain the LEV stability in their experiments.
In order to investigate the vortex dynamics and the stability of the leadingedge
vortex in particular, an accurate simulation method is developed to perform a CFD
simulation of a threedimensional ﬂapping 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 threedimensional
wing was modelled which was able to ﬂap around a base of which the location
can be varied. By varying the location of the centre of rotation, the inﬂuence of
the revolving strength (Rossby number) and the eﬀect 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 twodimensional sim
ulations (Bos et al., 2008) suggested that the wing kinematics may also have a
large inﬂuence on the ﬂapping performance in threedimensional hovering. Addi
tionally, (Lentink, 2008) showed interesting results concerning the stability of the
threedimensional leadingedge vortex depending on the Rossby number (equiv
7.2 Threedimensional ﬂapping wing simulations 137
F
X
F
Y
F
Z
θ(t)
φ(t)
α(t)
O
R
root
R
tip
start downstroke
start upstroke
midstroke
Stroke plane
F
drag
F
span
F
normal
Figure 7.1 Illustration of the wing motion and force deﬁnitions. Illustration of the wing
motion and force deﬁnitions. φ(t) corresponds to the stroke variation, α(t) to the geometrical angle of
attack and θ(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 diﬀerent 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 nonzero deviation
is applied. Diﬀerent deviation patterns are investigated, following the shape of
‘ﬁgureofO’, ‘ﬁgureofU’ and ‘ﬁgureofeight’.
The ﬂapping 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 eﬃcient, section 7.3 brieﬂy discusses the
validation and veriﬁcation of the ﬂow solver. The vortical ﬂow needs to be visu
alised in such a way that the resulting vortices are clearly visible. Diﬀerent vortex
identiﬁcation 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 Threedimensional ﬂapping wing simulations
In order to study the vortex dynamics and stability of the leadingedge vortex, the
ﬂow is solved using Computational Fluid Dynamics (CFD), of which the details
are described in chapter 2. The threedimensional ﬂapping wing is modelled in
order to provide a framework for comparison, which is still representative for true
insect ﬂight, 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 ﬂow phenomena of our interest. Additionally, the simulation
strategy is discussed in section 7.2.2.
138 Vortical structures in threedimensional ﬂapping ﬂight
7.2.1 Modelling and parameter selection
In general, most investigations concerning ﬂapping 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 ﬂy 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 inﬂuence on the forces and the ﬂow ﬁeld. In
addition, Luo & Sun (2005) showed that the airfoil corrugation, present in dragon
ﬂy wings, did not inﬂuence the force development signiﬁcantly. The length scales
of the corrugation are orders of magnitude smaller compared to the length scale
of the separated ﬂow region or the leadingedge vortex, such that signiﬁcant eﬀect
of corrugation on the ﬂow can be neglected.
Planform selection
The single wing span is ﬁxed to b
s
= 2.0 and the chord at midspan is c = 1.0.
The hinge around which the wing is able to ﬂap is ﬁxed 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 Roboﬂy was ﬁxed to 0.7 (Sane & Dickinson, 2002, Poelma
et al., 2006). Since the wing planform is chosen to be ellipsoidal, the wing sur
face is deﬁned by S = πab, where a = 0.5 and b = 1.0 are the semiminor and
semimajor axes, respectively, such that S = π/2. The average chord length of
this ellipsoidal planform is found to be c = S/b
s
= π/4. So, the three geomet
ric parameters important for the ﬂapping wing simulations are deﬁned: b
s
, S and c.
Kinematic models
The ﬂapping wing motion is prescribed by three diﬀerent motion angles deﬁning
the deviation angle θ(t), ﬂapping angle φ(t) and the angle of attack α(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
θ
= 0 and A
θ
= 20
◦
. A combined devi
ation and ﬂapping angle variation leads to a wing tip pattern. Depending on the
variation of the deviation angle, ﬁgure 7.2 shows the resulting ‘ﬁgureofO’, 7.2(a),
‘ﬁgureofeight’, 7.2(b), or ‘ﬁgureofU’, 7.2(c).
Realistic fruit ﬂy kinematics (Fry et al., 2003, Lentink & Dickinson, 2009b,
Bos et al., 2008) resembles an harmonically varying deviation angle, a ‘sawtooth’
shaped ﬂapping 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 twodimensional investigation (Bos et al.,
2008) showed that the eﬀect 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 threedimensional numerical study to investigate the eﬀects of the Rossby
number, Reynolds number, angle of attack and stroke kinematics, i.e. ‘trapezoidal’
7.2 Threedimensional ﬂapping wing simulations 139
Flapping angle, φ [
◦
]
D
e
v
i
a
t
i
o
n
a
n
g
l
e
,
θ
[
◦
]
[

]
0 20 40 60 80 100 120 140 160 180
80
60
40
20
0
20
40
60
80
upstroke
downstroke
(a) ‘ﬁgureofO’
Flapping angle, φ [
◦
]
D
e
v
i
a
t
i
o
n
a
n
g
l
e
,
θ
[
◦
]
[

]
0 20 40 60 80 100 120 140 160 180
80
60
40
20
0
20
40
60
80
upstroke
downstroke
(b) ‘ﬁgureofeight’
Flapping angle, φ [
◦
]
D
e
v
i
a
t
i
o
n
a
n
g
l
e
,
θ
[
◦
]
[

]
0 20 40 60 80 100 120 140 160 180
80
60
40
20
0
20
40
60
80
upstroke
downstroke
(c) ‘ﬁgureofU’
Figure 7.2 Diﬀerent wing tip patterns. Diﬀerent wing tip patterns as a result of the variation
in deviation with a combined ﬂapping motion. (a) ‘ﬁgureofO’. (b) ‘ﬁgureofeight’. (c) ‘ﬁgureofU’.
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 α = 15
◦
to α = 90
◦
with increments of α = 15
◦
. The current research
varies the amount of the ‘trapezoidal’ shape by varying the speed of rotation just
after stroke reversal from T
rot
= 0.10T to T
rot
= 0.25T, where T is the ﬂapping
period, such that T
rot
= 0.25T corresponds to fully harmonic angle of attack vari
ation. The ﬂapping angle was chosen to vary harmonically to isolate the eﬀects of
the deviation, the ‘trapezoidal’ angle of attack, the Rossby and Reynolds numbers.
Framework for comparison
In order to design a frame of comparison it is important to keep the following
three parameters ﬁxed: the dimensionless amplitude of the wing’s crosssection at
140 Vortical structures in threedimensional ﬂapping ﬂight
the radius of gyration A
∗
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
deﬁned as:
Re
R
g
=
4A
φ
fR
g
c
ν
, (7.1)
where A
φ
is the ﬂapping angle amplitude, f the ﬂapping frequency, c the average
chord length and ν the kinematic viscosity.
The kinematic viscosity is ﬁxed for three selected values, Re
R
g
= 100, 500 and
1000, provided that the wing kinematics and geometry are given. If the distance
of the rotation origin is varied, the wing tip radius changes, which is compensated
by the ﬂapping angle amplitude in order to keep the average Reynolds number
and the displacement at R
g
comparable. Therefore, the ﬂapping angle amplitude
is determined from (7.1), for every rotation radius.
The result of this scaling is a comparable average Reynolds number, Re
R
g
=
100, average velocity U
R
g
and displacement of the crosssection at the radius of
gyration, A
∗
R
g
= A
φ
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
φ
R/c. Concerning the baseline
case, with R
tip
= 2.5 the resulting amplitude of the crosssection at R
g
becomes
A
∗
R
g
≈ 2.2, which is of similar order as used in the twodimensional analysis in (Bos
et al., 2008).
In order to investigate the eﬀect of threedimensional wing kinematics in hov
ering ﬂight, which is the main subject of the present thesis, the hovering wing
kinematics is substituted into the expressions for the angular and centripetal co
eﬃcients, C
ang
and C
cen
, and the Rossby number Ro, equation (4.6) to ﬁnd the
following:
C
ang
=
2
π
A
φ
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 deﬁned by the wing geometry, whereas the
angular acceleration number C
ang
, depends on the wing kinematics.
Force and performance deﬁnitions
In order to determine the eﬀect of the diﬀerent motion and geometric parameters
on the forces and performance, proper deﬁnitions are necessary. Since the wing
rotates with a rotating reference frame, two diﬀerent force deﬁnitions are possible,
7.2 Threedimensional ﬂapping wing simulations 141
in the inertial and the rotating reference frame, which is shown in ﬁgure 7.1 in re
lation to the motion angles. Because the present research concerns hovering ﬂight,
the lift force is by deﬁnition 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 threedimensional
lift and drag are given by:
F
lift
= F
Y
, (7.5)
and
F
drag
= F
X
· sin(φ) −F
Z
· cos(φ). (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 coeﬃcients, C
L
and C
D
, are obtained by division
using the average dynamic pressure, q = 0.5ρU
2
ref
. Two performance indicators
are used, the lifttodrag 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 α, the
Reynolds number Re and the wing kinematic model. In order to investigate the in
ﬂuence of the wing kinematics, the shape of the angle of attack variation, reﬂected
by T
rot
and the deviation amplitude, A
θ
were varied. A systematic overview of all
cases is provided here.
Inﬂuence of wing stroke curvature
The stroke curvature is varied in order to investigate if there is a possible relation
between the angular acceleration, centripetal acceleration or the Rossby number
and the forces acting on the ﬂapping wing, under hovering conditions. The ﬂow is
solved for diﬀerent wing tip radii, equivalent to the Rossby number.
The wing tip radius is varied from fully revolving at R
tip
= 2.5 to a nearly
translating wing at R
tip
= 102. This range in stroke path curvature corresponds
to a changing Rossby number from Ro = 3.2 to 130. Lentink & Dickinson (2009b)
found that most insects and ﬁsh operate at a Rossby number close to Ro = 3.0,
which seems to be a biologically convergent solution for animals moving in ﬂuids.
Inﬂuence of Reynolds number and angle of attack
For two selected Rossby numbers Ro = 3.2 and Ro = 130, the midstroke angle
of attack is varied from α = 15
◦
to α = 90
◦
with increments of α = 15
◦
. This
provides insight in the force development as a function of angle of attack for a
fully revolving, i.e. ﬂapping, 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 eﬀects on the behaviour of the leadingedge vortex.
142 Vortical structures in threedimensional ﬂapping ﬂight
Rotation origin, R
tip
2.5 3.0 4.0 5.0 6.0 7.0 12.0 102.0
A
α
[
◦
]
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 ﬂapping wing. This matrix is used to
study the inﬂuence of the stroke curvature on the structure of the leadingedge vortex and corresponding
forces. The angle of attack at midstroke is varied from 90
◦
to 15
◦
together with the Reynolds number,
+: Re = 100, •: Re = 500, ◦: Re = 1000. The grid resolution was ﬁxed to 800k and the timestep
was chosen corresponding to Co
max
= 2.0.
Table 7.1 shows an overview of the variation of the wing stroke curvature, angle
of attack and Reynolds number.
Inﬂuence of the kinematic modelling
Besides the wing stroke curvature, Reynolds number and angle of attack, it remains
interesting to investigate the eﬀect of two kinematic parameters, the deviation,
which may cause a ‘ﬁgureofeight’ pattern and the ‘trapezoidal’ shape (deﬁned by
T
rot
) of the angle of attack variation. This ‘trapezoidal’ angle of attack variation
is deﬁned by
α =
_
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
_
A
α
sin(2πft) 0 ≤ t < T
rot
,
A
α
T
rot
≤ t <
1
2
T −T
rot
,
A
α
cos(2πft)
1
2
T −T
rot
≤ t < T
rot
+
1
2
T,
−A
α
T
rot
+
1
2
T ≤ t < T −T
rot
,
−A
α
cos(2πft) 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 diﬀerent values of T
rot
, the angle of attack is plotted
in ﬁgure 7.3. Note that the geometric angle of attack is given by α
geom
=
π
2
−α.
Table 7.2 shows diﬀerent deviation amplitudes A
θ
, in combination with a vary
ing, T
rot
, which determines the amount of ‘trapezoidal’ shape, as was already
illustrated in ﬁgure 7.2. The varying deviation angle amplitude, may cause dif
ferent wing tip patterns, i.e. ‘ﬁgureofO’, ‘ﬁgureofeight’ and the ‘ﬁgureofU’,
depending on the deviation frequency.
7.3 Flow solver accuracy 143
A
θ
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 ﬂapping 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
θ
. Two diﬀerent angle of attack amplitudes are used, 45
◦
and 60
◦
, which is
shown to result in maximal lift coeﬃcients. The Reynolds number is ﬁxed to Re = 100, the grid
resolution to 800k and the timestep was chosen corresponding to Co
max
= 2.0.
t/T []
α
[

]
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 inﬂuence
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 ﬂow solver, concerning highly unsteady
and vortical ﬂows, numerical comparisons are performed. A veriﬁcation 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
threedimensional simulations are constructed with GridPro
using a structured
approach. Grid reﬁnement is uniform and the cells are clustered close to the ﬂap
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 timestep indepen
dent, a veriﬁcation study is performed using the ﬂow around a threedimensionally
ﬂapping wing. The kinematics is according to the simple harmonic model. The
ﬂapping angle amplitude was ﬁxed to A
φ
= 63
◦
(1.1 rad), the midstroke angle
of attack was given by A
α
= 45
◦
and the wing tip radius corresponds to fully
revolving, the Rossby number was Ro = 3.2 and the average Reynolds number
Re
R
g
= 100. Table 7.3 gives an overview of the performed simulations by varying
144 Vortical structures in threedimensional ﬂapping ﬂight
Mesh resolution
100k 200k 400k 800k 1600k
C
o
m
a
x
2.0 + +
1.0 + + +• +• +
0.5 + +
Table 7.3 Simulation matrix: veriﬁcation. Veriﬁcation matrix showing the cases used for
veriﬁcation purposes, with varying mesh resolution (100k − 1600k) and timestep. The timestep is
reﬂected through the maximal Courant number, Co
max
, which varies from 2.0, 1.0 to 0.5. Two cases
are performed for two diﬀerent Reynolds numbers, +: Re = 100, •: Re = 1000. The kinematics of the
threedimensional wing is simple harmonics with A
φ
= 63
◦
, A
α
= 45
◦
and A
θ
= 0
◦
.
grid resolution and maximal Courant number. The smaller Co
max
, the smaller
the timestep. 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 ﬂow solver, the drag and lift coeﬃcients
are plotted in ﬁgure 7.4 for meshes from 100k to 1600k cells and Co
max
= 1.0. The
corresponding limit cycles are shown in ﬁgure 7.5. As can be clearly seen from
the ﬁgures, the ﬂow is periodic and the force coeﬃcients (lift and drag) appear
to be close for the grid resolutions considered. In order to assess spatial and
temporal convergence both timeaveraged lift and drag coeﬃcients are plotted with
increasing spatial and temporal resolution in ﬁgure 7.6. Figure 7.6(a) and 7.6(b)
shows a converging timeaveraged lift and drag coeﬃcients with increasing grid
resolution, the value at ∞ is determined by Richardson extrapolation (Ferziger &
Peric, 2002), see chapter 2. Temporal convergence is shown in ﬁgure 7.6(c), where
the timeaveraged lift and drag coeﬃcients are plotted with decreasing timestep.
In order to justify the choice for grid and temporal resolution for the three
dimensional ﬂapping 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 diﬀerences 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 suﬃciently ﬁne 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 timestep (Co
max
= 2.0, 1.0 and 0.5). Again, the diﬀerences
are small, even for the largest timestep, corresponding to Co
max
= 2.0, the error
is less than 0.2%.
Summarising, all generated grids provide suﬃciently accurate force coeﬃcients,
but to capture the far wake vortex dynamics at least 800k cells are required.
7.4 Vortex identiﬁcation methods for ﬂow visualisation 145
t/T []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 Threedimensional veriﬁcation: force coeﬃcients. Lift and drag coeﬃcients for the
veriﬁcation cases with varying grid size, 100k −1600k. The timestep is taken such that Co
max
= 1.0.
The kinematics of the threedimensional wing is simple harmonics with A
φ
= 63
◦
, A
α
= 45
◦
and
A
θ
= 0
◦
.
N ǫ
lift
[%] ǫ
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 coeﬃcients for varying grid sizes. Error values of
lift and drag coeﬃcients [%] with respect to the extrapolated values for varying grid sizes, ranging
from 100k to 1600k. The timestep was determined by a max Courant number of Co
max
= 1.0. The
kinematics of the threedimensional wing is simple harmonics with A
φ
= 63
◦
, A
α
= 45
◦
and A
θ
= 0
◦
.
Furthermore, at Co
max
= 2.0, the temporal errors are suﬃciently small. Therefore,
a grid resolution of 800k in combination with Co
max
= 2.0 was used for all three
dimensional ﬂapping wing simulations, described in this chapter.
7.4 Vortex identiﬁcation methods for ﬂow
visualisation
In order to study the vortex dynamics and the stability of the leadingedge vortex
in particular, a proper vortex identiﬁcation criterion is essential. Diﬀerent well
known techniques to detect and visualise vortices are based on the velocity gradient
tensor, which requires a complete velocity ﬁeld. Two diﬀerent vortex identiﬁcation
criteria are discussed, namely the magnitude of vorticity ω (Lu & Shen, 2008)
146 Vortical structures in threedimensional ﬂapping ﬂight
Drag coeﬃcient []
L
i
f
t
c
o
e
ﬃ
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 Threedimensional veriﬁcation: limit cycle. The limit cycles, constructed from the
lift and drag coeﬃcients clearly shows periodic behaviour for the veriﬁcation cases. The grid size was
ﬁxed to 800k. The timestep is taken such that Co
max
= 1.0. The kinematics of the threedimensional
wing is simple harmonics with A
θ
= 0 rad, A
φ
= 1.1 rad and A
α
= 0.785 rad.
Co
max
ǫ
lift
[%] 400k ǫ
drag
[%] 400k ǫ
lift
[%] 800k ǫ
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 coeﬃcients for varying temporal resolution. Error
values of lift and drag coeﬃcients [%] 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 threedimensional wing is simple harmonics with A
φ
= 63
◦
,
A
α
= 45
◦
and A
θ
= 0
◦
.
and the Q criterion (Hunt et al., 1988).
The ﬁrst criterion is based on the magnitude of vorticity ω, where ω = ∇×u.
If ω reaches a userdeﬁned threshold, that region is identiﬁed as a vortex. More
speciﬁcally, 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 threedimensional
ﬂows this may become a diﬃculty. In twodimensional ﬂow, however, ω 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 identiﬁed as a vortex. This second
invariant of ∇u is written as
Q =
1
2
_
Ω
2
−S
2
_
,
where the rate of strain tensor S is given by S =
1
2
(∇u +∇u
T
) and the vorticity
tensor by Ω =
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 identiﬁcation methods for ﬂow visualisation 147
Spatial resolution []
A
v
e
r
a
g
e
l
i
f
t
c
o
e
ﬃ
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
ﬃ
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
ﬃ
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
ﬃ
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
coeﬃcients for increasing spatial resolution and constant timestep, corresponding to Co
max
= 1.0.
The ﬁnal value, at ∞, is obtained using Richardson extrapolation. The temporal convergence is
illustrated in (c) and (d), showing the average drag and lift coeﬃcients for decreasing timestep at
two speciﬁc grid sizes, 400k and 800k.
rate. Therefore, a region where Q > 0 indicates a clear swirling ﬂow (as shown
by Chakraborty et al., 2005). It must be noted that Jeong & Hussain (1995)
found that Q > 0 is not a suﬃcient condition to have a pressure minimum in the
vortex core of that speciﬁc region. In most cases, however, a pressure minimum
does occur. By neglecting these unsteady and viscous eﬀects from the governing
NavierStokes equations the following relation can be obtained for the symmetric
tensor Ω
2
+S
2
:
Ω
2
+S
2
= −
1
ρ
∇(∇p),
where ρ is the ﬂuid density and p the pressure. In order to identify which vortex
criterion should be used, ﬁgure 7.7 shows isosurfaces around a ﬂapping wing.
The wing ﬂaps around a distance of 0.5 from the wing root and the ﬂapping
angles are varying harmonically. The isosurface is visualised at t = 0.25T, during
the downstroke. At t = 0.25T the leadingedge vortex is formed on the wing’s
148 Vortical structures in threedimensional ﬂapping ﬂight
(a) ω (b) Q
Figure 7.7 Comparison of the near wake ﬂow ﬁeld using diﬀerent vortex identiﬁcation
criteria. The spiralling leadingedge vortex is visualised using contour plots of the magnitude of
vorticity, ω and Q. The arrow shows the ﬂapping direction of a downstroke and the ﬂow is visualised
at midstroke. (a) shows the contour of the vorticity magnitude, ω = 5.0 and (b) the Q = 2.0 value.
The colours represent values of helicity, h = (u · ω)/(u ω), 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 leadingedge vortex, rolling up into a tip vortex, is
identiﬁed using a carefully chosen threshold of the vortex identiﬁcation criteria,
using the values ω = 5.0 and Q = 2.0, normalised by their maximal values. The
colours show the helicity which is deﬁned as h = (u· ω)/(u ω), within the range
of −1.0 ≤ h ≤ 1.0. A positive helicity, h > 0, means that the direction vector of
vorticity (ω = ∇×u) is aligned with the local ﬂow velocity.
In ﬁgure 7.7(a) it can be observed that ω shows not only the vortical struc
tures, but also the shear layers near the wing and between the vortices. This leads
to a thicker isosurface, such that detail of the vortical structures is lost. The Q cri
terion shows more detail, in ﬁgure 7.7(b), a smooth leadingedge vortex is shown,
rolling up into a tip vortex. Some of the previously shed vortices are still present.
Since the Q criterion oﬀers suﬃcient and adequate information about the local
ﬂow ﬁeld, e.g. a rotation dominated region is identiﬁed by Q < 0, this criterion is
used throughout the remainder of the present research, for all threedimensional
simulations.
7.5 Flapping wings at low Reynolds numbers
In order to provide insight into the vortex dynamics (for a purely harmonic ﬂap
ping motion) and its inﬂuence on the variation of forces, diﬀerent geometric and
kinematic parameters are systematically varied. First, the inﬂuence of the angle
of attack on the forces is brieﬂy discussed in section 7.5.1. In that section, the in
ﬂuence of an increasing midstroke angle of attack is brieﬂy discussed on the force
development. In order to investigate the eﬀect 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 diﬀerent
Reynolds number as well. The inﬂuence of the Reynolds number on the forces
and leadingedge 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 ﬂight simulations, section 7.6 deals with forward ﬂapping ﬂight with sim
ilar conditions as the twodimensional simulations performed by Bos et al. (2008),
Thaweewat et al. (2009).
7.5.1 The angle of attack in ﬂapping ﬂight
Previous studies showed that the angle of attack variation during the stroke in
ﬂuences the forces considerably. This was conﬁrmed by a recent twodimensional
investigation (Bos et al., 2008), concerning hovering ﬂight at fruit ﬂy conditions.
It was already mentioned that the angle of attack at midstroke is varied from
α = 15
◦
to α = 90
◦
. Note that α = 90
◦
implies that the wing keeps a constant up
right position during the entire stroke. Table 7.6 shows the timeaveraged lift and
drag coeﬃcient for varying angles. The Reynolds number is Re = 100 and Rossby
Ro = 3.2, corresponding to a ﬂapping wing with small radius of curvature. It is
seen that the maximal timeaveraged lift coeﬃcient occurs at a midstroke angle
of attack of α = 45
◦
. However, the average lifttodrag ratio obtains a maximal
value at an angle of attack of α = 30
◦
, from table 7.6.
Figure 7.8 shows a variation of the lift and drag coeﬃcients during a complete
ﬂapping cycle. It can be observed that the force variation is periodic and smooth.
The average drag is maximal and lift minimal for α = 90
◦
. Furthermore, it can be
seen in ﬁgure 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 ﬂapping
period.
Furthermore, it must be noted that the lift is nearly always nonnegative,
which means that during the hovering conditions, lift is being generated during
the complete stroke. The drag for α = 15
◦
shows two minor peaks within each half
stroke, due to shedding of the leadingedge vortex at low midstroke angle of attack.
While the lift is nearly identical for α = 45
◦
and α = 60
◦
, the drag is signiﬁcantly
lower for α = 45
◦
. In addition, it seems that the leadingedge vortex only grows
signiﬁcantly at higher (α ≥ 45
◦
) angles of attack. The diﬀerence between the
maximal (α = 45
◦
) and minimal (α = 15
◦
) lift is 57%. The next section will
discuss the drops in more detail, since these periods of lower lift appear to occur
at low angles of attack, independent of Rossby and Reynolds number.
7.5.2 Inﬂuence of ﬂapping stroke curvature
In order to investigate the forces and development of the leadingedge vortex, the
radius of curvature is increased to decrease the angular acceleration consequently.
The Rossby number was increased from Ro = 3.2 to Ro = 130, typical for revolving
150 Vortical structures in threedimensional ﬂapping ﬂight
α
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 coeﬃcients for Re = 100 and Ro = 3.2. Timeaveraged lift C
L
, drag C
D
and
lifttodrag C
L
/C
Dave
are shown as a function of the midstroke geometrical angle of attack for given
Re = 100 and Ro = 3.2, so a ﬂapping wing with small stroke curvature.
t/T []
L
i
f
t
c
o
e
ﬃ
c
i
e
n
t
[

]
α = 15
◦
α = 30
◦
α = 45
◦
α = 60
◦
α = 75
◦
α = 90
◦
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
4
(a)
t/T []
D
r
a
g
c
o
e
ﬃ
c
i
e
n
t
[

]
α = 15
◦
α = 30
◦
α = 45
◦
α = 60
◦
α = 75
◦
α = 90
◦
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 7.8 Variation of lift and drag coeﬃcients for a ﬂapping wing at Reynolds number
of Re = 100. The variation is shown for lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing at
Ro = 3.2 and a Reynolds number of Re = 100. The midstroke angle of attack is varied from ♦ : 15
◦
,
•: 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 Timeaveraged force coeﬃcients for varying Rossby numbers at Re = 100. The
variation of average lift (C
L
), drag (C
D
), lifttodrag ratio (C
L
/C
D
) are shown for Rossby numbers
from Ro = 3.2 to Ro = 130. The midstroke angle of attack was ﬁxed to α = 45
◦
and the Reynolds
number to Re = 100.
and translating wings, respectively. Table 7.7 shows the timeaveraged values for
lift, drag and lifttodrag ratio, with increasing Rossby number for α = 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 twodimensional 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 lifttodrag ratio is still signiﬁcant, 18.7%. Figure 7.9 shows the force
histories concerning the nearly translating wing, i.e. Ro = 130. When compared
with ﬁgure 7.8 (which applies to the revolving wing Ro = 3.2) it is seen that
both lift and drag variations are signiﬁcantly 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 coeﬃcients during the ﬂapping cycle and the eﬀect of Rossby number as it
increases. It is clear that the major loss in lift for high Rossby numbers, occurs at
midstroke t=0.25T and t=0.75T. The loss in drag, just after stroke reversal and
during midstroke are of similar magnitude.
Figure 7.11 shows the isosurfaces of Q = 1.0 at t = 0.25T, which is at the
midst of the downstroke. It can be clearly observed that the fully ﬂapping wing
shows a pronounced leadingedge vortex which spirals towards the tip to form a
tip vortex. However, the translating wing shows a leadingedge vortex which stays
symmetric with respect to the wing centre plane, feeding two wing tip vortices, at
the root and the tip. Additionally, ﬁgure 7.13 shows the streamlines to illustrate
the leadingedge and tip vortices in more detail.
An additional observation, while comparing ﬁgures 7.8 and 7.9 is that the lift
drops signiﬁcantly (75% at least) during midstroke (t = 0.2T) for an angle of
attack of α = 15
◦
. Figure 7.12 shows the Q = 1.0 isosurfaces of a translating
wing at Ro = 130 for α = 15
◦
and α = 45
◦
. It is clearly seen that the leading
edge vortex for α = 15
◦
is not yet fully developed, which results in the lower
152 Vortical structures in threedimensional ﬂapping ﬂight
t/T []
L
i
f
t
c
o
e
ﬃ
c
i
e
n
t
[

]
α = 15
◦
α = 30
◦
α = 45
◦
α = 60
◦
α = 75
◦
α = 90
◦
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
4
(a)
t/T []
D
r
a
g
c
o
e
ﬃ
c
i
e
n
t
[

]
α = 15
◦
α = 30
◦
α = 45
◦
α = 60
◦
α = 75
◦
α = 90
◦
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 7.9 Variation of lift and drag coeﬃcients for a translating wing at Ro = 130. Lift
(a) and drag (b) coeﬃcients induced by a ﬂapping wing at a Rossby number of Ro = 130, such that
the wing approximately translates. The Reynolds number remained ﬁxed at Re = 100. The angle of
attack amplitude is varied from ♦ : 15
◦
, •: 30
◦
, ▽ : 45
◦
, : 60
◦
, △: 75
◦
and : 90
◦
.
lift, compared to α = 45
◦
. It seems that the trend of the force development
with the angle of attack is similar for ﬂapping and translating wings, as long as
the scaling is appropriate, such that the dimensionless amplitude A
∗
R
g
, average
Reynolds number Re
R
g
and swept area A
swept
are comparable.
Summarising, it can be stated that a ﬂapping wing motion is of crucial impor
tance for lift generation at a small penalty of drag, compared to wing translation.
Additionally, the leadingedge 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 ﬂapping and translating wings
at α = 15
◦
, that the leadingedge vortex development is not signiﬁcant to increase
the lift, instead the lift decreases.
7.5.3 Inﬂuence of Reynolds number
In addition to the angle of attack and stroke curvature, a selection of Reynolds
numbers is used, Re = 100, Re = 500 and Re = 1000, relevant for insect aerody
namics of a fruit ﬂy (Sane & Dickinson, 2001, Birch & Dickinson, 2003), hawk
moth (Liu & Kawachi, 1998) and dragonﬂy (Isogai et al., 2004), respectively. The
timeaveraged lift and drag coeﬃcients are plotted in ﬁgure 7.14(a) for diﬀerent
angles of attack and diﬀerent Reynolds numbers. In addition, the results of a
variation in Rossby number are shown for Re = 100, 500 and 1000. Figure 7.14(b)
shows the power factor C
L
3/2
/C
D
as a function of lifttodrag ratio C
L
/C
D
.
From ﬁgure 7.14(a) it can be deduced that the overall lift coeﬃcients are signif
icantly higher for the ﬂapping (Ro = 3.2) compared to the translating (Ro = 130)
wing. The drag increases as well. At maximal lift, α = 45
◦
the diﬀerence between
ﬂapping (Ro = 3.2) and translating (Ro = 130) in lift coeﬃcient is 32.8%, 33.9%
7.5 Flapping wings at low Reynolds numbers 153
t/T []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 coeﬃcients for a ﬂapping wing with varying Rossby num
bers. Lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing around varying Rossby numbers,
Ro = 3.2 − 130. The amplitude of the angle of attack variation was ﬁxed such that at midstroke
α = 45
◦
. The average Reynolds number remained ﬁxed 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 diﬀerent Rossby numbers. Iso
surfaces of Q = 1.0 are shown for Rossby numbers Ro = 3.2 and Ro = 130. A timeframe is shown at
midstroke, t = 0.25T. The average Reynolds number was ﬁxed to Re = 100. Colours indicate helicity.
154 Vortical structures in threedimensional ﬂapping ﬂight
(a) α = 45
◦
(b) α = 15
◦
Figure 7.12 Vortex visualisation of the near wake for a translating wing at low and high
angle of attack. Isosurfaces of Q = 1.0 are shown for α = 15
◦
and α = 45
◦
for a Rossby number of
Ro = 130. A timeframe is shown at midstroke, t = 0.25T. The average Reynolds number was ﬁxed
to Re = 100. Colours indicate helicity.
(a) Ro = 3.2 (b) Ro = 130
Figure 7.13 Streamline visualisation of the near wake for diﬀerent Rossby numbers.
Streamlines are shown for Rossby numbers Ro = 3.2 and Ro = 130. A timeframe is shown at mid
stroke, t = 0.25T. The average Reynolds number was ﬁxed to Re = 100. Colours indicate helicity.
7.5 Flapping wings at low Reynolds numbers 155
Drag coeﬃcient []
L
i
f
t
c
o
e
ﬃ
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 lifttodrag ratio
C
L
/C
D
. In both (a) and (b), the results are shown for three diﬀerent 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
◦
, from right to left in (a) and from left to right in (b).
and 35.8% for increasing Reynolds numbers Re = 100, 500 and 1000, respectively.
Since the diﬀerence in lift increases (although slightly) with Re, ﬂapping is impor
tant at lower Reynolds numbers, but becomes slightly more important at higher
Reynolds number. The diﬀerence in drag is only signiﬁcant at larger angles of
attack. At α = 45
◦
the diﬀerences in timeaveraged drag coeﬃcient are negligi
ble when the variation in Reynolds number is concerned, while at α = 75
◦
the
diﬀerences in drag become signiﬁcant. However, considering a ﬂapping motion
(Ro = 3.2) with respect to translating (Ro = 130), an average diﬀerence in drag
of 7.5% is obtained. While lift is enhanced signiﬁcantly, combined with a small
drag penalty, there is still a large gain in lifttodrag, which is beneﬁcial in terms
of performance.
Another observation from ﬁgure 7.14(a) is that at large midstroke angles of
attack, the timeaveraged lift and drag coeﬃcients show marginal variations with
Reynolds number, for a translating (Ro = 130) wing. On the other hand, the
Reynolds number has a larger eﬀect on the lift and drag, while ﬂapping. This may
be explained by considering that the leadingedge vortex was found to be more
stable on a translating wing, compared to ﬂapping. Looking at ﬁgure 7.14(a),
the variations in lift and drag with Reynolds number are larger for lower Rossby
numbers. So, the structure of the leadingedge vortex strongly depends on the
Reynolds number in cases of large angular accelerations.
Figure 7.15 shows the isosurface of Q = 1.0 to visualise the leadingedge vortex
on a ﬂapping wing (Ro = 3.2) during the downstroke for both Reynolds numbers
Re = 100 and Re = 1000. As was already discussed, the leadingedge vortex be
comes slightly unstable with increasing Reynolds numbers, which is visualised by
irregularities in the isosurfaces. In addition, the streamlines for the corresponding
156 Vortical structures in threedimensional ﬂapping ﬂight
(a) Re = 100
(b) Re = 1000
Figure 7.15 Vortex visualisation of the near wake for diﬀerent Reynolds numbers, Ro =
3.2. Isosurface of Q = 1.0 are shown for two Reynolds numbers, Re = 100 and Re = 1000, for a
ﬂapping wing Ro = 3.2. A timeframe is shown at midstroke, t = 0.25T. Colours indicate helicity.
comparison are shown in ﬁgure 7.16. Besides the irregularities, the leadingedge
vortex clearly detaches at a smaller distance from the wing root for Re = 1000.
However, although the leadingedge vortex may be less stable, the lift increasing
eﬀects of the leadingedge vortex are larger for higher Reynolds numbers. Addi
tionally, for higher Reynolds numbers, the leadingedge vortex may burst as was
discussed by Lentink & Dickinson (2009b), without a signiﬁcant loss in lift. In or
der to illustrate the irregular motion at larger Reynolds numbers, i.e. Re = 1000,
ﬁgure 7.17 shows the vortical motion just before stroke reversal from down to
upstroke.
It was noted that the leadingedge vortex may play a more important role for
ﬂapping motion, compared to translation. Figure 7.18 shows the Q isosurfaces
for diﬀerent Reynolds numbers Re = 100 and Re = 1000, for a translating wing.
In case of the ﬂapping wing Ro = 3.2, ﬁgure 7.15 shows some irregularities of the
leadingedge vortex, with increasing Reynolds numbers. For a translating wing
Ro = 130, these irregularities are less pronounced. While the leadingedge vortex
detaches at a smaller distance from the wing tip, on a ﬂapping wing for increasing
Reynolds numbers, this is not the case for a translating wing.
Therefore, it seems plausible that the generation of a leadingedge vortex is
important for both ﬂapping Ro = 3.2 and translating Ro = 130 ﬂight. A Reynolds
number increase, leads to larger lift enhancement, but also to irregularities such
that the ﬂow 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 diﬀerent Reynolds numbers,
Ro = 3.2. Streamlines are shown for two Reynolds numbers, Re = 100 and Re = 1000, for a ﬂapping
wing Ro = 3.2. A timeframe is shown at midstroke, t = 0.25T. Colours indicate helicity.
(a) Re = 100
(b) Re = 1000
Figure 7.17 Vortex visualisation of the near wake for diﬀerent Reynolds numbers, Ro =
3.2. Isosurface of Q = 1.0 are shown for two Reynolds numbers, Re = 100 and Re = 1000, for a
ﬂapping wing Ro = 3.2. A timeframe is shown at the end of the downstroke, t = 0.5T. Colours
indicate helicity.
158 Vortical structures in threedimensional ﬂapping ﬂight
(a) Re = 100 (b) Re = 1000
Figure 7.18 Vortex visualisation of the near wake for diﬀerent Reynolds numbers, Ro =
130. Isosurface of Q = 1.0 are shown for two Reynolds numbers, Re = 100 and Re = 1000, for a
translating wing Ro = 130. A timeframe is shown at midstroke, t = 0.25T. Colours indicate helicity.
7.5.4 Inﬂuence of ‘trapezoidal’ angle of attack
In order to study the inﬂuence of the ‘trapezoidal’ angle of attack and later com
pared to an earlier twodimensional 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 deﬁned by a rotation timing T
rot
, which is
varied from 0.25 to 0.1, the ﬁrst 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 midstroke angle
of attack remained ﬁxed.
Table 7.8 shows the timeaverage lift, drag and lifttodrag 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 signiﬁcant
increase in lifttodrag ratio of 21.9%. Furthermore, it can be seen that the average
drag coeﬃcient, 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 ﬁgure 7.19 for varying
rotation duration during a full ﬂapping 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 eﬀective angle of attack. Since the wing rotates relatively quickly after
reversal, it reaches its midstroke 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 Timeaveraged force coeﬃcients for varying ‘trapezoidal’ shape of angle of
attack. The timeaveraged lift C
L
, drag C
D
and lifttodrag 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
ﬃ
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
ﬃ
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 coeﬃcients for a ﬂapping wing with ‘trapezoidal’ angle of attack. Lift
(b) and drag (c) coeﬃcients induced by a ﬂapping 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
α
= 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, eﬀect
applies to the drag coeﬃcient.
At every time instance, the lift increase and drag decrease are only marginal,
but overall the results are signiﬁcant, 21.9% lifttodrag enhancement. Therefore,
it can be stated that a ‘trapezoidal’ shape in angle of attack considerably increases
performance in threedimensional hovering ﬂapping ﬂight. In contrast to this, two
dimensional studies (Bos et al., 2008) showed opposite eﬀects, due to a premature
vortex shedding of the leadingedge vortex during the long period of high angle
of attack. The explanation for this discrepancy is that in the threedimensional
simulations the leadingedge vortex was found to remain more stable than in two
dimensional investigations. Therefore, it can be concluded that the leadingedge
vortex stability should be studied with a threedimensional approach.
160 Vortical structures in threedimensional ﬂapping ﬂight
7.5.5 Inﬂuence of deviation
The deviation amplitude A
θ
is used to tilt the twodimensional airfoil or three
dimensional wing out of the horizontal stroke plane. It may be used to generate
certain wing tip patterns, e,g. the wellknown ‘ﬁgureofeight’, which is present in
realistic fruit ﬂy kinematics (Fry et al., 2003). In (Bos et al., 2008) it was shown
that although deviation did not inﬂuence the timeaveraged values, the instanta
neous lift and drag variations are signiﬁcantly aﬀected. In order to investigate
the inﬂuence of deviation the amplitude A
θ
is varied from 0
◦
to 20
◦
. In addition,
three diﬀerent stroke patterns are studied, by varying the deviation frequency,
resulting in patterns that can be characterised as ‘ﬁgureofO’, ‘ﬁgureofU’ and
‘ﬁgureofeight’, which are shown in ﬁgure 7.2. Considering the ‘ﬁgureofeight’
patterns, two diﬀerent deviation directions are studied, corresponding to a vari
ation of A
θ
from 0
◦
to 20
◦
and from 0
◦
to −20
◦
. The reference velocities are
adapted correspondingly.
First, the deviation amplitude is varied from A
θ
= 0
◦
to A
θ
= 20
◦
, according
to the ‘ﬁgureofO’ wing tip pattern. Following ﬁgure 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 eﬀective angle of attack, which is therefore
subjected to an increase, decrease, decrease and another increase during the four
consecutive halfstrokes. Table 7.9 shows the timeaveraged force coeﬃcients for
a wing following this ‘ﬁgureofO’ motion, while the midstroke angle of attack is
ﬁxed to α = 45
◦
and the Reynolds number maintained at Re = 100. It is shown
that the timeaveraged lift coeﬃcient decreases signiﬁcantly with 9.8%. However,
this is fully compensated by a decrease in drag with the same amount such that
the diﬀerences in average lifttodrag ratio are negligible. The average lifttodrag
was obtained by using the average drag over up and downstroke. This was nec
essary, because of the asymmetry appearing in the average drag coeﬃcient. This
asymmetry in drag is the result of the asymmetrical variation in eﬀective angle of
attack, as was previously discussed. Figure 7.20 show the lift and drag variations
during a complete ﬂapping stroke.
Secondly, the results are considered for a ﬂapping wing following the ‘ﬁgure
ofU’ pattern, which is similar to the one used in (Bos et al., 2008). When using
this kinematic pattern, the wing moves down and up during every halfstroke, i.e.
the upstroke is identical to the downstroke. In table 7.10 is can easily be seen
that the diﬀerences in up and downstroke drag are negligible for all deviation
amplitudes. This is in contrast to the observations, considering the ‘ﬁgureofO’.
The timeaveraged drag coeﬃcient is constant for deviation amplitude variation,
while the lift coeﬃcient decreases with 9.4%, leading to a similar decrease in lift
todrag ratio. Because of symmetric (similar up and downstroke) ﬂapping, a
decrease of lifttodrag coeﬃcient is obtained, which is present, but not signiﬁcant
in comparison to the eﬀect of varying Rossby and Reynolds numbers.
The third pattern is governed by the ‘ﬁgureofeight’ wing tip motion. Al
though, the previously described deviation patterns only cause marginal eﬀects,
7.5 Flapping wings at low Reynolds numbers 161
A
θ
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 Timeaveraged force coeﬃcients for varying deviation using a ‘ﬁgureofO’
pattern. The timeaveraged lift C
L
, drag C
D
and lifttodrag ratio C
L
/C
D
are shown for a deviation
amplitude, varying from A
θ
= 0
◦
to 20
◦
, for a ‘ﬁgureofO’. The midstroke angle of attack was ﬁxed
at α = 45
◦
and the average Reynolds number was maintained at Re = 100.
t/T []
L
i
f
t
c
o
e
ﬃ
c
i
e
n
t
[

]
20
A
θ
= 0
A
θ
= 5
A
θ
= 10
A
θ
= 15
A
θ
= 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
ﬃ
c
i
e
n
t
[

]
A
θ
= 0
A
θ
= 5
A
θ
= 10
A
θ
= 15
A
θ
= 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 coeﬃcients for a ﬂapping wing with varying deviation using ‘ﬁgureof
O’. Lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing, such that Ro = 3.2 and Re = 100.
The deviation amplitude is varied from A
θ
= 0
◦
to A
θ
= 30
◦
, using an angle of attack amplitude of
A
α
= 45
◦
.
the ‘ﬁgureofeight’ pattern may cause signiﬁcant changes in forces. Two types of
‘ﬁgureofeight’ patterns are used, from A
θ
= 0
◦
to 20
◦
, this is called type 1 and
the second from A
θ
= 0
◦
to −20
◦
, type 2. To achieve a ‘ﬁgureofeight’ pattern,
the deviation frequency is twice the frequency of the other two wing tip patterns.
For both types of patterns, the eﬀective angle of attack variation consists of three
parts during each half stoke, see ﬁgure 7.2. Both up and downstroke follow exactly
the same, thus symmetrical, motion. The type 1 patterns starts each halfstroke
with a downward motion, than it goes up until it has to go done just before stroke
reversal. This wing tip motion leads to a consecutive increase, decrease and in
crease in eﬀective angle of attack, where the period of decrease is twice the period
of increase. The type 2 pattern follows precisely the inverse motion.
Table 7.11 shows the timeaveraged lift, drag and lifttodrag ratio’s for both
types of ‘ﬁgureofeight’ patterns. Since the variation of the eﬀective angle of at
tack is symmetric, the drag coeﬃcient is equal for up and downstroke. The results
162 Vortical structures in threedimensional ﬂapping ﬂight
of the type 1 motion pattern are remarkable. There is a considerable decrease in
both timeaveraged lift (52%) and drag (44%) when comparing A
θ
= 20
◦
with
the baseline case A
θ
= 0
◦
. Since the drag decrease is of similar magnitude as the
lift decrease, the loss in lifttodrag is limited to 15.5%, which is still signiﬁcant.
Figure 7.20 shows the lift and drag variations for this type 1 ‘ﬁgureofeight’ pat
tern, where the wing tip moves down, up and down, consecutively during each
halfstroke. It can be observed that the short period of downward motion at the
beginning of each stroke increases lift. On the other hand, the large period of
upward motion, decreases the eﬀective angle of attack for a relative long period,
leading to a signiﬁcant loss of lift, as is seen in the ﬁgure. For A
θ
= 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. α = 15
◦
, see section 7.5.1.
When considering the type 2 ‘ﬁgureofeight’ pattern, table 7.11 shows no sig
niﬁcant decrease in drag. The lift, however, decreases considerably, although sig
niﬁcantly smaller compared to the type 1 kinematic pattern, 53% versus 12.6%.
The motion of the type 2 deviation is apparently equally distributed, resulting
in only 12.6% less lift and no diﬀerences in drag, while increasing the deviation
amplitude. The maximal diﬀerence in lifttodrag ratio is therefore 12.0%. This
force balance is illustrated in ﬁgure 7.22, which shows the lift and drag during a
full stroke. At the beginning of a stroke, the eﬀective angle of attack is decreased,
leading to lower lift and lower drag. During midstroke, the eﬀective angle of
attack is increased, which is reﬂected in the higher lift and drag.
Summarising, it was shown that the variation in lift and drag can be signiﬁ
cantly inﬂuenced by introducing deviation in the stroke pattern, i.e. ‘ﬁgureofO’,
‘ﬁgureofU’ and ‘ﬁgureofeight’. The ‘ﬁgureofO’ patterns resulted in an asym
metric force variation, due to asymmetric modulation of the eﬀective angle of
attack. Lift and drag decrease with a similar amount, such that the lifttodrag
ratio was only marginally aﬀected. The timeaveraged drag was not inﬂuenced
by the symmetrical ‘ﬁgureofU’ pattern. The average lift, however, did decrease,
such that a loss in lifttodrag was observed. Two types of ‘ﬁgureofeight’ patterns
were investigated, diﬀering in the direction of motion. When the wing moved up
A
θ
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 Timeaveraged force coeﬃcients for varying deviation using a ‘ﬁgureofU’
pattern. The timeaveraged lift C
L
, drag C
D
and lifttodrag ratio C
L
/C
D
are shown for a deviation
amplitude, varying from A
θ
= 0
◦
to 20
◦
, for a ‘ﬁgureofU’ pattern. The midstroke angle of attack
was ﬁxed at α = 45
◦
and the average Reynolds number was maintained at Re = 100.
7.6 Flapping wings in forward ﬂight 163
A
θ
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 Timeaveraged force coeﬃcients for varying deviation using a ‘ﬁgureofeight’
pattern. The timeaveraged lift C
L
, drag C
D
and lifttodrag ratio C
L
/C
D
are shown for a deviation
amplitude, varying from A
θ
= 0
◦
to 20
◦
and A
θ
= 0
◦
to −20
◦
, for a ‘ﬁgureofeight’ pattern. The
midstroke angle of attack was ﬁxed at α = 45
◦
and the average Reynolds number was maintained at
Re = 100.
wards during midstroke, decreasing the eﬀective angle of attack for a long period,
the performance was limited in terms of lift and lifttodrag ratio. On the other
hand, if the wing moved downward during each midstroke, the performance was
similar to the ‘ﬁgureofU’ pattern.
The lift and drag are shown to be sensitive to diﬀerent stroke patterns, such
that the forces and performance can be easily modulated. Therefore, insects could
use stroke plane deviation in extreme hovering or manoeuvring conditions. These
ﬁndings are very interesting for the development of Micro Air Vehicles (MAV) as
well.
7.6 Flapping wings in forward ﬂight
In order to relate the results of the threedimensional ﬂow simulations to previously
conducted twodimensional studies (Thaweewat et al., 2009), a threedimensional
forward ﬂapping wing has been studied. Based on (Thaweewat et al., 2009), the
dimensionless wavelength is set to λ = 6.3 (maximal performance) in order to
justify a comparison. The midstroke angle of attack is varied from α = 0
◦
to 45
◦
,
while maintaining a Reynolds number of Re = 150. The threedimensional 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 diﬀerent cases. Additionally,
the average Reynolds number Re
R
g
is matched.
In table 7.12, the timeaveraged force coeﬃcients are shown for three diﬀer
ent ﬂapping situations: a twodimensional plunging airfoil, a threedimensional
translating wing (Ro = 130) and a threedimensional ﬂapping wing (revolving,
Ro = 3.2). For both ﬂapping and translating wings, the force coeﬃcients are com
164 Vortical structures in threedimensional ﬂapping ﬂight
t/T []
L
i
f
t
c
o
e
ﬃ
c
i
e
n
t
[

]
A
θ
= 0
A
θ
= 5
A
θ
= 10
A
θ
= 15
A
θ
= 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
ﬃ
c
i
e
n
t
[

]
A
θ
= 0
A
θ
= 5
A
θ
= 10
A
θ
= 15
A
θ
= 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 coeﬃcients for a ﬂapping wing with varying deviation using type 1 of
‘ﬁgureofeight’. Lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing, such that Ro = 3.2
and Re = 100. The deviation amplitude is varied from A
θ
= 0
◦
to A
θ
= 20
◦
, using an angle of attack
amplitude of A
α
= 45
◦
.
t/T []
L
i
f
t
c
o
e
ﬃ
c
i
e
n
t
[

]
A
θ
= 0
A
θ
= −5
A
θ
= −10
A
θ
= −15
A
θ
= −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
ﬃ
c
i
e
n
t
[

]
A
θ
= 0
A
θ
= −5
A
θ
= −10
A
θ
= −15
A
θ
= −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 coeﬃcients for a ﬂapping wing with varying deviation using type 2 of
‘ﬁgureofeight’. Lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing, such that Ro = 3.2
and Re = 100. The deviation amplitude is varied from A
θ
= 0
◦
to A
θ
= −20
◦
, using an angle of
attack amplitude of A
α
= 45
◦
.
7.6 Flapping wings in forward ﬂight 165
A
α
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 Timeaveraged force coeﬃcients in forward ﬂight. Two and threedimensional
timeaveraged force coeﬃcients for a ﬂapping wing in forward ﬂight at Re = 150, λ
∗
= 6.3. The
midstroke angle of attack is varied from 0
◦
to 45
◦
. A ﬂapping wing (Ro = 3.2) and a translating wing
(Ro = 130) are compared with a twodimensional plunging airfoil.
pared with the twodimensional results. Note that negative drag means thrust.
It is clear that the timeaveraged lift coeﬃcient decreases with increasing an
gle of attack amplitude. This is illustrated for the ﬂapping wing (Ro = 3.2)
in ﬁgure 7.23. Maximal lift occurs without wing rotation, but the resulting
thrust is minimal. The drag varies with angle of attack such that it reaches a
thrust optimum for A
α
= 30
◦
, which is the case for both twodimensional and
threedimensional ﬂapping. Concerning the threedimensional translating wing
(Ro = 130), the lift decreases with 28% compared to the twodimensional plung
ing airfoil. This is the case for midstroke values of the angle of attack. The
generation of tip vortices causes a loss of energy, which results in a lower lift coef
ﬁcient. The two tip vortices result in a symmetric ﬂow, such that for all angles of
attack, a relative equal amount of energy is lost, leading to a similar decrease in
lift, as is shown in table 7.12. Therefore, it can be stated that a threedimensional
wing, performing a twodimensional motion, leads to similar force results as if the
study was completely twodimensional.
The threedimensional ﬂapping wing (Ro = 3.2) generates larger lift coeﬃcients
compared to the translating case, but lower compared to the twodimensional
ﬂapping airfoil. The diﬀerence with the translating wing becomes smaller with
increasing angle of attack amplitude. Figure 7.24 shows a comparison of the ﬂow
ﬁelds for Ro = 130 and Ro = 3.2, without an angle of attack variation. For the
case without revolving (ﬁgure 7.24(a)), a smooth ring vortex is formed by the
166 Vortical structures in threedimensional ﬂapping ﬂight
t/T []
L
i
f
t
c
o
e
ﬃ
c
i
e
n
t
[

]
α = 0
◦
α = 15
◦
α = 30
◦
α = 45
◦
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
(a)
t/T []
D
r
a
g
c
o
e
ﬃ
c
i
e
n
t
[

]
α = 0
◦
α = 15
◦
α = 30
◦
α = 45
◦
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0.2
(b)
Figure 7.23 Force coeﬃcients for a threedimensional wing in forward ﬂight. Lift (a) and
drag (b) coeﬃcients for a threedimensional ﬂapping wing in forward ﬂight at Re = 150, λ
∗
= 6.3 and
midstroke angle of attack of A
α
= 45
◦
.
two counterrotating tip vortices. This vortex ring is shed and convected into the
wake. The ﬂapping wing (ﬁgure 7.24(b)), however, generates a more complex ﬂow
ﬁeld, induced by a spiralling leadingedge and tip vortex, even though the angle
of attack is zero. When the angle of attack is increased, ﬁgure 7.25 shows for
A
α
= 30
◦
that the wake becomes more smooth, which is governed by a decrease in
eﬀective angle of attack. The revolving wing (Ro = 3.2) induces a stable leading
edge vortex, leading to larger lift and thrust, compared to a translating (Ro = 130)
threedimensional wing.
Summarising, the conclusion can be drawn that a leadingedge vortex is very
important in threedimensional ﬂapping ﬂight. While the leadingedge vortex de
taches and convects in twodimensional plunging, it remains stably attached on
top of a threedimensional wing, increasing both lift and thrust. However, the
induced vortices are strongly dependent on the Rossby number, inﬂuencing the
forces accordingly. Although, the twodimensional forces are higher, compared to
the threedimensional cases, the force variation is similar. Therefore, performing
a twodimensional analysis may be representative to investigate threedimensional
ﬂapping wing aerodynamics.
7.7 Conclusions
This chapter deals with the results obtained by performing numerical simulations
of the ﬂow around a threedimensional ﬂapping wing. A numerical model has been
developed which solves the ﬂow around a threedimensional wing with complex
wing kinematics. The numerical code was veriﬁed by using a temporal and spatial
independence study.
Diﬀerent aspects, relevant to threedimensional ﬂapping wing aerodynamics,
7.7 Conclusions 167
(a) Ro = 130 (b) Ro = 3.2
Figure 7.24 Vortex visualisation of the near wake of a threedimensional ﬂapping wing
in forward ﬂight. Isosurface of Q = 1.0 are shown for a threedimensional ﬂapping wing in forward
ﬂight at Re = 150, λ
∗
= 6.3 and midstroke angle of attack of A
α
= 0
◦
. (a) shows the wing for
Ro = 130, while (b) shows the isosurfaces for Ro = 3.2. Colours indicate helicity.
(a) Ro = 130 (b) Ro = 3.2
Figure 7.25 Vortex visualisation of the near wake of a threedimensional ﬂapping wing
in forward ﬂight. Isosurface of Q = 1.0 are shown for a threedimensional ﬂapping wing in forward
ﬂight at Re = 150, λ
∗
= 6.3 and midstroke angle of attack of A
α
= 30
◦
. (a) shows the wing for
Ro = 130, while (b) shows the isosurfaces for Ro = 3.2. Colours indicate helicity.
168 Vortical structures in threedimensional ﬂapping ﬂight
have been studied. First, the ﬂow around a dynamically scaled model wing is
solved for diﬀerent angles of attack in order to study the force development and
vortex dynamics at small and large midstroke angle of attack. Secondly, the
Rossby number is varied at diﬀerent Reynolds numbers. A varying Rossby num
ber represents a variation in radius of stroke path and thus angular acceleration.
Thirdly, the threedimensional wing kinematics is varied by changing the shape in
angle of attack and by applying a deviation, which may cause a ‘ﬁgureofeight’
pattern. Finally, the threedimensional ﬂow is compared with the twodimensional
studies performed on ﬂapping forward ﬂight. All numerical simulations were dy
namically scaled, using the radius of gyration. The radius of gyration was used
in order to design a proper framework for comparison for the whole parameter
space investigated, twodimensional as well as threedimensional. The considered
parameters are subsequently investigated leading to the following results.
The eﬀect of a variation in angle of attack in this threedimensional study is
such that the maximal lift occurs at a midstroke angle of α = 45
◦
. However,
the performance, in terms of maximal lifttodrag ratio was found to be maximal
at α = 30
◦
. The ﬂapping motion induces a leadingedge vortex, which causes a
peak in lift halfway during each up and downstroke. This leadingedge vortex
appears to be strong at suﬃciently high angles of attack α = 30
◦
. At α = 15
◦
the
leadingedge vortex is not strong enough, causing a sudden decrease in both lift
and drag during each halfstroke.
Secondly, the eﬀects are studied of a varying stroke curvature, reﬂected by the
Rossby number. Both timeaveraged lift and drag decrease signiﬁcantly with in
creasing Rossby number. At Ro = 130, a nearly translating wing, the lift decreases
with 24.7%, compared to the ﬂapping wing with Ro = 3.2. The major decrease
in lift and drag occurs during midstroke, between t = 0.25T and t = 0.5T. Flow
visualisations showed that the leadingedge vortex is signiﬁcantly reduced in size
and strength for the translating with at Ro = 130, compared to the ﬂapping wing.
In addition, the leadingedge vortex rollsup to form two tip vortices instead of
one in case of the ﬂapping wings. This causes both lift and drag to be signiﬁcantly
lower at Ro = 130.
To study the eﬀect of three diﬀerent 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 timeaveraged lift and drag increases, but the diﬀerences become
smaller at Re = 1000. The eﬀect of a changing Reynolds number is negligible
for both lift and drag at high midstroke angles of attack translation. This is
caused by the larger importance of the leadingedge vortex for a ﬂapping wing at
Ro = 3.2. For both ﬂapping, Ro = 3.2, and translation Ro = 130, irregularities in
the leadingedge 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 leadingedge vortex bursts, causing a loss in performance.
Besides the inﬂuence of the angle of attack, Rossby and Reynolds numbers, the
7.7 Conclusions 169
eﬀects 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 pitchup motion at the beginning of each up and downstroke.
Therefore, the midstroke 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 signiﬁcant increase in timeaveraged lift of 10.8%. The lift increase
is accompanied by a decrease in drag of a similar amount, such that the liftto
drag ratio shows a signiﬁcant gain of 21.9%. Twodimensional studies showed
an opposite result which was caused by the fact that the leadingedge vortex
separated, while translating at a constant angle of attack. Threedimensional
eﬀects, however, lead to ﬁrm and stably attacked leadingedge vortex.
The second eﬀect of kinematic modelling is reﬂected by the presence of de
viation. Three diﬀerent wing tip patterns may be caused by deviation, such as
the ‘ﬁgureofO’, ‘ﬁgureofU’ and the ‘ﬁgureofeight’. The ‘ﬁgureofO’ wing tip
pattern leads to an asymmetric variation in eﬀective angle of attack, leading to
diﬀerences in drag for the up and downstroke. Nevertheless, the eﬀect of this
pattern on the timeaveraged forces coeﬃcients is negligible. On the other hand,
the ‘ﬁgureofU’ pattern does inﬂuence the forces. The eﬀective angle of attack is
signiﬁcantly decreased during midstroke. The timeaveraged lift coeﬃcient is de
creased by 9.4%. While drag is maintaining nearly constant, the lifttodrag ratio
is decreased by about 9%. The ‘ﬁgureofeight’ aﬀects the forces most, however,
it depends on the starting direction. Two types are considered, the ﬁrst starting
the downstroke with an downward deviation motion, whereas the type 2 starts
upwards. The eﬀective angle of attack is being inﬂuenced in such a way that the
type 1 ‘ﬁgureofeight’ decreases both lift and drag considerably. Lift decreases up
to 53%. Since drag decreases fast as well, the loss in lifttodrag ratio is limited to
15.5%. The type 2 ‘ﬁgureofeight’ 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 lifttodrag ratio decreases with about 12.0%.
In addition to the eﬀects of Reynolds number, Rossby number and wing kine
matics in hovering ﬂight, a preliminary study is performed to compare two and
threedimensional forward ﬂapping ﬂight. Both two and threedimensional 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 diﬀerent rotation amplitudes at a Reynolds number of Re = 150. For the
translating wing (Ro = 130), the threedimensional simulations result in 28% less
lift compared to the twodimensional case. This diﬀerence is mainly caused by
the loss in lift due to the generation of a tip vortex which is only present in the
threedimensional simulations. However, an increase in Rossby number resulted
in a signiﬁcant gain in lift. In combination with a higher thrust this observation
leads to the conclusion that a stable leadingedge vortex (induced by the revolving
motion) plays an important role in threedimensional ﬂapping aerodynamics.
CHAPTER 8
Inﬂuence of wing deformation by
ﬂexing
A preliminary study is performed to investigate the eﬀects of wing ﬂexing in ﬂap
ping wing aerodynamics. The eﬀects of a cosineshaped wing ﬂexing is analysed
in twodimensional forward ﬂapping ﬂight. For threedimensional hovering ﬂight,
the eﬀects of wing ﬂexing in spanwise and chordwise directions are discussed. In
twodimensional forward ﬂapping ﬂight, the ﬂexing of the airfoil shows similar
eﬀects as if wing rotation was applied, increasing its eﬃciency. Furthermore, in
threedimensional hovering, the ﬂexing reduces the strength of the leadingedge
vortex, compared to a rigid wing. This leads to an overall decrease in lift and
drag, this inﬂuence is larger for chordwise compared to spanwise ﬂexing.
Section 8.1 deals with the inﬂuence of ﬂexing for the twodimensional plunging
airfoil in forward ﬂight. Its inﬂuence on the ﬂow induced by a threedimensional
wing in hovering ﬂight is subject of study in section 8.2. The conclusions are
drawn in section 8.3.
8.1 Airfoil ﬂexing in twodimensional forward
ﬂapping ﬂight
In chapter 6, the inﬂuence on the forces and vortex patterns of a variation in
dimensionless wavelength, ﬂapping amplitude, angle of attack and stroke plane
angle was investigated in twodimensional forward ﬂapping ﬂight. These results
are extended by using preliminary simulations of deformation airfoils, subjected to
a predeﬁned ﬂexing. The boundary displacements due to the ﬂexing are deﬁned
172 Inﬂuence of wing deformation by ﬂexing
with respect to the initial boundary shape and varying in time such that the
maximal ﬂexing occurs at midstroke. The resulting airfoil shape is either a quarter
or halfcosine shape, corresponding to ǫ
f
= 0.25 or ǫ
f
= 0.5, respectively, as is
shown in ﬁgure 8.1. Besides the two diﬀerent deformation mode shapes, the ﬂexing
amplitude is varied from A
f
= 0.1 to A
f
= 0.4, which is similar to maximal 40%
of the chord length. This ﬂexing was imposed on a baseline ﬂapping motion
(see chapter 6) with a dimensionless wavelength of λ
∗
= 6.8 and a dimensionless
amplitude of A
∗
= 1.5. The rotation amplitude is ﬁxed to A
α
= 0
◦
and the average
Reynolds number was ﬁxed at Re = 150.
Table 8.1 shows the timeaveraged lift, drag and lifttodrag coeﬃcients for the
plunging airfoil. For both ﬂexing shapes the lift decreases with ﬂexing amplitude,
however, the timeaveraged lift decreases faster for ǫ
f
= 0.25. Considering this
quartercosine shaped ﬂexing, a small ﬂexing amplitude of A
f
= 0.1 already results
in a signiﬁcant loss in lift of 14%, which gradually increases up to 49.5% for
A
f
= 0.8, which is a very large deformation. The decrease in lift for the half
cosine shape, ǫ
f
= 0.5, is smaller compared to all ǫ
f
= 0.25 cases, for similar
ﬂexing amplitudes, but increases more rapidly with increasing ﬂexing amplitude.
On the other hand, the halfcosine deformation (ǫ
f
= 0.5) is more eﬃcient in
generation of thrust (negative drag). For example, the decrease in average lift for
A
f
= 0.4 and ǫ
f
= 0.5 is half the value obtained with ǫ
f
= 0.25, but instead, the
thrust is doubled. It appears that the halfcosine shaped ﬂexing increases thrust
signiﬁcantly, while the lift is less aﬀected compared to the quartercosine shape.
Another observation from table 8.1 is the presence of asymmetry in the time
averaged lift coeﬃcient for the quartercosine deﬂection. Chapter 6 shows a similar
asymmetric ﬂow behaviour concerning the airfoil rotation. It was shown that
Xcoordinate [m]
Y
d
i
s
p
l
a
c
e
m
e
n
t
[
m
]
ǫ
f
= 0.0
ǫ
f
= 0.25
ǫ
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 ﬂexing displacements
of a plunging airfoil are shown for diﬀerent shape ratio’s. A shape ratio ǫ
f
= 0.25 corresponds to a
quartercosine and ǫ
f
= 0.5 to a halfcosine shape.
8.1 Airfoil ﬂexing in twodimensional forward ﬂapping ﬂight 173
A
f
C
Lave
C
Ldownstroke
C
Lupstroke
C
D
0 1.956 (baseline) 1.758 2.154 0.058
ǫ
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
ǫ
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 Timeaveraged force coeﬃcients for a deforming plunging airfoil. The time
averaged lift and drag are shown for a deforming twodimensional airfoil, for two ﬂexing shapes, ǫ
f
=
0.25 and ǫ
f
= 0.5. The Reynolds number was ﬁxed to Re = 150, the dimensionless wavelength was
k = 6.8 and the rotation amplitude remained at A
α
= 0.
the force asymmetry is strongly related to the wake dynamics. Therefore, airfoil
ﬂexing using a quartercosine shape (ǫ = 0.25) may induce a similar eﬀect on the
wake dynamics as airfoil rotation. However, when applying the halfcosine airfoil
ﬂexing, the lift coeﬃcient is symmetric over a complete stroke. While the quarter
cosine increases the airfoil camber, the halfcosine shape redirects the trailingedge
towards the uniform freestream. This redirection of the trailingedge compensates
for the destabilising eﬀect of the airfoil rotation, such that a symmetric force is
the result of a symmetric wake pattern, which is beneﬁcial in terms of eﬃciency.
Figure 8.2 shows the drag coeﬃcients for both ﬂexing shapes. It is clear that the
thrust (negative drag) is more prominent for the halfcosine (ǫ = 0.5) ﬂexing mode.
The main regions of thrust enhancement are during midstroke, 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 ﬂow physics, ﬁgure 8.3 shows vorticity contours for both ﬂexing shapes with
an amplitude of A
f
= 0.2 at t = 0.25T, which is midway during the downstroke.
Comparing ﬁgure 8.3(a) with 8.3(b), it is apparent that the leadingedge vortex is
of similar size and strength for both ﬂexing shapes. However, the trailing edge of
the halfcosine 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 halfcosine (ǫ = 0.5) ﬂexing mode, increasing its thrust.
Summarising, this preliminary twodimensional investigation, to understand
the eﬀect of wing ﬂexing, has led to some interesting results. The applied wing
ﬂexing inﬂuences the ﬂow in a similar way as was obtained by applying airfoil
rotation, indicating a similar mechanism in terms of the eﬀects of angle of attack.
With the introduction of ﬂexing, the drag became negative to generate thrust in
forward ﬂight, and the lift decreased signiﬁcantly. The lift and drag development
strongly depends on the shape of the wing ﬂexing. The halfcosine ﬂexing shape
174 Inﬂuence of wing deformation by ﬂexing
t/T []
D
r
a
g
c
o
e
ﬃ
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) ǫ = 0.25
t/T []
D
r
a
g
c
o
e
ﬃ
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) ǫ = 0.5
Figure 8.2 Comparison of drag coeﬃcients for a ﬂexing airfoil. The drag coeﬃcients are shown
for the quartercosine (a) and halfcosine (b) shaped airfoil ﬂexing modes. The ﬂexing amplitude A
flex
is varied from 0.0 to 0.4, the latter corresponds to 40% of the chord length.
(a) ǫ = 0.25 (b) ǫ = 0.5
Figure 8.3 Vorticity contours. Vorticity contours around a plunging ﬂexing airfoil without ro
tation at Re = 150 are shown during midstroke at t = 0.25T. The ﬂapping amplitude was ﬁxed to
A
∗
= 1.5 and the dimensionless wavelength was set to λ
∗
= 6.8. The ﬂexing amplitude was set to
A
f
= 0.2 using both ﬂexing shapes, while the rotation amplitude was kept to zero A
α
= 0
◦
.
results in less lift than obtained using the quartercosine shape. Additionally, the
halfcosine ﬂexing mode also generates the largest thrust. This behaviour was
related to the fact that the halfcosine shape caused the generated vortices to
convect faster, since the trailingedge became aligned with the ﬂow.
8.2 Wing ﬂexing in threedimensional hovering ﬂight 175
8.2 Wing ﬂexing in threedimensional hovering
ﬂight
In addition to the investigation of ﬂexing eﬀects on an airfoil in forward ﬂapping
ﬂight, it was chosen to investigate a threedimensional wing as well. Hovering
conditions are assumed, since it was extensively studied in the previous chapter.
Besides the inﬂuence of the Rossby number, Reynolds number and the kinematic
modelling, applied under rigid wing conditions, it may be interesting to study the
additional eﬀects of wing deformation. The deformation of the wing is deﬁned by a
userdeﬁned ﬂexing function, leading to a realistic bending of the threedimensional
wing (Combes & Daniel, 2003a,b, Shyy et al., 2008a).
In order to investigate the eﬀects of threedimensional wing ﬂexing on the
forces and performance, the ﬂexing is deﬁned according to the model described
in section 4.3.3. The deformation model is hence similar to the one used for
the twodimensional airfoil in the previous section. Additionally, the ﬂexing was
independently applied to the spanwise and the chordwise directions. The three
dimensional wing is deformed such that its shape prescribes a quartercosine 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 ﬂexing 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 deﬂection at midstroke of 20% of the representative
ﬂexing length (wing span or chord, respectively). Figure 8.1 shows an illustration
of the ﬂexing 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 Timeaveraged force coeﬃcients for a deforming threedimensional wing. Time
averaged force coeﬃcients of a threedimensional ﬂexing wing. The spanwise ﬂexing amplitude is varied
from A
f
= 0.0 to A
f
= 0.4, while the chordwise ﬂexing amplitude varies from A
f
= 0.0 to A
f
= 0.2.
.
176 Inﬂuence of wing deformation by ﬂexing
t/T []
L
i
f
t
c
o
e
ﬃ
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
ﬃ
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 coeﬃcients for a spanwise and chordwise deforming ﬂapping wing. Lift
(a) and drag (b) coeﬃcients induced by a ﬂapping 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
midstroke remained at α = 45
◦
and the Reynolds number remained at Re = 100.
.
While varying the ﬂexing amplitude, the average Reynolds number is kept
constant at Re = 100 and the midstroke angle of attack is chosen to be α = 45
◦
.
Table 8.2 shows the timeaveraged force coeﬃcients for diﬀerent ﬂexing amplitudes,
using spanwise or chordwise deformation. The most important result is that both
timeaveraged lift and drag are only marginally inﬂuenced by the spanwise ﬂexing,
so where the wing tip bends. The maximal diﬀerence with the nondeforming case
(3.6%) occurs for 40% wing tip deﬂection, which is a signiﬁcant deformation.
The timeaveraged drag remains nearly unaﬀected as well, such that the maximal
diﬀerence in lifttodrag ratio is only 2.6%.
The chordwise deformation, on the other hand, inﬂuences the timeaveraged
forces to a large extent. As is easily seen, the timeaveraged 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
ﬂexing, such that the maximal diﬀerence in lifttodrag ratio is 24.8%.
The force variation, during a full ﬂapping stroke, is shown in ﬁgure 8.4. It
is clear that the largest diﬀerence in force coeﬃcients occurs during midstroke,
lowering the maximal lift coeﬃcient. It seems that the leadingedge vortex is ex
pelled by the wing deformation, with a maximal eﬀect at A
f
= 0.2 of chordwise
ﬂexing. When the lift decreases, the drag increases during midstroke. Addition
ally, ﬁgure 8.5 shows the streamlines to visualise the leadingedge vortex during
midstroke (t = 0.25T) for both spanwise and chordwise ﬂexing. For comparison,
also the case using no ﬂexing is shown. Compared to the case without ﬂexing the
leadingedge vortex detaches earlier (less close to the wing tip) for both ﬂexing
cases. The spanwise ﬂexing has less inﬂuence than the chordwise ﬂexing mode
shape. The inﬂuence of spanwise ﬂexing is minor, which was also reﬂected in the
8.3 Conclusions 177
timeaveraged force coeﬃcients, which were only marginally lower. However, the
chordwise ﬂexing induced the leadingedge vortex to burst almost at the middle
of the wing, leading to signiﬁcantly less lift, see table 8.2.
8.3 Conclusions
In this chapter, the result have been described of a preliminary investigation to
understand the eﬀects of wing ﬂexing. Therefore, a predeﬁned ﬂexing deformation
has been applied to a plunging airfoil in twodimensional forward ﬂight and to a
threedimensional ﬂapping wing in hovering ﬂight. First of all, the twodimensional
ﬂexing results in a similar eﬀect on the ﬂow and resulting forces as rotation of the
airfoil. With increasing ﬂexing amplitude, the larger eﬀective angle of attack leads
to the generation of negative drag or thrust. In addition, the inﬂuence of the ﬂexing
shape (quartercosine or halfcosine) of the airfoil was studied. A quartercosine
shaped ﬂexing results in signiﬁcantly higher thrust at less lift.
Secondly, a ﬂexing deformation has been applied to a threedimensional ﬂap
ping wing in hovering ﬂight. Two ﬂexing directions have been considered, span
wise and chordwise ﬂexing. While keeping the ﬂexing amplitude constant, the
chordwise ﬂexing aﬀects the ﬂow the most. For a chordwise ﬂexing amplitude of
A
f
= 0.2 the timeaveraged lift and drag coeﬃcients are signiﬁcantly decreased
and increased, respectively, such that the lifttodrag decreases with 24.8%. The
spanwise ﬂexing does not show a comparable impact, only 2.6%.
178 Inﬂuence of wing deformation by ﬂexing
(a) No ﬂexing
(b) Spanwise ﬂexing, A
f
= 0.2 (c) Chordwise ﬂexing, A
f
= 0.2
Figure 8.5 Streamlines on a ﬂexing wing in hovering ﬂight. Isosurface of Q = 1.0 are shown
for a threedimensional ﬂapping wing in hovering ﬂight at Re = 100 and midstroke angle of attack of
A
α
= 45
◦
at t = 0.25T. (a) shows the wing without ﬂexing, while (b) shows the wing using spanwise
ﬂexing and (c) using chordwise ﬂexing, 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 ﬂow around ﬂapping foils and wings. Diﬀerent
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 eﬃciently solve the ﬂow around ﬂapping
foils and wings at low Reynolds numbers, for ﬂow conditions corresponding to the
scale of ﬂying insects. The improved and implemented mesh motion technique,
is used to solve the ﬂow around ﬂapping wings in hovering and forward ﬂapping
ﬂight. Additionally, a preliminary study of the inﬂuence of wing ﬂexing has been
conducted as well.
The overall conclusions of this research are given in section 9.1. Secondly, the
conclusions on the assessment of diﬀerent mesh motion techniques are drawn in
section 9.2. Section 9.3 presents the conclusions of a ﬂapping wing under hovering
conditions, for twodimensional airfoil as well as for the threedimensional wing.
Section 9.4 shows the results for forward ﬂapping ﬂight, while section 9.5 describes
the conclusions of the preliminary study to understand the eﬀects of wing ﬂexing.
Additional recommendation are made in Section 9.6.
9.1 Overall conclusions
The following overall conclusions are drawn:
1. The ﬂow around a ﬂapping wing, at the scale relevant to insect ﬂight, can
be solved accurately using advanced mesh motion techniques in existing ﬂow
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 modiﬁed stress equation. How
ever, mesh motion based on radial basis function interpolation is much more
computationally demanding. Diﬀerent ways to improve its eﬃciency are
discussed.
3. The wing kinematic pattern has a large inﬂuence on the forces in two
dimensional hovering and forward ﬂight.
4. In forward ﬂight, thrust generation is the result of the forward tilting of the
wing.
5. In two and threedimensional ﬂapping ﬂight, the leadingedge vortex is a
very important lift enhancement mechanism.
6. The Rossby number, which describes the wing stroke curvature, has a large
inﬂuence on the timeaveraged lift force. A low Rossby number signiﬁcantly
increases lift as the result of an induced spiralling leadingedge vortex.
7. Wing ﬂexing may play an important role in insect ﬂight 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 ﬁrst method solves the Laplace equation with
a variable diﬀusion coeﬃcient, which is used to control the ﬁnal mesh quality.
Secondly, a modiﬁed 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 modiﬁed
stress equation, the mesh quality is not suﬃcient for ﬂapping wing cases,
where the rotation angles are large. However, these methods are very eﬃcient
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, diﬀerent radial basis functions, concerning the RBF mesh motion,
are compared. The RBF mesh quality provides superior mesh quality over
the Laplace and modiﬁed 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 ﬂapping ﬂight 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, diﬀerent methods are implemented to increase its eﬃciency. 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 ﬂapping ﬂight
9.3.1 Twodimensional hovering
The eﬀects of wing motion kinematics on the aerodynamic characteristics of hov
ering insect ﬂight have been investigated by means of twodimensional numerical
ﬂow simulations. The results of the present twodimensional study has provided
useful insights, which may be relevant also for the understanding of realistic three
dimensional insect ﬂight.
Four diﬀerent kinematic models, with diﬀerent complexity, have been analysed.
Two of these models, pure harmonic motion and the Roboﬂy experimental kine
matics have been used extensively in literature. The most prominent aspects of
the Roboﬂy kinematic model are the ‘sawtooth’ displacement and the ‘trapezoidal’
angle of attack. The third model represents the actual fruit ﬂy kinematics as ob
served in experiments and the last one was a modiﬁcation of the latter, chosen
to investigate the eﬀect of symmetry. The fruit ﬂy 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 quasisteady lift coeﬃcient was matched. The following conclusions
are drawn about the twodimensional hovering simulations:
1. It was found that the realistic fruit ﬂy wing kinematics result in signiﬁcantly
lower drag at similar lift compared with the simpliﬁed wing kinematic models
used in literature. The trend that the fruit ﬂy kinematics increases aerody
namic performance agrees with the predictions of the quasisteady theory,
but the numerical ﬂow simulations provide a more complete quantitative
analysis of the ﬂow behaviour.
2. It was shown that the diﬀerence in performance in terms of mean lifttodrag
ratio between the diﬀerent kinematic models was signiﬁcant. The mean
aerodynamic drag at equal lift of the fruit ﬂy models is about 49% lower
compared to the Roboﬂy model and about 29% lower with respect to the
harmonic model.
182 Conclusions and recommendations
3. The ‘sawtooth’ amplitude used in the Roboﬂy model has a small eﬀect on
the mean lift but the mean drag is aﬀected signiﬁcantly. 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 simpliﬁcation used by the Roboﬂy, the ‘trapezoidal’ angle
of attack, caused the LEV to separate during the translational phase. This
led to an increase in mean drag during each halfstroke. Also in this case
large accelerations at stroke reversal lead to a decrease in lifttodrag ratio
of 33%.
5. The extra ‘bump’ in angle of attack as used by the fruit ﬂy model is not
aﬀecting 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 signiﬁcant decrease in drag
which improves aerodynamic performance in the sense of lifttodrag ratio
by 15.6%.
6. The other realistic kinematic feature is the deviation, which is found to have
only a marginal eﬀect on the mean lift and mean drag. However, the eﬀective
angle of attack is altered such that the deviation leads to levelling of the force
distribution over the ﬂapping cycle.
9.3.2 Threedimensional hovering
A numerical model has been developed for solving the ﬂow around a threedimen
sional wing with complex wing kinematics. Diﬀerent aspects, relevant to three
dimensional ﬂapping wing aerodynamics, have been studied. First, the ﬂow around
a dynamically scaled model wing is solved for diﬀerent angles of attack in order to
study the force development and vortex dynamics at small and large midstroke
angle of attack. Secondly, the Rossby number is varied at diﬀerent Reynolds num
bers. A varying Rossby number represents a variation in stroke path curvature
and thus angular acceleration. Thirdly, the threedimensional wing kinematics is
varied by changing the shape in angle of attack and by applying a deviation, which
may cause a ‘ﬁgureofeight’ pattern. Finally, the threedimensional ﬂow is com
pared with the twodimensional studies performed on ﬂapping forward ﬂight. The
following conclusions are drawn with respect to the threedimensional simulations,
concerning hovering ﬂight:
1. The eﬀect of a variation in angle of attack is such that the maximal lift
occurs at a midstroke angle of α = 45
◦
. However, the performance, in
terms of maximal lifttodrag ratio was found to be maximal at α = 30
◦
.
The ﬂapping motion induces a leadingedge vortex which caused a peak in
lift halfway during each up and downstroke. This leadingedge vortex is
9.3 Conclusions on hovering ﬂapping ﬂight 183
strong at suﬃciently high angles of attack α > 30
◦
. At α = 15
◦
the leading
edge vortex is less strong, causing a sudden decrease in both lift and drag
during each halfstroke.
2. Both timeaveraged lift and drag decrease signiﬁcantly with increasing Ross
by number. At Ro = 130, a nearly translating wing, the lift decreases
with 24.7%, compared to the ﬂapping wing with Ro = 3.2. The major
decrease in lift and drag occurs during midstroke, between t = 0.25T and t =
0.5T. Flow visualisations showed that the leadingedge vortex is signiﬁcantly
reduced in size and strength for the translating wing at Ro = 130, compared
to the ﬂapping wing. In addition, the leadingedge vortex rollsup to form
two tip vortices instead of one in case of the ﬂapping wings. This causes
both lift and drag to be signiﬁcantly lower at Ro = 130.
3. It was seen that for increasing Reynolds number, both timeaveraged lift
and drag increases, but the diﬀerences become smaller. The eﬀect 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 leadingedge vortex for a ﬂapping
wing at Ro = 3.2. For both ﬂapping (Ro = 3.2) and translation (Ro = 130),
irregularities in the leadingedge 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 leadingedge vortex bursts, causing a
loss in performance.
4. The ‘trapezoidal’ shape in angle of attack causes a fast pitchup motion at
the beginning of each up and downstroke. Therefore, the midstroke 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 signiﬁcant increase in timeaveraged lift of 10.8%. The lift increase is
accompanied by a decrease in drag of a similar amount, such that the liftto
drag ratio shows a signiﬁcant gain of 21.9%. Twodimensional studies showed
an opposite result which was caused by the fact that the leadingedge vortex
separated, while translating at a constant angle of attack. Threedimensional
eﬀects, however, lead to a ﬁrm and stably attached leadingedge vortex.
5. Three diﬀerent wing tip patterns, caused by deviation, have been studied,
such as the ‘ﬁgureofO’, ‘ﬁgureofU’ and the ‘ﬁgureofeight’. The ‘ﬁgure
ofO’ wing tip pattern leads to an asymmetric variation in eﬀective angle of
attack, leading to diﬀerences in drag for the up and downstroke. Nonethe
less, the eﬀect of this pattern on the timeaveraged forces coeﬃcients is
negligible. On the other hand, the ‘ﬁgureofU’ pattern does inﬂuence the
forces in a nonbeneﬁcial way. The eﬀective angle of attack is signiﬁcantly
decreased during midstroke. The timeaveraged lift coeﬃcient is decreased
184 Conclusions and recommendations
by 9.4%. While drag is maintaining nearly constant, the lifttodrag ratio is
decreased by about 9%. The ‘ﬁgureofeight’ aﬀects the forces most, how
ever, it depends on the starting direction. Two types are considered, the
ﬁrst starting the downstroke with an downward deviation motion, whereas
the type 2 starts upwards. The eﬀective angle of attack is being inﬂuenced
in such a way that the type 1 ‘ﬁgureofeight’ decreases both lift and drag
considerably. Lift decreases up to 53%. Since drag decreases fast as well,
the loss in lifttodrag ratio is limited to 15.5%. The type 2 ‘ﬁgureofeight’
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
lifttodrag ratio decreases with about 12.0%.
9.4 Conclusions on forward ﬂapping ﬂight
9.4.1 Twodimensional forward ﬂapping
A numerical model for twodimensional ﬂow has been used to investigate the ef
fect of motion kinematics on the vortex dynamics around an ellipsoid foil subjected
to prescribed ﬂapping 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 twodimensional forward ﬂapping
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 ﬂapping 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 ﬂow is periodic.
9.4.2 Threedimensional forward ﬂapping
In addition to the investigation of the eﬀects of Reynolds number, Rossby number
and wing kinematics in hovering ﬂight, a preliminary study is performed to com
pare two and threedimensional forward ﬂight. Both two and threedimensional
simulations are dynamically scaled using the radius of gyration to justify compar
ison. The threedimensional simulations, concerning forward ﬂight 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 ﬂexing 185
1. For the case without rotation the threedimensional simulations result in
14.5% less lift compared to its twodimensional counterpart. This diﬀerence
is mainly caused by the loss in lift due to the generation of a tip vortex which
is only present in the threedimensional simulations.
2. An increase in rotation angle resulted in signiﬁcant higher lift, up to 70.5%,
which is the opposite as in the twodimensional plunging wing. Together
with the higher thrust, this leads to the conclusion that a stable leadingedge
vortex plays a more important role in threedimensional ﬂapping compared
to twodimensional 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 diﬀerent rotation amplitudes at a Reynolds number of Re = 150.
For the translating wing (Ro = 130), the threedimensional simulations re
sult in 28% less lift compared to the twodimensional case. This diﬀerence is
mainly caused by the loss in lift due to the generation of a tip vortex which
is only present in the threedimensional simulations. However, an increase in
Rossby number resulted in a signiﬁcant 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
threedimensional ﬂapping aerodynamics.
9.5 Preliminary conclusions on wing ﬂexing
A ﬂexing deformation has been applied to a plunging airfoil in twodimensional
forward ﬂight and to a threedimensional ﬂapping wing during hovering ﬂight.
Concerning the ﬂexible airfoil in forward ﬂight, a comparison is made with a
plunging airfoils, with additional rotation. The following conclusion are drawn:
1. The twodimensional ﬂexing has a comparable eﬀect on the ﬂow and forces as
rotation of the airfoil. With increasing ﬂexing amplitude, the larger eﬀective
angle of attack leads to the generation of negative drag or thrust. Besides,
the ﬂexing shape of the airfoil is important. A quartercosine shaped ﬂexing
results in signiﬁcantly higher thrust, while lift decreases.
2. The ﬂexing deformation was also applied to a threedimensional ﬂapping
wing in hovering ﬂight. Two ﬂexing directions were considered, spanwise and
chordwise ﬂexing. While keeping the ﬂexing amplitude constant, the chord
wise ﬂexing aﬀects the ﬂow signiﬁcantly. For a chordwise ﬂexing amplitude
of A
f
= 0.2 the timeaveraged lift and drag coeﬃcients are signiﬁcantly de
creased and increased, respectively, such that the lifttodrag decreases with
24.8%. The spanwise ﬂexing does not show a comparable inﬂuence, only
2.6%.
186 Conclusions and recommendations
9.6 Recommendations
The current thesis describes a comparison of diﬀerent mesh deformation methods
and development of an improved method, based on radial basis function interpo
lation. Additionally, these methods are used to solve the ﬂow around two and
threedimensional ﬂapping airfoils and wings under hovering and forward ﬂight
conditions. The following recommendation can be made for further research.
• The eﬃciency 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 signiﬁcantly reduced.
Additionally, the implementation of parallel iterative solver techniques may
increase its eﬃciency even further.
• The current implementation of the radial basis function mesh motion in
OpenFOAM
should be generally implemented in parallel. The current
method is only able to address up to 4 processors, depending on the mesh
partitioning.
• Using the RBF mesh motion solver, the ﬂow around multiple ﬂapping foils
or wings can be investigated. The current thesis describes the ﬂow ﬁelds in
duced by a single twodimensional foil or threedimensional wing. If multiple
wings are modelled, i.e. a dragonﬂy, the ﬂow will be more complex and the
force development may be aﬀected to certain extent.
• Investigate the eﬀects of transition and turbulence modelling on the ﬂow
ﬁelds. The current research considered scales at the laminar ﬂow regime. It
would be interesting to study the eﬀects of turbulence on the force develop
ment and vortex dynamics induced by a ﬂapping threedimensional wing.
• Investigate in more detail the vortex wake synchronisation in threedimen
sional forward ﬂight. Concerning twodimensional forward ﬂapping ﬂight,
this thesis describes the vortex wake synchronisation of a ﬂapping foil for
diﬀerent kinematic parameters. It would be interesting to study the vor
tex wake synchronisation of a threedimensional ﬂapping wing. The main
diﬃculty will be to deﬁne a proper framework to identify and quantify the
threedimensional vortex structures to study the pattern formation.
• The eﬀects of wing deformation could be investigated in more detail using
more complex ﬂexing models. The current thesis describes a preliminary
investigation of the eﬀects of predeﬁned wing ﬂexing. Further research could
introduce more degrees of freedom in the ﬂexing model, possibly based on
real insects. This may lead to advanced wing shapes, which result in optimal
aerodynamic eﬃciency, compared to rigid wings.
• It would be interesting to use ﬂuidstructure interaction methods to study
the inﬂuence of the ﬂow on the wing shapes. The wing shape is deformed
9.6 Recommendations 187
by inertial and aerodynamic forces. It will be interesting to couple these
forces with the ﬂow solver and the accompanying mesh motion method. The
main diﬃculty will be the coupling of ﬂow and structure, since the scaled
density of a fruit ﬂy wing is of similar order as the surrounding ﬂuid. Such
a strongly coupled problem is very sensitive and a converged solution is not
an easy objective.
• When advanced wing motions are desired, the possibility to implement im
mersed boundary methods needs to be explored. Using immersed boundary
methods it will become possible to model the clapandﬂing motion, when
two wings touches each other. Another application of the immersed bound
ary method will be the use of multiple wings and bodies with extreme body
motion to simulate advanced manoeuvring.
APPENDIX A
Grid generation for ﬂapping wings
To solve the ﬂow around ﬂapping wings using Computational Fluid Dynamics
(CFD), it is important to create high quality meshes. Three diﬀerent tools are used
for that purpose. The ﬁrst, blockMesh, is a mesh generation utility supplied with
OpenFOAM
. This utility creates a parametric mesh with grading and curved
edges. A practical use of blockMesh is limited to simple domains and geometries.
The second mesh generation software is Gambit, supplied with Fluent
, and is ca
pable of generating high quality meshes around complex geometries. Nevertheless,
using Gambit for structured meshes can be diﬃcult since deﬁning an appropriate
grading can be cumbersome. The third grid generation tool is the commercial
GridPro
package. GridPro
generates a structured grid with grading using a
block strategy around a complex geometry. The user generates a topology and the
grid solver creates a structured block such that the resulting mesh quality is high.
A.1 Introduction
This appendix deals with the mesh generation in order to obtain a high quality
initial mesh to solve for the ﬂow around a moving airfoil or wing. There are two
types of mesh generation, unstructured and structured. In principle, OpenFOAM
and Fluent
are capable of solving the discretised equations in an unstructured
way, so there are no problems using one of those types. Since the meshes need to
deform the initial mesh needs to be of very high quality, which can be more easily
obtained using a structured approach. Besides, the iterative solvers converge faster
on structured meshes. However, mesh generation around complex bodies is much
easier using unstructured meshing techniques, but since the present study deals
190 Grid generation for ﬂapping wings
b1 b2 b3
b4
b5 b6 b7
b8
p1 p2
p3 p4
X
Y
Figure A.1 The topology used by the
blockMesh utility. A blockMesh topology con
sists of points, lines and block. All need to be
speciﬁed manually, which can be cumbersome.
Figure A.2 A mesh generated by the
blockMesh utility. A mesh generated by
blockMesh, which can be obtained very fast,
but is limited to simpliﬁed geometries.
with simpliﬁed model airfoils and wings, it is decided to choose structured meshing
strategies. The meshes used in the present thesis were generated using one of the
earlier mentioned meshing tools: blockMesh, Gambit or GridPro
. These three
tools are described in sections A.2, A.2 and A.2, respectively.
A.2 BlockMesh
The opensource mesh generator, supplied with OpenFOAM
, is blockMesh. It is
an easy to use and robust utility, applicable for simpliﬁed cases. The computational
domain is speciﬁed by points, lines and blocks, see ﬁgure A.1. The connections be
tween the points (p1−p4) are deﬁned by lines, which can be curved. Additionally,
the domain consists of blocks (b1 − b8), speciﬁed by the lines, accordingly. The
relation between the points, lines and blocks need to be speciﬁed in a dictionary
ﬁle, within the OpenFOAM
case. The grading to deﬁne the mesh resolution can
be set on all topology lines. For a uniform grading the mesh around a block is
shown in ﬁgure A.2.
Since the manual creation of topology can be cumbersome, this mesh gener
ation utility is limited to simple computational domains. Therefore, the current
thesis used blockMesh to generate the meshes around a twodimensional block
for testing the mesh deformation methods (chapter 3). Additionally, the meshes
are generated, concerning the twodimensional channel cases, used to test the
OpenFOAM
ﬂow solver with vortex decay and convection (chapter 2).
A.3 Gambit 191
Figure A.3 Mesh around a circular cylinder, generated by Gambit. Meshes around complex
geometries is possible using Gambit, but clustering of cells is the main diﬃculty.
A.3 Gambit
The commercial ﬂow solver Fluent
also provides the mesh generator Gambit.
Gambit is able to generate structured and unstructured meshes, the latter con
taining only tetrahedral cells. It was already explained that the current thesis
makes only use of structured meshes, Gambit was used to generate these meshes
using hexahedral cells. Within Gambit, a strong graphical user interface is avail
able, such that meshes around complex CAD designed geometries are possible.
However, in order to generate a structured mesh with appropriate cell clustering
near the body at regions with large geometric gradients, it is invincible to result
with high mesh resolution throughout the computational domain, see ﬁgure A.3.
It is shown that a grid is generated using multiple blocks, generated manually,
in order to maintain high mesh quality. These multiple blocks, lead to an excess
of cells in regions where these are not necessary. This can be solved by gener
ating more blocks, which is not straightforward. Gambit is used to generate the
twodimensional meshes, which are used to test both Fluent
and OpenFOAM
using stationary and plunging cylinder ﬂows. The next grid generator, GridPro
,
is capable to generate the multiple blocks automatically, which will improve the
mesh quality considerably, especially for threedimensional cases.
A.4 GridPro
GridPro
is an advanced commercial multiblock structured mesh generator. The
usage is diﬀerent compared to other mesh generators, because the multiple blocks
are automatically generated to optimise mesh quality. Figure A.4 shows the topol
ogy, which needs to be generated, this approach is diﬀerent compared to the other
grid generators, blockMesh and Gambit. In ﬁgure A.4, the inner red squared
block is snapped to the adjacent boundary Γ
b
, and all other red squared regions
192 Grid generation for ﬂapping wings
Γ
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
generated the mul
tiple blocks. These blocks are shaped such
that the mesh quality is optimised.
Figure A.5 Grid around a circular
cylinder, generated by GridPro
. The
grid is generated by GridPro
according to
the userdeﬁned mesh topology, such that the
mesh quality is optimised.
represent structured grid blocks. An illustration of the resulting grid is shown in
ﬁgure A.5. Besides the high quality meshes, GridPro
is easy to use for complex
geometries. Additionally, for threedimensional cases, it is very important to put
the multiple blocks in an optimal way such that the cell clustering is only present
in the regions of interest. Therefore, the current thesis uses GridPro
to generate
the threedimensional meshes around a ﬂapping wing, as illustrated in ﬁgure A.6.
A.5 Conclusions
Three diﬀerent mesh generators have been described, blockMesh, Gambit and
GridPro
. The mesh generator, blockMesh, supplied with OpenFOAM
, is easy
to use for simpliﬁed problems. In the current thesis, blockMesh is used to gen
erate the twodimensional meshes, which were used to validate the ﬂow 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
twodimensional ﬂows around stationary and plunging cylinders. Using the third
grid generator, GridPro
, a userdeﬁned topology needs to be speciﬁed to optimise
the multiple blocks for the structured mesh generation. This topology procedure
is versatile, but diﬀerent compared to the other packages. GridPro
is used for
the threedimensional simulations around ﬂapping wings.
A.5 Conclusions 193
(a) 3D GridPro
mesh
(b) 3D GridPro
mesh, closeup at t = 0T (c) 3D GridPro
mesh, closeup at t = 0.5T
Figure A.6 Threedimensional grid around a wing, generated by GridPro
. (a) shows the
full computational domain which is used for mesh generation using GridPro
. A closeup of the mesh
near the threedimensional wing is shown for two diﬀerent time instances in (b) and (c), respectively
t = 0T and t = 0.5T.
APPENDIX B
Flow solver settings
B.1 Introduction
Within the current thesis, two diﬀerent ﬂow solvers have been used, the open
source OpenFOAM
and the commercial package Fluent
. For both ﬂow solvers,
the settings are described in this appendix. Section B.2 deals with Fluent
,
whereas section B.3 describes the settings used in OpenFOAM
.
B.2 Fluent
solver settings
Fluent
is a ﬁnite volume based CFD solver. This section deals with the dif
ferent ﬂow solver settings, that are necessary to reproduce the results from this
thesis. Fluent
is used for the twodimensional hovering simulations described in
chapter 5.
Solver Segregated
Space 2D/3D
Time ﬁrstorder implicit
Velocity formulation Absolute
Gradient option Cellbased
Viscous model laminar
Accuracy double precision
Table B.1 Solver settings
196 Flow solver settings
Pressure secondorder
PressureVelocity coupling PISO
Equations secondorder Upwind
Table B.2 Discretisation settings
Smoothing: Spring constant 0.1
Boundary node relaxation 0.3
Convergence tolerance 0.01
Number of iterations 20
Remeshing: min. cell volume 1.73e7  3e7 (grid dependent)
max. cell volume 0.488  1.39 (grid dependent)
max. cell skewness 0.4
Table B.3 Dynamic mesh settings
In order to solve the ﬂow, the user needs to specify which models are used, see
table B.1. The laminar viscous model is used which is quite misleading. No turbu
lence model is used so in fact a Direct Numerical Simulation (DNS) is performed
using the NavierStokes equations. For the highest accuracy the ‘double precision’
version is used. According to (Lentink, 2003, Bos et al., 2008) this high accuracy
is necessary since insect ﬂow might be very sensitive to initial conditions, under
certain circumstances.
The moving wings were studied using the ‘dynamic mesh’ module. Up to now
Fluent
is only capable of using this ‘dynamic mesh’ module in combination with
the ﬁrstorder implicit time integration. The validation cases without mesh motion
uses secondorder time integration. The multigrid settings worked ﬁne at default
settings. Table B.2 shows the discretisation settings. For space discretisation a
secondorder upwind scheme is used together with standard PISO scheme for the
pressurevelocity coupling.
The ‘dynamic mesh’ parameters are shown in table B.3. The mesh is moving
using two methods, smoothing and remeshing. Remeshing means a complete
examination of the mesh and adaptation of the nodes where needed. Smoothing
on the other hand holds the nodes together in such a way that they do not move
arbitrarily in any direction, but stay together in a way. Remeshing is deﬁned
by the maximal and minimal cell volumes. These values are bases on the grid
in that case. The maximal skewness is needed in order to keep the mesh quality
within acceptable range. A value of 0.4 turned out to give satisfying results. The
smoothing performs an iteration to smooth the mesh when it is updated. The
smoothness is given by a spring constant which holds the nodes together and a
boundary node relaxation which gives some freedom to boundary nodes to move.
Concerning unsteady cases, the solution is varying in time until the residuals
B.3 OpenFOAM
solver settings 197
reach a suﬃciently small value. These values are convergence criterion’s and can
be changed by the user. In this study the solution was considered converged as
the residuals reached a value of 1 · 10
−4
for every component. Per timestep a
ﬁxed number of 20 iterations was needed to converge in case of using the PISO
pressurevelocity coupling in case of the validation study. During the ﬂapping wing
simulations a ﬁxed number of 10 iterations was used. Furthermore the complete
solution is written to harddisk several times per ﬂapping period; the lift and drag
histories every timestep.
B.3 OpenFOAM
solver settings
Since Fluent
could not cope with the extreme threedimensional mesh deforma
tion, OpenFOAM
was explored and provided good results. OpenFOAM
(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 sourcecode is fully accessible, diﬀerent ﬂow solvers and utilities are
developed. For example, the unsteady NavierStokes 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 nonorthogonality). One of the
main strength’s of OpenFOAM
is the intuitive way of programming. Brieﬂy, two
diﬀerent illustrations are given, from the solver icoFoam and utility totalEnergy.
The following piece of code is taken from icoDyMFoam, which solves the unsteady
incompressible NavierStokes equations:
1 for (runTime++; !runTime.end(); runTime++)
2 {
3 Info<< "Time = " << runTime.timeName() << nl << endl;
4
5 fvVectorMatrix UEqn
6 (
7 fvm::ddt(U)
8 + fvm::div(phi, U)
9  fvm::laplacian(nu, U)
10 );
11
12 solve(UEqn == fvc::grad(p));
13
14 runTime.write();
15 }
Line 1 speciﬁes that the equations are solved for all timesteps until the speciﬁed
end time is reached, runTime.end(). Line 59 deﬁnes the implicit part of the
governing equation, which is the unsteady incompressible NavierStokes equation.
The implicit discretisation is performed by the fvm class such that the total system
of equations is properly constructed from the temporal, convection and diﬀusion
198 Flow solver settings
terms, i.e. ddt and div are explicit functions to generate the matrix system for
a given velocity, pressure and ﬂux ﬁeld. The explicit fvc is used, on line 12,
to equate the system with the source terms and the complete system is solved
using a chosen iterative solver. This piece of code needs to be placed within a
pressurevelocity coupling loop.
The next code example concerns the utility totalEnergy, which calculates the
total energy, integrated over the complete domain:
1 void Foam::calc(const argList& args, const Time& runTime, const fvMesh& mesh)
2 {
3 IOobject Uheader
4 (
5 "U",
6 runTime.timeName(),
7 mesh,
8 IOobject::MUST_READ
9 );
10
11 if (Uheader.headerOk())
12 {
13 Info<< " Reading U" << endl;
14 volVectorField U(Uheader, mesh);
15
16 Info << " Calculating totalEnergy" << endl;
17 dimensionedScalar totalEnergy = fvc::domainIntegrate(0.5*magSqr(U));
18 Info << " Total energy = " << totalEnergy.value() << endl;
19 }
20 else
21 {
22 Info << " No U" << endl;
23 }
24 }
Here, from line 124, the timeloop is deﬁned, calculating the total energy for every
time instance available. Line 39 reads the velocity ﬁeld from a given solution, at
the speciﬁc time directory. The actual calculation is performed at line 17, where
fvc::domainIntegrate is an implicit function, calculating the sum over all ﬁnite
volume cells in the domain of 0.5*magSqr(U), which is equivalent to 0.5u
2
. The
statement Info is a templated function, which is able to return strings, scalar
values and tensor ﬁelds back to the screen. These two illustrations are only two
examples. OpenFOAM
comes with a wide variety of solvers, utilities and tutorial
cases. If there is need for a speciﬁc application, the uses should take a look in the
sourcecode of OpenFOAM
to ﬁnd similar pieces of code.
All ﬂow solvers that are developed for this thesis, are based on icoDyMFoam,
slightly extended to use force output modiﬁcations, or to make use of modiﬁed
mesh motion solvers. icoDyMFoam solves the unsteady incompressible laminar
NavierStokes equations for a Newtonian ﬂuid. Therefore, the diﬀerent terms
of this equation need to be discretised accordingly, e.g. diﬀusion, convection.
Table B.4 shows the schemes that were used throughout this thesis. The convection
scheme was varied for validation purposes, but the Van Leer scheme was used for
the majority of numerical simulations.
In order to solve discretised governing equations, an iterative solver is used.
Three diﬀerent solvers were speciﬁed for the pressure equation, velocity equation
B.3 OpenFOAM
solver settings 199
Description Code Diﬀerencing scheme
General interpolation  secondorder linear
Temporal discretisation ddt(U) secondorder backward
Gradient discretisation div(phi,U) secondorder linear
Diﬀusion discretisation laplacian(nu,U) secondorder linear
Convection discretisation grad(p) Gamma, SuperBee, Koren,
Van Leer or linear
Table B.4 Diﬀerencing methods for diﬀerent terms in the transport equation. The tem
poral, gradient, diﬀusion and convection term, present in the general transport equations needs to be
discretised properly. To minimise temporal errors all chosen schemes are of secondorder, see (Weller
et al., 1998, Jasak, 1996, Jasak et al., 2004).
and mesh motion, respectively. Table B.5 shows the chosen solvers, combined with
the convergence criterion. For solving the equations for pressure and velocity, the
PISO coupling. Every timestep PISO evaluates an initial u and p, performs
multiple corrections (commonly twice in this thesis) until a convergence criterion
is met. An iterative method like PISO can be accelerated by Krylov subspace
methods (Saad, 2003). This means that the matrix A of the system Ay = b to be
solved is split as: A = M − N. M is used to precondition the problem, which
means that Ay = b is replaced by its preconditioned counterpart:
M
−1
Ay = M
−1
b. (B.1)
In this thesis, Incomplete Cholesky decomposition is used to precondition the
system as follows:
L
−1
AL
−T
˜ y = L
−1
b, y = L
−T
˜ y , (B.2)
where LL
T
is an Incomplete Cholesky decomposition of A and L
−T
= (L
T
)
−1
.
When Incomplete Cholesky (IC) decomposition is applied in combination with
Conjugate Gradient (CG), this is another Krylov subspace method for linear sys
tems Ay = b with a symmetric selfadjoint positive deﬁnite matrix A. Combining
Term iterative solver convergence criteria
pressure, p PCG with DIC precond 10
−6
velocity, u PBiCG with DILU precond 10
−5
mesh motion PCG with DIC precond 10
−8
Table B.5 Iterative solvers for the diﬀerent equations. The pressure and mesh motion equa
tions are solved using the preconditioned conjugate gradient (PCG) solver with an diagonal incomplete
Choleski (DIC) preconditioner. The pressurevelocity coupling equation is solved using the asymmetric
solver preconditioned BiStab conjugate gradient (PBiCG), with a diagonal incomplete LU decompo
sition (DILU) preconditioner.
200 Flow solver settings
IC and CG results in a new method called ICCG. If the matrix A is not selfadjoint,
one can apply BiConjugate Gradient method (BCG). Using this in combination
with IC results in BiCG. For every variable calculated the linear solver methods
and their solution tolerances are listed in table B.5. A more elaborate description
of iterative solvers is beyond the scope of this thesis. For more information, please
consult (Wesseling, 2001, Hirsch, 1988, Ferziger & Peric, 2002, Jasak et al., 2007).
Bibliography
Anderson Jr., J. D. (1991), Fundamentals of Aerodynamics, second edn,
McGrawHill Inc.
Aono, H., Liang, F. & Liu, H. (2008), ‘Near and farﬁeld aerodynamics in
insect hovering ﬂight: an integrated computational study’, The Journal of Exper
imental Biology 211, 239–257.
Baruh, H. (1999), Analytical Dynamics, McGrawHill Inc.
Batina, J. T. (1990), ‘Unsteady euler airfoil solutions using unstructured dy
namic meshes’, AIAA Journal 28 (8), 1381–1388.
Beckert, A. & Wendland, H. (2001), ‘Multivariate interpolation for ﬂuid
structureinteraction problems using radial basis functions’, Aerospace Science and
Technology 5(2), 125–134.
Birch & Dickinson, M. H. (2003), ‘The inﬂuence of wingwake interactions
on the production of aerodynamic forces in ﬂapping ﬂight.’, The Journal of Ex
perimental Biology 206, 2257–2272.
Birch, J. M. & Dickinson, M. H. (2001), ‘Spanwise ﬂow and the attachment
of the leadingedge vortex on insect wings’, Nature 412, 729–733.
Birch, J. M., Dickson, W. B. & Dickinson, M. H. (2004), ‘Force production
and ﬂow structure of the leadingedge vortex on ﬂapping wings at high and low
reynolds numbers.’, The Journal of Experimental Biology 207, 1063–1072.
Blom, F. J. (2000), ‘Considerations on the spring analogy’, International Jour
nal for Numerical Methods in Fluids 32, 647–669.
202 Bibliography
Blondeaux, P., Fornarelli, F. & Guglielmini, L. (2005a), Vortex struc
tures generated by a ﬁnitespan oscillating foil, in ‘43rd AIAA Aerospace Sciences
Meeting and Exhibit, Reno’, 200584.
Blondeaux, P., Fornarelli, F., Guglielmini, L., Triantafyllou, M. S. &
Verzicco, R. (2005b), ‘Numerical experiments on ﬂapping foils mimicking ﬁsh
like locomotion’, Physics of Fluids 17, 113601–12.
Boer de, A., van der Schoot, M. S. & Bijl, H. (2007), ‘Mesh deformation
based on radial basis function interpolation’, Computers & Structures 85, 784–795.
Bomphrey, R. J., Lawson, N. J., K., T. G. & Thomas, A. L. R. (2006),
‘Application of digital particle image velocimetry to insect aerodynamics: mea
surement of the leadingedge vortex and near wake of a hawkmoth’, Experiments
in Fluids 40, 546–554.
Bos, F. M., Lentink, D., van Oudheusden, B. W. & Bijl, H. (2008), ‘In
ﬂuence of wing kinematics on aerodynamic performance in hovering insect ﬂight’,
Journal of Fluid Mechanics 594, 341–368.
Bos, F. M., van Oudheusden, B. W. & Bijl, H. (2010a), ‘Mesh deforma
tion techniques for ﬂapping ﬂight’, Computer Methods in Applied Mechanics and
Engineering (Submitted).
Bos, F. M., van Oudheusden, B. W. & Bijl, H. (2010b), ‘Threedimensional
vortical structures in ﬂapping ﬂight’, Journal of Fluid Mechanics (Submitted).
Brodsky, A. K. (1994), The Evolution of Insect Flight, Oxford University Press
Inc.
Buhmann, M. D. (2000), ‘Radial basis functions’, Acta Numerica 9, 1–38.
Carr, J. C., Beatson, R. K., McCallum, B. C., Fright, W. R., McLennan,
T. J. & Mitchell, T. J. (2003), Smooth surface reconstruction from noisy
range data, First International Conference on Computer Graphics and Interactive
Techniques.
Chakraborty, P., Balachandar, S. & Adrian, R. J. (2005), ‘On the rela
tionships between local vortex identiﬁcation schemes’, Journal of Fluid Mechanics
535, 189–214.
Combes, S. A. & Daniel, T. L. (2003a), ‘Flexural stiﬀness in insect wings i
& ii’, Journal of Experimental Biology 206, 2979–2997.
Combes, S. A. & Daniel, T. L. (2003b), ‘Into thin air: contributions of
aerodynamic and inertialelastic forces to wing bending in the hawkmoth Manduca
sexta’, Journal of Experimental Biology 206, 2999–3006.
Bibliography 203
Degand, C. & Farhat, C. (2002), ‘A threedimensional torsional spring analogy
method for unstructured dynamic meshes’, Computers & Structures 80, 305–316.
Dickinson, M. H. (1994), ‘The eﬀects of wing rotation on unsteady aerody
namic performance at low Reynolds number.’, The Journal of Experimental Biol
ogy 192, 179–206.
Dickinson, M. H., Farley, C. T., Full, R. J., Koehl, M. A. R., Kram,
R. & Lehman, S. (2000), ‘How animals move: An integrated view’, Science
288, 100–106.
Dickinson, M. H. & G¨ otz, K. G. (1993), ‘Unsteady aerodynamic performance
of model wings at low reynolds numbers’, The Journal of Experimental Biology
174, 45–64.
Dickinson, M. H., Lehmann, F.O. & Sane, S. P. (1999), ‘Wing rotation
and the aerodynamic basis of insect ﬂight.’, Science 284, 1954–1960.
Dickson, W. B. & Dickinson, M. H. (2004), ‘The eﬀect of advance ratio on
the aerodynamics of revolving wings’, Journal of Experimental Biology 207, 4269–
4281.
Donea, J. (1982), ‘An arbitrary lagrangianeulerian ﬁnite element method for
transient ﬂuidstructure interactions’, Computer Methods in Applied Mechanics
and Engineering pp. 689–723.
Dong, H., Mittal, R., Bozkurttas, M. & Najjar, F. (2005), Wake structure
and performance of ﬁnite aspectratio ﬂapping foils, in ‘43rd AIAA Aerospace
Sciences Meeting and Exhibit, Reno’, 200581.
Dwight, R. P. (2004), Robust mesh deformation using the linear elasticity equa
tions, in H. Deconinck & E. Dick, eds, ‘Computational Fluid Dynamics 2006’,
Springer.
Ellington, C. P. (1984), ‘The aerodynamics of hovering insect ﬂight’, Phil.
Trans. Roy. Soc. London. B 305, 1–181.
Ellington, C. P., van den Berg, C., Willmott, A. P. & Thomas, A. L. R.
(1996), ‘Leadingedge vortices in insect ﬂight’, Nature 384, 626–630.
Farhat, C., Degand, C., Koobus, B. & Lesoinne, M. (1998), ‘Torsional
springs for twodimensional dynamic unstructured meshes’, Computer Methods in
Applied Mechanics and Engineering 163, 231–245.
Ferziger, J. H. & Peric, M. (2002), Computational Methods for Fluid Dynam
ics, third edn, SpringerVerlag Berlin.
Fry, S. N., Sayaman, R. & Dickinson, M. H. (2003), ‘The aerodynamics of
freeﬂight maneuvers in Drosophila’, Science 300, 295–298.
204 Bibliography
Ginsberg, J. H. (1998), Advanced Engineering Mechanics, second edn, Cam
bridge University Press.
Guglielmini, L. & Blondeaux, P. (2004), ‘Propulsive eﬃciency of oscillating
foils’, European Journal of Mechanics B/Fluids 23, 255–278.
Guilmineau, E. & Queutey, P. (2002), ‘A numerical simulation of vortex
shedding from an oscillating circular cylinder’, Journal of Fluids and Structures
16, 773–794.
Helenbrook, B. T. (2003), ‘Mesh deformation using the biharmonic operator’,
International Journal for Numerical Methods in Engineering 56 (7), 1007–1021.
Henderson, R. D. (1995), ‘Details of the drag curve near the onset of vortex
shedding’, Physics of Fluids 7 (9), 2102–2104.
Hirsch, C. (1988), Numerical Computation of Internal and External Flows, Vol
ume I: Fundamentals of Numerical Discretization, John Wiley & Sons.
Hover, F. S., Haugsdal, Ø. & Triantafyllou, M. S. (2004), ‘Eﬀect of angle
of attack proﬁles in ﬂapping foil propulsion’, Journal of Fluids and Structures
19, 37–47.
Hu, D. L., Chan, B. & Bush, J. W. M. (2003), ‘The hydrodynamics of water
strider locomotion’, Nature 424, 663–666.
Hunt, J. C., Wray, A. A. & Moin, P. (1988), Eddies, stream, and convergence
zones in turbulent ﬂows, in ‘Center for Turbulence Research Report’, pp. 193–208.
Isogai, K., Fujishiro, Saitoh, T., Yamamoto, M., Yamasaki, M. & Mat
subara, M. (2004), ‘Unsteady threedimensional viscous ﬂow simulation of a
dragonﬂy hovering’, AIAA Journal 42, 2053–2059.
Issa, R. I. (1986), ‘Solution of the implicitly discretized ﬂuid ﬂow equation by
operator splitting’, Journal of Computational Physics 62, 40–65.
Jakobsson, S. & Amoignon, O. (2007), ‘Mesh deformation using radial ba
sis functions for gradientbased aerodynamics shape optimization’, Computers &
Fluids 36, 1119–1136.
Jasak, H. (1996), Error analysis and estimation in the ﬁnite volume method with
applications to ﬂuid ﬂow, PhD thesis, Imperial College, University of London.
Jasak, H. (2009), Dynamic mesh handling in openfoam, in ‘47th AIAA
Aerospace Sciences Meeting, Orlando’, 2009341.
Jasak, H., Jemcov, A. & Maruszewski, J. P. (2007), Preconditioned linear
solvers for large eddy simulations, in ‘CFD 2007 Conference’, CFD Society of
Canada, Toronto.
Bibliography 205
Jasak, H. & Tukovi´c, Z. (2004), ‘Automatic mesh motion for the unstructured
ﬁnite volume method’, Journal of Computational Physics (Submitted).
Jasak, H., Weller, H. G. & Gosman, A. D. (1999), ‘High resolution nvd
diﬀerence scheme for arbitrary unstructured meshes’, International Journal for
Numerical Methods in Fluids 31, 431–449.
Jasak, H., Weller, H. G. & Nordin, N. (2004), ‘Incylinder CFD simulation
using a c++ objectoriented toolkit’, SAE Technical paper 2004010110.
Jeong, J. & Hussain, F. (1995), ‘On the identiﬁcation of a vortex’, Journal of
Fluid Mechanics 285, 69–94.
Johnson, A. A. & Tezduyar, T. E. (1994), ‘Mesh update strategies in par
allel ﬁnite element computations of ﬂow problems with moving boundaries and
interfaces’, Computer Methods in Applied Mechanics and Engineering 119, 73–94.
Juntasaro, V. & Marquis, A. J. (2004), ‘Comparative study of ﬂuxlimiters
based on must diﬀerencing scheme’, International Journal of Computational Fluid
Dynamics 18 (7), 569–576.
Juretic, F. (2004), Error analysis in Finite Volume CFD, PhD thesis, Imperial
College London.
Katz, J. (1981), ‘A discrete vortex method for nonsteady separated ﬂow over
an airfoil’, Journal of Fluid Mechanics 102, 315–328.
Knupp, P. M. (2003), ‘Algebraic mesh quality metrics for unstructured initial
meshes’, Finite Elements in Analysis and Design 39, 217–241.
Koren, B. (1993), ‘A robust upwind discretisation method for advection, dif
fusion and source terms’, Numerical Methods for AdvectionDiﬀusion Problems
p. 117.
Kuzmin, D. & Turek, S. (2004), ‘Highresolution femtvd schemes based on a
fully multidimensional ﬂux limiter’, Journal of Computational Physics 198, 131–
158.
Lehmann, F.O. (2004), ‘The mechanisms of lift enhancements in insect ﬂight’,
Naturwissenschaften 91, 101–122.
Lehmann, F.O., Sane, S. P. & Dickinson, M. (2005), ‘The aerodynamic
eﬀect of wingwing interaction in ﬂapping insect wings’, The Journal of Experi
mental Biology 208, 3075–3092.
Lentink, D. (2003), Inﬂuence of airfoil shape on performance in insect ﬂight,
Master’s thesis, Delft University of Technology.
206 Bibliography
Lentink, D. (2008), Exploring the Bioﬂuiddynamics of Swimming and Flight,
PhD thesis, Wageningen University.
Lentink, D. & Dickinson, M. H. (2009a), ‘Bioﬂuiddynamic scaling of ﬂap
ping, spinning and translating ﬁns and wings’, Journal of Experimental Biology
212, 2691–2704.
Lentink, D. & Dickinson, M. H. (2009b), ‘Rotational accelerations stabilize
leading edge vortices on revolving ﬂy wings’, Journal of Experimental Biology
212, 2705–2719.
Lentink, D. & Gerritsma, M. (2003), Inﬂuence of airfoil shape on performance
in insect ﬂight, in ‘43rd AIAA Fluid Dynamics Conference and Exhibit, Orlando’,
20033447.
Lentink, D., Muijres, F. T., DonkerDuyvis, F. J. & van Leeuwen, J. L.
(2008), ‘Vortexwake interactions of a ﬂapping foil that models animal swimming
and ﬂight’, The Journal of Experimental Biology 211, 267–273.
Lesoinne, M. & Farhat, C. (1996), ‘Geometric conservation laws for ﬂow prob
lems with moving boundaries and deformable meshes, and their impact on aeroe
lastic computations’, Computer Methods in Applied Mechanics and Engineering
134, 71–90.
Lewin, G. C. & HajHariri, H. (2003), ‘Modelling thrust generation of a
twodimensional heaving airfoil in a viscous ﬂow’, Journal of Fluid Mechanics
492, 339–362.
Lighthill, M. J. (1969), ‘Hydromechanics of aquatic animal propulsion’, Annual
Review of Fluid Mechanics 1, 413–446.
Liu, H. & Kawachi, K. (1998), ‘A numerical study of insect ﬂight’, Journal of
Computational Physics 146, 124–156.
L¨ohner, R. & Yang, C. (1996), ‘Improved ale mesh velocities for moving bod
ies’, Communications in Numerical Methods in Engineering 12 (10), 599–608.
Luo, G. & Sun, M. (2005), ‘The eﬀects of corrugation and wing planform on
the aerodynamic force production of sweeping model insect wings’, Acta Mech
Sinica 21, 531–541.
Lu, Y. & Shen, G. X. (2008), ‘Threedimensional ﬂow structures and evolution
of the leadingedge vortices on a ﬂapping wing’, The Journal of Experimental
Biology 211, 1221–1230.
Maxworthy, T. (1979), ‘Experiments on the weisfogh mechanism of lift gener
ation by insects in hovering ﬂight. part 1. dynamics of the ‘ﬂing”, Journal of Fluid
Mechanics 93, 47–63.
Bibliography 207
Michelson, R. C. (2008), ‘Test and evaluation of fully autonomous micro air
vehicles’, ITEA Journal 29, 367–374.
Mittal, R. & Iaccarino, G. (2005), ‘Immersed boundary methods’, Annual
Review of Fluid Mechanics 37, 239–261.
Panton, R. L. (2005), Incompressible Flow, John Wiley & Sons.
Patankar, S. V. & Spalding, D. B. (1972), ‘A calculation procedure for heat,
mass and momentum transfer in threedimensional parabolic ﬂows’, International
Journal of Heat and Mass Transfer 15, 1787.
Pedro, G., Suleman, A. & Djilali, N. (2003), ‘A numerical study of the
propulsive eﬃciency of a ﬂapping hydrofoil’, International Journal for Numerical
Methods in Fluids 42, 493–526.
Peskin, C. S. (2002), ‘The immersed boundary method’, Acta Numerica
pp. 479–517.
Poelma, C., Dickson, W. B. & Dickinson, M. H. (2006), ‘Timeresolved
reconstruction of the full velocity ﬁeld around a dynamicallyscaled ﬂapping wing’,
Experiments in Fluids 41 (2), 1–13.
Potsdam, M. A. & Guruswamy, G. P. (2001), A parallel multiblock mesh
movement scheme for complex aeroelastic applications, Technical report, AIAA
20010716.
Ramamurti, R. & Sandberg, W. C. (2002), ‘A threedimensional compu
tational study of the aerodynamic mechanisms of insect ﬂight’, The Journal of
Experimental Biology 205, 1507–1518.
Rendall, T. C. S. & Allen, C. B. (2008a), Eﬃcient mesh motion tech
niques using radial basis functions with data reduction algorithms, in ‘46th AIAA
Aerospace Sciences Meeting and Exhibit’.
Rendall, T. C. S. & Allen, C. B. (2008b), ‘Improved radial basis function
ﬂuidstructure coupling via eﬃcient localised implementation’, International Jour
nal for Numerical Methods in Engineering 78 (10), 1188–1208.
Rendall, T. C. S. & Allen, C. B. (2008c), ‘Uniﬁed ﬂuidstructure interpo
lation and mesh motion using radial basis functions’, International Journal for
Numerical Methods in Engineering 74 (10), 1519–1559.
Rhie, C. M. & Chow, W. L. (1983), ‘Numerical study of the turbulent ﬂow
past an airfoil with trailingedge separation’, AIAA Journal 21, 1525–1532.
Roe, P. L. (1986), ‘Characteristicbased schemes for the euler equations’, Annual
Review of Fluid Mechanics 18, 337.
208 Bibliography
Ruijgrok, G. J. J. (1994), Elements of airplane performance, Delft University
Press.
Saad, Y. (2003), Iterative Methods for Sparse Linear Systems, Society for In
dustrial Mathematics.
Sane, S. P. & Dickinson, M. H. (2001), ‘The control of ﬂight force by a
ﬂapping wing: lift and drag production’, The Journal of Experimental Biology
204, 2607–2626.
Sane, S. P. & Dickinson, M. H. (2002), ‘The aerodynamic eﬀects of wing
rotation and a revised quasisteady model of ﬂapping ﬂight’, The Journal of Ex
perimental Biology 205, 1087–1096.
Shyy, W., Lian, Y., Tang, J., Liu, H., Trizila, P. & Stanford, B. (2008a),
Computational aerodynamics of low reynolds number plunging, pitching and ﬂex
ible wings for mav applications, in ‘46th AIAA Aerospace Sciences Meeting and
Exhibit’.
Shyy, W., Lian, Y., Tang, J., Viieru, D. & Liu, H. (2008b), Aerodynamics
of low Reynolds number ﬂyers, Cambridge University Press.
Smith, M. J., Cesnik, C. E. S. & Hodges, D. H. (2000), ‘Evaluation of
some data transfer algorithms for noncontiguous meshes’, Journal of Aerospace
Engineering 13(2), 52–58.
Srygley, R. B. & Thomas, A. L. R. (2002), ‘Unconventional liftgenerating
mechanisms in freeﬂying butterﬂies’, Nature 420, 660–664.
Stepniewski, W. Z. & Keys, C. N. (1984), Rotarywing aerodynamics, Dover.
Sun, M. & Tang, J. (2002), ‘Unsteady aerodynamic force generation by a
model fruit ﬂy wing in ﬂapping motion’, The Journal of Experimental Biology
205, 55–70.
Sweby, P. K. (1984), ‘High resolution schemes using ﬂuxlimiters for hyperbolic
conservation laws’, SIAM Journal of Numerical Analysis 21, 995–1011.
Taylor, G. K., Nudds, R. L. & Thomas, A. L. R. (2003), ‘Flying and
swimming animals cruise at a strouhal number tuned for high power eﬃciency’,
Nature 425, 707–711.
Templin, R. J. (2000), ‘The spectrum of animal ﬂight: insect to pterosaurs’,
Progress in Aerospace Sciences 36, 393–436.
Tennekes, H. & Lumley, J. L. (1972), A ﬁrst course in turbulence, The MIT
Press.
Bibliography 209
Thaweewat, N., Bos, F. M., van Oudheusden, B. W. & Bijl, H. (2009),
Numerical study of vortexwake interactions and performance of a twodimensional
ﬂapping foil, in ‘47th AIAA Aerospace Sciences Meeting, Orlando’, 2009791.
Thomas, A. L. R., Taylor, G. K., Srygley, R. B., Nudds, R. L. & Bom
phrey, R. J. (2004), ‘Dragonﬂy ﬂight: freeﬂight and tethered ﬂow visualizations
reveal a diverse array of unsteady liftgenerating mechanisms, controlled primarily
via angle of attack’, The Journal of Experimental Biology 207, 4299–4323.
Triantafyllou, G. S., Triantafyllou, M. S. & Grosenbaugh, M. A. (1993),
‘Optimal thrust development in oscillating foils with application to ﬁsh propulsion’,
Journal of Fluids and Structures 7, 205–224.
Tukovi´c, Z. & Jasak, H. (2007), ‘Updated lagrangian ﬁnite volume solver for
large deformation dynamic response of elastic body’, Transactions of FAMENA
31, 1.
Usherwood, J. R. & Ellington, C. P. (2002), ‘The aerodynamics of revolving
wings III’, Journal of Experimental Biology 205, 1547–1576.
Van Den Berg, C. & Ellington, C. P. (1997), ‘The threedimensional leading
edge vortex of a ‘hovering’ model hawkmoth’, Phil. Trans. Roy. Soc. Lond. B Biol.
Sci. 352, 329–340.
Van Leer, B. (1979), ‘Towards the ultimate conservative diﬀerence scheme v.
a second order sequel to godunov’s method’, Journal of Computational Physics
32, 101.
Wang, Z. J. (2000a), ‘Two dimensional mechanism for insect hovering’, Physical
Review Letters 85(10), 2216–2219.
Wang, Z. J. (2000b), ‘Vortex shedding and frequency selection in ﬂapping ﬂight’,
Journal of Fluid Mechanics 410, 323–341.
Wang, Z. J. (2005), ‘Dissecting insect ﬂight’, Annual Review of Fluid Mechanics
37, 183–210.
Wang, Z. J. (2008), ‘Aerodynamic eﬃciency of ﬂapping ﬂight: analysis of a
twostroke model’, Journal of Experimental Biology 211, 234–238.
Wang, Z. J., Birch, M. B. & Dickinson, M. H. (2004), ‘Unsteady forces
and ﬂows in low reynolds number hovering ﬂight: twodimensional computations
vs robotic wing experiments’, The Journal of Experimental Biology 207, 461–474.
Wang, Z. J. & Przekwas, A. J. (1994), Unsteady ﬂow computation using
moving grid with mesh enrichment, Technical report, AIAA940285.
210 Bibliography
WeishFogh, T. & Jensen, M. (1956), ‘Biology and physics of locust ﬂight. i.
basic principles of insect ﬂight: a critical review’, Phil. Trans. Roy. Soc. London
239 (667), 415–458.
Weller, H. G., Tabor, G., Jasak, H. & Fureby, C. (1998), ‘A tensorial ap
proach to computational continuum mechanics using objectoriented techniques’,
Computers in Physics.
Wendland, H. (1996), Konstruktion und untersuchung radialer basisfunktionen
mit kompaktem tr¨ager, Technical report.
Wendland, H. (1998), ‘Error estimates for interpolation by compactly sup
ported radial basis functions of minimal degree’, Journal of Approximation Theory
93, 258–272.
Wendland, H. (1999), ‘On the smoothness of positive deﬁnite and radial func
tions’, Journal of Computational and Applied Mathematics 101, 177–188.
Wesseling, P. (2001), Principles of Computational Fluid Dynamics, Springer
Verlag Berlin.
White, F. M. (1991), Viscous Fluid Flow, second edn, McGrawHill Inc.
Williamson, C. H. K. (1988), ‘Deﬁning a universal and continuous strouhal
reynolds number relationship for the laminar vortex shedding of a circular cylin
der’, Physics of Fluids 31(10), 2742–2744.
Williamson, C. H. K. (1995), Fluid Vortices, Vol. 30 of Fluid Mechanics and
Its Applications, Kluwer Academic Publishers, chapter Vortex dynamics in the
wake of a cylinder.
Williamson, C. H. K. (1998), ‘A series in 1/
√
Re to represent the strouhal
reynolds number relationship of the cylinder wake’, Journal of Fluids and Struc
tures 12, 1073–1085.
Williamson, C. H. K. & Roshko, A. (1988), ‘Vortex formation in the wake
of an oscillating cylinder’, Journal of Experimental Biology 2, 355–381.
Willmott, A. P., Ellington, C. P. & Thomas, A. L. R. (1997), ‘Flow
visualization and unsteady aerodynamics in the ﬂight of the hawkmoth, Manduca
Sexta’, Phil. Trans. Roy. Soc. London 305, 303–316.
Wood, R. J. (2008), ‘The ﬁrst takeoﬀ of a biologically inspired atscale robotic
insect’, IEEE Trans. on Robotics 24, 341–347.
Young, J. & Lai, J. C. S. (2008), ‘Simulation and parameter variation of
ﬂappingwing motion based on dragonﬂy hovering’, AIAA Journal 46, 918–924.
Bibliography 211
Zhou, Y. C. & Wei, G. W. (2003), ‘High resolution conjugate ﬁlters for the
simulation of ﬂows’, Journal of Computational Physics 189, 159–179.
Zuijlen van, A. H. (2006), FluidStructure Interaction Simulations, Eﬃcient
Higher Order Time Integration of Partitioned Systems, PhD thesis, Delft Univer
sity of Technology.
Zuo, D., Peng, S., Chen, W. & Zhang, W. (2007), ‘Numerical simulation
of ﬂappingwing insect hovering ﬂight at unsteady ﬂow’, International Journal for
Numerical Methods in Fluids 53, 1801–1817.
Samenvatting
Wetenschappers in de biologie en techniek zijn altijd al gefascineerd geweest door
de vlucht van insecten en vogels. Gedurende een lange tijd bleef het aerodynamis
che mechanisme om de vlucht van insecten met ﬂappende vleugels te verklaren,
een raadsel. Tot een paar decennia terug. Experimenten lieten zien dat ﬂappende
vleugels een voorrand wervel veroorzaken, die aanzienlijke krachten veroorzaakt.
Het werd gevonden dat deze opgewekte krachten groter zijn dan bepaald met con
ventionele vliegtuig aerodynamica. Flappende vleugels produceren een lift en een
vooruit stuwende kracht, zodat insecten en zelfs kleine vogels, zoals de kolibrie,
stil kunnen blijven hangen en extreme manoeuvres uit kunnen voeren. Dankzij
deze veelzijdigheid vormen insecten en kleine vogels een krachtige bron van inspi
ratie voor de ontwikkeling van micro vliegtuigjes, kleine door de mens gemaakte
apparaatjes, inzetbaar voor ontdekkings en verkenningsmissies.
Verscheidene experimentele en numerieke visualisaties zijn uitgevoerd om de
kennis van de stroming rond ﬂappende vleugels te vergroten. Met deze kennis kun
nen micro vliegtuigjes worden ontworpen en verbeterd. Het eﬀect van de vleugel
beweging op de stroming en de krachten wordt nog steeds niet volledig begrepen.
Wij hebben twee en driedimensionale computersimulaties uitgevoerd, waarbij be
langrijke parameters voor de vleugelbeweging systematisch zijn gevarieerd. Om
de grenslaag en het zog goed in beeld te brengen, is het belangrijk om de kwaliteit
van het rekenrooster dicht bij de vleugel te behouden, in het bijzonder als de ro
tatiehoeken groot zijn. Daarom is het belangrijk om een nauwkeurige methode
te gebruiken om het rekenrooster te vervormen, die ook geschikt is voor grote
rotaties. Om in staat te zijn de stroming rond een ﬂappende vleugel uit te reke
nen, is het noodzakelijk om de techniek voor roostervervorming te optimaliseren.
Een belangrijk doel van dit proefschrift beschrijft een betrouwbare techniek voor
roostervervorming, in termen van nauwkeurigheid en eﬃci¨entie. Deze method is
tevens gebruikt om de stroming rond ﬂappende vleugels door te rekenen.
De stroming rond ﬂappende vleugels, op een schaal die relevant is voor de vlucht
van insecten, is sterk instationair, viskeus en wordt beschreven door de onsamen
214 Samenvatting
drukbare NavierStokes vergelijkingen. Verschillende dimensieloze getallen die het
stromingsgedrag karakteriseren, zijn beschreven, zoals het Strouhal en Reynolds
getal. Omdat de stroming bij het beschouwde Reynolds getal van Re = O(100)
zich laminair gedraagt, is er geen noodzaak om turbulentie te modelleren. Zo
doende mogen onze simulaties voor de laminaire stroming als een Directe Nu
merieke Simulatie (DNS) worden beschouwd.
Om de onsamendrukbare NavierStokes vergelijkingen op te lossen, is inten
sief gebruik gemaakt van het commerci¨ele pakket Fluent
en de openbron code
OpenFOAM
. Verschillende technieken voor roostervervorming zijn vergeleken.
Twee van deze methoden zijn gebaseerd op de Laplace vergelijking en een aangepaste
spanningsvergelijking. Beide methoden zijn erg eﬃci¨ent, omdat bestaande iter
atieve technieken kunnen worden gebruikt. Echter, de roosterkwaliteit is niet vol
doende voor gevallen met grote vleugelrotaties, wat het geval is bij het simuleren
van de vlucht van insecten. Zodoende is een nieuwe rooster deformatie techniek
ge¨ımplementeerd, gebaseerd op de interpolatie van radiale basis functies.
Deze techniek om rekenroosters te vervormen is gebaseerd op puntverplaatsin
gen, zodat de beweging van alle individuele interne roosterpunten wordt ge¨evalu
eerd. Er is geen connectiviteit van het rooster nodig, zodat deze methode een
voudig kan worden toegepast op ongestructureerde roosters. Om de eﬃci¨entie van
deze methode te verhogen, wordt een vergroving van de bewegende randpunten
toegepast. Dit verkleint het stelsel van vergelijkingen aanzienlijk, waardoor de
snelheid van de radiale basis functie methode, behoorlijk wordt verhoogd.
Na een discussie van de stromingsvergelijkingen, de eindige volume discretisatie
in OpenFOAM
en de vergelijking van verschillende technieken voor roosterver
vorming, beschrijven we de fysische en numerieke modellen. De onsamendrukbare
NavierStokes vergelijkingen zijn herschreven in een roterend referentie assens
telsel, zodat dimensieloze getallen die gerelateerd zijn aan de vleugel beweging
zijn afgeleid. Een belangrijk getal is het Rossby getal, dat een representatie is van
de kromming van het pad dat de vleugel aﬂegt.
Ten eerste is een tweedimensionale studie uitgevoerd om de invloed van ver
schillende vleugelbeweging modellen op de prestaties te onderzoeken. De con
dities voor stilhangende vlucht zijn hiervoor gebruikt. De resultaten laten zien
dat de ‘zaagtand’ ﬂap amplitude slechts een klein eﬀect heeft op de gemiddelde
liftkracht, maar dat de weerstand aanzienlijk wordt be¨ınvloed. De tweede model
vereenvoudiging, de ‘trapezium’ vorm van de invalshoek, leidt tot de loslating van
de voorrand wervel tijdens de translatie fase. Dit leidt tot een verhoging van
de gemiddelde weerstand tijdens elke halve ﬂap periode. De extra ‘bump’ van
de invalshoek, die aanwezig is bij de fruitvlieg vleugelbeweging, be¨ınvloedt de lift
niet beduidend. De laatste realistische vleugelbeweging karakteristiek, de deviatie,
heeft slechts een marginaal eﬀect op de gemiddelde lift en weerstandskrachten in
deze tweedimensionale studie. Desalniettemin verandert de eﬀectieve invalshoek
dusdanig dat de deviatie leidt tot een gelijkmatiger verdeling van de krachten.
Naast de tweedimensionale stroming voor stilhangende vlucht is een vergeli
jkbare studie uitgevoerd voor voorwaarts ﬂappende vlucht. Een numeriek model
215
voor een tweedimensionale stroming is gebruikt om het eﬀect van de vleugelbe
weging op de werveldynamica te onderzoeken voor een variatie van dimensieloze
golﬂengte, amplitude van de invalshoek en de hoek van het vlak waarin de vleugel
beweegt. Zowel translerende als roterende bewegingen zijn beschreven met simpele
harmonische functies, welke nuttig zijn om de parameter ruimte te onderzoeken,
ondanks de model vereenvoudigingen. Optimale voortstuwing met een ﬂappend
vleugelproﬁel bestaat voor elke variabele, zodat de aerodynamica waarschijnlijk
een range van wenselijke operationele condities selecteert. De condities voor op
timale voortstuwing liggen in een synchronisatie regio waarin een periodieke stro
ming bestaat.
Voorts zijn verschillende resultaten beschreven die relevant zijn voor de dried
imensionale stroming rond een ﬂappende vleugel. Ten eerste is de stroming rond
een dynamisch geschaalde vleugel numeriek berekend voor verschillende inval
shoeken om de ontwikkeling van de krachten en wervel dynamica te onderzoeken.
Daarnaast is het Rossby getal gevarieerd bij verschillende Reynolds getallen. Een
kleiner Rossby getal betekent een sterkere kromming van het pad dat de vleugel
aﬂegt, zodat de hoekversnelling ook hoger is. We laten zien dat een laag Rossby
getal gunstig is voor de stabiliteit van de voorrandwervel, zodat de liftkracht en
eﬃci¨entie worden vergroot. Ten derde is de driedimensionale vleugelbeweging
gevarieerd door de vorm van de invalshoek te veranderen. Ook is een deviatie
toegepast, wat kan leiden tot een achtvormige ﬁguur. De deviatie kan leiden tot
een geleidelijke verdeling van de krachten. Tenslotte is de driedimensionale stro
ming vergeleken met de tweedimensionale stroming voor een vleugel in voorwaarts
ﬂappende vlucht.
Het laatste onderwerp dat beschreven wordt in dit proefschrift, is het eﬀect van
vleugel vervorming. Een vooraf gedeﬁnieerde vleugelvervorming is toegepast op
een tweedimensionaal translerend vleugelproﬁel en een driedimensionale vleugel bij
condities voor stilhangende vlucht. Het vervormende vleugelproﬁel in vooruit ﬂap
pende vlucht leidt tot vergelijkbare resultaten als bij een starre vleugel, uitgebreid
met rotatie.
De huidige simulaties hebben geleid tot meer inzicht in hoe de prestaties van
een ﬂappende vleugel, representatief voor de vlucht van insecten en vogels, worden
be¨ınvloed door de vleugelbeweging en vervorming. Dit inzicht kan belangrijk zijn
voor het ontwerp en optimalisatie van micro vliegtuigjes.
Acknowledgements
This doctoral thesis presents the research that I have performed at the Aerody
namics Group of the Faculty of Aerospace Engineering at the Delft University of
Technology. After obtaining my MSc degree (2005) in numerical studies of ﬂap
ping foils, Professor Hester Bijl provided me the opportunity to investigate the
subject even further. I really enjoyed diving deep into the numerical techniques of
mesh motion, but also to apply these methods to real physical problems related
to ﬂapping wing aerodynamics. Thank you, Hester, for giving me the freedom
to shape my own research framework, providing me the opportunity to attend
international conferences and initiate several MSc projects.
Besides Hester Bijl, I would like to thank Bas van Oudheusden for his profes
sional insights and dedication. You have been a pleasant supervisor. I would also
thank all (former) PhD colleagues for the pleasant and inspiring working environ
ment. In particular, I would like to thank Alex Loeven en Peter Lucas for being
pleasant room mates and friends, having many scientiﬁc and general discussions.
Sander van Zuijlen provided lots of support concerning code development, thank
you for that.
I would like to thank the OpenFOAM community, for many stimulating dis
cussions on the workshops, conferences and online. In particular, I would like to
thank Professor Hrvoje Jasak and Henry Weller for many interesting and enlight
ening discussions. Hrvoje, thanks for your support and patience concerning my
programming skills. Additionally, I thank Dubravko Matijasevic for early imple
mentation of mesh motion based on radial basis function interpolation.
Finally, I say thanks to all my friends from Naaldwijk, Delft and beyond, for
the many joyful times to relax, drink beer or whisky and discuss many irrelevant
things. Special recognition goes to my parents, brother and familyinlaw for
their unconditional support, love and fun. Most importantly, I thank Marieke for
entering my life and having a lot of fun, together and with our beautiful son.
Frank Bos
Naaldwijk, January 2010
Curriculum Vitae
Frank Bos was born on March 17, 1980 in Naaldwijk, The Netherlands. He at
tended secondary school at the Interconfessionele Scholengemeenschap, Het West
land, in Naaldwijk from 1992 until he graduated the Atheneum in 1998. In 1998
he started his study at the Aerospace Engineering faculty of the Delft University
of Technology. He completed his Propedeuse year in 1999.
In order to obtain his Master of Science degree, he performed an internship
with a duration of 5 months in 2002 at the Queen’s University of Belfast, Aero
nautical department in the United Kingdom. He numerically investigated the
interaction between induced vortical ﬂow and a turbulent boundary layer. He
obtained his Master of Science degree at the Aerodynamics department in 2005,
entitled “Inﬂuence of wing kinematics on performance in insect ﬂight, a numerical
investigation”, supervised by dr.ir. David Lentink, dr.ir. Bas van Oudheusden and
Prof.dr.ir.drs. Hester Bijl. While studying, he performed several jobs at the Delft
University of Technology, related to project management and supervising students.
In August 2005 he started his Ph.D. project in the Computational Aerodynam
ics group supervised by Prof.dr.ir.drs. Hester Bijl. The results of this research on
the numerical simulations of ﬂapping foils and wings, are presented in this thesis.
He implemented mesh deformation, based on radial basis functions, in the open
source CFD code OpenFOAM
. In February 2010 he successfully defended this
thesis with accompanying propositions. He presented his work in several publica
tions and conference presentations. Additionally, he initiated MSc projects and
supervised several students.
Email: frank.m.bos@gmail.com
Linkedin: http://www.linkedin.com/in/fmbos
Copyright c 2009 by F.M. Bos All rights reserved. No part of this material protected by this copyright notice may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any other information storage and retrieval system, without written permission from the copyright owner. Printed by Ipskamp Drukkers B.V. in The Netherlands ISBN: 9789090251738 An electronic version of this thesis is available at http://repository.tudelft.nl
Numerical simulations of ﬂapping foil and wing aerodynamics
Mesh deformation using radial basis functions
Proefschrift
ter verkrijging van de graad van doctor aan de Technische Universiteit Delft, op gezag van de Rector Magniﬁcus Prof. ir. K.C.A.M. Luyben, voorzitter van het College voor Promoties, in het openbaar te verdedigen op woensdag 24 februari 2010 om 10:00 uur
door
Frank Martijn BOS
ingenieur luchtvaart en ruimtevaart geboren te Naaldwijk.
Dit proefschrift is goedgekeurd door de promotor: Prof. dr. ir. drs. H. Bijl
Copromotor: Dr. ir. B.W. van Oudheusden
Samenstelling promotiecommissie: Rector Magniﬁcus, Prof. dr. ir. drs. H. Bijl Dr. ir. B.W. van Oudheusden Prof. dr. ir. P.G. Bakker Prof. dr. ir. B. Koren Prof. dr. H. Jasak Prof. dr. FO. Lehmann Prof. dr. W. Shyy voorzitter Technische Universiteit Delft, promotor Technische Universiteit Delft, copromotor Technische Universiteit Delft Universiteit Leiden Centrum Wiskunde & Informatica Zagreb University University of Ulm The University of Michigan
This research was supported by The Netherlands Organisation for Scientiﬁc Research (NWO), NWOALW grant 814.02.019.
Voor mijn ouders .
.
hummingbirds. Because of this versatility. generates forces larger than obtained by using conventional aircraft aerodynamics. Several ﬂow visualisation experiments and numerical simulations have been performed to improve the understanding of ﬂapping wing aerodynamics in order to design and optimise Micro Air Vehicles. at the scale relevant to insect ﬂight. the eﬀects of wing kinematics on the ﬂow and forces is still not fully understood. For a long time. which is able to cope with large mesh deformations. e.Summary Both biological and engineering scientists have always been intrigued by the ﬂight of insects and birds. described by the unsteady incompressible NavierStokes equations. Experiments showed the presence of a vortex on top of the ﬂapping wings. Diﬀerent dimensionless numbers are discussed. The ﬂow around ﬂapping wings. i. especially at large rotations. but also to perform extreme manoeuvres. in terms of accuracy and eﬃciency. In order to capture the boundary layer and the near wake.e. it is important to maintain a high mesh quality near the moving wing. diﬀerent mesh motion techniques are compared and improved. We performed two.and threedimensional numerical simulations in order to systematically vary relevant parameters. small manmade ﬂyer’s to use in exploration and surveillance. The overall goal of this part of the research is to develop a reliable mesh deformation technique. the aerodynamic mechanism behind ﬂapping insect ﬂight was a complete mystery. Re = O(100). Strouhal and Reynolds numbers. an accurate mesh motion technique is necessary. Flapping wings produce both lifting and propulsive forces such that it becomes possible for insects and smaller bird species. there is no need for additional turbulence modelling. may be treated as a . Therefore. to stay aloft and hover. insects and smaller birds are an inspiration for the development of ﬂapping wing Micro Air Vehicles. Since the ﬂow at the considered Reynolds number. In order to incorporate a ﬂapping wing in our numerical model. characterising the ﬂow. is highly unsteady and vortical. to solve the ﬂow around ﬂapping wings. until several decades ago. such that our simulations. However. is laminar. related to the wing motion and ﬂow physics.g. assuming laminar ﬂow.
the ‘trapezoidal’ angle of attack. Therefore. on hovering ﬂight performance. which occur in insect ﬂight. Two existing methods are assessed. with increasing complexity. which is found to have only a marginal eﬀect on the mean lift and mean drag in this twodimensional study. Additionally. solving the Laplace and a modiﬁed stress equation. a coarsening is applied to the set of moving boundary points. ﬁnite volume discretisation in OpenFOAM and the assessment of the mesh motion solvers. the commercial software Fluent and the opensource code OpenFOAM have been used extensively. The second model simpliﬁcation. This decreases the size of the system of equations and associated computational eﬀort considerably. which means that the displacement of all individual internal mesh points are evaluated. No mesh connectivity information is necessary. so that it can be applied to unstructured polyhedral meshes. a new mesh motion solver is implemented.iv Summary Direct Numerical Simulation (DNS). which is based on the interpolation of radial basis functions. and stroke plane angles. However. a numerical model for twodimensional ﬂow was used to investigate the eﬀect of foil kinematics on the vortex dynamics around an ellipsoid foil subject to prescribed ﬂapping motion over a range of dimensionless wavelengths. The most important number is the Rossby number. Furthermore. The other realistic kinematic feature is the deviation. which represents the wing stroke path curvature. caused the leadingedge vortex to separate during the translational phase. and updated accordingly. This led to an increase in mean drag during each halfstroke. In order to solve the unsteady incompressible NavierStokes equations. the physical and numerical modelling are described. However. The results show that the ‘sawtooth’ amplitude has a small eﬀect on the mean lift but the mean drag is aﬀected signiﬁcantly. To increase its eﬃciency. This mesh motion solver is a point based method. First a twodimensional study is performed to investigate the eﬀects of diﬀerent wing kinematic models. angle of attack amplitudes. the eﬀective angle of attack is altered such that the deviation leads to levelling of the force distribution. Diﬀerent mesh motion solvers are compared. Both plunging and rotating motions are prescribed by simple harmonic functions which are useful for exploring the parametric space despite the model simplicity. dimensionless amplitudes. Both methods are very eﬃcient by using iterative solver techniques. The extra ‘bump’ in angle of attack as used by the fruit ﬂy model is not aﬀecting the mean lift to a large extent. The incompressible NavierStokes equations are rewritten in the rotating reference frame in order to identify dimensionless numbers related to the wing motion. such that only selected control points are used. based on a given boundary displacement. Optimal propulsion using ﬂapping foil exists for each variable which implies that aerodynamics might select a range of preferable operating condition. After the discussion of the governing equations. these mesh motion solvers are not able to maintain high mesh quality at large rotation angles. The conditions that give optimal propulsion lie in the synchronisation region in which the ﬂow is periodic. diﬀerent results relevant to threedimensional ﬂapping wing aero .
These results may be important to design and optimise Micro Air Vehicles. Concerning the ﬂexible airfoil in forward ﬂight. the ﬂow around a dynamically scaled model wing is solved for diﬀerent angles of attack in order to study the force development and vortex dynamics at small and large midstroke angle of attack. As in twodimensional studies. The present simulations have led to important insight to understand the inﬂuence of wing kinematics and deformation on the aerodynamic performance. subjected to additional rotation. A varying Rossby number represents a variation in stroke path curvature and thus angular acceleration. First. which may cause a ‘ﬁgureofeight’ pattern. the threedimensional wing kinematics is varied by changing the shape in angle of attack and by applying a deviation. the threedimensional ﬂow is compared with the twodimensional studies performed on ﬂapping forward ﬂight. Additionally. are described. by changing the eﬀective angle of attack. Therefore. leading to an increase in lift and eﬃciency. the deviation may inﬂuence the force distribution to a large extent. Finally.v dynamics. It is shown that a low Rossby number is beneﬁcial for the stability of the leadingedge vortex. a predeﬁned ﬂexing deformation is applied to a plunging airfoil in twodimensional forward ﬂight and to a threedimensional ﬂapping wing in hovering ﬂight. . the Rossby number is varied at diﬀerent Reynolds numbers. Thirdly. Secondly. a preliminary investigation is performed to show the eﬀect of wing ﬂexing. a similar behaviour was observed as for a rigid plunging airfoil.
.
. . 2. . 2. .5. .5.2. . . .1 Face interpolation schemes . . . . . . . . . . methods . . . . . . . . . . . . . .7.4 Objectives and approach . . . . . . . . 2. . . . . . . . . 2. . . . . 2. .5 Discretisation of an incompressible momentum equation . . . .4 Measures of cell quality .6 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Diﬀusion term . . . . 2. . . . . . . . . .2 Convection term . . . . . . . . . . . . .4 Temporal term . . . . . . . .7. . . . . . . .1 Introduction . . . . . . . .2 The NavierStokes equations . .5.Contents Summary 1 Introduction 1. . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . 2. . .2.1 Pressure equation and PressureVelocity coupling . . . . . . . 1. . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . .2 Physics of ﬂapping ﬂight . . . .2 Incompressible laminar ﬂow simpliﬁcations . . . . . 2. . . . . . .3 Experimental and numerical 1. . . 2. . . . . . . . . . . 2. . . . . . . . . . . . . .7. . . . . . . . . . . . . . . 1. . . . . . .5. . . . . . 2. . 2. .3 Dimensionless numbers . . . . . . . .1 Constitutive relations . .7 Solution of the NavierStokes equations . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Arbitrary Lagrangian Eulerian approach . . .3 Spatial and temporal discretisation . iii 1 1 2 7 9 13 13 16 17 18 18 19 20 21 23 23 24 25 26 27 28 29 30 2 Finite volume discretisation 2. . . . . .2 Procedure for solving the NavierStokes equations 2. . .
. . .10. . . . . . . . . . 3. 3.11 Conclusions . . . . . . .6 Conclusions . . . . . . . . .5 Improving computational eﬃciency . . . . 3. . . . . . . . .2. .4. . . . . . .3. . . 3. . . . . . . . . . . . . . . . . . .7 Conclusions . . . . . . . . .5. . . 4.1.1 Introduction .4 Dynamical scaling of ﬂapping wings . . . .2. . . . . . 2. 2. . . . . .1 Boundary point coarsening and smoothing . . . . .2 Kinematic modelling . . . . . 4. . . . 5. . . . .2 Inﬂuence of kinematic modelling . 3. . . . . . . . . . . . . . . . . . . . . . .1 Laplace equation with variable diﬀusivity . . 4. . . . . . . .4. . . . . . . . .1 Wing shape and planform selection . . . 3. . . 4. . 4. . . . . . . . 3. . . . . . . . . 2. 4. . . . . . . . . . . . . . . . 3. . .4. . 4. . 5. . . .9 Numerical ﬂow solvers . 3. . . . . . .2. . . .4 Comparison of mesh motion solvers . . . . . . . . . . . . . . . .4 Numerical implementation of the wing kinematics . . . . .2 Solid body rotation stress equation . . . . . . .2 Validation using cylinder ﬂows . . . . . .2 Iterative techniques and parallel implementations . 2. . . . . . .2 Numerical simulation methods . . .1 2D vortex decay and convection . . . . . . . . . .2 Diﬀerent mesh deformation techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Introduction . . . . . . . . . . . .1. . . . . . . . . Contents . . . . . . . .3 Flexing of a twodimensional block . .2. . . . . . . . . . . . . . . .3 Radial basis function interpolation .viii 2. . . . . . . . . . . . . . . . . . . . 5 Inﬂuence of wing kinematics in twodimensional hovering ﬂight 5. . . . . 2. . . . . . . . . . . . . . . . . . . . . .1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. .3 Modelling of active wing ﬂexing . . . . . 5. . . . . .1 Flow solver and governing equations . . . . . . . . . .10 Code validation and veriﬁcation . . . . . . . . . . . 4. . .10. . . . 32 33 34 35 39 46 49 50 52 53 54 55 58 59 59 62 64 66 67 69 70 73 73 75 78 79 80 82 83 86 88 89 92 93 94 94 94 96 96 97 3 Mesh deformation techniques for ﬂapping ﬂight 3. . . . . . . . . . .3. . . . . . . .3 Mesh quality measures . . . . .5. 4.2 Flapping of a threedimensional wing . . . . . . . . . . . . . .and threedimensional ﬂows . . . . . . . . . . 3. . . . . 5. . . . . . . . . . . . . . . . . . .5 Computational domain and boundary conditions . . . . . . . . . .1 Similarity and discrepancy between two. . . . . . . . . .6 Deﬁnition of force and performance coeﬃcients . . . . . . .2 Governing equations for ﬂapping wings . . . . . . . .3 Wing shape and kinematic modelling . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . .2 Mesh generation and boundary conditions . . .1 Translation and rotation of a twodimensional block 3.8 Swept volume calculation . . . . . . . . . . . 4 Physical and numerical modelling of ﬂapping foils and wings 4. . . . . . . . . 3. . . .3. . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . .3. . . . . . . . 4.
. . . 6. . .5 Flapping wings at low Reynolds numbers . . . 8. . . . . . . . . . . . . . . . .3 Force coeﬃcients and performance . . . . . . .4 Vortex identiﬁcation methods for ﬂow visualisation . . . .Contents 5. . . . . . . 5. . . . . . . . .2. . .3 Flow solver accuracy . . . . . . . . .3.3 Conclusions . . . . . . . . . . . . . 6. . . . . . . . . . . . . . . 5. .3. . . 6. . . . . . . . . . . . .5. . . . . . .6 Flapping wings in forward ﬂight .5 Discussion . . . .2 Wing ﬂexing in threedimensional hovering ﬂight . .3. . . . . . . . . 7.5. . . . 8 Inﬂuence of wing deformation by ﬂexing 8. 7. . . . . . . . . . . . . . . .2 Dynamical scaling of the wing model . . . . . 7. . . . . . . . . . . . . . 6. . . . . . . . . . . . . . . . 6. 5.5 Results and discussion . . . . . . . . . . . .2 Flapping foil parametrisation . . 7. . . . . . . . . .1 Airfoil ﬂexing in twodimensional forward ﬂapping ﬂight . . . . . . .2. . . . . .2 Inﬂuence of dimensionless amplitude .2 Simulation strategy and test matrix selection .5. . . . . . . . . . 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . . . . . .1 Overall model comparison . . . . . 7. . .1 The angle of attack in ﬂapping ﬂight . . . . . .3 Inﬂuence of angle of attack amplitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Numerical model . . . . . . . . . . . . . 6.4 Inﬂuence of ‘trapezoidal’ angle of attack . . . . 6. . . . . . .4. . . . . . . 7. . .4 5. . . . . . . . . . .5. . . . . . . . . . . . . . . .5. . . . . .5. . . . 8. . . . 6. . .2. . . . .3 Inﬂuence of Reynolds number . . . . . . Conclusions . . . . . . . .5 Inﬂuence of deviation . . . . . . . . . . . . . . . 7.1 Introduction . . . . . . . . . . . .2 Inﬂuence of ﬂapping stroke curvature . . .5. . . . . . . .1 Inﬂuence of dimensionless wavelength . . . . . 5.1 Modelling and parameter selection . . . . . . . . . . . . . . . . . . . . . . . . ix 99 99 100 101 102 103 105 105 109 118 121 122 122 124 125 126 126 127 129 130 130 132 135 135 137 138 141 143 145 148 149 149 152 158 160 163 166 171 171 175 177 5. . .2 Threedimensional ﬂapping wing simulations . . . . . . . . . . . . . . .3 Validation using harmonic wing kinematics Modelling insect wing kinematics . . . . . . . . . . . . . . . . 7. 5. . . . .7 Conclusions . . . . . . .4 Inﬂuence of stroke plane angle . . . . . . . . . . . . . . . . 7. . .6 Conclusions . . .3. . . . . . . . . . . . . . . ﬂapping foil . . . . . . . . . 6. . .1 Insect wing selection and model parameters 5. 7. . . 6. . . . . . . . . . . . . . . . . .4 Diﬀerent wing kinematic models . .5 6 Vortex wake interactions of a twodimensional 6. . . . . .3 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Introduction . . . . . . . . . . . . . . . . . Results and Discussion . . . . . 7. . . . . .3 Force and performance indicators . . . . .4.5. .2 Kinematic features investigation . . . . . . . . . . . . . . . . . . . . . . . . . 7. . 7 Vortical structures in threedimensional ﬂapping ﬂight 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . .
. . . . . . . . . . . A. . . . .5 Preliminary conclusions on wing ﬂexing . . . . . . .3 Conclusions on hovering ﬂapping ﬂight . . . . . . . Bibliography Samenvatting Acknowledgements Curriculum Vitae . . . . . . . . . . .2 BlockMesh . . . . . . . . . . . . . . . . . . . . . . . . . . A Grid generation for ﬂapping wings A. . . . . . . . . . . . . . . . . . . A. . . . . . . . . . . . . . . . . . . . .2 Fluent solver settings . . . . 9. . . . . . . . . . . . . . 9.x 9 Conclusions and recommendations 9. . . . . . . . . . 9. . . . . . . . . .4 GridPro .4 Conclusions on forward ﬂapping ﬂight . . . . . Contents 179 179 180 181 181 182 184 184 184 185 186 189 189 190 191 191 192 195 195 195 197 201 213 217 219 . . . . . . . . .4. . .3 OpenFOAM solver settings . . . . . . .1 Introduction . . . . . . . . . . . . . B Flow solver settings B. . . . . . . . . . . . . . . . 9. . . . . . . . . 9. . . . . . . .2 Threedimensional hovering . . . . . . . . . . .6 Recommendations .4. . . . . . . . . . . . . . . B. . . . . . . . . . . . . .1 Introduction . . . . . . . . .1 Twodimensional forward ﬂapping . . . . . . . . . . . . . . .1 Overall conclusions . . . . . . . . A. . . . . . . . . . . . . . . 9. . . . . .3 Gambit .3. . . . .3. . . . . . . . . . . . . . . . . B. . . . . . . . . . . . . . . . . . . . . .2 Conclusions on mesh motion techniques .2 Threedimensional forward ﬂapping 9. . . 9. . . . . . . . . . . . . .1 Twodimensional hovering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. . . . . . . . . . . . . 9. .5 Conclusions . . . . . .
1956) ﬂaps with about 200 times per second. insects adapted to leave the ground to take oﬀ into the thin air. since he was born 200 years ago. but also to perform extreme manoeuvres. e. the mainland bird species had travelled to the islands and on every diﬀerent island it adapted to the diﬀerences in environmental circumstances. whereas a small insect.g. see . hummingbirds. 2000) which developed the skill of ﬂight in order to migrate over large distances and to catch prey. such that it becomes possible for insects and even smaller bird species. Because of this versatility. inspired by his scientiﬁc observations during a voyage (18311836) around the world with his ship. This process has become known as natural selection. e. the Beagle. leading to a diﬀerent ﬂow behaviour. Flapping wings produce both lifting and propulsive forces.1 Motivation The year of writing.CHAPTER 1 Introduction 1. he published his world famous Origin of Species. At the Gal´pagos Archipelago. That book describes the natural selection. Birds and insects are both ﬂapping their wings at diﬀerent length scales. Birds are ancient descendants of feathered dinosaurs (Templin. to stay aloft and hover. tiny manmade ﬂyer’s to use in exploration and surveillance. Darwin discovered slightly diﬀerent bird a species living on the diﬀerent islands. The larger the animal. insects and smaller birds are a major inspiration of study to develop Micro Air Vehicles (MAV).g. millions of years ago. which is also applicable to the early era of ﬂight. At the age of 50. whereas. Long before the origin of dinosaurs and birds. the lower the need for ﬂapping wings. he only knew one species on the mainland of SouthAmerica. a fruit ﬂy (WeishFogh & Jensen. is known as the year of Charles Darwin (18091882). the Andean Condor only ﬂaps when it looses height in the thermal winds. Apparently. 2009.
. birds. 2009). see ﬁgure 1. when performing wind tunnel experiments it is not straightforward to extract the force data.2. the leadingedge vortex in particular. the construction of models needs to be very precise. diﬀerent aspects of ﬂapping wing aerodynamics are discussed. detailed experiments (Ellington et al.and threedimensional problems. 2002). In order to study this kind of ﬂows. either directly or indirectly. Additionally.. insects. is brieﬂy discussed.. which can be very costly as well.g. 2002. Large aeroplanes generate wingtip vortices. using Computational Fluid Dynamics (CFD). which can cause damage to a following aeroplane which encounters .1.3 deals with the diﬀerent approaches for analysing the ﬂow. Srygley & Thomas. 2008. Finally. To optimise the ﬂight performance of MAV’s it is important to get a thorough understanding of the complex ﬂow generated by its wings. Secondly.... Poelma et al. the ﬂow ﬁeld can not be visualised in much detail. as well as the inﬂuence of the wing kinematic modelling in two. from the ﬂow visualisation obtained by Particle Image Velocimetry (PIV) (Poelma et al. O(100 − 1000). Lentink & Dickinson.2 Introduction ﬁgure 1. experiments and numerical methods. Even when the most advanced techniques are used. performing CFD provides a suitable framework. On the other hand. The vortex dynamics. Sun & Tang. in view of the need of precision equipment and wind tunnel facilities.2 brieﬂy provides background information on the ﬂow physics concerned. Section 1. 2002.2 Physics of ﬂapping ﬂight In order to illustrate the necessity and diﬃculties with solving and visualising the ﬂow around ﬂapping foils and wings. Digital Particle Image Velocimetry (DPIV). The present study deals with the development and improvement of computational techniques to solve the ﬂow around ﬂapping wings at low Reynolds numbers. 1. Vortex generation in ﬂapping wing aerodynamics Vortex generation in nature is fairly common in ﬂows induced by aeroplanes. the forces and ﬂow visualisations are a direct result of the computations.4 describes the objectives and approach of the present study as well as the outline of this thesis. The ﬂapping wings induce complicated vortical structures which inﬂuence the forces and performance characteristics in hovering and forward ﬂight. One limitation of doing experiments is that they can be very expensive. 2009b) and numerical simulations (Wang et al. Sane & Dickinson. 1956. 2004. section 1. especially at smaller length scales (< 5 cm). but also by boats and trees. 1996. when performing numerical simulations. Since it is interesting to solve for the forces acting on a ﬂapping wing in combination with the vortical structures within the near wake. 2006. especially due to the reﬂections and shadows of the moving wings. researchers performed ﬂow visualisations (WeishFogh & Jensen. Bos et al. e. while section 1. 2006). Thaweewat et al.
generated by successive up and downstrokes. but this was not suﬃcient to explain the high lift generation of insects.2 Vortex induced force generation. (a) Wingtip vortex causes big disturbances in the wake. A water strider generates vortices with its long legs to create the necessary propulsion (Hu et al. (b) The U. patented Entomopter has four ﬂapping wings powered by chemicallyfuelled propulsion system (Michelson. 2008). ﬁsh and insects.S. (a) (b) (c) Figure 1. circulation and lift generation. It is a common story that ﬂies could not ﬂy according to conventional aircraft theory as developed by Lanchester (1907) and Prandtl (19141918). This mystery persisted until the discovery of the unsteady vortical ﬂow ﬁeld. 2003). (b) Entomopter. ﬁgure 1. alternating vortex rings were seen in the wake. Prandtl did develop a relation between the tip vortices. and especially the generation of the leadingedge vortex.g. which is able to perform a tethered takeoﬀ (Wood. e. bridges or struts in water. which is used by birds.2 Physics of ﬂapping ﬂight 3 (a) Wasp. (c) Delﬂy. limiting the time between two successive aeroplane approaches. (1997) performed smoke visualisation of the vortical ﬂow patterns induced by a hawkmoth.2. ﬁgure 1. (b) Vortex generation in insect ﬂight. At lower Reynolds numbers.. 2008). Another undesired eﬀect of vortex generation is ﬂow induced vibration of cables. which is especially interesting for intelligence and exploration. It was observed that the leadingedge vortex was stabilised by the radial ﬂow moving out towards the wing tip. vortex generation provides possibilities to generate forces. (a) Flying insect scale robotic model.1.2 shows induced propulsive vortices generated by a water strider. Figure 1. (c) The Delﬂy Micro is camera equipped and is able to hover (designed and developed at Delft University of Technology). ﬂapping MAV concepts can be used for hovering and low speed forward ﬂight. this vortex. On the other hand. (c) Willmott et al. Additionally.1 Diﬀerent ﬂapping wing Micro Air Vehicle concepts. .
4 Introduction ¯ CL = 1. Figure 1. 2009b). Lentink et al. operating at Re = 110.and threedimensional ﬂapping inﬂuences the shear layer direction and ﬂow accelerations. It was Ellington et al. depending on the kinematics. (1996) o who identiﬁed the presence of a leadingedge vortex (LEV) generated on top of the ﬂapping wing. notably advance ratio and dimensionless ﬂapping amplitude (Thaweewat et al. In ﬂapping foil aerodynamics the vortices are shed and form either a periodic or chaotic wake pattern. Leadingedge vortex The potential beneﬁt of vortices attached to the wing was already discussed by Maxworthy (1979) and Dickinson & G¨tz (1993). 2008).540 3 2 1 (a) x c [−] 0 1 2 3 (b) Figure 1. In order to understand the physics of ﬂapping wing aerodynamics.. (b) Threedimensional leadingedge vortex generated by a ﬂapping wing at Re = O(1000). (a) Twodimensional illustration of the wing kinematics and the resulting force vector generated by the ﬂapping airfoil at Re = O(100) (Bos et al. Wang . from (Bos et al..3(a) shows a twodimensional illustration of the wing kinematics of a fruit ﬂy..3 Forces and vortices in ﬂapping wing aerodynamics. it is important to obtain insight in how an insect moves its wing. 2008). It is thought that the kinematics in two.. Inﬂuence of insect wing kinematics on forces The relevance of experiments and ﬂow simulations of insect ﬂight has been found to depend on how reliably true insect wing kinematics are reproduced. 2009. increasing the lift force to values much higher than predicted by conventional wing theory. The origin of the leadingedge vortex is the rollup of shear layers. 2008). which will undoubtedly inﬂuence the development of the leadingedge vortex (Lentink & Dickinson. present in highly viscous ﬂows. It appears that the leadingedge vortex is more stable around a threedimensional ﬂapping wing compared to twodimensional ﬂapping foil situations. The stability of the helical threedimensional leadingedge vortex is still not yet fully understood and appears to heavily depend on the wing kinematics and Reynolds number. which is the case at low Reynolds numbers.
b) and Lentink & Gerritsma (2003) numerically investigated pure harmonic translational motion with respectively small and large amplitudes. rotation speed and angle of attack during translation are of great importance for the force development during each stroke. the wing maintains a constant velocity and angle of attack during most of the stroke. 1993).2 Physics of ﬂapping ﬂight 5 et al. may have on ﬂight performance. Slightly more complex fruit ﬂy kinematic models were used by Dickinson et al. (2004) showed that the angle of attack inﬂuences the ﬂapping foil propulsion eﬃciency to a large extent. such that including rotation is essential.1. Besides lift enhancement at certain reduced frequencies. deviation. (1999) and Sane & Dickinson (2001) with their Roboﬂy. angle of attack and the timing of the latter were explored. the results were compared with more realistic fruit ﬂy kinematics obtained from the observation of free ﬂying fruit ﬂies. (2004) and Sane & Dickinson (2001) showed that the kinematic modelling signiﬁcantly inﬂuences the mean force coeﬃcients and its distribution. including wing rotation. Furthermore. diﬀerent kinematic models have been employed to investigate the aerodynamic features of insect ﬂight. Both studies emphasised the importance of angle of attack to inﬂuence the propulsive eﬃciency. In literature. This results in the typical ‘sawtooth’ displacement and ‘trapezoidal’ angle of attack pattern of the Roboﬂy kinematic model. both the pure harmonic and the Roboﬂy model. Instead of performing a parameter study within the scope of one kinematic model. Using these models. They varied rotational parameters and showed that axisofrotation. Additionally. Wang (2000a. Hover et al. In addition to the harmonic models with pure translation (Dickinson & G¨tz. Harmonic wing kinematics. rotational parameters were investigated by Dickinson o (1994). they found periodic and aperiodic ﬂow solutions which are strongly related to the aerodynamic eﬃciency.. the eﬀect of amplitude. Based on observation of true insect ﬂight. where used by Pedro et al. Lentink & Gerritsma (2003) varied airfoil shape with amplitude and frequency ﬁxed at values representative to real fruit ﬂies. like parameter values and stroke patterns. The present thesis deals with diﬀerent kinematic models from literature. It further emphasises the need to critically assess the inﬂuence of kinematic model simpliﬁcations. with a relatively strong linear and angular acceleration during stroke reversal. Lewin & HajHariri (2003) performed a similar numerical study for heaving airfoils. This illustrates the appreciable eﬀects which details of the wing kinematics. For example. (2003) and Guglielmini & Blondeaux (2004) in their numerical models to solve for forward ﬂight. It was also shown that wing stroke models with only translational motion could not provide for realistic results. the objective of the present study is to compare the eﬀect . but large amplitudes lead to an increase of lift by a factor of 5 compared to static forces generated by translating airfoils. They concluded that the airfoil geometry choice is of minor inﬂuence. in order to investigate their inﬂuence on the aerodynamic performance (Bos et al. 2008). Wang (2000a.b) varied ﬂapping amplitude and frequency and showed that at a certain parameter selection the lift is clearly enhanced.
in case of advanced and symmetric rotation the forces were found to be similar in the twodimensional simulations compared to the threedimensional experiments.and threedimensional approaches are suﬃcient to warrant that a reasonable approximation of insect ﬂight can be obtained using a twodimensional approach. Bos et al. One of the major (and partially unresolved) issues in modelling of insect ﬂight and ﬂapping wing propulsion. which is present in three dimensions. In a recent paper Wang et al. as was mentioned by Lighthill (1969). wing ﬂexibility (ﬂuid structure interaction) and Reynolds number eﬀects. in relation to aerodynamic performance. of the available models as a whole. Additional important aspects are unsteady ﬂow mechanisms.6 Introduction (a) (b) (c) Figure 1. 2009. (2005). (2004) conﬁrmed that the similarities between two. First. depending on the kinematics a chaotic wake pattern may occur with unpredictable forces as the result.and threedimensional ﬂows To limit both the parametric space as well as the computational eﬀort. The similarity between two. (2005b) concluded that twodimensional studies over predict forces and performances since the energyloss. is the possibly restrictive applicability of twodimensional results to true insect ﬂight.. 2008. Also. (2005b) numerically investigated the wake structure behind ﬁnitespan wings at low Reynolds numbers. Blondeaux et al. They observed threedimensional vortical structures around ﬂapping wings with low aspect ratio. (2004) compared threedimensional Roboﬂy results with twodimensional numerical results. Dicko . Wang et al. is not resolved.and threedimensional ﬂow. 2004.. (a) Von K´rm´n vortex a a street behind a stationary cylinder at Re = 150. which is of great importance to both experiments and numerical simulations. one single and one vortex pair is generated each plunging period. This leads to better insights into the consequences of simpliﬁcations in kinematic modelling. many studies have been performed as twodimensional simulations (Thaweewat et al.4 Vortex wakes generated by cylinders and ﬂapping wings. Secondly it was observed that in both simulations and experiments the leadingedge vortex did not completely separate for amplitudetochord ratios between 35 (Dickinson & G¨tz. (b) Periodic vortex wake behind a plunging airfoil at Re = 110. (2005) and Blondeaux et al. twodimensional analysis has often been applied to obtain insight into the aerodynamic eﬀects of wing kinematics and geometry. 2003). Both Dong et al. 1993. Lewin & HajHariri.. (c) Chaotic vortex wake behind a plunging airfoil at Re = 110. it may reveal the importance of certain speciﬁc features of the stroke pattern. Wang et al. Notwithstanding the possible discrepancy between two. Dong et al.
like delayed stall and wake capture enhance the lift force beyond values predicted by quasisteady theory. and revealed the complex nature of insect ﬂight aerodynamics. Drosophila Melanogaster.2. amplitudes and the very low Reynolds number involved (Re < 1000 for a large number of insects and Re ≈ 110 for the fruit ﬂy. (1996) discovered that these lift increasing mechanisms are ampliﬁed by the generation of a leadingedge vortex (LEV). (1999) and Wang (2000b) applied the quasisteady theory to compare with unsteady forces.b). the ﬁrst part of this thesis makes extensive use of twodimensional simulations. 1993. In the second part. hence revealing that important unsteady and vortical ﬂow phenomena play a major role in insect ﬂight. 1994. 1. The current research deals with amplitudes that are in this range. In this section. diﬀerent methods were used to solve and visualise the ﬂow around ﬂapping insect wings. 1999. and with the above justiﬁcation.3 Experimental and numerical methods In literature. from experimental methods to computational ﬂuid dynamics simulations. as visualised by WeishFogh & Jensen (1956) using tethered locusts and by Willmott et al. The lift increasing eﬀect of the leadingedge vortex strongly depends on the kinematics of the ﬂapping wing (Dickinson et al. Lentink & Dickinson. from realistic fruit ﬂy measurements to threedimensional simulations using a representative model wing.3 Experimental and numerical methods 7 inson. 2001. 2000b. 2009a. In order to understand insect ﬂight performance Dickinson et al. It was shown that this leadingedge vortex arises during the translational part of the wing motion rather than during the rotational ﬂip between up and down stroke. in particular). Dickinson et al. 1994). various threedimensional simulations were performed using limited variations wing kinematics. In several studies (Dickinson & G¨tz.. Experimental investigations and quasi steady theory Several experimental studies considered the ﬂight performance of insects. Ellington (1984) indicated that the lift in insect ﬂight is signiﬁcantly higher than expected on the basis of quasisteady aerodynamics. (2004) performed free ﬂight and tethered experimental visualisations using butterﬂies and dragonﬂies to show the complicated vortical structures. (1997) using a hawkmoth (Manduca Sexta). Srygley & Thomas (2002) and Thomas et al. 2002. diﬀerent methods will be brieﬂy addressed. Sane & Dickinson.1. Dickinson. In view of the excessive computational expense required to perform accurate threedimensional simulations. o 1999) it was conﬁrmed that important aspects. The ﬂow induced by the motion of insect wings is highly unsteady and vortical. The quasisteady approach was revised by Sane & Dickinson (2002) to include ro .. see ﬁgure 1. Ellington et al. More recently. This unsteady and vortical ﬂow behaviour is a consequence of the high relative frequencies. Wang.
. 2007. or even complete remeshing (Young & Lai. Several experimental studies have been performed with the aim of characterising the unsteady aerodynamics of insect ﬂight. 2005). such as ﬂapping wings. In order to cope with high rotation rates. dragonﬂy: Young & Lai (2008). Mittal & Iaccarino. Poelma et al. as in the manoeuvre clapandﬂing (WeishFogh & Jensen. it remains diﬃcult to capture all the relevant details of the ﬂow using only experimental techniques. An appealing approach. therefore. one can use either immersed boundary methods (Peskin. when two wings touch. 2006). one will undoubtedly need methods like overset. Numerical simulations Notwithstanding important advances in experimental techniques for nonintrusive ﬂow ﬁeld analysis. This conﬁrms the restricted applicability of the quasisteady theory due to lack of unsteady mechanisms like rotational lift and wake capture. Particle Image Velocimetry in particular (Bomphrey et al. the computational eﬀort involved in threedimensional studies is presently still extremely demanding. 2006. To perform numerical simulations around moving objects. In an immersed boundary method. Jasak.. in relation to speciﬁc insect geometries (moth: Liu & Kawachi (1998). the producer of the . 2009). which is an opensource framework to solve the NavierStokes equations on threedimensional 1 OpenFOAM is a registered trade mark of OpenCFD OpenFOAM software. 2002. Nevertheless.8 Introduction tational eﬀects but even then the results require further improvement. (2004)). immersed boundary or remeshing techniques. Limited. an integrated computational study was performed by Aono et al. even not for ﬁxed boundaries (Mittal & Iaccarino. 2008. Dickinson et al. Sun & Tang (2002). deforming mesh techniques (Boer de et al. 2005). 1956). The commonly used mesh motion techniques result in high quality meshes as long as the rotation of the moving boundaries is limited. Isogai et al.. A number of numerical studies on full threedimensional conﬁgurations have been reported. (2008) who developed a code to incorporate two wings and a body using overset mesh techniques. is to supplement experiments with numerical ﬂow simulations. mesh motion based on radial basis function (RBF) interpolation is implemented in this thesis and improved in terms of accuracy and eﬃciency.5. but the mean drag is underestimated. the moving boundary is projected on a ﬁxed Cartesian background grid. it was chosen to assess and improve existing mesh motion techniques. Zuo et al. which is not allowed to deform.. fruit ﬂy: Ramamurti & Sandberg (2002). Although. see ﬁgure 1. This modern mesh motion technique is incorporated in OpenFOAM 1 . the conservation of mass and momentum in current immersed boundary methods is not obvious. Together with the unavailability of an accurate ﬂow solver with parallel support. 2007). (1999) investigated the ﬂow around a ﬂapping Roboﬂy model which moves in oil to meet the same ﬂow conditions as the real fruit ﬂy encounters (reproduction of Reynolds number in particular). According to Sane & Dickinson (2001) the mean lift is well predicted by quasisteady theory. Besides interpolation issues.
whereas the Eulerian part takes care of the ﬂuid ﬂow through the mesh. while (b) shows the mesh motion obtained by using radial basis function interpolation. where the mesh is ﬁxed to the ﬂuid or material. 1.4 Objectives and approach Flapping ﬂight aerodynamics is governed by many parameters. Reynolds number. Two illustrations of mesh motion solutions.4 Objectives and approach 9 (a) (b) Figure 1. a ﬁxed mesh becomes inconvenient. The Lagrangian contribution allows the mesh to move and deform according to the boundary motion. when the ﬂow domain moves or deforms in time due to a moving boundary.1. However. Arbitrary LagrangianEulerian formulation The governing equations to solve the ﬂow are generally discretised using the Eulerian description. If the material or ﬂuid deforms. (a) shows a Laplacian mesh motion. This method is commonly used to discretise the governing equations encountered in structure mechanics. 1982). the mesh deforms with it. like advance ratio. This code is thoroughly tested and used for ﬂapping foil and wing simulations. etc. In order to perform accurate numerical simulations it is important to use an eﬃcient code which is capable to solve for . At the time of writing. unstructured grids of polyhedral cells with full parallel support. where the ﬂuid is allowed to ﬂow through the ﬁxed mesh.5 Diﬀerent mesh motion solvers. because it requires the explicit tracking of the domain boundary. This is in contrast to the Lagrangian formulation. Therefore. This method incorporates and combines both Lagrangian and Eulerian frameworks. the ALE method has become the standard implementation in most popular codes to solve for the ﬂow around moving boundaries while the mesh deforms accordingly. wing kinematics. the Arbitrary Lagrangian Eulerian (ALE) formulation is used to discretise the ﬂow equations on moving and deforming meshes (Donea.
These mesh deformation techniques are described and assessed with respect to accuracy in chapter 3. validate and verify the numerical solver with the implemented and improved mesh motion technique. Before proceeding to the numerical results of the ﬂow around ﬂapping foils and wings. To meet the objectives the current thesis is structured in the following chapters. which is the subject of chapter 2. an improved mesh motion technique. Chapters 5 and 6 deal with the numerical investigations of twodimensional ﬂow around a ﬂapping foil in hovering and forward ﬂight conditions. improve current mesh motion techniques and implementation. For large wing translations and rotations the numerical grid needs to deform accordingly to maintain high accuracy of the ﬂow solver. 3. This method is used to study the complex vortical patterns to identify optimal strategies in ﬂapping foil and wing aerodynamics. It is . using realistic wing kinematics. in terms of accuracy and eﬃciency. the overall goal of this research is to develop a reliable mesh deformation technique. the solution procedure is described together with a brief discussion about the opensource framework OpenFOAM . This mesh deformation technique is implemented and used for ﬂapping foil and wing aerodynamics. study the threedimensional structure of vortical patterns. In order to satisfy this aim. Approach and outline In order to solve for the ﬂow around ﬂapping foils and wings.10 Introduction various conditions. using an accurate and eﬃcient framework. the following objectives are deﬁned: 1. In order to solve the governing equations for ﬂuid ﬂow. solve for the ﬂow around twodimensional ﬂapping foils to study the wingwake interaction as well as the inﬂuence of wing kinematics. respectively. 2. to solve the ﬂow around ﬂapping wings. That chapter deals with the discretisation of the diﬀerent terms as well as a deﬁnition of mesh quality. which is thoroughly validated and veriﬁed. based on radial basis function interpolation. Within the code. 4. The mesh motion technique is used by an incompressible unsteady CFD solver to solve for the ﬂow around a threedimensional ﬂapping wing on dense meshes in parallel. a ﬁnite volume discretisation is used. diﬀerent mesh deformation techniques are incorporated. 5. solve for the ﬂow around threedimensional ﬂapping wings to assess the importance of parameters like ﬂapping amplitude. it is important to discuss the numerical modelling for ﬂapping ﬂight in chapter 4. Therefore. In addition to the already implemented mesh deformation techniques. a method based on radial basis function interpolation is discussed. is implemented in the opensource framework OpenFOAM . especially the leadingedge vortex. Furthermore. frequency or Reynolds number.
1. performance and wake patterns. Additionally.and threedimensional ﬂapping ﬂight. It will be seen that accurately solving the ﬂow around a ﬂapping wing is not an easy task when the wing performs complex rotational motion. which is discussed in chapter 7. chapter 8 presents the preliminary results of a ﬂexing wing in two. . Complex vortical structures induced by a model ﬂapping wing can be accurately solved and analysed. These twodimensional results provide good insight what to expect of the threedimensional ﬂow around a ﬂapping wing. The conclusions and recommendations can be found in chapter 9.4 Objectives and approach 11 found that the kinematic modelling has a large inﬂuence on forces.
.
four important aspects of numerical . since the ﬂow is dominated by the convection of vortices. Therefore. the boundary conditions and the solution procedure are presented. Fluent was already tested by Bos et al. Important aspects of numerical simulations Before describing the used methods in detail.CHAPTER 2 Finite volume discretisation A secondorder ﬁnite volume discretisation of the incompressible NavierStokes equations on arbitrary polyhedral meshes is described. (2008). it was found that the Van Leer ﬂux limiter provides the most accurate results. 2. This ﬁnite volume approach is applicable to general commercial and noncommercial CFD codes.1 Introduction Important aspects concerning numerical simulations. Furthermore. the ﬂow around stationary and transversely oscillating cylinders showed that the code of OpenFOAM solves the ﬂow in detail. A numerical simulation needs to be performed in an accurate and concise way. For test problems involving vortex decay and convection. Spatial and temporal convergence was proved as well. In addition to the mesh quality measures nonorthogonality and skewness. like stability and convergence. are being described. The commercial code Fluent and the opensource code OpenFOAM have been used for the simulations described in this thesis. diﬀerent properties of a CFD simulation. such that this chapter deals with the validation of OpenFOAM using problems relevant for low Reynolds number insect ﬂight. are addressed.
1 shows an example of a structured and an unstructured grid. the numerical grid. Wesseling (2001) and Ferziger & Peric (2002) described these methods in more detail. the interpolation from cell centres to cell faces and how to approximate the surface and volume integrals. the time is discretised as well. the cell ordering is fairly straightforward such that the ﬂow solver uses this fact to solve the system in a more eﬃcient way. hexahedral (four corners) to polyhedral (arbitrary number of corners) cells. is the generation of a numerical grid. When using (block) structured grids. When using the ﬁnite volume method. Figure 2. the discretisation method. which is necessary to perform unsteady simulations. the governing equations. 2002). Besides the type of grid.14 Finite volume discretisation (a) Structured (b) Unstructured Figure 2. and the solution method to solve the system. concerning a CFD simulation. structured. Meshes can be generated in an structured way (a) and using unstructured methods (b). the cell shape can be varied from tetrahedral (three corners in two dimensions). Traditionally. simulations are discussed. However. needs to be described. as is the case in the current thesis. In the ﬁeld of CFD. There are three types of grids. Besides the spatial discretisation. blockstructured and unstructured grids (Ferziger & Peric. This is the more important asset of unstructured grids. The third aspect. In order to solve the incompressible NavierStokes equations. a structured grid is favourable in terms of accuracy and eﬃciency of the ﬂow solver. . for less complex geometries. ﬁnite element and ﬁnite volume method.1 Diﬀerent mesh generation methods. a division of the computational domain in a ﬁnite amount of cells. a suitable method to discretise these equations needs to be chosen. the ﬁnite diﬀerence. ﬁnite element methods are used for structural problems whereas numerical simulations related to ﬂuid ﬂow are mostly solved with ﬁnite volume methods. A drawback of a (block) structured grid is that it is more diﬃcult to create around complex geometries (commonly encountered in engineering problems). three methods are commonly used.
When an appropriate iterative solver is used.1 Introduction 15 Finally. When performing the iterations of the numerical process it should be the case that the numerical errors are not ampliﬁed. One of the advantages of the ﬁnite volume approach is that conservation is guaranteed for every small control volume and therefore. it should be suﬃciently small. discretisation method and the choice of grid. especially when using a ﬁnite volume approach. the last aspect is boundedness. the system of discretised equations can be easy or diﬃcult to solve. stability is the third important aspect. When the solution converges to a gridindependent solution. However. The second important property of a numerical scheme is convergence.e. it may happen that the exact solution is not approximated with decreasing timestep. from (Ferziger & Peric. In that case the solution process is called stable. conservation and boundedness. it is important to discuss diﬀerent properties of the numerical solution method. the choice of numerical method is usually a tradeoﬀ. i. Since the governing equations in ﬁnite volume formulation are conservative. In order to check the consistency of the complete numerical scheme. when the cell size approaches zero. Considering a steady problem. where the same computation is repeated on subsequently reﬁned meshes. convergence. consistency. this property should be respected by the discretised equations. Depending on the governing equations. The solution of the discretised system of equations should tend to the exact solution of the governing diﬀerential equations as the mesh spacing tends to zero. 2002). the mass ﬂux of a conserved quantity through a speciﬁed system should be zero. Finally. The fourth property of a numerical scheme is conservation. The diﬀerence between the discretised and the exact solution is called the truncation error. like concentration or . so commonly the empirical approach is followed. The following properties are relevant concerning numerical simulations. The discrete system of equations needs to be solved up to a certain convergence criterion. For general engineering problems the stability of the numerical process is strongly dependent on the timestep. when the method is not stable. The ﬁrst important property is consistency. Properties of numerical solution methods In order to solve the governing equations in a satisfactory manner. applied in general commercial and noncommercial CFD solvers. Certain variables in the governing equations contain physical bounds. Therefore. the fourth aspect of a CFD simulation is the iterative solver. The discretisation should become exact when the mesh resolution tends to inﬁnity. the solution process is said to be converged. Convergence is diﬃcult to prove theoretically for real engineering applications.2. a grid and timestep convergence study has to be performed using increasing grid resolution and decreasing timestep. a convergence criterion needs to be applied for the inner (within the linear system) and outer iterations (to couple the nonlinear parts and perform nonorthogonal corrections). Since it is often not possible to ﬁnd a numerical method which outperforms on all aspects. depending on the temporal discretisation scheme. limiting the iterative solver. stability. for the complete computational domain as a whole. without sources or sinks.
5 a general transport equation will be discretised to show how to deal with the diﬀerent terms. When the numerical process respects these physical bounds. The used opensource solver. For mass conservation.7. the spatial and temporal discretisation methods are described in section 2. (2.3.. These equations are derived from conservation of mass. 2005). 1991. Additionally. The validation and veriﬁcation discussion is the subject of section 2. like diﬀusion and convection. In this thesis. momentum and energy within an inﬁnitesimally small spatial control volume. Finally. the solution procedure to solve the incompressible NavierStokes equations will be dealt in section 2. OpenFOAM . The nabla ∇ operator is deﬁned in three dimensions as ∇= ∂. To solve the governing equations. which has been an authority in the ﬁeld of computational ﬂuid dynamics for decades. ∂. like discretisation schemes. Fluent is a generalpurpose CFD solver. providing support for code development. provides the source code and there is a big user community. the commercial ﬂow solver Fluent and the noncommercial opensource code OpenFOAM are used. some small test problems are deﬁned in order to test the inﬂuence of diﬀerent numerical settings. Fluent and OpenFOAM . 2. a brief discussion about the treatment of boundary conditions is provided in section 2.2 The NavierStokes equations The governing equations for viscous ﬂuid ﬂow are a coupled set of nonlinear partial diﬀerential equations (Anderson Jr. ∂x ∂y ∂z . Section 2. Two major drawbacks of a commercial solver are the unavailability of the source code and the potential lack of suﬃcient support from the company or the user community in code development. In order to validate and verify the CFD codes for our problem.10. ∂. two diﬀerent CFD codes. Panton.16 Finite volume discretisation density and all other nonnegative variables. the method is called bounded. grid resolution and timestep size.9 provides a brief description of the background and usage of both CFD codes.1) ∂t Here. These numerical solution procedures are present in the used ﬂow solvers. ρ [kg/m3 ] is the density and u [m/s] the ﬂow velocity vector.11. the major conclusions of this chapter are summarised in section 2. + + . the following continuity equation is obtained: ∂ρ + ∇ • (ρu) = 0. In section 2. like skewness and nonorthogonality in 2. When the discretised transport equation and corresponding boundary conditions are fully explained. followed by a discussion about the cell quality measures.4. The remainder of this chapter deals with a description of the governing equations of ﬂuid ﬂow in section 2.6.2.
with λ [W/(m·K)] the heat conduction transport coeﬃcient. compressible viscous ﬂows. For a Newtonian ﬂuid. the stress tensor.1). the heat conduction is described using Fourier’s law: q = λ∇T. ∂t (2. The governing equations (2. T ).2 The NavierStokes equations 17 Secondly. only for very simpliﬁed problems there exists an analytical solution. Here I represents the identity tensor. for momentum conservation the following expression can be derived (neglecting gravity and additional body forces): ∂(ρu) + ∇ • (ρuu) = ∇ • σ. Without other restrictions.2) where σ [N/m2 ] is the surface stress tensor. the equation of state is speciﬁed. turbulence research. σ.1 Constitutive relations In order to close the system of equations (2. these equations can be diﬃcult to solve and simpliﬁcations can be made if applicable to the concerning problem. q [W/s] is the heat ﬂux vector and Q [J·m3 /kg] equals the nett energy generation. For compressible ﬂow calculations also the energy conservation equation is speciﬁed: ∂(ρe) + ∇ • (ρeu) = ∇ • (σu) − ∇ • q + ρQ. p [N/m2 ] is the pressure and µ [Ns/m2 ] is the dynamic viscosity.2.2) and (2. these equations are used for high and low speed ﬂows. Additionally. (2. The NavierStokes equations are nonlinear.2. which makes them diﬃcult to solve. 2. necessary in viscous ﬂuid ﬂow. such as the perfect gas law: p = ρRT. However.3) in combination with additional turbulence modelling can be used in a wide variety of engineering problems. To close the energy equation.3) where e [J/kg] is the total speciﬁc energy (including kinetic and potential energy). .2) and (2. in which T [K] is the temperature and R [J/(mol·K)] the speciﬁc gas constant. (2. constitutive relations are needed. The constitutive relation for the total speciﬁc energy yields as follows: e = e(p.1). which is deﬁned for a Newtonian ﬂuid as 2 σ = − p + 3 µ∇ • u I + µ ∇u + ∇uT . are called the NavierStokes equations. The full set of equations describing unsteady. multiphase ﬂows and a lot of other applications. ∂t (2.3).
. 2003. this system of equations is closed such that there is no need to use the energy equation and additional turbulence modelling. A ﬂow can be assumed to be incompressible.2. 2. (2.5) the following nondimensional form of the incompressible continuity and momentum equations is obtained: ∇ • u∗ = 0. 2003. ∇p ∂u + ∇ • (uu) = − + ν∇2 u. Therefore. these relevant timescales are. (2.. x.9) Tvisc Uref Lref = . as follows: u∗ = u . the incompressible laminar NavierStokes equations are solved for Reynolds numbers ranging from Re = 100 to 1000.10) ν Tconv These dimensionless numbers represent order estimates for timescale ratios in the ﬂow. the relative relevance of the diﬀerent terms in equations (2. In case of incompressible ﬂow. which is considered to be incompressible (Lentink. two main dimensionless numbers are identiﬁed as relevant parameters. Therefore.4) (2.5) is revealed by making those equations dimensionless. the Strouhal (St) and Reynolds number (Re): St St = fref Lref Tconv = .3 Dimensionless numbers In general. When substituting equation (2. u. ∂t ρ (2.10). ρref (2. t. x∗ = x . Uref Tmotion (2. such that ρ∗ = 1. 2008)) and thermal expansion eﬀects can be neglected. L p∗ = p 2 . ρref · Uref ρ∗ = ρ . the timescale Re = . Bos et al. respectively.2. In (2. the density is constant.9) and (2.6) into equations (2.3 times the speed of sound (Lentink. p and ρ are scaled with their reference values.4) and (2.18 Finite volume discretisation 2. The incompressible NavierStokes equations are: ∇ • u = 0. Uref t∗ = t · fref .5) with ν = µ/ρ [m2 /s] being the kinematic viscosity. the main variables.6) The star (*) is used to indicate the dimensionless variables.7) 1 2 ∗ ∂u∗ + ∇ • (u∗ u∗ ) = −∇p∗ + ∇ u . Bos et al.8) ∂t Re In these equations.4) and (2. 2008) and laminar (Williamson. 1995).2 Incompressible laminar ﬂow simpliﬁcations The current research deals with the ﬂow around ﬂapping wings at insect scale. when the velocity is lower than 0. and (2. For an incompressible ﬂow.
Tvisc and the relevant timescale of the body motion. combined with the constitutive relations. Discretisation is performed assuming a linear variation of scalar variable φ across . consequently. face surfaces and points. the reference values need to be chosen appropriately. the governing equations need to be approximated over these cells.2. Tmotion .2. completely ﬁlling the domain. i. which will be explained in section 2.3 Spatial and temporal discretisation 19 for convective transport. OpenFOAM uses a collocated variable arrangement (Ferziger & Peric. with centres P and N from ﬁgure 2. the governing equations are signiﬁcantly simpliﬁed. Derived boundary ﬁelds. solving the unsteady incompressible laminar ﬂow can be seen as performing a Direct Numerical Simulation.2 shows an arbitrary polyhedral control volume VP with centre P and neighbouring centre N .5. The full NavierStokes equations. 1996): (x − xP )dV = 0. Since time can be interpreted as a parabolic coordinate (Patankar & Spalding.3 Spatial and temporal discretisation This section deals with the spatial and temporal discretisation of the governing mathematical equations. are applicable to all kind of ﬂows. starting at the initial solution. which can only be solved analytically for extremely simpliﬁed model problems. dependent on the maximal Courant number. are deﬁned in the face centre.e. however. Faces which are not shared are boundary faces. Tconv . viscous transport. 1972). In order for the dimensionless numbers to have a proper physical meaning. VP Every two cells. these simpliﬁcations need to be justiﬁed by the concerned ﬂuid ﬂow problem. at suﬃciently low Reynolds numbers. The timestep may vary. 2002). Therefore. After the domain is discretised into a set of control volumes. needs to be discretised throughout the entire computational domain. However. such that the costs for solving may be reduced. like pressure and velocity. It was seen that ﬂuid ﬂow is governed by nonlinear partial diﬀerential equations. share an internal face whose geometric centre is denoted by f and has an outward pointed normal vector Sf . which is used to march linearly in time. The computational point xP is located at the centroid of the computational cells. like surface normal gradients or face ﬂuxes. Space. it is suﬃcient to specify the initial timestep. When the ﬂow is considered laminar and incompressible. Concerning ﬂapping wing physics at lower Reynolds numbers (100 ≤ Re ≤ 1000) the ﬂow inherently is incompressible and laminar. which means that every control volume centre is used to store the values of all variables. The ﬁnite volume approach needs a domain subdivision into a ﬁnite number of convex polyhedral control volumes without overlap. where the computational costs strongly depend on the desired resolution and solution methods. which is found from the following relation (Jasak. 2. Figure 2.
11) where O((x − xP )2 ) represents the secondorder truncation error. 1996). a cell.2 Discretisation of the computational domain using ﬁnite volume cells. d. The vector from the cell centre to the neighbouring cell centre N is df .4 Measures of cell quality Since the accuracy of the numerical solution heavily depends on the interpolation from cell to face centre. The faces of cell P are directed with the unit normal vector Sf and may have an arbitrary number of corners. An arbitrary polyhedral control volume is constructed around a centre P and with volume VP . 2. We will brieﬂy describe the cell quality based on nonorthogonality and skewness. which is not within the scope of the present thesis.3(a) by the angle αN between the face normal vector Sf and the line connecting the two cell centres. Both truncation errors can be expanded using a full Taylor series expansion. including local mesh reﬁnement. ∂t ∆t (2. 1996). similar to the spatial truncation O(∆x2 ). Using a Taylor series approximation.12) With this linear temporal behaviour of φ the truncation error is secondorder O(∆t2 ). From (Jasak. Ferziger & Peric. the following expression is obtained: φ(x) = φP + (x − xP ) · (∇φ)P + O((x − xP )2 ). this method can be used for complex unstructured threedimensional meshes. which will both be used to assess the performance of mesh motion solvers in chapter 3. one can imagine that the cell quality is very important. (2. Since this discretisation approach is able to cope with arbitrary polyhedral cell volumes. This angle needs to be as small as possible in order to minimise the truncation . This scalar variable φ can be seen as pressure or a velocity component. For the temporal variation of this scalar variable φ a similar expression can be found: ∂φ(t) φ(t + ∆t) − φ(t) = + O(∆t). Jasak. cell nonorthogonality is deﬁned in ﬁgure 2. 2002. but can be found in (Wesseling. First of all. 2001.20 Sf f Finite volume discretisation N df P rP z y VP x Figure 2.
Figure 2. From (Jasak.13) Here. The second quality criterion is the cell skewness.3(b). u [m/s] the transport velocity and ν = µ/ρ [m2 /s] is the kinematic viscosity. error of the diﬀusion term. When the line connecting the two neighbouring face centres does not coincide with the connecting face centre. The cell skewness is deﬁned by the vectors m and d. This expression contains a temporal.2. d where m and d are deﬁned in ﬁgure 2. which forms the basis for the incompressible NavierStokes equations. con . ρ [kg/m3 ] is the reference density. see ﬁgure 2. Cell nonorthogonality is deﬁned as the angle between the face normal vector Sf and the direction vector between two cell centres P and N . ∂t ρ (2. 2.5 Discretisation of an incompressible momentum equation 21 Sf f m N Sf αN P f d N d fi P (a) Cell nonorthogonality. This partial diﬀerential equation has the following form: ∇p ∂u + ∇ • (uu) − ∇ • (ν∇u) = . the cell is skewed. Twodimensional representation of cell nonorthogonality (a) and cell skewness (b) as a measure for the ﬁnite volume cell quality. Assessing cell skewness is important.3 Quality measures using cell nonorthogonality and cell skewness. 1996). (b) Cell skewness.3(b). since the interpolation from cell centre to face centre strongly depends on this quality criterion as will later be seen in this chapter.5 Discretisation of an incompressible momentum equation This section deals with the temporal and spatial ﬁnite volume discretisation of the incompressible momentum equation. The degree of skewness is deﬁned by: ψ= m .
the integral form of the incompressible NavierStokes equations is obtained by integrating over a control volume. For this. (x − xP )dV ) · (∇φP ) Similar. dS • a. Before dealing with the discretisation of the diﬀerent terms of equation (2. given by: ∂u ∂t Finite volume discretisation ∇ • (uu) ∇ • (ν∇u) : temporal term.14) This equation is solved in both CFD codes used. (2. In the remainder of this section. like: ∇ • adV VP = SCV dS • a.. a secondorder approximation is obtained.14) are elaborated in more detail. Using Gauss’ theorem. which may represent the diﬀerent velocity components.16) . When substituting equation (2. the diﬀerent terms of equation (2. The divergence and gradient terms are evaluated using Gauss’ theorem (Panton.14) it is important to discuss the evaluation of the volume. ρ (2. f Sf = ≈ dSf • af .22 vection and a diﬀusion term. f (2. the scalar variable φ is used. surface. Anderson Jr. necessary to understand the evaluations of the convection and diﬀusion terms. CV : VCV ∂u dV + ∂t VCV ∇ • (uu)dV − ∇ • (ν∇u)dV = VCV VCV ∇p dV. : diﬀusion term. divergence and the gradient integrals. Fluent and OpenFOAM . the volume integral of the divergence of a vector a can be written as the sum of all faces. : convection term. such that the result is a multiplication of the scalar value multiplied by the cell volume. 1991).11) into the volume integral. Using the ﬁnite volume approach. which deﬁnes a relation between the volume and the surface integrals. φ(x)dV VP = VP (φP + (x − xP )(∇φ)P )dV dV + ( VP VP = φP ≈ φP V P . 2005.15) where Sf is the face surface area. for the surface integral the following yields: f dS · a = Sf · af .
2. 1993).4 (Sweby.16) and a linearization . dSφ. A linear interpolation is performed using the following expression: φf = fx φP + (1 − fx )φN . The SuperBee. as: ∇φdV VP = SCV dSφ. Discretisation of the gradient integral of a scalar variable φ can be written. the ﬂux φf is interpolated from the cell centre to the face centres. (1999) applied extra face interpolation schemes. f Sf = ≈ dSf .1 Face interpolation schemes Similar to the divergence term.5 Discretisation of an incompressible momentum equation 23 Here. or Van Leer (Van Leer. Koren and Van Leer limiter are shown using the Sweby diagram in ﬁgure 2.2. CV is the control volume with surface normal vectors Sf and af is a vector interpolated to the cell faces using a secondorder linear interpolation method. In OpenFOAM . f 2.2 Convection term When the volume integral of the convection term from equation (2. Jasak et al.5. using Gauss’ theorem. fx = f D/P D. Section 2. the following relation can be derived using equation (2. 1986). So. the Koren limiter (Koren. The purpose of ﬂux limiters is to limit the gradient of the solution in order to avoid spurious oscillations and to improve the stability of the scheme. 1984). both divergence and gradient volume integrals can be reduced to a summation of the corresponding vector or scalar variable over the cell faces.13) is considered. 1984).10 shows a comparison of the results using diﬀerent ﬂux limiters on the solution of a model problem of vortex decay and convection. Sweby. The linear interpolation factor fx is deﬁned as the ratio of two distances. 1979. The standard face interpolation scheme is obtained by central diﬀerencing. which is illustrated in ﬁgure 2. like upwind blending using a gamma coeﬃcient and diﬀerencing using a ﬂux splitting limiter such as SuperBee (Roe.5.φf .5.
Gamma.13) is discretised and approximated using linearization as ∇ • (ν∇φ)dV VP = f Sf · (ν∇φ)f νf (S · ∇φ)f . (S · ∇φ)f = m d . method (midpoint.g.5 0 0 0.5 r 2 2. Van Leer.5 r 2 2. which represents the ratio of successive gradients on the mesh. e.5 1 0.5.5 1 0.4 Diﬀerent ﬂux splitting limiters. given by F = Sf · (u)f . The other term (S · ∇φ)f is obtained on a nonorthogonal mesh by the following expression: φN − φP + k · (∇φ)f . 1984). 2. 1986). 1979.5 3 φ(r) 3 2.5 1 1.3 Diﬀusion term The volume integral of the diﬀusion term from equation (2.5 1 1.5 3 Finite volume discretisation 3 2. The scalar variable φ needs to be interpolated using a secondorder interpolation method in combination with a ﬂux limiter. Diﬀerent ﬂux limiters are employed. 1993) and (c) Van Leer (Van Leer. the terms (S · ∇φ)f and νf need to be approximated using a proper method.5 0 0 0. F is the mass ﬂux. Sweby.24 3 2. (b) Koren (Koren. The face viscosity νf is obtained by interpolation from cell centre to faces. The ﬂux splitting scheme are a function of r. (a) SuperBee (Roe.5 2 1. least squares): ∇ • (uφ)dV VP = f Sf · (uφ)f Sf · (u)f φf F φf .5 2 φ(r) 1. linear.5 1 1. = f = f Here.5 0 0 0.5 1 0. = f Here. Flux splitting schemes are used to limit the gradient of the solution in order to avoid spurious wiggles.5 2 φ(r) 1.5 r 2 2.5 3 (a) SuperBee (Roe) (b) Koren (c) Van Leer Figure 2.
ts
2.5 Discretisation of an incompressible momentum equation
φD
25
φf Sf φP φU U P Flow direction f D P d f αN m N k
Figure 2.5 Variation of the ﬂux φ. The value of φ at the face f is determined as a function of upstream and downstream values.
Figure 2.6 Cell nonorthogonality treatment. Illustration of the cell nonorthogonality correction which is used on meshes with large skewness and nonorthogonality.
Here d is the vector between two adjacent cell centres and m is parallel to d with magnitude of the surface normal vector Sf . The decomposition of Sf is shown in ﬁgure 2.6 and derived in Jasak (1996), Juretic (2004) such that the following relation holds: Sf = m + k, where k is orthogonal to the surface normal vector S.
2.5.4
Temporal term
Since the unsteady NavierStokes equations are solved, a proper discretisation of the temporal scheme is necessary. The time derivative represents the temporal rate of change of φ which needs to be discretised using new and old time values. This time diﬀerence is deﬁned using prescribed timestep size ∆t such that: φn+1 = φn + ∆t, where φn and φn+1 are the scalar variable φ at the old and new time instances, respectively. Two implicit time discretisation methods are considered, one ﬁrstorder and and one secondorder scheme. The ﬁrstorder discretisation is simply the temporal diﬀerence: ∂φ φn+1 − φn = , ∂t ∆t and the secondorder discretisation, see (Wesseling, 2001, Hirsch, 1988), is given by: 1 3 n+1 φ − φn + 2 φn−1 ∂φ = 2 . ∂t ∆t This implicit scheme is referred to as the secondorder backward diﬀerencing scheme, where φn−1 is the oldold value of φ. Consequently, the corresponding
26 volume integrals obey the following relations:
Finite volume discretisation
CV
φn+1 − φn ∂φ dV = VP , ∂t ∆t
3 n+1 2φ 1 − 2φn + 2 φn−1 VP . ∆t
CV
∂φ dV = ∂t
(2.17)
Note, that these relations are only valid on ﬁxed meshes and constant timesteps. According to (Wesseling, 2001, Hirsch, 1988), the explicit ﬁrstorder time integration method may be unstable if the Courant number is larger than 1, where the Courant number is deﬁned as u · ∆t Co = . ∆x Implicit methods are in general more stable, compared to (semi) explicit methods, such that in the current research the implicit ﬁrst and secondorder backward scheme have been used. While the implicit methods are bounded and stable, a predeﬁned maximal Courant number Comax is used to vary the corresponding 3 timestep during the simulation. In that case, the coeﬃcients 2 , 2 and 1 in (2.17) 2 should be elaborated to incorporate the ratio of the old and current timesteps. 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 deﬁned 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. zerogradient boundary condition, deﬁning the solution gradient to be zero. This condition is known as a Neumanntype condition, ∂φ/∂n = a, 2. ﬁxedvalue boundary condition, deﬁning a speciﬁed value of the solution. This is a Dirichlettype condition, φ = b, 3. symmetry boundary condition, treats the conservation variables as if the boundary was a mirror plane. This condition deﬁnes that the component of the solution gradient normal to this plane should be ﬁxed to zero. The parallel components are extrapolated from the interior cells, 4. movingwallvelocity boundary condition is used on a moving boundary to keep the ﬂux zero, using the Arbitrary Lagrangian Eulerian approach. For external ﬂow simulations, a distinction is made between the outer and the inner boundaries, the latter corresponds to the moving wing or body. To minimise the eﬀects of the outer boundaries it is desirable to specify a symmetry boundary
2.7 Solution of the NavierStokes equations
27
condition (3) at those ﬁxed boundaries, unless a freestream is speciﬁed. In case of forward ﬂapping ﬂight, two domain boundaries are deﬁned as inﬂow and outﬂow, respectively. At the inﬂow boundary the velocity is deﬁned as ﬁxedvalue (2) and the pressure as zerogradient (1). On the other hand, at the outﬂow boundary, the pressure has to be ﬁxedvalue and the velocity zerogradient (Hirsch, 1988, Wesseling, 2001). On a stationary wall the noslip condition needs to be guaranteed, therefore a ﬁxedvalue (u = 0) is speciﬁed for the velocity in combination with a zerogradient for the pressure. If the boundary of the wall moves, than the proper boundary condition is the movingwallvelocity (4) which introduces an extra velocity in order to maintain the noslip condition and ensures a zero ﬂux through the moving boundary.
2.7
Solution of the NavierStokes equations
Previously, the diﬀerent terms to discretise the general momentum equation (2.13), were described. This section brieﬂy deals with the discretisation of the NavierStokes equations and the solution procedure. The incompressible laminar NavierStokes equations were given by (2.4) and (2.5): ∇ • u = 0, ∂u ∇p + ∇ • (uu) = − + ν∇2 u. ∂t ρ There are two items, requiring special attention, namely the nonlinear term present in the momentum equation and the pressurevelocity coupling (Ferziger & Peric, 2002). The nonlinear term in these governing equations, ∇ • (uu), can be solved either by using a solver for nonlinear systems or by Newton linearization. Previously, it was seen that the convection term can be written as: ∇ • (uu)dV
VP
=
f
Sf · (u)f (u)f F(u)f
=
f
= ap up +
N
aN uN ,
where ap , aN and F are still depending on u. ap and aN represent the diagonal and oﬀdiagonal 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 nonlinear systems, this convection term is linearised such that existing velocity ﬁelds will be used to calculate the matrix coeﬃcients ap and aN .
28
Finite volume discretisation
2.7.1
Pressure equation and PressureVelocity coupling
Since the pressure depends on the velocity and viceversa, a special treatment of this interequation coupling is needed. In order to derive the pressure equation, a semidiscrete formulation of the momentum equation is written as: ap up = 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. Following 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 contribution. The convection part includes the matrix coeﬃcients 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) = − aN uN +
N
u0 . ∆t
Additionally, the discretised continuity equation (2.4) is given by: ∇•u =
f
Sf • uf = 0,
(2.19)
Now equation (2.18) is rewritten to ﬁnd an expression for up : up = H(u) 1 − ∇p. ap ap (2.20)
The velocities on the face of a ﬁnite volume cell can be expressed as the interpolated value on the face of equation (2.20): uf = H(u) ap − 1 ap (∇p)f .
f
(2.21)
f
This equation will be used to determine the face ﬂuxes. If equation (2.21) is substituted into equation (2.19), the following pressure equation can be obtained: ∇•
1 ap ∇p
= ∇•
H(u) ap
.
(2.22)
The Laplacian operator is discretised using existing methods, which are previously explained. Combining equations (2.18) to (2.22), the ﬁnal form of the discretised NavierStokes equations can be written as: ap up = H(u) − Sf (p)f ,
f
(2.23)
a brief description is given. are lagged. 2. from the previous stage.22) can be formed. the face ﬂuxes are guaranteed to be conservative (Ferziger & Peric.22) is satisﬁed. Using the predicted velocity. since the actual pressure gradient is not yet calculated. which means that these equations are solved in sequence.7. 2.24) it can be observed that both equations are coupled through the pressure and velocity. Pressure solution stage. The last equation in this sequence is (2.20) in an explicit fashion.23) and (2. 1972) based algorithms.2. known from the previous timestep. The PISO algorithm consists of the following steps: 1. In unsteady simulations all other interequation couplings. 1986) or SIMPLE (Patankar & Spalding. For incompressible unsteady ﬂow . The momentum equation (2. please consult (Jasak. f (2. 2002). Ferziger & Peric. Juretic. (2. Explicit velocity correction stage. which are consistent with the new pressure ﬁeld. Using this pressure equation a better approximation of the new pressure ﬁeld can be obtained. For the discretised form of the NavierStokes equations (2.23) is solved using the pressure gradient. it has become possible to describe the solution procedure to obtain the solution of the NavierStokes equations. from stage 2. which determines the conservative ﬂuxes. is replaced by a better pressure ﬁeld. 2004). Fluent and OpenFOAM used the PISO scheme for transient ﬂows and the SIMPLE scheme for steady ﬂows. 3. besides the pressurevelocity equations. equation (2.24) additionally.7 Solution of the NavierStokes equations and Sf • f 29 1 ap (∇p)f = f Sf • H(u) ap . the H(u) term can be constructed such that the pressure equation (2. Since the PISO scheme was used. the velocity ﬁeld has to be corrected accordingly.25) f When equation (2. The innerequation coupling is established using either PISO (Issa.25). which requires special attention. this system is solved in a segregated manner. the face ﬂux is calculated using: F = Sf • u = Sf • H(u) ap − 1 ap (∇p)f . Momentum predictor stage. Furthermore. 2002. from stage 1.23) provides an approximation of the new velocity ﬁeld. This is performed using (2. such that they are included in the PISO loop. 1996. Since the approximated pressure ﬁeld.2 Procedure for solving the NavierStokes equations After dealing with the discretisation of the NavierStokes equations in combination with the PISO algorithm. Both ﬂow solvers. Since a simultaneous approach would be too computationally demanding. For more detailed information.
The Lagrangian contribution allows the mesh to move and deform according to the boundary motion. like pressure. commonly used to satisfy conservation on deforming meshes. the mesh deforms with it. 2. because it requires the explicit tracking of the domain boundary. all remaining equations of the system are solved. go back to step 2. the solution sequence can be summarised as follows: 1. as well as a new set of conservative ﬂuxes. 3. when the ﬂow domain moves or deforms in time due to a moving boundary. a ﬁxed mesh becomes inconvenient. special attention is necessary to describe the modiﬁcations to the discretised equations dealing with the Arbitrary Lagrangian Eulerian approach. However. 1982). Using the new conservative ﬂuxes. 5. At the time of writing. update the turbulent viscosity at this stage. If the material or ﬂuid deforms. Initialisation of all ﬁelds. This procedure results in solution ﬁelds for all solved variables. whereas the Eulerian part takes care of the ﬂuid ﬂow through the mesh. where the mesh is ﬁxed to the ﬂuid or material. This method incorporates and combines both Lagrangian and Eulerian frameworks. 6.7. the ALE method has become the standard implementation in most popular CFD codes to solve for the ﬂow around moving boundaries while the . the Arbitrary Lagrangian Eulerian (ALE) formulation is used to discretise the ﬂow equations on moving and deforming meshes (Donea. Start the simulation to obtain the velocity and pressure values at the new timestep. velocity and possible turbulence variables. If turbulence modelling is included. Since the present thesis deals with deforming mesh problems. This method is commonly used to discretise the governing equations encountered in structural mechanics. including pressure and velocity. where the ﬂuid is allowed to ﬂow through the ﬁxed mesh (Ferziger & Peric. 2002). This is in contrast to the Lagrangian formulation. using the initial condition. 4. The pressure and velocity ﬁelds are obtained for the current timestep. 2. Unless the ﬁnal time is reached.3 Arbitrary Lagrangian Eulerian approach The governing equations to solve the ﬂow are generally discretised using the Eulerian description. Therefore. Iterate through the PISO loop until the predeﬁned tolerance of the pressurevelocity system is reached. Create and solve the momentum predictor equations using the obtained face ﬂuxes.30 Finite volume discretisation with additional turbulence modelling.
e. otherwise inconsistency could introduce numerical errors.7 Solution of the NavierStokes equations 31 mesh deforms accordingly. the secondorder accurate three time levels backward scheme was used throughout this research. In time. For c eﬃciency. The current research uses . The ﬂuid mass ﬂux m is obtained as part of the solution. dV − ∂t VCV SCV Using the current ﬁnite volume discretisation. the momentum equation (2.26) where VCV is the arbitrary volume and us the velocity of the moving surface. the maximal Courant number varies during the simulation. If equation (2. 1996): ∂ n · us dS = 0. using either an implicit ﬁrstorder Euler method or a secondorder backward scheme (Tukovi´ & Jasak. Note.2. The mass ﬂux through the face is given by mf = nf • uf Sf and ˙ ˙ the cell face volume change by Vf = nf • usf Sf . (2.8. the temporal dimension is marched using a variable timestep. that the previous equation is derived for a constant timestep. the computational domain is split into a ﬁnite number of polyhedral cells with varying shape and volume. where us represents the cell face velocity. satisfying mass ˙ conservation. φ f f f f + = where the subscript P denotes the cell values and f represents the values at the face centres. old and oldold values. 2007). corresponding to a maximal Courant number.26) is discretised in space and time the following relation is obtained for a constant timestep: n+1 n−1 n−1 n−1 n 3ρn+1 φn+1 VP − 4ρn φn VP + ρP φP VP P P P P 2∆t ˙ (mn+1 − ρn+1 Vfn+1 )φn+1 ˙f f f f n+1 n+1 (ρΓφ )n+1 Sf nn+1 • (∇φ)n+1 + sn+1 VP . the new. n and n−1 are. The superscripts n+1. see section 2. It is very important to determine the volume face ﬂux in a consistent way such that it equals the swept volume calculation. it is important to determine the volume face ﬂux such that it satisﬁes the Space Conservation Law. 2002) or Geometric Conservation Law (Lesoinne & Farhat. 2009).14) for a scalar ﬁeld φ can be derived on a moving mesh as ∂ ∂t ρφdV + VCV SCV ρn·(u−us )φdS− SCV ρΓφ n·∇φdS = VCV Sφ (φ)dV. since the mesh is deforming. The temporal discretisation scheme should be similar to the one used in the momentum equation. i. In general. respectively. The cells do not overlap and completely ﬁll the domain (Jasak. Furthermore. The relationship between the rate of change of the volume VCV and the velocity us of the boundary surface S is deﬁned by the socalled Space Conservation Law (SCL) (Ferziger & Peric.
the polyhedral cells and faces are decomposed into tetrahedral cells. described in the previous section. tn and tn+1 . it may occur that the volume becomes warped.32 Finite volume discretisation tn+1 f P z y f V x tn (a) (b) Figure 2. The swept volume V of a decomposed face is shown in (b). (a) shows the decomposition of a polyhedral cell into tetrahedral volumes and faces. The ﬁnite volume cell decomposition is used to form a tetrahedral mesh used for pointbased mesh motion solvers and to calculate the swept volumes. If a polyhedral face is swept from one timestep to the next.7(a). 2. Therefore.7 Finite volume cell decomposition to calculate swept volumes. The polyhedral face is decomposed into triangles. . using its centroid.8 Swept volume calculation The swept volume is deﬁned as the volume swept by a face of a polyhedral cell between two subsequent timesteps. Therefore. such that a volume calculation will not be trivial. a deﬁned maximal Courant number leading to a varying timestep. φ f f f f where the new and old timesteps are respectively given by ∆tn+1 = tn+1 − tn and ∆tn = tn − tn−1 . the following relation is derived for a nonconstant timestep: 1+ ∆tn+1 ∆tn+1 + ∆tn n+1 ρn+1 φn+1 VP − 1 + P P ∆tn+1 ∆tn n ρn φn VP P P + + = 1+ (∆tn+1 )2 ∆tn (∆tn+1 + ∆tn ) f n−1 n−1 n−1 ρP φP VP ˙ (mn+1 − ρn+1 Vfn+1 )φn+1 ˙f f f n+1 n+1 (ρΓφ )n+1 Sf nn+1 • (∇φ)n+1 + sn+1 VP . This calculation is necessary in order to satisfy the Space Conservation Law. see ﬁgure 2.
Vp4 and Vp5 correspond to diﬀerent tetrahedron volumes. i. Therefore. For completeness.9 Numerical ﬂow solvers 33 which is illustrated in ﬁgure 2.2. The pressurevelocity coupling in incompressible ﬂow simulations was obtained using the iterative PISO scheme (Ferziger & Peric. To obtain the total swept volume of the decomposed triangle Vtriangle . using either one of the two diagonals in the right side face of the swept volume shown in ﬁgure 2. Since this prism may be warped.e. One tetrahedron is shown in the ﬁgure. Fluent is a wellknown. The accuracy was set to doubleprecision and the initial conditions were chosen to be uniform. The convergence criterion for . 6 1 (Vp1 + Vp4 + Vp5 ). necessary for ﬂuid ﬂow. the volume is calculated as the sum of three tetrahedron volumes. but the remaining volume of the prism contains two more tetrahedrons. are brieﬂy discussed. which exploits the ﬁnite volume approach. The spatial discretisation was secondorder upwind and the time discretisation was ﬁrstorder implicit Euler (Hirsch. 2.7(b).7(b). the CFD codes. These two additional tetrahedrons can be constructed in two diﬀerent unique ways. 1988). The swept volume of a polyhedral face f is equal to the sum of the swept volumes of the diﬀerent decomposed triangles. this method is implemented in OpenFOAM . used throughout the present research. 2 This swept calculation is successfully validated on test cases using unsteady ﬂow with mesh motion and proved to be suﬃciently accurate. Two diﬀerent CFD codes are used. This section deals with a brief elaboration of the computer software. the average of both V1 and V2 is taken as V2 = Vtriangle = 1 (V1 + V2 ). 2002). 2006).7 described the solution procedure to solve the NavierStokes equations. easy to use and proven CFD solver. The boundary condition on the body was set to noslip.7(b). one commercial (Fluent ) and one opensource package (OpenFOAM ). which is the only method for which the dynamic mesh module is implemented by Fluent . the main settings that we used. As is illustrated in ﬁgure 2. Therefore. see (Zuijlen van. 6 where the same base tetrahedron Vp1 is used and Vp2 . which need to be accurately calculated.9 Numerical ﬂow solvers Section 2. the swept volume of such a triangle is similar to a prism with a triangleshaped bottom area. the total swept volume using the two diﬀerent unique tetrahedron decompositions is: V1 = and 1 (Vp1 + Vp2 + Vp3 ). Vp3 .
One of the main assets of the code is that the user writes the code in an intuitive way himself. c 2009) and Large Eddy Simulations (Jasak. relevant to ﬂapping insect ﬂight in (Lentink & Gerritsma. 2002). 2003. The commercial ﬂow solver Fluent has already been tested earlier speciﬁcally for low Reynolds number ﬂows. making this code very versatile. but not for low Reynolds number ﬂows. All terms are discretised using standard secondorder central diﬀerencing. Bos et al. Zuo et al. . which is used to solve partial diﬀerential equations. mesh motion (Jasak.10 Code validation and veriﬁcation This section deals with the validation and veriﬁcation of the CFD solvers that were used throughout the current research. Jasak et al. 2007). the iterative solvers and their corresponding convergence criteria need to be speciﬁed. while the pressurevelocity coupling equation employs its asymmetric counterpart preconditioned bistab conjugate Gradient (PBiCG) solver. which is derived from the complete system of equations. has been tested extensively for ﬂuidstructure interaction (Tukovi´ & Jasak. The convergence criterion is based on the residual. Additionally. the NavierStokes equations. such that the residual Res is deﬁned by: Res = b − Ax The pressure equation is solved using a preconditioned conjugate Gradient (PCG) iterative solver.. 1979). Therefore. however. turned out to be the linear scheme with the Van Leer limiter (Van Leer. The other code used in this research. 1996. OpenFOAM . OpenFOAM .10 will show that the best method to discretise the convection term.. Thaweewat et al. 2004. in combination with a variable timestep corresponding a maximal Courant number (Wesseling. 2007). 2008. 2007). e.. Ferziger & Peric. except for the convection term. 2001. written in C++.. on a ﬁnite volume mesh containing polyhedral cells. which is written as Ax = b. 2009). is a general objectoriented toolbox. for our low Reynolds number problems. without the need to dig deep in the underlying code. Appendix B summarises the used discretisation schemes and iterative solvers combined with the convergence criteria for both ﬂow solvers Fluent and OpenFOAM . Concerning the temporal discretisation scheme. Jasak et al. The preconditioning method varies from incomplete Choleski to incomplete LU decomposition (Wesseling.. the implicit secondorder backward scheme is used. 2007.34 Finite volume discretisation the iterative method was satisﬁed with mass and momentum residuals dropping O(10−5 ) in magnitude. Juretic. Section 2. this section only deals with the validation and veriﬁcation of the OpenFOAM ﬂow solver for test problems relevant for low Reynolds number ﬂapping insect ﬂight.g. 2001. 2.
2. 2003): Vθ = t−m f (η). two vortex cases are considered. This problem is relevant for low Reynolds number ﬂapping ﬂight.27) where Vθ represents the radial velocity. followed by the two vortex simulations to discuss the results. which solves for the unsteady. described in section 2.8 shows the velocity proﬁle for diﬀerent values of m to show the eﬀect on the compactness of the vortex. the most accurate scheme needs to be determined from linear differencing. Within the context of the present research. 2005) is obtained. on the other hand. Since this ﬂow solver was used for low Reynolds number ﬂows with vortex convection it is necessary to choose the correct face interpolation scheme for the convection term. Vortex deﬁnition When a vortex is used for code validation it is important that a well conﬁned deﬁnition is used. the compactness of the vortex is increased with increasing m. With decreasing m. the curve proﬁle becomes steeper until the asymptotic behaviour is lost at negative m.1. 2005. two test cases are investigated. Concerning the convection term. Figure 2. SuperBee splitting.5.10 Code validation and veriﬁcation 35 The opensource toolbox OpenFOAM provides a framework of ﬁnite volume based functions in order to build a speciﬁc application for solving partial differential equations. the accuracy of the code is assessed using unsteady ﬂows around static and plunging cylinders in section 2. For validation purposes a vortex corresponding to m = 1. the vortex should have ﬁnite and monotonic velocity and vorticity proﬁles. . Conversely. one concerning a decaying vortex.e. the other deals with the vortex convection. Therefore. m is a compactness coeﬃcient and η a similarity parameter. Koren splitting and Van Leer splitting. although maximal angular momentum (Panton.5 after which a region with counterrotation appears.5 the wellknown LambOseen vortex is the result. since these ﬂows are dominated by unsteady vortical structures.2. incompressible ﬂow using deforming meshes. gamma diﬀerencing. In this ﬁgure some interesting characteristics can be observed.1 2D vortex decay and convection To validate the code. the face interpolation for diﬀusion and source terms will be kept ﬁxed at secondorder linear (central) interpolation. the decay and convection of a Taylor vortex. i. When m is chosen to be 0.10. until m = 1. (2. After the selection of the proper ﬂow solver settings. In order to assess the accuracy of these interpolation schemes. which is not of the desired compactness.10.10. as was brieﬂy discussed in section 2. First a proper vortex deﬁnition is described. an application is developed. Zhou & Wei.5. but analogue to (Zhou & Wei. Besides the ﬂuid and ﬂow properties it is necessary to specify diﬀerent interpolation schemes for the diﬀerent terms of the governing equations. a Taylor vortex (Panton.2. The following deﬁnition is used from (Panton. 2005).
The mesh size was ﬁxed to (100 x 100) and the timestep is varying to meet a maximal Courant number of Comax = 1.5.36 Finite volume discretisation Figure 2.. The temporal term is discretised using a secondorder backward scheme. 1986). To study the eﬀect of the face interpolation of the convection term. which are both considered to be ﬁne enough. Koren limiter (Koren.28) Vθ = 2e 2 λe−2λ . 2005).9. where λ is a function of the similarity parameter η: λ= η 2 (2) . Vortex decay As a ﬁrst test problem. The Reynolds number . provides a better representation of vortical ﬂows like in low Reynolds number vortex shedding problems (Panton. a standard OpenFOAM solver is used. from (Panton. icoFoam. a Taylor vortex is considered. 1999). ranging from −0. This squared domain is discretised with a Cartesian ﬁnite volume grid of 100 x 100 mesh cells as is shown in ﬁgure 2. which decays in a twodimensional squared domain with dimensions (5 x 5). 1979) and standard linear interpolation. which is not necessary for this problem. 2003). The vortex deﬁnition. except for the convection term. Furthermore. This m values determines the compactness of the vortex. without mesh motion.0.5 to 2 in steps of 0.8 Diﬀerent velocity proﬁles of a similar vortex deﬁnition. 2005) and is given by: 2 1 (2. SuperBee limiter (Roe.5 provides the most compact vortex. Van Leer limiter (Van Leer. To solve the incompressible NavierStokes equations. all other terms. When the compactness coeﬃcient m is ﬁxed to 1. provides diﬀerent vortex proﬁles for varying m. where m = 1. which shows the vorticity of the vortex at t = 0. 1993). are discretised using a secondorder linear interpolation.5 the velocity proﬁle of this Taylor vortex can be derived (Panton. the following schemes are used: Gamma (Jasak et al. 2005).
Any excess or lack in numerical diﬀusion may become visible.0). One major drawback of these methods is that the vortex looses symmetry after 5 s. The Koren and Van Leer limiters are slightly more .10(c). without the presence of a convection freestream. The velocity proﬁles and total energy are monitored to identify the accuracy of the ﬂow solver.and Y direction for the diﬀerent face interpolation schemes. 0.01 and the velocity vector to u = (1. 0. Besides the evolution of the velocity proﬁle the total energy is shown in ﬁgure 2. The total energy is calculated as N Etot = i=1 0.10 Code validation and veriﬁcation 37 Figure 2. is that the total energy is increasing for the SuperBee ﬂux limiter. where i represents the cell index. Starting from this initial Taylor vortex solution the ﬂow diﬀusion is solved.10(b) show the initial and ﬁnal velocity variations in X.2. ﬁgure 2. The ﬁrst important observation. The velocity proﬁle as well as the total energy of both Gamma and linear interpolation schemes are similar.10. N the total number of cells and ui the velocity in cell i.0. which is a measure for the diﬀusion.5ui 2 . from ﬁgure 2. without a signiﬁcant amount of diﬀusion. This scheme clearly introduced a large amount of negative numerical diﬀusivity which causes the Taylor vortex to grow.10(b).9 Initial solution of a Taylor vortex on an Cartesian mesh. Results Figure 2.0. The ﬂow solver solves for 40 seconds such that the temporal eﬀect of the diﬀerent face interpolation schemes on the shape and magnitude of the velocity proﬁles can be compared.10(a) and 2. which is not physical. is ﬁxed to Re = 100 by setting the kinematic viscosity to ν = 0.
94 0 2 4 6 Gamma Koren SuperBee van Leer Linear (b) 10 8 Time [t] 12 14 16 (c) Figure 2.6 U velocity [m/s] 1 3 Initial solution Gamma Koren SuperBee van Leer Linear 1 V velocity [m/s] 0.10(c).5 4 3.02 1 0.5 Y coordinate [m] 5 4. such that the vortex was convected through the entire domain.96 0.38 6 5. shown in ﬁgure 2.5 1 3 1 0.6 0.0. see ﬁgure 2. The Reynolds number was ﬁxed to Re = 100 by setting a kinematic viscosity to ν = 0. the discretisation of the face interpolation is varied and all other discretisation terms are ﬁxed to secondorder linear interpolation.08 1.5 5 4 X coordinate [m] 5. while (b) provides v(x). The Cartesian mesh resolution was set to (400 x 100) and the max Courant number was equal to Comax = 1. Similar to the vortex decay problem.01 and an inlet velocity . for diﬀerent face interpolation schemes.2 0.10 Velocity and total energy variations of a decaying Taylor vortex.2 0.12 1. The Van Leer limiter is called shape preserving and provides good results for both vortex decay and convection as will be seen in the next section.04 1.5 6 (a) 1.2 0.11. but provide similar results. Vortex convection The second validation case concerns a Taylor vortex. The ﬂow solver solved the governing equations for 20 s. at t=20 s.1 Total energy [m2 /s2 ] 1. (a) shows u(y). diﬀusive.98 0.6 Finite volume discretisation Initial solution Gamma Koren SuperBee van Leer Linear 3.2 0.6 0. which is convected through a channel with dimensions (20 x 5).5 4.06 1. (c) shows the decaying total energy due to numerical diﬀusion.
2004). so a Reynolds number Re > 47 needs to be chosen. This Taylor vortex solution is used as the initial solution for the convection validation case. smearing the vortex.0. and more symmetrical vortices (Juntasaro & Marquis.and Y direction through the vortex core at t = 10 s.10 Code validation and veriﬁcation 39 Figure 2. the Van Leer limiter slightly outperforms the Koren limiter. to u = (1. 2004. Furthermore. depending on the convection velocity.12(b) show. Additionally. This eﬀect increases with time and therefore these two methods are not appropriate to study the vortical wake patterns in insect ﬂight. the Koren and Van Leer limiters are very close. Extra diﬀusion is clearly visible after t=14 s when the vortex approaches the outlet boundary. 2004). When looking in real detail to these results. Results First of all. since the integrated energy is close for all schemes. which is not relevant for unsteady insect ﬂight. 1998).11 Initial solution of a Taylor vortex on an Cartesian mesh. At Reynolds numbers less than Re = 47 the ﬂow exhibits steady behaviour (see Williamson. representing a channel. respectively. The secondorder linear and Gamma schemes lead to an overshoot of the velocity in Y direction. such that this scheme was used throughout this study.12(a) and 2. the SuperBee scheme possibly still has negative numerical diﬀusion. 0. the eﬀect of diﬀerent face interpolation schemes is compared.2.12(c) shows that the total energy in the entire domain is decreasing with time for all schemes. when taking a Reynolds . another observation can be made. the velocity in X. Again. as was seen in the results for vortex decay. The Van Leer limiter leads to smoother. On the other hand. two cylinder example problems are deﬁned at suﬃciently low Reynolds numbers. except for the SuperBee scheme. without overshoots and with a proper symmetry preservation of the vortex. Kuzmin & Turek. 0.0. Besides the possibly negative numerical diﬀusion in the SuperBee scheme (Juntasaro & Marquis.10.0). At suﬃciently large grid resolution and timestep size. ﬁgure 2. the total energy does not provide more information about the accuracy of the scheme. 2.2 Validation using cylinder ﬂows To validate the accuracy of the ﬂow solver for unsteady and vortical ﬂow. ﬁgure 2. Since convection induces physical diﬀusion.
The velocity in Xdirection u(y) is shown in (a) at time t=10 s. with periodic force histories. Figure 2. vortical ﬂow around ﬂapping wings.2 50.6 0. the ﬂow becomes turbulent and additional turbulence modelling becomes necessary. (c) shows the total energy for the convected vortex. The main parameter selected for comparison is the timeaveraged drag coeﬃcient. in the range 100 ≤ Re ≤ 200. where D is the cylinder diameter.12 Velocity proﬁles of a convected Taylor vortex. one concerns the ﬂow around a stationary cylinder at Re = 150 and the other involves a transversely oscillating cylinder at Re = 185 (Guilmineau & Queutey.2 0.9 Total energy [m2 /s2 ] 50.40 5 4.5 4 Y coordinate [m] 3.1 50 0 2 4 6 (b) Gamma Koren SuperBee van Leer Linear 8 10 12 14 16 18 20 Time [t] (c) Figure 2.4 0.2 1 U velocity [m/s] 1.5 0 0. 2002). above and below the cylinder.6 Gamma Koren SuperBee van Leer Linear 0.3 0. number larger than about Re = 185. Therefore.4 0. v(x).14 shows the computational domain used for both validation cases.2 0. Since the main objective of this research is to solve for unsteady.5 3 2.4 50. (b) illustrates the velocity proﬁle in Y direction as a function of the Xcoordinate.5 50.5 0.5 2 1.3 V velocity [m/s] 0.4 0. the boundaries are located at 10D before.4 1. The ﬂapping wing simulations are performed in the laminar ﬂow regime Re = O(100).5 12 Finite volume discretisation Gamma Koren SuperBee van Leer Linear 13 14 16 15 X coordinate [m] 17 18 (a) 51 50.6 50.8 1. The outﬂow . two validation cases were selected.7 50.3 50.8 50.1 0 0. the characteristics of that kind of ﬂow needs to be present in the validation cases.5 1 0.1 0. which is welldocumented in literature.
13(a) shows the relation between the Strouhal and Reynolds number.183 is obtained for a stationary cylinder at Re = 150. A Strouhal number of St = 0. according to (Williamson. Bos et al. dealing with the ﬂow around a static cylinder at Re = 150 is inherently laminar and unsteady. (b) relates the viscous drag coeﬃcient (◦). e. Flow around a stationary circular cylinder The ﬁrst case. where D is the cylinder diameter. boundary is located at a distance of 40D.18 0.4 1.2 0. Henderson (1995) performed a spectral element numerical study which is used as the baseline reference for this case. ﬁgure 2.2 0 101 102 Reynolds number [] 41 103 (a) (b) Figure 2. The . resulting in a periodic vortex wake.13(b) shows the results from an extensive study performed by Henderson (1995) to identify a relation between drag coeﬃcient and Reynolds number. The ﬂow is from left to right and the inlet boundary is located 10D upstream.g.17 0. Figure 2. which is related to the vortex shedding frequency.10 Code validation and veriﬁcation 0. the outlet 40D downstream and the upper and lower boundaries are located 10D from the cylinder surface.183 for Re = 150 at which the stationary cylinder validation case is performed.8 0.6 0. Previous studies.16 0. 1998).14 0.2 1 0.19 Strouhal number [] Drag coeﬃcient [] 0. Lentink & Gerritsma (2003). Figure 2.12 40 60 80 100 120 140 160 180 200 Reynolds number [] 2 1. (2008).6 1.13 Relationships between Reynolds number. The dot (•) shows a Strouhal number of St = 0. pressure drag coeﬃcient (•) and total drag coeﬃcient (△) to the Reynolds number (Henderson. with sizes 25k.2. showed that boundary eﬀects are minimal at this domain size.15 0. This grid. Strouhal number and drag coefﬁcient.4 0.8 1. (a) shows the relation between Strouhal number. 50k and 100k is used to validate the accuracy of the ﬂow solver. Additionally.13 0. 1995). and the Reynolds number (Williamson.14 Computational grid around a cylinder. 1998).
25k and Comax = 2. the timeaveraged drag coeﬃcient is determined and compared with literature in table 2. it can be seen that the Strouhal number matches the value from Henderson (1995) even more closely than the timeaveraged drag coefﬁcient. The drag and lift coeﬃcient are respectively deﬁned as: CD = D 1 2 2 ρUref . it can be observed that the drag coeﬃcient of the coarsest case. When the grid is reﬁned and the timestep decreased. The timestep is systematically decreased according to a maximal Courant number corresponding to Comax = 2.1. which reveals the presence of the Von K´rm´n vortex street a a behind the stationary cylinder. From these periodic forces. the grid size is varied from 25k.15 Vorticity visualisation of the Von K´rm´n vortex street. 1. Vorticity ω = ∇ × u is used to identify the vortical structures. it is seen that the solution decreases asymptotically.2. which are clearly visible at a Reynolds number of Re = 150. The alternating vortex shedding pattern leads to periodic force variations which can be visualised using CL CD limit cycles.19. the calculated timeaveraged Strouhal number (shedding frequency) is shown in table 2. 50k and 100k. The drag on the ﬁnest grid with the smallest timestep is about 2.16. These limit cycles are determined by taking the periodic part of the force histories as shown in ﬁgure 2. of which the frequency depends on the Reynolds number.63%. Results To illustrate the ﬂow behaviour.0.42 Finite volume discretisation Figure 2.0. obtained for diﬀerent mesh resolutions (25k.25.333. ﬁgure 2. 5. has the largest diﬀerence with literature. CL = L 1 2 2 ρUref . In addition. for example. From that table. From this table.15 shows the instantaneous vorticity (ω = ∇×u) contours.0. The diﬀerences of the mean Strouhal number with literature ranges from . 0. The results. (2.5 and 0. The ﬂow around a a a stationary cylinder shows a periodic vortex street.7% larger compared to the value obtained by Henderson (1995). 50k and 100k) are shown in ﬁgure 2. In practical applications such a vortex street exists behind struts in water.29) In order to investigate the temporal and spatial convergence of the solution. resulting drag coeﬃcient at Re = 150 is found to be CD = 1.
The periodicity of the ﬂow around a stationary circular cylinder at Re = 150 is illustrated by a CL CD limit cycle. which is considered suﬃciently small. The extrapo .0 25k 50k 100k Henderson (1995) 1. As can be seen. From table 2.6 1. is the extrapolated values.6 0.46 Grid Size=25k Grid Size=50k Grid Size=100k (a) Comax = 2.385 1. ﬁgure 2.369 (+2.2.4 1.64%.385 1. If the ﬂow solver is developed in a numerically consistent way.372 (+2.34 1.28% to 1. 3.42 1.370 1. 25k. 2002).0 (b) Comax = 1. (a).4 0.4 1.34 1. Comparison of the timeaveraged drag coeﬃcient for diﬀerent grids.25. and Comax = 1.2.5 and 0. 1. Drag coeﬃcient.2 0. 50k and 100k.1 Drag comparison for the ﬂow around a stationary cylinder.42 1.2 0 0. In order to illustrate if the solution converges. which should be the case.44 1.0.36 1. Besides a comparison solely with literature it is important to investigate the convergence of the solution with increasing grid resolution and decreasing timestep size.1 and 2.9%) 1.4 0.5%) 1.408 1. Additionally.46 Grid Size=25k Grid Size=50k Grid Size=100k 0.2 0 0. the solution decreases asymptotically.372 1.376 1. it is clear that the ﬂow solver produces results which are suﬃciently close to the values obtained from literature. and diﬀerent timesteps corresponding to a maximal Courant number.38 Drag coeﬃcient [] 1.392 1.38 Drag coeﬃcient [] 1.0 (b).377 Comax 0.4 0.2 0.6 1.10 Code validation and veriﬁcation 43 0.0 Figure 2. using Richardson’s extrapolation (Ferziger & Peric. the ﬂow solution should converge to an asymptotically value with increasing spatial and temporal resolution. The last value in these two ﬁgures. Comax = 2.373 1.7%) Table 2.4 0. 0. the grid convergence of the solution can be observed for Comax = 2.25 1. CD Mesh size 2.6 Lift coeﬃcient [] Lift coeﬃcient [] 0.36 1.393 1.0.0.5 0.44 1.333 Richardson 1.0 1.16 LiftDrag limit cycles of the ﬂow around a stationary cylinder.17 shows the timeaveraged drag coeﬃcient for increasing mesh resolution (for each timestep) and for decreasing timestep (for each mesh).381 1.379 (+3.381 1.
25 1.187 0.0 1.44 Finite volume discretisation 1.39 1.5 and 2.4 1. The Richardson extrapolated values are plotted at ∞. 0.188 0.188 (+2.187 (+2. 1.64%) 0.30) and should be 2 for a secondorder discretisation scheme. (b) shows the drag coeﬃcient with decreasing timestep size for the diﬀerent grid sizes. the value of p lies between 1.188 0. St Mesh size 2.37 1.0 0.30) where φextrap is the extrapolated value.37 Grid Size=25k Grid Size=50k Grid Size=100k 4 Spatial resolution ∞ 1. lated value is obtained using the following expression: φextrap = φf ine + φf ine − φcoarse . the order of the scheme p can be obtained using (2.2 Strouhal number comparison for the ﬂow around a stationary cylinder.2 it can be deduced that for both secondorder spatial and temporal schemes.0 25k 50k 100k Henderson (1995) 0. Strouhal number.28%) Table 2.36 1 2 4 8 Temporal resolution ∞ (a) (b) Figure 2.4 1. Comparison of the timeaveraged Strouhal number for diﬀerent grids. .39 1.186 (+1. However.38 1.25.73%) Comax 1.41 1.0. 25k. (a) shows the drag coeﬃcient with increasing grid reﬁnement level for diﬀerent timesteps.38 1. Theoretically.188 0.17 Timeaveraged drag coeﬃcient as a function of spatial and temporal resolution for the stationary cylinder.0 0. and diﬀerent timesteps corresponding to a maximal Courant number.188 (+2. 2002) which is not the case for the cylinder simulations. φf ine and φcoarse are the two most accurate solutions available.189 (+3.1 and 2. Comax = 2. 2p − 1 (2.73%) 0.188 0.41 Drag coeﬃcient [] 1.19%) 0.5 0.36 1 2 Drag coeﬃcient [] Comax Comax Comax Comax = = = = 2. 50k and 100k.189 (+3. A ﬂow solver is numerically consistent if the ﬂow solution converges with increasing grid resolution and decreasing timestep size. This is true for uniform Cartesian meshes (Ferziger & Peric.28%) 0. from table 2.188 0.5 0.5 and 0.25 0.183 0.0.
5 0 0. For all meshes.58 0.18 Timeaveraged lift convergence with grid resolution and timestep size for the stationary cylinder.0. 1.56 0. ﬁgure 2.7% to 3.0 1. are smaller than 1.2.25 0.5 Force coeﬃcient [] Drag 1 Lift 0. it can be determined that the diﬀerences varies from 2.5 and 0.62 0.0 0.5 0.54 1 2 4 Spatial resolution ∞ Comax Comax Comax Comax = = = = 2.0 to result in an accurate solution.0% for maximal Courant numbers Comax = 1.5 0. the grid size is 25k and the timestep was varied according to a maximal Courant number of Comax = 2.5% compared to literature with decreasing mesh resolution.58 0.64 0.25. Flow around a transversely oscillating circular cylinder The last validation case concerns the ﬂow around a transversely oscillating cylinder .1.54 1 2 4 8 Temporal resolution Grid Size=25k Grid Size=50k Grid Size=100k 45 ∞ (a) (b) Figure 2. 25k. it seems suﬃcient to consider the mesh of 50k and Comax = 1.62 0. (a) shows the average lift coeﬃcient amplitude with increasing grid reﬁnement level for diﬀerent timesteps. (b) shows the timeaveraged lift coeﬃcient amplitude with decreasing timestep for the diﬀerent grid sizes.64 Lift coeﬃcient amplitude [] Lift coeﬃcient amplitude [] 0.25. 0.0 1.0 0.25 Additionally to the temporal and spatial convergence of the drag coeﬃcient.0. 1.56 0. 0. Concerning the extrapolated values of the drag coeﬃcient.6 0.10 Code validation and veriﬁcation 0. Lift and drag coeﬃcients for a stationary circular cylinder case at Re = 150. these settings were used throughout the present research.5 and 0.6 0. Comax Comax Comax Comax = = = = 2. the diﬀerences in drag coeﬃcient compared to the extrapolated values.5 0 50 100 150 Time [s] Figure 2. from table 2.18 shows the convergence of the lift coeﬃcient amplitude in order to prove the consistency of the ﬂow solver. Therefore.0.19 Forces around a stationary cylinder.
Figure 2.20(b).0. the mesh quality should be high in terms of nonorthogonality and skewness. The value for all timesteps was within 2% compared to the extrapolated value. zerogradient (Neumann).2. (2. The frequency was set to fe = 0. 2. From this ﬁgure it is obvious that the ﬂow solution converges with decreasing timestep. 0. which is considered to be suﬃciently accurate. which employs amplitudes of several chord lengths. An amplitude of 0. both mesh quality measures. This statement is conﬁrmed if the timeaveraged drag coefﬁcient is plotted in ﬁgure 2. Although. The discretisation concerns arbitrary polyhedral meshes. four diﬀerent types of boundary conditions were speciﬁed. with D the cylinder diameter. Using those boundary conditions.8 times the natural shedding frequency of a stationary cylinder at a Reynolds number Re = 185. using the present ﬂow solver. performed by Guilmineau & Queutey (2002). Comax = 2. Results To assess the accuracy of the ﬂow solver applied to ﬂapping wings.20(a) shows the limit cycle results. the discretised NavierStokes equations can be solved using a PISO pressure velocity coupling in combination with an Arbitrary Lagrangian Eulerian (ALE) approach if dynamic meshes are used. for the 50k grid. ﬁxedvalue (Dirichlet). Guilmineau & Queutey (2002) found a drag coeﬃcient of CD = 1. . symmetry and movingwallvelocity. corresponding to 0. 1. such that this method can easily be applied to a wide variety of problems with complex geometries. the following values are used to show that the ﬂow solver converges to an asymptotic solution for an oscillating cylinder as well.2D. The oscillating direction is perpendicular to the freestream direction.11 Conclusions This chapter has presented the ﬁnite volume discretisation of the incompressible laminar NavierStokes equations. therefore that mesh is also used for this test case.5 and 0. The cylinder motion is deﬁned as y(t) = −Ae sin(2πfe t).25 were considered. with secondorder temporal and spatial discretisation.31) where the amplitude is set to Ae = 0. the results of an oscillating cylinder case were compared with literature. To obtain accurate and eﬃcient results.0 was found to be suﬃcient.46 Finite volume discretisation at Re = 185 using a numerical study.154. but suﬃcient to investigate the moving wing capabilities of the numerical model. The diﬀerent terms of the governing equation were discretised using secondorder schemes and diﬀerent ﬂux splitting methods were described concerning the face interpolation. a timestep corresponding to Comax = 1.0. Previously conducted simulations on stationary cylinder ﬂow showed that a mesh of 50k provides a suﬃciently accurate solution. In order to solve the ﬂow on a computational grid.2D is relatively small for insect aerodynamics.
Fluent has already been tested for various ﬂows in literature.2.3 1. Vortex decay and convection were used to study the inﬂuence of the face interpolation scheme with diﬀerent ﬂux limiters.0 0.1 1. stationary and transversely oscillating cylinder ﬂows were used to successfully prove spatial and temporal convergence. this chapter presented a validation of OpenFOAM for a number of relevant test cases.5 0.0 1.26 0.275 1. In principle.265 1. concerning vortex decay and convection. relevant for ﬂapping insect ﬂight.25 Drag coeﬃcient [] 1.20 Forces for the ﬂow around an oscillating cylinder. (a) shows the CD CL limit cycles for diﬀerent timesteps on a 50k mesh.11 Conclusions 0.15 0. OpenFOAM too. but not for low Reynolds ﬂows. The Richardson extrapolated values are plotted at ∞.27 47 0. It was found that the Van Leer ﬂux limiter provides the most accurate results. In addition.28 1.2 1.05 0. (b) shows the timeaveraged drag coeﬃcient with decreasing timestep for the 50k mesh. The present research used the commercial CFD solver Fluent and the opensource CFD code OpenFOAM . Therefore.2 0.25 Drag coeﬃcient [] 1.2 1.35 1. .05 0 Co Co Co Co = = = = 2.255 1 2 4 8 Temporal resolution ∞ (a) (b) Figure 2. the described discretisation method and solution procedure is applicable to diﬀerent commercial and noncommercial CFD solvers.15 Lift coeﬃcient [] 0. both were brieﬂy described.1 0. It is concluded that the opensource solver OpenFOAM provides an accurate and eﬃcient framework to investigate the ﬂow around ﬂapping wings at low Reynolds numbers.
.
a mesh deformation routine based on the interpolation of radial basis functions is introduced. two techniques are applied. In order to increase the eﬃciency of this method. . A test case. Mech. based on solving the Laplace and solid body rotation stress equations. based on boundary coarsening and smoothing of the radial basis function. Engrg. In order to use mesh deformation techniques to investigate ﬂapping wing aerodynamics it is necessary to maintain a high mesh quality for relevant wing kinematics. The main diﬃculty is to maintain high mesh quality when the wing exhibits large translations and rotations. compared to a function with compact support. Diﬀerent mesh deformation techniques are compared in order to identify their applicability for cases with ﬂapping wings. (January 2010). Additionally. it is shown that this method can be used for mesh deformation for threedimensional ﬂapping wings and can handle ﬂexing boundaries. The mesh quality. Appl. The radial basis function method can be used with diﬀerent basis functions with global or compact support. concerning a moving twodimensional block. is found to be highest when the thin plate spline is used as a basis function. is used to show that the radial basis function method provides superior mesh quality compared to the Laplace mesh motion solver. Meth. In addition to existing mesh motion methods. based on skewness and nonorthogonality. but is computationally more expensive.CHAPTER 3 Mesh deformation techniques for ﬂapping ﬂight Submitted to Comput. A globally supported basis function results in the highest average mesh quality.
The disadvantages of the immersed boundary method are the diﬃculty to capture the boundary layer and to meet the requirements for mass and momentum conservation. it is possible to solve this mathematical case by changing the boundary conditions as if the boundary was deforming or to deform the complete mesh. which deﬁnes a moving boundary on a stationary Cartesian background mesh. like the rigid body motion. 2002). In literature. using the discretisation methods from chapter 2. it is important that the internal mesh preserves its validity (no negative cell volumes) and quality (cell orthogonality and skewness). existing methods will be compared and an improved mesh motion solver is explored and incorporated. However. In order to assess the quality of a mesh motion solver. a computational mesh is necessary.1 Introduction In order to solve the governing equations. the equations are iteratively solved on the computational domain. For . Preservation of high mesh quality is necessary to solve the ﬂow in an accurate and eﬃcient way. e. The second method to deal with a deforming boundary. Therefore. If the shape of the domain boundary is timevarying. if this happens to be known beforehand. current mesh motion techniques are not fully suitable to cope with the mesh deformation around an object which moves with a large change in rotation. the one used in the current thesis. Examples of such interaction problems are ﬂuidstructureinteraction cases like blood ﬂow through arteries or deforming ﬂags. where the ﬂow is being inﬂuenced by a changing boundary shape. The quality of the resulting mesh is deﬁned by the nonorthogonality and skewness of the ﬁnite volume cells. Using prescribed initial and boundary conditions. This may be caused by imposed external eﬀects. In order to deal with moving objects. The ﬁrst method is called the immersed boundary method (Peskin. A robust mesh motion solver is deﬁned such that it needs little to no userinput. An appropriate implementation of this method is not trivial. Additionally. In engineering. there are numerous computational ﬂuid dynamics (CFD) problems in which the ﬂow solution involves geometrically deforming boundaries. the computational mesh points are moved in order to keep track of the changing location of boundary points. robustness is used to identify if a method is userfriendly. eﬃciency and robustness. is the use of a mesh motion solver which moves the internal mesh points. Eﬃciency is a measure of the used computation time to calculate the displacements of the mesh points at the new timestep. quality.50 Mesh deformation techniques for ﬂapping ﬂight 3. three diﬀerent aspects need to be formulated. Current CFD solvers incorporate diﬀerent mesh motion techniques in order to change the location of the internal mesh points according to the varying domain shape. several mesh deformation methods have been presented using different approaches to calculate the motion of the computational mesh points. or a prescribed body deformation. When using a mesh motion solver. a ﬂapping wing. A simpliﬁed type of interaction is that which concerns a oneway coupling. The boundaries of the computational domain can be either ﬁxed or deforming.g.
which assumes static equilibrium for small deformations of a linear elastic solid (the mesh is treated as if it was a solid). is called the spring analogy (Batina. which acts as a stiﬀness of the system of equations. Helenbrook. which signiﬁcantly improves the mesh quality for meshes subjected to large boundary translations and rotations. high resulting mesh quality was only achieved by specifying a problem speciﬁc spring stiﬀness. Another choice of equations is made by Johnson & Tezduyar (1994) who used the pseudosolid equation. for example Transﬁnite Interpolation (Wang & Przekwas. Depending on the method. Jasak et al. the parc allel implementation of these methods is fairly straightforward. Concerning the governing partial diﬀerential equations. The latter method is often used in the Arbitrary Lagrangian Eulerian formulation of ﬁnite element codes. Farhat et al.1 Introduction 51 structured meshes. Radial basis functions (RBF) are commonly used in literature to interpolate . possibly in combination with mesh reﬁnement. Therefore. Therefore. The most popular mesh deformation method. 1990) where the pointtopoint connection of every two neighbouring mesh points is represented by a linear spring. 2004).. (1998). 2007. used in the ﬁnite volume code of OpenFOAM (Weller et al. 2000).. However. Other mesh deformation techniques involve solving a partial diﬀerential equation on the complete ﬁeld of internal mesh displacements for given boundary point displacements. Using an additional mapping (Wang. the focus is put on mesh deformation techniques which can be applied to unstructured meshes containing arbitrary polyhedral cells. Degand & Farhat (2002) proposed a method to incorporate torsional springs to improve the robustness of this method. 2004. this method proved to lack robustness. Dwight (2004) modiﬁed this method to incorporate rigid body rotation. One major drawback of these methods is that they all fail in maintaining high mesh quality when the boundary points move with high rotation angles. In (Boer de et al. They interpolated the displacements of the boundary points along grid lines through the entire computational mesh to ﬁnd the displacements of all interior mesh points. Since.3. as was observed by (Blom. this inﬂuences the eﬃciency. applicable to both structured and unstructured meshes. based on the use of radial basis function (RBF) interpolation to obtain the mesh point displacements.. 2003) are often used o in combination with a constant or variable distancebased diﬀusion coeﬃcient to improve the mesh quality. especially on arbitrarily unstructured meshes. already available in existing CFD codes (Jasak & Tukovi´. Since these methods solve a partial diﬀerential equation on the complete ﬁeld of internal mesh points the existing iterative solvers can be used. These methods are perfectly suitable for structured but unsuitable for unstructured grids. 1998. the Laplace and biharmonic operators (L¨hner & Yang. 2000b) the mesh quality can be improved signiﬁcantly if the boundary is subjected to signiﬁcant rotation and deformation. a new mesh deformation method was developed and incorporated. Additionally. Jasak.. Bos et al. a variable diﬀusivity ﬁeld needs to be deﬁned. there are eﬃcient techniques available to deform the mesh. 2010a) it was shown that radial basis function interpolation could improve mesh quality considerably. 2009). unstructured meshes are used for complex geometries. 1994). 1996.
section 3. 2007. the idea was born to interpolate the boundary mesh to all computational mesh points.2 Diﬀerent mesh deformation techniques When a moving mesh problem is considered. described in chapter 2. radial basis functions were only applied to mesh motion concerning the boundaries in multiblock meshes (Potsdam & Guruswamy. A preliminary study was performed by Boer de et al. geophysics. error estimation. a distinction can be made between the motion of the boundary points and the motion of the internal (ﬂuid) points. they have been used in computer graphics. the shape of the computational domain is varying in time. Both Laplace and pseudosolid mesh motion techniques are commonly used within the OpenFOAM community. two techniques are implemented to improve its eﬃciency. or it is part of the solution. This is the subject of section 3. which is . two more relevant test cases were considered.b) have been carried out to improve the eﬃciency of mesh motion based on radial basis function interpolation. In addition to the simpliﬁed moving block. The displacement of the boundary points can be considered to be given.6. They noted that applying this method to all mesh points would be too computationally expensive. The application of radial basis functions is very wide. Since the application of RBF’s to interpolate from and to the boundary mesh was very accurate and eﬃcient.2. Rendall & Allen. Therefore. 2001). (2007). either it is externally deﬁned. Since mesh deformation using RBF interpolation results in high quality meshes even with large body rotation angles. but also in coupled simulations as in ﬂuidstructureinteraction. 2007) used radial basis function interpolation to couple two nonmatching meshes at the interface of a ﬂuidstructure interaction computation. a prescribed rigid body motion. based on the Laplace equation with variable diﬀusivity and a modiﬁed pseudosolid equation. In order to assess the mesh quality. An RBF interpolation function is used to transfer the known boundary point displacements to the ﬂuid boundary mesh. Since the radial basis function mesh motion method is computationally expensive.e. the conclusions are drawn in section 3. In this chapter two existing mesh deformation techniques are compared. 3. Only recent studies (Jakobsson & Amoignon. (Boer de et al.52 Mesh deformation techniques for ﬂapping ﬂight scattered data.4. section 3. The resulting mesh quality of the diﬀerent mesh motion methods is studied using a twodimensional test case of a block which translates and rotates. because of their good approximation properties which is discussed by Buhmann (2000). Finally. Previously. i. The mesh quality is investigated using a visualisation and histograms of the skewness and nonorthogonality criterion.3 discusses two diﬀerent criteria. with radial basis function mesh motion.5 deals with two techniques to increase its eﬃciency. one using a threedimensional ﬂapping wing and the other involves a twodimensional ﬂexing airfoil.. 2008c. based on nonorthogonality and skewness. These diﬀerent mesh motion methods are described in section 3.
2002). The internal point motion inﬂuences the solution only through the discretisation errors (Ferziger & Peric. one c could also have used an exponentially decreasing diﬀusion coeﬃcient or a diﬀusion coeﬃcient related to the mesh deformation energy. This variable diﬀusion coeﬃcient can be chosen such that a region next to the deforming or moving boundary closely moves with the boundary. other type of equations were used.2. since it decreases the robustness of the method. To maintain robustness. m = 2. 3.1) The resulting mesh quality strongly depends on the chosen γ(r) function. which depends on the distance from the moving boundary. The internal point motion can be calculated using diﬀerent methods. . decreasing diﬀusion coeﬃcient. However. provided that the ALE formulation is correctly implemented. rm (3. appears to be very problem dependent and thus optimisation of γ(r) seems not cost eﬀective. The current research uses a γ(r) function like equation (3. the internal points need to be moved in order to maintain mesh quality and validity. which is a variant of the linear stress equation (Dwight. as will be shown in the next section. When the mesh motion is governed by the Laplace equation. it is also possible to deﬁne γ(r) for every internal mesh cell for all timesteps independently. 2004).3. Ideally. such that existing iterative solvers can be used eﬃciently. the given boundary point motion may be arbitrary and nonuniform. This. a user input is not desired. The resulting mesh contains less cell quality deterioration next to the boundary. In addition to the freedom of choosing a diﬀusion function. 2004). namely the Laplace equation and the solid body rotation stress (SBR Stress) equation. where x is the displacement ﬁeld and γ the diﬀusion coeﬃcient. in the current work we use a quadratically. however. The nature of the Laplace equation is that the point displacements will be largest close to the moving boundary and small at large distance.2 Diﬀerent mesh deformation techniques 53 the case in ﬂuidstructure interaction problems.1 Laplace equation with variable diﬀusivity One can think of a deforming computational domain as if it was a solid body undergoing internal stresses given by the PiolaKirchhoﬀ stressstrain equation (Baruh. Additionally. which decreases with the radius r from the deforming boundary as follows: γ(r) = 1 .1). This leads to the following deﬁnition of the Laplace equation: ∇ • (γ∇x) = 0. Therefore. According to the given boundary point motion. 1999). A mesh motion method based on one of these equations is computationally cheap since the resulting matrix system is sparse. this method needs the speciﬁcation of a variable diﬀusivity. That equation is nonlinear and thus expensive to solve using existing numerical techniques. which was found to provide eﬃcient and a smooth mesh motion (Jasak & Tukovi´.
The constants can be related to Young’s modulus. which is treated as if it was a linear solid. E. (3. the following solid body rotation stress equation is obtained: ∇ • (γ∇x) + ∇(γ(∇x − ∇xT )) − λ tr(∇x) = 0. In order to deal with large rotations.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. meaning the material contraction ratio as it stretches. which are a e property of the elastic material.54 Mesh deformation techniques for ﬂapping ﬂight 3. 1994) usually applied to the points of multiblocks.4) ǫ = (∇x + ∇xT ). may be written as ∇ • σ = f. In Dwight (2004) an extra term was added to obtain the following strain relation: ǫ= 1 (∇x + ∇xT + ∇xT · ∇x). µ= . a newly implemented mesh motion solver is based on radial basis function interpolation. such that it can be used for ﬂapping wing simulations.4) does not allow for rotations. it is also possible to explicitly deﬁne the point motion using interpolation techniques. Although equation (3. The equation of linear elasticity. where x is the position of an internal mesh point. 2 (3. (3. 1999). 2004). there is nothing against changing this strain equation such that rigid body rotations are allowed.3) where tr is the trace and λ and µ are Lam´ constants (Baruh.4 it is shown that both previously described methods maintain high mesh quality for problems with limited boundary rotation. which is given by the following constitutive relation: σ = λ tr (ǫ)I + 2µǫ. The stress tensor σ is given in terms of the strain.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.5) Combining equations (3.2).5). such that standard iterative techniques can be used. 2 deﬁnes the relative change in length. However. like the transﬁnite interpolation (Wang & Przekwas. In section 3.6) where γ is a similar diﬀusion coeﬃcient as in equation (3.2) where σ is the stress tensor and f the acting force vector.1). Equation (3. as E νE . Solving the Laplace or the SBR Stress equation leads to a sparse system of equations. λ= (1 + ν)(1 − 2ν) 2(1 + ν) where ν is Poisson’s ratio. like the preconditioned Conjugate Gradient (PCG) method. . valid for small displacements. together with λ = −E and µ = E. (3.3) and (3. The following equation: 1 (3. (3.
which is diﬃcult to solve using standard iterative techniques. The values for the coeﬃcients γj and the linear polynomial can be obtained by solving the system: Φbb Qb γ ∆xb = (3.3 Radial basis function interpolation In the current work we use radial basis function interpolation to ﬁnd the displacements of the internal ﬂuid points for given boundary displacements.8) leads to a dense matrix system. j=1 which holds for all polynomials p with a degree less or equal than that of polynomial q. The interpolation function s(x) describing the displacement of all computational mesh points. a linear polynomial can be used (Beckert & Wendland.2. Here ∆xbj contains the known discrete values of the boundary point displacements. The possibilities of solving the system in a more eﬃcient way are discussed in section 3. Together with the additional requirements: Nb γj p(xbj ) = 0.. the γj values can be determined (Boer de et al. The polynomial q is deﬁned by the coeﬃcients γj which can be deﬁned by evaluating the interpolation function s(x) in the known boundary points: s(xbj ) = ∆xbj . Nb is the number of boundary points and φ is a given basis function as a function of the Euclidean distance x. 2007). If the basis functions are conditionally positive deﬁnite of order m ≤ 2.2 Diﬀerent mesh deformation techniques 55 3. is approximated by a sum of basis functions: Nb s(x) = j=1 γj φ(x − xbj ) + q(x). 2007).5.8) . The minimal degree of polynomial q depends on the choice of the basis function φ (Boer de et al. In general. A unique interpolant is given if the basis function is a conditionally positive deﬁnite function.3. Φbb an nb × nb matrix contains the evaluation of the basis function φbi bj = φ(xbi − xbj ) and can be seen as a connectivity matrix connecting all boundary points with all internal ﬂuid points. 2001). . (3. by doing a LU decomposition. ybj .. it needs to be solved directly. zbj ]. (3. We only applied basis functions that satisfy this criterion.7) where the known boundary value displacements are given by xbj = [xbj . Therefore. β the four coeﬃcients of the linear polynomial q. Qb is an (nb × (d + 1)) matrix with row j given by [ 1 xbj ]. A consequence of using a linear polynomial is that rigid body translations are exactly recovered. 0 β QT 0 b where γ is containing all coeﬃcients γj . q is a polynomial.
Instead. the proper radial basis function needs to be chosen to satisfy the need for robustness. no partial diﬀerential equation needs to be solved and the evaluation of all internal boundary points is straightforward to implement in parallel. Radial basis functions with compact support From literature. such that no mesh connectivity is necessary. In contrast to the Laplace and SBR Stress methods. Functions with compact support have the following property: φ(x/r) = f (x/r) 0 0 ≤ x ≤ r. a very large support radius leads to a dense matrix system. 4. which are then used for step two. 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. various radial basis functions are available. (3. it must be noted that larger values for the support radius lead to more accurate mesh motion. which is considerably smaller than other techniques using the mesh connectivity. only the internal mesh points inside a circle (twodimensional problem) or a sphere (threedimensional problem) with radius r around a centre xj are inﬂuenced by the movement of the boundary points. 6}. Two types of radial basis functions can be distinguished: functions with compact and functions with global support. with Nin the total number of mesh points. where f (x/r) ≥ 0 is scaled with a support radius r. since no mesh connectivity is needed. which are suitable for data interpolation.8) gives the values of the necessary coeﬃcients γ and β. When choosing r. 1996). When a support radius is used.1 various radial basis functions with compact support are shown using the scaled variable ξ = x/r. x > r.9) is transferred to the mesh motion solver to update all internal points accordingly. the evaluation using equation (3.8) is ((Nb + 4) x (Nb + 4)). Solving the system (3. 2. while a low support radius results in a sparse system which can be solved eﬃciently using common iterative techniques. Every internal mesh point is moved based on its calculated displacement. The result of (3.7). The last four are a series of functions based on the thin plate spline .9) ∆xinj = s(xinj ). which is a dimension higher than the total number of boundary points.9). In table 3. this method is not using a variable diﬀusion coeﬃcient which has to be tuned by the user.56 Mesh deformation techniques for ﬂapping ﬂight When the coeﬃcients in γ and β are obtained they are used to calculate the values for the displacements of all internal ﬂuid points ∆xinj using the evaluation function (3. On the other hand. The size of the system of this method (3. This interpolation function is equal to the displacement of the moving boundary or zero at the outer boundaries. Concerning robustness. The ﬁrst four are based on polynomials (Wendland. The mesh connectivity techniques encounter systems of the order (Nin x Nin ).
. 2007). the computational domain. i. .e. There are two possible CTPS C 2 continuous functions which are distinguished by subscript a and b. TPS MQB IMQB QB IQB Gauss f (x) x2 √ log(x) a2 + x2 1+x e 1 1+x2 −x2 1 a2 +x2 2 Table 3. which creates C n continuous basis functions with n ∈ {0. Table 3. nr. 1 2 3 4 5 6 7 8 RBF Name CP C 0 CP C 2 CP C 4 CP C 6 CTPS C 0 CTPS C 1 2 CTPS Ca 2 CTPS Cb f (ξ) (1 − ξ)2 (1 − ξ)4 (4ξ + 1) (1 − ξ)6 ( 35 ξ 2 + 6ξ + 1) 3 (1 − ξ)8 (32ξ 3 + 25ξ 2 + 8ξ + 1) (1 − ξ)5 8 80 2 3 1 + 3 ξ − 40ξ + 15ξ 4 − 3 ξ 5 + 20ξ 2 log(ξ) 2 3 4 1 − 30ξ − 10ξ + 45ξ − 6ξ 5 − 60ξ 3 log(ξ) 1 − 20ξ 2 + 80ξ 3 − 45ξ 4 − 16ξ 5 + 60ξ 4 log(ξ) Table 3.g. functions with global support are not equal to zero outside a certain radius. as will be discussed in section 3. Boer de et al. Taken from Wendland (1996). 1996).5. which can be improved by multiplication with a smoothing function. Radial basis functions with global support cover the whole interpolation space.2 Radial basis functions with global support. 9 10 11 12 13 14 RBF Name Thin plate spline Multiquadratic Biharmonics Inverse Multiquadratic Biharmonics Quadric Biharmonics Inverse Quadric Biharmonics Gaussian Abbrev. 2} (Wendland. neural networks. 1.. Note that ξ = x/r. Ref. nr. computer graphics (Carr et al.. Radial basis functions with compact support are nonzero within the range of the support radius r.2 shows six radial basis functions with global support which are commonly used in e. Radial basis functions with global support In contrast to the functions with compact support.3. 2003) and for data transfer in ﬂuidstructure interaction computations (Smith et al.2 Diﬀerent mesh deformation techniques 57 Ref.1 Radial basis functions with compact support. but cover the whole interpolation space. 2000. Radial basis functions with global support generally lead to dense matrix systems.
4 the skewness and nonorthogonality deﬁnition were introduced. It is important that the ideal mesh motion solver maintains high quality in terms of skewness and nonorthogonality after mesh deformation. The skewness and nonorthogonality are written to scalar ﬁelds such that they can be used for postprocessing. 1998). (2007) compared the resulting mesh quality using all radial basis functions from table 3. Therefore. 2003). the absolute and the relative implementation. These mesh quality measures are based on the cell properties such as size. resulting in lower computation costs (see section 3. . leading to optimal initial mesh quality.58 Mesh deformation techniques for ﬂapping ﬂight The MQB and IMQB methods use a shape parameter a. 1999. This method is very eﬃcient since the direct matrix solve.3 Mesh quality measures In section 2. When using the relative implementation it is important to use diﬀerent techniques to decrease the number of boundary points. this is shown in section 3. like a 180◦ rotation. Wendland.In this method the inverse is calculated at every timestep and the motion is deﬁned with respect to the previous timestep.1 and 3. Absolute and relative radial basis function interpolation In principle there are two diﬀerent ways to implement this RBF mesh motion method. the functions based on a continuous thin 2 2 plate spline. the mesh quality measures. the relative method is used when very large boundary displacements occur. contain a Cartesian grid around a square box. skewness and nonorthogonality. shape and skewness. The coeﬃcient arrays γ and β are calculated and used to calculate the internal point displacements at all timesteps. Ca and Cb and ﬁnally the globally supported thin plate spline. is only performed initially. More information about RBF’s and their error and convergence properties can be found in (Buhmann. orientation. A large value of a gives a ﬂat sheetlike function. which controls the shape of the radial basis function. The value of a is typically chosen in the range 10−5 − 10−3 . The best results were obtained using the continuous polynomial C 2 . which is more expensive than the evaluation. C 1 . 2000. the mesh quality is limited since the coeﬃcients are not deﬁned with respect to the previous timestep. 3.4 these six radial basis functions are applied to our test problem and the best one is used throughout the current thesis. On the other hand.5). it is much cheaper to use the absolute method. First. The absolute method performs a direct solve of the system (3. In order to compare the quality of the diﬀerent meshes after mesh motion it is important to elaborate on how to interpret those two mesh quality measures (Knupp. In section 3. are discussed to compare the mesh quality for the diﬀerent mesh motion solvers. Boer de et al. whereas a small value of a gives a narrow conelike function.2. The test cases used to compare the mesh quality of the diﬀerent mesh motion solvers.4. For reasonably small boundary displacements.8) only once at the beginning of the simulation.
the more stable. the eﬀect of diﬀerent radial basis functions is discussed. In both cases.4. On the other hand. 3.4. In addition. three diﬀerent kinds of mesh motion solvers are described. In the next section. the block is rotated around its translating centre with 57. The outer boundary points are kept ﬁxed. The domain size of the test problem is limited to 25D x 25D and the size of the moving block is 5D x 1D. solving the solid body rotation (SBR Stress) equation and based on interpolation using radial basis functions (RBF).2 deals with the mesh motion around a threedimensional ﬂapping wing. The ﬁrst simple case is a twodimensional block which performs a combination of translation and rotation.1. followed by an example of a twodimensional ﬂexible moving boundary in section 3. the simulation will diverge within a couple of iterations.0 rad). 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.3. 3.2. Therefore. both the average and minimal value of the skewness and nonorthogonality mesh quality measures are used to compare the mesh motion solvers. The higher the average quality of the mesh. After this simple model problem. The initial computational mesh is shown in ﬁgure 3. a lower value means a higher mesh quality. and that the mesh nonorthogonality. The resulting mesh .4.5D in both Xand Y direction. should be within 0◦ and 90◦ .0 ≤ 90◦ (3.0 0◦ ≤ fskewness ≤ fnon−ortho ≤ 1.3. based on solving the Laplace equation. section 3. This section introduces with three numerical test cases which were performed to investigate the diﬀerences in mesh quality obtained with the diﬀerent mesh motion solvers.10) When assessing the mesh quality it is important to analyse the maximal and average values. accurate and eﬃcient the computation will be.4 it is shown that mesh skewness should be within 0 and 1. the grid spacing corresponds to 1D in order to obtain a Cartesian grid as can be seen in the ﬁgure.4 Comparison of mesh motion solvers In section 3. The twodimensional block is initially centred and translates 2.1 Translation and rotation of a twodimensional block The ﬁrst test case considers a combined motion of translation and rotation to compare the mesh motion solvers under these conditions. Additionally. the average value of the mesh quality measure will provide an indication of the average quality of the mesh.3◦ (1.4 Comparison of mesh motion solvers 59 In section 2. Mesh motion simulations are performed using the diﬀerent mesh motion solvers and a variation of the radial basis function. such that the inﬂuence of the moving boundary points on all internal points could be studied independently. which is an angle. desirable mesh quality bounds are: 0.
which leads to the motion of all internal mesh points. decreasing from the moving boundary. the newly implemented RBF mesh motion solver is used with ﬁve 2 2 diﬀerent functions. The domain size is (25D x 25D) around a block with size (5D x 1D). applicable to this test case. it is best to apply a mesh motion solver. quality. As can be seen in ﬁgure 3.60 Mesh deformation techniques for ﬂapping ﬂight Figure 3. From . In cases where large rotation angles occur. Before proceeding to the results. Using those ﬁelds. the maximal and average values are obtained as well as a complete visualisation of those mesh quality measures. with these standard methods the mesh quality near the moving boundary is low. coping with the boundary deformation. need to be described. Finally.e. a quadratically decreasing diﬀusivity coefc ﬁcient. the mesh deformation is largest near the boundary. after the combined translation and rotation. (2007). especially near the leading and trailing edges. CP C 2 .2(a). the boundary conditions need to be set for the motion solver. Figure 3. Following Jasak & Tukovi´ (2004) and Jasak (2009). at every unit spacing. In both cases. The mesh motion method which solves the Laplace equation with a quadratically decreasing diﬀusion coeﬃcient. i. CTPS Cb and TPS. so the outer boundary points are kept ﬁxed.1 Initial mesh around a moving block. CTPS Ca . based on the assessment performed by Boer de et al. The initial hexahedral mesh with optimal quality around a moving block. a grid point is places such that an optimal hexahedral mesh is obtained. combined with mesh quality histograms. some special settings. is assessed using the ﬁelds of skewness and nonorthogonality. As seen in the ﬁgure. which is not desirable. is simply not robust enough to obtain high mesh quality when the boundary rotates. CTPS C 1 . Secondly. First. when using the Laplace and SBR Stress mesh motion solver it is necessary to specify the diﬀusivity coeﬃcient which decreases proportionally to the distance from the moving boundary points. was chosen. The gain in mesh quality by using the SBR Stress method is marginal.2 shows the cell nonorthogonality at maximal mesh deformation for the Laplace and the SBR Stress motion solvers. the boundary conditions on all outer boundary points are set to Dirichlet type with value 0. Laplace and SBR Stress motion solvers.
The cell nonorthogonality is compared of the Laplace (a) and the SBR Stress (b) motion solvers.2(a) and 3.5D in both Xand Y direction and rotate 57. The rotation of the boundary is very large. since the maximal skewness and maximal nonorthogonality are large.2(b) it can be seen that the cells with a high nonorthogonality occur at a distance from the moving boundary. ﬁgure 3. with the RBF mesh motion. occurs in the outer regions of the mesh and all cells are dealing with the boundary displacement. Still. The results of the Laplace and SBR Stress mesh motion solver are very similar. The boundary points translate over a distance of 2. such that it moves according to the body motion. Table 3. ﬁgure 3. including diﬀerent RBF’s. CTPS Ca and CTPS Cb ) and one with global support.3(b) shows the resulting mesh. four RBF’s with compact 2 2 support are used (CP C 2 . the mesh quality near the body surface. This method is very robust but computationally expensive as will be shown in section 3. CTPS C 1 . respectively fsmax = 0. within 6. and the mesh remains valid.1.4 Comparison of mesh motion solvers 61 (a) Laplace (b) SBR Stress Figure 3. which is obtained using the relative implementation of the RBF mesh motion method. Figure 3. In order to neglect the eﬀect of the support .1%. The nonorthogonality is visualised for a twodimensional block with a combined motion of translation and rotation.3(a) shows the resulted mesh obtained using RBF interpolation using a thin plate spline function.3 shows a quantitative comparison of the resulting mesh quality obtained with diﬀerent mesh motion solvers. When comparing the cell nonorthogonality with ﬁgure 3. needs improvement.2(b) it may be seen that most of the cell deformation.95 and fomax = 72.2 Cell nonorthogonality of Laplace and SBR Stress mesh motion solvers. Improvement can be obtained by specifying a constraint to an inner mesh region. the thin plate spline TPS. The maximal and averaged values of both skewness and nonorthogonality are compared.5.0 rad) around the translating centre.3. 180◦ . especially near the leading and trailing edges. Additionally. Concerning the RBF method.3◦ (1. The mesh quality calculated by the Laplace method is low. lower (blue) is better.
the basis function thin plate spline provides the highest mesh quality for both skewness and nonorthogonality. it is set to r = 75 which is about 3 times the domain size such that the results in table 3.0 rad) for (a) and 180◦ (3. in addition to the nonorthogonality visualisations.3 are independent of the support radius for r > 75. The nonorthogonality is visualised for a twodimensional block with a combined motion of translation and rotation. Overall. The rotation is 57. The C 2 and TPS RBF’s result in an maximal skewness of respectively −32% and −45% compared to the Laplace motion solver. the RBF mesh motion results in a smooth proﬁle. respectively.4 and 3.3 Cell nonorthogonality of relative and absolute RBF mesh motion solvers. Concerning the average orthogonality. Especially. the globally sup . while the diﬀerence of the other RBF’s is only about 10%. which emphasises the fact that all internal cells are coping with the mesh motion.14 rad) for (b).62 Mesh deformation techniques for ﬂapping ﬂight (a) RBF absolute (b) RBF relative Figure 3.3◦ (1. maximal and averaged.5D in both X. which are considered to result in the worst and best mesh quality. The cell nonorthogonality is shown for the relative and absolute versions of the RBF mesh motion solver. lower (blue) is better. the C 2 RBF is outperformed by the TPS RBF. which is the only mesh motion solver resulting in a lower value compared to the Laplace method. The boundary points translate a distance of 2. 3. For both nonorthogonality and skewness.4.5 show the histograms of cell nonorthogonality and skewness for the Laplace and RBF mesh motion solver. The RBF method provides high mesh quality. radius.2 Flapping of a threedimensional wing It was shown that high mesh quality was obtained for a simpliﬁed twodimensional test case. by using radial basis function interpolation. ﬁgure 3. (a) shows the absolute and (b) the relative implementation. for both C 2 and TPS compared to the other functions. Finally.and Y direction. Similar results are shown in the table for the average skewness and maximal orthogonality. such that this globally supported function was used for the current investigations.
3.11 0.5 (17%) (+3%) (+2%) (+5%) (27%) (fo )ave 20.4 Comparison of mesh motion solvers 63 Method Laplace SBR Stress CP C 2 CTPS C 1 2 CTPS Ca 2 CTPS Cb TPS (fs )max 0.103 0.08 (fs )ave () (4%) (28%) (+17%) (+14%) (+22%) (41%) (fo )max 72.5.86 0.4 and 3. while ﬁgure 3.4 31.0 (+16%) (+59%) (+56%) (+63%) (6%) 0. Results are shown for the Laplace. respec . since it has global support.6(b) presents the mesh quality halfway of the upstroke.065 0.6 at midstroke.88 0.0 52. SBR Stress and RBF mesh motion solver.81 0. the latter using diﬀerent RBF’s. Figure 3.52 (32%) (10%) (15%) (7%) (45%) 0. This was already illustrated for the moving twodimensional block in ﬁgures 3.96 (+0. without any user input.95 (baseline) 0.1 () 21. the mesh around a ﬂapping wing is shown in ﬁgure 3.5 73.051 Table 3.3 Comparison of mesh quality for diﬀerent mesh motion solvers. The mean and maximal values of the skewness fs and nonorthogonality fo are compared at maximal displacement and rotation of the twodimensional rectangular block.4 32.9 31.1 () 75. ported thin plate spline (TPS) provided high quality and robust mesh deformation.65 0. 90 80 70 Number of cells 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 Nonorthogonality 80 90 Number of cells 90 80 70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 Nonorthogonality 80 90 (a) Laplace (b) RBF Figure 3.4 Cell nonorthogonality histograms for Laplace and RBF mesh motion solvers.1 (+4%) 59. Concerning the RBF mesh motion solver. To show the threedimensional capabilities of the RBF mesh motion solver.7%) 0.6 74.8 19. showing the histograms of the nonorthogonality and skewness.8 76. The TPS was used as radial basis function. Mesh quality histograms show the variation in nonorthogonality for Laplace (a) and RBF mesh motion solvers (b) using the TPS.09 0.6(a) shows the nonorthogonality during the downstroke. the cells close to the wing take a larger part of the deformation compared to the Laplace method.3 (+6%) 23. From the ﬁgure it is clear that a large part of the near wake is deformed in order to deal with the threedimensional wing motion.105 0.
The .9 Skewness 350 300 250 200 150 100 50 1 0 0 0. the RBF mesh motion solver is tested by employing the ﬂexing of a twodimensional moving boundary.5 Cell skewness histograms for Laplace and RBF mesh motion solvers.4 0. tively. the eﬀects of wing ﬂexibility was incorporated by deﬁning the wing ﬂexing using harmonic functions.6(b) it is seen that the mesh during both the upstroke and downstroke is symmetric.. performing a ﬂuidstructureinteraction simulation is too computationally demanding for a full threedimensional ﬂapping wing.1 0.9 Skewness 1 (a) Laplace (b) RBF Figure 3. This is caused by the fact that the radial basis function interpolation determines the new internal mesh points with respect to the initial mesh. The motion of the twodimensional block can be decomposed into translation. These techniques are described in section 3. the model problem of a moving block is used.6 0. When comparing ﬁgure 3. to mimic realistic insect wing deformation (Shyy et al. But ﬁrst.8 0.7 0.2. As discussed in subsection 3.3 0.6 0.3 Flexing of a twodimensional block Within the context of the present research.2 0. 2008a). such that the initial mesh is recovered after every ﬂapping period. Therefore. the mesh deformation for threedimensional cases may become very expensive.4 0. Therefore. such that it is necessary to implement techniques to improve its eﬃciency.3.2 0.6(a) and 3.8 0.1 0. 3.7 0.5 0. rotation and ﬂexing.4. Mesh quality histograms show the variation in skewness for Laplace (a) and RBF mesh motion solvers (b) using the TPS. while the RBF mesh motion technique leads a smoother decline of both quality measures in the histograms.3 0. all deﬁned with respect to the initial conﬁguration.5. The Laplace method results in a small number of cells with large values for both nonorthogonality and skewness. In order to show that the RBF mesh motion method is able to deal with a ﬂexing boundary. the radial basis function mesh motion contains a direct system solve and an evaluation to determine the displacement of all internal mesh points.64 500 450 400 Number of cells 350 300 250 200 150 100 50 0 0 Mesh deformation techniques for ﬂapping ﬂight 500 450 400 Number of cells 0.5 0.
25T (b) t = 0.0. 0. A larger domain will undoubtedly lead to high quality meshes when using RBF mesh motion. 2 · cf α = Aα · sin(2πf t). lower (blue) is better.5 which is about 10% of the ﬂexible boundary length. which is mainly caused by the ﬁxed points on the outer boundary and the small computational domain.75T . df = (0. (a) shows the mesh quality at t = 0.0) with a ﬂexing amplitude of Af = 0.0.25T and (b) at t = 0. like was the case with a rigid airfoil. where Af is the ﬂexing amplitude vector. 2.7 shows the resulting mesh deformation at t = 0. f the frequency and t the time. respectively during the downstroke and upstroke. where. 0.0.0. At and Aα represent the translation and rotation amplitude vectors. The ﬂexing was deﬁned such that the main ﬂexing direction is perpendicular to the ﬂexible boundary surface. the amplitudes of both translation and rotation were ﬁxed to At = (2.0) and Aα = (0.0).25T and t = 0.3. dc is the direction vector inplane of the ﬂexing surface.75T Figure 3. cf the length of the ﬂexing surface and df represents the direction vector of the ﬂexing.0. some high nonorthogonality can be observed within a region of 1 − 2 block lengths. Still. translation and rotation are deﬁned by: xt = At · sin(2πf t).5. . 0. Figure 3. respectively. 1. 1. obtained with the RBF mesh motion solver with the globally supported thin plate spline (TPS). The ﬂexing of the boundary is deﬁned by combining two harmonic functions like: xf = Af · cos(2π x • dc ) · sin(2πf t) • df . It can be seen that the whole mesh is deformed by the RBF mesh motion solver. Cell nonorthogonality of a mesh around a threedimensional ﬂapping model wing.4 Comparison of mesh motion solvers 65 (a) t = 0.6 Cell nonorthogonality for a threedimensional wing. In this model problem of a moving twodimensional block.75T where T = 1/f is the motion period. cf = 5.5.
(a) shows the mesh deformation at t = 0. Table 3. equation (3. 2. the total number of operations for a direct solve and the RBF evaluation is given in the table.and threedimensional meshes both the system solve and the evaluation procedures may become very computationally expensive. 3.66 Mesh deformation techniques for ﬂapping ﬂight (a) (b) Figure 3.4 shows the total number of internal and boundary points as a function of Nb . especially when direct methods are used. equal to Nb . A ﬂexible rectangular block which is translating.5 Improving computational eﬃciency When using direct methods. The computational domain is square shaped with an equal number of points on all edges of the boundary.25T and (b) at t = 0. For large two. using the given boundary points xbi and the in step 1 computed coeﬃcient vectors γ and β. standard direct solvers require O(Nb ) operations.8) for given boundary points xbi and corresponding displacements ∆xbi to ﬁnd the coeﬃcient vectors γ and β. rotating and ﬂexing using the RBF mesh motion solver with the globally supported thin plate spline (TPS). Therefore. for both the two. In order to illustrate the increasing costs with increasing number of mesh points. Additionally.7 Deforming mesh around a ﬂexible block using RBF mesh motion. This evaluation leads to a computational cost of order O(Nb Nin ).and threedimensional Cartesian grid is considered with a uniform mesh distribution. a two.and threedimensional example. The ﬂexing is deﬁned by simple harmonic functions. The upper diagonal block matrix Φbb is in general a dense symmetric matrix of 3 size (Nb x Nb ).75T . which . The method consists of two computationally expensive steps: 1. at all Nin (of order O(Nin )) internal mesh points. Solving the system of equations (3.7). Evaluation of the radial basis function summation. increases fast with an increased number of boundary and internal points. the computational costs of the new mesh motion solver based on radial basis function interpolation.
special treatment is necessary concerning the system solve. This can be achieved in two diﬀerent ways.12) Router − Rinner xi represents the coordinate of the i−th inner mesh point. x= From table 3. based on greedy algorithms are applied by Rendall & Allen (2008a). It seems reasonable to reduce the number of moving boundary points by performing a sound selection procedure. the outer boundary points can be neglected. which is deﬁned (Jakobsson & Amoignon. (3. Concerning the threedimensional case. More advanced coarsening techniques. 3. step 1. 0 ≤ x ≤ 1. So when a complex threedimensional case. like a ﬂapping wing. 2007) as x ≤ 0. by reducing the number of boundary nodes or using advanced direct solver techniques. Rinner and Router are two radii. the eﬃciency of the RBF method is improved.1 Boundary point coarsening and smoothing xi − Rinner  . 1. shows that the computational costs for both a direct solve and evaluation scales 3 with Nb for the twodimensional case. Computational costs for solving the system and evaluation the RBF’s on all internal mesh points. x is given by: . to evaluate this smoothing function at that particular location in space. by the notion that all outer boundary points are ﬁxed in general.11) ψ(x) = 0. is considered.3. Therefore. where. 1 − x2 (3 − 2x). Especially when the body displacement follows a rigid body motion.g. not all boundary points are necessary.5 Improving computational eﬃciency 67 Internal points 2D 3D 2 Nb 3 Nb All boundary points 4Nb 2 6Nb Direct solve 3 64Nb 6 216Nb RBF evaluation 3 4Nb 5 6Nb Table 3. Therefore. which is a factor Nb more expensive than the evaluation. This is achieved by specifying a smoothing function such that the RBF contribution reduces to zero at the outer boundary. where ξ is the coarsening factor.and ydirection. First. a coarsening technique was incorporated which selects a boundary point for every ξ points.4 Computational costs for solving the system and evaluate the RBF’s. (3. a major part of the computational cost is spent at solving the system.4 it is seen that the total computation costs will decrease if a constraint is put on the number of mesh points. Illustration of a twodimensional and threedimensional uniform square shaped Cartesian mesh with an equal number of boundary nodes Nb in x. the computational costs for the direct system solve are a factor Nb larger compared to the costs for the RBF evaluation. Secondly.5. e. x ≥ 1.
Concerning the RBF mesh motion solvers. per timestep. Additionally. the mesh resolution is increased.8(a) shows the computing times with increasing grid resolution for the Laplace. the system to be solved. Figure 3. the three described variants are used. If the absolute and relative RBF methods are compared. It is clear that the mesh motion solvers based on solving a partial diﬀerential equation are very fast. Figure 3. the number of boundary points used in the system solving is still growing nonlinearly. but outside Router the value of the RBF becomes zero. 190 hours. since standard iterative techniques can be used for these sparse systems. only contains the control points on the moving boundary. if the coarsening and smoothing techniques are applied the computing times are of similar order compared with the fast Laplace mesh motion. These threedimensional simulations are performed for grid sizes of 100k. such that all ﬁxed outer boundary points may be neglected. Figure 3.8(b) shows superlinear curves for both solving the ﬂow equations and the RBF mesh motion. 103. selected by the coarsening function. the absolute implementation.68 Mesh deformation techniques for ﬂapping ﬂight between those radii this smoothing function is decreasing from 1. Therefore. it is interesting to address the computing times concerning the diﬀerent mesh motion solvers.0 to 0.0. concerning a threedimensional ﬂapping wing simulation (6 ﬂapping periods). 200k.9) is multiplied by φ(x) the corrected RBF evaluation is obtained by: ∆xinj = s(xinj ) • ψ(x). relative implementation and the relative method in combination with the previously dealt coarsening and smoothing techniques. 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. ﬁgure 3. This is considered to be very eﬃcient. While keeping the coarsening ratio ﬁxed.13) Within Rinner the contribution of the RBF evaluation remains unaltered. 400k. it must be noted that mesh coarsening and a smoothing function are applied for acceleration. Validation showed that a mesh resolution of 800k provided accurate results for ﬂapping wing aerodynamics. All simulations were performed on four CPU cores of an AMD Opteron 280 cluster. When the RBF evaluation function (3. For high resolution meshes. SBR Stress and RBF mesh motion solvers. 48.8(b) shows that the computing time used for RBF mesh motion is less than 10% of the computing time for the complete timestep. the order of computing times can be decreased further by increasing the coarsening ratio. respectively. 32. In principle. . In addition. at least an order of magnitude more. On the other hand.8(b) shows the computing times. it is observed that these methods require very large computing times. The nonlinear behaviour of the ﬁnal curve is caused by the choice of the coarsening ratio to select the moving boundary points. (3. 800k and 1600k cells and needed a total computing time of about 8.
the SBR Stress equation ( ). the two diﬀerent phases of the radial basis function mesh motion. RBF relative method (×) and RBF relative method in combination with coarsening and smoothing techniques (•). Concerning the coarsening. Meshes are used from 100k to 1600k. which is about reducing the number of boundary points by applying coarsening and smoothing techniques.4. A possible better choice would be to use parallel direct techniques. a mesh motion problem leads in general to an illconditioned system.4.8 Computing times for diﬀerent mesh motion solvers. section 3. leading to a system which is more eﬃcient to solve using existing iterative solver techniques. the illconditioned system cannot be eﬃciently solved using iterative techniques.. which occur in this type of problems. (b) shows the computing times for one timestep of a threedimensional ﬂapping wing simulation. 3. which is diﬃcult to solve directly and iteratively. solving (I) and evaluation (II) can be very expensive. RBF absolute method ( ).5 Improving computational eﬃciency 104 70 60 Computing time [s] 103 Computing time [s] 50 40 30 20 10 10 0 10 0 69 Mesh motion Flow equation Total timestep 102 101 101 Grid spacing 102 0 0 1 2 4 8 Grid spacing 16 (a) Twodimensional (b) Threedimensional Figure 3. are caused by the boundary point locations. 2007) to approximate the system of equations (3. Diﬀerent preconditioning techniques are described in literature (Boer de et al. However. as long as the resulting mesh is of suﬃciently high quality.5. the combination of cell clustering on the moving surface and the large distance to the outer boundary. causes large diﬀerences between points in Φbb . was about O(1010 ). Despite the preconditioning. In general moving mesh applications. it is not necessary to solve (I) and (II) exactly. (a) shows a comparison of computing times for the mesh motion solvers based on the Laplace equation (◦).3. with a signiﬁcant gain in computing time. Furthermore.2 Iterative techniques and parallel implementations As was shown in the previous section. a parallel version of LAPACK. A diﬀerent way to improve the eﬃciency of the computation has already been applied in section 3. complex greedy algorithms (Rendall & . since the internal point motion can be arbitrary.8). The condition number in case of the model problem of a moving square. The high condition numbers. The times are subdivided by solving the RBF mesh motion and the ﬂow equations. The computing times for the Laplace equation (◦) and the SBR Stress equation ( ) are nearly identical. available in the linear algebra packages SuperLU and ScaLAPACK.
. Furthermore. The diﬀusivity. the way OpenFOAM deals with the ﬁnite volume implementation. The ﬁrst method solves the Laplace equation with a variable diﬀusion coeﬃcient. in both cases. 2007). using all control points and corresponding coeﬃcients. reordering can be applied for further enhancement. the evaluation can be easily implemented in parallel since it only involves a matrixvector multiplication. αi and βi . both based on solving a partial diﬀerential equation. Then every processor only performs the evaluation of the internal points. the solid body rotation stress (SBR Stress) mesh motion uses the diﬀusivity. concerning a parallel implementation. the internal point displacements are obtained by evaluating the radial basis functions. The major diﬃculty. Currently. are compared. A new mesh motion solver is incorporated in OpenFOAM . Especially when using the thin plate spline (TPS) or the continuous polynomial C 2 . this is being implemented in OpenFOAM . 2008a) may be applied to select the necessary boundary points such that the eﬃciency is increased even further. concerning the RBF mesh motion. to inﬂuence the quality of the mesh. is deﬁned to decrease quadratically with the distance from the moving boundary.70 Mesh deformation techniques for ﬂapping ﬂight Allen. which uses the interpolation of radial basis functions (RBF).8) is to decrease the condition number by only taking the boundary points with a low mutual distance (Boer de et al. The cell nonorthogonality and skewness are compared. For given boundary point displacements the internal mesh displacements are obtained by solving a system of equations to obtain an array of interpolation coeﬃcients. Secondly. is that all processor partitions need to know which control point belongs to itself and which to the other partitions. acting as a stiﬀness. Besides solving a partial diﬀerential equation the motion of mesh points can be deﬁned using interpolation techniques. Finally. As with the Laplace equation.. which are distributed over all partitions. 2007). Using those coeﬃcients. 3.6 Conclusions In this chapter two diﬀerent mesh motion techniques were described which are commonly used within the code of OpenFOAM . of that particular partition. This new mesh motion technique does not need any information about the mesh connectivity and can be applied to arbitrary unstructured meshes containing polyhedral cells. Additionally. Another method to increase the eﬃciency to solve the system (3. the linear stress equation was modiﬁed to include rigid body rotations in order to cope with the severe mesh deformation present in ﬂapping wing simulations. in addition. The RBF mesh quality provides superior mesh quality over the Laplace and SBR Stress mesh motion solvers. which is used to control the ﬁnal mesh quality. the speed of the evaluation (II) can be increased by various fast evaluation algorithms (Boer de et al. The three mesh motion solvers are tested using a case of a twodimensional rectangular block which moves through a Cartesian mesh. diﬀerent radial basis functions.
After this elaborate discussion it is concluded that. the mesh quality is high in terms of low skewness and nonorthogonality. The TPS has global support. which reduces the system of equations considerably. The RBF mesh motion was successfully tested on simple test problems and for a threedimensional ﬂapping wing with the possibility to incorporate a ﬂexing moving boundary. whereas the C 2 basis function has compact support. Secondly. So a coarsening algorithm selects those control points. diﬀerent methods are implemented to increase its eﬃciency. Therefore. concerning the threedimensional wing simulations. . First of all.3. Since the RBF mesh motion technique encounters a dense system of equations.6 Conclusions 71 as radial basis functions. a subset of the moving boundary points was selected. it is justiﬁed to neglect the outer (ﬁxed) boundary points. a smoothing function is used to decrease the RBF contribution to zero at the outer domain boundaries. because not all points are necessary when the body performs a rigid body motion. the globally supported TPS should be used in combination with the coarsening and smoothing techniques to increase the efﬁciency of the RBF mesh motion method.
.
1993. undergo signiﬁcant interactions with the environmental ﬂuid in which they move. Additionally. Animals that ﬂy or swim. Both geometry and wing kinematics are dynamically scaled in order to design a sound framework for comparison. 4. the force coeﬃcients are used in conjuncture with the lifttodrag ratio to assess the ﬂapping wing performance. Reynolds and Rossby numbers) are identiﬁed after writing the NavierStokes equations in a rotating reference frame.. Taylor et al. 2003) at certain scales. the physical and numerical modelling of ﬂapping wing and foils is described. representing occasional wing deformation. . To systematically study the aerodynamics around ﬂapping wings.. which are equivalent from the ﬂuiddynamic perspective (Triantafyllou et al. These equations are used to deduce the dimensionless numbers relevant for insect ﬂight. using the radius of gyration.1 Introduction This chapter deals with the physical and numerical modelling of ﬂapping wings. which describes the wing motion. The kinematic model. for the mathematical analysis of swimming or ﬂying it is important to formulate the governing equations and accompanying boundary conditions in an appropriate form. Therefore. a model planform and kinematic model is deﬁned. The relevant dimensionless numbers (Strouhal. in hovering as well as in forward ﬂight conditions. consists of a rigid body motion appended by a ﬂexing.CHAPTER 4 Physical and numerical modelling of ﬂapping foils and wings In this chapter.
The Reynolds number is deﬁned as the ratio between the inertial and viscous forces present in a ﬂuid. 2005. Here Φ0 is half of the total ﬂapping angular amplitude.. The ﬁrst use of the Strouhal number was in the context of the natural vortex shedding behind a stationary cylinder in a uniform ﬂow. A diﬀerent approach to dimensionless numbers is to deﬁne them as the ratio of time or length scales. J = U/4Φ0 f R. This deﬁnition of the Strouhal number. Additionally. White.. the Strouhal number is commonly used to characterise . 2003. travelled by the wing. see chapter 2.74 Physical and numerical modelling of ﬂapping foils and wings The Reynolds number In order to improve our understanding of biological ﬂows. 2008b). 2003. like the ﬂows around ﬂapping wings or ﬁns. This expression is very similar to the reduced frequency (Shyy et al. the Strouhal number plays an important role in ﬂapping wing aerodynamics as well. i. In general engineering applications. Lentink & Gerritsma. Williamson (1988) found a universal relation between the Reynolds and Strouhal number based on the observed vortexshedding frequency in the laminar ﬂow regime. the Strouhal number can be very useful. The typical deﬁnition of the Strouhal number is the ﬂapping frequency times ﬂapping amplitude divided by a reference ﬂow velocity. Bos et al. For example. provided that the wing ﬂaps in a stroke plane perpendicular to the forward velocity (Lentink & Gerritsma.e. deﬁning implicitly the complexity of the mathematical model needed to solve the problem. which is easier to understand and to apply to a ﬂow problem. like shear layers and vortex generation. For moving bodies and especially ﬂapping wings. which is commonly used to relate the two velocities due to either ﬂapping and forward ﬂight. the Strouhal number is proportional to the maximum value of the induced angle of attack.e. compared to the inertial component. the single wing span. such that it identiﬁes what kind of propulsive mechanism applies to the ﬂapping wing. the Reynolds number determines whether the ﬂow behaves turbulent or laminar. instead of the ratio between two distinct forces (Tennekes & Lumley. Stc = f c/U .g. f the ﬂapping frequency and R represents the distance from root to tip of the wing. like the Reynolds (Re) and Strouhal (St) number in particular (e. in forward ﬂapping ﬂight. will be more pronounced. as deﬁned by Ellington (1984).. such that viscous phenomena. 2008). Panton. based on the ﬂapping amplitude. It is a property of the ﬂow. For ﬂapping wings or oscillating bodies. 2003. the Strouhal number can be deﬁned based on the imposed oscillation frequency and amplitude. 1972. is closely related to the advance ratio. such as the Von K´rm´n vortex street behind a bluﬀ body. the forces in the ﬂow are dominated by the viscous term. Another deﬁnition was introduced by Dickinson (1994) and Wang (2000b) using the average chord length as reference. Re = O(100). it is of importance to make extensive use of dimensionless numbers. when a ﬂapping wing operates at a very low Reynolds number.. Thaweewat et al. i. which is the ratio between the forward and ﬂapping distance. For example. In that case the Reynolds number can be deﬁned as the ratio between the convection time over the diﬀusion time. a a The Strouhal number Besides the Reynolds number. Taylor et al. 2009). 1991).
only one ﬂapping wing is considered. Furthermore. 4. whereas the reduced frequency number is used in ﬂapping wing problems. Additionally. relative to the mean chord. like the operation of insects and ﬁsh.2. followed by the conclusions of this chapter in section 4. the ﬂow can be considered to be incompressible since the Mach number (a measure for compressibility) is typically M a = U/a = O(10−3 ) (Brodsky. In view of simplicity. Section 4. and restated here: ∇ • u = 0. corresponding to the operating conditions of fruit ﬂies. In section 2. and 1000. ∂t ρ (4. speciﬁc attention is given to the appropriate deﬁnition of the governing dimensionless numbers to investigate the ﬂow around ﬂapping wings.2 Governing equations for ﬂapping wings Concerning ﬂapping ﬂight in nature. In section 4. the governing equations are formulated for forward and hovering ﬂapping ﬂight.4) and (2.1) 1 ∂u + ∇ • (uu) = − ∇p + ν∇2 u.4.5). Equations and other assumptions As previously discussed. where U [m/s] is the reference velocity and a [m/s] the speed of sound. This implies that no interaction between two wings or with the body are included. Secondly. In accordance with Lentink & Dickinson (2009a).2) . 500. the reciprocal of the Strouhal number is known as the dimensionless wavelength λ∗ = U/f c which is often used in studies concerning forward ﬂapping ﬂight as it seems reasonably intuitive in that it corresponds to the distance travelled over one ﬂapping period. followed by the dynamical scaling of ﬂapping ﬂight in section 4. the present study deals with a model wing which is a simpliﬁed representation of a ﬂying insect wing operating at Reynolds numbers.3.6 describes the force and performance deﬁnitions. which allows that the induced vortical ﬂow can be studied in more detail. under hovering as well as forward ﬂight conditions. In order to perform a sound and valid comparison it is important to maintain constant dimensionless numbers while kinematic parameters or ﬂow properties are varied to study their inﬂuence. house ﬂies and bumblebees. Nevertheless.2 the incompressible NavierStokes equations were deﬁned by equation (2. the considered ﬂapping kinematics that result in large rotation rates put the current numerical techniques to a signiﬁcant challenge. the model wing selection and the deﬁnition of the kinematic model parameters are described in section 4. (4. 1994).2 Governing equations for ﬂapping wings 75 the vortex shedding. it is appropriate to use dimensionless numbers to study the eﬀect of ﬂapping characteristics on the aerodynamic performance. Re = 100.4. As a prelude to the numerical solution of the governing equations. respectively.7.5. the mesh generation in combination with the boundary conditions is brieﬂy dealt with in 4.
Using Computational Fluid Dynamics (CFD) techniques.76 Physical and numerical modelling of ﬂapping foils and wings These equations are derived by analysis of the forces on an inﬁnitely small ﬂuid element in an inertial reference frame. has to be zero. approximately. the reference frame approach is only used to identify certain dimensionless numbers that are related to the rotation of the threedimensional wing. which means that the velocity. Ωwing is the angular velocity of the rotating reference frame. 1998. This relation results in a noslip condition in the rotating reference frame. When ﬂight under hovering conditions is considered. is deﬁned as (Ginsberg. The corresponding velocity transformation. In the present study. 2008) with respect to the inertial frame (XY Z). in all directions. The resulting boundary condition on the ﬂapping wing will be such that the eﬀective velocity will be zero. a noslip condition (White. the actual ﬂow computations are made with respect to the inertial reference frame. resulting in the following expression: uwing = uWING − (utrans + Ωwing × r). where uWING is the ﬂow velocity at the wing in the inertial reference frame.e. derived from realistic insect data (e. 1991) needs to be speciﬁed. representing the ﬂapping motion of the wing. since the reference frame moves with the boundary.. where uXY Z corresponds to the velocity in the inertial reference frame. these equations are solved on a discretised computational domain in combination with appropriate initial and boundary conditions. the initial velocity ﬁeld is zero as well as the boundary conditions at the outer domain boundary. which the case in hovering ﬂight. 2003). i. In the present study.g. utrans is the translating velocity of the reference frame itself and can be used to specify the translation of the insect body. This is accomplished by deﬁning the mathematical velocity on the moving boundary to be equal to the actual wing motion. The boundary condition needs to take care of the rotation of the reference frame. 1999) uXY Z = uxyz + (utrans + Ωwing × r). The rotating reference frame is attached with its origin to the joint around which the wing rotates. to make the velocity at the moving boundary equal to zero. Fry et al. the translation velocity of the rotating reference frame is assumed to be zero. which moves according to a speciﬁc kinematic model. At the boundary of the ﬂapping wing. which is ﬁxed to the ﬂapping wing and moves accordingly. whereas uxyz represents the velocity in the local rotating reference frame. i. relative to the wing. Additionally. The accelerations in the inertial and rotating reference . Momentum analysis in rotating reference frame A diﬀerent 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).e. r is the distance from a rotating point to the origin. Baruh. it is interesting to study the accelerations (Lentink.
Baruh.2) such that the following transformed NavierStokes equations are obtained (for the sake of simplicity. Ω needs to be related to the ﬂapping motion. Dt ρ (4. 1999): aXY Z = axyz + (aang + acen + acor ).2 Governing equations for ﬂapping wings Y y x O X 77 Z z Figure 4. This will be elaborated in detail in section 4. Now the expressions for velocity and accelerations in the rotating reference frame are substituted into the NavierStokes equations (4. The rotational reference frame xyz is moving with the wing and obtained by rotating the inertial reference frame XY Z by three orientation angles. 1998. in addition to the already described Reynolds and Strouhal numbers. These new dimensionless numbers will become available if the diﬀerent . uxyz [m/s] is the velocity in the rotating frame. In order to explore the diﬀerent acceleration terms. and Coriolis acor [m/s2 ] accelerations are respectively deﬁned as aang acen acor ˙ = Ω × r. the subscripts are dropped): 1 Du ˙ + (Ω × r) + (Ω × (Ω × r)) + (2Ω × u) = − ∇p + ν∇2 u. which strongly depends on the wing kinematics.3. aang [m/s2 ].1 Illustration of the rotational reference frame. centripetal. acen [m/s2 ]. The angular.1) and (4.4.3) This transformed equation describes the momentum balance for a ﬂuid particle close to the wing (in the boundary layer) in the rotating reference frame. = Ω × (Ω × r). The main parameter in these three diﬀerent accelerations is the angular velocity Ω [rad/s]. With this approach it is possible to derive dimensionless numbers representing the diﬀerent acceleration terms. frames are related using the following (Ginsberg. = 2Ω × uxyz .
2R c Ω (4. other dimensional numbers can be identiﬁed. ∇∗ = c · ∇. Furthermore. Ωc By the rotation amplitude Ω. 4. 2001. ˙ ΩR c 2 Uref . r∗ = .3 Wing shape and kinematic modelling When investigating the threedimensional aerodynamics around ﬂapping wings. related to the rotation of the reference frame. rewriting gives: 1 Du 1 1 1 1 ˙ · + · (Ω × r) + · (Ω × (Ω × r)) + · (2Ω × u) = −∇p + · ∇2 u. ρref [kg/m3 ] is the reference density (constant in incompressible ﬂows). In addition to the Reynolds and Strouhal numbers. 1998) or . especially the velocity and acceleration due to the wing rotation. these dimensionless numbers strongly depend on the wing kinematics. St Dt Cang Ccen Ro Re where the additional dimensionless number are respectively deﬁned as Cang = 2 Uref . 2003). or to opt for a more generic approach. the eﬀect of diﬀerent kinematics and ﬂow features may be related to these dimensionless numbers. t∗ = t. in order to fully understand them. R [m] is a radius length. one may try to simulate the geometry and conditions of speciﬁc species.4) Ccen = (4. which ref is deﬁned as f = Uc . where c the average chord length.5) (4. a fruit ﬂy (Sane & Dickinson. L [m] is the reference ˙ length. Ω∗ = . respectively.g.3) are scaled as u∗ = ˙ Uref r p Ω u Ω ˙ . which is still attached to the wing. Birch & Dickinson. f · c Dt Uref Uref Uref Uref c where Uref [m/s] is the reference velocity. Ω∗ = . hawkmoth (Liu & Kawachi. Previous investigations of speciﬁc insect species have been reported for e. Therefore. f [1/s] the ﬂapping frequency.78 Physical and numerical modelling of ﬂapping foils and wings terms in equation (4. p∗ = 2 . ˙ Uref c Ω R ρref · Uref Ω leading to (dropping the stars for simplicity): ˙ Uref Du ΩRc ˙ Ω2 Rc Ωc ν · + 2 ·(Ω×r)+ 2 ·(Ω×(Ω×r))+ ·(2Ω×u) = −∇p+ ·∇2 u. Ω [ rad /s] and Ω [rad/s] are the average rotational velocity and acceleration.6) Ro = Uref .
7).9) which is obtained by rewriting equation (4.4.5 and c = 1. Radius of gyration In order to deﬁne a sound framework of comparison for diﬀerent threedimensional and twodimensional simulations.8) and (4. Section 4. the average chord length can be obtained by integration of the chord distribution c(r) along the wing span from root to tip: 1 R c= c(r)dr.2 shows the wing planform and corresponding parameters for a fruit ﬂy wing and a ellipsoidal model wing.4.1 Wing shape and planform selection The general model wing is described by an ellipsoid in the threedimensional wing reference frame: x 2 y 2 z 2 + + = 1.2 deals with the kinematic modelling. (4.7). . b and c are the semiaxes of the ellipsoidal wing. In order to obtain a wing which has a single wing span.7) a b c where a.3. c2 (4.3 Wing shape and kinematic modelling 79 a dragonﬂy (Isogai et al.9) leads to an average chord length of c = π/4 for c = 1.05.0. the kinematic model was appended with an active wing ﬂexing component.. The present study follows the second approach and considers a threedimensional ellipsoidal model wing (Bos et al.0 and a thickness of 10% of the chord. a similar ellipsoid was used by Wang et al. the semiaxes are chosen as a = 0.. Since the planform is analytically given by equation (4. The numerical implementation of the wing kinematics is the subject of section 4. Assuming that the chord c(r) is represented by y(z) the following relation is used to calculate the average chord: c(r) = b2 1 − r2 .1 brieﬂy describes the wing shape and morphology. while section 4. 4.3. a similar elliptical crosssection is applied for z = 0). Previous studies show that speciﬁc insect features. bs = 2. which is described in section 4. it is necessary to deﬁne all reference parameters.3.3. In addition. like the corrugated wing planform (Luo & Sun. Figure 4. 2005) are of minor inﬂuence on the resulting ﬂuid behaviour. 2008). The threedimensional elliptical planform is shown in ﬁgure 4.3.3.8) R 0 where R [m] is the radius of the wing tip and c(r) [m] the chord distribution along the wing span.0.0 (for the twodimensional airfoil. both deﬁning the planform.2 in comparison to a more realistic representation of the fruit ﬂy wing planform. Evaluation of equation (4. The general kinematic modelling consists of a translation and a rotation component. (4. (2004) for their twodimensional research. a maximal chord length cmax = 1. b = 0. 2004).
.80 Axis of rotation Physical and numerical modelling of ﬂapping foils and wings Axis of rotation Rtip Rroot y z c(r) Rg O S dr dr r Rg r Rg (b) Ellipsoidal shape S bs Rroot y z c(r) Rg Rtip bs O (a) Fruit ﬂy shape Figure 4. Luo & Sun (2005) compared the ﬂow induced by ﬂapping wings with diﬀerent aspect ratios and found that the radius of gyration provided a reliable framework for force comparison when the ﬂapping velocity is varied. The planform is deﬁned by the chord variation c(r). especially when the comparison of diﬀerent kinematic models is concerned (Bos et al. Lentink. Previous twodimensional studies . (4. 4. 2008). 2008) and is calculated as 1 R 2 r c(r)dr. this is the location where the resulting lift acts.2 Model wing geometry and planform. the single wing span bs and the planform surface area S. of which the location is varied to study the eﬀect of the angular accelerations. 2005. The radius of gyration. the purpose of the present research is also to investigate the threedimensional ﬂow around ﬂapping wings at low Reynolds numbers. Rg is deﬁned as the weighted second moment of inertia (Luo & Sun. R the distance from the rotation origin to the wing tip and c(r) represents the chord distribution along the wing. As the local velocity of each crosssection varies during ﬂapping. In threedimensional simulations the wing revolves around the origin O.3.2. r the spanwise coordinate.2 Kinematic modelling Besides the numerical interest in the development and improvement of mesh motion techniques. at a representative crosssectional area of the wing. described in chapter 3. Additionally. Schematic illustration of the geometry and planform of a fruit ﬂy (a) and ellipsoidal model wing (b).10) Rg = S 0 Here S is the wing planform. Besides the average chord length c and the single wing span bs the radius of gyration is another important geometric parameter. the spanwise reference location is chosen to be at the radius of gyration. According to Ellington (1984). introduced in section 4. The radius of gyration Rg is used to deﬁne a sound framework for comparison between mutual threedimensional and twodimensional simulations. the scale at which insects operate.
Figure 4. 2003). In addition. from purely harmonic motion to complex realistic fruit ﬂy kinematics. The kinematic wing motion is deﬁned by the variation of three independent attitude angles. Furthermore.. as is shown in ﬁgure 4.3 Schematic illustration of the governing ﬂapping angles. In this threedimensional model the three degrees of freedom of the wing motion are deﬁned as the ﬂapping angle.. like harmonic motion. φ(t) corresponds to the stroke variation. the centre of rotation (which is equivalent to the Rossby number Ro) or the angle of attack amplitude.. which is measured in a plane normal to the stroke plane. the shape of the fruit ﬂy angle of attack clearly shows a plateau of constant value.. showed the tight relationship between the aerodynamic forces or performance and the kinematic model (Wang et al. which are considered. Figure 4. Using twodimensional numerical techniques. 2009). 2008. The deviation may be used to create a ‘ﬁgureofeight’ pattern which is present in realistic fruit ﬂy kinematics (Fry et al. 2003). Figure 4. e. diﬀerent kinematic models have been studied. the angle of attack. in the mean stroke plane.1T).. Since an accurate threedimensional numerical method has been developed to solve for the ﬂow around ﬂapping wings.. with respect to the horizontal plane and the deviation θ(t). which is ex . 2004) compared to (b) the realistic fruit ﬂy kinematics (Fry et al. α(t) to the geometrical angle of attack and deviation θ(t). 2004. the angle of attack shows a dip (at t=0.4(b) shows that the realistic fruit ﬂy kinematics is characterised by an asymmetry in ﬂapping angle. simpliﬁed kinematics was used.4. (2008) showed that the ﬂapping wing performance may be inﬂuenced by the speciﬁc features of a particular kinematic model. Besides the use of an idealisation of the insect wing planform.4 shows the least and the most complex of the kinematics models.3 Wing shape and kinematic modelling Y 81 α(t) Rtip Stroke plane O θ(t) φ(t) start upstroke midstroke Rroot X start downstroke Z Figure 4.3. to study diﬀerent parameters independently. α(t). In literature. Thaweewat et al. see ﬁgure 4.3.g. which is measured in a plane normal to the stroke plane. lowering the eﬀective angle of attack. φ(t). the eﬀect of diﬀerent kinematic models on the aerodynamic performance can be evaluated. Bos et al. Bos et al. Flapping wing motion is governed by three angles. and angle of attack.4 illustrates (a) the harmonic model (Wang et al.
the deﬁnition of the harmonic model is described accordingly. is deﬁned by a sine function with respect to 90◦ and the deviation angle. the realistic fruit ﬂy kinematics in characterised by the presence of deviation which may result in a ‘ﬁgureofeight’ pattern (Shyy et al. 4. Remind that this amplitude is half times the value used in literature (e.82 Physical and numerical modelling of ﬂapping foils and wings ploited by Sane & Dickinson (2001.. 2003).. (4. Bos et al. Fry et al. The ‘trapezoidal’ shaped angle of attack is implemented using a piecewise continuous function and the deviation by deﬁning a nonzero harmonically variation of θ(t). (4. 2008). 2005. such as the previously described plateau in angle of attack. 2002). which is deﬁned from stroke reversal to midstroke. α(t). simpliﬁed (harmonic) kinematics may be interesting for Micro Air Vehicles (MAV) implementation. Φ0 in Ellington. Additionally. 2008b) In order to illustrate the variation of the motion angles.3. 2003. Lehmann et al.12) In general. 2003. The corresponding angular velocities are found by taking the timederivatives of equation (4. 2008. Aθ is the amplitude of the deviation angle which causes the socalled ‘ﬁgureofeight’ pattern and is sinusoidal shaped (Fry et al. is given by a pure sine: φ(t) = Aφ · cos(2πf t).11). The geometric angle of attack. φ(t).3 Modelling of active wing ﬂexing In addition to the rigid body motion as shown in ﬁgure 4. ˙ ˙ θ(t) = Aθ · (2πf ) · cos(2πf t).g. The ﬂapping angle.. 2 θ(t) = Aθ · sin(2πf t). ˙ φ(t) = −Aφ · (2πf ) · sin(2πf t)... is described by a cosine function. θ(t).. 1984). Lehmann et al. which is the initial position under hovering conditions. Also. a ﬂexing displacement of the wing surface is deﬁned. Lu & Shen. (2005) using a Roboﬂy. Aφ is the ﬂapping amplitude. The ﬂexing displacement is deﬁned with respect to the initial wing position and can be written .11) Here. 2008). As will be shown in chapter 5 (Bos et al. f is the ﬂapping frequency and Aα represents the amplitude of the angle of attack with respect to π/2. modelled by a ‘trapezoidal’ shape and the presence of deviation (Birch & Dickinson. Since a full ﬂuid structure interaction (FSI) simulation is too expensive and beyond the scope of the current research.3 it is possible to deﬁne an extra displacement concerning ﬂexing of the wing. α(t) = −Aα · (2πf ) · cos(2πf t). π α(t) = − Aα · sin(2πf t). the ﬂapping wing performance may be signiﬁcantly inﬂuenced by modiﬁcations of the basic kinematics. this harmonic model is crude but fairly reasonable representation of the ﬂapping motion of a fruit ﬂy.
rota . At every timestep the location of the boundary points is determined leading to three distinct displacement arrays.1 0.9 1 t/T [] φ. as x(t) = Af · cos 2πx0 ǫf · sin(2πf t). The cosine shaped wing ﬂexing is deﬁned by ǫf . representing pure harmonic variations of the ﬂapping angle φ(t) (•). the angle of attack. an extra ‘bump’ and a degree of ‘trapezoidal’ in angle of attack and the presence of deviation. (a) shows the least complex kinematic model. which corresponds to the cosine ratio. The realistic fruit ﬂy kinematics is characterised by an asymmetric variation of ﬂapping angle. which is the most complex kinematics available.5 0. where the ﬁrst cosine function deﬁnes the wing shape and the sine function represents the timevariation.6 0.3 Wing shape and kinematic modelling 100 80 60 40 φ. the wing kinematics is calculated beforehand and applied to the numerical ﬂow solver.5 (b). 4.8 0.3 0. ǫf = 0. In general.8 0.e. due to translation.5 shows a plunging airfoil incorporating ﬂexing for ǫf = 0.2 0. α.4 Numerical implementation of the wing kinematics In previous sections the physical kinematic modelling was described. The translational component is not used in the current threedimensional simulations. in the twodimensional simulations.7 0. the wing kinematics can be decomposed into a translation. The current section deals with the numerical implementation of this particular wing kinematics.7 0. In addition to the translation and rotation.5 means that the shape is like a half cosine function. a limited number of two.4. a rotation and a ﬂexing (deformation) component. which is limited to hovering and forward ﬂow conditions with stationary position of the rotation origin.1 0. However. The variation of realistic fruit ﬂy kinematics (Fry et al.4 0.5 0. the ﬂapping motion is deﬁned by a translation in combination with one rotation angle.4 Comparison of the harmonic and fruit ﬂy kinematic models.. Af is the ﬂexing amplitude vector and x0 the location of the initial boundary points. angle of attack α(t) (◦) and deviation θ(t) ( ). θ[◦ ] 20 0 20 40 60 80 0 0.9 1 t/T [] 83 (a) Harmonic kinematics (b) Fruit ﬂy kinematics Figure 4.3 0. 2003) is shown in (b). Figure 4.3. i. In general.2 0.and threedimensional simulations using a predeﬁned wing ﬂexing have been performed. α. θ[◦ ] 100 80 60 40 20 0 20 40 60 80 0 0.25 (a) and ǫf = 0.4 0.6 0.
z(t)).5 Illustration of a ﬂexing twodimensional airfoil. a cosine function is used to deﬁne the motion in the inertial reference frame. Translation The displacement array of the boundary points.25 (b) Half cosine shape. Rold and Rnew are the rotation transformation matrices at respectively the old and new time instances.5 Figure 4. Here ∆xrot is the boundary displacement due to rotation. y(t). Both ﬁgures (a) and (b) show the upstroke (left) and downstroke (right). such that: ∆xrot = xnew − xold .ǫf = 0. The twodimensional ﬂexing airfoil is modelled by a timevarying cosine shape. tion and ﬂexing. The sine function in this deﬁnition is used when the wing needs to move according to an ordinary rigid body motion (Ginsberg.25 (a) or a half cosine. ǫf = 0. ǫf = 0. (4. At and ft are respectively the translation amplitude and frequency vectors.14) (4. is calculated at subsequent (old and new) timesteps with respect to the initial mesh: xold = Rold · x0 . Rotation The second boundary point displacement array.84 Physical and numerical modelling of ﬂapping foils and wings (a) Quarter cosine shape. where x(t) = (x(t). and xnew = Rnew · x0 . This airfoil shape is either a quarter cosine. ǫf = 0. 1998). due to rotation.13) .5 (b). When a ﬂapping motion is desired. which is due to translation is obtained from x(t) = At · sin(2πft t).
3 Wing shape and kinematic modelling 85 and The initial boundary points are given by x0 . ff the ﬂexing frequency and x0 represents the initial boundary points of the ﬂapping wing at t = 0.18) which is substituted into equation (4. 1998.16) RY (t) = sin(φ(t)) 0 cos(φ(t)) cos(α(t)) RZ (t) = sin(α(t)) 0 −sin(α(t)) cos(α(t)) 0 0 0 . This wing ﬂexing is deﬁned as xﬂex (t) = Af · cos(2πx0 ) · sin(2πff t). According to (Ginsberg. Y . the ﬂapping wing is able to perform a ﬂexing motion as well. due to rotation: ∆xrot = [Rrot (tnew ) − Rrot (told )](x0 − r0 ). After a general sequence of matrix multiplications. Y and Zaxis. spanwise and chordwise. accelerating the ﬂuid from top to bottom. the total rotation matrix is obtained by combining equations (4. respectively. which is varied systematically in chapter 7. these three matrices are deﬁned as 1 0 0 (4. where Af is the ﬂexing amplitude vector. This deﬁnition permits ﬂexing in each orientation of a threedimensional wing. rotation and ﬂexing.14) to ﬁnd the displacement of the boundary points at the old and new times.e. Flexing In addition to the translation and rotation. (4. the following is obtained: ∆xb = ∆xtrans + ∆xrot + ∆xﬂex . These rotation transformation matrices consist of three diﬀerent components.or Zaxis correspond to the deviation angle. due to a rotation around the X.17) Here. θ(t).4. 1 (4.3. it is clear that a hovering wing ﬂaps in the XZ plane. α(t). (4. 1999). Using the motion convention.13) and (4. Baruh. 0 1 0 (4. shown in ﬁgure 4.17) as Rrot (t) = RX (t) · RY (t) · RZ (t). i. 0 sin(θ(t)) cos(θ(t)) cos(φ(t)) 0 −sin(φ(t)) . the rotation around the X.15) to (4. and the angle of attack. When combining the previous results for the displacements due to translation.15) RX (t) = 0 cos(θ(t)) −sin(θ(t)) . φ(t). . ﬂapping angle.19) where r0 is the direction vector of the initial rotation origin.
the reference crosssection of the threedimensional wing is positioned at the radius of gyration. 2008. where T is the ﬂapping period. angular accelera˙ tion Ω and average chord length c it is necessary to deﬁne an appropriate reference velocity. Besides the angular velocity Ω. evaluation ˙ gives Ω = 4Aφ f and Ω = 2πf Ω.2 diﬀerent dimensionless numbers. First.22) it is straightforward to ﬁnd Uref by multiplying the expression for Ω (4. This approach has two major advantages.4 Dynamical scaling of ﬂapping wings In section 4. it is desirable to start the numerical simulation at maximal ﬂapping angle. but in case of forward ﬂight conditions. were identiﬁed in order to scale the NavierStokes equations. Rg .22) where uRg (t) is the absolute velocity at Rg which can be decomposed into three components u(t). yielding good conditions for numerical convergence. Cang = Uref /ΩR c. yielding high mesh quality at the extreme wing positions.and Zdirection. in order to minimise the initial acceleration of the mesh. 0 (4. (4. 2008. Lentink. Therefore.. As previously described and in accordance with (Bos et al. there is an additional freestream . (4. 2005). 4. governing ﬂuid ﬂow. To de2 2 ˙ ﬁne these rotational dimensional numbers. As shown. Ccen = Uref /Ω2 R c and Ro = Uref /Ω c. This timeaveraged velocity follows from: Uref = 1 T T 0 uRg dt = 1 T T u2 (t) + v 2 (t) + w2 (t)dt. Ω ˙ and its timederivative Ω for the particular kinematics implemented: Ω= 1 ˙ Ω= T 1 T T 0 T 0 ˙ φdt = 1 T 1 T T T 0  − Aφ · (2πf ) · sin(2πf t)dt = 4Aφ f. the initial velocity is as small as possible. it was chosen to use a rigid body motion until the maximal ﬂapping angle was reached.86 Physical and numerical modelling of ﬂapping foils and wings This total displacement ﬁeld for the boundary points can be applied to the mesh motion solver.20) ¨ φdt = 0  − Aφ · (2πf )2 · cos(2πf t)dt = 8πf 2 Aφ = 2πf Ω. Luo & Sun. v(t) and w(t) in respectively X. and the reference velocity is calculated at that particular location.23) These relations hold if the wing is the only driving force behind the resulting ﬂow velocity. in general this occurs at T /4. the mesh deformation is symmetric. concerning rotational motion. it is important to identify the average rotation amplitude. available in most commercial and noncommercial CFD codes. Numerical initialisation If a ﬂapping wing is considered. (4.20) by Rg : Uref = 4Aφ f Rg .21) ˙ where the ﬂapping velocity φ was taken from equation (4. Secondly. Y .12). Using equation (4.
The Reynolds number for hovering conditions and using Rg can be written: Re = Uref c 4Aφ f Rg c = . Aswept . 2008b. Additionally. which is illustrated in ﬁgure 4.. 2008b). Aswept .4. the Reynolds number at Rg .. J. ν ν The area that is swept by the revolving wing (Usherwood & Ellington. (4. the dimensionless amplitude at Rg . Thaweewat et al. 2002). with f the ﬂapping frequency. U∞ . For completeness and consistency..6. which describes the forward speed with respect to the ﬂapping velocity at a certain radius. two other important parameters are kept constant as well. R: λ∗ U∞ . 1984). A∗ is kept constant for all relevant simulations. 2009).6 Schematic illustration of the kinematic parameters in forward ﬂight. In order to create an appropriate framework for comparison. is given by λ∗ = U∞ /f c. c which is a measure for the dimensionless translation of the selected crosssectional area. The dimensionless amplitude at the radius of gyration Rg is deﬁned as A∗ = Aφ Rg .24) In forward ﬂight another important parameter is the advance ratio (Shyy et al. = J= 4Aφ f Rg 4A∗ where the reduced frequency (Ellington. is obtained by subtracting the area swept by the wing tip from the swept . According to (Shyy et al.4 Dynamical scaling of ﬂapping wings 90◦ − β 87 Aα 2A∗ sinβ tan−1 (2Stsinβ) λ∗ Figure 4. 2003) for the representing simulations. velocity. a constant A∗ leads to similar wingwake interactions (Birch & Dickinson. Lentink & Gerritsma (2003) showed that the following relation can be used as a good approximation for both hovering (U∞ = 0) and forward ﬂight conditions: 2 Uref = U Rg ≈ U∞ + (4Aφ f Rg )2 . which is an implicit result of keeping Rg constant and the area swept by the wing. λ∗ is also known as the dimensionless wavelength. λ∗ .
it is deduced that keeping the swept area constant is similar to maintaining a constant Froude eﬃciency (Stepniewski & Keys. Additionally.6) as Ro = Rg Uref = . If Ro is varied. 2Aφ · (Rtip − Rroot ) · (Rtip + Rroot ).and threedimensional ﬂapping wing simulations. generating thrust by moving the ﬂuid. For a translating wing. which is inversely proportional to the Coriolis acceleration. Ωc c (4. It is necessary to use a suﬃciently large computational domain to minimise disturbances. 2008). the value for Ro is inﬁnite. However. since those values are readily available from literature. the determination of Ro would be easier if deﬁned as Ro = Rtip / c. where Rtip is the wing tip radius. from nearly translating to strongly revolving. the Rossby number is given by Ro = Rg / c. the Rossby number is equivalent to the single wing aspect ratio ARs = Rtip / c. it is eﬃcient . 1984. From (Lentink. When generating a grid around a twodimensional thin and ellipseshaped airfoil. a computational domain is necessary to contain the mesh on which the governing partial diﬀerential equations are solved.0 for insect and ﬁsh (Lentink. 2Aφ · bs · (Rtip + Rroot ).5 Computational domain and boundary conditions In this section the general approach for mesh generation is described. domain size and boundary conditions for these simulations in more detail. 6 and 7. explain the speciﬁc mesh generation. Furthermore. for a rotating wing.88 Physical and numerical modelling of ﬂapping foils and wings area by the wing root: Aswept = = = 2 2 2Aφ · (Rtip − Rroot ). where Rtip and Rroot are the radii at respectively the wing tip and root. Lentink. that has been applied for the two. the Rossby number. 2008). 2008). the eﬀect of diﬀerent rotation origins can be investigated. a geometric wing characteristic. which are dealing with respectively twodimensional hovering. the radius of gyration divided by the average chord length. needs to be obtained by rewriting equation (4. in general Ro = 3. 4. Ro is ﬁnite. 1984). like in real insects. it may be diﬃcult to obtain the radius of gyration. for Rroot = 0. When performing a Computational Fluid Dynamics (CFD) study. Ro. If the wing planform is complex. forward and threedimensional hovering ﬂight. with appropriate boundary conditions in order to obtain large convergence rates and the correct solution. The distance from Rtip to Rroot is identiﬁed as the single wing span bs . Chapter 5.25) When the reference crosssectional area is located at the radius of gyration (Ellington.
to use an Otype mesh. which are necessary (Wesseling. Thaweewat et al. leading to a small inﬂow and outﬂow at the boundaries of the computational domain. 2008. Depending on the ﬂapping conﬁgurations. which is shown in ﬁgure 4. The meshes for the twodimensional simulations were generated using Gambit software. GridPro . Γleft and Γright .4. Using conformal mappings (Bos et al. Γfront .. The twodimensional domain uses the Otype topology (a). it was possible to generate a high quality mesh around a thin ellipsoidal wing in a threedimensional boxshaped computational domain.6 Deﬁnition of force and performance coeﬃcients 89 Γtop Γleft Γwing Γleft Γwing Γback Γright Γright Y Y X Γfront Γbottom Z X (a) Twodimensional (b) Threedimensional Figure 4. and the cylindershaped outer boundaries. Γright . Γbottom .6 Deﬁnition of force and performance coeﬃcients Besides ﬂow ﬁeld analysis. while the threedimensional ﬂapping wing simulations uses the boxed topology (b). Generation of a threedimensional structured mesh around a wing is not an easy task. 6 and 7) it .and threedimensional simulations (chapters 5. Γleft .7(b). the outer boundaries. When simulating hovering ﬂight. 4. Γtop . 2001) when forward ﬂight conditions are simulated. using advanced visualisation techniques.7 Computational domain and boundary conditions. In order to make a sound comparison of forces and performance for mutual two. Γback . 2009). Using GridPro . outﬂow or symmetry planes. Γwing . the ﬂapping wing still induces a signiﬁcant amount of downwash.. Therefore. are set to inﬂow. an automated topology mesh generator is used. a high quality mesh is generated between the ellipseshaped wing boundary. the resulting forces acting on the airfoils and wings are of primary importance to assess aerodynamic performance. The outer boundary is split into Γleft and Γright in order to be able to specify inﬂow and outﬂow boundary conditions. which is shown in ﬁgure 4. see appendix A. which can be very cumbersome using manual procedures used in programs like Gambit or Gridgen .7(a).
S [m] the wing surface and dS [m] represents an inﬁnitely small surface area element. the lift and drag are constructed from the forces in X. In hovering conditions. Y . hovering or forward ﬂight. together with µ this forms the wall shear stress. If threedimensional ﬂapping in hovering ﬂight without deviation is considered. such that the wing is not moving in the horizontal plane. hovering or forward ﬂapping ﬂight. as shown in ﬁgure 4. The centre of the axes coincides with the origin of rotation of the threedimensional wing. is important to properly deﬁne force and performance coeﬃcients. i. ∂n where Ftot [N] is the total force vector. the general determination of the force coeﬃcients is explained in combination with performance characteristics. the lift force Flift is deﬁned in the vertical direction and equal to FY . where φ(t) is the ﬂapping angle. If the motion includes a deviation velocity. the drag force derivation is more elaborate but similar. when the main ﬂapping direction is around the Y axis.8. a twodimensional derivation is trivial.e. The total force vector is integrated over the wing surface and contains a pressure and a viscous contribution.90 PSfrag Physical and numerical modelling of ﬂapping foils and wings FY Z Y ˙ φ X FZ FX Rg Figure 4.or Zdirection.8 Forces on a general threedimensional ﬂapping wing.or threedimensional. depending on the type of threedimensional motion. Two diﬀerent force deﬁnitions can be deﬁned. FY and FZ are shown in ﬁgure 4. however.3. The term ∂u/∂n is the gradient of the velocity vector with respect to the normal vector to the wall. the deﬁnitions of the forces. p [N/m2 ] is the pressure and µ [Ns/m2 ] the dynamic viscosity. is deﬁned in opposite direction of the ﬂapping wing motion. Depending on the type of motion. Fdrag (t) = FX (t) · sin(φ(t)) − FZ (t) · cos(φ(t)). The drag force Fdrag . In this section. The forces are deﬁned in the threedimensional inertial reference frame. Besides the lift and drag. Forces In general three dimensions. there is a force in the . two. FX . These forces are calculated using the following expression: Ftot = S pdS − µ S ∂u dS. the following lift and drag variations are found: Flift (t) = FY (t).
the ratio between timeaveraged lift coeﬃcient CL and timeaveraged drag coeﬃcient CD is used to characterise performance. The lift is averaged over the complete ﬂapping period. opposite to the freestream velocity: Flift (t) = FX (t). Commonly the forces are made dimensionless using the dynamic pressure based on the average velocity. the forces are deﬁned as: CD = Fdrag . The average lifttodrag ratio. the same inertial reference frame is used as shown in ﬁgure 4. however. q·S where CD and CL are the drag and lift coeﬃcients and S the wing surface. This spanwise force is dominated by the viscous wall shear stress and therefore small compared to the lift and drag forces.4. In the case of forward ﬂapping ﬂight the lift force is deﬁned in the positive Xdirection and the drag in the negative Y direction. perpendicular to the horizontal plane (in hovering ﬂight). 2008). the ﬂapping velocity Uﬂap and the deviation velocity Udev . The reference velocity contains the freestream velocity U∞ . while for the drag the absolute values are used for averaging. CL /CD ave is chosen as an indicator of aerodynamic performance..3.6 Deﬁnition of force and performance coeﬃcients 91 direction of the spanwise coordinate. such that the sign ﬂips at stroke reversal. Fdrag (t) = −FY (t). The drag is opposed to the ﬂapping motion. It is justiﬁed to neglect this spanwise force. Hence. 0 where the integration is evaluated over one ﬂapping cycle with period T [s]. also known as the glide number in aerospace engineering. also because it has little relevance to performance. which is nonzero in forward ﬂapping ﬂight. but the direction of the uniform ﬂow is from top to bottom in direction of the negative Y axis. These force averages are obtained by integration of CL and CD . In forward ﬂight conditions. The main ﬂapping direction is still around the Y axis. With the strong variation in velocity. q·S CL = Flift . Performance The force coeﬃcients are the major parameters used to assess the inﬂuence of the diﬀerent wing motion models. In addition. When the average lift coeﬃcients of the diﬀerent kinematic models are matched. it is deemed more appropriate to scale the forces with the mean dynamic pressure itself (Bos et al. The mean dynamic pressure q is deﬁned as: 2 q = 1/2ρUref = 1/2ρ · 1 T T (U∞ + Uﬂap (t))2 + (Udev (t))2 dt. . The complete system needs to be rotated around the Zaxis in order to get a horizontal orientation of the freestream.
The radius of gyration is used as a reference crosssection for both twodimensional ﬂapping foil as well as threedimensional ﬂapping wing simulations. The corresponding coeﬃcients are obtained by averaging the dynamic pressure. 1994). the radius of gyration can be easily obtained. 4. Diﬀerent ﬂapping wing kinematic models are brieﬂy addressed. Most importantly. Two important dimensionless numbers were identiﬁed. 3/2 CL /CD . the governing NavierStokes equations are written in a rotating reference frame. The Rossby number is a way to describe the radius of curvature and thus the angular accelerations in dimensionless terms. from realistic fruit ﬂy kinematics to a fully harmonic model. In order to analyse the ﬂow around ﬂapping wing. In addition to the forces. These numbers are common in general ﬂuid ﬂow.92 Physical and numerical modelling of ﬂapping foils and wings the lifttodrag ratio is corrected for any diﬀerences in lift. (see Ruijgrok. In order to systematically study the aerodynamics around ﬂapping wings at the scale of insects. Additionally. a high lifttodrag ratio eﬀectively means low drag at equal lift. a deformation of the wing is deﬁned in order to study the eﬀects of wing ﬂexing. to design a sound framework for comparison it is necessary to dynamically scale the wing kinematics for all numerical simulations. the Reynolds and Strouhal number. has been used to assess the required power for a certain amount of lift (Wang. Using that planform. namely the Rossby number. but for ﬂapping ﬂight the deﬁnition has been slightly changed. a model wing planform has been deﬁned with an ellipsoidal planform. the average Reynolds number and the area swept by the wing result in comparable values. In addition to the rigid body rotations.7 Conclusions This chapter dealt with the physical and numerical modelling of threedimensional ﬂapping wings and twodimensional ﬂapping airfoils. which has proved to be proper reference. Therefore. which is used to dynamically scale the wing kinematics. while the drag is opposed to the ﬂapping velocity. the power factor. . 2008). the lifttodrag ratio is used to assess the ﬂapping wing performance. This ﬂexing motion is deﬁned with respect to the initial wing position and has a timevarying cosine shape. related to the wing rotation. Analysis of the vortical ﬂow around the wings and foils is primarily performed by plotting the force coeﬃcient. This is achieved by scaling the motion parameters such that the dimensionless amplitude. This leads to an extra important dimensionless number. The lift is deﬁned in vertical direction.
whereas the deviation is most likely used for levelling the forces over the cycle. 594. With increasing complexity. simpliﬁed models are compared with averaged representations of the hovering fruit ﬂy wing kinematics. To facilitate the comparison. as well as the resulting lift and drag forces were studied. For this. a Roboﬂy model and two more realistic fruit ﬂy models are considered. all dynamically scaled to Re = 110. In addition. The inﬂuence of diﬀerent wing kinematic models on the aerodynamic performance of a hovering insect is investigated by means of twodimensional timedependent NavierStokes simulations. (2008). Details of the vortex dynamics. light is shed on the eﬀect of diﬀerent speciﬁc characteristic features of the insect wing motion. vol.CHAPTER 5 A 2D investigation of the inﬂuence of wing kinematics in hovering ﬂight J. The simulation results reveal that the fruit ﬂy wing kinematics result in forces that diﬀer signiﬁcantly from those resulting from the simpliﬁed wing kinematic models. pp. the parameters of the models were selected such that their mean quasisteady lift coeﬃcient were matched. The angle of attack variation used by fruit ﬂies increases aerodynamic performance. . 341368. a harmonic model. Fluid Mech.
Dickinson.2 Inﬂuence of kinematic modelling The relevance of (experimental or numerical) simulations of insect ﬂight has been found to depend on how reliable true insect wing kinematics are reproduced. Notwithstanding the possible discrepancy between twodimensional and threedimensional ﬂow. and with the above justiﬁcation. Dong et al. In view of the excessive computational expense required for accurate threedimensional simulations. a similar amplitude range was used in the present o research. (2005) and Blondeaux et al. 1994). (2005) and Blondeaux et al. (2004) conﬁrmed that the similarities between two. This chapter deals with the evolution of the forces and the wake originated by a ﬂapping foil in hovering conditions. Both Dong et al. is resolved. (2005b) concluded that twodimensional studies overpredict forces and performances since the energyloss. In a twodimensional simulation our mesh resolution can be higher compared to a threedimensional simulation. in view of the limitation of computational resources.1 Similarity and discrepancy between two. 5. the present study was restricted to twodimensional simulations. Reynolds numbers. 5.1. (2005b) numerically investigated the wake structure behind ﬁnitespan wings at low Reynolds numbers.and threedimensional approaches are suﬃcient to warrant that a reasonable approximation of insect ﬂight can be obtained using a twodimensional approach. which is present in three dimensions. First. Wang et al. Wang et al. which can then be investigated more thoroughly in three dimensions. twodimensional analysis has often been applied to obtain insight into the aerodynamic eﬀects of choices in kinematics.94 Inﬂuence of wing kinematics in twodimensional hovering ﬂight 5. in case of advanced and symmetric rotation the forces were found to be similar in the twodimensional simulations compared to the threedimensional experiments. airfoil crosssection. Secondly it was observed that in both simulations and experiments the leadingedge vortex did not fully separate for amplitudetochord ratios between 35 (Dickinson & G¨tz. 1993. (2004) compared threedimensional Roboﬂy results with twodimensional numerical results. twodimensional simulations are performed to get insight in the complicated ﬂow structures. They observed that the ﬂapping wings with low aspect ratio generates threedimensional vortical structures as was mentioned by Lighthill (1969). etc.1 Introduction In order to investigate the full ﬂow around a threedimensional ﬂapping wing.1. (2004) and Sane & Dickinson (2001) showed that the kinematic modelling . This showed that twodimensional simulations are useful to obtain a better understanding of the ﬂow features.and threedimensional ﬂows In a recent paper Wang et al.
It further emphasises the need to critically assess the inﬂuence of kinematic model simpliﬁcations. Also.b) and Lentink & Gerritsma (2003) numerically investigated pure harmonic translational motion with respectively small and large amplitudes. Hover et al. were used by Pedro et al. rotao tional parameters were investigated by Dickinson (1994).. (1999) and Sane & Dickinson (2001) with their Roboﬂy. diﬀerent models from literature were considered. may have on ﬂight performance. such that including rotation is essential. the results were compared with more realistic fruit ﬂy kinematics obtained from the observation of free ﬂying fruit ﬂies (Fry et al. It was also shown that wing stroke models with only translational motion could not provide for realistic results. Slightly more complex fruit ﬂy kinematic models were used by Dickinson et al. Wang (2000a. it can reveal the importance of certain speciﬁc .5. the objective of the present study is to compare the eﬀect of the available models as a whole. Using these models. in order to investigate their inﬂuence on the aerodynamics. (2004) showed that modelling the angle of attack inﬂuences the ﬂapping foil propulsion eﬃciency to a large extent. with a relatively strong linear and angular acceleration during stroke reversal. This results in the typical ‘sawtooth’ displacement and ‘trapezoidal’ angle of attack pattern of the Roboﬂy kinematic model. which is of great importance to both experiments and numerical simulations. like parameter values and stroke patterns. the eﬀect of amplitude. both the pure harmonic and the Roboﬂy model. Based on observation of true insect ﬂight. This illustrates the appreciable eﬀects which details of the wing kinematics. Besides lift enhancement at certain reduced frequencies. Wang (2000a.1 Introduction 95 signiﬁcantly inﬂuences the mean force coeﬃcients and its distribution. Lewin & HajHariri (2003) performed a similar numerical study for heaving airfoils. In addition to the harmonic models with pure translation (Dickinson & G¨tz. they found periodic and aperiodic ﬂow solutions which are strongly related to the aerodynamic eﬃciency. including wing rotation. (2003) and Guglielmini & Blondeaux (2004) in their numerical models to solve for forward ﬂight. Harmonic wing kinematics. They varied rotational parameters and showed that axisofrotation. diﬀerent kinematic models have been employed to investigate the aerodynamic features of insect ﬂight. Both studies emphasised the importance of angle of attack modelling to inﬂuence the propulsive eﬃciency.b) varied ﬂapping amplitude and frequency and showed that at a certain parameter selection the lift is clearly enhanced. For example. 1993). They concluded that the airfoil choice is of minor inﬂuence. rotation speed and angle of attack during translation are of great importance of the force development during each stroke. This leads to better insights in the consequences of simpliﬁcations in kinematic modelling. deviation. angle of attack and the timing of the latter were explored. but large amplitudes lead to an increase of lift by a factor of 5 compared to static forces generated by translating airfoils. In literature. 2003). In the present study. Lentink & Gerritsma (2003) varied airfoil shape with amplitude and frequency ﬁxed at values representative to real fruit ﬂies. Furthermore. the wing maintains a constant velocity and angle of attack during most of the stroke. Additionally. Instead of performing a parameter study within the scope of one kinematic model.
5. In brief. which solves the governing incompressible NavierStokes equations on a twodimensional computational mesh. the fourth model is a slightly simpliﬁed version of the latter. Therefore. Henderson (1995) and Williamson (1995) showed that for circular cylinders. 1994). The second model contains the kinematics implemented by Dickinson et al. which solves the NavierStokes equations under the assumption of incompressible ﬂow. (2003). At the considered Reynolds number. (1999) for their Roboﬂy at UC Berkeley (presently CalTech).1 Flow solver and governing equations To simulate the ﬂow around moving wings with predeﬁned motions the commercial CFD solver Fluent was used. the actual modelling of the insect parameters is discussed in 5. these kinematic models are constructed such that their mean quasisteady lift coeﬃcients are comparable such that our performance comparison is justiﬁed. which supports this assumption. Finally. This basis of comparison is veriﬁed aposteriori from the force results of the actual simulations. In section 5.2 Numerical simulation methods The diﬀerent kinematic models are implemented in the commercial ﬂow solver Fluent . The resulting model has been validated using stationary and moving circular cylinders and veriﬁed using harmonically moving wings. 5.3. observed fruit ﬂy model. In addition. The results of the numerical simulations obtained with the diﬀerent kinematic models are treated in section 5. Re = O(100). The twodimensional timedependent NavierStokes equations are solved using the ﬁnite volume method. in relation to aerodynamic performance.4 and concluding remarks are given in 5. This study considers four diﬀerent wing kinematic models with varying degree of complexity. based on data measured by Fry et al.2.96 Inﬂuence of wing kinematics in twodimensional hovering ﬂight features of the stroke pattern. The mass and momentum equations are solved in a ﬁxed inertial reference frame incorporating a moving mesh following the Arbitrary Lagrangian Eulerian (ALE) formulation (Ferziger & Peric. All these kinematic models are dynamically scaled at a Reynolds number of Re = 110 which corresponds to the ﬂight conditions of the fruit ﬂy. These models are implemented in a generalpurpose Computational Fluid Dynamics (CFD) code. the transient incompressible laminar . the ﬁrst model describes the wing motion using basic harmonics as derived by Wang (2000a). 2002). The third model is a representation of the real kinematics used by a hovering fruit ﬂy (Drosophila Melanogaster ). In addition.2 the numerical simulation methods are described. the transition from laminar to turbulent ﬂow occurs at Re = 180 ± 5. The outline of this chapter is as follows. assuming incompressible ﬂow which is justiﬁed since the Mach number of ﬂapping insect ﬂight is typically O(10−3 ) (see Brodsky. the ﬂow is assumed to be laminar.5.
1 Otype mesh topology with boundary conditions on Γ1 . while remeshing takes place in the outer ﬁeld Ω2 . 2000b). a symmetry boundary condition was applied at Γ3 for numerical reasons. This complicates the creation of a high quality mesh..2.2) are used. which is shown schematically in ﬁgure 5. At the body surface a noslip boundary condition is applied.56 x 0.e. an Otype computational domain is used. The intermediate interface Γ2 divides the mesh into two separate ﬁelds. Additional solver settings can be found in (Bos et al. i. In order to create this body conformal mesh (see ﬁgure 5.Γ2 and Γ3 .58 Ω1 Ω2 Γ1 Γ2 Γ3 y 0.42 Figure 5.2 Mesh generation and boundary conditions In order to compute the ﬂow around the moving airfoils.5.2 Numerical simulation methods 97 y 0. The geometric surface gradient is high. The computational domain is divided into two parts: Ω1 and Ω2 for the inner and outer mesh respectively. The body surface Γ1 is located in the centre of the computational domain. generation of a high quality mesh is not as straightforward as for a cylinder. The complete inner mesh moves according to the wing kinematics. For the wing. corresponding respectively to the inner conformal mesh (Ω1 ) and the outer mesh (Ω2 ). high cell orthogonality. The inﬂuence of this symmetry condition has been investigated and found to be suﬃciently small. Figure 5. 2008.1) and (2. It has the reference length L which corresponds to the wing chord length. such that a freestream is absent. 2003). NavierStokes equations (2. The described computational setup was thoroughly validated using the ﬂow around stationary and moving circular . the ﬂow around the wing is not aﬀected by the mesh regeneration. especially at the leading and trailing edges.2 Body conformal moving mesh around a 2% ellipsoid airfoil.4 x 0.2) a conformal mapping was applied (see Wang. appendix B).1. which is modelled as an ellipse of 2% thickness. Since the moving wing simulations concern hovering insect ﬂight. Since remeshing occurs at a distance of 25 to 30 body lengths away from the wing. The outer boundary Γ3 is located at 25L such that the inﬂuence of the far ﬁeld boundary condition is negligible (Lentink & Gerritsma. 5.
The grid resolution near the wing. The linear displacement of this cell is y and yref is the original length of this cell. simulating 18 ﬂapping periods needed approximately 10 days on one serial AMD Athlon 2500+ CPU. In order to minimise the interpolation errors from one timestep to the next it is important to analyse the inﬂuence of the relative cell displacements. it was shown that a relative displacement of 10% in both rotational and translational direction leads to accurate results with diﬀerences in drag coeﬃcients remaining below 5%. Therefore. From the relative displacements in rotational and translational direction follow the constraints for the size of the timestep in order to keep the interpolation errors within limits. up to 1 chord length. The relative displacements in rotational and translational direction are deﬁned as ǫr = and ǫy = △α αref △y 2fe Ae N △t = . amplitude and number of cells on the surface. yref yref Here α corresponds to the angular displacement of the reference cell. (2008) (appendix D) investigated the mesh and timestep independence for the nominal solver settings using harmonic wing kinematics for hovering ﬂight.. Ae and N correspond respectively to. Additionally. . Bos et al. At this mesh.98 Inﬂuence of wing kinematics in twodimensional hovering ﬂight t + dt y x Figure 5. In (Bos et al. fe . The airfoil simulations were performed on a mesh of 50000 cells with 2000 timesteps within one motion period. appendix C). the frequency. One run. Furthermore. 2008. The computational eﬀorts are acceptable: 2000 timesteps within one excitation period.. which is illustrated for the rotational motion in ﬁgure 5. cylinders (Bos et al. while αref is the original radial length of this cell. appendix C). the motion of a reference cell was investigated.3 αref + dα αref t Relative cell displacement in rotation. the size of the ﬁrst cell at the wing surface varies between 2% and 50% of the wing thickness at the leadingedge and in the middle of the proﬁle respectively. 2008.3. was 8800 (176x50) cells such that the leading and trailing edge vortices where captured with at least 1000 cells.
5.3 0. 5. 2004).82. Further details of the validation and veriﬁcation studies can be found in (Bos et al. our force distribution looks similar for both cases.5 1 0. Our forces are normalised with the maximum of the quasisteady force. The amplitude was 2. Moreover. A common procedure is to deﬁne an equivalent twodimensional geometry.7 0. (2004) (◦).8 times the chord length.3 Validation using harmonic wing kinematics The main numerical parameters.8 0. while . just as in (Wang et al.3 0. which corresponds to Re = 75. Generally.. 2004) the drag in ﬁgure 5.7 0. Comparison of lift (a) and drag (b) coeﬃcients using harmonic wing kinematics with A = 2.8 0. The mean lift and drag coeﬃcients are 0.5 1 0.2. Only just after stroke reversal our computation ﬁnds a larger lift and drag which is probably the result of diﬀerent numerical dissipation properties of both codes.5. (2004).9 1 t/T [] Drag coeﬃcient [] Lift coeﬃcient [] 0.1 0.3 Modelling insect wing kinematics 1. In similarity to (Wang et al. 1. (2004) for similar but not entirely identical conditions. are used to validate our results with those obtained by Wang et al.84.4 Comparison of force coeﬃcients between the present simulations and Wang et al.4 0.5 0 0.4 0.4(a) is deﬁned to be positive in the direction opposite to the horizontal motion.4 shows the lift and drag coeﬃcients for validation purposes.5 0 0. with a moving wing according to harmonic kinematics. a mesh size of 50000 cells and 2000 timesteps within one excitation period. 1. within the context of comparing results of diﬀerent stroke patterns. Re = 75 for the present study (•) and obtained by Wang et al. compared to 0.5 1 0 (b) 99 0. appendix C and D).44 obtained by Wang et al.2 0. which is a diﬀerence of only 2% in lift and drag and therefore.3 Modelling insect wing kinematics In order to derive the twodimensional kinematic models the threedimensional degrees of freedom need to be converted to their twodimensional counterparts. the present numerical method is proved to be accurate.6 0.1 0..47 for our simulation. Figure 5. 2008.8.6 0.9 1 t/T [] Figure 5..5 1 0 (a) 1. (2004).5 0. A twodimensional case was selected.5 0.2 0. the computations were considered to be suﬃciently accurate.
1 in terms of wing selection and model parameters.3. Re = O(100). In the present study. Wang et al. From (Sane & Dickinson.3.3.3. a diﬀerent argument for the selection of the projection . This twodimensional setup is derived in section 5.5 (see Sane & Dickinson. α(t) to the geometrical angle of attack and θ(t) to the deviation from the horizontal stroke plane. shown in ﬁgure 5. with respect to the horizontal plane and the deviation from the horizontal plane θ.3. where R is the wing span. Lentink & Gerritsma (2003) found this airfoil an acceptable choice to model insect wings at low Reynolds numbers. as is shown in ﬁgure 5. 2001).. 5. 2001.100 Inﬂuence of wing kinematics in twodimensional hovering ﬂight Figure 5. Dickinson et al.5 Illustration of the main motion directions. the angle of attack α. maintaining the characteristic aspects of the wing motion.6.. 2003).2 and 5. 1999). In this threedimensional model the three degrees of freedom of the wing motion are deﬁned as the angular displacement φ in the mean stroke plane. φ(t) corresponds to the stroke variation.1 Insect wing selection and model parameters The computational approach is applied to investigate the inﬂuence of diﬀerent kinematic wing motion models on the aerodynamic performance. The deviation causes a ‘ﬁgureofeight’ pattern which is present in real fruit ﬂy kinematics (see Fry et al. The dynamical scaling and the force deﬁnitions are described respectively in 5. The twodimensional projection is to be deﬁned at a representative spanwise location such that the motion is conﬁned to an arc around the wing root. (2004) used this distance to derive their twodimensional model. The twodimensional airfoil shape is chosen to be a 2% thick ellipsoid. Birch & Dickinson (2003) found strongest vorticity at a spanwise location of 0. Therefore.65R from the wing root. The diﬀerent kinematic models are illustrated using the Roboﬂy experimental setup.
3.. based on the mean chord length. Apparently.42 · 10−4 m4 . the conversion from threedimensional angles to nondimensional displacements is given by: φ · Rg θ · Rg x= . 0 (5.6 Force deﬁnition on the twodimensional airfoil. the spanwise location was selected to be at the radius of gyration where the mean lift acts (Ellington. The centre of rotation is deﬁned in the aerodynamic centre which lies at the quarter chord point of the mean chord. For the radius of gyration the following value was obtained Rg = 0.0167 m2 the wing tip radius R=0.0882 m.1) c c where Rg is the radius of gyration. Finally. This leads to a value for the mean chord length of c = 0. 1984). 5. 2003): U= 1 T T u2 + v 2 dt.0667 m and the moment of inertia Icg =40. Lref .6396 · R. In view of providing completeness on the threedimensional setup. Both the displacement x and the deviation y have been made dimensionless with the mean chord c. 2003) was used. A deﬁnition of the mean chord length based on the moment of inertia around the wing root was proposed.082 m.254 m. 2004) the current crosssection is less than 2% closer to the wing root. the location of the wing base xbase =0.3 Modelling insect wing kinematics Fy Fr 101 Fx M xac α xcg c Figure 5.2 Dynamical scaling of the wing model Since the ﬂapping of the wings induces highly unsteady ﬂow the relevant ﬂow and motion parameters have to be scaled dynamically. When comparing this distance to the value used by (Wang et al. y= . The period of the motion is used to average the relevant ﬂow velocity (Lentink & Gerritsma. Another important parameter to be deﬁned is the reference length. the used values are for the wing surface S=0.5. the location of centre of gravity xcg =0. location (Lentink & Gerritsma. (5. Considering that the local velocity of each crosssection varies during ﬂapping. the mean lift acts nearly at the location where the vorticity is maximal.2) .
CL . (5. while the drag is taken equal to the horizontal force Fx . In addition. Both are given by u = ∂x/∂t and v = ∂y/∂t.102 Inﬂuence of wing kinematics in twodimensional hovering ﬂight Here T [s] is the period . Since the average lift coeﬃcients of the diﬀerent kinematic models are matched. From (5. where the integration is evaluated over one ﬂapping cycle. The mean dynamic pressure q is deﬁned as q = 1/2ρU 2 = 1/2ρ · 1 T T 0 The deﬁnition of the drag and lift forces is shown in ﬁgure 5. the ratio between timeaveraged lift coeﬃcient. it is deemed more appropriate to scale the forces with the mean dynamic pressure itself. The lift is equal to the vertical force Fy . 5. With the strong variation in velocity. φ and θ the threedimensional kinematic angles for the displacement and deviation. u represents the nondimensional velocity in the stroke plane and v the nondimensional deviation velocity. q·c q·c ∂x ∂t 2 + ∂y ∂t 2 dt. The average lifttodrag ratio. a high lifttodrag ratio eﬀectively means low drag at equal lift. CL = . Substituting equation (5. however. These force averages are obtained by integration of CL and CD .3) and (5. The force coeﬃcients are the major parameters used to assess the inﬂuence of the diﬀerent wing motion models. the lifttodrag ratio is corrected for any diﬀerences in lift. The lift is averaged over the complete period. while for the drag the averages are per half stroke. Therefore. where t = t/T is the dimensionless time. the forces are deﬁned as Fx Fy CD = .1) into (5. Hence. also known as the glide number in aerospace engineering. CD . We ﬁxed the Reynolds number to Re = 110.6. is used to characterise performance. deﬁned positive in the positive xdirection. .4) it can be observed that the Reynolds number Re depends solely on the frequency f for a given displacement φ(t) and deviation θ(t). CL /CD ave is chosen as an indicator of aerodynamic performance.3. and timeaveraged drag coeﬃcient. The Strouhal number St is not to be varied independently.3) 1 0 ( ∂φ )2 + ( ∂θ )2 ∂t ∂t (5.3 Force and performance indicators where CD and CL are the drag and lift coeﬃcients. Commonly the forces are made dimensionless using the dynamic pressure based on the average velocity.4) Here f = 1/T is the frequency.2) and evaluating. the following relations for the Reynolds and Strouhal numbers were derived: Re = and St = Uc f Rg c = · ν ν fc c = · Rg U 1 ( 0 ∂θ ∂φ 2 ) + ( )2 ∂t ∂t 1 .
In order to investigate the fact that the observed fruit ﬂy kinematics lacks an exact symmetry in the wing stroke pattern. a symmetrical model was constructed. referred to as the symmetric fruit ﬂy model. Two of these models. appendix A). the mean lift coeﬃcient is predicted well using this theory..7(d). are shown in ﬁgure 5. 2003) and is therefore considered as the most realistic fruit ﬂy kinematic model.5. 2008. Neither the displacement. four diﬀerent kinematic models. In view of the limitations of the quasisteady theory.3 Modelling insect wing kinematics 103 5.4 it is shown that the computed mean lift coeﬃcient of the numerical simulations are reasonably well matched for all models. Within this model the motion is identical for the downstroke and upstroke. angle of attack nor deviation is symmetric during the ﬂapping period. diﬀerent kinematic models were constructed. just as in the harmonic model. 1999). angle of attack and deviation. The ﬁrst 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. Analysing those aspects leads to a better understanding of how the fruit ﬂy may beneﬁt from kinematic features which are absent in the simpler models.4. which are not fully present in the quasisteady theory. The second model takes the wing kinematics as used in the Roboﬂy model (Dickinson et al. in section 5. is derived from measurements on real fruit ﬂies (Fry et al. Like the realistic . This model does include the deviation which results in a ‘ﬁgureofeight’ pattern.. which provides an aposteriori justiﬁcation of our choices for the model parameters. shown in ﬁgure 5.7(b) it is shown that the ﬂip from down to upstroke is postponed to the end of the translational phase which results in the ‘sawtooth’ shape of the displacement. with diﬀerent degree of complexity. 2004). displayed in ﬁgure 5. and reveals the relevance of including these aspects in theoretical models. have been analysed. Although according to Sane & Dickinson (2001) the mean drag is strongly inﬂuenced by the unsteady ﬂow physics.7(a). The third model represents the actual fruit ﬂy kinematics as observed in experiments and the last one was a modiﬁcation of the latter. However..7(c).3. The deviation is zero. The third model. The displacement. In order to facilitate the comparison the model parameters are chosen based on matching the mean quasisteady lift coeﬃcient (Bos et al. Large accelerations at stroke reversal are the result. The characteristic shapes of each model are described. Subsequently they are used to investigate the inﬂuence of the models on the force histories and the performance in section 5. In ﬁgure 5.. Using quasisteady theory. such that their quasisteady lift coeﬃcients are matched within 1%. chosen to investigate the eﬀect of symmetry in the wing motion. the pure harmonic motion and the Roboﬂy experimental kinematics have appeared in literature. For the symmetric models this force is equal to the resultant force. the diﬀerence between predicted and simulated values is expected to exceed this 1% tolerance.4 Diﬀerent wing kinematic models Since the main purpose of this study is to investigate the inﬂuence of wing kinematics on the aerodynamic performance during hovering fruit ﬂy ﬂight.
5 0.1T to t = 0. θ[◦ ] 100 80 60 40 20 0 20 40 60 80 0 (d) 0.7 Roboﬂy model.9 1 t/T [] Kinematic angles of the diﬀerent kinematic models.1 0.6 0.8 0.7 0.9 1 t/T [] φ. fruit ﬂy model this symmetric model includes a timedependent deviation such that the observer sees a ‘ﬁgureofeight’ pattern of the wing. (d) simpliﬁed fruit ﬂy model. (c) fruit ﬂy model. This ‘trapezoidal’ shape of α is characteristic for the Roboﬂy and may be inﬂuencing the performance. The most obvious peculiarity of the realistic fruit ﬂy models is the extra ‘bump’ in angle of attack just after stroke reversal.2 0.4 0.7 0. •: displacement angle φ.3 0.2 0.9 1 t/T [] 20 0 20 40 60 80 0 (b) 0.6 0.8 0. α and θ for each model it becomes possible to identify certain important diﬀerences.2 0. Neither of those last two realistic kinematic models can be described by using simple analytical functions without losing signiﬁcant detail.4 0. φ. α.4T the angle of attack ﬂattens at a value of almost 40◦ .1 0. The Roboﬂy initially has a larger gradient in time of the angle of attack compared to the harmonic case. : deviation angle θ. ◦: angle of attack α. .6 0.3 0.1 0.5 0.6 0.104 100 80 60 40 φ.5 0. θ[◦ ] 20 0 20 40 60 80 0 (c) 0. During translation from about t = 0.4 0.1 0. (b) Figure 5. θ[◦ ] 0. α. Although the Roboﬂy model clearly shows similarities with the fruit ﬂy models the latter has some typical additional features.7 0.3 0.5 0. When comparing the motion parameters.7 0.8 0.3 0.4 0.7(a) and (b).8 0. θ[◦ ] 20 0 20 40 60 80 0 (a) Inﬂuence of wing kinematics in twodimensional hovering ﬂight 100 80 60 40 φ. see ﬁgure 5. α. α.9 1 t/T [] 100 80 60 40 φ. (a) Harmonic model.2 0.
In order to assess the eﬀect of these kinematic features in isolation. The ﬁrst is an overall comparison of the complete kinematic models. The diﬀerences of the obtained mean lift coeﬃcient are signiﬁcantly smaller than the diﬀerences in lifttodrag ratios. as baseline. This last model is used to investigate the inﬂuence of deviation on the force histories and performance. This implies that strong translational and rotational accelerations occur at stroke reversal. the harmonic model. the Roboﬂy model. the realistic fruit ﬂy model and the simpliﬁed fruit ﬂy model. It is also observed that the deviation is negative for a certain period during the upstroke. in terms of actual vortex dynamics.7(d). 5. The harmonic and Roboﬂy models lack deviation. the comparison is made using the simplest model.4. which characterises aerodynamic performance. In the second study the eﬀect of the characteristic features identiﬁed above. but instead of ﬂattening α. the fruit ﬂy wing α descends to the ‘bump’.4 Results and Discussion In the previous section it was observed that the most interesting aspects of the Roboﬂy kinematic model are the ‘sawtooth’ displacement and the ‘trapezoidal’ angle of attack. as well as the resulting lift and drag histories are studied in 5. The mean drag. are considered in more in detail.1. During stroke reversal the gradient of α matched the harmonic model closer than the Roboﬂy with its high gradients. Therefore. It follows the same high angular velocity. After the ‘bump’ the angle of attack more or less matches the plateau found in Roboﬂy but starts to increase earlier.7(c)). the deviation of the realistic fruit ﬂy is averaged to derive the simpliﬁed fruit ﬂy model. as well as the average lifttodrag ratio.4. extra ‘bump’ in angle of attack and the presence of deviation. Therefore.1 the mean force coeﬃcients are given for the four complete models. The deviation of the fruit ﬂy model is asymmetric during the complete cycle. The more realistic fruit ﬂy models are characterised by a ‘bump’ in angle of attack and the presence of deviation. Hereto this baseline model is subsequently modiﬁed by adding respectively the ‘sawtooth’ displacement. and lift coeﬃcients are given.4 Results and Discussion 105 compared to the Roboﬂy (ﬁgure 5. for each halfstroke.7(b) and (c)). Results of two comparative studies were presented. which is described in section 5.1 Overall model comparison In table 5. the conclusions on the performance comparison are considered to be signiﬁcant. but also during each half stroke (ﬁgure 5. see ﬁgure 5. The results of this comparison.4.‘trapezoidal’ angle of attack. This is likely to inﬂuence the performance since the eﬀective angle of attack is altered due to deviation. so no ‘ﬁgureofeight’ is present. The mean drag for the harmonic and Roboﬂy models is substantially higher . the harmonic model. 5.5.2.
8 0.1 0.8 0. .466 1.448 1.6 0. : realistic fruit ﬂy model. kinematic model harmonic Roboﬂy realistic fruit ﬂy simpliﬁed fruit ﬂy Table 5.387 1.5 0.577 1.7 0.1 0.9 1 t/T [] Figure 5.848 2.417 1.9 1 t/T [] Drag coeﬃcient [] 8 6 4 2 0 2 4 6 8 0 (b) 0.6 0.1 0.106 Inﬂuence of wing kinematics in twodimensional hovering ﬂight 5 4 Lift coeﬃcient [] 3 2 1 0 1 0 (a) 0. ◦: Roboﬂy model.5 0. 8 6 Drag coeﬃcient [] 4 2 0 2 4 6 8 0 (a) 0.3 0.540 1.6 0.4 0.4 0.5 0.596 CL /CD ave 0. ◦: Roboﬂy model.4 0. : realistic fruit ﬂy model. •: harmonic model.8 0.2 0.4 0.2 0. ▽: simpliﬁed fruit ﬂy model.805 0.7% −8.1 CL 1.3 0.9 1 t/T [] Lift coeﬃcient [] 5 4 3 2 1 0 1 0 (b) 0. ▽: simpliﬁed fruit ﬂy model.839 2.1 0.8 Lift coeﬃcient histories of the baseline kinematic models.8 0.7 0.5% Timeaveraged force coeﬃcients using the complete baseline models.012 −CD up 1.483 1.6 0.115 −29% −49% baseline −1.9 1 t/T [] Figure 5.7 0.335 1.5 0.3 0.0% baseline −5.454 −3.7 0.3 0. •: harmonic model.2 0.9 Drag coeﬃcient histories of the baseline kinematic models.132 1.6% CD down 1.2 0.
454 1 2 3 (d) x c [−] 0 3 2 1 x c [−] 0 1 2 3 Figure 5. (d) symmetric fruit ﬂy model. (a) harmonic model. (c) realistic fruit ﬂy model.5.10 Force vectors during each halfstroke.417 1 2 3 (b) x c [−] 0 3 2 1 x c [−] 0 1 2 3 ¯ CL = 1.483 3 2 1 (a) ¯ CL = 1.54 3 2 1 (c) ¯ CL = 1. (b) Roboﬂy model.4 Results and Discussion 107 ¯ CL = 1. .
49%. Therefore. compared to the realistic fruit ﬂy model. the fruit ﬂy models perform better than the less complex models. it can be observed that within the model assumptions. It can be seen in ﬁgure 5. ﬁgure 5.11 Vorticity contours around a ﬂapping airfoil.108 Inﬂuence of wing kinematics in twodimensional hovering ﬂight (a) harmonic model (b) realistic fruit ﬂy model Figure 5. the mean drag coeﬃcient of the simpliﬁed fruit ﬂy is not symmetric.10 (force vectors).1. the mean drag contribution of the leadingedge vortices (LEV) is higher. compared to the fruit ﬂy models.10.11 shows the vorticity contours of the realistic fruit ﬂy model compared with the harmonic model. These performance increases are the result of the lower drag coeﬃcients in both fruit ﬂy models due to certain beneﬁcial . The diﬀerent kinematic patterns are also illustrated in ﬁgure 5. the drag during the upstroke is about 57% higher than during the downstroke. corresponding to negative vorticity values). When comparing the lifttodrag ratios in table 5. Vorticity contours are shown for t = 0. the realistic fruit ﬂy model results in a signiﬁcant decrease in drag of 29% at comparable lift.1T (blue: clockwise. Furthermore. The ‘sawtooth’ shaped Roboﬂy displacement could possibly play an important role as is discussed in the next section. Compared to the harmonic model. which shows the resultant force vectors during a full stroke for those baseline kinematic models. which is attributed to the complex vortex dynamics.7(c). i. This drag increasing eﬀect is even larger in case of the Roboﬂy model due to the ‘trapezoidal’ angle of attack. the average value during a complete stroke matches the mean drag coeﬃcient obtained with the realistic fruit ﬂy model.e. Figure 5. Nevertheless. The decrease in eﬀective angle of attack in the realistic fruit ﬂy model is also enlarged by the presence of the ‘bump’. This is also illustrated in ﬁgure 5.8 and 5.7(a) that the eﬀective angle of attack is higher in the harmonic case.9 (lift and drag histories) and ﬁgure 5. The diﬀerence with the Roboﬂy model is even larger.
13(b). This leads to a larger shear layer to form a stronger vortex.1T for the harmonic model and the one with the appended ‘sawtooth’ shaped displacement. see ﬁgure 5.1T and t = 0. The current results provide insight into the eﬀects of certain speciﬁc kinematic features. In order to determine the eﬀect of this shape the harmonic model is extended by this ‘trapezoidal’ angle of attack. the purely harmonic model was appended with the Roboﬂy displacement and the results were compared with the ones obtained using the original harmonic model. From ﬁgure 5.13. This leads to the nonzero mean horizontal force along a complete . The lift peaks are almost equal but the drag peaks are signiﬁcantly larger for the ‘sawtooth’ case. A surprising and unexpected observation is the asymmetry in the periodic force distribution for the ‘trapezoidal’ angle of attack notwithstanding the symmetry of the kinematics. one has to be cautious when extrapolating these results to real ﬂying ﬂies since in reality not every ﬂapping period displays exactly the same kinematic proﬁle. the Roboﬂy uses a ‘trapezoidal’ shape for the angle of attack.5. Next. 5. the force histories during one full stroke are shown in ﬁgure 5. Figure 5. at t = 0.1T . The stronger LEV at the beginning of the downstroke in the ‘sawtooth’ case is most likely caused by the higher velocity gradient. The results are compared with those obtained with the original harmonic model. the drag peak is larger than the lift peak. In addition. The lift and drag coeﬃcients are plotted in ﬁgure 5. On the other hand.4. respectively.3% with respect to the harmonic case. see ﬁgure 5.14(a) and (b) the vorticity contours are plotted at t = 0.2 Kinematic features investigation Inﬂuence of ‘sawtooth’ displacement used by the Roboﬂy The ‘sawtooth’ shaped displacement of the Roboﬂy is investigated in isolation to assess its inﬂuence on the force histories and the aerodynamic performance. which are repeated since the motion is symmetric. the individual inﬂuences of the diﬀerent interesting kinematic shapes are studied.13 it is observed that compared to the harmonic model the global force histories look similar. The larger mean drag is reﬂected in the integrated values in table 5.4 Results and Discussion 109 kinematic features.15. at the end of the halfstroke the wing decelerates faster in the ‘sawtooth’ case which results in a lower strength in the LEV. Inﬂuence of ‘trapezoidal’ angle of attack used by the Roboﬂy In combination with the ‘sawtooth’ displacement.2. Due to this larger drag during each stroke.2. In ﬁgures 5. This also explains the larger mean drag compared to the harmonic model which can be read from table 5. Since the wing orientation is almost vertical. From ﬁgure 5. Therefore.14 it can be seen that the LEV is stronger for the ‘sawtooth’ case which explains the higher drag peak.12(a) shows the force vectors acting on the wing during the up and downstroke.12(a) for the force vectors.4T . However. which shows a decrease of 24. Two force peaks are observed close to t = 0. the ‘sawtooth’ shaped displacement leads to a lower lifttodrag ratio.
(c) harmonic model with ‘extra bump’ α. . (d) harmonic model with ‘deviation’ θ. (b) harmonic model with ‘trapezoidal’ α.366 3 2 1 (a) ¯ CL = 1.110 Inﬂuence of wing kinematics in twodimensional hovering ﬂight ¯ CL = 1. (a) harmonic model with ‘sawtooth’ φ.12 Force vectors during each halfstroke.323 1 2 3 (d) x c [−] 0 3 2 1 x c [−] 0 1 2 3 Figure 5.351 1 2 3 (b) x c [−] 0 3 2 1 x c [−] 0 1 2 3 ¯ CL = 1.483 3 2 1 (c) ¯ CL = 1.
8 0.6 0. Lift (a) and drag (b) histories to investigate the inﬂuence of the ‘sawtooth’ displacement compared to the harmonic model.3 0.7 0.3 0.2 0.1 0. corresponding to negative vorticity values) . Vorticity contours are shown for t=0.4 0. ◦: harmonic α.1 0. •: harmonic φ.θ.14 Vorticity contours around a ﬂapping airfoil.6 0.8 0.2 0.5. θ and Roboﬂy φ.5 0.4 0.4 Results and Discussion 111 6 Drag coeﬃcient [] 0.α.9 1 t/T [] (b) Lift coeﬃcient [] 4 2 0 2 4 6 0 (a) 6 4 2 0 2 4 6 0 0.7 0. (a) harmonic model (b) harmonic model with ‘sawtooth’ displacement Figure 5.9 1 t/T [] Figure 5.1T (blue: clockwise.5 0.13 Lift and drag coeﬃcients.
8%) CD down −CD up 1.3 %) (+15.1 0. θ + Roboﬂy φ φ. harm. .930 0. harm. α and θ α.7 0.323 CL (baseline) (−7.5 0.733 1.112 Inﬂuence of wing kinematics in twodimensional hovering ﬂight kinematic model harm. φ and Roboﬂy α.807 1.8 0.15 Lift and drag coeﬃcients.221 1.9 1 t/T [] (b) Lift coeﬃcient [] 4 2 0 2 4 6 0 (a) 6 4 2 0 2 4 6 0 0. •: harmonic φ.6 %) (−8.7 0.848 2.9%) (−8.3 0.3 0.2 Timeaveraged force coeﬃcients to investigate the inﬂuence of kinematic shapes.9 1 t/T [] Figure 5. θ + fruit ﬂy.4 0.969 1.608 0. harm. α φ. θ 1.4 0. φ. 6 Drag coeﬃcient [] 0.302 1.5 0.α.738 (baseline) (−24.0%) (−10.θ.483 1. ◦: harmonic α. α + fruit ﬂy.6 0. Each characteristic shape is varied with respect to the harmonic motion model.8 0.537 0.2 0.2 %) Table 5.240 2. Lift and drag histories to study the inﬂuence of the ‘trapezoidal’ angle of attack compared to harmonic model.483 1.250 2. θ + Roboﬂy α φ.3 %) (−33. harm.6 0.2 0.804 0.9%) (0.366 1.351 1.1 0.776 CL /CD ave 0.839 2.
see ﬁgure 5. Since large angle of attacks cause high velocity gradients over the leadingedge. one observes stronger and more pronounced vortices in the wake of the ‘trapezoidal’ case. Using ﬁgure 5.2. From ﬁgure 5.15 it is clear that at the beginning of a stroke the lift peak of the ‘trapezoidal’ case is larger. This leads to a signiﬁcant performance decrease of 33.12(c) . at the end of each stroke. the angle of attack is larger at the early start of a stroke compared to the harmonic model. compared to the harmonic model. the mean lift is slightly decreased whereas the mean drag is increased.12(b).16 Vorticity contours around a ﬂapping airfoil. This can be explained as follows. To make comparison plausible the symmetric ‘bump’ variation used in the simpliﬁed fruit ﬂy model is used to compare results with the harmonic model. Taking a closer look at ﬁgure 5. In the ‘trapezoidal’ case the wing reaches the maximum angle of attack earlier in the stroke.16 this is illustrated at the beginning of the upstroke using vorticity contours.4 Results and Discussion 113 (a) harmonic model (b) harmonic model with ‘trapezoidal’ α Figure 5. Another interesting result is the low second peak in the lift. This results therefore in a lower second peak since the LEV has decreased in size and strength. Vorticity contours are shown for t=0. The LEV is larger in case of the ‘trapezoidal’ angle of attack.17(b). see table 5. since the complex vortex dynamics are nonlinear and asymmetric. corresponding to negative vorticity values) stroke cycle. Altogether. Inﬂuence of extra ‘bump’ in angle of attack used by the fruit ﬂy The fruit ﬂy models are subject to an extra ‘bump’ in angle of attack. larger vortices occur in the beginning of a stroke. This could indicate a larger amount of vortex shedding during the period when the angle of attack is nearly constant. Figure 5.3% due to the ‘trapezoidal’ angle of attack variation. the force distributions are not.5. Although this model is symmetric. Therefore.6T (blue: clockwise.
This is also the main reason for the lower drag during the downstroke. It is also noted that this case results in asymmetric force distributions as was the case when using the ‘trapezoidal’ angle of attack.20 shows the vorticity at the beginning of the upstroke at the time of the ‘bump’. From ﬁgure 5. Vorticity contours are shown for t=0. The decrease in eﬀective angle of attack as a result of the ‘bump’ is considerable compared to the harmonic case. From table 5.18 it is observed that the extra ‘bump’ generates an extra lift peak at the beginning of the downstroke. This causes the loss in lift just after stroke reversal in case of the ‘bump’ angle of attack compared to the harmonic model. On the other hand the drag is slightly increased during the upstroke such that the mean lifttodrag ratio is still increased with more than 15. The change in angle of attack due to the extra ’bump’ is shown when ﬁgure 5. However.2 it is seen that using this feature the mean lift does not change signiﬁcantly. the drag during the downstroke is very much aﬀected. compared to the harmonic case. present in the . In ﬁgure 5. the LEV provides nearly complete lift since the wing orientation is approximately horizontal. A decrease of at least 30% in mean drag is found. Figure 5.19(a) and (b) are compared.6%.17 Vorticity contours around a ﬂapping airfoil.18 the lift and drag forces are shown for the harmonic model with and without the symmetric ‘bump’ in angle of attack. corresponding to negative vorticity values) shows the force vectors during up and down stroke. The LEV is larger compared to the case with the extra ‘bump’ in angle of attack. Inﬂuence of wing deviation used by the fruit ﬂy The last important characteristic of the kinematics is the deviation.114 Inﬂuence of wing kinematics in twodimensional hovering ﬂight (a) harmonic model (b) harmonic model with ‘trapezoidal’ α Figure 5.4T (blue: clockwise. The same was found for the Roboﬂy case. for the case with the ‘bump’ in angle of attack. Therefore.
4 0.2 0.θ. corresponding to negative vorticity values) . ◦: harmonic α.5. •: harmonic φ.19 Vorticity contours around a ﬂapping airfoil.7 0. φ and fruit ﬂy α.1T (blue: clockwise.5 0.6 0.2 0.1 0.4 0.8 0.5 0.6 0. Vorticity contours are shown for t=0.3 0. (a) harmonic model (b) harmonic model with ‘extra bump’ in α Figure 5.3 0.α. Lift (a) and drag (b) histories to study the inﬂuence of the extra ‘bump’ in angle of attack.9 1 t/T [] (b) Lift coeﬃcient [] 4 2 0 2 4 6 0 (a) 6 4 2 0 2 4 6 0 0.9 1 t/T [] Figure 5.8 0.1 0.4 Results and Discussion 115 6 Drag coeﬃcient [] 0.7 0.18 Lift and drag coeﬃcients.
8% and the mean drag is almost not aﬀected by the presence of deviation.1T .2. The inﬂuence of the deviation is relatively large since the deviation increases the eﬀective angle of attack considerably just after stroke reversal.20 Vorticity contours around a ﬂapping airfoil. The mean lift is decreased by 10. Compared to the harmonic model. It is also revealed that the force distributions remain symmetric. The ﬂow dynamic mechanism for this is shown in the vorticity visualisations of ﬁgure 5. about 2% − 4% diﬀerence in both strokes. Vorticity contours are shown for t=0. which is higher compared to the harmonic case. At the end of a stroke the wings move up again which leads to a decrease in eﬀective angle of . the deviation causes a slightly stronger LEV at t = 0.116 Inﬂuence of wing kinematics in twodimensional hovering ﬂight (a) harmonic model (b) harmonic model with ‘extra bump’ in α Figure 5. This motion perpendicular to the stroke plane is illustrated in ﬁgure 5. Since deviation could introduce a large velocity component perpendicular to the stroke plane. see table 5.6T (blue: clockwise. the eﬀective angle of attack is highly aﬀected. However.6T ) and end (t = 0.12(d) which also shows the force vectors.1T and t = 0. described by the wing tip instead of wing motion solely in the stroke plane.21 shows the force coeﬃcients during one ﬂapping period with deviation added to the harmonic model. The mean lift and drag are not strongly inﬂuenced by the deviation.22 which shows the vorticity at the beginning of the stroke. Figure 5. Just after stroke reversal a lift peak occurs. corresponding to negative vorticity values) realistic and simpliﬁed fruit ﬂy model. This deviation causes a ‘ﬁgureofeight’ pattern. The large inﬂuence of the deviation on the variation of the lift force is observed at the start (t = 0.9T ) of each stroke. 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.4T and t = 0.
2 0.22 Vorticity contours around a ﬂapping airfoil.4 0.2 0.9 1 t/T [] Figure 5.21 Lift and drag coeﬃcients.5 0. ◦: harmonic α. Lift (a) and drag (b) histories to study the inﬂuence of the deviation compared to harmonic model.6 0.9 1 t/T [] (b) Lift coeﬃcient [] 4 2 0 2 4 6 0 (a) 6 4 2 0 2 4 6 0 0. •: harmonic φ.5 0.4 Results and Discussion 117 6 Drag coeﬃcient [] 0. Vorticity contours are shown for t=0.5. corresponding to negative vorticity values).7 0.α.1 0.1T (blue: clockwise. φ and fruit ﬂy θ.6 0.4 0.1 0. .8 0.7 0.8 0. (a) harmonic model (b) harmonic model with deviation Figure 5.3 0.3 0.θ.
. with diﬀerent complexity. This leads to the suggestion that a fruit ﬂy may use the deviation to level the wing loading over a ﬂapping cycle. the deviation is levelling the force distributions while the mean lift and drag are almost unaﬀected. attack. The third model represents the actual fruit ﬂy kinematics as observed in experiments and the last one is a modiﬁcation of the latter. 2004). Two of these models. corresponding to negative vorticity values).118 Inﬂuence of wing kinematics in twodimensional hovering ﬂight (a) harmonic model (b) harmonic model with deviation Figure 5.6T (blue: clockwise.23(a) and (b) show LEV’s of comparable strength for both cases.5 Conclusions The eﬀect of wing motion kinematics on the aerodynamic characteristics of hovering insect ﬂight was investigated by means of twodimensional numerical ﬂow simulations. The fruit ﬂy models are characterised by a ‘bump’ in angle of attack and the presence of deviation. chosen to investigate the eﬀect of symmetry.23 Vorticity contours around a ﬂapping airfoil. Figure 5. The most prominent aspects of the Roboﬂy kinematic model are the ‘sawtooth’ displacement and the ‘trapezoidal’ angle of attack. Summarising. The results of the present twodimensional study may provide useful insights in the understanding of real threedimensional insect ﬂight (Wang et al. Threedimensional studies are needed to investigate to what extent this eﬀect is also present in real insect ﬂight. To facilitate the comparison all models are dynamically scaled at Re = 110 and constructed such that their mean quasisteady lift . 5. pure harmonic motion and Roboﬂy experimental kinematics have appeared in literature. have been analysed using twodimensional timedependent NavierStokes simulations. Vorticity contours are shown for t=0. Four diﬀerent kinematic models.
may lead to drag reduction or force levelling compared to harmonic kinematics. First. The trend that the fruit ﬂy kinematics increases aerodynamic performance agrees well with the predictions of the quasisteady theory. caused the LEV to separate during the translational phase. extra ‘bump’ in angle of attack and the presence of deviation. To investigate which aspects of the kinematic shapes are the most important.6%. It was shown that the diﬀerence in performance in terms of mean lifttodrag ratio between the diﬀerent kinematic models was signiﬁcant. This leads to a signiﬁcant decrease in drag which improves aerodynamic performance in the sense of lifttodrag ratio by 15. Hereto the harmonic model was extended by respectively the ’sawtooth’ displacement.3%. they were compared to the harmonic model. In particular they indicate that kinematic features. During the beginning of the up and downstroke the ‘bump’ decreases the angle of attack such that the wing orientation is almost horizontal. the ‘trapezoidal’ angle of attack. It was found that the realistic fruit ﬂy wing kinematics result in signiﬁcantly lower drag at similar lift compared with the simpliﬁed wing kinematic models used in literature. The other realistic kinematic feature is the deviation. The extra ‘bump’ in angle of attack. the eﬀective angle of attack is altered such that the deviation leads to levelling of the force distribution. . found in fruit ﬂy kinematics.3%. ‘trapezoidal’ angle of attack. like the extra ‘bump’ in angle of attack and deviation. as used by the fruit ﬂy model. Also in this case large accelerations at stroke reversal lead to a decrease in lifttodrag ratio of 33. the eﬀect of the characteristic features has been studied. as well as the resulting lift and drag histories were studied. the mean drag at comparable lift is increased by 24. The results from the present study show that special features of insect ﬂight have an appreciable eﬀect on the accuracy of performance models of insect ﬂight. Due to the high acceleration during stroke reversal of the ‘sawtooth’ shaped amplitude. The actual vortex dynamics. However.5 Conclusions 119 coeﬃcient was matched. This led to an increase in mean drag during each halfstroke. an overall comparison of the complete kinematic models was given. is not aﬀecting the mean lift to a large extent. The second model simpliﬁcation used by the Roboﬂy.5. which is found to have only a marginal eﬀect on the mean lift and mean drag. Therefore. but the numerical ﬂow simulations provide a more complete quantitative analysis of the ﬂow behaviour. The results showed that the ‘sawtooth’ amplitude used in the Roboﬂy model has a small eﬀect on the mean lift but the mean drag is aﬀected signiﬁcantly. The mean aerodynamic drag at equal lift of the fruit ﬂy models is about 49% lower compared to the Roboﬂy model and about 29% lower with respect to the harmonic model.
CHAPTER
6
Vortex wake interactions of a twodimensional forward ﬂapping foil
AIAA paper 2009791.
A twodimensional numerical investigation is performed to study the vortical ﬂow around a ﬂapping foil that models an animal wing, ﬁn, or tail in forward motion. The vortex dynamics and performance are studied to determine the inﬂuence of foil kinematics. The baseline kinematic model is prescribed by harmonic functions which can be characterised by four variables, the dimensionless wavelength, the dimensionless ﬂapping amplitude, the amplitude of geometric angle of attack, and the stroke plane angle. The foil motion kinematics has a strong inﬂuence on the vortex dynamics, in particular on the vortexwake 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 leadingedge vortex (LEV) is not signiﬁcantly advantageous for the force enhancement during the full stroke. Plots of eﬃciency versus the independent variable show that, for symmetric kinematics, the largest eﬃciency is achieved at an intermediate value of each variable within the parameter range considered, where periodic ﬂow occurs.
122
Vortex wake interactions of a twodimensional ﬂapping foil
6.1
Introduction
The ﬂow around a ﬂapping wing, ﬁn, or tail is highly unsteady and governed by the dynamics of the generated vortices (WeishFogh & Jensen, 1956, Dickinson et al., 2000). An experimental study (Lentink et al., 2008) showed that these shed vortices interact with each other and organise themselves, similarly to an oscillating cylinder as described by Williamson & Roshko (1988), into speciﬁc 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 ﬂow is the result of a match between the driving frequency and the natural shedding frequency which is referred to synchronisation of the ﬂow (Williamson & Roshko, 1988, Lentink & Gerritsma, 2003). The wake will be aperiodic if synchronisation of the vortexwake does not occur. The synchronisation band organisation for the ﬂapping 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 ﬂapping foil displays a variation of the natural shedding frequency as a function of angle of attack (Katz, 1981, Dickinson & G¨tz, 1993). o 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 timeaveraged lift over a complete ﬂapping period. Several studies have shown that the wing beneﬁts 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 ﬂapping motion, plunging and pitching, at a Reynolds number of Re = 150 which corresponds to the ﬂight of a small insect, e.g. a fruit ﬂy. Here only the near wake of the foil is studied. The motivation for this is that performance of a ﬂapping foil is inﬂuenced mainly by near wake dynamics. The objective of the present study is to investigate the inﬂuence of diﬀerent foil kinematics on the vortexwake structure, force coeﬃcients, and performance.
6.2
Flapping foil parametrisation
The baseline kinematic model is based on harmonic motion, such as used by Wang (2000b,a). The ﬂow around a ﬂapping foil and the foil kinematics can be characterised by dimensionless parameters. The method used to make the governing 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 inﬂuence of diﬀerent foil kinematic parameters on the vortexwake 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 amplitude of translation A [m], the amplitude of the sinusoidal foil rotation Aα (◦ ), the forward velocity of the foil U∞ [m/s], the chord length of the foil c [m] and the stroke plane angle β (◦ ). The deﬁnition of the dimensionless parameters is schematically illustrated in ﬁgure 6.1 and described in more detail below. The dimensionless wavelength λ∗ represents the number of chord lengths travelled during one ﬂapping period: U∞ . λ∗ = fc The dimensionless amplitude A∗ represents the ratio of amplitude of the foil translation and the chord length of the foil: 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∗ and the dimensionless wavelength λ∗ : A∗ fA = ∗ . St = U∞ λ It corresponds to the maximum induced angle of attack Aαind at midstroke due to the translation of the ﬂapping motion of the foil. The mean velocity U [m/s] is obtained by averaging the velocity components over one ﬂapping period: U= 1 T
T 0
(U∞ + UﬂapX )2 + (UﬂapY )2 dt .
Here T [s] is the period, UﬂapX [m/s], and UﬂapY [m/s] the velocity of the foil kinematics in X and Y directions respectively. The timeaveraged Reynolds number Re becomes Uc , Re = ν where ν [m2 /s] is the kinematic viscosity and changed for every computation to match the Reynolds number. For the basic model the dimensionless wavelength λ∗ , the dimensionless amplitude A∗ , the amplitude of geometric angle of attack Aα , and the stroke plane angle β are chosen as independent variables. The Reynolds number is kept constant at Re = 150. In order to study the inﬂuence of the kinematics, each parameter is varied from the baseline model, deﬁned by λ∗ = 6.8, A∗ = 1.5, Aα = 15◦ , and β = 90◦ . The dimensionless wavelength was varied from λ∗ = 24, 20, 12, 10, 7.9, 6.8, 6.3, 6.0, 5.7, 5.3, 4.5, 4.0 to 3.0. The dimensionless amplitude is varied within the range of 0.5 ≤ A∗ ≤ 3.0 with a 0.5 increment. The amplitude of angle of attack varies from 0◦ ≤ Aα ≤ 45◦ with a 15◦ increment. The inﬂuence of the stroke plane angle is
the angle amplitude Aα . see ﬁgure 6. 2 Using q. qc CM = M . . In the cases that the stroke plane angle diﬀers from 90◦ . Cl = CL · sinβ + CD · cosβ . qc2 Projecting the lift coeﬃcient CL and drag coeﬃcient CD onto the y. (b) The twodimensional relation between two inertial coordinate systems. The xyplane is tilted at an angle β.1.and xaxes we obtain the foil lift coeﬃcient Cl and the foil drag coeﬃcient Cd respectively. The downstroke phase is ﬁlled by dark blue and the upstroke by light blue. the resulting ﬂapping motion is asymmetric.3 Force coeﬃcients and performance In this study two inertial coordinate systems are used.124 Vortex wake interactions of a twodimensional ﬂapping foil Y β 90 − β ◦ Aα 2A∗ · sin(β) U∞ 2A (tan−1 )2Stsin(β) y λ∗ x β X (a) (b) Figure 6. investigated for two diﬀerent angle amplitudes Aα = 15◦ and 45◦ in combination with 15◦ ≤ β ≤ 90◦ with a 15◦ increment. (a) Illustration of the foil parameters in forward ﬂight: the dimensionless wavelength λ∗ . The XY plane has the Xaxis in the direction of the freestream velocity and the Y axis in vertical direction. The drag force D is the component of the total aerodynamic force parallel to the forward velocity of the foil and is positive when directed in the positive Xdirection. 6. the stroke plane angle.1 Schematic illustration of the foil kinematics in forward ﬂight. In this frame of reference the observer is ﬁxed relative to the undisturbed air. the force and moment coeﬃcients are scaled using the average dynamic pressure q [N/m2 ]: 1 2 ρU . In the present study. and the stroke plane angle β. The ﬂight direction is from right to left. the force and moment coeﬃcients are deﬁned as: q= CD = D . the dimensionless amplitude ∗ A . The lift force L is the component of the total aerodynamic force perpendicular to the forward velocity of the foil and is positive when it is in the positive Y direction. the Strouhal number St. qc CL = L .
d · Uﬂap dt − 1 T 0 (6. and (6..2. The comparative assessment of the aerodynamic performance of the diﬀerent kinematic models is based on the mechanical eﬃciency of the foil motion. 2008. The result of this conformal mapping can be seen in ﬁgure 6. (6. d [N] the foil drag. The inner Otype mesh of 50000 cells is surrounded by a ring of tetrahedral cells. T [s] the ﬂapping period. Wang. U∞ [m/s] the freestream velocity. (6. 6. θ) is concentrated around the leading and trailing edges.1). The constant µ and θ correspond to confocal ellipses and hyperbolas respectively.3) Here −D [N] represents thrust. Uﬂap [m/s] the translational velocity of the foil in the stroke plane. more information on this mapping can be found in (Bos et al.1) η= Preq where 1 T T 0 Peﬀ = −D · U∞ . whereas the outer mesh is remeshed every timestep. The inner mesh is able to move. Those twodimensional simulations are performed on mesh resolutions of about 50000 cells. In order to obtain a good quality mesh.2) T Preq = − M · ωﬂap dt .4 Numerical model Cd = −CL · cosβ + CD · sinβ . The radius of the inner computational domain is chosen to be 25 chord lengths. a 2% thick ellipsoid shape with unit chord length represents the foil. A uniform grid in (µ.6. This type of grid is suitable for the problem since the vorticity is strongest near the edge of the foil. or tail and is relevant to the required power of locomotion. 125 Note that a negative drag coeﬃcient CD means thrust which is necessary in forward locomotion whereas the foil drag coeﬃcient Cd indicates the ﬂuid force that the animal must overcome for translational motion of its wing. These elliptical coordinates can be transformed to Cartesian coordinates via a conformal mapping: x + iy = cosh(µ + iθ) . The eﬃciency η [%] is the ratio between the eﬀective propulsive power Peﬀ [Nm/s] and the required power Preq [Nm/s] which are given in (6. elliptical coordinates (µ. M [N·m] the moment about the centre of rotation and ωﬂap [rad/s] represents rotational velocity of the foil. such that the inﬂuence of far ﬁeld boundary condition can be neglected.3) respectively: Peﬀ · 100% . Note that we have neglected inertial cost of mechanical work done by the foil. ﬁn.4 Numerical model In the present thesis. (2008).2). (6. . θ) are used following Bos et al. 2000a).
7 the LEV’s are shed before stroke reversal. Thrusting modes are found for 12 ≥ λ∗ because of generated LEV’s which pull the foil toward in forward direction. thrusting mode. The resulting wake patterns have been classiﬁed using a symbolic code of letters and numbers developed by Williamson & Roshko (1988) that describes the combination of pairs (P) and single (S) vortices shed during each ﬂapping cycle. and 6.1 by decreasing dimensionless wavelength which is equivalent to an increase in ﬂapping frequency at a constant ﬂight velocity. (a) The Otype body conformal mesh with a grid size of 50000 cells is moving within a ring of tetrahedral cells. 6. 6. For dimensionless wavelengths 12 ≥ λ∗ ≥ 5. 6. 6.5.e.2 The body conformal moving mesh around a 2% ellipsoid foil.2. The numerical results are provided in table 6.1 Inﬂuence of dimensionless wavelength The wake pattern and vortex behaviour are studied as a function of the dimensionless wavelength in the range of 24 ≥ λ∗ ≥ 3. The moment when a LEV is shed from the wing is deﬁned as the moment when its core passes the trailing edge. i. (b) The closeup of the mesh at the foil surface shows that the grid is concentrated around the leading and trailing edge. Our results are similar to the experimental results found by Lentink et al. The averaged aerodynamic force coeﬃcients in table 6.4 are obtained using three simulation periods.5 Results and discussion The simulations start with the ﬂuid at rest in which the initial velocity vector is zero. Thus the lift and drag are a function of the eﬀective angle of attack Aαeﬀ which leads to positive drag.1. The amount of LEV’s . At high dimensionless wavelengths λ∗ = 24 and 20 the numerical results give no strong vortices shedding from the foil in relation to the foil oscillation.126 Vortex wake interactions of a twodimensional ﬂapping foil 30 20 10 Y 0 10 20 30 0 3 20 10 X 0 10 20 30 (a) (b) Figure 6. Note that the eﬃciency is only calculated when the drag is negative. (2008) using a soapﬁlm tunnel.3.
772 1.5.440 1.580 1.577 1.002 0. The asymmetry in the lift coeﬃcient is a result of the asymmetry in wake pattern.0 127 pattern no vortices no vortices 2P+2S 2P+2S 2P+2S 2P+2S 2P+2S 2P+S 2P+S 2P P+S P+S Aperiodic CL 0.478 0.617 η 4.0 and 1.004 0. and TEV’s shed from the foil is decreasing with the dimensionless wavelength because the vortices have less time to develop and shed.320 0.186 0. 6.8 6.218 0.034 CLupstroke 0.2 Inﬂuence of dimensionless amplitude Six dimensionless amplitudes are chosen to investigate the inﬂuence of this parameter.934 3. Therefore the foil does not fully beneﬁt from the attachment of LEV’s.0 7.633 1. shear layers from the foil organise themselves into a 2S pattern. the LEV’s stay attached to the foil relatively longer at lower dimensionless wavelengths.502 0.66 13.9 6.095 1. The LEV’s increase in size and strength due to increasing eﬀective angle of attack.47 13. Therefore.494 CLdownstroke 0.43 9.1 Inﬂuence of dimensional wavelength.927 1.005 0.418 0. Nevertheless.624 1. see ﬁgure 6.71 10.719 1. For medium dimensionless amplitude A∗ = 1.7 5.16 11.5 Results and discussion λ∗ 24.712 1.925 1. so that the forces of this case are varying with relative small changes from period to period. As a result of this.5 the eﬀective angle of attack is high enough to .2 the numerical results are given for the six cases.001 0.495 1.021 CD 0.0 5.35 11.6. no vortices are formed on the foil due to the low eﬀective angle of attack.009 0.937 2. and β = 90◦ .071 0.70 12.833 1.4.48 13. At low dimensionless amplitude A∗ = 0.481 0.028 0.287 0.568 0.704 1. In table 6. The numerical results are shown for 13 diﬀerent values for the dimensionless wavelength. see ﬁgure 6. This means that the foil cannot produce lift enhancement just before the end of each halfstroke whether the LEV is shed before or after stroke reversal.0 20.002 0.0 3. Aα = 15◦ .092 0.254 2.252 0.3.302 0.786 1.241 0.3 4.342 0. A∗ = 1.003 0.185 0.06 Table 6.00 12. As a result.0 10.5.3 6.05 13.100 1. the foil produces higher lift and thrust during each halfstroke for decreasing dimensionless wavelength. the force distributions have a sinusoidal shape because the foil cannot produce force enhancement.0 12.004 0.324 0.103 0.5 4.874 2. It is observed for cases when vortices are formed on the foil that the lift changes its direction before stroke reversal. A further decrease in dimensionless wavelength results in stronger vortexwake interactions which lead to an aperiodic wake at λ∗ = 3.289 0.5.
70 13.064 CL downstroke 0.252 0.830 CL upstroke 0. λ∗ = 4. Aαeff = 49◦ (f ) Aperiodic.198 0.373 1. λ∗ = 6.151 1.5 3. Aαeff = 39◦ (c) 2P+S.71 7. λ∗ = 3. Aαeff = 43◦ (d) 2P.0.561 1.8. Aα = 15◦ . λ∗ = 24.5. Vorticity contours of various wake patterns for decreasing dimensionless wavelength λ∗ .147 1.638 1.0 2.8.5 1. All images are taken at t = 0.0 pattern 2S 2P+2S 2P+2S Aperiodic Aperiodic Aperiodic CL 0.13 Table 6. The ﬂow is from left to right.302 η 12. Aαeff = 46◦ (e) P+S.0 1. Aαeff = 53◦ Figure 6. and β = 90◦ .086 0. and β = 90◦ . λ∗ = 6.005 0.376 1.3. Aα = 15◦ . λ∗ = 5.47 9.5.770 1. Aαeff = 7◦ (b) 2P+2S.217 0.002 0.128 Vortex wake interactions of a twodimensional ﬂapping foil (a) No vortices.274 0.35T . .2 Numerical results of the kinematics for six diﬀerent dimensionless amplitudes.3 Vorticity contours for decreasing wavelength. A∗ 0. A∗ = 1. λ∗ = 6.038 0.002 0.624 1.5 2.109 0.598 1.94 8.959 CD 0.633 1.
58 Table 6.282 1.004 CL downstroke 2.47 28.5.5 0 0.3 Numerical results of the kinematics for four diﬀerent angle amplitudes.3 shows numerical results for diﬀerent angle of attack amplitudes.3. Aα 0 15 30 45 pattern 2P+S 2P+2S 2P 2P CL 0. P+S = 3.4 Force coeﬃcients to study the inﬂuence of wavelength.2 Drag coeﬃcient [] k k k k k k = 24. Aperiodic 0.348 0.0. A∗ = 1. Aα = 0◦ . A peak performance of 28.5. the reverse trend is found for thrust. Strong foilvortex interactions lead to an aperiodic wake pattern causing aperiodic force coeﬃcients. vortices with a diameter larger than chord length are formed.0. the eﬀective angle of attack is lower. results in an asymmetric 2P+S pattern.850 1. The foil rotation leads to thrust generation due to the frontal surface area for the pressure diﬀerence acting toward in forward direction (Lentink & Gerritsma.050 0.624 1. and ◦ β = 90 . However. This results in decreasing lift in each halfstroke for increasing angle amplitude.094 0.2 0.18% is obtained which is considerably larger com .6 0.3 Inﬂuence of angle of attack amplitude Table 6.8. 2P+2S = 6.5. 2P+2S = 6. and β = 90◦ .579 CD 0.5 1 1.8 1 2 0 0. Aperiodic 1. No thrust is generated for this setting.18 14.0.5.0.005 0.001 0.5 Results and discussion 129 8 6 Lift coeﬃcient [] 4 2 0 2 4 6 8 0 0.588 CL upstroke 1. form a LEV which leads to lift enhancement and thrust.252 0. The plunging kinematic model.4 t/T [] 0.5 k k k k k k = 24. nonzero angle of attack amplitude.5.3. Once the foil is allowed to rotate. 2003).633 1. For high dimensionless amplitude A∗ ≥ 2. Lift (a) and drag (b) histories of diﬀerent wake patterns for six dimensionless wavelengths. P+S = 3. λ∗ = 6. A∗ = 1.089 η 13.8. 2P+S = 5. Some of these vortices are hit by the foil during stroke reversal.6 t/T [] 0.4 0.6. 2P+S = 5. No vortices = 6.097 0.8 1 (a) (b) Figure 6.8. 2P = 4. 2P = 4.0. Aα = 15◦ .216 0. which gives positive mean lift over a period. No vortices = 6. The LEV in the upstroke is weaker than those generated in the downstroke.5 1 0. 6.
240 CLupstroke 1. At Aα = 45◦ . At this low eﬀective angle of attack the foil produces lower lift and thrust.252 0. The orientation of the mean lift depends on initial .264 0. due to the foil rotation.084 0. shear layers which are generated by the foil.053 3. This is because during the downstroke of asymmetric kinematics the foil undergoes a greater relative velocity.344 1.082 2.600 0.36  Table 6.056 0. Therefore.004 0. Numerical results of the kinematics for six diﬀerent stroke plane angles in combination with two diﬀerent angle amplitudes.262 1.4.296 CD 0.988 1.518 1.632 1.045 η 13.242 0.063 0. This is because the thrust component of the resulting aerodynamic force is high compared to the normal component. nonzero average lift exists only as a result of an asymmetry in wake pattern.624 2.120 2.5.626 1.768 CLdownstroke 1.844 3. λ∗ = 6.4 Inﬂuence of the stroke plane angle. 6.892 2.156 0.182 0.500 1.5 pared to other cases.8 and A∗ = 1.303 0. From the baseline kinematics the stroke plane angle is tilted backward by 15◦ . The diﬀerence in relative velocity between up and downstroke also aﬀects the drag contribution in a similar way.160 0.183 3.58 0.619 0.087 0.002 0.868 0.062 0.47 5.089 0.282 1. Also the negative lift in the upstroke is decreasing in magnitude. A similar trend is found for both angle amplitudes that the lift coeﬃcient is increasing for decreasing stroke plane angle during the downstroke until the ﬂow becomes aperiodic.103 0.464 0. 6.579 0.4 Inﬂuence of stroke plane angle The stroke plane angle causes an asymmetry in the kinematics. However.402 2. no signiﬁcant vortices are formed on the foil because the eﬀective angle of attack is low Aαeﬀ = 9◦ at midstroke.624 1.42 14.633 2.005 0.5 Discussion In the symmetric kinematics.5.130 Vortex wake interactions of a twodimensional ﬂapping foil β 90 75 60 45 30 15 90 75 60 45 30 15 Aα 15 15 15 15 15 15 45 45 45 45 45 45 pattern 2P+2S 2P+S 2P Aperiodic Aperiodic Aperiodic 2P 2P+S 3P+S P+3S P+2S Aperiodic CL 0.521 2.124 3. form themselves into a 2P pattern. Here the results for the inﬂuence of the stroke plane angle with two diﬀerent angle amplitudes Aα = 15◦ and Aα = 45◦ are shown in table 6.588 1.092 1. the averaged lift is mainly generated during the downstroke.
∗ λ 2π (6. Figure 6. . The peak eﬃciency of 28.4 0.6 0.4). the ratio of A∗ and λ∗ as: St = A∗ 1 > · tan(Aαgeo + Aαstall ) .6 shows plots of eﬃciency versus the independent motion parameters.2 0.5 1 P+S 2P+2S 2P 2P+S No vortices 0. There is an important limitation in forward ﬂapping locomotion. conditions. it is thought that the results could also shed light on the Micro Air Vehicle (MAV) design. Besides.6. where the timeaveraged lift coeﬃcient is plotted against wake pattern.5 Results and discussion 131 0. The grey bands indicate a symmetric wake pattern in which the nearly zero mean lift is obtained. To begin with. The vortex synchronisation diagram for all models is shown in ﬁgure 6. (b) Vortexwake synchronisation A∗ − λ∗ diagram.5(a).5(b).2 4 3. the (absolute) eﬀective angle of attack should be high enough to form a LEV in order to generate force enhancement.5 Aperiodic P+S 2P 2P+S 2P+2SNo vortices 0 0 5 10 λ∗ 15 20 25 (a) Mean lift coeﬃcient of symmetric kinematics.5 3 2. This approximately restricts the values of St. In symmetric kinematics there is an optimal value for each variable. see ﬁgure 6. (a) The mean lift coeﬃcient over a complete period of symmetric kinematics as a function of wake pattern. The results are summarised in ﬁgure 6.4) which is illustrated by the dashed line. The angle amplitude and stroke plane angle are kept constant at Aα = 15◦ and β = 90◦ . i. the wake and consequently the forces become aperiodic which will inﬂuence the stability and controllability of the MAV’s. We have added the operating conditions of insects belonging to the order Diptera.6 0.6(a).e.5 Aperiodic Present study Diptera CL 0 A∗ 2 1. (b) Vortexwake synchronisation A∗ − λ∗ diagram of our sinusoidal ﬂapping wing. Figure 6.18% could conﬁrm that the wing rotation plays an important role in the unsteady aerodynamic force production.5 Inﬂuence of the kinematics on the vortex wake pattern and force generation. (c). The dash line represents our theoretical estimate of the boundary governed by equation (6.4 0. When the wing operates outside the synchronisation region. (b) and.
(b) dimensionless amplitude.5 4 (a) 30 (b) 15 Aα = 15◦ Aα = 45◦ 2P+S 2P 2P+2S 20 10 η 2P+2S 10 η 5 2P 0 0 10 20 30 40 50 0 60 Aα 65 70 75 80 β 85 90 95 100 (c) (d) Figure 6. The results are in satisfactory agreement with the comparable experiments. 6. and stroke plane angles at the Reynolds number of 150. Although such an attempt at classifying vortex patterns can lead to confusion due to the shedding. (c) angle amplitude and (d) stroke plane angle.5 Aperiodic 3 P+S 3.5 1 1. . or merging of tiny vortices. Both plunging and rotating motions are prescribed by simple harmonic functions which are useful for exploring the parametric space despite the model simplicity. dimensionless amplitudes. it is suitable for straightening out the shedding vortices in our simulations.132 Vortex wake interactions of a twodimensional ﬂapping foil 15 15 10 10 η Aperiodic η 2P+2S 0 2 4 6 λ 8 ∗ 10 12 14 0 2P+2S 5 5 2P 2P+S 0 0.6 Inﬂuence of ﬂapping kinematics on the eﬃciency.6 Conclusions A numerical model for twodimensional ﬂow was used to investigate the eﬀect of foil kinematics on the vortex dynamics around an ellipsoid foil subjected to prescribed ﬂapping motion over a range of dimensionless wavelengths. angle of attack amplitudes. The resulting wake patterns behind the foil are categorised using the concept of Williamson & Roshko (1988). tearing. (a) dimensionless wavelength.5 A∗ 2 2. Eﬃciency as a function of the independence variables.
However. which implies that aerodynamics might select a range of preferable operating conditions. the synchronisation band was not investigated completely. .6. Since the computational costs are high and the parameters cannot be varied continuously. the present study is beneﬁcial for understanding the inﬂuence of wing kinematics on the performance characteristics.6 Conclusions 133 Optimal propulsion using ﬂapping foil exists for each variable. The conditions that give optimal propulsion lie in the synchronisation region in which the ﬂow is periodic.
.
Finally.1 Introduction To understand the aerodynamic performance of ﬂapping wings at low Reynolds numbers. the Reynolds number and the stroke kinematics. A varying Rossby number represents a variation in the radius of the stroke path and thus the magnitude of the angular acceleration. namely the angle of attack. (January 2010). Fluid Mech. Results are obtained by performing numerical simulations of the threedimensional ﬂow around a ﬂapping wing. have been studied. the threedimensional ﬂow is compared with the twodimensional studies performed on ﬂapping forward ﬂight. relevant for threedimensional ﬂapping wing aerodynamics. 7. Secondly. the Rossby number. the Rossby number is varied at diﬀerent Reynolds numbers. a ‘ﬁgureofO’ or a ‘ﬁgureofU’ pattern. the threedimensional wing kinematics is varied by changing the shape in angle of attack and by applying a deviation. Diﬀerent aspects. relevant for insect ﬂight. it is important to obtain insight into the vortex .CHAPTER 7 Vortical structures in threedimensional ﬂapping ﬂight Submitted to J. First. which may result in a ‘ﬁgureofeight’. Thirdly. the ﬂow around a dynamically scaled model wing is solved for diﬀerent angles of attack in order to study the force development and vortex dynamics at small and large midstroke angles of attack. A parameter study is performed to investigate the performance in ﬂapping ﬂight and to get insight into the vortex dynamics and force generation.
Dickinson et al. By varying the location of the centre of rotation. Bos et al.. Ellington et al. However. 2002) or ﬂapping (Dickinson et al. Dickinson. (Lentink. 1993. 1997). 2002. 2004.136 Vortical structures in threedimensional ﬂapping ﬂight dynamics and its inﬂuence on force development. using plates at diﬀerent spanwise locations to block the spanwise ﬂow. Lehmann.. which could suggest that a spanwise ﬂow may play an important role concerning the LEV stability in insect ﬂight (Ellington et al. In addition. Shyy et al. 2008) showed interesting results concerning the stability of the threedimensional leadingedge vortex depending on the Rossby number (equiv . In order to gain insight into the threedimensional ﬂow ﬁeld induced by the ﬂapping wings. Lentink & Dickinson. 1994. which rotates around its base. 2002. Wang. Lentink & Dickinson. Recent twodimensional simulations (Bos et al. Lentink & Dickinson (2009b) discussed that the stability of the LEV growth speciﬁcally might be increased by the spanwise ﬂow through the LEV core.a. 1996. a threedimensional wing was modelled which was able to ﬂap around a base of which the location can be varied. 2001. Van Den Berg & Ellington. driven by either the dynamic pressure gradient on the wing’s surface.. the inﬂuence of the revolving strength (Rossby number) and the eﬀect of the tip vortices can be studied. several twodimensional studies have been performed (Dickinson & G¨tz. 2008).. In order to investigate the vortex dynamics and the stability of the leadingedge vortex in particular. The most important feature in ﬂapping wing aerodynamics has been established to be the generation of a stable leadingedge vortex (LEV) on top of the wing. Additionally. 2004) model wing.. The spiral leadingedge vortex generated by a translating swept or delta wing is stabilised by the induced spanwise ﬂow. while a threedimensional LEV remains stably attached to a threedimensional revolving (Usherwood & Ellington. 1996). 1996. but they did not completely explain the LEV stability in their experiments.. o It was shown that the leadingedge vortex generated by a twodimensional moving foil is shed after several travelled chord lengths.. the LEV stability may be strengthened by a reduction of the eﬀective angle of attack as a result of the tip vortex generation (Birch & Dickinson. Van Den Berg & Ellington. Birch & Dickinson (2001) showed no signiﬁcant eﬀect of the spanwise ﬂow on the LEV strength and stability. 2004.. 1997). 1999. Previously conducted research addressed a possible analogy between the LEV on ﬂapping wings and the LEV generated by swept and delta wings (Ellington et al. 2008) suggested that the wing kinematics may also have a large inﬂuence on the ﬂapping performance in threedimensional hovering. Additionally. an accurate simulation method is developed to perform a CFD simulation of a threedimensional ﬂapping wing.. which is responsible for the unexpectedly large force augmentation in hovering insect ﬂight (Maxworthy. 1999. 1996. 2009b). the kinematics is varied from simple harmonics by adding a deviation and ‘trapezoidal’ shaped angle of attack. Birch et al. 2008b). Bos et al... 1979. 2009b. Birch et al. Those results indicate that threedimensional ﬂow eﬀects are essential for the LEV stability. Srygley & Thomas. Based on the discussion about LEV stabilisation due to wing revolving (Usherwood & Ellington. 2005. the centrifugal acceleration of the boundary layer or the induced velocity ﬁeld of the spiral vortex lines (Ellington et al.. 2010b).
kinematic models and the simulation strategy.1 Illustration of the wing motion and force deﬁnitions. Diﬀerent vortex identiﬁcation methods are described in 7.4. φ(t) corresponds to the stroke variation. while the conclusions are summarised in 7.2. Diﬀerent deviation patterns are investigated. this is the subject of section 7. the Rossby number is systematically varied for diﬀerent Reynolds numbers and midstroke angles of attack. a selection of geometric and kinematic parameters is made to systemically investigate the ﬂow phenomena of our interest. alent to the stroke path curvature) and the Reynolds number. the results are discussed in 7. which is still representative for true insect ﬂight. α(t) to the geometrical angle of attack and θ(t) to the deviation from the horizontal stroke plane.2. In order to show that the CFD method is accurate and eﬃcient.6. Therefore. The vortical ﬂow needs to be visualised in such a way that the resulting vortices are clearly visible. . 2008).3 brieﬂy discusses the validation and veriﬁcation of the ﬂow solver. of which the details are described in chapter 2. following the shape of ‘ﬁgureofO’. The threedimensional ﬂapping wing is modelled in order to provide a framework for comparison. 7.7.1.5 and 7. ‘ﬁgureofU’ and ‘ﬁgureofeight’.. Additionally.7. the ﬂow is solved using Computational Fluid Dynamics (CFD). section 7.2. In view of limiting computing resources.2 Threedimensional ﬂapping wing simulations 137 FY Fnormal α(t) Rtip Fspan Stroke plane O Rroot FX θ(t) φ(t) start downstroke Fdrag start upstroke midstroke FZ Figure 7. Furthermore.2 Threedimensional ﬂapping wing simulations In order to study the vortex dynamics and stability of the leadingedge vortex. In agreement with (Bos et al.2. the simulation strategy is discussed in section 7. The ﬂapping wing modelling is described in section 7. Illustration of the wing motion and force deﬁnitions. the kinematic model is extended with a ‘trapezoidal’ shaped angle of attack and a nonzero deviation is applied. which also addresses wing geometry.
2006)... respectively. a ‘sawtooth’ shaped ﬂapping angle and a ‘trapezoidal’ shaped angle of attack (Sane & Dickinson. Bos et al. Planform selection The single wing span is ﬁxed to bs = 2. 2008) resembles an harmonically varying deviation angle. the three geometric parameters important for the ﬂapping wing simulations are deﬁned: bs . 7. 2002. 2008).. such that signiﬁcant eﬀect of corrugation on the ﬂow can be neglected. 2002. 2008) showed that the eﬀect of the ‘trapezoidal’ angle of attack was most prominent. i. The average chord length of this ellipsoidal planform is found to be c = S/bs = π/4. did not inﬂuence the force development signiﬁcantly. ‘trapezoidal’ .138 Vortical structures in threedimensional ﬂapping ﬂight 7. the deviation is varied harmonically with an amplitude between Aθ = 0 and Aθ = 20◦ . angle of attack and stroke kinematics.. ‘ﬁgureofeight’.5. Reynolds number. where a = 0. 7. Dickson & Dickinson. In addition. 2003. since Lentink & Gerritsma (2003) showed that airfoil shape was of minor inﬂuence on the forces and the ﬂow ﬁeld. present in dragonﬂy wings. shortly after stroke reversal (Bos et al. while the rotational distance of the Roboﬂy was ﬁxed to 0. Lentink & Dickinson.2.0. Poelma et al. The deviation angle is the angle with respect to the horizontal stroke plane. or ‘ﬁgureofU’.5 from the wing root. such that the wing tip radius becomes Rtip = 2.7 (Sane & Dickinson. During the stroke. ﬂapping angle φ(t) and the angle of attack α(t). ﬁgure 7. From the discussion in section 7. The current research uses a model wing with an ellipsoidal shaped planform with 10% thickness. S and c.2 shows the resulting ‘ﬁgureofO’.1 it can be concluded that there is need for a detailed threedimensional numerical study to investigate the eﬀects of the Rossby number.5 and b = 1.. Kinematic models The ﬂapping wing motion is prescribed by three diﬀerent motion angles deﬁning the deviation angle θ(t).2(a). Realistic fruit ﬂy kinematics (Fry et al. 2008). Since the wing planform is chosen to be ellipsoidal. 7.0 and the chord at midspan is c = 1.1 Modelling and parameter selection In general. 2004) with an incidental ‘bump’. Depending on the variation of the deviation angle.e. Luo & Sun (2005) showed that the airfoil corrugation.2(b). the wing surface is deﬁned by S = πab. as described in chapter 4 and (Bos et al. So. 2009b. A twodimensional investigation (Bos et al. A combined deviation and ﬂapping angle variation leads to a wing tip pattern. most investigations concerning ﬂapping 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 ﬂy wing.0 are the semiminor and semimajor axes. The length scales of the corrugation are orders of magnitude smaller compared to the length scale of the separated ﬂow region or the leadingedge vortex. The hinge around which the wing is able to ﬂap is ﬁxed to a distance of 0. such that S = π/2..2(c).
θ [◦ ] [] 80 upstroke 80 60 40 20 0 20 40 60 80 0 downstroke upstroke downstroke 20 40 60 80 100 120 140 160 180 Flapping angle.2 Threedimensional ﬂapping wing simulations 139 Deviation angle.25T . The current research varies the amount of the ‘trapezoidal’ shape by varying the speed of rotation just after stroke reversal from Trot = 0. the (geometric) angle of attack is varied from α = 15◦ to α = 90◦ with increments of α = 15◦ . The current study varied the Rossby number from Ro = 3. (a) ‘ﬁgureofO’. θ [◦ ] [] 60 40 20 0 20 40 60 80 0 Deviation angle. φ [◦ ] (a) ‘ﬁgureofO’ (b) ‘ﬁgureofeight’ Deviation angle.7. (b) ‘ﬁgureofeight’.25T corresponds to fully harmonic angle of attack variation. Diﬀerent wing tip patterns as a result of the variation in deviation with a combined ﬂapping motion. φ [◦ ] upstroke (c) ‘ﬁgureofU’ Figure 7. 2009a. the Rossby and Reynolds numbers. In addition to the variation of Reynolds number from Re = 100. which is relevant for vortex induced propulsion in nature (Lentink & Dickinson.10T to Trot = 0. 500 and 1000. the ‘trapezoidal’ angle of attack. to a nearly translating wing.2.2 Diﬀerent wing tip patterns. θ [◦ ] [] 80 60 40 20 0 20 40 60 80 0 downstroke 20 40 60 80 100 120 140 160 180 Flapping angle. shape and deviation. such that Trot = 0.b). φ [◦ ] 20 40 60 80 100 120 140 160 180 Flapping angle. (c) ‘ﬁgureofU’. Ro = 130. Framework for comparison In order to design a frame of comparison it is important to keep the following three parameters ﬁxed: the dimensionless amplitude of the wing’s crosssection at . The ﬂapping angle was chosen to vary harmonically to isolate the eﬀects of the deviation. where T is the ﬂapping period.
Since the wing rotates with a rotating reference frame.5 the resulting amplitude of the crosssection at Rg becomes A∗ g ≈ 2. which is compensated by the ﬂapping angle amplitude in order to keep the average Reynolds number and the displacement at Rg comparable.1) ν where Aφ is the ﬂapping angle amplitude. whereas the angular acceleration number Cang . Force and performance deﬁnitions In order to determine the eﬀect of the diﬀerent motion and geometric parameters on the forces and performance.4) Here. are deﬁned by the wing geometry. proper deﬁnitions are necessary.2) (7. f the ﬂapping frequency. . For the baseline case. maximal values.140 Vortical structures in threedimensional ﬂapping ﬂight the radius of gyration A∗ g .1). Concerning the baseline R case.2. the wing tip radius changes. average velocity U Rg and displacement of the crosssection at the radius of gyration. (7. two diﬀerent force deﬁnitions are possible. the ﬂapping angle amplitude is determined from (7. R 1 and the area swept by the wing Aswept . The result of this scaling is a comparable average Reynolds number. the average Reynolds number. U R and A∗ = Aφ R/c. which is the main subject of the present thesis.58. If the distance of the rotation origin is varied. R π c Rtip = ARs . and the Rossby number Ro. Cang and Ccen . 2008). Using Rg = S 0 r2 c(r)dr. A∗ g = Aφ Rg /c. for every rotation radius. which is of similar order as used in the twodimensional analysis in (Bos R et al. based on the radius of gyration is deﬁned as: 4Aφ f Rg c ReRg = . 500 and 1000. where Rtip = 2. the radius of gyration is determined from the rotation origin to the tip r = 0 to Rtip . and the Rossby number Ro.5 the radius of gyration becomes Rg = 1. the hovering wing kinematics is substituted into the expressions for the angular and centripetal coeﬃcients.. c Rtip = ARs . c (7. ARs is the single wing aspect ratio. the Reynolds number at the radius of gyration ReRg . The kinematic viscosity is ﬁxed for three selected values. c the average chord length and ν the kinematic viscosity. equation (4.3) (7. In order to investigate the eﬀect of threedimensional wing kinematics in hovering ﬂight. It remains clear that both the centripetal Ccen . ReRg = 100. occurring at the R wing tip are still varying like ReR . ReRg = 100. with Rtip = 2. Using (4. On the other hand. provided that the wing kinematics and geometry are given.23).6) to ﬁnd the following: R Cang Ccen = = Ro = 2 Aφ Rg = A∗ g . depends on the wing kinematics. Therefore.
The ﬂow is solved for diﬀerent wing tip radii. In order to investigate the inﬂuence of the wing kinematics. Because the present research concerns hovering ﬂight. 2008) in chapter 4. the Rossby number Ro (due to varying rotation origin). 1994). (7.2. Additionally. This range in stroke path curvature corresponds to a changing Rossby number from Ro = 3. to study its eﬀects on the behaviour of the leadingedge vortex. the lift force is by deﬁnition vertical and thus equal to FY .5) and Fdrag = FX · sin(φ) − FZ · cos(φ).1 in relation to the motion angles. CL /CD (see Ruijgrok. the force coeﬃcients.e. the angle of attack amplitude α. including a variation in angle of attack.5 to a nearly translating wing at Rtip = 102. Lentink & Dickinson (2009b) found that most insects and ﬁsh operate at a Rossby number close to Ro = 3.0. Re = 500 and Re = 1000. (7. and translating wing. CL /CD and the 3/2 power factor.7.6) The force in spanwise direction is not used throughout our analysis. As discussed previously (Bos et al. derived by a decomposition of FX .2 to 130. the Reynolds number Re and the wing kinematic model. which seems to be a biologically convergent solution for animals moving in ﬂuids. the drag force is opposite to the motion direction. Inﬂuence of Reynolds number and angle of attack For two selected Rossby numbers Ro = 3. q = 0. also known as the glide factor. which is shown in ﬁgure 7. the midstroke angle of attack is varied from α = 15◦ to α = 90◦ with increments of α = 15◦ . under hovering conditions. A systematic overview of all cases is provided here.2 and Ro = 130. CL and CD . i. are obtained by division 2 using the average dynamic pressure. Inﬂuence of wing stroke curvature The stroke curvature is varied in order to investigate if there is a possible relation between the angular acceleration. FY and FZ in the rotating reference frame. . centripetal acceleration or the Rossby number and the forces acting on the ﬂapping wing. the threedimensional lift and drag are given by: Flift = FY . On the other hand. the Reynolds number is varied from Re = 100. Two performance indicators are used. Aθ were varied.2 Threedimensional ﬂapping wing simulations 141 in the inertial and the rotating reference frame. 7. reﬂected by Trot and the deviation amplitude.. the shape of the angle of attack variation.5ρUref . equivalent to the Rossby number. The wing tip radius is varied from fully revolving at Rtip = 2. since that force is small compared to the lift and drag. ﬂapping. the lifttodrag ratio.2 Simulation strategy and test matrix selection The following variables are varied throughout the current research. This provides insight in the force development as a function of angle of attack for a fully revolving. Therefore.
25 recovers a fully harmonic angle of attack variation.1 Simulation matrix: wing stroke curvature origin. i. Trot .1 shows an overview of the variation of the wing stroke curvature.e. the deviation. Simulation matrix to vary the rotation origin of the ﬂapping wing. ‘ﬁgureofO’. angle of attack and Reynolds number. depending on the deviation frequency. The varying deviation angle amplitude.2 shows diﬀerent deviation amplitudes Aθ . Note that the geometric angle of attack is given by αgeom = π − α.0. 2 (7. which determines the amount of ‘trapezoidal’ shape. the angle of attack is plotted in ﬁgure 7.0 5. as was already illustrated in ﬁgure 7. Aα [◦ ] 1 2T Trot ≤ t < 1 T − Trot . 2 Trot + 1 T ≤ t < T − Trot . •: Re = 500. The angle of attack at midstroke is varied from 90◦ to 15◦ together with the Reynolds number.3. − Trot ≤ t < Trot + 1 T.2. +: Re = 100. 2 T − Trot ≤ t < T. For diﬀerent values of Trot . This ‘trapezoidal’ angle of attack variation is deﬁned by Aα sin(2πf t) Aα Aα cos(2πf t) α= −Aα −Aα cos(2πf t) 0 ≤ t < Trot . which may cause a ‘ﬁgureofeight’ pattern and the ‘trapezoidal’ shape (deﬁned by Trot ) of the angle of attack variation. ‘ﬁgureofeight’ and the ‘ﬁgureofU’.7) Here Trot is the rotation duration. .0 12. such that Trot = 0. in combination with a varying. angle of attack and Reynolds number. The grid resolution was ﬁxed to 800k and the timestep was chosen corresponding to Comax = 2.0 7. 2 Table 7. it remains interesting to investigate the eﬀect of two kinematic parameters.0 6.0 Rotation origin. ◦: Re = 1000. This matrix is used to study the inﬂuence of the stroke curvature on the structure of the leadingedge vortex and corresponding forces. Reynolds number and angle of attack. may cause different wing tip patterns. Table 7.0 + + + + + + + + + + + + 102. Inﬂuence of the kinematic modelling Besides the wing stroke curvature. Rtip 4.5 +•◦ +•◦ +•◦ +•◦ +•◦ +•◦ 3.142 Vortical structures in threedimensional ﬂapping ﬂight 90 75 60 45 30 15 2.0 +•◦ +•◦ +•◦ +•◦ +•◦ +•◦ Table 7.
130 120 110 100 α [] 90 80 70 60 50 0 0.3 Angle of attack variation with a ‘trapezoidal’ shape.9 1 t/T [] Trot Trot Trot Trot = = = = 0.6 0.7 0.2 and the average Reynolds number ReRg = 100. The ﬂapping angle amplitude was ﬁxed to Aφ = 63◦ (1. The Reynolds number is ﬁxed to Re = 100.3 Flow solver accuracy In order to test the accuracy of the used ﬂow solver. Table 7.8 0. a veriﬁcation study is performed using the ﬂow around a threedimensionally ﬂapping wing. The rotation duration is varied from Trot = 0. The deviation is varied by the deviation amplitude Aθ .20 0.3 gives an overview of the performed simulations by varying .2 0. 45◦ and 60◦ .4 0. In order to show that the numerical solution is grid and timestep independent. the grid resolution to 800k and the timestep was chosen corresponding to Comax = 2. The meshes for these threedimensional simulations are constructed with GridPro using a structured approach.25 Figure 7. the midstroke angle of attack was given by Aα = 45◦ and the wing tip radius corresponds to fully revolving. The kinematics is according to the simple harmonic model.2 Simulation matrix: kinematic modelling.15 0. More detailed information on grid generation can be found in appendix A.0.20 0. by decreasing the maximum Courant number. which is shown to result in maximal lift coeﬃcients. Simulation matrix to vary the rotation duration and the deviation of the ﬂapping stroke. 7. concerning highly unsteady and vortical ﬂows.3 0.15 0.5 0. numerical comparisons are performed. A veriﬁcation is performed by varying the grid resolution (grid independence study) and the timestep size. the Rossby number was Ro = 3.7.3 Flow solver accuracy 143 0. Grid reﬁnement is uniform and the cells are clustered close to the ﬂapping wing boundary. the amount of this shape is systematically varied by Trot .25 to Trot = 0. To investigate the inﬂuence of a ‘trapezoidal’ shape in angle of attack.25 0. Two diﬀerent angle of attack amplitudes are used.10 in order to get a ‘trapezoidal’ shaped angle of attack.1 rad).10 0 + + + + 5 + Aθ 10 15 + + 20 + Trot + + + + Table 7.1 0.10 0.
2002). but preferably 800k mesh cells are desired.0 and 0. less than 4%. Summarising.0. where the timeaveraged lift and drag coeﬃcients are plotted with decreasing timestep. which varies from 2. i.5 Comax + Mesh resolution 200k 400k 800k + + + +• +• + + 1600k + Table 7.5). the smaller the timestep.5.2%. The smaller Comax .0 0.0 to 0.6(c). This may be explained by the fact that the forces are mainly dependent on the near wake. The spatial grid independence study was performed for a maximal Courant number of Comax = 1. •: Re = 1000. Table 7. 1. the value at ∞ is determined by Richardson extrapolation (Ferziger & Peric.0.0 1.6(b) shows a converging timeaveraged lift and drag coeﬃcients with increasing grid resolution. the ﬂow is periodic and the force coeﬃcients (lift and drag) appear to be close for the grid resolutions considered. the drag and lift coeﬃcients are plotted in ﬁgure 7.5 show the spatial and temporal errors in average lift and drag with the Richardson extrapolated values. In order to assess the accuracy of the ﬂow solver.0 and the temporal convergence for two grid sizes of 400k and 800k cells.4 and 7.0 to 0.4 shows that even the diﬀerences in lift and drag for 100k mesh cells and Comax = 1. the error is less than 0. with several chord lengths from the wing.0. on which the forces depend. Again. Comax . the diﬀerences are small. table 7. However.144 Vortical structures in threedimensional ﬂapping ﬂight 100k 2.0 are reasonably small. are suﬃciently ﬁne to capture the near wake. even for the largest timestep.0. from 100k to 1600k. all generated grids provide suﬃciently accurate force coeﬃcients.3 Simulation matrix: veriﬁcation. The corresponding limit cycles are shown in ﬁgure 7. but to capture the far wake vortex dynamics at least 800k cells are required. with varying mesh resolution (100k − 1600k) and timestep. The grid resolution was varied from 100k to 1600k cells and the Comax from 2.5.e. Two cases are performed for two diﬀerent Reynolds numbers. all grid resolutions considered. .4 for meshes from 100k to 1600k cells and Comax = 1. corresponding to Comax = 2. The kinematics of the threedimensional wing is simple harmonics with Aφ = 63◦ . Aα = 45◦ and Aθ = 0◦ . The timestep is reﬂected through the maximal Courant number. 2008) at least 400k. Temporal convergence is shown in ﬁgure 7. Apparently. As can be clearly seen from the ﬁgures. in order to visualise the vortices in the far wake (Bos et al.6(a) and 7.6. grid resolution and maximal Courant number. Table 7. Figure 7. +: Re = 100. In order to assess spatial and temporal convergence both timeaveraged lift and drag coeﬃcients are plotted with increasing spatial and temporal resolution in ﬁgure 7.. see chapter 2. In order to justify the choice for grid and temporal resolution for the threedimensional ﬂapping wing simulations. Veriﬁcation matrix showing the cases used for veriﬁcation purposes. 1.5 shows the errors in lift and drag with respect to the Richardson extrapolated values for both 400k and 800k with decreasing timestep (Comax = 2.5.
4 Threedimensional veriﬁcation: force coeﬃcients. Furthermore. which requires a complete velocity ﬁeld. 100k − 1600k.2 0.1 0.66 2. Lift and drag coeﬃcients for the veriﬁcation cases with varying grid size.13 Table 7. Aα = 45◦ and Aθ = 0◦ . ranging from 100k to 1600k.4 0. the temporal errors are suﬃciently small.49 0. The timestep is taken such that Comax = 1.5 0. N 100k 200k 400k 800k 1600k ǫlift [%] 3.3 0.9 t/T [] 1 3 0 100k 200k 400k 800k 1600k 2 Drag coeﬃcient [] 1 0 1 2 145 100k 200k 400k 800k 1600k 0.5 1 0.4 Error values of lift and drag coeﬃcients for varying grid sizes.7 0. Diﬀerent wellknown techniques to detect and visualise vortices are based on the velocity gradient tensor. The timestep was determined by a max Courant number of Comax = 1.7. at Comax = 2. a proper vortex identiﬁcation criterion is essential.5 0. described in this chapter.8 0.23 0.4 Vortex identiﬁcation methods for ﬂow visualisation In order to study the vortex dynamics and the stability of the leadingedge vortex in particular. namely the magnitude of vorticity ω (Lu & Shen.6 0.52 1.0.5 Lift coeﬃcient [] 2 1.8 0. 2008) . The kinematics of the threedimensional wing is simple harmonics with Aφ = 63◦ .50 0. a grid resolution of 800k in combination with Comax = 2.27 0.19 2.7 0.3 0.0.9 t/T [] 1 (a) (b) Figure 7.4 0.4 Vortex identiﬁcation methods for ﬂow visualisation 3 2. 7.12 ǫdrag [%] 3.2 0. Therefore.0 was used for all threedimensional ﬂapping wing simulations. Aα = 45◦ and A θ = 0◦ .32 1. Two diﬀerent vortex identiﬁcation criteria are discussed.0.5 0 0 0.6 0.1 0. Error values of lift and drag coeﬃcients [%] with respect to the extrapolated values for varying grid sizes. The kinematics of the threedimensional wing is simple harmonics with Aφ = 63◦ .
shear layers.082 0. 400k and 800k. a positive value of Q > 0 is a measure for any excess of rotation rate (in terms of vorticity) with respect to the strain . Aφ = 1. 1988).5 1 0.067 0. 1988) is the second invariant of the local velocity gradient tensor ∇u.. More speciﬁcally. where ω = ∇ × u. The Q criterion (Hunt et al.0 0. 2 where the rate of strain tensor S is given by S = 1 (∇u + ∇uT ) and the vorticity 2 1 tensor by Ω = 2 (∇u − ∇uT ).183 0.0 to Comax = 0.146 3 2. The kinematics of the threedimensional wing is simple harmonics with Aθ = 0 rad. this criterion may lead to undesired contours of e. For Q > 0 the region is identiﬁed as a vortex.. corresponding to Comax = 2.5 ǫlift [%] 400k 0. The timestep is taken such that Comax = 1. ω is the common vortex visualisation method (Bos et al.043 0.5 Threedimensional veriﬁcation: limit cycle. there is a concentration of vorticity.0 1. within this particular region. Shown are the errors for two grid sizes.097 0.1 rad and Aα = 0. however.785 rad. If ω reaches a userdeﬁned threshold. Since shear layers and curved streamlines also are a source of vorticity.053 0.011 ǫdrag [%] 800k 0. constructed from the lift and drag coeﬃcients clearly shows periodic behaviour for the veriﬁcation cases.024 ǫlift [%] 800k 0.5 Error values of lift and drag coeﬃcients for varying temporal resolution. The grid size was ﬁxed to 800k. Comax 2. 2008).5 3 2 1 1 0 Drag coeﬃcient [] 2 3 Figure 7. that region is identiﬁed as a vortex.159 0.g. In twodimensional ﬂow. and the Q criterion (Hunt et al. Aα = 45◦ and Aθ = 0◦ .087 0. This second invariant of ∇u is written as Q= 1 Ω2 − S2 . Hence..5 0 Vortical structures in threedimensional ﬂapping ﬂight 0.5. The limit cycles.5 Lift coeﬃcient [] 2 1. Especially in threedimensional ﬂows this may become a diﬃculty. The ﬁrst criterion is based on the magnitude of vorticity ω.013 ǫdrag [%] 400k 0.022 Table 7.0. The kinematics of the threedimensional wing is simple harmonics with Aφ = 63◦ . Error values of lift and drag coeﬃcients [%] with respect to the extrapolated values for varying temporal resolutions.
7.4 Vortex identiﬁcation methods for ﬂow visualisation 1. ρ where ρ is the ﬂuid density and p the pressure.03 2.224 1.. The temporal convergence is illustrated in (c) and (d). The ﬁnal value. Therefore.97 (a) Average drag coeﬃcient [] 2.7 shows isosurfaces around a ﬂapping wing.25T .02 2. corresponding to Comax = 1. The isosurface is visualised at t = 0. rate.04 2.22 1.015 12 4 8 16 Temporal resolution [] ∞ Average lift coeﬃcient [] 400k 800k (b) 1. a pressure minimum does occur.03 147 Average drag coeﬃcient [] ∞ 2.216 1.218 1.0.19 12 4 16 8 Spatial resolution [] 1.02 2.05 2.01 2 1. a region where Q > 0 indicates a clear swirling ﬂow (as shown by Chakraborty et al.23 1.214 1.22 1. during the downstroke.2 2.98 12 4 8 16 Spatial resolution [] ∞ 1. showing the average drag and lift coeﬃcients for decreasing timestep at two speciﬁc grid sizes.23 1. 2005). ﬁgure 7. By neglecting these unsteady and viscous eﬀects from the governing NavierStokes equations the following relation can be obtained for the symmetric tensor Ω2 + S2 : 1 Ω2 + S2 = − ∇(∇p). It must be noted that Jeong & Hussain (1995) found that Q > 0 is not a suﬃcient condition to have a pressure minimum in the vortex core of that speciﬁc region. 400k and 800k.035 2. (a) and (b) are showing the average lift and drag coeﬃcients for increasing spatial resolution and constant timestep.222 1. The wing ﬂaps around a distance of 0.25T the leadingedge vortex is formed on the wing’s . At t = 0.025 2.18 1.226 1.212 1.99 1.228 1.5 from the wing root and the ﬂapping angles are varying harmonically.21 1.24 Average lift coeﬃcient [] 1.21 400k 800k 12 4 8 16 Temporal resolution [] ∞ (c) (d) Figure 7. In order to identify which vortex criterion should be used. however. In most cases.6 Spatial and temporal convergence. at ∞. is obtained using Richardson extrapolation.
The leadingedge vortex. The Q criterion shows more detail. Some of the previously shed vortices are still present. In that section. within the range of −1. the vortices from the previous stroke should be visible as well. e. normalised by their maximal values.g. The spiralling leadingedge vortex is visualised using contour plots of the magnitude of vorticity. within the range of −1.1. ω and Q. a smooth leadingedge vortex is shown. (a) shows the contour of the vorticity magnitude. in ﬁgure 7. In ﬁgure 7. The colours represent values of helicity. is identiﬁed using a carefully chosen threshold of the vortex identiﬁcation criteria.0. rolling up into a tip vortex. Since the Q criterion oﬀers suﬃcient and adequate information about the local ﬂow ﬁeld. This leads to a thicker isosurface.7(a) it can be observed that ω shows not only the vortical structures. rolling up into a tip vortex. diﬀerent geometric and kinematic parameters are systematically varied. The arrow shows the ﬂapping direction of a downstroke and the ﬂow is visualised at midstroke. but also the shear layers near the wing and between the vortices. In order to investigate the eﬀect of the Rossby number. a rotation dominated region is identiﬁed by Q < 0. h = (u · ω)/(u ω).148 Vortical structures in threedimensional ﬂapping ﬂight (a) ω (b) Q Figure 7.0 value. h > 0. The colours show the helicity which is deﬁned as h = (u · ω)/(u ω). ω = 5.0.0 and Q = 2. upper surface and rolls up into a tip vortex.0. the radius . using the values ω = 5.5 Flapping wings at low Reynolds numbers In order to provide insight into the vortex dynamics (for a purely harmonic ﬂapping motion) and its inﬂuence on the variation of forces.0 and (b) the Q = 2.7(b). First. 7.5. this criterion is used throughout the remainder of the present research. A positive helicity.7 Comparison of the near wake ﬂow ﬁeld using diﬀerent vortex identiﬁcation criteria. the inﬂuence of an increasing midstroke angle of attack is brieﬂy discussed on the force development.0 ≤ h ≤ 1.0 ≤ h ≤ 1. means that the direction vector of vorticity (ω = ∇ × u) is aligned with the local ﬂow velocity. for all threedimensional simulations. the inﬂuence of the angle of attack on the forces is brieﬂy discussed in section 7. such that detail of the vortical structures is lost.
. Figure 7. In addition. due to shedding of the leadingedge vortex at low midstroke angle of attack.6 deals with forward ﬂapping ﬂight with similar conditions as the twodimensional simulations performed by Bos et al. this is performed for diﬀerent Reynolds number as well. The average drag is maximal and lift minimal for α = 90◦ .8 that the maximal force values occur at approximately halfway through the down and upstroke. the average lifttodrag ratio obtains a maximal value at an angle of attack of α = 30◦ . since these periods of lower lift appear to occur at low angles of attack.5.6 shows the timeaveraged lift and drag coeﬃcient for varying angles.4 and 7.5. Note that α = 90◦ implies that the wing keeps a constant upright position during the entire stroke. typical for revolving .2. The Rossby number was increased from Ro = 3.2 to Ro = 130. concerning hovering ﬂight at fruit ﬂy conditions.5.1 The angle of attack in ﬂapping ﬂight Previous studies showed that the angle of attack variation during the stroke inﬂuences the forces considerably.5. the radius of curvature is increased to decrease the angular acceleration consequently. corresponding to a ﬂapping wing with small radius of curvature. the kinematic model is varied by considering a ‘trapezoidal’ shape adaptation and the addition of deviation in section 7.6.75T . The drag for α = 15◦ shows two minor peaks within each halfstroke. (2009). t = 0. Thaweewat et al.2 Inﬂuence of ﬂapping stroke curvature In order to investigate the forces and development of the leadingedge vortex. (2008). from table 7. where T is the ﬂapping period. The Reynolds number is Re = 100 and Rossby Ro = 3.8 shows a variation of the lift and drag coeﬃcients during a complete ﬂapping cycle. Additionally. respectively. section 7.7. it must be noted that the lift is nearly always nonnegative. This was conﬁrmed by a recent twodimensional investigation (Bos et al. 2008).5 Flapping wings at low Reynolds numbers 149 of curvature is subsequently varied in section 7.5. While the lift is nearly identical for α = 45◦ and α = 60◦ . Furthermore. It is seen that the maximal timeaveraged lift coeﬃcient occurs at a midstroke angle of attack of α = 45◦ . The inﬂuence of the Reynolds number on the forces and leadingedge vortex stability is assessed in 7.2. lift is being generated during the complete stroke. Furthermore.3. 7. The next section will discuss the drops in more detail. which means that during the hovering conditions. It can be observed that the force variation is periodic and smooth. Table 7. independent of Rossby and Reynolds number. The diﬀerence between the maximal (α = 45◦ ) and minimal (α = 15◦ ) lift is 57%. It was already mentioned that the angle of attack at midstroke is varied from α = 15◦ to α = 90◦ . 7. However.5. In addition to the variety of hovering ﬂight simulations. the drag is signiﬁcantly lower for α = 45◦ .5. it can be seen in ﬁgure 7.25T and t = 0. it seems that the leadingedge vortex only grows signiﬁcantly at higher (α ≥ 45◦ ) angles of attack.
436 0.977 0.2 and a Reynolds number of Re = 100.034 1. : 60◦ .000 0. The variation is shown for lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing at Ro = 3.7 0.028 1.9 1 t/T [] (a) (b) Figure 7. •: 30◦ . ▽ : 45◦ . 4 3.333 0.224 0.6 Force coeﬃcients for Re = 100 and Ro = 3.5 0.9 1 t/T [] α α α α α α = = = = = = 15◦ 30◦ 45◦ 60◦ 75◦ 90◦ 6 Drag coeﬃcient [] 4 2 0 2 4 6 0 α α α α α α = = = = = = 15◦ 30◦ 45◦ 60◦ 75◦ 90◦ 0.329 2. .5 0 0. drag CD and lifttodrag CL /CD ave are shown as a function of the midstroke geometrical angle of attack for given Re = 100 and Ro = 3.2 0. △: 75◦ and : 90◦ .2 0.6 0.750 2.1 0.3 0.5 0.6 0.3 0.667 0.4 0.2.5 2 1.127 1.8 0.2. The midstroke angle of attack is varied from ♦ : 15◦ .397 Table 7.722 0.339 0.5 Lift coeﬃcient [] 3 2.957 CL /CD ave 0. so a ﬂapping wing with small stroke curvature.8 Variation of lift and drag coeﬃcients for a ﬂapping wing at Reynolds number of Re = 100.178 0.4 0.150 Vortical structures in threedimensional ﬂapping ﬂight αgeom 90 75 60 45 30 15 CL 0.5 1 0.7 0.5 0 0.000 0.526 CD downstroke 3.546 3.325 2. Timeaveraged lift CL .739 2.963 CD upstroke 3.8 0.1 0.543 3.703 1.
However. The midstroke angle of attack was ﬁxed to α = 45◦ and the Reynolds number to Re = 100.e.883 CD upstroke 2. occurs at midstroke t=0.058 1. It is clear that both lift and drag are decreasing with increasing Rossby number.2T ) for an angle of attack of α = 15◦ . At Ro = 130 the wing nearly performs a twodimensional motion leading to a decrease in lift of 24.8 and 7. feeding two wing tip vortices. This is due to the loss in energy by the tip vortices which was also studied by (Blondeaux et al.997 0. Ro = 130. When compared with ﬁgure 7..11 shows the isosurfaces of Q = 1. Figure 7.572 0.3 130 1.2 3.922 CL (baseline) (−3.75T.8 (which applies to the revolving wing Ro = 3.023 0.e.1 6. decreasing curvature of the stroke path. The decrease in drag is small.933 1. the translating wing shows a leadingedge vortex which stays symmetric with respect to the wing centre plane.899 1.904 1.8 5.9%) (−9.9 is that the lift drops signiﬁcantly (75% at least) during midstroke (t = 0.886 1.877 CL /CD ave 0. Figure 7.5 Flapping wings at low Reynolds numbers 151 Ro 3.981 1.2 to Ro = 130. The loss in drag. i. with increasing Rossby number for α = 45◦ . just after stroke reversal and during midstroke are of similar magnitude.537 0. An additional observation.892 1.908 1.25T and t=0.4%) (−18.2%) (−8.4%) (−11. 18.903 1.9%) (−24. It is clear that the major loss in lift for high Rossby numbers.552 0.4 7.224 1. The variation of average lift (CL ). Additionally.13 shows the streamlines to illustrate the leadingedge and tip vortices in more detail. Table 7. Figure 7.7% in comparison to the baseline case with Ro = 3. drag (CD ).0 at t = 0. drag and lifttodrag ratio.7.5%) (−22.7%) Table 7. ﬁgure 7.490 (baseline) (−1.7 shows the timeaveraged values for lift.2%) (−18.0%) (−13. which is at the midst of the downstroke.12 shows the Q = 1. It can be clearly observed that the fully ﬂapping wing shows a pronounced leadingedge vortex which spirals towards the tip to form a tip vortex.175 1.7%) (−13. Figure 7. which results in the lower .2.028 2. such that the decrease in lifttodrag ratio is still signiﬁcant.593 0.2) it is seen that both lift and drag variations are signiﬁcantly lower when the Rossby number is large.105 1. at the root and the tip.935 1.500 0. lifttodrag ratio (CL /CD ) are shown for Rossby numbers from Ro = 3.6 8.603 0.943 0.916 1.6%) (−16. respectively.9 15.915 1. while comparing ﬁgures 7. 2005a).6%) (−5.983 1.0 isosurfaces of a translating wing at Ro = 130 for α = 15◦ and α = 45◦ .25T .034 1.9 shows the force histories concerning the nearly translating wing.10 shows the variation of the lift and drag coeﬃcients during the ﬂapping cycle and the eﬀect of Rossby number as it increases.525 0.7%) CD downstroke 2. and translating wings. i.7 Timeaveraged force coeﬃcients for varying Rossby numbers at Re = 100.7%.0%) (−17. It is clearly seen that the leadingedge vortex for α = 15◦ is not yet fully developed.
14(a) for diﬀerent angles of attack and diﬀerent Reynolds numbers.1 0.8 0.6 0.14(b) shows the power factor CL /CD as a function of lifttodrag ratio CL /CD . it was shown for both ﬂapping and translating wings at α = 15◦ . Re = 100. The drag increases as well. This leadingedge vortex is larger and more stable at angles of attack larger than about 30◦ . lift. compared to α = 45◦ . 7.5 0. Figure 7. α = 45◦ the diﬀerence between ﬂapping (Ro = 3.2) and translating (Ro = 130) in lift coeﬃcient is 32. The angle of attack amplitude is varied from ♦ : 15◦ . 2004).6 0. In addition. as long as the scaling is appropriate. such that the dimensionless amplitude A∗ g . average R Reynolds number ReRg and swept area Aswept are comparable.. it can be stated that a ﬂapping wing motion is of crucial importance for lift generation at a small penalty of drag. It seems that the trend of the force development with the angle of attack is similar for ﬂapping and translating wings. Summarising.14(a) it can be deduced that the overall lift coeﬃcients are significantly higher for the ﬂapping (Ro = 3. The Reynolds number remained ﬁxed at Re = 100. •: 30◦ .5 0. The timeaveraged lift and drag coeﬃcients are plotted in ﬁgure 7.4 0. respectively. Birch & Dickinson. Re = 500 and Re = 1000. such that the wing approximately translates.2 0. 500 and 1000. that the leadingedge vortex development is not signiﬁcant to increase the lift.3 0. △: 75◦ and : 90◦ .5. compared to wing translation. a selection of Reynolds numbers is used.2 0.5 0 0. 33.5 0 Vortical structures in threedimensional ﬂapping ﬂight Drag coeﬃcient [] α α α α α α = = = = = = 15◦ 30◦ 45◦ 60◦ 75◦ 90◦ 6 4 2 0 2 4 6 0 α α α α α α = = = = = = 15◦ 30◦ 45◦ 60◦ 75◦ 90◦ 0. At maximal lift. the leadingedge vortex is important for the gain in lift. Lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing at a Rossby number of Ro = 130. Additionally. ▽ : 45◦ . 2003).7 0.9 1 t/T [] (a) (b) Figure 7.5 1 0.8%.2) compared to the translating (Ro = 130) wing.9 1 t/T [] 0. From ﬁgure 7.3 0. : 60◦ . the results of a variation in Rossby number are shown for Re = 100.8 0. instead the lift decreases. relevant for insect aerodynamics of a fruit ﬂy (Sane & Dickinson. 2001.5 2 1.4 0.3 Inﬂuence of Reynolds number In addition to the angle of attack and stroke curvature. 1998) and dragonﬂy (Isogai et al.9% 3/2 .5 Lift coeﬃcient [] 3 2.9 Variation of lift and drag coeﬃcients for a translating wing at Ro = 130.1 0. hawkmoth (Liu & Kawachi. At smaller angles of attack.152 4 3.7 0.
9.4.2 (b) Ro = 130 Figure 7.2 0. A timeframe is shown at midstroke.2 3.6 0.4 8.25T .9 130 0. Lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing around varying Rossby numbers.5 1 0. •: 3.8. ▽ : 6.3 0.8 6. Ro = 3.2.8 0. The amplitude of the angle of attack variation was ﬁxed such that at midstroke α = 45◦ .7 0.5 0.5 0 0. .5 0.4 0.9 130 4 3 Drag coeﬃcient [] 2 1 0 1 2 3 4 0 Ro Ro Ro Ro Ro = = = = = 3.2 0. : 8. Colours indicate helicity.5 Flapping wings at low Reynolds numbers 153 3 2.4 0. t = 0.2 3.9 1 t/T [] Ro Ro Ro Ro Ro = = = = = 3.4 8.1 0. The average Reynolds number remained ﬁxed at Re = 100.2 and Ro = 130.5 Lift coeﬃcient [] 2 1.9 1 t/T [] (a) (b) Figure 7.5 0 0.8 0.8 6.0 are shown for Rossby numbers Ro = 3.2 − 130.1 0.3 0. (a) Ro = 3.6 0.7. △: 130. The Rossby number is varied from ◦ : 3. Isosurfaces of Q = 1.10 Variation of force coeﬃcients for a ﬂapping wing with varying Rossby numbers.11 Vortex visualisation of the near wake for diﬀerent Rossby numbers. The average Reynolds number was ﬁxed to Re = 100.7 0.
Colours indicate helicity. . A timeframe is shown at midstroke.154 Vortical structures in threedimensional ﬂapping ﬂight (a) α = 45◦ (b) α = 15◦ Figure 7. Streamlines are shown for Rossby numbers Ro = 3.13 Streamline visualisation of the near wake for diﬀerent Rossby numbers. A timeframe is shown at midstroke.2 and Ro = 130.12 Vortex visualisation of the near wake for a translating wing at low and high angle of attack. The average Reynolds number was ﬁxed to Re = 100. (a) Ro = 3.25T .2 (b) Ro = 130 Figure 7.0 are shown for α = 15◦ and α = 45◦ for a Rossby number of Ro = 130. The average Reynolds number was ﬁxed to Re = 100. t = 0. t = 0. Isosurfaces of Q = 1.25T . Colours indicate helicity.
14(a) is that at large midstroke angles of attack.6 0. As was already discussed.5 4 0 Re=100.6 0. Re = 100. The angle of attack varies from 90◦ to 15◦ .7 0. In both (a) and (b). Ro=130 Re=500. On the other hand. for a translating (Ro = 130) wing.6 1. an average diﬀerence in drag of 7. considering a ﬂapping motion (Ro = 3.8 0. 500 and 1000.2 Re=500. ﬂapping is important at lower Reynolds numbers. Ro=3.8 0. Figure 7.2 0 0.3 0. Looking at ﬁgure 7. Ro=3. the structure of the leadingedge vortex strongly depends on the Reynolds number in cases of large angular accelerations. the timeaveraged lift and drag coeﬃcients show marginal variations with Reynolds number.7.9 1 Glide ratio [] Drag coeﬃcient [] (a) Figure 7. while ﬂapping.2 Re=500.6 0. Ro=3. (b) illustrates the power factor CL /CD as a function of the lifttodrag ratio CL /CD . and two Rossby numbers. At α = 45◦ the diﬀerences in timeaveraged drag coeﬃcient are negligible when the variation in Reynolds number is concerned.2 3 3.5 Re=100.2 0. while at α = 75◦ the diﬀerences in drag become signiﬁcant. Another observation from ﬁgure 7.2 Re=1000.4 0.5 0.5 0 0. the results are shown for three diﬀerent Reynolds numbers. and 35.5 Flapping wings at low Reynolds numbers 1. from right to left in (a) and from left to right in (b).4 0. which is beneﬁcial in terms of performance.8% for increasing Reynolds numbers Re = 100.8 0.4 Lift coeﬃcient [] Power factor [] 1. The diﬀerence in drag is only signiﬁcant at larger angles of attack. which is visualised by irregularities in the isosurfaces. Ro=130 Re=1000. Since the diﬀerence in lift increases (although slightly) with Re. Ro=130 Re=1000.1 0. However. Ro = 3.15 shows the isosurface of Q = 1. Ro=3. Ro=130 1 1.14 (b) Force and performance polars. While lift is enhanced signiﬁcantly. respectively.5 2 2.2 1 0.2 Re=100. Ro=3.14(a).2 and Ro = 130.2) with respect to translating (Ro = 130). the variations in lift and drag with Reynolds number are larger for lower Rossby numbers.2 0. the leadingedge vortex becomes slightly unstable with increasing Reynolds numbers. there is still a large gain in lifttodrag.2 Re=1000. the streamlines for the corresponding .4 0.0 to visualise the leadingedge vortex on a ﬂapping wing (Ro = 3. (a) shows the force polars as a function of angle 3/2 of attack amplitude. Ro=3. but becomes slightly more important at higher Reynolds number. combined with a small drag penalty. 500 and 1000. In addition.5% is obtained. Ro=130 155 1 0.2 Re=100. So. This may be explained by considering that the leadingedge vortex was found to be more stable on a translating wing. compared to ﬂapping.2) during the downstroke for both Reynolds numbers Re = 100 and Re = 1000. Ro=130 Re=500. the Reynolds number has a larger eﬀect on the lift and drag.
Figure 7. .e. but also to irregularities such that the ﬂow at low Ro is more sensitive to changes in Reynolds number. Isosurface of Q = 1. although the leadingedge vortex may be less stable.16.2. Additionally. However.2 and translating Ro = 130 ﬂight. this is not the case for a translating wing. for a ﬂapping wing Ro = 3.156 Vortical structures in threedimensional ﬂapping ﬂight (a) Re = 100 (b) Re = 1000 Figure 7. ﬁgure 7. compared to translation. While the leadingedge vortex detaches at a smaller distance from the wing tip. In order to illustrate the irregular motion at larger Reynolds numbers. it seems plausible that the generation of a leadingedge vortex is important for both ﬂapping Ro = 3. with increasing Reynolds numbers.2. Therefore. Colours indicate helicity. Besides the irregularities.2.15 Vortex visualisation of the near wake for diﬀerent Reynolds numbers. It was noted that the leadingedge vortex may play a more important role for ﬂapping motion. comparison are shown in ﬁgure 7. without a signiﬁcant loss in lift. For a translating wing Ro = 130. the leadingedge vortex may burst as was discussed by Lentink & Dickinson (2009b).15 shows some irregularities of the leadingedge vortex. In case of the ﬂapping wing Ro = 3. on a ﬂapping wing for increasing Reynolds numbers. for a translating wing. for higher Reynolds numbers.0 are shown for two Reynolds numbers. leads to larger lift enhancement. t = 0.18 shows the Q isosurfaces for diﬀerent Reynolds numbers Re = 100 and Re = 1000. A timeframe is shown at midstroke. these irregularities are less pronounced. A Reynolds number increase. Ro = 3.17 shows the vortical motion just before stroke reversal from down to upstroke. i.25T . Re = 100 and Re = 1000. ﬁgure 7. the lift increasing eﬀects of the leadingedge vortex are larger for higher Reynolds numbers. Re = 1000. the leadingedge vortex clearly detaches at a smaller distance from the wing root for Re = 1000.
Re = 100 and Re = 1000. t = 0. .5 Flapping wings at low Reynolds numbers 157 (a) Re = 100 (b) Re = 1000 Figure 7. (a) Re = 100 (b) Re = 1000 Figure 7. A timeframe is shown at the end of the downstroke. Ro = 3. Re = 100 and Re = 1000.2. Ro = 3.2. A timeframe is shown at midstroke. Colours indicate helicity.25T . Isosurface of Q = 1.2. for a ﬂapping wing Ro = 3.17 Vortex visualisation of the near wake for diﬀerent Reynolds numbers.2. Streamlines are shown for two Reynolds numbers.0 are shown for two Reynolds numbers. for a ﬂapping wing Ro = 3. t = 0.7. Colours indicate helicity.5T .16 Streamline visualisation of the near wake for diﬀerent Reynolds numbers.
whereas the latter corresponds to a strong rotation during stroke reversal.4 Inﬂuence of ‘trapezoidal’ angle of attack In order to study the inﬂuence of the ‘trapezoidal’ angle of attack and later compared to an earlier twodimensional study (Bos et al. 7. generated in both up. the Reynolds number and midstroke angle of attack remained ﬁxed. it reaches its midstroke angle of attack early in the stroke. for a translating wing Ro = 130. are within 1. Isosurface of Q = 1.8 shows the timeaverage lift. accompanied by a decrease in drag.158 Vortical structures in threedimensional ﬂapping ﬂight (a) Re = 100 (b) Re = 1000 Figure 7.19 for varying rotation duration during a full ﬂapping stroke.5.0%. drag and lifttodrag values..10. which is varied from 0. At low rotation duration Trot = 0. the angular acceleration just after stroke reversal is large. using a ‘trapezoidal’ shaped angle of attack variation.and downstroke. Ro = 130. Colours indicate helicity. The time variation of both lift and drag is shown in ﬁgure 7. so drag generation is symmetric during a complete stroke. 2008). The most important observation is that with decreasing rotation duration. Figure 7. The average drag decreases with a similar amount.8%. Furthermore. compared .2 shows the angle of attack as a function of the rotation duration Trot . increasing angular acceleration during stroke reversal.0 are shown for two Reynolds numbers. it can be seen that the average drag coeﬃcient. Table 7. the shape of the angle of attack is varied.25 to 0. Various experimental and numerical studies have been conducted.1. While varying the rotation duration.e. a gain in average lift is obtained of 10. the ‘trapezoidal’ shape is deﬁned by a rotation timing Trot . i.25T . Since the wing rotates relatively quickly after reversal.9%. leading to a signiﬁcant increase in lifttodrag ratio of 21. t = 0. A timeframe is shown at midstroke. the ﬁrst representing a harmonic variation. As previously described. which may be caused by a fast decrease in eﬀective angle of attack.3 in section 7. which leads to an increase in lift.2. Re = 100 and Re = 1000.18 Vortex visualisation of the near wake for diﬀerent Reynolds numbers.
The explanation for this discrepancy is that in the threedimensional simulations the leadingedge vortex was found to remain more stable than in twodimensional investigations.5 Lift coeﬃcient [] 2 1.6 0.5 1 0.493 CL /CD ave 0. 2008) showed opposite eﬀects. At every time instance. but opposite.536 2.9 1 t/T [] Trot Trot Trot Trot = = = = 0.740 2.25 0.8%) (+10. Lift (b) and drag (c) coeﬃcients induced by a ﬂapping wing.7. The rotation duration is varied from Trot = 0. Therefore. The average Reynolds number was maintained at Re = 100.15 0. using an angle of attack amplitude of Aα = 60◦ . 3 2.8 0.8%) CD downstroke 2. which leads to the integrated gain in lift of 10.2%) (+16. This corresponds to respectively harmonic variation to a ‘trapezoidal’ shape.496 CD upstroke 2.3 0.127 1. The timeaveraged lift CL . eﬀect applies to the drag coeﬃcient. A similar.9 1 t/T [] Trot Trot Trot Trot = = = = 0.19 Force coeﬃcients for a ﬂapping wing with ‘trapezoidal’ angle of attack.2 0.6 0.10 0.25 to Trot = 0.5 Flapping wings at low Reynolds numbers 159 Trot 0. drag CD and lifttodrag ratio CL /CD are shown for a rotation duration.4 0.2 0.20 0.9% lifttodrag enhancement.6%) (+21.20 0.215 1.5 0. the lift increase and drag decrease are only marginal. .8 0.9%) Timeaveraged force coeﬃcients for varying ‘trapezoidal’ shape of angle of Table 7. Therefore.5%) (+7.500 (baseline) (+9.411 0.15 0.5 0 0.448 0.10 1.5 0 0.1 0.8 attack.629 2.10.10 0. 21.625 2.25 (a) (b) Figure 7.248 CL (baseline) (+4.15 0.8%.7 0.4 0.539 2. In contrast to this.7 0.25. such that Ro = 3.10.20 0. varying from Trot = 0. twodimensional studies (Bos et al.1 0.479 0.3 0.25 to Trot = 0.25 5 4 Drag coeﬃcient [] 3 2 1 0 1 2 3 4 5 0 0. it can be concluded that the leadingedge vortex stability should be studied with a threedimensional approach. with the harmonic case Trot = 0.2 and Re = 100. it can be stated that a ‘trapezoidal’ shape in angle of attack considerably increases performance in threedimensional hovering ﬂapping ﬂight. but overall the results are signiﬁcant. due to a premature vortex shedding of the leadingedge vortex during the long period of high angle of attack.750 2.. This causes a long period of lift enhancement.177 1.5 0.
160
Vortical structures in threedimensional ﬂapping ﬂight
7.5.5
Inﬂuence of deviation
The deviation amplitude Aθ is used to tilt the twodimensional airfoil or threedimensional wing out of the horizontal stroke plane. It may be used to generate certain wing tip patterns, e,g. the wellknown ‘ﬁgureofeight’, which is present in realistic fruit ﬂy kinematics (Fry et al., 2003). In (Bos et al., 2008) it was shown that although deviation did not inﬂuence the timeaveraged values, the instantaneous lift and drag variations are signiﬁcantly aﬀected. In order to investigate the inﬂuence of deviation the amplitude Aθ is varied from 0◦ to 20◦ . In addition, three diﬀerent stroke patterns are studied, by varying the deviation frequency, resulting in patterns that can be characterised as ‘ﬁgureofO’, ‘ﬁgureofU’ and ‘ﬁgureofeight’, which are shown in ﬁgure 7.2. Considering the ‘ﬁgureofeight’ patterns, two diﬀerent deviation directions are studied, corresponding to a variation of Aθ from 0◦ to 20◦ and from 0◦ to −20◦ . The reference velocities are adapted correspondingly. First, the deviation amplitude is varied from Aθ = 0◦ to Aθ = 20◦ , according to the ‘ﬁgureofO’ wing tip pattern. Following ﬁgure 7.2, the wing moves consecutive down and up during the downstroke and up and down during the upstroke. Since a downward motion increases the eﬀective angle of attack, which is therefore subjected to an increase, decrease, decrease and another increase during the four consecutive halfstrokes. Table 7.9 shows the timeaveraged force coeﬃcients for a wing following this ‘ﬁgureofO’ motion, while the midstroke angle of attack is ﬁxed to α = 45◦ and the Reynolds number maintained at Re = 100. It is shown that the timeaveraged lift coeﬃcient decreases signiﬁcantly with 9.8%. However, this is fully compensated by a decrease in drag with the same amount such that the diﬀerences in average lifttodrag ratio are negligible. The average lifttodrag was obtained by using the average drag over up and downstroke. This was necessary, because of the asymmetry appearing in the average drag coeﬃcient. This asymmetry in drag is the result of the asymmetrical variation in eﬀective angle of attack, as was previously discussed. Figure 7.20 show the lift and drag variations during a complete ﬂapping stroke. Secondly, the results are considered for a ﬂapping wing following the ‘ﬁgureofU’ pattern, which is similar to the one used in (Bos et al., 2008). When using this kinematic pattern, the wing moves down and up during every halfstroke, i.e. the upstroke is identical to the downstroke. In table 7.10 is can easily be seen that the diﬀerences in up and downstroke drag are negligible for all deviation amplitudes. This is in contrast to the observations, considering the ‘ﬁgureofO’. The timeaveraged drag coeﬃcient is constant for deviation amplitude variation, while the lift coeﬃcient decreases with 9.4%, leading to a similar decrease in lifttodrag ratio. Because of symmetric (similar up and downstroke) ﬂapping, a decrease of lifttodrag coeﬃcient is obtained, which is present, but not signiﬁcant in comparison to the eﬀect of varying Rossby and Reynolds numbers. The third pattern is governed by the ‘ﬁgureofeight’ wing tip motion. Although, the previously described deviation patterns only cause marginal eﬀects,
20
7.5 Flapping wings at low Reynolds numbers
161
Aθ 0 5 10 15 20 1.224 1.204 1.180 1.151 1.104
CL (baseline) (1.6%) (3.6%) (5.9%) (9.8%)
CD downstroke 2.030 2.067 2.072 2.056 1.996
CD upstroke 2.030 1.934 1.850 1.767 1.676
CL /CD ave 0.603 0.602 0.601 0.602 0.601 (baseline) (0.15%) (0.23%) (0.12%) (0.28%)
Table 7.9 Timeaveraged force coeﬃcients for varying deviation using a ‘ﬁgureofO’ pattern. The timeaveraged lift CL , drag CD and lifttodrag ratio CL /CD are shown for a deviation amplitude, varying from Aθ = 0◦ to 20◦ , for a ‘ﬁgureofO’. The midstroke angle of attack was ﬁxed at α = 45◦ and the average Reynolds number was maintained at Re = 100. 3 2.5 Lift coeﬃcient [] 2 1.5 1 0.5 0 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 t/T [] Aθ Aθ Aθ Aθ Aθ = = = = = 0 5 10 15 20 4 3 Drag coeﬃcient [] 2 1 0 1 2 3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 t/T [] Aθ Aθ Aθ Aθ Aθ = = = = = 0 5 10 15 20
(a)
(b)
Figure 7.20 Force coeﬃcients for a ﬂapping wing with varying deviation using ‘ﬁgureofO’. Lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing, such that Ro = 3.2 and Re = 100. The deviation amplitude is varied from Aθ = 0◦ to Aθ = 30◦ , using an angle of attack amplitude of Aα = 45◦ .
the ‘ﬁgureofeight’ pattern may cause signiﬁcant changes in forces. Two types of ‘ﬁgureofeight’ patterns are used, from Aθ = 0◦ to 20◦ , this is called type 1 and the second from Aθ = 0◦ to −20◦ , type 2. To achieve a ‘ﬁgureofeight’ pattern, the deviation frequency is twice the frequency of the other two wing tip patterns. For both types of patterns, the eﬀective angle of attack variation consists of three parts during each half stoke, see ﬁgure 7.2. Both up and downstroke follow exactly the same, thus symmetrical, motion. The type 1 patterns starts each halfstroke with a downward motion, than it goes up until it has to go done just before stroke reversal. This wing tip motion leads to a consecutive increase, decrease and increase in eﬀective angle of attack, where the period of decrease is twice the period of increase. The type 2 pattern follows precisely the inverse motion. Table 7.11 shows the timeaveraged lift, drag and lifttodrag ratio’s for both types of ‘ﬁgureofeight’ patterns. Since the variation of the eﬀective angle of attack is symmetric, the drag coeﬃcient is equal for up and downstroke. The results
162
Vortical structures in threedimensional ﬂapping ﬂight
of the type 1 motion pattern are remarkable. There is a considerable decrease in both timeaveraged lift (52%) and drag (44%) when comparing Aθ = 20◦ with the baseline case Aθ = 0◦ . Since the drag decrease is of similar magnitude as the lift decrease, the loss in lifttodrag is limited to 15.5%, which is still signiﬁcant. Figure 7.20 shows the lift and drag variations for this type 1 ‘ﬁgureofeight’ pattern, where the wing tip moves down, up and down, consecutively during each halfstroke. It can be observed that the short period of downward motion at the beginning of each stroke increases lift. On the other hand, the large period of upward motion, decreases the eﬀective angle of attack for a relative long period, leading to a signiﬁcant loss of lift, as is seen in the ﬁgure. For Aθ = 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. α = 15◦ , see section 7.5.1. When considering the type 2 ‘ﬁgureofeight’ pattern, table 7.11 shows no signiﬁcant decrease in drag. The lift, however, decreases considerably, although signiﬁcantly smaller compared to the type 1 kinematic pattern, 53% versus 12.6%. The motion of the type 2 deviation is apparently equally distributed, resulting in only 12.6% less lift and no diﬀerences in drag, while increasing the deviation amplitude. The maximal diﬀerence in lifttodrag ratio is therefore 12.0%. This force balance is illustrated in ﬁgure 7.22, which shows the lift and drag during a full stroke. At the beginning of a stroke, the eﬀective angle of attack is decreased, leading to lower lift and lower drag. During midstroke, the eﬀective angle of attack is increased, which is reﬂected in the higher lift and drag. Summarising, it was shown that the variation in lift and drag can be signiﬁcantly inﬂuenced by introducing deviation in the stroke pattern, i.e. ‘ﬁgureofO’, ‘ﬁgureofU’ and ‘ﬁgureofeight’. The ‘ﬁgureofO’ patterns resulted in an asymmetric force variation, due to asymmetric modulation of the eﬀective angle of attack. Lift and drag decrease with a similar amount, such that the lifttodrag ratio was only marginally aﬀected. The timeaveraged drag was not inﬂuenced by the symmetrical ‘ﬁgureofU’ pattern. The average lift, however, did decrease, such that a loss in lifttodrag was observed. Two types of ‘ﬁgureofeight’ patterns were investigated, diﬀering in the direction of motion. When the wing moved up
Aθ 0 5 10 15 20 1.224 1.187 1.167 1.137 1.108
CL (baseline) (3.0%) (4.6%) (7.1%) (9.4%)
CD downstroke 2.030 2.003 2.012 2.014 2.021
CD upstroke 2.030 1.994 2.001 2.002 2.007
CL /CD ave 0.603 0.593 0.580 0.565 0.548 (baseline) (1.7%) (3.8%) (6.3%) (9.0%)
Table 7.10 Timeaveraged force coeﬃcients for varying deviation using a ‘ﬁgureofU’ pattern. The timeaveraged lift CL , drag CD and lifttodrag ratio CL /CD are shown for a deviation amplitude, varying from Aθ = 0◦ to 20◦ , for a ‘ﬁgureofU’ pattern. The midstroke angle of attack was ﬁxed at α = 45◦ and the average Reynolds number was maintained at Re = 100.
7.6 Flapping wings in forward ﬂight
163
Aθ 0 5 10 15 20 5 10 15 20 1.224 1.107 0.934 0.747 0.575 1.220 1.205 1.151 1.070
CL (baseline) (9.5%) (23.7%) (39.0%) (53.0%) (0.3%) (1.5%) (6.0%) (12.6%)
CD downstroke 2.030 1.859 1.621 1.370 1.129 2.067 2.057 2.056 2.017
CD upstroke 2.030 1.852 1.616 1.365 1.125 2.064 2.047 2.040 2.003
CL /CD ave 0.603 0.596 0.576 0.545 0.510 0.600 0.586 0.560 0.530 (baseline) (1.2%) (4.5%) (9.5%) (15.5%) (0.5%) (2.8%) (7.1%) (12.0%)
Table 7.11 Timeaveraged force coeﬃcients for varying deviation using a ‘ﬁgureofeight’ pattern. The timeaveraged lift CL , drag CD and lifttodrag ratio CL /CD are shown for a deviation amplitude, varying from Aθ = 0◦ to 20◦ and Aθ = 0◦ to −20◦ , for a ‘ﬁgureofeight’ pattern. The midstroke angle of attack was ﬁxed at α = 45◦ and the average Reynolds number was maintained at Re = 100.
wards during midstroke, decreasing the eﬀective angle of attack for a long period, the performance was limited in terms of lift and lifttodrag ratio. On the other hand, if the wing moved downward during each midstroke, the performance was similar to the ‘ﬁgureofU’ pattern. The lift and drag are shown to be sensitive to diﬀerent stroke patterns, such that the forces and performance can be easily modulated. Therefore, insects could use stroke plane deviation in extreme hovering or manoeuvring conditions. These ﬁndings are very interesting for the development of Micro Air Vehicles (MAV) as well.
7.6
Flapping wings in forward ﬂight
In order to relate the results of the threedimensional ﬂow simulations to previously conducted twodimensional studies (Thaweewat et al., 2009), a threedimensional forward ﬂapping wing has been studied. Based on (Thaweewat et al., 2009), the dimensionless wavelength is set to λ = 6.3 (maximal performance) in order to justify a comparison. The midstroke angle of attack is varied from α = 0◦ to 45◦ , while maintaining a Reynolds number of Re = 150. The threedimensional wing ∗ motion is harmonic and scaled such that the dimensionless amplitude ARg , based on the radius of gyration Rg is comparable between diﬀerent cases. Additionally, the average Reynolds number ReRg is matched. In table 7.12, the timeaveraged force coeﬃcients are shown for three diﬀerent ﬂapping situations: a twodimensional plunging airfoil, a threedimensional translating wing (Ro = 130) and a threedimensional ﬂapping wing (revolving, Ro = 3.2). For both ﬂapping and translating wings, the force coeﬃcients are com
5 1 0.164 Vortical structures in threedimensional ﬂapping ﬂight 3 2.6 0. such that Ro = 3.5 0 0. 3.7 0.21 Force coeﬃcients for a ﬂapping wing with varying deviation using type 1 of ‘ﬁgureofeight’. The deviation amplitude is varied from Aθ = 0◦ to Aθ = 20◦ .2 and Re = 100. The deviation amplitude is varied from Aθ = 0◦ to Aθ = −20◦ .4 0. such that Ro = 3.22 Force coeﬃcients for a ﬂapping wing with varying deviation using type 2 of ‘ﬁgureofeight’.6 0. Lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing.1 0.4 0.7 0.9 1 t/T [] Aθ Aθ Aθ Aθ Aθ = = = = = 0 5 10 15 20 (a) (b) Figure 7.5 0.9 1 t/T [] Aθ Aθ Aθ Aθ Aθ = = = = = 0 −5 −10 −15 −20 (a) (b) Figure 7. .5 Lift coeﬃcient [] 2 1.5 3 Lift coeﬃcient [] 2.8 0.8 0.5 0.3 0.1 0.5 2 1. using an angle of attack amplitude of Aα = 45◦ .2 and Re = 100.7 0.4 0.5 0.7 0.8 0.5 0 0.3 0.5 0.8 0.9 1 t/T [] Aθ Aθ Aθ Aθ Aθ = = = = = 0 5 10 15 20 3 2 Drag coeﬃcient [] 1 0 1 2 3 0 0.3 0.2 0.2 0.2 0.5 1 0.6 0.9 1 t/T [] Aθ Aθ Aθ Aθ Aθ = = = = = 0 −5 −10 −15 −20 4 3 Drag coeﬃcient [] 2 1 0 1 2 3 4 0 0.1 0.2 0.6 0. Lift (a) and drag (b) coeﬃcients induced by a ﬂapping wing.5 0 0.1 0. using an angle of attack amplitude of Aα = 45◦ .3 0.4 0.5 0 0.
028 0. a smooth ring vortex is formed by the Ro = 3.580 0. The threedimensional ﬂapping wing (Ro = 3.0%) (28. Therefore. leads to similar force results as if the study was completely twodimensional. Maximal lift occurs without wing rotation.231 0.429 0.214 CL /CD ave 35.12. a relative equal amount of energy is lost.9%) (21.2. it can be stated that a threedimensional wing.2 Ro = 130 2D . as is shown in table 7.12 Timeaveraged force coeﬃcients in forward ﬂight.964 5. Two.969 1.8%) (20.148 4.403 0. The two tip vortices result in a symmetric ﬂow. The diﬀerence with the translating wing becomes smaller with increasing angle of attack amplitude.316 0.401 0. λ∗ = 6.571 6. Concerning the threedimensional translating wing (Ro = 130). which results in a lower lift coefﬁcient.710 1.119 3.23. Figure 7. the lift decreases with 28% compared to the twodimensional plunging airfoil. For the case without revolving (ﬁgure 7. It is clear that the timeaveraged lift coeﬃcient decreases with increasing angle of attack amplitude.366 0. The drag varies with angle of attack such that it reaches a thrust optimum for Aα = 30◦ .2) generates larger lift coeﬃcients compared to the translating case. Note that negative drag means thrust. leading to a similar decrease in lift.904 1. The midstroke angle of attack is varied from 0◦ to 45◦ .376 1.24 shows a comparison of the ﬂow ﬁelds for Ro = 130 and Ro = 3. but lower compared to the twodimensional ﬂapping airfoil.414 1.834 1.824 Table 7.24(a)).673 27. A ﬂapping wing (Ro = 3.3. which is the case for both twodimensional and threedimensional ﬂapping.0676 0.544 1. but the resulting thrust is minimal.897 2.604 CLave (baseline) (baseline) (baseline) (baseline) (28.016 0.270 0.7. This is the case for midstroke values of the angle of attack.2%) (27. performing a twodimensional motion.419 1.920 2.175 0. The generation of tip vortices causes a loss of energy. pared with the twodimensional results.2) in ﬁgure 7.4%) (27.719 2.211 3. without an angle of attack variation.and threedimensional timeaveraged force coeﬃcients for a ﬂapping wing in forward ﬂight at Re = 150. This is il