You are on page 1of 116

Dissertation presented to the Instituto Tecnológico de Aeronáutica, in partial

fulfillment of the requirements for the degree of Master of Engineering of the


Professional Master’s Course in Aeronautical Engineering in the Program of
Aeronautical and Mechanical Engineering

Ana Luísa Manzaro da Costa

A NUMERICAL INVESTIGATION ON WAVY LEADING

EDGE APPLICATION IN TRANSONIC FLOW CONTROL

Dissertation approved in its final version by the signatories below:

Prof. Dr. Adson Agrico de Paula


Advisor

MSc Grace Rodrigues de Lima


Co-advisor

Prof. Dr. Pedro Teixeira Lacava


Prorector of Graduate Studies and Research

Campo Montenegro
São José dos Campos, SP – Brasil
2019
Cataloging-in Publication Data
Documentation and Information Division
Manzaro da Costa, Ana Luísa
A Numerical Investigation on Wavy Leading Edge Application in Transonic Flow Control / Ana
Luísa Manzaro da Costa.
São José dos Campos, 2019.
115f.

Dissertation of Master of Engineering – Professional Master’s Course in Aeronautical Engineering


in the Program of Aeronautical and Mechanical Engineering – Instituto Tecnológico de Aeronáutica,
2019. Advisor: Prof. Dr. Adson Agrico de Paula. Co-advisor: MSc Grace Rodrigues de Lima.

1. Palavra-chave conforme seção 3.2. 2. Palavra-chave2. 3. Palavra-chave3. I. Instituto Tecnológico


de Aeronáutica. II. A Numerical Investigation on Wavy Leading Edge Application in Transonic Flow Control

BIBLIOGRAPHIC REFERENCE
MANZARO DA COSTA, Ana Luísa. A Numerical Investigation on Wavy Leading Edge
Application in Transonic Flow Control. 2019. 115f. Dissertation of Master of Engineering
in Aeronautical Engineering – Instituto Tecnológico de Aeronáutica, São José dos Campos.

CESSION OF RIGHTS
AUTHOR’S NAME: Ana Luísa Manzaro da Costa
PUBLICATION TITLE: A Numerical Investigation on Wavy Leading Edge Application in
Transonic Flow Control
PUBLICATION KIN/YEAR: Dissertation / 2019

It is granted to Instituto Tecnológico de Aeronáutica permission to reproduce copies of this


dissertation and to only loan or sell copies for academic and scientific purposes. The author
reserves other publication rights and no part of this dissertation can be reproduced without her
authorization.

__________________________________
Ana Luísa Manzaro da Costa
Rua Pedro de Toledo, 205, apt. 62
CEP: 12243-740, São José dos Campos - SP
iii

A NUMERICAL INVESTIGATION ON WAVY LEADING

EDGE APPLICATION IN TRANSONIC FLOW CONTROL

Ana Luísa Manzaro da Costa

Dissertation Committee Composition:

Prof. Dr. Adson Agrico de Paula Chairperson/Advisor - ITA


MSc Grace Rodrigues de Lima Co-advisor - Embraer
Prof. Dr. Roberto Gil Annes da Silva Internal Member - ITA
PhD Marcello do Areal Souto Ferrari External Member - Embraer

ITA
iv

To my mother, father and sisters,


for the unlimited love and support.

To my grandparents,
Neyde and Mario Manzaro,
forever in my heart.
v

Acknowledgements

I am most fortunate to say there are many people who deserve my sincerest gratitude
for their contribution in this work.
Regarding the technical aspects of the dissertation, I’ll begin by thanking my advisor,
Prof. Dr. Adson Agrico de Paula, for suggesting such an interesting topic and providing the
greatest part of the related literature. With your enthusiasm, sharing your passion for humpback
whales was inevitable. Just as fundamental was my co-advisor, Grace Rodrigues de Lima, MSc,
who guided and supported me through the entire process, and without whom I would be simply
lost. I am also grateful for your frequent requests on status updates, which put me back on track
every time, and for sharing your career experiences and valuable advices, which deeply
enriched my own professional and personal path.
As for the numerical investigation itself, I must thank a few colleagues from Embraer.
I am deeply indebted to Pedro Alberto Godinho Ciloni, MSc, for setting, performing and post-
processing all CFD simulations reported in this thesis; and to Marcello do Areal Souto Ferrari,
PhD, for walking me through an incredible mesh generation process. I will never be able to
repay you two for selflessly sharing such a great amount of time and knowledge with me. I
would also like to recognize the aid of the following: Daniel Martins da Silva, MSc,
Maximiliano Alberto Fernandes de Souza, MSc, Kelvin Cristofalo de Morais and Gilberto
Bueno Luque Filho.
Even though the experimental investigation had to be cancelled, I want to extend my
appreciation to Prof. Dr. Roberto Gil Annes da Silva, Bruno Ricardo Massucatto Padilha, MSc,
and Vinícius Arquimedes Sepetauskas, colleagues from ITA, for the effort spent on seeing it
through.
I would also like to express my gratitude to Embraer, ITA and Fundação Casimiro
Montenegro Filho for the opportunity and resources that enabled me to become a Master of
Engineering.
And last but definitely not least, I must thank those whose moral support was nothing
short of vital to me. My parents, Silvana and Jairo, and my sisters, Natália and Isadora, for the
lifetime of love, encouragement and comfort that lead me to this accomplishment. Victor, for
trying to help in every possible way and never giving up on me. And my dearest friends, the
ones I’ll make sure get to read this page, for always cheering me up and for making this journey
tremendously easier.
vi

"Never limit yourself because of others’ limited imagination;


never limit others because of your own limited imagination”.
(Mae Jemison)
vii

Resumo

A extraordinária capacidade de manobra das baleias jubarte tem sido atribuída aos tubérculos
encontrados em suas nadadeiras peitorais. Bordos de ataque ondulados, inspirados em tais
tubérculos, se mostraram benéficos como um meio de controle passivo de escoamento para
corpos aerodinâmicos em regime subsônico. Este trabalho pretende contribuir para a crescente
pesquisa na aplicação de tubérculos em regime transônico, por meio de uma investigação
numérica do desempenho de uma asa infinita RAE 2822 com bordo de ataque ondulado. Este
primeiro estudo em um perfil supercrítico foi conduzido com o uso de mecânica dos fluidos
computacional. As simulações foram executadas na faixa de velocidade de 0.5 a 0.86 Mach,
em três regimes de turbulência e em três coeficientes de sustentação, em um total de 117 casos
de teste. Análises de aumento de arrasto, decomposição de arrasto e topologia do escoamento
foram empregadas para avaliar o efeito dos tubérculos no coeficiente de arrasto. Os resultados
mostram que o bordo de ataque ondulado induz uma distribuição de pressão periódica ao longo
da envergadura. Gradientes adversos de pressão mais fortes são encontrados atrás dos vales, o
que causa uma antecipação da formação de ondas de choque ao longo da corda nas seções
transversais de vales em comparação às seções de pico. Na maioria dos casos de teste, os
arrastos viscoso, invíscido e total são aumentados pelo bordo de ataque ondulado. Ainda assim,
os casos de teste em que os tubérculos trouxeram uma diminuição no arrasto estão dentro do
envelope de número de Reynolds, número de Mach e coeficiente de sustentação
experimentados por aeronaves comerciais em cruzeiro – um resultado promissor para a
aplicação de bordo de ataque ondulado em controle de escoamento transônico.
viii

Abstract

The extraordinary maneuverability of humpback whales has been attributed to the tubercles
found on their pectoral flippers. Tubercle-inspired wavy leading edge has been shown to be
beneficial as a means of passive flow control for aerodynamic bodies in subsonic regime. This
work aims to contribute to the recently growing research on the application of tubercles in
transonic flow control, by numerically investigating the performance of a RAE 2822 infinite
wing with wavy leading edge. This first study on a supercritical profile has been conducted
using computational fluid dynamics. Simulations were performed in the 0.5 to 0.86 Mach range,
at three different Reynolds numbers and at three lift coefficients, in a total of 117 test cases.
Drag rise, drag component breakdown and flow topology analyses were employed to assess the
effect of tubercles in drag coefficient. Results show that the wavy leading edge leads to a
periodic spanwise pressure distribution. Stronger adverse pressure gradients are found behind
troughs, which causes anticipated chordwise shock wave formation at trough cross-sections in
comparison to peak cross-sections. In the majority of test cases, inviscid, viscous and total drag
are increased by the wavy leading edge. Nevertheless, test cases in which tubercles led to a
decrease in drag are within the Reynolds number, Mach number and lift coefficient conditions
experienced by commercial aircraft during cruise – a promising result for wavy leading edge
application in transonic flow control.
ix

List of Figures

Figure 1 – Biologically inspired aerodynamic design (WAHL, 2016) .................................... 21

Figure 2 – Humpback whale breaching and detailed view of tubercled pectoral flippers

(CONFORTH, 2013). ............................................................................................................... 22

Figure 3 – WhalePower Corporation tubercle-enhanced industrial ceiling fan (left) and wind

turbine concept (right) (WHALEPOWER CORPORATION, 2018). ...................................... 23

Figure 4 – Tubercle-inspired rear wing (SOMERFIELD, 2014) ............................................. 24

Figure 5 – Total drag component breakdown (adapted from TORENBEEK, 1982) ............... 27

Figure 6 – Boundary Layer separation (HOUGHTON et al., 2012) ........................................ 30

Figure 7 – Bubble trapped between separation and reattachment points (HOUGHTON et al.,

2012). ........................................................................................................................................ 30

Figure 8 – Characteristic short bubble pressure distribution (RUSSELL, 1979) ..................... 31

Figure 9 – Pressure coefficients of short and long laminar separation bubbles (RINOIE, 2009)

.................................................................................................................................................. 31

Figure 10 – Flow regimes according to Mach number ranges (ANDERSON, 2001) .............. 32

Figure 11 – Progression of shock waves with increasing Mach number in transonic flow

(MASON, 2018). ...................................................................................................................... 34

Figure 12 – Variation of drag coefficient with freestream Mach number (ANDERSON, 2001).

.................................................................................................................................................. 35

Figure 13 – Weak shock and its effect on the boundary layer (HOUGHTON et al., 2012). ... 36

Figure 14 – Differences between near-normal shock with local flow separation (left) and

lambda shock with complete flow separation (right) (HOUGHTON et al., 2012). ................. 37
x

Figure 15 – Lift, drag and aerodynamic efficiency data for the smooth (solid line) and tubercled

flippers, evidencing significant improvements by the addition of tubercles, Re = 500,000, 𝑀∞

= 0.2 (MIKLOSOVIC et al., 2004). ......................................................................................... 38

Figure 16 – Lift coefficient dependence on tubercle geometry, Re = 183,000, 𝑈∞ = 1.8 m/s

(LEVSHIN et al., 2006). ........................................................................................................... 39

Figure 17 – Airfoil geometry effect on wavy leading edge performance, Re = 120,000, 𝑈∞ =

25 m/s (HANSEN et al., 2009). ................................................................................................ 40

Figure 18 – Force measurement results for control foil and tubercled test foil (STANWAY,

2008). ........................................................................................................................................ 43

Figure 19 – Pressure contours and streamlines for a NACA 63-021 model with straight (left)

and tubercled (right) leading edge, Re = 1,000,000, α = 10° (PATERSON et al., 2003) ........ 44

Figure 20 – Flow visualization of airfoils at 12° (pre-stall) and 24° (post-stall) angle of attack,

Re = 183,000, 𝑈∞ = 1.8 m/s (LEVSHIN et al., 2006) ............................................................ 45

Figure 21 – Oil flow visualization and its interpretation for baseline (α = 15°) and tubercled (α

= 20°) configurations, Re = 170,000, 𝑈∞ = 12 m/s (ZVERKOV et al., 2008). ...................... 46

Figure 22 – Separated flow regions (green areas) over the suction surface of the smooth and

wavy airfoils in pre-stall regime, Re = 120,000, 𝑈∞ = 25 m/s (ROSTAMZADEH et al., 2014).

.................................................................................................................................................. 46

Figure 23 – Normalized streamwise vorticity, Re = 120,000, 𝑈∞ = 25 m/s, α = 2°

(ROSTAMZADEH et al., 2014)............................................................................................... 48

Figure 24 – Time-averaged streamlines showing deflection of oncoming flow (left) and

secondary flow (right), Re = 120,000, α = 20° (SKILLEN et al., 2015). ................................. 49

Figure 25 – Pressure coefficient comparison between baseline (yellow) and wavy (blue for

peaks and green for troughs) NACA 0012 configurations, Re = 1,000,000, 𝑀∞ = 0.8, α = 0°

(PADILHA, 2017). ................................................................................................................... 50


xi

Figure 26 – Tubercle effect on pressure distribution, with an indication of lambda-shock

suppression for the wavy configuration, Re = 1,000,000, 𝑀∞ = 0.7, α = 2° (SEPETAUSKAS

et al., 2018). .............................................................................................................................. 50

Figure 27 – Schlieren visualization of baseline (left) and tubercled (right) wings, Re = 820,000,

𝑀∞ = 0.81, α = 0° (ASGHAR et al., 2017) ............................................................................. 51

Figure 28 – Pressure coefficient contours for (a) straight and (b) tubercled leading edge wings,

Re = 820,000, 𝑀∞ = 0.8, α = 0° (PEREZ et al., 2018) ........................................................... 51

Figure 29 – Turbulent Kinetic Energy (TKE) computational results for straight and tubercled

wings, Re = 820,000, 𝑀∞ = 0.8, α = 0° (PEREZ, 2018). ....................................................... 52

Figure 30 – RAE 2822 airfoil normalized in XZ-plane............................................................ 55

Figure 31 – WLE geometry definition and positioning relative to the baseline leading edge. 56

Figure 32 – SLE (left) and WLE (right) RAE 2822 wings (only four wavelengths are shown).

.................................................................................................................................................. 56

Figure 33 – Far field definition (left) and detailed views of the wing (right) for the WLE

geometry (only four wavelengths shown). ............................................................................... 58

Figure 34 – Leading edge detailed view for the SLE (left) and WLE (right) grids. ................ 58

Figure 35 – Coarse, Medium, Fine and Finest grids for SLE geometry................................... 61

Figure 36 – Dimensionless viscous wall distance (y+) contour for SLE and WLE wings (top

view) at 𝑅𝑒 = 6.5𝑥106, 𝑀∞ = 0.72 and 𝐶𝐿 = 0.5. ............................................................... 65

Figure 37 – Coarse, Medium and Fine grids 𝐶𝐷 convergence (left) with zoomed in view (right)

at validation case flow conditions. ........................................................................................... 67

Figure 38 – Coarse, Medium and Fine grids 𝐶𝐿 convergence (left) with zoomed in view (right)

at validation case flow conditions. ........................................................................................... 67

Figure 39 – Finest grid 𝐶𝐷 and 𝐶𝐿 convergence at validation case flow conditions. .............. 67
xii

Figure 40 – 𝐶𝐿 and 𝐶𝐷 convergence for Coarse grid with periodic and symmetric boundary

conditions at validation case flow conditions. .......................................................................... 68

Figure 41 – 𝜏𝑥 contours for Coarse grid with periodic and symmetric boundary conditions at

validation case flow conditions. ............................................................................................... 69

Figure 42 – Re6.5 grid 𝐶𝐷 convergence (left) with zoomed in view (right) at validation case

flow conditions. ........................................................................................................................ 70

Figure 43 – Re6.5 grid 𝐶𝐿 convergence (left) with zoomed in view (right) at validation case flow

conditions. ................................................................................................................................ 71

Figure 44 – Comparison of 𝐶𝑃 distribution along the chord at validation case flow conditions.

.................................................................................................................................................. 71

Figure 45 – 𝐶𝐷 convergence for 2D and 3D simulations (left) with zoomed in view (right) at

𝑅𝑒 = 6.5𝑥106, 𝑀∞ = 0.72 and 𝐶𝐿 = 0.7. ............................................................................. 72

Figure 46 – Drag rise curves for SLE and WLE wings at 𝑅𝑒 = 1𝑥106. ................................. 73

Figure 47 – Drag rise curves for SLE and WLE wings at 𝑅𝑒 = 6.5𝑥106. .............................. 73

Figure 48 – Drag rise curves for SLE and WLE wings at 𝑅𝑒 = 2.2𝑥107. .............................. 74

Figure 49 – Percentage drag increment at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.1.................................... 78

Figure 50 – Percentage drag increment at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.3.................................... 79

Figure 51 – Percentage drag increment at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.5.................................... 80

Figure 52 – Percentage drag increment at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.1................................. 81

Figure 53 – Percentage drag increment at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.3................................. 82

Figure 54 – Percentage drag increment at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.5................................. 83

Figure 55 – Percentage drag increment at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.1................................. 84

Figure 56 – Percentage drag increment at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.3................................. 85

Figure 57 – Percentage drag increment at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.5................................. 86

Figure 58 – Mach number contours at 𝑅𝑒 = 1𝑥106, 𝑀∞ = 0.68 and 𝐶𝐿 = 0.1. ................... 87


xiii

Figure 59 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom) at

𝑅𝑒 = 1𝑥106, 𝑀∞ = 0.68 and 𝐶𝐿 = 0.1. ................................................................................ 88

Figure 60 – Mach number contours at 𝑅𝑒 = 1𝑥106, 𝑀∞ = 0.68 and 𝐶𝐿 = 0.5. ................... 89

Figure 61 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom) at

𝑅𝑒 = 1𝑥106, 𝑀∞ = 0.68 and 𝐶𝐿 = 0.5. ................................................................................ 90

Figure 62 – Mach number contours at 𝑅𝑒 = 1𝑥106, 𝑀∞ = 0.76 and 𝐶𝐿 = 0.5. ................... 91

Figure 63 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom) at

𝑅𝑒 = 1𝑥106, 𝑀∞ = 0.76 and 𝐶𝐿 = 0.5. ................................................................................ 92

Figure 64 – Mach number contours at 𝑅𝑒 = 1𝑥106, 𝑀∞ = 0.78 and 𝐶𝐿 = 0.5. ................... 93

Figure 65 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom) at

𝑅𝑒 = 1𝑥106, 𝑀∞ = 0.78 and 𝐶𝐿 = 0.5. ................................................................................ 94

Figure 66 – Mach number contours at 𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.76 and 𝐶𝐿 = 0.3................. 95

Figure 67 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom) at

𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.76 and 𝐶𝐿 = 0.3. ............................................................................. 96

Figure 68 – Mach number contours at 𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.80 and 𝐶𝐿 = 0.3................ 98

Figure 69 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom) at

𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.80 and 𝐶𝐿 = 0.3. ............................................................................. 99

Figure 70 – Mach number contours at 𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.82 and 𝐶𝐿 = 0.3............... 100

Figure 71 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom) at

𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.82 and 𝐶𝐿 = 0.3. ........................................................................... 101

Figure 72 – Mach number contours at 𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.78 and 𝐶𝐿 = 0.5............... 103

Figure 73 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom) at

𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.78 and 𝐶𝐿 = 0.5. ........................................................................... 104

Figure 74 – Mach number contours at 𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.82 and 𝐶𝐿 = 0.5............... 106


xiv

Figure 75 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom) at

𝑅𝑒 = 2.2𝑥107, 𝑀∞ = 0.82 and 𝐶𝐿 = 0.5. ........................................................................... 107


xv

List of Tables

Table 1 – Grid Refinement Parameters. 60

Table 2 – Final grid parameters 61

Table 3 – Performance comparison cases matrix 63

Table 4 – SLE wing validation case (adapted from COOK et al., 1979). 64

Table 5 – Grid independence study results. 66

Table 6 – 𝐶𝐿 and 𝐶𝐷 comparison to validation case. 70

Table 7 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 1𝑥106. 74

Table 8 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 6.5𝑥106. 75

Table 9 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 2.2𝑥107. 76

Table 10 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.1. 77

Table 11 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.3. 78

Table 12 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.5. 79

Table 13 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.1. 81

Table 14 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.3. 82

Table 15 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.5. 83

Table 16 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.1. 84

Table 17 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.3. 85

Table 18 – Effect of tubercles on drag coefficient (Δ𝐶𝐷) at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.5. 86


xvi

List of Abbreviations

AIAA American Institute of Aeronautics and Astronautics

CAD Computer-Aided Design

CATIA CAD software developed by Dassault Systèmes

CFD Computational Fluid Dynamics

CFD++ CFD solver developed by Metacomp Technologies, Inc.

EFD Experimental Fluid Dynamics

ICEM CFD Mesh generation software developed by ANSYS, Inc.

ITA Technological Institute of Aeronautics (Instituto Tecnológico de

Aeronáutica)

NACA National Advisory Committee for Aeronautics

NASA National Aeronautics and Space Administration

PIV Particle Image Velocimetry

RAE Royal Aircraft Establishment

SA-RC-QCR Spalart-Allmaras model with rotation/curvature correction and

quadratic constitutive relation

SLE Straight leading edge

Tecplot Data visualization software developed by Tecplot, Inc.

TKE Turbulent Kinetic Energy

UAV Unmanned Air Vehicle

V&V Verification and Validation

WLE Wavy leading edge


xvii

List of Symbols

𝐴 Amplitude

𝑐 Local speed of sound

𝑐̅ Mean chord length

𝐶𝐷 Drag coefficient

𝐶𝐷𝑖 Inviscid drag coefficient

𝐶𝐷𝑣 Viscous drag coefficient

𝐶𝑓 Skin friction coefficient

𝐶𝐿 Lift coefficient

𝐶𝑀 Moment coefficient

𝐶𝑃 Pressure coefficient

𝐷 Drag

𝑒 Internal energy

𝐿 Lift

𝑙 Characteristic length

𝑀 Local Mach number

𝑀𝑑𝑑 Drag-divergence Mach number

𝑀∞ Freestream Mach number

𝑃 Local pressure

𝑝 Pressure normalized by the dynamic pressure (ρU²)

𝑃∞ Freestream pressure

𝑞∞ Dynamic pressure

𝑅𝑒 Reynolds number
xviii

𝑆 Reference area

𝑈∞ Freestream velocity (x-component)

𝑢 Velocity vector normalized by a velocity scale (U)

𝑢̅ Mean (time-average) component (𝑢)

𝑢′ Fluctuating component (𝑢)

𝑢𝑓 Friction velocity

𝑡 Time normalized by the convective time scale (L/U)

𝑉 Local flow velocity

𝑉∞ Freestream velocity

x Distance along the flat plate

Δ𝑦 First cell height

𝑦+ Dimensionless wall distance

𝛻 Spatial gradient operator normalized by a length scale (L)

𝛼 Angle of attack

𝛿 Boundary layer thickness

𝜆 Wavelength

𝜇∞ Dynamic viscosity

𝜌∞ Freestream density

𝜏𝑤 Local wall shear stress


xix

Table of Contents

1 INTRODUCTION .......................................................................................................... 21

2 THEORETICAL FUNDAMENTALS AND LITERATURE REVIEW ................... 26


2.1 Fundamentals of Aerodynamics ............................................................................ 26
2.1.1 Aerodynamic Coefficients ........................................................................................ 26
2.1.2 Reynolds Number ..................................................................................................... 27
2.1.3 Boundary Layer and Flow Separation ...................................................................... 29
2.1.4 Mach Number ........................................................................................................... 32
2.1.5 Transonic Regime ..................................................................................................... 33
2.1.6 Shock-Wave/Boundary-Layer Interactions .............................................................. 35
2.2 The Wavy Leading Edge ........................................................................................ 38
2.2.1 Tubercle Geometry ................................................................................................... 39
2.2.2 Airfoil Geometry ...................................................................................................... 40
2.2.3 Flow Dimension ....................................................................................................... 41
2.2.4 Reynolds Number ..................................................................................................... 42
2.2.5 Flow Topology ......................................................................................................... 44
2.2.6 Flow Control Mechanism ......................................................................................... 47
2.2.7 Transonic Regime Effect .......................................................................................... 49
2.2.8 Aims and objectives of this thesis ............................................................................ 53

3 METHODOLOGY ......................................................................................................... 54
3.1 Physical modeling ................................................................................................... 54
3.2 Airfoil and Tubercle Geometry ............................................................................. 55
3.3 Computational Domain .......................................................................................... 56
3.3.1 Baseline and Tubercled Wings Design ..................................................................... 56
3.3.2 Grid Generation Strategy .......................................................................................... 57
3.4 Simulation Setup ..................................................................................................... 62
3.5 Verification and Validation ................................................................................... 63

4 RESULTS AND DISCUSSION ..................................................................................... 65


4.1 Solution Verification and Validation .................................................................... 65
4.1.1 Dimensionless Viscous Wall Distance ..................................................................... 65
4.1.2 Grid Convergence ..................................................................................................... 66
xx

4.1.3 Comparison to Experimental Data ........................................................................... 70


4.1.4 CL Driver Convergence............................................................................................ 72
4.2 Wavy Leading Edge Effect in Transonic Regime ................................................ 73
4.2.1 Drag Rise .................................................................................................................. 73
4.2.2 Drag Breakdown ....................................................................................................... 76
4.2.3 Flow Topology ......................................................................................................... 87

5 CONCLUSIONS ........................................................................................................... 108

REFERENCES ..................................................................................................................... 111


21

1 Introduction

The principle of flow control is the manipulation of a flow field in order to achieve a
particular design objective. Over the past few decades, a variety of flow control techniques have
been developed and implemented in the aircraft industry – from vortex generators on wings for
flow separation delay to chevrons on exhaust nozzles for noise mitigation.
Flow control methods can be classified according to energy expenditure. Both given
examples are classified as passive, that is, they do not involve any extra power input.
Conversely, active techniques require supplementary power to operate actuation devices, such
as pneumatic systems for boundary layer suction and blowing.
Regarding overall performance, active flow control devices have an advantage over
passive methods, since they can be turned on and off as needed and aim for optimized operation
in dynamic flight conditions. Passive devices, on the other hand, rely on geometric
modifications and thus may cause performance degradation when operating out of the design
condition. Nevertheless, the implementation of active control methods is considerably more
complex than that of passive mechanisms, mostly due to their testing and manufacturing
complexity (COLLIS et al., 2004). Hence, the lower cost and relative simplicity of passive
flow control devices is still appealing even in modern aircraft design.
Not seldom, the inspiration for passive geometric modifications comes from observing
and reproducing nature’s time-tested engineering solutions, a line of research known as
biomimicry. As depicted in Figure 1, the addition of winglets to wings, for instance, comes
from the morphology of wingtip feathers found in birds of prey.

Figure 1 – Biologically inspired aerodynamic design (WAHL, 2016)


22

Biomimetic research has also brought to light the remarkable swimming pattern of
humpback whales. Known as the most “acrobatic” of baleen whales (LEATHERWOOD et al.,
1988), the humpback whale presents exquisite turning, surfacing and diving capabilities. It is
also equipped with the largest pectoral flippers among cetaceans, reaching almost a third of its
body length. As observed by researchers, the flippers were found to be accountable for the
greatest part of said maneuverability, allowing for sharply executed banked turns (EDEL;
WINN, 1978).
In addition to their singular proportions, the flippers of the humpback whale present
unique protuberances on its leading edge surface, as shown in Figure 2. It was suggested that
these protuberances or tubercles also contributed to the maneuverability of the humpback whale
(BUSHNELL; MOORE, 1991). In fact, studies indicated leading edge tubercles could act as
lift enhancing devices, making the whales capable of maintaining high lift values even at high
angles of attack (FISH; BATTLE, 1995).

Figure 2 – Humpback whale breaching and detailed view of tubercled pectoral flippers
(CONFORTH, 2013).

The growing interest on tubercles has given rise to promising findings regarding their
potential application in flow control. Airfoils with tubercled or wavy leading edge show
increased momentum exchange in the boundary layer, contributing to separation delay and
consequent maximum lift and stall angles (FISH; BATTLE, 1995; MIKLOSOVIC et al., 2004).
23

Tubercles also promote softer stall characteristics and higher post-stall lifts (JOHARI et al.,
2007), and have been shown to increase pre-stall lift in some cases (HANSEN et al., 2009;
JOHARI et al., 2007; MIKLOSOVIC et al., 2004; STANWAY, 2008). Hence, potential
application include lifting surfaces which experience high adverse pressure gradients, have high
lift requirements, operate near the stall angle and experience tip stall (HANSEN, 2012).
The wavy leading edge concept has already been implemented in the industry. The
WhalePower Corporation commercializes tubercle-enhanced industrial ceiling fans and has
wind turbines under development, as shown in Figure 3. High volume, low speed (HVLS) fans
with tubercled blades have been demonstrated to reduce power consumption in about 20%
(MCLEOD, 2010), being capable of operating at lower speeds to provide the same amount of
airflow as conventional fans, while also being considerably less noisy. Tubercle-enhanced wind
turbines show increased electrical generation at moderate wind speeds when compared to
conventional designs (WIND ENERGY INSTITUE OF CANADA, 2008), additionally
overcoming major limitations in wind power generation such as operating in low unsteady or
turbulent winds and reduction of noise associated to tip-stall (DEWAR, 2014).

Figure 3 – WhalePower Corporation tubercle-enhanced industrial ceiling fan (left) and wind
turbine concept (right) (WHALEPOWER CORPORATION, 2018).

Tubercled rotary wings can also be applied to aircraft. When high forward speed is
required, helicopter rotors need to operate at high angles of attack. Since the velocity of the
flow relative to the retreating blade is considerably lower, this blade is susceptible to dynamic
stall (TANG; DOWELL, 1992). A wavy leading edge would aid increasing the stall angle and
reducing the severity of stall, contributing to reduced fatigue and increased service life
(HANSEN, 2012).
Besides rotary wings, it has been shown (SUDHAKAR et al., 2017) that small UAVs
(Unmanned Air Vehicles) with tubercled fixed wings present higher stall angle, lower drag and
24

higher lift to drag ratio when compared to vehicles with unmodified leading edge. Operating in
low-speed laminar or transitional regime, UAVs are likely to experience boundary layer
separation at small angles of attack, which becomes critical during take-off and landing
(MUELLER; DELAURIER, 2003). Incorporating slats and flaps to such small aircraft could
outweigh the potential benefits; tubercles, on the other hand, could reduce the minimum stall-
speed without adding a significant increase in drag or weight, being easily implemented
(HANSEN, 2012).
The wavy leading edge design has also been recently implemented on Formula One
cars, as shown in Figure 4. The top flap is an inverted wing located on the rear-deck of racecars,
used to provide downforce during cornering and reduce drag through straights (BOLZON,
2015). Top flaps with wavy leading edge allow for increased stall angle and augmented
performance for a greater range of angles of attack (MADIER, 2014).

Figure 4 – Tubercle-inspired rear wing (SOMERFIELD, 2014)

Even though wavy leading edge applications seems to be on the rise, they are still
restrained to the subsonic regime of the humpback whale. Bolzon (2015), however, suggests
there is potential for tubercles to be implemented in transonic and supersonic flight, where wave
drag could be significantly reduced. For an undulated leading edge, spanwise pressure
distribution fluctuates with a spatial periodicity (SKILLEN et al., 2015; WATTS; FISH, 2001),
presenting alternating spanwise regions with higher and lower pressure values than those of a
smooth wing. Since pressure and critical Mach number are related, tubercled wings will have
25

varying critical Mach number along span. This fluctuation could possibly lead to a delayed and
less intense drag rise associated to the critical Mach number.
This hypothesis is particularly relevant for transonic wings, which have subsonic cruise
speed limited by the onset of drag rise (WHITCOMB, 1965). However, research on wavy
leading edge performance in transonic regime is still very limited. Further understanding on
how tubercles interact with the complex dynamics of the transonic regime is yet to be explored,
which is the motivation of the present work.
The flow physics in modern aircraft design can be achieved by two methods:
Experimental Fluid Dynamics (EFD) and Computational Fluid Dynamics (CFD). Proper setup
precautions taken, EFD is capable of providing highly reliable data for complex problems
without the inherent bias of mathematical approximations and assumptions (STUMPE, 2018).
Nevertheless, besides time and cost constraints that may restrict the number of test cases,
operational limitations often make EFD platforms like wind tunnels unable to reproduce the
desired flow conditions (STUMPE, 2018). Overcoming such obstacles, current CFD modeling
and simulation enables the evaluation of virtually unlimited flow conditions at a much lower
cost (KUZMIN, 2018).
Taking this methodology comparison into consideration, in this thesis, the aerodynamic
effect of the wavy leading edge in transonic regime is investigated through computational fluid
dynamics (CFD). In order to provide a substantial contribution to this field of research, this
numerical investigation will focus on evaluating the effects of Reynolds and Mach numbers on
wavy leading edge application as a means of transonic flow control.
This work is divided into five chapters. This first chapter gives an introductory overview
on the background and current application of wavy leading edge, as well as the motivation and
objectives of this research. The second chapter follows with the theoretical fundamentals of
aerodynamics employed in the investigation, along with a bibliographic review of the research
made on wavy leading edge thus far, highlighting its most relevant aspects. The adopted
methodology and development are detailed in chapter 3, while results are presented and
discussed in chapter 4. The last chapter summarizes the results and conclusions obtained,
additionally suggesting directions for future works on wavy leading edge application in
transonic flow control.
26

2 Theoretical Fundamentals and Literature Review

In order to investigate the wavy leading edge phenomena in transonic flow, an overview
of the relevant aerodynamic effects involved in this flow regime is given, followed by a
summary of the research on wavy leading edge application and its findings thus far.

2.1 Fundamentals of Aerodynamics

2.1.1 Aerodynamic Coefficients

As a standard practice, aerodynamic characteristics are given in terms of dimensionless


coefficients. By using a ratio of forces instead of the forces themselves, these coefficients enable
comparison between designs regardless of geometric characteristics. The aerodynamic
coefficients used in this work are defined as below:

𝑃 − 𝑃∞
Pressure coefficient, 𝐶𝑃 : 𝐶𝑃 =
𝑞∞
𝜏𝑤
Skin friction coefficient, 𝐶𝑓 : 𝐶𝑓 =
𝑞∞
𝐿
Lift coefficient, 𝐶𝐿 : 𝐶𝐿 =
𝑞∞ 𝑆
𝐷
Drag coefficient, 𝐶𝐷 : 𝐶𝐷 =
𝑞∞ 𝑆

Since the aim of this work is to evaluate transonic performance, a breakdown of drag
sources in external fluid dynamics is given in Figure 5. According to the decomposition given
by Torenbeek (1982), aerodynamic drag is divided into inviscid and viscous drag. Inviscid drag
comprises lift-dependent drag due to vortex shedding (vortex drag) and drag due to
compressibility effects, in the form of wave drag due to lift (induced drag) and due to volume
(inviscid form drag). Viscous drag, on the other hand, depends on the surface roughness
(friction drag), shape (viscous form drag) and boundary layer growth (wake drag). The sum of
friction and viscous form drag composes profile drag.
27

Figure 5 – Total drag component breakdown (adapted from TORENBEEK, 1982)

2.1.2 Reynolds Number

One of the most important similarity parameters in aerodynamics, the Reynolds number
(𝑅𝑒) measures the ratio of inertial forces to viscous forces in a flow. Its definition is as follows:

𝜌∞ 𝑉∞ 𝑙
𝑅𝑒 = (2.1)
𝜇∞
Laminar flow results from dominant viscous forces, whereas turbulent flow results from
dominant inertial forces. By giving the ratio of inertial forces to viscous forces in a flow, the
Reynolds number can be understood as a measure of the balance between laminar and turbulent
flow. At a sufficiently low Reynolds number, the viscosity is dominant and the flow stays
laminar. As the Reynolds number increases, inertial forces become more significant and the
flow transitions to turbulent.
Given the difference between laminar and turbulent flows, the variation of the Reynolds
number can strongly modify the flow topology to which the airfoil is subjected. To aid the
understanding of the Reynolds number influence on flow dynamics around airfoils, Carmichael
(1981) provides a discussion on the different flow regimes typically encountered in different
Reynolds number ranges. A summary of such discussion is given below:

 𝑅𝑒 < 30,000: at the lower end of this regime, the flow is completely viscous, having
no application in airfoil aerodynamics. At the upper end, the boundary layer is still
28

completely laminar and artificial boundary layer tripping is difficult to achieve. At


higher lift coefficients, the flow produces laminar flow separation without
reattachment.
 30,000 < 𝑅𝑒 < 3,000,000: extensive laminar flow is possible throughout the
entire Reynolds number range. At the lower end of the range, laminar flow
separation can transition to turbulent reattachment, however forming an extensive
laminar separation bubble that degrades airfoil performance significantly. Artificial
boundary layer tripping is possible, but combinations of airfoil thickness and camber
are limited. As the Reynolds number gets higher, the influence of the laminar bubble
diminishes and airfoil performance increases considerably, as well as artificial
transition to turbulent flow gets easier. At 𝑅𝑒 = 1,000,000, the bubble can still
cause a drag coefficient penalty of about 0.001, but becomes completely
unimportant for 𝑅𝑒 > 4,000,000.
 3,000,000 < 𝑅𝑒 < 40,000,000: extensive natural laminar flow is still possible,
however limited to specific design and flight conditions. Very low drag coefficients
are possible, and the laminar separation bubble is no longer a problem. Given the
correct design methods, the turbulent boundary layer on the aft part of the wing can
stay attached through very severe adverse pressure gradients. At the upper end of
this Reynolds number range, strong favorable pressure gradients are required in
order to obtain extensive natural laminar flow, and artificial boundary layer
stabilization has significant payoff in performance.
 𝑅𝑒 > 40,000,000: boundary layer flow is mostly turbulent, and as the Reynolds
number increases (over 109), drag becomes primarily turbulent friction drag.

Typically, for aerodynamic airfoils the flow will be completely laminar for 𝑅𝑒 <
100,000. Between 103 and 106, the boundary layer is likely to present laminar, transitional and
turbulent state, and above this range, the flow tends to become completely turbulent. The value
at which transition occurs is called the critical Reynolds number. Nevertheless, it is important
to notice that flow transition from laminar to turbulent is not defined by the Reynolds number
alone. The Reynolds number can be obtained from the non-dimensional formulation of the
incompressible Navier-Stokes equations for a Newtonian fluid, which can be written in terms
of the Lagrangian derivative:
29

∇. 𝑢 = 0 (2.2)
𝜕𝑢 1 2
+ (𝑢. ∇)𝑢 = ∇𝑝 + ∇ 𝑢 (2.3)
𝜕𝑡 𝑅𝑒
1 2 3 4

In this form, the left-hand side of equation 2.3 represents the inertial terms – local
acceleration (1) and convective acceleration (2), while the right-hand side is defined by the
pressure gradient (3) and the viscous force (4), the latter in terms of the Reynolds number.

At a sufficiently low Reynolds number, the viscous force term (4) of equation 2.3 is
dominant, making the left-hand side of the equation – which ultimately represents flow
instabilities – insignificant. Under these circumstances, the flow is laminar, as previously stated.
On the other hand, as the Reynolds number gets higher, the left-hand side of the equation starts
to affect the flow dynamics, causing instabilities.
In this high-Reynolds-number scenario, however, the pressure gradient also has a
considerable effect on flow characteristics. Since the pressure gradient is determined by the
geometry of the body, the flow behavior is therefore influenced by both the Reynolds number
and geometric characteristics – for instance, the flow around different airfoils may present
completely distinct flow dynamics at the same Reynolds number. This is of particular interest
to wavy leading edge research applied to aircraft, since the entire wing geometry is affected by
the insertion of tubercles.

2.1.3 Boundary Layer and Flow Separation

The boundary layer is the portion of the flow in the immediate vicinity of the
aerodynamic body where the effects of viscosity are dominant. Flow separation occurs when
the boundary layer is subjected to a strong adverse pressure gradient, that is, when pressure
increases rapidly in the flow direction, as shown in Figure 6. Under the influence of an adverse
pressure gradient, fluid elements along a streamline are slowed down. Inside the boundary layer,
where the velocity is already small due to friction forces, the fluid eventually comes to a stop
somewhere downstream and reverses its direction, as a result detaching from the surface of the
object (ANDERSON, 2001).
30

Figure 6 – Boundary Layer separation (HOUGHTON et al., 2012)

This separated flow leads to a drastic loss of lift (stalling) and increase in drag, more
specifically pressure drag. Typically, thick airfoils present trailing-edge stall – separation
occurs closer to the trailing edge, and loss of lift is somewhat smooth and gradual, while thin
airfoils present leading-edge stall – abrupt loss of lift with separation near the leading edge.
Thin airfoil flow separation is also distinguished by the existence of separation bubbles, as
shown in Figure 7.

Figure 7 – Bubble trapped between separation and reattachment points (HOUGHTON et al.,
2012).

Given the disturbances generated by flow separation, a separated laminar boundary


layer may undergo transition to turbulent, with characteristic thickening. Turbulent boundary
layers also withstand much stronger adverse pressure gradients when compared to the laminar
case. Thus, sufficient thickening and a sustainable adverse pressure gradient may allow the
31

separated turbulent boundary layer to reattach to the surface (HOUGHTON et al., 2012). This
results in a bubble of fluid to be trapped between the separation and reattachment points,
distance between which a constant pressure plateau is formed (Figure 8).

Figure 8 – Characteristic short bubble pressure distribution (RUSSELL, 1979)

The laminar separation bubble can be either short – of the order of 1% of chord length,
or long – from a few percent of chord length to almost the entire chord. Short bubbles remain
small with increasing airfoil incidence, and exert very little influence on the pressure
distribution over the airfoil. Long bubbles, on the other hand, rapidly increase in length with
increasing incidence, and exert a large effect on the value of the peak suction near the leading
edge, presenting a relatively flattened pressure distribution instead (Figure 9).

Figure 9 – Pressure coefficients of short and long laminar separation bubbles (RINOIE, 2009)
32

2.1.4 Mach Number

The Mach number (𝑀) is a dimensionless number defined as the ratio of the flow speed
to the local speed of sound:

𝑉
𝑀= (2.4)
𝑐

By definition, at Mach 1 the flow velocity is equal to the speed of sound, which implies
subsonic speed below this value and supersonic speed above it. When a body moves through a
fluid, the air molecules in the vicinity are disturbed and move around it. For sufficiently low
subsonic speeds, air density remains constant; whereas for higher speeds the air is compressed
and local air density is changed, affecting the resulting aerodynamic force acting on the body.
Analogously to the viscosity regimes defined by ranges of the Reynolds number, given
the Mach number there are typically four different compressibility regimes to be found –
subsonic, transonic, supersonic and hypersonic, as presented in Figure 10.

Figure 10 – Flow regimes according to Mach number ranges (ANDERSON, 2001)

Given the geometry of airfoils, the flow expands and accelerates as it moves around
them, which means the local Mach number is increased. In subsonic flow (Figure 10(a)), the
local Mach number remains less than unity everywhere in the flow field, and the streamlines
33

are smooth with continuous varying properties. This condition is usually achievable up to
freestream Mach number 𝑀∞ <0.8 (ANDERSON, 2001).
With increasing subsonic freestream speed, at some point over the surface of the body,
Mach 1 is exceeded and the flow transitions to supersonic. This transition is marked by an
abrupt increase in drag. In this regime, the flow is said to be transonic, with regions where both
subsonic and supersonic flow occur (Figure 10(b)). This mixed flow regions are called sonic
pockets, which initiate as soon as the local Mach number equals the unity and terminates in a
weak shock wave downstream, which decelerates the flow back to subsonic condition. If the
freestream Mach number is already supersonic, a bow shock is formed in front of the airfoil
and the flow behind it is subsonic. This subsonic flow accelerates over the airfoil to supersonic
values and terminates in a trailing edge shock (Figure 10(c)).
In the supersonic regime, the flow stays supersonic even after being decelerated by an
oblique shock wave (Figure 10(d)). In order to minimize drag, supersonic aerodynamic bodies
are generally sharp edged. Behind the shock, streamlines remain straight and parallel, but take
the direction of the wedge surface. Further downstream, expansion waves may also occur,
decreasing pressure inversely to shock waves.
Finally, the hypersonic flow is achievable when the freestream speed is above Mach 5.
In this regime, the shock wave moves closer to the body, forming a thin shock layer (Figure
10(e)). In this layer, viscous interaction and chemical reactions may take place, resulting in
extreme heating typical to space vehicles aerodynamics.
Since this thesis evaluates the performance of wavy leading edge in transonic flow,
further detail on this regime is given in the following subsection.

2.1.5 Transonic Regime

Transonic flow occurs when there are mixed subsonic and supersonic flow regions in
the flow field (ANDERSON, 2001). Figure 11 shows the development of the flow with
increasing Mach number.
At 𝑀∞ = 0.5, the flow is entirely subsonic. The freestream Mach number at which sonic
flow is first achieved over the aerodynamic body is 𝑀∞ = 0.72 – by definition, this is the
critical Mach number (𝑀∗ ) for this flow. As the freestream Mach number exceeds the critical
Mach number, a region of supersonic flow is developed, further being brought back to subsonic
speed by a shock wave (𝑀∞ = 0.77). With increasing freestream Mach number, the shock wave
34

moves aft and becomes stronger, and a second shock wave is bound to occur on the lower
surface as well (𝑀∞ = 0.82).

Figure 11 – Progression of shock waves with increasing Mach number in transonic flow
(MASON, 2018).

Transitioning from subsonic to supersonic flow is distinguished by a large increase in


drag. Figure 12 exemplifies the variation of the drag coefficient (𝐶𝐷 ) as a function of 𝑀∞ . Point
c depicts the already defined critical Mach number, past which the flow becomes supersonic
and the drag coefficient starts increasing. However, it is only after 𝑀∞ is increased past point e
that the drag coefficient rises by a much larger factor. This value of 𝑀∞ is defined as the drag-
divergence or drag-rise Mach number (𝑀𝑑𝑑 ), and the rapid drag increase is known as transonic
drag rise. The drag rise is associated to the shock wave required to bring the flow back to
subsonic speed. Shock waves introduce wave drag to the flow and induce boundary layer
thickening, eventually leading to flow separation (point f in Figure 12). Finally, past point g,
which was believed to be the sound barrier, the flow is entirely supersonic and drag coefficient
begins to decrease as transonic instabilities fade (ANDERSON, 2001).
35

Figure 12 – Variation of drag coefficient with freestream Mach number (ANDERSON, 2001).

For commercial aircraft cruising in transonic regime, particular attention is given to


point e in Figure 12. Beyond the drag-rise Mach number, the additional power and fuel
consumption required to further increase cruise speed increases drastically, and for that reason
the subsonic cruise speed of high performance aircraft is limited by the onset of transonic drag
rise (WHITCOMB, 1965). The use of wing sweepback delays this onset, however for practical
amounts of sweepback (approximately 35°) the onset still occurs well below the speed of sound
(WHITCOMB, 1965). Another method for increasing the cruise speed further is to adopt
supercritical airfoils, which are designed to delay the drag rise by causing the shock wave to
move to the rear part of the airfoil at the design condition, corresponding to a considerably
higher drag-divergence Mach number (WHITCOMB, 1965).
As previously mentioned, the drag rise is associated to the formation of shock waves.
To provide understanding the influence of shock waves on flow behavior, a brief discussion on
transonic boundary layer interactions is given in the following subsection.

2.1.6 Shock-Wave/Boundary-Layer Interactions

Since the undisturbed free-stream conditions downstream the trailing edge must be met,
shock waves are responsible for reducing the supersonic stream back to subsonic, and this speed
reduction comes with an abrupt increase in pressure. Hence, shock waves impose a strong
adverse pressure gradient on the boundary layer, which leads to its thickening and possible
36

separation from the surface - the latter effect more probable for a laminar boundary layer, which
may also transition to turbulent. Each shock-wave/boundary-layer interaction leads to a
different effect on the external flow.
With a relatively weak shock, the diffused pressure rise may cause only a gradual
thickening of the boundary layer, which in turn results in a family of weak compression waves
to develop ahead of the main shock, giving it a diffused base. Immediately behind the shock,
the boundary layer tends to get thinner, accelerating the flow – which may become supersonic
again. This process may be repeated several times before the mainstream flow settles to become
entirely subsonic (HOUGHTON et al., 2012), as shown in Figure 13. This condition is not
usually associated with boundary-layer separation.

Figure 13 – Weak shock and its effect on the boundary layer (HOUGHTON et al., 2012).

As the freestream Mach number increases, the first shock becomes much stronger,
resulting in a much lower Mach number behind it, as depicted in. Given a sufficiently high
freestream flow speed, a subsequent shock is not likely to occur. The rate of thickening of the
boundary layer increases and is followed by a local separation and reattachment of the boundary
layer at the base of the normal shock. In this scenario, transition to turbulence is expected to
occur behind the shock.
With still greater freestream Mach number, the pressure rise at the shock can cause
laminar boundary layer separation ahead of the main shock position. When separation takes
place, the flow suffers a sharp change in direction accompanied by an oblique shock that joins
the main shock at some distance from the surface. The resulting shock is called lambda shock
due to its geometry, as shown in Figure 14. Under these conditions, it is unlikely that the
37

boundary layer will reattach. The sudden separation is usually associated with an equally
sudden decrease in lift coefficient, a phenomenon known as shock stall (HOUGHTON et al.,
2012).

Figure 14 – Differences between near-normal shock with local flow separation (left) and
lambda shock with complete flow separation (right) (HOUGHTON et al., 2012).

Finally, since the turbulent boundary layer is far less susceptible to the effects of an
adverse pressure gradient, separation is not expected to happen for freestream Mach numbers
less than 1.3. Without separation, boundary layer thickening is rapid and no lambda formation,
secondary shock or thinning of the boundary layer behind the shock is expected. If separation
does occur, a lambda shock develops and the mainstream flow gets similar to that of a fully
separated laminar boundary layer.
38

2.2 The Wavy Leading Edge

Despite a few promising investigations at the time, it was not until the results from
Miklosovic et al (2004) that the wavy leading edge phenomena aroused interest among the
scientific community. Two models of idealized humpback whale flippers were built based on
the symmetrical NACA 0020 airfoil – one with a smooth leading edge and the other with a
sinusoidal one, mimicking the tubercles. Through wind tunnel measurements in low (500,000)
Reynolds numbers, it was shown that the addition of leading-edge tubercles delayed stall angle
by 40% while increasing lift and decreasing drag. Moreover, post-stall regime presented lower
drag for the tubercled flipper (MIKLOSOVIC et al., 2004).

Figure 15 – Lift, drag and aerodynamic efficiency data for the smooth (solid line) and
tubercled flippers, evidencing significant improvements by the addition of tubercles, Re =
500,000, 𝑀∞ = 0.2 (MIKLOSOVIC et al., 2004).

Following Miklosovic et al. (2004), several researches provided significant advances in


understanding the wavy leading edge phenomena as a flow control mechanism. The following
subsections summarize the past fourteen years of research regarding its most relevant aspects,
in order to both present the state of the art and the unexplored gaps in the field.
39

2.2.1 Tubercle Geometry

Wavy leading edge geometry is parameterized in terms of the wavelength (λ) and
amplitude (A) of tubercles as functions of chord lengths (𝑐̅). Levshin et al. (2006) conducted
the first investigations on the tubercle geometry effect on aerodynamic performance. Low
(183,000) Reynolds number water tunnel measurements compared a baseline 634-021 airfoil to
a tubercled foil with amplitude and wavelength of leading-edge protuberances ranging from 2.5
to 12% and 25 to 50% of the mean chord length (𝑐̅), respectively. It was found that the amplitude
of the protuberances have a much stronger influence on the performance of the airfoils than the
wavelength. Furthermore, as depicted in Figure 16, the smallest amplitude (8S) presented a
better overall performance in comparison to the other wavy configurations (8M, 8L).

Figure 16 – Lift coefficient dependence on tubercle geometry, Re = 183,000, 𝑈∞ = 1.8 m/s


(LEVSHIN et al., 2006).

Hansen et al. (2009) observed the same tendency by carrying out experimental tests of
full-span NACA 0021 and NACA 65-021 based models, at 𝑅𝑒 = 120,000 for a freestream
velocity 𝑈∞ = 25 m/s. For both airfoils, in the pre-stall regime decreasing amplitude increases
CLmax and stall angle, whereas in post-stall larger amplitudes contribute to smoother stall
characteristics and higher lift values. Drag values are roughly similar among baseline and
tubercled airfoils at lower angles of attack (α < 8°). As the stall angle is approached (8°< α <
15°), the tubercles act to increase drag when compared to the baseline airfoil, with lower drag
associated to smaller amplitudes and wavelength. In post-stall regime, (α > 15°), tubercles
40

achieve a favorable effect in terms of drag, but with larger amplitudes presenting lower drag
coefficients.
The amplitude-to-wavelength ratio (A/λ) is also indicated as a relevant tubercle
parameter (HANSEN, 2012). Configurations with higher amplitude-to-drag ratio, which can be
understood as tubercle sweep angles, increase span flow and enable the generation of
streamwise vortices (DE PAULA, 2016). Ultimately, the peak vorticity near the leading edge
is in direct proportion to the amplitude-to-wavelength ratio (ROSTAMZADEH et al., 2014).

2.2.2 Airfoil Geometry

In the same study discussed on the previous topic, Hansen et al. (2009) provided
dedicated results on airfoil geometry effects. Figure 17 gives the lift-to-drag ratio (L/D) for the
baseline and tubercled configurations of wings based on NACA 0021 and NACA 65-021 foils.
For both airfoils, incorporating tubercles results in higher post-stall lift-to-drag ratios, whereas
in pre-stall, the unmodified airfoil is generally more efficient - the baseline NACA 0021
presents better performance over the tubercled versions for the entire pre-stall range. However,
for the NACA 65-021 in the 8º<α<14º range, a higher L/D ratio is noted for both tubercled
configurations when compared to the unmodified airfoil.

Figure 17 – Airfoil geometry effect on wavy leading edge performance, Re = 120,000, 𝑈∞ =


25 m/s (HANSEN et al., 2009).

Van Nierop et al. (2008) proposes the insertion of tubercles alters the pressure
distribution on wings. Since airfoil geometry is directly related to the pressure distribution, it is
41

expected to have an effect on wavy leading edge performance as well. Stanway (2008) also
suggests that thinner foils operating at higher Reynolds numbers are more likely to experience
increase in maximum lift. However, most research on the field do not venture far from the
morphology of humpback whale flippers, focusing on thick airfoils such as NACA 0021.
Hansen et al. (2009) findings show that airfoil geometry effect on wavy leading edge
performance is certainly an aspect of research that needs to be further investigated. De Paula
(2016) resumed this investigation by performing experimental tests in three sets of airfoil
thickness – NACA 0012, NACA 0020 and NACA 0030, each consisting of one smooth plus
three wavy configurations, at Reynolds number varying between 50,000 and 290,000. Results
show that at higher Reynolds number, thinner airfoils present leading edge stall characteristics,
enabling for lower aerodynamic deterioration in pre-stall regime when compared to the smooth
configuration – confirming the hypothesis made by Stanway (2008).

2.2.3 Flow Dimension

Infinite and finite wings present different flow characteristics due to the three-
dimensional effect of the wing tip, which changes the pressure coefficient and the adverse
pressure gradients along span (DE PAULA, 2016). Thus, distinct performances of tubercled
leading edge can be expected for full and semi-span wings.
Miklosovic et al. (2007) justified the semi-wing superior performance by proposing the
three-dimensional flow induced by the wavy leading edge triggered early separation for the
full-span wing. On the other hand, it enhanced the effectiveness of the half-span wing by
inhibiting spanwise stall progression, allowing for an extended operational envelope with
minimal performance penalties.
However, Miklosovic et al. (2007) full and semi-span experiments were conducted in
different Reynolds numbers. Stanway (2008) states it is important to take into account the
Reynolds number effect on the stall characteristics of the airfoil. In order to isolate Reynolds
number effects and thus identify the impact of flow three-dimensionality, Hansen et al. (2009)
evaluated wavy leading edge geometries for full and half-span NACA 0021 based models, at a
Reynolds number of 120,000. In contrast with results from Miklosovic et al. (2007), it was
observed that the performance of full-span and semi-span wings is similar. Additionally, the
study suggests that the tubercled wings with sweep and/or taper could be benefited in
performance since these geometries carry significantly more spanwise crossflow.
42

Besides comparing infinite to finite wings, Abrantes et al. (2017) also explored
rectangular versus tapered and swept versus unswept wings, for a Reynolds number ranging
between 80,000 and 200,000 at a maximum freestream velocity of 𝑈∞ = 30.6 m/s. At higher
Reynolds number (200,000), the characteristic aerodynamic deterioration of tubercled wings at
pre-stall regime is minimized for all finite wing configurations, with the swept (rectangular and
tapered) tubercled wing presenting the best overall performance when compared to its smooth
counterpart. This distinguished behavior is attributed to the wing tip stall characteristic of swept
wings, for which tubercles act as an efficient flow control mechanism delaying tridimensional
flow separation.

2.2.4 Reynolds Number

As discussed in Section 2.1.1, the Reynolds number can be seen as a measure of the
momentum level on the boundary layer over the airfoil, and it exerts considerable influence on
flow topology (ANDERSON, 2001). Therefore, although wavy leading edge improves
humpback whales performance, it is not certain that such improvement will be achieved in
every flow condition.
The first investigation on the Reynolds number effects on wavy leading edge
performance was carried out by Stanway (2008). Two hydrofoils were designed - a control foil
with smooth leading edge and a test foil with tubercles, both with a NACA 0020 cross section.
The Reynolds number ranged from 44,000 to 120,000.
Figure 18 presents the force measurements for both airfoils, with cases I, II, III and IV
being defined by the corresponding Reynolds number, respectively 44,648; 59,530; 89,295 and
119,060, with corresponding freestream velocity ranging between 0.75 and 2.00 m/s. For the
smooth foil, Stanway (2008) draws attention to the fact that for angles greater than 12° the lift
curve is no longer linear, and flow separation near the tip of the foil causes partial stall – a
mechanism that is Reynolds-number dependent. For the tubercled foil, the same tendency is
observed, with higher Reynolds number providing better lift curves as an indication of less
viscous separation at higher speeds, as expected. Maximum lift is smaller for the test foil in all
cases, except for case IV, which presents the highest Reynolds number. On the other hand, stall
is delayed and smoothened for the test foil in all cases.
43

Figure 18 – Force measurement results for control foil and tubercled test foil (STANWAY,
2008).

In apparent contradiction, Miklosovic et al. (2007) recorded increased maximum lift and
delayed stall on similar three-dimensional foils at 𝑅𝑒 = 50,000. However, said results
exhibited short-bubble type leading edge stall and corresponding bubble-bursting immediate
loss of lift, while Stanway (2008) experiment presents characteristics of trailing edge stall for
both control and test foils. Such observations lead to the conclusion that the effect of tubercles
depends on stall type, and is therefore sensitive to Reynolds number. As comparison implies,
foils operating in a short bubble leading edge stall regime will benefit more from the addition
of tubercles than those operating in a trailing edge stall regime – as suggested by Stanway
(2008) and confirmed by De Paula (2016).
Motivated by the flow characteristics encountered by humpback whales, the vast
majority of researches focus on low Reynolds number applications of tubercles. The effect of
the Reynolds number on turbulence levels typical to commercial aircraft flight has yet to be
assessed, and is one of the main objectives of the present work. Furthermore, as Stöppler (2017)
suggests, transonic investigations on the wavy leading edge concept should be conducted in
Reynolds numbers higher enough to avoid the influence of a laminar boundary layer on shock
behavior.
44

2.2.5 Flow Topology

In order to optimize wavy leading edge performance, it is necessary to understand how


the surrounding flow is affected by the insertion of tubercles. The first findings of Watts and
Fish (2001) by performing non-viscous numerical simulation identified that lower pressure
regions are found between tubercles instead of behind their peaks. However, boundary layer
development and streamwise vorticity were not assessed since the effects of viscosity were
neglected.
A following study addressed this limitation by evaluating the flow over a NACA 63-
021 tubercled airfoil using unsteady Reynolds-averaged Navier-Stokes (RANS) simulation
(PATERSON et al., 2003). Results (Figure 19) show that vorticity is generated on troughs in
chord wise direction, increasing velocity downstream of peaks and thus reducing the adverse
pressure gradient, which leads to a delayed flow separation in peak regions and anticipated flow
separation in trough regions. This study was the first to discuss the idea of momentum exchange
between peaks and troughs due to vorticity generation.

Figure 19 – Pressure contours and streamlines for a NACA 63-021 model with straight (left)
and tubercled (right) leading edge, Re = 1,000,000, α = 10° (PATERSON et al., 2003)

Tuft experiments by Levshin et al. (2006) enabled the first results from flow
visualizations to endorse the simulations from Paterson et al. (2003). Figure 20 shows that in
pre-stall (α = 12°) the flow is attached over the first three quarters of the baseline foil, while for
the tubercled foil the flow is detached over at least half the chord length. In this image, the flow
separation appears to originate from the troughs between tubercles. In post-stall (α = 24°), the
45

flow over the baseline airfoil is completely detached, whereas the tubercled configuration keeps
a locally attached flow over the leading edge.

Figure 20 – Flow visualization of airfoils at 12° (pre-stall) and 24° (post-stall) angle of attack,
Re = 183,000, 𝑈∞ = 1.8 m/s (LEVSHIN et al., 2006)

Different flow visualization techniques have been investigated since then, obtaining
similar results for flow topology and clear indication of streamwise vortices being generated in
the troughs. Custodio (2008) performed a study using dye flow visualization, while Stanway
(2008) applied Particle Image Velocimetry (PIV) to capture the components of velocity at a
plane parallel to the suction side of the airfoil. Zverkov et al. (2008) carried out oil flow
visualizations, and as depicted in Figure 21, the study reveals yet another difference between
the flow topology of the baseline and wavy configurations. For a Reynolds number of 170,000,
near the stall angle (α = 15°) the baseline foil presents a long separation bubble, and after that
the stall is characterized by full separation at the leading edge. On the other hand, the tubercled
foil at α = 20° presents three-dimensional separation bubbles downstream of the troughs while
the flow downstream of the peaks is kept attached, which explains the delay in leading edge
stall. Furthermore, by means of measurements regarding the velocity profiles of the boundary
layer, the results also indicate an early laminar-turbulent transition at troughs.
46

Figure 21 – Oil flow visualization and its interpretation for baseline (α = 15°) and tubercled (α
= 20°) configurations, Re = 170,000, 𝑈∞ = 12 m/s (ZVERKOV et al., 2008).

Investigating the streamwise vorticity identified by experimental visualization methods,


Rostamzadeh et al. (2014) brought back the aid of numerical simulations to the research. As
shown in Figure 22, for a NACA 0021 airfoil at 𝑅𝑒 = 120,000, the simulations identified the
patterns of three-dimensional laminar separation bubbles for different tubercled configurations
in pre-stall regime. In addition, attached flow regions downstream of the peaks were found to
justify higher lift values in post-stall regime.

Figure 22 – Separated flow regions (green areas) over the suction surface of the smooth and
wavy airfoils in pre-stall regime, Re = 120,000, 𝑈∞ = 25 m/s (ROSTAMZADEH et al., 2014).
47

2.2.6 Flow Control Mechanism

There are multiple explanations for the changes in flow behavior and consequent
aerodynamic performance due to the insertion of tubercles. Hansen (2012) provides a summary
of the most relevant in literature:
 Increased boundary layer momentum exchange by vortex generation (CUSTODIO,
2008; FISH; BATTLE, 1995; JOHARI et al., 2007; MIKLOSOVIC et al., 2004;
PEDRO; KOBAYASHI, 2008)
 Compartmentalization of the flow along the span minimizing tip stall (FISH; BATTLE,
1995; MIKLOSOVIC et al., 2007; PEDRO; KOBAYASHI, 2008)
 Non-uniform separation characteristics (FISHER; LAUDER, 2006; JOHARI et al.,
2007; PEDRO; KOBAYASHI, 2008; VAN NIEROP et al., 2008; ZVERKOV et al.,
2008)
 Alteration of the pressure distribution over the airfoil (VAN NIEROP et al., 2008)
 Vortex lift (CUSTODIO, 2008; MIKLOSOVIC et al., 2007)

As pointed out by Hansen (2012), such explanations are mutually inclusive and indicate
that several benefits can occur simultaneously. However, there is limited explanation
concerning the failure to provide performance enhancement in some scenarios.
The first suggestion for the existing flow control mechanism was an analogy between
conventional wing vortex generators and tubercles, meaning the vortices enabled increase in
momentum exchange thus keeping the flow attached (FISH; BATTLE, 1995). This analogy
was contested by Miklosovic et al. (2007), by suggesting that the streamwise vortices
established a physical barrier more analogous to a wing fence to prevent spanwise stall
progression. Van Nierop et al. (2008) also claimed that since the wavelength and amplitude of
tubercles are much larger than boundary layer thickness, it was not possible for tubercles to act
as vortex generators. Instead, it was proposed that the trigger was more related to the geometric
difference between profiles at peaks and troughs. Hence, since the sections have different chord
lengths but similar thicknesses, the pressure gradient is higher for a trough, which explains
separation beginning in this region. Furthermore, Van Nierop et al. (2008) proposed that the
separation behind peaks is further delayed by a non-uniform downwash component, which
implies lower effective angles of attack at peaks.
48

For post-stall increased performance, Custodio (2008) suggests that between tubercle
peaks the flow generates a pair of counter-rotating streamwise vortices, which increase toward
troughs and merge in this region, creating suction areas that help in lift generation. In pre-stall
regime, however, the suction generated vortices are not strong enough to compensate flow
separation at troughs.
According to Rostamzadeh et al. (2014), streamwise vorticity can be generated by
“skew-induced” and “stress-induced” mechanisms. The first occurs when existing vorticity
turns in the streamwise or transverse direction, while the latter arises from the anisotropy of
turbulence. As shown in Figure 23, for the flow over a tubercled foil, flow skewness is
distinguished by the change in the streamlines’ curvature as they approach the leading-edge.
Here, two types of streamwise vortices can be identified. Three-dimensional bubbles forming
at tubercle troughs drive the formation of primary vortices, adding streamwise vorticity to the
flow downstream of the bubbles, while secondary vortices are formed at the region near the
trailing edge, establishing a delta-shaped flow separation.

Figure 23 – Normalized streamwise vorticity, Re = 120,000, 𝑈∞ = 25 m/s, α = 2°


(ROSTAMZADEH et al., 2014).

Regarding the stress-induced mechanism, Skillen et al. (2015) shows that the flow
behind trough areas experience higher adverse pressure gradients than areas behind peaks,
which causes flow separation with consequent reattachment and the appearance of a laminar
separation bubble, as depicted in Figure 24. This higher pressure gradient is attributed to a low
pressure area and a reduced chord length, as suggested by Fish and Lauder (2006). As the flow
accelerates between tubercles a suction area is generated, leading to a secondary flow.
49

Figure 24 – Time-averaged streamlines showing deflection of oncoming flow (left) and


secondary flow (right), Re = 120,000, α = 20° (SKILLEN et al., 2015).

2.2.7 Transonic Regime Effect

As discussed in the previous subsections, the effect of implementing wavy leading edge
to wings in subsonic regime has been explored by a multitude of studies, however other regimes
still lack investigation. The very first experimental study of the wavy leading edge phenomena
in transonic regime was performed by Stöppler (2017), by using NACA 0020 based models and
capturing pressure distributions via pressure sensitive paint (PSP) in wind tunnel testing, for a
Reynolds number of 1,000,000. Even though technical issues prevented absolute results to be
considered valid, this study instigated the use of PSP in wavy leading edge research.
Padilha (2017) performed a similar study with NACA 0012 based models instead, for
the same Reynolds number. Figure 25 summarizes three points of interest in his results. Point
A shows the pressure coefficient peak values are similar for both foils, although the peak
location is further downstream for the wavy configuration. The shock position (point B)
presents little variation among the three lines, but the pressure coefficient variation due to the
shock is notably smaller for the wavy configuration. This promising behavior is kept throughout
the profile, as shown by point C.
50

Figure 25 – Pressure coefficient comparison between baseline (yellow) and wavy (blue for
peaks and green for troughs) NACA 0012 configurations, Re = 1,000,000, 𝑀∞ = 0.8, α = 0°
(PADILHA, 2017).

Additional results on the same NACA 0012 based models from Padilha (2017) are
provided by Sepetauskas et al. (2018). As shown in Figure 26, for 𝑀∞ = 0.7 and α = 2°, a
pressure drop typical to a lambda-shock formation occurs for the smooth wing; on the other
hand, the tubercled wing presents a different flow behavior, indicating a suppressed shock. As
discussed in Section 2.1.6, the lambda-shock develops in laminar flow and leads to flow
separation. Sepetauskas et al. (2018) suggest the shock-induced laminar separation might have
been suppressed by the counter-rotating vortices generated by the tubercles.

Figure 26 – Tubercle effect on pressure distribution, with an indication of lambda-shock


suppression for the wavy configuration, Re = 1,000,000, 𝑀∞ = 0.7, α = 2° (SEPETAUSKAS
et al., 2018).
51

Schlieren flow visualization experiments conducted by Asghar et al. (2017) also


employed the NACA 0012 as a reference airfoil in transitional Reynolds numbers, ranging from
700,000 to 820,000. Compared to the smooth counterpart, the shock was weakened and
displaced aft-wards for the tubercled wing, as shown in Figure 27.

Figure 27 – Schlieren visualization of baseline (left) and tubercled (right) wings, Re =


820,000, 𝑀∞ = 0.81, α = 0° (ASGHAR et al., 2017)

Perez et al. (2018) continued these experiments by performing a numerical investigation


using computational fluid dynamics. Computed results were validated against the experimental
data from Ashgar et al. (2017) with satisfactory agreement. As shown in Figure 28, spanwise
pressure distribution is periodic up to 35% of the chord length, becoming essentially two-
dimensional past this point. Furthermore, as previously observed by Padilha (2017), the shock
position was shifted further downstream for the tubercled foil.

Figure 28 – Pressure coefficient contours for (a) straight and (b) tubercled leading edge wings,
Re = 820,000, 𝑀∞ = 0.8, α = 0° (PEREZ et al., 2018)
52

Regarding the shock-wave/boundary-layer interaction, the straight leading edge wing


presents shock-induced boundary layer separation, as shown in Figure 29. For the tubercled
wing, however, this interaction seems to be strongly dependent on the spanwise position
downstream of the peak, valley and mean line of the foil, the latter being equivalent to the
baseline wing in chord length and airfoil shape. The tubercled wing seems to withstand flow
separation for the peak and mean line locations, as the shock appears to be less intense. In the
valley region, on the other hand, the effect of the shock on the boundary layer is even stronger
than for the straight wing, which can be inferred from the wider wake and higher turbulent
kinetic energy levels. This behavior agrees with previously discussed studies indicating both
momentum exchange between peaks and troughs (PATERSON, 2003; CUSTODIO, 2008;
FISH; BATTLE, 1995; JOHARI et al., 2007; MIKLOSOVIC et al., 2004; PEDRO;
KOBAYASHI, 2008) and higher adverse pressure gradients being encountered at troughs
(VAN NIEROP et at., 2008). Such differences in flow topology corresponded to a drag
reduction in the order of 10% and an increase in drag-divergence Mach number by 1.5% for the
tubercled foil.

Figure 29 – Turbulent Kinetic Energy (TKE) computational results for straight and tubercled
wings, Re = 820,000, 𝑀∞ = 0.8, α = 0° (PEREZ, 2018).
53

2.2.8 Aims and objectives of this thesis

The literature review in this work has shown that the vast majority of the research on
tubercles has focused on flow conditions similar to those of the humpback whale, that is, low
Reynolds number up to 500,000 in subsonic flow. For various airfoil and tubercled geometries
in laminar-turbulent transitional or completely laminar Reynolds numbers, tubercles have been
found to be consistently beneficial in the post-stall regime, also delaying flow separation and
enabling softer or delayed stall.
The performance of wavy leading edge in transonic regime had mostly been theorized
thus far. In the past few years, however, research on transonic regime has been getting attention,
along with the aforementioned promising results. However, to date, transonic investigation has
only been performed for the symmetric NACA 0012 profile, which is not designed for transonic
speeds. With the intent to expand the flow conditions to those typical to commercial aircraft,
this work contributes to research by adopting a supercritical airfoil, providing insight on the
applicability of wavy leading edge as a passive flow control device in the aircraft industry.
As presented in the literature review, the wavy leading edge is affected by multiple
parameters, including tubercle and airfoil geometry, flow dimension and flow conditions.
Taking that into consideration, it is preferred to isolate a few parameters at a time to understand
the flow control mechanism at hand. For that reason, this study will investigate a single tubercle
geometry incorporated to an infinite straight wing, leaving only flow conditions to be explored.
As previously discussed, both Reynolds number and Mach number have major influence over
flow characteristics in transonic regime. Therefore, the aim of this thesis is to evaluate the effect
of Reynolds and Mach numbers on wavy leading edge transonic performance.
In order to achieve this objective, a numerical investigation using computational fluid
dynamics (CFD) is performed, comparing an infinite RAE 2822 wing with wavy leading edge
to its reference counterpart with straight leading edge. To investigate the Reynolds number
effect, simulations are performed at Reynolds numbers of 1x106, 6.5x106 and 2.2x107; the latter
corresponding to a typical Reynolds number encountered by transonic aircraft in commercial
flight. For the Mach number effect, simulations comprehend a Mach range from 0.5 to 0.86.
Wavy leading edge application is analyzed in terms of drag rise, drag coefficient component
breakdown, computational flow visualization and pressure distribution.
54

3 Methodology

This chapter presents the CFD analysis process adopted in the present work, which
comprises the mathematical modeling, definition of the geometry, the computational domain,
the simulation setup and the applicable verification and validation (V&V) assessment. Since
the aim of this project is to evaluate the wavy leading edge performance rather than the
principles of CFD itself, the theory behind this numerical methodology will not be stressed in
detail.

3.1 Physical modeling

As remarked by the Nobel laureate physicist Richard Feynman, turbulence is “the most
important unsolved problem of classical physics” (FEYNMAN, 1977). To solve turbulent
flows, most CFD applications use the Reynolds-Averaged Navier-Stokes (RANS) equations.
In this form, the instantaneous Navier-Stokes equations are decomposed into a mean (time-
averaged) component (𝑢̅) and a fluctuating component (𝑢′), for which the mean is equal to zero
(𝑢̅′ = 0) (WILCOX, 2006). For a compressible, viscous flow assumed as stationary, the
continuity, momentum and energy RANS equations are given respectively as follows:

𝜕 𝜌̅ 𝜕
+ (𝜌̅ 𝑢𝑗 ) = 0 (3.1)
𝜕𝑡 𝜕𝑡
𝜕 𝜕 𝜕𝑝̅ 𝜕
(𝜌̅ 𝑢𝑖 ) + (𝜌̅ 𝑢𝑖 𝑢𝑗 ) + − (𝜏̅̅̅ − ̅̅̅̅̅̅̅̅̅
𝜌𝑢𝑖′′ 𝑢𝑗′′ ) = 0
𝜕𝑡 𝜕𝑥𝑗 𝜕𝑥𝑖 𝜕𝑥𝑗 ⏟ 𝑖𝑗 (3.2)
2

𝜕𝑒̅ 𝜕 𝜕 𝜕 ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
𝜌𝑢𝑖′′ 𝑢𝑗′′ 𝜕𝑞̅𝑗
+ [(𝑒̅ + 𝑝̅)𝑢𝑗 ] + ̅̅̅̅̅̅̅
(𝜌𝑢 ′′
𝑗 ℎ) − [ 𝑢𝑖 (𝜏̅̅̅ ̅̅̅̅̅̅̅̅̅
𝑖𝑗 − 𝜌𝑢

′′ ′′ ′′
𝑖 𝑢𝑗 ) + 𝑢𝑖 (𝜏𝑖𝑗 − )] + =0 (3.3)
𝜕𝑡 𝜕𝑥𝑗 𝜕𝑥𝑗 𝜕𝑥𝑗 ⏟ 2 𝜕𝑥𝑗
1
3

As highlighted by terms 1, 2 and 3 in the equations above, the Reynolds decomposition


introduces three unknowns to the Navier-Stokes equations, namely the Reynolds stress tensor,
the Reynolds heat flux and the Reynolds dissipation respectively. With these additional
unknowns, for a closed system with solvable equations to be possible, a turbulence model must
be employed.
55

Although a few turbulence models are available and applicable for the flow
characteristics of this thesis, the one-equation Spalart-Allmaras model (SPALART;
ALLMARAS, 1992) was adopted. However, considering the high-Reynolds-number transonic
flow investigated, the effects of turbulence may be too severe for the conventional Spalart-
Allmaras model, and according to Spalart et al. (2015), modifications of the closure model can
provide a better physical representation. Thus, a Spalart-Allmaras model with two variants -
rotation/curvature correction and quadratic constitutive relation is applied, which will be
referred to as SA-RC-QCR. This form of the Spalart-Allmaras model attempts to account for
rotation and curvature effects (SHUR et al., 2000), while also introducing anisotropy to the
Reynolds stress to improve the physical representation of the boundary layer under complex
flow phenomena (SPALART et al., 2015).

3.2 Airfoil and Tubercle Geometry

As mentioned in section 2.2.8, a supercritical airfoil is chosen as the baseline geometry


for the wing, hereinafter referred to as SLE (straight leading edge). Figure 30 presents the RAE
2822 supercritical airfoil, a choice based on its well-documented performance in the transonic
regime (COOK et al., 1979).

Figure 30 – RAE 2822 airfoil normalized in XZ-plane

For the tubercled geometry, hence WLE (wavy leading edge), the adopted configuration
is the A3λ11, which presented the best overall performance for the entire low-Reynolds number
range investigated by De Paula (2016). This configuration corresponds to a sinusoidal wave
placed on the leading edge according to Figure 31, defined by an amplitude (A) of 3% and a
wavelength (λ) of 11% of the chord length (𝑐̅). This positioning aims to keep the average chord
length, average profile thickness and surface area of the tubercled wing equal to those of the
baseline wing.
56

Figure 31 – WLE geometry definition and positioning relative to the baseline leading edge.

3.3 Computational Domain

3.3.1 Baseline and Tubercled Wings Design

With both airfoil and tubercle geometries defined, the baseline and tubercled wings are
designed in CATIA V5, a Computer Aided Design (CAD) software largely used for CFD
applications. The RAE 2822 airfoil coordinates (COOK et al., 1979) are imported to CATIA
V5 and connected with a spline. For the SLE wing, this profile is simply extruded to the desired
span length. For the WLE wing, the curve that defines the wavy leading edge is also imported,
and a lofted surface is generated using it as a guide curve. Figure 32 illustrates the resulting
baseline and tubercled geometries. Although a finite span is represented, lateral walls are placed
on each side of the wings as to ensure there is no endplate three-dimensional effect on the flow,
keeping the two-dimensional flow condition of infinite wings.

Figure 32 – SLE (left) and WLE (right) RAE 2822 wings (only four wavelengths are shown).

The baseline chord length is 1m. For the WLE wing, this results in maximum and
minimum chord lengths of 1.015m and 0.985m respectively, with a spanwise wavelength of
0.11m on the leading edge tubercles. For both wings, the sharp trailing edge is cut off to a blunt
57

thickness of 3mm, to avoid excessive mesh skewness in this region. Along with the tubercle
geometry described in section 3.2, the geometric characteristics herein detailed were defined as
to ensure both wings had the same reference area (𝑆), which simplifies the comparison in terms
of aerodynamic coefficients.

3.3.2 Grid Generation Strategy

The subsequent step in preprocessing the CFD simulation is the discretization of the
flow domain, which is known as grid generation or meshing. The selected meshing software
was ANSYS ICEM CFD 19.1, for its large application in CFD analysis at Embraer, where the
entire study was developed.
Appropriate meshing is essential for an accurate and fast converging solution in CFD.
In 3D analyses, there are two major types of grids: structured hexahedral and unstructured
tetrahedral. While the latter has the advantage of being highly automated and more efficient in
handling complex geometries, it does so by sacrificing control and solution accuracy (JARVIS,
2017). In opposition, despite being a much more time-consuming task, structured grid
generation provides the user with complete control over grid parameters (JARVIS, 2017). This
is of particular interest when different geometric solutions are being compared.
With the main objective of ensuring similar discretization between the baseline and
tubercled wings, the structured hexahedral grid was the adopted meshing method of this thesis.
The mesh generating process was performed with the aid of GMA – Geometry and Mesh
Automation – a system developed by Embraer and ESSS for automatic and parametric
geometry and grid generation (OLIVEIRA et al., 2008). With respect to hexahedral meshing,
the system consists of a replay file (.tcl) written in ICEM CFD that generates supporting entities
(points and curves), performs block cutting, calculates bunching and outputs the grid; with
embedded functions that provide robust high-quality discretization for CFD applications
(SOUZA et al., 2016). Further details on the method can be found in Ferrari (2018).
Considering the geometry of a wing, the GMA requires two sizing parameters: the
smallest leading edge element and the boundary layer first cell height. The rest of the grid is
defined by prescribing growth ratios; hence, grid refinement can be obtained by simply
changing these grid parameters (SOUZA et al., 2016). Figures 33 and 34 illustrate the grid
resulting from the GMA method.
58

Figure 33 – Far field definition (left) and detailed views of the wing (right) for the WLE
geometry (only four wavelengths shown).

Figure 34 – Leading edge detailed view for the SLE (left) and WLE (right) grids.

As shown in Figure 33 – Far field definition (left) and detailed views of the wing (right)
for the WLE geometry (only four wavelengths shown).Figure 33, the far field boundary is
placed 50𝑐̅ upstream, downstream, above and below the wing. The spanwise depth of ten
wavelengths (1.1m) is adopted. Originally, the intent was to use a four-wavelength span
(0.44m), which has been observed to be a good practice for wavy leading edge modeling
considering periodic boundary conditions at the lateral walls (ROSTAMZADEH et al., 2014;
SKILLEN et al., 2015). However, as it will be discussed in section 4.1, solution convergence
issues required a longer span to be employed.
59

The mesh shown in Figures 33 and 34 is a generic representation of the boundary layer
definition and refinement. The actual grid parameters were adjusted according to the boundary
layer thickness (𝛿 ) and required first cell height (Δ𝑦).
The boundary layer thickness is estimated from the one-seventh-power law combined
with empirical data for turbulent flow over a flat plate (ÇENGEL; CIMBALA, 2004), as shown
in Equation 3.4:

𝛿 0.38
≅ 1 (3.4)
𝑥
(𝑅𝑒𝑥 )5

For the boundary layer first cell height (Δ𝑦), Equations 3.5 through 3.8 were employed:

0.059
𝐶𝑓 ≅ 1 (3.5)
(𝑅𝑒𝑥 )5
𝜌𝑈 2 𝐶𝑓
𝜏𝑤 = (3.6)
2
𝜏𝑤
𝑢𝑓 = √ (3.7)
𝜌

+
𝜌𝑢𝑓 Δ𝑦 𝜇y +
𝑦 = → Δ𝑦 = (3.8)
𝜇 𝜌𝑢𝑓

The skin friction coefficient (𝐶𝑓 ) from Equation 3.5 is estimated from the one-seventh-
power law combined with empirical data for turbulent flow over a flat plate, similarly to the
boundary layer thickness (ÇENGEL; CIMBALA, 2004). Equations 3.6 and 3.7 are solved
within Equation 3.8, which gives the definition of the dimensionless viscous wall distance, or
simply y+. Since a thoroughly explanation of this parameter is not in the scope of this thesis, it
can be understood as a measure of how coarse or fine a grid must be near the surface for a
particular flow pattern and modeling strategy in order to correctly solve the boundary layer.
Considering the RANS approach with the SA-RC-QCR turbulence model and turbulence wall
integration with no wall functions enabled, that means, resolving the viscous sublayer, the
boundary layer first cell height (Δ𝑦) must satisfy 𝑦 + ≈ 1 (SPALART et al., 2015), which will
be demonstrated in section 4.1.
60

As presented in section 2.2.8, three Reynolds numbers are investigated in this thesis:
1x106, 6.5x106 and 2.2x107. The smallest Reynolds number corresponds to the maximum
achievable in a transonic wind tunnel used for NACA 0012 investigations (SEPETAUSKAS,
2017; PADILHA, 2017; FILHO et al., 2018). The intermediate value is that of the validation
case, which is presented in section 4.1, and the highest is a typical Reynolds number
encountered by transonic aircraft in commercial flight. From Equations 3.4 through 3.8, it can
be noticed that both the boundary layer thickness (𝛿 ) and first cell height (Δ𝑦) depend on the
Reynolds number – the higher the Reynolds number, the thinner the boundary layer is and the
smaller the first cell must be.
With the intent of performing a grid independence study, four different meshes for the
SLE geometry under the validation case Reynolds number (6.5x106) were generated, varying
in refinement levels according to Table 1 and illustrated in Figure 35.

Table 1 – Grid Refinement Parameters.

Leading edge Boundary Boundary


Total Trailing
smallest layer layer first Spanwise
Grid element edge
element size thickness cell height nodes
count nodes
(µm) (mm) (µm)
Coarse 8.6 x 106 300 25 4.0 128 15
Medium 11.9 x 106 250 25 3.9 160 18
Fine 16.1 x 106 208 25 3.8 200 22
Finest 23.1 x 106 174 25 3.7 250 26
61

Figure 35 – Coarse, Medium, Fine and Finest grids for SLE geometry.

The same procedure should have been followed for the other Reynolds numbers,
however solution convergence issues limited the scope of simulations, as it will be discussed in
section 4.1. For this reason, only one mesh for each of the three Reynolds number was used,
namely Re1, Re6.5 and Re22. The parameters that define each mesh are shown in Table 2. To
properly capture shock wave formation, the boundary layer thickness (𝛿 ) in all three grids was
defined to be two times larger than the largest value estimated by Equation 3.4, which
corresponds to Re1 mesh with 𝛿 ≈ 24𝑚𝑚. Leading edge smallest element size, the number of
spanwise and trailing edge nodes were also kept the same among grids.

Table 2 – Final grid parameters

Leading edge Boundary Boundary


Total Trailing
smallest layer layer first Spanwise
Grid element edge
element size thickness cell height nodes
count nodes
(µm) (mm) (µm)
Re1 7.91 x 106 300 50 23 220 15
Re6.5 8.74 x 106 300 50 4.0 220 15
Re22 9.27 x 106 300 50 1.3 220 15
62

3.4 Simulation Setup

All numerical simulations were performed using the commercial software Metacomp
Technologies’ CFD++, available at Embraer in its 18.1 release. After importing the mesh
generated in ICEM CFD, boundary conditions must be applied to each surface of the domain.
Wing surfaces were regarded as adiabatic viscous (no-slip) walls with Solve to Wall checked
for wall integration, which brings the requirement of 𝑦 + ≈ 1 mentioned previously. Since the
effect of tubercles is being investigated for an infinite wing, the left and right sides of the far
field were initially established as periodic boundary conditions. This boundary condition is
used when the geometry and the expected pattern of the flow have periodically repeating nature,
allowing for a considerable reduction in total grid elements count and thus reducing
computational costs (ANSYS, 2016). However, solution verification issues required symmetric
boundary conditions to be employed, which will be discussed in section 4.1. As for the
remaining four surfaces of the far field, an Inflow/Outflow-Characteristics-based boundary
condition is applied.
In Information Set Entry menu, the Compressible PG Navier-Stokes/Euler solver is
selected, and as discussed in section 3.1, turbulence is solved by RANS equations closed with
the SA-RC-QCR turbulence model. Remaining options in this menu were left as default.
Reference quantities were set as pressure-based dimensional in SI units. Simulations
were performed with the second order Harten-Lax-van Leer-Contact (HLLC) Riemann solver
with min-mod flux limiting, and multigrid acceleration was used to improve solution time and
accuracy. In the Help Set Numerics menu, the transonic regime was chosen under the Perfect
Gas tab. Turbulence control settings were left as default.
The above-described settings were applied to all test cases, with only flow conditions
left to be defined according to each case. For performance comparison cases, simulations are
set according to the matrix presented in Table 3. This matrix enables drag rise analyses to be
carried out for the SLE and WLE infinite wings. In this type of analysis, the Mach number is
swept through a subsonic-to-transonic range for a given lift coefficient (𝐶𝐿 ), the resulting drag
coefficient (𝐶𝐷 ) is computed against each Mach number and a curve similar to Figure 12 is
developed. This procedure is executed for each Reynolds number and four lift coefficients to
explore different flight phases. Considering both SLE and WLE geometries, the total number
of cases run to draw drag rise curves was 234.
63

Table 3 – Performance comparison cases matrix

Re M CL
0.5
0.6
0.65
0.68
0.70
1 x 106 0.72 0.1
6.5 x 106 0.74 0.3
2.2 x 107 0.76 0.5
0.78
0.80
0.82
0.84
0.86

To reach each of the desired lift coefficient (𝐶𝐿 ), the CL Driver Wizard from CFD++
was employed. The total number of iterations at each drag rise run was set as 20000, with the
first and second CL Driver checkpoints set at 6000 and 6500 iterations respectively.

3.5 Verification and Validation

According to the AIAA Guide for Verification and Validation (V&V) (AIAA, 1998),
verification is “the process of determining that a model implementation accurately represents
the developer’s conceptual description of the model”, while validation consists in “determining
the degree to which a model is an accurate representation of the real world” solution. In that
sense, it is necessary to distinguish errors from uncertainties. An error can be defined as a
recognizable deficiency in any phase or activity of modeling and simulation that is not due to
lack of knowledge, whereas uncertainty is as a potential deficiency that is due to lack of
knowledge (AIAA, 1998). In other words, verification accounts for the errors in the
mathematical representation of the conceptual model, while validation accounts for the errors
and uncertainties in the physical representation of the mathematical model (SLATER, 2008).
There are four major sources of error in CFD simulations: computer programming,
temporal discretization, iterative convergence and spatial discretization (AIAA, 1998).
Computer programming is covered by the use of CFD++, an industry-standard software that
has been verified and validated for a wide range of applications. Temporal discretization is not
64

applicable either, since the simulations performed are steady state only. Iterative convergence
is the main source of error in this study, as it will be discussed in section 4.1. Spatial
discretization is assessed by grid convergence studies, which were performed in the early
development of the thesis, but were disregarded in the final stages due to solution convergence
issues, as it will also be discussed in section 4.1.
As for uncertainties, the major source in CFD analysis lies in the prediction of
turbulence (SLATER, 2008). The assessment of uncertainties is accomplished by comparing
CFD results to experimental data, aiming to validate the simulation for specific conditions and
thus provide credibility for predicting flow behavior in similar scenarios (SLATER, 2008).
With this purpose, the SLE wing was simulated using Re6.5 mesh under the flow conditions
given in Table 4, which were assessed for the RAE 2822 airfoil in wind tunnel measurements
by Cook et al. (1979). At such flow conditions, a shock is expected to take place at the suction
side of the wing.

Table 4 – SLE wing validation case (adapted from COOK et al., 1979).

Re M α
6.5 x 106 0.725 2.92°

To date, however, experimental data is only available for the baseline RAE 2822 wing,
which makes the validation of the tubercled wing results unfeasible. Furthermore, ideally, the
validation case should be examined under different turbulence models (SLATER, 2008). In this
thesis, however, only the RANS equations closed with SA-RC-QCR turbulence model is
adopted. Since the aim of this work is to investigate wavy leading edge performance by
comparison between a tubercled wing and a baseline wing under the exact same conditions, it
was deemed appropriate to disregard the assessment of the turbulence model, focusing on the
differences between the predicted results for each wing rather than on absolute values.
65

4 Results and Discussion

This chapter is divided in two subsections. The first one inspects the solution regarding
its accuracy and reliability, while the second one explores the effect of tubercles in transonic
regime in terms of drag rise, drag component breakdown and flow topology analyses.
Results were obtained with the aid of scripts that either extracted data directly from
CFD++ into predefined plots or imported CFD++ data into Tecplot 360 EX 2016, a data
visualization and analysis tool widely used for post-processing CFD simulations.

4.1 Solution Verification and Validation

4.1.1 Dimensionless Viscous Wall Distance

For all test cases, the dimensionless viscous wall distance (y+) ranged from 0.1 to 0.8 along
wing surfaces, as exemplified in Figure 36. Thus, the requirement of 𝑦 + ≈ 1 for adopting the
SA-RC-QCR turbulence model solving the viscous sublayer has been met.

Figure 36 – Dimensionless viscous wall distance (y+) contour for SLE and WLE wings (top
view) at 𝑅𝑒 = 6.5𝑥106 , 𝑀∞ = 0.72 and 𝐶𝐿 = 0.5.
66

4.1.2 Grid Convergence

At first, as mentioned in section 3.3, simulation grids were described with four
wavelengths of span (0.44m) and periodic boundary layers to properly employ the concept of
infinite wings. At this stage, a grid independence study was conducted for the SLE geometry
subjected to the validation case flow conditions from Table 4, from which partial results are
presented and discussed in this section.
Table 5 shows lift and drag coefficients results for each mesh, considering the mean
value of the last 500 iterations. The same flow conditions and simulation parameters were
adopted for all four meshes, except for the number of iterations. For Coarse, Medium and Fine
grids, simulations were run up to 3000 iterations. For Finest grid, 7000 iterations were
determined.

Table 5 – Grid independence study results.

Grid CL CD
Coarse 0.752 0.0131
Medium 0.747 0.0130
Fine 0.744 0.0131
Finest 0.744 0.0131
Cook et al. (1979) 0.743 0.0127

Table 5 also provides experimental results obtained by Cook et al. (1979). Comparing
final values alone, it would seem that grid convergence was reached and both Fine and Finest
meshes are refined enough to provide accurate and reliable results. Figures 37 to 39 reveal,
however, that iteration convergence was not reached for any of the grids.
67

Figure 37 – Coarse, Medium and Fine grids 𝐶𝐷 convergence (left) with zoomed in view
(right) at validation case flow conditions.

Figure 38 – Coarse, Medium and Fine grids 𝐶𝐿 convergence (left) with zoomed in view (right)
at validation case flow conditions.

Figure 39 – Finest grid 𝐶𝐷 and 𝐶𝐿 convergence at validation case flow conditions.

Figure 37 shows that 𝐶𝐷 has converged to a variation of less than two drag counts (Δ𝐶𝐷 )
in the last 1000 iterations. Yet 𝐶𝐿 is still increasing with no signs of converging, as seen in
68

Figure 38. For the Finest grid, as illustrated in Figure 39, even at 7000 iterations 𝐶𝐿 deviates by
a large amount. Comparison between SLE and WLE cannot be made if the solution is iteration-
dependent.
Several attempts to obtain solution convergence were made. Varying mesh parameters
such as increasing boundary layer thickness and decreasing first cell height had no effect, nor
did increasing the number of iterations. Since a shock is expected to occur at the flow conditions
from Cook et al. (1979), the possibility of an actual transient effect (ANSYS, 2016) was
considered. Thus, a simulation was run at subsonic conditions – 𝑀∞ = 0.2, 𝑅𝑒 = 1𝑥106 , α =
0°; still the same fluctuation in aerodynamic coefficients was observed. Finally, by changing
boundary conditions at the lateral walls from periodic to symmetric, the fluctuation was damped
to acceptable levels, as shown in Figure 40.

Figure 40 – 𝐶𝐿 and 𝐶𝐷 convergence for Coarse grid with periodic and symmetric boundary
conditions at validation case flow conditions.

Both solutions were compared in terms of different variables to investigate such


behavior, and chordwise wall shear stress (𝜏𝑥 ) contours revealed that the periodic boundary
condition failed to represent an infinite wing. This is evidenced by the non-parallel lines
highlighted in Figure 41, which do not occur for the symmetric boundary conditions and are
not expected for infinite wings with no spanwise crossflow.
69

Figure 41 – 𝜏𝑥 contours for Coarse grid with periodic and symmetric boundary conditions at
validation case flow conditions.

Even though previous studies indicated that a span of four wavelengths were enough for
applying periodic boundary conditions to tubercled wings (ROSTAMZADEH et al., 2014;
SKILLEN et al., 2015), there was no consideration regarding the chord-to-span ratio of the
wing. Lockard et al. (2015) investigated the influence of spanwise boundary conditions on slat
noise CFD analysis, simulating a short wing with a span of 𝑐̅/9 and another one with 2.9𝑐̅. The
study suggested that shorter span-to-chord ratios could interfere on the spanwise coherence of
acoustic waves, leading to slightly different results between both wings. Unlike what has been
observed in the present work, however, there was no indication of large fluctuations of
coefficients through iterations nor of chordwise wall shear stress in Lockard et al. (2015).
In an attempt to investigate this effect, a new grid (which became Re6.5) was designed
with a spanwise extent of 1.1𝑐̅ and simulated with both periodic and symmetric conditions at
the validation case flow conditions, but resulted in the same behavior observed in Figures 40
and 41. A longer span was not considered due to time constraints – the longer the span, the
larger the total number of elements in the grid, consequently requiring longer simulation time.
Considering the intent of running 312 simulations to evaluate drag rise, it was decided to adopt
symmetric boundary conditions with 1.1𝑐̅ span length and compromise grid convergence
analysis, resulting in the three final grids presented in Table 2.
70

4.1.3 Comparison to Experimental Data

Comparison of Re6.5 grid to experimental data in terms of 𝐶𝐿 and 𝐶𝐷 is given in Table


6. Computed values for Re6.5 grid consider the mean value of the last 1,000 iterations in a total
of 20,000 iterations. Results from a 2D CFD analysis (SLATER, 2002) are also displayed. This
analysis was performed using NASA’s WIND-US 4.0 platform with the Spalart-Allmaras
turbulent model, and flow conditions were adjusted to 𝑀∞ = 0.729 and α = 2.31° for a better
match with experimental results (SLATER, 2002).

Table 6 – 𝐶𝐿 and 𝐶𝐷 comparison to validation case.

Grid CL CD
Re6.5 0.747 0.0133
Cook et al. (1979) 0.743 0.0127
Slater (2002) 0.731 0.0121

Although results for Fine and Finest grids were closer to those from Cook et al. (1979),
the lack of solution convergence must be taken into account. For Re6.5 grid, Figures 42 and 43
show that the solution has converged.

Figure 42 – Re6.5 grid 𝐶𝐷 convergence (left) with zoomed in view (right) at validation case
flow conditions.
71

Figure 43 – Re6.5 grid 𝐶𝐿 convergence (left) with zoomed in view (right) at validation case
flow conditions.

The validation case was also analyzed in terms of the pressure coefficient (𝐶𝑃 ) around
the airfoil, which is illustrated in Figure 44. It can be noticed that the shock is anticipated in
chordwise direction for both CFD analyses in comparison to experimental data, additionally
taking into account that the CFD analysis by Slater (2002) adjusted both Mach number and
angle of attack for a better patch with the experimental results by Cook et al. (1979). The
pressure distribution on the suction side of the SLE wing overshoots and deviates from the other
two up to the shock; nevertheless, the shock intensity seems to agree well between Re6.5 and
Cook et al. (1979), despite the shift in position. Given such considerations and adding the
overall agreement of results from Table 6, the solution is considered validated for the SLE wing
using the Re6.5 grid. Since this is the only consolidated experimental data available, this
validation is assumed extendable to the entire range of transonic flow conditions investigated
in this thesis.

Figure 44 – Comparison of 𝐶𝑃 distribution along the chord at validation case flow conditions.
72

4.1.4 CL Driver Convergence

As presented in section 3.4, in order to reach the desired lift coefficient for drag rise analysis,
the CL Driver Wizard was employed. For still unknown reasons, the fluctuations observed in
the previous subsections is worsened for this setup, with large oscillations even after 20,000
iterations. Considering simulation time constraints, increasing the number of iterations beyond
this already excessive mark would compromise the development of the thesis.
To account for the oscillations, the mean value of the last 5,000 iterations was
considered to compose the drag coefficient at each run. To verify whether this solution was
reliable, 2D simulations were run with the planform of the Re6.5 grid and compared to their 3D
counterparts. Figure 45 illustrates this procedure for a single case, at 𝑅𝑒 = 6.5𝑥106 , 𝑀∞ =
0.72 and 𝐶𝐿 = 0.7.

Figure 45 – 𝐶𝐷 convergence for 2D and 3D simulations (left) with zoomed in view (right) at
𝑅𝑒 = 6.5𝑥106 , 𝑀∞ = 0.72 and 𝐶𝐿 = 0.7.

In this example, considering the mean value of the last 5,000 iterations, the drag
coefficient obtained was 0.0113 for both 2D and 3D cases. Given the appreciable agreement in
results, this procedure is assumed adequate to assess drag coefficient values obtained in drag
rise simulations assisted by the CL Driver Wizard.
73

4.2 Wavy Leading Edge Effect in Transonic Regime

4.2.1 Drag Rise

Figures 46 through 48 present the drag rise curves obtained for 𝐶𝐿 = 0.1, 𝐶𝐿 = 0.3 and
𝐶𝐿 = 0.5, for the Reynolds numbers of 1x106, 6.5x106 and 2.2x107.

Figure 46 – Drag rise curves for SLE and WLE wings at 𝑅𝑒 = 1𝑥106 .

Figure 47 – Drag rise curves for SLE and WLE wings at 𝑅𝑒 = 6.5𝑥106 .
74

Figure 48 – Drag rise curves for SLE and WLE wings at 𝑅𝑒 = 2.2𝑥107 .

From Figures 46 through 48, the effects of the wavy leading edge are only
distinguishable for 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.1 (Figure 46). Thus, to aid the assessment of
results, Tables 7 through 9 present the effect of tubercles in terms of drag increments (Δ𝐶𝐷 )
from the reference wing, where positive values mean the drag coefficient is higher for the
tubercled configuration at the same flow conditions.

Table 7 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 1𝑥106 .

Lift Coefficient (𝑪𝑳 )


0.1 0.3 0.5
0.50 0 (0%) +1 (+1%) +2 (+2%)
0.60 +1 (+1%) +1 (+1%) +4 (+3%)
0.65 +1 (+1%) +2 (+2%) +5 (+4%)
0.68 +2 (+1%) +3 (+2%) +5 (+5%)
Mach Number

0.70 +2 (+2%) +3 (+3%) +9 (+9%)


0.72 +3 (+2%) +4 (+3%) +12 (+10%)
0.74 +4 (+3%) +5 (+4%) +10 (+9%)
0.76 +4 (+4%) +5 (+4%) +15 (+9%)
0.78 +5 (+3%) +6 (+4%) +27 (+10%)
0.80 +5 (+2%) +8 (+3%) +20 (+5%)
0.82 +7 (+2%) +10 (+3%) +21 (+3%)
0.84 +11 (+2%) +11 (+2%) +21 (+3%)
0.86 +11 (+2%) +12 (+1%) +19 (+2%)
75

At 𝑅𝑒 = 1𝑥106 , Table 7 indicates tubercles increase drag for all Mach number and lift
coefficient conditions. Furthermore, an increase in lift coefficient further increases the penalty
in drag due to the addition of tubercles, a tendency that is also observed for increasing Mach
number up to 𝑀∞ = 0.76 for 𝐶𝐿 = 0.1, and up to 𝑀∞ = 0.78 for 𝐶𝐿 = 0.3 and 𝐶𝐿 = 0.5.
Beyond those Mach numbers, however, this tendency is suppressed, as the percentage increase
in drag due to tubercles is slightly diminished.

Table 8 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 6.5𝑥106 .

Lift Coefficient (𝑪𝑳 )


0.1 0.3 0.5
0.50 +0 (+0%) +0 (+0%) +1 (+1%)
0.60 +0 (+0%) +0 (+0%) +1 (+1%)
0.65 +0 (+0%) +0 (+0%) +1 (+1%)
0.68 +0 (+0%) +0 (+1%) +1 (+1%)
Mach Number

0.70 +0 (+0%) +0 (+0%) +1 (+1%)


0.72 +0 (+0%) +0 (+0%) +1 (+1%)
0.74 +0 (+0%) +1 (+1%) +1 (+1%)
0.76 +0 (+0%) +1 (+1%) +1 (+1%)
0.78 +0 (+0%) +1 (+0%) +14 (+6%)
0.80 +1 (+0%) +1 (+0%) -5 (-1%)
0.82 +2 (+0%) +1 (+0%) -3 (-0%)
0.84 +2 (+0%) +2 (+0%) +1 (+0%)
0.86 +3 (+0%) +2 (+0%) +11 (+1%)

According to Table 8, at 𝑅𝑒 = 6.5𝑥106 the increase in drag is negligible for 𝐶𝐿 = 0.1


and 𝐶𝐿 = 0.3. At 𝐶𝐿 = 0.5, this increase is minimal up to 𝑀∞ = 0.76, with a sudden peak at
𝑀∞ = 0.78. At 𝑀∞ = 0.80 and 𝑀∞ = 0.82, however, the wavy leading edge results in a slight
decrease in drag for the first time.
This interesting effect is reproduced for even more test cases at 𝑅𝑒 = 2.2𝑥107 , as shown
in Table 9. Similarly to the results obtained at 𝑅𝑒 = 6.5𝑥106 , at 𝐶𝐿 = 0.5 there is a small drag
increment that peaks at 𝑀∞ = 0.78, and at 𝑀∞ = 0.80 and 𝑀∞ = 0.82 a decrease in drag is
observed. Also analogously to 𝑅𝑒 = 6.5𝑥106 , there is little to no drag increment for 𝐶𝐿 = 0.1
and 𝐶𝐿 = 0.3, with the additional benefit that past 𝑀∞ = 0.78 there are six test cases in which
drag is decreased due to the addition of tubercles.
76

Table 9 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 2.2𝑥107 .

Lift Coefficient (𝑪𝑳 )


0.1 0.3 0.5
0.50 +0 (+0%) +0 (+1%) +1 (+1%)
0.60 +0 (+0%) +0 (+1%) +1 (+1%)
0.65 +0 (+0%) +0 (+1%) +1 (+1%)
0.68 +0 (+1%) +0 (+1%) +1 (+1%)
Mach Number

0.70 +0 (+1%) +0 (+1%) +1 (+1%)


0.72 +0 (+1%) +0 (+1%) +1 (+1%)
0.74 +0 (+1%) +0 (+1%) +1 (+1%)
0.76 +0 (+0%) +1 (+1%) +5 (+5%)
0.78 +0 (+0%) +1 (+0%) +12 (+6%)
0.80 -2 (-1%) -5 (-2%) -1 (-0%)
0.82 -4 (-1%) -4 (-1%) -6 (-1%)
0.84 +2 (+0%) -4 (-1%) +8 (+1%)
0.86 +2 (+0%) -4 (-0%) +9 (+1%)

From the above discussed results, it can be noticed that the effectiveness of tubercles is
improved with increasing Reynolds numbers, similarly to the studies performed at subsonic
flows presented in section 2.2.4. The effect of Mach number, however, is not as explicit. At
𝑅𝑒 = 1𝑥106 , increasing Mach number results in further increase in drag up to a certain
threshold, past which this effect seems to be reversed as the percentage drag increase
diminishes. At 𝑅𝑒 = 6.5𝑥106 and 𝑅𝑒 = 2.2𝑥107 , the initial increase in drag is much more
subtle, and even slight drag reductions are observed past the threshold. This behavior is further
explored in the following subsections.

4.2.2 Drag Breakdown

In this section, the wavy leading edge is assessed in terms of its effect in viscous (𝐶𝐷𝑣 )
and inviscid (𝐶𝐷𝑖 ) drag. Considering viscous compressible flow around infinite wings, these
two components can be translated as wave and profile drag respectively, according to the drag
decomposition given in section 2.1.1.
Tables 10 through 18 present the breakdown of drag components for each test case. The
left-hand side of these tables provides the drag breakdown for the baseline configuration, while
the right-hand side gives the drag increment due to the wavy leading edge. To aid the analysis,
the effect of tubercles in percentage drag increment is also illustrated in Figures 49 through 57.
77

The left-hand side of Table 10 shows that the effect on drag from increasing Mach
number is according to expected for a constant Reynolds number: inviscid drag increases and
viscous drag decreases. Furthermore, comparing the magnitude of both drag components, the
total drag is mostly viscous up to 𝑀∞ = 0.76. Above 𝑀∞ = 0.78, the inviscid drag increases
rapidly and becomes responsible for the pronounced increase in total drag, as a result of the
wave drag associated to shock wave formation.
From the percentage drag increment in Figure 49, the viscous drag due to tubercles
consistently increases with increasing Mach number, reaching an increment of 10%. As for
inviscid drag, this unfavorable effect is also observed up to 𝑀∞ = 0.74, when tubercles result
in an increase of nearly 4%. Above 𝑀∞ = 0.76, however, this percentage is reduced, and from
𝑀∞ = 0.80 the wavy leading edge accounts for only 1% of increase. Since above 𝑀∞ = 0.78
inviscid drag greatly exceeds viscous drag, this percentage reduction is reflected in total drag
as well.
According to Table 10 and Figure 49, although the wavy leading edge adversely affects
both inviscid and viscous drag components, the percentage increase in inviscid drag does not
follow the same trend of the percentage increase in viscous drag. Above 𝑀∞ = 0.78, the
inviscid drag is predominantly wave drag. Thus, even though the wavy leading edge still results
in an increase in drag, this increase becomes less prominent with shock wave formation.

Table 10 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.1.

SLE Wing WLE Wing


𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫 𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫
0.50 25 81 105 +0 (0%) +0 (+0%) +0 (+0%)
0.60 28 79 107 +0 (+1%) +0 (+1%) +1 (+1%)
0.65 31 78 108 +0 (+1%) +1 (+1%) +1 (+1%)
0.68 33 76 109 +1 (+2%) +1 (+1%) +2 (+1%)
Mach Number

0.70 35 76 110 +1 (+2%) +1 (+2%) +2 (+2%)


0.72 37 74 111 +1 (+3%) +2 (+2%) +3 (+2%)
0.74 39 73 112 +2 (+4%) +2 (+3%) +4 (+3%)
0.76 49 70 120 +2 (+3%) +3 (+4%) +4 (+4%)
0.78 85 66 150 +2 (+2%) +3 (+5%) +5 (+3%)
0.80 170 59 229 +2 (+1%) +3 (+6%) +5 (+2%)
0.82 296 54 349 +3 (+1%) +4 (+8%) +7 (+2%)
0.84 462 53 515 +5 (+1%) +5 (+10%) +11 (+2%)
0.86 661 60 720 +5 (+1%) +6 (+10%) +11 (+2%)
78

Figure 49 – Percentage drag increment at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.1.

The effect of tubercles at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.3 is very similar to that at 𝐶𝐿 = 0.1,
as shown in Table 11 and Figure 50. In this case, the percentage drag increase due to tubercles
is slightly higher, with viscous drag increment reaching almost 14%, inviscid drag increment
around 6% and total drag increment peaking at 4%.

Table 11 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.3.

SLE Wing WLE Wing


𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫 𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫
0.50 27 80 108 +0 (+1%) +0 (+0%) +1 (+1%)
0.60 31 79 109 +1 (+2%) +1 (+1%) +1 (+1%)
0.65 33 77 110 +1 (+3%) +1 (+1%) +2 (+2%)
0.68 35 76 111 +1 (+3%) +1 (+2%) +3 (+2%)
Mach Number

0.70 37 75 112 +1 (+4%) +2 (+2%) +3 (+3%)


0.72 38 75 113 +2 (+6%) +2 (+2%) +4 (+3%)
0.74 41 73 114 +2 (+5%) +2 (+3%) +5 (+4%)
0.76 48 71 119 +2 (+5%) +3 (+4%) +5 (+4%)
0.78 88 67 154 +3 (+3%) +3 (+5%) +6 (+4%)
0.80 192 60 252 +5 (+2%) +4 (+6%) +8 (+3%)
0.82 364 54 418 +5 (+1%) +5 (+10%) +10 (+3%)
0.84 564 52 616 +5 (+1%) +6 (+12%) +11 (+2%)
0.86 791 55 846 +4 (+1%) +8 (+14%) +12 (+1%)
79

Figure 50 – Percentage drag increment at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.3.

At 𝐶𝐿 = 0.5, as previously presented in Table 7, the drag increment due to the wavy
leading edge is significantly greater, reaching a 10% increase. According to Table 12 and Figure
51, this increase is primarily attributed to inviscid drag. The percentage increase in inviscid
drag reaches 20% at 𝑀∞ = 0.72. However, as observed at 𝐶𝐿 = 0.1 and 𝐶𝐿 = 0.3, this trend is
reduced as the Mach number further increases.

Table 12 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.5.


SLE Wing WLE Wing
𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫 𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫
0.50 36 79 114 +2 (+5%) +0 (+0%) +2 (+2%)
0.60 37 76 114 +3 (+8%) +1 (+1%) +4 (+3%)
0.65 43 75 117 +3 (+8%) +2 (+2%) +5 (+4%)
0.68 40 73 113 +3 (+8%) +2 (+3%) +5 (+5%)
Mach Number

0.70 38 72 111 +7 (+18%) +3 (+4%) +9 (+9%)


0.72 44 72 115 +9 (+20%) +3 (+4%) +12 (+10%)
0.74 52 70 122 +7 (+14%) +3 (+4%) +10 (+9%)
0.76 86 67 154 +12 (+13%) +3 (+4%) +15 (+9%)
0.78 199 61 260 +23 (+12%) +4 (+6%) +27 (+10%)
0.80 383 56 439 +14 (+4%) +6 (+11%) +20 (+5%)
0.82 561 53 614 +14 (+3%) +7 (+13%) +21 (+3%)
0.84 758 51 809 +13 (+2%) +8 (+15%) +21 (+3%)
0.86 993 49 1042 +6 (+1%) +12 (+26%) +19 (+2%)
80

Figure 51 – Percentage drag increment at 𝑅𝑒 = 1𝑥106 and 𝐶𝐿 = 0.5.

From the results obtained at 𝑅𝑒 = 1𝑥106 , the wavy leading edge leads to an increase in
both viscous and inviscid drag. As the lift coefficient increases, the drag increment is aggravated
mainly due to inviscid drag. In addition, the adverse effect of tubercles on viscous drag is
intensified with increasing Mach number, a tendency that is also observed on inviscid drag up
to a certain Mach number threshold. Beyond this threshold, however, this effect in inviscid drag
is diminished as the Mach number increases.
At 𝑅𝑒 = 6.5𝑥106 , the detrimental effect of tubercles on both inviscid and viscous drag
is greatly reduced, as shown in Tables 13 through 15 and Figures 52 through 54. The viscous
drag increment is nearly null for the entire Mach number and lift coefficient ranges, in
opposition to what has been observed at 𝑅𝑒 = 1𝑥106 . As for inviscid drag, up to 𝑀∞ = 0.72
an increase in Mach number has little effect on the percentage drag increase produced by the
wavy leading edge, regardless of the lift coefficient. Above 𝑀∞ = 0.72, however, the effect of
tubercles at 𝐶𝐿 = 0.1 and 𝐶𝐿 = 0.3 is considerably different from that at 𝐶𝐿 = 0.5.
Similarly to the results obtained at 𝑅𝑒 = 1𝑥106 , at 𝐶𝐿 = 0.1 (Table 13 and Figure 52)
and 𝐶𝐿 = 0.3 (Table 14 Figure 53) the trend in percentage inviscid drag increase is diminished
above 𝑀∞ = 0.74. Also analogously to 𝑅𝑒 = 1𝑥106 , at 𝐶𝐿 = 0.5 (Table 15 and Figure 54) the
percentage inviscid drag increment reaches a distinct peak before being reduced. In this case,
however, the wavy leading edge leads to a slight decrease in both inviscid and total drag at
81

𝑀∞ = 0.80 and 𝑀∞ = 0.82, a result that supports the hypothesis that tubercles have the
potential to reduce wave drag in transonic regime (BOLZON, 2015).

Table 13 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.1.

SLE Wing WLE Wing


𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫 𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫
0.50 18 63 81 +0 (+1%) +0 (+0%) +0 (+0%)
0.60 21 62 82 +0 (+1%) +0 (+0%) +0 (+0%)
0.65 23 61 83 +0 (+1%) +0 (+0%) +0 (+0%)
0.68 24 60 84 +0 (+1%) +0 (+0%) +0 (+0%)
Mach Number

0.70 26 60 85 +0 (+1%) +0 (+0%) +0 (+0%)


0.72 27 59 87 +0 (+1%) +0 (+0%) +0 (+0%)
0.74 29 59 88 +0 (+1%) +0 (+0%) +0 (+0%)
0.76 40 57 97 +0 (+0%) +0 (+0%) +0 (+0%)
0.78 76 54 130 +0 (+0%) +0 (+0%) +0 (+0%)
0.80 163 50 212 +1 (+0%) +0 (+0%) +1 (+0%)
0.82 295 47 342 +1 (+0%) +0 (+0%) +2 (+0%)
0.84 463 47 510 +1 (+0%) +0 (+1%) +2 (+0%)
0.86 667 51 718 +3 (+0%) +0 (+0%) +3 (+0%)

Drag Breakdown ‒ Re = 6.5x106, CL = 0.1


+2.0%
Percentage Drag Increment (%ΔCD)

Viscous Drag
Inviscid Drag
+1.5%
Total Drag

+1.0%

+0.5%

0.0%
0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85
Mach Number (M)

Figure 52 – Percentage drag increment at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.1.


82

Table 14 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.3.

SLE Wing WLE Wing


𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫 𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫
0.50 20 63 83 +0 (+1%) +0 (+0%) +0 (+0%)
0.60 23 62 84 +0 (+1%) +0 (+0%) +0 (+0%)
0.65 24 61 85 +0 (+1%) +0 (+0%) +0 (+0%)
0.68 26 60 86 +0 (+1%) +0 (+0%) +0 (+1%)
Mach Number

0.70 27 60 87 +0 (+1%) +0 (+0%) +0 (+1%)


0.72 28 59 88 +0 (+1%) +0 (+0%) +0 (+1%)
0.74 31 59 89 +0 (+1%) +0 (+0%) +1 (+1%)
0.76 37 58 95 +0 (+1%) +0 (+0%) +1 (+1%)
0.78 77 55 132 +0 (+1%) +0 (+0%) +1 (+0%)
0.80 181 51 232 +1 (+1%) +0 (+0%) +1 (+0%)
0.82 353 48 401 +1 (+0%) +0 (+1%) +1 (+0%)
0.84 555 47 602 +1 (+0%) +0 (+0%) +2 (+0%)
0.86 784 50 833 +2 (+0%) +0 (+0%) +2 (+0%)

Figure 53 – Percentage drag increment at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.3.


83

Table 15 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.5.


SLE Wing WLE Wing
𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫 𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫
0.50 26 62 88 +1 (+3%) +0 (+0%) +1 (+1%)
0.60 27 61 88 +1 (+3%) +0 (+0%) +1 (+1%)
0.65 32 60 92 +1 (+3%) +0 (+0%) +1 (+1%)
0.68 31 60 91 +1 (+3%) +0 (+0%) +1 (+1%)
Mach Number

0.70 29 59 88 +1 (+3%) +0 (+0%) +1 (+1%)


0.72 33 59 92 +2 (+3%) +0 (+0%) +1 (+1%)
0.74 42 58 100 +1 (+2%) +0 (+0%) +1 (+1%)
0.76 76 56 132 +1 (+1%) +0 (+0%) +1 (+1%)
0.78 178 53 230 +14 (+8%) +0 (+0%) +14 (+6%)
0.80 365 49 414 -6 (-2%) +1 (+2%) -5 (-1%)
0.82 548 47 595 -3 (-1%) +0 (+1%) -3 (-0%)
0.84 744 47 791 +1 (+0%) +0 (+0%) +1 (+0%)
0.86 972 48 1020 +11 (+1%) +0 (+0%) +11 (+1%)

Figure 54 – Percentage drag increment at 𝑅𝑒 = 6.5𝑥106 and 𝐶𝐿 = 0.5.

At 𝑅𝑒 = 2.2𝑥107 , according to Tables 16 through 18 and Figures 55 through 57, the


effect of tubercles on viscous and inviscid drag is comparable to what has been observed at
𝑅𝑒 = 6.5𝑥106 , with an additional improvement above the Mach number thresholds at each lift
coefficient. At 𝐶𝐿 = 0.1 (Table 16 and Figure 55), the wavy leading edge causes a 1% decrease
in drag at 𝑀∞ = 0.80 and 𝑀∞ = 0.82. At 𝐶𝐿 = 0.3 (Table 17 and Figure 56), a 2% decrease is
84

obtained at 𝑀∞ = 0.80 and a slight drag reduction is sustained up to 𝑀∞ = 0.86. At 𝐶𝐿 = 0.5


(Table 18 and Figure 57), drag is decreased in 1% at 𝑀∞ = 0.82. Particular interest must be
given for the drag reduction obtained from the wavy leading edge at 𝐶𝐿 = 0.3 and 𝐶𝐿 = 0.5,
since commercial transport aircraft typically cruise inside an envelope of 0.76 < 𝑀∞ < 0.86
and 0.3 < 𝐶𝐿 < 0.5 (ROSKAM, 1998).

Table 16 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.1.

SLE Wing WLE Wing


𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫 𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫
0.50 15 53 68 +0 (+1%) +0 (+0%) +0 (+0%)
0.60 17 52 69 +0 (+1%) +0 (+0%) +0 (+0%)
0.65 19 51 69 +0 (+1%) +0 (+0%) +0 (+0%)
0.68 20 50 70 +0 (+1%) +0 (+0%) +0 (+1%)
Mach Number

0.70 21 50 71 +0 (+1%) +0 (+0%) +0 (+1%)


0.72 22 50 72 +0 (+1%) +0 (+0%) +0 (+1%)
0.74 24 49 73 +0 (+1%) +0 (+0%) +0 (+1%)
0.76 34 48 82 +0 (+0%) +0 (+0%) +0 (+0%)
0.78 71 46 117 +0 (+0%) +0 (+0%) +0 (+0%)
0.80 158 42 200 -2 (-1%) +0 (+1%) -2 (-1%)
0.82 303 40 344 -4 (-1%) +0 (+0%) -4 (-1%)
0.84 460 40 499 +2 (+0%) +0 (+0%) +2 (+0%)
0.86 668 43 711 +2 (+0%) +0 (+0%) +2 (+0%)

Figure 55 – Percentage drag increment at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.1.


85

Table 17 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.3.

SLE Wing WLE Wing


𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫 𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫
0.50 17 53 69 +0 (+1%) +0 (+0%) +0 (+1%)
0.60 19 52 70 +0 (+1%) +0 (+0%) +0 (+1%)
0.65 20 51 71 +0 (+1%) +0 (+0%) +0 (+1%)
0.68 21 51 72 +0 (+1%) +0 (+0%) +0 (+1%)
Mach Number

0.70 22 50 72 +0 (+1%) +0 (+0%) +0 (+1%)


0.72 23 50 73 +0 (+1%) +0 (+0%) +0 (+1%)
0.74 25 49 74 +0 (+1%) +0 (+0%) +0 (+1%)
0.76 31 49 79 +0 (+1%) +0 (+0%) +1 (+1%)
0.78 71 46 117 +0 (+0%) +0 (+0%) +1 (+0%)
0.80 170 43 214 -5 (-3%) +0 (+0%) -5 (-2%)
0.82 342 40 382 -4 (-1%) -0 (-1%) -4 (-1%)
0.84 545 39 585 -4 (-1%) +1 (+1%) -4 (-1%)
0.86 784 42 826 -4 (-1%) -0 (-0%) -4 (-0%)

Figure 56 – Percentage drag increment at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.3.


86

Table 18 – Effect of tubercles on drag coefficient (Δ𝐶𝐷 ) at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.5.


SLE Wing WLE Wing
𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫 𝚫𝑪𝑫𝒊 𝚫𝑪𝑫𝒗 𝚫𝑪𝑫
0.50 22 52 74 +1 (+2%) +0 (+0%) +1 (+1%)
0.60 23 51 74 0 (+2%) 0 (0%) +1 (+1%)
0.65 24 50 75 +1 (+2%) +0 (+0%) +1 (+1%)
0.68 25 50 75 +1 (+2%) +0 (+0%) +1 (+1%)
Mach Number

0.70 26 50 75 +0 (+3%) +0 (+0%) +1 (+1%)


0.72 27 49 76 +1 (+3%) +0 (+0%) +1 (+1%)
0.74 32 49 81 +1 (+3%) +0 (+0%) +1 (+1%)
0.76 68 47 115 +4 (+6%) 0 (0%) +5 (+5%)
0.78 155 45 200 +12 (+8%) 0 (0%) +12 (+6%)
0.80 340 42 382 -1 (-0%) +0 (+0%) -1 (-0%)
0.82 532 40 572 -7 (-1%) +0 (+1%) -6 (-1%)
0.84 727 40 766 +8 (+1%) -0 (-0%) +8 (+1%)
0.86 962 41 1003 +9 (+1%) 0 (0%) +9 (+1%)

Drag Breakdown ‒ Re = 2.2x107, CL = 0.5


Percentage Drag Increment (%ΔCD)

+8.0%
Viscous Drag
Inviscid Drag
+6.0%
Total Drag

+4.0%

+2.0%

0.0%

-2.0%
0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85
Mach Number (M)

Figure 57 – Percentage drag increment at 𝑅𝑒 = 2.2𝑥107 and 𝐶𝐿 = 0.5.


87

4.2.3 Flow Topology

The analysis of drag coefficient presented in the previous subsections revealed a few
interesting results for the wavy leading edge in transonic regime. In order to interpret them, the
effect of tubercles on flow topology is explored in this section.
Figure 58 shows the Mach contours for the cross-sections of both wings at 𝑅𝑒 = 1𝑥106 ,
𝑀∞ = 0.68 and 𝐶𝐿 = 0.1. Since the airfoil of the tubercled wing varies periodically spanwise,
three distinct cross-sections are displayed: trough, peak and mean; where the mean-section
corresponds to the cross-section of the baseline wing. In this flow condition, the highest local
Mach number reaches around 𝑀 = 0.89 for the four cross sections.
Further exploring this test case, Figure 59 presents the pressure coefficient contours and
chordwise pressure distribution at the same flow condition. It can be seen that the wavy leading
edge generates a distinct periodic spanwise pressure distribution over the wing up to 40% of
the chord length, being dissipated further downstream. In addition, stronger adverse pressure
gradients are encountered at troughs, leading to small regions of separation near the trailing
edge.

Figure 58 – Mach number contours at 𝑅𝑒 = 1𝑥106 , 𝑀∞ = 0.68 and 𝐶𝐿 = 0.1.


88

Figure 59 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom)
at 𝑅𝑒 = 1𝑥106 , 𝑀∞ = 0.68 and 𝐶𝐿 = 0.1.
89

At 𝐶𝐿 = 0.5, the effect of the wavy leading edge on flow topology is more substantial.
In this case, the highest local Mach number reaches 𝑀 = 1.01 for the baseline wing and at the
mean-section of the tubercled wing, 𝑀 = 0.98 at the peak-section and 𝑀 = 1.27 at the trough-
section. Figure 60 and Figure 61 show this peak in Mach number occurs at the leading edge,
and it is related to a much stronger adverse pressure gradient occurring only at the trough.
Further downstream of the trough, the flow goes through compression and decelerates back to
subsonic speed. Then, the effect of the wavy leading edge on chordwise pressure distribution
begins to dissipate, becoming inappreciable past 40% of the chord length.

Figure 60 – Mach number contours at 𝑅𝑒 = 1𝑥106 , 𝑀∞ = 0.68 and 𝐶𝐿 = 0.5.


90

Figure 61 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom)
at 𝑅𝑒 = 1𝑥106 , 𝑀∞ = 0.68 and 𝐶𝐿 = 0.5.
91

As exemplified by the flow topology at 𝐶𝐿 = 0.1 and 𝐶𝐿 = 0.5, the difference in


chordwise pressure distribution near the leading edge between peak, mean and trough-sections
is amplified with increasing lift coefficient. In addition, at 𝐶𝐿 = 0.5 the flow is compressed and
rapidly decelerated behind troughs. This behavior could explain the trend in inviscid drag seen
in the analysis of drag component breakdown. In section 4.2.2, it was noticed that the inviscid
drag consistently increased with increasing lift coefficients at all Reynolds numbers up to a
certain Mach number threshold. This trend can be related to the increase in pressure drag
associated with the rapid flow deceleration that occurs behind troughs.
At 𝑀∞ = 0.76 and 𝐶𝐿 = 0.5, a shock wave occurs and leads to boundary layer
thickening, as it can be seen in Figure 62. The streamlines over pressure coefficient contours in
Figure 63 shows the shock wave leads to local flow separation, followed by reattachment and
another separated region closer to the trailing edge. It can be seen that the shock is anticipated
chordwise due to the wavy leading edge, an effect that is accompanied by a consequent
anticipated and considerably larger separated region at the trailing edge. This effect contributes
to an increase in viscous form drag, as observed in section 4.2.2.

Figure 62 – Mach number contours at 𝑅𝑒 = 1𝑥106 , 𝑀∞ = 0.76 and 𝐶𝐿 = 0.5.


92

Figure 63 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom)
at 𝑅𝑒 = 1𝑥106 , 𝑀∞ = 0.76 and 𝐶𝐿 = 0.5.
93

At 𝑀∞ = 0.78 and 𝐶𝐿 = 0.5, the shock wave is anticipated chordwise at all cross-
sections for the tubercled wing, as shown in Figures 64 and 65. This shock leads to complete
flow separation, and it can be seen that a larger wake occurs at the trough-section. This behavior
is worsened with increasing Mach number and lift coefficient, thus explaining the ever-
increasing viscous drag due to the wavy leading edge at 𝑅𝑒 = 1𝑥106 .

Figure 64 – Mach number contours at 𝑅𝑒 = 1𝑥106 , 𝑀∞ = 0.78 and 𝐶𝐿 = 0.5.


94

Figure 65 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom)
at 𝑅𝑒 = 1𝑥106 , 𝑀∞ = 0.78 and 𝐶𝐿 = 0.5.
95

The remaining of this section will focus on the design space of commercial aircraft at
cruise flight, that is, at 𝑅𝑒 = 2.2𝑥107 , with Mach number ranging from 0.76 to 0.86 lift
coefficient between 𝐶𝐿 = 0.3 and 𝐶𝐿 = 0.5.
As presented in Figure 56, at 𝐶𝐿 = 0.3 the wavy leading edge causes around 1% increase
in inviscid drag coefficient up to 𝑀∞ = 0.76. Figure 66 shows the shock wave formation is
similar for all cross-sections at 𝑀∞ = 0.76 and 𝐶𝐿 = 0.3. Furthermore, in Figure 67 it can be
seen the spanwise periodic pressure distribution of the tubercled wing is aligned with the
pressure distribution of the baseline wing, and no separation occurs. Such observations
corroborate to similar inviscid and viscous drag coefficients for both wing geometries.

Figure 66 – Mach number contours at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.76 and 𝐶𝐿 = 0.3.


96

-1.4
-1.2
-1.0
-0.8
-0.6
Pressure coefficient (CP)

-0.4
-0.2
0.0
0.2
0.4
0.6 SLE
WLE - mean
0.8
WLE - trough
1.0
WLE - peak
1.2
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
x/c

Figure 67 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom)
at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.76 and 𝐶𝐿 = 0.3.
97

At 𝑀∞ = 0.80 and 𝐶𝐿 = 0.3, the wavy leading edge leads to a decrease in drag
coefficient. Figure 68 gives the Mach contours at this flow condition, in which near normal
shock waves are formed in the upper and lower surfaces of the wings. Shock structures are
indistinguishable between configurations, however shock-induced boundary layer thickening is
noticeably greater at the trough and smaller at the peak-position of the tubercled wing. The
periodic spanwise pressure distribution for the tubercled wing is aligned with that of the
baseline wing, as shown in Figure 69, but the flow separation region is quite different for both
configurations. The baseline wing presents a distinct chordwise shock-induced local separation
region, which is followed by reattachment downstream and separation at the trailing edge. On
the other hand, the tubercled wing exhibits separation regions that varies periodically spanwise.
It can be noticed that behind troughs the flow is completely separated downstream the shock
wave, whereas behind peaks large reattachment regions are encountered. Furthermore, although
the shock wave is anticipated chordwise behind troughs, it is delayed further downstream
behind peaks. Thus, results indicate that the decrease in drag coefficient is a consequence of the
periodicity of the pressure distribution and the combined effects of troughs and peaks on shock
wave formation and flow separation.
A decrease in drag is also obtained at 𝑀∞ = 0.82 and 𝐶𝐿 = 0.3. In this flow condition,
it can be seen in Figure 70 that the wavy leading edge results in a considerably thinner wake
region downstream of the shock wave. Figure 71 shows that complete separation occurs behind
the shock wave at nearly the same chordwise position for both configurations. A bow-shaped
spanwise pressure distribution is distinguished on the verge of flow separation, and the reversed
flow region presents two counter-rotating vortices in an asymmetric formation. In addition, in
this case the shock is delayed chordwise for the entire span of the wing with wavy leading edge.
98

Figure 68 – Mach number contours at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.80 and 𝐶𝐿 = 0.3.


99

Figure 69 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom)
at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.80 and 𝐶𝐿 = 0.3.
100

Figure 70 – Mach number contours at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.82 and 𝐶𝐿 = 0.3.


101

Figure 71 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom)
at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.82 and 𝐶𝐿 = 0.3.
102

At 𝐶𝐿 = 0.5, the two cases of interest are those at 𝑀∞ = 0.78 and 𝑀∞ = 0.82, which
correspond to the worst and best results obtained with the addition of tubercles at this lift
coefficient. According to Figure 72, at 𝑀∞ = 0.78 a near normal shock wave is formed at the
suction side of the wings, with a slightly larger wake encountered for the baseline configuration.
In Figure 73, it can be seen that local flow separation is marked by large reversed flow regions
on both wings, however only the tubercled configuration presents the two-counter rotating
vortices in this case. Furthermore, the periodic spanwise pressure distribution has an anticipated
chordwise development, resulting in a stronger pressure gradient over the entire wing. This
stronger gradient, in turn, leads to a chordwise anticipated shock wave formation, which is
responsible for the increase in drag coefficient observed in section 4.2.2.
103

Figure 72 – Mach number contours at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.78 and 𝐶𝐿 = 0.5.


104

Figure 73 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom)
at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.78 and 𝐶𝐿 = 0.5.
105

At 𝑀∞ = 0.82 and 𝐶𝐿 = 0.5, Figure 74 shows that for all cross-sections a stronger near
normal shock wave is formed at the suction side of the wings, accompanied by a weaker shock
at the pressure side. It can be seen that the baseline configuration is marked by a large flow
separation region with distinct recirculation. The wing with wavy leading edge, on the other
hand, presents a much smaller separation region, with a recirculation region being
distinguishable only at the peak-section. Furthermore, unlike previous test cases with flow
separation, in this flow condition the trough-section presents a thinner wake than the peak-
section. Figure 75 shows that a secondary adverse pressure gradient region is formed behind
troughs, and unlike the test case at 𝑀∞ = 0.78, the periodic spanwise pressure distribution of
the tubercled wing develops further downstream in comparison to the baseline configuration.
In addition, the wavy leading edge results in a significantly delayed chordwise shock wave
formation over the entire span of the wing, further contributing to a decrease in drag coefficient.
106

Figure 74 – Mach number contours at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.82 and 𝐶𝐿 = 0.5.


107

Figure 75 – Pressure coefficient contours (top) and chordwise pressure distribution (bottom)
at 𝑅𝑒 = 2.2𝑥107 , 𝑀∞ = 0.82 and 𝐶𝐿 = 0.5.
108

5 Conclusions

This work investigated the application of leading edge tubercles as a means of transonic
flow control, focusing on the effect of Mach and Reynolds numbers on the resulting drag
coefficient increment.
The adopted methodology was a numerical investigation using computational fluid
dynamics, in order to evaluate a wide range of test cases. Simulations were carried out for a
baseline and a tubercled RAE 2822 infinite wings in three different Reynolds numbers, with
Mach number ranging from 0.5 to 0.86 and at three different lift coefficients.
At 𝑅𝑒 = 1𝑥106 , results indicated tubercles provoked an overall increase in drag when
compared to the baseline wing, an effect that was worsened with increasing lift coefficient. On
the other hand, increasing Reynolds number consistently enhanced the transonic performance
of the tubercled wing in comparison to its smooth counterpart. This Reynolds number effect
had also been observed in subsonic regime for wings with trailing edge stall characteristics
(STANWAY, 2008).
In the subsonic range, the wing with wavy leading edge presented higher or equal drag
coefficient. Increasing Mach number further increased the drag increment up to a certain Mach
number threshold, which was around 𝑀∞ = 0.78 for all simulation conditions. Above
𝑀∞ = 0.78, this trend was reverted and the percentage penalty in drag due to tubercles was
diminished, with a few test cases even resulting in drag reduction.
From drag component breakdown analysis, results revealed that increasing flow speed
adversely affects the viscous drag increment due to tubercles at 𝑅𝑒 = 1𝑥106 . Nevertheless, this
effect was dissipated at 𝑅𝑒 = 6.5𝑥106 and 𝑅𝑒 = 2.2𝑥107 . The percentage increase in inviscid
drag was also negatively affected by increasing Mach number up to a certain threshold, which
was between 𝑀∞ = 0.72 and 𝑀∞ = 0.78 for all test cases. Above the Mach number threshold,
however, the percentage penalty in drag was reduced. In the test cases that presented drag
coefficient reduction, this behavior supports the hypothesis of tubercles having the potential to
reduce wave drag in transonic regime (BOLZON, 2015).
Flow visualization enlightened the three-dimensionality of the numerical results. The
periodicity of the spanwise pressure distribution was observed, in accordance to previous works
in subsonic (WATTS; FISH, 2001; SKILLEN et al., 2015; DE PAULA, 2016) and transonic
109

(PEREZ et al., 2018) regimes. Furthermore, in all test cases, stronger adverse pressure gradients
were found at troughs (VAN NIEROP et al., 2008).
Chordwise pressure distributions showed that the shock wave occurred at distinct
chordwise locations between peak, trough and mean cross-sections; which also affected the
flow topology behind the shock wave. In the majority of test cases, tubercles resulted in
anticipated shock wave chordwise formation in the entire span of the wing. Nevertheless,
similarly to previous investigations performed in transonic regime (PADILHA, 2017;
SEPETAUSKAS, 2018; PEREZ et al., 2018), in some of the simulated conditions the shock
wave was located further downstream in comparison to the baseline wing, leading to a decrease
in drag.
Results obtained at 𝑅𝑒 = 2.2𝑥107 , 0.8 < 𝑀∞ < 0.82 and 0.3 < 𝐶∞ < 0.5 are of
particular interest to the aircraft industry. Even though the greatest drag reduction was of only
6 drag counts (-1%), those flow conditions are within the typical envelope experienced by
commercial aircraft in cruise flight – a promising result towards considering the wavy leading
edge as a means of transonic flow control.
In this study, a single airfoil and wavy leading edge configuration was analyzed. Future
studies would benefit from investigating different tubercle and airfoil geometries, aiming to
broaden the design space in which tubercles enhance transonic performance. Geometric
parameters were demonstrated to have a huge impact on tubercle performance in subsonic flows
(LEVSHIN et al., 2006; HANSEN et al., 2009; HANSEN, 2012; ROSTAMZADEH et al.,
2014; DE PAULA, 2016). In transonic flow, however, the effect of airfoil and tubercle
geometries has not been explored yet.
Although this investigation presented interesting results, there are limitations to the
extent of their validity. In the Verification and Validation phase, the numerical solution
presented issues regarding iterative convergence. A compromise solution between simulation
time and number of iterations was made by adopting 20,000 iterations, a number much greater
than expected for the type of flow being analyzed. Comparison between 2D and 3D simulations
showed the 3D solution oscillated around the 2D solution. In addition, a mean value of the last
5,000 iterations was considered to account for oscillations. Nevertheless, this oscillatory
behavior introduces inaccuracy and uncertainties to the results and cannot be ignored.
Another aspect of the Verification and Validation phase must be given attention to. The
numerical solution were compared to consolidated experimental data (COOK et al., 1979),
however at a single test case that was considered as validation for the entire test matrix. Even
though comparisons between numerical solutions with the same setup are less prone to bias,
110

this assumption might compromise the results obtained. Experimental investigations would
definitely enable a better assessment of the validity of the numerical method, as well as enhance
the understanding of the flow control phenomena involved.
111

References

ABRANTES, T. T.; CRUZ, A. A. R.; DE PAULA, A. A.; KLEINE, V. G., BÜTTNER, F.


The Wing Three-Dimensional Effects on Wavy Leading Edge Performance. 35th AIAA
Applied Aerodynamics Conference, AIAA AVIATION Forum, Denver, Colorado, 2017.

AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS. AIAA Guide for


the Verification and Validation of Computational Fluid Dynamics Simulations. AIAA G-
077-1998, 1998.

ANDERSON, JR. J.D. Fundamentals of Aerodynamics. McGraw-Hill Education, 5th


edition, 2010.

ANSYS, Inc. ANSYS Fluent 17.0 Theory Guide. ANSYS, Inc., 2016.

ASGHAR, A.; PEREZ, R. E.; FERCHICHI, M. Effect of Leading Edge Tubercles on


Transonic Performance of Airfoils. 35th AIAA Applied Aerodynamics Conference, AIAA
AVIATION Forum, Denver, Colorado, 2017.

BOLZON, M. D.; KELSO, R. M.; ARJOMANDI, M. Tubercles and Their Applications.


Journal of Aerospace Engineering, v. 29, n. 1, Apr. 2015.

BUSHNELL, D. M.; MOORE, K. J. Drag reduction in nature. Annual Review of Fluid


Mechanics, v. 23, p. 65-79, Jan. 1991.

ÇENGEL, Y. A.; CIMBALA, J. Fluid Mechanics: Fundamentals and Applications.


McGraw-Hill Education, 1st edition, 2004.

CHARMICHAEL, B. H. Low Reynolds number airfoil survey. NASA Report, v. 1, 1981.

COLLIS, S. S.; JOSLIN, D. R.; SEIFERT, A.; THEOFILIS, V. Issues in Active Flow
Control: Theory, Control, Simulation and Experiment. Progress in Aerospace Sciences, v.
40, p. 237-289, May 2004.

CONFORTH, J. Vava’u Humback Whale Breach. Available at:


<https://www.outdoorphotographer.com/vavau-humpback-whale-breach-3/>. Accessed on: 18
Aug. 2018.

COOK, P. H.; MCDONALD, M. A., FIRMIN, M.C.P. Aerofoil RAE 2822 - Pressure
Distributions, and Boundary Layer and Wake Measurements. Experimental Data Base for
Computer Program Assessment, AGARD Report AR-138, 1979.

CUSTODIO, D. The effect of Humpback whale-like leading edge potuberances on


hydrofoil performance. 2007. 76 p. Thesis (MSc in Mechanical Engineering) – Worcester
Polytechnic Institute, Worcester, MA, 2008.
112

DE PAULA, A. A. The airfoil thickness effects on wavy leading edge phenomena at low
Reynolds number regime. 2016. 341 p. Thesis (PhD in Energy and Fluids Mechanical
Engineering) - Escola Politécnica, University of Sao Paulo, Sao Paulo, 2016.

DEWAR, S. The Physics of Tubercle Airfoils and Their Application to Wind Turbines.
WhalePower Corporation, 2014. Available at: < https://whalepowercorp.wordpress.com/the-
science/>. Accessed on: 12 Sep. 2018.

EDEL, R. K.; WINN, H. E. Observations on underwater locomotion and flipper movement of


the humpback whale, Megaptera novaeangliae. Marine Biology, v. 48, n. 3, p. 279-287, 1978.

FERRARI, M. A. S. Automatic Method for Multiblock Structured Grid Generation. 2018


AIAA Aerospace Sciences Meeting, AIAA SciTech Forum, 2018.

FEYNMAN, R. The Feynman Lectures on Physics: Mainly Mechanics, Radiation and Heat.
Basic Books, 6th ed., New York, NY, 1977.

FILHO, G. B. L.; COSTA, A. L. M.; LIMA, G. R.; DE PAULA, A. A. A Numerical


investigation of the wavy leading edge phenomena at transonic regime. 2018 AIAA
Aerospace Sciences Meeting, AIAA SciTech Forum, 2018.

FISH, F. E; BATTLE, J. M. Hydrodynamic design of the humpback whale flipper. Journal of


Morphology, v. 225, p. 51-60, 1995.

GROVES, C. E. Dissertation Defense Computational Fluid Dynamics Uncertainty


Analysis for Payload Fairing Spacecraft Environmental Control. 2014. 180p. Thesis
(PhD in Fluid Mechanics and Thermodynamics) - NASA Kennedy Space Center, Cocoa
Beach, FL, 2014.

HANSEN, K. L.; KELSO, R. M.; DALLY, B. B.; BATILL, S. M. The effect of leading edge
tubercle geometry on the performance of different airfoils. Proceedings of the 7th World
Conference on Experimental Heat Transfer, Fluid Mechanics and Thermodynamics,
Krakow, Poland, 2009.

HANSEN, K. L. Effect of Leading Edge Tubercles on Airfoil Performance. 2012. 298 p.


Thesis (PhD in Fluid Mechanics) – University of Adelaide, Adelaide, 2012.

HOUGHTON, E. L.; CARPENTER, P. W.; COLLICOTT, S. H.; VALENTINE, D. T.


Aerodynamics for Engineering Students. 6th edition, Butterworth-Heinemann, England,
2012.

JARVIS, R. H. Quality and control - why structured grids are still in style. Available at: <
http://www.engineersjournal.ie/2017/09/26/quality-and-control-why-structured-grids-are-still-
in-style/>. Accessed on: 03 Oct. 2018.

JOHARI H.; HENOCH, C.; CUSTODIO, D.; LEVSHIN, A. Effects of leading-edge


protuberances on airfoil performance. AIAA Journal, v. 45, n. 11, p. 2634–2642, 2007
113

KUZMIN, D. Introduction to Computational Fluid Dynamics. Lecture, Institute of


Applied Mathematics, University of Dortmund. Available at: <http://www.mathematik.uni-
dortmund.de/~kuzmin/cfdintro/lecture1.pdf>. Accessed on: 20 Oct. 2018.

KUZMIN, D. A Guide to Numerical Methods for Transport Equations. Friedrich–


Alexander University Erlangen–Nürnberg, Germany, 2010.

LEATHERWOOD, S.; REEVES, R. R.; PERRIN, W. F.; EVANS, W. E. Whales, dolphins


and porpoises of the eastern north pacific and adjacent arctic waters: a guide to their
identification. Dover Publications, New York, 1988.

LEVSHIN, A.; CUSTODIO, D. Effects of leading edge protuberances on airfoil performance.


Proceedings of the 36th AIAA Fluid Dynamics Conference and Exhibit, San Francisco,
California, 2006.

LOCKARD, D. P.; CHOUDHARI, M. M.; BUNING, P. G. Influence of Spanwise Boundary


Conditions on Slat Noise Simulations. 21st AIAA/CEAS Aeroacoustics Conference, AIAA
AVIATION Forum, Dallas, Texas, 2015.

MADIER, D. (2014). Aerodynamic & mechanical updates 2014 volume II. Formula One
Technical Files, 2014. Available at: <http//www.f1-
forecast.com/index.php?option=com_content&view=article&id=84&Itemid=502&lang=en>.
Accessed on: 14 Aug. 2018.

MASON, W. H. Configuration Aerodynamics. Virginia Tech, Blacksburg, VA, 2018.

MCLEOD, M. Whale of an Idea: WhalePower’s Humpback-inspired Tubercle Technology


marks next evolution in airfoil design. Available at: <https://www.design-
engineering.com/features/whale-of-an-idea/>. Accessed on: 10 Aug. 2018.

MIKLOSOVIC, D.S.; MURRAY, M.M.; HOWLE, L.E.; FISH, F.E. Leading-edge tubercles
delay stall on Humpback whale (Megaptera Novaeangliae) Flippers. Physics of Fluids, v. 16,
n. 5, p. 39-42, 2004.

MIKLOSOVIC, D.S.; MURRAY, M.M.; HOWLE, L.E. Experimental evaluation of


sinusoidal leading edges. Journal of Aircraft, v. 44, n. 4, p. 1404-1407, 2007.

MUELLER, T. J.; DELAURIER, J. D. Aerodynamics of Small Vehicles. Annual Review of


Fluid Mechanics, v. 35, p. 89-111, 2003.

OLIVEIRA, G. L.; SANTOS, L. C. C.; MARTINS, A. L.; BECKER, G. G. A Tool for


Parametric Geometry and Grid Generation for Aircraft Configurations. Proceedings of the
26th ICAS Congress, 2008.

PADILHA, B. R. M. Desenvolvimento De Metodologias De Prototipagem Rápida Para O


Projeto De Modelos De Ensaio Em Túnel De Vento Transônico. 2017. 115p. Thesis (MSc
in Space Science and Technologies) - Instituto Tecnológico de Aeronáutica, Sao Jose dos
Campos, 2017.
114

PATERSON, E.G.; WILSON, R.V.; STERN, F. General-purpose parallel unsteady RANS


CFD code for ship hydrodynamics. IIHR Hydroscience and Engineering Report No. 432,
College of Engineering, The University of Iowa, 2003.

PEDRO, H.T.C; KOBAYASHI, M.H. Numerical study of stall delay on humpback whale
flippers. 46th AIAA Aerospace Sciences Meeting and Exhibit. Reno, Nevada, 2008.

PEREZ, R. E.; ASGHAR, A. Numerical Study of the Effects of Leading Edge Tubercles on
Transonic Performance of Airfoils. 2018 Applied Aerodynamics Conference, AIAA
AVIATION Forum, Atlanta, Georgia, 2018.

RINOIE, K.; OKUNO, M.; SUNADA, Y. Airfoil Stall Suppression by Use of a Bubble Burst
Control Plate. AIAA Journal, v. 47, n. 2, p. 322-330, 2009.

ROSKAM, J. Airplane Flight Dynamics and Automatic Flight Controls: Part I. Design,
Analysis and Research Corporation (DARcorporation), 2nd edition, 1998.

ROSTAMZADEH, N.; HANSEN, K. L; KELSO, R. M.; Dally, B. B. The formation


mechanism and impact of streamwise vortices on NACA 0021 airfoil's performance with
undulating leading edge modification. Physics of fluids, v. 26, n. 10, 2014.

RUSSELL, J. M. Length and Bursting of Separation Bubbles: A Physical Interpretation.


Science and Technology of Low Speed and Motorless Flight, NASA Conference
Publication 2085, Part 1, p. 177-202, 1979.

SEPETAUSKAS, V. A.; PADILHA, B. R.; DE PAULA, A. A.; DA SILVA, R. G. A. Wavy


Leading Edge Phenomena on Transonic Flow Regime. 2018 Flow Control Conference, AIAA
AVIATION Forum, Atlanta, Georgia, 2018.

SHUR, M. L.; STRELETS, M. K.; TRAVIN, A. K.; SPALART, P. R. Turbulence Modeling


in Rotating and Curved Channels: Assessing the Spalart-Shur Correction. AIAA Journal, v.
38, n. 5, p. 784–792, 2000.

SKILLEN, A.; REVELL, A.; PINELLI, A.; PIOMELLI, U.; FAVIER, J. Flow over a Wing
with Leading-Edge Undulations. AIAA Journal, v. 53, n. 2, p. 464-472, 2015.

SLATER, J. W. NPARC Alliance Validation Archive – RAE 2822 Transonic Airfoil.


NASA, 2002. Available at:
<https://www.grc.nasa.gov/WWW/wind/valid/raetaf/raetaf04/raetaf04.html>. Accessed on:
11 Aug. 2018.

SLATER, J. W. NPARC Alliance CFD Verification and Validation. NASA, 2008.


Available at: <https://www.grc.nasa.gov/www/wind/valid/homepage.html>. Accessed on: 10
Oct. 2018.

SOMERFIELD, M. Bite Size Tech: McLaren MP4-29 new tubercles inspired rear wing.
2014. Available at: < http://www.somersf1.co.uk/2014/07/bite-size-tech-mclaren-mp4-29-
new.html>. Accessed on: 5 Sep. 2018.
115

SOUZA, R. F.; MESTRINER, M. C.; SOUZA, M. A. F.; FERRARI, M. A. S.;


BREVIGLIERI, C.; SPODE, C. Aerodynamic and Static Aeroelastic Numerical Simulations
for the 6th AIAA CFD Drag Prediction Workshop, AIAA AVIATION Forum Conference,
2016.

SPALART, P. R.; ALLMARAS, S. R. A One-Equation Turbulence Model for Aerodynamic


Flows. 30th Aerospace Sciences Meeting and Exhibit, 1992.

SPALART, P. R.; SHUR, M. L.; STRELETS, M. KH.; TRAVIN., A. K. Direct Simulation


and RANS Modeling of Vortex Generator Flow. Flow, Turbulence and Combustion, v. 95,
issue 2-3, p. 335-350, 2015.

STANWAY, M. J. Hydrodynamic effects of leading edge tubercles on control surfaces


and in flapping foil propulsion. 2008. 101 p. Thesis (MSc in Aerospace Engineering) -
Massachusetts Institute of Technology, Cambridge, 2008.

STÖPPLER, S. The Experimental Investigation of the Wavy Leading Edge Phenomena


in a Transonic Regime. 2017. 101 p. Thesis (MSc in Aerospace Engineering) – University of
Stuttgart, Stuttgart, 2017.

STUMPE, J. Symbiosis: Why CFD and wind tunnels need each other. Aerospace America,
v. 56, n. 6, p. 30-33, 2018.

SUDHAKAR, S.; KARTHIKEYAN, N.; VENKATAKRISHNAN, L. Influence of leading


edge tubercles on aerodynamic characteristics of a high aspect-ratio UAV. Aerospace
Science and Technology, v. 69, p. 281-289, 2017.

TANG, D. M.; DOWELL, E. H. Chaotic stall response of helicopter rotor in forward flight.
Journal of Fluids and Structures, v. 6, issue 3, p. 311-335, 1992.

TORENBEEK, E. Synthesis of Subsonic Airplane Design. Delft University Press, Kluwer


Academic Publishers, 1982.

VAN NIEROP, E. A.; ALBEN, S.; BRENNER, M.P. How Bumps on Whale Flippers Delay
Stall: An Aerodynamic Model. Physical Review Letters, v. 100, n. 5, 2008.

WATTS, P.; FISH, F. E. The Influence of Passive, Leading Edge Tubercles on Wing
Performance. Proceedings of the Twelfth International Symposium on Unmanned
Untethered Submersible Technology, Durhan New, Hampshire, 2001.

WAHL, D. C. Biologically inspired product design. Tryarchy Press, 2016. Available at:
<https://hackernoon.com/biologically-inspired-product-design-1161497707c>. Accessed on:
23 Aug. 2018.

WHALEPOWER CORPORATION. WhalePower Tubercle Technology Applications.


Toronto, Canada, 2018. Available at: <https://whalepowercorp.wordpress.com/applications/>.
Accessed on: 5 Sep. 2018.

WHITCOMB, R.; CLARK, L. An Airfoil Shape for Efficient Flight at Supercritical Mach
Numbers. Tech. Rep. NASA TM X-1109, NASA, 1965.
116

WILCOX, D. C. Turbulence Modeling for CFD. DCW Industries, 3rd edition, 2006.

ZVERKOV, I.; ZANIN, B.; KOZLOV, V. Disturbances growth in boundary layers on


classical and surface wings. AIAA Journal, v. 46, n. 12, p. 3149-3158, 2008.

You might also like