Master of Science Thesis

Noise Prediction of a NACA 0015

Airfoil with Gurney Flap
A study based on Time-Resolved Particle Image Velocimetry

Xueqing Zhang

December 22, 2015

Ad
 Noise Prediction of a NACA 0015
Airfoil with Gurney Flap
A study based on Time-Resolved Particle Image Velocimetry

Master of Science Thesis

For obtaining the degree of Master of Science in Aerospace

Engineering at Delft University of Technology

Xueqing Zhang

December 22, 2015

Faculty of Aerospace Engineering · Delft University of Technology

Delft University of Technology

Copyright c Aerospace Engineering, Delft University of Technology

All rights reserved.
 DELFT UNIVERSITY OF TECHNOLOGY
DEPARTMENT OF AERODYNAMICS

The undersigned hereby certify that they have read and recommend to the Faculty
of Aerospace Engineering for acceptance the thesis entitled “Noise Prediction of a
NACA 0015 Airfoil with Gurney Flap” by Xueqing Zhang in fulfillment of the
requirements for the degree of Master of Science.

Dated: December 22, 2015

Supervisors:
Dr. ir. Andrea Sciacchitano

Prof. dr.-Ing. Fulvio Scarano

Dr. ir. Daniele Ragni

Dr. ir. Bas W. van Oudheusden

 Preface

This thesis is dedicated to my family, whose unconditional love is never absent to back
me up, and whose own sedulous pursuits of professional excellence always encourage me
to move on. In memory of my grandfather, who set the paragon of a qualified aerospace
engineer for me to follow. I am grateful to Teng Wang, for his help in digital signal
processing and for every moment we have spent together.
Many gratitudes to Dr.Andrea Sciacchitano, for offering me the opportunity to work on
aeroacoustics and for being an excellent mentor throughout the process. It has been an
honour to work under his supervision. Si me lo permite, quisiera darle mis más sinceras
felicitaciones y mis mejores deseos para esta nueva etapa en su vida, ¡un bébé!
Furthermore, I would like to thank Dr.Stefan Pröbsting, for the patient guidance
and discussions in aeroacoustics, as well as for sharing the perceptions he harbours
for this world that offered me a viewpoint I could otherwise ignore with my eastern
cultural background. I would also like to thank all the technicians in the aerodynamics
department for their friendly help and icecreams.
To all who shared this journey, thank you! Special mentions are due to Ilke and Zhenni,
without whom my journey at TU Delft would never have begun; to Alberto, Ananth,
Bo, Bowei, Jiale, Jiggar, Koen, Linfeng, Mudit, Remco, Shengling, Shruthi, Teng, Yifei,
Ziyuan and many others with whom I had the pleasure of crossing our paths; to Xin
Liao and Xing Meng, who are far away but always near.
Xueqing Zhang (Caddie)
张 雪晴
Delft, December 2015

MSc. Thesis Xueqing Zhang

vi Preface

People who design machines and airplanes, no matter how much they be-
lieve that what they do is good, the winds of time eventually turn them into tools
of industrial civilization. They’re cursed dreams. Beautiful yet cursed dreams.
— Hayao Miyazaki

“设 计 飞 机 和 机 械 的 人， 无 论 他 们 的 意 图 是 多 么 善 良， 时 代 之 风 会 把 它 转
化为机械文明的工具，从来都不是无害的，都是被诅咒的梦想。”
— 宫崎 骏

Xueqing Zhang M.Sc. Thesis

 Abstract

Air traffic noise, especially during the aircraft take-off and landing, has been uni-
versally acknowledged as a nuisance, for which reason, the International Civil Avia-
tion Organization(ICAO) has ratified articles and established specifications aiming
at the attenuation of air traffic noise, such as the Annex 16. The implementation
of large bypass ratio turbo-fan engines on civil aircrafts and the application of de-
signs such as engine nacelle and chevron nozzle have shifted the attention of noise
control to the deployment of high-lift devices during take-off and landing. Source
control is proved to be effective in the reduction of aeronautical noise. Traditional
aeroacoustic assessments rely mainly on microphone (array) measurements and
instantaneous flow fields obtained from CFD. However, the former fails in provid-
ing the mechanism of sound generation and the latter becomes prohibitive in the
operating Reynolds number range of real aircrafts.
Previous investigations have validated the causality between the unsteady trans-
verse velocity in the wake of the high-lift devices and the far-field pressure fluctua-
tion. This thesis is a continuation of the previous study, focusing on investigating
the feasibility of noise prediction of high-lift devices based on experimental data
obtained by time-resolved PIV . A model of NACA 0015 airfoil with Gurney flap
of the height of 6% the chord length has been investigated by means of 2D time-
resolved PIV in combination with simultaneous microphone measurements. The
Gurney flap is a simplified model of the high-lift devices with more complex struc-
ture in real application. The experimental parameters were selected such that the
source of sound can be treated as compact and consequentially guaranteed the

MSc. Thesis Xueqing Zhang

viii Abstract

applicability of both the distributed and the lumped formulation of the Curle’s
aeroacoustic analogy.
In the first step, the time evolving pressure field was deduced from the velocity
field measured by 2D TR-PIV, by the Poisson-based solver for quasi-2D incom-
pressible flows. Pressure on the surfaces of the airfoil then constitutes the source
terms of the distributed formulation of Curle’s analogy and the corresponding far-
field noise was evaluated. In the second step, the aerodynamic loads of the airfoil
were evaluated from the velocity field by means of momentum balance, which were
then used in the lumped formulation for noise prediction. The spanwise coherence
length evaluated from the cross-plane flow visualizations was applied for the cor-
rection of 3D effect.
Both formulations of the Curle’s analogy yield noise evaluation of fair agreement
with the simultaneous microphone measurements. All the calculated far-field noise
power spectra reproduce the peak at Kármán vortex shedding frequency, which
agrees well with the microphone measurements. The pressure fluctuations yielded
by distributed formulation of the Curle’s analogy are somehow damped due to
the Gaussian smoothing applied to the velocity fields in the process of pressure
reconstruction for the elimination of spurious spatial derivatives. Such damping
effect can be circumvented by the lumped formulation, since the pressure is only
evaluated on control volume boundaries in the locating in the far-wake, where 3D
motions are weaker than in the region in direct proximity of the airfoil trailing
edge. The high frequency components in the computed acoustic spectra are the
most affected by the experimental and numerical errors. The coherence length
correction reduce the over-estimation of the predicted noise levels with respect to
the microphone measured ones. However, there still exists room for improvement
in the procedure of coherence length correction.
Concludingly, this thesis proved TR-PIV an effective approach in the noise pre-
diction of airfoil with Gurney flap and stepped out the first step in PIV-based
aeroacousitic investigations of the high-lift devices. Application of the approach
on high-lift devices with more complex configurations would be a prospective con-
tinuation of the current study.

Xueqing Zhang M.Sc. Thesis

 Abstract (in Chinese)

基于时间解析粒子图像测速技术(PIV)的NACA0015翼
型Gurney襟翼噪声预测

摘 要

航空噪声，尤其是在飞机起飞着陆阶段的近场噪声，随着民用飞机数量的
剧增，日益引起世界范围的关注。 为此，在国际民用航空组织（ICAO）制定了
一系列飞机噪声适航条例，《国际民用航空公约》附件16便是其中之一。 由于大
涵道比涡扇发动机在民航客机中的广泛使用以及消声短舱、 V-型花瓣喷嘴等结
构上的改进大大地降低了发动机噪声，机体噪声飞机噪声中所占比例日益增大，
其中飞机起降阶段增升装置噪声是机体噪声的重要组成部分。 飞机降噪通常采
用声源控制的方法主动降噪，此方法需要根据噪声产生机理改进增升装置结构，
然而，传统的麦克风及阵列测量无法解释发声机理，而对于飞机设计雷诺数范
围的瞬时流场计算目前仍是数值模拟无法企及的。
前人研究中已经证实了增升装置尾流中横向速度波动与远场压力扰动的
因果关系。 本文基于这一结论，对基于时间分辨PIV速度测量数据进行增升
装置噪声预测的可行性进行了分析。 本文使用Gurney襟翼作为增升装置的简
化模型，对安装有高度为6%弦长的Gurney襟翼的NACA0015翼型同时进行二
维PIV和远场麦克风测试。 通过调整麦克风安装位置使得模型可以被看作致密
声源，从而远场噪声既可以通过对机翼表面偶极子源进行积分直接求解（下称：
分布式Curle声类比）
，也可以由速度场导出的气动载荷间接获得（下称：集总

MSc. Thesis Xueqing Zhang

x Abstract (in Chinese)

式Curle声类比）
。
本文首先运用针对准二维不可压流的泊松方程由二维PIV测量的速度场导出
流场压力分布。进而由机翼表面压力波动通过分布式Curle声类比导出远场噪声。
除此之外，本文还通过动量方程解出机翼在每个瞬时的气动载荷，并由此运用
集总式Curle声类比求解远场噪声。 本文通过将两种形式的声类比导出的远场压
力扰动与麦克风测试结果对比，对两种情况的适用性进行了评估。 此外，通过
三维立体PIV在Gurney襟翼下游展向平面上的速度测量，可以导出展向相关长
度，从而对二维PIV的噪声预测结果进行修正。
在导出的压力场中可以观察到压力极小值点与涡量场中涡核位置相对应。
在实验过程中，可以清晰地听到音调噪音，且这一音调噪音所对应的频率峰值
可以在所有由速度场导出的噪声频谱中观察到，而这一峰值频率恰好对应卡门
涡街的漩涡脱落频率。 两种形式的Curle声类比预测的远场噪声都与相应位置的
麦克风测试有较好的吻合，所有由PIV预测出的噪声功率谱都再现了卡门旋涡脱
落频率位置的峰值。 为了避免流场测量误差对压力梯度带来的杂散，在求解压
力场之前首先需要对PIV测得的速度场进行高斯平滑处理，而这一过程使得近场
波动幅值衰减，从而导致分布式Curle声类比预测的远场压力扰动幅值下降。 相
比之下，由于导出气动载荷的控制体边界远离三维流动效应显著的区域，集总
式Curle声类比求解过程中避免了平滑处理的过程。 由本文中PIV导出的噪声频
谱观察到，噪声中高于音调噪音频率的高频组分受试验及数值的影响较其他组
分大。 相关长度修正减少了基于二维的噪声预测与麦克风测试之间的偏差，但
本文的校正过程仍有改进空间。
综上所述，本文验证了时间解析PIV在Gurney襟翼噪声预测上的可行性，迈
出了基于PIV的增升装置噪声研究的第一步，此方法在增升装置的噪声预测中将
具有可观的应用前景。

Xueqing Zhang M.Sc. Thesis

 Table of Contents

Preface v

Abstract vii

Abstract (in Chinese) ix

List of Figures xv

List of Tables xxi

1 Introduction 1
1.1 Noise and noise control . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Aeronautical noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 Gurney flap aerodynamics and aeroacoustics 9

2.1 Gurney flap aerodynamics . . . . . . . . . . . . . . . . . . . . . . . . 10

2.2 Aeroacoustic analogies . . . . . . . . . . . . . . . . . . . . . . . . . . 12

MSc. Thesis Xueqing Zhang

xii Table of Contents

2.3 Lighthill’s analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.4 Curle’s analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3 Data reduction techniques 27

3.1 Statistical data analysis . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.1.1 Causality-correlation . . . . . . . . . . . . . . . . . . . . . . . 28

3.1.2 Spectral density . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.1.3 Coherence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Evaluation of pressure from velocity field . . . . . . . . . . . . . . . . 31

3.3 Evaluation of aerodynamic forces from velocity field . . . . . . . . . . 35

4 Experimental techniques 39

4.1 Microphone Measurement . . . . . . . . . . . . . . . . . . . . . . . . 40

4.1.1 Microphone sensitivity . . . . . . . . . . . . . . . . . . . . . . 41

4.1.2 Microphone placement . . . . . . . . . . . . . . . . . . . . . . 42

4.1.3 Data reduction procedure . . . . . . . . . . . . . . . . . . . . 43

4.2 Particle image velocimetry . . . . . . . . . . . . . . . . . . . . . . . . 49

4.2.1 Seeding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.2.2 Illumination . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2.3 Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.2.4 Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.2.5 Image interrogation . . . . . . . . . . . . . . . . . . . . . . . 55

5 Experimental set-up 59

5.1 Facilities and models . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

Xueqing Zhang M.Sc. Thesis

Table of Contents xiii

5.1.1 Wind tunnel facility . . . . . . . . . . . . . . . . . . . . . . . 60

5.1.2 Airfoil and trailing edge models . . . . . . . . . . . . . . . . . 61

5.2 Experimental set-up for synchronized planar PIV and microphone mea-
surements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3 Experimental set-up for stereoscopic PIV measurements . . . . . . . . 68

5.4 PIV image processing . . . . . . . . . . . . . . . . . . . . . . . . . . 72

5.4.1 Planar PIV image processing . . . . . . . . . . . . . . . . . . 72

5.4.2 Stereoscopic PIV image processing . . . . . . . . . . . . . . . 73

6 Results and discussion 75

6.1 Instantaneous velocity fields and reconstructed pressure fields . . . . . 76

6.1.1 Overall flow visualization . . . . . . . . . . . . . . . . . . . . . 76

6.1.2 Pressure reconstruction . . . . . . . . . . . . . . . . . . . . . 77
6.2 Acoustic prediction based on the distributed formulation of the Curle’s
analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

6.2.1 Prediction of the far-field acoustic pressure based on velocity

measurements with FOV 2 . . . . . . . . . . . . . . . . . . . . 83
6.2.2 Prediction of the far-field acoustic pressure based on velocity
measurements with FOV 1 . . . . . . . . . . . . . . . . . . . . 86
6.3 Acoustic prediction based on the lumped formulation of the Curle’s
analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

6.3.1 Evaluation of aerodynamic forces with FOV 2 . . . . . . . . . . 89

6.3.2 Prediction of the far-field acoustic pressure with FOV 2 . . . . 90

6.4 Correction of acoustic prediction based on length of coherence . . . . . 95

6.4.1 Calculation of spanwise coherence length . . . . . . . . . . . . 95

6.4.2 Correction to the full coherence prediction . . . . . . . . . . . 99

7 Conclusions and recommendations 105

MSc. Thesis Xueqing Zhang

xiv Table of Contents

7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

7.2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

Bibliography 109

A Gaussian curve fitting for the evaluation of coherence length 117

B On the different shedding frequency obtained from the planar and

stereoscopic PIV measurements 121

Xueqing Zhang M.Sc. Thesis

 List of Figures

1.1 Drawing by Keely McGann, winning artist of the 2009 drawing contest
organized by ICAO on the theme: Aviation in a Green Environment. . . 2

1.2 Decibel - Loudness Comparison Chart [46] . . . . . . . . . . . . . . . 4

1.3 Example of AFN source localizations of a landing A340 with high-lift

device and landing gears (from flight test campaign)[62] . . . . . . . . 5

1.4 Source of noise on a typical wing[62] . . . . . . . . . . . . . . . . . . 6

1.5 Gurney flap on Rennspeed modified Porsche . . . . . . . . . . . . . . 7

1.6 Gurney flap on F1 framework . . . . . . . . . . . . . . . . . . . . . . 7

2.1 Visualization of counter-rotating vortices downstream the trailing edge

for 4% Gurney flap with α = 0◦ ,4◦ and 8◦ [64] . . . . . . . . . . . . . 11

2.2 Source-listener locations standard 3D Green’s function [27] . . . . . . . 19

2.3 Theoretical directivity patterns for far-field sound pressure levels radi-
ated from (a) monopole, (b) dipole, (c) lateral quadrupole, and (d)
longitudinal quadrupole sound sources. [59] . . . . . . . . . . . . . . . 22

2.4 Root mean square of the source terms retrieved from PIV measurements
of a 2D cavity by Koschatzky et al. (2010)[36] . . . . . . . . . . . . . 24

2.5 Power spectra of sound signal from a microphone and of sound com-
puted from PIV at the same locations with the Curle’s analogy and the
vortex sound theory by Koschatzky et al. (2010)[36] . . . . . . . . . . 24

MSc. Thesis Xueqing Zhang

xvi List of Figures

2.6 Power spectra computed using Curle’s analogy with the microphone
measurement. Predictions based on full span coherence assumption and
on the measured spanwise coherence length by Lorenzoni et al.(2012)
[43] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.1 Control volume and Control surface around the cylinder [37] . . . . . . 34

3.2 Definition of Inviscid and Viscous region on the Control surface [37] . . 34

3.3 Construction of simply-connected domain[50] . . . . . . . . . . . . . . 37

4.1 The technical chart, the dimensions and the equivalent schematic of a
Linear X-M53 microphone[31] . . . . . . . . . . . . . . . . . . . . . . 45

4.2 Electronic scheme of a condenser microphone . . . . . . . . . . . . . . 46

4.3 Directivity at different frequencies of a Linear X-M53 microphone[31] . 46

4.4 Original frequency response of a Linear X-M53 microphone[31] . . . . 47

4.5 Corrected frequency response of a Linear X-M53 microphone[31] . . . . 48

4.6 G.R.A.S. Pistonphone 42AA . . . . . . . . . . . . . . . . . . . . . . . 48

4.7 PIV working principle[57] . . . . . . . . . . . . . . . . . . . . . . . . 49

4.8 Timing diagram for PIV recording based on Double Frame/Single Ex-
posure frame straddling mode [44] . . . . . . . . . . . . . . . . . . . 55

4.9 Procedure of evaluating velocity from PIV images[16] . . . . . . . . . 55

4.10 Sub-pixel interpolation of correlation peak[15] . . . . . . . . . . . . . 57

5.1 Schematic of vertical low turbulence wind tunnel (V-tunnel) in TU Delft

[55] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

5.2 Test sections for stereoscopic PIV measurements(left) and synchronized

acoustic-planar PIV measurements(right) designed by J.Shah [63] . . . 62

5.3 NACA 0015 and Gurney flap models designed by J.Shah [63] and man-
ufactured at TU Delft . . . . . . . . . . . . . . . . . . . . . . . . . . 62

Xueqing Zhang M.Sc. Thesis

List of Figures xvii

5.4 Photograph of the experimental set-up for synchronized planar PIV and
microphone measurements . . . . . . . . . . . . . . . . . . . . . . . . 63

5.5 Schematic of set-up for synchronised planar PIV and microphone mea-
surements(side view) . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

5.6 Schematic of set-up for synchronised planar PIV and microphone mea-
surements(top view) . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

5.7 Two versions of Field of View for two-camera recording: FOV1(left)

and FOV2(right) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

5.8 Photograph of the experimental set-up for stereoscopic PIV measurements 69

5.9 Schematic of set-up for stereoscopic PIV measurements . . . . . . . . 71

5.10 Velocity field obtained by camera 1(left) and camera 2(right) . . . . . 73

5.11 Combined velocity field from two-camera measurement . . . . . . . . . 74

6.1 Non-dimensionalized instantaneous transverse velocity (u-component)

at different time instant: GF6-AOA4-V20-FOV2 . . . . . . . . . . . . 79

6.2 Non-dimensionalized instantaneous streamwise velocity (v-component)

at different time instant: GF6-AOA4-V20-FOV2 . . . . . . . . . . . . 80

6.3 Instantaneous vorticity at different time instant: GF6-AOA4-V20-FOV2 81

6.4 Reconstructed pressure (non-dimensionalised by free stream dynamic

pressure) at different time instant: GF6-AOA4-V20-FOV2 . . . . . . . 82

6.5 A snapshot of the instantaneous transverse velocity, streamwise veloc-

ity, vorticity and pressure evaluated from GF6-AOA4-V20-FOV1 . . . . 87

6.6 Power spectra (∆f = 2.4Hz) of far-field acoustic pressure evaluated by

distributed formulation of Curle’s analogy from PIV measurement (GF6-
AOA4-V20-FOV2) and (GF6-AOA4-V20-FOV2) based on full span co-
herence assumption in comparison with the simultaneous microphone
measurement at the same location . . . . . . . . . . . . . . . . . . . 88

6.7 An excerpt of time history of lift and the components of lift evaluated
from PIV measurement:GF6-AOA4-V20-FOV2 . . . . . . . . . . . . . 90

6.8 An excerpt of time history of drag and the components of drag evalu-
ated from PIV measurement:GF6-AOA4-V20-FOV2 . . . . . . . . . . 91

MSc. Thesis Xueqing Zhang

xviii List of Figures

6.9 Power spectra (∆f = 2.4Hz) of lift and drag evaluated from PIV
measurement:GF6-AOA4-V20-FOV2 . . . . . . . . . . . . . . . . . . . 92
6.10 An excerpt of time history of far-field acoustic pressure evaluated by
lumped formulation of Curle’s analogy based on full span coherence
assumption from PIV measurement (GF6-AOA4-V20-FOV2) in com-
parison with the simultaneous microphone measurement at the same
location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.11 Power spectra (∆f = 2.4Hz) of far-field acoustic pressure evaluated
by lumped formulation of Curle’s analogy based on full span coherence
assumption from PIV measurement (GF6-AOA4-V20-FOV2) in com-
parison with the simultaneous microphone measurement at the same
location on GF6 and GF0 model. . . . . . . . . . . . . . . . . . . . . 94
6.12 Stereoscopic PIV measurement: GF6-AOA4-V20, z = 15mm (First
row: mean velocity, Second row: velocity RMS, Third row: a snapshot
of instantaneous velocity, Left column: transverse u-component, Right
column: streamwise v-component) . . . . . . . . . . . . . . . . . . . 97

6.13 Coherence length of the transverse (u-component) and streamwise (v-

component) velocity calculated from stereoscopic PIV measurement. . 98

6.14 Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by dis-

tributed formulation of Curle’s analogy at the location of Microphone
1: correction based on measured spanwise coherence of transverse (u)
and streamwise (v) velocity . . . . . . . . . . . . . . . . . . . . . . . 100

6.15 Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by dis-

tributed formulation of Curle’s analogy at the location of Microphone
2: correction based on measured spanwise coherence of transverse (u)
and streamwise (v) velocity . . . . . . . . . . . . . . . . . . . . . . . 100

6.16 Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by dis-

tributed formulation of Curle’s analogy at the location of Microphone
3: correction based on measured spanwise coherence of transverse (u)
and streamwise (v) velocity . . . . . . . . . . . . . . . . . . . . . . . 101

6.17 Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by dis-

tributed formulation of Curle’s analogy at the location of Microphone
4: correction based on measured spanwise coherence of transverse (u)
and streamwise (v) velocity . . . . . . . . . . . . . . . . . . . . . . . 101

6.18 Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by lumped

formulation of Curle’s analogy at the location of Microphone 1: cor-
rection based on measured spanwise coherence of transverse (u) and
streamwise (v) velocity . . . . . . . . . . . . . . . . . . . . . . . . . . 102

Xueqing Zhang M.Sc. Thesis

List of Figures xix

6.19 Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by lumped

formulation of Curle’s analogy at the location of Microphone 2: cor-
rection based on measured spanwise coherence of transverse (u) and
streamwise (v) velocity . . . . . . . . . . . . . . . . . . . . . . . . . . 102

6.20 Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by lumped

formulation of Curle’s analogy at the location of Microphone 3: cor-
rection based on measured spanwise coherence of transverse (u) and
streamwise (v) velocity . . . . . . . . . . . . . . . . . . . . . . . . . . 103

6.21 Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by lumped

formulation of Curle’s analogy at the location of Microphone 4: cor-
rection based on measured spanwise coherence of transverse (u) and
streamwise (v) velocity . . . . . . . . . . . . . . . . . . . . . . . . . . 103

A.1 Gaussian curve fit of the spanwise coherent coefficients . . . . . . . . . 118

A.2 Extrapolated Gaussian curve and the predicted length of coherence . . 120

B.1 Comparison of the power spectra (∆f = 2.4Hz) of the transverse

velocity (u-component) measured by planar PIV and stereoscopic PIV
(GF6-AOA4-V20). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

B.2 Feedback mechanism between flow and acoustic field.[49] . . . . . . . 123

B.3 The first three resonance modes in an open cylindrical tube. The hori-
zontal axis is pressure.[73] . . . . . . . . . . . . . . . . . . . . . . . . 124

B.4 Power spectra (∆f = 2.4Hz) of the background noise of V-tunnel with
empty plexiglass-Kevlar test section and incoming flow at 20m/s. . . . 125

B.5 An example of longitudinal acoustic pressure distribution. [−0.203, 0]∪

[0.406, 0.609]: solid wall; [0, 0.406]: acoustic liner. SP L = 130dB,
M a = 0.335,f = 500Hz, Z = 0.33 − 0.98i[34] [76] . . . . . . . . . . 126

MSc. Thesis Xueqing Zhang

xx List of Figures

Xueqing Zhang M.Sc. Thesis

 List of Tables

5.1 Synchronized planar PIV and microphone measurement: experimental

parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5.2 Synchronized planar PIV and microphone measurements: experimental

devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3 Synchronized planar PIV and microphone measurements: recording pa-
rameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.4 Stereoscopic PIV measurements: recording parameters . . . . . . . . 70

6.1 Value of t − te for microphones with respect to the sources distributed

on the airfoil surface . . . . . . . . . . . . . . . . . . . . . . . . . . 83

MSc. Thesis Xueqing Zhang

xxii List of Tables

Xueqing Zhang M.Sc. Thesis

 Chapter 1

Introduction

Coughing, sputtering, roaring, plume of engine washing downwards and wings

splitting the stream of the air, and there lifting off a Leviathan, these probably
are what pop instantly into people’s mind when talking about aviation. The
cacophony played a symphony as it marked the opening of the golden era of
aviation, and it continued to roar through the 20th century, casting enchantment
that rouses awe towards the mighty metallic machine. Even nowadays, the throaty
roar across the sky can still capture the attention like nothing else can. However,
while the world is enjoying the amenity brought about by the proliferation of air
traffic, the concomitant issues, such as the noise from the operation of aircraft,
are not equally as romantic as the legend of aviation.
Early in the 1960s, the noise and disturbance caused by civil aviation has roused
serious concern with the introduction of jet aircraft as a public transport and
introduced heated social discussions. Based on the recommendations made in
series of conferences, the International Civil Aviation Organization (ICAO) called
for an international conference in Buenos Aires, September 1968, to consider
the problem of aircraft noise in the vicinity of airports and sought to establish
international specifications and guidance material relating to aircraft noise. The
new standards regarding aircraft noise eventually became effective in 1971, which

MSc. Thesis Xueqing Zhang

2 Introduction

is known as the first SARPS of Annex 16 - Aircraft Noise.[45] Contemporarily,

articles on environmental impacts, noise emission included, were added to the
airworthiness and certification process of aircraft by the Noise Control Act
of 1972 and by a 1973 amendment to the Clean Air Act [7]. In deference of
these regulations, designers and manufacturer of aviation industry nowadays are
responsible for the environmental consequences of the aviation operations and
are actively seeking for approaches of noise abatement. And thus, a fundamental
understanding of the mechanism behind the emittance of aeronautical noise is
critical for tackling the challenge.

Figure 1.1: Drawing by Keely McGann, winning artist of the 2009 drawing contest organized
by ICAO on the theme: Aviation in a Green Environment.

Xueqing Zhang M.Sc. Thesis

1.1 Noise and noise control 3

1.1 Noise and noise control

Noise is often referred to as unwanted sound, from the quiet but annoying sort such
as the barking of dog, to the loud and harmful ones, such as jet engine in close
quarter. The physics of the noise is the same as that of any other sound: in the case
of fluid (gas, plasma or liquid), the sound source creates vibration that propagates
longitudinally in the surrounding medium as sound waves. Sound waves are weak
perturbations that cause the local pressure in the medium to deviate from the
equilibrium one. This deviation is referred to as sound pressure, noted as p0 .
Loudness, commonly used to refer the auditory sensation, is a subjective measure
resulting from psychological factor and physical strength of the sound.[32] The
physical strength of the sound is often measured by sound pressure level, which
is a logarithmic measure of the sound pressure with respect to a reference value.
Denoted by SP L and measured in decibel dB, the sound pressure level is defined
as:
 
P
SP L = 20log10 (1.1)
P0

where P is the root mean square of the local sound pressure, P0 is the reference
sound pressure. In air, P0 = 20µP a, which is the threshold of human hearing.[21]
Figure 1.2 illustrates the decibel of a variety of sources, and the uneasiness and
damage caused to human hearing.
In free space, the sound pressure of a spherical sound wave decreases with the
inverse of the distance travelled, and thus the sound intensity decreases with the
inverse square, known as the ”inverse square law”. The propagation of noise is
also influenced by the convection of the medium, the obstacle presented on the
path, and the absorption effect due to viscosity. Approaches of noise abatement
therefore fall into two categories: acoustic insulation and source control. Since the
efficacy of acoustic insulation, with the installation of sound absorbing materials, is
relatively low and is often confined to small scale (such as room-acoustic), source
control is more effective and practical in the case of aviation operations. The
source control is fulfilled by adjusting the geometric or kinetic parameters such as
to altering the noise generating flow features, and thus the understanding of the

MSc. Thesis Xueqing Zhang

4 Introduction

mechanism of noise generation in specific flow fields is indispensable.

Figure 1.2: Decibel - Loudness Comparison Chart [46]

1.2 Aeronautical noise

Fixed-wing aircraft powered by jet engine is prototyped for airlines at the moment.
Two major noise sources are identified for this configuration of aircraft: the jet
engine and the airframe. Early generation of turbo-jets were major source of
aeronautical noise, because of the increasing power generated for takeoff. The
noise from the engine is positively related to exhaust velocity, the relation of which
is described by the eighth power law (Lighthill,1952). With the development and
implementation of the large by-pass ratio turbo-fan engines on the civil aircrafts,
the exhaust velocity of the engine is lowered. Additional designs such as engine
nacelles with acoustic liner and chevron nozzle further contribute to the engine

Xueqing Zhang M.Sc. Thesis

1.2 Aeronautical noise 5

noise attenuation. Nowadays, the engine noise has been markedly reduced to
the level comparable or even lower than the level of airframe noise. To reduce the
overall noise level of the aircraft, attention should be shifted to the noise generated
by airframe. The main noise generating airframe elements are:[62]

1. turbulent flow over a gap

2. flow past an airfoil with a flap

3. flow past an airfoil with a slat

4. flow past landing gear

Figure 1.3: Example of AFN source localizations of a landing A340 with high-lift device
and landing gears (from flight test campaign)[62]

MSc. Thesis Xueqing Zhang

6 Introduction

Figure 1.4: Source of noise on a typical wing[62]

1.3 Research objectives

Since the broadband noise generated by the deployment of high-lift devices

(slat and flap) of the aircraft during takeoff and landing consists the main
components of the airframe noise of aviation [58] and is universally considered
as a nuisance especially to the residents in proximity to the airport, an effective
way to attenuating the aeronautical noise is controlling the source of the sound,
which requires an insight into the mechanism of sound generated by fluid flow,
or aeroacoustics. According to the landmark paper in aeroacoustics in 1952 by
M.J.Lighthill [41], the noise generation is attributed to the fluctuations in the
unsteady fluid patterns.
Although there are already mature techniques in measuring the radiated sound,
such as the microphone array measurement, the information about the source
itself, such as the flow patterns and the pressure distribution in the source region,
are still missing pieces of the puzzle and are beyond the reach of the traditional
measurement techniques. With the development computers, lasers and cameras,
particle image velocimetry (PIV) is capable in capturing the detail unsteady
features in the flow field with high temporal and spatial resolution, and hence,
can be a powerful tool in studying the source of sound.

Xueqing Zhang M.Sc. Thesis

1.4 Thesis Outline 7

This study aims at the development and validation of a technique which predicts
the far-field noise based on the velocity field measured by the PIV system in
the vicinity of the solid body immersed in the flow, based on the measurements
acquired by time-resolved PIV and microphones. The outcome is a first step to
the source control of the airframe noise.
A Gurney flap is a lift-enhancing device of simple configuration, typically with
the shape of a small tab protruding perpendicularly into the flow. It was first
implemented on race car by Daniel Gurney to enhance the turning capability and
nowadays is utilized on wind turbines for lift increment [61]. In this preliminary
design of the implementation of PIV in flap noise measurement, the Gurney flap
serves as a simplified model of the flap on the aircraft, such as to facilitate the
model manufacturing and experimental set-up.

Figure 1.6: Gurney flap on F1 frame-

Figure 1.5: Gurney flap on Rennspeed work
modified Porsche

1.4 Thesis Outline

Chapter 2 starts with an introduction of the Gurney flap aerodynamics and then
gives an overview of the aeroacoustic analogies. In chapter 3, the mathematical
tools applied in the data reduction procedure of this thesis work are introduced,

MSc. Thesis Xueqing Zhang

8 Introduction

following which, the bridging techniques between the flow measurements and the
application of acoustic analogies are presented. Chapter 4 continues with a descrip-
tion of the experimental techniques employed in the aerodynamic and aeroacoustic
measurements in the course of the present study, before the detailed experimental
set-ups are described in chapter 5. Results are presented in chapter 6 and the
conclusions are reached in chapter 7.

Xueqing Zhang M.Sc. Thesis

 Chapter 2

Gurney flap aerodynamics and

aeroacoustics

Aerodynamics is a study of the motion of the air, particularly its interaction with
the solid body submerged in the flow, which not only enables the calculation of
forces or energy that can be harnessed, but also is essential in understanding the
methodology behind the design and construction of airborne crafts and fluid ma-
chines.
Aeroacoustics studies the mechanism of noise generation aerodynamically, either
by the turbulent motion (e.g. noise emitted by high speed air jet), or by the in-
teraction of aerodynamic forces with the surfaces (e.g. ’singing’ telephone wires).
Although the scientific theory of the noise generation by aerodynamic flows has
not been fully established, yet practical analysis that relates the hydrodynamic
flow phenomena and the wave-like propagating disturbances has been investigated
extensively for various flow conditions.
This chapter first gives an overview of the flow features and lift-enhancing mech-
anism of the Gurney flap in section 2.1. Thereafter, a general introduction to
aeroacoustic analogies is given in section 2.2 followed by detailed information on
Lighthill’s analogy and Curle’s analogy.

MSc. Thesis Xueqing Zhang

10 Gurney flap aerodynamics and aeroacoustics

2.1 Gurney flap aerodynamics

Since the first implementation on the race car by Daniel Gurney in the 1970s,
the flow properties of the Gurney flap are widely investigated in literatures.
Lieback(1978) [40] was the pioneer of investigating the aerodynamic performance
by experiments. His experiments on a Newman airfoil with 1.25% the chord length
proved the Gurney flap to be lift-enhancing. At the same time, though counter-
intuitive, the Gurney flap was found to be drag-reducing. Lieback further sug-
gested the height of the Gurney flap should be kept between 1% to 2% of the
chord length to optimize the aerodynamic benefits, while the flaps with height
exceeding 2% chord length will substantially increase the drag. A more detailed
study on the scaling of Gurney flap by Giguère concludes that the Gurney flap
height should be less than the local boundary layer thickness to avoid the increase
in drag.[52]
The efficacy of Gurney flap in improving the lift characteristics of airfoils, wings
and aircraft models has been confirmed with the measurements from various stud-
ies, such as [48], [33] and [39]. Pressure and velocity measurements have helped
shedding light on the mechanism responsible for the lift-enhancement. However,
compared to the widely observed improvements in aerodynamic performance, the
mechanism is not equally unveiled. Pressure and velocity measurements have
helped shedding light on the issue. Jeffrey et al. (2000) [33] attributed the lift-
enhancement for subsonic airfoils to two major causes:

1. The Gurney flap decelerates the flow upstream, and consequently the pressure
difference of the lower surface and upper surface of the airfoil is increased at
the trailing edge. In such a way, Gurney flap adds to the effective camber of
the airfoil.

2. The long wake downstream the Gurney flap increases the trailing edge suction
and thus delays the flow separation on the upper surface of the airfoil.

Either cause leads to the increase in airfoil circulation, which results in the increase
in lift. Notably, Jeffrey et al. suggested in the same article that the flow structure

Xueqing Zhang M.Sc. Thesis

2.1 Gurney flap aerodynamics 11

in the wake downstream the Gurney flap contains a pair of counter-rotating

vortices, which is first observed by Lieback(1978) [40].

Figure 2.1: Visualization of counter-rotating vortices downstream the trailing edge for 4%
Gurney flap with α = 0◦ ,4◦ and 8◦ [64]

MSc. Thesis Xueqing Zhang

12 Gurney flap aerodynamics and aeroacoustics

The development of time-resolved PIV (TRPIV) facilitates the detailed investi-

gation of flow patterns. Troolin et al. (2006) [64] examined the evolving flow
patterns downstream a NACA 0015 airfoil with Gurney flap with TRPIV, which
also reported the counter-rotating vortices. In this article, they further proposed
the existence of two distinct vortex shedding modes.

1. Dominant mode: this mode resembles the Kármán vortex shedding.

2. Second mode: this mode is the result of the intermittent ejection of the
circulating fluid in the cavity formed by the Gurney flap and the airfoil.

According to Troolin et al., the second shedding mode induces negative normal
velocity in the wake of the airfoil and thus contributes to the increase in circula-
tion. However, the observation of the second shedding mode has, curiously, not
been reported by other researchers until the writing of this thesis.
For further information and detailed theories about Gurney flap, one can refer to
[69], which is a compendium summarizing the aerodynamic characteristics, lift-
enhancing mechanism, and optimum design of Gurney flap, as well as the appli-
cation of Gurney flap on airfoils, wings and aircraft models ranging from subsonic
to supersonic.

2.2 Aeroacoustic analogies

Aeroacoustic analogies describe the dependence of the flow generated sound field
and its causes, the fluctuations in the flow. The notion of aeroacoustic analogy
(also mentioned as ’acoustic analogy’) was first introduced by Sir James Lighthill
in 1952, with his ’On sound generated aerodynamically’ [41], which placed the
noise from jet engine under the scientific scrutiny and later is acknowledged as the
origin of modern aeroacoustics. [38]
In several decades, aeroacoustic analogy has burgeoned into a big family. Despite
the difference in the choice of field variables, the way of interpreting source terms

Xueqing Zhang M.Sc. Thesis

2.2 Aeroacoustic analogies 13

and the mathematical complexity of numerical application, the analogies can be

rearranged in the general form of Equation 2.1.

Lf = g (2.1)

In the general form, f is the field variable used by the analogy, such as density,
velocity or enthalpy. L is the wave operator that relates the field variables in such
a way that the fluctuation of these variables propagates like waves. g is the source
term that has to be defined a priori, or the equation of aeroacoustic analogy should
be solved iteratively to increase the accuracy of the source term. In most cases,
instead of using the true acoustic source distributions, the source term is defined
by collecting from the field variable terms left over by the wave operator on the
left hand side.
The aeroacoustic analogies can be sorted into three categories based on the field
variables applied. The complexity of the application of the following analogies
grows in the order they are presented below.[65]

1. Density-based analogies

• Lighthill’s analogy
• Curle’s analogy
• Powell’s analogy
• The Ffowcs Williams-Hawkings analogy

2. Phi-based analogies (convected wave equation analogies)

• Phillip’s analogy
• Lilley’s analogy

3. Enthalpy-based analogies

• Howe’s analogy

MSc. Thesis Xueqing Zhang

14 Gurney flap aerodynamics and aeroacoustics

• Doak’s analogy

Trailing edge noise is an important topic in aeroacoustics, where sound is produced

by the interaction of unsteady flow with a sharp-edged body such as a flap. The
first investigation of trailing edge noise at low Mach flow was performed by Powell
in 1959 on a rigid flat plate moving at zero incidence [53], following which various
studies and theories were published. These theories are summarized in [30]and
are roughly categorized according to the assumptions made of the way in which
the fluctuating flow interacts with the sharp edge to produce sound. The theories
roughly fall into two categories:

1. Theories based on Lighthill’s analogy

2. Theories designed for particular model or relating the sound field to the
pressure distribution model on or near the edge of the plate

Gurney flap noise is a type of trailing edge noise and falls into the realm of the
first category, where the far-field radiations determined in terms of the turbulent
velocity field. Examples of the theories in this category are developed by Ffowcs
Williams and Hall[74], Crighton and Leppington[8] and Howe [29].
Since the present work is a first step towards the prediction of trailing edge noise
of the airfoil with Gurney flap, the Curle’s analogy, a special (and simpler) case of
Ffowcs Williams-Hawkings analogy, is chosen to educe the far-field noise from the
PIV measured velocity field for the simplicity of numerical application. Therefore
for the reason of brevity, only the physical background and working equations of
the Lighthill’s and Curles analogy are presented. Note that the Lighthill’s analogy
is the antecedent of Curle’s analogy. For information of other acoustic analogies
and for detailed mathematical derivations, one can refer to[41], [9], [17], [27], [28]
and [65].
Since both Lighthill’s analogy and Curle’s analogy are density-based aeroacoustic
analogy, the field variable f in Equation 2.1 is substituted with ρ0 , the fluctuation
of the density ρ. In the case of current project, it is equivalent to choose the
pressure perturbation (sound pressure) p0 as field variable. The argument goes as

Xueqing Zhang M.Sc. Thesis

2.3 Lighthill’s analogy 15

follows.
Apart from the problems where energy is transferred by heat conduction, such as
frictional dissipation of sound, highly non-linear events caused by the shock waves
or sound generation by heat sources(e.g. combustion), it is sufficient to consider
the flow to be homentropic.[28] Namely, the entropy s of the fluid is uniform and
constant, and thus the energy equation can be substituted by:

s = s0 = constant (2.2)

For perfect gas, the speed of sound is determined by:

s
∂p p
c0 = = γRT (2.3)
∂ρ |s=s0

where R is the gas constant, γ = Cp /Cv is the specific heats ratio and T is the
absolute temperature in K.
Under the homentropic condition, local pressure and density obey the Poisson
relation: p ∝ ργ . Combined with Equation 2.3, results in a linear relation between
small fluctuation in density and pressure:

p0 = c20 ρ0 (2.4)

In the following sections, both ρ0 and p0 are used as field variable.

2.3 Lighthill’s analogy

Lighthill’s analogy establishes an analogous relation between the acoustic ra-

diation from relatively small regions of turbulent flow embedded in an infinite
homogeneous fluid, and the sound produced and propagating in an ideal ,
stationary medium driven by a stress distribution. This relation is described with
a non-homogeneous wave equation, known as the Lighthill’s equation. The stress
distribution serves as the source term of the non-homogeneous wave equation,
denoted as Tij , namely, the Lighthill’s stress tensor.

MSc. Thesis Xueqing Zhang

16 Gurney flap aerodynamics and aeroacoustics

Derivation of Lighthill’s equation

The Lighthill’s equation is retrieved from the rearrangement of Navier-Stokes equa-
tion. The mass and momentum conservation of a compressible flow described by
Navier-Stokes equations in summation convention read:
∂ρ ∂ρvi
+ = 0, (2.5)
∂t ∂xi

∂vi ∂vi ∂(pδij − σij )

ρ + ρvj =− (2.6)
∂t ∂xj ∂xj

where σij is the (i, j)th component of the viscous stress tensor. For a Stokesian
gas, it is defined by the constitutive equation reads:
 
∂vi ∂vj 2 ∂vk
σij = µ + − δij (2.7)
∂xj ∂xi 3 ∂xk

where µ is the viscosity of the fluid. Multiplying Equation 2.5, adding the result
to Equation 2.6, the result of which upon adding and subtracting c20 ∂ρ/∂xi , yields:

∂ρvi ∂ρ ∂Tij
+ c20 =− (2.8)
∂t ∂xi ∂xi
where

Tij = ρvi vj + δij [(p − p0 ) − c20 (ρ − ρ0 )] − σij (2.9)

is the Lighthill’s stress tensor. Differentiate Equation 2.5 with respect to t, take the
divergence of Equation 2.8, and then subtract the results to obtain the Lighthill’s
equation:
 2
∂ 2 Tij

∂ 2 2 0
− c 0 ∇ ρ = (2.10)
∂t2 ∂xi ∂xj

The subscripts ”0” used in p0 and ρ0 in Equation 2.9 denotes the constant refer-
ence value, which is taken the value of corresponding property at large distances
from the sound generating turbulent region. The reason for this arrangement

Xueqing Zhang M.Sc. Thesis

2.3 Lighthill’s analogy 17

is that, at the distance sufficiently far away from the turbulent region, the flow
recovered to be homogeneous, with uniform equilibrium state: speed of sound c0 ,
density ρ0 and pressure p0 .

Approximation of Lighthill’s stress tensor

A first step towards solving the Lighthill’s equation is determining the value of Tij .
However, the pressure fluctuation term in Tij cannot be determined until Equation
2.10 is solved, which is equivalent of solving the Navier-Stokes equations governing
the flow problem and is virtually impossible. Therefore approximation should be
made according to the flow condition under investigation.
Fortunately, many cases of aeronautical interest are of high Reynolds number,
which renders the contribution of viscous term σij negligible when compared to
the Reynolds stress term ρvi vj . In addition, as stated beforehand, when heat
transfer is negligible in the flow, the flow field can be considered homentropic,
resulting in the elimination of δij [(p − p0 ) − c20 (ρ − ρ0 )].
At large distance from the turbulent region, density fluctuation should behave in
the way of acoustic wave and obeys d’Alembert equation:
 2 
∂
2
− c0 ∇ ρ 0 = 0
2 2
(2.11)
∂t
which indicates ρvi vj equals zero in the far-field. Furthermore, when Mach
number is sufficiently low, the pressure variations in the vicinity of a localized
flow are substantially uninfluenced by the compressibility, and hence ρ = ρ0 can
be approximated inside the flow.
At fairly low Mach number, there exists an intermediate region between the
localized flow and the sound field. The treatment can be the same as in the
localized flow, according to the argument given by Goldstein (1976)[17]:

...the pressure fluctuations have some of the properties of both the localized
fluctuations and those in the sound field. Thus, in this region, the pressures are as
weak as in the sound field, but the distances involved are small enough so that the
effects of finite propagation speed, and hence of compressibility, can be neglected.

MSc. Thesis Xueqing Zhang

18 Gurney flap aerodynamics and aeroacoustics

Consequently, the Lighthill’s stress tensor can be approximated as:

Tij ≈ ρ0 vi vj (2.12)

Solution to Lighthill’s equation for unbounded flow

When no solid boundaries are present, or the presence of the solid boundaries does
not influence the sound field to an appreciable extent, the solution to Equation
2.10 is the summation of the influence of sources in the dimensions of space and
time. By solving the corresponding equation expressed with Green’s function G,
the solution of which is the influence of a certain source at a given location and
time instant.
 2 
∂ 2 2
− c0 ∇ G = δ(x − y)δ(t − τ ) (2.13)
∂t2
The solution to Equation 2.13 is the standard 3D Green’s function:
r
δ(t − τ − c0
)
G(x, y, t, τ ) = (2.14)
4πr
where r = |x − y|. Equation 2.14 indicates the influence exerted by an elementary
source, positioned at y and at time τ on the listener, positioned at x and at time t.
Detailed execution of the mathematical procedures in solving such PDEs by means
of Green’s function can be found in reference[19]. The general solution given by
Lighthill to Equation 2.10 results in:
∂2
Z  
0 1 r dV
ρ (x, t) = Tij y, t − (2.15)
4πc20 ∂xi ∂xj Vy c0 r

With the approximations on Lighthill’s stress tensor (Equation 2.12), and the
relation between fluctuation in pressure and density (Equation 2.4), the solution
reduces to:
1 ∂2
Z  
0 r dV
p (x, t) = ρ0 vi vj y, t − (2.16)
4π ∂xi ∂xj Vy c0 r

Discussion
Lighthill’s equation treats the sound sources convecting with the same speed of
the flow in a way as if they were moving in the stationary medium. It incorporates

Xueqing Zhang M.Sc. Thesis

2.3 Lighthill’s analogy 19

Figure 2.2: Source-listener locations standard 3D Green’s function [27]

the sound generation, convection with the flow, propagation with different speed,
and dissipation due to conduction and viscosity.
Under the flow conditions:

1. there are no mass, heat, force or momentum source distributions

2. the perturbation of sound propagation on the hydrodynamic parameters is

small enough to be neglected

3. the spatial gradients of the speed of sound and the static hydrodynamic
parameters are, at most, of the order of the perturbation

Lighthill’s equation is an exact consequence of mass and momentum conservation.

Approximations are only introduced after the derivation of the Lighthill’s equation
to make the solution plausible. The solution given by Lighthill (Equation 2.15)
is applicable when no solid boundary present in the flow or the influence of solid
boundary is negligible, such as subsonic cold air jet. The solution displays a
quadrupole structure of the sound source, which is relatively inefficient in sound
radiation, as discussed in the next section.

MSc. Thesis Xueqing Zhang

20 Gurney flap aerodynamics and aeroacoustics

2.4 Curle’s analogy

The solution given to Equation 2.10 by Lighthill is for unbounded flow, which does
not include the influence of solid boundaries submerged in the flow, although at the
time Lighthill has remarked on the importance of solid boundary in the generation
of sound. In fact, in many cases of technological interest, solid boundaries play
a vital role in sound generation, such as high-lift devices, propellers, helicopters,
windmills and turbofans.
The presence of solid boundaries in the flow influence the sound radiation in two
ways, the mechanisms of which lie respectively in theories of classical acoustics
and aeroacoustics:

• The solid boundaries reflect and diffract the sound generated by the volume
quadrupole sources that appear in Lighthill’s solution.

• The solid boundaries support the distribution of dipole sources; in the case
where the solid boundaries are in motion, monopole sources can be generated.

The limitation was overcome by N. Curle in 1955 [9]. The solution given to the
Lighthill’s equation by Curle takes the effect of solid boundaries into account and
is later known as the Curle’s equation.

Solution to Lighthill’s equation with effect of solid boundaries

The classical formulation of Curle’s equation is for solid boundaries which are rigid,
fixed and non-permeable, and thus the interaction between fluid and structure does
not involve deformation and displacement.
Integrating Equation 2.10 by means of 3D Green’s function Equation 2.14, the
difference between the solution given by Curle and that by Lighthill lies in the
boundary condition. The general solution in the presence of solid boundaries

Xueqing Zhang M.Sc. Thesis

2.4 Curle’s analogy 21

writes:

t+ t+
∂ 2G
Z Z Z Z  
0 ∂Tij ∂G
p (x, t) = Tij dV dτ + G − Tij ni dSdτ
−∞ Vy ∂yi ∂yj −∞ ∂Vy ∂yj ∂yj
Z t+ Z 
∂ρ0

2 0 ∂G
−c0 ρ −G ni dSdτ
−∞ ∂Vy ∂yi ∂yi
(2.17)

When the surface of the solid body is rigid and stationary, the solution of far-field
pressure can be written as:

1 ∂2
Z Z
0 Tij 1 ∂ Pij
p (x, t) = dV − ni dS (2.18)
4π ∂xi ∂xj Vy r t=te 4π ∂xj δVy r t=te

where Pij = p0 δij −σij is the generalized pressure tensor formed by the combination
of pressure fluctuation and viscous stress. te is the retarded time.
The volume integral and surface integral of Equation 2.18 (respectively the first
and second term on the right hand side) are often referred to as quadrupolar and
dipolar source terms.
It can be noted that the presence of solid boundaries engenders a dipole source.
This phenomenon is caused by the prevention of pressure recovery by fluid
acceleration due to the solid boundaries. The flow in turn exerts a fluctuating
force on the solid body, in the form of time-varying surface pressure. The
fluctuating force is directly related to a dipole acoustic source and causes a higher
level of noise emission compared to the free turbulence.
Figure 2.3 shows the radiation patterns of monopole, dipole and quadrupole
sources. It can be seen that the monopole source is the most efficient for low
Mach number and is omnidirectional, while the emission of dipole and quadrupole
is directional due to mutual cancellation. Take dipole as an example: a dipole is
a separation of a sink and a source. The line connecting the two coupled source
in dipole is called the axis of dipole. Along the axis, the direction of the dipole is
defined as pointing from sink to source. The positions on the line perpendicular

MSc. Thesis Xueqing Zhang

22 Gurney flap aerodynamics and aeroacoustics

to the axis of dipole do not perceive any disturbance from the source.
As mentioned previously the fluctuating of aerodynamic force is directly related
to the dipole source. It can be further argued that, the strength and direction of a
dipole source is equivalent to the strength and direction of the force exerted by the
solid body on the external fluid. [75] It can be thus concluded that for an airfoil,
the strongest noise emission can be expected on the direction perpendicular to
the chord.

Figure 2.3: Theoretical directivity patterns for far-field sound pressure levels radiated from
(a) monopole, (b) dipole, (c) lateral quadrupole, and (d) longitudinal quadrupole sound
sources. [59]

Curle’s analogy under compact-body assumption

Simplifications can be made under certain flow conditions, and with the intro-
duction of compactness. When the characteristic dimension of the solid body
L (for airfoil, it can be chord length)is much smaller compared to the emitted
acoustic wavelength λ, the noise source is said to be acoustically compact. Since
under this condition, the acoustic wave propagates much faster than the flow car-

Xueqing Zhang M.Sc. Thesis

2.4 Curle’s analogy 23

rying the sound generating vortices. The compact relation can be written as:
L/λ << M << 1, which indicates that at low Mach number, the flow induced
sound source are more likely to be compact.
For low Mach number (M < 0.05), the compressibility effect of the flow can be
neglected. Furthermore, for chord Reynolds number (105 ), though the viscosity
plays an important role in the near wall region, the overall effect can still be ne-
glected when compared to the normal stresses. Then, the contribution of viscous
term σij in the dipolar source can also be neglected [70].In addition, under such
circumstances, the magnitude of the quadrupolar source is M 2 order lower than
the dipolar term and therefore can be neglected [17]. Thus the solution of far-field
pressure fluctuation reduces to:

p0 δij
Z
0 xj ∂
p (x, t) = − ni dS (2.19)
4πc0 |x| ∂t δVy r t=te

With the assumption of compact body, it is possible to further assume a uniform

retarded time te = t − R/c0 on the integral surface, and Equation 2.19 can be
further simplified with the surface integral of pressure substituted by the instan-
taneous aerodynamic forces exerted on the airfoil section:
Rj ∂Fj (te )
p0 (x, t) = − (2.20)
4πc0 R2 ∂t
Equation 2.20 is known as the Gutin’s principle for compact and rigid body [54],
where Rj is the j th component of the position vector originating from the airfoil
leading edge pointing to the listener and Fj is the aerodynamic force.

Discussion
Curle’s analogy is the solution to the Ligththill’s equation Equation 2.10 when
the effect of rigid, stationary solid boundaries is taken into consideration. The
presence of solid boundaries engenders a dipolar source term in the solution,
which radiates sound more efficiently than the quadrupolar source introduced by
free turbulence. Especially, at low Mach number, the contribution of quadrupolar
source can be neglected at the presence of dipolar source.
When the source satisfies the condition of compactness, with the application of

MSc. Thesis Xueqing Zhang

24 Gurney flap aerodynamics and aeroacoustics

Gutin’s principle, the term of the integral of surface pressure in the dipolar source
term can be substituted by the instantaneous aerodynamic forces. Since there are
ready techniques for evaluating or measuring fluctuating aerodynamic loads, the
simplification of Curle’s analogy not only facilitates the numerical application,
but also lowers the demand for experimental set-up.

Figure 2.4: Root mean square of the source terms retrieved from PIV measurements of a
2D cavity by Koschatzky et al. (2010)[36]

Figure 2.5: Power spectra of sound signal from a microphone and of sound computed
from PIV at the same locations with the Curle’s analogy and the vortex sound theory by
Koschatzky et al. (2010)[36]

Xueqing Zhang M.Sc. Thesis

2.4 Curle’s analogy 25

Figure 2.6: Power spectra computed using Curle’s analogy with the microphone measure-
ment. Predictions based on full span coherence assumption and on the measured spanwise
coherence length by Lorenzoni et al.(2012) [43]

Aeroacoustic predictions with Curle’s analogy based on PIV measurements have

already been carried out successfully. Haigermoser (2009) [20] derives the acoustic
emission of a cavity flow using planar TRPIV measurements in a water tunnel.
Koschatzky et al. (2010) [36] applied Curle’s analogy and vortex sound theory to
the PIV measured velocity field of an open, two-dimensional shallow cavity, where
Curle’s analogy is superior in predicting broad band noise and inferior in tonal
and harmonic component. Lorenzoni et al. (2012) [43] investigates the acoustic
emission from vortex-structure interaction with the TRPIV measurements on a
rod-airfoil configuration, where the Curle’s analogy is applied in both distributed
( Equation 2.19) and lumped ( Equation 2.20) formulation to acquire the far-field
noise.

MSc. Thesis Xueqing Zhang

26 Gurney flap aerodynamics and aeroacoustics

Xueqing Zhang M.Sc. Thesis

 Chapter 3

Data reduction techniques

The following sections offer a general introduction on the techniques for analysing
the experimental data. First, the mathematical background of statistical data
analysis methods are introduced, including the causality-correlation, the spectral
density calculation and the variance and covariance. Then, the techniques for
evaluation of pressure field and aerodynamic loads are presented, which are the
bridging techniques between the PIV measured velocity and the application of
Curle’s analogy.

3.1 Statistical data analysis

This section gives a brief introduction to the data analysis techniques applied in
the current study for analysing the relationship between near-/far-field.

MSc. Thesis Xueqing Zhang

28 Data reduction techniques

3.1.1 Causality-correlation

Cross-correlation is a standard method that detects the relationship between as-

sumed excitation and observed response. The cross-correlation coefficient mea-
sures the existence of such interrelationship and their relative strength, which is
calculated by [23]:

Cψ,p (x, τ ) hψ 0 (x, t)p0 (t + τ )i

Rψ,p (x, τ ) = =p (3.1)
σψ (x)σp hψ 0 (x, t)2 i hp0 (t)2 i

where ψ 0 (x, t) is the zero-mean part of the near-field quantity, such as velocity
fluctuation u0 and v 0 , measured at position x and time t; p0 represents the acoustic
pressure measured in the far-field; τ is the time shift between between the far-field
pressure and near-field quantity.
The cross-correlation is similar to the covariance of two functions except for the
terms in denominator, σψ (x) and σp , which are root-mean-square (RMS) of the
near and far field variables. The normalization with the RMS results in bounding
the value of cross-correlation coefficient Rψ,p (x, τ ) within [−1, 1]. The bounding
value 1 or -1 is taken when the near and far field signals display a perfect linear
relationship. Non-linear relationship or data scattering due to measurement noise
will force the value Rψ,p to zero.
It should be noted that correlation does not necessarily imply causality, while the
latter should be validated by other proof. In the current study of far-field noise
generated by an airfoil with Gurney flap, the proof of causality relation is the
Curle’s analogy described by Equation 2.18. Although Equation 3.1 is given in a
time domain, yet it is equivalent to the spatial correlation between near and far
field variables in this study, since the retarded time and the spatial distance are
related with a constant speed of sound.
Since the experimental data is acquired at discrete time instants tn , Equation 3.1
should be applied in discrete form:

N
1 X 0
Cψ,p (x, τ ) = [ψ (x, tn )p0 (tn + τ )] (3.2)
N n=1

Xueqing Zhang M.Sc. Thesis

3.1 Statistical data analysis 29

with
v
u
u1 X N
σψ (x) = t ψ 02 (x, tn ) (3.3)
N n=1

Series of synchronized PIV and microphone measurements of cylinder wake [22],

rod-airfoil configuration [23], cold jet [24] and high-lift device [25] have proved
the validity of the causality correlation method in identifying the flow structure
responsible for noise generation. Though Henning et al. (2014) [26] proposed the
ambiguity of noise source localization with the correlation technique, the relation
between near-/far-field is not vitiated.
Caution should be taken when investigating the non-linear system with cross-
correlation. In certain cases, signals strongly related by non-linear dynamics can
be blinded by cross-correlation coefficient suggesting little correlation due to the
zero quadratic moments. The phenomenon is beyond the scope of discussion of
the current study. More information about the correlation in non-linear system
can be found in [4].

3.1.2 Spectral density

Any signal that can be described as an amplitude varying in time represents a

certain physical phenomenon. When these signals are projected onto the frequency
domain and are represented in the form of power spectra, often, the underlying
properties, such as harmonics, become more evident and thus provide insight
into the mechanisms that generate the signal. The power spectrum describes
the power distribution over the different frequencies. When the signal is from
measurement of turbulent velocity field, the study of power spectrum facilitates
revealing the periodicity and dominating time-scale of the vortex structures, and
hence the revelation of the mechanism of phenomena related to the vortices.

The spectral density can be defined in three equivalent ways [3]:

MSc. Thesis Xueqing Zhang

30 Data reduction techniques

1. via correlation functions,

2. via finite Fourier transforms,

3. via filtering-squaring-averaging operations.

The discussion in this section will focus on the first two approaches.
Assume the input and output signal are mono-dimensional continuous time signals
x(τ ) and y(τ ), then the general formulation of cross-correlation function is written
as:
Z T
1 2
Rxy (τ ) = lim x(t)y(t + τ )dt (3.4)
T →∞ T − T2

when x(τ ) = y(τ ), Rxy (τ ) becomes the auto-correlation function Rxx (τ ) (or
Ryy (τ )).
The Fourier transform of the correlation function yields the Power Spectral Den-
sity:
Z
Sxy (f ) = Rxy (τ )e−2iπf τ dτ (3.5)

where Sxy (f ) is the two-sided cross-spectral density function and is an even func-
tion of frequency f . Here, the value of f can be both positive and negative.
Since the negative frequency and the positive frequency is equivalent, one-sided
cross-spectral density can be defined for f ≥ 0 as:
(
2Sxy (f ) f >0
Φxy (f ) = (3.6)
Sxy (f ) f =0

The auto-spectral density Φxx (or Φyy ) can be defined similarly. Equation 3.5 is
often referred to as the Wiener-Khinchine relations in honour of the two mathe-
maticians N. Wiener and A. I. Khinchine.
Alternatively, the spectral density can be derived based on finite Fourier trans-
form, which is better for engineering application since the measurements in real
world can only be acquired in finite time interval 0 ≤ t ≤ T . The cross-spectral
density is defined as:
1 ∗
Sxy (f, T, k) = X (f, T )Yk (f, T ) (3.7)
T k
Xueqing Zhang M.Sc. Thesis
3.2 Evaluation of pressure from velocity field 31

where Xk (f, T ), Yk (f, T ) are the Fourier transforms of the sample records xk (t),
yk (t), with Xk∗ (f, T ) the complex conjugate of Xk (f, T ).
To extrapolate Sxy (f, T, k) to the infinite time range, expected value of Sxy (f, T, k)
should be used for statistical consistency:

Sxy (f ) = lim E[Sxy (f, T, k)] (3.8)

T →∞

The definition given by Equation 3.7 and Equation 3.8 is often referred to as the
Welch’s method, named after P. D. Welch.

3.1.3 Coherence

Coherence, similar to cross-correlation, evaluates how well the two signals are
related, except that cross-correlation evaluates the phase lag of the two signals in
the time domain, while coherence evaluates the dependency of the average phase
angle for each frequency components of the two signals.
The coherence function is defined based on the cross-spectral and auto-spectral
density function, which can be used to examine the relation between two signals,
such as the investigation of the power transfer between input and output of a linear
system or estimation of the causality.
The coherence between two signals is defined as:

2 |Φxy (f )|2
γxy (f ) = (3.9)
Φxx (f )Φyy (f )

Φxy (f ), Φxx (f ) and Φyy (f ) are the cross- and auto-spectral density as defined in
subsection 3.1.2.

3.2 Evaluation of pressure from velocity field

Equation 2.19 and Equation 2.20 both require the pressure distribution on the
airfoil. The pressure is related to the velocity by the Navier-Stokes equations and

MSc. Thesis Xueqing Zhang

32 Data reduction techniques

the spatial gradients of pressure can be expressed in either the Eulerian ( Equation
3.10) or the Lagrangian formulation ( Equation 3.11) as:

 
∂u
∇p = −ρ + (u · ∇)u − ν∇2 u (3.10)
∂t

 
Du
∇p = −ρ − ν∇2 u (3.11)
Dt

For the data acquired from PIV measurements, the pressure distribution needs
to be reconstructed from the velocity field. Two strategies are available for the
computation of the pressure field:

1. direct spatial integration [2],

2. in-plane Poisson [18].

In strategy (1), the pressure gradients are integrated with spatial marching scheme
from the boundaries where Dirichlet boundary condition is assigned. The method
introduces the accumulation of random errors along the integration path and thus
a dependence of the accuracy on the path. Liu and Katz (2006) [42] have de-
veloped an omni-directional virtual boundary integration scheme to integrate the
acceleration while minimizing the effect of local random error. Strategy (2) takes
the in-plane divergence of Equation 3.10 or Equation 3.11 such that the equation
about pressure is rearranged into a Poisson equation:

∂ 2p ∂ 2p
+ = −ρfxy (3.12)
∂x2 ∂y 2

where the source term fxy is a function of velocity and can be decomposed into
the component of 2D and 3D. For high Reynolds number, the viscous term can be

Xueqing Zhang M.Sc. Thesis

3.2 Evaluation of pressure from velocity field 33

neglected [66], and thus:

Du
fxy = f2D + f3D = −ρ∇xy
" Dt  2 #
2
∂u ∂v ∂u ∂v
= +2 +
∂x ∂x ∂y ∂y
 
∂(∇xy · u) ∂w ∂u ∂w ∂v
+ + (u · ∇)(∇xy · u) + +
∂t ∂x ∂z ∂y ∂z
(3.13)

Test cases applying Equation 3.12 can be found in literatures such as [11] [12].
The different approaches (Lagrangian and Eulerian) yield different performance.
Violato et al. (2011) [68] compare the two approaches on a rod-airfoil configu-
ration and offer guidelines to the choice of experimental parameters according to
the approach adopted. According to the authors, the Lagrangian approach is less
sensitive to measurement noise. De Kat and van Oudheusden (2012) [10] fur-
ther assessed the experimental viability of the approaches in the reconstruction
of instantaneous pressure field, giving emphasis on the flow features of convecting
vortices.

MSc. Thesis Xueqing Zhang

34 Data reduction techniques

Figure 3.1: Control volume and Control surface around the cylinder [37]

Figure 3.2: Definition of Inviscid and Viscous region on the Control surface [37]

Xueqing Zhang M.Sc. Thesis

3.3 Evaluation of aerodynamic forces from velocity field 35

3.3 Evaluation of aerodynamic forces from velocity field

With the pressure field reconstructed in section 3.2, there is still one step
before the application of the lumped form Curle’s analogy ( Equation 2.20),
the evaluation of instantaneous aerodynamic forces exerted on the solid body.
This section introduces two method for evaluating aerodynamic loads from PIV
measured velocity field.

Evaluation of aerodynamic loads based on reconstructed pressure

Kurtulus et al. (2007) [37] estimate the unsteady forces on a square cylinder from
the velocity field measured by planar TRPIV. In this article, the aerodynamic
forces are evaluated by integrating the momentum balance equation within a con-
trol volume enclosing the square cylinder (Figure 3.1). The general integral form
of the momentum balance is written as:
∂ V~
ZZZ ZZ ZZ ZZ
~
F (t) = −ρ dV − ρ ~ ~
(V · ~n)V dS − p~ndS + τ̄¯~ndS (3.14)
V ∂t S S S

In this case the planar PIV measurement cannot resolve the out-of-plane velocity,
only 2D components are considered. In addition, with the outer boundary of the
control volume sufficiently far from the body surface, the viscous effect is neglected,
resulting in:
" # Z Z " ∂u # I " 2 # I " #
D ∂t
dxdy −u dy + uvdx −pdy
= −ρ ∂v
+ρ 2
+ (3.15)
L V ∂t
dxdy S −uvdy + v dx S pdx
The authors of [37] proposed a robust procedure for the evaluation of the pressure
terms on the control surface in Equation 3.15. As shown in Figure 3.2, the part
of control surface hit by the wake is defined as ’viscous region’, where pressure is
obtained by integrating Equation 3.11, but neglecting the viscous contribution.
To minimize the effect of error propagation in the process of spatial integration,
outside the wake, pressure is estimated by Bernoulli equation. In such a way,
pressure in the wake region can be resolved by integrating from both edges of
the wake, and the end point of integration, lying in the inviscid region, can be
compared with the value given by Bernoulli equation. The discrepancy between

MSc. Thesis Xueqing Zhang

36 Data reduction techniques

the integration result and the Bernoulli equation can then be used to redistribute
the discrepancy along the viscous region with a linear weighting function. The
above mentioned technique is derived and applied to the incompressible scenario,
an extension of this method to the application of compressible flow can be found
in [66], where a test-case with the supersonic flow around a bi-convex airfoil is
presented.

Evaluation of aerodynamic loads directly from velocity measurements

There are alternative methods for evaluation of time-dependent forces which cir-
cumvent the step of pressure reconstruction and depend only on the velocity mea-
surements obtained in an arbitrarily chosen region enclosing the solid body. Three
expressions for a solid body, moving or steady, immersed in an incompressible,
viscous and rotational flow are presented and compared by Noca et al.(1999)[50].
The three expressions, all derived from the integral form of momentum equation
(Equation 3.14), are referred to as the ”Impulse equation”, the ”Momentum equa-
tion” and the ”Flux equation”. According to the authors, the volume integrals
over the control volume required in the force evaluation of the ”Impulse equation”
and the ”Momentum equation” can be prohibitive as the velocity field within the
boundary layers of the solid body is not well-resolved. The ”Flux equation” cir-
cumvents the problem by adding an additional constraint of the velocity field being
divergence free (∇·u = 0). In such a way, only surface integrals are involved in the
equation. The simply-connected domain is constructed by umbilicus to transform
the volume integrals to surface integrals, as shown in Figure 3.3. When the solid
body is non-permeable and fixed, the ”Flux equation” can be written as:
I
F= n̂ · γf lux dS (3.16)
S(t)

with
1 1 1
γf lux = u2 I − uu − u(x ∧ ω) + ω(x ∧ u)
2  N − 1 N −1 
1 ∂u ∂u ∂u
− x· I −x + (N − 1) x (3.17)
N −1 ∂t ∂t ∂t
1
+ [x · (∇ · T)I − x(∇ · T)] + T
N −1
Xueqing Zhang M.Sc. Thesis
3.3 Evaluation of aerodynamic forces from velocity field 37

Figure 3.3: Construction of simply-connected domain[50]

where n̂ is the surface normal, N is the spatial dimension, ω is the vorticity, x is the
position vector, I is the unit tensor and T is the viscous stress tensor. The authors
of [50] recommended that the forces be evaluated from the data across the region
of the wake where the vorticity and viscous effects are present since it is probable
that the surface integration of the product of x and ∂u/∂t may not converge in
the far-wake. Although in 2-D measurements, the divergence-free condition may
not be satisfied for the ”Flux equation”, yet concerning the spurious divergence of
velocity field generated by the measurements in the boundary layers of the solid
body, the ”Flux equation” is still strongly preferred in such situation.

MSc. Thesis Xueqing Zhang

38 Data reduction techniques

Xueqing Zhang M.Sc. Thesis

 Chapter 4

Experimental techniques

For engineers, theories are not the central objectives, nor an end in itself,
but a means to the utilitarian end: the incarnation of the idea of people.
Quotes of statement made by a British engineer to the Royal Aeronautical So-
ciety in 1922 acknowledged the gap between scientific theories and engineering [67]:

Aeroplanes are not designed by science, but by art in spite of some pretence and
humbug to the contrary. I do not mean to suggest for one moment that engineering
can do without science, on the contrary, it stands on scientific foundations, but
there is a big gap between scientific research and the engineering product which
has to be bridged by the art of the engineer.

Experimental research is one of the art of engineer, which not only is effective in
validating the prediction made by scientific theories, but also provide information
that are out of the reach of modern theories, furthermore, can reversely produce
analytical concepts and way of thinking, as with the case of Durand and Lesley’s
idea of propulsive efficiency of a propeller in front of an obstruction.

MSc. Thesis Xueqing Zhang

40 Experimental techniques

In the history of aerodynamics, it is often most fruitful theories and experiments

are employed in interactive proximity, such as the Durand-Lesley propeller test,
the development of laminar-flow airfoil and the recognition of the relationship be-
tween dynamic response and flying qualities etc.. Because of development of more
advanced measurement devices, the magic of the art of engineers has been en-
hanced.
Upon the topic of Gurney flap aerodynamics and aeroacoustics, a various of ex-
perimental techniques have been applied for the flow visualization and acoustic
measurements. In the current study, particle image velocimetry (PIV) measure-
ments are designed to measure the velocity field and from which derive the related
field quantities as described in section 3.2 and 3.3. Microphone measurements are
used to acquire the noise spectra in the far-field, which can serve as a comparison
to the quantities derived from PIV measurement. The quantities can thus be com-
pared in the domain of both time and frequency, with the techniques described in
section 3.1.
This section gives a brief introduction on PIV and microphone measurement tech-
niques, the knowledge of which is indispensable for the design of experimental
set-up. In section 4.1, the microphone measurements and related data reduc-
tion techniques are described. In section 4.2 the history, the working principle,
the main components and the velocity evaluation techniques of PIV system are
illustrated.

4.1 Microphone Measurement

Microphone measurement is a direct way to characterizing the acoustic field. A

microphone is an acoustic-to-electric transducer that converts the sound pressure
into electrical signal. Microphones can be further classified according to the quan-
tity measured, the generation of electric signal, or the directivity. For the current
study, the LinearX-M53 microphones are applied in experiments for low noise and
low distortion measurement. Since the model is condenser microphone, the fol-
lowing introduction will be focused on condenser microphones with the example

Xueqing Zhang M.Sc. Thesis

4.1 Microphone Measurement 41

offered by Linear X-M53. For detailed information on working principle of other

types of microphones, and the recording techniques for various objectives, one can
refer to [14].
Figure 4.2 shows the electronic scheme of a typical condenser microphone. A
diaphragm and a backplate form a capacitor, the capacity of which is defined by:
A
C= (4.1)
d

where  is the dielectric constant of the substance between the two plates, d and
A are the dimensions of the capacitor.
The diaphragm is excited to vibrate by the impinging pressure fluctuation caused
by the acoustic wave, and thus the distance between the two plates d is changing
with the variation of impinging pressure on the diaphragm. When the charge stored
by the capacitor Q is kept constant, according to the capacity law, the voltage
V between the two plates responses linearly and continuously to the pressure
fluctuation:
Q Q
V = = d (4.2)
C A
The resistor functions as a high pass filter. In such a way, the output voltage
measures only the fluctuating part of the pressure signal.

The following sections expatiate on the microphone sensitivity and placement,

as well as the data reduction procedure to convert the voltage back to acoustic
pressure. These aspects are essential for succeeding in capturing sound with mi-
crophone. For measurements of good quality, there are other factors one needs to
consider, such as background noise and reflection etc. (refer to [47]), which are
beyond the discussion of this article for the concern of brevity.

4.1.1 Microphone sensitivity

Microphone sensitivity defines the ratio between the output voltage and the input
pressure signal. The sensitivity is depicted in two dimensions:

MSc. Thesis Xueqing Zhang

42 Experimental techniques

• Directional sensitivity
Directional sensitivity, or directionality, describes the microphone’s sensitiv-
ity to sound from various directions.

• Frequency sensitivity
Frequency sensitivity, or angular sensitivity, describes the amplitude response
of microphone to signal of different frequencies.

Figure 4.3 shows the directional sensitivity of Linear X-M53. It can be observed
that in audible frequency range (20 Hz to 20 kHz), the model is essentially omni-
directional. When moved to higher frequencies, the sensitivity lobe moves to the
front and the directionality becomes hypercardioid.

Figure 4.4 shows the frequency sensitivity of Linear X-M53. It can be observed
that from 50 Hz to 10 kHz, the ratio between output and input signal is roughly
constant. The reason behind this phenomenon is due to the schematic of the
condenser microphone. In the range of frequency below 50 Hz, the amplitude
of fluctuation is large due to the large amount of energy contained by the low
frequency components. The displacement of diaphragm is insufficient to measure
the amplitude of fluctuation in this frequency range, resulting in the low sensitivity.
In the frequency range above 10 kHz, the response first increases then quickly drops
because of the insufficiency of the dynamic response. The time lag introduced by
the diaphragm-condenser system causes out-of-phase cancellation.
Microphone response can be corrected to acquire a constant response with the
frequency, which requires a source in anechoic environment with known constant
amplitude and position.

4.1.2 Microphone placement

To determine the microphone placement in the experiments, several characteris-

tics should be taken into consideration, such as source directivity, near-field effect,
background noise and reflections. The scope of this study focuses on the valida-
tion of method of acoustic prediction by PIV where microphone measurements

Xueqing Zhang M.Sc. Thesis

4.1 Microphone Measurement 43

only offer a comparison of acoustic pressure at certain locations and thus does not
include the mapping of acoustic source. Therefore the major concern of micro-
phone placement lies in avoiding the near-field effect while having sufficient signal
to noise ratio.
Microphones placed in the near-field region measure the radiation sound and the
pseudo-sound at the same time, resulting in the loss of representation of the sound
propagating to the far-field. To avoid this effect, acoustic measurements are usu-
ally conducted in the far-field, in which the microphone is placed at least one
acoustic wavelength and two source dimensions away from the source.
On the basis of far-field measurement, the microphone should be placed in the
region close to the source, such that the sound level of the source is strong enough
to dominate the background noise as well as the reflections. The signal-to-noise
difference is preferred to be above 10 dB. In occasions where 10 dB is not achiev-
able, a lesser difference is tolerable as long as the interfering noise can be separated
from the data.[47]

4.1.3 Data reduction procedure

Data recorded by microphones is in voltage. To derive the sound pressure in

P a from the voltage, calibration with piston phone calibrator is required. The
G.R.A.S. Pistonphone 42AA used in this study produces a sound source of 114
dB at 250 Hz. Before the measurement, the microphone first record a 30-second
signal of the sound generated by the piston phone. The recording can be used to
derive the transfer function of the measuring system.
Equation 4.2 shows the linear relation between sound pressure input and voltage
output. Further, it can be seen in Figure 4.4 that the output-to-input ratio remains
roughly constant in range between 50 Hz and 10 kHz. Additionally, frequency out
of this range falls out of the interest of this study. The reason is that below
50 Hz, the sound is basically background noise; above 10 kHz, the amount of
energy contained by this frequency range is several orders smaller than in the
range of lower frequency. Thus, a constant output-input ratio can be assumed for
all frequencies, as in Figure 4.5.

MSc. Thesis Xueqing Zhang

44 Experimental techniques

From Equation 1.1, the root mean square of sound pressure can be calculated by:

SP L
P = P0 · 10 20 (SP L = 114dB) (4.3)

And with the root mean square of voltage V acquired from the microphone record-
ing, the output-input ratio H can be thus written as:
V
H= (4.4)
P

Xueqing Zhang M.Sc. Thesis

4.1 Microphone Measurement 45

Figure 4.1: The technical chart, the dimensions and the equivalent schematic of a Linear
X-M53 microphone[31]
MSc. Thesis Xueqing Zhang
46 Experimental techniques

Figure 4.2: Electronic scheme of a condenser microphone

Figure 4.3: Directivity at different frequencies of a Linear X-M53 microphone[31]

Xueqing Zhang M.Sc. Thesis

4.1 Microphone Measurement 47

Figure 4.4: Original frequency response of a Linear X-M53 microphone[31]

MSc. Thesis Xueqing Zhang

48 Experimental techniques

Figure 4.5: Corrected frequency response of a Linear X-M53 microphone[31]

Figure 4.6: G.R.A.S. Pistonphone 42AA

Xueqing Zhang M.Sc. Thesis

4.2 Particle image velocimetry 49

4.2 Particle image velocimetry

Particle image velocimetry (PIV) is a non-intrusive measurement technique for

research and diagnostics of flow field. The velocity measurement of PIV is based
on the measurement of the displacement of tracer particles δx carried by the flow
during a short time interval δt. The tracer particles are seeded into the flow and
illuminated by a thin light sheet or a small light volume. The light scattered by the
particles is recorded with high speed cameras onto two subsequent image frames,
from which the displacement of the particles can be determined. And thus the
local velocity of the flow can be calculated by:

δx
U= (4.5)
δt

Figure 4.7: PIV working principle[57]

The idea of PIV is derived from the laser speckle velocimetry used by several
research groups in late 1970s, and was greatly developed in the following decades.
The availability of high-speed lasers and cameras allows for variations upon the
original idea, such as:

MSc. Thesis Xueqing Zhang

50 Experimental techniques

• Standard planar PIV

Standard planar PIV measures two velocity components in a 2D plane with
a single camera.

• Stereoscopic PIV
Stereo-PIV measures three velocity components in a 2D plane with two cam-
eras.

• Tomographic PIV
Tomo-PIV measures three velocity components in a volume using two, or
more cameras.

• Time-Resolved PIV
Time-Resolved PIV benefits from the advances in CMOS camera technology
to enable the measurement with high time-resolution.

A comprehensive review on PIV techniques and related applications can be found

in [57]. The following sections will mostly focus on time-resolved planar and stereo
PIV, since they are employed in the experiments of this study.

4.2.1 Seeding

Tracer particles play a vital role in PIV. There are essentially three requirements
for the particles:

• The particles should exactly follow the motion of the fluid flow.

• The particles should not alter the local flow patterns.

• The particles should scatter light efficiently.

To choose the proper tracer particle that meets the demands, the mechanical
properties and the scattering properties of the particle should be considered.

Xueqing Zhang M.Sc. Thesis

4.2 Particle image velocimetry 51

Mechanical properties
For very small particles in low speed motion, the quasi steady viscous term, namely
the Stokes drag dominates the particle dynamics. The difference between the
particle velocity V and that of the surrounding fluid U is used to describe how
well the tracer particle can follow the flow. This difference is often referred to as
slip velocity and is defined as:

2 ρp − ρf dV
V − U = a2 (4.6)
9 µ dt

where a = 12 dp is the radius of the tracer particle, µ is the dynamic viscosity of

fluid, ρp and ρf are the density of particle and fluid respectively.
When (ρp −ρf )(ρf ) << 1, the particle is said to be neutrally buoyant. Equation 4.6
indicates that the motion of fluid is well represented by the tracer particles when
the buoyancy neutral condition is satisfied and the diameter of the particle is small.
For gas flows, (ρp −ρf )/ρf << 1 cannot be achieved. Thus typical values of density
ratio (ρp − ρf )/ρf = O(103 ) and small particle diameter (0.5µm < dp < 5µm) are
employed in gas flows.
Since the term dV/dt is difficult to characterize, the ratio between the particle time
response τp and the characteristic flow time scale τf , namely the Stokes number is
used as an indication of the tracer particle’s ability to follow the flow.
τp
St = (4.7)
τf

where
1 2 ρp − ρf
τp = d (4.8)
18 p µ
Theoretically, Stokes number should be St << 1. In gas flow, for practical
limitations, St < 0.1 is used.

Scattering properties
The scattering efficiency of the tracer particles depends on:

MSc. Thesis Xueqing Zhang

52 Experimental techniques

• ratio between the refractive index of particle and fluid np /nf

• wavelength of incoming light λ−4

• particle diameter dp

Wavelength is restricted by laser and camera sensor and thus cannot be chosen
freely. To acquire a high refractive index ratio, material of high index of refraction
is preferred. Furthermore, according to Mie’s theory, particle of large diameter is
advantageous.

It can be seen that the indications in the choice of particle diameter given by me-
chanical properties and scattering properties are contradictory. The most common
approach is to choose the largest particle that can follow the flow accurately.

4.2.2 Illumination

Due to the working principle of PIV measurements, there are several requirements
on the illumination. First, the particles should be sufficiently illuminated such that
the scattered light can be captured by the camera sensor. Second, the illumination
should be in short duration, such that the particle images do not appear as streaks.
Third, the light provided by the light source can be shaped into thin sheet that
only particles lying within can be imaged in focus.
For the requirements mentioned above, in PIV setups, lasers are often used as the
light source due to their ability to provide light beam that is:

• of high energy in short pulses,

• with low chromatic aberrations,

• collimated with approximately plane wave front.

Xueqing Zhang M.Sc. Thesis

4.2 Particle image velocimetry 53

Lasers differ in the laser material, geometrical dimensions, power consumption,

pulse energy, pulse repetition rate and wavelength of emitted light. In this
study,Nd:YLF (lasers based on neodymium-doped yttrium lithium fluoride crys-
tals) is used. The light is emitted primarily at 1053 nm and then turned into
visible green light of 526 nm by crystal frequency doubler. Currently the pulse
energy emitted by Nd:YLF laser ranges between 10 to 30 mJ, at a repetition rate
of 1 to 5 kHz.
A set of optics is used to transmit the light and to form the light sheet. Typically,
the light beam is shaped into a thin light sheet with the combination of spherical
lenses and cylindrical lenses. Either, it is a combination of cylindrical concave
lens (expanding the beam) followed by two orthogonal cylindrical convex lenses
(converging into light sheet); or, it can be a combination of a concave spherical
lens, followed by a convex spherical lens.

4.2.3 Imaging

The imaging system is characterized by the focal length f , the aperture number f#
and the magnification factor M . Focal length and magnification factor are related
to image distance di and object distance do in the thin lens approximation, by:
1 1 1 di
= + ;M = (4.9)
f di do do
The aperture number (also referred to as f-number), is defined as the ratio of focal
length f and the aperture diameter D. However, in practice, f# is determined
by the magnification factor and the thickness of the light sheet. The relation is
explained as follows.
The focal depth δz is the depth within which the particles are in focus, and is de-
scribed by the wavelength of the light source λ, the f-number and the magnification
factor as:
 2
M +1
δz = 4.88λf#2 (4.10)
M
δz should be larger or equal than the light sheet thickness ∆z, which is about 1mm
in practice.

MSc. Thesis Xueqing Zhang

54 Experimental techniques

There is a trade-off in the choice f-number, since a large f-number allows a higher
depth of field while a small f-number enhances the light intensity collected by the
sensor. This issue is far from trivial for recordings with high repetition rate where
the power per pulse decreases.
The diameter of particle image dτ depends on the geometric optics effect and the
diffraction effect:
q
dτ = (M dp )2 + (ddif f )2 (4.11)

where

ddif f = 2.44λ(1 + M )f# (4.12)

Typically, the term (M dp ) is much smaller than ddif f and hence can be neglected.
The particle image diameter is a parameter concerning the accuracy of velocity
measurements. When the particle image occupies less than one pixel, it is referred
to as peak locking, which lowers the accuracy of measurement since the particle
positions are locked to integer values and cannot be detected with subpixel accu-
racy.

4.2.4 Recording

High-speed cameras used in PIV measurements usually store the images in memory
on board during the measurement and later transfer to the computer. For the
current study, cameras with CMOS sensors are used. The CMOS cameras available
offer frame rates of over 5000 frames per second for full mega pixel resolution
(1024 × 1024 pixels). Higher frame rates can be reached when sensor is cropped.
For example, at 512×512 pixel resolution, 10,000 frames per second can be reached.
In the current study, Double Frame/Single Exposure acquisition is used. The
camera shuttering time and the laser pulses are synchronised in so-called frame
straddling mode, as shown in Figure 4.8, where Tshut is the camera shutting and
data transfer time, δt is the laser pulse separation and ∆t is the acquisition time
of an image couple.

Xueqing Zhang M.Sc. Thesis

4.2 Particle image velocimetry 55

Figure 4.8: Timing diagram for PIV recording based on Double Frame/Single Exposure
frame straddling mode [44]

4.2.5 Image interrogation

To evaluate the velocity of the flow field, the displacement of particles during the
pulse separation δt should be evaluated from the two subsequent image frames.
The steps and rules to follow are described as follows.

Figure 4.9: Procedure of evaluating velocity from PIV images[16]

MSc. Thesis Xueqing Zhang

56 Experimental techniques

1. Image windowing
Since the displacement of particles is usually not uniform on the entire region,
the image is first partitioned into small windows, inside which the displace-
ment is assumed to be uniform. To optimize between the pulse separation
and window size, there are a few rules:

• The maximum in-plane displacement should be smaller than 1/4 of the

window size.
• The maximum out-of-plane displacement should be less than 1/4 of the
light sheet thickness.
• The maximum in-plane variation of particle image displacement for par-
ticles inside the same window should not exceed the particle image di-
ameter.

2. Cross-correlation analysis
Cross-correlation is applied the image pair to determine the particle displace-
ment in each window. Let I(x, y, t) and I 0 (x, y, t + δt) the two interrogation
window at time t and time t+δt as shown in Figure 4.9. The cross-correlation
function Φ is computed in its discrete form reads:
PI PJ
i=1 j=1 IA (i, j) · IB (i + m, j + n)
Φ(m, n) = qP P (4.13)
I J 2
PI PJ 2
i=1 j=1 IA (i, j) · i=1 j=1 IB (i, j)

The m and n corresponding to the largest Φ define the displacement in integer

number of pixels.

3. Sub-pixel interpolation
The displacement in pixels defined by correlation coefficient Φ defines only
discrete pixel locations. In order to reach sub-pixel precision, the correlation
peak is interpolated, typically by means of Gaussian interpolator, around the
maximums as shown in Figure 4.10. There are additional peaks besides the
main peak on the correlation coefficient plot due to noise and correlation of
non-paired particles. Assume the value of highest peak be Φ1 , the second
highest peak Φ2 , and the correlation signal-to-noise ratio can be defined as
shown in Equation 4.14.

Xueqing Zhang M.Sc. Thesis

4.2 Particle image velocimetry 57

Figure 4.10: Sub-pixel interpolation of correlation peak[15]

Φ1
SN = (4.14)
Φ2
Usually, small interrogation window is preferred for higher vector resolution, espe-
cially in the case where fine flow patterns are involved. However, this requirement
give rise to the problem of the reduction in image density NI and hence the infor-
mation contained in one interrogation window. Another disadvantage with small
interrogation window is particles escaping the computation domain, namely Loss-
of-pairs. The above mentioned two defects both lead to rise in the uncertainty of
particle image displacement.
The Multi-pass interrogation is a solution to this issue. The method is an iterative
process, in which a relatively large interrogation window is used to produce an
initial velocity approximation; then, the approximated velocity is used to shift (or
deform) the correlation windows. In each iteration, the window size decreases and
corrections are applied to the location of correlation peak.
Peak-locking should be avoided for reliable PIV measurement, since particle image
much smaller than pixel size leads to discontinuous velocity field. Theoretically,
f# should be increased for larger particle image. However, in the case where the

MSc. Thesis Xueqing Zhang

58 Experimental techniques

pulse duration is short due to high acquisition frequency, small f# is required for
the sake of image contrast. In such case, a slight defocusing of the camera lenses
helped to avoid this effect in practice.

Xueqing Zhang M.Sc. Thesis

 Chapter 5

Experimental set-up

Given the experimental techniques presented in chapter 4, this chapter furthers

the discussion on the experimental set-up for the aeroacoustic investigation of the
Gurney flap. The experiments can be divided into two categories: (i) the syn-
chronized planar PIV and acoustic measurement, from which noise is evaluated
from the measured velocity field and then is compared with the simultaneous mi-
crophone recording, and (ii) the stereoscopic PIV measurement, which takes into
consideration of 3D effect that is beyond the reach of planar PIV measurement
and in such a way correcting the predicted noise level.
The facilities and models to fulfil the above mentioned measurements are intro-
duced in section 5.1. The detailed experimental parameters and arrangements are
then presented in the two sections that follow: section 5.2 and section 5.3. In
section 5.4, a brief description on retrieving velocity field from PIV raw images is
presented. This study is a continuation of the work of J.Shah [63]. For this reason,
the airfoil and Gurney flap model, the wind tunnel, and the experimental condi-
tions were remained the same as in the previous work. The modifications to the
previous set-up is due to the choice of PIV field of view (FOV), which is enlarged
to enable the flow visualization on both sides of the airfoil for the reason of acous-
tic predictions. In addition, the FOV of the planar PIV measurements are shifted

MSc. Thesis Xueqing Zhang

60 Experimental set-up

upstream and downstream in different rounds of measurements for the parametric

study of the optimum FOV location for acoustic related PIV measurements.

5.1 Facilities and models

All experiments of this study are carried out with the facilities and equipment
available on campus of Delft University of Technology. The following sections give
a brief introduction on the experimental facilities and models. The wind tunnel
and test sections used for this study are introduced in subsection 5.1.1, following
which, the airfoil and trailing edge models are presented in subsection 5.1.2.

5.1.1 Wind tunnel facility

The wind tunnel test was carried out at the vertical low turbulence wind tun-
nel (V-tunnel) located in the Low Speed Laboratory (LSL) at Delft University of
Technology. The V-tunnel has an open test section with a vertical inflow coming
from below, with an operating velocity range of 5 - 45 m/s. Figure 5.1 shows the
schematics of the V-tunnel.
A circular exit with a diameter of 0.6m leads the inflow to the open section. In this
study, due to the requirements of model installation and acoustic measurement, a
transition duct was installed on top of the circular exit to accommodate the square
test section sized 0.4 × 0.4m2 .
The reasons for choosing V-tunnel for this study are: (i) the low turbulence level
of the inflow (below 0.1%) due to the high contraction ratio between the settling
chamber and the open-section exit, such as to provides incoming flow with high
flow quality, and (ii) the relatively low operating noise level, which makes the tun-
nel suitable for accommodating acoustic measurements in absence of an anechoic
wind tunnel. For PIV measurement alone, the closed test section was built from
plexiglass. For synchronized PIV and acoustic measurement, two side walls of the
test section were substituted with Kevlar, which is transparent for acoustic waves

Xueqing Zhang M.Sc. Thesis

5.1 Facilities and models 61

Figure 5.1: Schematic of vertical low turbulence wind tunnel (V-tunnel) in TU Delft [55]

[13] but impermeable (at least to some degree) to air flow.

5.1.2 Airfoil and trailing edge models

A NACA 0015 airfoil with detachable Gurney flap sized 0%, 2%, 4% and 6% the
chord length were designed and manufactured for the wind tunnel experiments.
The airfoil have a chord length of 20cm and a span of 40cm. The detachable
trailing edge were made from plexiglass to allow the passage of laser light, by
which means the velocity field on both pressure and suction side of the trailing
edge can be captured simultaneously.
The 0% (or ’clean’) and 6% Gurney flap ( Figure 5.3) were chosen for this PIV-
based noise investigation by evaluating the results of the previous study given
by J.Shah. In addition, roughness elements (height around 0.841mm) made of

MSc. Thesis Xueqing Zhang

62 Experimental set-up

Figure 5.2: Test sections for stereoscopic PIV measurements(left) and synchronized
acoustic-planar PIV measurements(right) designed by J.Shah [63]

carborundum grains were added at quarter-chord to trip the boundary layer. In

such a way, at the trailing edge, on both sides of the airfoil, the boundary layer is
turbulent. Until this point, the experimental conditions remain the same as those
in reference [63].

Figure 5.3: NACA 0015 and Gurney flap models designed by J.Shah [63] and manufactured
at TU Delft

Xueqing Zhang M.Sc. Thesis

5.2 Experimental set-up for synchronized planar PIV and microphone
measurements 63

5.2 Experimental set-up for synchronized planar PIV and

microphone measurements

The schematic diagrams showing

the side view and top view of the
synchronized planar PIV and mi-
crophone measurements set-up can
be found respectively in Figure 5.5
and Figure 5.6. The measure-
ments were taken for the NACA
0015 airfoil models with Gurney
flap of height 0%c and 6%c, with
an incoming flow at velocity of
20m/s, correspondingly, the chord
Reynolds number Rec of 2.7 × 105
and Mach number M a∞ of 0.059.
The aim of the planar PIV mea-
surements is to capture the flow
patterns around and downstream
the Gurney flap with high tempo-
ral and spatial resolution, the mea-
surements of which can be used for
the reconstruction of pressure in
the flow field and the prediction of
far-field noise.
The NACA 0015 model was placed
Figure 5.4: Photograph of the experimental
vertically in the wind tunnel with
set-up for synchronized planar PIV and micro-
geometric angle of attack set to 4◦ .
phone measurements
The laser light produced by the Nd:
YLF laser was shaped in to a thin light sheet (thickness around 1mm) and hit the
location around the mid-span of the model to illuminate the field of view (FOV).
For the reason of spatial resolution, two Photron Fastcam SA1.1 cameras with

MSc. Thesis Xueqing Zhang

64 Experimental set-up

CMOS sensors, placed facing each other on both sides of the test section, were
used for image acquisition. The FOVs of the two cameras have an overlap of 25%
(see Figure 5.7), which is used to stitch the two FOVs together in the procedure
of image processing. For the concern of high temporal resolution, the sensors of
the cameras were cropped to 704 × 704 to reach an acquisition frequency at 5kHz.
The flow is seeded with water-glycol based fog with size of particles dp ≈ 1µm.
PIV images are acquired in double frame mode at the acquisition frequency of

Parameter Value Unit

Airfoil chord length, c 20 cm
Gurney flap height, h 0(0%c, or clean), 12(6%c) mm
Angle of attack, α 4 ◦

Incoming flow velocity, V∞ 20 m/s

Incoming flow chord Reynolds number, Rec 2.7 × 105 –
Incoming flow Mach number, M a∞ 0.059 –

Table 5.1: Synchronized planar PIV and microphone measurement: experimental parame-
ters

Illumination Nd: YLF laser

Seeding Water-glycol based fog
Cameras Photron Fastcam SA1.1 with CMOS sensors
Microphone Linear X-M53

Table 5.2: Synchronized planar PIV and microphone measurements: experimental devices

5kHz, with a delay between two frames of 50µs. Detailed parameters of PIV
recording can be found in Table 5.3.
Simultaneous velocity and acoustic measurements were performed since previous
works (reference [63]) found the quality of causality correlation to be poor when
velocity and acoustic measurements are not performed simultaneously. The two
types of measurements can be synchronized by recording the trigger input of the
CMOS cameras. Usually, the duration of microphone measurement is much longer
than that of the PIV measurement and thus, the trigger input can be used to crop
from microphone measurement the portion of recording that is simultaneous with

Xueqing Zhang M.Sc. Thesis

5.2 Experimental set-up for synchronized planar PIV and microphone
measurements 65

the PIV measurement.

The microphones should be placed in the far-field, which means the distance be-
tween the microphone and the source of sound (in this case can be assumed as the
airfoil trailing edge) has to be larger than the wave length of the sound as well
as two times the acoustic dimension (in this case, the chord length of the airfoil).
The former was estimated as 0.85m according to the measurements in reference
[63] and [64]. The latter is 0.4m. For structural reasons in the laboratory, the
distance between the airfoil trailing edge and the microphones was chosen as 1.3m
in practice.

Parameter Value Unit

Field of View (FOV), for each camera 80 × 80 mm2
Field of View (FOV), combined 80 × 140 mm2
Sensor size 704 × 704 pixels
Magnification factor, M 0.176 –
Focal length, f 105 mm
(both cameras)
Numerical aperture, f# 2.8 –
(both cameras)
PIV acquisition frequency, fs 5 kHz
Pulse separation, dt 50 µs
Particle image diameter,dτ 4.278(0.21)1 µm (pixels)
Digital image resolution 8.8 pixels/mm
Recording mode Double frame –
Number of images, N 5773 –
Microphone placement 1.3 m
(distance from the trailing edge)
Microphone acquisition frequency, fmic 51.2 kHz

Table 5.3: Synchronized planar PIV and microphone measurements: recording parameters

1
In experiments, cameras were slightly defocused to make the particle image larger such as to avoid
peak-locking

MSc. Thesis Xueqing Zhang

 Planar PIV set-up: Side view
1 NACA 0015 with Gurney flap 7 Microphone #2
2 Laser sheet 8 Microphone #3
6
3 Laser head 9 Microphone #4
4 Mirrors and lenses 10 Circular-to-square flow duct
5 Test section 11 Incoming flow 7
6 Microphone #1

5mm
4 8
1.3m

3 20cm

9
2
1

10
11

Figure 5.5: Schematic of set-up for synchronised planar PIV and microphone measurements(side view)
 Planar PIV set-up: Top view
1 NACA 0015 with Gurney flap 5b Test section: plexiglass
2 Laser sheet 6,7,8,9 Microphone #1,2,3,4
3 Laser head 12 Camera #1
4 Mirrors and lenses 13 Camera #2
5a Test section: Kevlar

12
5b
5a 6,7,8,9
4

3 40cm

40cm

2
1
13

Figure 5.6: Schematic of set-up for synchronised planar PIV and microphone measurements(top view)
68 Experimental set-up

1.5c
1.2c

FOV Camera #2
FOV Camera #2
FOV Camera #1
FOV Camera #1

80mm
80mm

0.8c
0.5c
80mm
80mm

Figure 5.7: Two versions of Field of View for two-camera recording: FOV1(left) and
FOV2(right)

5.3 Experimental set-up for stereoscopic PIV measure-

ments

The experimental parameters and devices for the stereoscopic PIV measurements
are identical as those applied for planar PIV measurements (refer to Table 5.1 and
5.2), except that the stereoscopic PIV measurements were not synchronised with
acoustic measurements and therefore the microphones were not used.
The aim of stereoscopic measurements is to assess the spanwise coherence of the
flow structures. The spanwise coherence can be used for the correction of 3D effect
in noise generation, since noise predictions based on planar PIV measurements are
essentially based on 2D assumptions.

Xueqing Zhang M.Sc. Thesis

5.3 Experimental set-up for stereoscopic PIV measurements 69

In the stereoscopic PIV measurements, the sheet of laser sheet was placed normal
to the free stream. The distance between the laser sheet and the Gurney flap is
15mm, which is approximately one Gurney flap height downstream the Gurney
flap. Velocity components of all three dimensions were captured, from which the
spatial derivatives in the spanwise and transverse dimension can be resolved.
The front and top views of the set-up of the stereoscopic PIV measurements are
illustrated in Figure 5.9. Detailed recording parameters of the PIV system can be
found in Table 5.4.

Figure 5.8: Photograph of the experimental set-up for stereoscopic PIV measurements

MSc. Thesis Xueqing Zhang

 Parameter Quantity Unit
Field of View (FOV), for each camera 70 × 70 mm2
Sensor size 704 × 704 pixels
Magnification factor, M 0.201 –
Laser sheet position 15 mm
(distance downstream the trailing edge)
Focal length, f 105 mm
(both cameras)
Numerical aperture, f# 2.8 –
(both cameras)
PIV acquisition frequency, fs 5 kHz
Pulse separation, dt 30, 35 µs
Recording mode Double frame –
Number of images, N 5773 –

Table 5.4: Stereoscopic PIV measurements: recording parameters

Xueqing Zhang M.Sc. Thesis

Stereo PIV set-up: Front view

Camera 1 Camera 2

15mm
Laser sheet
NACA0015 with Gurney flap
(view from pressure side)

Planar PIV set-up: Top view

FOV Suction side

Camera 1 Camera 2
Pressure side

Plexiglass

Mirrors and lenses

Laser head

Figure 5.9: Schematic of set-up for stereoscopic PIV measurements

72 Experimental set-up

5.4 PIV image processing

DAVIS 8.2 provides various pre/post-processing techniques to retrieve the velocity

field from PIV raw images. The following subsections offer a brief summary on
the image processing procedures of planar and stereoscopic PIV measurements.

5.4.1 Planar PIV image processing

Pre-processing of the raw images lies in minimum intensity subtraction from the
time series such as to remove the unwanted reflections. Sequential cross-correlation
algorithm with multi-pass iterations of decreasing interrogation window sizes (from
64 × 64, 50% overlap, 2-pass, square-shaped window to 32 × 32 to 75% overlap,
3-pass, round-shaped window) is applied for vector field calculation.
Vector post-processing is applied to remove the outliers. The spurious velocity vec-
tors are detected by means of the normalized median test that yields a ‘universal’
probability density function for the residual and hence a single threshold value can
be applied to variety of flow conditions without a priori knowledge of the flow char-
acteristics to detect spurious vectors. The method is developed by Westerweel and
Scarano in 2005 [71]. The velocity fields evaluated by DAVIS 8.2 are shown in Fig-
ure 5.10.
To obtain the combined velocity field, the two velocity fields are first trimmed to
remove the verge effect (measurements of poor quality at the verge of the camera
sensor). The trimmed velocity fields are then concatenated with MATLAB pro-
gramme, which detects the overlap region and averages the vector values given by
the two vector fields. An example of the combined velocity field is given by Figure
5.11.

Xueqing Zhang M.Sc. Thesis

5.4 PIV image processing 73

Figure 5.10: Velocity field obtained by camera 1(left) and camera 2(right)

5.4.2 Stereoscopic PIV image processing

The image processing of the stereoscopic PIV measurements basically follows the
same procedures as that of the planar PIV, except that the cross-correlation al-
gorithm applied should be stereo cross-correlation algorithm. In this study, the
interrogation window size set for multi-pass iterations for stereoscopic PIV image
processing is: 75% overlap, 2-pass, from 64 × 64 square-shaped to 16 × 16 round-
shaped.
Correct calibration is essential for PIV velocity measurements. An additional step
in calibration of stereoscopic PIV is referred to as the ’self-calibration’, developed
and validated by Wienke in 2005 [72]. Self-calibration involves computation of
a disparity map based on the real particle images taken by camera 1 and 2 at
the same time (in practice, very short pulse separation, e.g. 1µs). The cross-
correlation of the two images determines whether the calibration plate coincides
with the light sheet. The true position of the light sheet and the mapping func-
tions are then corrected according to the disparity vectors, and thus improve the
calibration of the stereoscopic PIV.

MSc. Thesis Xueqing Zhang

74 Experimental set-up

Figure 5.11: Combined velocity field from two-camera measurement

Xueqing Zhang M.Sc. Thesis

 Chapter 6

Results and discussion

This chapter presents the results of the velocity measurements, acoustic measure-
ments and the deducted far-field acoustics with the implementation of different
formulation of the Curle’s analogy. To start with, section 6.1 presented the veloc-
ity fields measured by the time-resolved planar PIV, from which, the pressure in
the flow field is evaluated by in-plane Poisson solver. On the basis of evaluated
pressure fields, in section 6.2, the far-field acoustic pressure is evaluated at each
location of the four microphones with the distributed formulation of the Curle’s
analogy. The far-filed acoustic pressure is evaluated based on the PIV measure-
ments obtained from both FOV 1 and FOV 2, and the results are compared.
Section 6.3 continues with the far-field acoustic prediction, but with the lumped
formulation of Curle’s analogy. The section starts with the evaluation of aerody-
namic loads. Thereafter, the lumped formulation is applied and the predictions are
compared to the simultaneous microphone measurements both in time series and
in power spectra. Finally, in section 6.4, the 2D acoustic predictions are corrected
with the length of coherence in order to narrow the discrepancies between the PIV
and microphone measurements.

MSc. Thesis Xueqing Zhang

76 Results and discussion

6.1 Instantaneous velocity fields and reconstructed pres-

sure fields

The velocity fields and reconstructed pressure fields presented in this section are
retrieved from the PIV measurement at free stream velocity V∞ = 20m/s and with
the field of view FOV 2 (see Figure 5.7).

6.1.1 Overall flow visualization

Previous study (reference [63], p.74) has shown that for airfoil with Gurney flap,
the velocity fluctuation and the far-field microphone measurement share the same
peak frequency on power spectra. Hence, the Kármán vortex shedding period
Tshed = 0.32ms can be evaluated from the peak frequency 314.9Hz in the power
spectra of microphone measurement (see Figure 6.11, subsection 6.3.2).
The sequence of instantaneous snapshots of the transverse velocity (u-component)
and the streamwise velocity (v-component) in one shedding period (0 to 0.32ms)
are presented respectively in Figure 6.1 and 6.2. The shadow region (as shown in
Figure 5.11), which is caused by refraction of laser light when passing through the
airfoil trailing edge, is filled with velocity values evaluated from nearest neighbour-
interpolation. Such interpolation is justifiable for the following reasons:

• According to the experimental parameters for boundary layer characteriza-

tion given in reference [63], measurement of flow in the boundary layer is
beyond the resolution of the experimental set-up of the current study.

• Outside the boundary layer, the flow is essentially irrotational and effectively
inviscid [1], and thus the relation between pressure and velocity is governed
by Bernoulli equation.

• The Gurney flap moves the stagnation point downstream [69] and the spatial
variation of pressure on the suction side is small within the last 3% of the
trailing edge. Consequently, the spatial variation in velocity is expected to
be small.

Xueqing Zhang M.Sc. Thesis

6.1 Instantaneous velocity fields and reconstructed pressure fields 77

It can be observed from Figure 6.2 the velocity deficit in the wake region down-
stream the Gurney flap. The alternation of positive and negative transverse veloc-
ity (u-component) observed in Figure 6.1 indicates the existence of large vortical
structures.
Correspondingly, the sequence of vorticity at the same time instants are shown in
Figure 6.3, where red (counter-clockwise vorticity) and blue (clockwise vorticity)
blobs indicate the counter-rotating vortices. The vortices shed at the trailing edge
are conveyed downstream at an average spacing of 5cm with a velocity around
16.7m/s, which is approximately 83% of the free stream velocity.

6.1.2 Pressure reconstruction

The in-plane pressure is reconstructed from the PIV measured velocity by Poisson
solver as described in section 3.2 (Equation 3.12 and 3.13). As for the velocity
measured by planar PIV, the spanwise velocity component is not measured and
the flow is assumed as 2D flow. Consequently, the source term is reduced to:

 2  2
∂u ∂v ∂u ∂v
fxy = +2 + (6.1)
∂x ∂x ∂y ∂y
As shown in Equation 6.1, the source term is a combination of spatial velocity gra-
dients. However, the velocity gradients retrieved from planar PIV measurement
in the region directly downstream the Gurney flap are unreliable due to the oc-
currence of high intensity 3D motion. Such velocity gradients can cause spurious
oscillations in the solution and may spoil the whole pressure field. To avoid this
problem, the velocity field is smoothed with the Gaussian kernel filter available in
Matlab R , with standard deviation σ = 10 and window size [10 10].
To facilitate the explanation about the boundary conditions, the outer boundary
of the domain is again divided into ’viscous region’ and ’inviscid region’ as in sec-
tion 3.3. The criterion applied in this case is the local value of enstrophy, defined
as:

ε = |→
−
ω |2 (6.2)

MSc. Thesis Xueqing Zhang

78 Results and discussion

In this case, the threshold for viscous region is set as ε > 1.5 × 104 [Hz 2 ].
Dirichlet boundary condition is assigned to the inviscid region of the boundary,
where pressure is calculated by Bernoulli equation:
1 1
p = P∞ + ρV∞2 − ρ(u2 + v 2 ) (6.3)
2 2
Neumann boundary condition is assigned to the airfoil surface and to the viscous
region of the outer boundary. The pressure gradient in the Neumann condition is
defined as in Equation 3.10.
The sequence of instantaneous snapshots of pressure fields corresponding to the
previously shown velocity and vorticity are presented in Figure 6.4. The pressure
downstream the Gurney flap exhibits a coherence pattern as well as a distinguish-
able convection process as that exhibits by the vorticity. The locations of the local
minima of the pressure coincide with the locations of the cores of vortices as shown
in Figure 6.3, while the local maxima indicate the stagnation region caused by flow
impinging the solid surface (upstream the Gurney flap) or the interaction of the
vortices (downstream the Gurney flap).

Xueqing Zhang M.Sc. Thesis

Figure 6.1: Non-dimensionalized instantaneous transverse velocity (u-component) at different time instant:
GF6-AOA4-V20-FOV2
Figure 6.2: Non-dimensionalized instantaneous streamwise velocity (v-component) at different time in-
stant: GF6-AOA4-V20-FOV2
Figure 6.3: Instantaneous vorticity at different time instant: GF6-AOA4-V20-FOV2
Figure 6.4: Reconstructed pressure (non-dimensionalised by free stream dynamic pressure) at different time
instant: GF6-AOA4-V20-FOV2
6.2 Acoustic prediction based on the distributed formulation of the Curle’s
analogy 83

6.2 Acoustic prediction based on the distributed formula-

tion of the Curle’s analogy

This section presents the results of acoustic prediction obtained by the imple-
mentation of the distributed formulation of the Curle’s analogy (Equation 2.19),
based on the PIV measurements performed on both FOV 2 (more information
downstream of the Gurney flap) and FOV 1 (more information upstream of the
Gurney flap). Comparison between the two aeroacoustic computations as well as
to the far-field microphone measurements are presented in power spectra of 2.4Hz
resolution.

6.2.1 Prediction of the far-field acoustic pressure based on velocity

measurements with FOV 2

As shown in Equation 2.19, the pressure fluctuation at the location of listener

(or microphone) is the sum of influence received at time instant t by the listener
and emitted at (different) retarded time instant(s) te by the sources. Therefore,
information about the retarded time for each source location is required before
summing up the influence of sources at the location of listener.
Since both the listener and the sources are fixed in this case, the value t−te = r/c0
is a constant value for a certain source with respect to a certain listener. The values
of t − te of this case are evaluated for the solid surface locations within the PIV
field of view and are summarized in Table 6.1.

Listener min.(t − te )[ms] max.(t − te )[ms] mean.(t − te )[ms] max.(∆(t − te ))[ms]

Microphone 1 4.3 4.4 4.4 0.1
Microphone 2 4.0 4.0 4.0 0
Microphone 3 3.8 3.9 3.8 0.1
Microphone 4 3.9 4.0 4.0 0.1

Table 6.1: Value of t − te for microphones with respect to the sources distributed on the
airfoil surface

MSc. Thesis Xueqing Zhang

84 Results and discussion

It can be observed that the largest difference in retarded time is 0.1ms along the
surface of airfoil. Under the sampling rate of the current study (5000Hz), the time
interval between two snapshots of velocity field is 0.2ms. For the source locations
with a different retarded time from the mean value, the treatment can either be
interpolation of the local time series of pressure to obtain the value in between the
two snapshots, or approximating with the value of the nearest snapshot in time.
With further observation of the mean value of (t − te ), it can be inferred that
source locations with a value of (t − te ) different from the mean value only account
a small portion of airfoil surface. Concerning the simplicity of application, each
microphone is assigned with the respective mean value of (t − te ) for the calcula-
tion of retarded time. The observation of retarded time inversely testifies that the
experimental set-up for the current study satisfies the compact source assumption
as described in section 2.4.
The resulting power spectra of the acoustic prediction based on the PIV mea-
surement on FOV 2 are presented in Figure 6.6 (blue curve). As comparison, the
measurements by corresponding microphones are plotted on the same schematic
(red curve).
The spectra of microphone measurements and the acoustic prediction all feature
a local peak at 315Hz in correspondence to the vortex shedding frequency, with
the drift of peak frequency no more than 10Hz. The first harmonic 630Hz can
be observed but not obvious for microphone measurements, yet it is more remark-
able on the spectra predicted from PIV measurement. The noise level at the peak
frequency is slightly over-estimated by PIV measurement for listener locations at
microphone 1, 2 and 3. The over-estimation of noise level can be attributed to the
2D flow assumption, or full spanwise coherence assumption. However, the over-
estimation in noise level is not observed for microphone 4. Thus, the full spanwise
coherence assumption is only one (though important) of the factors that may influ-
ence the prediction of noise level. In other words, the extent of over-estimation in
noise level should be larger if the full coherence assumption is the only contributing
factor. The reduce in estimated noise level can possibly be ascribed to:

• The omission of velocity/pressure fluctuations on the airfoil surface upstream

the FOV 2.

Xueqing Zhang M.Sc. Thesis

6.2 Acoustic prediction based on the distributed formulation of the Curle’s
analogy 85

• The damping of fluctuation amplitude due to smoothing of velocity field.

To reduce the influence of the above mentioned two factors, one can either move
the field of view upstream, or avoid the smoothing of velocity field by applying
the lumped formulation of Curle’s analogy. The viability of the two options is
discussed respectively in subsection 6.2.2 and section 6.3.
Less agreement between the PIV prediction and the microphone measurements
is retrieved in broadband components. The microphone spectra feature a rise in
amplitude in frequency range below 100Hz. The low frequency components in
the microphone spectra have been shown to be caused by the wind tunnel oper-
ation noise, concluded from the preliminary background measurements and the
measurements on the airfoil model without Gurney flap (Figure 6.11,green curve).
Self-evidently, the background noise of wind tunnel operation cannot be revealed
by acoustic computation based on velocity measurements.
The sharp decay in magnitude observed at 1445Hz in Figure 6.6 has no physi-
cal meaning. The artificial magnitude reduction is mainly due to the numerical
smoothing of the signal in the calculation of the time derivatives. Comparison
of the effect of the discrete method used for time derivatives was presented by
Lorenzoni (2008)[44], suggesting that the central scheme (which is applied in the
present study) is more effective in magnitude reduction since a wider kernel is used
for the computation of the differences.
The amplitude of computed acoustic spectra does not decay in the high frequency
range as in the spectra of microphone measurements since the high frequency com-
ponents are more susceptible to experimental and numerical noise which can be
attributed to:

• The effect of 3D motion with high intensity directly downstream the Gurney
flap.

• Inaccuracy of the experimental data.

• Energy fold-back due to finite sampling frequency.

MSc. Thesis Xueqing Zhang

86 Results and discussion

6.2.2 Prediction of the far-field acoustic pressure based on velocity

measurements with FOV 1

As mentioned in subsection 6.2.1, the omission of velocity/pressure fluctuations

upstream the FOV 2 may influence the predicted noise level. This subsection
presents the effect on noise prediction when moving the PIV field of view more
upstream.
The velocity field measured by PIV on FOV 1 is again smoothed with the Gaus-
sian kernel filter as for the case of FOV 2 , with standard deviation σ = 10 and
window size [10 10]. However, in this case, the threshold for viscous region is set as
ε > 2 × 106 Hz 2 . The reason for increasing the threshold is to guarantee sufficient
Dirichlet boundary such as to prevent the numerical instabilities of the system.
Otherwise, since the downstream edge of the physical domain is too close to the
Gurney flap, the local enstrophy is too high to allow any Dirichlet boundary con-
dition on this edge. Example snapshots of the velocity, vorticity and reconstructed
pressure at the same time instant are presented in Figure 6.5.
The resulting power spectra of the acoustic prediction based on the PIV measure-
ment on FOV 1 are presented in Figure 6.6 (green curve). Instead of yielding a
higher noise level than the prediction made from FOV 2, the noise level predicted
from FOV 1 is lower. This reduction in amplitude of pressure fluctuation can be
attributed to the deliberate increase in the threshold of ’viscous region’. The extra
Dirichlet boundary condition assigned to the outer boundary while stabilizes the
solution, also exerts damping effect on the fluctuation.
It can be observed that moving the PIV field of view upstream while maintaining
the spatial resolution (and hence the size of the field of view) is not preferable for
pressure reconstruction, since the downstream edge the physical domain is sub-
merged in the near-wake of the Gurney flap, the condition of which incommodes
the choice of boundary conditions and exacerbates the stability of Poisson solver.

Xueqing Zhang M.Sc. Thesis

Figure 6.5: A snapshot of the instantaneous transverse velocity, streamwise velocity, vorticity and pressure
evaluated from GF6-AOA4-V20-FOV1
Figure 6.6: Power spectra (∆f = 2.4Hz) of far-field acoustic pressure evaluated by distributed formulation
of Curle’s analogy from PIV measurement (GF6-AOA4-V20-FOV2) and (GF6-AOA4-V20-FOV2) based on
full span coherence assumption in comparison with the simultaneous microphone measurement at the same
location
6.3 Acoustic prediction based on the lumped formulation of the Curle’s analogy
89

6.3 Acoustic prediction based on the lumped formulation

of the Curle’s analogy

As discussed in section 6.2, due to the damping effect of velocity fields smoothing
as treatments for PIV measurements noise and the local 3D motions, the far-
field noise level is unavoidably under-estimated by the distributed formulation of
Curle’s acoustic analogy with the pressure field reconstructed by Poisson solver.
To circumvent the above-mentioned issues, given that the source of sound has been
proved to be compact under the experimental set-up of this study, in this section,
the acoustic prediction results from the implementation of the lumped formulation
of the Curle’s analogy (Equation 2.20) are presented.

6.3.1 Evaluation of aerodynamic forces with FOV 2

The first step towards the application of Equation 2.20 is the evaluation of
aerodynamic loads of the airfoil. The velocity measurement on FOV 2 is chosen
since the outer boundaries of the control volume should preferably be away from
the airfoil surface to avoid the influence of local 3D motions on the reconstructed
pressure.
Lift and drag of the airfoil model are estimated with the momentum balance
equation (Equation 3.15). The pressure in the momentum balance is evaluated by
the method proposed by Kurtulus et al. (2007) as described in section 3.3. In this
case, the threshold for viscous region is set as ε > 1.5 × 104 Hz 2 as in subsection
6.2.1.
The excerpts of lift and drag in time series are plotted respectively in Figure 6.7
and 6.8, with the aerodynamic loads plotted in black and the first, second and
third term in Equation 3.15 plotted respectively in green, blue and red. The plots
also show the mean value of the aerodynamic loads, plotted in magenta.

Periodicity can be observed in the reconstructed aerodynamic loads and the com-
ponents. For lift, the fluctuations are majorly attributed to the volume integral

MSc. Thesis Xueqing Zhang

90 Results and discussion

Figure 6.7: An excerpt of time history of lift and the components of lift evaluated from
PIV measurement:GF6-AOA4-V20-FOV2

of the time derivatives of the velocity and the surface integral of the convection,
while for drag, the major components of fluctuations are the two surface integral
terms. However, it would be arbitrary to neglect any of the three components in
both lift and drag, when looking closely at the order of magnitude and considering
their contribution to the phase of the overall fluctuation.
The resulting power spectra of the aerodynamic loads evaluated from the PIV
measurement on FOV 2 are presented in Figure 6.9. Both the lift and the drag
peak at the frequency of 314.9Hz, which is consistent with the shedding frequency.

6.3.2 Prediction of the far-field acoustic pressure with FOV 2

Based on the aerodynamic loads evaluated from the PIV measurement, Equation
2.20 is implemented to reconstruct the far-field acoustic pressure. The distance

Xueqing Zhang M.Sc. Thesis

6.3 Acoustic prediction based on the lumped formulation of the Curle’s analogy
91

Figure 6.8: An excerpt of time history of drag and the components of drag evaluated from
PIV measurement:GF6-AOA4-V20-FOV2

from the airfoil trailing edge and the certain microphone is used to calculate the
uniform retarded time for that specified microphone. The values of t − te of the
microphones coincide with the average values listed in Table 6.1.
The excerpts of the far-field acoustic pressure in time series, both the micro-
phone measurements and the reconstruction from PIV measurements based on
full spanwise coherence assumption are shown in Figure 6.10. The two measure-
ments roughly share the same period and phase. However, due to the many stages
in between, such as the smoothing of velocity fields, the computation of time and
spatial derivatives and the acoustic analogy, the maximum coefficients obtained
from the cross-correlation of the far-field pressure measured by microphone and
that evaluated by PIV is barely 0.19.
As expected, the amplitude predicted by PIV measurements is larger than that
measured by the microphones at all four locations of microphones, since in reality
the flow is intrinsically 3D and the spanwise coherence length is shorter than the
span.

MSc. Thesis Xueqing Zhang

92 Results and discussion

Figure 6.9: Power spectra (∆f = 2.4Hz) of lift and drag evaluated from PIV
measurement:GF6-AOA4-V20-FOV2

Figure 6.11 shows the power spectra of the far-field acoustic pressure. The blue
curve and the red curve show respectively the predictions of PIV and the mea-
surements of microphones, and both peak at frequency around 315Hz, which has
been observed in the power spectra of the prediction given by distributed form
of Curle’s analogy (Figure 6.6) and is consistent with the peak frequency of aero-
dynamic loads. The latter consistency is expected, since at the Mach number of
the current study (M a∞ = 0.059), the Doppler effect is negligible, the frequency
reaching the receiver should be equal to the frequency emitted by the source.
The extra green curve in Figure 6.11 offers a comparison of the power spectra
of the clean NACA 0015 (microphone recordings performed at same locations),
which consists of only broadband components. The absence of frequency peak on
the clean NACA 0015 power spectra serves as a side proof that the tonal noise
generated by the model with Gurney flap should be attributed to the presence of
the Gurney flap.

Xueqing Zhang M.Sc. Thesis

Figure 6.10: An excerpt of time history of far-field acoustic pressure evaluated by lumped formulation of
Curle’s analogy based on full span coherence assumption from PIV measurement (GF6-AOA4-V20-FOV2)
in comparison with the simultaneous microphone measurement at the same location
Figure 6.11: Power spectra (∆f = 2.4Hz) of far-field acoustic pressure evaluated by lumped formulation
of Curle’s analogy based on full span coherence assumption from PIV measurement (GF6-AOA4-V20-FOV2)
in comparison with the simultaneous microphone measurement at the same location on GF6 and GF0 model.
6.4 Correction of acoustic prediction based on length of coherence 95

6.4 Correction of acoustic prediction based on length of

coherence

It can be observed from Figure 6.6 and Figure 6.11 that, based on the full span-
wise coherence assumption, where the intensity and phase of the fluctuations are
presumably the same along the span, the PIV measurements over-estimate the
far-field acoustic pressure. In fact, it is reasonable to assume the intensity of the
fluctuation to be uniform along the span except for the region close to the test
section walls where horse shoe vortex develops. However, the assumption of uni-
form phase angle along the span could be the dominating factor leading to the
overprediction as much as 20dB in the frequency range of lumped formulation of
the Curle’s analogy.
Since the planar PIV measurements only offers the intensity and phase information
at the location of the thin light sheet, assumptions should be made on the pressure
fluctuations out of the measuring plane. The coherence function (as introduced in
subsection 3.1.3) provides a good measure of the dependency of the average phase
angle at two locations. An equivalent coherence length can be defined by choos-
ing a critical value for the coherence function. When the value of the coherence
function is above the chosen critical value, the fluctuation can be regarded as in
the same phase, and otherwise, the phase angle is completely independent.
The method is originally proposed by Kato et al.[35] and is proven to work rea-
sonably well by literatures such as [43], [51] and [56].

6.4.1 Calculation of spanwise coherence length

The measurements by the stereoscopic PIV set-up(see section 5.3) are used for the
assessment of the spanwise length of coherence. The mean velocity, RMS of the
velocity fluctuation and a snapshot of the instantaneous cross-plane velocity fields
of u-component and v-component are shown in Figure 6.12.
The length of coherence is evaluated for both transverse (u) and streamwise (v)
components. The spanwise coherence was calculated along the line parallel to the
airfoil trailing edge, where the maximum RMS of velocity fluctuation locates. In

MSc. Thesis Xueqing Zhang

96 Results and discussion

Figure 6.12, the trailing edge of the airfoil locates at x = 0, while the lines for
calculation of coherence are marked with grey dash dots.
For the calculation of the coherence of fluctuations at two spatial locations, Equa-
tion 3.9 is adapted as:
| hû0 (z0 , f )û0∗ (z, f )i |2
γ 2 (z, f ) = (6.4)
| hû0 (z0 , f )û0∗ (z0 , f )i hû0 (z, f )û0∗ (z, f )i |
In this case, z0 is chosen as the mid-span location.
The equivalent coherence length is dependent on the frequency f . In the current
study, a Gaussian function is fitted for all the frequency components around the
spanwise location z = z0 (for detailed information, see Appendix A). The critical
value of coherence function is chosen as γc = e−2 . The coherence length l is thus
defined as the width of the coherence curve at the critical value.
As reported by literatures such as [35] and [43], the equivalent length indicates
high values only at the frequency associated with large vortex shedding phenom-
ena (in this case, the shedding frequency 314.9Hz) and should drops drastically
below or above the tone. However, for some reasons, the coherence length peaks at
271Hz based on the stereoscopic measurements of this study. For the reason that
the coherence length is later applied for the correction of the planar PIV acoustic
prediction, it is rescaled such that the peak coherence length locates at 314.9Hz.
The length of coherence is plotted with respect to frequency in Figure 6.13.
The occurrence of 271Hz frequency in the coherence length is incongruous when
compared to the shedding frequency measured by planar PIV, however, it is con-
sistent with the stereoscopic PIV measurement, since under the set-up of stereo-
scopic PIV, the shedding frequency is measured as 271Hz. The experimental data
obtained by J.Shah[63], who applied similar experimental set-up, also reports dis-
crepancy in shedding frequencies measured by planar and stereoscopic PIV set-ups.
The phenomenon can be a result of change in test section material in two set-ups.
The discussion on the discrepancy of shedding frequency is dilated in Appendix B.

Xueqing Zhang M.Sc. Thesis

Figure 6.12: Stereoscopic PIV measurement: GF6-AOA4-V20, z = 15mm (First row: mean velocity,
Second row: velocity RMS, Third row: a snapshot of instantaneous velocity, Left column: transverse
u-component, Right column: streamwise v-component)
Figure 6.13: Coherence length of the transverse (u-component) and streamwise (v-component) velocity
calculated from stereoscopic PIV measurement.
6.4 Correction of acoustic prediction based on length of coherence 99

6.4.2 Correction to the full coherence prediction

The values of the coherence length are used for the corrections of the 2D predictions
of the far-field acoustic pressure, which corresponds to adding to each specified
frequency a correction value equals to:
 
L
−10log10 (6.5)
l(f )

where L is the full span length, and l(f ) is the coherence length at the specified
frequency f .
The corrections of both the coherence length calculated from u-component(green
prisms) and that from v-components(black diamonds) are applied. Figure 6.14 to
6.17 illustrate the results of corrections on the acoustic prediction from distributed
formulation of Curle’s analogy, and figure 6.18 to 6.21 show the results of correc-
tions on the acoustic prediction from lumped formulation of Curle’s analogy.
It can be observed that for both formulations of the Curle’s analogy, the coherence
length correction improves the acoustic prediction in the right tendency, that is
to decrease the amplitude. The corrected predictions by distributed formulation
yield amplitudes lower than the microphone measurements, while for the predic-
tion by lumped formulation, the over-estimation is not completely removed by the
correction.
However, the results of coherence length correction presented in this section is
only illustrative of the viability of this method. Within the scope of current study,
the effectiveness of the coherence length correction cannot be evaluated, since
the above-mentioned shift in shedding frequency observed in the stereoscopic PIV
measurements has rendered the evaluation of coherence length questionable.

MSc. Thesis Xueqing Zhang

100 Results and discussion

Figure 6.14: Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by distributed for-
mulation of Curle’s analogy at the location of Microphone 1: correction based on measured
spanwise coherence of transverse (u) and streamwise (v) velocity

Figure 6.15: Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by distributed for-
mulation of Curle’s analogy at the location of Microphone 2: correction based on measured
spanwise coherence of transverse (u) and streamwise (v) velocity

Xueqing Zhang M.Sc. Thesis

6.4 Correction of acoustic prediction based on length of coherence 101

Figure 6.16: Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by distributed for-
mulation of Curle’s analogy at the location of Microphone 3: correction based on measured
spanwise coherence of transverse (u) and streamwise (v) velocity

Figure 6.17: Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by distributed for-
mulation of Curle’s analogy at the location of Microphone 4: correction based on measured
spanwise coherence of transverse (u) and streamwise (v) velocity

MSc. Thesis Xueqing Zhang

102 Results and discussion

Figure 6.18: Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by lumped for-
mulation of Curle’s analogy at the location of Microphone 1: correction based on measured
spanwise coherence of transverse (u) and streamwise (v) velocity

Figure 6.19: Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by lumped for-
mulation of Curle’s analogy at the location of Microphone 2: correction based on measured
spanwise coherence of transverse (u) and streamwise (v) velocity

Xueqing Zhang M.Sc. Thesis

6.4 Correction of acoustic prediction based on length of coherence 103

Figure 6.20: Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by lumped for-
mulation of Curle’s analogy at the location of Microphone 3: correction based on measured
spanwise coherence of transverse (u) and streamwise (v) velocity

Figure 6.21: Power spectra (∆f = 2.4Hz) of acoustic pressure predicted by lumped for-
mulation of Curle’s analogy at the location of Microphone 4: correction based on measured
spanwise coherence of transverse (u) and streamwise (v) velocity

MSc. Thesis Xueqing Zhang

104 Results and discussion

Xueqing Zhang M.Sc. Thesis

 Chapter 7

Conclusions and recommendations

An aeroacoustic investigation on NACA 0015 with Gurney flap has been performed
based on the measurements of time-resolved PIV. This chapter first highlights the
main findings of this study in section 7.1, following which, the recommendations
for the future investigations are given in section 7.2.

7.1 Conclusions

The main findings from the literature study, the experimental campaigns and the
analysis of the present work, as listed below:

1. The Kármán vortex shedding is the dominating vortex shedding mode of the
Gurney flap. The interaction between the vortices and solid boundaries forms
dipolar sources resided on the solid surface, which is the major source of the
noise generated by trailing edge with Gurney flap. The frequency of the tonal
noise is consistent with the frequency of vortex shedding. The power spectra
of the fluctuations of velocity, aerodynamic loads and far-field pressure did
not indicate the existence of the second shedding mode for this configuration

MSc. Thesis Xueqing Zhang

106 Conclusions and recommendations

as the one proposed by Troolin et al.[64]. (chapter 2, chapter 6)

2. The in-plane pressure was reconstructed from PIV measured velocity with the
in-plane Poisson solver proposed by Gurka et al.[18] for both the measure-
ments on FOV 1 and FOV 2. Coherence pattern and distinguishable con-
vection can be observed downstream the Gurney flap in the reconstructed
pressure field. In addition the locations of the local minima of the recon-
structed pressure are consistent with the locations of vortex cores identified
in the vorticity field. (chapter 6: section 6.1)

3. The far-field acoustic pressure predicted by the distributed formulation of the

Curle’s analogy based on the in-plane pressure reconstruction yielded results
that were slightly over-estimated but in fair agreement with the simultane-
ous microphone measurements. However, the noise level predicted under the
full span coherence assumption was lower than expected. The phenomenon
can be attributed to the damping of fluctuations because of the Gaussian
smoothing. For the case of FOV 1, the deliberate rise in enstrophy thresh-
old for more Dirichlet boundary conditions also contributed to the damping
effect.(chapter 6: section 6.2)

4. The noise level predicted by the lumped formulation is higher than that of
the distributed formulation, and at all four microphone locations, the PIV
measurements yielded over-estimation in noise level when compared to the
simultaneous microphone measurements. Such over-estimation was expected
under the assumption of full span coherence. The lumped formulation of the
Curle’s analogy was implemented to avoid the smoothing of velocity field, and
therefore avoiding the concomitant damping of the near-field fluctuations.
(chapter 6: section 6.2, 6.3)

5. The over-estimated noise level under the full spanwise coherence assumption
was corrected with the method of equivalent coherence length, proposed by
Kato et al.[35]. The correction reduced the noise level on the power spectra
for all frequencies. The correction was in the right direction, but the effec-
tiveness cannot yet be estimated in this study for the reason of the observed
inconsistency in shedding frequencies measured by planar PIV and stereo-
scopic PIV. The values of coherence length were rescaled before applied to

Xueqing Zhang M.Sc. Thesis

7.2 Recommendations 107

the noise level correction. (chapter 6:section 6.4)

7.2 Recommendations

The present study proved the applicability and revealed the limitation of the PIV
techniques in the acoustic investigations of an airfoil with Gurney flap. Hereby
are recommendations on the further researches on the topic:

• Pressure reconstruction based on PIV measurements

For planar PIV measurements, before the implementation of the in-plane
Poisson solver proposed by Gurka et al.(Equation 3.12), it is preferable that
the velocity fields first be smoothed to remove the spurious spatial derivatives
caused by experimental noise. And for the planar PIV set-up, it is advisable
that the outer boundaries of the field of view should avoid the region with
high vortex intensity and strong 3D motions. An alternative solution can be
the use of tomographic PIV, which allows 3D flow visualization. In such a
way, the 3D components in the source term of Equation 3.13 can be resolved.

• Evaluation of far-field acoustic pressure

Under the experimental condition of planar PIV, lumped formulation of the
Curle’s analogy is preferable as it avoids resolving the pressure fluctuations
on the solid surfaces. Although the far-field pressure fluctuation can be ob-
tained with both formulations of the Curle’s analogy, yet with the effect of
the unresolved 3D motions which dominates the near-wake region, the fluctu-
ations of pressure on the solid surface are damped in the process of in-plane
pressure reconstruction.

• Evaluation of the aerodynamic loads from PIV measurements

For the evaluation of aerodynamic loads for the application of the lumped
formulation of the Curle’s analogy, this study used the method of momen-
tum balance proposed by Kurtulus et al.(Equation 3.14), which involves the
pressure evaluation on the integral surfaces. Alternatively, the aerodynamic

MSc. Thesis Xueqing Zhang

108 Conclusions and recommendations

forces can be computed directly from the velocity fields with the ”flux equa-
tion” proposed by Noca et al. (Equation 3.16), circumventing completely the
pressure reconstruction. Simultaneous PIV and balance (with high frequency
response) measurements is a continuation of the current study, from which
the accuracy of the two approaches can be evaluated.

• Materials of the test sections

The materials of the test sections used in planar PIV and stereoscopic PIV
measurements should remain the same, since the change in material leads
to change in acoustic impedance of the side walls, and the vortex shedding
frequency is found to be shifted by the transverse sound wave in the flow duct.
Test sections of same materials can avoid the discrepancy in measured vortex
shedding frequencies and thus the rescaling of coherence length. Although
plexiglass is the preferable test section material for the PIV set-up, yet it is
advisable that at least one side wall of the test section be made of Kevlar in
order to avoid the flow induced resonance.

• Application to high-lift devices with complex structure

With the feasibility of PIV based acoustic prediction proved on the Gurney
flap, a prospective extension can be the application of the PIV technique to
the noise prediction of high-lift devices with structures that are more complex.
This continuation may facilitate the design of quieter lift enhancing devices.

Xueqing Zhang M.Sc. Thesis

 Bibliography

[1] J. Anderson Jr. Fundamentals of Aerodynamics. McGraw Hill, 2001.

[2] T Baur and J Köngeter. Piv with high temporal resolution for the deter-
mination of local pressure reductions from coherent turbulence phenomena.
In 3rd international workshop on particle image velocimetry, Santa Barbara,
CA, USA, 1999.

[3] Julius S Bendat and Allan G Piersol. Random data: analysis and measurement
procedures, volume 729. John Wiley & Sons, 2011.

[4] Stephen A Billings. Nonlinear system identification: NARMAX methods in

the time, frequency, and spatio-temporal domains. John Wiley & Sons, 2013.

[5] RD Blevins. The effect of sound on vortex shedding from cylinders. Journal
of Fluid Mechanics, 161:217–237, 1985.

[6] YN Chen. Criteria for the cross-flow-induced tube vibrations in tube bank
heat exchangers. In Vibration in nuclear plant. Proceedings of international
conference held at Keswick, UK in May 1978, 1979.

[7] National Research Council. Improving the Continued Airworthiness of Civil

Aircraft: A Strategy for the FAA’s Aircraft Certification Service. The National
Academies Press, Washington DC, 1998.

[8] DG Crighton and FG Leppington. Scattering of aerodynamic noise by a semi-

infinite compliant plate. Journal of Fluid Mechanics, 43(04):721–736, 1970.

MSc. Thesis Xueqing Zhang

110 BIBLIOGRAPHY

[9] N Curle. The influence of solid boundaries upon aerodynamic sound. In

Proceedings of the Royal Society of London A: Mathematical, Physical and
Engineering Sciences, volume 231, pages 505–514. The Royal Society, 1955.

[10] R De Kat and BW Van Oudheusden. Instantaneous planar pressure determi-

nation from piv in turbulent flow. Experiments in fluids, 52(5):1089–1106,
2012.

[11] Roeland de Kat and Bas W van Oudheusden. Instantaneous planar pressure
from piv: analytic and experimental test-cases. In Proceedings of the 15th in-
ternational symposium on applications of laser techniques to fluid mechanics,
Lisbon, Portugal, 2010.

[12] Roeland De Kat, Bas W Van Oudheusden, and Fulvio Scarano. Instantaneous
planar pressure field determination around a square-section cylinder based on
time resolved stereo-piv. In Proceedings of the 14th International Symposium
on Applications of Laser Techniques to Fluid Mechanics, Lisbon, Posrtugal,
07-10 July, 2008, paper No. 1259. Calouste Gulbenkian Foundation, 2008.

[13] M Debrouwere. An assessment of acoustically transparent wind tunnel walls.

Master of Science Thesis, Delf university of technology, Aerospace department,
2013.

[14] John M Eargle. Handbook of recording engineering. Springer Science & Busi-
ness Media, 2012.

[15] Scarano F. Experimental aerodynamics. Lecture Notes, 2013.

[16] LaVision GmbH. Piv image evaluation.

http://www.piv.de/piv/measurement_principle/page_1.php. Accessed:
2015-09-29.

[17] Marvin E Goldstein. Aeroacoustics. New York, McGraw-Hill International

Book Co., 1976. 305 p., 1, 1976.

[18] Roi Gurka, Alex Liberzon, D Hefetz, D Rubinstein, and U Shavit. Computa-
tion of pressure distribution using piv velocity data. In Workshop on particle
image velocimetry, volume 2, 1999.

Xueqing Zhang M.Sc. Thesis

BIBLIOGRAPHY 111

[19] Richard Haberman. Applied partial differential equations with fourier series
and boundary value problems. AMC, 10:12, 2004.

[20] Christian Haigermoser. Application of an acoustic analogy to piv data from

rectangular cavity flows. Experiments in fluids, 47(1):145–157, 2009.

[21] Colin H Hansen and David A Bies. Engineering noise control. Spon, 1995.

[22] Arne Henning, Kristian Kaepernick, Klaus Ehrenfried, Lars Koop, and An-
dreas Dillmann. Investigation of aeroacoustic noise generation by simultane-
ous particle image velocimetry and microphone measurements. Experiments
in fluids, 45(6):1073–1085, 2008.

[23] Arne Henning, Lars Koop, and Klaus Ehrenfried. Simultaneous particle image
velocimetry and microphone array measurements on a rod-airfoil configura-
tion. AIAA journal, 48(10):2263–2273, 2010.

[24] Arne Henning, Andreas Schröder, Lars Koop, and J Agocs. Causality correla-
tion analysis on a cold jet by means of simultaneous piv and microphone mea-
surements. In 6th AIAA/CEAS Aeroacoustics Conference, Stockholm (Swe-
den), vol AiAA-2010-3753, 2010.

[25] Arne Henning, Björn Wrede, and Reinhard Geisler. Aeroacoustic investi-
gation of a high-lift device by means of synchronized piv and microphone
measurements.

[26] Arne Henning, Björn Wrede, and Andreas Schröder. About the ambiguity
of noise source localization based on the causality correlation technique. In
Proceedings of the 17th international symposium on applications of laser tech-
niques to fluid mechanics, Lisbon, Portugal, 2014.

[27] Avraham Hirschberg and Christophe Schram. A primitive approach to aeroa-

coustics. In Sound-Flow Interactions, pages 1–30. Springer, 2002.

[28] Michael S Howe. Theory of vortex sound, volume 33. Cambridge University
Press, 2003.

[29] MS Howe. The influence of vortex shedding on the generation of sound by

convected turbulence. Journal of Fluid Mechanics, 76(04):711–740, 1976.

MSc. Thesis Xueqing Zhang

112 BIBLIOGRAPHY

[30] MS Howe. A review of the theory of trailing edge noise. Journal of Sound
and Vibration, 61(3):437–465, 1978.

[31] LinearX Systems Inc. M53 Measurement Microphone, 2000.

[32] American National Standards Institute. American National Psychoacoustical

Terminology S3.20. American Standard Association, 1973.

[33] David Jeffrey, Xin Zhang, and David W Hurst. Aerodynamics of gurney flaps
on a single-element high-lift wing. Journal of Aircraft, 37(2):295–301, 2000.

[34] MG Jones, WR Watson, and TL Parrott. Benchmark data for evaluation of

aeroacoustic propagation codes with grazing flow. AIAA paper, 2853:2005,
2005.

[35] Chisachi Kato, AKIYOSHI Iida, Yasushi Takano, HAJIME Fujita, and
MASAHIRO Ikegawa. Numerical prediction of aerodynamic noise radiated
from low mach number turbulent wake. AIAA paper, (93-0145), 1993.

[36] Valentina Koschatzky, Jerry Westerweel, and Bendiks Jan Boersma. Com-
parison of two acoustic analogies applied to experimental piv data for cavity
sound emission estimation. In Proceedings of the 16th AIAA/CEAS Aeroa-
coustic Conference, pages 7–9, 2010.

[37] DF Kurtulus, F Scarano, and L David. Unsteady aerodynamic forces estima-

tion on a square cylinder by tr-piv. Experiments in Fluids, 42(2):185–196,
2007.

[38] LD Landau and EM Lifshitz. Fluid mechanics, vol. 6. Course of Theoretical

Physics, pages 227–229, 1987.

[39] YC Li, JJ Wang, GK Tan, and PF Zhang. Effects of gurney flaps on the
lift enhancement of a cropped nonslender delta wing. Experiments in fluids,
32(1):99–105, 2002.

[40] Robert H Liebeck. Design of subsonic airfoils for high lift. Journal of aircraft,
15(9):547–561, 1978.

Xueqing Zhang M.Sc. Thesis

BIBLIOGRAPHY 113

[41] Michael J Lighthill. On sound generated aerodynamically. i. general theory.

In Proceedings of the Royal Society of London A: Mathematical, Physical and
Engineering Sciences, volume 211, pages 564–587. The Royal Society, 1952.

[42] Xiaofeng Liu and Joseph Katz. Instantaneous pressure and material accelera-
tion measurements using a four-exposure piv system. Experiments in Fluids,
41(2):227–240, 2006.

[43] V Lorenzoni, M Tuinstra, and F Scarano. On the use of time-resolved particle

image velocimetry for the investigation of rod–airfoil aeroacoustics. Journal
of Sound and Vibration, 331(23):5012–5027, 2012.

[44] Valerio Lorenzoni. Aeroacoustic Investigation of Rod-Airfoil Noise based on

Time-Resolved PIV. Master’s thesis, Delft University of Technology, the
Netherlands, 2008.

[45] David Mackenzie. ICAO: a history of the international civil aviation organi-
zation. University of Toronto Press, 2010.

[46] American Scientist Magazine.

http://www.americanscientist.org/Libraries/images/
201473114311210514-2014-09CoxF3.jpg. Accessed: 2015-11-03.

[47] Thomas J Mueller. Aeroacoustic measurements. Springer Science & Business

Media, 2002.

[48] Roy Myose, Michael Papadakis, and Ismael Heron. Gurney flap experiments
on airfoils, wings, and reflection plane model. Journal of Aircraft, 35(2):206–
211, 1998.

[49] Tomomichi Nakamura, Shigehiko Kaneko, Fumio Inada, Minoru Kato, Ku-
nihiko Ishihara, Takashi Nishihara, and Njuki W Mureithi. Flow-induced vi-
brations: Classifications and lessons from practical experiences. Butterworth-
Heinemann, 2013.

[50] F Noca, D Shiels, and D Jeon. A comparison of methods for evaluating time-
dependent fluid dynamic forces on bodies, using only velocity fields and their
derivatives. Journal of Fluids and Structures, 13(5):551–578, 1999.

MSc. Thesis Xueqing Zhang

114 BIBLIOGRAPHY

[51] RM Oreslli, Julio R Meneghini, and Fabio Saltara. Two and three-dimensional
simulation of sound generated by flow around a circular cylinder. In 15th
AIAA/CEAS Aeroacoustics Conference, AIAA, volume 3270, 2009.

[52] G Philippe, D Guy, and L Jean. Gurney flap scaling for optimum lift-to drag
ratio. AIAA j, 35(12):1888–1890, 1997.

[53] Alan Powell. On the aerodynamic noise of a rigid flat plate moving at zero
incidence. The Journal of the Acoustical Society of America, 31(12):1649–
1653, 1959.

[54] Alan Powell. Theory of vortex sound. The journal of the acoustical society of
America, 36(1):177–195, 1964.

[55] S Probsting. Coherent structures at the serrated trailing-edge of a naca 0012.

Master of Science Thesis, Delf university of technology, Aerospace department,
2012.

[56] Stefan Pröbsting, Fulvio Scarano, Matteo Bernardini, and Sergio Pirozzoli.
On the estimation of wall pressure coherence using time-resolved tomographic
piv. Experiments in fluids, 54(7):1–15, 2013.

[57] Markus Raffel, Christian E Willert, Jürgen Kompenhans, et al. Particle image
velocimetry: a practical guide. Springer, 2013.

[58] M Roger. Aeroacoustics: Some theoretical background—the acoustic analogy.

Computational Aeroacoustics, pages 2006–05, 2006.

[59] Daniel A Russell, Joseph P Titlow, and Ya-Juan Bemmen. Acoustic

monopoles, dipoles, and quadrupoles: An experiment revisited. American
Journal of Physics, 67(8):660–664, 1999.

[60] T.Watanabe S. Hayama. In Proceedings of the 73rd JSME’s Annual Meeting

(in Japanese) pp.155 á156, 1995.

[61] S Salcedo, F Monge, F Palacios, F Gandı́a, A Rodrı́guez, M Barcala, and Crtra

Ajalvir. Gurney flaps and trailing edge devices for wind turbines. EWEC.
Athens: EWEA, 2006.

Xueqing Zhang M.Sc. Thesis

BIBLIOGRAPHY 115

[62] Christophe Schram and Lilla Koloszár. Validation and improvement of air-
frame noise prediction tools. Innovation for Sustainable Aviation in a Global
Environment: Proceedings of the Sixth European Aeronautics Days, Madrid,
30 March-1 April, 2011, page 144, 2012.

[63] J Shah. Aeroacoustics and flow dynamics of an airfoil with a gurney flap using
tr-piv. Master of Science Thesis, Delf university of technology, Aerospace
department, 2015.

[64] DR Troolin, EK Longmire, and WT Lai. Time resolved piv analysis of flow
over a naca 0015 airfoil with gurney flap. Experiments in Fluids, 41(2):241–
254, 2006.

[65] Seppo Uosukainen. Foundations of acoustic analogies. VTT Publications 757,

2011.

[66] Bas W van Oudheusden, Fulvio Scarano, Eric WM Roosenboom, Eric WF

Casimiri, and Louis J Souverein. Evaluation of integral forces and pressure
fields from planar velocimetry data for incompressible and compressible flows.
Experiments in Fluids, 43(2-3):153–162, 2007.

[67] Walter G Vincenti. What engineers know and how they know it, volume 141.
Baltimore: Johns Hopkins University Press, 1990.

[68] Daniele Violato, Peter Moore, and Fulvio Scarano. Lagrangian and eulerian
pressure field evaluation of rod-airfoil flow from time-resolved tomographic
piv. Experiments in fluids, 50(4):1057–1070, 2011.

[69] JJ Wang, YC Li, and K-S Choi. Gurney flap—lift enhancement, mechanisms
and applications. Progress in Aerospace Sciences, 44(1):22–47, 2008.

[70] Meng Wang, Jonathan B Freund, and Sanjiva K Lele. Computational predic-
tion of flow-generated sound. Annu. Rev. Fluid Mech., 38:483–512, 2006.

[71] Jerry Westerweel and Fulvio Scarano. Universal outlier detection for piv data.
Experiments in Fluids, 39(6):1096–1100, 2005.

[72] B Wieneke. Stereo-piv using self-calibration on particle images. Experiments

in Fluids, 39(2):267–280, 2005.

MSc. Thesis Xueqing Zhang

116 BIBLIOGRAPHY

[73] Wikipedia. Open cylinder resonance.

https://upload.wikimedia.org/wikipedia/commons/4/48/
OpenCylinderResonance.svg. Accessed: 2015-12-01.

[74] JE Williams and LH Hall. Aerodynamic sound generation by turbulent flow in

the vicinity of a scattering half plane. Journal of Fluid Mechanics, 40(04):657–
670, 1970.

[75] JE Ffowcs Williams. Hydrodynamic noise. Annual Review of Fluid Mechanics,

1(1):197–222, 1969.

[76] X Zhang. Finite element numerical simulation of straightforward method for

liner impedance eduction. Bachelor of Science Thesis (in Chinese), Beihang
University, School of Jet Propulsion, 2013.

Xueqing Zhang M.Sc. Thesis

 Appendix A

Gaussian curve fitting for the evaluation of

coherence length

The purpose of the present appendix is to offer a brief mathematical description of

the Gaussian curve fit procedure applied in the computation of spanwise coherence
length, as mentioned in subsection 6.4.1.
The standard Gaussian equation is written as:
(x−µ)2
y = g(x) = Ae− 2σ 2 (A.1)

Based on the considerations of the properties of the coherence, the coherence

coefficients ranges from 0 to 1, and thus in this case A = 1. In addition, the
value of coherence coefficients are calculated at a specified spanwise location with
respect to the mid-span location (z = 0), such that at z = 0, the coherence
coefficient reaches the maximum value 1, and thus µ = 1. The Gaussian equation
used in the curve fitting hence takes a simplified form of:
x2
y = g(x) = e− 2σ2 (A.2)

In this case, y takes the values of coherence coefficients calculated basd on the
measurements of stereoscopic PIV, and x the spanwise locations, with z = 0 the

MSc. Thesis Xueqing Zhang

118 Gaussian curve fitting for the evaluation of coherence length

location of the mid-span location.

As can be observed in Figure A.1, besides the main peak at the mid-span location
where coherence coefficient takes the value of unity, there are secondary peaks.
However, Gaussian function should be fitted with the data points that constitute
only the main peak. In order to identify the portion of data corresponding to the
main peak automatically for the data at each frequency, the first derivatives of the
curve at the specified frequencies have been investigated. A Matlab R programme
searches for the spanwise locations where the first derivatives change signs. Be-
tween the nearest sign-changing locations on both sides of the mid-span (marked
with magenta lines on Figure A.1), the portion of data is taken for the curve fit-
ting.
The present Gaussian curve fitting problem can be converted to problem of solv-

Figure A.1: Gaussian curve fit of the spanwise coherent coefficients

ing a overdetermined system of linear equations, and thus the linear least square
approach can be applied to determine the value of standard deviation σ.
Take log of both side, then Equation A.2 becomes:
x2
ln(y) = − (A.3)
2σ 2
Xueqing Zhang M.Sc. Thesis
 119

1
then let Y = ln(y), a = 2σ 2
, X = x2 , the system can be rewritten as linear
functions:

Yi = aXi + εi (A.4)

Finding the optimized slope a is equivalent in minimizing the objective function:

minQ(a) (A.5)
a

for Q(a) = ε2i = (Yi − aXi )2 . In other word, a is the solution of the following
P P

minimization problem. By using the calculus:

∂S
=0 (A.6)
∂a
it can be shown that a that minimize the objective function is:
P
Yi ki
â = P 2 (A.7)
Xi
p
Correspondingly, the σ that optimizes the curve fitting is σ̂ = −1/2â. The
coherence length is then approximated with the distance between the mid-span
location (z = 0) and location where the Gaussian function reaches the critical
value. Different critical values were taken by authors of the references mentioned
in subsection 6.4.1 and currently there exists no rule of thumb for the choice of this
critical value. In the present study, e−2 is chosen as the critical value, such that the
Gaussian function reaches the critical value at x = 2σ. Since the range [−2σ 2σ]
contains 95% of the total integral of the Gaussian function, thus 2σ corresponds
to the 95% confidence level and expresses a conventional heuristic in the empirical
sciences. Whether the value of e−2 is the suitable for the flow under investigation
is still a topic open for further investigations.
An example of the fitted Gaussian function is plotted in Figure A.2 with the fre-
quency specified at f = 314Hz. The blue diamonds show the calculated coherent
coefficients at each spanwise location with respect to the mid-span, the red line
plots the fitted Gaussian curve within the FOV of the stereoscopic PIV, and the
dotted line is the extrapolation of the Gaussian function to the spanwise locations
outside the FOV. The extrapolation is only necessary for the frequencies (such as
f = 314Hz as shown) at which the value of coherence coefficients does not drop
below the specified critical value within the FOV.

MSc. Thesis Xueqing Zhang

120 Gaussian curve fitting for the evaluation of coherence length

Figure A.2: Extrapolated Gaussian curve and the predicted length of coherence

Xueqing Zhang M.Sc. Thesis

 Appendix B

On the different shedding frequency

obtained from the planar and stereoscopic
PIV measurements

As has been mentioned in subsection 6.4.1, the velocity fluctuations measured by

stereoscopic PIV exhibit a frequency peak, corresponding to the vortex shedding
frequency, of 271Hz in the power spectra, which deviates from the 314.9Hz mea-
sured by planar PIV and the far-field microphones. A comparison of the power
spectra obtained from the measurements of the two different PIV set-ups is given
in Figure B.1. The present appendix offers a qualitative discussion on the mecha-
nism of the observed frequency shift.
As described in chapter 5, two side walls of the test section for simultaneous pla-
nar PIV measurements are made of Kevlar, which is permeable for acoustic waves,
while the side walls for stereoscopic PIV measurements are all plexiglass in order
to facilitate the pass of laser light. The change in material essentially leads to
the change in acoustic impedance of the test section walls, which may result in
resonance when the following two conditions are satisfied:

MSc. Thesis Xueqing Zhang

 On the different shedding frequency obtained from the planar and stereoscopic
122 PIV measurements

• Frequency condition: the frequency of excitation coincides with the nat-

ural frequency of any acoustic mode of the flow duct.

• Energy condition: the energy input to the acoustic mode exceeds the
energy dissipation of this mode in the acoustic field.

Figure B.1: Comparison of the power spectra (∆f = 2.4Hz) of the transverse velocity
(u-component) measured by planar PIV and stereoscopic PIV (GF6-AOA4-V20).

In the experimental campaign, at the same incoming flow rate and with the same
model, the experiments with all-plexiglass test section were audibly noisier than
those with plexiglass-Kevlar test section. A possible explanation can be inferred
that the shift in vortex shedding frequency is the result of the effect of sound
on vortex shedding. Figure B.2 shows the schematic of the feedback mechanism
between the flow and acoustic field.
The influence of a transverse sound wave on vortex shedding frequency of a rigid
circular cylinder in a duct has been explored by Blevins(1985) [5], the findings of
which can be summarized as:

Xueqing Zhang M.Sc. Thesis

 123

Figure B.2: Feedback mechanism between flow and acoustic field.[49]

1. In absence of the sound, vortex shedding is not a steady harmonic process,

but wanders 1 − 2% around the nominal vortex shedding frequency. Applied
sound reduces the bandwidth of vortex shedding and when the strength of the
sound reaches a certain threshold, the frequency wander can be completely
eliminated (in some literatures, this phenomenon is also referred to as ’locked-
in’ effect).

2. Sound applied near the vortex shedding frequency shifts the vortex shed-
ding towards the sound frequency. Maximum shifts of 8% in frequency
were achieved with the sound of higher frequency than the vortex shedding.
Blevins further claimed a tendency for the preferred vortex-shedding fre-
quency to decrease when sound is applied and expected a larger shift by
applying sound with frequency lower than the shedding frequency.

3. The entrainment produced by sound is due to the induced velocity of the

sound wave rather than by the sound pressure. The greatest influence can
be achieved when the cylinder is placed at a node of a sound pressure (an
anti-node of the acoustic velocity).

4. Turbulence in the free stream tends to suppress the influence of sound.

Sound-induced entrainment can occur only if the acoustic velocity is above
the background turbulence velocities.

5. Correlation of vortex shedding along the cylinder span increases substantially

when sound is applied at the vortex shedding frequency.

MSc. Thesis Xueqing Zhang

 On the different shedding frequency obtained from the planar and stereoscopic
124 PIV measurements

From Figure B.1, it can be observed that the width of the frequency peak
measured by stereoscopic PIV is narrower than the one measured by planar
PIV. And the increase in amplitude could be the result of energy input from the
excitation. In addition, the V-tunnel provides incoming flow with turbulent level
below 0.1%, which is very low. However, the cross-section of the test section used
in the current study is 40 × 40cm2 , correspondingly, the frequency of the first
transversal mode is 425Hz, which is larger than and faraway from the natural
shedding frequency 314.9Hz. And thus, the transversal sound wave is not the
reason and there should be other causes.

The investigations on the reso-

nance of heat exchanger have re-
vealed several similar phenomena
in longitudinal wave in the flow
duct. Chen reported the lock-in ef-
fect caused by longitudinal acoustic
waves on vortex shedding back in
1979[6]. The location of the anti-
node of acoustic velocity was also
discussed for longitudinal modes
by Hayama et al.(1995) [60]. Yet
shift in vortex shedding frequency
caused by longitudinal wave mode
is not found by the author of this
thesis in the literatures about vi-
brations in tube bank heat ex-
changers. Figure B.3 shows the
shapes of the first three longitudi-
nal modes in an open cylindrical
Figure B.3: The first three resonance modes
tube.
in an open cylindrical tube. The horizontal
In the current study, the total
axis is pressure.[73]
height (test section and part of the
transition duct) of the flow duct with the cross-section of 40 × 40cm2 is around

Xueqing Zhang M.Sc. Thesis

 125

0.8m, which corresponds to the frequency of the first longitudinal mode of 215Hz.
This longitudinal frequency may have contributed to the shift in vortex shedding
frequency of the Gurney flap.
A local maxima at 215Hz is distinguishable, but not evident on the power spec-
trum of the background noise of the wind tunnel in operation (Figure B.4). How-
ever, the background noise was measured with the installation of the plexiglass-
Kevlar test section, while the background noise measurement on all-plexiglass test
section is missing. If the longitudinal mode of 215Hz is the mode that shifts the
vortex shedding in the all-plexiglass case, a more discernible peak is expected on
the background noise power spectrum of that case. Another issue that remains to
be clarified is what the energy input is, or excitation of the 215Hz longitudinal
mode.

Figure B.4: Power spectra (∆f = 2.4Hz) of the background noise of V-tunnel with empty
plexiglass-Kevlar test section and incoming flow at 20m/s.

Despite the issues to be clarified in the mechanism of frequency shift in this ap-
pendix, a conclusion can still be reached that all-plexiglass test section, although

MSc. Thesis Xueqing Zhang

 On the different shedding frequency obtained from the planar and stereoscopic
126 PIV measurements

Figure B.5: An example of longitudinal acoustic pressure distribution. [−0.203, 0] ∪

[0.406, 0.609]: solid wall; [0, 0.406]: acoustic liner. SP L = 130dB, M a = 0.335,f =
500Hz, Z = 0.33 − 0.98i[34] [76]

favoured for PIV set-up, is not as preferable in the acoustic measurements or vor-
tex shedding related measurements.
Figure B.5 offers an example of the longitudinal acoustic pressure distribution in
a cylindrical tube with different wall conditions at different longitudinal locations.
The sound pressure level applied at the inlet of the tube x = −0.203m is 130dB,
with the flow speed at the center of the tube of M a = 0.335. It can be ob-
served that due to the presence of the acoustic liner in the region [0, 0.406]m, the
sound pressure level exhibits a drop up to 10dB (black diamond in Figure B.5),
corresponding to a reduction by 2/3 in sound pressure. While for the solid wall
region [−0.203, 0]m and [0.406, 0.609]m, the sound pressure level almost remains
the same.
It can be inferred that the application of liner can be an effective way of avoiding
resonance by attenuating the energy input of a certain transverse or longitudinal
acoustic mode. It can also be inferred that the acoustic impedance of Kevlar is

Xueqing Zhang M.Sc. Thesis

 127

larger than that of the plexiglass. Hence, the replacement of Kevlar, at least on
one of the four sides of the test section, would be desirable for avoiding the effect
of sound on vortex shedding.

MSc. Thesis Xueqing Zhang

 On the different shedding frequency obtained from the planar and stereoscopic
128 PIV measurements

Xueqing Zhang M.Sc. Thesis