Effect of Regular Surface Perturbations on Flow Over

an Airfoil
Arvind Santhanakrishnan∗ and Jamey D. Jacob†
Dept. of Mechanical Engineering, University of Kentucky, Lexington, KY 40506

This paper presents an investigation on the effect of introducing large-scale roughness

through static curvature modifications on the low speed flow over an airfoil. The surfaces of
a standard Eppler 398 airfoil have been modified with regular perturbations or “bumps” of
the order of 2%c for this purpose. While the actual E398 airfoil is not a suitable candidate
for low Re cases due to extensive prevalence of boundary layer separation, it is expected
that the bumps would exercise passive flow control by promoting early transition to tur-
bulence, thereby reducing the extent of separation and improving the performance. The
fluid dynamic mechanism behind this separation control methodology is not clearly un-
derstood, however. Smoke-wire flow visualization is performed for qualitative observation
of the separation region in both the perturbed and unmodified airfoil geometry cases. At
higher Re values, pressure probe measurements are made to quantify the wake momentum
deficits. Unsteady 2D PIV measurements are employed to understand the near-wall flow-
field behavior . The size and strength of vortical structures formed in the separating shear
layer are examined, along with measurements on the laminar separation bubble. All the
experiments are conducted for chord based Re values ranging from 25,000 to 500,000.

Nomenclature

α Angle of attack
δ Boundary layer height
b Span
c Chord length
k Roughness height
M Mach number
Re Reynolds number
Rec Reynolds number based on chord
U∞ Freestream speed
w Width of bump
FSTI Freestream turbulence intensity

I. Introduction
he aerodynamic characteristics of low Reynolds number airfoils are fundamentally different from those
T seen in typical aerospace applications. Subsonic aerodynamics, not a major area of study until the recent
past, promises tremendous potential in the development of small, robust and high performance aircraft:
Unmanned Aerial Vehicles (UAVs), Remotely Piloted Vehicles (RPVs) and Micro Aerial Vehicles (MAVs).
These are particularly useful for defense applications such as surveillance, communication links, ship decoys
and detection of biological, chemical or nuclear materials.1 Another important application of these vehicles
has been identified in space or planetary exploration, especially in extreme low density environments such as
in Mars. These vehicles present extreme constraints to the airfoil design process in the form of (a) extreme
operating conditions (cruise velocity, altitude, density); and (b) very small aspect ratios. The mission profiles
∗ Graduate Student; Student Member, AIAA; asant0@engr.uky.edu.
† Associate Professor; Senior Member, AIAA; jdjacob@uky.edu.

1 of 9

American Institute of Aeronautics and Astronautics

tend to incorporate entirely different regimes in terms of their speed, altitude and maneuvering requirements.
For example, RPVs need to be operative at both normal and very high altitudes (where the density of air is
low). From a fluid dynamist’s point of view, the performance of an aircraft is essentially controlled by the
development of the boundary layer on its surface and its interaction with the mean flow. This interaction
decides the pressure distribution on the airfoil surface, and subsequently the aerodynamic loads on the wing.
In order to obtain the highest levels of performance efficiencies for mission varying aircraft, it is necessary
to either: (a) alter the boundary layer behavior over the airfoil surface—flow control methods of interest
here, and/or (b) change the geometry of the airfoil real time for changing freestream conditions—adaptive
wing technology.2 The starting point towards achieving deterministic design of low speed airfoils lies in
understanding the physics of the “fluid dynamic problem” - in essence.

A. The Low Re Problem

In the case of airfoils with low chord lengths as in micro and meso aerial vehicles, the Reynolds number
of the flow remains low—ranging from 15,000 to 500,000. Fig. 1, from Lissaman3 illustrates the Reynolds
number and speed envelope (in terms of the Mach number) of aerial vehicles, both natural and man-made.
The zone of our interest (for UAVs, RPVs and MAVs) spans a large Re–M bandwidth because of their
widely varying operating range. These vehicles are usually operated at high altitudes, where the kinematic
viscosity is sufficiently increased by the low ambient density. They are frequently required to perform at
Reynolds number values below half a million. The predominance of viscous effects in this regime causes the
production of high drag forces and limits the maximum lift coefficient.
At a critical Re value of about 70,000 - performance transition occurs and lift-to-drag ratio improves.4
The range of 30,000 to 70,000 is of interest to MAV designers. The choice of airfoil is very important in
this range as a small increase in thickness can have significant effects on laminar separation. The flow over
the airfoil at these low Reynolds numbers is laminar, and slight changes in flow speed can have significant
effects in the lift-to-drag ratio.5 The largest contribution to the total drag is from the pressure or form drag
component, which arises due to the viscous influence of the boundary layer on the primarily inviscid pressure
field. The drag problem is worsened when the flow separates, as the form drag increases tremendously. It can
thereby be seen that the ultimate solution for drag reduction in highly viscous flows lies in the development
of efficient methods for controlling flow separation. It is herein that flow control research proves to be of
immense value.

Figure 2. L/D variation with Re (The inset shows

Figure 1. Re versus M for a wide range of flying
a low Re airfoil, LNV109A, undergoing gross sepa-
objects (from Lissaman 1983)
ration at Re=25,000.)

B. Roughness Application
In certain cases when the pressure gradient imposed on the flow is not too adverse, transition and reattach-
ment may occur after laminar separation, and the resultant turbulent boundary layer is found to be more
resistant to flow separation.6 This provides a reasonable justification for separation control by means of pro-
moting early transition in laminar flows, thereby reducing the otherwise imminent form drag. Experimental
observations show that “rough” airfoils perform better than the “smooth” surface airfoils at low Re values, as

2 of 9

American Institute of Aeronautics and Astronautics

shown in Fig. 2 (original curve adapted from McMasters and Henderson7 ). The turbulence promoting devices
(or turbulators) may range from passive methods such as mechanical roughness elements (strips, bumps), to
active methods such as acoustic excitation, surface vibration. These methods introduce large disturbances
in the flow so as to cause bypass transition, and are hard to analyze.8 One of the earliest studies of the effect
of surface protuberances on airfoil and wing characteristics can be found in NACA reports.9, 10 The chord
based Re of the flow in these experiments was approximately 3.1 million; and the effects of variations in
shape, span length, height and position of protuberances were considered. Four protuberances were placed
on the upper surface at leading edge, front spar, maximum thickness and rear spar locations. It was observed
that the loss of lift was directly proportional to the height of protuberances (order of 1/100-1/500 of chord
length). At higher angles of attack, the protuberances had an adverse effect, especially when moved closer to
the leading edge. Most of the work in roughness-related research has been aimed to understand the effects of
icing on unsteady flow over an airfoil.11, 12 Drag reduction by employing boundary layer trips was reported
by Lyon and coworkers.13 They observed that thicker trips showed slightly better performance than thinner
trips, and simple 2D trips provided the same advantage (if any) as complex 3D trips. The effects of large
distributed surface roughness on airfoil boundary layer development and transition to turbulence has been
investigated for Re values of 0.5,1.25 and 2.25 million by Kerho et al.14 Hot wire measurements were con-
ducted for a NACA 0012 airfoil with hemispherical disturbances of 0.35 mm height taped up to a maximum
chordwise extent of 0.5 inches. They examined a variety of roughness ranging in heights lesser and greater
than the boundary layer thickness. They observed that the roughness promoted the growth of a transitional
boundary layer, which required substantial chordwise extent (downstream of the roughness) to become fully
turbulent. The fluctuating streamwise velocity and turbulence intensity in the roughness-induced boundary
layer was found to be lower than the smooth case. In general, longer the chordwise extent of the roughness
and larger the roughness dimensions, length of the transitional region was found to decrease.

C. Present Flow Control Approach

The proposed method of flow control here is in introducing “large-scale” roughness to the airfoil surface,
such that the resultant shape would have a minor change in curvature. The basis of this concept stems from
the design of an inflatable-rigidizable wing for the University of Kentucky BIG BLUE student project. A
detailed description of the project and the airfoil selection process can be found elsewhere.15 The structure
of the wing had to accommodate internal baffles required for the inflation process. Due to this manufacturing
constraint, the Eppler 398, a relatively thick airfoil, was selected. The E398 has a design Rec value of around
2 million, tailored to application in HPVs (Human Powered Vehicles). The baffles inside the wing gave rise
to a modified E398 profile with regular surface perturbations (Fig. 3). The radius of the bumps was of the
order of 2%c. While covering the airfoil with a membrane (to mimic the smooth profile) and adding a trailing
edge extension were considered, it was decided to leave the airfoil unskinned to keep the flow tripped at all
times along the surface. It is interesting to note that this bumpy profile has a blunt trailing edge.

Figure 3. Modified E398 “bumpy” cross section (shown within the unmodified “smooth” E398 profile.)

When using roughness elements to alter flowfield behavior, the effects of changing the following parameters
should be considered: (a) Rec , (b) imposed pressure gradient (angle of attack), (c) roughness placement,
(d) number of roughness elements, (e) geometric roughness configurations, and (f) height of roughness with
respect to the boundary layer. In the present case, factor (c) translates to chordal/spanwise bump location,
while factor (e) translates to size and shape of bumps and “inter-bump” spacing. In this paper, the effects
of variations in factors (a) to (d) will be considered. For flow control to be of any advantage, the following
recommendations are available in the literature: (i) the roughness height (k) should be small as compared
with the boundary layer height (δ), i.e. k/δ ≤ 1, (ii) roughness location prior to the region of separation is
“optimal”.16
It is important to note at this juncture that considering the flow over the bumps in the E398 profile
as a roughness-induced effect would not be accurate. Specifically, one is faced with the question: what

3 of 9

American Institute of Aeronautics and Astronautics

would be the length scale to safely consider “roughness” as a “curvature” related problem (and vice versa)?
It is intuitive to expect that both these effects have some similarity in their mechanism of affecting the
fluid, and that there should be a limiting length scale when both these effects become one and the same.
There has not been reported work on the effects of “large-scale” roughness or “small-scale” curvature on
low Re flow over airfoils, and this presents a challenging problem that involves understanding the governing
instability and transition mechanisms. A study that is closely related to the surface modification herein is
the delay of stall on humpback whale flippers due to presence of leading edge tubercules.17 The tubercules
considered in the above study were approximately large sinusoidal protuberances extending with varying
amplitude throughout the leading edge of the flipper. Wind tunnel lift-drag measurements were conducted
on a modified NACA 0020 airfoil (representative of the flipper) for an Rec range from 5.05x103 to 5.20x105 ,
and an angle of attack range from −2 ◦ to 20 ◦ . It was found that the stall angle was delayed by about
40% when compared to the “smooth” baseline case, with increase in lift and decrease in drag. This provides
the motivation to examine a potential passive flow control application of “large-scale” roughness in low Re
flows.

II. Experimental Setup

A. Wind Tunnel
The ideal and perturbed (or “bumpy”) E398 airfoils were tested in the wind tunnel facilities at the University
of Kentucky. For flow visualization and wake survey tests, a wind tunnel with a 0.6 m x 0.6 m test section
and length of 1.2 m, driven by a 50 hp motor was used (FSTI < 0.5 %). Another low-turbulence open-circuit
blow-down wind tunnel was used for PIV measurements. This is driven by a 7.5 hp motor; and has a 0.2
m x 0.4 m test section with a length of 1.5 m. Seeding the flow for PIV was less problematic in the smaller
test section of the latter wind tunnel.

B. Test Airfoils
Models of the bumpy and ideal E398 airfoils were constructed in the UK stereo-lithography lab, with c=0.3 m
(as in the full scale inflatable wing design) and span b=0.6 m (which is the width of the tunnel test section).
These models were used for the flow visualization and wake measurements. To understand the effect of
introducing regular perturbations on the airfoil, the simplest case of having a single bump (or “unibump”)
centered at 10%c (k=0.1265 cm=0.9855 %c; w=0.8529 cm=6.645%c) on the E398 airfoil was also considered.
For PIV measurements, the ideal, unibump, and bumpy models (c =0.128 m, b=0.2 m) were made by rapid
prototyping; with the profiles as shown in Fig. 4. The ideal E398 airfoil model surface was smoothed using
commercial quality sandpaper. The experiments conducted on all the above models spanned a wide Re range
from 25,000 to 500,000. In all the experiments, the model was mounted such that the flow over the airfoil
was completely two-dimensional. The chord c was used as the length scale for Re calculation (Re and Rec
mean the same, and would be used interchangeably hereon).

(a) Ideal. (b) Unibump. (c) Bumpy.

Figure 4. Model cross sections used for PIV.

C. Flow Visualization
Smoke-wire flow visualization as described in Batill and Mueller18 was used for qualitative observation of
the separation region. A stainless steel and tungsten wire (0.006 in diameter) was doped with a model train

4 of 9

American Institute of Aeronautics and Astronautics

smoke mixture containing mineral oil, oil of anise and blue dye. This wire was mounted on a stand placed
in front of the wing. On introducing a current to this wire, the oil evaporates due to Joule heating, making
smoke trails around the wing. A Sony XC-55BB camera and a Matrox Pulsar frame grabber package was
used to capture images. These tests were conducted for Rec values ranging from 25,000 to 100,000, and for
an angle of attack range from 0 ◦ to 12 ◦ .

D. Seven-Hole Pressure Probe

A seven hole probe manufactured by Aeroprobe Inc. was employed to quantify the time-averaged pressures
in the wake of the airfoil. The data acquisition system consists of a seven-hole probe, an 8 port multi-channel
pressure scanner, a two-axis traverse, a multi-axis stepping motor controller, a CIO-DAS08 Analog/Digital
Board, and a ESPIO signal conditioner. The seven-hole probe system is capable of automatically measuring
(a) all three-components (u, v, w) of velocity, (b) static and stagnation pressures on a pre-defined user grid.
The measurement plane was normal to the freestream direction and located at midspan (Fig. 5). The probe
was positioned at a distance of 30%c from the trailing edge. The computer controlled traverse was used to
move the probe in the vertical direction and acquire data for a total of 100 grid points, distributed evenly
across a total distance which included 100 mm above and 150 mm below the upper surface of the airfoil
respectively. The tests were conducted for Rec values from 150,000 to 500,000, and for an α range from −4 ◦
to 12 ◦ with 2 ◦ increments.

Figure 5. Wake survey setup.

E. Particle Image Velocimetry (PIV)

The laser sheet for PIV measurements was generated from a 50 mJ double-pulsed Nd:YAG laser with a
maximum repetition rate of 15 Hz. Pulse separations were varied from 40 µs to 1 µs based upon the
freestream velocity. A 10 bit Kodak Megaplus CCD camera with a 1008x1018 pixel array was used for
capturing images. Uniform seeding was accomplished using 1 micron oil droplets inserted upstream. For each
PIV run, 122 images were recorded for processing resulting in a minimum of 61 vector and vorticity fields from
which to generate mean flow field and statistics. The Wall Adaptive Lagrangian Parcel Tracking (WaLPT)
DPIV algorithm was employed for processing the raw data. A detailed description of the algorithm and its
formulation can be found elsewhere.19 In the LPT algorithm, seeding is tracked as fluid parcel markers and
tracks both their translations and deformations. During this tracking, fluid particles registered by individual
CCD pixels are advected with individually estimated velocities and total accelerations. The velocity field
needed to initialize the LPT process is obtained from a standard DPIV algorithm which uses multiple passes,
integer window shifting, and adjustable windows. Both the LPT and DPIV algorithms employ a rigorous
peak-detection scheme to determine velocity vectors and use the local velocity gradient tensor to identify
spurious velocity vectors. We have found that the LPT algorithm works well in the flow field of a vortex
which is characterized by high deformation rates where DPIV algorithms are plagued by biasing and limited
dynamic range. No smoothing algorithms or other post-processing techniques are employed on the data.
Vorticity, being a component of the velocity gradient tensor, is calculated spectrally at each grid point as an
intrinsic part of the LPT algorithm. To increase near-wall accuracy, image parity exchange (IPX) developed

5 of 9

American Institute of Aeronautics and Astronautics

by Tseui and Savas20 was used. The image is extended across the wall interface by mirroring and reversing
it, allowing the calculation of a velocity vector at the wall. The raw images were processed as image pairs in
32x32 interrogation areas to give 61 tensors containing the flow information, i.e the velocities and velocity
gradients. The sampling rate of the present PIV system was limited to a nominal value of 10 Hz.

III. Results and Discussion

A. Flow Visualization
The results of some of the tests are shown in Fig. 6. At Re = 25 · 103 and α = 0◦ , separation starts very
close to the leading edge for the ideal E398 profile, and there is no reattachment. Also, the flow streamlines
adjacent to the surface in the separation region are well demarcated, suggesting the flow is laminar prior
to experiencing separation. For the same (Re, α) and chordwise position, the bumpy profile (Fig. 6(b))
shows attached flow, and the streamlines adjacent to the surface are not distinctly clear. This is due to
the bumps tripping the flow to promote transition—earlier than in the previous case. The very low Re and
the adverse pressure gradient near the maximum thickness point causes the flow to separate in the bumpy
profile, however. It can be observed that the position of the separation point is shifted further downstream
(of the laminar separation point in Fig. 6(a)) due to addition of the bumps. The disturbance level posed by
the bumps is not enough to fully promote transition ahead of the maximum thickness point in the bumpy
profile, and the unstable laminar-turbulent boundary layer separates from the surface. At Re = 50 · 103 and
α = 4◦ (Fig. 6(c)), the flow separates earlier upstream for the ideal profile, compared to the Re = 25 · 103
case (Fig. 6(a)), as expected. The bumpy profile shows marked control in the separation extent (Fig. 6(d));
more than the equivalent lower (Re, α) case discussed earlier. The higher flow momentum and angle of
attack, coupled with the bumps - produce a higher disturbance level in the flow (as compared with the lower
(Re, α) case), and flow is tripped closer to the leading edge.

(a) Ideal, Re = 25 · 103 , α = 0◦ . (b) Bumpy, Re = 25 · 103 , α = 0◦ .

(c) Ideal, Re = 50 · 103 , α = 4◦ . (d) Bumpy, Re = 50 · 103 , α = 4◦ .

Figure 6. Separation region flow visualization.

For all (Re, α) cases observed, the separated shear layer did not reattach in both the ideal and bumpy

6 of 9

American Institute of Aeronautics and Astronautics

profiles. A laminar separation bubble was not expected at these low Rec values, in line with the observations
of Lissaman3 ((Rec )min = 50 · 103 for “closed” bubble formation). Among the perturbed wing cases, the
extent of separation is lesser for a higher Reynolds number and higher angle of attack. At smaller α, the
separation region difference is also noticable for Re = 25 · 103 and Re = 100 · 103 (not shown). For the very
low range of Re considered here, the critical roughness height expressed by Rek needs to be on the order of
1000, which is thus fairly large compared to the chord length.21 The flow visualization was limited to values
of Rec lesser than 100,000.

B. Wake Survey
At higher speeds where flow visualization was not possible, wake surveys were made. The results of two of
the high Re test cases are shown in Fig. 7; note that the red curves correspond to ideal profile, while the
blue curves correspond the bumpy profile. The quantity of interest in here is the momentum deficit in the
wake, given by the difference between freestream speed U∞ and velocity at the region of interest u. From
Fig. 2, it can be expected that the ideal or “smooth” E398 profile would start to show better performance at
Re values greater than 0.5 million. This is evidenced by the ideal profile showing slightly lesser momentum
deficits than the bumpy profile at an Re value of 0.25 million (Fig. 7(a)). At an Re value of 0.5 million
(closer to the design point of E398), the bumpy profile shows drastic momentum deficits as compared to
the smooth profile (Fig. 7(b)). The latter case approaches the rough airfoil borderline performance limit
suggested by McMasters and Henderson.7

(a) Re = 25 · 104 (b) Re = 50 · 104 .

Figure 7. Momentum deficits.

C. Flowfield Investigation (PIV)

One of the main objectives of this research was in understanding the actual mechanism behind the observed
separation control in the perturbed airfoil. For this purpose, near wall PIV measurements were conducted for
the three models in consideration: (a) ideal (or “smooth”), (b) unibump, and (c) bumpy - profiles (Fig. 4).
To increase the available spatial resolution, several windows were used for the ideal and bumpy profiles ((a)
and (c)), while one window was used to examine the flow close to the bump in profile (b). A summary of
the complete test matrix with the actual zoom window extents can be seen in table 1. The discussion hereon
would be divided into observations in terms of: (a) separation control - limited to ideal and bumpy profiles,
(b) effect of Rec , and (c) effect of pressure gradient (α).

IV. Conclusion
This had been a brief example of some of the more advanced options available for LATEX. Please see the
documentation for each package for extended discussion or usage.

7 of 9

American Institute of Aeronautics and Astronautics

 Figure 8. Wake survey setup.

Table 1. Test Matrix of PIV Runs

Profile Set Zoom Window (%c) Re · 103 α (deg)

Ideal 1 (a) 22 - 48, (b) 13 - 39 18, 36, 54 (a) 7, (b) 10
2 46 - 72 18, 36, 54 7
3 65 - 91 18, 36, 54 7
4 80 - 100 18, 36, 54 7
Bumpy 1 0 - 28 18, 36, 54 7, 10
2 26 - 54 18, 36, 54 7, 10
3 52 - 80 18, 36, 54 7, 10
4 74 - 100 18, 36, 54 7, 10
Unibump - 2.09 - 17.76 18, 36, 54 0, 7
17.26 18, 36, 54 15

8 of 9

American Institute of Aeronautics and Astronautics

 References
1 Mueller, T. and DeLaurier, J., “Aerodynamics of Small Vehicles,” Annual Rev. of Fluid Mech., Vol. 35, 2003, pp. 89–111.
2 Stanewsky, E., “Adaptive Wing and Flow Control Technology,” Progress in Aerospace Sciences, Vol. 37, 2001, pp. 583–
667.
3 Lissaman, P., “Low-Reynolds-Number Airfoils,” Annual Rev. of Fluid Mech., Vol. 15, 1983, pp. 223–239.
4 Carmichael, B., “Low Reynolds Number Airfoil Survey,” NASA CR–165803, 1981.
5 Jacob, J., “On the Fluid Dynamics of Adaptive Airfoils,” Proc. ASME International Mechanical Engineering and Expo-

sition, ASME, Anaheim, CA, 1998.

6 Gad-El-Hak, M., “Flow Control: The Future,” Journal of Aircraft, Vol. 38, 2001, pp. 402–418.
7 McMasters, J. and Henderson, M., “Low-Speed Single Element Airfoil Synthesis,” Technical Soaring, Vol. 6, 1980,

pp. 1–21.
8 Gad-El-Hak, M., “Control of Low-Speed Airfoil Aerodynamics,” AIAA Journal, Vol. 28, 1990, pp. 1537–1552.
9 Jacobs, E., “Airfoil Section Characteristics as Affected by Protuberances,” NACA TR–446, 1932.
10 Jacobs, E., “Wing Characteristics as Affected by Protuberances of Short Span,” NACA TR–449, 1932.
11 Zaman, K. and Potapczuk, M., “The Low-Frequency Oscillation in the Flow Over a NACA 0012 Airfoil with an Iced

Leading Edge,” Low Reynolds Number Aerodynamics, edited by T. Mueller, Springer-Verlag, New York, 1989, pp. 271–282.
12 Bragg, M. B., Cummings, M. J., and Henze, C. M., “Boundary-layer and heat-transfer measurements on an airfoil with

simulated ice roughness,” AIAA Paper 1996–0866, 34th AIAA Aerospace Sciences Meeting, Reno, NV, Jan. 1996.
13 Lyon, C. A., Selig, M. S., and Broeren, A. P., “Boundary layer trips on airfoils at low Reynolds numbers,” AIAA Paper

1997–0511, 35th AIAA Aerospace Sciences Meeting, Reno, NV, Jan. 1997.
14 Kerho, M. F. and Bragg, M. B., “Airfoil Boundary Layer Development and Transition with Large Leading-Edge Rough-

ness,” AIAA Journal, Vol. 35, No. 1, 1997, pp. 75–84.

15 Usui, M., Simpson, A., Smith, S., and Jacob, J., “Development and Flight Testing of a UAV with Inflatable-Rigidizable

Wings,” AIAA Paper 2004-1373, 42nd AIAA Aerospace Sciences Meeting, Reno, NV, Jan. 2004.
16 Huebsch, W., “Dynamic Surface Roughness for Aerodynamic Flow Control,” AIAA Paper 2004–0587, 42nd AIAA

Aerospace Sciences Meeting, Reno, NV, Jan. 2004.

17 Miklosovic, D., Murray, M., Howle, L., and Fish, F., “Leading-Edge tubercules delay stall on humpback whale (Megaptera

novaeangliae) flippers,” Phys. Fluids, Vol. 16, No. 5, May 2004, pp. 39–42.
18 Batill, S. and Mueller, T., “Visualization of Transition in the Flow Over an Airfoil Using the Smoke Wire Technique,”

AIAA Journal, Vol. 19, 1981, pp. 340–345.

19 Sholl, M. and Savas, O., “A Fast Lagrangian PIV Method for Study of General High-Gradient Flows,” AIAA Paper

1997–0493, 35th AIAA Aerospace Sciences Meeting, Reno, NV, Jan. 1997.
20 Tsuei, L. and Savas, O., “Treatment of Interfaces in Particle Image Velocimetry,” Experiments in Fluids, Vol. 29, 1999,

pp. 203–214.
21 Braslow, A., Hicks, R., and Harris, R., “Use of Grit-Type Boundary-Layer-Transition Trips on Wind Tunnel Models,”

NASA TN D-3579, 1966.

9 of 9

American Institute of Aeronautics and Astronautics