Optics and Photonics Journal, 2013, 3, 233-239
http://dx.doi.org/10.4236/opj.2013.33038 Published Online July 2013 (http://www.scirp.org/journal/opj)
Laminar-Turbulent Boundary Layer Transition Imaging
Using IR Thermography
Brian K. Crawford, Glen T. Duncan Jr., David E. West, William S. Saric
Department of Aerospace Engineering, Texas A&M University, College Station, USA
Email: crawford.briank@tamu.edu
Received May 1, 2013; revised June 1, 2013; accepted June 8, 2013
Copyright © 2013 Brian K. Crawford et al. This is an open access article distributed under the Creative Commons Attribution Li-
cense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
ABSTRACT
Experimental techniques for imaging laminar-turbulent transition of boundary layers using IR thermography are pre-
sented for both flight and wind tunnel test environments. A brief overview of other transition detection techniques is
discussed as motivation. A direct comparison is made between IR thermography and naphthalene flow visualization. A
technique for obtaining quan titative transition location is presented.
Keywords: IR Thermography; Laminar; Turbulent; Transition; Flight Test; Wind Tunnel; Fluid Flow
1. Introduction
Over the past several years, laminar flow research has
been ongoing at the Texas A&M Flight Research Labo-
ratory (FRL) using both the Swept Wing In Flight Test-
ing (SWIFT) and Swept Wing In Flight Testing Excres-
cence Research (SWIFTER) models. Both models were
designed to be mounted to the hardpoint of a Cessna O-
2A Skymaster aircraft, and use IR thermography as the
primary measurement technique for laminar-turbulent
transition imaging. Figures 1 and 2 show SWIFTER
mounted under the port wing. SWIFT has the same outer
dimensions and is mounted on the same pylon when it is
flown.
IR thermography exploits the fact that when there ex-
ists a temperature differential between the surface of an
airfoil and the ambient fluid, the turbulent region will
tend to equalize to the ambient temperature faster due to
its higher shear stress. This difference in temperature can
be imaged using IR thermography [1-3]. The two models
differ in how they achieve this differential. SWIFT is
cooled by cold soak ing at 3200 m MSL until it equalizes
to ambient temperature. The aircraft then descends rap-
idly through warmer air. SWIFTER, on the other hand,
has an internal heating sh eet that is used to warm the sur-
face above ambient. As such, SWIFTER does not require
a high-altitude cold soak. Additionally, since SWIFTER
does not rely on an externally applied temperature dif-
ferential, it is also capable of utilizing IR thermography
when installed in a wind tunnel which operates under
adiabatic conditions. An example of SWIFTER in the
Klebanoff-Saric Wind Tunnel (KSWT) is shown in Fig-
ure 3.
Figure 1. Cessna O-2A Skymaster with SWIFTER on the port outboard hardpoint with the IR camera port circled in green.
C
opyright © 2013 SciRes. OPJ
B. K. CRAWFORD ET AL.
234
Figure 2. SWIFTER as viewed from the pilot.
U
Figure 3. SWIFTER Mounted in the KSWT along with the
2. Motivation
a significant portion of a transport air-
ent transition de-
te
su
local pressure
flu
sualization involves apply-
in
pl
t can be used to image shear stress
ov
ary, most existing techniques for boundary
la
3. Theory
fferences in shear stress, laminar and turbu-
exploited experimentally by utilizing a
he
SC8100 IR Camera.
Laminarization of
craft wing would result in an estimated overall drag sav-
ings of 15% [4]. This would result in considerable fuel
savings for any such aircraft; however, in order to make
reliable laminar flow a reality, it must be studied and
quantified in a laboratory environment.
Several techniques for laminar-turbul
ction have been developed and implemented over the
last several decades. Some of the more common tech-
niques include hot-film anemometry, surface-mounted
microphones, sublimation based flow visualization, shear
sensitive paint, and oil-film interferometry. A review of
options for laminar-turbulent transition is given in [5].
Hot-film anemometry provides local shear stress mea-
rements. These local shear stress measurements can be
used to detect whether flow is laminar or turbulent over
the sensor as shown in [6]. However, to image an entire
transition front, many sensors are needed. This becomes
unfeasible when mm accuracy is desired.
Surface-mounted microphones measure
ctuations. Turbulent flow can be detected by a large
rise in the broadband oscillations in pressure. However,
much like hot-films, the number of microphones required
to image an entire transition front to sub centimeter ac-
curacy becomes infeasible.
Sublimation based flow vi
g a coating of naphthalene or similar chemical to the
surface under test and allowing the chemical to sublimate
off during the experiment. The turbulent region subli-
mates much faster than the laminar region, and as such,
the resulting pattern of remaining naphthalene indicates
what part of the flow is laminar. This technique can pro-
vide excellent images, but requires complete sublimation
and re-application fo r every test point. Referen ce [7] pro-
vides examples of this technique in a flight environment.
For oil film interferometry, multiple drops of oil are
aced on the model. When the model is brought up to
speed, the oil shears proportionally with the local shear
stress. The shearing of the oil is then measured optically,
and the variation in shear can be used to detect regions of
laminar vs. turbulent flow. The down side is that this
technique relies on conditions being constant for the du-
ration of the test, and must be re-applied for every test.
Additionally, oil residue can pose an issue in a clean
tunnel environment. Furth er details on this technique can
be found in [8].
Shear sensitive pain
er the entire surface. However, shear sensitive paints
that are capable of measuring relatively low shear stress
are very fragile, and must be re-applied somewhat fre-
quently. As shear sensitive paint technology improves,
this may become a practical technique for imaging lami-
nar-turbulent transition in many environments. Reference
[9] provides more information on the use of shear sensi-
tive paint.
In summ
yer transition have limited fidelity, as is the case with
hot-films and surface microphones or require extended
time on very stable conditions, naphthalene flow visuali-
zation and oil film interferometry. In order to detect tran-
sition fronts rapidly with high fidelity, a technique such
as infrared thermography is required.
Because of di
lent boundary layers have markedly different convection
coefficients. As such, a heated surface will cool much
faster under the influence of a turbulent boundary layer
than under a laminar boundary layer; visa-versa for a
cooled surface.
This can be
ated or cooled substrate with a thin, insulating exterior
coating. Since the coating is heated or cooled at a rea-
Copyright © 2013 SciRes. OPJ
B. K. CRAWFORD ET AL. 235
sonably uniform rate from below by the substrate, but is
under the influence of a very discontinuous convection
rate from above, a strong temperature gradient is devel-
oped at the point of transition. This gradient can then be
imaged via the use of an IR camera. A notional diagram
of this process is shown in Figure 4.
4. Experimental Setup
concerned, the experiment
4.1. Cooled Model
from milled aluminum. The model
As far as IR thermography is
requires three main components; an IR camera, a heated
or cooled model, and a suitable thin, insulating coating
on the model.
SWIFT is constructed
itself has no specific provisions for cooling. Instead, the
model is cold soaked at 3200 m MSL until the IR image
shows that it has equalized to the local atmospheric tem-
perature. This typically takes around 20 minutes. Once
the model is cool, the aircraft then rapidly descends
through the warmer air at lower altitudes. At a 300 - 600
m/min descent rate, the standard lapse rate of the atmos-
phere is sufficient to induce the necessary temperature
differential. However, the necessary descent rate will
vary based on model construction and local atmospheric
conditions. Reference [10] contains addition information
about the SWIFT model.
4.2. Heated Model
Figure 4. Construction of heated model (not to scale) (top)
Temperature distribution ne ar transition (bottom).
nstructed from milled aluminum. SWIFTER is also co
Due to structural considerations, the model has two dis-
tinct regions of wall thickn esses; a forward an d aft region .
In the forward region, the skin is between 12.7 mm and
25.4 mm. In the aft region, the skin is 3.2 mm thick. Thus,
the heating sheet for the model is split into two regions,
one forward and one aft, which run on separate control
loops in order to keep the two regions matched in tempe-
rature. The forward and aft heating sheets have a maxi-
mum power of 200 W (792 W/m2) and 300 W (350 W/m2),
respectively.
The heating sheet itself was constructed using pre-
sheathed heating wire. This particular wire is typically
used to warm tile floors in houses from below. The heat-
ing wire is bonded onto the inside of the test surface of
the model via RTV silicone, as shown in Figure 5. In the
forward region, the wire is bonded in at 12.7 mm spacing,
while the aft region is spaced at 25.4 mm. It is covered
by a layer of 12.7 mm polyurethane foam to help direct
as much of the heat as possible through the test surface.
The temperature controller is connected to a pair of
surface mounted resistance temperature detectors (RTDs)
placed on the inside surface of the test section inside
each of the two heating sheets. There is also a reference
RTD placed on the inside of the unheated side of the
model. During a test, the controller is generally set to
drive the test side RTDs to 5˚C higher than the reference
RTD. Since these measurements are on the inside of the
model, they do not match the external surface tempera-
tures exactly. However, they do maintain a surface tem-
perature that is consistent between the two regions, and
nominally a constant offset from the atmospheric tem-
perature. For this imaging technique, the exact tempera-
Figure 5. Heating wire (blue) bonded into SWIFTER usin
RTV (red) with the insulation (black) pulled back. g
Copyright © 2013 SciRes. OPJ
B. K. CRAWFORD ET AL.
236
ture offset is not important, as long as the model is a few
degrees warmer than ambient. Additional details about
the SWIFTER model can be found in [11].
4.3. IR Camera
Early experiments with SWIFT utilized a FLIR SC3000
IR camera. The SC3000 has a 320 × 240 pixel sensor
with a maximum frame rate of 60 Hz. It operates in the 8
- 9 micron wavel ength band.
Later experiments with SWIFT, and all experiments
with SWIFTER utilize a FLIR SC8100 IR camera. The
SC8100 has a 1024 × 1024 pixel sensor capable of sam-
pling at 132 Hz. It has a temperature resolution of better
than 25 mK, and oper ates in the 3 - 5 micron wavelength
band. For these experiments, the frame rate is typically
set to 20 Hz in order to reduce data overhead. Addition-
ally, the system does not respond rap idly enough to war-
rant a higher frame rate. Figure 6 shows a comparison
between the two cameras.
In the flight environment, the camera is mounted fac-
ing outwards through an opening in the rear port window,
as seen just above the port wing strut in Figure 1. In the
U
U
wind tunnel, the camera is mounted facing through an
opening in the test section viewing window, as shown in
Figure 3. In both cases, an opening in the window is
utilized rather than an IR transparen t window. This is due
to the substantial cost of an IR transparent window large
enough to fit the camera lens.
4.4. Paint Selection
Figure 6. SWIFT using the SC8100 w/17 mm lens (top) vs.
the SC3000 w/17 mm lens (bottom).
ly, the model must be coated with
phy has been used in three principle con-
iscussed in Sec-
As mentioned previous
a thin insulating layer in order to hold a strong tempera-
ture gradient. Additionally, the coating must have as low
of a reflectivity in the IR band as possible. This is par-
ticularly important in flight where the exhaust flare from
the front engine and the warm earth reflect brightly in the
image otherwise. In order to meet these parameters,
Sherwin Williams F93 mil-spec aircraft paint in luster-
less black is currently used on SWIFTER.
SWIFT has a flat black powder coated surface applied
over aluminum. This coating holds a gradient well, but is
fairly reflective in the IR band.
Initially SWIFTER had a flat black powder coat as
well. However, that particular coating was not thick en-
ough to hold a strong temperature gradient at the surface;
as such, the images were completely washed out. Addi-
tionally, the powder coat was very reflective in the IR
band. The new aircraft paint is very flat in the IR band.
The new paint was applied over the existing powder coat
to a total thickness on the order of 300 microns. The
thicker paint results in a much crisper image.
5. Results
IR thermogra
figurations at the TAMU FRLand KSWT. The first con-
figuration used SWIFT as a cooled flight model. This
was followed recently by SWIFTER as a heated flight
model as well as a heated wind tunnel model.
5.1. Cooled Flight Model (SWIFT)
Cooling the model via cold soaking, as d
tion 4.1, results in images such as shown in Figures 6
and 7, where the laminar region is cooler (darker), and
the turbulent region is warmer (lighter). Figure 6 shows
a comparison between the SC8100 and the SC3000. The
extra resolution is capab le of showing finer details of the
transition front. When the 50 mm lens is installed on the
SC8100 and aimed at the lower forward region, as shown
in Figure 7, even more detail is apparent, specifically a
much finer wedge structure. In all SWIFT images, one
spanwise and three chordwise darker regions are visible.
These correspond to the internal structural members of
the model, which do not heat up as quickly as the skin.
Additionally, there is a pair of curved lines correspond-
ing to the reflection of the fuselage and radio antenna in
Copyright © 2013 SciRes. OPJ
B. K. CRAWFORD ET AL. 237
Figure 7. SWIFT using the SC8100 w/50 mm lens.
the mular
owder coat.
oak is that it does not require any addition
eq
sadvantage.
only neces-
d Tunnel Model (SWIFTER)
Since SWIFTER can be mounted in the wind tunnel, it
odel. This is due to the reflectivity of the partic
pThe primary advantage to cooling a model via high-
altitude cold s
uipment or coatings on the model, other than a suitable
paint job. However, every set of tests requires a 20 min-
ute cold soak. Given a typical descent rate during a test, a
high speed descent from 3200 m to 900 m allows for
around 5 - 10 test points before the model must be cold
soaked again depending on atmospheric conditions. For a
typical sortie with an O-2A Skymaster, two dives are
practical, res ulting in 10 - 20 test poi nt s pe r fl ight.
5.2. Heated Flight Model (SWIFTER)
Actively heated models do not share this di
For a heated model, a high speed descent is
sary in order to hit high speed test points. Additionally,
no appreciable recycle time is required between dives. As
such, 6 dives of 5 - 10 test points are typical. This resu lts
in 30 - 60 test points per flight. The only real disadvan-
tage is that the extra complexity of a heating sheet must
be integrated into the model. However, compared to the
cost of flight time saved by increased testing efficiency,
this additional start-up cost is quickly recovered. An
example image is shown in Figure 8. Note that when the
model is heated, the image coloration is inverted com-
pared to the cooled case. In Figure 8, the vertical seam
between the forward and aft regions is visible due to
transient effects.
5.3. Heated Win
U
U
Figure 8. SWIFTER using the SC8100 w/ 50mm lens Full
image (top), and zoomed in on mid-span (bottom).
lization.
provides an opportunity to compare IR thermography to
he more ubiquitous naphthalene flow visuat
Naphthalene performs well at resolving fine structures in
the flow, which IR thermography has not generally been
able to match. However, given a sufficiently high resolu-
tion camera, such as the SC8100 and the appropriate
choice of coating, comparable results can be obtained, as
shown in Figure 9. The fine streaking in the forward
region is clearly visible in both images. One of the dis-
advantages to naphth alene is that it is very susceptible to
the uniformity of the coating applied, as shown by the
waviness in the structures shown in the image. Addition-
ally, the main motivating benefit of IR thermography is
Copyright © 2013 SciRes. OPJ
B. K. CRAWFORD ET AL.
238
U
U
Figure 9. SWIFTER transition fronts in the wind tunnel at
a repeated condition using IR thermography (top), and
naphthalene flow visualization (bottom).
second, while
aphthalene would require a complete re-application. Be-
phy,
e data.
age as
pix-
el
its quick turnaround time. When the transition front
changes, the image responds in under a
n
cause of this fact, an IR thermography test can easily
capture over 200 test points in an hour, while a typical
day of naphthalene runs involves only 3 test points.
5.4. Post-Processing of Transition Images
Once images have been captured via IR thermogra
the next step is to process them into quantitativ
The top panel of Figure 10 shows a typical IR im
filtered and colorized by FLIR’s Examine-IR software.
The outlined area is the region of interest for the experi-
ment. The bottom panel shows the image after coordinate
trans- formation, transition detection, and filtering .
In order to generate this post processed image, the raw
data file is first converted into temperature values, and
dead pixels are corrected. Any remaining hot or cold
s are then corrected in turn. The columns of the image
are then normalized with respect to their neighbors in
order to remove a large portion of the sensor noise in-
herent to the construction of this particular type of cam-
U
x/c [%]
Median
Median
MAD
U
Figure 10. FLIR filtered image (top) compared to post-
processed image (bottom).
t regions mentioned earlier.
hese processes result in a reasonably clean image,
era. This also corrects for the temperature differential
between the forward and af
T
which is then cropped and transformed to make vertical
lines constant in chord and horizontal lines constant in
span. The image is then duplicated into a pair of images.
The second image is then spatially low-passed and sub-
tracted from the first. This filtering removes any slow-
varying features such as structural members, non-uni-
form heating due to internal electronics and aircraft ex-
haust flare. The resulting image is shown as the green
channel in the final image.
The gradient vector of this image is then calculated at
every pixel and then dotted with the vector of character-
Copyright © 2013 SciRes. OPJ
B. K. CRAWFORD ET AL.
Copyright © 2013 SciRes. OPJ
239
edges propagate along. This
st
of
obust method for
rbulent transition fronts in a wid
s without perturbing the flow.
e Air Force Re-
) through General Dynam-
as provided by AFOSR, the
REFERENCES
[1] S. Zuccher ared Thermography
Investigationsnic Boundary Lay-
istic lines that the turbu lent w
rongly selects for the wedge structure while suppress-
ing spurious noise and non-transition related features,
such as those caused by sun/shadows on the model. A
first guess is then made at the transition front by picking
the location with the largest result of this dot product at
each row in the image. This guess is then used to narrow
the region of the image where transition is believed to be
occurring. The reduced region is then re-evaluated,
searching for the strongest gradient. The process is re-
peated until this region is collapsed into a line, which
corresponds to the center of the high gradient region.
This line is then used to separate the image into laminar
and turbulent regions. The lamin ar region is then colored
red, while the turbulent region is colored blue. This,
when combined with th e earlier mentioned green chan nel,
results in the post processed image shown in Figure 10.
Finally, a chord scale is added to the image in black,
the median transition location is marked in purple, and
the median minus the median absolute deviation (MAD)
the transition location is marked in orange. All values
are presented in percent chord location. This provides a
robust and repeatable metric to quantify the transition
location that is not subject to the bias of a user manually
tracing transition fronts. Work is currently ongoing to
obtain additional quantitative data from these images
such as frequency content of the chordwise streaking due
to the influence of crossflow vortices.
6. Conclusion
IR thermography provides a fast and r
imaging laminar-tu
riety of environmente va-
The
quality of the results is on par or better than the industry
standard of naphthalene flow visualization, and requires
only a small fraction of the time. Additionally, IR ther-
mography responds to ch anging conditions in well under
a second. Lastly, IR thermography data can be post pro-
cessed into quantitative transition fronts.
7. Acknowledgements
Funding for SWIFTER is provided by th
search Laboratories (WPAFB
ics IT. Funding for SWIFT w
AFRL under the AEI program, and the Northrop Grum-
man Corporation. The authors would like to acknowl-
edge our test pilots, Roy Martin, Dr. Donald Ward, Dr.
Celine Kluzek, Lee Denham, and Lt Col Aaron Tucker;
the staff of the Oran W. Nicks Low Speed Wind Tunnel,
the Klebanoff-Saric Wind Tunnel, and the Texas A&M
Flight Research Lab; and our A&P Mechanic, Cecil
Rhodes.
nd W. S. Saric, “Infra
in Transitional Superso
ers,” Experiments in Fluids, Vol. 44, 2008, pp. 145-157.
doi:10.1007/s00348-007-0384-1
[2] W. S. Saric, H. L. Reed and D. W. Banks, “Flight Testing
of Laminar Flow Control in Hig
h-Speed Boundary Lay-
oughness
-Experiments,”
ers,” The RTO Applied Vehicle Technology Panel (AVT)
SpecialistsMeeting, Prague, 4-7 October 2004.
http://www.cso.nato.int/Main.asp?topic=5
[3] A. L. Carpenter, W. S. Saric and H. L. Reed, “R
Receptivity in Swept-Wing Boundary Layers
International Journal of Engineering Systems Modeling
and Simulation, Vol. 2, No. 9, 2010, pp. 128-138.
doi:10.1504/IJESMS.2010.031877
[4] D. Arnal and J. P. Archambaud, “Laminar-Turbulen
Transition Control: NLF, LFC, HLt
FC,” Advances in La-
Reed, “Investigation of a Health-
-
gs of
minar-Turbulent Transition Modeling, VKI Lecture Se-
ries, Brussels, 2008.
[5] D. N. Mavris, W. S. Saric, H. Ran, M. J. Belisle, M. J.
Woodruff and H. L.
Monitoring Methodology for Future Natural Laminar
Flow Transport Aircraft,” ICA S Paper 1.9.3, Nice, 2010.
[6] K. L. Chapman, M. S. Reibert, W. S. Saric and M. N.
Glauser, “Boundary-Layer Transition Detection and Struc-
ture Identification through Surface Sheer-Stress Meas-
urements,” Proceedings of the 36th AIAA Aerospace Sci-
ences Meeting and Exhibit, Reno, 12-15 January 1998.
[7] A. Ahmed, W. H. Wentz Jr. and R. Nyenhuis, “Natural
Laminar Flow Flight Experiments on a Swept-Wing Bu
siness Jet,” Proceedings of the AIAA 2nd Applied Aero-
dynamics Conference, Seattle, 21-23 August 1984.
[8] A. Drake, “Oil Film Interferometry for Boundary Layer
Measurements in Aircraft Development,” Proceedin
the 24th AIAA Aerodynamic Measurement Technology
and Ground Testing Conference, Portland, 28 June-1 July
2004. doi:10.2514/6.2004-2114
[9] M. McQuilling, M. Wolff, S. Fonov, J. Crafton and R.
Sondergaard, “An Experimental Investigation of Low-
icron-Sized Discrete
f Step Excrescences on
Pressure Turbine Blade Suction Surface Stresses Using
S3F,” Proceedings of the 44th AIAA Aerospace Sciences
Meeting and Exhibit, Reno, 2006.
[10] A. L. Carpenter, “In-Flight Receptivity Experiments on a
30-Degree Swept-Wing Using M
Roughness Elements,” Ph.D. Thesis, Texas A&M Uni-
versity, College Station, 2009.
[11] G. T. Duncan Jr., B. K. Crawford and W. S. Saric, “Flight
Experiments on the Effects o
Swept-Wing Transition,” Proceedings of the 48th Applied
Aerodynamics Symposium, Saint-Louis, 25-27 March 2013.