Research Article Vol. 55, No.

34 / December 1 2016 / Applied Optics D1

Pulsed micro-laser line thermography on

submillimeter porosity in carbon fiber reinforced
polymer composites: experimental and numerical
analyses for the capability of detection
HAI ZHANG,1,* HENRIQUE FERNANDES,1,2 FRANK BILLY DJUPKEP DIZEU,1 ULF HASSLER,3 JULIEN FLEURET,1
MARC GENEST,4 CLEMENTE IBARRA-CASTANEDO,1 FRANÇOIS ROBITAILLE,5 SIMON JONCAS,6 AND
XAVIER MALDAGUE1
1
Department of Electrical and Computer Engineering, Computer Vision and Systems Laboratory, Laval University, 1065 Av. de la Médecine,
Quebec G1 V 0A6, Canada
2
Department of Mechanical Engineering, Federal University of Uberlandia, 2121 Av. Joao Naves de Avila, Uberlandia 38400-902, Brazil
3
Fraunhofer Development Center X-Ray Technologies (EZRT), Department of Application Specific Methods and Systems (AMS), Fraunhofer IIS,
Flugplatzstrasse 75, 90768 Fuerth, Germany
4
Aerospace Portfolio, Structures, Materials and Manufacturing, National Research Council Canada, 1200 Montreal Road, Ottawa K1A 0R6,
Canada
5
Department of Mechanical Engineering, University of Ottawa, 161 Louis Pasteur, Ottawa K1N 6N5, Canada
6
Department of Automated Manufacturing Engineering, École de Technologie Supérieure, 1100 Rue Notre-Dame Ouest, Montreal H3C 1K3,
Canada
*Corresponding author: hai.zhang.1@ulaval.ca

Received 3 June 2016; revised 5 August 2016; accepted 5 August 2016; posted 8 August 2016 (Doc. ID 267654); published 12 September 2016

In this article, pulsed micro-laser line thermography (pulsed micro-LLT) was used to detect the submillimeter
porosities in a 3D preformed carbon fiber reinforced polymer composite specimen. X-ray microcomputed tomog-
raphy was used to verify the thermographic results. Then, finite element analysis was performed on the corre-
sponding models on the basis of the experimental results. The same infrared image processing techniques were
used for the experimental and simulation results for comparative purposes. Finally, a comparison of experimental
and simulation postprocessing results was conducted. In addition, an analysis of probability of detection was
performed to evaluate the detection capability of pulsed micro-LLT on submillimeter porosity. © 2016
Optical Society of America

OCIS codes: (120.4290) Nondestructive testing; (110.3080) Infrared imaging; (140.3070) Infrared and far-infrared lasers.

http://dx.doi.org/10.1364/AO.55.0000D1

1. INTRODUCTION composites due to its advantages of fast detection, cheap equip-

Along with the development of composite materials, micro-
sized defects in carbon fiber reinforced polymer composites
(CFRP) are becoming a new challenge for quality control, es-
pecial in porosity and cracks. Therefore, the detection of sub-
millimeter porosity in CFRP is increasingly essential [1,2].
Compared to microscopic inspection, which is the most
often used in composite defect detection, nondestructive testing
(NDT) techniques have the advantages of nondestruction, larger-
area inspection, fast detection, and effectiveness on subsurface
defects [3]. X-ray microcomputed tomography (micro-CT) is
an often used NDT technique for composites [4]. However,
micro-CT equipment is usually expensive, energy hungry, and
difficult to move for in-line detection [5,6]. Infrared thermogra-
phy (IRT), as a NDT technique, is increasingly used for
ment, simple setup, and mobility for in-line detection [7–10].
The IR camera is the most essential apparatus for IRT. A fast-
cooled IR camera is usually expensive (e.g., the latest version
FLIR SC7000 of the camera used in this article is $100,000 this
year). However, the price decreases rapidly year by year, and an
IR camera without the fast-cooled function already works for
a majority of applications (e.g., FLIR S65HS is just $6000
this year) [11]. Other apparatuses used for IRT are cheaper
(e.g., the fiber-coupled laser diode system used in this article
was $10,000 in 2005, and a flash lamp used for IRT is just
$150 this year) [12]. Therefore, infrared thermography has a
significant price advantage.
Flying spot laser thermography uses a laser, which is scanned
over the surface. Changes in the heat conductivity then lead to

1559-128X/16/3400D1-10 Journal © 2016 Optical Society of America

 D2 Vol. 55, No. 34 / December 1 2016 / Applied Optics Research Article

changes in the thermal footprint, which is used for crack de-

tection. This method was suggested for the first time in the late
1960s [13]. A good overview about crack detection with laser-
excited thermography can be found elsewhere [14]. Rashed et al.
[15] first tried to detect cracks with laser-excited thermography
using modern spatial-resolving IR cameras. The influences of
crack opening and depth to the measurement results were
discussed using the 2D finite element method (FEM). On
the basis of this work, a combination of thermographic and
ultrasonic techniques, using the laser for both thermal and
acoustical excitation, was suggested [16,17].
IRT on subsurface submillimeter flaw detection in compo-
sites was poorly documented in the literature, which may be
due to the IR camera resolution limitation and the difficulty
for the heat to transmit through composites [18]. Large-sized
porosities in composites can be detected using classical IRT
techniques including pulsed thermography and lock-in ther-
mography [1,2]. However, these techniques cannot detect sub-
millimeter porosity because the distance between the specimen
and the camera is usually small (approximately 4.5 cm) when
using a microlens, which is difficult for the experimental con-
figurations. Flying spot laser thermography used to detect the
subsurface submillimeter flaws in previous composites has been
reported [1,2]. However, the results revealed that the technique
cannot work on micro-sized flaws. The authors proposed a new
micro-laser line thermography (micro-LLT) technique [1],
which was verified to be suitable for subface submillimeter
flaws in composites [2]. Fig. 1. (a) Complete stitched T-joint sample, and (b) flat surface
In this article, pulsed micro-laser line thermography (pulsed view.
micro-LLT) was used to detect a 10 mm × 152 mm zone of a
3D preformed CFRP. Micro-CT slices and pulsed micro-LLT
images were compared to verify the thermographic results. the fiber insertion process was completed, the resin infusion
Cold image subtraction (CIS) and principal component ther- process was initiated.
mography (PCT) were used for the image processing of the The specimen is shown in Fig. 1. It contains six stitching
thermographic results. Then, the corresponding models were lines. The purpose of the stitching is to consolidate the T-joint
built and finite element analysis (FEA) was performed. The structure and reduce dry core. The sample measures 152 mm in
same image processing techniques were used for the corre- length, 148 mm in width, 63 mm in height, and 5 mm in
sponding simulation for comparative purposes, which was thickness (excluding the T-stringer), as shown in the figure.
poorly documented in the literature. Finally, an experimental A 10 mm × 152 mm zone, which is shown in Fig. 1(b), was
and simulation comparison of pulsed micro-LLT was con- detected using pulsed micro-LLT and micro-CT.
ducted. In addition, a probability of detection (PoD) analysis B. Pulsed Micro-laser Line Thermography
on the basis of the hit/miss method was performed to evaluate The authors proposed micro-laser line thermography (pulsed
the detection capability of pulsed micro-LLT. micro-LLT) [1], which was shown to function well for subsur-
face submillimeter flaws in composites [2]. In this article,
pulsed micro-LLT was used to detect the subsurface submillim-
2. MATERIALS AND METHODS eter porosities in a 3D preformed CFRP specimen for the
A. Specimen evaluation of detection capabilities.
The specimen selected for evaluation was a T-joint 3D pre- Figure 2(a) shows the experimental setup for the pulsed micro-
formed CFRP. It was sewn using stacked TC-06-T 3K carbon LLT. In the setup, a midwave IR camera “FLIR Phoenix” at a
fiber. The 3D architecture was woven using 3K/12K carbon frame rate of 55 fps (640 × 512 pixels) was used to record the
fiber. A continuous row of stacked 12K tow fiber was used temperature profiles. A microlens with a magnification of 1 ×
for insertion. A toughened epoxy resin infusion system was was mounted on the IR camera to identify and characterize
selected. The specimen was fabricated using 3D preform con- the porosities. Table 1 shows the IR camera technical specifica-
sisting of multiple layers of woven fabric. The noodle for the tions. A diode laser with a wavelength of 805 nm was used as the
T-joint insertion was preshaped through compaction. During heating source. A convex lens was used to focus the laser beam,
processing, the twisted round-shaped stacked 12K carbon fiber and a waveform generator was used to generate the laser line.
tows were placed into the molding tool and compacted to a A beam expander and cylindrical lens were already used to
triangular shape as the tool was bolt clamped together. After convert a laser spot with a radius of around 0.9 mm to a laser
 Research Article Vol. 55, No. 34 / December 1 2016 / Applied Optics D3

(approximately 4.5 cm). Therefore, a galvanometer scanning

mirror with a frequency of 600 Hz was used to generate the
laser line in this article. Figure 2(b) shows the pulsed micro-
LLT schematic setup. The position of the laser line can be
changed by modulating the mirror angle, which avoids block-
ing the capture of the camera [Fig. 2(a)]. The laser line is ap-
proximately 10 mm in length and 3 mm in width, which has a
power of 2.9 W measured by the laser power meter. The heat-
ing time (pulse length) is 0.5 s, and the acquisition time is
5 s (≈300 frames).

C. Infrared Image Processing

Infrared image processing is becoming increasingly essential
for IRT. In this article, different image processing techniques
were used.
1. Cold Image Subtraction
Cold image subtraction is intended to reduce the effects of fixed
artifacts in a thermographic sequence [20], for example, reflec-
tions from the environment such as residual heating coming
from the lamps and even the reflection from the camera that
appears during the acquisition. Since these artifacts are more or
less constant during the whole acquisition, including before
heating when the image is cold, this image or the average of
several images can be subtracted before heating, so their effect
can be reduced [2]. Figure 3 shows an example with an aca-
demic aluminum plate.
In Fig. 3, the cold image is affected by noise due to a non-
uniform correction (NUC) that was badly performed. This
happens when an old NUC is used. The noise can be seen
in the raw sequence as well (the 2nd image in Fig. 3), although
Fig. 2. Pulsed micro-LLT setup: (a) the experimental setup, and it is less evident since the temperature is higher.
(b) the schematic setup. CIS can be considered as a preprocessing step to improve the
quality of the sequence. In the subsequent processing, more
advanced algorithms, such as phase pulse thermography (PPT),
line in the literature [19]. However, this setup does not func- etc., are used.
tion properly for micro-sized flaws because the cylindrical lens
will block the capture of the camera, which is caused by
2. Principal Component Thermography
the small distance between the camera and the specimen
Principal component thermography, originally proposed by
Rajic in 2002 [21], extracts the image features and reduces un-
Table 1. FLIR Phoenix (MWIR) Technical Specifications desirable signals. It relies on singular value decomposition,
which is a tool to extract spatial and temporal data from a
Technical Specification Explanation/Value
matrix in a compact manner by projecting original data onto
Detector type indium antimonide (InSb) a system of orthogonal components known as empirical
Spectral range 1.5–5.0 μm orthogonal functions (EOFs). The first EOF will represent
Cold filter bandpass 3.0–5.0 μm standard
the most important characteristic variability of the data; the
Pixel resolution 320 (H) × 256 (V) pixels or
640 (H) × 512 (V) pixels
Detector size 30 × 30 μm for 320 × 256
25 × 25 μm for 640 × 512
Well capacity 18 M electrons for 320 × 256
11 M electrons for 640 × 512
Integration time 320 × 256:9 μs to full frame time
640 × 512: < 50 μs to full frame time
Sensor assembly f ∕# f ∕2.5 standard, f ∕4.1 optional
Sensor cooling stirling closed cycle cooler
Spec. performance <25 mK
Max. frame rate with 320 × 256:345 fps in full frame
DTS electronics 640 × 512:100 fps in full frame
Max. frame rate with 320 × 256:120 fps in full frame
Fig. 3. Example with an academic aluminum plate to explain cold
RTIE electronics 640 × 512:30 fps in full frame
image subtraction.
 D4 Vol. 55, No. 34 / December 1 2016 / Applied Optics
second EOF will contain the second most important variability,
and so on. Usually, original data can be adequately represented
with only a few EOFs. Typically, an infrared sequence of 1000
images can be replaced by 10 or less EOFs [22].
3. ANALYSIS OF THE RESULTS
A. Experimental Results
In this article, the thermographic results were compared to the
corresponding micro-CT slices for verification. The zone of in-
terest for pulsed micro-LLT analysis is the laser line coverage
zone, which approximately has a 3 mm × 10 mm area.
Figure 4 shows the comparison of the micro-CT slices and
the pulsed micro-LLT results from the position of 69 mm,
Fig. 4. Comparison of micro-CT and pulsed micro-LLT results
from the position of 69 mm: (a) micro-CT slice on the surface,
(b) micro-CT slice from the depth of 54 μm, (c) pulsed micro-
LLT image after CIS, and (d) pulsed micro-LLT image after PCT.
Research Article
which is the distance from the porosity to the top of the speci-
men [Fig. 1(b)].
Figure 4(a) shows the surface slice from the micro-CT.
Figure 4(b) shows the fourth slice of the same detected zone
from micro-CT, which is 54 μm in depth. Porosities A (marked
in red) can be detected by pulsed micro-LLT after CIS [shown
in Fig. 4(c)] and PCT [shown in Fig. 4(d)]. PCT provides a
clearer result than CIS.
Figure 5 shows the comparative results from the position of
108 mm. Porosity B from the depth of 72 μm can be detected
after CIS and PCT using pulsed micro-LLT.
Fig. 5. Comparison of micro-CT and pulsed micro-LLT results
from the position of 108 mm: (a) micro-CT slice on the surface,
(b) micro-CT slice from the depth of 72 μm, (c) pulsed micro-
LLT image after CIS, and (d) pulsed micro-LLT image after PCT.
 Research Article Vol. 55, No. 34 / December 1 2016 / Applied Optics D5

Similarly, porosities C and D can also be detected after CIS

and PCT, which are shown in Figs. 6 and 7, respectively.
In Fig. 6, porosity C is 90 μm in depth, and the correspond-
ing detected zone is 87 mm from the top of the specimen. In
Fig. 7, the detected zone for porosity D is 66 mm from the top
of the specimen, and the depth is 0.108 mm.
Figure 8 shows the comparison of the micro-CT slices and
the pulsed micro-LLT images from the position of 129 mm.
Porosities E from the depth of 36 μm can be detected after
CIS and PCT using pulsed micro-LLT, which is similar to the
previous results. However, porosity F (marked in blue), which
is located at a depth of 0.162 mm, can only be detected after
PCT. Porosities with the depths of 0.18 mm or more cannot be
detected after CIS or PCT. The experimental results from the
position of 141 mm are shown in Fig. 9.

Fig. 7. Comparison of micro-CT and pulsed micro-LLT results

from the position of 66 mm: (a) micro-CT slice on the surface,
(b) micro-CT slice from the depth of 0.108 mm, (c) pulsed micro-
LLT image after CIS, and (d) pulsed micro-LLT image after PCT.

In Fig. 9, porosity G at a depth of 90 μm can be detected

Fig. 6. Comparison of micro-CT and pulsed micro-LLT results
from the position of 87 mm: (a) micro-CT slice on the surface,
(b) micro-CT slice from the depth of 90 μm, (c) pulsed micro-
LLT image after CIS, and (d) pulsed micro-LLT image after PCT.
after CIS and PCT. However, porosity H (marked in blue) at a
depth of 0.18 mm cannot be detected using any of the image
processing techniques. The reason why porosity H cannot be
detected is that the fiber covers this porosity, which blocked the
heating diffusion [23]. Another submillimeter porosity at the
same depth of 0.18 mm, which is not covered by the fiber,
cannot be found in the specimen. However, the experimental
results for porosity F can also indicate that the depth limitation
of the technique is approximately 0.162 mm. FEA on the basis
of the corresponding models, which is useful to verify the
estimation, was performed in this article [23,24].
As a conclusion, the porosities located at depths of 36 μm to
0.108 mm can be detected after CIS and PCT. The porosity at
the depth of 0.162 mm can only be detected after PCT, which
 D6 Vol. 55, No. 34 / December 1 2016 / Applied Optics
Fig. 8. Comparison of micro-CT and pulsed micro-LLT results
from the position of 129 mm: (a) micro-CT slice on the surface,
(b) micro-CT slice from the depth of 36 μm, (c) micro-CT slice from
the depth of 0.162 mm, (d) pulsed micro-LLT image after CIS, and
(e) pulsed micro-LLT image after PCT.
can be considered as the detection limitation. Porosities at a
depth of 0.18 mm and more cannot be detected. PCT shows
clearer results than CIS. A potential way to greatly enhance the
detection capacity is to greatly increase the laser power.
Research Article
Fig. 9. Comparison of micro-CT and pulsed micro-LLT results
from the position of 141 mm: (a) micro-CT slice on the surface,
(b) micro-CT slice from the depth of 90 μm, (c) micro-CT slice from
the depth of 0.18 mm, (d) pulsed micro-LLT image after CIS, and
(e) pulsed micro-LLT image after PCT.
However, there is a risk that the specimen would be damaged.
It is possible to increase the maximum detection depth by
modulating the laser power, which should be on the basis of
different characteristics of materials.
 Research Article Vol. 55, No. 34 / December 1 2016 / Applied Optics D7

B. Finite Element Analysis 2hc qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

Finite element analysis was performed to simulate the corre- z 1
x 
2x − ag 
a − ag − 2x  hy − hc ;
a − 2ag
sponding models on the basis of the position of 129 mm

(Fig. 8, model A) and the position of 141 mm (Fig. 9, ag a
<x< ; (2)
model B). The models were implemented into COMSOL 2 4
Multiphysics 5.0 as the user-defined material models to verify
the nonlinear behavior of heat transmission. Then, the simu- where hy is the yarn thickness, a is the dimension of the
lation data was exported to MATLAB for further image models in either the fill or warp directions (warp or fill faces,
processing. CIS and PCT were also performed on the simula- respectively), ag is the gap width between two adjacent
tion images. Finally, a comparison of experimental and simu- transverse yarns, and
lation results with the same image processing techniques was

 
hy πag
conducted. hc  sin 1 : (3)
2 a
1. Modeling The material properties used for the models are shown in
The models needed for finite element discretization were Table 2 [23].
developed from the micro-CT measurements. Figure 10 shows The zones covered by the laser line in models A and B are
the micro-CT measurements and the corresponding models. shown in Figs. 10(c) and 10(d), respectively. The lengths of the
The geometrical models were developed on the basis of the fibers are 5 mm and 4 mm, respectively, and their width is
method proposed by Ito and Chou [25]. The models consist of 0.8 mm. The thickness of the fibers is 90 μm. The fill yarn
four intertwined yarns surrounded by the isotropic matrix. cross section is an ellipse with major axis 0.8 mm and minor
There are two warp yarns in the longitudinal direction and axis 0.09 mm. The warp yarn curves are described by Eq. (1),
two fill yarns in the transverse direction, respectively. Each yarn where hy  0.1 mm and a  5 mm. The 4 mm fibers were
is a unidirectional composite in the material coordinate system rotated by 20°. Two concave surfaces face each other. The cen-
with orthotropic properties [26]. The equations proposed by ters of mass lay on the z-axis. Two fibers were embedded
Ito and Chou for the yarn geometry were used to develop
the models. The warp yarn path curve on the fill face is
described by Table 2. Material Properties


 Epoxy Fiber Porosity

hy 2π a a
z 0
x  sin x ; − <x< ; (1) Thermal 0.2 {60,4,4} 0.004
2 a 4 4 conductivity (k) W/(m·K) W/(m·K) W/(m·K)
Density (ρ) 1200 kg∕m3 1500 kg∕m3 0
and the fill yarn cross-section curve on the same face by Heat capacity (Cρ ) 1000 J/(kg·K) 1000 J/(kg·K) 1004.5 J/(kg·K)

Fig. 10. Micro-CT measurements and the corresponding models: (a) model A, micro-CT measurements; (b) model B, micro-CT measurements;
(c) model A, topview; (d) model B, topview; (e) model A, sideview, and (f ) model B, sideview.
 D8 Vol. 55, No. 34 / December 1 2016 / Applied Optics
in cubic epoxy with the volumetric dimension of 6 mm ×
2.5 mm × 0.6 mm laid along the x-axis [23]. The parameters
of the submillimeter porosities E, F, G, and H are shown in
Figs. 10(e) and 10(f ).
The laser line power is 2.9 W, which is the same as the laser
line power in the experimental setup. The power density is
2.9 W∕
3 mm × 10 mm. The heating time is 0.5 s, and
the cooling time is 1.5 s. In order to match the IR camera
frame rate (≈55 fps), the simulation frames were set to
0.02 s (50 Hz).
2. Temperature Distributions
The simulation results can reveal the temperature distribution
directly. Figures 11(a) and 11(b) show the mesh of the models.
Figures 11(c) and 11(d) show the surface temperature distribu-
tion at 0.5 s. Porosities E can be identified from the temper-
ature distribution results. However, porosities F, G, and H
cannot be inspected. The reason may be the restricted heating
transmission. Figures 11(e) and 11(f ) show the temperature
distributions from the sideview at 0.5 s, which reveal that it
Fig. 11. (a) Model A, mesh; (b) model B, mesh; (c) model A, surface temperature distribution; (d) model B, surface temperature distribution;
(e) model A, temperature distribution from the sideview; and (f ) model B, temperature distribution from sideview.
Research Article
is difficult to transmit the heating onto porosities F and H.
The temperature distribution images can directly reveal the
trend of the heating transmission. However, it cannot provide
more reliable information for the detection capability. For com-
parative purposes, CIS and PCT were also used for simulation,
which was previously poorly documented in the literature.
3. Postprocessing Results
The simulation data was exported to MATLAB for further
image processing, and the corresponding data structural trans-
formation was performed.
Figures 12(a) and 12(b) show the simulation results of
model A after CIS and PCT, respectively. In Fig. 12(a), poros-
ities E can be clearly inspected. However, porosity F cannot be
identified, which is similar to the corresponding experimental
result after CIS. In Fig. 12(b), porosities E can also be in-
spected. The position of porosity F is abnormal in the simu-
lation image after PCT. However, it is difficult to conclude
that porosity F can be inspected in the image. On the contrary,
porosity F in the corresponding experimental image after PCT
 Research Article Vol. 55, No. 34 / December 1 2016 / Applied Optics D9

Table 3. Thermo-PoD of the Porosities on the Basis of

Hit/Miss Data
Submillimeter Porosity Experiment Simulation
Depth (μm) Size (μm) CIS PCT CIS PCT
36 72 1 1 1 1
54 54 1 1 1 1
72 126 1 1 1 1
90 108 1 1 1 1
108 90 1 1 1 1
162 72 0 1 0 1
180 216 0 0 0 1

Fig. 12. Simulation results of the models A and B after the corre-
sponding image processing: (a) model A, CIS; (b) model A, PCT;
(c) model B, CIS; and (d) model B, PCT.
is clearer. The simulation result may be affected by the model
mesh, which leads to the phenomenon and restricts the usage of
PCT on FEA.
Figures 12(c) and 12(d) show the simulation images of the
model B after image processing. The corresponding experimen-
tal results after image processing reveal porosity G more clearly
than the simulation images. Porosity H cannot be inspected
in the experimental results. However, it can be identified
in Fig. 12(d), which is the corresponding simulation image
after PCT.
C. Thermo-PoD
Probability of detection, as a quantitative reliability measure-
ment method, has been established for decades to evaluate
the detection capability of NDT techniques [27]. It is already
widely used for traditional NDT techniques, but only a brief
study for IRT was reported in the literature [28]. Usually, two
methods can be used to obtain data for the formation of PoD.
One involves a data method that uses quantitative stimulus re-
sponses as the reference. The other method uses hit/miss data,
where “1” indicates that a defect is found, and “0” indicates that
a defect is not found [29]. In this article, a statistical thermo-
PoD on the basis of the hit/miss method was performed, which
is shown in Table 3 [30].
In Table 3, “1” means that the porosity is detected by the
corresponding image processing technique, and “0” means the
corresponding porosity is not detected. Table 3 reveals the de-
tection capability. The minimum size is 54 μm, and the
maximum depth is 0.18 mm.
A thermo-PoD on the basis of simulation data contributes
to the study. However, the mesh of the corresponding models
affects the image processing results severely, which is shown in
Fig. 12. Therefore, a thermo-PoD on the basis of simulation is
unreliable when the detected defects are at a submillimeter
level. The simulation results in Fig. 12 reveal that clearer results
can be obtained by PCT than CIS, but the defect size also
affects the detection capability. Therefore, a thermo-PoD curve
on the basis of both detection depth and size is needed for
further PoD analysis.
The PoD curve is a generally accepted method for character-
izing and quantifying the detection capability and reliability of a
NDT technique, which requires a mass of experimental data
[31]. Therefore, robotic pulsed micro-LLT is necessary, which
will need the corresponding specific image reconstruction al-
grorithm [32]. This provides an orientation for further study.
4. CONCLUSIONS
In this article, the detection capability of pulsed micro-LLT
was studied. The technique was used to detect the subsurface
submillimeter porosities in a 3D preformed CFRP specimen.
Micro-CT was used to verify the thermographic results. CIS
and PCT were used to process the results. Then, a FEA was
performed on the basis of image processing techniques for
comparative purposes, which was previously poorly docu-
mented in the literature. Finally, PoD analysis was performed
on the basis of experimental and simulation postprocessing
results.
As a conclusion, PCT can provide clearer results than CIS.
A FEA on the basis of image processing presents promising re-
sults. However, the mesh of the models affects the processing
results severely, especially for the submillimeter flaw. The de-
tection capability is limited to 54 μm in size and 0.18 mm
in depth.
Lock-in micro-LLT may be able to detect flaws at a deeper
depth [33], which is also useful for PoD analysis. The quanti-
tative PoD curves on the basis of both the response data
method and the hit/miss method are promising and merit
further study. Robotic micro-LLT could greatly reduce the ex-
periment time, which will need the corresponding image
reconstruction algorithms.
Funding. Consortium for Research and Innovation in
Aerospace in Québec (CRIAQ) (COMP-501 - CRDPJ
408010-10); Governments of Québec and Bavaria, Ministère
des Relations Internationales and Ministry of External Affairs
(13.311); Natural Sciences and Engineering Research
Council of Canada (NSERC); Canada Research Chair in
Multipolar Infrared Vision (MiViM); Conselho Nacional
de Desenvolvimento Científico e Tecnológico (CNPq)
(166285/2015-3).
 D10 Vol. 55, No. 34 / December 1 2016 / Applied Optics
Acknowledgment. The authors would like to acknowl-
edge the support provided by the industrial partners: Bell
Helicopter (Canada) Inc., Bombardier Inc., Hutchinson Inc.,
Delastek Inc., Texonic Inc., and the CTT group. The authors
would also like to thank Wolfgang Holub from Fraunhofer
EZRT (Germany) and Lingyao Yu and Yunlong Sheng from
the Department of Physics, Physical Engineering and Optics,
Center for Optics, Photonics and Lasers, Laval University
(Canada) for their assistance.
REFERENCES
1. H. Zhang, M. Genest, F. Robitaille, X. Maldague, L. West, S. Joncas,
and C. Leduc, “Infrared thermography, ultrasound C-scan and micro-
scope for non-destructive and destructive evaluation of 3D carbon
fiber materials: a comparative study,” Proc. SPIE 9485, 94850X
(2015).
2. H. Zhang, U. Hassler, M. Genest, H. Fernandes, F. Robitaille, C.
Ibarra-Castanedo, S. Joncas, and X. Maldague, “Comparative study
on submillimeter flaws in stitched T-joint carbon fiber reinforced poly-
mer by infrared thermography, microcomputed tomography, ultrasonic
C-scan and microscopic inspection,” Opt. Eng. 54, 104109 (2015).
3. B. Liu, Y. Cao, H. Zhang, Y. Lin, W. Sun, and B. Xu, “Weak magnetic
flux leakage: a possible method for studying pipeline defects located
either inside or outside the structures,” NDT&E Int. 74, 81–86 (2015).
4. A. Da Silva, M. Leabad, C. Driol, T. Bordy, M. Debourdeau, J.-M.
Dinten, P. Peltié, and P. Rizo, “Optical calibration protocol for an
x-ray and optical multimodality tomography system dedicated to
small-animal examination,” Appl. Opt. 48, D151–D162 (2009).
5. P. J. Schilling, B. R. Karedla, A. K. Tatiparthi, M. A. Verges, and P. D.
Herrington, “X-ray computed microtomography of internal damage in
fiber reinforced polymer matrix composites,” Compos. Sci. Technol.
65, 2071–2078 (2005).
6. P. Wright, X. Fu, I. Sinclair, and S. Spearing, “Ultra high resolution
computed tomography of damage in notched carbon fiber epoxy com-
posites,” J. Compos. Mater. 42, 1993–2002 (2008).
7. H. Fernandes, H. Zhang, and X. Maldague, “An active infrared
thermography method for fiber orientation assessment of fiber-
reinforced composite materials,” Infrared Phys. Technol. 72, 286–292
(2015).
8. H. Fernandes, C. Ibarra-Castanedo, H. Zhang, and X. Maldague,
“Thermographic non-destructive evaluation of carbon fiber-reinforced
polymer plates after tensile testing,” J. Nondestruct. Eval. 34, 1–10
(2015).
9. B. Yousefi, S. Sojasi, C. I. Castanedo, G. Beaudoin, F. Huot, X. P.
Maldague, M. Chamberland, and E. Lalonde, “Mineral identification
in hyperspectral imaging using sparse-PCA,” Proc. SPIE 9861,
986118 (2016).
10. B. Yousefi, S. Sojasi, C. I. Castanedo, G. Beaudoin, F. Huot, X. P.
Maldague, M. Chamberland, and E. Lalonde, “Emissivity retrieval
from indoor hyperspectral imaging of mineral grains,” Proc. SPIE
9861, 98611C (2016).
11. http://www.flir.com/.
12. http://www.apolloinstruments.com/.
13. E. J. Kubiak, “Infrared detection of fatigue cracks and other near-
surface defects,” Appl. Opt. 7, 1743–1747 (1968).
14. C. Gruss, F. Lepoutre, and D. Balageas, “Nondestructive evalua-
tion using a flying-spot camera,” in 6th European Conference on
Non-Destructive Testing (ECNDT), Nice, France, 1994, pp. 657–661.
Research Article
15. A. Rashed, D. P. Almond, D. Rees, S. Burrows, and S. Dixon,
“Crack detection by laser spot imaging thermography,” AIP Conf.
Proc. 894, 500–506 (2007).
16. S. E. Burrows, A. Rashed, D. P. Almond, and S. Dixon, “Combined
laser spot imaging thermography and ultrasonic measurements for
crack detection,” Nondestr. Test. Eval. 22, 217–227 (2007).
17. J. Schlichting, C. Maierhofer, and M. Kreutzbruck, “Crack sizing by
laser excited thermography,” NDT&E Int. 45, 133–140 (2012).
18. T. Rockstroh and J. Mazumder, “Infrared thermographic temperature
measurement during laser heat treatment,” Appl. Opt. 24, 1343–1345
(1985).
19. T. Li, D. P. Almond, and D. A. S. Rees, “Crack imaging by scanning
laser-line thermography and laser-spot thermography,” Meas. Sci.
Technol. 22, 035701 (2011).
20. C. Alard and R. H. Lupton, “A method for optimal image subtraction,”
Astrophys. J. 503, 325–331 (1998).
21. N. Rajic, “Principal component thermography for flaw contrast en-
hancement and flaw depth characterisation in composite structures,”
Compos. Struct. 58, 521–528 (2002).
22. H. Fernandes, H. Zhang, C. Ibarra-Castanedo, and X. Maldague,
“Fiber orientation assessment on randomly-oriented strand compo-
sites by means of infrared thermography,” Compos. Sci. Technol.
121, 25–33 (2015).
23. H. Zhang, L. Yu, U. Hassler, H. Fernandes, M. Genest, F. Robitaille, S.
Joncas, W. Holub, Y. Sheng, and X. Maldague, “An experimental and
analytical study of micro-laser line thermography on micro-sized flaws
in stitched carbon fiber reinforced polymer composites,” Compos. Sci.
Technol. 126, 17–26 (2016).
24. H. C. Fernandes, H. Zhang, K. Morioka, C. Ibarra-Castanedo, F.
López, X. P. Maldague, and J. R. Tarpani, “Infrared thermography
for CFRP inspection: computational model and experimental results,”
Proc. SPIE 9861, 98610H (2016).
25. M. Ito and T.-W. Chou, “An analytical and experimental study
of strength and failure behavior of plain weave composites,”
J. Compos. Mater. 32, 2–30 (1998).
26. E. Barbero, P. Lonetti, and K. Sikkil, “Finite element continuum dam-
age modeling of plain weave reinforced composites,” Composites B
37, 137–147 (2006).
27. A. P. Berens, “NDE reliability data analysis,” in ASM Handbook
(1989), Vol. 17, pp. 689–701.
28. Y. Duan, P. Servais, M. Genest, C. Ibarra-Castanedo, and X. P.
Maldague, “ThermoPoD: a reliability study on active infrared
thermography for the inspection of composite materials,” J. Mech.
Sci. Technol. 26, 1985–1991 (2012).
29. Y. Duan, S. Huebner, U. Hassler, A. Osman, C. Ibarra-Castanedo,
and X. P. Maldague, “Quantitative evaluation of optical lock-in and
pulsed thermography for aluminum foam material,” Infrared Phys.
Technol. 60, 275–280 (2013).
30. B. Liu, H. Zhang, H. Fernandes, and X. Maldague, “Quantitative evalu-
ation of pulsed thermography, lock-in thermography and vibrother-
mography on foreign object defect (FOD) in CFRP,” Sensors 16,
743–754 (2016).
31. Y. Duan, “Probability of detection analysis for infrared nondestructive
testing and evaluation with applications including a comparison with
ultrasonic testing,” Ph.D. thesis (Université Laval, 2014).
32. B. Yousefi, S. M. Mirhassani, A. AhmadiFard, and M. Hosseini,
“Hierarchical segmentation of urban satellite imagery,” Int. J. Appl.
Earth Obs. Geoinf. 30, 158–166 (2014).
33. H. Zhang, H. Fernandes, L. Yu, U. Hassler, M. Genest, F. Robitaille, S.
Joncas, Y. Sheng, and X. Maldague, “A comparative study of exper-
imental and finite element analysis on submillimeter flaws by laser and
ultrasonic excited thermography,” Proc. SPIE 9861, 98611A (2016).