Structural Health Monitoring

shootperchUrban and Civil

Nov 26, 2013 (3 years and 6 months ago)

Structural Health Monitoring
The online version of this article can be found at:
DOI: 10.1177/1475921706057982
2006 5: 59Structural Health Monitoring
D. Lecompte, J. Vantomme and H. Sol
Crack Detection in a Concrete Beam using Two Different Camera Techniques
Published by:
can be found at:Structural Health MonitoringAdditional services and information for Alerts:
What is This?
- Feb 14, 2006Version of Record >>
by guest on November 25, 2013shm.sagepub.comDownloaded from
by guest on November 25, 2013shm.sagepub.comDownloaded from
Crack Detection in a Concrete Beam
using Two Different Camera Techniques
* J.Vantomme
and H.Sol
Royal Military Academy,Department of Materials and Construction
Renaissancelaan 30,PO Box 1000,Brussels,Belgium
Free University of Brussels,Mechanics of Materials and Constructions
Pleinlaan 2,PO Box 1050,Brussels,Belgium
The study presents an application of two different optical measurement techniques for the detection of
cracks at the surface of a realistically sized concrete beam subjected to flexural loading conditions.
Both techniques provide displacement measurements in a discrete number of points on the surface.
Based on these displacement fields,deformations are calculated by means of the Green–Lagrange
strain expression.The study deals with the relationship between cracks and the concept of
deformation and it examines which of the two methods presented appears to be the most suitable for
crack prediction or detection.The study shows that it is possible to detect the appearance and
evolution of cracks,even before the cracks become visually detectable,with both methods and reveals
their complementarities.
Keywords concrete beam  damage-detection  optical measurements  non-contact
1 Introduction
The present study is defined in the framework
of the FWO-Flanders-project Nr G.0266.01,
where the objective is to evaluate the behavior of
a realistically sized prestressed concrete beam,
during static and dynamic loading,using a four-
point bending disposition in which the two
bending points (position of the loading jacks) are
decoupled (see Figure 6).
Classical experimental validation techniques
that use strain gauges,extensometers,etc.,do not
allow a correct apprehension of the local damage
mechanisms,which lead to the failure of the
structure,due to the heterogeneous nature of
concrete.A more relevant modeling of these
mechanisms can be assisted by a full field
measuring technique focused on a given region of
interest of the loaded specimen.The question is
whether it is possible to predict crack develop-
ment before a crack becomes visually detectable
and whether the crack evolution during the
process can be monitored.
In this article,two different measurement
techniques are used to measure the displacements
of a discrete number of points on the surface of a
beam during loading.
The first technique is based on the compar-
ison of a number of digital images,taken during
the loading of the beam.It uses a given speckle
pattern,exhibited by the surface of the specimen
or obtained by a paintspray,combined with
*Author to whom correspondence should be addressed.
Copyright ￿ 2006 SAGE Publications,
Vol 5(1):0059–10
[1475-9217 (200603) 5:1;59–10 10.1177/1475921706057982]
Copyright ￿ 2006 SAGE Publications,
Vol 5(1):0059–10
[1475-9217 (200603) 5:1;59–10 10.1177/1475921706057982]
by guest on November 25, 2013shm.sagepub.comDownloaded from
digital image correlation (DIC) and a charge-
coupled device (CCD) camera.It will be referred
to as the ‘DIC technique’ [1].
The second technique is based on the locali-
zation in space of a number of light emitting
diodes (LEDs),fixed to the surface of the beam.
This technique uses the principle of space inter-
section combined with three linear CCD cameras
and a number of infrared (IR) light emitting
diodes.It will be referred to as the ‘LED-CCD
technique’ [2].
The article presents a theoretical overview of
both techniques,examines the measurement prin-
ciples,and discusses the obtained results with
respect to the determination of cracks.
2 Measurement Techniques
2.1 Digital Image Correlation (DIC)
The DIC technique is an optical-numerical
measuring technique,which offers the possibility
to determine displacement and deformation fields
at the surface of objects under any kind of
loading,based on a comparison between images
taken at different loading steps.This technique
has been used in different technological domains
and many of its applications have been reported
[3–6].It allows to study qualitatively as well as
quantitatively the mechanical behavior of materi-
als under certain loading conditions.
A measurement consists of taking several
pictures of the object of interest during loading,
with a CCD camera.Every pixel of the camera
stores a certain gray scale value from 0 to 255,
corresponding to the intensity of the light
reflected by the surface of the tested specimen.
The objective is to obtain an image of the object
with a varied and distinctive gray value pattern,
to enable differentiation between different square
groups of pixels (subsets).
The concept behind the DIC-software match-
ing algorithm is that the gray value distribution
of a subset in the image of the undeformed object
should correspond to the gray value distribution
of the same material area in the image of the
deformed specimen.To measure a displacement
field,the image is divided into a number of
subsets.The size of a subset can for example
be 77,1111,or 1515 pixels.The image
correlation routine permits to locate every subset
of the initial image in the deformed image.
Subsequently,the software determines the displa-
cement values of the centers of the subsets,which
yield an entire displacement field.
Figure 1 depicts the sequence of taking a
picture of an object before and after loading,
storing the images onto a PC through a frame
grabber,performing the correlation of both
images (i.e.,locating the different undeformed
subsets in the deformed image) and finally
calculating the corresponding displacement of the
centers of the subsets,which finally yields the
desired displacement field.
2.2 LED-CCD Technique
The camera system is based on three linear
CCD units (three circular apertures in Figure 2).
Each unit is composed of a 2048 pixel linear
CCD,optics,and a processing board.The optics
are composed of an infrared window,a normal
lens,and a cylindrical lens,which compresses the
viewing area to a line.The processor of each unit
calculates the position of the peak of a light
source on the CCD.
The basic unit acts like a goniometer.Each
unit is capable of measuring the angle that a light
source makes with the optical axis,which is
parallel to the z-axis in Figure 3.This light
source should be small and yet powerful enough
Figure 1 Working principle of the DIC-system.
60 Structural HealthMonitoring 5(1)
by guest on November 25, 2013shm.sagepub.comDownloaded from
to exceed the light level of the environment.
IR-emitting diodes are used for this task.They
are small and can emit enough IR light.The IR
windows ensure that environmental light does not
influence the measurement.
The three optical goniometers are mounted
on a rigid structure (Figure 2).Using triangula-
tion techniques,one 3D position can be calcu-
lated from three 1D positions or angles.In
Figure 3,a schematic top view of the system is
shown.Cameras 1 and 3 provide the Z-and the
Y-position.Camera 2 mainly measures the
X-position.Measuring the position of a LEDbefore
and after loading of an object,on which the LED
is attached,yields its displacement in 3D space.
Since the cylindrical lens projects the 2D
viewing area on a 1D line,not more than one IR
marker can be observed at the same time.To
track several markers simultaneously,the markers
need to be flashed sequentially.Since the cameras
are mounted on a rigid structure,calibration only
has to be performed once in the factory.
3 Measurement of Deformation
As the objective is to measure crack damage,
the first question is how to quantify this damage
based on displacement measurements.Therefore,
a concept from continuum mechanics is used
The behavior of a material can be character-
ized by the displacement of a discrete set of
points at the surface of the material.Figure 4
shows an object before and after deformation.
The points P and Q delimit a line segment dX
in the undeformed configuration.P
and Q
delimit the same line segment dx
in the deformed
configuration.The translation vector is called u.
In vector notation this means:
x þ
dx ¼


The Taylor expansion of the last term of the
second member yields:
x þ
dx ¼

þ r
uð Þ 

þ   ð2Þ
Figure 2 Coordinate measurement system K600 (by
Figure 4 Deformation principle in continuummechanics.
Figure 3 Measurement by spatial intersection.
Lecompte et al.Crack Detection in a Concrete Beam 61
by guest on November 25, 2013shm.sagepub.comDownloaded from
Knowing that:
x ¼

Equation (2) becomes:
dx ¼
dXþ r
uð Þ 
dXþ   ð4Þ
or in finite difference notation
￿x ¼
￿Xþ r
uð Þ 
￿Xþ   ð5Þ
When projecting these vectors onto an orthogon-
ally normalized coordinate system one obtains:

¼ I þ

þ   ð7Þ
where F is the deformation gradient tensor,which
is defined as follows:
dx ¼ F  dX ð8Þ
F ¼
@ Xþuð Þ
@ Xþuð Þ
@ Yþvð Þ
@ Yþvð Þ
¼ I þ
The gradient deformation tensor is a linear
operator that transforms a rectangle into a
parallelogram.x and y,X and Y,and thus u and
v are measured.The remaining unknowns are the
four members of the deformation gradient tensor.
Figure 5 shows four arbitrary data points
(centers of facets or LEDs) in the initial and
deformed configuration.The result covers the
deformation of the zone between Points 1,2,3,
and 4.
It is possible to calculate the strain compo-
nents from the deformation gradient tensor:
E ¼
 F I
or in two-dimensional space:
E ¼
Once the deformation gradient tensor F is calcu-
lated,the chosen strain tensor E can be deter-
mined.The Green–Lagrange strain is an objective
strain expression,which means that it takes rigid
body motions into account.Finally,the eigenva-
lues and eigenvectors of the strain tensor are
calculated to obtain the major and the minor
principal strain values and their corresponding
In practice,strain is calculated between four
adjacent data points (four LEDs in the case of
Figure 5 Deformation principle in discrete 2D case.
62 Structural HealthMonitoring 5(1)
by guest on November 25, 2013shm.sagepub.comDownloaded from
the LED-CCD system or the centers of four
subsets in the case of the DIC system).When the
initial positions and the displacement components
of these four data points are known,the discrete
form of Expression (10) is used to calculate the
strain value in the region between the four data
points considered.The distance between the
initial positions of the data points is called the
gauge length.
It is clear that in the present case,the strain
value calculated does not represent an evenly
distributed deformation but is more a measure of
damage due to the appearance of cracks.
4 Experimental
4.1 Introduction
The experimental principle consists in loading
the prestressed concrete beam up to different
loading values and subsequently unloading it.
After the static loading and unloading procedure,
a dynamic analysis of the beam – which is not
discussed here – is performed.
The measurements discussed in the present
article are performed during one of the loading–
unloading cycles shown in Figure 7.Four tests
had already been performed where maximum
loads of 45,65,76,95kN,respectively were
imposed per jack.This study is focused on the
exploitation of the images that were taken during
the fifth cycle.Figure 6 shows an image of the
experimental setup and Figure 7 represents the
deflection curve during the eight consecutive
loading and unloading cycles.
4.2 Setup of the Two
Measuring Systems
A given zone is selected to perform the static
displacement measurements.Figures 8 and 9
show the location of the IR-LEDs (represented
Figure 7 Load–deflection diagram.
Figure 6 Experimental setup.
Lecompte et al.Crack Detection in a Concrete Beam 63
by guest on November 25, 2013shm.sagepub.comDownloaded from
by the dots).Forty-nine of those markers are
used and placed over the entire height of the
beam.To enhance the resolution of the 2D CCD
camera,only the lower part of the zone covered
by the LEDs is taken into account.In Figure 8,
this zone is represented by the rectangle at the
bottom flange.
Both measurement systems (Figure 10)
provide displacement measurements in a discrete
number of points.For the LED-CCD system,
these points are represented by the IR-LEDs
(encircled in Figure 11).Only the diodes located
at the lower flange are used for the comparison.
As for the DIC-system,the data points are
represented by the centers of the facets (white
dots in Figure 11).
Figure 12 represents the six different zones
in which finally the deformation/damage is deter-
mined by the two techniques.During the previous
loading cycles,several cracks already occurred.
They are put in evidence by the black lines.
Zones 1,3,and 5 exhibit such cracks.
At the end of the loading (95kN),a visual
inspection of the beam is performed.It leads to
the detection of two new cracks in Zones 2–4
(Figure 13).This should be detected by both
measuring systems.
Figure 9 Field of view of the 2D CCD-camera with LED
Figure 10 Image of both camera systems.
Figure 8 Experimental setup with zones of interest.
Figure 13 Appearance of two new cracks during the
5th loading step up to 95kN per jack.
Figure 12 Zoom in on the zone of interest with
Figure 11 Zoom in on the zone of interest.
64 Structural HealthMonitoring 5(1)
by guest on November 25, 2013shm.sagepub.comDownloaded from
5 Results and Discussion
5.1 Results from the
LED-CCD System
During the loading of the beam,a number of
8000 displacement measurements are performed
for each LED with this system.
Figures 14 and 15 represent the horizontal
and vertical displacements of the lower seven
LEDs during the loading of the beam.The
purpose of showing these graphs is to demon-
strate that it is difficult to retrieve any informa-
tion concerning damage or deformation based on
the displacement values.
However,when calculating the major princi-
pal strains in the different zones during loading,
more information becomes visible.Figure 16
represents the major principal strain values in the
six different zones between 0 and 95kN.It is
important to know that the orientation corre-
sponding to the major strain is the horizontal
direction.It is,as one would expect,perpendicu-
lar to the crack direction.
The accuracy of the determined displacements
depends on the distance between the system and
the measured object.For a distance between 1.5
and 3m,which is the case in the present study,a
standard deviation on the measured displace-
ments of 15mm is obtained.The standard
deviation of the calculated strains is 50mstrain.
Because of the large number of measurements in
time,it is possible to quantify the accuracy in
terms of ‘standard deviation’ and not in terms of
a more severe ‘uncertainty value’.The standard
deviation is calculated based on a piecewise linear
regression of the displacement and strain curves.
At a load of 40kN,it becomes clear where
the original cracks are located.The higher strain
values are found in Zones 1,3,and 5.In these
zones,a major principal strain value of
200mstrain is calculated.This value corresponds
very well with the elastic limit of the reinforcement
steel.Beyond this value (marked by the thick
black horizontal line in Figure 16),the curves
show a steeper ascending path.Figure 16 clearly
Figure 14 Horizontal displacement of lower seven
Figure 16 Major principal strain (continuous) in the
different zones during loading (measured by Krypton
Figure 15 Vertical displacement of lower seven LEDs.
Lecompte et al.Crack Detection in a Concrete Beam 65
by guest on November 25, 2013shm.sagepub.comDownloaded from
shows when yielding of the reinforcement bars
occur and cracks begin to form in every zone.
Between a load of 80 and 90kN,an impor-
tant increase in the strain plot is observed
in Zone 2.This corresponds to the appearance
of a new crack in that zone,as can be seen in
Figure 13.
When exceeding a load of 90kN,the curves
corresponding to Zones 3 and 4 ascend more
steeply.This corresponds to the existence of a
new crack between Zones 3 and 4.The strain in
both zones is influenced by the new crack.
At a load of 90kN,the slope of the curve
corresponding to Zone 6 becomes more impor-
tant.During the visual inspection however,no
new crack was detected.
The major principal strains based on the
measured displacements of the LEDs are princi-
pally the result of the opening of a crack
in a given region.The ductile strain of concrete
is 100mstrain,beyond which cracks start to
appear.The strains that exceed this value are no
longer a measure of continuous deformation,but
a measure of the crack opening.Hence,the
increasing strain in Figure 16 has to be attributed
to the widening of the crack in the corresponding
To reveal the correlation between the calcu-
lated major principal strain and the size of the
cracks,a region with only one crack has to be
chosen.When more than one crack occurs in
the same region,it is no longer possible to derive
the width of every single crack from the calcu-
lated strain.At the end of the loading (at 95kN),
a single visible crack is present in
Zones 1,2,and 5.The measured crack openings
are 0.28,0.16,and 0.21mm respectively.Every
crack is measured in the middle of the bottom
flange with a crack microscope with an accuracy
of 0.01 mm.
The ratio between the crack width and the
corresponding strain value in Zones 1,2,and 5
makes it possible to estimate the crack width in
Zone 6.
Based on the results in Table 1,the strain
value of 380 mstrain in Region 6 is
associated with a crack opening with a width
between 33 and 39mm,which is not visible with
the naked eye.
5.2 Results from the DIC System
With this technique,a picture is taken every
10kN load.The displacements are measured in a
field of 1400 data points.This means that
1400 subsets are used in the correlation process.
Finally,the displacement components in the
centers of these subsets are determined.The
uncertainty on the measured displacements is a
value expressed in pixels and is accepted to be
0.07 pixel.This means that the uncertainty
expressed in units of length depends on the
resolution of the images and thus depends on the
experiment.In the present case,a resolution
of 0.79mm/pixel is obtained.Therefore,the
uncertainty on the displacements expressed in
units of length is 55mm.For this system,it is
more appropriate to express the accuracy in
terms of the uncertainty on the measured displa-
cements and not in terms of standard deviation.
First of all,the mean values of the displace-
ment components of the data points (white
markers in Figure 11) located in the vicinity of
the IR-LEDs are calculated.Subsequently,the
major principal strain values are determined as
for the LED-CCD data.The results are shown in
Figure 17.It is clear that this plot is less detailed
than the one based on the LED-CCD data
because of the limited number of load steps at
which the necessary pictures are taken.
Table 1 Comparison between measured crack width and strain.
crack width (mm)
strain (strain)
Crack width/strain
Region 1 0.28 2700 10.3e 5
Region 2 0.16 1810 8.8e 5
Region 5 0.21 2230 9.4e 5
66 Structural HealthMonitoring 5(1)
by guest on November 25, 2013shm.sagepub.comDownloaded from
For the determination of the accuracy on the
calculated strains,basic error analysis was used.
The uncertainty on the strain value depends on
the uncertainty of the displacements but also
on the chosen gauge length (i.e.,the distance over
which the strain is calculated).The smaller
the gauge length,the higher the uncertainty on
the strains.For the strains in Figure 17,a gauge
length of 150mm is chosen (similar to the
horizontal distance between the LEDs).This
means that the uncertainty on the strains in
Figure 17 is 500mstrain.
However,the advantage and the complemen-
tarity of the DIC technique become clear in
Figures 18–20.These figures show plots of the
calculated major principal strain in the entire
zone of interest.For these plots,the entire
displacement field is used (all the centers of the
subsets) and the gauge length is chosen to be
much smaller than in the case of Figure 17.This
means that crack information can be revealed
more locally because it is less smeared out over
the corresponding zone.The disadvantage is that,
because of the smaller gauge length,the uncer-
tainty on the strains is even higher.However,the
benefit of these plots is their capacity to locate
the cracks in the corresponding zones.
Figure 18 shows the existence of the three
initial cracks.The situation corresponds to a load
of 80kN.
Figure 19 displays the major principal strains
at a load of 90kN.A fourth crack appears in
Zone 2.This can also be noticed in Figure 16,
Figure 17 Major principal strain (every 10kN load)
in different zones during loading (measured by DIC
Figure 18 Major principal strain at 80kN.
Figure 19 Major principal strain at 90kN.
Figure 20 Major principal strain at 95kN.
Lecompte et al.Crack Detection in a Concrete Beam 67
by guest on November 25, 2013shm.sagepub.comDownloaded from
where the moment of crack forming can be
determined very precisely.However,the determi-
nation of the exact location of the crack is only
possible in the full field plot.
Figure 20 corresponds to the end of the
loading with a load of 95kN.A last clear crack
appears between Zones 3 and 4.This can again
be verified in Figure 16,where an increase in the
slope of the corresponding curves becomes clear.
Finally,an augmentation of the curve corre-
sponding to Zone 6 becomes noticeable at the end
of the loading (in Figure 16).Visual inspection did
not reveal a crack in the sixth zone.However,in
Figure 20,a change in the strain field in Zone 6
becomes clear as well.It means that a combination
of both techniques reveals the existence of a begin-
ning crack,not yet visible with the naked eye.
5.3 Comparison between
Both Systems
The number of measurements performed by
the LED-CCD system in the same period of time
is much higher than the number of pictures taken
with the DIC system.It is of course possible to
increase the number of pictures.However,due to
the time-consuming post-processing of the DIC
technique,one has to limit the frequency with
which the pictures are taken.To give an idea,
during the loading cycle from 0 to 95kN (which
took 30
) about 9000 measurements are nearly
instantly performed with the LED-CCD system.
Only 11 images over that same period of time are
taken and used for post-processing by the DIC
system.The processing of these 11 pictures took
about 40
with a Pentium III processor.
Due to the limited number of IR-markers,
the obtained information with the LED-CCD
system is not very local.Every change in the
material is extrapolated to the motion of only a
few markers.The DIC technique offers numeri-
cally the possibility to increase the number of
data points for which the displacement compo-
nents are determined.It therefore generates a
more local view of the deformation.
The present study clearly shows that the
LED-CCD system should be used to indicate the
larger region in which a crack occurs.
Furthermore,the calculated strain values give a
good idea about the actual crack width,located
in the corresponding region.The DIC system on
the other hand can be used to more precisely
locate the crack in that same region.
As a conclusion,one can state that both
measuring techniques are complementary.On the
one hand,the LED-CCD technique permits to
perform a large number of measurements over a
given period of time in a few data points.On the
other hand,the DIC technique enables the user
to obtain a dense field of data points in which
the displacement values are expressed in a few
number of time steps.
The results for the LED-CCD technique have been kindly
put to our disposal by Krypton Industrial Metrology and
Prof Dr Ir G.De Roeck from the KUL.
This project is supported by the Belgian National Science
1.Aramis Users Manual 2000.Gom (GOM– Gesellschaft
fu¨r Optische Mebtechnik mbH),Gom mbH,
2.Krypton Industrial Metrology,Research Park Z1,
Interleuvenlaan 86,B-3001 Leuven,Belgium,
3.Chevalier,L.,Calloch,S.,Hild,F.and Marco,Y.(2001).
Digital image correlation used to analyze the multiaxial
behavior of rubber-like materials.European Journal of
Mechanics – A/Solids,20(2),169–187.
4.Wang,Y.and Cuitin˜o,A.M.(2002).Full-field measure-
ments of heterogeneous deformation patterns on poly-
meric foams using digital image correlation.International
Journal Solids and Structures,39(13–14),3777–3796.
5.Abanto-Bueno,J.and Lambros,J.(2002).Investigation
of crack growth in functionally graded materials using
digital image correlation.Engineering Fracture
6.De Roover,C.,Vantomme,J.,Wastiels,J.and
Taerwe,L.(2003).DIC for deformation assessment:
A case study.European Journal of Mechanical and
Environmental Engineering,48(1),13–20.
7.Geers,M.G.D.,De Borst,R.and Brekelmans,W.A.M.
(1996).Computing strain fields from discrete displace-
68 Structural HealthMonitoring 5(1)
by guest on November 25, 2013shm.sagepub.comDownloaded from