APPLICATION OF DIGITAL IMAGE CORRELATION (DIC) TO

MONITOR THE SOLID DEFORMATION AND OPENING OF

MULTIPLE FRACTURES SIMULTANEOUSLY: APPLICATION TO
THE ITAIPU DAM
Étore F. de Faria *, Ana Carolina P dos Santos†, and Luis Saucedo-Mora3‡
*
ITAIPU Binacional, Divisão de Obras Civis - Setor de Segurança de Barragens e Lab. Concreto
Av. Tancredo Neves, 6731- Conjunto 85856970 - Foz do Iguaçu, PR - Brasil

Keywords: Dams, Fracture, DIC, Energy

Abstract. In plain or reinforced concrete structures the different cracks reflect the
internal, mechanical and chemical, processes affecting the material. And it denotes an
evolution of the material to adapt itself to the loading or environmental conditions at
each stage of its service life. In massive concrete structures such as dams the cracks
given by differential expansion may have economical consequences. For that reason a
monitoring technique such as Digital Image Correlation (DIC), capable to catch those
cracks in a wide region, is of great interest to prevent the consequences with an early
diagnostic. Digital Image Correlation is a contactless and non-destructive monitoring
technique, consisting on a set of photographs correlated with a reference to calculate the
displacement field. In this paper, the methodology developed by Saucedo-Mora et al.1
(2017) is applied to correct the dataset and to obtain the cracking information in a
specific region of the Itaipu Dam. If the crack tip is captured in the photograph, the
measurement can include the fracture energy released by each crack, the mode of the
failure and the cracks active in each measurement with its opening. This methodology
allows the continuous measurement in a wide region obtaining information of the
structure. Which can be used for a continuous automatic evaluation of the structural
integrity, used for both: an identification of the weak regions in the full structure with a
mapping, and the measurement of the material mechanisms failing. This information can
be used to identify locally its more appropriate repair.

1 INTRODUCTION
The structures are exposed to degradation during its service life, which may end in a
lack of structural integrity with safety and economical consequences. To avoid this
scenario it’s needed a monitoring technique capable to detect at an early stage those
defects. This problem is not trivial, since there are two open questions; what to measure
and where to measure. Usually the measurements are done on demand, focusing on the
regions where a severe defect is detected. This is a local approach hard to be applied in

†
Universidade Federal da Integração Latino Americana – UNILA
‡
Institute Eduardo Torroja of Construction Sciences (IETcc)/University of Oxford

1
 Étore F. de Faria , Ana Carolina P dos Santos and Luis Saucedo-Mora

large structures where a visual inspection is not feasible in the full structure. This is the
case, among others, of the dams. And specifically of the Itaipu dam due to its size. Here is
where the DIC technique developed by Saucedo-Mora1 gain importance. This can detect
and measure automatically cracks appearing inside a region photographed, and combined
with for example drones, it can be applied fast and affordable in the full structure. This
will allow the operator to have a continue data acquisition with the software scanning the
surface of the dam, identifying any defect. Also, the DIC technique developed can detect
openings of 10 µm in a picture of 2x2m, which is much more precise than the visual
inspections. Another potentiality is that there are no spatial restrictions, since the cracks
can appear in any position photographed, and the software will locate them and start the
monitoring. This is the big drawback of the conventional techniques that can only
measure locally the crack opening after the crack is detected by visual inspection. It
converts this technique in the only one capable to monitor the structure before the cracks
visually appear, measuring the appearance and locations of the cracks and their evolution
automatically. Also, since the full field displacement is measured in a wide region, there
is a large amount of information that can be used for monitoring, not only the crack
opening.

Digital Image Correlation is a contactless and non-destructive monitoring technique

capable to measure the displacement field of a wide region. It compares two photographs,
a reference and a deformed, finding the new position of the sub-matrices of the reference
picture in the deformed, considering the grey scale value of the images. Those sub-
matrices are called windows and their displacement is the local displacement of this
region in the picture. Finally those displacements are translated into millimetres through
the pixel size, which is calculated through a known length in the photograph (i.e. if we
photograph something which is 1000mm long and in the picture it occupies 500 pixels,
the pixel size is 1000/500=2mm/pixel). With this ratio we can measure directly in the
picture in mm.

The software developed implements the motion correction algorithm owned by this
software, and thanks to that it can be used with mobile platforms for the image acquisition,
such as drones. Also it is self-configured, and automatically offers the best resolution that can
be achieved with the image taken, as well as the optimized parameters for the calculation.
From the picture acquisition, the full process is completely autonomous, and can be used
without technical knowledge about the DIC theory. With friendly results easy to interpret
thanks to the post-process packages included in the software. The technique, because
measures the displacement field continuously in a wide region, is ideal for early crack
detection and global inspections. And with the big amount of information obtained, ad-hoc
post-process packages can be developed considering the maintenance requirements of the
structure.

Figure 1: The full automatic image based calculation of the structural integrity, with the type of image used
as an input of the calculation.

2
 Étore F. de Faria , Ana Carolina P dos Santos and Luis Saucedo-Mora

The motion correction module developed, and integrated in the software, permit the
technology to be applied with a wide range of acquisition methods to reach inaccessible
places in large structures.

Figure 2: The dataset before and after correction with 2 types of typical spurious movements in large
structures, showing clearly, after correction, the cracks and the main deformation of the structure. Both noisy
datasets after correction become the same, because both are the same beam under the same loading conditions.
The software is developed in 3 stages, in the first one the photographs are obtained with a
camera; used by hand, in a tripod or mounted in a drone. Then the DIC and motion correction
algorithms will compute and correct the displacements of the full region measured.
Additionally mobile devices such as drones can be used to do maps of images with a certain
overlap to extract accurate information of inaccessible places of the structure or scans large
surfaces.
The different algorithms make the software versatile both in the acquisition of the images
and the results produced. The final results are easy to interpret and can be focused to the
specific requirements of each structural maintenance.

Figure 3: A real software result and examples of 3 post-processing packages to calculate the crack opening,
the energy released in Mode I and Mode II and the penetration of a pre-stressed wire in a pre-stress reinforced
beam. This data can be used by other packages of the software to perform a risk analysis, evaluate the safety of
the structure, to manage the repairs and different durability calculations.
The paper is structured as follows; in the section 2 the photographs to be used in the
analysis are presented with the regions measured and the restrictions. Then in section 3, the
software is applied to photographs of a crack that appear in the surface of the structure, and
finally in section 4 the crack openings are measured from the DIC results.

3
 Étore F. de Faria , Ana Carolina P dos Santos and Luis Saucedo-Mora

2 MEASUREMENTS IN THE SURFACE OF THE DAM

Two pictures were taken of the surface of the dam, where a crack was previously found
and a monitoring system with two indicators, where the opening is measured with a
calliper, was installed. There is a difference of 2 years between the photographs, so the
software will measure the opening of the crack between 2016 and 2018, considering 2016
as the reference. The pictures are shown in Figure 4, and were taken with a regular
camera, without tripod or precise precautions to keep the same sample to camera distance
or the camera angle. Both inserts errors in the measurement, as Figure 2 shows, but are
corrected with the software. The original resolution of the images was of 4032x3024
pixels in RGB, with a pixel size of 1.2 mm, and they were converted into 8bits in grey
scale, reducing the resolution to 1000x750 pixels for the analysis.

Figure 4: The two measurements for the analysis 2016 and 2018 in grey scale.

The pictures are not ideal for the application of the technique, since there are deep
changes in certain regions of the surface. The main one is the degradation of the white
paint located at the centre, also affecting the different positioning of the label as well as
the dark stain around the white paint. Figure 5 shows the regions crossed by the crack
with a good correlation chosen by the algorithm.

Figure 5: Inclusion and exclusion regions of the picture for the analysis.

4
 Étore F. de Faria , Ana Carolina P dos Santos and Luis Saucedo-Mora

Those drawbacks are more unlikely to happen in regions without previous treatments, as
for example a point in the middle of the dam’s wall. Also, by taking regular pictures (i.e.
every 6 months) the changes in the surface of the structure are more progressive and allow us
to run DIC on them. In the Figure 5, the region with deep changes in the photographs is
shown in red, then the 2 regions that the software localized to run DIC are in green. The
yellow points are the centroids of the windows used in the measurement (50x50 pixels with
an overlap of 25 pixels, which are the yellow squares). During the correction, the perimetric
windows are sacrificed, so the final region measured is the one shown in grey in Figure 5.

3 RESULTS

After the measurement, the data is treated to correct the dataset and to extract the
displacement and strain information. When the crack tip is photographed, this information
can be used for the calculation of the energy released by the crack. But in this case, both
the displacement and strain fields are used for the crack identification and the
measurements of the overall behaviour.

Figure 6: Results of the top region measured, the crack is remarked in black.

Figure 6 shows the results of the top region measured. In Figure 6a the jump of the x
displacement field shows the 2 cracks crossing the region, Figure 6c shows the y
displacement of the region, which is a deflection, so the region is in bending. This is
clearly shown in Figure 6e with the deformed shape of the measured region.

5
 Étore F. de Faria , Ana Carolina P dos Santos and Luis Saucedo-Mora

The strains of Figure 6b,d,f are not homogeneous, which is because of the presence of
the crack. It introduces a discontinuity that creates compressions and tensions in the
surrounding area. At the same time, this is a consequence of the bending measured with
the y displacement in Figure 6e. It will be discussed further in the next section, but
regarding to this bending, as well as the opening of the crack, both are very small and are
far from showing a lack of structural integrity. But the application shows the capabilities
of the technique to measure small defects with a very simple and affordable data
acquisition.

Figure 7: Results of the bottom region measured, the crack is remarked in black.

This analysis of the region is the same done in Figure 6. Here there is also a bending
and a crack opening. Again the crack introduces a discontinuity in the field. In Figure 7b
the x strain concentration shows the crack position, but it differs a little from the exact
position due to the window’s size and the posterior interpolation. It means that when a
jump in the displacement field or a concentration of strains are measured, the real crack
should be within a distance equal to the window used in the correlation. In this case it is
61.25 mm.

4 ANALYSIS OF THE RESULTS

From the measurements of Figures 6 and 7 the crack opening can be calculated, among
other information. In this case, Figures 8 and 9 show the x and y displacements along the
central white line, with their openings.

6
 Étore F. de Faria , Ana Carolina P dos Santos and Luis Saucedo-Mora

Figure 8: x and y displacements of the top region measured. The sign criterion is: the x positive grows to the
right and the y positive grows to the bottom.

In Figure 8a and c, the general trend of the displacement x should be linear, the
nonlinearities are mainly related to the cracks, and in a small amount are due to errors in
the measurement and small discontinuities. With this we can identify the 2 cracks that
cross the region. The one of the left is closing, 0.360 mm, and the one on the right has
opened 0.34 mm, both in those 2 years. The rest is close to linearity, so the material
doesn’t have serious discontinuities. In Figure 8b and d the y displacement is plotted. The
deflection of the region is clear, which denotes a general bending of the surroundings of
the structure. It can be given by aspects far from the measurements done in this paper, but
the most significant should be the crack, which introduces an inhomogeneity. The
difference between the centre and the sides of the measured region in y is of 0.70 mm
over the 200 mm measured.

Figure 9: x and y displacements of the bottom region measured. The sign criterion is: the x positive grows to
the right and the y positive grows to the bottom.

7
 Étore F. de Faria , Ana Carolina P dos Santos and Luis Saucedo-Mora

In the case of Figure 9, the bottom region measured of the crack is less homogeneous
than the one of the top. In Figure 9a and c, the x displacement is far from the linear
expected behaviour of the linear trend. Especially, with the crack on the right, and a
discontinuity on the left part of the measured region. The heterogeneities are caused by
the main crack and small cracks within the region, located in the jumps of the datasets in
Figures 9c and d. The main crack has opened 0.685 mm, and the other discontinuities
located have generally opened or closed less than 0.1 mm. The y displacement shows
again a higher deflection of the region crossed by the crack compared with its
surrounding area. In this case the maximum difference is of 1.377 mm with the 400 mm
measured. Both measurements locate the main discontinuity of the region measured, as
happened in Figure 8 with the top part.

5 CONCLUSIONS
The DIC technique developed has been successfully applied in the Itaipu dam. It opens
a new perspective in the full monitoring of dams with the information that can be
extracted with this non-destructive and contactless affordable technique. Considering the
limitations of the pictures used, the crack could be identified and measured, as well as its
surrounding area. So, the results show that the technique works with the kind of cracks
that appear in the Itaipu dam, as well as with the photograph acquisition restrictions
onsite. This technique has the potential to be applied in large structures. As well, the
precision, versatility and cost of the technique makes it interesting for structures such as
the Itaipu dam, where a huge surface needs to be monitored for an early identification of
the defects. Reducing uncertainties and ensuring the integrity knowledge that a structure
of this importance requires.

REFERENCES
[1] L. Saucedo Mora, M.C. Andrade, C. Lopez Hombrados, J. Barroso and A. Bragado,
Contactless safety evaluation of damaged structures through energetic criteria, Struct
Control Health Monit., (2017). https://doi.org/10.1002/stc.2060