See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/235956848

FDTD modelling in high-resolution 2D and 3D GPR surveys on a reinforced

concrete column in a double wall of hollow bricks

Article in Near Surface Geophysics · February 2013

DOI: 10.3997/1873-0604.2012047

CITATIONS READS

14 1,564

3 authors:

Domenica De Domenico Davide Campo

Università degli Studi di Messina Sandberg
19 PUBLICATIONS 107 CITATIONS 18 PUBLICATIONS 45 CITATIONS

SEE PROFILE SEE PROFILE

Antonio Teramo
Università degli Studi di Messina
54 PUBLICATIONS 221 CITATIONS

SEE PROFILE

All content following this page was uploaded by Domenica De Domenico on 05 June 2014.

The user has requested enhancement of the downloaded file.

Near Surface Geophysics, 2013, 11, 29-40  doi:10.3997/1873-0604.2012047

FDTD modelling in high-resolution 2D and 3D GPR surveys on

a reinforced concrete column in a double wall of hollow bricks
D. De Domenico*, D. Campo and A.Teramo
Seismological Observatory, University of Messina, Messina, Italy

Received November 2011, revision accepted September 2012

ABSTRACT
A structural characterization of a reinforced concrete column in a double wall of hollow bricks,
carried out within diagnostics testing through 2D and 3D high-frequency GPR surveys and aimed
at the evaluation of the seismic vulnerability of a building is proposed. It highlights how the use of
FDTD simulation allows the limits of these surveys to be overcome, with reference to the difficulty
of the processing and interpretation steps in order to identify rebars beyond the hollow bricks. Their
reflection pattern, being complex, makes it very hard to recognize the rebars placed parallel to the
brick hollow axis direction.

INTRODUCTION detailed structural diagnostic surveys (Binda et al. 2008; Zanzi

GPR is a geophysical technique that, over the last years, has and Lualdi 2008) of relevant interest in building seismic vul-
become a significant diagnostic tool in Non-Destructive nerability evaluation.
Testing (NDT) for civil engineering, as a support to invasive As far as the numerical modelling is concerned, it is regarded
diagnostics, due to a high-resolution, fast acquisition, immedi- to be a powerful tool for the comprehension of the interaction
ate and continuous graphic rendering of the signal during the between electromagnetic waves and materials (i.e., Zeng and
investigation (McCann and Forde 2001; Bungey 2004; Klysz McMechan 1997) and efforts to model the exact GPR pulse have
et al. 2006). been made (Giannopoulos 2005; Giannopoulos and Warren
There are a large number of works in literature about the dif- 2007). Tests in reinforced concrete or masonry buildings are
ferent employments of the radar technique in civil engineering: reported in literature (Millard et al. 1998; Shaari et al. 2004;
concrete steel reinforcement location is the most common appli- Diamanti et al. 2008) taking into account different source polar-
cation because of high rebar conductivity (there is a good sum- izations as well (Shaari et al. 2010).
mary in Xian-Qi et al. 2009) and in some cases it is supported by This paper deals with a specific application of GPR method-
automatic detection through a neural network (Bungey et al. ology to a two-storey building and more specifically to a rein-
2003; Shaw et al. 2005) or aimed at diameter estimation (Shaw forced concrete (r.c.) column, in a double wall of hollow bricks
et al. 2003; Utsi and Utsi 2004; Lualdi and Zanzi 2006; Che Way in order to characterize the column structural configuration,
et al. 2009). The assessment of structural conditions of buildings through 2D and 3D profile acquisition. It was possible to achieve
or seismic vulnerability evaluation performed through GPR is an effective identification of the column, with its dimensions, the
another field that has been studied by many authors: Barrile and longitudinal rebars and the concrete cover thickness but the pres-
Pucinotti (2005) reported a case study of seismic vulnerability ence of the hollow bricks obstructs the detection of the brackets
assessment in a reinforced concrete building, whereas Bottari et that are placed parallel to hollow brick axes. Through an FDTD
al. (2003) applied 3D GPR to study the seismic susceptibility (finite-difference time-domain) approach (Yee 1966; Taflove and
damage level of a masonry building. Hagness 2005), some numerical tests based on the GPR survey
Over the last few decades, significant developments of the and next destructive test results were performed and were aimed
GPR methodology have been focused both on the implementa- at clarifying the complex EM response and suggest more suitable
tion of 3D gathers (Grasmueck 2003; Lualdi et al 2003) and acquisition, processing and interpretation strategies to overcome
numerical modelling. As for the first it is more precise, accu- the limits of 2D and 3D surveys.
rate and faster than the usual 2D profile and moreover the
arrangement of new software for 3D imaging allows an easier THE GPR SURVEYS
identification of buried objects. In particular, with high-fre- Survey purposes
quency GPR systems (>2 GHz) it is possible to perform The GPR surveys carried out, which have allowed some struc-
tural elements of the building to be identified, refer to a wide-
*
[email protected] spread construction typology characterized by columns delimit-

© 2013 European Association of Geoscientists & Engineers 29

30 D. De Domenico, D. Campo and A.Teramo

ed by two walls of hollow bricks, due to the need for sound and
heat insulation.
The significance and the resolution level required for such
surveys on the different configurations detected inside the build-
ing in the study, which differ in column or beam size, rebar
number and diameter and thicknesses of the double wall of hol-
low bricks, are related to the need to acquire the level of knowl-
edge of structures required by Italian seismic rules and regula-
tions for a seismic vulnerability evaluation of existing buildings.
Figure 1 shows a detail of the reinforced concrete column
embedded in a double wall of hollow bricks, where the total
thickness of the wall was the only element that could have been
acquired by a direct measure (49 cm); the thickness of each wall
of hollow bricks was unknown; the column location and size
were assumed only with reference to the geometry and distribu-
tion of the emerging beams of the floor covering. For the per-
formed GPR surveys the actual column dimensions, the hollow
brick wall thickness, the number of longitudinal rebars and the
cover concrete thickness were estimated and then confirmed by FIGURE 1
a destructive survey that provided information about the real Location of the surveys: a) front showing the position of the door, the r.c.
dimensions of the hollow bricks, the same hollow inside them column, the 2D GPR profiles (scan 1, scan 2) and 3D grid; b) plan of the
and the presence and the pitch of the brackets. wall section and column geometry. The total thickness of the wall was
These surveys have also formed the basis by which a semei- the only known element, the others were estimated trough the GPR sur-
otic analysis was arranged in order to identify the structural ele- vey and then confirmed by a destructive test. The linear unit of measure-
ments on which to perform invasive procedures fixed by the ment is in cm.
Italian seismic code.
0.025 ns and a sample frequency more than ten times the antenna
Data acquisition central frequency at about 40 GHz and 312 samples per scan.
The GPR survey acquisition steps were performed using a co- For the spatial sampling, the lowest trace increment allowed
polarized perpendicular configuration to obtain the TM by the acquisition software, which is 2 mm for 2D-scans and
(Transverse-Magnetic) polarization. This dipolar configuration, 4 mm in the in-line direction for 3D-scans, was chosen. These
also known as the perpendicular broadfire configuration, is values adhere to the fundamental Nyquist sampling criteria,
obtained by orienting the axis of the dipolar transmitter and avoiding spatial aliasing, described by the following relationship
receiver antennas parallel to the longitudinal column rebars and (Jol 2009):
perpendicular to the antenna acquisition direction; since the scat-
tering properties of rebars are strongly polarization dependent Δx< v/6fc
(Radzevicius and Daniels 2000), in this way the energy reflected
from cylindrical metallic objects (the rebars) is maximized where fc is the antenna central frequency and v is the velocity of
because it is in agreement with the theorem of reciprocity the medium. In order to validate this choice, at worst, the veloc-
(Balanis 1997), the receiving dipole antenna is more sensitive to ity of bad quality concrete (0.10 m/ns) is taken into account; this
electric fields that are parallel to its axis. Indeed, the electromag- value implies Δx < 0.007 m.
netic wave polarization affects the scattering properties of the The first acquisition, denoted by scan 1 in Fig.1, was carried
cylindrical objects: the polarization of the incident field on the out horizontally, 1.3 m above the floor, on the partition wall
receiver antenna is determined by the polarization of the radiated between two rooms in the building, on the left side of the door
field by the transmitter antenna and the degree of depolarization (Fig.1) to highlight the column width through the vertical edge
due to scattering on buried objects. Most commercial antennas are rebar identification. After this, in order to recognize the horizon-
dipole or bowtie, like those used for this study that radiate energy tal rebars and the brackets, a second scan, (denoted by scan 2 in
linearly polarized with the main component of the electric field, Fig.1) was performed vertically, starting from 30 cm above the
oriented along the dipole axis (Capizzi and Cosentino 2008). floor, with the GPR antenna in an intermediate position between
All the acquisitions were performed in reflection mode using the two hyperbolic reflections caused by the rebars, identified in
a GPR system with a central frequency of 2.3 GHz produced by the first scan.
Mala Geoscience, setting a total time window of 8 ns, regarded to Finally, a 3D survey (shown with GRID in Fig. 1) was real-
be long enough for the purpose of this study, a time increment of ized in correspondence to the column, using a grid of 80 cm long

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
 FDTD modelling in high-resolution 2D and 3D GPR surveys 31

column vertical rebars. In order to optimize the information

return, the raw data were processed by removing the DC drift,
performing a time-zero adjustment and recovering the amplitude
losses due to geometrical spreading with divergence compensa-
tion, without any other gain function. Furthermore, the data were
Butterworth band-pass filtered, setting the two cut-off frequency
values, lower and upper, at about 1400 MHz and 4200 MHz
respectively (using mean values extracted from the all trace spec-
tra) to eliminate the low- and high-frequency noise (Fig. 2b).
Fitting the rebar hyperbola diffractions, a migration step with
a constant velocity v = 0.155 m/ns was performed (Fig. 2c) by
using the frequency-wave number migration algorithm (Stolt
1978), assuming the same velocity to calculate the depth in the
radargrams.
Analysing the processed data in the first step (Fig. 2b), the
horizontal band in the first nanosecond identifies the first hollow
brick wall. This band is not continuous, because of an anomalous
zone located in the first 30 cm, up to 2 ns, which is related to the
r.c. door jamb (marked with a blue ellipse); the other elements
shown by blue arrows, are two reflections referable to vertical
joints between two successive bricks; these reflections are not
detected in the last part of the radar section probably due to a
better adherence of the bricks. Between 0.8–1.2 m, at about 2 ns,
the profile shows two clear hyperbolic reflections that could be
the column vertical rebars (indicated with the red arrows);
moreover, above them between 1.2–1.4 ns, the concrete column
surface is recognizable as a linear reflection with small hyper-
FIGURE 2 bolic edges (red ellipse in Fig. 2b). Figure 2(c) shows the
GPR scan 1: a) raw data: the red arrows show the two hyperbolae prob- migrated radargram. In this picture it is possible to highlight the
ably related to the column vertical bars; b) elaboration highlights some thickness of the first wall of hollow bricks, 8 ± 0.1 cm, the col-
details of the wall: the blue arrows show vertical joints between two umn width (about 32 ± 2 cm), the rebars, between 13–15 cm
successive bricks, the blue ellipse indicates the r.c. door jamb; as above, depth and the concrete cover thickness, the thickness between
between 0.8–1.2 m, at about 2 ns, two clear hyperbolic reflections are the rebars and the column surface, about 2.3 ± 0.4 cm.
marked with two red arrows and, above them, at about 1.2 ns, the con- Once the column was located, scan 2 (Fig.1) was performed
crete cover surface is recognizable as a linear reflection with small in a mid-position between the two hyperbolic reflections, shown
hyperbolic edges red ellipse); c) migration processing and evaluation of above. The raw data profile (Fig. 3a) highlights many near-sur-
the concrete cover thickness (2.3± 0.4 cm) between the column surface face elements including five couples of hyperbola branches,
and the rebars. denoted by red arrows and placed at regular intervals. These
reflections could be attributed to the brick hollows (partially
single profiles, in the two perpendicular directions, with 10 cm overlapped by the horizontal reflection of the wall surface)
line spacing between the separate profiles. The adopted scan whose hyperbola branches are so close and dense as to create a
increment, greater than the necessary one by the spatial sampling very complex pattern but no reflection related to horizontal
theorem, 7 mm as above-mentioned, may determine spatial alias- rebars is apparently evident. Applying the same processing of
ing, requiring a high degree of interpolation for 3D rendering but scan 1 and focusing the reflections with the fk-migration algo-
in this case it is appropriate owing to the linearity of the subsur- rithm (Fig.3b), the shape and geometry of the single brick hol-
face targets (rebars) that have a known orientation (Grasmueck et lows are recognized in the first ns for a thickness of 8 ± 0.1 cm,
al. 2003) and it significantly reduces the acquisition times. like in the foregoing profile, which is best highlighted in the
enlargement of Fig. 3(c), so it is possible to associate the previ-
Data processing and interpretation ous five couples of hyperbola branches to the horizontal joints of
Owing to the high-impedance contrast between metal and con- the bricks. In the radar signal beyond the brick reflections, other
crete, some reflections are already clearly visible between 2–4 ns chaotic anomalies and multiple reflections are detectable but
and in the raw data scan (Fig. 2a). In particular, there are two despite the column closeness, it is hard to distinguish any ele-
hyperbolic reflections (shown by red arrows) produced by the ment referable to rebars.

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
32 D. De Domenico, D. Campo and A.Teramo

In order to obtain a 3D representation of the wall, the grid only the two vertical rebars; neither anomaly ascribed to the
was realized, acquiring horizontal and vertical profiles in brackets in this case is recognizable, because the horizontal
accordance with the setup described above to make the identi- bands, probably caused by the brick hollow multiple reflec-
fication of the column rebars easier. The same processing steps tions, are still visible beneath the first wall of bricks both in
described above for all the 2D scans, were applied to the data, correspondence of the column and, on its sides, in the inter-
with the addition of a background removal filter and the differ- space between the two hollow brick walls.
ence of the used migration algorithm, in this case a 3D-fk
migration with the same constant value of 0.155 m/ns. The time NUMERICAL MODELLING FOR GPR DATA
slices obtained from the three-dimensional reconstruction, at SIMULATION
different depths (Fig. 4a–f) allowed an effective recognition of Theoretical background
the wall elements. The reflections due to the hollows of the Since the GPR data of the scans performed perpendicularly to
hollow brick appear as horizontally stretched light lines with the directions of the brick hollows brought about a difficulty in
amplitude values different from those relative to the reflections interpretation and, consequently, an impossible bracket identifi-
caused by the mortar between the brick rows that appear as cation, a suitable numerical modelling was adopted to under-
dark lines instead (Fig. 4a,c). It is possible to detect the brick stand, through a simulation of the acquired data, the link between
texture details (Fig. 4b,d), not recognizable in the 2D surveys the GPR signals and the probed region, modelled as a complex
and the signal amplitude difference between the horizontal of structures and interfaces, parametrized in terms of electromag-
joints between the bricks of two different rows (filled with netic properties.
mortar) and the vertical ones among adjacent bricks in the same Since the EM response is very complex in some cases, know-
row (without mortar). Beyond the first wall of hollow bricks, at ing how the electromagnetic waves propagate is of quite sig-
1.10 ns (Fig. 4e) the column shape and width are recognizable nificant interest, synthetic data could be used to understand how
and, finally, at 1.87 ns (Fig. 4f), the time slice shows clearly a GPR system detects the spatial variability of subsurface elec-
tromagnetic properties, improving data elaboration and interpre-
tation.
Various authors have proposed different numerical modelling
approaches: frequency- domain methods (Powers and Olhoeft
1994; Zeng et al.1995), pseudo-spectral methods (Carcione
1996; Casper and Kung 1996; Lui and Fan 1999), integral meth-
ods (Ellefsen 1999), ray-based methods (Goodman 1994; Cai
and McMechan 1995); the most popular for GPR data simulation
is the finite-difference time-domain (FDTD) technique, suitable
for its simplicity and precision in complex inhomogeneous mate-
rial modelling and solving Maxwell’s equations with time and
space discretization (Kunz and Luebbers 1993).
In particular, the exploding reflector approach used allows a
simulation of a zero-offset section to be made and, starting at
t = 0 time, all the points belonging to a reflective object are the
source of a Huygens elementary wave, with amplitude propor-
tional to the reflection coefficient in the case of normal inci-
dence (Sandmeier 2010). This approach shows some limitations
due to both the lack of acquisition of the direct wave between
the transmitter and receiver and the assignment of the same
polarity to the waves emitted by the opposite sides of an inter-
face (How-Wei and Tai-Min 1998). In such a context, these
lacks are considered as minor for the purpose of understanding
the complex reflection pattern of a radargram and more practi-
cally, the reason why it is impossible to identify the column
FIGURE 3 brackets despite the high resolution of the used antenna and the
GPR scan 2: a) raw data; the red arrows show the five couples of hyper- strong reflection coefficient of the metal.
bola branches connected with the mortar joint between the brick rows In this approach the medium is modelled in terms of the rela-
clearly distinguishable even without processing; b) elaboration high- tive dielectric permittivity εr, which controls the propagation
lights the details of hollow bricks and their joints, better highlighted in c) velocity v, the relative magnetic permeability μr and the electrical
through the red outlines. No element referable to brackets is visible. conductivity σ, which control the attenuation of the EM wave.

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
 FDTD modelling in high-resolution 2D and 3D GPR surveys 33

FIGURE 4
3D grid obtained by horizontal
and vertical profiles, processed
and migrated. Time slices of the
hollow brick wall at different
times showing different elements:
a) t = 0.45 ns, in light tones the
upper surfaces of the brick hol-
lows, in the black ones, the mortar,
layers among the brick rows;
b) t = 0.75 ns, traces of the brick
texture; c) t = 0.90 ns, lower sur-
faces of the hollows; d) t = 1.00 ns,
lower surface of the bricks, traces
of the brick texture; e) t = 1.10 ns,
column outline; f) t = 1.87 ns,
vertical column bars. The depth is
calculated with a mean velocity
value of 0.155 m/ns.

These assumptions are expressed by the following well-known TESTS AND RESULTS
simplified expressions in low- loss media: The goal of this study is the evaluation of the EM response for
a hollow brick and how waves propagate inside it through
v = c/(μrεr)1/2(1) numerical modelling. By means of the FD modeller in Reflexw
software (Sandmeier 2010) 11 different models and simulations
α ≈ σ/2 (μ/ε)1/2(2) were carried out: 6 models reproducing the geometry and elec-
tromagnetic properties of a brick with a single square hollow of
where α is the attenuation constant, c is the EM waves velocity 6 different sizes, 3 models with two adjacent hollows at
in the free space, μ is the magnetic permeability (μ = μ0 μr where decreasing distances and the last two relative to a wall of hol-
μ0 ≈ 4π10-7 H/m is the magnetic permeability of the free space) low bricks located in front of a concrete column without or with
and ε is the dielectric permittivity (ε = ε0 ε r, with ε0≈8.854 10-12 F/m, rebars, respectively. The last model reflects approximately the
the permittivity of the free space). real situation, where the dimensions of bricks, hollows, col-

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
34 D. De Domenico, D. Campo and A.Teramo

of the model as a region with a high-conductivity gradient

(Sandmeier 2010).
The first simulation was carried out using a 0.3 m x 0.3 m square,
which represents a single hollow, characterized by the following
electromagnetic properties, ε r = 1 and σ = 0 (which is air obviously)
and located a 0.05 m depth in a brick with εr = 4 and σ = 0.001
(Fig. 5a). In the synthetic radargram (Fig. 5b) the black arrows show
the upper and lower interfaces, respectively at about 1 ns and 3 ns
and two further horizontal reflections at about 5 ns and 7 ns, under
the lower interface, which are identifiable as multiple reflections.
Moreover, the hyperbolic reflections, caused by the hollow corners,
are visibly more pronounced at the top than at the bottom.
Using the same numerical model, a fixed point source simula-
tion was performed. In this way, the visualization of wave propa-
gation is possible through several snapshots (Fig. 6a–f), which
allow how and why the reflections of the hollows so strongly
affect objects located beyond them to be understood. This phe-
nomenon is caused by the entrapment of the wave inside the
hollow in continuous reflections and inversions of polarity on its
inner surfaces and by a partial transmission of it beyond the hol-
low. In the radargrams this is expressed by an additional number
of reflections of the same object (or part of it) that are superim-
posed with what is actually placed beyond the hollow.
The previous 2D synthetic scan (Fig. 5b equivalent to Fig. 7a)
is compared with the other 5 synthetic profiles relative to square
hollows, obtained by decreasing the side by 5 cm at a time
(Fig. 7b–f), located in the same medium with the same electro-
magnetic properties. It was observed that, for the same time
window, the multiple hyperbolic reflections, decreasing the hol-
FIGURE 5 low size, tend to be superimposed and the cusps of the hyperbo-
a) Model of a single square hollow with a side of 0.3m (ε = 1, μ = 1 and las converge towards the hollow (Fig. 7f).
σ = 0) into a brick (ε = 4, μ = 1 and σ = 0.001); b) synthetic radargram, Afterwards three other simulations (Fig. 8) were carried out
where the upper and lower interfaces of the brick hollow and their mul- on the basis of two hollow models in a homogeneous medium
tiple reflections are indicated by the black arrows, while the red square characterized by the electromagnetic properties of brick, and a
shows the hollow position. ‘layer’ of air placed beneath; such simulations were performed
gradually by reducing the distance between the two square hol-
umns and rebars are from the destructive tests. Both the electri- lows with a side of 5 cm. It was noticed that, for a hollow dis-
cal and magnetic properties result from the comparison of lit- tance equal to 13 cm (Fig. 8a), the reflections generated from the
erature works (Cuiñas and Sanchez 2002; Jol 2009) to the hollow corners interfere with the brick-air interface, still recog-
processing of the GPR scans; all the media simulated are nizable as a ‘layer’, while, for a hollow distance equal to 5 cm
assumed to be non-magnetic and of very low conductivity, (Fig. 8b), the hyperbolic branches of the edges interfere with
apart from rebars. each other and with the same brick-air interface making it similar
To reproduce as accurately as possible the EM response to an isolated object rather than a continuous band. Reaching the
obtained with the GPR surveys mentioned above, we used the actual distance of 1 cm between the two hollows (Fig. 8c), the
Ricker type wavelet, as described by Yelf (2004), setting the interference between the hyperbola reflections, together with
central frequency of the pulse as 2300 MHz, a time window of that produced by the interface, create an artefact, which could be
8 ns and a fake trace increment of 2 mm as the real one. Only the wrongly interpreted as an object and in this case it could be eas-
y-component of the electric field was simulated because, in fact ily mistaken for a rebar.
a GPR transmitter generates electromagnetic radiation with the Having characterized the influence of small hollows in bricks
electric field mainly in the y direction (Shaari et al. 2010). on the high-frequency electromagnetic impulse, the influence of
Moreover, boundary conditions were imposed choosing a the hollow brick wall, with dimensions based on the foregoing
‘big exponential absorbing range’ to eliminate disturbing reflec- destructive tests on the identification of the r.c. column (εr = 3
tions from the boundaries: this condition defines the boundary and σ = 0.001) through two other numerical tests, was subse-

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
 FDTD modelling in high-resolution 2D and 3D GPR surveys 35

FIGURE 6
Simulation of EM wave propaga-
tion in the same model of
Fig. 5(a) with a point source
(located in x = 0.50 m; y = 0); the
hollow shape is depicted with the
dashed blue square, the wavefront
with the dashed red line; a) the
waves propagate from a source in
the medium outside the hollow;
b) the wave is partially reflected
by an upper hollow interface
(reflection at about 1 ns in
Fig. 5b) and partially transmitted
inside the hollow with an increase
of velocity; c) the waves reach the
bottom of the hollow sketching
its shape; d) the wave reflected by
the lower hollow interface backs
to the top; (e) the wave is partially
reflected again by the upper inter-
face towards the bottom; f) waves
are reflected and partially trans-
mitted by the lower hollow inter-
face creating in this way the mul-
tiple reflections as shown in
Fig. 5(b).

quently investigated: in the first one, rebars were not included 0.68 m, could be misinterpreted as a rebar. In order to highlight
(Fig. 9a), while in the second one, circular objects were placed at possible reflections below the bricks, the following sequence to
a regular distance of 15 cm from each other, with the electromag- process this profile was applied: a predictive deconvolution (8 ns
netic characteristics of iron, 6 mm diameter, to simulate the pres- of autocorrelation range, corresponding to the whole trace, filter
ence of column brackets (Fig. 9b). length of 1 ns and lag of 0.5 ns), a Butterworth band-pass filter
Comparing the two raw data files, without any processing (1100 MHz lower cut-off, 3400 MHz upper cut-off), a diffraction
step, in the second profile (Fig. 9b), all metal objects produce migration calibrated on electrical parameters of the model
different anomalies with highly variable intensity but only some (Fig. 10a). Afterwards a wave envelope, where the instantaneous
of them are associated to hyperbolic reflections (shown by amplitude is calculated by the Hilbert transformation, owing to
arrows); the other ones (shown by circles) are less clear and, in which gives an overview of the energy distribution of the traces
the interpretation step, they would be hardly recognized as rebars and the reflectivity strength, it could facilitate the determination
without prior knowledge, while the artefact at a distance close to of signal rebar reflections (Fig. 10b).

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
36 D. De Domenico, D. Campo and A.Teramo

In this regard it is to be observed that, even if the legibility of reflections are superimposed with those of rebars, so both are
the radargram is improved by the removal of multiple reflections removed by the deconvolution.
caused by the hollows, in particular with the effectiveness of the Therefore, the deconvolution and the envelope are pointless,
deconvolution, only the most intense anomalies are left visible, it is better to concentrate attention on the correct interpretation of
the same shown by the arrows in Fig. 9(b). Some rebars that are the artefacts to distinguish them from the real object.
still invisible, lie in a different position compared to the elements On the basis of the indications obtained from the analysis of
of the hollow brick wall (hollows, joints...), whose multiple synthetic data, the 3D surveys at a depth corresponding to the

FIGURE 7
Comparison among the EM
responses related to hollows,
with decreasing dimensions:
a) 30 m x 30 cm (as in Fig. 5b);
b) 25 m x 25 cm; c) 20 cm x
20 cm; d) 15 m x 15 cm; e)
10 cm x 10 cm; f) 5 x 5 cm, into
bricks. Decreasing the hollow
size, the number of multiple
reflections increases and they
tend to overlap until disappearing
in Fig. 7(f).

FIGURE 8
Models and relative synthetic pro-
files of the brick layer with two
square hollows distant a) 13 cm b)
8 cm and c) 1 cm, respectively,
placed on an air layer, built to
detect the lower interface of the
bricks, highlighted with red
dashed circles, which is recogniz-
able as a layer for the model a),
while, for the others, tends to
appear as a distinct object.

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
 FDTD modelling in high-resolution 2D and 3D GPR surveys 37

FIGURE 9
Models and synthetic radargrams
of hollow bricks on a layer of: a)
concrete; b) reinforced concrete
with 6 mm diameter metallic ele-
ments. The rebars beyond the
hollow brick wall generate anom-
alies but only some of them are
connected with hyperbolic reflec-
tions (shown by arrows); the
other ones (shown by circles) are
less clear and, in the interpreta-
tion step, they would be hardly
recognized without prior knowl-
edge of the wall geometry.

FIGURE 10
Further elaborations of the syn-
thetic radargram of Fig. 9(b): a)
application of predictive decon-
volution and diffraction migration
calibrated with the permittivity
values of the different materials
used in the numerical model; b)
wave envelope. Only the most
intense anomalies, shown by the
arrows in Fig. 9(b), are left visi-
ble, the others disappear, the
same shown by the circle and less
clear in Fig. 9(b).

actual position of the column bracket, beyond the first brick wall particular, another three simulations relative to the same model
and before the longitudinal rebar visible at 1.87 ns (Fig. 4f) were (Fig. 9b) of the hollow brick wall that bounds the r.c. column
revised and analysed better. In particular, in the time slice relative were realized but using different central frequencies, with the
to 1.75 ns (Fig. 11a), where the column edges are indicated by same processing described above (including predictive deconvo-
black lines, four horizontal anomalies (indicated by black arrows) lution, Butterworth band-pass and wave envelope). From the
with an intensity slightly different from those of multiple hollows comparison of the results obtained with the different signals, it
or brick joints are recognizable. In the x section shown in was found that: at 1 GHz (Fig. 12a) the resolution is such that
Fig. 11(b) the same four anomalies are well identified (denoted by brick and bar reflections are indistinguishable; at 1.6 GHz
black circles), whereas the others are close to the joints position (Fig. 12b), the bricks appear as a horizontal band where no
and therefore are indistinguishable. Such reflections could be details are visible but under it there are numerous anomalies,
related to the horizontal rebars, whose irregular spacing no greater some referable to metallic elements and corresponding to the
than the true one equal to 20 cm as obtained by the destructive test, correct distance between them but others due to artefacts; the
highlights that some rebars would not be likewise distinguishable 2.3 GHz (Fig. 12c) is the same as the one in Fig. 10(b), where
because of their position compared to the brick elements, as some but not all rebar anomalies are clear; and finally at 4 GHz
already noticed in the synthetic radargram shown in Fig. 10(b). (Fig. 12d) the very high resolution allows the clear definition of
Given that at a 2.3 GHz frequency the influence of hollows in the plaster thickness, invisible to the lower frequencies falling
the brick wall is significant, a further simulation was performed into their near-field region and the rebar reflections are stronger,
by varying the central frequency of the synthetic GPR pulse, to without any additional information with respect to the result
identify the most suitable frequency for this type of survey. In shown in Fig. 12(c). Therefore, the most suitable frequency for

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
38 D. De Domenico, D. Campo and A.Teramo

FIGURE 11
3D grid obtained by horizontal
and vertical profiles, processed
and migrated. a) Time slice of the
wall at 1.75 ns, at a 0.13 m depth:
the black lines indicate the col-
umn edges, the black arrows
show the anomalies hypothetical-
ly referable to the brackets. b)
The x section at 0.21m: the same
four anomalies are well identified
(denoted by the black circles),
whereas the others are close to
the joints position and therefore
are indistinguishable.

FIGURE 12
Comparison among EM respons-
es with different GPR signal fre-
quencies relative to the same
model of Fig. 9(b); each synthetic
scan was processed through a
predictive deconvolution and a
wave envelope. a) 1 GHz, the
resolution is such that brick and
bar reflections are indistinguish-
able; b) 1.6 GHz, the bricks
appear as a horizontal band where
no details are visible but under it
there are numerous anomalies
referable to the bars and artefacts;
c) 2.3 GHz like in Fig. 10 (b);
d) 4 GHz, the very high resolu-
tion allows the clear definition of
the plaster thickness of the wall
and the bar anomalies are more
concentrated, without any further
information compared to the
result shown in Fig. 12(c).

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
 FDTD modelling in high-resolution 2D and 3D GPR surveys 39

the rebar detection could be the 1.6 GHz that displays the greater REFERENCES
number of anomalies under the hollow brick wall referable to Balanis C.A. 1997. Antenna Theory, Analysis and Design. pp.129–132.
John Wiley & Sons, New York.
rebars but others due to artefacts even if the processing steps
Barrile V. and Pucinotti R. 2005. Application of radar technology to
must be properly developed. Nevertheless, the resolution of this reinforced concrete structures: A case study. NDT&E International 38,
frequency is not as high as the 2.3 GHz, to evaluate the concrete 596–604.
cover thickness, which is an important element for seismic vul- Binda L., Saisi A. and Zanzi L. 2008. Radar Investigation and diagnosis
nerability assessments. of historic masonry. Proceedings of the 12th International Conference
on Structural Faults and Repair, 10–12 June 2008, Edinburgh, on
CD-ROM.
CONCLUSIONS Bottari A., De Domenico D., Giannino F., Marino A. and Teramo A.
GPR surveys are currently used to locate buried objects such as 2003. On a GPR survey to characterize the structural configuration of
steel reinforcements and to characterize the structural element a masonry building. 3rd International Conference on NDT Chania,
geometry. The 2D and 3D GPR surveys carried out in a rein- Crete, Expanded Abstracts, 99–103.
Bungey J.H. 2004. Sub-surface radar testing of concrete: A review.
forced concrete column embedded in a double wall of hollow
Construction Building Material 18, 1–8.
bricks, have allowed the effective identification of longitudinal Bungey J.H., Shaw M.R., Millard S.G. and Molyneaux T.C.K. 2003.
rebars but not the transverse ones, the brackets, placed parallel to Location of steel reinforcement in concrete using ground penetrating
the direction of the longitudinal brick hollow axes. radar and neural networks. In: Structural faults and repair, (ed. M.C.
Through FDTD modelling, it was found that in each brick Forde), p. 8. London: Engineering Technics Press.
Cai J. and McMechan G.A. 1995. Ray-based synthesis of bistatic
hollow, multiple reflections arise interfering strongly with those
ground-penetrating radar profiles. Geophysics 60, 87–96.
generated by the adjacent hollows, due to its geometry and pres- Capizzi P. and Cosentino P.L. 2008. GPR multi-components data analy-
ence of air, generating artefacts. The particular geometric con- sis. Near Surface Geophysics 6, 87–95.
figuration of a hollow brick and its non-homogeneity from an Carcione J.M. 1996. Ground penetrating radar: Wave theory and numerical
electromagnetic point of view, cause strong conditioning for the simulation in lossy anisotropic media. Geophysics 61 (6), 1664–1677.
Casper D.A. and Kung K.S. 1996. Simulation of ground-penetrating
high-frequency GPR signal (whose wavelength is comparable to
radar waves in a 2-D soil model. Geophysics 61(4), 1034–1049.
the hollow size), so the reflections of objects, placed beyond the Che Way C., Chen Hua L. and Hung Sheng L. 2009. Measurement
hollow bricks, are inevitably affected. This phenomenon is par- radius of reinforcing steel bar in concrete using digital image GPR.
ticularly evident in the case of small and limited size objects, Construction and Building Materials 23, 1057–1063.
such as rebars and occurs only for acquisitions perpendicular to Cuiñas I. and Sanchez M.G. 2002. Permittivity and Conductivity
Measurements of Building Materials at 5.8 GHz and 41.5 GHz.
the hollow axis direction.
Wireless Personal Communications 20 (1), 93–100.
The modelling has shown the GPR response for the different Diamanti N., Giannopoulos A., Michael C. and Forde M.C. 2008.
elements of the hollow brick wall: the joints, the hollows and the Numerical modelling and experimental verification of GPR to investi-
division between them, allowing a correct interpretation of the gate ring separation in brick masonry arch bridges. NDT&E
hollow bricks. The migration and deconvolution processes, even International 41, 354–363.
Ellefsen K.J. 1999. Effects of layered sediments on the guided wave in
though improving the legibility of the radargram by the removal
crosswell radar data. Geophysics 64, 1698–1707.
of multiple reflections caused by the hollows and, unfortunately, Giannopoulos A. 2005. Modelling ground penetrating radar by GprMax.
also by the rebars, leave only the most intense anomalies visible, Construction and Building Materials 19, 755–762.
the same shown in the unmigrated section. Giannopoulos A. and Warren C. 2007. Numerical modelling of commer-
Knowing the typical reflection of the hollow brick, it is pos- cial GPR antennas. 13th European Meeting of Environmental and
Engineering Geophysics, 3–5 September, Istanbul.
sible to recognize the artefacts and focusing on the interpretation
Goodman D. 1994. Ground-penetrating radar simulation in engineering
of the correct depth of the rebars, it is possible to obtain informa- and archaeology. Geophysics 59, 224232.
tion about them but a complete identification is often impossible Grasmueck M., Weger R. and Horstmeyer H. 2003. How dense is dense
depending on their position compared to the joints and the hol- enough for a ‘real’ 3D GPR Survey? SEG Expanded Abstracts 22,
low separators. 1180.
How-Wei C. and Tai-Min H. 1998. Finite-difference time-domain simu-
More complicated investigations might be arranged, for
lation of GPR data. Journal of Applied Geophysics 40, 139–163.
example, using grids with a thicker spacing or antennas with a Jol M.H. 2009. Ground Penetrating Radar: Theory and Applications.
lower central frequency, to obtain a low level of influence of hol- Elsevier 9, 46
lows, as already confirmed by tests carried out at different fre- Klysz G., Ferrieres X., Balayssac J.P. and Laurens S. 2006. Simulation
quencies with FDTD, even if in this case the decrease in resolu- of direct wave propagation by numerical FDTD for GPR coupled
antenna. NDT&E International 39, 338–347.
tion could cause the loss of significant information like the con-
Kunz K.S. and Luebbers R.J. 1993. The Finite Difference Time Domain
crete cover thickness or the rebar pitch when it is too narrow. In Method for Electromagnetics. CRC Press.
both cases it is necessary to assess the relationship between Lualdi M. and Zanzi L. 2006. Analytical and experimental evaluation of
advantages (full characterization of rebars) and disadvantages the Radar Cross Section of concrete reinforcements. Proceedings of
(high-acquisition time in the first case and resolution loss in the the 2006 NDE Conference on Civil Engineering, St. Louis (MO),
14–18 August 2006.
second one).

© 2013 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2013, 11, 29-40
 40 D. De Domenico, D. Campo and A.Teramo

Lualdi M., Zanzi L. and Binda L. 2003. Acquisition and processing penetrating radar and neural networks. NDT&E International 38(3),
requirements for high quality 3D reconstructions from GPR investiga- 203–212.
tions, CD-ROM. Proceedings of the International Symposium Non Shaw M.R., Molyneaux T.C.K., Millard S.G., Taylor M.J. and Bungey
Destructive Testing in Civil Engineering NDT-CE 2003, September J.H. 2003. Assessing bar size of steel reinforcement in concrete using
16–19, Berlin. ground penetrating radar and neural networks. Non-Destructive
Lui Q.H. and Fan G. 1999. Simulations of GPR in dispersive media using Testing and Condition Monitoring 45(12), 813–816.
a frequency-dependent PSTD algorithm. IEEE Transactions on Stolt R.H. 1978. Migration by Fourier Transform. Geophysics 43, 23–48.
Geoscience and Remote Sensing 37, 2317–2324. Taflove A. and Hagness S.C.. 2005. Computational Electrodynamics.
McCann D.M. and Forde M.C. 2001. Review of NDT methods in the The Finite Difference Time-Domain Method, Third Edition. Artech
assessment of concrete and masonry structure. NDT&E International House, Boston.
34(2), 71–84. Utsi V. and Utsi E. 2004. Measurements of reinforcement bar depths and
Millard S.G., Shaw M.R., Giannopoulos A. and Soutsos M.N. 1998. diameters in concrete. Proceedings of the 10th International
Modelling of Subsurface Pulsed Radar for Non-destructive Testing of Conference on Ground Penetrating Radar, Delft, 21–24, June 2004.
Structures. Journal of Material in Civil Engineering 10, 188. Xian-Qi H., Zi-Qiang Z., Qun-Yi L. and Guang-Yin L. 2009. Review of
Powers M.H. and Olhoeft G.R. 1994. Modelling dispersive ground pen- GPR rebar detection PIERS. Proceedings of PIERS (Progress in
etrating radar data. Proceedings of the 5th International Conference Electromagnetics Research Symposium) 2009, Beijing, China, 804–13.
on Ground-Penetrating Radar, Waterloo, Ontario 1994, 173–183. Yee K.S. 1966. Numerical solution of initial boundary value problems
Radzevicius S.J. and Daniels J.J. 2000. Ground penetrating radar polari- involving Maxwell’s equations in isotropic media. IEEE Transactions
zation and scattering from cylinders. Journal of Applied Geophysics on Antennas and Propagation 14(3), 302–307
45, 111–125. Yelf R. 2004. Where is the true time zero? Proceedings of the 10th
Sandmeier K. 2010. Reflex version 5.5, user’s manual. International Conference on Ground-Penetrating Radar, Delft, The
Shaari A., Ahmad R.S. and Chew T.H. 2010. Effects of antenna target Netherlands 2004, 279–282
polarization and target medium dielectric contrast on GPR signal from Zanzi L. and Lualdi M. 2008. Recenti progressi nella tecnologia GPR e
non-metal pipes using FDTD simulation. NDT&E International 43(5), loro impatto sulle applicazioni per la diagnostica. Il Giornale delle
403–408. Prove non Distruttive Monitoraggio Diagnostica 4, 47–53.
Shaari A., Millard S.G. and Bungey J.H. 2004. Modelling the propaga- Zeng X. and McMechan G.A 1997. GPR characterization of buried tanks
tion of a radar signal through concrete as low-pass filter. NDT&E and pipes. Geophysics 62, 797
International 37, 237–42. Zeng X., McMechan G.A., Cai J. and Chen H.W. 1995. Comparison of
Shaw M.R., Millard S.G., Molyneaux T.C.K., Taylor M.J. and Bungey ray and Fourier methods for modelling monostatic ground-penetrating
J.H. 2005. Location of steel reinforcement in concrete using ground radar profiles. Geophysics 60, 1727–1734.

24-27 November 2013 | Al Ain, UAE | www.eage.org

Second International Conference on

Call for Papers deadline
Engineering 15 June 2013
Geophysics
ICEG13-V2H.indd 1 09-01-13 16:38

View publication stats