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

net/publication/230905679

Ground penetrating radar and microwave tomography 3D applications for the

deck evaluation of the Musmeci bridge in Potenza, Italy

Article in Journal of Geophysics and Engineering · August 2011

DOI: 10.1088/1742-2132/8/3/S04

CITATIONS READS

21 2,203

7 authors, including:

Massimo Bavusi F. Soldovieri

Italian National Research Council Italian National Research Council
49 PUBLICATIONS 527 CITATIONS 569 PUBLICATIONS 8,676 CITATIONS

SEE PROFILE SEE PROFILE

Antonio Loperte Antonio Di Cesare

Italian National Research Council Università degli Studi della Basilicata
79 PUBLICATIONS 1,034 CITATIONS 51 PUBLICATIONS 762 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Antonio Di Cesare on 19 March 2020.

The user has requested enhancement of the downloaded file.

Home Search Collections Journals About Contact us My IOPscience

Ground penetrating radar and microwave tomography 3D applications for the deck evaluation

of the Musmeci bridge in Potenza, Italy

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2011 J. Geophys. Eng. 8 S33

(http://iopscience.iop.org/1742-2140/8/3/S04)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 79.23.39.36
The article was downloaded on 23/08/2011 at 17:27

Please note that terms and conditions apply.

IOP PUBLISHING JOURNAL OF GEOPHYSICS AND ENGINEERING
J. Geophys. Eng. 8 (2011) S33–S46 doi:10.1088/1742-2132/8/3/S04

Ground penetrating radar and microwave

tomography 3D applications for the deck
evaluation of the Musmeci bridge in
Potenza, Italy
Massimo Bavusi1 , Francesco Soldovieri2 , Rosario Di Napoli2 ,
Antonio Loperte1 , Antonio Di Cesare3 , Felice Carlo Ponzo3 and
Vincenzo Lapenna1
1
IMAA-CNR, C. da S. Loja, 85100 Zona Industriale, Potenza, Italy
2
IREA-CNR, Via Diocleziano, 198, Napoli, Italy
3
DiSG-Università degli Studi della Basilicata, Potenza, Italy
E-mail: bavusi@imaa.cnr.it, loperte@imaa.cnr.it, lapenna@imaa.cnr.it, soldovieri.f@irea.cnr.it,
dinapoli.r@irea.cnr.it, felice.ponzo@unibas.it and antodice@yahoo.it

Received 14 January 2011

Accepted for publication 18 May 2011
Published 23 August 2011
Online at stacks.iop.org/JGE/8/S33

Abstract
An extensive experimental and numerical investigation has been carried out to assess the status
of the ‘Ponte sul Basento’ (1967–1976), in the town of Potenza (Basilicata region, southern
Italy), better known as the Musmeci bridge. Architecturally, the bridge is a considerable
reinforced 20th century concrete structure that was designed and built by the Italian architect
Sergio Musmeci (1926–1981). Moreover, the bridge represents an important element of the
infrastructural network, linking the city centre to the Potenza-Sicignano highway, crossing the
Basento river and the railway close to the main train station of the city. Recently, due to ageing
and continuous and significant traffic, the bridge started to be affected by several problems
such as water infiltration. Within the presented study, a widespread ground penetrating radar
(GPR) survey has been designed to investigate the geometrical characteristics of the bridge
deck (Gerber saddles, internal stiffening walls, pillar supports) and detect the presence of
defects or damage due to water infiltration and traffic fatigue. Concerning this, a 900 MHz 3D
GPR survey has been performed along a zone of one of the lanes on the road surface.
Moreover, a second 1500 MHz 3D survey has been carried out at the bottom of the bridge deck
in order to gain detailed information about an important structural element of the bridge, the
Gerber saddle. Both results have been processed following two approaches: the first a
classical time-domain processing session based on commercial software and the use of
migration; the second in microwave tomography, an advanced frequency domain automatic
PC-based inversion algorithm. In this paper, we present a comparative interpretation of both
kinds of processed results, and provide considerations about the investigated structures.

Keywords: ground penetrating radar, processing, microwave tomography, bridge deck,

reinforced concrete
(Some figures in this article are in colour only in the electronic version)

1742-2132/11/030033+14$33.00 © 2011 Nanjing Geophysical Research Institute Printed in the UK S33

M Bavusi et al
1. Introduction
Most of the civil structures and infrastructures (bridges, dams,
viaducts, tunnels, etc) of the last century have been built using
reinforced concrete. This is a relatively young material which,
in the last 20 years, has shown ageing problems related to
environmental factors such as water infiltration, reinforcement
corrosion with delamination and carbonation of the concrete
cover. These problems have become more worrying for
important structures such as bridges, where the heavy traffic
conditions can determine a more rapid deterioration compared
to other structural typologies. A series of non-destructive
testing (NDT) techniques have been planned in order to
investigate such problems (McCann and Forde 2001, Dérobert
et al 2002, Scott el al 2003).
Ground penetrating radar (GPR), when used at suitably
high frequencies (from 900 to 2500 MHz), can be considered
an NDT technique, providing precious information about the
presence of ‘embedded’ objects such as reinforced rebars
(Shaw et al 2005, Che et al 2009), and embedded ‘defects’
such as fractures and voids, by using special antennas (Huston
et al 2000, Forest and Utsi 2004, Utsi et al 2008). Moreover,
the GPR technique can help determine the concrete moisture
content (Shaari et al 2004, Hugenschmidt and Loser 2008, Lai
et al 2009, 2010).
Bridge inspection work can benefit from the GPR
technique (Scott et al 2003, Hugenschmidt and Mastrangelo
2006, Belli and Wadia-Fascetti 2008, Hugenschmidt et al
2010), and in the absence of all original project documentation
and particularity of the considered structure (inner rebars,
tendons, boundary conditions, anchors and other internal
element), the bridge inspection can also benefit from the
application of the GPR technique for the 3D survey design
and processing (Hugenschmidt et al 2010).
This work concerns GPR measurements carried out at
‘Ponte sul Basento’ (1967–1976) in the town of Potenza
(southern Italy), better known as the ‘Musmeci bridge’,
derived from the name of its designer Sergio Musmeci (1926–
1981). Such measurements have been performed in the
framework of an extensive experimental and numerical study
within the ‘Integrated System for Transport Infrastructures
surveillance and Monitoring by Electromagnetic Sensing
(ISTIMES)’ and ‘Ponte Musmeci Potenza: Analisi del
degrado dei materiali e progettazione interventi, Parte 1 e
2’ projects. The bridge is a considerable structure for the
city due to its relevant architectural aspects and its operative
function. This study was requested by the bridge’s owners
following the growth in traffic volume in recent years, in order
to assess the structural safety levels. Due to the particular
shape of the structure and the incomplete original project
documentations, several destructive testing (DT) and non-
destructive testing (NDT) techniques, based on traditional
technologies (extraction of carrots, pacometer test, combined
SonReb method, sclerometer-ultrasonic waves, termocamera,
etc) and innovative ones (GPR, optical fibres, laser-scanner
3D survey, etc), have been combined.
A GPR survey composed of two 3D datasets has been
carried out on a part of the Musmeci bridge; this investigation
S34
is aimed at evaluating its deterioration/conservation status,
particularly concerning its connections (Gerber saddle
system), where water infiltration can occur more easily, and
detecting some structural elements such as stiffening walls
and pillar supports. The first 3D dataset was gathered in
February 2010 and consisted of 26 parallel profiles 30 m long
and spaced at 10 cm, covering an entire external lane and
carried out using a 900 MHz antenna. The second 3D dataset,
composed of 29 radargrams 2 m long and spaced at 10 cm, was
gathered in May 2010 using a 1500 MHz antenna on the lower
surface of the bridge deck, in order to adequately characterize
the joint connection (Gerber saddle) between two consecutive
structural elements of the deck. The in situ measurements
have been planned in order to avoid interruption of the traffic
conditions as much as possible.
The achievement of interpretable radargrams and 3D GPR
data requires good survey design, good operator’s field skill
as well as a reliable processing procedure. The most common
way to process the radargram consists of a series of numerical
procedures (geometrical correction, filtering, gain correction,
migration, etc) acting as filtering and focusing of the raw
data. On the basis of his or her skill, the operator acts
sequentially on the data in order to geometrically correct the
radargram, enhance the amplitudes, suppress the noise, focus
the reflectors and convert time in depth. This processing
approach can achieve good results, but is constrained by
the skill of the operator. In other words, the final result
depends on the number, kind, order and features of the
applied filters and, most important of all, on the operator
choices. A number of commercial processing software
packages can be deployed, providing satisfactory results for
most applications. However, an accurate processing system
that avoids subjectiveness problems and exploits all spatial
resolution is needed, especially in GPR concrete scanning
where the small dimensions of the embedded objects and the
dielectric features of the concrete pose a challenge in terms of
processing and interpretation.
Then, a valid processing alternative can be relied on an
automatic PC-based procedure, ensuring objectiveness and
speed, and microwave tomography (MT) can be this suitable
processing approach (Soldovieri and Solimene 2010). The
application of MT has provided good results in cultural
heritage problems (Leucci et al 2007, Bavusi et al 2010),
archaeology (Soldovieri and Orlando 2009) and other realistic
applications (Soldovieri et al 2007). This novel approach is
based on more accurate formulation of the electromagnetic
scattering phenomenon model and robust reconstruction
procedures, when compared to the ones usually adopted
in the classic GPR processing approach (Soldovieri et al
2007, Bavusi et al 2010). Following this strategy, after the
classical processing session, both datasets have been inverted
using MT. In particular, MT has been applied on each 2D
radargram of every dataset. Then inverted radargrams have
been interpolated in order to build a data volume after being
visualized through the ore significant constant depth slices.
The results have been compared with the respective classical
processed data.
This paper is organized as follows. Section 2 provides
an overview of the description of the Musmeci bridge and

(a)
Figure 1. (a) Location of the Musmeci bridge. (b) Detail of a span.
(a)
(b)
Figure 2. The Musmeci bridge: (a) longitudinal cross section of a span with the main elements of the span and deck; (b) front figure of the
entire bridge formed by four spans.
is devoted to the description of the survey design for both
datasets. In section 3, both processing strategies are described
and applied to the data. Section 4 is devoted to the evaluation
and comparison of the processing performances.
2. The bridge and survey design
The ‘Ponte sul Basento’ or ‘Musmeci bridge’ represents an
important flyover linking the city to the Potenza-Sicignano
road, which crosses the Basento river and the Napoli–Potenza–
Taranto railway (figure 1(a)).
The bridge was designed by Sergio Musmeci in the 1960s
and represents a true work of art in reinforced concrete which
influenced the architectonic culture of the 20th century. In fact,
the complex shape of the inferior vault (figure 1(b)), based on
the concept of uniform stress on all surfaces, makes it a unique
architectonic piece, designed when computer assisted design
was in its infancy.
From a structural point of view, the deck of the
Musmeci bridge constitutes a box section made of reinforced
concrete held every 17.30 m by a continuous reinforced
Ground penetrating radar and microwave tomography 3D applications
(b)
concrete vault equivalent to four arches with wheelbases of
69.20 m and a free span amongst the supports of 58.80 m
(figure 2(a)); the entire bridge has an overall length of
69.20 m × 4 = 277 m (figure 2(b)).
The box section of the deck has longitudinal and
transversal stiffening walls (internal sects), each 3.46 m long
(figure 3(a)). Its transversal cross section shows an ellipsoidal
shape in the lower side with a maximum thickness of 1.5 m
and width of 16 m (figure 3(b)). The continuity of the deck is
interrupted by a Gerber saddles system (figures 3(c) and (d)),
which realizes simply laid beams (each 10.38 m) and other
beams (each 24.22 m long) lying on supports (17.30 m distant,
see figure 2(a)) and cantilevers of 3.46 m. The road pavement
is formed by 4 cm of binder and 3 cm of asphalt.
Since its construction (1971–1975), the ‘Ponte sul
Basento’ has remained substantially unchanged until the
present day. However, the traffic growth that has taken place in
Potenza since the 1980s represents an important factor of stress
for the structure, together with ageing problems affecting the
reinforced concrete and the rain water infiltration phenomena,
especially in the Gerber saddles (figure 4).
S35
M Bavusi et al

(a) (d )

(b)

(c)

Figure 3. Deck structure of the Musmeci bridge: (a) inner view showing the square grid of reinforced longitudinal and transversal stiffening
walls located every 3.46 m; (b) transversal cross section of the deck with its reinforced sects every 3.46 m; (c) longitudinal cross section of
the deck with the internal rebars and Gerber system saddle (in the circle); (d) Gerber saddle detail with internal rebars.

(a) (b)

(c) (d )

Figure 4. Details of the Musmeci bridge: (a) central span; (b) Gerber saddle; (c) detail of the Gerber saddle with concretions; (d) detail of
the Gerber saddle with rebar corrosion and cover breaking.

S36
 Ground penetrating radar and microwave tomography 3D applications

(a)

(c)
(b)

Figure 5. First GPR survey area: (a) the location with respect to the structure of the bridge; (b) detail of the surveyed lane; (c) survey design
with all traces of the 26 radargrams.

In order to investigate the bridge deck, a GPR survey Table 1. Three-layer model used in classical migration and MT
was performed along a zone of the lane on the second span inversion.
in the highway direction (figures 5(a) and (b)). Figure 5(c) Thick Velocity
depicts the detail of the lane survey design. In particular, Layer (m) Material (m ns−1) ε
30 parallel, 26 m long radargrams, spaced at 0.1 m were Shallow 0.15 Asphalt and 0.15 4
gathered along the trait of the lane using a GSSI SIR 3000 reinforced concrete
acquisition unit on a 900 MHz antenna mounted on a cart Medium 0.40 Mainly void 0.30 1
provided by the survey wheel set providing a trace increment Deep 0.10 Reinforced concrete 0.1 9
of 0.015 m. In order to keep the antenna path as rectilinear
as possible, the survey was run very slowly by checking the 3.1. Classical processing approach
position of the antenna with respect to a metric tape that
was opportunely tightened. Moreover, the starting and final Classical processing involved the following operation:
positions of the centre antenna were referred to references • standard 2D processing and migration carried out on each
preventively drawn on the asphalt. The remaining positioning radargram of the dataset;
errors were corrected in the processing phase. • resampling of each radargram and interpolation in order
Moreover, in order to obtain a more detailed investigation to build a 3D dataset;
of the Gerber saddle, a second 3D survey was performed on • extraction of significant time/depth slices at proper
the lower surface of the bridge deck corresponding to the time/depth;
Gerber saddle (figure 6(a)). In particular, 29 parallel, 2 m • the time/depth-slice misalignment corrected by using a
long, with 0.1 m spacing radargrams have been gathered specific algorithm.
using the GSSI SIR 3000 acquisition unit equipped with a In particular, for the first dataset on the road surface:
1500 MHz antenna mounted on the survey cart with an encoder
• the standard 2D processing included zero-time correction,
set to provide a trace increment of 0.006 m (figures 6(b) and removed header gains, AGC gain and time cut;
(c)). The lower margin of the Gerber saddle falls at 1.0 m • the migration has been based on a three-layer velocity
along each radargram. model suggested by the velocity analysis together with
some considerations about the box structure of the bridge
deck (table 1);
3. Data processing and results • the resampling has been performed with a trace increment
of 0.03 m in order to reduce the data number and
The processing has been carried out following two strategies: consequently the computational time required during the
the classical processing routine and the MT approach. 3D interpolation.

S37
M Bavusi et al

(a)

(b)
(c)

Figure 6. Second GPR survey area: (a) the location with respect to the structure of the bridge; (b) detail of the surveyed area; (c) survey
design with all traces of the 29 radargrams.

(a)

(b)

Figure 7. Processed radargram (#19) of the first GPR survey (a), with the interpretation of its main reflectors on the basis of the original
deck cross section (b).

As an example of the radargrams, figure 7(a) shows radargram permits the main structural elements such as the
the processed radargram #19 with the interpretation of the upper horizontal flat reflector (0–2.0 ns) to be referable to
elements of the deck plan in cross section (figure 7(b)). The the asphalt. This layer contains evidence of the Gerber

S38

(a) (b)
Figure 8. Constant time slices in arbitrary units obtained by interpolating all 26 radargrams: (a) time slice at 0.8 ns (0.06 m); (b) time slice
at 2.5 ns (0.20 m); (c) time slice at 3.6 ns (0.40 m); (d) time slice at 4.8 ns (50 cm). ab: absorption zone, Gs: Gerber saddle, ps: pillar
support, sw: stiffening wall, ls: longitudinal stiffening wall.
saddles appearing as absorptive zones, probably due to the rain
water infiltration as denoted by the presence of concretions
and corrosion of rebar (see figure 4). Then, a next layer,
visible between 2.0 and 3.2 ns, producing several steep dipping
reflectors can be referred to a reinforced concrete layer with its
rebars. Moreover, a layer formed mainly by ‘void’ is located
between 3.2 and 5.8 ns. This layer shows vertical reflections
probably due to the stiffening walls and the larger pillar
supports forming the box structure depicted in figure 4(a).
Then, the layer visibly close to 6.0 ns can be related to a
concrete plate.
Deeper, a similar box structure, formed by a continuous
signal with reverse polarity with respect to the upper one,
seems to occur. On the basis of the civil engineering
information about the structure, this lowest continuous signal
may be due to the presence of multiple reflections, although
the complexity of the investigated structure does not allow us
to foresee its polarity at this stage.
The processed radargrams have been interpolated in order
to obtain a data volume of the investigated area. Then, four
time/depth slices in arbitrary units have been built with the
aim of pointing out the main structural elements of the bridge
(figure 8). A specific algorithm has been applied to the
data to correct misalignments in the radargrams appearing
Ground penetrating radar and microwave tomography 3D applications
(c) (d )
as a ‘staggering’ in the linear features in the time slices. At
0.8 ns, it is possible to see the Gerber as absorptive regions
(figure 8(a)). Moreover, other concentrated absorptive zones
can be related to water infiltration in the asphalt layer. Then,
at 2.5 ns the stiffening walls also appear (figure 8(b)), and
at 3.6 ns a new linear feature parallel to the survey direction
appears (figure 8(c)). It can be interpreted as a longitudinal
reinforcement sect (see figure 4(a)). Finally, in depth, no
important features can be seen (figure 8(d)). In fact, the visible
structures are just the ‘shadows’ produced by the overlooking
structures.
The second dataset (on the bottom of the deck) has been
processed as follows.
• Zero-time correction, remove header gains, AGC gain,
time cut, FK filtering, migration. Hyperbolae adaptation
velocity analysis provided a velocity average value of
0.08 m ns−1 (ε = 14), which is consistent with a wet
concrete. In fact, the second dataset was gathered after
a rainy period, and the Gerber saddles are a preferential
route for water infiltration.
This dataset required an FK filtering with respect to the
processing of the first one in order to mitigate the diffraction
and dipping reflectors produced by the rebars.
S39
M Bavusi et al
(c)
(a)
Figure 9. (a) Processed radargram (#10) of the second GPR survey compared with the real-section view and (b) the original plan in the
cross section (c).
Figure 9(a) shows processed and interpreted radargram
#10 compared with the real geometry of the Gerber saddle
visible on the southern side of the bridge (figure 9(b)) and the
original cross section (figure 9(c)). A series of point shallow
reflectors is related to the transversal rebars. Moreover, it is
possible to detect several strong reflectors related to the inner
longitudinal rebars and a central absorptive zone referable
to the neoprene rubber layer (figure 9(a)). Furthermore, at
0.3 m depth, the signal shows multiple reflections of the same
structures.
No other relevant absorptive zone liable to water
infiltration is detected. This evidence, apparently in
disagreement with the previous 3D dataset collected on the
road surface and that shows strong absorptive zones close to
the Gerber saddle, can be explained if we consider that the two
datasets were collected at different times, with the second one
carried out in a spring month, such as May.
These features slightly change by moving from one
radargram to another. A good way to visualize the
reflector changes in cross direction is to build constant time
slices. In figure 10 four time slices extracted respectively
at 0.05 m (a), 0.08 m (b), 0.14 m (c) and 0.20 m (d) are
shown. They represent a subset of the gathered data of size
2.9 × 1.4 m. In spite of the selected trace increment set to
S40
(b)
0.006 m during the acquisition, the data quality was not
good, probably because of the less than perfect contact
between the survey wheel and the surface to be scanned.
In fact, the overturned position of the antenna together with
the harshness of concrete produced a mispositioning error
of scans and consequently a possible local undersampling.
The mispositioning correction restored the alignment between
several radargrams, but does not compensate for the
undersampling, although with due caution an interpretation
can be made.
At 0.05 m depth (figure 10(a)), several strong horizontal
linear reflectors appear below the trace of the Gerber
saddle. They can be related to the rebars of the lower
part of the Gerber saddle (A in figure 9(b)), while they
disappear above. Another feature is a middle-reflective zone
between 1.3 and 1.5 m along the x-axis, referable to other
rebars parallel to the survey direction. These features are
confirmed in the constant time-slices of figure 10 showing
the reflection intensities in arbitrary units, in particular at
0.08 m (figure 10(b)) where they appear more defined. At
0.14 m depth (figure 10(c)), the time slice cuts only the
neoprene rubber layer zone. The saddle zone does not provide
reflections, while the central linear reflector is visible only
in the central-lower portion of the diagram. At a depth of

(a)
(c)
Figure 10. Time slices in arbitrary units built starting from the second dataset: (a) time slices at 1.6 ns (0.05 m); (b) time slices at 2 ns
(0.08 m); (c) time slices at 5 ns (0.14 m); and (d) time slices at 5.8 ns (0.20 m).
0.20 m (figure 10(d)), any structure is well delineated even if
some stronger reflectors can be seen. They can be due to the
diagonal reinforcements of figure 9.
3.2. Microwave tomography
The microwave tomographic approach has already been
presented in many papers (Leucci et al 2007, Persico et al
2010). Here we recall the details of the approach (see step
1, below) in order to make the technical content of the paper
consistent.
Due to the 3D nature of the problem, we applied a strategy
based on a 2D slice based inversion approach (Solimene et al
2007). This strategy consists of the following steps.
• Step 0: pre-processing of the data.
• Step 1: tomographic reconstruction of the 2D datasets.
• Following steps 0 and 1, data were transferred from 2D to
3D.
• Step 2: superimposition and interpolation of the 2D
reconstructions to obtain a 3D representation.
Step 0. The pre-processing of the data. As said above, the
measurements have been collected by a commercial GPR in
the time domain. In particular, the recorded data account for
the total field (given as the contribution of the direct wave, the
field reflected by the interface and the field backscattered by
the targets) in the time domain. As mentioned previously, the
Ground penetrating radar and microwave tomography 3D applications
(b)
(d )
inverse scattering approach works differently with the only
field backscattered by the targets in the frequency domain.
Therefore, pre-processing is necessary to turn the GPR
measurements into data suitable for the inversion algorithm.
The first step is to ‘gate’ the first part of all the time
domain traces, which corresponds to erasing the direct, surface
and first layers, wave contributions; this step roughly provides
an estimation of the scattered field. The gating consists of
constraining the initial part of all the traces of the radargram
equal to zero. Finally, after choosing the time-zero gate, we
Fourier transformed the gated data to the frequency domain.
Step 1. Tomographic reconstruction of the 2D profiles. In
this step, the MT approach is applied to the vertical profiles
(radargrams) with the aim of giving focalized 2D images of
the targets (rebars, etc).
The geometry of the problem consists of two half-spaces
(air and concrete) separated by a planar interface at z = 0
(figure 11). The upper half-space is made up of free space
with (dielectric permittivity and magnetic permeability ε0 and
μ0, respectively). The lower half-space (representative of the
structure to be investigated) is assumed to be homogeneous and
with relative dielectric permittivity εb , electrical conductivity
σ b and no magnetic properties.
The incident field source is a time-harmonic (time
dependence exp(j2π ft)) filamentary y-directed electric current
(TM-polarization) of infinite extent and invariant along the
y-axis. The targets are invariant along the y-axis and their
S41
M Bavusi et al
Figure 11. Geometry of the problem.
cross section is contained in the rectangular investigation
rectangular domain D = [−a, a] X [zmin , zmin + 2b]. Thus
a two-dimensional geometry is considered (Liseno et al 2004).
A multi-monostatic observation configuration is assumed,
where the locations of the transmitting and receiving antennas
are assumed to be coincident. The radiation is multi-frequency
in the band [f min, f max]. The unknowns of the problem
are the relative dielectric permittivity profile εD (x, z) and the
conductivity profile σD (x, z) inside the investigation domain
D. The problem is then recast in terms of the so-called unknown
contrast function, defined as
εeq (x  , z ) − εeqb
χ (x  , z ) = (1)
εeqb
(a)
(b)
Figure 12. MT inversion results of the first GPR survey in terms of contrast function normalized with respect to the absolute maximum (a),
with the interpretation of its main reflectors on the basis of the original deck cross section (b).
S42
 
where εeq (x  , z ) = ε0 εD (x  , z ) − j σD2πf
(x ,z )
, while εeqb =
ε0 εb − jσb /2πf denotes the equivalent dielectric permittivity
in the investigated medium.
Under the Born approximation (Slaney et al 1984, Chew,
1995, Hugenschmidt et al 2010, Persico et al 2010) the
unknown-data relationship is provided by the integral equation

Es (xs , ω) = ks 2
Ge (xs , ω, r )Einc (xs , ω, r )χ (r  ) dr  (2)
D
where Es (datum of the problem) denotes the scattered electric
field probed at the air–concrete interface at the point xs , and ks
is the wave number in the lower half-space. The scattered field
is collected within an observation domain at the air–concrete
interface for xs ranging from−xM to xM .
The external Green’s function Ge and the incident field
Einc (the incident (unperturbed) electric field in the lower half-
space, i.e. the electric field in the absence of the scattering
objects) are the key ingredients in defining the relevant linear
to be inverted in equation (2). Their expressions for the
considered half-space geometry are given under the spectral
form in Persico et al (2010).
The problem of reconstructing the targets inside D
amounts to solving the integral relationship (2) where the
unknown is the contrast function, while the ‘datum’ is
the scattered field. The inversion of the linear integral
equation is performed thanks to the truncated singular value
decomposition (TSVD) that allows us to obtain a stable
solution to the problem (Slaney et al 1984, Chew 1995, Persico
et al 2010).
 Ground penetrating radar and microwave tomography 3D applications

(a) (b) (c)

Figure 13. Depth slices built starting from inverted radargrams of the first dataset. (a) Depth slices at 0.06 m; (b) depth slices at 0.20 m; and
(c) depth slices at 0.40 m. Data are shown in terms of contrast function normalized with respect to the absolute maximum.

Step 2. 3D representation. The third step aims at achieving
a pseudo 3D representation of the investigated structure via
the approach presented in Solimene et al (2007), where the
volumetric cube is obtained by superposing and interpolating
the 2D tomographic reconstruction of the vertical profiles.
In the surveys, the GPR system exploits antennas radiating
a linearly polarized field; this entails that the reconstruction
results will not be of comparable quality along the two axes in
the plane of movement of the antennas.
In order to simplify the readability of the dataset, several
depth slices can be extracted at proper depths. The depth
slices can suffer from ‘staggering’ due to the misalignment of
the radargrams during the acquisition due to several factors.
The entire procedure has been applied at both datasets to
produce inverted depth slices. In particular, for the first
dataset, the procedure has been applied to three sub-datasets.
In fact, each radargram has been divided into three layers
due to the high contrast in the dielectric permittivities of

S43
M Bavusi et al
(a)
(c )
Figure 14. Depth slices built starting from the inverted radargrams of the second dataset at 0.05 m depth. (a) Depth slice at 0.05 m; (b)
depth slice at 0.08 m; (c) depth slice at 0.14 m; and (d) depth slice at 0.20 m. Data are shown in terms of contrast function normalized with
respect to the absolute maximum.
several materials constituting the bridge deck. The model of
dielectric permittivities is the same, used in order to migrate
the radargrams in the classical processing approach (table 1).
Figure 12(a) shows the results achieved by applying the
inversion approach to radargram #19 of the first dataset in terms
of the contrast function normalized with respect to the absolute
maximum. Actually, the radargram is formed by a puzzle of
three layers inverted separately using three different dielectric
permittivities, as given in table 1. The layered radargram
allows us to detect the main structural features such as the
upper and lower concrete plates, the stiffening walls, the pillar
supports and, finally, the Gerber saddles (figure 12(b)).
In figure 13, three depth slices built by interpolating
all inverted radargrams show the horizontal distribution of
the normalized contrast function at three specified depths.
Moreover, a specific algorithm has been designed and
applied to cancel the mispositioning error. From top to
bottom it is possible to observe several features: with
reference to figure 13(a), the Gerber saddles appear absorptive,
in agreement with the previous classical processed data.
Moreover, an important reflective feature appears at about
6 m along the y-axis. It can be related to the presence of
the pillar support. A less defined reflection referable to the
other pillar support appears at about 22 m along the y-axis.
Figure 13(b) shows a deeper layer (0.20 m) where reflectors
referable to the stiffening walls appear very defined. Actually,
the first one appears thicker than expected for the pillar support.
Finally, in depth (0.40 m), a new longitudinal feature appears,
in agreement with the one observed in the classical processed
data (figure 13(c)).
S44
(b)
(d )
The second dataset has also been inverted by using MT
and, as well as for the classic processing, a relative dielectric
permittivity of 14 has been used. Then, some depth slices were
built and were subjected to the same mispositioning correction.
Figure 14 shows the depth slices with normalized contrast
function at 0.05 m (a), 0.08 m (b), 0.14 m (c) and 0.20 (d) m.
At 0.05 m depth (figure 14(a)), several strong horizontal linear
reflectors appear both below and above the trace of the Gerber
saddle, especially in the left portion of the diagram. Between
1.3 and 1.5 m along the x-axis, a strong reflector referable to
other rebars parallel to the survey direction appears. As well
as figure 10(b), these features are confirmed in the depth slice
at 0.08 m (figure 14(b)), where they appear better defined. At
0.14 m depth (figure 14(c)), all the described features appear
to be still well delineated. Finally, at 0.20 m (figure 14(d)),
it is possible to see some horizontal linear reflectors and a
well-defined vertical reflective zone.
4. Discussion and conclusion
In this work, GPR surveys have been carried out in order to
investigate the deck structure of the Musmeci bridge in Potenza
town (Basilicata region, southern Italy). In particular, two 3D
GPR datasets have been achieved above and below the bridge
deck, respectively. The first one, gathered in February 2010,
was carried out using a 900 MHz central frequency antenna
with the aim of providing information about the main structural
elements of a part of the bridge.
The second one, carried out in May 2010, made use of a
1500 MHz central frequency antenna with the aim of achieving

detailed and high-resolution information/images of the Gerber
saddle, which is one of the main structural elements of the
bridge.
Both datasets have been processed following two different
strategies: (1) a classical GPR 2D processing routine followed
by an interpolation to build a 3D dataset and significant
time slices; (2) a 2D MT inversion procedure applied on
each 2D profile, followed by 3D interpolation and pseudo-
visualization, and constant depth-slice extraction.
Due to the structural complexity and inhomogeneity of the
bridge deck, the first dataset required the use of a three-layer
velocity model for both processing approaches to achieve the
reconstructions.
A processed radargram allowed us to observe the internal
deck box structure composed of reinforced concrete plates
and stiffening walls. Moreover, the pillar supports and Gerber
saddles that appear absorptive due to water infiltration have
been identified. The first dataset radargrams appear affected
by multiple reflections on the walls and bottom of the deck
box structure. Nevertheless, radargrams provided significant
information in agreement with the bridge plan both in section
and time-slice views. In the latter, important absorptive zones,
related to water infiltration, have been identified in the asphalt
and in the correspondence of the Gerber saddles. A simple MT
approach has been adapted to work in a layered medium by
using the same three-layer velocity model. The reconstructed
tomographic images showed a generalized improvement of
spatial resolution, allowing us to easily detect stiffening walls
and pillar supports. Horizontal features such as upper and
lower plates are less defined than absorptive features such as
Gerber saddles. In the depth-slice view, these characters are
preserved: stiffening walls appear well defined with respect
to the classical processed data, whereas absorptive zones and
Gerber saddle images are less defined.
Concerning the second dataset, GPR revealed a useful
tool to identify internal structures of the Gerber saddle in
good agreement with the bridge plan. The data quality of
the second dataset was lower compared to the first one due to
contact problems between the survey wheel, used to identify
the movement abscissa of the survey, and the surface, and the
low accessibility of the measurement surface, and accordingly
mispositioning errors and local undersampling arise. Depth
slices have been produced by using two processing approaches
where the same mispositioning correction has been applied.
Due to the complex shape of the Gerber saddle, the presence
of rebars along several directions and the presence of different
materials, both processing procedures, based on the movement
of the linearly polarized antenna along only a linear direction,
exhibit some limitations in the complete definition of this
structure in particular. In general, MT allowed us to
obtain more detailed images, but the classical processing
approach is able to delineate the structure with an acceptable
quality.
In conclusion, the two processing approaches used in
this work can be considered complementary, although the
two procedures did not always highlight the same features
in the same way: generally, the absorptive zones are better
detected by the classical processing procedure, while MT
Ground penetrating radar and microwave tomography 3D applications
defined the constructive details better as internal stiffening
walls and rebars. Moreover, MT was more robust in the case
of poor data quality, such as the second dataset. However, an
improvement of the MT routine could be of interest in the case
of layered structures such as bridges.
Acknowledgments
The research leading to these results has received funding from
the European Community’s Seventh Framework Programme
(FP7/2007-2013) under Grant Agreement no 225663 Joint
Call FP7-ICT-SEC-2007-1, and from Soprintendenza per
i Beni Architettonici e Paesaggistici della Basilicata and
Direzione Regionale per i beni Culturali e Paesaggistici della
Basilicata within the project ‘Ponte Musmeci Potenza: Analisi
del degrado dei materiali e progettazione interventi—Parte 1
e 2’. Moreover, the authors would like to thank the owners
of the Musmeci bridge, ASI Consortium, for their interest in
these research activities, and the Municipality of Potenza for
granting authorization to access and work on the structure.
References
Bavusi M, Soldovieri F, Piscitelli S, Loperte A, Vallianatos F and
Soupios P 2010 Ground-penetrating radar and microwave
tomography to evaluate the crack and joint geometry in
historical buildings: some examples from Chania, Crete,
Greece Near Surf. Geophys. 8 377–87
Belli K and Wadia-Fascetti S 2008 Model based evaluation of
bridge decks using ground penetrating radar Comput. Aided
Civ. Infrastruct. Eng. 23 3–16
Che W C, Chen H L and Hung S L 2009 Measurement radius of
reinforcing steel bar in concrete using digital image GPR
Constr. Build. Mater. 23 1057–63
Chew W C 1995 Waves and Fields in Inhomogeneous Media
(Piscataway, NJ: IEEE)
Dérobert X, Aubagnacb C and Abrahama O 2002 Comparison of
NDT techniques on a post-tensioned beam before its autopsy
NDT&E Int. 35 541–8
Forest R and Utsi V 2004 Non destructive crack depth
measurements with ground penetrating radar 10th Int. Conf. on
Ground Penetrating Radar (Delft, The Netherlands,
21–24 June 2004) pp 799–802
Hugenschmidt J and Loser R 2008 Detection of chlorides and
moisture in concrete structures with ground penetrating radar
Mater. Struct. 41 785–92
Hugenschmidt J and Mastrangelo R 2006 GPR inspection of
concrete bridges Cem. Concr. Compos. 28 384–92
Hugenschmidt J, Kalogeropoulos A, Soldovieri F and Prisco G 2010
Processing strategies for high-resolution GPR concrete
inspections NDT&E Int. 43 334–42
Huston D, Hu J Q, Maser K, Weedon W and Adam C 2000 GIMA
ground penetrating radar system for monitoring concrete
bridge decks J. Appl. Geophys. 43 139–46
Lai W L, Kind T and Wiggenhauser H 2010 Using ground
penetrating radar and time-frequency analysis to characterize
construction materials NDT&E Int. 44 111–20
Lai W L, Kou S C, Tsang W F and Poon C S 2009 Characterization
of concrete properties from dielectric properties using ground
penetrating radar Cem. Concr. Res. 39 687–95
Leucci G, Persico R and Soldovieri F 2007 Detection of fractures
from GPR data: the case history of the Cathedral of Otranto
J. Geophys. Eng. 4 452–61
S45
M Bavusi et al
Liseno A, Soldovieri F and Pierri R 2004 Improving a
shape reconstruction algorithm with thresholds and
multi-view data AEÜ Int. J. Electron. Commun.
58 118–24
McCann D M and Forde M C 2001 Review of NDT methods in the
assessment of concrete and masonry structures NDT&E
Int. 34 71–84
Persico R, Soldovieri F and Utsi E 2010 Microwave tomography for
processing of GPR data at Ballachulish J. Geophys.
Eng. 7 164–73
Scott M, Rezaizadeha A, Delahazab A, Santosc C G, Moored M,
Graybeale B and Washerf G 2003 A comparison of
nondestructive evaluation methods for bridge deck assessment
NDT&E Int. 36 245–55
Shaari A, Millard S G and Bungeyb J H 2004 Modelling the
propagation of a radar signal through concrete as a low-pass
filter NDT&E Int. 37 237–42
Shaw M R, Millard S G, Molyneauxc T C K, Taylord M J and
Bungeyb J H 2005 Location of steel reinforcement in concrete
using ground penetrating radar and neural networks NDT&E
Int. 38 203–12
S46
View publication stats
Slaney M, Kak A C and Larsen L E 1984 Limitations of imaging
with first-order diffraction tomography IEEE Trans. Microw.
Theory Tech. 32 860–74
Soldovieri F, Hugenschmidt J, Persico R and Leone G 2007 A linear
inverse scattering algorithm for realistic GPR applications
Near Surf. Geophys. 5 29–42
Soldovieri F and Orlando L 2009 Novel tomographic based
approach and processing strategies for GPR measurements
using multifrequency antennas J. Cultural
Heritage 10S e83–92
Soldovieri F and Solimene R 2010 Ground Penetrating Radar
Subsurface Imaging of Buried Objects (Vienna, Austria: Radar
Technology IN-TECH) pp 105–26
Solimene R, Soldovieri F, Prisco G and Pierri R 2007
Three-dimensional microwave tomography by a 2-D
slice-based reconstruction algorithm IEEE Geosci. Remote
Sens. Lett. 4 556–60
Utsi V, Birtwisle A and Cook J 2008 Detection of subsurface
reflective cracking using GPR Structural Faults &
Repair-2008: 12th Int. Conf. (Edinburgh, UK, 10–12 June
2008) pp 74–5