Cassidy2011 - GPR V Ultrasonic Echo

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Journal of Applied Geophysics 74 (2011) 263–276

Contents lists available at ScienceDirect

Journal of Applied Geophysics


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j a p p g e o

Void detection beneath reinforced concrete sections: The practical application of


ground-penetrating radar and ultrasonic techniques
Nigel J. Cassidy a,⁎, Rod Eddies b, Sam Dods b
a
Applied and Environmental Geophysics Group, School of Physical and Geographical Sciences, Keele University, Staffordshire, ST5 5BG, United Kingdom
b
Fugro Aperio Limited, Focal Point, Newmarket Road, Bottisham, Cambridge, CB25 9BD, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: Ground-penetrating radar (GPR) and ultrasonic ‘pulse echo’ techniques are well-established methods for the
Received 17 February 2011 imaging, investigation and analysis of steel reinforced concrete structures and are important civil engineering
Accepted 6 June 2011 survey tools. GPR is, arguably, the more widely-used technique as it is suitable for a greater range of problem
Available online 15 June 2011
scenarios (i.e., from rebar mapping to moisture content determination). Ultrasonic techniques are
traditionally associated with the engineering-based, non-destructive testing of concrete structures and
Keywords:
GPR
their integrity analyses (e.g., flaw detection, shear/longitudinal velocity determination, etc). However, when
Ultrasonic used in an appropriate manner, both techniques can be considered complementary and provide a unique way
Void detection of imaging the sub-surface that is suited to a range of geotechnical problems. In this paper, we present a
Reinforced concrete comparative study between mid-to-high frequency GPR (450 MHz and 900 MHz) and array-based, shear
wave, pulse-echo ultrasonic surveys using proprietary instruments and conventional GPR data processing and
visualisation techniques. Our focus is the practical detection of sub-metre scale voids located under steel
reinforced concrete sections in realistic survey conditions (e.g., a capped, relict mine shaft or vent).
Representative two-dimensional (2D) sections are presented for both methods illustrating the similarities/
differences in signal response and the temporal–spatial target resolutions achieved with each technique. The
use of three-dimensional data volumes and time slices (or ‘C-scans’) for advanced interpretation is also
demonstrated, which although common in GPR applications is under-utilised as a technique in general
ultrasonic surveys. The results show that ultrasonic methods can perform as well as GPR for this specific
investigation scenario and that they have the potential of overcoming some of the inherent limitations of GPR
investigations (i.e., the need for careful antenna frequency selection and survey design in order to image
through the rebar meshes). More importantly, we show that standard GPR data collection, processing and
visualisation techniques can be used with both types of data without users needing to change existing
operational protocols or survey criteria.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction resolution and mode of operation, their broader practical use beyond
traditional NDT investigations is less common (i.e., geological, environ-
The use of elastic and electromagnetic wave propagation methods mental and contaminated land applications). In terms of concrete
for the non-destructive testing (NDT) and practical investigation of investigation, GPR has an excellent reputation for being able to image
building structures (concrete sections, brickwork, masonry, etc) is steel reinforcing bars (e.g., Al-Qadi and Lahouar, 2005; Chang et al.,
common across civil engineering. Techniques such as ultrasonic “echo” 2009; De Souza, 2005; Hugenschmidt, 2002; Hugenschmidt and
(or pulse-echo), impact echo (IE) and ground-penetrating radar (GPR) Kalogeropoulos, 2009; Utsi and Utsi, 2004) voids and tendon ducts
are all well-established characterisation and analysis tools with ground- (Giannopoulos et al., 2002; Maierhofer, 2003) and, more pertinently,
penetrating radar being, arguably, the most broad-based and widely voids beneath concrete sections (Chen and Scullion, 2008). Unfortu-
applicable of these techniques. Impact echo and pulse-echo ultrasonic nately, GPR is less successful than ultrasonic methods for the detection
techniques (using either shear waves or longitudinal/compressional of fine voids/fractures within concrete (i.e., millimetre-scale delamina-
waves) are primarily used for the NDT-based testing of engineered tion and cracks) and often requires careful antenna frequency selection
concrete structures and although pulse-echo ultrasonic methods can be and planning (e.g., survey line orientation and antenna polarisation) in
considered as comparable to GPR in terms of application, scope, order to successfully penetrate beneath the internal reinforcing bar
meshes. To do so, a degree of a priori survey knowledge is usually
required (e.g., rebar diameter, mesh size, placement geometry/
⁎ Corresponding author. orientation, depth, etc) and although this information may have been
E-mail address: n.j.cassidy@esci.keele.ac.uk (N.J. Cassidy). recorded at the time of construction, survey practitioners commonly

0926-9851/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jappgeo.2011.06.003
264 N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276

have to collect data ‘blind’ where the target and subsurface parameters both the GPR (900 MHz antennae) and ultrasonic systems. The test
are unknown or, at best, poorly defined. slab forms part of the NSGG (Near-Surface Geophysics Group of the
A number of authors have comparatively evaluated all three Geological Society) shallow geophysics test facility at the University of
methods in an engineering investigation context where internal Leicester (Area 9, test site 1 — http://www.nsgg.org.uk/test-sites) and
faults/features such as delaminations, thickness variations, voids, consists of a 2 m-by-6 m, 0.3 m deep concrete slab (well-mixed
corrosion and tension duct grouting defects have all been included in concrete with an ~ 6 mm gravel aggregate) overlying uniform, dense,
the tested concrete sections (e.g., Aggelis et al., 2010; Beutel et al., 2008; low permeability boulder clay soils of conductivity in the order of
Dilek, 2007; Maierhofer et al., 2003; Martina et al., 2001; Muldoon et al., 10–50 mS/m. Across one half of the slab a single layer of 10 mm
2007; Shaw and Xu, 1997). In the recent study by Beutel et al. (2008) for diameter, 0.2 m centred, steel, orthogonal rebar mesh has been
example, shear wave based ultrasonic pulse-echo devices have shown incorporated into the slab at a depth of approximately 0.15 m. The
to be effective at detecting concrete slab thickness variations in the mesh elements are aligned with the edges/sides of the slab although
range of 0.1–0.3 m with a resolution accuracy of less than 10 mm. Their their exact x–y positioning is unknown. A 0.4 m long, 0.6 m wide and
equivalent GPR tests (1.5 GHz antenna) showed similar levels of 0.5 m deep, air-filled target void sits directly beneath the slab in a
thickness evaluation accuracy but, unfortunately, the article did not central position (see survey area in Fig. 1) and is constructed from a
discuss how well either method performed when imaging the grout sealed, thin-walled plastic tank buried into the underlying boulder
defects. Muldoon et al. (2007) investigated the use of GPR, ultrasonics clay soils. The concrete slab was laid directly over the underlying soils/
and IE for tension duct void location and concluded that ultrasonic void (i.e., no membrane or shuttering) and, as such, the basal interface
tomography, direct GPR imaging and the spectral analysis of impact can be considered as ‘in contact’ with the substrate (i.e., is not
echo amplitudes all provided successful methods for detecting grouting- impervious) but discontinuous from a wave propagation perspective.
related voids. What these studies ultimately show is that IE and The slab has been in situ for over 15 years and the upper surface has
ultrasonic methods work very well when used specifically for detecting become weathered with sub-metre scale patches of light scaling (or
faults/voids within uniform concrete sections and where the surface flaking) a few millimetres deep.
conditions are favourable for the coupling of the elastic/ultrasonic The test slab and void target are intended to represent the size,
waves (i.e., new and/or unweathered surfaces). For more ‘non- construction and investigation environment one would typically
engineering’ geotechnical and hazard-related applications, such as expect for a concrete cap on a small relict mine shaft/vent or the early
locating relict mine shafts/vents and karst dissolution features under development of surface-breaching voids under foundation concrete
older concrete structures, there is limited published research on the floors. Such features are common in areas of karst or evaporite
practical use of these ultrasonic echo-based techniques and particularly dissolution and shallow mining (Pringle et al., 2008; Styles et al.,
so when compared to the popularity of ground-penetrating radar. In this 2005) and the problem is of significant concern to many areas of
paper, we aim to address this deficit by comparing the performance of industrialised Europe, often as a consequence of goafing or mine roof
both methods for the location and imaging of sub-metre scale, open failure.
voids situated under weathered, steel reinforced concrete sections
(i.e., as would expected for a mine shaft ‘cap’). Although our test case is 2.2. Data collection
idealised (because the target criteria are well-constrained), we have
attempted to make the study as realistic as possible by using The GPR and ultrasonic surveys were collected across an
conventional surveying geometries and data collection/processing approximately North–South/East–West orientated orthogonal grid
principles (e.g., bi-static, reflection-mode data collection on a 0.2 m aligned to the edges of the concrete test slab (Fig. 1). Because the
spaced orthogonal grid) and proprietary GPR/ultrasonic survey equip- centre position and approximate width of the target void was known
ment. The survey data are collected ‘blind’ with the operator having before data collection, the boundary of the 1 m × 1.4 m survey area
limited a priori knowledge of the sub-surface conditions (e.g., size, grid was laid out equidistance from the mid-point. This was the only a
shape, layout of the rebar mesh, etc) and only a very basic idea of the priori information used in the survey (to avoid collecting unnecessary
target geometry (i.e., the concrete slab's approximate size and depth). data) and, therefore, the overall size, orientation, alignment, depth
As such, this paper represents an important practical study into the and shape of both the reinforcing bar mesh and target void were
performance of both techniques in an application are where GPR initially unknown to the operator on site. This ensured that the survey
dominates over ultrasonic methods. design and data collection approach mimicked, as close as possible,
The paper is divided into the following sections: firstly, the typical ‘blind’ survey conditions encountered in this type of
features of the test site will be described, including the configuration investigation. Line spacings were set to 0.2 m (approximately 1.5λ
of the GPR and ultrasonic surveys, plus a detailed explanation of the for the 900 MHz GPR survey) with individual traces being collected at
steps used to process the ultrasonic data and how this compares to 0.02 m increments along each line (for all methods). These values
GPR data processing approaches. After this, representative two- were selected on the basis of providing a practical balance between
dimensional (2D) sections will be presented for both methods adequate data coverage (for the approximate target width) and a
illustrating the similarities/differences in signal response and the reasonable speed of manual data collection. Although the 0.2 m line
temporal–spatial target resolutions derived from the section in- spacing does not provide very high-density data coverage, as would
terpretations. Next, the use of time slices (or ‘C-scans’) for visualising be expected for high-resolution 3D data sets, it does provide enough
the ultrasonic data will be illustrated, which although familiar for GPR coverage to create meaningful time-slices for the size of the target's
data, is rare for more general, non-engineering ultrasonic pulse-echo footprint. Data was collected in manual “step mode” with the survey
applications (Iyer et al., 2005). Finally, the practical performance of grid marked out accurately on the top of the concrete slab. This
both methods will be discussed and how the developed approaches ensured a spatial accuracy of +/−0.5 cm for the antenna/array central
relate to practical sub-surface investigation. locations. Individual survey lines (L1–L12) were collected in a
conventional North-to-South (L8–L12) and West-to-East (L1–7)
2. Materials and methods direction with the GPR data being used in its default bi-static, co-
planar, common-offset, broadside, reflection mode (transmitter
2.1. Test site surveys leading) with the E-field polarisation being parallel to the antenna
axis (Fig. 1). This resulted in an antenna polarisation that is either
Fig. 1 shows the general layout of the reinforced concrete test parallel or perpendicular to the orthogonal rebar mesh, depending on
‘slab’, the survey area geometry and images of data collection with survey line orientation. Ultrasonic data was collected in the same
N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276 265

Fig. 1. Target void layout, survey area and data collection across the reinforced concrete ‘test’ slab.

manner but with the exception of having separate Longitudinal and ❑ Dewow correction
Transverse orientated surveys for each line (Fig. 1). ❑ Time-zero correction
❑ Background removal across all traces for the full time window
(mean trace subtraction)
2.2.1. GPR surveys ❑ Dynamic correction to zero-offset (to correct for the additional
A Sensors and Software, PulseEkko 1000™ GPR unit with 450 MHz two-way travel time associated with the non-zero, TX–RX antenna
and 900 MHz shielded antenna was used for all GPR surveys with each separation — 0.25 m for the 450 MHz antenna and 0.17 m for the
trace collected manually at defined intervals and processed using 900 MHz). A constant velocity of 0.11 m/ns was used for all
conventional methodologies (Cassidy, 2008). The average of thirty- corrections based on a combination of ground-truthing (travel
two trace stacks were recorded at each collection point with each time to the base of the slab), common-midpoint (CMP) velocity
manually triggered trace having a temporal increment of 0.1 ns and a analysis and hyperbolic matching of the rebar diffractions.
total time window of 40 ns (approximately 2 m total depth of ❑ Bandpass filtering (frequency-domain Butterworth filter with a
investigation). The choice of both 450 MHz and 900 MHz antenna pass band of 200–900 MHz for the 450 MHz data and 400–
was based on the anticipated depth of investigation (b2 m), likely 1500 MHz for the 900 MHz data).
target size (a sub-metre isolated volume) and data collection time ❑ Time cut at 16 ns (equates to a depth of 0.8 m with a uniform
constraints. 1.2 GHz antenna surveys were initially considered, but velocity of 0.11 m/ns).
the additional data collection time required for the more densely
spaced data (0.1 m line spacing and 0.01 trace increments) was The processed, but unmigrated, sections were used for initial data
deemed prohibitive. interpretation and then migrated using a standard Stolt, 2D migration
The proprietary ReflexW v5.5 software package (Sandmeier algorithm with a uniform velocity of 0.11 m/ns. The final migrated
Software, 2010) was used for post-collection data processing and sections were then combined into a 3D data volume for time-slicing
consisted of the following standard steps for the 2D GPR sections: (see later section).
266 N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276

2.2.2. Ultrasonic surveys the data were stored in a raw state in the form of individual sections (i.e.,
An Acoustic Control Systems Ltd, A1220 MONOLITH low-frequency no filtering, gains, editing or other pre-processing steps) in order to
ultrasonic ‘flaw detector’ system was used for all the ultrasonic surveys mimic the GPR surveys. The arrangement of the array-based transducer
with a stated operating frequency of 55 kHz. The system consists of a head means that the data collection geometries and survey approaches
dedicated data logger/control unit connected via fibre-optic cables to a are, in principle, comparable to the GPR surveys allowing appropriate
transmitter–receiver (Tx–Rx) transducer head. The head contains 24 comparisons to be made. The in-phase transmitter array produces a
individual ‘dry contact’ piezoelectric shear wave generating transducers propagating wavefront that, to a reasonable approximation, can be
arranged into a dual 2 × 12 element array (Fig. 2). Each individual considered as being spherical and spatially uniform. As such, conven-
piezoelectric element is spring loaded to allow an equal coupling force tional seismic/elastic wave propagation theory can be used to assess the
to be applied to the survey surface using light hand pressure. The velocity, attenuation, wavelength, pulse width and approximate fresnel
transmitted shear wave pulse is generated by ‘firing’ each of the 12 zone related footprint of the propagating ultrasonic wave (Aki and
transmitter elements simultaneously, creating an ‘in-phase’ shear wave Richards, 1980). However, the orientation of the Tx–Rx array, and the
motion that is aligned along the long axis of the array (Acoustic Control polarisation of the shear wave particle motion that it produces, means
Systems, 2004). that the ultrasonic surveys are more analogous to the parallel-endfire
During the surveys, the array can be orientated to give either a and perpendicular-endfire modes of GPR antenna configuration than
Transverse (T) or Longitudinal (L) polarisation geometry (Fig. 2) with the default perpendicular-broadside mode of most proprietary GPR
data being collected in a manual step mode with the average of 16 systems (Annan, 2005). Although this affects how the individual
individual pulse stacks recorded at each survey position. The data electromagnetic and elastic (stress) fields are orientated in space and
control unit's software has been specifically adapted to collect full time, from a practical point of view, the wave propagation/reflection
waveform data (rather than just absolute values, which is common for principles are complementary (Carcione and Cavallini, 1995) and,
many engineering NDT applications) with a 10-bit analogue-to-digital therefore, the data processing and visualisation approaches needed to
conversion capacity. This equates to a ~120 dB dynamic range, which is interpret the data can be considered as being equivalent.
comparable to most commercial GPR systems. In practice, this means a The data collected along each survey line were stored as individual
signal penetration of up to ~2 m for most concrete investigations with a sections (L1–12) and then converted into SEGY format for processing
target resolution of 0.1–0.4 m depending on the velocity of the concrete with the same ReflexW v5.5 software package (Sandmeier Software,
and the degree of material heterogeneity (Acoustic Control Systems, 2010) that was used for GPR surveys. One interesting initial
2004). The system is predominantly used for commercial NDT-style observation is that the frequency spectrums of the recorded re-
concrete investigations (Lange et al., 1998; Shevaldykin et al., 2002) but flections from the base of the concrete slab are lower than the stated
less so for geotechnical applications. As such, the data control, handling central frequency of the source pulse. This is illustrated in Fig. 3 where
and processing commands of the proprietary software are biased the effective operational central frequency is approximately 40 kHz
towards the rapid, on-site evaluation of concrete integrity and the with a bandwidth of about 25–55 kHz (when compared to the stated
collection of multiple single traces and/or 2D sections. In this instance, central frequency of 55 kHz). This is similar to GPR surveys where the
however, the data collection and storage parameters were set-up so that true spectrum of the recorded signals is often ‘down-shifted’ from the

Fig. 2. Design, construction and layout of the Acoustic Control Systems (ACS) A1220 Monolith Ultrasonic flaw detection instrument including the data control unit and the
24-element Tx–Rx transducer head array.
N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276 267

reflection to be generated (Annan, 2005). Fresnel zones can be easily


calculated for both EM and elastic waves (Aki and Richards, 1980;
Balanis, 1989) but the theory assumes that wave propagation is
uniform and purely spherical. This may be a reasonable assumption
for the ultrasonic surveys but the shielded, dipole antennae of the GPR
system produce propagating EM wavefronts that are spatially non-
uniform. As such, their fresnel zone areas/shapes are more complex to
calculate (Balanis, 1997) and tend to be elongated into ellipses whose
long axes are orientated perpendicular to the axis of the antenna. This
complexity has led to the ‘footprint’ concept in GPR, which essentially
describes the area (at a given depth) illuminated by the GPR wave that
returns the majority of the reflected energy. For simple dipole
antennae used in common-offset, co-planar, bi-static surveys (as in
this study), the footprint area is approximately the same size as the 1st
fresnel zone for a planar, horizontal interface (Annan, 2005), although
it is not exactly the same shape. Therefore, the 1st fresnel zone concept
Fig. 3. Frequency spectrum of the recorded ultrasonic pulse reflected from the 0.3 m deep,
can be used as approximate comparative measure of the horizontal
concrete/air interface situated at the base of the uniform, non-reinforced concrete slab. target resolution of both the GPR and ultrasonic surveys. This is
illustrated in Fig. 5 where the calculated 1st fresnel zones at 0.3 m (the
depth of the slab's basal interface) are overlain on the survey area and
target void's cross-sectional footprint.
stated antenna central frequency due to variable surface/antenna On the reasonable assumption of a uniform average velocity
coupling and material attenuation effects. Similar factors are likely to through the slab of 0.11 m/ns for the GPR wave and 2000 m/s for the
be responsible for the observed reduction in the ultrasonic spectrum ultrasonic, the 1st fresnel zone diameters are; 0.4 m (450 MHz GPR),
as the weathered surface of the slab and permeable nature of the aged 0.28 m (900 MHz GPR) and 0.18 m (ultrasonic). Consequently, there
concrete are far removed from the pristine conditions of freshly laid is only going to be a relatively small region in the centre of the survey
materials. Regardless of cause, what this change in spectral attribu- where the footprint of the target void is a significant proportion of the
tives does signify is that any estimates of wavelength, resolution, 450 MHz GPR wave's first fresnel zone. Outside of this region, the
fresnel zone size, etc., at depth must take into account the drop in reflected EM waves will become more incoherent/scattered and,
frequency observed in the signal. therefore, difficult to identify as unique reflections in the GPR sections.
To be consistent with the GPR surveys, and therefore allow This problem is less severe for the 900 MHz GPR and ultrasonic
conventional data processing approaches to be used on both data sets, surveys due to their smaller fresnel zones, but there will still be an
the ultrasonic data have been processed using the ‘standard’ flow element of signal scattering and incoherence from the edges/corners
procedures in the ReflexW v5.5 software. The raw 2D data were of the target void.
corrected for DC shift and time-zero variations then filtered (band-
pass, background removal and simple 2D x–t average filtering) to 3. Survey results
produce the processed 2D data for each longitudinal and transverse
section (Fig. 4). Initial interpretations were made with these 2D 3.1. 2D sections: GPR and ultrasonic
processed sections and then the data were processed further to
provide the 3D data volume for time-slicing (see later section — Fig. 6 shows the processed 900 MHz (migrated and unmigrated)
Section 3.2). A number of different advanced processing techniques and 450 MHz (unmigrated) GPR data collected along line L10, which
were also attempted on a trail-and-error basis, including deconvolu- runs centrally across the long axis of the target void. The strong,
tion, migration, attribute analysis, etc., with the aim of emphasising coherent, uniform diffraction hyperbolae are easily identified and
the target features in the sections (i.e., the slab's basal reflection) prior indicative of the GPR response from regularly spaced, steel reinforcing
to creating the 3D data volume. However, none of these methods meshes. The slab's basal reflection is also clear with a loss of signal
improved the coherence or visual integrity of the features to any continuity and strength as the survey passes over the edges of the
significant degree except for the vectoral addition (i.e., a ‘vector sum’) target void. Strong, broad, coherent, hyperbolic-shaped diffractions/
of the two individually polarised longitudinal and transverse sections reflections associated with the top and sides of the target void are also
along each line. Given the polarised nature of the propagating shear easily identified, particularly in the 900 MHz section, and the data
waves, this is a logical processing step as it produces a more accurate migrates well with the majority of the hyperbolic features being
representation of the true shear wave response of the target in the collapsed to their source points.
horizontal plane; i.e., the orthogonally orientated ShX and ShY shear For experienced users, the 900 MHz section is easy to interpret but
wave components are combined as a vector sum to give the total what is more interesting is the comparison between it and the
magnitude of the Sh motion in the plane of the slab's basal interface — corresponding 450 MHz section. As would be expected from the
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 x 2  2
Sh = Sh + Syh . larger footprint of the 450 MHz antenna, the observable reflection
from the top of the void is spatially smaller, less coherent and more
Before the results from each survey are described, it is worth poorly defined. The diffractions from the rebar mesh are as strong and
considering what the approximate wave propagation and target identifiable as in the 900 MHz section but have ‘tails’ that extend
resolution parameters are for each method (Table 1). For regularly beyond the two-way travel time of the slab's basal reflector, masking
shaped, yet isolated, three-dimensional objects (such as the target its presence in the section. The size of the rebar mesh grid (0.2 m) also
void) the cross-sectional area ‘illuminated’ by the propagating means that the area of the mesh illuminated by the 450 MHz wave's
incident wave becomes an important parameter as it determines 1st fresnel zone is proportionally greater than that of the 900 MHz
how much coherent energy is reflected back to the receiving antenna/ survey. As a consequence, relatively more energy is reflected back
array (Daniels, 2004). To a first-order approximation, a target's cross from the mesh to the receiving antenna and less is transmitted
sectional area must be at least half the area of the propagating wave's through to base of the slab. It is also interesting to note that there is a
first fresnel zone, at a given depth, in order for a coherent observable strong asymmetry to the form and amplitude of the recorded rebar
268 N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276

Fig. 4. Data processing methodology for the ultrasonic survey data.

Table 1
diffractions in the 450 MHz data, which can lead to the section being Approximate wave propagation and target resolution parameters for the ultrasonic,
interpreted as having falsely dipping features. This phenomenon has 900 MHz GPR and 450 MHz GPR surveys at depth corresponding to the base of the
also been observed in GPR data where there are there are shallow, reinforced concrete ‘test’ slab (0.3 m).

strongly contrasting, regularly spaced, point or axis-linear target GPR GPR Ultrasonic
sources close to each other in the sub-surface (e.g., Cassidy and 450 MHz 900 MHz
Millington, 2009; Cassidy et al., 2009). If the separation distance Effective frequency (at slab basal interface) 450 MHz 900 MHz 40 kHz
between the targets is of the same order as the wavelength of the Velocitya 0.11 m/ns 0.11 m/ns 2000 m/s
propagating GPR signal, and the data is collected along the line in the Pulse widthb 3 ns 1.5 ns 0.4 ms
Wavelength (at slab basal interface) 0.24 m 0.12 m 0.05 m
same direction, then the asymmetry of the diffractions (and therefore
Vertical target resolutionc 0.12 m 0.06 m 0.025 m
the relative dip direction of the features) appears to depend on 1st fresnel zone diameter 0.4 m 0.28 m 0.18 m
whether the transmitter is leading the Tx–Rx pair or the receiver. This a
Estimated from the two-way travel time of the basal reflection at a ground-truthed
is illustrated in Fig. 7 where the 450 MHz GPR surveys along line L10 depth of 0.3 m, plus CMP and hyperbolic matching velocity analysis (GPR data only).
were collected in a forward orientation (Tx leading) and reverse b
Average measured pulse width of slab basal reflection.
orientation (Rx leading). The asymmetrical left- and right-handed dip c
Estimated from the ½ wavelength distance at the depth of slab basal interface.
N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276 269

quence of the EM field asymmetry of the shielded, bi-static, common-


offset transmitter–receiver pair, data collection irregularities or other
source/antenna effects is debatable, but it does have important
practical consequences for grid-based data collection geometries and
the subsequent generation of 3D volumes for time-slicing. If the
surveys are collected in a zig-zag mode (i.e., up one line and down the
next) without swapping the antenna Tx–Rx pair around, then at any
specific distance signal amplitudes and phases could become
mismatched between adjacent sections. This can result in poorly
rendered data when the 2D sections are migrated and interpolated
into a 3D volume. Even with data collected systematically in one
direction only, there is always the potential of some degree of
handedness/asymmetry to the data, loss of accuracy and interpreta-
tional uncertainty in the resultant 3D data volumes. Ultimately, this
illustrates that care must be taken when planning 3D GPR survey
geometries, selecting antenna frequencies and quality controlling the
data post collection.
Fig. 8 shows the processed 900 MHz (unmigrated) and 450 MHz
(unmigrated) GPR data collected across the un-voided section of the
concrete test slab (Line no. L8). As with the previous example, the
diffraction hyperbolae from the rebars are clear in both data sets, as is
the reflection from the base of the concrete slab in the 900 MHz
section. In this instance, however, the reflection is more uniform,
coherent and continuous along its length with only minor disruption
to its form in the centre (most likely from side-swipe off the top/edges
of the target void). The 450 MHz section shows similar asymmetrical
features to its L10 counterpart but with no evidence of the reflection
from the top of the void. In fact, the swamping effect of the strong
diffractions almost obliterates the slab's basal reflection to the point
where it is very difficult to identify in the section.
Fig. 5. Approximate 1st fresnel zone footprint areas at the base of the reinforced concrete
For brevity, the rest of the 2D GPR sections will not be shown
‘test’ slab (at a depth of 0.3 m) for the ultrasonic, 900 MHz GPR and 450 MHz GPR waves.
individually but they do show similar features to the illustrated
of the rebar diffractions is clear and dominates the whole section. sections with strong diffractions from the rebar mesh and variability
There is even a subtle change in the symmetry of the void's in the coherence of the slab's basal reflection. Likewise, only the
diffraction/reflection. Whether this asymmetrical effect is a conse- corresponding two-dimensional L10 (central section) and L8 (un-

Fig. 6. Processed 900 MHz (migrated and unmigrated) and 450 MHz (unmigrated) GPR data collected centrally across the long axis of the target void (Line L10). Depth estimates are
based on a uniform electromagnetic wave velocity of 0.11 m/ns.
270 N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276

Fig. 7. Processed 450 MHz (unmigrated) GPR data collected across centrally across the long axis of the target void (Line L10) in both forward (Tx leading–Rx trailing) and reverse
(Rx leading–Tx trailing) antenna orientations. Depth estimates are based on a uniform electromagnetic wave velocity of 0.11 m/ns.

Fig. 8. Processed 900 MHz (unmigrated) and 450 MHz (unmigrated) GPR data collected across the un-voided section of the concrete test slab (Line L8). Depth estimates are based on
a uniform electromagnetic wave velocity of 0.11 m/ns.
N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276 271

voided section) from the ultrasonic surveys are shown to highlight the strong, short, discontinuous reflection in the same position that
similarities/differences between the voided and un-voided data sets. arrives slightly later in two-way travel time. It is more prominent in
Fig. 9 illustrates the processed ultrasonic data (longitudinal, the longitudinal orientated data set and could be a polarised reflection
transverse and vector sum) collected centrally across the long axis from the upper long edge and/or side of the void. From a practical
of the target void (Line L10) plus a comparison between the vector perspective, the loss of coherence in the basal reflection is indicative
sum section and the 900 MHz GPR data. of a change in the physical nature of the sub-surface materials (i.e., a
For the same investigation depth window (0.8 m), the shorter discontinuity in the elastic properties) which, being associated with
wavelength and tighter pulse width of the ultrasonic wave results in a target void, can then be used as a ‘marker’ for its presence in the
higher-resolution image of the sub-surface when compared to the subsequent 3D data interpretations.
GPR data. Both techniques show a high degree of commonality in the The processed ultrasonic data collected across the un-voided
form, depth, continuity and coherence of the basal reflector but there section of the concrete test slab (Line L8) is shown in Fig. 10 along
is a distinct lack of any defined response from the steel reinforcing with a comparison to its corresponding 900 MHz GPR section. The
mesh grid, despite the potential shear wave velocity contrast between continuous, coherent, strong basal reflector is easily seen with no loss
the rebars and the concrete. The basal interface produces a prominent, of signal strength along its length. There is no also discernable
tight, coherent, flat-lying reflection in the ultrasonic section that is evidence of side-swipe or edge effects from the upper top/sides of the
enhanced in the vector sum data. The top of the target void is also target void. This is mostly likely to be a consequence of the smaller,
evident through the loss of reflected signal amplitude/coherence in more regularly shaped fresnel zone footprint of the ultrasonic wave
the slab's basal reflector in the centre of the section. There is also a not overlapping with the edges of the target void in this particular

Fig. 9. Processed ultrasonic data collected centrally across the long axis of the target void (Line L10) and a comparison between the 900 MHz GPR and ultrasonic data for the same
line. Depth estimates are based on a uniform ultrasonic shear wave velocity of 2000 m/s.
272 N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276

Fig. 10. Processed ultrasonic data collected across the un-voided section of the concrete test slab (Line L8) and a comparison between the 900 MHz GPR and ultrasonic data for the
same line. Depth estimates are based on a uniform wave velocity of 2000 m/s for the ultrasonic surveys and 0.11 m/ns for the GPR surveys.

survey line. There are also shallow, weak diffraction hyperbolae in the away data volume to illustrate the form of the reflections/diffractions
section that may relate to a response from the steel reinforcing mesh. from the steel reinforcing mesh and concrete slab base.
Their depths would be consistent with the corresponding rebar This technique is common in GPR surveys and has the advantage of
signatures in the GPR data but it is impossible to prove their casual providing a virtual 3D view without affecting the data through
nature definitively. What is clear from both techniques is that, again, additional processing. Its interpretational worth is illustrated well in
direct comparisons can be made between the depth, form, coherence the 900 MHz GPR ladder diagram where the sagging form of the rebar
and continuity of the basal reflectors and, therefore, conventional mesh is visible (highlighted in yellow) in the upper 0.2 m of the
GPR-based 3D processing and data visualisation techniques can be concrete slab. The continuous, planar nature of the slab's basal
used appropriately to view and interpret the data. reflection is also highlighted in both the GPR and ultrasonic diagrams
with the loss of reflection coherency and signal amplitude across the
3.2. 3D data and time slices: GPR and ultrasonic top of the void being made more apparent. Although a powerful
visualisation tool, ladder diagrams do not provide the user with a
The most basic 3D data visualisation approach is to combine the practical way of visualising stratigraphical or depth-related data in an
individual sections into a ‘ladder diagram’ where the sections are shown accurate ‘map’ form. Time-slicing (or C-scans) is a popular way to
together in their correct relative spatial positions and ‘cut outs’ or achieve this and is common to many GPR application areas. However,
‘translucency’ effects used to emphasise certain parts of the data care must be taken when choosing the data interpolation and slicing/
volume. This is illustrated in Fig. 11 where the processed ultrasonic and processing parameters as it is easy to corrupt the rendering of a 3D
900 MHz unmigrated GPR sections are used to create a pseudo-3D, cut- volume. Getting the relative time-zero positions, and the depth verses
N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276 273

3.2.1. 3D data processing procedures


Standard 3D GPR data processing procedures (via the ReflexW
v.5.5 3D processing software option) were used to generate the time-
slices with common parameters used on both the 900 MHz migrated
GPR and ultrasonic data. In each instance;
1) Time slice data extraction: A 0.12 m thick slice of data was extracted
from each of the 2D processed sections with an initial two-way
travel time that corresponds to the earliest first arrival time of the
basal reflector for the whole data set. For the ultrasonic data, for
instance, this equated to a travel time slice of 0.3–0.42 ms
extracted equally from each section (L1–L12).
2) 3D data interpolation: Interpolation of the extracted 2D sections
into a regularly spaced (x–y) time-slicing grid having a 0.02 m grid
point increment. A value of 0.02 m was chosen as it avoids spatial
aliasing yet keeps the data volume to manageable proportions,
computationally. The interpolation algorithm used a 0.2 m-wide
interpolation template (or raster) with an interpolation weight
equal to the square of the distance. In combination with the
filtering step described below, this approach ensures that the
interpolation does not a) ‘over smooth’ the data, with
the subsequent loss of signal coherence, or b) under-interpolate
the data leaving it looking too ‘stripey’ along the location of the
original 2D section lines. The 0.2 m raster size gives an interpo-
lation area just slightly bigger than the 1st fresnel zone of both
techniques and ensures that the interpolation area is not only
appropriate from a practical sense, but also that the footprint of
each sample point overlaps adjacent lines. It is also small enough
to make sure that the interpolation does not overemphasise the
influence of values close to the survey line intersection points.
3) 3D data spatial filtering: x–y average filtering across a 0.06 m-by-
Fig. 11. Pseudo-3D, cut-away ‘ladder diagrams’ for the ultrasonic and GPR data volume
illustrating the form of the reflections/diffractions from the steel reinforcing mesh and
0.06 m spatial template to smooth and balance the data, post
concrete test slab base. The approximate depth ranges of the two time slices (upper and interpolation.
lower) are also shown. 4) 3D data temporal filtering: Conversion of the positive–negative
amplitude data to absolute values and then averaged over the
specific time window for both the upper and lower time-slices (i.e.,
0.6 ms for the ultrasonic data). This averaging approach not only
smoothes the data in time, avoiding noisy time slices, but also
two-way travel time correlations correct across individual traces/
provides amplitude values that are indicative of the true EM/
section (or at least as accurate as possible), is particularly important.
ultrasonic energy being reflected back from the region of interest.
Timing variations can result in signal phase-amplitude inconsistencies
that, in turn, can cause mismatches across reflections from otherwise These 3D data processing steps are standard for GPR surveys and
continuous interfaces, resulting in incoherent and un-interpretable although getting the exact processing parameters correct (e.g., raster
time-slice data. This can become problematic for high-resolution data size, etc) is something of a hit-and-miss affair, the approach taken is
sets with fine, narrowly-spaced interfaces of strong material property consistent with the good practice and experience gained from over a
contrasts (or irregular forms) and is probably why time/depth slicing decade of 3D GPR research, particularly in the archaeological discipline
is uncommon in nonNDT style-ultrasonic surveys (i.e., geotechnical or (Conyers and Goodman, 1997). Fig. 12 shows both the upper and lower
geohazard-related surveys). In this instance, it is fortunate that the time-slices of both data sets with the colours indicating the absolute
depth/velocity of the concrete slab is known (and essentially uniform) signal amplitude for each specific slice. The limits of the colour scale
and its upper and lower surfaces are topographically flat, relatively have been set at two standard deviations from the mean amplitude
smooth and continuous. As such, time-slices can be specifically value to show the range of reflected energy with a 93% confidence
targeted at the features of interest (i.e., the base of the slab) with interval. This not only reduces the effect of isolated, high-amplitude data
travel time windows that are set to be neither too long (resulting in ‘spikes’ but also provides a consistent way of the viewing the reflected
‘bluring’ of the features) nor too short (giving noisy, incoherent data). energy distribution of each slice relatively.
For both the 900 MHz GPR and ultrasonic data sets, two separate Note that the corresponding 3D data sets and time-slices for the
but concordant time-slice windows were chosen to highlight a) the 450 MHz GPR surveys have not been illustrated. This is not because
spatial form of the reflection response from the base of the slab and the data quality is too poor to include, or that the processing methods
top of the target void (the upper slice in Figs. 11 and 12) and b) a are inappropriate for this frequency. Instead, it is a consequence of the
lower, underlying slice to capture the form of any deeper signal dominating effect of the rebar diffractions that override the basal
responses from the sides and/or corners of the void (Figs. 11 and 12). reflection (as discussed previously). Their influence is so strong that
To maintain commonality between the two data sets, the duration of the corresponding time-slices lack the necessary coherence to provide
each time-slice widow was designed to be approximately 1.5 times any additional interpretational benefit over the 900 MHz GPR data.
the reflected pulse width (to capture the full reflected signal) and As would be expected, the upper time-slices provide the most
extend sequentially from the earliest first arrival time of the basal coherent reflection response as they were intended to capture the
reflection. Consequently, the upper time slice covers an approximate region encompassing the base of the slab and the top of the void. Both
depth of 0.3–0.36 m and the lower slice 0.36–0.42 m (i.e., a total of data sets show a well-defined, rectangular, low-amplitude zone in the
0.12 m). middle of the slice that correlates well with the position of the target
274 N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276

Fig. 12. Upper and lower time slices for the ultrasonic and 900 MHz (migrated) GPR data with colours indicating the absolute signal amplitude for each specific slice. Upper and lower
colour scale limits have been set at two standard deviations from the mean amplitude value. The location of the target void feature is indicated by the white dashed line.

void. There is a distinct amplitude contrast between the lower- regions (i.e., depths). The relative ‘tightness’ of the reflected wavelet,
amplitude voided region and the higher amplitude un-voided region plus the lack of any additional diffractions/interference from the sides
that surrounds it. This contrast illustrates the relatively strong return of the void, results in a clear change in the nature of the signal and the
of coherent energy from the slab's basal interface and the loss of ability to designate a well-defined depth range to the slab's basal
energy over the void. Despite the potential effect of signal interference interface (0.3–0.36 m in this case). Obviously, we know this is true
from the sides/corners, the boundary between the two regions is well from the known geometry of the slab but given that the survey was
defined, particularly in the GPR data. Therefore, the true spatial (x–y) collected under realistic and practical site conditions (i.e., collected
form of the top of the void can be confidently mapped, even with this blind with no a priori knowledge of the target depth), it shows that a
relatively coarse line spacing. Although the ultrasonic data looks more high degree of stratigraphic accuracy can be obtained with the
‘striped’, which is most likely a consequence of the smaller, more ultrasonic technique. In contrast, the GPR data shows a greater degree
regular footprint area, it does not hinder the overall interpretation to of reflected signal variation with a high amplitude ‘hot spot’ in the
any significant degree. It is also interesting to note that there appears centre of the survey area. The corresponding depth of the time-slice
to be no observable correlation between the GPR and ultrasonic (0.36–0.42 m) is too shallow for this energy to be from the base of the
amplitude data in the un-voided region. This suggests a lack of any void and is, instead, most likely to be a consequence of side-swipe and
casual link between the elastic and dielectric properties across the late-time signal interference from the void's edges/sides. In some
slab's basal interface. ways, this to be expected as the duration of the 900 MHz wavelet
In the lower time-slice, the lack of any significant coherent signal pulse is longer than its ultrasonic counterpart and the antenna
variation (or contrast) in the amplitude of the reflected ultrasonic footprint more irregular, as discussed previously. In practical terms,
signal is expected, but interesting, as it illustrates the distinct change this means that defining a precise depth range for the slab's basal
in the nature of the energy return from the upper and lower time-slice reflection is more problematic with the GPR survey (when compared
N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276 275

to the ultrasonic) and, therefore, the interpretational accuracy is through complex rebar meshes) and can produce results that are
always likely to be compromised. equally accurate from an interpretational point of view.
What this research ultimately shows is that despite the NDT focus
4. Discussion of most proprietary concrete testing ultrasonic instruments, when
used in full-waveform pulse-echo mode, the method can be
What the previous results section illustrates is that the ultrasonic considered as comparable to GPR in terms of application, scope,
technique is a perfectly capable investigation tool for non-NDT, resolution and mode of operation. Consequently, standard GPR data
geotechnical applications, particularly when imaging through rein- collection, processing and visualisation techniques can be applied to
forced concrete. Not only has it been able to spatially define the form complementary ultrasonic data without users needing to change
of the buried target void under the slab but also the depth of its upper operational protocols or survey procedures. This is an important
interface to reasonably high degree of precision. Nevertheless, there practical outcome of this work as it means that both GPR and
are some particular aspects of both surveys that warrant further ultrasonic surveys can be used across a much wider range of
discussion. Despite GPR being a popular and almost ubiquitous tool geotechnical application areas with results that can be considered
for rebar mapping, it does have some limitations when it comes to comparable and interpretationally consistent.
more general reinforced concrete investigation. The results from the
450 MHz surveys have shown that the selection of antenna frequency
is important (in order to ‘image’ through the rebar mesh) and that Acknowledgments
care must be taken with the mode and configuration of the survey
geometries. In this instance, the higher-frequency, 900 MHz survey The authors gratefully acknowledge the financial support provided
was capable of imaging through the ‘flat’ regular orthogonal steel by the Royal Society (Industrial Fellowship) and the Engineering and
mesh of 0.2 m spacing. However, the test-slab example has a Physical Sciences Research Council (EPSRC grant EP/004032/1). We
relatively simple, well-defined structure but, unfortunately, real would like to thank the University of Leicester for the use of the NSGG
engineering concrete is not that straightforward with meshes that test site and the staff at Fugro Aperio, Cambridge, for the loan of the
can overlap, have different sizes/spacings, be laid at different angles Ultrasonic equipment.
and have multiple depths/layers, etc. In these cases, the complexity of
the structural elements almost excludes GPR from anything other
References
than a very simple method for mesh identification. However,
ultrasonic techniques, when used in pulse-echo mode and combined Acoustic Control Systems, 2004. Technical Passport Operation Manual, A1220 Monolith,
Low-Frequency Ultrasonic Flaw Detector. Acoustic Control Systems, Moscow.
with conventional 2D/3D data processing and visualisation tech-
Aggelis, D.G., Kordatos, E.Z., Soulioti, D.V., Matikas, T.E., 2010. Combined use of
niques (such as time-slicing or C-scans), can provide an appropriate, thermography and ultrasound for the characterization of subsurface cracks in
accurate and complimentary investigation technique that overcomes concrete. Construction and Building Materials 24, 1888–1897.
these limitations. It is also worth noting that the ultrasonic data in this Aki, K., Richards, P.G., 1980. Quantitative Seismology, Theory and Methods, Vol I and II.
Freeman and Co., San Francisco, CA, U.S.A.
study is unmigrated and, as such, does not truly represent the depth- Al-Qadi, I.L., Lahouar, S., 2005. Measuring rebar cover depth in rigid pavements
corrected 3D results in a conventional GPR/seismic data processing with ground-penetrating radar. Construction 2005, Transportation Research
sense. Relatively basic migration routines were attempted but no Record, pp. 81–85.
Annan, A.P., 2005. Ground-penetrating radar. In: Butler, D.K. (Ed.), Near-Surface
interpretational benefit was gained from the process. In more Geophysics. Society of Exploration Geophysicists, Tulsa, OK, U.S.A., pp. 357–434.
complex environments, migration would be advised (preferably 3D) Balanis, C.A., 1989. Advanced Engineering Electromagnetics. John Wiley & Sons,
but it has been shown here that migration is not necessarily a New York.
Balanis, C.A., 1997. Antenna Theory: Analysis and Design, second edition. John Wiley &
compulsory processing step. As is common in many GPR surveys, Sons, New York.
migration does not always provide interpretational benefits (Cassidy, Beutel, R., Reinhardt, H., Grosse, C.U., Glaubitt, A., Krause, M., Maierhofer, C., Algernon, D.,
2008) and, as such, the additional time and data processing effort Wiggenhauser, H., Schickert, M., 2008. Comparative performance tests and validation
of NDT methods for concrete testing. Journal of Nondestructive Evaluation 27, 59–65.
expended is unnecessary in this case.
Carcione, J.M., Cavallini, F., 1995. On the acoustic-electromagnetic analogy. Wave
Motion 21, 149–162.
5. Conclusion Cassidy, N.J., 2008. Processing, modelling and analysis. In: Jol, H. (Ed.), Ground
Penetrating Radar Theory and Applications. Elsevier, pp. 141–176.
Cassidy, N.J., Millington, T.M., 2009. The application of finite-difference time-domain
In this paper, the application and operational performance of NDT- modelling for the assessment of GPR in magnetically lossy materials. Journal of
based, pulse-echo, ultrasonic concrete techniques (using shear waves) Applied Geophysics 67 (4), 296–308.
has been compared to the more popular method of ground-penetrating Cassidy, N.J., Gertisser, R., Nuzzo, L., Charbonnier, S., Preece, K., 2009. Evaluating block-
and-ash flows : a GPR facies approach from Merapi Volcano, Central Java, Indonesia,
radar (GPR) for the investigation of a real-world, geotechnical problem 2009. Proceedings of the 5th International Workshop on Advanced Ground
— the location of relict mine shafts/vents or karst dissolution features Penetrating Radar, Granada, pp. 78–82.
under steel-reinforced concrete sections. Complementary GPR and Chang, C.W., Lin, C.H., Lien, H.S., 2009. Measurement radius of reinforcing steel bar in
concrete using digital image GPR. Construction and Building Materials 23 (2),
ultrasonic surveys were collected across a 0.3 m-thick, steel-reinforced 1057–1063.
concrete test ‘slab’ that covered a sub-metre scaled target void buried in Chen, D.H., Scullion, T., 2008. Detecting subsurface voids using Ground-coupled
the underlying clayey soils. To make the study as realistic as possible, Penetrating Radar. Geotechnical Testing Journal 31 (3), 217–224.
Conyers, L.B., Goodman, D., 1997. Ground-Penetrating Radar. An Introduction for
and ensure data compatibility, the surveys were collected ‘blind’ Archaeologists. AltaMira Press, London, U.K.
(i.e., without a detailed a priori knowledge of the target depths and/or Daniels, D.J., 2004. Ground Penetrating Radar — 2nd Edition (Radar, Sonar, Navigation
sub-surface conditions) and across the same survey grid. Data was and Avionics Series 15. Institute of Electrical Engineers, London.
De Souza, T., 2005. Ground penetrating radar as an alternative to radiography. Insight
processed using proprietary software with standard GPR/seismic data 47 (7), 414–415.
processing techniques and converted into 3D data volumes for Dilek, U., 2007. Ultrasonic pulse velocity in nondestructive evaluation of low quality
visualisation with time-slicing (‘C-scans’) and ladder diagrams. The and damaged concrete and masonry construction. Journal of Performance of
Constructed Facilities 21 (5), 337–344.
results show that despite the relative popularity of GPR, ultrasonic
Giannopoulos, A., Macintyre, P., Rodgers, S., Forde, M.C., 2002. GPR detection of voids in
methods can perform equally well for this type of specific investigation post-tensioned concrete bridge beams. SPIE Proceedings of the Ninth International
scenario (i.e., void location beneath weathered, steel-reinforced Conference on Ground Penetrating Radar (GPR2002), pp. 376–381.
concrete). In fact, the ultrasonic surveys have shown the potential for Hugenschmidt, J., 2002. Concrete bridge inspection with a mobile GPR system.
Construction and Building Materials 16, 147–154.
overcoming some of the inherent limitations of GPR (i.e., the need for Hugenschmidt, J., Kalogeropoulos, A., 2009. The inspection of retaining walls using GPR.
careful antenna frequency selection and survey design in order to image Journal of Applied Geophysics 67 (4), 335–344.
276 N.J. Cassidy et al. / Journal of Applied Geophysics 74 (2011) 263–276

Iyer, S.R., Sinha, S.K., Schokker, A.J., 2005. Ultrasonic C-scan imaging of post-tensioned Pringle, J.K., Stimpson, I.G., Toon, S.M., Caunt, S., Lane, V.S., Husband, C.R., Jones, G.M.,
concrete bridge structures for detection of corrosion and voids. Computer-Aided Cassidy, N.J., Styles, P., 2008. Geophysical characterisation of derelict coalmine
Civil and Infrastructure Engineering 20, 79–94. workings and mineshaft detection: a case study from Shrewsbury, UK. Near Surface
Lange, Y.V., Moujitski, V.F., Shevaldykin, V.G., Kozlov, V.N., Samokrutov, A.A., 1998. Non- Geophysics 6, 185–194.
destructive testing of multilayer structures and concrete. Insight 40 (6), 400–403. Sandmeier Software, 2010. Reflex-Win v. 5.5. Sandmeier Software, Karlsruhe, Germany.
Maierhofer, C., 2003. Nondestructive evaluation of concrete infrastructure with ground Shaw, P., Xu, A., 1997. Pulse-echo methods in inspection of concrete structures. Insight
penetrating radar. Journal of Materials in Civil Engineering 15 (3), 287–297. 39 (3), 180–185.
Maierhofer, C., Brink, A., Röllig, M., Wiggenhauser, H., 2003. Detection of shallow voids Shevaldykin, V., Samokrutov, A., Kozlov, V., 2002. Ultrasonic low-frequency transducers
in concrete structures with impulse thermography and radar. NDT&E International with dry dot contact and their applications for evaluation of concrete structures.
36, 257–263. IEEE Ultrasonics Symposium Proceedings 1 & 2, 793–798.
Martina, J., Broughton, K.J., Giannopolous, A., Hardy, M.S.A., Forde, M.C., 2001. Styles, P., McGrath, R., Thomas, E., Cassidy, N.J., 2005. The use of microgravity for cavity
Ultrasonic tomography of grouted duct post-tensioned reinforced concrete bridge characterisation in karstic terrains. Quarterly Journal of Engineering Geology &
beams. NDT&E International 34, 107–113. Hydrogeology 38, 155–169.
Muldoon, R., Chalker, A., Forde, M.C., Ohtsu, M., Kunisue, F., 2007. Identifying voids in Utsi, V., Utsi, E., 2004. Measurement of reinforcement bar depths and diameters in
plastic ducts in post-tensioning prestressed concrete members by resonant concrete. Proceedings of the Tenth International Conference on Ground Penetrating
frequency of impact-echo, SIBIE and tomography. Construction and Building Radar (GPR2004), pp. 659–662.
Materials 21, 527–537.

You might also like