Review of moisture measurements in civil engineering with ground penetrating radar – Applied methods and signal features [original]

Review

Review of moisture measurements in civil engineering with ground

penetrating radar – Applied methods and signal features

Tim Klewe

⇑

, Christoph Strangfeld

, Sabine Kruschwitz

a,b

Bundesanstalt für Materialforschung und -prüfung, Unter den Eichen 87, 12205 Berlin, Germany

Technische Universität Berlin, Straße des 17. Juni 135, 10623 Berlin, Germany

highlights

A detailed literature review is given regarding the different signal features for moisture measurement in civil engineering.

Applied signal features can be grouped into amplitude, time and frequency features.

An Overview of applied signal features is presented regarding their group, wave type, preceding survey method and use for moisture content estimation.

Quantitative estimations deliver good results in laboratory conditions, on site measurements underly a lot of uncertainties and mostly allow only

qualitative estimations.

Most reviewed publications consider only one signal feature, thus future approaches may include multiple features for further development.

article info

Article history:

Received 7 July 2020

Received in revised form 15 December 2020

Accepted 3 January 2021

Available online 2 February 2021

Keywords:

Ground Penetrating Radar (GPR)

Moisture

Civil engineering

Signal features

abstract

When applying Ground Penetrating Radar (GPR) to assess the moisture content of building materials, dif-

ferent medium properties, dimensions, interfaces and other unknown influences may require specific

strategies to achieve useful results. Hence, we present an overview of the various approaches to carry

out moisture measurements with GPR in civil engineering (CE). We especially focus on the applied signal

features such as time, amplitude and frequency features and discuss their limitations. Since the majority

of publications rely on one single feature when applying moisture measurements, we also hope to

encourage the consideration of approaches that combine different signal features for further

developments.

Ó2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license

(http://creativecommons.org/licenses/by/4.0/).

Contents

1. Introduction . . . ........................................................................................................ 2

2. Fundamentals of GPR. . . . . . . . . . . . . . . ..................................................................................... 2

2.1. Wave propagation and received signals. . . . . ........................................................................... 2

3. Configurations and their parameters . . ..................................................................................... 3

4. Moisture content of building materials ..................................................................................... 3

5. Determination of material moisture based on GPR . . . . . . . . . . .................................................................. 4

5.1. Travel time and velocity features. . . . . . . . . . ........................................................................... 5

5.2. Amplitude and attenuation features . . . . . . . ........................................................................... 5

5.3. Features in the frequency domain . . . . . . . . . ........................................................................... 6

5.4. Inversion-based methods ........................................................................................... 6

5.5. Problems and limitations of on site investigations . . . . . . . . . . . . . . . ........................................................ 6

6. Summary and conclusion. . . . . . . . . . . . ..................................................................................... 7

Declaration of Competing Interest . . . . ..................................................................................... 7

Acknowledgments . . . . . . . . . . . . . . . . . ..................................................................................... 7

References . . . . ........................................................................................................ 8

https://doi.org/10.1016/j.conbuildmat.2021.122250

0950-0618/Ó2021 The Author(s). Published by Elsevier Ltd.

This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

⇑

Corresponding author.

E-mail address: [email protected] (T. Klewe).

Construction and Building Materials 278 (2021) 122250

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

1. Introduction

In civil engineering (CE), all constructions are facing several

types of degradation, which cause expensive costs of repair. Fre-

quently occurring degradations are the migration of chloride ions

[1], leading to corrosion [2], alkali-silica reaction [3], mould [4],

frost-thaw cycle, micro cracking [5], spalling, salt efflorescence

[6], etc. All these degradation processes are promoted by an

increased amount of material moisture and affect building materi-

als such as concrete, steel, wood, stones, bricks, and screed [7].

Constructions in the field of transport infrastructure, building con-

structions, or hydraulic structures are mainly made of reinforced

concrete [8]. Therefore, the moisture content in the porous cement

matrix of concrete is an important and crucial parameter. How-

ever, a reliable and non-destructive investigation of the moisture

content is still very complex and challenging, while the costs of

repair rise continuously. Especially in the field of building con-

structions, ageing pipe systems are becoming a high burden for

insurance. In 2018, 2:9 billion Euro of damage was caused by piped

water, accounting for the largest share (49%) of building insurance

claims in Germany [9]. In the event of damage, the accurate deter-

mination and localization of water ingress is essential to plan for

and perform efficient renovations.

There are various methods to approach this task, ranging from

destructive (Darr test, calcium carbide method) to non-

destructive procedures (electrical, capacitive, radiometric, thermal

and hygromatic) [10,11]. While destructive methods are time and

cost intensive, most non-destructive measurements are compara-

tively easy to perform, however, without the use of complex pro-

cessing and interpretation procedures, they deliver only

qualitative results.

Ground Penetrating Radar (GPR) is a fast and widely used mea-

surement method based on electromagnetic waves. Although

Radio Detection And Ranging (Radar) is commonly applied for air-

craft and ship detection [12], it is used in many other applications

like weather forecasts [13] or collision avoidance in the automotive

industry [14]. Apart from that, it recently received attention from a

wider audience with its application to prove the existence of water

on Mars [15]. Over the last decades, many publications already

demonstrated the high sensitivity of GPR for water, especially in

geophysics [16,17]. However, also in CE it is increasingly used for

nondestructive moisture measurements (e.g. [18–22]).

A variety of publications have shown multiple ways to assess

moisture with GPR. Different building materials and inspection

purposes require individual methods concerning the antenna con-

figuration or the used frequency. The applied setup as well as the

underlying structure highly influence the subsequent interpreta-

tion of the recorded data, which is why also numerous signal fea-

tures have been established. Due to the many possibilities to

perform moisture measurements with GPR, this review gives an

overview of the various methods and their limitation in the con-

text of CE. It shall serve as a basis for inexperienced readers, as

well as for advanced users, to encourage further development.

Approaches that combine different features or even methods, like

data fusion and machine learning, could also benefit from this

overview.

2. Fundamentals of GPR

GPR signals are electromagnetic waves whose dynamic behav-

ior is mathematically expressed by Maxwell’s equations [23,24].

The wave propagation in investigated objects is highly dependent

on the particular electrical properties of the underlying material.

Those properties are described by dielectric permittivity

, electri-

cal conductivity

, and magnetic permeability

While an influence of

is neglectable in the majority of GPR

applications,

and

are of high importance [25]. A low electrical

conductivity

provides the best conditions for an effective use

of GPR since the energy is primarily maintained in low-loss mate-

rials. Less attenuation allows signals to penetrate deeper into a

given material, which results in a better signal-to-noise ratio for

low-lying regions of interest. The velocity of wave propagation in

low-loss and nonmagnetic materials can be calculated with Eq.

(1) [25,26].

¼c

ﬃﬃﬃﬃ

pð1Þ

As the speed of light c

is constant, the wave velocity is a function of

the relative permittivity e

(with e

and the electric permittivity

in vacuum e

¼8:854 10

6As

). The relative permittivity describes

the dielectric characteristics of materials and is defined as a com-

plex value, where the real and imaginary part are associated with

energy storage and dissipation, respectively [16]. While the real

part of most building materials is considered as constant and dom-

inant between 10 MHz and 1000 MHz, the frequency-dependent

loss becomes relevant as you get closer to the water relaxation fre-

quency between 10 GHz and 20 GHz [25,27]. According to the pop-

ular models of Debye [28] and Cole–Cole [29] this results in a

decreasing real part and an increasing imaginary part of e

for

higher frequencies. Following Eq. 1smaller real parts of e

cause

higher velocities. However, this effect is not measurable in common

GPR applications on building materials, since the high frequency

components of transmitted broadband impulses also experience

more attenuation due to an increased imaginary part. The frequency

dependent attenuation leads to a shift towards lower frequencies

that can be observed with rising water contents (see Section 5.3).

Furthermore, the enhanced scattering of higher frequencies on

heterogeneities leads to additional damping, which limits the possi-

ble penetration depth [30,31].

The real value of

is 1 for air and approximately 81 for pure

water in liquid phase. For most of dry soils and building materials

is in the range of 2–9 [26] and increases with a higher water con-

tent (e.g.

is 10 to 20 for wet concrete). Due to these significant

differences, the water content is considered as the main influence

on dielectric properties of building materials, which highly justifies

the use of GPR for moisture measurements [32].

A related method that also utilizes the influence of water on the

electric permittivity and the resulting attenuation of electromag-

netic waves is the application of microwave measurements. A

review of the various concepts is presented by Okamura [33].

2.1. Wave propagation and received signals

The occurrence of reflection, refraction, and transmission on

boundaries between two different permittivities is an important

behavior to describe the propagation of electromagnetic waves.

Fig. 1. Principle of GPR reflection and underlying signal paths between a

transmitter T

and a receiver R

: A – direct air wave, G – direct ground wave, R –

reflected wave, C – critical refracted wave (according to [16,25]).

T. Klewe, C. Strangfeld and S. Kruschwitz Construction and Building Materials 278 (2021) 122250

The simplified ray-based wave paths describing these principles

are shown in Fig. 1. The amount of reflected energy (R, black) of

a transmitted pulse is dependent on the difference between the

permittivities

and

of two considered materials. Expressed as

a factor in the range of 1 and 1, the reflection coefficient ris cal-

culated as follows:

r¼ﬃﬃﬃﬃﬃ

pﬃﬃﬃﬃﬃ

ﬃﬃﬃﬃﬃ

pþﬃﬃﬃﬃﬃ

pð2Þ

Another important wave path is taken by the direct air (A, grey) and

direct ground (G, grey) wave, which propagate in their respective

medium on the shortest way between transmitter and receiver.

With a small antenna spacing, both waves superpose and form

the direct wave (DW) that is seen first in the received signal. A qual-

itative example of such a recorded energy-signal in time domain (A-

Scan) is shown in Fig. 2. The DW is followed by a reflection wave

(RW), with a certain peak-to-peak amplitude A

pp;RW

and a time dif-

ference

between DW and RW. These basic signal features, as

well as the peak-to-peak amplitude of the DW, can already serve

as good indicators for moisture measurements on building materi-

als, which will be further discussed in Section 5. Similar to optics,

refraction is described by Snell’s law that deals with the relation-

ship of occurring angles on interfaces. For reasons of clarity, the

occurring refraction from e

to e

is not depicted in Fig. 1. Critical

refractions and reflections (C, grey) also exist. They often superpose

with the reflected wave R, which makes a separate investigation dif-

ficult. In general, the received A-Scan is always a superposition of all

different wave paths, whereas their individual amplitude and sepa-

ration in time and space depend on the spacing between transmit-

ter and receiver as well as on the target depth and on the elevation

of the antenna [31]. Different GPR configurations, therefore, have a

considerable influence on the received signals and their interpreta-

tion. The commonly used set-ups for moisture measurements in

building materials and their underlying parameters are discussed

in the following section.

3. Configurations and their parameters

GPR configurations are divided into two types: reflection and

transillumination measurements. Transillumination is rarely per-

formed in civil engineering since a suitable back wall is often inac-

cessible (e.g tunnels, foundations, containments). Also, the

transmission between a pair of boreholes, as it is applied by geo-

physicists for soil moisture measurements [24,34], is commonly

not used to investigate building materials. This paper, therefore,

focuses on reflection configurations. Here, the common–offset

reflection survey, as shown in Fig. 3, is the simplest and mostly

applied configuration. As the name of the method suggests, the

spacing sbetween a single transmitter and a receiver is kept con-

stant for every measurement position on a survey line. The most

significant parameter is the applied center frequency, which

describes the dominant frequency of the transmitted broadband

pulse. As mentioned in Section 2, it influences the general attenu-

ation and, therefore, the penetration depth. When it comes to mul-

tilayered mediums or rebar localization, the spatial resolution

becomes more relevant since it rises with shorter wavelengths.

This creates a trade-off between resolution and penetration depth

when choosing the most suitable center frequency. GPR frequen-

cies used for material moisture measurements are mainly between

100 kHz and 10 GHz. In case of concrete, frequencies between

1 GHz to 3 GHz yield a good trade-off between spatial resolution

and penetration depth. Other essential parameters are the antenna

spacing s, the antenna orientation, the bit resolution, recording

time window, sampling frequency, the distance between each

measurement and the parallel line spacing (when performing more

than one survey line) [24]. The suitable choice of these values

highly depends on the particular test object and purpose.

When using GPR in reflection configuration, one rarely has pre-

cise information about the underlying layer depth. In that case, a

profound water content estimation through the use of velocity

data is impossible since the traveled distance is unknown. Here,

multi-offset configurations like common-midpoint (CMP) and wide

angle reflection and refraction (WARR), as shown in Fig. 4, can be

used. While the CMP method uses a fixed midpoint with a stepwise

increasing spacing between transmitter and receiver, WARR con-

figurations often have one source at a constant location with mul-

tiple receivers in defined distances. WARR can also be performed

by using a single receiver (or transmitter) that is moved further

from the corresponding transmitter (or receiver) with every mea-

surement. The series of reflection waves (B-Scan) collected by

multi-offset techniques form a hyperbola from which velocity

and layer depth can be estimated [35–37]. After that, the obtained

vertical velocity model allows a layer specific permittivity

estimation.

4. Moisture content of building materials

As discussed in Section 2, the permittivity of an investigated

material heavily depends on the moisture content and, therefore,

Fig. 2. Example of a recorded energy-signal in time domain (A-Scan) including the

direct wave (DW) and reflection wave (RW). Basic signal features are the peak-to-

peak amplitudes A

pp;DW

and A

pp;RW

located at the first peaks of DW and RW. The

time difference

between both peaks is often considered as the travel time.

Fig. 3. Principle of a line measurement with GPR in common-offset configuration and its most important signal paths: Direct air wave, direct ground wave and the reflected

waves.

T. Klewe, C. Strangfeld and S. Kruschwitz Construction and Building Materials 278 (2021) 122250

changes when water is adsorbed. Water adsorption or desorption

generally requires an open porosity as well as a hydrophobic

behavior. Several building materials possess these material proper-

ties such as: concrete, screed, mineral rock wool, calcium silicate

plate, etc. In these, the adsorbed water is located in completely

filled pores and partially saturated pores [38].

Concrete is a highly heterogeneous material. Every concrete

mixture has its own composition of water, cement, minerals,

aggregates, additives, etc. This defines the resulting material prop-

erties, e.g. comprehensive strength, the dry density, and resistance

to water and chlorides. Furthermore, the resulting pore system is

also influenced significantly by the concrete composition as well

as by the ambient temperature and humidity. On the long-term,

the pore system might vary due to carbonation, micro-cracking,

and sulphate attack. Thus, every pore system and hence every con-

crete is unique, and a general calibration function does not exist.

Furthermore, the properties of the pore water are not time con-

stant. Chloride and sulphate ingress change the surface tension of

water, which changes the moisture content as well as the relative

permittivity. In conclusion, quantitative moisture measurements

of concrete would require intensive calibration for all the influenc-

ing factors mentioned above. Also in case of assumed perfect dry

concrete samples (all evaporable pore water is released), no uni-

form permittivity exists. Every cement mixture contains different

amounts of chemically bound water, calcium, aluminum, iron,

etc. which influence the final permittivity.

The open porosity varies between 5% and 25% in most cases. The

pore sizes are distributed over several magnitudes and are

described by a representative diameter. Micro-pore diameters are

between 0.4 nm and 2 nm, so called meso-pores are in the range

of 2 nm to 100 nm, and macro-pores range from 100 nm up to

100

m[39]. In these pores, the material moisture is located in liq-

uid phase at the pore surface/ fringe and in vapor phase in the pore

center. Occurring transportation processes are a two-phase flow

including both phases, vapour and liquid. The total porosity and

the distribution of the different pore sizes determine the capability

of adsorption and desorption of water. The function which relates

the water vapour partial pressure, i.e. relative humidity and the

resulting material moisture is the so-called sorption isotherm.

In most cases, the moisture content of concrete is between

0.15 M% and 10 M%. It is defined as the ratio between the mass

of moisture divided by the mass of the dry material. Besides this

definition, sometimes the moisture content is based on the mass

of the wet material or the volumetric moisture content is given

[40]. For comparison, these values must be converted to each

other. In the following, we use the definition based on the mass

of the dry material. Obviously, this requires the knowledge of the

dry mass. One common approach is oven drying to evaporate the

mobile water in the pores [41]. Depending on the sample size, this

might take up to several days. If mass constant is reached, the sam-

ple is considered as dry and all evaporable/ mobile water is

released. Nevertheless, the cement matrix still contains a lot of

chemically bound water. This water is immobile and never con-

tributes to the moisture transport [11]. However, if the oven tem-

perature is too high, this chemically bound water is partially

released as well, which leads to systematic deviations from the

‘‘true” moisture content. In practice, various drying temperatures

are established for different materials (cement based, calcium-

sulphate based, clay minerals, etc.) and different testing purposes.

Therefore, we will emphasize to consider clearly the drying

method and a suitable temperature (in case of oven drying) when

using moisture content values for calibration.

5. Determination of material moisture based on GPR

After the reflected energy-signals are recorded (Fig. 2), there are

two possible ways to estimate the underlying water content h. The

first is illustrated by the steps A and B in Fig. 5, where the

h-estimation is based on a preceding

-estimation. This is an appro-

priate approach, since the electric permittivity

strongly depends

on the material moisture [42]. However, such two-step calcula-

tions may increase the risk of error accumulation and therefore,

already require a precise determination of

. With a known

petrophysical models derived from empirical studies can be

applied to calculate hð

Þ. Topp’s equation [43] is the most com-

monly used model in geophysics to estimate soil moisture

[16,17]. Therefore, it is also applied in civil engineering to derive

hof subgrades below pavement and railway structures

[20,44,45]. It provides a basic polynomial approximation for hð

Þ,

which only regards the real part of

. For asphalt mixes, Fernandes

et al. [46] applied a modification of the complex refractive index

model (CRIM) derived by Wang and Lytton [47] to estimate the

moisture content. CRIM offers a more complex approach, which

requires further knowledge about the volumetric fraction of differ-

ent material components. An overview of various petrophysical

models and further literature can be found in [48].

Fig. 4. GPR with multiple offsets: common-midpoint (CMP, left) and wide angle reflection and refraction (WARR, right) configuration.

Fig. 5. Principle of signal processing to estimate the moisture content hby using the reflected energy-signal of GPR.

T. Klewe, C. Strangfeld and S. Kruschwitz Construction and Building Materials 278 (2021) 122250

When there is no established model for the investigated mate-

rial, like for concrete, site-specific relationships based on the col-

lected data serve as a suitable way to evaluate h[49]. However,

those fitted models always require an accurate reference of the

underlying water contents. Here, the gravimetric method, as dis-

cussed in the previous section, provides results by drying and

weighing collected samples (e.g. drilling cores) according to ASTM

D2216-19 [41].

Such references also enable the second possible way to estimate

h, which is shown by step C in Fig. 5. This approach skips the

estimation by formulating a direct relationship between the

received GPR signals and the obtained moisture data. Both

approaches, AB and C, require characteristic signal features that

correlate with h. Those features commonly applied in civil engi-

neering can be divided into time, amplitude, and frequency fea-

tures. Their related methods and limitations will be discussed in

the following sections.

5.1. Travel time and velocity features

Measuring the time

tthat it takes for a transmitted and

reflected impulse to travel back to its origin is by far the oldest sig-

nal feature in radar technology [50]. It is generally used to detect

objects and to calculate their distance dwith Eq. 3by means of

the known wave velocity

¼2d

tð3Þ

When performing moisture inspections with GPR, the permittivity

of an investigated medium and hence the propagation velocity

(Eq. 1) is obviously unknown. Therefore, a known reflector depth

is one possibility to calculate

and eby measuring

t. It is usually

performed in laboratory studies where the dimensions of test spec-

imens are given, like in the works of Fernandes et al. [46], Lai et al.

[22] and Laurens et al. [21]. Such set-ups allow an accurate determi-

nation of edepending on the water content hwith a normal com-

mon–offset survey. This approach is also applicable to

multilayered structures, as shown by Grote et al. [20]. Here,

tis

the time passing between two reflections from the top and bottom

of a specific layer. In singlelayered mediums,

tis often defined as

the time difference between the first peak of the direct and the

reflection wave (see Fig. 2), since the true time-zero is not available

[51]. In the study of Viriyametanont et al. [52], the authors even

propose a delay correction when using the direct wave as a time ref-

erence. This is motivated by a measurable influence of the underly-

ing moisture content on the direct wave velocity, which can vary in

near surface areas. The correction is calculated with a given trans-

mitter to receiver offset and the wave velocity obtained from vari-

ous rebar reflections in known depths.

In practice, usually the propagation velocity, as well as the

reflector depth, are unknown. In this case, ASTM D6432-11 [37]

suggests two techniques to obtain

from measured travel times.

The first is called hyperbolic geometry and can be performed in

common–offset configuration. It requires point reflectors such as

pipes or rebars that cause a hyperbolic pattern in the radargram

(B-Scan), from which the propagation velocity can be derived. Che-

ung and Lai [53] recently used this technique to detect water-pipe

leakages by observing significant velocity drops compared to

undamaged dry areas. A quantitative approach was studied by

Koyan et al. [54] who presented a migration-based velocity analy-

sis on multiple rebar reflections, however, this was done with

known reflector depths to conduct a time-zero correction. Here,

occurring uncertainties were mainly caused by interferences due

to a small spacing between the rebars and a limited vertical reso-

lution for shallow rebar depths.

The second way to obtain

without a known dis called velocity

sounding which describes the use of the CMP or WARR survey

method, as mentioned in Section 3. CMP measurements were

applied by Cai et al. [45] and Grote et al. [20] to obtain a vertical

velocity characterization for railway and pavement structures,

respectively. The use of such multi-offset configurations also

allows greater spacing between transmitter and receiver. There-

fore, the travel time of the direct wave gets more relevant while

it is barely considered in common–offset configuration. Klysz and

Balayssac [19] suggest the analysis of direct wave travel times with

WARR configuration to estimate the mositure content of concrete

covers. However, it is important to mention that only near surface

moisture can be measured with this technique, since the penetra-

tion depth of the ground wave is limited [55].

5.2. Amplitude and attenuation features

Quantitative estimation of underlying permittivities by analyz-

ing the received signal amplitudes is very challenging compared to

the use of travel times [20]. The theoretical approach is given with

Eq. 2, which suggests the evaluation of occurring reflections on a

boundary of two different permittivities. Besides the need of one

given

, this method often leads to inaccuracies due to several

other influences on the amplitude. In complex mediums, like con-

crete, the propagating waves experience a lot of scattering through

heterogeneously distributed electrical properties [25].An

increased moisture content leads to the presence of more water

filled pores and thus to more heterogeneities and a higher conduc-

tivity. Furthermore, attenuation grows with longer travel paths

and rising frequencies, which distorts an amplitude-based

estimation even more.

The use of an air-launched and, therefore, air-coupled antenna

configuration makes it possible to analyze the reflection that

occurs on the air-material interface. Air, as the overlying medium,

has a known permittivity

air

¼1 and causes comparatively low

losses. Thus, the permittivity of the underlying material

can

be calculated with the following equation:

¼1þ

1

ð4Þ

Here, the reflection coefficient is defined as r¼A

with the

reflected air-surface amplitude A

, that is scaled by the highest pos-

sible amplitude A

. This maximum can be easily obtained with the

use of a metallic reflector in the same distance as the investigated

surface. Benedetto et al. [44] used this technique to map the spatial

variation of soil moisture as a preinvestigation for pavement appli-

cations. However, this approach only covers moisture near the sur-

face and ignores possible permittivity changes in deeper areas. To

capture those, Maser and Scullion [56] also analyzed the reflections

from the asphalt-bottom to estimate the moisture content of the

underlying base layer. Obviously, this method encounters all the

aforementioned uncertainties with the additional risk of error accu-

mulation by the use of previously estimated asphalt permittivities.

Due to the numerous unknown influences for on site investiga-

tions, most of the works that study the correlation between ampli-

tudes and water content were performed under laboratory

conditions. A controlled environment also allows the use of addi-

tional metallic reflectors below the investigated specimens to

emphasize the occurring reflections or to generate reference

amplitudes for scaling/ normalization. As a signal feature, Hugen-

schmidt and Loser [57] used the quotient between the reflected

amplitudes at the concrete surface and those at an aluminum sheet

below the bottom. The authors observed a qualitative correlation

with the water content of concrete as well as with ingressed chlo-

rides that increase the underlying conductivity and therefore cause

T. Klewe, C. Strangfeld and S. Kruschwitz Construction and Building Materials 278 (2021) 122250

Loading more pages...