Condition Assess of GFRP-Retrofitted Concrete Cylinders Using Electromagnetic Waves (2008) - Thesis PDF

Condition Assessment of GFRP-Retrofitted
Concrete Cylinders Using Electromagnetic Waves

by
Tzu-Yang Yu
Submitted to the Department of Civil and Environmental Engineering
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in the field of Structures and Materials
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2008
c Massachusetts Institute of Technology 2008. All rights reserved.

Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Civil and Environmental Engineering
May 8, 2008
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Oral Buyukozturk
Professor of Civil and Environmental Engineering
Thesis Supervisor
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Daniele Veneziano
Chairman, Departmental Committee for Graduate Students
Condition Assessment of GFRP-Retrofitted Concrete

Cylinders Using Electromagnetic Waves
by
Tzu-Yang Yu
Submitted to the Department of Civil and Environmental Engineering
on May 8, 2008, in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy in the field of Structures and Materials
Abstract
The objective of this study is to develop an integrated nondestructive testing (NDT)
capability, termed FAR NDT (Far-field Airborne Radar NDT), for the detection of
defects, damages, and rebars in the near-surface region of glass fiber reinforced polymer (GFRP)-retrofitted concrete cylinders through the use of far-field radar measurements (electromagnetic or EM waves). In this development, two far-field monostatic ISAR (inverse synthetic aperture radar) measurement schemes are identified
for collecting radar measurements, and the backprojection algorithm is applied for
processing radar measurements into spatial images for visualization and condition
assessment. Reconstructed images are further analyzed by mathematical morphology to extract a numerical index representing the feature of the image as a basis
for quantitative evaluation. The components of the development include dielectric
modeling of materials, laboratory radar measurements, numerical simulation, and
image reconstruction. It is found that using the developed technique the presence of
near-surface defects can be detected by the oblique incidence measurements. Radar
signals in the frequency range of 8 GHz to 18 GHz are found effective for damage
detection in the near-surface region of the specimens. Numerical simulation using the
finite-difference time-domain (FDTD) method is conducted to understand the propagation and scattering of EM waves from the defects and inclusions in two-dimensional
and three-dimensional GFRP-concrete models. The FDTD simulation is capable of
predicting the far-field response of GFRP-concrete cylinders and beneficial to better understanding the pattern of field measurements in the application of the FAR
NDT technique. Dielectric properties of materials are investigated for their use in
numerical simulation and for improving the precision of reconstructed images. Reconstructed images of GFRP-concrete cylinders with and without artificial features
(rebar and defect) clearly indicate the presence of these features. Normal incidence
scheme is found to be effective for rebar detection, and the oblique incidence scheme
can discover near-surface defects such as GFRP debonding and delamination. The
proposed FAR NDT technique is found to be capable of detecting near-surface defects in GFRP-concrete cylinders and potentially applicable for the field condition as3
sessment of GFRP-retrofitted reinforced concrete and other reinforced concrete civil

infrastructure systems.
Thesis Supervisor: Oral Buyukozturk
Title: Professor of Civil and Environmental Engineering
Acknowledgments
It is Professor Oral Buyukozturk (Course I) who led me into the research field of condition assessment of concrete structures using electromagnetic waves through working
on a research project several years ago, on which my doctoral dissertation is essentially
based. It is also through working on this project, my research interests on various
topics were subsequently developed/discovered. His guidance, encouragements, and
supports are indispensable to me for making this dissertation possible. I am deeply
indebted to the time he spent with me at nights and on weekends, and to his tolerance
in the process of forging my research attitude and enhancing my research capabilities.
It would not have been possible for me to accomplish this work without his training
on many aspects. For that, I truly appreciate this precious opportunity he gave me
at MIT.
Professor Jerome J. Connor (Course I) was kind enough for supervising my Masters thesis when I came to MIT in 2001 and for joining my thesis committee in 2005.
It is very difficult not to be encouraged and inspired by him in every discussion we
have made through his infectious passion on research and teaching.
I am also indebted to the late Professor Jin Au Kong (Course VI) for his leading
me into the intriguing world of electromagnetism and for his valuable suggestions
made in my committee meetings. His extraordinary sense of humor proportionally
reflects the magnitude of his knowledge. His research philosophy inspires me and has
made me a good friend of SAM ever since. His sudden decease on March 13, 2008,
is an unmeasured loss to me and everyone who knows him, while his lecturing and
words still vividly survive in our memories.
It is my pleasure to have Dr. Tomasz M. Grezgorczyk (Course VI) serving on
my thesis committee. His insightful suggestions to the electromagnetic problems I
have encountered in conducting this research are most helpful and valuable. Dr.
Grezgorczyk has also been very supportive to the completion of this research in many
aspects. I am deeply grateful for his willing to guide me in exploring the world of
electromagnetism.
5
I should like to take this opportunity to express my gratitude to Professor Michael

C. Forde (University of Edinburgh, Scotland) serving as a member on the thesis
committee. His constructive suggestions and questions I have received during his
stay at MIT in 2004 and during my thesis defense in 2008 are valuable to the further
improvement of this thesis. I am grateful for his supporting this research on many
aspects and for his sharing his perspectives and thoughts on many critical problems
in civil engineering.
Special thanks go to Dr. Antonis Giannopoulos for the use of GPRMax2D/3D
and his suggestions in the numerical simulation.
Many productive and interesting discussions with Professors E. Kausel and D.
Veneziano (both in Course I) are greatly appreciated.
I would like to extend my thanks to a number of people for their help of various
kinds; to Dennis Blejer and Alex Eapen (MIT Lincoln Laboratory) for their help
on laboratory radar measurements and data interpretation; to Patricia Dixon and
Cynthia Stewart when I was in need of help in 2004; to Donna Hudson for her help
on proposal budgets; to Patricia (Patty) Glidden, Kris Kipp, Jeanette Marchocki,
and Donna Beaudry for their everyday relentlessly greetings on the aisles in Building
1. This journey would have been much more colder without their warm smiles.
I have been lucky enough to make many friends in Course I; O. Gunes, C. Au, E.
Karaca, R. Sudarshan, J. Pei, A.E. Sew, M.A. Nikolinakou, J.A. Ortega, K. Ishimaru,
S. Cheekiralla, P. Dohnalek, C. Tuakta, S. Lin, I. Tsai, J. Park, Y. Moriyama, I.(Aki)
Choo, D. Lau, as well as in other Departments including K. Lee (Course VIII), J. Chen
(Course VI), and M. Nikku (Course VIII). The time with my classmates including
Marc, Bora, Carmen, Jason, Tashan, Vimal, Luca, Chinghuei, and Sakda was also
memorisable.
Last but not least, I will always have special gratitude and love for my grandma,
Shun Jen, my father, Jr-Shen Yu, my mother, Kuei-Yin Shu, my brother, Shun-Hwa
Yu. Their endless, unconditional supports warm my heart as always. Finally, I want
to dedicate this thesis to my wife, Kaiwen Chen, who has been my Muse on many
aspects ever since she walked into my life.
6
Contents
1 Introduction and Research Motivation
23
1.1
Research Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
1.2
Research Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
1.3
Organization of the Dissertation . . . . . . . . . . . . . . . . . . . . .
31
2 Literature Review
2.1
2.2
35
Nondestructive Testing (NDT) Techniques . . . . . . . . . . . . . . .
36
2.1.1
Optical Methods . . . . . . . . . . . . . . . . . . . . . . . . .
42
2.1.2
Acoustic Methods . . . . . . . . . . . . . . . . . . . . . . . . .
44
2.1.3
Thermal Methods . . . . . . . . . . . . . . . . . . . . . . . . .
48
2.1.4
Radiographic Methods . . . . . . . . . . . . . . . . . . . . . .
51
2.1.5
Magnetic and Electrical Methods . . . . . . . . . . . . . . . .
53
2.1.6
Microwave and Radar Methods . . . . . . . . . . . . . . . . .
56
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
3 Numerical Simulation
63
3.1
Maxwells Curl Equations and Linearly Polarized EM Waves . . . . .
65
3.2
Finite Difference Time Domain Solution and Yees Algorithm . . . . .
68
3.3
Absorbing Boundary Condition Perfectly Matched Layer . . . . . .
73
3.4
Stability Criteria in Discretization . . . . . . . . . . . . . . . . . . . .
74
3.4.1
Discretization in Space . . . . . . . . . . . . . . . . . . . . . .
75
3.4.2
Discretization in Time . . . . . . . . . . . . . . . . . . . . . .
76
Two-Dimensional and Three-Dimensional Simulations . . . . . . . . .
77
3.5
3.6
3.7
3.5.1
Validation of the Code . . . . . . . . . . . . . . . . . . . . . .
77
3.5.2
Actual Far-Field Simulation . . . . . . . . . . . . . . . . . . .
82
Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . .
85
3.6.1
Damage Detection in Normal Incidence . . . . . . . . . . . . .
86
3.6.2
Damage Detection in Oblique Incidence . . . . . . . . . . . . .
86
3.6.3
Effect of Defect Width in Normal Incidence . . . . . . . . . .
88
3.6.4
Effect of Defect Depth in Normal Incidence . . . . . . . . . . .
89
3.6.5
Rebar Detection in Normal Incidence . . . . . . . . . . . . . .
89
3.6.6
2D and 3D Responses . . . . . . . . . . . . . . . . . . . . . . 107
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4 Laboratory Radar Measurements
111
4.1
Experimental Program . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.2
Manufacturing of the Specimens . . . . . . . . . . . . . . . . . . . . . 116
4.3
Experimental Configuration and Parameters . . . . . . . . . . . . . . 118
4.4
4.5
4.6
4.3.1
Monostatic ISAR Normal Incidence Scheme . . . . . . . . . . 120
4.3.2
Monostatic ISAR Oblique Incidence Scheme . . . . . . . . . . 121
Calibration of Laboratory Radar Measurements . . . . . . . . . . . . 121

4.4.1
PEC Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.4.2
Lossy Dielectric Specimen and Its Optical Model . . . . . . . 124
Frequency-Angle Measurements . . . . . . . . . . . . . . . . . . . . . 127

4.5.1
Monostatic ISAR Normal Incidence Scheme . . . . . . . . . . 133
4.5.2
Monostatic ISAR Oblique Incidence Scheme . . . . . . . . . . 152
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
5 Image Reconstruction
159
5.1
Single Scattering and Synthetic Aperture Radar . . . . . . . . . . . . 160
5.2
Inverse Synthetic Aperture Radar . . . . . . . . . . . . . . . . . . . . 172
5.3
Backprojection Algorithms . . . . . . . . . . . . . . . . . . . . . . . . 175

5.3.1
Range Compression . . . . . . . . . . . . . . . . . . . . . . . . 175
5.3.2
Backprojection Processing . . . . . . . . . . . . . . . . . . . . 178

8
5.3.3
Support Band Analysis Method of Stationary Phase . . . . . 179
5.3.4
Advantages of Backprojection Algorithms . . . . . . . . . . . 187
5.4
Implementation and Coding Procedure . . . . . . . . . . . . . . . . . 187
5.5
Effects of Aperture Size and Bandwidth
5.6
. . . . . . . . . . . . . . . . 193
5.5.1
Aperture Size . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
5.5.2
Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
6 Dielectric Modeling of GFRP-concrete Systems

6.1
201
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.1.1
Definition and Physics of Dielectric Properties . . . . . . . . . 202
6.1.2
Dielectric Spectroscopy and Dielectric Dispersion . . . . . . . 211
6.1.3
Energy Storage and Dissipation Mechanisms . . . . . . . . . . 212
6.1.4
Dielectric Properties in Microwave and Radar NDT . . . . . . 216
6.2
Approaches for the Determination of Dielectric Properties . . . . . . 216
6.3
Integrated Methodology for the Determination of Dielectric Properties 219
6.4
6.5
6.6
6.7
6.3.1
Time Difference of Arrival (TDOA) . . . . . . . . . . . . . . . 220
6.3.2
Root-searching Optimization Scheme . . . . . . . . . . . . . . 222
6.3.3
Validation of the Methodology . . . . . . . . . . . . . . . . . . 227
Modeling Approach for the Dielectric Properties of Materials . . . . . 233

6.4.1
Internal Field Approach versus External Field Approach . . . 233
6.4.2
Geometrical Analysis . . . . . . . . . . . . . . . . . . . . . . . 242
Dielectric Properties of Water . . . . . . . . . . . . . . . . . . . . . . 247

6.5.1
Free Water
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
6.5.2
Bound Water . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Dielectric Properties of GFRP . . . . . . . . . . . . . . . . . . . . . . 263

6.6.1
Epoxy Resin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
6.6.2
E-glass Fabric . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
6.6.3
GFRP Layer/Sheet . . . . . . . . . . . . . . . . . . . . . . . . 267
Dielectric Properties of Concrete
. . . . . . . . . . . . . . . . . . . . 272
9
6.8
6.7.1
Determination of Volumetric Fractions . . . . . . . . . . . . . 275
6.7.2
Dielectric Modeling of Oven-Dried Hydrated Cement Paste . . 277
6.7.3
Challenges in the dielectric modeling of concrete . . . . . . . . 280
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
7 Condition Assessment of GFRP-concrete Systems FAR NDT
285
7.1
Components of FAR NDT . . . . . . . . . . . . . . . . . . . . . . . . 286
7.2
Physical Inspection Far-field ISAR Measurements . . . . . . . . . . 289
7.3
Numerical Processing Image Reconstruction . . . . . . . . . . . . . 294

7.3.1
Physical Meaning of the Scattering Signals in the Images . . . 294
7.3.2
Progressive Image Focusing . . . . . . . . . . . . . . . . . . . 295
7.4
Image Resolutions and Damage Detectability . . . . . . . . . . . . . . 299
7.5
Pattern Recognition Damage Detection . . . . . . . . . . . . . . . . 300
7.6
7.5.1
Local Index - Maximum Amplitude . . . . . . . . . . . . . . . 301
7.5.2
Global Index - Mathematical Morphology . . . . . . . . . . . 303
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
8 Conclusions
319
A Phase Velocity of Love Waves in A Layer Underlain by A Half Space

Medium
325
B Analytical Approach to Several Plane Wave Incidence Problems
331
B.1 Reflection Coefficient and Reflectivity . . . . . . . . . . . . . . . . . . 331

B.2 A Two-dimensional Three-layer Model . . . . . . . . . . . . . . . . . 334
B.3 A Three-dimensional Infinite Dielectric Cylinder Model . . . . . . . . 336
B.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
10
List of Figures
1-1 Several FRP strengthening/repair scenarios of concrete structures . .
25
1-2 Intact multi-layer cementitious composite systems . . . . . . . . . . .
26
1-3 Damaged multi-layer cementitious composite systems . . . . . . . . .
26
1-4 Detected air voids at various locations in a bridge box-girder wall and
their repair . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
1-5 Several failure modes in GFRP-concrete systems . . . . . . . . . . . .
28
1-6 Modeling of construction defects and structural damages using artificial

anomaly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
1-7 Organization of the dissertation . . . . . . . . . . . . . . . . . . . . .
32
2-1 Procedure of NDT Techniques . . . . . . . . . . . . . . . . . . . . . .
37
3-1 Configuration grid in a two-dimensional domain . . . . . . . . . . . .
70
3-2 Derivation of global and local ABCs . . . . . . . . . . . . . . . . . . .
74
3-3 Quantization error Er (kx) vs. kx . . . . . . . . . . . . . . . . . .
76
3-4 Geometry of a 2D model for validation . . . . . . . . . . . . . . . . .
79
3-5 Gaussian current source and the reflection from PEC . . . . . . . . .
80
3-6 Sinusoidal current source and the reflection from PEC . . . . . . . . .
81
3-7 Gaussian current source and the reflection from PEC . . . . . . . . .
82
3-8 Gaussian current source and the reflection from PEC Close-Up . . .
83
3-9 Theoretical curve and the FDTD solution of the amplitude reflection
coefficient of a 2D dielectric plate . . . . . . . . . . . . . . . . . . . .
84
3-10 Relative difference/error between the theoretical curve and the FDTD
solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11
85
3-11 Sinusoidal current source and the reflection from a 2D dielectric plate
86
3-12 Numerical domain for simulating actual far-field radar measurements
87
3-13 Incident field, total field, and net reflection field of a 9 GHz signal . .
88
3-14 Scattered (net reflection) field of Ez Intact, lossless concrete cylinder
91
3-15 Total field of Ez Intact, lossless concrete cylinder . . . . . . . . . .
92
3-16 Total field of Ez Intact, lossy concrete cylinder . . . . . . . . . . . .
93
3-17 Scattered field of Ez , = 0 Damaged, lossless concrete cylinder . .
94
3-18 Scattered field of Ez , = 0 Damaged, lossy concrete cylinder . . .
95
3-19 Scattered field of Ez , = 30 Intact, lossless concrete cylinder . . .
96
3-20 Scattered field of Ez , = 30 Damaged, lossless concrete cylinder .
97
3-21 Thirteen 2D models for studying the effect of defect width in normal
incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
98
3-22 Variation of reflected electric field with respect to different defect widths
in normal incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . .
98
3-23 Sixteen 2D models for studying the effect of defect depth in normal
incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
99
3-24 Variation of reflected electric field with respect to different defect depths
in normal incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . .
99
3-25 Total field of Hx Plain concrete cylinder with a center rebar . . . . 100
3-26 Total field of Hy Plain concrete cylinder with a center rebar . . . . 101
3-27 Total field of Ez Plain concrete cylinder with a center rebar . . . . 102
3-28 Scattered field of Hx Plain concrete cylinder with a center rebar . . 103
3-29 Scattered field of Hy Plain concrete cylinder with a center rebar . . 104
3-30 Scattered field of Ez Plain concrete cylinder with a center rebar . . 105
3-31 Scattered field of Ez Plain concrete cylinder with a center rebar . . 106
3-32 A 3D lossless dielectric cylinder model . . . . . . . . . . . . . . . . . 107
3-33 Comparison between the 2D and 3D responses of a PEC cylinder model108
4-1 Specimens CON and GFRP . . . . . . . . . . . . . . . . . . . . . . . 114
4-2 Specimens CRE and 4RE
. . . . . . . . . . . . . . . . . . . . . . . . 114
12
4-3 Specimens AD1 and AD2 . . . . . . . . . . . . . . . . . . . . . . . . . 115

4-4 Specimen AD3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4-5 Compact RCS antenna range facility in the MIT LL [Courtesy of the
MIT LL] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4-6 Schematic of the compact RCS/antenna range facility in the MIT LL
125
4-7 Far-field monostatic ISAR normal incidence scheme. Note that the
angular zero is referred to a selected point on the cylinder. . . . . . . 126
4-8 Far-field monostatic ISAR oblique incidence scheme. Note that the
angular zero is in alignment with the axis of the cylinder. . . . . . . . 127
4-9 PEC specimen Aluminum tube [Courtesy of the MIT Lincoln Laboratory] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4-10 Frequency-angle response of the Aluminum tube Amplitude (dBsm),
X-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4-11 Frequency-angle response of the Aluminum tube Amplitude (dBsm),
Ku-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4-12 Mean amplitudes and their standard deviations of the reflection response at different frequencies Aluminum tube, X-band . . . . . . . 129
4-13 Mean amplitudes and their standard deviations of the reflection response at different incident angles Aluminum tube, X-band . . . . . 130
4-14 Mean amplitudes and their standard deviations of the reflection response at different frequencies Aluminum tube, Ku-band . . . . . . 130
4-15 Mean amplitudes and their standard deviations of the reflection response at different incident angles Aluminum tube, Ku-band . . . . 131
4-16 Correlation coefficients of the reflection response at different frequencies Aluminum tube, X-band . . . . . . . . . . . . . . . . . . . . . . 131
4-17 Correlation coefficients of the reflection response at different frequencies Aluminum tube, Ku-band . . . . . . . . . . . . . . . . . . . . . 132
4-18 Lossy dielectric specimen Plexiglass rod
. . . . . . . . . . . . . . . 132
4-19 Frequency-angle response of the plexiglass specimen Amplitude (dBsm),

Ku-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
13
4-20 Simulated and measured responses Plexiglass rod, HH (upper level)

and VV (lower level) polarizations [Courtesy of the MIT LL] . . . . . 134
4-21 Frequency-angle response of specimen CON (plain concrete) Amplitude (dBsm), X-band . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4-22 Frequency-angle response of specimen CON (plain concrete) Amplitude (dBsm), Ku-band . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4-23 Frequency-angle response of the specimen GFRP (plain concrete with
GFRP) Amplitude (dBsm), X-band . . . . . . . . . . . . . . . . . . 137
4-24 Comparison between the power responses of specimens CON and GFRP,
X-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4-25 Photos of the specimen CON showing surface roughness . . . . . . . . 140
4-26 Misalignment between the cylinder specimen and the Styrofoam tower
in the normal incidence scheme . . . . . . . . . . . . . . . . . . . . . 142
4-27 Frequency-angle response of specimen AD1B Amplitude (dBsm), Xband . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4-28 Frequency-angle response of specimen AD1F Amplitude (dBsm), Xband . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4-29 Comparison between the power responses of the intact and damaged
sides of the specimen AD1, X-band . . . . . . . . . . . . . . . . . . . 144
4-30 Photos showing the concave region in the specimen . . . . . . . . . . 145
4-31 Frequency-angle response of specimen AD2 Amplitude (dBsm), X-band147
4-32 Frequency-angle response of specimen AD3 Amplitude (dBsm), X-band147
4-33 Frequency-angle response of specimen CRE Amplitude (dBsm), X-band148
4-34 Comparison between the power responses of specimens CON and CRE,
X-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4-35 Frequency-angle response of specimen 4RE (rebar 1) Amplitude (dBsm),
X-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
X-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
14

X-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
X-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4-39 Frequency-angle response of the intact side of the specimen AD1
Amplitude (dBsm), X-band . . . . . . . . . . . . . . . . . . . . . . . 153
4-40 Frequency-angle response of the damaged side of the specimen AD1
4-41 Frequency-angle response of the damaged side of the specimen AD2
5-1 Configuration of SAR measurements . . . . . . . . . . . . . . . . . . 170
5-2 Monostatic radar and its footprint . . . . . . . . . . . . . . . . . . . . 171
5-3 rs , r0 , and rj in the far-field region . . . . . . . . . . . . . . . . . . . 171
5-4 Configuration of ISAR measurements . . . . . . . . . . . . . . . . . . 184
5-5 Two cases of the reflection of radar signals . . . . . . . . . . . . . . . 185
5-6 Time shifting error . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
x , ky ) and their relationship . . . . . . . . . . . . . . . 186
5-7 Variables in I(k
5-8 Far-field ISAR measurement of the specimen AD1 HH polarization,
i = 15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
5-9 Complex form of the far-field ISAR measurement . . . . . . . . . . . 189
5-10 Desired sidelobe pattern with different sidelobe levels (SLLs) ranging
from 30dB to 90dB . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
5-11 Weighted complex form of the far-field ISAR measurement . . . . . . 191
5-12 Shifted 1D DFT of the weighted complex signal . . . . . . . . . . . . 192
5-13 Poynting vector of the radar . . . . . . . . . . . . . . . . . . . . . . . 192
5-14 Projected image of the transformed far-field ISAR measurement . . . 193
5-15 Full-bandwidth far-field ISAR measurements with an azimuth vector
of (14.8 , 15 , 15.2 ) . . . . . . . . . . . . . . . . . . . . . . . . 194
15

of (14.8 , 15 , 15.2 ) . . . . . . . . . . . . . . . . . . . . . . . . 195
of (14.6 , 14.8 , 15 , 15.2 , 15.4 ) . . . . . . . . . . . . . . . 196
of (14.4 , 14.6 , 14.8 , 15 , 15.2 , 15.4 , 15.6 ) . . . . . . 197
5-19 Full-bandwidth half-aperture backprojection image i [150 : 1 :
1] (i = 0.2 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
5-20 Full-bandwidth full-aperture backprojection image i [150 : 1 :
150] (i = 0.2 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5-21 Sub-bandwidth full-aperture backprojection image [8, 11] (GHz) . . 198
5-22 Sub-bandwidth full-aperture backprojection image [8, 10] (GHz) . . 199
5-23 Sub-bandwidth full-aperture backprojection image [8, 9] (GHz) . . . 199
6-1 Dielectric dispersion of several types of polarization (Modified after
Knight and Nur (1987) [135]) . . . . . . . . . . . . . . . . . . . . . . 213
6-2 A two-dimensional free-space transmission model . . . . . . . . . . . 220
6-3 Overview of the proposed methodology . . . . . . . . . . . . . . . . . 221
6-4 Transmitting horn and the network analyzer for the free-space measurement [Courtesy of the MIT Lincoln Laboratory (MIT LL)] . . . . 223
6-5 Transmitting and receiving horns for the free-space measurement [Courtesy of the MIT LL] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
6-6 Error surfaces of the dielectric measurements collected from Teflon,
Lexan, Bakelite, and Portland cement concrete . . . . . . . . . . . . . 225
6-7 Error surfaces of the dielectric measurements collected from Teflon,
Lexan, Bakelite, and Portland cement concrete . . . . . . . . . . . . . 226
6-8 Convergence of estimates of 0r at different frequency bandwidths using
TDOA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
6-9 Normalized phase velocity vs. loss factor 00r . . . . . . . . . . . . . . . 233
6-10 Behavior of the Cole-Cole model (s = 2, = 1) . . . . . . . . . . . 240
16
6-11 Behavior of the Davidson-Cole model (s = 2, = 1) . . . . . . . . . 241

6-12 The Cole-Cole diagram of Cole-Coles model . . . . . . . . . . . . . . 243
6-13 The Cole-Cole diagram of Davidson-Coles model . . . . . . . . . . . 244
6-14 The Cole-Cole diagram and the equivalent circuit of Frohlichs model
244
6-15 The Cole-Cole diagram and the equivalent circuit of Cole-Coles model 247
6-16 The Cole-Cole diagram and the equivalent circuit of Davidson-Coles
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
6-17 Performance of the Cole-Cole model for free water . . . . . . . . . . . 253
6-18 Considered sorption model in the pore structure of hcp . . . . . . . . 254
6-19 Jellium distance in the sorption model . . . . . . . . . . . . . . . . . 257
6-20 Used chemical potential accounting for the formation of the bound
water on the surface of hcp . . . . . . . . . . . . . . . . . . . . . . . . 258
6-21 Measurements and model prediction of non-porous and porous specimens of hydration products . . . . . . . . . . . . . . . . . . . . . . . 259
6-22 Calculated bonding potential of free water molecules . . . . . . . . . 260
6-23 Calculated relaxation time distribution over the bound water later . . 261
6-24 Performance of the curve-fitted model for the specific surface area of
hcp by nitrogen adsorption . . . . . . . . . . . . . . . . . . . . . . . . 264
6-25 Performance of the Cole-Cole model for epoxy . . . . . . . . . . . . . 266
6-26 Performance of the Cole-Cole model for E-glass fabric . . . . . . . . . 268
6-27 Performance of the Cole-Cole model for E-glass fabric Real part and
imaginary part . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
6-28 Unidirectional GFRP layer . . . . . . . . . . . . . . . . . . . . . . . . 269
6-29 Performance of six mixing models for GFRP-epoxy . . . . . . . . . . 271
6-30 Relationship between the w/c ratio and product of the w/c ratio and
dielectric constant of oven-dried cement paste specimens . . . . . . . 279
6-31 Curve-fitting results of the oven-dried hcp in the frequency range of 3
GHz to 24 GHz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
7-1 Overview of the FAR NDT . . . . . . . . . . . . . . . . . . . . . . . . 287
17
7-2 Normal and oblique incidence inspection schemes of the FAR NDT
technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
7-3 Defect and rebar detection . . . . . . . . . . . . . . . . . . . . . . . . 289
7-4 Specular dominant circumstance . . . . . . . . . . . . . . . . . . . . . 290
7-5 Specular recessive circumstance . . . . . . . . . . . . . . . . . . . . . 290
7-6 Computed far-field distances at various frequencies and two antenna
apertures
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
7-7 Inspection procedure of the FAR NDT technique
. . . . . . . . . . . 292
7-8 Bridge inspection Beam and column . . . . . . . . . . . . . . . . . 293

7-9 Two types of frequency bandwidth integration . . . . . . . . . . . . . 296
7-10 Improvement of image resolutions Progressive image focusing Frequency integration, HH polarization . . . . . . . . . . . . . . . . . . . 297
7-11 Prediction error of cross-range resolution formulae . . . . . . . . . . . 298
7-12 Progressive image focusing Frequency integration using shifting center frequency, HH polarization, = 15 . . . . . . . . . . . . . . . . . 309
7-13 Improvement of image resolutions progressive image focusing Angular
integration, HH polarization . . . . . . . . . . . . . . . . . . . . . . . 310
7-14 Comparison of images of the intact and damaged surfaces of the specimen AD1 at different incident angles (30 10 ) . . . . . . . . . . . 311
7-15 Comparison of images of the intact and damaged surfaces of the specimen AD1 at different incident angles (10 30 ) . . . . . . . . . 312
7-16 Description of the used far-field ISAR measurements and specimens
for damage detection . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
7-17 Maximum amplitudes of the backprojection images of the specimen
AD1 Full bandwidth (8GHz12GHz), HH polarization . . . . . . . 313
7-18 Differential maximum amplitudes of the backprojection images of the
specimen AD1 Full bandwidth (8GHz12GHz), HH polarization . . 314
7-19 An eight-node element for morphological operations . . . . . . . . . . 314
7-20 Backprojection images of the specimen AD1 HH polarization, = 15 315
7-21 Backprojection images of the specimen AD1 VV polarization, = 15 315
18
7-22 Variation of nE with respect to nthv of the intact-side of the specimen

AD1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
7-23 Variation of nE with respect to nthv of the damaged-side of the specimen AD1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
7-24 Backprojection images and their feature-extracted version of the specimen AD1 (HH polarization, full bandwidth, = 15 ) . . . . . . . . 317
7-25 Original nE () curves of the intact and damaged surfaces of the specimen AD1 (HH polarization, full bandwidth) . . . . . . . . . . . . . . 317
7-26 Filtered nE () curves of the intact and damaged surfaces of the specimen AD1 (HH polarization, full bandwidth, L = 3) . . . . . . . . . . 318
A-1 A layer underlain a solid half space . . . . . . . . . . . . . . . . . . . 326
B-1 A two-dimensional two-layer model with infinite boundary (TE waves) 332
B-2 TE and TM waves and incident wave vector . . . . . . . . . . . . . . 334
B-3 A two-dimensional three-layer model (TE waves)
. . . . . . . . . . . 335
B-4 A three-dimensional infinite dielectric cylinder impinged by plane waves337
19
20
List of Tables
2.1
NDT techniques utilizing mechanical waves . . . . . . . . . . . . . . .
38
2.2
Electromagnetic spectrum . . . . . . . . . . . . . . . . . . . . . . . .
39
2.3
NDT Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
2.4
Types of mechanical wave . . . . . . . . . . . . . . . . . . . . . . . .
45
2.5
Microwave and radar NDT techniques . . . . . . . . . . . . . . . . . .
57
2.6
Microwave and radar NDT applications in civil engineering . . . . . .
59
4.1
Designed specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.2
Used materials and their suppliers . . . . . . . . . . . . . . . . . . . . 117
4.3
Manufacturing standards . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.4
Statistical parameters of the frequency-angle amplitude measurement

of the Aluminum tube (PEC) . . . . . . . . . . . . . . . . . . . . . . 123
4.5
Statistical parameters of the frequency-angle response of the plexiglass

rod specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.6
Peak RCS of simulated and measured responses of the plexiglass rod

specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.7
Signal contents of the radar measurements of laboratory specimens . 135
4.8
Maximum and minimum powers (dBsm) of specimens CON and GFRP,

X-band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.9
Maximum and minimum RCS (dBsm) of the specimens AD1 . . . . . 141
4.10 Configuration of the specimen 4RE . . . . . . . . . . . . . . . . . . . 148

5.1
Comparison of several SAR modes . . . . . . . . . . . . . . . . . . . . 161

21
6.1
Types of magnetic materials . . . . . . . . . . . . . . . . . . . . . . . 206
6.2
Types of dielectric materials . . . . . . . . . . . . . . . . . . . . . . . 208
6.3
Types of materials defined by conductivity . . . . . . . . . . . . . . . 210
6.4
Civil engineering applications of dielectric properties . . . . . . . . . 216
6.5
Thicknesses of the specimens . . . . . . . . . . . . . . . . . . . . . . . 227
6.6
TDOA measurements and dielectric constants 0r of the specimens . . 229
6.7
Loss factors 00r of the specimens . . . . . . . . . . . . . . . . . . . . . 229
6.8
Estimated 0r by TDOA using different frequency bandwidths (GHz) . 231
6.9
Comparison of internal field approach and external field approach . . 242
6.10 Measured complex permittivity of water at 20 C . . . . . . . . . . . . 252

6.11 Fitted parameters in Debyes and Cole-Coles models of free water . . 253
6.12 Averaged statistical thickness, t (
A), of adsorbed water layer on hcp . 259
6.13 Relaxation time, w (ps), of free water molecules . . . . . . . . . . . . 260
6.14 Specific surface area of hcp [83] . . . . . . . . . . . . . . . . . . . . . 263
6.15 Comparison of two GFRP-epoxy systems . . . . . . . . . . . . . . . . 271
6.16 Performance of six mixing models for a GFRP-epoxy system . . . . . 272
6.17 Volumetric fractions of cementitious composites . . . . . . . . . . . . 277
6.18 Density of cement paste specimens [83] . . . . . . . . . . . . . . . . . 277
6.19 Parameters in the oven-dried hcp model . . . . . . . . . . . . . . . . 280
7.1
Image resolution formulae . . . . . . . . . . . . . . . . . . . . . . . . 297
22
Chapter 1
Introduction and Research
Motivation
All difficult things have their origin in that which is easy, and great things
in that which is small.
DaoDeJing, Lao Tzu ( 600 B.C.)
Deterioration of manmade structures is an inevitable and non-stopping process, no

matter how carefully structures are maintained or reserved. Among all manmade
structures, civil infrastructures such as buildings, bridges, tunnels, dams, pipelines,
roads, airports suffer from the severe attacks from the environment during their design
lifespan which may range from 30 to 50 years (bridges designed for 120 years). In order
to meet the expected lifespan and performance for civil infrastructures, strengthening
and repair of concrete structures has become an important issue for public safety and
for effective infrastructure management. Engineering technologies are developed and
introduced for extending the service life of concrete structures by means of restoring
their design capacity for continuous use and/or upgrading them for possible future
challenges from the environment, and for meeting the demand for increased service
load conditions.
Strengthening and repairing of concrete structures can be conducted either internally or externally. Internal strengthening techniques such as injection techniques use
23
adhesive materials to fill in interconnecting cracks and voids in concrete. In this type
of techniques, targeted infiltration/penetration depth of injected adhesive may not
be easy to achieve, and the effectiveness of construction sometimes remains questionable. External strengthening techniques, on the other hand, are more effective than
internal techniques and can achieve a significant level of strengthening, especially for
column-type structures. At present time, the use of fiber reinforced polymer (FRP)
composites as an externally bonded element to confine the concrete in order to secure
the integrity of concrete structures has been proven, both theoretically and practically,
to be an effective strengthening/repair approach. FRP composite jacketing systems
have emerged as an alternative to traditional construction, strengthening, and repair of reinforced concrete columns and bridge piers. A large number of projects,
both public and private, have used this technology and escalating deployment is expected, especially in seismically active regions. Integration of the new FRP composite
with the existing concrete substrate results in the formation of a new structural system. Differences in the material properties of the two structural components (FRP
and concrete) pose challenging problems of predicting the behavior of the integrated
structural system. Extensive research effort has been devoted to this active field
as reported in the literature on structural engineering, and composite materials and
construction.
Compared with the traditional materials such as steel, general advantages offered
by FRP composites include high strength-to-mass ratio, high stiffness-to-mass ratio,
ease in handling, and resistance to corrosion. Ample research activities and applications of FRP strengthening in civil engineering have been reported in research papers
[44] and reports [193, 49] in conjunction with developed design codes [53, 87, 198]
and industrial manuals [253, 55, 152].
Typical FRP composites used in civil engineering applications are carbon FRP
(CFRP), glass FRP (GFRP) and aramid FRP (AFRP). Among these strengthening
materials, GFRP composites have been widely adopted since the 1990s [228, 202,
145, 225], particularly in retrofitting of reinforced concrete (R/C or RC) members,
such as slabs [145], beams [45], walls [132] and columns [258] [160] to increase/restore
24
their mechanical capabilities (flexural, shear and compressive). GFRP composites

can also be applied to masonry [250], metal [56] and wood structures [62]. Figure 1-1
demonstrates several strengthening/repair scenarios on concrete structures. In this
dissertation, GFRP strengthened and retrofitted concrete columns are considered and
modeled by GFRP-retrofitted concrete cylinder specimens. Figure 1-2 and Figure 1-3
show the schematics of intact and damaged GFRP-concrete column systems.
FRP sheet/layer
for axial &

flexure
for shear
for flexure
(a) Beam strengthening
(b) Column strengthening
for flexure
for shear
(c) Slab strengthening
(d) Wall strengthening
Figure 1-1: Several FRP strengthening/repair scenarios of concrete structures

Prior to the strengthening/repair construction, it is important to know
(1) the location(s) for strengthening and (2) the level of strengthening at a specific
location in order to properly and effectively strengthen the structure. Therefore, appropriate and field applicable nondestructive testing (NDT) or evaluation techniques
need to be introduced to assess the current condition (level of damage) of the deteriorated/damaged structure.
25
FRP-epoxy
Epoxy
Steel
reinforcement
Concrete
(a) FRP-wrapped concrete
column
(b) FRP-wrapped reinforced

concrete column
Figure 1-2: Intact multi-layer cementitious composite systems

FRP-epoxy
Epoxy
Steel
reinforcement
Concrete
Concrete cracking
Delamination
(a) FRP-retrofitted concrete
column
(b) FRP-retrofitted reinforced

concrete column
Figure 1-3: Damaged multi-layer cementitious composite systems

After the strengthening/repair construction, the integrated concrete structure with the externally bonded GFRP composite forms a multi-layer composite system. Construction defects and structural/environmental damages may occur within
the GFRP-retrofitted concrete structures, and especially, in the vicinity of FRPconcrete interfaces. construction defects such as air voids/pockets being trapped
between the GFRP wrap and the concrete substrate may be encountered. The presence of these air voids creates a region at which the shear stresses are discontinuous.
The stress discontinuity will further encourage the formation and development of FRP
debonding at and in the vicinity of the interface of FRP and concrete under associ26
ated loading conditions. Therefore, in practice, the defected areas must be identified
and repaired. Figure 1-4 shows the detected air voids and their repair in a bridge
rehabilitation project (the Jamestown Bridge, Rhode Island).
(a) Found air void defect with the

removal of GFRP sheet
(b) Air void found near the corner
(c) Two found air voids
(d) Repaired air void defect
Figure 1-4: Detected air voids at various locations in a bridge box-girder wall and
their repair
FRP-concrete interface and concrete conditions cannot be fully revealed until
physical removal of the FRP composite layer unless the member has already been
subjected to apparent substantive damage. Partial or complete removal of the FRP
composite layer for observation of the damage may pose a danger of structural collapse. A FRP-retrofitted concrete beam or column could appear safe without showing any sign of substantial damage underneath the FRP composite while containing
a severely deteriorated region. Such scenario could happen when the structure has
27
undergone a modest seismic event that has significantly damaged the FRP-concrete
system while the system has not reached the failure stage. Failures of damaged FRPconcrete systems are often brittle, involving delamination of the FRP, debonding of
concrete layers, and shear collapse, which can occur at load levels lower than the
predicted theoretical strength of the retrofit system. Research results from largescale retrofitted RC (reinforced concrete) beam tests [260, 34, 196] have also shown
that failures of these systems may take place through various possible mechanisms,
depending on the concrete grade, rebar provision, and properties of FRP. Identified
failure modes include: (1) concrete crushing before steel yielding; (2) steel yielding followed by concrete crushing; (3) steel yielding followed by FRP rupture; (4)
shear failure; (5) concrete cover delamination; and (6) debonding in the vicinity of
the FRP/epoxy/concrete bond interface. Brittle debonding has been particularly observed [227, 34, 9]. Similar failure modes can occur to FRP-strengthened RC columns.
Figure 1-5 illustrates several failure modes in two GFRP-concrete specimens and one
GFRP-RC column. Gradual debonding of the FRP composite (structural damage)
(a) Global shear cracking in

a CFRP-wrapped
concrete specimen
(Au and Buyukozturk, 2005)
(b) Local concrete crumbling in a GFRPwrapped concrete specimen

(Au and Buyukozturk, 2005)
(c) GFRP rupture and

concrete crumbling in
a GFRP-wrapped
concrete column
(Sheikh and Yau, 2002)
Figure 1-5: Several failure modes in GFRP-concrete systems

under service load conditions may result in premature failures of the retrofitted system, leading to the total collapse of the structure. Thus, there is a need for inspection
28
of debonding in such multi-layer systems using NDT techniques in field conditions.

This dissertation deals with the development of a nondestructive testing/evaluation
(NDT/E) technique for concrete columns/brideg piers retrofitted by externally wrapped
glass fiber reinforced polymer (GFRP) composites using far-field radar measurements.
In this development, research components including analytical approach, experimental measurements, numerical simulation, and image reconstruction are described in
detail in the following chapters. Research objectives, approach and the organization
of the dissertation are given in the following.
1.1
Research Objective
The objective of this research is to develop a microwave-based NDT/E capability

for the distant assessment of the physical condition of GFRP-retrofitted concrete
structures with emphasis on the detection of anomalies and delaminations in the
GFRP-concrete interface region and in the concrete cover areas. The research aims at
the development of new knowledge on the interaction of GFRP-concrete systems with
electromagnetic waves and an image reconstruction capability for physical imagery. A
knowledge-based interpretation algorithm for an effective NDT/E technique as a basis
for the condition assessment of GFRP-retrofitted concrete structures is developed.
1.2
Research Approach
The research approach to achieve the objective consists of four components:

1. Numerical simulation Transmission and reflection of radar signals can be numerically simulated by the propagation and scattering of EM waves in a digital
environment. Special models of the GFRP-concrete cylinders are constructed
and impinged by plane EM waves using finite difference time domain (FDTD)
methods. The purpose of the numerical simulation is to identify and study major system design parameters (measurement scheme, measurement frequency,
and incident angle) in the development of a distant microwave-based NDT/E
29
technique. Parametric study on the effects of artificial defect and dielectric

properties in the reflected radar signals is conducted.
2. Laboratory experimentation Physical radar measurements were made from
laboratory GFRP-wrapped concrete cylinders manufactured with and without
artificial defects and rebars. Artificial defects including a cube, a thin-plate,
and a strip (made of Styrofoam whose electromagnetic properties are same as
air) were embedded at the interface between the GFRP layer and the concrete core. The selection of regular shapes for artificial defects represents the
simplification to the complex shapes of real defects. Figure 1-6 shows this concept. The purpose of conducting radar measurements on laboratory specimens
is twofold: First, a forward study can be performed using laboratory specimens whose defects are known. Radar measurements made on intact (without
defect) and damaged (with defect) specimens can be evaluated based on their
raw measurements and finally processed by image reconstruction as a basis for
comparison with the real specimens. Secondly, the laboratory configuration
provides a noise-free environment for the reflection radar measurements of the
specimens. The removal of background noise is advantageous for better distinguishing the signal due to the presence of defect at a higher signal-to-noise ratio
(SNR). Such measurements are insightful for damage detection and convenient
to deal with in terms of the need of denoising.
3. Image reconstruction In this research, collected distant radar signals (measurements) were processed by tomographic reconstruction (TR) methods in order
to reconstruct the spatial profile of GFRP-wrapped concrete cylinders for condition assessment. Fast backprojection algorithm was applied for implementing
TR methods and for developing numerical codes. The purpose of signal processing is to establish an effective and efficient transformation to visualize radar
measurements.
4. Modeling of Dielectric Properties For the use of radar signals (electromagnetic
(EM) waves) on probing materials, knowledge is needed regarding the dielectric
30
GFRP-epoxy layer
GFRP-epoxy layer
Styrofoam anomaly
Concrete
Delamination between GFRPepoxy and concrete
Concrete
A detected air void
width
depth
The multi-layer cylinder
structure
Structural cracks
Artificial defects
Figure 1-6: Modeling of construction defects and structural damages using artificial
anomaly
properties of the medium in which EM waves travel. Wave-medium interactions including transmission, reflection, and scattering of EM waves can not be
understood and simulated without the knowledge of the dielectric properties of
materials. For the materials considered in this research, dielectric properties
are modeled in the frequency range of 8 GHz to 18 GHz for water, epoxy resin,
E-glass fabric, GFRP, and oven-dried hydrated cement paste. This modeling
effort is important to an accurate numerical simulation, as well as in the image reconstruction for better locating defects (to be further explained in the
chapters).
1.3
Organization of the Dissertation
This dissertation is organized in the following manner. Figure 1-7 illustrates the
relations among the chapters in this dissertation.
Chapter 2 reviews the current development of NDT techniques for the condition assessment of FRP-retrofitted concrete structures. Different NDT methods are
compared for their current or potential use on GFRP-retrofitted concrete structures.
31
Literature review
Far-field Airborne Radar NDT
Numerical
simulation using
FDTD
Far-field monostatic
ISAR measurements of
GFRP-retrofitted
concrete columns
Laboratory
radar
measurements
Backprojection
processing and image
reconstruction using farfield ISAR
measurements
Image
reconstruction
Dielectric
modeling of
materials
Condition assessment
Figure 1-7: Organization of the dissertation
Chapter 3 illustrates the numerical simulation results of electromagnetic waves

propagation and scattering using finite difference time domain (FDTD) methods.
Principles and implementation guidelines of FDTD methods are also introduced.
Effects of system design parameters including measurement scheme, measurement
frequency, and incident angle are studied.
Chapter 4 reports the radar measurements of GFRP-retrofitted concrete cylinders made in the MIT Lincoln Laboratory. Far-field ISAR (inverse synthetic aperture
radar) measurements of GFRP-concrete cylinders were collected by a horn antenna
operating in monostatic mode. Linearly polarized continuous wave (CW) radar signals in the frequency range of 8 GHz to 18 GHz were used in probing the GFRPconcrete cylinders. Azimuthal (angular) range of 60 degrees were explored. Collected
ISAR measurements were represented in the frequency-angle format.
32
Chapter 5 introduces the data processing approach for image reconstruction.

Backprojection algorithms are used for processing far-field ISAR measurements into
spatial images. Reconstructed rangecross-range images are also provided in this
chapter.
Chapter 6 addresses the dielectric modeling of the materials encountered in
GFRP-concrete systems, including water, epoxy, GFRP, and oven-dried hydrated
cement paste. Dielectric models applicable in the frequency range of 8 GHz to 18
GHz are developed based on the dielectric measurements of the materials reported in
the literature.
Chapter 7 addresses the condition assessment methodologies for inspecting the
near-surface defects/damages in GFRP-retrofitted concrete cylinders using reconstructed images.
Chapter 8 summarizes the research findings and discusses possible research topics
for future work.
Appendix A provides the derivation of the phase velocity of Love waves in a
layer underlain by a half space medium, as the basis for condition assessment of
multi-layered systems using the phase velocity of Love waves.
Appendix B addresses the analytical investigation of several two-dimensional
EM scattering problems in which a multi-layered dielectric medium is impinged by
plane waves.
33
34
Chapter 2
Literature Review
Read not to contradict and confute, not to believe and take for granted,
but to weight and consider.
Essay on Of Studies, Francis Bacon (1561 1626)
Condition assessment of materials and structures in a noninvasive manner is called

nondestructive testing (NDT) or nondestructive evaluation (NDE). While the information regarding the condition of structural systems (e.g., bridge, buildings, dam,
tunnel) designed for public use is not only crucial to the operation of the structure
but also vital to public safety. Therefore, various NDT techniques have been developed for obtaining such information. Reported NDT techniques demonstrate the
research results and findings based on different approaches for assessing the condition of a material/structure system. The purpose of this chapter is to review these
techniques.
In this chapter, general description of NDT is first given. Current NDT techniques
including optical NDT, acoustic NDT, thermal NDT, magnetic/electrical NDT, radiographic NDT, and microwave/radar NDT are reviewed with emphases on (1) physical
principles regarding each NDT technique for understanding the characteristics (advantages and constraints) of each technique, and (2) reported or potential application
of these techniques on GFRP-wrapped concrete systems.
35
2.1
Nondestructive Testing (NDT) Techniques
American Society for Nondestructive Testing (ASNT) defines NDT as

The testing of a specimen that determines its serviceability without damage
that could prevent its intended use.
As indicated in the definition, NDT aims at providing information regarding the
serviceability of a specimen (or material or structure) without damaging the specimen.
Such a non-invasive scheme is favored for pertaining not only to the serviceability
but also to the sustainability of the specimen under investigation. The specimen
inspected by NDT techniques is sometimes termed the material under testing (MUT)
which includes laboratory specimens and actual structures. General description of
NDT is described in the following.
1. Purpose and Effectiveness of NDT Inspection The purpose of NDT
inspection is to understand the MUT via obtaining the information regarding
its material properties through the knowledge of wave-medium or field-medium
interaction. Serviceability of the MUT is based on the interpretation of material properties. Therefore, the effectiveness of NDT inspection in a particular
application theoretically depends on the significance of such interaction and
practically depends on the design of instrumentation. Should there be minor
or no interaction between the chosen wave or field and the target MUT, the selected NDT method is theoretically ineffective. Should the signal-to-noise ratio
(SNR) obtained through certain instrumentation be too small to be detectable,
the selected NDT method is practically ineffective.
2. Procedure of NDT techniques Figure 2-1 illustrates the general inspection procedure of NDT techniques.
3. Classification of NDT Techniques All NDT techniques are based on certain types of physical law for probing and manifesting the characteristics (material properties or geometrical properties) of the MUT. Current NDT techniques
are distinguished by the following features.
36
Selection of NDT methods
Experimental configuration
and installation
Man-made source
(Instrumentation)
Natural or man-made
source for manifestation
Generation of manifesting agent

for probing the MUT
Natural source
Measuring the
response of the MUT
(Manifestation-medium
interaction)
Interpretation of the measurement
(Signal processing)
Condition assessment
Figure 2-1: Procedure of NDT Techniques

(a) Underlying physical principle The physical principle behind each
NDT technique is the relationship between the transmitted/produced waves
(dynamically) or fields (statically) and the material properties of the MUT.
The essential concept is to interpret the change in the MUT through the
change in the received waves/fields.
(b) Type of wave or field Selected waves or fields for probing the MUT
can be mechanical, thermal, electric, magnetic, or electromagnetic. These
waves/fields can be produced by man-made devices (e.g., electronic transducer and radar antenna) or by natural sources (e.g., thermal radiation
and radiative decay) (Figure 2-1). It is obviously that one NDT technique
utilizing mechanical waves is different from another technique utilizing
electromagnetic waves. Table 2.1 shows several NDT methods utilizing
mechanical waves. Table 2.2 lists the content of the electromagnetic spec37
trum.
Table 2.1: NDT techniques utilizing mechanical waves
Technique
Frequency (Hz)
Impact-echo and pulse-echo
5 104 5 105
Acoustic emission (AE)
105 106
Ultrasound
Pulse velocity
2 104 2.5 105
Spectral analysis of surface waves
108
(c) Measurement device Information resulting from wave-medium interactions is the response of a MUT and can be collected by electronic devices.
Typically, an electronic device capable of manifesting a MUT (producing
the manifesting agent) are equally capable of collecting the response of
a MUT (reciprocal theorem). However, recent advances on novel measurement technology encourages the development of hybrid NDT whose
manifesting agent is produced by one device and response is measured by
another. Therefore, the use of a measurement device other than the one
producing manifesting agent is considered a different NDT technique.
(d) Instrumentation type Information must be collected and transformed
by devices through instrumentation configuration (equipment). Several
instrumentation types such as single-input-single-output and single-inputmultiple-output are possible for different interpretation schemes leading to
the materialization of different NDT techniques.
For generality the classification of NDT techniques in this thesis is made based
on the underlying physical principles behind each NDT technique, rather than
elaborating the details of each NDT technique.
4. Content of NDT techniques While the NDT techniques are rapidly developing and expanding, their physical principles can be generally classified into
38
Table 2.2: Electromagnetic spectrum

Class
Frequency Wavelength,
(Hz)
(m)
Energy (eV)
Gamma
rays
3 1019 1011 1014

3 1022
> 1.24 105
X rays
3 1016 108 1011

3 1019
1.24102 1.24 Dislocations,

105
cipitates
Ultraviolet
(UV)
7.51014 7 107 108

3 1016
3.1 1.24 102
Visible
light
4.31014 (7 4) 107
7.5 1014
1.77 3.1
Infrared
(IR)
3 1011 103 4 107

4.3 1014
1.24 103 1.77 Texture,

residual
stresses,
cracking,
grain size,
inclusions,
fiber
fracture, delaminations, porosity
Microwaves 3 109 103 101

3 1011
Radio freq.
(RF)
3 10 107 101
3 109
39
Type of inhomogeneity
pre-
1.24 103
1.24 105
Texture,
residual
stresses, cracking,
delaminations
1.24 1013
1.24 105
Cracking,
nations
delami-
the following categories. Reported examples are provided under each category.
Optical methods Including visual inspection [13], surface coating
(e.g., photoelastic coating [41] and brittle coating), , fluorescent penetrants
(e.g., dye penetrant [213], volatile liquid, and filtered particle), Moire interferometry [284], optical holography [100], shearography [97], laser speckle
metrology [218], optical heterodyne interferometry [19], borescope [204],
fiber optical sensors [109], and machine vision [110].
Acoustic methods Including impact-echo [231], sonics [144], ultrasonics [59, 247], acoustic emission [182], acousto-ultrasonics [281], electromagnetic acoustic transducer (EMAT) [161, 187], laser-ultrasonics [275],
ultrasonic holography [136], acoustic microscopy [162], acoustography and
vibro-acoustography [230, 189], and sonic signature analysis [112].
Thermal methods Including thermoelectric probe [71], impulse thermography [167], infrared thermography (IRT) [14], ultraviolet fluorescence
(UVF) [165], and emission spectroscopy [92].
Magnetic and electrical methods Including magnetic field [170],
magnetic particle [249], eddy current [222], nuclear magnetic resonance
(NMR) [146], Kirlian photography [172], Barkhausen effect [264], scanning
electron microscopy (SEM) [98], and impedance spectroscopy [128].
Radiographic methods Including X-ray radiography [200, 94], gammaray radiography [101], neutron radiography [116], proton radiography [127],
synchrotron radiation [86], and Compton backscatter [46].
Microwave and radar methods Including microwave radiometry
[273], waveguides, coaxial probes, radar horn antenna [35, 195].
Miscellaneous Including mechanical impedance (hardness testing [61]),
embedded fiber-optic strain gauge [232], chemical, liquid penetrant testing,
replication microscopy, SQUID (Superconducting Quantum Interference
Device) [276], and spark testing [131].
40
Table 2.3 summaries these methods classified by the elements mentioned previously in this section. These methods are briefly described in the following
sections. While some NDT methods are applied in fluid dynamics and aerodynamics, this thesis only concerns solid medium as the MUT. Review of NDT
methods in civil engineering can also be found at [133] and [175].
Table 2.3: NDT Techniques
Technique
Type
wave/field
of
Optical
EM waves in
optical spectrum
Human eye, lasers,

and optical cameras
Acoustic
Mechanical
waves
Electronic
transducers
Mechanical
properties
Thermal
EM waves in
thermal radiation spectrum
Ultraviolet and infrared cameras
Thermal
properties
Magnetic/
electrical
Magnetic field
and electric field
Radiographic X-rays, Gammarays, neutrons

Microwave
and radar
Measurement
device
Material
properties
Magnetic
and electrical
properties
Photographic films
Radiographic
properties
EM waves in Waveguides, coax- Dielectric

microwave special probes, radar properties
trum and radio antennas
frequency range
41
2.1.1
Optical Methods
Principles
Optical methods includes the earliest NDT method (visual inspection) and some modern methods (e.g, laser speckle metrology and machine/computer vision), all relying
on the electromagnetic radiation in the optical spectrum range, which can be detected
by the human eye. The optical spectrum contains visible lights of wavelengths in air
from approximately 400 nm to 700 nm (corresponding to frequencies from 750 THz to
428 THz) as the manifestation agent. Geometrical optics and physical optics can be
applied for analyzing the propagation of visible light, while the former approach assumes zero-wavelength of light and considers no diffraction. However, the ray tracing
technique in geometrical optics still serves as a convenient tool for providing insights
in some problems. In physical optics, the scalar wave equation governs the motion of
light waves in an isotropic medium [31].
2 (z, t) =
1 2 (z, t)
c2 t2
(2.1)
where z is the traveling distance of light waves, t is time, (z, t) is the space-time
function of light waves (or wave function), and c is the speed of light waves in free
space (c= 3 108 m/s). To account for the interference and diffraction phenomena
of light waves during propagation, the Fresnel transform is used for determining the
light intensity distribution on an observation plane (, ) from a source plane (x, y).

Z Z
ieit
1 + cos
g(, ) =
f (x, y)eik rdxdy
r
2
(2.2)
where i is the imaginary number, is the wavelength, is the radian frequency, r is

the observation vector between the source point (x, y) and the observation point (, )
with length r, is the angle between normal vector, and k is the wave vector with
magnitude k (wave number). The Fresnel transform of light intensity distribution in
the source plane f (x, y) is the observed g(, ). Fresnel transform is also the basis for
42
synthetic aperture radar (SAR) processing. Further investigation of optical analysis

can be found at [31] and [265].
The first appearance of qualitative visual inspection technique may be dated long
back in history without a record, while the practice develops and matures with time.
Visual inspection is still by far the predominant technique for the assessment of
transportation infrastructures such as bridges, although its reliability for routine
and in-depth inspections can be doubtful [163]. Results of evaluation using human
eye are qualitative in nature and could be very subjective. The quality of such
evaluation strongly relies on inspectors experience and judgement. As such, machine/computer/digital vision-based evaluation is introduced to eliminate the factor
of subjectiveness and provides quantitative analysis. Signal analysis of recorded optical images using artificial intelligence approach (e.g., fuzzy set theory, expert system
[110], neural network, and genetic algorithm) and computer vision approach provides
alternatives to physical approach.
Optical NDT for GFRP-concrete Systems

At the present time and to the authors knowledge, there is no literature of optical
NDT for GFRP-concrete systems, although field engineers do use visual inspection
for the detection of significant GFRP debonding and delamination in practice. However, visual inspection is inherently not applicable to the condition assessment of
GFRP-concrete systems for the ineffectiveness of optical lights in penetrating GFRP
sheets/layers. In some circumstances, significant GFRP debonding, GFRP delamination, and concrete cracking may incur discoloring on the surface of GFRP, which is
detectable to optical NDT. However, at this stage of failure, severe damage levels are
reached without precaution. Additionally, similar discoloring of FRP could also due
to the on-site variation in the curing of epoxy, causing potential false-alarm detection
for optical NDT. Some optical NDT utilizes computer/mechine vision analysis incorporating artificial intelligence approach offers in-depth information through extensive
simulation. But the approach is not physically sound, and the solution may not be
unique.
43
In summary, constraints of optical NDT methods for GFRP-concrete systems

include: (1) physical limitation of visual lights in penetrating the opaque GFRP
layer, and (2) insufficient credibility of optical signal (e.g., discoloring) for damage
indication. These constraints hinder the use of optical NDT on GFRP-wrapped
concrete systems for condition assessment.
2.1.2
Acoustic Methods
Principles
Acoustic methods are classified as the NDT techniques using mechanical waves as
the manifesting agent to investigate MUTs. The use of mechanical waves, including
body wave and surface wave, leads to the dynamic/vibrational response of MUTs.
The governing equation of all acoustic methods from a microscopic point of view is
the Navier equation of motion. For homogeneous and isotropic materials [171],
2u
( + ) ( u) + u + b = 2
t
2
(2.3)
E
E
and =
are the Lame constants, u = ui xi
(1 + )(1 2)
2(1 + )
is the displacement field, and b is the body force field per unit mass. The traction
where =
boundary condition is needed for solving the observable displacement field at the
boundary.

( u) n
+
ui uj
+
xj
xi

n
=f
(2.4)
where n
is the normal vector at the boundary and f is the prescribed function acting
on the boundary. Displacement fields at the boundary in different conditions (different mechanical properties; , ) are measured in order to retrieve the variation in the
u
2u
mechanical properties of materials. Velocity fields
and acceleration fields 2 can
t
t
be determined either analytically when displacement fields u are obtained by evaluating the Naviers equation, or numerically when displacement fields are measured over
a period of time. Table 2.4 provides the relationship between wave velocity and mechanical properties of materials. It is clear that changes in the mechanical properties
44
(E, , ) of a MUT are reflected in the variation of these wave velocities. In Table
2.4, the direction of particle movement is determined with respect to the direction of
wave propagation. Also, denotes the density of the MUT, E the Youngs modulus,
and the Poissons ratio. Another dispersive wave mode existing between body and
surface waves is the Lamb wave whose wave velocities (phase and group) can only be
numerically evaluated from the Rayleigh-Lamb equations [32].
Table 2.4: Types of mechanical wave
Class
Particle
movement
Wave velocity
Name
Parallel
E(1 )
vp =
(1 + )(1 2)
1/2
[255]
P-wave,
compressional
wave,
dilational wave
Body

wave
Orthogonal
Elliptical
orbit (symmetrical
Surface mode)
wave
Horizontally
E
vs =
2(1 + )
vR
=
1/2
[255]

1/2
0.87 + 1.12
E
1+
2(1 + )
S-wave, shear wave,

distortional wave,
equivolumnial
wave,
rotational
wave
Rayleigh wave
[267]
vL (See Appendix A)
Love wave
polarized
shear mode
Mechanical responses of the MUT need to be collected by electronic devices (transducers) which translate mechanical responses into electrical signals for further analysis. Practical issues such as coupling between transducers and the surface of materials,
multiple propagation paths, frequency-dependent characteristics of materials complicate the interpretation result.
45
Acoustic NDT for GFRP-concrete Systems

In principle, acoustic NDT such as acoustic emission is applicable to brittle materials
like concrete [26]. Since the integration of concrete with GFRP layers also forms a
brittle system, acoustic NDT is theoretically applicable to GFRP-concrete systems.
Mirmiran et al. (1999) [1] applied acoustic emission (AE) technique to GFRPconfined concrete cylinder specimens to study the correlation between AE signals and
the stress state in concrete. AE transducers were mounted on the surface of the specimens by a highly viscous coupling agent in order to ensure the close contact condition.
The used dominant peak frequency of AE transducers was 150 kHz. GFRP-confined
concrete column specimens with different lengths, cross sections, jacket types, and
jacket thickness were manufactured and subjected to cyclic compressive loadings.
The Felicity and Kaiser effects of AE signals were also discussed. They found that
the AE activity can be correlated to the extent of damage within the specimen. Higher
AE activities were observed on specimens with longer dimensions and thicker jackets. Although the frequency content of AE signals is a function of AE transducers
frequency response, they considered spectral analysis ineffective for evaluating the
condition of GFRP-concrete columns.
Kundu et al. (1999) [247] studied the scanned ultrasonic images of concrete plate
specimens attached by GFRP and CFRP composites. The GFRP-concrete system
was formed by gluing a GFRP sheet with epoxy and mounting it onto a concrete plate.
A circular delamination region of about 50 mm diameter was introduced between
the GFRP layer and the concrete substrate for inspection. The ultrasonic signals
were generated with frequencies from 200 kHz to 800 kHz. Both monostatic and
bistatic configurations of ultrasonic transducers were used in the applied two scanning
modes; Lamb wave scanning (L-scan) and longitudinal wave scanning (pulse-echo
and C-scan). It was found that the longitudinal scanning mode is not effective for
distinguishing intact regions from damaged ones, while the Lamb wave scanning mode
showed the damaged region (with delamination) as a bright spot in the produced
image. They further concluded that the insensitivity of Lamb wave scanning mode to
46
the small variations in the epoxy and concrete properties has made the Lamb wave
scanning mode superior for GFRP-concrete systems.
Mirmiran and Wei (2001) [183] used ultrasonic pulse velocity (UPV) to investigate
the extent and progression of damage in concrete cylinders with and without Eglass FRP tube, also used as a tool for damage indication. The GFRP-concrete
system was formed by filling concrete into GFRP tubes. Cyclic compressive loadings
were applied on the specimens whose mid-section was attached by two ultrasonic
sensors on two opposite sides of the mid-section. They found that the UPV in GFRPconcrete systems were sensitive not only at lower stress levels but also after concrete
had significantly cracked, when compared with plain concrete specimens. Unlike the
generally increasing trend of the AE signal in a GFRP-concrete system, the UPV
signal exhibited a fluctuating pattern at different loading stages. They considered
such difference as an opportunity to complement UPV with AE for the condition
assessment of GFRP-concrete systems.
Bastianini et al. (2001) [80] used ultrasonic pulse amplitude (UPA) to locate
the debonding defects in a polyurethane slab reinforced on both sides with GFRP, a
concrete cylinder specimen wrapped with CFRP composites, and masonry columns
wrapped with CFRP. An ultrasonic transducer was required in close contact condition
with the surface of MUTs for effective measurements of UPA. They found that the
use of UPA is rather independent from the MUT and from the defect nature.
In summary, the features of acoustic NDT methods for GFRP-concrete systems
include: (1) mechanical waves can penetrate through the GFRP layer in GFRPconcrete systems. Surface waves such as Lamb waves are found effective on detecting
the presence of unseen delamination, due to the change of GFRP thickness [247];
(2) properties of ultrasound waves traveling within GFRP-concrete systems can be
used for locating defects within the systems [183, 80]; (3) the use of coupling agent is
required to assure the contact condition between transducers and the surface of MUTs,
suggesting that acoustic NDT is essentially a contact technique; (4) limited to the
size of acoustic transducers, inspection must be conducted on a point-by-point basis;
and (5) interpretation of results from reflection measurement is difficult, especially
47
for heterogeneous materials like concrete. These constraints pose difficulties for the
field application of acoustic NDT on GFRP-wrapped concrete systems for condition
assessment.
2.1.3
Thermal Methods
Principles
Thermal methods measure the thermal radiation (radiometry) or temperature (thermometry) emitted/reflected from the surface of the MUT, as well as from the ambient surroundings. Thermal radiation is the inherent electromagnetic radiation of
materials above absolute zero temperature, which is a small fraction of the entire
electromagnetic spectrum, containing electromagnetic waves ranging from ultraviolet
light (3nm 400 nm) to far infrared radiation (300m 1000m). The radiant
energy flux at a given temperature and at a given wavelength is determined by [122]
Z
e(, T ) =
e (, T )d
(2.5)
where e (, T ) is the distribution function of radiative flux at wavelength (monochromatic emissive power), and T is the Kelvin (absolute) temperature. Since a MUT
emits a unique distribution of energy in wavelength at a given temperature, as the
result of the thermal properties of MUT, such distribution (at a given temperature)
can be used for determining the thermal properties of the MUT and, furthermore,
for distinguishing one material from another. The relationship among e , , and T
can be demonstrated by the Plank law of emission for a black body which yields the
maximum value of e that a material can attain.
eb =
2~c2
h
i
~c
e kB T 1
(2.6)
where eb is the e of a black body, ~ is Plancks constant (~ = 6.626068761034 J s),

c is the speed of light, and kB is Boltzmanns constant (kB = 1.3806503 1023 J/K).
48
Experimentally, the Stefan-Boltzmann law states the strong dependency of the radiant
energy flux on T as
eb (T ) = T 4
(2.7)
where is the Stefan-Boltzmann constant ( = 5.6704 108 W/m2 K 4 )). Since the
radiant energy flux of a MUT is a unique signature at a given temperature and significantly changes with varying temperatures, it is used for characterizing the MUT. As
a NDT method, thermal radiation is practically more applicable for field application
over other heat transfer modes including heat conduction (Fouriers equation) and
heat convection (Newtons law of cooling). The evaluation of heat transfer process is
governed by the heat diffusion equation. The heat diffusion equation for incompressible medium without convection is [122]
kT +
q
T
= c
t
t
(2.8)
where k is the thermal conductivity ( watts

), T is the Kelvin temperature (K), q is the
mK
kg
), t is time (sec), is the density ( m
heat flux ( watts
3 ), c is the specific heat capacity
m2
kJ
( kgK
) [122]. Techniques for solving inverse heat transfer problems can be found in
textbooks such as [201]. The applicability of this technique is based on the availability
of heat (cold) source, thermal gradient, and induced heating [257]. Anomalies such
as air voids between two material layers will obstruct the thermal radiation from the
bottom layer and alter the thermal image.
Thermal NDT for GFRP-concrete Systems
Thermal NDT has been a popular inspection technique for FRP-concrete systems.
Many of the applications are CFRP-wrapped systems [210, 166, 197] and some GFRPwrapped systems [262].
Halabe et al. (2002) [262] used IR cameras on several laboratory specimens and
composite bridge decks for delamination detection. The GFRP-concrete system in
their study was a concrete bridge column wrapped with GFRP. Their laboratory tests
suggested that delaminations in a composite deck with a depth of the order 0.5 in (1.3
49
cm) and thickness of the order of 0.05 in (0.13 cm) could be detected. However, issues
affecting the thermal contrast between delaminations and the background, including
surface defects and oil spills, shadow from adjacent objects, and inspection angle,
could interfere with the results.
Brown and Hamilton III (2003) [33] applied IR cameras and heating lamps on
concrete block specimens mounted with GFRP/vinylester, CFRP/epoxy, and CFRP/
polyurethane composites for detecting the delamination within the FRP-concrete
systems. The heating lamps for pre-warming purpose were placed 12 in from the
surface of the specimen, and the IR camera was located 24 in from the specimen. A
certain amount of time (30 sec or 60 sec) was needed for recording the cooling process
on the surface of the specimen. They found that defects residing under single-layer
CFRP are well detected, however, the IR detectability decreases when the thickness
of the FRP increases. They also found that GFRP layers with a thickness of 0.3 in
0.4 in are difficult to detect in their experiments.
In summary, the features of thermal NDT methods for GFRP-concrete systems include: (1) the presence of unseen delamination within GFRP layers or in the interface
between GFRP and concrete may not be detectable from the surface measurement by
thermal sensors (e.g., IR camera) as reported by some researchers [33]; (2) thermal
NDT relies on the thermal contrast between damaged and intact regions, triggered
by either natural (sun) or man-made (heating blanket or cooling liquid) sources. Although thermal NDT is basically a distant inspection technique, surface preparation
may require direct access to the surface of MUTs; and (3) variations in the ambient thermal condition and surface emissivity of materials affect the resolution of the
imagery and the interpretation. Controlling ambient temperature could be a troublesome and difficult task in some practical situations.
50
2.1.4
Radiographic Methods
Principles
Radiographic methods are the technique of obtaining a shadow image of a MUT using penetrating radiation (radiation particles) including X-rays (or X-ray photons),
gamma-rays (or gamma-ray photons), neutrons, positrons, protons, or electrons as
manifesting agent. Among these radiation particles, gamma-ray photons are emitted
by the disintegration of a radioactive isotope (radioactive decay), while other manmade particles are generated by accelerator or circuit, radioactive, reactor neutron
sources. The images produced by radiographic methods are known as radiographs,
and the contrast in a radiograph indicates the differential absorption of penetrating radiation in the MUT. In this section, characteristics of penetrating radiation
are illuminated using X-rays as an example, since it is widely used in medical and
engineering applications. Information regarding the use of penetrating radiation in
NDT applications can be found at [25] and [43]. A typical X-ray instrumentation
scheme consists of (1) X-ray tube (to generate X-rays), (2) collimators (to restrict
X-rays within desired area), and (3) image receptor (to capture transmitted X-rays).
Further information about X-rays can be found in [54].
Generation and Features of X-rays

X-rays are generated when an incident electron beam impinges on a solid matter.
The electrons of the matter are ejected (photoelectric absorption), scattered (coherent scattering), recoiled (Compton scattering), released (pair and triplet production),
and absorbed (photointegration) by the incident electron beam, depending on the energy the electron beam carries [42]. X-rays or X-ray photons are produced during the
process of these interactions. Discovered by W.C. Rontgen in 1895, X-rays are characterized by (1) They are invisible, highly penetrating rays traveling in straight lines.
Access to opposite sides of the MUT is required in order to capture the transmitted
rays; (2) They are electrically neutral and immune to electromagnetic fields; (3) They
are polyenergetic rays which can be generated at various energy levels; and (4) They
51
can ionize the MUT through penetration and excitation. Such ionization produces
chemical and biological changes in the MUT, meaning that protecting personnel from
X-ray radiation is vital.
X-ray Interaction with Matter

The absorption and scattering of X-rays in matter are responsible for the attenuation
of X-rays in matter. Quality radiographs are usually obtained with minimum X-ray
scattering. The attenuation (absorption) coefficient of X-rays for a narrow, wellcollimated beam is defined by
dI
= dz
I
(2.9)
where dI/I is the relative decrease in intensity, is the linear attenuation (or absorption) coefficient, and dz is the unit traveling distance of X-rays. Integrating Eq. 2.9
provides Lamberts law (also known as Beers law) given by
Iz = I0 ez
(2.10)
where Iz is the intensity at a distance z and I0 is the intensity at the surface of the
MUT. For a broad diverging X-ray beam, the linear absorption coefficient is defined
by
= T h + ph + j + +
(2.11)
where T h is the absorption coefficient accounting for the Thomson scattering, ph for
the photoelectric effect, j for Compton scattering (absorption and scattering), for
pair and triplet production, and for photodisintegration [103]. These coefficients
represent the radiographic properties of the MUT.
X-ray diffraction is one of the two phenomena (with x-ray fluorescence) associated with the onset of x-radiation. It is the elastic scattering of x-ray photons by
atoms in a periodic lattice structure of materials. The phase information of the scattered monochromatic x-rays provides meaningful information about the structure of
materials and can be further used to identify the material.
52
Radiographic NDT for GFRP-concrete Systems

At present time, only a limited amount of research has been devoted to the use of
radiographic NDT for concrete (cementitious) materials. Masad et. al. (2002) [72]
applied x-ray diffraction on identifying the distribution (sizes and locations) of air
voids in asphalt mix specimens. Daigle et. al. (2005) [159] compared the performance of x-ray diffraction and ultrasound on concrete specimens based on several
factors including cost and image resolution. Their research results suggest that the
cost, safety concern, and experiment preparation of x-ray diffraction may limit its
applicability in field, although it provides high resolution images.
To the authors knowledge, there is no literature of radiographic NDT application
for GFRP-concrete systems. The reason is simply because that radiographic NDT is
essentially a laboratory method, and GFRP-concrete systems are situated in an open
environment, making the application of the method in field and the protection of
inspectors from excessive radiographic exposure (contamination) extremely difficult.
In addition, the requirement to access on both sides of the MUT and the safety
concern make radiographic NDT a near-contact technique, although radiographic
rays are theoretically suitable for distant inspection. It is the operational and safety
issues that limit the use of radiographic NDT for GFRP-concrete systems. The future
application of radiographic NDT for GFRP-concrete systems may be conducted in
laboratory, but it will not be possible to make radiographic NDT a field technique
unless the previously addressed concerns can be resolved.
2.1.5
Magnetic and Electrical Methods
Principles Magnetic Methods

Magnetic methods are described by the magnetic particle inspection (MPI) in this
thesis as an example. In MPI methods, the constitutive relation connecting magnetic
flux density B(webers/m2 ), magnetic permeability , and magnetic field strength H
53
(amperes/m) for homogeneous, isotropic, non-frequency dispersive media is read [138]

B = H
(2.12)
In an imperfect MUT with a crack, a discontinuity is created between two surfaces

of a crack, resulting in a distorted magnetic field termed the magnetic leakage field.
The constitutive relation becomes
B = 0 H + M
(2.13)
where magnetization M (amperes/m) accounts for the contribution of the magnetic

leakage and is used for detecting defects. For ferromagnetic materials, M H.
Therefore,
B
= 0 M
(2.14)
Such variation in magnetic fields can be related to the surface current density k M
within an area A which is defined by
P
k M = lim
A0
qi v i
A
(2.15)
where qi is the moving charge on the surface, and v i is the velocity vector of qi . Considering a Gaussian surface of negligible thickness with area A across the boundary
of two regions (i and j) consisting the crack, Gausss law provides
Z
k M A =
j M dV
(2.16)
where j M = M is the equivalent current distribution, and dV is the unit volume.

Also,
Z
k M A =

M dV = n
M j M i A
where n
is the normal vector pointing from region i to region j. It is seen that the
surface current accounts for the total difference in magnetization between two regions
54
creating a crack if k M is measured.

MPI is performed by (1) cleaning the surface of the MUT in order to allow magnetic particles capture the geometry of the MUT after magnetization, (2) magnetizing
the MUT, (3) spreading magnetic particles over the surface of the MUT, (4)illuminating the surface for inspection based on the distribution of magnetic fields indicated
by sprayed magnetic particles, and (5) demagnetizing the MUT [43].
Magnetic NDT for GFRP-concrete Systems

In order to successfully detect defects by MPI, the surface of the MUT must be cleaned
and the MUT magnetized. Since neither GFRP-epoxy composites nor concrete can
be magnetized, magnetic NDT methods are inapplicable to the condition assessment
of GFRP-concrete systems. Although steel reinforcements are magnetizable, they
are buried under concrete cover, and their cracking is not concerned in this research.
Additionally, magnetic NDT methods provide only the surface information of the
MUT, thus, not relevant for the purpose of in-depth investigation.
Principles Electrical Methods

Electric methods are characterized by eddy current testing (ECT) which in this thesis.
This section introduces the physical principles of ECT as an example for electrical
NDT methods.
When a coil carrying alternating currents approaches the surface of a MUT, an
eddy current flowing in a closed loop is formed on the surface of the MUT, causing
a back electromotive force (EMF), f , opposing the in the coil. The magnitude of f
is changed when defects are presented in the MUT, and is used for indicating the
presence of defects. Faradays law provides [138]
I
f=
C
v
E+ B
c

dl
(2.17)
where C is the closed loop containing the considered magnetic flux density B, E is the
electric field strength (volts/m) flowing through the closed loop C, v is the velocity
55
of a given point in the middle of an unit length dl on the loop, and c is the speed of
light. The first part in the R.H.S. (right hand side) of the above equation results is
the result of the time-varying vector potential, and the second part is the result of the
closed loops motion. The second part diminishes if the loop is stationary. With the
constitutive relation E = 1 D, where D is the electric displacement (coulombs/m)
and the electrical permittivity, f is related to the variation of the MUTs electrical
properties.
ECT is conducted by the use of inspection devices typically containing an induction coil circuit and an oscilloscope for illuminating the result. Contact inspection
of ECT is not required, but the depth of penetration is usually restricted. Distance
between the ECT device and the surface of a MUT, the frequency of the alternating
current inside the loop, and design of the coil can be changed so as to achieve desired
results [102].
Electrical NDT for GFRP-concrete Systems

As an example of electrical NDT methods, ECT is only applicable to electrically
conductive materials. Variation in the electrical properties of GFRP-concrete systems
due to the presence of defects (GFRP delamination, GFRP debonding, or concrete
cracking) is not significantly detectable since GFRP and concrete are not highly
conductive materials. The contrast between intact region and cracked region in a
GFRP-concrete system due to the inclusion of air gaps can be amplified if the presence
of water within the crack is found. Although the near-surface condition of GFRPconcrete systems can be evaluated by ECT, the restricted lift-off distance classifies
ECT as a near-contact method which is not feasible for distant inspection in practice.
2.1.6
Microwave and Radar Methods
Principles
Microwave and radar (Radio Detection and Ranging) NDT methods are based on
Maxwells equations which govern the spatial and temporal relationships between
56
electric field and magnetic field. The coupling between electric field and magnetic
field is the kernel to the propagation of electromagnetic (EM) fields which are used
as the manifesting agent in microwave and radar NDT methods. Although the terms
microwave NDT [8] and radar NDT [4] are both used by various researchers,
differences can be distinguished between them as listed in Table 2.5. By this definition,
radar NDT also includes those used in geophysical and geotechnical applications
(subsurface radar [69], georadar, ground penetrating radar or GPR), although those
applications are not the focus of this thesis. The formulation of EM fields in microwave
Table 2.5: Microwave and radar NDT techniques

Microwave
Radar
Spectrum
0.3 GHz 300 GHz
0.003 GHz110 GHz
Wavelength
1 m 102 m
3102 m 2.72 102 m
Measurement
device
Waveguides, coaxial probes,

cavity resonators
Bow-tie antennas, horn antennas
and radar NDT applications is typically carried out by excluding the source (EM field
radiator or radar antenna) within the domain of interest, unless the radiation pattern
of the antenna is an issue. For this reason, the source-free Maxwells equations are
used to derive the EM wave propagation equation as follows [138].
B
t
D
H =
t
B =0
E =
D =0
(2.18)
(2.19)
(2.20)
(2.21)
Rearranging the above with the constitution relationships provided in the Section of
Magnetic and Electrical Methods and using the vector identity ( X) =
57
(X)2 X where X is a vector field, two Helmholtz wave equations are obtained.
2E
=0
t2
2H
2 H 2 = 0
t
2 E
(2.22)
(2.23)
The use of microwave and radar NDT covers a wide variety of applications in
civil engineering, which is summarized in the Table 2.6. As can be seen in Table 2.6,
EM signals used for probing underground and underwater objects are generally of
frequencies lower than 1 GHz 2 GHz for the sake of better penetration into lossy
dielectric media such as wet soils, while higher frequencies are usually adopted for
detecting anomalies (air voids, rebars, debonding, delamination) to achieve better
resolution. It should be noted that Table 2.6 only demonstrates the explored frequency ranges in various applications. Frequencies outside the range listed in Table
2.6 are not excluded from their potential use in each application.
Microwave and Radar NDT for GFRP-concrete Systems
Several radar NDT techniques for assessing the condition of FRP-retrofitted/wrapped
concrete structures are reported in the literature.
Li and Liu (2001)[149] applied a bistatic radar NDT system to detect air voids in
the interface region between GFRP-epoxy layer/jacket and concrete surfaces. Radar
measurements were conducted at 10 GHz using a pair of dielectric lenses for signal
focusing. The focusing distance between the dielectric lenses and the specimen was
10 in. (25 cm) at 10 GHz. The dielectric focus lenses enabled revealing of localized
response from the GFRP-concrete specimen, with a spatial resolution as small as 4 in.
(10.16 cm) in diameter. Artificial defects of size 0.07 in2 (0.43 cm2 ) were detected at
10 GHz. The defined imagery was constructed by assembling the reflection coefficient
at each spatial point, and used for condition assessment.
Feng et al. (2002)[185] used a horn antenna and a waveguide reflectometer for
detecting air voids in the interface region of GFRP-wrapped concrete specimens.
Dielectric lenses were also introduced to mechanically focus plane waves on a local58
Table 2.6: Microwave and radar NDT applications in civil engineering

Application
Frequency range
References
Material characterization
0.1 MHz 18 GHz
[5] [168] [29] [176] [2]

[248] [73] [29] [141]
[30] [99] [237] [3] [8]
Bridge pier scour detection
0.1 GHz 1 GHz
[192] [181] [88]
Bridge deck assessment
0.05 kHz 16 GHz
[64] [60] [17] [164]
Void and crack detection in 0.5 GHz 1.5 GHz

stone and masonry structures
[39] [27] [169] [38]
Damage detection in sluices
0.5 GHz 0.9 GHz
[168]
Pavement thickness detection
10 MHz 2.5 GHz
[246] [173] [229] [40]

[6]
Underground object detection
0 Hz 0.6 GHz
[199] [93]
Tunnel linings
0 Hz 1 kHz
[4]
Channel walls
0.2 GHz 0.9 GHz
[216]
Railway tracks
0.5 GHz 1.5GHz
[37] [259] [158]
Fatigue cracks detection in 24.13 GHz 33.6 [211]

steel structures
GHz
Debonding in composites
8.5 GHz 40 GHz
[283] [190] [149] [11]

[48]
Concrete cover thickness detection
1.5 GHz
[134]
Concrete and cement cracking
1.5 GHz 10 GHz
[35] [157] [217] [167]

[191]
Post-tensioned concrete beam
1.5 GHz
[274]
Rebar detection
1 GHz 12 GHz
[217] [226] [233]
Structural testing and remote

sensing
5.75 GHz 10 GHz
[57] [234] [205]
59
ized spot in order to enhance the strength of reflected signals. The focusing distance
between the dielectric lenses and the specimen was 2.5 in. (6.4 cm) in X-band. Reflection coefficients were used for evaluating the difference between intact and damaged
responses. Artificial defects as small as 0.105 cm in concrete were detected at the
frequency 12.4 GHz. The imagery was the assembling of reflection coefficients at
different points on the surface of the specimen. Later, Kim et al. (2004)[279] proposed the use of planar slot antenna arrays for detecting air voids in concrete panel
and block specimens. The slot antenna consisted of two arrays: transmitting array
and receiving array, operating in 5.2 GHz with a resolution of 1 in. (2.5 cm) and
penetration depth of 10 in. (25 cm). The focusing distance between the slot antenna
and the specimen was 1.2 in. (3 cm). The imagery for damage detection was generated by integrating reflection responses of the receiving slots to obtain an equivalent
electric current distribution within the target. This distribution was decomposed by
the Hankel function of the second kind in the case of cylindrical waveguide problems.
The coefficients operating with the Hankel function were considered as the focusing
operator. This development was based on the use of cylindrical slot antenna arrays
applicable for circular column structures.
Akuthota et al. (2004) [11] utilized an open-ended rectangular waveguide probe
and near-field radar measurements for detecting disbonds and delaminations between
CFRP (carbon FRP) laminate and the concrete substrate. Transmitted EM waves
were linearly polarized with orientation orthogonal to the unidirectional carbon fibers
in order to achieve strong penetration through the CFRP laminate. Artificial defects
with sizes in the millimeter range were inserted between the CFRP laminate and the
concrete substrate. Inspection distances were 0.06 in.0.12 in. (0.150.3 cm) at 10
GHz and 0.12 in.0.27 in. (0.30.7 m) at 24 GHz as the authors reported in their
work.
Although EM waves can be used in distant ranges (far-field), most of the work on
this particular problem adopted contact or near-contact (near-field) measurements.
Tradeoffs between near-field and far-field techniques were discussed in the literature
[142, 215]. FRP debonding, delamination, and air voids in the interface regions of
60
FRP-concrete systems were investigated by near-field inspections using different hardware systems including horn antenna, slot antenna arrays, and waveguide probe. High
frequencies ranging from 5 GHz to 24 GHz were used in order to capture small defects.
Reflection coefficients were calculated and used for damage assessment. Visualization
of the results was generally performed by assembling the localized response at each
inspection location to form a spatial profile in the reported techniques.
2.2
Summary
This chapter reviews current nondestructive testing (NDT) methods at large and their
applicability to the particular structural system (GFRP-concrete system) considered
in this thesis. Features of these NDT techniques are summarized as follows.
1. Optical NDT Physical limitation of visual lights on penetrating the opaque
GFRP layer, as well as the insufficient credibility of optical signal for damage
indication, constrain the use of optical NDT for GFRP-concrete systems.
2. Acoustic NDT Acoustic waves (Lamb wave) and ultrasounds are found
effective on detecting the presence of unseen delamination, owning to the change
of GFRP thickness in affecting the propagation of waves. However, acoustic
NDT is basically a near-contact, point-by-point inspection technique which is
not ready for the field inspection of GFRP-concrete systems.
3. Thermal NDT The presence of unseen delamination within GFRP layers
or in the interface between GFRP and concrete may not be detectable from the
surface measurement by thermal cameras as reported. Variations in the ambient
thermal condition and surface emissivity of materials affect the resolution of
the imagery and the interpretation. Controlling ambient temperature could be
difficult in some practical situations. Although thermal NDT is inherently a
distant inspection technique, surface preparation may require direct access to
the surface of the system, making it not very preferable for the field application
of GFRP-concrete systems.
61
4. Radiographic NDT The cost and safety concerns of radiographic NDT

made it rather a laboratory technique but a field technique. These concerns
also unfavor the use of radiographic NDT for the field inspection of GFRPconcrete systems.
5. Magnetic and Electrical NDT The non-magnetic and non-conductive
nature of GFRP layers and concrete makes the use of magnetic and electrical
NDT on GFRP-concrete systems inherently less effective, especially for targeting GFRP debonding and GFRP delamination.
6. Microwave and radar NDT Microwave and radar NDT is in nature a
field applicable method since EM waves attenuate sightly when traveling in
free space and are capable of penetrating into non-metallic, lossy media such
as concrete. Current microwave and radar NDT techniques for GFRP-concrete
systems are either distant methods for surface inspection or near-field methods
for in-depth inspection. While the in-depth inspection capability can be offered
by some near-field methods, the need for mechanical focusing devices (e.g.,
lens, antenna arrays) is not practical for civil engineering application. On the
other hand, surface information provided by distant methods is not critical for
GFRP-concrete systems.
From the results of literature review, it is found that a distant NDT technique that is
capable of conducting in-depth inspection of GFRP-concrete systems in field conditions is currently lacking. Microwave and radar NDT is identified to be applicable
for developing such technique, based on the following reasons: (1) its distant inspection capability when compared with acoustic, magnetic, and electrical NDT, (2) its
capability of penetrating into non-metallic, lossy media such as concrete when compared with optical and thermal NDT, (3) it is less vulnerable to temperature variation
in field conditions when compared with thermal NDT, and (4) it is not limited by
safety issues in field conditions when compared with radiographic NDT. According to
these findings, it is concluded that microwave and radar NDT is potentially applicable
for the distant, in-depth condition assessment of GFRP-concrete systems.
62
Chapter 3
Numerical Simulation
Whatever philosophical standpoint one may adopt today, from every point
of view the erroneousness of the world in which we think we live is the
surest and firmest fact that we can lay eyes on: we find reasons upon reasons for it which would like to lure us to hypothesis concerning a deceptive
principle in the essence of things.
Beyond Good and Evil, Friedrich Nietzsche (18441900)
Understanding the transmission and reflection of radar signals in GFRP-retrofitted

concrete columns is equivalent to studying the propagation and scattering of electromagnetic (EM) waves in a dielectric medium. In such investigation, it is beneficial
to develop the knowledge of radar response from GFRP-concrete columns through
numerical simulation such that a close link between radar response and structural
integrity can be established. On the other hand, complexities of the dielectric properties of concrete can also be studied by the forward modeling in numerical simulation.
The main interest of this simulation work is to investigate the time-domain far-field
response of scattered EM fields from GFRP-concrete systems. Among many numerical simulation techniques, the finite difference time domain (FDTD) method using
Yees algorithm [277] was chosen for its superior performance and capability of generating the space-time response of EM wave propagation and scattering. GprMax, a
63
GNU/GPL code based on the FDTD method, was used in producing the simulation
results [95].
Conceptually, the radar can collect measurements reflected from the GFRP-concrete
column (cylinder) at any angle. Should we define a plane containing the longitudinal
axis of the cylinder, the profile of the cylinder will be rectangular in shape (oblique
incidence); should we define another plane perpendicular to the longitudinal axis of
the cylinder, the profile of the cylinder will be circular (normal incidence). These
two scenarios represent two different measurement schemes. The coinciding use of a
radar for transmitting and receiving signals leads to the operation mode of monostatic
radar. It can be expected that the reflection response of GFRP-concrete cylinders
depends on the selection of measurement scheme, measurement (incident) frequency, and incident angle.
The purpose of this chapter is to investigate the effects of system parameters

(measurement scheme, measurement frequency, and incident angle) in the reflection
response of GFRP-concrete cylinders with artificial features (near-surface defect and
rebars). In the two-dimensional simulation of this work, GFRP-concrete cylinders
are consequently modeled as rectangles and circles. The knowledge developed in
numerical simulation can benefit the design of an efficient and effective experimental
program (Chapter 4 Laboratory Radar Measurements).
In the time-domain far-field region the incident EM waves are approximately planar. The use of linearly polarized EM waves leads to the decoupling of Maxwells curl
equations into transverse electric (TE) and transverse magnetic (TM) wave modes
[138]. In this chapter, the FDTD solutions to both TE and TM waves are provided, as
the basis for the explicit evaluation to Maxwells curl equations and for numerical programming. Computational issues including stability criteria and absorbing boundary
conditions are also addressed. Description to numerical models is provided, followed
by the results of parametric study and their discussion.
64
3.1
Maxwells Curl Equations and Linearly Polarized EM Waves
In linear, isotropic, nondispersive, lossy, source-free problems, Maxwells curl equations are written as [138]

H
1
= E + H
t

E
1
E
=
H
t

(3.1)
(3.2)
= magnetic field strength (amperes/m or A/m), = magnetic permeability

where H
of the medium (henrys/m or H/m), E = electric field strength (volts/m or V/m), =
equivalent magnetic loss (ohms/m or /m), = electric permittivity of the medium
(farads/m or F/m), and = electric conductivity of the medium (1/ohms/m =
siemens/m or S/m). In construction materials which are in general non-magnetically
lossy, contribution of can be neglected. In such problems, the medium in which
electromagnetic waves propagate is fully characterized by its dielectric properties
= H(
r, t) and E = E(
r, t), where r is the three(, , and ). Meanwhile, H
and E)
are concerned,
dimensional spatial vector at the point where the EM fields (H
and t is the one-dimensional temporal variable at the time instant when the EM
fields are evaluated. Note that Maxwells equations can be expressed in integral or
differential form and in different coordinate systems (e.g., Cartesian, cylindrical, and
spherical). In brief, the four-dimensional space-time behavior of EM waves is governed
by Maxwells equations, regardless of their representation.
65
In Cartesian coordinate systems, Eqs.(3.1) and (3.2) are written as

1 Ez Ey
Hx
=
t
y
z

Hy
1 Ex Ez
=
t
z
x

Hz
1 Ey Ex
=
t
x
y

1 Hz Hy
Ex
=
+ Ex
t
y
z

Ey
1 Hx Hz
=
+ Ey
t

z
x

1 Hy Hx
Ez
=
+ Ez
t
x
y
(3.3)
(3.4)
(3.5)
(3.6)
(3.7)
(3.8)
In the far-field region where EM waves are planar, variation of the EM wave components in the direction normal to the plane on which EM waves propagate diminishes.
In other words, the EM wave components in the direction perpendicular to the plane
is defined are constant. Suppose we choose the x-y plane
on which the wave vector (k)
as the wave propagation plane (k = xkx + yky ), we have

= 0. This plane-wave
z
condition eliminates the

terms in Eqs.(3.3), (3.4), (3.6), and (3.7), leading to
z

1 Ez
Hx
=
t
y

1
Hy
Ez
=
x

Ex
1 Hz
=
+ Ex
t
y

Ey
1
Hz
=
+ Ey
t

x
(3.9)
(3.10)
(3.11)
(3.12)
Taking the time derivative of Eqs.(3.19) and (3.22) yields

2 Hz
1 Ey
Ex
=
t2
x t
y t

2
Ez
1 Hy
Hx
Ez
=
t2
x t
y t
t
66
(3.13)
(3.14)
With Eqs.(3.9)(3.12), Eqs.(3.13) and (3.14) become

2

Hz 2 Hz
2 Hz
1
Ey Ex
+
=
+
t2

x2
y 2
x
y
2

Ey Ex
2
2
+
=0
+
2 Hz
x2 y 2
t
x
y

2 Ez
1 2 Ey 2 Ex
Ez
=
t2
x2
y 2
t

2
2
2
Ez = 0
+ 2 2
2
x
y
t
t
(3.15)
(3.16)
It can be shown that two EM wave propagation modes can be defined by properly
grouping the six governing equations (Eqs.(3.9), (3.10), (3.19), (3.11), (3.12), and
(3.22)) into two groups. Eqs.(3.17), (3.18), and (3.13) form the propagation mode of
transverse electric (TEz) waves of waves or TE waves, governing equations are

Ex
1 Hz
=
+ Ex
t
y

1
Hz
Ey
=
+ Ey
t

x

Hz
1 Ey Ex
=
t
x
y
(3.17)
(3.18)
(3.19)
It is clear that TE waves can be fully described by Ex , Ey , and Hz . In radar terminology, transmitted EM waves with an electric field horizontally polarized are called
H-polarized waves. Orientation of the electric field is defined with respect to the
plane formed by the wave vector k and the characteristic axis (e.g., the longitudinal
axis of a cylinder). When the received electric field is also horizontally polarized,
the radar response and its associated images are called HH-polarized. In this case,
TE waves are HH-polarized signals. Similarly, Eqs.(3.20), (3.21), and (3.14) form the
propagation mode of transverse magnetic (TM) waves since it is the magnetic field
The governing equations of TM waves
Hz normal/transverse to the wave vector k.
67
are

1 Ez
Hx
=
t
y

Hy
1
Ez
=
x

Ez
1 Hy Hx
=
+ Ez
t
x
y
(3.20)
(3.21)
(3.22)
TM waves can be fully described by Hx , Hy , and Ez . When TM waves are transmitted

and received in vertical polarization, VV-polarized signals are obtained. In this case,
TM waves are identical to VV-polarized signals. In this dissertation, elliptically
polarized EM signals are not used, nor are the switched transmission and receiving
modes (HV- and VH-polarized) considered. Both TE and TM waves are linearly
polarized since the variation of the electric field is linear. HH- and VV-polarized
signals are used in the laboratory radar measurements.
3.2
Finite Difference Time Domain Solution and

Yees Algorithm
While the governing equations of TE and TM waves are sufficient for analytically
studying the EM wave propagation and scattering problems, they are not directly
applicable for numerical application. Spatial and temporal discretization is necessary for implementing Maxwells curl equations in the discrete domain established by
computing machines.
Among current discretization methods in numerical analysis, the finite difference
method in time domain or the finite difference time domain (FDTD) method is chosen
for simulating the propagation and scattering of radar signals in space and time. The
main reason for choosing the FDTD method are (1) its explicit evaluation of Maxwells
curl equations, which avoids potential numerical instability, and (2) its accuracy of
approximation. Details of the FDTD method can be found in [251].
The central difference scheme is selected for its higher order of accuracy over
68
the forward and backward difference schemes. It can be shown that, with Taylors
theorem, for a one-dimensional function f (x), the first-order derivative of f (x) by
several difference schemes is
Forward difference: f (x) = f (x + h) f (x)
f (x)
f 0 (x) = O(h).
h
Backward difference: f (x) = f (x) f (x h)

Central difference: f (x) =
f (x)
f 0 (x) = O(h).
h
f (x + h) f (x h)
f (x)
f 0 (x) = O(h2 ).
2
h
When the spacing h is small enough, the error of the central difference scheme decreases faster than the forward and backward difference schemes. Should we list the
table of Stirling numbers [104], it is obvious that, given sufficient continuity of f (x),
the central difference is superior to the other two since the forward and backward
differences use only one side of the Stirling numbers in approximating f (x), while the
central difference uses both sides. Second and third orders of derivative of f (x) by
the central difference are
Second derivative
2
d2 f
d df
f (x + h) 2f (x) + f (x h)
f
=
2 =
2
dx
dx dx
h
h2
2
f = f (x + h) 2f (x) + f (x h)
Third derivative

d3 f
d d2 f
3f
f (x + 2h) 2f (x + h) + 2f (x h) f (x 2h)
=
=
3
2
3
dx
dx dx
h
2h3
f (x + 2h) 2f (x + h) + 2f (x h) f (x 2h)
3f =
2
Replacing x with t leads to the expression of discretization in time domain.
As for two-dimensional functions, consider f = f (x, y) in a grid as shown in Figure
3-1 and its Taylors series expanding at node 0 in the x direction:
1 2 f
f
1 3 f
2
3
f (x, y) = f (x0 ) +
(x x0 ) +
(x x0 ) +
(x x0 ) + ...
2
3
x x0
2! x x0
3! x x0
(3.23)
Using a uniform spacing h in both x and y directions, x = y = h. Since x3 = x0 h
69
y
10
11
12
x
Figure 3-1: Configuration grid in a two-dimensional domain
and x1 = x0 + h, we have
f
x
f
f (x1 , y1 ) = f1 = f0 +
x
f (x3 , y3 ) = f3 = f0

2 f

h
+

x2 0
0

2 f

h+ 2
x 0
0
h2 3 f h3
3
+ ...
2
x 0 6
h2 3 f h3
+ 3
+ ...
2
x 0 6
(3.24)
(3.25)
When h is small, higher order terms become insignificant.

f
f3 = f0
x
f
f1 = f0 +
x

2 f

h+ 2
x 0
0

2 f

h+ 2
x 0
0
h2
2
h2
2
(3.26)
(3.27)
Then we have
f
f1 f3
=
x 0
2h
2 f
f1 2f0 + f3
=
x2 0
h2
(3.28)
(3.29)
(3.30)
70
f
2 f
Similarly,
and
can be obtained.
y 0
y 2 0
f
f2 f4
=
y 0
2h

2
f
f2 2f0 + f4
=

y 2 0
h2
(3.31)
(3.32)
in the governing equations of TE and TM waves,

Replacing f (x, y) with E and H
the FDTD solution to Maxwells curl equations are reached.
As for three-dimensional problems, Yees algorithm (grid and notation) is widely

used in most FDTD solutions and is adopted in this dissertation [277]. In Yees
notation, unit spacings in space and time are chosen to be
x y z
, 2 , 2,
2
and
t
.
2
We
and
use dummy indices i, j, and k for denoting the spatial location of E and H
Note that i and j are not
a dummy index n for the time variable in E and H.
the imaginary number, nor is k the wave vector here. Following Yees notation, (xi ,
yj + y, zk + z, tn ) is written as (i, j + 12 , k + 21 ,n), and the Ex at (i, j + 12 , k + 12 ,n)
n
is Ex i,j+ 1 ,k+ 1 .
2
Considering Eq.(3.17) as an example, its FDTD expression is

Hz
+ Ex
y
n+ 1
n 1
Ex i,j+21 ,k+ 1 Ex i,j+21 ,k+ 1
2
2
2
2
=
t
n
" n
#
Hz i,j+1,k+ 1 Hz i,j,k+ 1

1
n
2
2
+ i,j+ 1 ,k+ 1 Ex i,j+ 1 ,k+ 1
2
2
2
2
i,j+ 1 ,k+ 1
y
Ex
1
=
t

(3.33)
n
Linear interpolation (semi-implicit approximation) is used for Ex i,j+ 1 ,k+ 1 .
2
n+ 1
n 1
Ex i,j+21 ,k+ 1 + Ex i,j+21 ,k+ 1
n
2
2
2
2
Ex i,j+ 1 ,k+ 1 =
2
2
2
71
(3.34)
n+ 1
n 1
Rearranging Eq.(3.33) with respect to Ex i,j+21 ,k+ 1 and Ex i,j+21 ,k+ 1 gives
2
1+
i,j+ 1 ,k+ 1 t
2
2i,j+ 1 ,k+ 1
2
n 1
Ex i,j+21 ,k+ 1
2
i,j+ 1 ,k+ 1 t
2
n+ 1
Ex i,j+21 ,k+ 1 =
2
2
2i,j+ 1 ,k+ 1
2 n 2
n

H
H
1
1
z
z
t
i,j+1,k+ 2
i,j,k+ 2
(3.35)
+
i,j+ 1 ,k+ 1
y
2
n+ 1
The explicit expression of a time-stepping Ex i,j+21 ,k+ 1 is found after dividing both
2
2
!
i,j+ 1 ,k+ 1 t
2
2
. Finally, the used FDTD expression of
sides in Eq.(3.35) by 1 +
2i,j+ 1 ,k+ 1
2
2
n+ 1
2

Ex i,j+ 1 ,k+ 1 is obtained.
2
i,j+ 1 ,k+ 1 t
2
2
1 2i,j+ 1 ,k+ 1
n 1
n+ 1
2
2
2
Ex i,j+ 1 ,k+ 1 =
Ex i,j+21 ,k+ 1 +
i,j+ 1 ,k+ 1 t
2
2
2
2
1 + 2 2 1 2 1
i,j+ ,k+
2 2
t
H n

i,j+ 1 ,k+ 1
z i,j+1,k+ 1 Hz i,j,k+ 1
2
2
2
2
1
1 t
y
1 + i,j+ 2 ,k+ 2
2i,j+ 1 ,k+ 1
2
(3.36)
Other electric and magnetic fields can be obtained following same procedure. Only
the results, as used in this numerical simulation work, are presented here.
i 1 ,j+1,k+ 1 t
1
2i 1 ,j+1,k+ 1
n+ 1
1
2
2
2

Ey n12
Ey i 1 ,j+1,k+ 1 =
i 2 ,j+1,k+ 12
2
2
1 + i 12 ,j+1,k+ 12 t
2i 1 ,j+1,k+ 1
2
2
n
t
H n

i 1 ,j+1,k+ 1
z i,j+1,k+ 1 Hz i1,j+1,k+ 1
2
2
2
2
1
1 t
y
i
,j+1,k+
1 +
2
2
2i 1 ,j+1,k+ 1
2
72
(3.37)
i 1 ,j+ 1 ,k+1 t
n+ 1
Ez i 12,j+ 1 ,k+1
2
t
1
2i 1 ,j+ 1 ,k+1

i 1 ,j+ 1 ,k+1
1
2
2
2
2
Ey n12 1
i 2 ,j+ 2 ,k+1
1
1
1
1
i 2 ,j+ 2 ,k+1
1 + i 2 ,j+ 2 ,k+1
1 + 2 1 1
i 2 ,j+ 2 ,k+1
2i 1 ,j+ 1 ,k+1
2
2
n
n
n
n
Hy i,j+ 1 ,k+1 Hy i1,j+ 1 ,k+1 Hx i 1 ,j+1,k+1 Hz i 1 ,j,k+1
2
2
2
2
x
y
2
(3.38)
n+1
n
Hx i 1 ,j+1,k+1 = Hx i 1 ,j+1,k+1 +
2
i 1 ,j+1,k+1
n+ 1
n+ 1
Ey i 12,j+ 3 ,k+1 Ez i 12,j+ 1 ,k+1
2
2
2
2
(3.39)
1
1

!
n+ 2
n+ 2

E
Ez i 1 ,j+ 1 ,k+1
t
z i+ 12 ,j+ 21 ,k+1
2
2
i,j+ 1 ,k+1
x
n
n+1
Hy i,j+ 1 ,k+1 = Hy i,j+ 1 ,k+1 +
2
n+1
n
Hz i,j+1,k+ 1 = Hz i,j+1,k+ 1 +
i,j+1,k+ 1
2
n+ 1
n+ 1
Ey i+ 12,j+1,k+ 1 Ey i 12,j+1,k+ 1
2
2
2
2
x
2
n+ 1
n+ 1
Ex i,j+23 ,k+ 1 Ex i,j+21 ,k+ 1
2
(3.40)
!
(3.41)
The leapfrog scheme in Yees algorithm proves to be efficient and reliable in the
space-time simulation of EM waves.
3.3
Absorbing Boundary Condition Perfectly Matched

Layer
In the numerical simulation of EM wave propagation and scattering in free space, the
use of reflectionless boundaries is necessary in order to simulate the infinite physical
environment (open space) in a finite numerical environment (computational domain).
73
Such boundary conditions are termed absorbing boundary condition (ABC). ABC
can be derived from either the integral form or the differential form of the reflectionless condition, leading to the global and local ABCs (Figure 3-2). The computational burden of global ABCs on storing and processing massive data makes local
ABCs attractive and efficient in most applications. Types of local ABCs include
(1) predictor-based ABC (radiation boundary condition): Merewether (1971) [115],
(2) operator-based ABC [78] [186] [18] [111], and (3) material-based ABC (perfectly
matched layer or PML) [24]. In this simulation work, PML is chosen for its demonstrated efficiency in various applications.
Maxwells equations
Assumed outgoing wave field

solutions (E, H)
Reflectionless condition of EM
fields at the artificial boundary
Integral form of the

reflectionless condition
Differential form of the

reflectionless condition
Global ABCs
Local ABCs
Figure 3-2: Derivation of global and local ABCs
3.4
Stability Criteria in Discretization
In the discretization process of continuous functions spatial and temporal intervals

must be chosen with care to avoid numerical dispersion and divergence throughout the
computation. To do so, stability criteria are applied for determining the spatial and
temporal intervals. Considered stability criteria in the simulation work are introduced
in this section.
74
3.4.1
Discretization in Space
Quantization criterion for the chosen spatial interval of sinusoidal functions (EM
waves) can be evaluated in the following manner. Consider a one-dimensional E(x, t)
as
E(x, t) = sin (kx x)
(3.42)
where k = the wave vector (rad/m), x = spatial variable (m), = radian frequency
(rad/sec), and t = temporal variable (sec). Following Yees notation for discretization,
we have
n
E(x, t) = E(i, n) = E i = sin (k ix nt)
(3.43)
The second-order space derivative of E(i, n) and its finite difference approximation
are
2E
= k 2 sin (kx t)
x2
(3.44)
2
[cos (kx) 1] sin (k ix nt)
x2
(3.45)
Eq.(3.45) will be identical to Eq.(3.44) if
(kx)2
= cos (kx) 1
2
(3.46)
is valid. Should we define the approximation error as

cos (kx) 1 +
Er (kx) =
cos (kx)
(kx)2
2
100%
(3.47)
a relationship between kx and Er (kx) can be obtained to evaluate the accuracy

of discretization in space. Figure 3-3 shows the kx-Er (kx) curves. For example,
to achieve Er (kx) 5%, the criterion for x is
x <
0.9
9c
=
k
20f
75
(3.48)
where x is in m/rad, c is the speed of light (m/s), and f is the temporal frequency
0.07
0.06
0.05
Er(k x), %
Er(k x), %
4
3
0.03
0.02
0.01
1
0
0.04
0.2
0.4
0.6
k x
0.8
0.1
0.2
k x
0.3
Figure 3-3: Quantization error Er (kx) vs. kx
(1/s or Hertz or Hz). Note that the factor 0.9 in Eq.(3.48) is determined in Figure
3-3. For EM waves of frequency f = 10 GHz, x must be less than 0.043 m (1.69 in).
The approximation error for E(x, t) will be 5% if x= 0.043 m (1.69 in) at f = 10
GHz.
3.4.2
Discretization in Time
According to the sampling theorem which is defined by the folding or Nyquist frequency, the necessary sampling rate in time, t, is determined by
t <
1
2fmax
(3.49)
where fmax is the maximum frequency content (Hz) (Nyquist frequency) that is allowed to ensure the signal can be fully reconstructed at sampling rate t [Hamming
1989]. For instance, given a maximum frequency f = 10 GHz, the sampling rate must
be less than 51011 sec. or 5102 ns.
Another criterion on the coupling between t and spatial increments (x, y,
z) is the Courant-Friedirchs-Levy (CFL) or the Courant stability criterion which is
76
defined by [214]
1
t < q
1
1
+ y
+
c x
1
z
For uniform discretization in three-dimensional problems we have t <

t <
c 2
(3.50)
x
,
c 3
and
for two-dimensional problems. Derivation of the CFL stability criterion
can be found at [252]. For example, t must be less than 3.538103 ns. for x=
0.0015 m.
The quantization approach applied on x can be used on t. The second-order
time derivative of E(x, t) and its finite difference approximation are
2E
= 2 sin (kx t)
t2
2
[cos (t) 1] sin (k ix nt)
t2
(3.51)
which means
t =
0.9
2f
(3.52)
for an approximation error Er(t) 5% at f = 10 GHz. This gives a t =

2.8651011 sec or 2.865102 ns.
3.5
Two-Dimensional and Three-Dimensional Simulations
In this research, two-dimensional and three-dimensional simulations of EM wave propagation and scattering are conducted using GprMax [132]. The code was developed
based on the principles previously introduced in this chapter.
3.5.1
Validation of the Code
The code is validated by four examples whose theoretical solutions are available; (1)
reflection response from a two-dimensional (2D) perfectly electric conductor (PEC)
plate due to a Gaussian current source, (2) reflection response from the 2D PEC
77
plate due to a continuous/sinusoidal current source, (3) reflection response from a 2D

dielectric plate due to a Gaussian current source, and (4) reflection response from the
2D dielectric plate due to a continuous current source. Geometrical configuration of
the four examples is provided in Figure 3-4.
In Figure 3-4, a 0.3m-by-0.3m computational domain filled with air (0r = 1, 0r =
1) was first built with uniform spatial discretization of x = y = 0.0025 m (0.098
in.). At the right boundary a target plate was inserted, which is made of PEC (0r =
3000(>> 1), 0r = (>> 1)) or dielectric(0r = 525, 0r =1). Eight layers of PML ABC,
whose total thickness is 0.02 m (0.79 in.), were deployed on four boundaries (top,
bottom, left, and right) of the square domain to simulate the reflectionless effect in
an infinite space. A current source (transmitter) was placed at the location (0.075 m,
0.15 m). A target plate made of PEC or dielectric was located at the right boundary
of the domain, simulating an infinite surface (x [0.25m, 0.3m], y [0m, 0.3m]). The
reflection response was collected at the receivers location (0.15 m, 0.15 m) between
the current source and the target plate. The receiver was 0.1 m (3.94 in.) from
the target plate to the right and 0.075 m (2.95 in.) from the source to the left.
Total response collected at the receiver was comprised of the incident wave from the
transmitter and the reflected wave from the target plate. Later, the net reflection
response was extracted from the total response, considering the time-delay between
the transmitter and the receiver, and compared with theoretical values. The wave
vector was k = xkx + yky , and the current source was in the z direction. This led to
the use of TM waves (Hx , Hy , and Ez ) in the examples. Time interval was determined
by the CFL stability criterion to be 5.891012 sec or 5.89103 ns, and the total
duration of simulation was 1108 sec or 10 ns, leading to a time series response with
1,696 points collected at the receiver location. In these examples, the impulse current
source is a Gaussian function of the following form.
"
1
I(t, f ) = exp t
f
2 #
(3.53)
where = 2(f )2 . The continuous/sinusoidal current source is a sinusoidal function

78
PML layer
PEC or dielectric
free space
0.3 m
Hertzian
current
dipole
0.02 m
0.15 m
0.075 m
0.15 m
0.25 m
Transmitter
Receiver
0.3 m
Figure 3-4: Geometry of a 2D model for validation
of the following form.

I(t, f ) =
tf
sin (2f t)
4
(3.54)
The theoretical response of EM waves reflected from an infinite interface can be

evaluated by the amplitude reflection coefficient, r.
0 0
1
r = r0 r0
r r 1
(3.55)
For non-magnetic media, 0r = 1. In the considered examples, the 2D target plate

extends to the boundaries of the square domain, making it equivalently an infinite
plate to any sources within the domain. The amplitude reflection coefficient of a PEC
plate is
0
0
r >> 1 1

r=0
r0 = 1
r >> 1 1
r
79
(3.56)
with a phase difference of . As for dielectric plates with a 0r ranging from 5 to
25, r ranges from 0.38 to 0.67. Comparison of impulse response is performed at the
peak value of the impulse and continuous responses. These results are provided in
the following.
1. 2D PEC plate subjected to a Gaussian current source The current
source (solid line) and the net reflection (dashed line) are shown in Figure
3-5. In Figure 3-5, the difference in the peak amplitudes is around 0.06%, an
insignificant difference. The amplitude reflection coefficient is 0.9968 by FDTD,
and the theoretical value is 1. The time delay between the two peaks is 114t
= 0.2017 m 0.2 m which is twice the distance between the receiver and the
plate. In this example, the FDTD solution is in excellent agreement with the
theoretical value.
1.5
Source
Reflection from PEC
1.2165
1
Ez (V/m)
0.5
-0.5
-1
1.2172
-1.5
215 t
0.1
0.2
0.3
0.4
0.5
Time (sec)
0.6
0.7
0.8
0.9
1
-8
x 10
329 t
Figure 3-5: Gaussian current source and the reflection from PEC
2. 2D PEC plate subjected to a sinusoidal current source Figure 3-6

shows the sinusoidal current source and the reflection response from the PEC
plate. A constant time delay of 114t is observed between two curves, representing twice the traveling distance between the receiver and the plate. The
80
difference in the peak values of incident and reflected amplitudes is less than
0.1%, also suggesting an excellent agreement between the FDTD solution and
theoretical prediction.
2
1.5
1
Ez (V/m)
0.5
0
-0.5
-1
-1.5
-2
Source
Reflection from PEC
0
0.1
0.2
0.3
0.4
0.5
Time (sec)
0.6
0.7
0.8
0.9
1
-8
x 10
Figure 3-6: Sinusoidal current source and the reflection from PEC
3. 2D dielectric plate subjected to a Gaussian current source Figure 3-7

shows the impulse current source and the reflection responses from a 2D dielectric plate with five values of the dielectric constant 0r of the plate; 5, 10, 15, 20,
and 25. In Figure 3-7, the time delay representing twice the traveling distance
between the receiver and the plate (0.2017 m) is almost identical to the actual
value (0.2 m). Figure 3-8 shows the close-up of the reflection responses. It is
found that the greater the dielectric constant the higher the reflection amplitude, as Eq.(3.55) predicts. As for the amplitude reflection coefficient, Figure
3-9 shows the theoretical curve and the FDTD solution. Error or relative difference is calculated and provided in Figure 3-10. It is found that the maximum
error (2.4%) is less than 3%.
4. 2D dielectric plate subjected to a sinusoidal current source Figure
3-11 shows the sinusoidal current source and the reflection response from a 2D
81
Ez (V/m)
0.5
0
Source
'r =5
-0.5
'r =10
'r =15
-1
'r =20
'r =25
-1.5
215 t
Time (sec)
5
-9
x 10
329 t
Figure 3-7: Gaussian current source and the reflection from PEC
dielectric plate with a dielectric constant 0r = 25. The error is also less than 3%.
From the four examples provided in this section, the accuracy of the FDTD simulations is validated, both qualitatively and quantitatively. Hence, it is believed that
the used FDTD code is reliable.
3.5.2
Actual Far-Field Simulation
An incident wave impinging on a finite specimen with boundaries modeled within

the domain of consideration produces multiple reflections and transmissions. When
analyzing such time domain response as a forward study, one can choose to investigate
the content of the signal with respect to certain reflection/transmission by windowing
the time series. In the considered problem, it is the first complete reflection (reflected
from the backside of the specimen) that is concerned. While multiple reflections could
be also advantageous in the analysis, it is not convenient to distinguish/determine
the amplitude and phase differences attributed to presence of a local defect in the
specimen from the total response. Therefore, the windowed time domain response is
82
0.8
'r =5
0.6
'r =10
'r =15
E z (V/m)
0.4
'r =20
'r =25
0.2
0
-0.2
-0.4
-0.6
1.5
2.5
Time (sec)
3
-9
x 10
Figure 3-8: Gaussian current source and the reflection from PEC Close-Up
analyzed.
The monostatic radar operation defined in the developed NDT technique and in
laboratory radar measurements indicates a coincidental location for both the source
(transmitter) and the receiver. As the result of it, difficulties including angular coupling between the source and the receiver and a required computational domain in
order to contain the source and the receiver are avoided. In laboratory radar measurements the monostatic radar is placed in alignment with the center of cylinder
specimens and perpendicular to the longitudinal axis of cylinder specimens. Such
configuration is also followed in numerical simulation.
While the option of an ideal equivalent Huygens surface (surface equivalence theorem) is not available in GprMax, the infinite range distance for perfect plane waves
is not possible to be modeled in the simulation using GprMax. Rather, considering
that in the proposed NDT technique the radar is configured beyond the far-field distance but located at infinite range from the structure, it is decided to place the radar
in the far-field region (approximate plane waves) in the simulation. This way, the
simulation will be honestly reproducing actual far-field measurements.
Figure 3-12 shows the computational configuration. The size of the numerical
83
Reflection coefficient
0.65
0.6
0.55
0.5
0.45
0.4
Theoretical
FDTD solution
0.35
0.3
10
15
'
20
25
Figure 3-9: Theoretical curve and the FDTD solution of the amplitude reflection
coefficient of a 2D dielectric plate
domain is determined so as to satisfy the far-field distance. Snapshots of the incident
fields are shown in Figure 3-13 using a 9 GHz frequency sinusoidal signal as example.
The purpose of performing two-dimensional (2D) simulations is twofold; first,

2D models are convenient to built and results are efficient to obtain. Secondly, the
negligence of the third dimension in 2D models removes the influence from the third
dimension in the simulation results and, thus, purifies the parametric study by considering only the parameters within the 2D plane. In other words, scattering/reflection
responses, owing to the incidence of plane waves, from the third dimension are eliminated.
In the simulation, the normal and oblique incidence schemes are both considered,
where in the normal incidence scheme the concrete cylinder specimens with/without
defect are defined as a circle and in the oblique incidence scheme they are described
as a rectangle. All models are situated in free space (air), as the physical condition
suggested in laboratory radar measurements. The defect is placed at the surface of
the concrete core, which is characterized by its width, depth, and length. System
parameters include incident frequency, incident angle, defect width, defect depth,
defect length, and concrete permittivity. Frequency range is from 8 GHz to 18 GHz,
84
2.5
Relative difference (%)
2.4
2.3
2.2
2.1
1.9
10
15
'
20
25
Figure 3-10: Relative difference/error between the theoretical curve and the FDTD
solution
and the models are constructed in analogy to the GFRP-concrete specimens AD1 and
AD2 whose dimensions are previously provided in the chapter on laboratory radar
measurements.
In this section, time-domain responses in the neighborhood of the 2D models in
the normal and oblique incidence schemes are produced to illustrate the scattering response due to the presence of a near-surface defect and rebar(s) in concrete (dielectric)
and metallic (PEC) specimens.
3.6
Simulation Results
In the following cases, only scattered Ez is provided and discussed in order to save
the space. The use of monostatic radar suggests that only the reflection response in
the line-of-sight of radar can be perceived. In the following results, incident waves
are introduced from the left boundary of the numerical domain, implying that the
monostatic radar is located on the left boundary. Additionally, the line-of-sight of
the radar intersects with the left boundary at Y = 0.2 m, the center of left boundary.
85
2
1.5
1
Ez (V/m)
0.5
0
-0.5
-1
-1.5
-2
Source
Reflection from dielectric
0
0.1
0.2
0.3
0.4
0.5
Time (sec)
0.6
0.7
0.8
0.9
1
-8
x 10
Figure 3-11: Sinusoidal current source and the reflection from a 2D dielectric plate
3.6.1
Damage Detection in Normal Incidence
In the numerical simulation of normal incidence responses for damage detection, the
interest is to know whether the scattered signal due to the presence of a near-surface
defect can be detected. Figure 3-14 shows the net reflection field of Ez from an intact,
lossless concrete cylinder in the normal incidence scheme. It is noticed that, in Figure
3-14, a strong reflection (specular return) from the surface of the cylinder is observed.
This specular return will present at all incident angles in the normal incident scheme
due to the circular geometry of the cylinder in such scheme, suggesting a permanent
influence of specular return on the normal incidence responses. Scattering signals due
to the features (defect or rebar) of the cylinder will be mixed with the specular return
in the observation of the radar. A low signal-to-noise ratio (SNR) can be expected.
3.6.2
Damage Detection in Oblique Incidence
In the oblique incidence scheme, it is of interest to know if the near-surface defects

can be revealed by the reduction/removal of specular returns in such scheme. In this
section, the FDTD simulation explains and demonstrates this behavior using two
incident angles 0 and 30 .
86
source
receiver
PEC or dielectric
y
free space
0.6 m
0.3 m
0.075 m
0.15 m
5.7 m
6m
Figure 3-12: Numerical domain for simulating actual far-field radar measurements
The concrete cylinder in the oblique incidence scheme was modeled as a rectangle
in 2D space, and the cubic defect AD1 was used. Four cases were considered at
the incident angle of = 0 ; (1) intact, lossless concrete cylinder, (2) intact, lossy
concrete cylinder, (3) damaged, lossless concrete cylinder, and (4) damaged, lossy
concrete cylinder. Among them, cases (1) and (2) were used as the background signals
for cases (3) and (4) for extracting the scattered fields due to defect. Figures 3-15
and 3-16 show the total fields of Ez of cases (1) and (2), respectively. The extracted
scattered fields of Ez of cases (3) and (4) are provided in Figures 3-17 and 3-18.
In
Figures 3-17 and 3-18, the scattered Ez due to defect was found in the line-of-sight
direction of the monostatic radar. However, given the intensity of specular returns in
Figures 3-15 and 3-16, the existence of such scattered Ez is difficult to assert.
Two cases were considered at the incident angle of = 30 ; (1) intact, lossless
concrete cylinder and(2) damaged, lossless concrete cylinder. It is found that, in
Figure 3-19, the specular return from the cylinder is no longer perceivable by the
monostatic radar. Meanwhile, the scattered field becomes detectable by the radar as
shown in Figure 3-20 since the background noise (specular return) has been removed.
It is understood that the geometry of near-surface defects is related to the selection
of incident angle even after the removal of specular returns. While greater incident
angles (>30 ) may demonstrate satisfactory results, they pose a difficulty when they
are applied in field applications; greater incident angles need shorter inspection dis87
0
-0.5
-1
Net reflected E z (V/m)
Total reflected E z (V/m)
Incident Ez (V/m)
1
0.5
0.5
1.5
2.5
Time (ns)
3.5
4.5
0.5
1.5
2.5
Time (ns)
3.5
4.5
0.5
1.5
2.5
Time (ns)
3.5
4.5
2
1
0
-1
-2
0.5
0
-0.5
Figure 3-13: Incident field, total field, and net reflection field of a 9 GHz signal
tances than what smaller ones need. Therefore, it is of interest to explore the use of
smaller incident angles. In the cases considered here, the angular range [0 , 30 ] is to
be explored.
3.6.3
Effect of Defect Width in Normal Incidence
In studying the effect of defect width in normal incidence thirteen 2D models are
constructed as shown in Figure 3-21. As one can observe in the pattern of variation
among these models, the width (opening) of the artificial defect decreases from the
benchmark defect (physically embedded in the specimen AD1) (1.5 in.) to the extreme case of 0.12 in. The purpose of this part of parametric study is to learn how
the variation of defect width affects the reflection amplitude in a monostatic radar
measurement scenario. In Figure 3-21, two frequencies, 8 GHz and 9 GHz, were considered. It is found that, in general, the use of higher frequencies leads to a greater
88
reflected electric field, while the measurement frequencies have their own region of
defect width in which the reflected electric field is suddenly decreased.
3.6.4
Effect of Defect Depth in Normal Incidence
For the study of defect depth another sixteen 2D models are prepared as shown in
Figure 3-23. In this group of models it is the depth of the defect decreasing from
1 in. to 0.06 in. The purpose of this study is to learn how the variation of defect
depth affects the reflected electric field. In Figure 3-24, the reflection coefficient is the
ratio between the reflected electric field and the incident electric field. In Figure 3-24,
two frequencies of signals were used; 8 GHz and 9 GHz. It is found that these two
frequencies demonstrate different reflection patterns with respect to the variation of
defect depth. In other words, detectability of near-surface defects of this kind depends
on the use of incident frequency. Variation of the reflected electric field displays a
sinusoidal pattern with respect to the changes of defect depth.
3.6.5
Rebar Detection in Normal Incidence
Three simulation cases were prepared; (1) plain concrete cylinder, (2) plain concrete
cylinder with a center rebar, and (3) plain concrete cylinder with four rebars configured as the rebars in the specimen 4RE. In these cases, the concrete cylinder was
modeled as a circle in 2D space, and the rebar was a small circle in the center of the
cylinder. Case (1) was used as a background signal for extracting the scattering signal
due to rebar(s). Dielectric constant of concrete was taken to be 5, and it is considered
lossless (nonconductive)). Dimensions of the cylinders and rebars were same as the
ones used in physical laboratory specimens when described in a 2D domain. The
rebars were modeled as perfectly electric conductor. Simulation time was determined
in order to incorporate the instants right before the impinging of incident waves.
The total fields (Hx , Hy , and Ez ) of the time-domain response of case (2) are first
provided as shown in Figures 3-25, 3-26, and 3-27. Scattered fields due to the presence
of rebar(s) in cases (2) and (3) are extracted from the total fields using the case (1)
89
response as the background signal. Figures 3-28, 3-29, and 3-30 show the scattered
fields of case (2).
Figures 3-25, 3-26, and 3-27 demonstrate different patterns of
wave scattering due to the presence of a point PEC inside a dielectric circle, as the
FDTD simulation result suggested. The scattered fields reveal the scattering response
of the point PEC. It is observed that the amplitude of scattered Ez (due to the rebar)
in the line-of-sight direction of the radar is not significant. This confirms the finding
on the laboratory measurements on specimens PC and CRE; difference between the
reflection responses with and without a rebar embedded at a depth of 3 inches inside
a concrete cylinder is not strong in the normal incident scheme, when compared with
the amplitude of specular returns. Nonetheless, this difference leads to a greater
reflection response of the concrete specimen with rebar, also to be consistent with the
laboratory measurement results.
When four rebars are present at different depths inside a concrete cylinder, it
is found that the amplitude of scattered Ez (due to the rebars) in the line-of-sight
direction of the radar is increased. This is attributed to the decrease of embedment
depth (from 3 in. to 0.25 in.) and the presence of adjacent rebars in the concrete
cylinder. Figure 3-31 shows the scattered Ez of case (3). It is noticed that strong
reflections are present not only in the line-of-sight direction of the radar, but also
in other directions. The snapshot images at time instants 21.4 ns, 21.7 ns, and 22
ns in Figure 3-31 clearly demonstrate this phenomenon. The strong reflections in
other directions are caused by adjacent rebars next to the rebar on the line-of-sight
direction. These reflections are the reason why the normal incidence scheme is capable
of detecting rebars outside the line-of-sight direction.
90
Ez at time = 18.7 ns
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-14: Scattered (net reflection) field of Ez Intact, lossless concrete cylinder
91
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-15: Total field of Ez Intact, lossless concrete cylinder
92
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-16: Total field of Ez Intact, lossy concrete cylinder
93
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-17: Scattered field of Ez , = 0 Damaged, lossless concrete cylinder
94
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-18: Scattered field of Ez , = 0 Damaged, lossy concrete cylinder
95
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-19: Scattered field of Ez , = 30 Intact, lossless concrete cylinder
96
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-20: Scattered field of Ez , = 30 Damaged, lossless concrete cylinder
97
width
(1.42, 1)
(1.30, 1)
(1.18, 1)
(1.06, 1)
(0.94, 1)
(0.83, 1)
(0.71, 1)
(0.59, 1)
(0.47, 1)
(0.35, 1)
(0.24, 1)
(0.12, 1)
depth
(width, depth)
= (1.5 in., 1 in.)
Figure 3-21: Thirteen 2D models for studying the effect of defect width in normal
incidence
0.3
0.25
0.2
0.15
f = 8 GHz
f = 9 GHz
0.1
0.017
0.018
0.019
0.02
0.021
0.022
Defect width (m)
0.023
0.024
0.025
Figure 3-22: Variation of reflected electric field with respect to different defect widths
in normal incidence
98
(1.5, 0.94)
(1.5, 0.86)
(1.5, 0.83)
(1.5, 0.77)
(1.5, 0.71)
(1.5, 0.65)
(1.5, 0.59)
(1.5, 0.53)
(1.5, 0.47)
(1.5, 0.41)
(1.5, 0.35)
(1.5, 0.3)
(1.5, 0.24)
(1.5, 0. 18)
(1.5, 0.12)
(1.5, 0.06)
Figure 3-23: Sixteen 2D models for studying the effect of defect depth in normal
incidence
0.3
0.25
0.2
0.15
0.1
f = 8 GHz
f = 9 GHz
0.005
0.01
0.015
Defect depth (m)
0.02
0.025
Figure 3-24: Variation of reflected electric field with respect to different defect depths
in normal incidence
99
Hx at time = 18.7 ns
Hx at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Hx at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-25: Total field of Hx Plain concrete cylinder with a center rebar
100
Hy at time = 18.7 ns
Hy at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Hy at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-26: Total field of Hy Plain concrete cylinder with a center rebar
101
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-27: Total field of Ez Plain concrete cylinder with a center rebar
102
Hx at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Hx at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-28: Scattered field of Hx Plain concrete cylinder with a center rebar
103
Hy at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Hy at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-29: Scattered field of Hy Plain concrete cylinder with a center rebar
104
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-30: Scattered field of Ez Plain concrete cylinder with a center rebar
105
Ez at time = 19 ns
0.2
0.3
0.2
0.3
0.1
0.2
0.3
0.1
0.2
0.2
0.3
0.2
0.3
0.1
0.3
0.2
Y (m)
0.2
0.3
0.2
0.3
0.1
0.3
0.2 0.3
X (m)
0.3
Y (m)
0.1
0.2
0.3
0.1
0.2
Ez at time = 22 ns
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
0.1
Y (m)
0.1
0.2
0.2
0.3
0.1
0.3
0.1
0.3
0.1
0.2
Y (m)
Y (m)
Y (m)
0.3
0.1
0.3
Y (m)
0.2
0.1
0.2
0.2
0.3
0.1
Y (m)
0.1
Y (m)
0.1
Y (m)
Y (m)
0.1
0.2
0.3
0.1
0.2 0.3
X (m)
0.1
0.2 0.3
X (m)
Figure 3-31: Scattered field of Ez Plain concrete cylinder with a center rebar
106
3.6.6
2D and 3D Responses
A 3D PEC cylinder model is shown in Figure 3-32, whose height is 12 in. and diameter
is 6 in. A plane wave source is initiated by the introduction of Hyguens surface such
that the longitudinal axis of the cylinder is parallel to the wavefront of the plane wave.
The reflected electric field of this model is compared with the one from the 2D version
of the model with identical dielectric and geometrical properties. Figure 3-33 shows
the comparison between the 2D and 3D responses of the PEC cylinder model. It is
found that the 3D response is weaker than the 2D response since the 3D model has
less area for reflection. With more variables (e.g., dielectric properties, local defects)
involved, the difference between 2D and 3D response can also be expected.
Figure 3-32: A 3D lossless dielectric cylinder model
3.7
Summary
Findings in this chapter are summarized and discussed in the following.

In the FDTD simulation of normal incidence measurements, direct reflection
(specular return) is dominant, leading to a low signal-to-noise ratio (SNR) (Figure 3-14). Near-surface defects are unlikely to be detected in such measurement
scheme.
107
2
1.5
1
Ez (V/m)
0.5
0
-0.5
-1
Source
2D FDTD -- PEC
3D FDTD -- PEC
-1.5
-2
0.1
0.2
0.3
0.4
0.5
Time (sec)
0.6
0.7
0.8
0.9
1
-8
x 10
Figure 3-33: Comparison between the 2D and 3D responses of a PEC cylinder model
In the FDTD simulation of oblique incidence measurements, specular returns
are dramatically reduced (Figures 3-19 and 3-20). This suggests a high SNR
for detecting the presence of near-surface defects in the far-field region.
When four rebars are present at different depths inside a concrete cylinder, it
is found that the amplitude of scattered Ez (due to the rebars) in the line-ofsight direction of the radar is increased. This is attributed to the decrease of
embedment depth (from 3 in. to 0.25 in.) and the presence of adjacent rebars in
the concrete cylinder. Strong reflections are present not only in the line-of-sight
direction of the radar, but also in other directions. These reflections are the
reason why the normal incidence scheme is capable of detecting rebars outside
the line-of-sight direction.
3D reflection response is different from 2D reflection response in a way that 3D

reflection is weaker than 2D reflection. While 2D models offer certain computational advantages (less computational cost), 3D models are needed for better
understanding the radar response (electromagnetic reflection) of the considered
GFRP-concrete systems.
108
With these findings, laboratory radar measurements (Chapter 4) are planned

to further investigate the system design parameters such as measurement scheme,
measurement frequency, and the range of incident angles and to understand how the
actual radar response of GFRP-concrete systems in a laboratory environment should
be.
109
110
Chapter 4
Laboratory Radar Measurements
Every effect obviously has its cause, which can be retracted from cause to
cause into the abyss of eternity; but every cause does not have its effect to
the end of time.
Philosophical Dictionary, Francis-Marie Arouet (Voltaire) (16941778)
A crucial experiment is one which gives a result consistent with one
hypothesis and inconsistent with another.
Anonymous
This chapter reports the laboratory radar measurements collected on concrete and
GFRP-retrofitted concrete cylinders with artificial features (defect and rebar). As
found in Chapter 3 Numerical Simulation, measurement scheme, measurement
frequency, and incident angle play a key role in the reflection response of GFRPconcrete cylinders. The purposes of conducting radar measurements on laboratory
specimens are multiple; (1) a forward study can be performed using laboratory specimens whose defects are known. Radar measurements made on intact (without defect)
and damaged (with defect) specimens can be evaluated based on their raw measurements and finally compared after being processed into images. (2) the laboratory
configuration provides a noise-free environment for the reflection radar measurements
of the specimens. The removal of background noise is advantageous for better distinguishing the signal due to the presence of defect at a higher signal-to-noise ratio
111
(SNR). Such measurements are insightful for damage detection and convenient to
deal with in terms of the need of denoising; and (3) the radar measurement on a
plain concrete cylinder can provide insights to the variation of dielectric properties of
concrete and is advantageous for the dielectric modeling of concrete (Chapter 6).
GFRP-retrofitted concrete cylinder specimens were manufactured in the Concrete
Laboratory of the Department of Civil and Environmental Engineering at MIT. Physical radar measurements were performed under the supervision of Dennis Blejer at
the MIT Lincoln Laboratory (MIT LL). In this chapter, the design of experimental program, manufacturing of laboratory specimens, and experimental configuration
and parameters are first described. Selected radar measurements are reported in this
chapter for the discussion of damage detection and rebar detection.
4.1
Experimental Program
An experimental program aiming at investigating the effect of unseen near-surface

anomaly in the far-field monostatic ISAR measurement of GFRP-wrapped concrete
specimens is prepared. In this work, there are three types of anomaly that are targeted
for detection.
1. GFRP Delamination and Debonding GFRP delamination occurs when
wrapped multi-layer GFRP sheets deteriorate due to the application of service
loads and/or environmental effects. GFRP debonding happens either during
construction (trapped air voids between GFRP sheets and irregular surface of
concrete during the wrapping of GFRP sheets), or due to the application of
service loads and/or environmental effects. This type of anomaly (Artificial
Defect, AD) is simulated by the introduction of Styrofoam pieces between
GFRP sheets and the surface of concrete.
2. Reinforcement Detection of the location of reinforcement is important to
the rehabilitation and repair of corroded R/C structures. Steel rebars are embedded in GFRP-wrapped concrete specimens to simulate the detection of re112
inforcement in such structure.

Table 4.1 lists all laboratory specimens in the experimental program in this research.
All specimens are casted in cylinder forms.
Table 4.1: Designed specimens
Spec. Rad. H
Freq.
(in) (in) band
Azi.
(deg.)
Scheme
Description
Figure
CON
12
X, Ku
30
Normal
Concrete cylinder
4-1
CRE
12
X, Ku
30
Normal
Concrete
cylinder 4-2
with a rebar in the
center
GFRP 3
12
30
Normal
Concrete
cylinder 4-1
wrapped with GFRP
4RE
12
15
Normal
Concrete
cylinder 4-2
with four rebars
AD1B 3
12
X, Ku
30
Normal,
Oblique
GFRP-con. cylinder 4-3

with AD1, intact side
AD1F 3
12
X, Ku
30
Normal,
Oblique

with AD1, damaged
side
AD2
15
45
Normal,
Oblique

with AD2, damaged
side
AD3
15
45
Normal

with AD3, damaged
side
It is noteworthy to point out that scaled GFRP-wrapped concrete specimens differ

from GFRP-wrapped concrete columns in the following aspects:
1. Size (radius and height) The radius of the GFRP-concrete specimens is
approximately 3 inches/in. (or 7.62 centimeters/cm), and the specimens are
designed in two heights; 12 in. and 15 in. The radius and height of actual
concrete columns vary in a wide range of values, depending on the level of
113
12 in
(30.4 cm)
12 in
(30.4 cm)
6 in
(15.2 cm)
6.2 in
6 in
(15.2 cm)
CON
GFRP
Figure 4-1: Specimens CON and GFRP
12 in
(30.4 cm)
15 in
(38.1 cm)
1.5 in
(3.8 cm)
6 in
(15.2 cm)
135
0.1 in
(0.25 cm)
45
45
1 in
(2.5 cm)
4RE
CRE
Figure 4-2: Specimens CRE and 4RE

114
135
0.5 in
(1.27 cm)
1.5 in
(3.8 cm)
3 in
(7.6 cm)
12 in
(30.4 cm)
15 in
(38.1 cm)
1 in
(2.5 cm)
0.2 in
(0.5 cm)
6.2 in
6 in
(15.2 cm)
3 in
(7.6 cm)
1.5 in
(3.8 cm)
AD1
AD2
Figure 4-3: Specimens AD1 and AD2
15 in
(38.1 cm)
6.2 in
6 in
(15.2 cm)
1 in
(2.5 cm)
1 in
(2.5 cm)
AD3
Figure 4-4: Specimen AD3

115
needed bearing capacity. In any case, they are much greater than the values of
the specimens used in this study.
2. Number of GFRP sheets/layers Only one sheet/layer of GFRP composite
is used in this work. In practice, the use of more than one layer of GFRP
composite is usual.
3. Reinforcement Steel reinforcement is considered separate from the artificial
defect in this study in order to distinguish their contributions in the far-field
monostatic ISAR measurements. Such design allows us to simplify the problem
by investigating their individual effects.
4.2
Manufacturing of the Specimens
The concrete and GFRP-wrapped concrete specimens were manufactured in the Concrete Laboratory of the Department of Civil and Environmental Engineering at MIT.
In this section, descriptions of the materials used in the experimental work are provided. Material suppliers are also indicated. Materials used in the manufacturing of
GFRP-wrapped concrete specimens and their suppliers are listed in Table 4.2. Standards by ASTM (American Society for Testing and Materials) and material suppliers
for the quality control of the materials used in the manufacturing of concrete and
GFRP-concrete specimens are indicated in Table 4.3. Equipments used in the manufacturing of concrete specimens and GFRP-epoxy sheets include a concrete mixer
for rotationally blending material components, an iron plate for casting concrete into
cylinder forms, a plastic rectangular plate for molding E-glass fabric with epoxy, and
other auxiliary accessories such as handheld shovels. The manufacturing of GFRPwrapped concrete specimens involves the following steps.
1. Calculation of the needed fractions of each material component in concrete
(water, cement, fine aggregate, and coarse aggregate) and GFRP sheets (Eglass fabric and epoxy resin)
116
Table 4.2: Used materials and their suppliers

Material
Brand
Supplier
Cement
Dragon r Portland Cement

Type I
Dragon Products Company,

Thomaston, Maine
Fine
aggregate
Walbro r Mortar Sand
B Vitalini Inc.,
Massachusetts
Milford,
Coarse
aggregate
Walbro r PEA Gravel
B Vitalini Inc.,
Massachusetts
Milford,
Water
Batch water
M.I.T.
Styrofoam Foamular 250 r Extruded Owens

Corning
Polystyrene Insulation
Toledo, Ohio
Corp.,
E-glass
fabric
Tyfo r SEH-51 Composite
Fyfe Co. LLC, San Diego,

California
Epoxy
resin
Tyfo r S Epoxy
Fyfe Co. LLC, San Diego,

California
Table 4.3: Manufacturing standards

Material
Standard
General
ASTM C125-03 Standard Terminology Relating to Concrete

and Concrete Aggregates
Aggregates
ASTM C33-03
Aggregates
Standard
Specification
for
Concrete
ASTM C136-04 Standard Test Method for Sieve Analysis of

Fine and Coarse Aggregates
ASTM C127-04 Standard Test Method for Density, Relative Density (Specific Gravity), and Absorption of Coarse
Aggregate
Water
ASTM C1602/C1602M-05 Standard Specification for Mixing

Water Used in the Production of Hydraulic Cement Concrete
Concrete
ASTM C192/C192M-05 Standard Practice for Making and

Curing Concrete Test Specimens in the Laboratory
GFRP sheet
Fyfe 01-05 Quality Control Manual for the Tyfo r FIBRWRAP System
117
2. Preparation of the needed amount (in weight) of materials for each batch in the
mixing of concrete and GFRP
3. Preparation of the equipments and concrete cylinder forms (two types of forms
are used; paperboard and iron)
4. For concrete specimens with anomaly:
Attaching artificial defects onto the inside surface of the form at a prescribed location for artificially damaged concrete specimens
Attaching and fixing rebar(s) in the prescribed location(s) inside the cylinder form for concrete specimens with rebar(s)
5. Brushing the inside surface of concrete cylinder forms with lubricant (oil) in
order to ensure a clean separation between the form and the concrete
6. Mixing concrete patch, casting and consolidating concrete cylinder specimens
7. Removal of the forms of concrete specimens after one day
8. Curing concrete cylinder specimens in water tank for 28 days
9. Mixing E-glass fabrics with epoxy resin in a rectangular plate to form GFRP
sheets
10. Brushing the surface of concrete specimens using epoxy resin
11. Wrapping one layer of GFRP sheet onto the concrete specimen
12. Curing the GFRP sheet in room temperature for one day
4.3
Experimental Configuration and Parameters
Laboratory radar measurements were performed in the MIT Lincoln Laboratory using
the compact RCS (Radar Cross Section)/antenna range facility. The experimental
set-up consists of a horn antenna, stepped-frequency radar, network analyzer systems,
118
and a Harris dual-shaped reflection system (Model 1606) designed for conducting farfield measurements which are made at a range beyond the far-field distance. The
far-field distance, dff , is generally defined by [243, 137]
dff =
2D2
(4.1)
where D is the maximum dimension of radar antenna apertures, and is the minimum wavelength of radar signals. The reflection system is used for transforming
cylindrical/spherical waves (by the radar antenna) into plane waves in the limited
space of laboratory, which is not needed in field measurements where the far-field
distance requirement is typically satisfied. Design of reflection systems can be found
at [16]. The picture and schematic of the facility are provided in Figures 4-5 and
4-6.
The compact RCS/antenna range facility can achieve a 6-ft (20-meter) quite
Figure 4-5: Compact RCS antenna range facility in the MIT LL [Courtesy of the
MIT LL]
119
zone with the installation of Chebyshev absorbers (shown in Figure 4-5) covering the
floors, walls, and the support tower of the laboratory. The facility can make fully
polarimetric RCS measurements and achieve high signal-to-noise ratio (SNR) for a
large frequency bandwidth ranging from 0.7 GHz (Ultra High Frequency, UHF) to
100 GHz (Extremely High Frequency, EHF).
As found in Chapter 3 Numerical Simulation, a time-harmonic/continuous
wave (CW) radar signal at 8 GHz frequency whose wavelength is 0.0375 m (1.48 in)
can produce scattering signals due to the presence of a near-surface cubic-like defect
in a dielectric medium. Since the use of higher frequency CW signals can better detect
defects than lower frequency CW signals, CW radar signals in the frequency range
of 8 GHz to 18 GHz (8GHz 12GHz: X-band, and 12GHz 18GHz: Ku-band) at an
interval of 0.02 GHz were used in the laboratory radar measurements. Azimuth angle
ranging from 30 to 30 with 0 corresponding to perpendicular incidence ( = 90
were explored at an angular interval of 0.2 . Linearly polarized radar signals were
transmitted and received (HH- and VV-polarized). The time needed for collecting
the steady-state response per frequency per angle was 0.14 second [254].
As also found in Chapter 3 Numerical Simulation, two measurement schemes
were considered in this research; normal incidence scheme and oblique incidence
scheme, which are to be further explained in the following sections.
4.3.1
Monostatic ISAR Normal Incidence Scheme
Defined monostatic ISAR normal incidence scheme is illustrated in Figure 4-7. In

Figure 4-7, the synthetic aperture plane is always perpendicular to the longitudinal
axis of the cylinder. For cylinders of perfect geometry the measurements should be
independent of azimuth angle, suggesting a theoretically circumferential symmetry
of ISAR normal incidence measurements. Additionally, the monostatic ISAR normal incidence measurements capture the specular returns from the cylinder at all
azimuth/incident angles since the radar is aligned with the center of the cylinder.
The default range of azimuth angle in the measurements is [60 , 120 ] [ 3 ,
2
],
3
equivalently n [30 , 30 ] where n is normalized to the center of the aperture.

120
or
4.3.2
Monostatic ISAR Oblique Incidence Scheme
Defined monostatic ISAR oblique incidence scheme is sketched in Figure 4-8. In Figure 4-8, the longitudinal axis of the cylinder is on the synthetic aperture plane. For
cylinders of perfect geometry the measurements are not circumferentially symmetrical. Unlike in the normal incidence scheme, the monostatic ISAR oblique incidence
measurements capture the specular returns from the cylinder only when equals
90 ( 2 ) which is the case of normal/perpendicular incidence. The difference in the
pattern of specular returns is believed to favor the revealing of scattering signals due
to defects. The default range of azimuth angle in the measurements is [60 , 120 ]
, which is equivalent to a range of [30 , 30 ], or equivalently n [30 , 30 ] where
n is normalized to the angle at perpendicular incidence ( = 90 ).
4.4
Calibration of Laboratory Radar Measurements
Laboratory radar measurements were calibrated using a perfect electric conductor

(PEC) tube specimen made of Aluminum and a lossy dielectric rod specimen made of
plexiglass in the normal incidence scheme. Two approaches applied in this calibration
are described as follows.
1. Second-order statistics In the normal incidence scheme, the circumferential
symmetry in the measurements can be used to evaluate the measurement quality and the geometrical perfectness of a cylinder specimen. For this purpose,
second-order statistics (the use of mean , standard deviation s , and correla of a
tion coefficient ) is applied. In the frequency-angle measurement A(f, )
specimen featuring angular symmetry, its statistical properties are:
The mean amplitude at one angle over all frequencies should be identical
or close to the mean amplitude at another angle over all frequencies.
[A(f, i )] = [A(f, j )]
where A denotes the amplitude response.
121
(4.2)
The standard deviation of amplitude and phase responses at one frequency

over all angles equals or close to zero.
=0
s [A(fi , )]
(4.3)
=0
s [(fi , )]
(4.4)
where denotes the phase response.

The correlation coefficient of amplitude and phase responses evaluated
between any two angles over all frequencies equals or close to unity.
[A(
, i ), A(
, j )] = 1
(4.5)
[(f, i ), A(f, j )] = 1
(4.6)
These parameters quantitatively evaluate the circumferential/angular symmetry in the frequency-angle measurements of cylinder specimens in the normal
incidence scheme.
2. Optical backscattering analogy The far-field radar reflection response of a
dielectric circular/spherical scatterer can be analogically simulated by its optical
model. The optical backscattering model by Inada (1974) [118] and Stephens
et al. (1975) [124] is used in this analogy. This calibration was conducted by
Dennis Blejer at the MIT Lincoln Laboratory, and the result is provided here
at his courtesy.
4.4.1
PEC Specimen
An Aluminum tube specimen of 12 in. (15.24 cm) height, 3 in. (7.62 cm) radius , and
0.25 in. (0.635 cm) thickness used as an example of PEC was prepared by the laboratory staff in the MIT Lincoln Laboratory for the purpose of calibrating the compact
RCS/antenna range facility. Figure 4-9 (a) shows a picture of the tube specimen.
The scattering signals showing in Figure 4-9 (b) indicates a strong specular return
122
from the PEC specimen. The far-field monostatic ISAR normal incidence responses
(amplitude) of this PEC specimen were collected in X- and Ku-band for both HH
(TE) and VV (TM) polarizations, as shown in Figures 4-10 and 4-11.
Figure 4-12
(a) and (b) show the mean and standard deviation of amplitudes at X-band. Figure
4-13 (a) and (b) show the mean and standard deviation of phases at X-band. Figure
4-14 (a) and (b) show the mean and standard deviation of amplitudes at Ku-band.
Figure 4-15 (a) and (b) show the mean and standard deviation of phases at Ku-band.
Correlation coefficients of the radar measurements in X- and Ku-band are computed
and shown in Figures 4-16 and 4-17.
Table 4.4 lists the statistical parameters of
the far-field ISAR normal incidence frequency-angle amplitude measurements of the

PEC tube. It is found that (1) the fluctuation of [A(f, i )] indicates the changes of
and s [(fi , ]
are close
the surface properties of the PEC tube; (2) both s [A(fi , ]
to zero; and (3) both [A(
, i ), A(
, j )] and [(
, i ), A(
, j )] are close to unity.
It is found that, although a PEC specimen was used, there exists some differences
Table 4.4: Statistical parameters of the frequency-angle amplitude measurement of
the Aluminum tube (PEC)
Parameter
Mean
[A(f, i )]
Std. dev.
s [A(fi , )]
Corr. coef.
[A(f, i ), A(f, j )]
Theo.
value
0
1
Measured value
X-band: 2.48 1.8(HH) / 2.55 1.9(VV)
Ku-band: 1.85 0.75 (HH) / -2.18-1 (VV)
X-band: 0.124 0.277(HH) / 0.052 0.396(VV)
Ku-band: 0.239 0.549(HH) / 0.156 0.658(VV)
X-band: 0.99838(HH) / 0.99497(VV)
Ku-band: 0.98606(HH) / 0.96593(VV)
between theoretical values and measured values in Table 4.4. The differences are
attributed to the following factors:
Geometrical imperfectness of the specimen The manufactured tube
specimen may not be a perfect cylinder in shape, owing to the limited precision of manufacturing. Local surface roughness of the tube specimen may also
contribute to the reflection response in steady-state.
123
Alignment error The center of the cylinder specimen is supposed to coincide

with the center of the Styrofoam tower in order to satisfy the condition of normal
incidence scheme in Figure 4-7. Should this alignment condition of centers is
not met, the specular return from the cylinder will be reduced.
Background noise Measurements made in the steady-state could amplify the
background noise from the imperfect reflectionless boundary condition. While
this type of error cannot be completely eliminated, its contribution is considered
to be secondary when compared with other factors.
4.4.2
Lossy Dielectric Specimen and Its Optical Model
A lossy dielectric cylinder specimen made of plexiglass of 7.335 in. (18.63 cm) height
and 1 in. (2.54 cm) radius, as shown in Figure 4-18 (a), was prepared by the MIT LL
staff for calibration purpose. The laser scattering picture shows not only the specular return, but also the back reflection and surface wave from the lossy dielectric
specimen in Figure 4-18 (b). The plexiglass rod were manufactured with fibers oriented in parallel to the axis of the rod. The measured Ku-band amplitude response
is shown in Figure 4-19. In this measurement set, heterogeneity of the lossy dielectric specimen is due to the nonuniform distribution of fibers within the specimen,
resulting in the fluctuating reflection response shown in Figure 4-19 The second-order
statistical parameters of the amplitude response are listed in Table 4.5 in which the
circumferential heterogeneity of the specimen is suggested by an increased s [A(fi , )].

An optical model for estimating the backscattering waves and the surface creeping
waves of a lossy dielectric cylinder was developed by D. Blejer (MIT LL) and used
for the far-field ISAR normal incidence measurements of the plexiglass rod [118, 124].
This model produced the simulated far-field response of the plexiglass rod with an
dielectric constant (r =2.6) and an electrical conductivity (=0.03 mho/m) for the
plexiglass rod. Both simulated and measured (Figure 4-19) responses were processed
by a backprojection algorithm (to be further explained in Chapter 5 Image Reconstruction) to render the range-crossrange imagery as shown in Figure 4-20. In
124
Reflector 2
GFRP-concrete
specimen
Plane wave
Network
analyzer
Reflector 1
Stepped-freq.
radar and horn
antenna
Styrofoam
tower
Figure 4-6: Schematic of the compact RCS/antenna range facility in the MIT LL
Table 4.5: Statistical parameters of the frequency-angle response of the plexiglass rod
specimen
Parameter
Theo.
value
Mean [A(f, i )]
Std. dev.
s [A(fi , )]
Corr. coef.
1
[A(f , i ), A(f , j )]
Measured value
Ku-band: 8.5 4.7 (HH) / 6.8 3.25 (VV)
Ku-band: 0.683 1.808(HH) / 0.609 1.64(VV)
Ku-band: 0.98937(HH) / 0.98204(VV)
125
Horn antenna
x
GFRP-concrete
specimen
q = 60~120
qint/2
r
GFRP-concrete
specimen
Styrofoam
tower
Top view
Front view
Figure 4-7: Far-field monostatic ISAR normal incidence scheme. Note that the angular zero is referred to a selected point on the cylinder.
Figure 4.6, the images produced by measured radar response using HH- and VVpolarized signals approximate the ones by simulated far-field optical response. This
indicates a good agreement between theory (optics) and measurement (the compact
RCS/antenna range facility). Table 4.6 also lists the peak RCS of simulated and
measured responses. It is found that the measured response demonstrates satisfactory approximation to the theoretical values predicted by an optical model, both
qualitatively (Figure 4-20) and quantitatively (Table 4.6). This example, along with
the PEC specimen, validates the reliability of laboratory radar measurements used in
this dissertation.
Table 4.6: Peak RCS of simulated and measured responses of the plexiglass rod
specimen
Polarization
Simulated RCS (dBsm)
Measured RCS (dBsm)
HH
-11.7
-11.2
VV
-7.2
-7.1
126
Horn antenna
x
f = 60~120
GFRP-concrete
specimen
qint/2
r
Styrofoam
tower
GFRP-concrete
specimen
Front view
Top view
Figure 4-8: Far-field monostatic ISAR oblique incidence scheme. Note that the angular zero is in alignment with the axis of the cylinder.
4.5
Frequency-Angle Measurements
Selected laboratory radar measurements of the GFRP-concrete cylinder specimens

are reported in this section. Only the amplitude response is provided. The measured reflection responses of GFRP-concrete cylinders are the combined result of the
following factors:
Variation in the dielectric properties of the cylinder The magnitude
of reflected radar signals is a function of the dielectric properties of the medium
(cylinder). Variation in the dielectric properties in different parts of the cylinder
can result in the change of reflected radar signals.
Variation in the surface roughness of the cylinder Theoretically, the
incident angle should be the differential angle between the incident wave vector
and the normal vector perpendicular to the surface of the cylinder on the lineof-sight of the radar. Should there be any change in the surface geometry
(roughness or smoothness) of the cylinder, such condition is not preserved,
leading to the change of reflected radar signals.
Magnitude of misalignment In the normal incidence scheme, the incident
wave vector is supposed to be perpendicular to the normal vector of the surface
127
Specular
(a) Picture of the Aluminum tube
12
(b) Light scattering from the Aluminum tube
Figure
4-9:HHPEC
Aluminum tube
fname:
CYL03,
Pol., specimen
max = 4.77dBsm
tory]
0
11.5
[Courtesy of the MIT Lincoln Labora12
11.5
-2
-2
11
10.5
-4
10
-6
9.5
9
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
8.5
8
60
-8
8.5
70
80
90
(deg)
(deg)
100
110
120
-10
(a) HH (TE)
8
60
70
80
90
(deg)
(deg)
100
110
(b) VV (TM)
Figure 4-10: Frequency-angle response of the Aluminum tube Amplitude (dBsm),

X-band
128
120
-10
fname: CYL03K, VV Pol., max
fname: CYL03K, HH Pol., max = 5.92 dBsm

18
Frequency (GHz)
16
-4
15
-6
14
-8
13
70
80
90
(deg)
(deg)
100
110
120
-2
16
-4
15
-6
14
-8
13
12
60
-10
70
(a) HH (TE)
80
90
(deg)
(deg)
100
110
(b) VV (TM)
Figure 4-11: Frequency-angle response of the Aluminum tube Amplitude (dBsm),

Ku-band
-0.5
HH (TE)
VV (TM)
HH (TE)
VV (TM)
0.35
-1
Standard deviation,
12
60
17
-2
Mean, (dBsm)
Frequency (GHz)
17
5.76 dBsm
18
-1.5
-2
-2.5
-3
-3.5
0.3
0.25
0.2
0.15
0.1
10
Frequency (GHz)
11
12
(a) [A]
10
Frequency (GHz)
11
12
(b) s[A]
Figure 4-12: Mean amplitudes and their standard deviations of the reflection response
at different frequencies Aluminum tube, X-band
129
120
-10
HH (TE)
VV (TM)
-1.9
0.96
0.95
Standard deviation,
Mean, (dBsm)
-2
-2.1
-2.2
-2.3
0.94
0.93
0.92
0.91
-2.4
0.9
-2.5
-30
HH (TE)
VV (TM)
0.89
-20
-10
0
(deg)
10
20
30
-30
-20
-10
10
20
30
(deg)
(a) [A]
(b) s[A]
at different incident angles Aluminum tube, X-band
0.65
HH (TE)
VV (TM)
0.6
0.55
Standard deviation,
Mean, (dBsm)
-1
-1.5
-2
0.5
0.45
0.4
0.35
0.3
0.25
-2.5
12
HH (TE)
VV (TM)
0.2
13
14
15
16
Frequency (GHz)
17
18
(a) [A]
12
13
14
15
16
Frequency (GHz)
17
18
(b) s[A]
at different frequencies Aluminum tube, Ku-band
130
-0.8
HH (TE)
VV (TM)
0.55
Standard deviation,
Mean, (dBsm)
-1
-1.2
-1.4
-1.6
-1.8
0.5
0.45
0.4
HH (TE)
VV (TM)
-2
-30
-20
-10
0
(deg)
10
20
30
0.35
-30
-20
-10
10
20
30
(deg)
(a) [A]
(b) s[A]
at different incident angles Aluminum tube, Ku-band
(a) [A] HH
(b) [A] VV
Figure 4-16: Correlation coefficients of the reflection response at different frequencies

Aluminum tube, X-band
131
(a) [A] HH
(b) [A] VV
Figure 4-17: Correlation coefficients of the reflection response at different frequencies

Aluminum tube, Ku-band
Specular
7.335 in
Back
reflection
Surface
wave
2 in
(a) Schematic of the plexiglass rod
(b) Light scattering from the

plexiglass rod
[Courtesy of the MIT LL]
Figure 4-18: Lossy dielectric specimen Plexiglass rod
132
fname: CYL08K, HH Pol., max
4.33 dBsm
18
17
-4
15
-6
14
-8
13
12
60
70
80
90
(deg)
(deg)
100
110
120
-10
17
-2
16
1.81 dBsm
18
Frequency (GHz)
Frequency (GHz)
-2
16
-4
15
-6
14
-8
13
12
60
70
80
(a) HH (TE)
90
(deg)
(deg)
100
110
120
(b) VV (TM)
Figure 4-19: Frequency-angle response of the plexiglass specimen Amplitude

(dBsm), Ku-band
of the cylinder. This condition will be violated if the center of the cylinder
does not coincide with the center of the measurement tower in laboratory radar
measurements. Misalignment will be further explained later in this chapter.
Analysis and discussions are provided on the topics of damage detection and rebar
detection in both normal and oblique incidence schemes.
4.5.1
Monostatic ISAR Normal Incidence Scheme
Amplitude responses of the specimens measured in the far-field monostatic ISAR

normal incidence scheme are provided in this section, including the measurements
of specimens CON, GFRP, CRE, 4RE, AD1B, AD1F, AD2, and AD3. Table 4.7
describes the signal content of each specimen, as well as the use of each specimen
in the discussion. In Table 4.7, DD denotes Damage Detection and RD denotes
Rebar Detection.
In what follows, effects of the presence of GFRP layer and the heterogeneity of
concrete in the far-field monostatic ISAR normal incidence measurements (or normal
incidence measurements) are discussed. The use of normal incidence measurements
133
-10
Simulated: lo rhi= 2.6, = 0.03
Measured
= 22.9183 , [flo,fhi] = [12,18] GHz,
0.8
-10
0.8
-10
0.6
-20
0.6
-20
0.4
-30
0.4
-30
0.2
-40
0.2
-40
-50
-50
-0.2
-60
-0.2
-60
-0.4
-70
-0.4
-70
-0.6
-80
-0.6
-80
-90
-0.8
-1
-1
meters (m)
Range
meters
Range (m)
-0.5
0
meters
0.5
-90
-0.8
-1
-1
-0.5
Cross-range (m)
Simulated: r = 2.6, = 0.03
0.8
-10
0.8
-10
0.6
-20
0.6
-20
0.4
-30
0.4
-30
0.2
-40
0.2
-40
-50
-50
-0.2
-60
-0.2
-60
-0.4
-70
-0.4
-70
-0.6
-80
-0.6
-80
-90
-0.8
-0.8
-1
-1
-0.5
0
meters
0.5
Range
meters(m)
meters
0.5
Measured
Range (m)
0
meters
Cross-range (m)
-1
-1
-90
-0.5
0
meters
0.5
Cross-range (m)
Cross-range (m)
Figure 4-20: Simulated and measured responses Plexiglass rod, HH (upper level)
and VV (lower level) polarizations [Courtesy of the MIT LL]
134
for damage and rebar detections is discussed. Measurements on the mechanicallydamaged are also reported and discussed.
Table 4.7: Signal contents of the radar measurements of laboratory specimens
Specimen Signal content
Usage
CON
Splain concrete
DD, RD
GFRP
Splain concrete + SGFRP
DD
CRE
Splain concrete + Sone center rebar
RD
4RE
Splain concrete + Sfour near-surface rebars
RD
AD1B
Splain concrete + SGFRP
DD
AD1F
Splain concrete + SGFRP + Sartificial defect AD1
DD
AD2
DD
AD3
DD
Effects of GFRP and Concrete Heterogeneity

For the understanding of effects of GFRP and concrete heterogeneity in normal incidence measurements, two specimens (CON, GFRP) are considered. Specimen CON
is a plain concrete cylinder, and the specimen GFRP is a plain concrete cylinder
wrapped with one layer of GFRP sheet. Radar measurements made on the specimen
CON are shown in Figures 4-21 and 4-22. Figure 4-23 shows the radar measurements
of the specimen GFRP.
135
fname: CYL05, HH Pol., max = -3.34 dBsm

12
fname: CYL05, VV Pol., max
3.77 dBsm
12
11.5
11.5
-2
-2
11
10.5
Frequency (GHz)
Frequency (GHz)
11
-4
10
-6
9.5
10.5
-4
10
-6
9.5
9
-8
-8
8.5
8.5
8
60
70
80
90
(deg)
(deg)
100
110
120
8
60
-10
70
(a) HH (TE)
80
90
(deg)
(deg)
100
110
120
-10
(b) VV (TM)
Figure 4-21: Frequency-angle response of specimen CON (plain concrete) Amplitude

(dBsm), X-band
fname: CYL05K, HH Pol., max
1.77 dBsm
18
17
16
-4
15
-6
14
12
60
-8
70
80
90
(deg)
(deg)
100
110
120
17
-2
13
2.37 dBsm
18
Frequency (GHz)
Frequency (GHz)
-2
16
-4
15
-6
14
-8
13
-10
(a) HH (TE)
12
60
70
80
90
(deg)
(deg)
100
110
(b) VV (TM)
Figure 4-22: Frequency-angle response of specimen CON (plain concrete) Amplitude

(dBsm), Ku-band
136
120
-10
fname: CYL06, HH Pol., max = -3.35 dBsm

12
3.79 dBsm
12
11.5
11.5
-2
-2
11
10.5
Frequency (GHz)
Frequency (GHz)
11
-4
10
-6
9.5
9
10.5
-4
10
-6
9.5
9
-8
-8
8.5
8
60
8.5
70
80
90
(deg)
(deg)
100
110
120
8
60
-10
(a) HH (TE)
70
80
90
(deg)
(deg)
100
110
(b) VV (TM)
Figure 4-23: Frequency-angle response of the specimen GFRP (plain concrete with
GFRP) Amplitude (dBsm), X-band
In Figures 4-21 and 4-22, the circumferentially symmetry of the reflection response
from the plain concrete cylinder is observed. Unlike other homogeneous media, the
heterogeneity of concrete as a cementitious composite material produces a locationdependent radar response which also varies with respect to the incident angle. This
heterogeneity is sensible by the chosen radar signal in the X- and Ku-band whose
half-wavelength ranges from 1.87 cm (0.74 in, 8GHz) to 0.83 cm (0.32 in, 18GHz)
and are comparable to the size of aggregates used in concrete. Table 4.8 lists the
maximum and minimum power response (in dBsm) values of specimens CON and
GFRP for both HH and VV polarizations.
Table 4.8: Maximum and minimum powers (dBsm) of specimens CON and GFRP,
X-band
Polar.
CON(plain concrete)
GFRP(plain concrete with GFRP)
HH
Max.=0.6808 / Min.=0.2503
Max.=0.68 / Min.=0.2286
VV
Max.=0.6479 / Min.=0.216
Max.=0.6464 / Min.=0.2101
The effect of the GFRP wrap in the normal incidence scheme can also be illustrated
137
120
-10
by comparing the radar measurements of specimens CON and GFRP. Figure 4-24
shows the selected power responses (dBsm) of specimens CON and GFRP in HH
polarization. It can be seen that the presence of the GFRP wrap slightly reduces the
level of radar reflection. Same result is observed in VV polarization. In Figure 4-24,
local fluctuations in the power response of specimens CON and GFRP is the result of
not only the surface roughness of the cylinder but also the material heterogeneity of
the cylinder, while global variation may be caused by the misalignment of cylinder on
the measurement tower. Figure 4-25 shows the surface roughness (porous surface) of
the specimen CON. Figure 4-26 illustrates the weakening of reflected power response
due to misalignment; the misaligned specimen does not produce a perfectly normal
reflection after the rotation of the tower. It is noteworthy to point out that the
measured power response is the combined result of all factors.
Damage Detection
In the damage detection using far-field monostatic ISAR normal incidence measurements six specimens (CON, GFRP, AD1B, AD1F, AD2, AD3) are considered. Figures
4-27 and 4-28 are the far-field radar responses of the intact and damaged (with an
artificial defect) sides of the specimen AD1 in X-band. In Figure 4-27, while no defect is presented, variation of the power response in the circumferential direction is
still observed. Table 4.9 lists the extreme values of the measurements on the specimen AD1. Although the presence of a local defect globally increases the power level
of far-field ISAR normal incidence measurements as the experimental data suggest,
it should be noted that these extreme values do not occur at same locations. The
angular pattern of the power fluctuation is better illustrated in Figure 4-29.
In Figure 4-29, fluctuation of power response curve representing the specimen
GFRP-con. (intact side of AD1) is attributed to the surface roughness and the material heterogeneity of the specimen. There is a descending region around 20 and
30 in the GFRP-con. curve. A careful investigation on the specimen AD1 reveals
a concave concave on the surface of the specimen as shown in Figure 4-30. The distribution of this concave region matches the descending region in the curve around
138
Ref. power (dBsm) Ref. power (dBsm) Ref. power (dBsm) Ref. power (dBsm)
Ref. power (dBsm)
[ 8 GHz ]
Con.w.GFRP
Con.
1
0.5
0
-30
-20
-10
0
[ 9 GHz ]
10
20
30
-20
-10
0
[ 10 GHz ]
10
20
30
-20
-10
0
[ 11 GHz ]
10
20
30
-20
-10
10
20
30
0
10
Incident angle (deg.)
20
30
1
0.5
0
-30
1
0.5
0
-30
1
0.5
0
-30
[ 12 GHz ]
1
0.5
0
-30
-20
-10
Figure 4-24: Comparison between the power responses of specimens CON and GFRP,
X-band
139
= 30
= 0
= 0
= 30
Figure 4-25: Photos of the specimen CON showing surface roughness
140
30 . On the other hand, the curve representing the damaged side of the specimen
AD1 has a descending region around 0 , while there is no any surface irregularity
at the location corresponding to 0 except the embedded artificial defect underneath
the GFRP wrapping. The descending becomes most obvious at frequencies around
10 GHz to 11 GHz whose wavelengths (1.18 in. to 1.07 in.) are approximate to the
depth of the defect (1 in) (the half-wavelength of radar signals between 8 GHz and
12 GHz ranges from 0.73 in. to 0.49 in.). This phenomenon suggests a possibility of
discovering local near-surface defect in GFRP-retrofitted concrete columns, although
other factors leading to a false alarm result must be excluded, such as surface irregularity. Radar measurements collected in Ku-band are reported in the Appendix.
Table 4.9: Maximum and minimum RCS (dBsm) of the specimens AD1
Intact side of AD1
Damaged side of AD1
HH
Max.=0.5439 / Min.=0.3327
Max.=0.6615 / Min.=0.3908
VV
Max.=0.5321 / Min.=0.3479
Max.=0.6412 / Min.=0.4027
141
Monostatic
radar
Center of the
specimen
Styrofoam
tower
Misalignment
Misalignment
Cylinder
specimen
Center of the
tower
(a) Before rotation
(b) After rotation
Figure 4-26: Misalignment between the cylinder specimen and the Styrofoam tower
in the normal incidence scheme
fname: CYLAD1BV, HH Pol., max = -5.29 dBsm

12
fname: CYLAD1BV, VV Pol., max = -5.48 dBsm

12
11.5
11.5
-2
-2
11
10.5
-4
10
-6
9.5
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
8.5
8
-30
-8
8.5
-20
-10
0
(deg)
(deg)
10
20
30
-10
(a) HH (TE)
8
-30
-20
-10
0
(deg)
(deg)
10
20
(b) VV (TM)
Figure 4-27: Frequency-angle response of specimen AD1B Amplitude (dBsm), Xband

142
30
-10
fname: CYLAD1FV, HH Pol., max = -3.59 dBsm

12
fname: CYLAD1FV, VV Pol., max

12
11.5
3.86 dBsm
0
11.5
-2
-2
11
10.5
-4
10
-6
9.5
9
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
8.5
8
-30
-8
8.5
-20
-10
0
(deg)
(deg)
10
20
30
-10
8
-30
-20
-10
(a) HH (TE)
0
(deg)
(deg)
10
20
(b) VV (TM)
Figure 4-28: Frequency-angle response of specimen AD1F Amplitude (dBsm), Xband

Far-field monostatic ISAR normal incidence measurements of specimens AD2 and
AD3 are also reported here, as shown in Figures 4-31 and 4-32, both in X-band. Since
the intact side of the specimen AD2 was not measured, there is no available baseline
to be used as the intact response. The intact response from the specimen AD1 is not
relevant here because the size of AD2 is different from AD1. Nonetheless, in Figure
4-31, a weaker power response is found, centering at the 0 where the delamination
defect is embedded. A reasonable postulation suggests that this weakening in power
response is due to the scattering of the defect, after excluding the possibility of surface
irregularity via carefully checking the surface of the specimen AD2. Firm conclusion
can be obtained should the intact side of AD2 be measured for comparison. Similar
constraint occurs to the measurement on the specimen AD3. In Figure 4-32, a significant weaker region is observed at 0 where a strip defect is placed at the interface
between GFRP and concrete along the longitudinal axis of the cylinder. Surface irregularity has also been excluded after inspecting the specimen. No baseline data
can be used for classifying the fluctuations of power response into different causes,
although it seems obvious that the weaker region is caused by the presence of a strip
defect. In this case it is suggested that a reliable assessment should be conducted
143
30
-10
Ref. power (dBsm) Ref. power (dBsm) Ref. power (dBsm) Ref. power (dBsm)
Ref. power (dBsm)
[ 8 GHz ]
0.5
0.4
0.3
-30
GFRP-con.w.AD1
GFRP-con.
-20
-10
0
[ 9 GHz ]
10
20
30
-20
-10
0
[ 10 GHz ]
10
20
30
-20
-10
0
[ 11 GHz ]
10
20
30
-20
-10
10
20
30
0
10
20
30
0.5
0.4
0.3
-30
0.4
-30
0.4
-30
[ 12 GHz ]
0.6
0.4
-30
-20
-10
Figure 4-29: Comparison between the power responses of the intact and damaged
sides of the specimen AD1, X-band
144
= 0
[Max. depth of the concave region is

about 0.25 in. (0.635 cm).]
= 30
Figure 4-30: Photos showing the concave region in the specimen
145
via other supportive evidences, rather than solely based on the power reflection response. It is noted that the damage detection using far-field monostatic ISAR normal
incidence measurements relies not only on the half-wavelength of CW radar signals
(whether the signal can be interfered by the defect) but also on how the detectability
is defined. While the half-wavelength of a 8 GHz signal, 0.73 in., seems to be capable
of detecting a 1-in.-depth defect (specimen AD1), there is no clear indication of the
presence of defect in the reflection response at 8 GHz. Rather, clear indication is
observed in the neighborhood of 10 GHz and 11 GHz (Figure 4-27). However, care
must be taken when defining the indication using far-field monostatic ISAR normal
incidence measurements. Questions such as how much difference should be used as
a positive sign for the presence of defect? need to be answered in advance.
Another issue further complicating the detectability in field conditions is the presence of ambient noise. Without field measurements, it is not clear how the ambient,
background electromagnetic noise affects far-field monostatic ISAR normal incidence
measurements. Performance of de-noising techniques using compact (wavelets) and
non-compact (e.g., Fourier series) basis functions remains unknown in this particular
application.
Rebar Detection
In the rebar detection using far-field monostatic ISAR normal incidence measurements three specimens (CON, CRE, and 4RE) are considered. Figure 4-33 shows the
frequency-angle power reflection response/measurement of the specimen CRE (plain
concrete cylinder with a rebar at the center), and Figure 4-34 provides a comparison
between the power responses of specimens CON and CRE. Knowing that the curve
of the specimen CON is not perfectly ideal in Figure 4-34, the power response of the
specimen CRE is in average greater than the one of the specimen CON. As expected,
this is due to the presence of a PEC rebar in the center of the cylinder. The difference
is not significant since the rebar is embedded approximately 3 inches from the surface.
In the case of the specimen 4RE (four rebars buried at different depths from the
surface of a plain concrete cylinder), significant fluctuations of the power response are
146
fname: CYLAD2V, HH Pol., max = -3.71 dBsm

12
fname: CYLAD2V, VV Pol., max
3.14 dBsm
12
11.5
11.5
-2
-2
11
10.5
-4
10
-6
9.5
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
-8
8.5
8.5
8
-40
-20
0
(deg)
(deg)
20
40
-10
-40
-20
(a) HH (TE)
0
(deg)
(deg)
20
40
-10
(b) VV (TM)
Figure 4-31: Frequency-angle response of specimen AD2 Amplitude (dBsm), Xband

fname: CYLAD3, HH Pol., max = -2.43 dBsm
12
fname: CYLAD3, VV Pol., max
2.34 dBsm
12
11.5
11.5
-2
-2
11
10.5
-4
10
-6
9.5
9
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
8.5
-8
8.5
8
-40
-20
0
(deg)
(deg)
20
40
-10
8
-40
-20
(a) HH (TE)
0
(deg)
(deg)
20
(b) VV (TM)
Figure 4-32: Frequency-angle response of specimen AD3 Amplitude (dBsm), Xband
147
40
-10

12
2.96 dBsm
12
11.5
11.5
-2
-2
11
10.5
-4
10
-6
9.5
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
-8
8.5
8.5
8
60
70
80
90
(deg)
(deg)
100
110
120
8
60
-10
70
80
(a) HH (TE)
90
(deg)
(deg)
100
110
(b) VV (TM)
Figure 4-33: Frequency-angle response of specimen CRE Amplitude (dBsm), Xband

expected due to the configuration of local PEC scatterers (the four rebars). Configuration of the rebars is illustrated in Figure 4-2 and summarized in Table 4.10. Figures
4-35, 4-36, 4-37, and 4-38 show the power reflection responses of the specimen 4RE at
four angular regions; namely, RE1RE4. Four data ranges are consequently defined,
centering at the location where one rebar is buried. Mean RCS or power reflection
response at each data range is also reported.
It is found that the power reflection
Table 4.10: Configuration of the specimen 4RE

Rebar
Angle
Date range
Embed.
depth (in)
Mean RCS (dBsm)
135
[150 120 ]
0.25
HH=0.6375/VV=0.6265
45
[60 30 ]
1.5
HH=0.4830/VV=0.5388
45
[30 60 ]
HH=0.4833/VV=0.4801
135
[120 150 ]
0.5
HH=0.5281/VV=0.527
response exhibits a clear two-dimensional sinusoidal pattern of change when power

attenuation in concrete is small (the rebar is buried very close to the surface, RE1).
148
120
-10
Ref. power (dBsm)

Ref. power (dBsm) Ref. power (dBsm) Ref. power (dBsm)
Ref. power (dBsm)
[ 8 GHz ]
0.45
0.4
0.35
-30
Con.w.1 rebar
Con.
-20
-10
0
[ 9 GHz ]
10
20
30
-20
-10
0
[ 10 GHz ]
10
20
30
-20
-10
0
[ 11 GHz ]
10
20
30
-20
-10
10
20
30
0
10
20
30
1
0.5
0
-30
1
0.5
0
-30
1
0.5
0
-30
[ 12 GHz ]
1
0.5
0
-30
-20
-10
Figure 4-34: Comparison between the power responses of specimens CON and CRE,
X-band
149
fname: CYLRE1, HH Pol., max = -1.51 dBsm

12
fname: CYLRE1, VV Pol., max
2.11 dBsm
12
11.5
11.5
-2
-2
11
10.5
-4
10
-6
9.5
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
-8
8.5
8.5
8
-150
-145
-140
-135
-130
(deg)
(deg)
-125
-120
8
-150
-10
-145
(a) HH (TE)
-140
-135
-130
(deg)
(deg)
-125
-120
-10
(b) VV (TM)
Figure 4-35: Frequency-angle response of specimen 4RE (rebar 1) Amplitude

(dBsm), X-band
12
1.55 dBsm
12
11.5
11.5
-2
-2
11
10.5
-4
10
-6
9.5
9
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
8.5
8
-60
-8
8.5
-55
-50
-45
(deg)
(deg)
-40
-35
-30
-10
8
-60
-55
-50
(a) HH (TE)
-45
(deg)
(deg)
-40
-35
(b) VV (TM)

(dBsm), X-band
150
-30
-10

12
4.26 dBsm
12
11.5
11.5
-2
-2
11
10.5
-4
10
-6
9.5
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
-8
8.5
8.5
8
30
35
40
45
(deg)
(deg)
50
55
60
8
30
-10
35
(a) HH (TE)
40
45
(deg)
(deg)
50
55
60
-10
(b) VV (TM)

(dBsm), X-band
12
3.35 dBsm
12
11.5
11.5
-2
-2
11
10.5
-4
10
-6
9.5
9
Frequency (GHz)
Frequency (GHz)
11
10.5
-4
10
-6
9.5
9
-8
8.5
8
120
-8
8.5
125
130
135
(deg)
(deg)
140
145
150
-10
(a) HH (TE)
8
120
125
130
135
(deg)
(deg)
140
145
(b) VV (TM)

(dBsm), X-band
151
150
-10
When the power attenuation is significant (the rebar is buried at a depth), this pattern
becomes blur and unclear. This attenuation issue determines whether such pattern
can be used for rebar detection or not. The specimen representing the most extreme
case of power attenuation for rebar detection is the specimen CRE in which the rebar
is buried at a depth of 3 inches from the surface. There is almost no indication of
any sinusoidal pattern in the power response of the specimen CRE in Figure 4-33, as
can be observed in Figures 4-354-38.
In Table 4.10, the mean power response of the frequency-angle measurement in
each data range is also provided. It is expected that the deeper the rebar is buried from
the surface of concrete, the weaker the power response should be. This understanding
is in good agreement with the power response in HH polarization and generally correct
with the power response in VV polarization.
4.5.2
Monostatic ISAR Oblique Incidence Scheme
The amplitude and phase responses of the specimens measured in far-field monostatic
ISAR oblique incidence scheme are reported in this section. The original purpose for
using this measurement scheme is to alleviate the effect of specular returns in the
reflection response. Considered specimens include AD1 (GFRP-concrete cylinder
with a cubic defect) and AD2 (GFRP-concrete cylinder with a delamination defect)
(refer to Table 4.1 and previous figures for a detailed description to these specimens).
Both the intact surface (without artificial defect) and the damaged surface of the
specimen AD1 were measured, while only the damaged surface of the specimen AD2
was measured. Only X-band measurements are available. Specimens with rebars are
not measured in the oblique incidence scheme.
Figures 4-39 and 4-40 show the oblique incidence measurements of the intact and
damaged sides of the specimen AD1. It is seen that, the specular return dominates
the reflection response at azimuth angles around = 0 (2 2 ). The specular
return is reduced to a significant degree when the incident angle deviates from 0 ,
and the alleviation of specular return reveals the hidden scattering signals due to the
presence of defect. The spreading signals in Figure 4-40 indicates the presence of the
152
near-surface artificial defect in the specimen AD1 for both HH and VV polarizations.
Figure 4-40 demonstrates the effectiveness of the use of the oblique incidence scheme in
eliminating the specular return at different incident angles. The radar measurements
on the specimen AD2 in Figure 4-41 also shows scattering signals similar to those
obtained for the specimen AD1, suggesting the presence of the artificial defect in the
specimen AD2.
It should be noted that, the scattering signals due to defect are not continuously
distributed outside the specular-dominant zone in the oblique incident measurements.
Rather, variation of power amplitude with respect to the incident angle is found.
This is attributed to the sensitivity of monostatic ISAR measurements to the geometry
of a scatterer (the specimen or the defect) in the oblique incidence scheme; not only
the specular return but also the scattering signals varies with . This suggests the
existence of a set of incident angles for better inspection.
fname: CYLAD1FH, HH Pol., max = -5.09 dBsm

12
12
11.5
11.5
11
-5
10.5
10
-10
9.5
9
Frequency (GHz)
Frequency (GHz)
11
-15
-5
10.5
10
-10
9.5
9
-15
8.5
8.5
8
-30
-20
-10
0
(deg)
10
20
30
-20
8
-30
-20
-10
0
(deg)
10
20
Figure 4-39: Frequency-angle response of the intact side of the specimen AD1
Amplitude (dBsm), X-band
153
30
-20
fname: CYLAD1BH, HH Pol., max = -6.41 dBsm

12
11.5
11.5
-5
10.5
10
-10
9.5
9
11
Frequency (GHz)
11
Frequency (GHz)
12
-5
10.5
10
-10
9.5
9
-15
8.5
-15
8.5
8
-30
-20
-10
0
(deg)
10
20
30
8
-30
-20
-20
-10
0
(deg)
10
20
30
-20
Figure 4-40: Frequency-angle response of the damaged side of the specimen AD1
12
12
11.5
11.5
-5
11
10.5
10
-10
9.5
9
Frequency (GHz)
Frequency (GHz)
11
-15
8.5
8
-5
10.5
10
-10
9.5
9
-15
8.5
-40
-20
0
(deg)
(deg)
20
40
-20
(a) HH (TE)
8
-40
-20
0
(deg)
(deg)
20
(b) VV (TM)
Figure 4-41: Frequency-angle response of the damaged side of the specimen AD2
154
40
-20
4.6
Summary
This chapter reports the laboratory radar measurements conducted in far-field monostatic ISAR normal and oblique incidence schemes. Steady-state, complex responses
in terms of amplitude and phase were collected for each harmonic frequency (f ) and
azimuth angle ( in normal incidence or in oblique incidence). The frequency-angle
measurements were obtained. Research findings are summarized as follows.
Far-field monostatic ISAR normal incidence measurements
1. Variations in the material composition and the dielectric properties of the
components of concrete lead to the fluctuation of measured reflection power
responses at differen incident angles. The measured power responses essentially collect the specular returns from the specimens in the normal
incidence scheme. Additionally, changes in the geometry of the cylinder
specimen can also result in the variation of power response.
2. Presence of the externally bonded GFRP sheet slightly reduces the total
power response of the GFRP-wrapped concrete cylinder (Figure 4-24).
This is attributed to the lossness of GFRP in the steady-state response,
which consumes a small amount of incident energy.
3. Presence of a local, artificial defect (cavity) in the GFRP-retrofitted concrete cylinder may produce a weakened power response at the right incident
angle (Figure 4-27), due to the local scattering of signals at the edges of the
rectangular defect. However, considering other possible causes such as the
geometry of the cylinder, it is not recommended to use power response as
the only information for asserting the presence of defect. In other words,
changes in the reflected power response can be attributed to the presence
of defect or to other causes, considering the strong influence of specular
returns in the normal incidence scheme.
4. Alignment accuracy is important in the far-field monostatic ISAR norma
incidence measurements. Misalignment (Figure 4-26) can cause global,
155
significant change of the power response. This is mainly due to the use of
monostatic radar.
5. From the results on specimens AD1, AD2, and AD3, it is believed that
a reliable assessment should be achieved via the use of power reflection
response (frequency-angle measurement) and other supportive evidences;
power reflection response in the normal incidence scheme may be used as
an indicator, but not as a conclusive tool.
6. A sinusoidal pattern of change is observed in the frequency-angle measurements of the specimen 4RE, accounting for the scattering from the rebars.
This pattern decays when the depth of rebar location increases.
7. Based on the findings in the normal incidence scheme, it is believed that
the far-field ISAR normal incidence measurements can not be exclusively
used for determining the presence of near-surface defects, due to the strong
influence of specular returns in the measurements. On the other hand, normal incidence measurements of the specimen 4RE (concrete cylinder with
four rebars) demonstrate a clear pattern different from the measurements
of specimens without rebar. This suggests the feasibility of using the normal incidence scheme for rebar detection.
Far-field monostatic ISAR oblique incidence measurements
1. In Figures 4-39 and 4-40, the specular return dominates the reflection
response at azimuth angles around = 0 (2 2 ). The specular return
is reduced to a significant degree when azimuth angle deviates from 0 ,
and the alleviation of specular return reveals the hidden scattering signals
due to the presence of defect.
2. Since the monostatic ISAR measurements are sensitive to to the geometry
of a scatterer (the specimen or the defect) in the oblique incidence scheme,
not only the specular return but also the scattering signals varies with the
incident angle . This suggests the existence of a set of incident angles for
better inspection.
156
3. From the Figures 4-39 and 4-40, it is found that the importance of alignment has been reduced in the oblique incidence measurements. It is because that the measurement surface is no longer always normal to the
line-of-sight of the radar;. Although misalignment in the neighborhood
of perpendicular incidence ( 0 ) can reduce the intensity of scattering
signals due to defect, it is not the angular range which the oblique incidence scheme rely on for damage detection. Other than the perpendicular
incidence region, misalignment will not produce a significant impact to the
intensity of defect signals since the background specular return has been
removed. This feature offers the oblique incidence scheme another benefit
as it is more robust than the normal incidence scheme.
4. Based on the findings in the oblique incidence scheme, it is believed that
the far-field ISAR oblique incidence measurements can be used for detecting near-surface defects such as GFRP debonding and delamination. The
alleviation of specular returns in such measurement scheme encourages the
revealing of scattering signals due to defect.
For the purpose of condition assessment, collected frequency-angle measurements
(far-field monostatic ISAR measurements) of GFRP-concrete cylinders are further
processed by the image reconstruction algorithm in order to render a spatial profile
of the cylinders. Principles of signal processing are described in Chapter 5 Image
Reconstruction.
157
158
Chapter 5
Image Reconstruction
Between the matter of the physical image and its object there is a very
different relation: they resemble each other. What should we understand
by this?
The Imaginary, Jean-Paul Sartre (19051980)
The purpose of this chapter is to provide a theoretical background and principles of

the image reconstruction method used for processing the collected far-field ISAR measurements (Chapter 4 Laboratory Radar Measurements) in this research. The
image reconstruction method is based on tomographic reconstruction (TR) methods
[203, 129] and implemented by the fast backprojection algorithm [176, 194, 278].
In this chapter, classical formulation of synthetic aperture radar (SAR) and the
backprojection algorithm in the fields of radar engineering and remote sensing is
provided for demonstrating the process of transforming inverse SAR (ISAR) signals
to backprojection images from a theoretical perspective. This image reconstruction
processing is the basis for producing the images used in Chapter 7 Condition
Assessment of GFRP-concrete Systems FAR NDT.
Credits for developing fundamental equations in SAR [261, 47, 121]and the backprojection algorithm [68, 239, 67, 7, 194, 240] should go to the researchers in both
fields. The authors contributions lay in (1) deriving the form of backprojection images using the ISAR-processed signal with a support band analysis, and (2) identifying
159
the importance of dielectric properties in the formulation regarding image resolution.

In this chapter, a brief introduction to traditional radar and SAR/ISAR measurements is first provided, followed by the principle of backprojection algorithms. A
support band analysis using the method of stationary phase for determining a band
region in which the spatial spectrum of backprojection images is non-trivial. Coding procedure for the implementation of backprojection algorithms is also provided.
Effects of aperture size and frequency bandwidth in the performance of the image
reconstruction method are discussed.
5.1
Single Scattering and Synthetic Aperture Radar
Consider the planar scattering problem in a domain s containing N point scatterers

and an incident wave with unit amplitude to be [261]

1
exp ik r
r
inc (
r) =
(5.1)
where ki = kix x kiy y is the incident wave vector, r is the relative position vector
from the radar to any observation point, and i = 1 is the imaginary number. The
scattered field from scatterer j at rj and observed at r is [261]

sj ks , ki
scat,j (
r, rj ) = s,j (
r, rj ) =
|
r rj |
exp (ik |
r rj |) inc (
r)
(5.2)
where ki = kix x kiy y is the incident wave vector, r is the relative position vector from

the radar to any observation point, i = 1 is the imaginary number, sj = sj ks , ki

is the scattered amplitude at scatterer j due to an incident wave of ki , observed at
r. When N scatterers are presented, the total scattered field observed at r is the
summation of the scattered fields from all scatterers. It can be approximated to be
scat (
r) = s (
r) =
N
X
j=1
sj
ks , ki
|
r rj |
160

exp (ik |
r rj |) inc (
r)
(5.3)
where ks = ksx x + ksy y is the scattering direction vector. ks = ki when the radar
operates in monostatic mode. Note that in Eq.(5.3) interactions among scatterers are
neglected as the result of approximation (independent scattering approximation).
The spatial resolution in traditional radar imaging is limited to its frequency and
bandwidth, while the amplitude and phase of scattered waves can both be measured.
To remedy this disadvantage, the concept of synthetic aperture radars (SAR) is introduced in which the spatial resolution is improved by phase compensation (focusing).
Several major SAR modes are listed and compared in Table 5.1. In Table 5.1, flight
path is the shape of synthetic aperture, and footprint is the area of radar investigation. Other SAR modes and their derivatives can be found in the radar literature.
[180, 58, 240] The configuration of SAR measurements is illustrated in Figure 5-1. For
Table 5.1: Comparison of several SAR modes

SAR mode
Flight path
Footprint
Spotlight
Straight line
Circular
Strip
Straight line
Strip
Inverse
Circular
Circular
SAR measurements using monostatic radar located at rs = (xs , ys , zs ), conducting at

constant zs and at constant flight speed v Figure 5-1, we have
zs = Rs cos i = h
(5.4)
ys = Rs sin i = h tan i
(5.5)
xs = xs0 + vt
(5.6)
where Rs = |
rs | is the range distance, i is the inclination angle with respect to the z
axis, h is the height of the radar, and xs0 = xs (t = 0) is the initial x position of the
monostatic radar. For ISAR measurements conducting at constant zs and at constant
161
flight speed v, we have

zs = Rs cos i = h
(5.7)
vt
= Rs sin i sin(t)
Rs sin i

vt
xs = Rs sin i cos
= Rs sin i cos(t)
Rs sin i
ys = Rs sin i sin
(5.8)
(5.9)
v
is the angular velocity of the radar. Assuming a constant propRs sin i
agation time of radar signals in the measurements, the distance variables (e.g., rs ,
where =
rj ) are exchangeable with propagation time. Therefore, inc (

rs , rj ) = inc (, t) and
s (
rs , rj ) = s (, t). Assuming that G contains N scatterers, the received scattered
wave due to the presence of N scatterers is
s (, t) =
N
X
s,j (, t)
(5.10)
j=1
where
s,j (, t) =
fj (ks , ki )
exp [ikRs,j ] inc (
rj )
Rs,j
Rs,j = |
rs rj |
(5.11)
(5.12)
and fj (ks , ki ) = fj (ki , ki ) is the backscattering amplitude of the nth scatter. Without losing generality, consider the case of single scatterer and its response in the farfield region (|
r rj | |
r| = r for amplitude approximation and |
r rj | r + rj cos
for phase approximation). The size of the circular footprint is assumed to be Gaussian
amplitude with radius g (Figure 5-4(b)). The incident wave on the footprint surface
is
inc (
rs , rj ) =
w(
rj )
exp [iki Rs,j ]
Rs,j
(5.13)
where w(
rj ) is a Gaussian amplitude distribution function at the scatterer rn =
(xn , yn , zn ) G , Rs,j is the distance between rj and the radar location rs and
p
Rs,j = |
rs rj | = (xj xs )2 + (yj ys )2 + (zj zs )2 . The projected G on x y
162
plane is an ellipse with characteristic dimensions gx and gy as shown in Figure 5-2.
Rs
L
(5.14)
sec i Rs
L
(5.15)
gx =
gy =
Considering the footprint domain located at zj = 0, the Gaussian amplitude distribution function is

(xj xs )2 (yj ys )2
w(xj , yj , 0) = w(xj , yj ) = exp

gx
gy
(5.16)
Therefore,
s (, t) =
N
X
fj (ki , ki )
j=1
2
Rs,j

(xj xs )2 (yj ys )2
exp [2ikRs,j ] exp
gx2
gy2
(5.17)
The signal can be focused at a location r0 = (x0 , y0 , z0 ) by multiplying s (, t) by a

reference signal
0 (
r0 , , t) = exp [2ikRs,0 ]
(5.18)
where Rs,0 = |
rs r0 |. The focused signal becomes
F (
r0 , , t) = s (, t)0 (
r0 , , t)
(5.19)
Integrating F (
r0 , , t) over [c B, c + B] and t [t0 , tf ] provides the final
focused signal C(
r0 ) at r0 .
Ztf
C(
r0 ) =
c Z+ B
dt
t0
c B
Ztf
dF (
r0 , , t) =
c Z+ B
dt
t0
ds (, t)0 (
r0 , , t)
(5.20)
c B
where [to , tf ] and [c B, c +B] are the time window and frequency window of the
flight path in which s (, t) is collected and processed, respectively. In the far-field
163
region, Rs,j is approximately Rs . C(

r0 ) becomes
1
C(
r0 ) = 2
Rs
Z Z
dtd
N
X
fj (ki , ki ) exp [2ik(Rs,j Rs,0 )]
j=1

(xj xs )2 (yj ys )2
exp
gx2
gy2
(5.21)
The reason 0 (
r0 , , t) can focus C(
r0 ) at r0 is due to the phase 2k(Rs,j Rs,0 ).
(Rs,j Rs,0 )
r0 ) projects the scattering from a scatterer
= 0 when rn
= r0 . This way C(
near r0 and creates an image centered at r0 . The integral of C(
r0 ) is inherently
computationally-expensive since a two-dimensional integral over and t is required
for each r0 . To alleviate this disadvantage, approximation on the phase is typically
performed in order to utilize computationally-efficient transforms like fast Fourier
transform (FFT). Considering in the region where rn
= r0 , Rs,j
= Rs,0 . Also, in the
far-field region, Rs,0
= Rs (Figure 5-3).
= Rs since Rs,j
Rs,j Rs,0 =
2
2
2
2
2
2
Rs,0
Rs,j
Rs,j
Rs,0
Rs,j
Rs,0
=
=
Rs,j + Rs,0
2Rs,0
2Rs
(5.22)
The phase in C(
r0 ) becomes
2
2
Rs,j
Rs,0
(
r0 , , t) = 2k(Rs,j Rs,0 )
=
Rs

k
=
[ x2j x20 + yj2 y02 + zj2 z02
Rs
+2xs (x0 xj ) + 2ys (y0 yj ) + 2zs (z0 zj )]
k
[x (xj + x0 ) y (yj + y0 ) z (zj + z0 )
Rs
(5.23)
(5.24)
(5.25)
where
x = x0 xj
(5.26)
y = y0 yj
(5.27)
z = z0 zj
(5.28)
164
Considering the configuration of SAR measurements, the phase (

r0 , , t) can be
represented by

(
r0 , , t) = k aj + bj t
(5.29)
where
1
[x [2xs0 (xj + x0 )]
Rs
+y [2ys (yj + y0 )] + z [2zs (zj + z0 )]]
aj =
bj = 2x
v
t =
t
Rs
(5.30)
(5.31)
(5.32)
With these relations, the focused signal C(

r0 ) becomes
c
C(
r0 ) =
vRs
vt
ZRs
c + B
Zc
dk
c B
c
dt
N
X

fj (ki , ki ) exp ik aj + bj t
j=1

(xj xs0 tRs )2 yj2
exp
2
gx2
gy
(5.33)
The flight speed v of the monostatic radar is constrained by the size of the footprint
G (gx and gy ) and the Gaussian amplitude distribution function. In the configuration
of SAR measurements (Figure 5-1), the maximum movement of radar is related to
the size of the footprint gx by
gx xs0
Rs
|xs0 + vt| gx
vt
gx xs0
Rs
Rs
165
(5.34)
Since gx =
Rs ,
L
xs0
vt
L
Rs
Rs
xs0

t
L
Rs
xs0
L
Rs
xs0
L
Rs
(5.35)
Assuming that xj
r0 ) is then
= xs0 + tRs , the integrand in C(
c
C(
r0 ) =
vRs
c
=
vRs
c + B
Zc
c B
c
xs0
L Z Rs
c + B
Zc
dk
c B
c
dt
N
X
fj (ki , ki )
j=1
xs0

L
Rs
2

yj
exp ik aj + bj t exp 2 (5.36)
gy
N
X
2
yj
dk
fj (ki , ki ) exp 2
gy
j=1
xs0
L Z Rs

dtexp ik aj + bj t (5.37)
xs0
L
Rs
First, we evaluate the integrand over t.

xs0
L Z Rs

dtexp ik aj + bj t = exp[ikaj ]
xs0
L
Rs
166
xs0
L Z Rs
xs0
L
Rs
dtexp[ikbj t]
xs0
exp[ikaj ]
Rs
exp[bj t]| L
=
xs0
ikbj

L
Rs

xs0
exp[ikaj ]
exp ikbj
exp ikbj
exp ikbj
=
ikbj
L
L
Rs

exp[ikaj ]
xs0
=
2i sin kbj
exp ikbj
ikbj
L
Rs

xs0
2 exp[ikaj ]
sin kbj
exp ikbj
=
kbj
L
Rs
(5.38)
Therefore,
c
C(
r0 ) =
vRs
c + B
Zc
c B
c
N
X
2
yj
dk
fj (ki , ki ) exp 2
gy
j=1
2 sin kbj
L
kbj

xs0
exp[ikaj ] exp ikbj
Rs
(5.39)
Since C(
r0 ) is a correlation integral of F (
r0 , , t), it indicates the likeliness of seeing
a scatterer within the footprint. The correlation is defined by aj and bj . The aj
dependence is a exponentially decaying as a function of x, y, and z. The bj
sin x
dependence is 2x = 2 (x0 xj ) and has a
form. To positively assert the
x
presence of a scatterer, we need

sin kbj
L

1
(5.40)
and
kbj
L
bj
L
k
(5.41)
Since bj = 2x = 2(x0 xj ), xj should be within the neighborhood of

2 |x0 xj |
L
L
L
=
=
k
2
2
167
(5.42)
L
The azimuth or cross-range resolution (xr ) of SAR measurements is now where L is
2
L
the antenna size, improved from gx = Rs to . For the radar antenna operating in
L
2
the range of X-band (8 GHz12 GHz) and Ku-band (12 GHz18 GHz), L is roughly
1
m as an example, the far-field distance is
0.3 m. Take f = 12 GHz and =
40
10
approximately 10 m. The cross-range resolution xr is improved from
=0.83 m to
12
L
=0.15 m, a 5.53 times improvement. This example demonstrates the effectiveness
2
of SAR measurements on improving cross-range resolution xr .
c
Should we use the center frequency fc (or c = 2fc or kc =
) for the wave
c
number variable (k = ) and have the radar antenna start from xs0 = 0, we have
c

1
1
2
bj since k = 2. The correlation integrand C(

2 sin kbj
=
2 sin
r0 )
kbj
L
kc bj
L
can be written to be
C(
r0 ) =
c
vRs
B
kc +
Z c
kc
B
c
2
2 2 sin
bj
N
X
yj
L
dk
fj (ki , ki ) exp 2
gy
kc bj
j=1

2
2 2 sin
bj
N
X
y
c
L
j
=
fj (ki , ki ) exp 2
vRs j=1
gy
kc bj

exp[ikaj ]
Baj
2 sin
c
aj

(5.43)
The aj dependence is also constrained by

sin
Baj
c

1
Baj
c
aj
c
B
(5.44)
Knowing that
aj =
1
[x [(xj + x0 )] + y [2ys (yj + y0 )] + z [2zs (zj + z0 )]]
Rs
(5.45)
Assuming that the radar inspection is performed on a similar level with the footprint,
zs is then much smaller than ys . Consequently, x << ys and z << ys . aj is
168
dominated by ys .
1
[y [2ys (yj + y0 )]]
aj
=
Rs
1
2ys y
=
Rs
(5.46)
The aj dependence is then

1
c
2ys y
Rs
B
c Rs
y
2B ys
Since
(5.47)
(5.48)
ys
= sin i 1,
Rs
c
1
c Rs
2B ys
2B sin i
c
|y0 yj | sin i
2B
y = |y0 yj |
(5.49)
(5.50)
This is the limit of the range resolution of SAR measurements. In view of the form
of C(
r0 ), the SAR point response S(x, y sin i ) is usually written as

S(x, y sin i ) = sinc
y sin i
r

sinc
x
xr

(5.51)
sin x
.
x
For traditional radar imaging, the range resolution r is
where sinc(x) =
r gy =
Rs
sec i Rs =
L
L cos i
(5.52)
1
10
m, L =0.3 m, Rs =10 m, and i
=
= 0 (therefore zs << ys ), gy
=
40
12
c
0.83 m. For B= 4 GHz, the range resolution for SAR imaging is
=0.0375 m,
2B
showing a 22 times improvement. Additional improvement on r can be expected,
Use =
providing more bandwidth B. This example demonstrates the effectiveness of SAR

measurements on improving range resolution r .
169
z
Radar position
rs = ( xs , ys , zs )
Flight path
ki
y
Footprint domain G
x
(a) Synthetic aperture radar (SAR)
Footprint domain G
gx
Radar position
gy
Flight path
x
(b) SAR Plane view
Figure 5-1: Configuration of SAR measurements
170
z
rs = ( xs , ys , zs )
Monostatic radar
ki
i
Footprint domain G
rj = ( x j , y j , z j )
gx
gy
x
Figure 5-2: Monostatic radar and its footprint
z
rs = ( xs , ys , zs )
Monostatic radar
rs rj
rs r0
Footprint domain G
y
nth scatter
focused location
r0 = ( x0 , y0 , z0 )
rj = ( x j , y j , z j )
Figure 5-3: rs , r0 , and rj in the far-field region
171
5.2
Inverse Synthetic Aperture Radar
In this research, far-field ISAR measurements are adopted. ISAR is a circular mode
of SAR. Configurations of SAR measurements is illustrated in Figure 5-4. In ISAR
measurements, an incident wave vector is considered.
ki = sin i cos i x sin i sin i y cos i z
(5.53)
where i is the angle between ki and the z axis (Figure 5-4(a)) and i the angle
between ki and the x axis (Figure 5-4(b)). The radar is located at the range distance
of Rs and denoted by a position vector rs = (xs , ys , zs ) where
xs = Rs sin i cos i
(5.54)
ys = Rs sin i sin i
(5.55)
zs = Rs cos i = h
(5.56)
and Rs = |
rs |. Assuming a Gaussian amplitude distribution for the circular footprint
with radius g containing N scatterers (Figure 5-4(b)), the scattered field s () is
s () =
N
X
fj (ki , ki )
j=1
Rs2

(xj xs )2 (yj ys )2
exp [2ikRs ] exp
gx2
gy2
(5.57)
and the phase (

r0 , ) is
k
[x (xj + x0 ) y (yj + y0 ) z (zj + z0 )
(
r0 , )
=
Rs
(5.58)
In the far-field region, Rs sin i is usually much greater than the radius of G, g (Figure
5-4). Therefore, the phase (
r0 , ) is approximated by
(
r0 , )
=
2k
[xs x + ys y + zs z]
Rs
172
(5.59)
The focused signal C(

r0 ) can be written as
1
C(
r0 ) = 2
Rs
Z Z
N
X

2ik
(xs x + ys y + zs z)
dtd
fj (ki , ki ) exp
R
s
j=1

(5.60)

vt
, [c B, c + B], and c is the radian center frequency.
in which t 0,
Rs
Knowing that
ri = vt rdi = vdt
(5.61)
and r = Rs sin i , dt is transformed to di by

dt =
Rs sin i
di
v
i [0, 2]
(5.62)
(5.63)
Also,
d = cdk
B
B
k [kc
, kc +
]
c
c
(5.64)
(5.65)
where kc is the center wavenumber or radian period (c = f where is the time

c
period, f is the time frequency. By analogy, k =
is the radian period.) The
focused signal C(
r0 ) becomes
cRs sin i
C(
r0 ) =
v
B
kc +
Z c
Z2
N

X
dk di
fj ki , ki
j=1
0
B
c
exp [2ik (sin i cos i x + sin i sin i y + hz)]
kc
173
(5.66)
The circular flight path of ISAR measurements suggests

x = r cos
(5.67)
y = r sin
(5.68)
where is the angular variable locally defined for (x, y) and r =
p
x2 + y 2 .
Substitute these relations into C(

r0 ).
cRs sin i
C(
r0 ) =
v
B
kc +
Z c
Z2
N

X
fj ki , ki
dk di
j=1
0
B
c
exp [2ik (sin i cos(i )r + hz)]
kc
(5.69)
Integration over i provides

B
kc +
Z c
cRs sin i
C(
r0 ) =
v
Z2
N

X
dk di
fj ki , ki
j=1
0
B
kc
c
exp [2ik (sin i cos(i )r + hz)] (5.70)
B
kc +
Z c
N

X
cRs sin i
=
dk
fj ki , ki exp [2ikhz]
v
j=1
B
kc
c
Z2
di exp [2ikr sin i cos(i )] (5.71)

0
cRs sin i
=
v
B
kc +
Z c
dk
kc
B
c
N
X
fj
ki , ki exp [2ikhz] (2)J0 (2kr sin i ) (5.72)
j=1
174
where J0 is the Bessel function of zero order. The integral representation of the Bessel
function is
1
Jn (x) =
2
h
n i
d exp ix cos in + i
2
(5.73)
J0 (x) can be defined by an infinite power series.

J0 (x) = 1
x2
x4
x6
+
+
22 22 42 22 42 62
(5.74)
The combined range and cross-range resolution, r, is constrained by

2kr sin i r
c
1
1
=
4 sin i
2f 2nf sin i
(5.75)
c
2
and = . f is the frequency resolution, and nf is the length of the
f
discrete frequency vector (f = nf f ). Considering the previous example in which
1
=
m (f = 12 GHz) and assuming zs = h << ys (i
= ), the spatial resolution of
40
2
ISAR measurements is approximately 0.00625 m (0.25 in), a figure much smaller than
since k =
the range and cross-range resolutions of traditional radar and SAR measurements.
This proves the efficacy of ISAR over traditional radar measurements.
5.3
5.3.1
Backprojection Algorithms
Range Compression
In this section, derivation of backprojection algorithms is provided. Recall the timeindependent SAR point response S(x, y sin i ) as

S(x, y sin i ) = sinc
y sin i
r

sinc
x
xr

(5.76)
It is equivalent to express S(x, y sin i ) in S(

rs,j ) in which rs,j is the distance of the
point scatterer in the slant plane.

S(
rs,j ) = sinc
175
Rs,j

(5.77)
where Rs,j = |
rs,j |. The time-dependent S(
rs,j ) can be written by
1
S(
rs,j , t) = 2
Rs,j
c Z+ B
d exp [it]
(5.78)
c B
2Rs,j
The range compression on S(
rs,j , t) is conducted by shifting the time t to t = t
,
c
leading to
2Rs,j
1
S(
rs,j , t
) = S(
rs,j , t) = 2
c
Rs,j
c Z+ B

2Rs,j
d exp i t
c
(5.79)
c B
which gives

B
S(
rs,j , t) = 2 exp ic t sinc B t
(5.80)
Rs,j

2Rs,j
where the term t
shifts the origin of traveling time to the time receiving the
c
reflected signals. By shifting the origin of time to the center of the scatterer, the range
distance Rs,j has been compressed in the way that the center of the reconstructed
image coincides with the center of the scatterer.
In this time-shifting process, it is assumed that the traveling speed of radar signals
is the speed of light in free space (c = 3 108 m/s), which is only true when the
propagation and scattering of radar signals entirely occur in free space. However, in
the circumstance of scattering from multi-layered dielectrics, some received signals
are reflected from in-depth scatterers/interfaces as the result of multiple reflections in
c
multi-layered dielectrics, whose traveling speed is 0 for non-magnetic dielectrics,
r
where 0r is the dielectric constant or relative permittivity. Consider two cases of the
reflection of radar signals in Figure 5-5. Two radar antennas transmit signals toward
a PEC (perfect electric conductor) plate of thickness d. In the propagation path of
radar signals, there is only Medium 1 for Antenna A, and there are Media 1 and 2 for
Antenna B. Assume non-magnetic dielectrics (relative permeativity 0r1 = 0r2 = 1)
and Medium 1 to be free space (0r1 = 1). Considering only the first reflection from
176
the PEC plate, the difference in the traveling time of Antennas A and B is
p
d
t = 2 0r2 1
c
(5.81)
Should the surface of the PEC plate facing the radar be the center of a scatterer,
the time difference t becomes significant when (1) 0r2 is large, or (2) thickness d
is large. Since the footprint is assumed to be a Gaussian amplitude distribution,
the correlation decays as the amplitude of a Gaussian distribution decays from its
peak. Consequently, this time difference will project the center of the scatterer to a
shifted position on the time axis as illustrated in Figure 5-6, leading to the shifting
of scattering signals in reconstructed images. This creates an error for locating the
center of defect using reconstructed images. To rectify or minimize this t error,
accurate estimation on the dielectric properties of target material is necessary, leading
to the need for a comprehensive dielectric modeling of multi-layered cementitious
composites.
Express the radar location vector rs by its position on the synthetic aperture, ,
as shown in Figure 5-7. It is clear that
= |
rs |i = Rs i
(5.82)
The total range-compressed focused signal collected at on the aperture, D(, t), is
the integration of the range-compressed focused signal received from all scatterers in
the domain s . In cylindrical coordinate systems,
ZRs
Z2
D(, t) =
d
rj dj G(
rj , j )S(
rs,j , t)
0
(5.83)
where G(
rj , j ) is a scattering amplitude density function at (
rj , j ). Or it can be
177
written to be
xZmax yZmax
q
D(, t) =
dx
dy G(x, y)S( ((xs xj )2 + (ys yj )2 ), t)
xmin
ymin
(5.84)
in Cartesian coordinate systems. D(, t) is the range-compressed focused signal for

2Rs,j
the radar located at on the aperture, whose time being shifted by
.
c
5.3.2
Backprojection Processing
The backprojection processing can be conducted by defining the backprojected signal

BBP (, t) as
BBP (, t) = CBP
D(, t)
t
(5.85)
where CBP is the backprojection coefficient hereby defined in order to yield an ideal
bandpass transfer function and
CBP
1
=
i
4
c
2
(5.86)
when an ideal point scatterer with unit cross section at rj is considered and the
frequency integration is performed by . When the frequency integration is carried
out by f ,
CBP
1
=
2i
4
c
2
(5.87)
D(, t)
is defined continuously, while D(, t) is usually discrete. To prevent a pot
D(, t)
tential discontinuity in
from occurring, a matched filter M (t) is applied on
t
the derivative.
D(, t)
=
t
t
Zt
dt0 D(, t)M (t t0 ) =
Zt
0
178
M (t t0 )
dt0 D(, t)
t
(5.88)
This convolution operation is also advantageous for computational efficiency since

M (t t0 )
can be calculated in advance. In the frequency domain, convolution is
t
performed by multiplying a phase factor. Finally, the backprojection image is given
by
RZs int
I(
r, ) = y
d BBP (, t)
(5.89)
0
which is the integration of BBP (, t) over the entire aperture length Rs int . I(
r, ) is
a two-dimensional, spatial image (range vs. cross-range) of the structure.
5.3.3
Support Band Analysis Method of Stationary Phase
The purpose of support band analysis is to define a band support region in which
the spatial spectrum of backprojection image I(x, y) is guaranteed non-trivial, using
the method of stationary phase [22]. This analysis is performed in the following
paragraphs.
Consider a time-dependent SAR point response S(
rs,j , t
1
S(
rs,j , t) = 2
Rs,j
c Z+ B
2Rs,j
) as
c

2Rs,j
d exp i t
c
(5.90)
c B
The scattering amplitude density function G(x, y) is (x xj )(y yj ). The rangedcompressed signal D(, t) is
1
D(, t) =
2
2Rs,j
c Z+ B

2Rs,j
d exp i t
c
c B
fZmax

1
2Rs,j
= 2
df exp i2f t
Rs,j
c
fmin
179
(5.91)
(5.92)
Therefore,
D(, t)
i2
= 2
t
Rs,j
fZmax

2Rs,j
df f exp i2f t
c
fmin
(5.93)
The backprojected signal BBP (, t) is

BBP (, t) =
4
c
2
1
2
Rs,j
fZmax

2Rs,j
df f exp i2f t
,
c
fmin
(5.94)
and the backprojection image I(x, y) is

I(x, y) =
4
c

2 RZs int fZmax

yf
2Rs,j
d
df 2 exp i2f t
Rs,j
c
0
fmin
(5.95)
where Rs,j is a function of since Rs (or rs ) depends on . The spatial spectrum of

I(x, y) is obtained by
x , ky ) = 1
I(k
(2)2
xZmax yZmax
dy I(x, y) exp [i (2f t kx x ky y)]
dx
ymin
xmin
(5.96)
This leads to
2 xZmax yZmax RZs int fZmax
yf
x , ky ) = 2
I(k
dx
dy
d
df 2
c
Rs,j
xmin
ymin
0
fmin

2Rs,j
exp i2f t
+ i (2f t kx x ky y)
c
(5.97)
Consider the far-field ISAR measurements being conducted at a constant elevation

x , ky ) are illustrated in Figure 5-7. I(k
x , ky ) is
(zs = constant). The variables in I(k
the reconstructed image assembled by r = {
r(x, y)|(x, y) BP } in account of the
scattering from the scatterer j at rj observed by the monostatic radar at rs whose
position is defined by . BP is the domain of the image to be reconstructed, typically
180
x , ky ) is
to be rectangular. The phase factor (x, y, , f ) in I(k

(x, y, , f ) = 4f t 2f
2Rs,j
c

(kx x + ky y)
(5.98)
where
q
2
t=
(x xs )2 + (y ys )2
c
q
Rs,j =
(xs xj )2 + (ys yj )2
cos i
i
ys = Rs sin i =
sin i
i
xs = Rs cos i =
(5.99)
(5.100)
(5.101)
(5.102)
Assuming that the monostatic radar operates at constant range Rs , the radar position
(xs , ys ) is only a function of i . Replacing 2f by , the phase factor (x, y, i , )
can be written to be
q
4
(x, y, i , ) =
(x Rs cos i )2 + (y Rs sin i )2
c
q
2
(Rs cos i xj )2 + (Rs sin i yj )2 kx x ky y

c
(5.103)
where the wave numbers kx and ky remain unchanged in the four-dimensional integral over (x, y, i , ). The principle of stationary phase suggests that the condi x , ky ). The first derivatives of
tion of (x, y, i , ) determines the condition of I(k
181
(x, y, i , ) are
x Rs cos i
=
q
x
c
kx
(5.104)
4
y Rs sin i
=
q
ky
y
c
(5.105)
(x Rs cos i ) + (y Rs sin i )
4Rs (sin i cos i )

= q
i
c (x Rs cos i )2 + (y Rs sin i )2
4Rs (cos i sin i )
q
c (Rs cos i xj )2 + (Rs sin i yj )2
q
4
=
q
c
2
(Rs cos i xj )2 + (Rs sin i yj )2

c

The stationary point at which = 0 occurs when
,
,
,
x y i
the stationary point occur at (
x, y, i ,
) which are determined by
ckx
4
x = Rs cos i + s

2 (y Rs sin i )
ckx
1
4
cky
y = Rs sin i + s 4

2 (x Rs cos i )
cky
1
4
(x2 + y 2 ) (x2j + yj2 ) y yj
sin i
2Rs (x xj )
x xj
4(x2 + y 2 ) (x2j + yj2 ) 4y yj
cos i =
sin i
2Rs (4x xj )
4x xj
cos i =
182
(5.106)
(5.107)

= 0. Let
(5.108)
(5.109)
(5.110)
(5.111)
Since cos i =
xs
ys
and sin i =
, the condition of
= 0 is further expressed by
Rs
Rs
i
(x2 + y 2 ) (x2j + yj2 ) y yj
xs =
ys
2(x xj )
x xj

ys
1
i = tan
xs
and the condition of
(5.112)
= 0 provides
xs =
4(x2 + y 2 ) (x2j + yj2 ) 4y yj
ys
2(4x xj )
4x xj

ys
1
i = tan
xs
(5.113)
x , ky ) in order to obtain
The stationary points must fall in the integration limits of I(k
x , ky ). In other words, (
non-zero values of I(k
x, y) > 0. The conditions on (
x, y)
provide the constraints on (kx , ky ). Condition x > 0 provides
s
kx >
where C1 =
C12 4
1 + C12 c
(5.114)
Rs cos i
. Condition y > 0 provides
Rs sin i y
s
ky >
C22 4
1 + C22 c
(5.115)
Rs sin i
. Therefore, the stationary point is inside the limits of
Rs cos i x
integration if and only if (kx , ky ) where
where C2 =
= {(kx , ky )|fmin <
and k =
c
k < fmax , kx >
4
C12
4
, ky >
2
1 + C1 c
kx2 + ky2 . is the band support region for I(kx , ky ).
183
C22 4
} (5.116)
1 + C22 c
z
Flight path
rs = ( xs , ys , zs )
Ending position
Starting position
ki
y
Footprint domain G
x
(a) Inverse synthetic aperture radar (ISAR)
Flight path
Footprint domain G
g
y
Starting position
Synthetic
aperture
Ending position
int
x
(b) ISAR Plane view
Figure 5-4: Configuration of ISAR measurements
184
Medium 1 ( r1 , r1 )
Rs , j
A
PEC
Rs , j d
Medium 1 ( r1 , r1 )
Medium 2
( r2 , r2 )
Figure 5-5: Two cases of the reflection of radar signals
t = 2 r2 1
Actual origin
Projected origin
Figure 5-6: Time shifting error
185
d
c
rs , j = rs rj
rj
rs
= rs i
x
r = ( x, y )
x , ky ) and their relationship

Figure 5-7: Variables in I(k
186
5.3.4
Advantages of Backprojection Algorithms
Main advantages of the backprojection algorithm include: (1) lower and localized
artifact levels than frequency-domain algorithms, (2) easy adjustment to an approximate inverse formula for perturbed problems, (3) readily for parallel computing with
limited interprocessor communications, and (4) simple motion compensation by timeshift operation [177, 194].
As a result of the advantages mentioned above, sub-images with different bandwidths and angular ranges can be available before the physical inspection (far-field
ISAR measurements) is completed. This enables the technique for different purposes
of inspection. For instance, sub-images of narrow bandwidths and angular ranges
(although poor resolution) are advantageous for preliminary inspections due to their
rapid inspection and processing. Final images which utilize full bandwidth and full
angular range data are useful for detailed inspections once a suspicious local area
can be determined. The superposing process of sub-images provides not only the
gradually-improved image resolutions (range and cross-range), but also the evolution
of scattering signals in the image. For example, angular sensitivity of defect scattering signals is revealed by processing ISAR measurements at each sub-aperture. It is
evident that understanding the pattern (evolution, convergence) of defect scattering
signals in backprojection images is beneficial to studying actual defects or damages.
Knowledge about the needed frequency bandwidth and angular range for revealing
certain types of defects in backprojection images can also be established based on
such understanding.
5.4
Implementation and Coding Procedure
In this dissertation, the backprojection processing is implemented mainly by the

r (MATrix LABoratory) language (by The MathWorks), a high-performance
Matlab
technical computing environment, and some by the FORTRAN (FORmula TRANsr is a scripting language similar to the
lator) 90 (or F90) language. While Matlab
C++ language, its extensibility (Toolboxes), efficiency (matrix computations, graph187
ics generation), and ability to directly communicate with other computing languages
(e.g., FORTRAN and C++) potentially ease the effort for future modifications and
expansions. The original codes were developed by the MIT LL and modified by the
author.
In what follows, implementation of the backprojection processing is described in
the coding procedure using a sample measurement signal from the specimen AD1
(8GHz12GHz, HH polarization, i = 15 , oblique incidence) in which basic SAR
processing steps are also included.
1. Input the one-dimensional (1D), steady-state, complex far-field ISAR measurement/signal (amplitude and phase) collected at the azimuth angle i and over
the frequency range of [fmin , fmax ]. The length of the signal is n. Amplitude
(A = A(fi )) and phase ( = (fi )) of the sample far-field ISAR data used are
shown in Figure 5-8.
-20
Phase (rad)
Amplitude (dBsm)
2
-22
-24
-26
-28
1
0
-1
-2
-30
-3
8
9
10
11
Frequency (GHz)
12
9
10
11
Frequency (GHz)
12
Figure 5-8: Far-field ISAR measurement of the specimen AD1 HH polarization,

i = 15
2. Combine the amplitude and phase of the signal to obtain the complex form of
the signal by
A
(fi ) = 10 20 exp [i]

188
(5.117)
The combined complex signal is shown in Figure 5-9
0.08
0.1
0.06
0.08
0.04
Imaginary part
Real part
0.06
0.04
0.02
0
0.02
0
-0.02
-0.02
-0.04
-0.04
-0.06
-0.08
-0.06
8
9
10
11
Frequency (GHz)
12
9
10
11
Frequency (GHz)
12
Figure 5-9: Complex form of the far-field ISAR measurement
3. Perform SAR processing by choosing a desired sidelobe pattern. Reduction of

sidelobe levels (SLLs) is achieved by applying a weighting function to the raw
data as shown in Figure 5-8). A weighting function used in this dissertation is
wj = 1 +
2m
l=1

lj
fl cos 2
2
(5.118)
where m is the order the weighting function and the shape function
"
#1
2 # "
2 2
a + 2 m 12
1
l
l2
fl =
1
2 1
m1
2
m
(l 1)! (l+1)!
a2 + 2 l 21
(5.119)
SLL2 1) and SLL is the sidelobe levels. Figure 5-10

wj
shows the normalized weighting function P
with different SLLs ranging
[wj ]
from 30dB to 90dB. The SLLs in dB is determined by SLLdB = 10SLL/20 . The
where a = log(SLL +
weighted complex signal is shown in Figure 5-11.

4. Prepare an empty range-cross-range image as a base image for backprojection.
Default size of the base image is 1 m-by-1 m. Since the size of used GFRP189
0.014
30dB
40dB
50dB
60dB
70dB
80dB
90dB
0.012
Weighting
0.01
0.008
0.006
0.004
0.002
0
50
100
Data point
150
200
Figure 5-10: Desired sidelobe pattern with different sidelobe levels (SLLs) ranging
from 30dB to 90dB
wrapped concrete specimens is less than 0.5 m, 1 m-by-1 m is capable of covering
the specimen and the scattering signals. Another size, 2 m-by-2 m, is also
used in some cases. Increase the image resolutions by oversampling the grids
to prepare for the coordinate transformation after performing one-dimensional
(1D), complex discrete Fourier transform (DFT) of the weighted data. The
used oversampling ratio (rOS ) is to be greater than 23 = 8 for less interpolation
errors during the coordinate transformation. 24 = 16 and 25 = 32 have been
used for obtaining satisfactory results. Perform the center-shifting operation
again on the obtained 1D Fourier conjugate of the elongated data. The amount
nDF T
of shifting index is 1 +
. The length of the oversampled/elongated signal
2
for 1D complex DFT is determined by
nDFT = 2log2 (nrOS )
(5.120)
5. Shift the center of the weighted data of length n to the origin by truncating
190
-8
x 10
0
Imaginary part
Real part
-8
x 10
1
0
-1
-1
-2
-3
-2
-4
-3
8
9
10
11
Frequency (GHz)
12
9
10
11
Frequency (GHz)
12
Figure 5-11: Weighted complex form of the far-field ISAR measurement
the data points of indices [nc : n] and moving it to [1 : n nc + 1] of an

n
elongated/oversampled data with length n0 = rOS n, where nc = + 1. The
2
other half of the weighted data, [1 : nc 1], is placed at the end of the elongated
data. Perform zero-padding for the increased points of the elongated data.
Conduct 1D complex discrete Fourier transform on the elongated data collected
at (i , f ). The 1D Fourier conjugate (also complex) of the elongated data is
obtained. The shifted 1D DFT of the complex signal is shown in Figure 5-12.
6. Prepare the radiation pattern projection of the radar from its local coordinate
system to the image plane. The radiation pattern at the local coordinate system
of the radar is defined by the Poynting vector Pl = (Pxr , Pr , Ph )T as shown in
Figure 5-13. The nominal Poynting vector of the radar at the local coordinate
system is Pl = (0, 1, 0)T or Pr = 1 (far-field). The coordinate transformation
matrix from the local coordinate system of the radar to the image plane is
cos i sin i 0
1
0
0
Pg = sin i cos i 0 0 cos i0 sin i0

0
0
1
0 sin i0 cos i0
191

Pl
(5.121)
-7
-6
x 10
x 10
0
0
-0.5
Imaginary part
Real part
-2
-4
-6
-8
-1
-1.5
-2
-10
-2.5
-12
-3
1000
2000
3000
Data point
4000
1000
2000
3000
Data point
4000
Figure 5-12: Shifted 1D DFT of the weighted complex signal
i is the grazing angle. i is the azimuth angle. In the case

2
considered in this research, i = .

2
where i0 =
Ph
z
Pr
i
Image plane
Pxr
i
x
Figure 5-13: Poynting vector of the radar
7. Multiply the shifted 1D DFT signal with the phasor R which is

f
R = exp i4 Rs,j
c
(5.122)
and with the transformed Poynting vector based on the azimuth angle. Form
the BP image by projecting the multiplied signal onto the base image using the
192
range distance of each pixel (Rs,j ) (with respect to the center of the image) and
the azimuth angle (i ). The projected image is shown in Figure 5-14.
Figure 5-14: Projected image of the transformed far-field ISAR measurement
5.5
5.5.1
Effects of Aperture Size and Bandwidth

Aperture Size
As shown in Figure 5-14, the 1D signal at a given azimuth angle generates an image
with constant scattering signals along the cross-range direction since physically no
angular variation can be discovered if the measurement is only made at one azimuth
angle. Should we expand the azimuth vector of input measurements, for example,
from (15 ) to (14.8 , 15 , 15.2 ) (the far-field ISAR measurements are shown
in Figure 5-15) and follow the same procedure described in the previous section,
the projected image becomes Figure 5-16. In Figure 5-16, it is found that variation
of the scattering signals in the cross-range direction occurs as expected. Projected
images using azimuth vectors (14.6 , 14.8 , 15 , 15.2 , 15.4 ) and (14.4 ,
14.6 , 14.8 , 15 , 15.2 , 15.4 , 15.6 ) are shown in Figures 5-17 and 518, respectively.
Ultimately, in this example, the half-aperture and full-aperture
measurements produce images shown in Figures 5-19 and 5-20, respectively.

193
12
12
-1
11.5
11.5
2
-2
-3
10.5
-4
10
-5
9.5
-6
-7
11
Frequency (GHz)
Frequency (GHz)
11
10.5
10
9.5
-1
-8
8.5
-9
8
-0.3
-0.2
-0.1
0.1
0.2
-2
8.5
8
-0.3
0.3
-3
-0.2
-0.1
0.1
0.2
0.3
Figure 5-15: Full-bandwidth far-field ISAR measurements with an azimuth vector of

(14.8 , 15 , 15.2 )
5.5.2
Bandwidth
The effect of different bandwidths on reconstructed BP images is studied by using

the same measurement as shown in Figure 5-8. The full-bandwidth of the collected
far-field ISAR measurement is 4 GHz (8 GHz 12 GHz). The bandwidth of collected
measurements represents the probing ability of the radar in the range direction. As
a result of it, wider bandwidths offer better detectability in the range direction.
Consequently, range resolution of the radar and of BP images can be improved or
reduced by the use of an increasing or decreasing bandwidth.
Figures 5-21, 5-22, and 5-23 show the full-aperture BP images using different
bandwidths (in a descending order). It is clear that reducing the bandwidth of input
measurements results in poorer range resolution in the BP images. Equivalently,
better detectability in the range direction can be provided, given wider bandwidths
of the measurement.
194

(14.8 , 15 , 15.2 )
5.6
Summary
In this chapter, theoretical background of SAR and ISAR processing for image reconstruction is provided. Fast backprojection algorithm is used for numerically implementing the concept of tomographic reconstruction in this work. The implementation
and coding procedure is also described with far-field ISAR measurement examples.
Effects of aperture size and bandwidth in reconstructed images are addressed. Research findings are summarized as follows.
As demonstrated in this chapter, ISAR processing offers a flexibility for improving the resolution of radar measurements by incorporating additional measurements with respect to frequency (range and cross-range resolutions) and
incident angle (cross-range resolution). The basis of ISAR processing is the
movement of radar and the integration of measurements at different azimuth
angles. The bandwidth of measurements is proportional to the resolution in
the range direction, and the aperture size decides the focusing capability in the
cross-range direction.
The importance of dielectric properties of materials is identified in this chapter.
195

(14.6 , 14.8 , 15 , 15.2 , 15.4 )
While the dielectric properties of the structure are not required as input for
rendering images containing near-surface defects, they become important when
the center of backprojection is deep inside the dielectric medium; the backprojected center of scattering signals in reconstructed images will be shifted from
its actual location if the difference in traveling time is not accounted (defect
locating error). This leads to the research need for better understanding the
dielectric properties of GFRP-concrete systems including concrete, a porous cementitious composite, as described in Chapter 6 Dielectric Modeling of
GFRP-concrete Systems.
196

(14.4 , 14.6 , 14.8 , 15 , 15.2 , 15.4 , 15.6 )
Figure 5-19: Full-bandwidth half-aperture backprojection image i [150 : 1 : 1]

(i = 0.2 )
197
Figure 5-20: Full-bandwidth full-aperture backprojection image i [150 : 1 :

150] (i = 0.2 )
Figure 5-21: Sub-bandwidth full-aperture backprojection image [8, 11] (GHz)
198
-0.4
Power (W/m 2)
Cross-range (m)
0.05
-0.2
0
0.04
0.03
0.02
0.01
0.2
0.4
0.2
0.4
-0.4
-0.2
0
0.2
Range (m)
0.4
0
-0.2
-0.4
Cross-range (m)
-0.4
0.2
0.4
-0.2
Range (m)
199
200
Chapter 6
Dielectric Modeling of
GFRP-concrete Systems
Different men can be affected by one and the same object in different
ways, and one and the same man can be affected in different ways by one
and the same object at different times.
Ethics, Prop. LII., Baruch de Spinoza (16321677)
Any use of electromagnetic waves for investigating materials must deal with the dielectric properties of materials, this includes the case of radar NDT for construction
materials. As indicated by Maxwells equations materials are characterized by electric
permittivity , electric conductivity , magnetic permeability , and magnetic conductivity . Materials are described by these properties and classified accordingly.
The frequency-dependence of these properties is termed dielectric dispersion. The
essential goal of dielectric modeling aims at the development of mathematical models/equations for the determination of dielectric properties at different measurement
frequencies.
In Chapter 3 Numerical Simulation concrete is assumed as a homogeneous,
isotropic, and lossless dielectric medium. Knowing the heterogeneity, multi-phase
nature of concrete, it is realized such assumption is highly simplified. To improve
the accuracy of material modeling in numerical simulation, appropriate dielectric
201
models must be developed. Additionally, in Chapter 5 Image Reconstruction,

it is also identified that the knowledge of dielectric properties can further improve
the precision of backprojection processing when deep defects are to be detected and
located. Therefore, the purpose of this chapter is to model the dielectric dispersion of
considered materials (water, epoxy, E-glass fabric, GFRP, and oven-dried hydrated
Portland cement paste) in the frequency range of 8 GHz to 18 GHz using reported
experimental measurements in literature [75, 107, 153, 96, 113, 119, 107, 154, 271].
In view of the novel use of dielectric properties in civil engineering, definition
and physics of dielectric properties and dielectric dispersion are introduced as the
background information. Applications of dielectric properties in civil engineering are
provided, indicating the increasing importance of dielectric properties in the profession. A geometrical analysis technique used for dielectric modeling is introduced.
Dielectric models of considered materials (water, epoxy, E-glass fabric, GFRP, and
oven-dried hydrated cement paste) are provided. Challenges in the dielectric modeling
of concrete as a multi-phased cementitious composite are also addressed.
6.1
Background
Dielectric properties of a material can be used to determine other material properties such as moisture content, bulk density, bio-content, chemical concentration, and
stress-strain relationship. Such knowledge can be utilized for research and application
in food science, medicine, biology, agriculture, chemistry, electrical devices, defense
industry (security), and various engineering disciplines including civil engineering.
This section provides the background information regarding the impacts of dielectric
properties of materials in general and in the NDT problem of the GFRP-concrete
systems.
6.1.1
Definition and Physics of Dielectric Properties
Dielectric properties of materials can be interpreted both microscopically and macroscopically. Microscopically, dielectric properties represent the polarization ability of
202

molecules in the material corresponding to an externally applied electric field E.
Macroscopically, dielectric properties are the relationship between the applied elec (coulombs/m2 ), both
tric field strength E (volts/m2 ) and the electric displacement D
externally measured.
Dielectric properties are the collective term of electric permittivity (farads/m
or F/m), electric conductivity (siemens/m or S/m), magnetic permeability (henrys/m or H/m), and magnetic conductivity (ohms/m or O/m). Materials are
described and classified by these properties into various types such as metals and
dielectrics. In this research work, only isotropic materials are considered, meaning
that all these properties are described by first-order tensor. In other words, modeling
using second-order (or rank-two) tensors for anisotropic materials is not considered
in this research. These quantities can be real or complex, depending on the nature
of the material. When alternating current (a.c.) fields are applied on the material
under investigation, generally these quantities are complex. Definition of dielectric
properties of materials is introduced in the following paragraphs.
Complex electrical permittivity (F/m) describes the ability a material

interacts with an applied electric field. It is defined as the ratio between the
(coulombs/m2 or C/m2 ) and the electric field E
electric displacement field D
(volts/m or V/m). Generally,
= E
D
(6.1)
where = () for dielectric materials. It represents the ability a material to

permit an electric field passing through the material. For non-metallic materials
(or dielectrics) their frequency-dependent response subject to applied a.c. fields
is the result of the molecular polarizability [28]. Their delayed and attenuated
response is also observed and described as the dielectric dispersion phenomenon
attributing to several polarization phenomena in the microscopic level [113]. To
account for these energy absorption and losses an imaginary part is needed
in the dielectric description of the material property. Therefore, the complex
203
electrical permittivity (or complex permittivity) is defined as

= 0 + i(00 ) = 0 i00
(6.2)
where 0 is the real part of representing the ability a material can store the
energy carried by the electromagnetic field transmitting through it, and 00 is
the imaginary part representing energy absorption and loss. The negative sign
definition of 00 is applied since energy dissipation/loss occurs to the materials
considered in this research. Positive sign of 00 , on the other hand, suggests
energy creation.
The measured values of 0 and 00 mainly depend on measurement frequency and
temperature, while in some cases, as well as on pressure [126]. For example,
the Debye equations provide a frequency-dependent representation of 0 and 00 ,
satisfying the Kramers-Kronig relations [140], as [65]
s
1 + ( )2
(s )
00 () =
1 + ( )2
0 () = +
(6.3)
(6.4)
where = 2f is the angular frequency (rad/s or rad-Hz), f is the temporal

frequency (1/s or Hz), is the permittivity measured by the alternating current
(a.c.) field at frequency = (infinite frequency), s is the permittivity
measured by the direct current (d.c.) field at frequency = 0 (static frequency),
and is the characteristic relaxation time (s/rad). The reported relaxation time
is usually represented by t =
(s). Materials whose response can be described
2
by the Debye equations are called the Debye material. Details about the Debye
equations will be introduced in the modeling section of this chapter.
A dimensionless representation is also used for defining the complex permittivity. The complex relative permittivity is defined by
r =
0 00

=
= 0r i00r
0
0
204
(6.5)
where 0 is the electrical permittivity of free space and 0 = 8.85 1012 F/m.
0r is the dimensionless dielectric constant and 00r is the dimensionless loss
factor. It is only the dimensionless nature leading to the name dielectric
constant since 0r is not a constant when considered over a range of frequencies.
Different symbols were adopted for the dielectric constant such as [113] and
[75]. In this research, the representation is adopted.
The ratio between 0r and 00r is the loss tangent (or dissipation factor).
tan =
This dimensionless representation
00r
00
=
0r
0

, tan
0
(6.6)
is simpler than and advantageous
over the original (0 , 00 ) representation because it clearly shows how the material
is different from free space (or vacuum).
Complex magnetic permeability (H/m) describes the ability a material
interacts with an applied magnetic field. It is the ratio between the magnetic
(webers/m2 or W/m2 or Tesla) and the magnetic field H
(amflux density B
peres/m or A/m).
= H
(6.7)
is a scalar for isotropic materials. The complex magnetic permeability (or

complex permeability) is used when magnetic losses are present in the material.
= 0 i00
(6.8)
where 0 and 00 are the real and imaginary parts of the complex permeability
(henrys/m, or H/m), respectively. The negative imaginary part of suggests
the energy dissipation. A dimensionless, relative complex permeability can be
further defined as
r =
0 i00
= 0r i00r
0
(6.9)
where 0 is the permeability of free space and 0 = 4 107 H/m. Definition

205
of various materials based on is listed in Table 6.1. Threshold values were

suggested by Epstein (1962) [79]. It was reported in the literature that DebyeTable 6.1: Types of magnetic materials
Description
Criterion
Ferromagnetic
Magnetic Paramagnetic
Diamagnetic
Non-magnetic
r > 10
1 < r < 10
r < 1
r = 0r = 1, 00r = 0
type models can be used for modeling the complex permeability of magnetic
materials [79].
Apparent electrical permittivity a (F/m) is defined by accounting for the
direct current (d.c.) conductivity loss in the representation of complex permittivity in this research. Since the imaginary part of the complex permittivity
represents the dielectric loss of the material, which is only part of the total energy dissipation from EM waves transmitting through the material. The total
energy dissipation in the material can be expressed by the dissipated or absorbed
power using the complex Poynting vector theorem [52] and the Maxwells equations. The complex Poynting vector P is
1
P =
2

dS
E H
(6.10)
where S is a closed surface and dS is a vector element of area directed outward

is the conjugate of the magnetic field H
(A/m). The
from the volume V , H
divergence theorem provides
I

dS =
E H

dV
E H
(6.11)
From vector calculus we have

= E H
H
E
E H
206
(6.12)
and = i D
+ J,
Ohms law
With Maxwells equations E = i B
and B
= H,
the complex Poynting vector becomes
J = E,
1
2
I
S

dS = 1
E H
2

H
+ i E E E dV (6.13)
i H
Since = 0 i00 and = 0 i00 , reorganizing the real and imaginary parts
of the complex Poynting vector gives
1
2
I
S
Z

1
H
00 H
H
dV +
i0 H
E H dS =
2 V
Z

0
1
E E i00 E E E E dV
2 V
(6.14)
Rearranging the above equation provides

I

1
Re
E H
dS =
2 S
Z

H
+ 00 E E + E E d(V )
00 H
2 V
I

1
E H
Im
dS =
2 S
Z
H
0 E E dV
0 H
2 V
(6.15)
(6.16)
where dS is the vector pointing toward the closed surface S and tangent to
the surface boundary. The real part of the complex Poynting vector is the
dissipated energy Pdis absorbed by the material, which consists of three parts;
magnetic loss, electric loss, and conductivity loss. Since 00 = 0 for non-magnetic
materials, Pdis becomes
Pdis
=
2
Z
V

+
E E dV =
2
00
00e E E dV
(6.17)
and the effective dielectric loss factor is defined as

00e = 00 +
207
(6.18)
Equivalently, an effective conductivity can be defined as

e = 00e = 00 +
(6.19)
where = s is the electrical conductivity (S/m) measured by the d.c. current

at static frequency ( = 0). This conductivity is the real part of the complex
conductivity. With the effective dielectric loss factor at hand, an apparent
complex permittivity is defined.

s
a = 0 i00e = 0 i 00 +
(6.20)
Definitions of materials based on this apparent complex permittivity are listed

in Table 6.2.
Table 6.2: Types of dielectric materials
Description
Perfect dielectric
Imperfect dielectric
Criterion
0
> 0, 00e = 0
0 > 0, 00e > 0
Complex electrical conductivity (S/m) describes the ability a material

to conduct an applied electric current, represented by the ratio between the
current flux density J (amperes/m2 or A/m2 ), and the electric field E (V/m2 ).
J = E
(6.21)
where is the scalar electrical conductivity (S/m) for isotropic materials. Following the previously shown energy treatment for the definition of the complex
permittivity, the complex electrical conductivity (or complex apparent conductivity) is defined as
a = ia = i (0 i00e ) = e0 + i 00
208
(6.22)
where
e0 = 00e = 00 + s
(6.23)
00 = 0
(6.24)
The defined effective conductivity is actually the real part of the complex apparent conductivity.
e0 = 00e = e
(6.25)
The d.c. conductivity, an frequency-independent term, is part of the apparent

complex conductivity in the definition. The d.c. conduction effect is significant
only in low-frequency or high-temperature situations, while it is insignificant in
microwave frequency range since in e0
00 > s
(6.26)
By constructing this relationship the definition regarding conductivity in complex apparent permittivity is connected to the complex apparent conductivity.
This also suggests that the behavior of this complex apparent conductivity is,
by definition, similar to the one the complex apparent permittivity exhibits. For
the example of the Debye model, the real and imaginary parts of the complex
apparent conductivity are
a = e0 + i 00
s
( )2 ( s )
0
e = s +
= s +
1 + ( )2
1 + ( )2
( )1 ( s)
( s )
00
= +
= +
2
1 + ( )
1 + ( )2
(6.27)
(6.28)
(6.29)
where is the permittivity measured by the alternating current (a.c.) field at

frequency = (infinite frequency), and s is the permittivity measured by
the direct current (d.c.) field at frequency = 0 (static frequency). Comparing
209
e0 with 00e provides

s = ( s )
(6.30)
The relaxation time can be determined by

=
s
s
(6.31)
The Debye-type behavior of complex electrical conductivity was observed by

Grant (1958) in his measurements. It is evident that the loss tangent can also
be expressed by
tan =
00e
e0
e0
00e
=
=
=
0
00
0
00
(6.32)
Those with a high conductivity are considered to be conductors such as metals,

and those without conductivity are insulators such as glass. Nonconductive
materials are also termed loss-less. Materials with slight conductivity are called
low-loss. Table 6.3 summaries various materials defined by conductivity.
Table 6.3: Types of materials defined by conductivity

Description
Criterion
Note
Lossy
0 < tan
When the electrical conductivity is present in the material
(1) tan 1
General definition
(2) tan < 0.2
For concrete of dielectric

constant in the range of 2.5
and 8 [10]
Low-loss
(3) 104 tan [188]

103
Lossless
(4) tan < 5 104
For electronic packing materials [123]
tan = 0
When the electrical conductivity is not present in the

material
210
In this dissertation, the complex electrical permittivity is used as the characteristic

parameter in the dielectric modeling of materials for the following reasons.
All the materials considered in this research are not magnetically dispersive,
suggesting constant magnetic properties of the materials.
Complex electrical conductivity is not directly measured from the received EM
waves, but can be indirectly related to the complex electrical permittivity.
Therefore, emphasis of the dielectric modeling is placed on the development of dielectric models for determining complex permittivity. Only isotropic, non-magnetic
dielectrics are considered, suggesting that the complex permittivity is a tensor of rank
zero.
6.1.2
Dielectric Spectroscopy and Dielectric Dispersion
From the perspective of electromagnetism materials can be generally classified into

metal and dielectric. Metals are usually perfect conductors, in which spontaneous
response occurs when subject to the application of an external electric field, meaning
that the creation and diminishing of internal field in metals occur without delay in
time. It is because that the molecular dipoles in metals can freely rotate to align with
the direction of the external field instantaneously without consuming extra energy.
Consequently, the dielectric properties of metals are frequency-independent.
Dielectric materials (or dielectrics), on the other hand, display delayed response
when subject to an external electric field. It is because the rotation of molecular
dipoles in dielectrics suffers from the inherent friction around the dipoles. This
hindered rotation [65] further leads to a hysteretic response which serves as the
energy absorption and storage mechanism of dielectrics, making dielectrics an insulating material or an insulator as called by Faraday (1965) [81].
The frequency-dependence behavior of complex permittivity of dielectric materials
is called the dielectric dispersion. This behavior can be explained by different types
of dispersion theories such Debye dispersion [65], Drude dispersion [147], Lorentz
211
dispersion [15], and other multiple poles dispersions. Among these dispersion theories, Debye and Drude dispersions characterize single-pole response, while Lorentz
dispersion describes double-pole response. In the frequency range of investigation (8
GHz18 GHz) the considered materials in this research exhibit single-pole response
in their Cole-Cole diagram (to be explained later), hence the use of multiple poles
dispersion theories is excluded.
The single pole dispersion theory or the single relaxation time theory was developed by several pioneer researchers in the early development of this field [65, 50, 51,
90, 147]. The physical meaning of the relaxation time is explained as follows: When
a dielectric is subjected to the application of an external field, an internal field is
generated inside the dielectric. When the external field is removed, the field inside
the dielectric will decrease gradually. The time the internal field drops to 1/e of
its original magnitude is called the relaxation time and usually denoted by . For
gaseous substances there is no relaxation behavior since all the charges are free to
rotate upon the application of external field. However, for liquid and solid dielectrics,
the free rotation of bond charges is dampened by the friction around the molecules,
as the hindered rotation hypothesis proposed by Debye (1929) [65].
The study of the dispersion behavior of dielectrics is termed dielectric spectroscopy. The response of dielectrics is usually represented by an effective complex
permittivity, characterized by the frequency-dependent fluctuation of the complex
permittivity, and originated at several energy storage and dissipation mechanisms
which the fluctuation is attributed to.
6.1.3
Energy Storage and Dissipation Mechanisms
A typical dielectric response of the real part of the complex permittivity is provided
in Figure 6-1. Two fundamental mechanisms dominate the response of a material
when it is placed in an electric field; the energy storage and dissipation mechanisms.
The energy storage mechanism is a capacitance effect, caused by the polarizibility of
the material or the molecular/molar polarizibility [28], which can be explained by the
following polarization phenomena.
212
'
Frequency
(Hz)
AM/FM radio
Microwave
Infrared
UV
Visual light
Figure 6-1: Dielectric dispersion of several types of polarization (Modified after

Knight and Nur (1987) [135])
1. Electronic and atomic polarizations - These polarizations are attributed

to the displacement of binding electrons/atoms from their equilibrium position
on a molecular level. These polarization processes are rapid and considered as
frequency-dependent with characteristic frequency ranging between 1012 1016
Hz (infrared, visual light, and ultraviolet (UV)).
2. Dipolar/orientational polarization - The dipolar process is the rotation

and alinement of permanent dipoles along electric field lines. Its characteristic
frequency ranges between 109 1011 Hz (microwave).
3. Interfacial/ionic polarization - The interfacial polarization is also called

the Maxwell-Wagner polarization. This process happens at the internal interfaces between two materials with different conductivities with characteristic
frequency less than 106 Hz (amplitude modulation (AM)).
213
In Figure 6-1, it is clear that, in the frequency range of interest (8 GHz18 GHz),
only the dipolar polarization is dominant. The energy dissipation/loss mechanism
in materials is attributed to two main components; conductivity dissipations and
dielectric loss.
1. Conductivity dissipations - Conductivity dissipation process occurs when
free-charge conduction is present in a dielectric. The dissipated energy is transformed into the heat of conduction. There are two types of conductivity dissipation: [270]
(a) d.c. conductivity or conductivity in the interfacial polarization At low
frequencies including direct current (d.c.) condition, the conduction in
materials may be attributed to bond conduction [245] and d.c. hopping
conduction [77]. To account for the electrical conduction, the imaginary
part (loss factor) of complex permittivity needs to add one more term in
the Debyes dispersion equation as

s
00
() = () i () +
s
0
(6.33)
where s is the d.c. (static) electrical conductivity when = 0 (S/m)

[219]. Modified dispersion equations by [50] and [63] can also be treated
the same way to account for d.c. conduction. d.c. conductivity is critical
at low frequency and high temperature situations.
(b) Conductivity in electrode polarization The polarization effect takes place
at electrodes when a dielectric displays significant bulk conductivity. Unlike the Maxwell-Wagner effect, the measured electrical conductivity increases at low frequencies. It is because of the presence of a high-impedance
layer on the electrode surface. The reasons for such layer may be attributed
to imperfect contact between the electrode and the dielectric. It is postulated that at low frequencies the conduction has sufficient time to transfer
the applied field through the dielectric and the electrode layer, resulting
in an increase in the measured capacitance. Electrode polarization occurs
214
in the frequency range up to 10 MHz and the conductivity induced by

electrode polarization can be modeled as
() =
(e )2
1 + 2 (c + e )2
(6.34)
where is the a.c. electrical conductivity when = (S/m), e is the

relaxation time of the electrode layer (s), c is the relaxation time of the
material (s) [270].
(c) Viscous conduction effect It occurs in high frequency region (around 40
GHz) as suggested by [270].
2. Dielectric loss - Field energy is dissipated by the friction between molecules/atoms
when the bound-charge molecular/atomic displacement takes place. The dissipated energy is transformed into the heat of friction due to the drift of electrons
or free ions in the applied field. The dielectric loss is proportional to the total
measured a.c. conductivity minus the d.c. conductivity.
The loss factor representing energy loss mechanism has two loss components. It
is possible to distinguish these two loss components experimentally by using the
Kramers-Kronig relations which depicts the interdependence of 0 () and 00 () [147].
Further information about dielectric polarizations and theory can be found at [65, 90,
113, 236, 245, 151].
In the case where perfect dielectric is considered, electrical conductivity is not
present or no free-ion conduction occurs. However, most materials used in civil engineering are not perfect dielectric. When an imperfect dielectric is subject to an
external field, the resulting total current density flowing through the dielectric is consisted of three parts; the instantaneous current density due to capacitance (dielectric
storage), the polarization current density due to the polarization of the material, and
the conduction current density from the electrical conductivity of the material. Electrically conductive materials are also called the lossy material and the magnitude of
electrical conductivity is called the lossness.
215
6.1.4
Dielectric Properties in Microwave and Radar NDT
Applications of dielectric properties in civil engineering have been increased during

the past decade. The contrast in the dielectric properties of different materials is
useful for distinguishing the presence of an inclusion in a matrix dielectric, as well
as for monitoring the change of material property during a period of time. In civil
engineering, reported applications of dielectric properties are summarized in Table
6.4.
Table 6.4: Civil engineering applications of dielectric properties
Application
Reference
Quality control of concrete/cement curing
[117, 263, 30, 139]
Mechanical strength determination
[114]
Corrosion detection of steel reinforcement
[224]
Pavement thickness detection
[6]
Structural health monitoring
[162]
NDT and damage detection
[66, 35, 168, 167, 88, 217, 185, 4]
6.2
Approaches for the Determination of Dielectric Properties
Dielectric properties can be determined by various existing approaches, depending on

the chosen experimental configuration. These approaches include
1. Approach I Use total current density J (A/m2 ), geometrical capacitance C0
(F), electric potential difference V0 (V), and phase difference (rad) [74].
() =
J
(sin i cos )
C0 V0
(6.35)
This is the parallel plate capacitor technique which is based on a capacitor model
for determining the permittivity of the material. It requires the specimen to
216
possess certain shape with flat surfaces on the two sides contacting the two
parallel plates. The constraints on the shape of specimen and measurement
condition limit the use of the technique, making it a laboratory method rather
than an in-situ method.
2. Approach II Use average electric displacement field vector E (V/m) and electric field [12].
1
() =
E 2
|V |2 d
(6.36)
3. Approach III Use capacitance C (F) and equivalent resistance R (O/m).

0 ()
= Q = CR
00 ()
(6.37)
where Q is the cavity quality factor or the Q-parameter [236, 244]. This is the
resonate cavity technique which is based on a resonator/oscillator model. It
provides accurate results than other broadband techniques. Only results at one
frequency are available at a time, suggesting the need for significant measurement efforts when a wide range of frequency response is needed. Additionally,
it cannot measure sample sizes greater then the cavity volume/capacity.
4. Approach IV Use waveguide wavelengths (m) and resonant frequencies
(rad/s) [184].
0
() =
2

+
2
=
2
g
2
c
(6.38)
This is a waveguide technique based on the transmission line model. Waveguides

can only operate in designed frequency bands associated with certain wave
propagation modes. Several different samples are needed when the measurement
is conducted over a large frequency range. Inaccuracy in the measurements
may occur due to the air gap between the waveguide and the specimen [178].
Waveguide techniques can be tedious in terms of sample preparation when the
designated frequency band is not available in advance.
217
5. Approach V Use average absorption microwave power W (W/m3 ) and average

electric field |E| (V/m) [70].
00 () =
2W
|E|2
(6.39)
6. Approach VI Use the amplitude and phase measurements of transmitted EM

waves through plate specimens (usually the reflection is needed as well). The
amplitude transmission coefficient T0 is defined by

p
p
d
d
0
0
T0 = exp r tan
exp i2 r
0
0
(6.40)
where d is the thickness of the plate specimen (m), 0 is the wavelength of the
incident wave in free space (m). For low-loss dielectrics (tan 104 103 )
[188], the specimen thickness d should be less than the effective wavelength
0
within the dielectric; d < e = 0 , for the ease of phase estimation. In the
r
case where d > e is inevitable or preferred, such as the use of high frequencies,
the real phase of EM waves transmitted through the specimen is
= s 2 + 1 =
2 p 0
r d
0
(6.41)
where s is an integer, and 1 h0, 2i. For the estimation of unknown dielectric
constant 0r , two plate specimens with different thicknesses are prepared. The
phase measurements provides two equations.
p
0
(s1 2 + 1 )
0r =
2d1
p
0
0r =
(s1 2 + 2 )
2d2
(6.42)
Solving for the integers s1 and s2 leads to the evaluation of 0r . With the amplitude measurements T0 (d1 ) and T0 (d2 ), the loss tangent tan can be determined.
Therefore, the complex permittivity is found.
218
6.3
Integrated Methodology for the Determination of Dielectric Properties
In this research work, an integrated methodology for the determination of complex permittivity of dielectric materials based on transmission-only, coherent, widebandwidth free-space measurements [36]. This proposed methodology, as a timedomain method, integrates the time difference of arrival (TDOA) measurement of
plate specimens and a root-searching optimization scheme using parametric system
identification (PSI) and the error sum of squares (SSE) criterion for the unique determination of complex permittivity.
Consider the normal incidence of a TE wave upon a plate specimen in the configuration as shown in Figure 6-2. In Figure 6-2, a dielectric plate specimen is placed
in the middle of two horn antennas in free space. Construction materials including
Portland cement concrete (PCC) slabs and a glass fiber reinforced polymer (GFRP)
sheet are subjected to the free space measurement. Validation of this methodology
is performed using Teflon, Lexan, and Bakelite whose dielectric properties are welldocumented in literature.
Procedure of the integrated methodology consists of the following steps: (1) collection of free-space measurements of complex transmission coefficients, (2) estimation
of the real part of complex permittivity using TDOA, and (3) determination of the
imaginary part of the complex permittivity using PSI and SSE. The overview of the
proposed methodology is provided in Figure 6-3. In the transmission model shown in
Figure 6-2 and considering the normal incidence of a transverse electric (TE) wave,
the complex transmission coefficient T is given by [138]
T =
4 exp [i(k1z k0z )d1 ]

(1 + p01 )(1 + p10 )(1 + R01 )R10 exp i2k1z d1
(6.43)
k1z
1
0 k1z
1 p01
where k0z = 0 0 , k1z = 1 1 , p01 =

=
= , R01 =
,
k0z
p10
1 k0z
1 + p01
d1 = thickness of the specimen, and 1 = 01 i001 = 0 (0r i00r ). For nonmagnetic
specimens, = 0 . Components of the methodology are further described in the
219
Network
Analyzer
Region 0
Transmitting
horn antenna
0 , 0
Region 1
Region 2
1* , 1*
0 , 0
Receiving
horn antenna
d0
d1
d0
--z-direction
Figure 6-2: A two-dimensional free-space transmission model
following sections.
6.3.1
Time Difference of Arrival (TDOA)
The TDOA technique is used to estimate the dielectric constant of a low-loss (less
conductive) material using experimental measurements of transmission coefficients.
This technique is conceptually based on the same two-dimensional model for EM
wave propagation shown in Figure 6-2. Under the assumptions of normal incidence,
first-peak response, and minor transmission losses, the time difference of arrival of an
EM plane wave due to the presence of the specimen is [241]
t =

d p 0
r 1
c
(6.44)
where t = additional propagation time (time difference of arrival) between the

transmitting and receiving horn antennas when the specimen is present compared to
220
Free-space measurements
of complex transmission
coefficients
Unique combination of the real and

imaginary parts of complex permittivity
Real part of
complex
permittivity, r'
Estimation of real part of complex

permittivity using TDOA
Parametric System Identification

and Error Sum of Squares (SSE)
Imaginary part of
complex
permittivity, r"
2.2
r'
0.5
2.1
0.4
log
(ESS)
log(SSE)
Loss Factor
0.8
0.6
r"
0.3
0.4
0.2
1.9
0.2
0.1
r'
0
1.8
'
Dieletric Constant
r
10
0.2
0.4
0.6
0.8
Loss Factor
Figure 6-3: Overview of the proposed methodology
the measurement when the specimen is not present, d = thickness of the plate/slab
specimen, and c = speed of light in free-space. The estimation of t is achieved by
processing the measured transmission coefficient from frequency-domain (recorded by
the network analyzer) to time-domain using inverse Fourier transformation. Eq.6.44
can be used as a tool for assessing the dielectric constant of the specimen by estimating
its time difference of arrival using a set of experimentally measured transmission
coefficients. With Eq. (6.44), the real part of complex permittivity or dielectric
constant is found.
0r

=
ct
1+
d
2
(6.45)
The accuracy of the t estimation depends on the bandwidth of the signal being processed and the accurate measurement of the specimen thickness. Note that Eqs.(6.44)
and (6.45) must be modified when the loss factor is significant. This estimate will
be the basis for the identification of the imaginary part (or loss tangent) of complex
221
permittivity in the root-searching optimization scheme.
6.3.2
Root-searching Optimization Scheme
Complex permittivity cannot be explicitly represented in terms of the transmission
coefficient T or the scattering parameter (S-parameter),S21

, which is the forward
transmission gain measured by the transmitting and the receiving radar antennas.
Furthermore, the exact solution for the complex permittivity is not straightforward
due to the multiple roots associated with Eq.(6.43) for lossy materials. In order to
resolve the problem, a root-searching procedure involving the use of PSI and the SSE
criterion is proposed. PSI refers to the use of a mathematical model to characterize the
behavior of a system [266], and in this application it is used to provide the theoretical
estimation of transmission coefficients. Thereafter, SSE criterion is introduced for
evaluating the optimal combination of complex permittivity.
Experimental measurements were conducted in coherent condition where mea
, equals the complex transmission coefficient, T .
sured complex S-parameter, S21
Coherent condition provides a non-distorted phase in the measurement of S21

(be-
tween antennas 2 (receiver) and 1 (transmitter)) such that amplitude attenuation

within the specimen does not contribute to the measured transmission coefficient.
Experimental facilities for the free-space measurement are shown in Figures 6-4 and
6-5.
In coherent condition, the magnitude of T expressed in decibel (dB), TdB , is
related to S21
by
TdB = 10 log S21

S21
(6.46)
where TdB is a real number, and S21

is the complex conjugate of S21
. TdB is the
measured transmission coefficient or TdB = T m Eq. (6.43) provides the form of

theoretical/predicted transmission coefficient collected in decibel, or T = T p . With
the correct value of complex permittivity, the predicted T p will be identical to the
measured T m . The root-searching optimization scheme is introduced to determine
the closest value of T p to T m by minimizing the SSE between T p and T m . PSI
(parametric system identification) is carried out by generating a set of T p from the
222
(a) Transmitting horn antenna
(b) Network analyzer
Figure 6-4: Transmitting horn and the network analyzer for the free-space measurement [Courtesy of the MIT Lincoln Laboratory (MIT LL)]
possible combinations of 0r and 00r . From the possible combinations of 0r and 00r , those
resulting in minimum difference between T m (measurement) and T p (theory) will be
the estimates closest to the actual values.
SSE (error sum of squares) criterion is applied in this root-searching procedure.
Each combination of the estimated 0r and 00r provides a corresponding SSE value,
which is calculated by
SSE
(0r , 00r )
n
X
|T m () T p (, 0r , 00r )|
(6.47)
j=1
where n is the number of measurement frequencies. It is obvious that SSE evaluates

the absolute error in an energetic manner; the closest prediction possess the minimum
energy with the measurement. With the SSE criterion, an error surface is generated
for various combinations of 0r and 00r . Figure 6-6 shows the error surfaces generated by the free-space measurements of four dielectric materials including Teflon,
Lexan, Bakelite, and Portland cement concrete. Discussion of the result is provided
in the validation section. Using the estimated dielectric constant (from TDOA) a
corresponding error curve containing various combinations of loss factor can be lo223
(a) Transmitting horn
(b) Plate specimen
(c) Receiving horn
Figure 6-5: Transmitting and receiving horns for the free-space measurement [Courtesy of the MIT LL]
cated from the error surface, as shown in Figure 6-7. The most possible loss factor
is determined by selecting the one with minimum error on the curve. Hence, the
root-searching procedure is accomplished.
224
1
6
5
0.8
0.8
0.4
2
Loss Factor
Loss Factor
4
4
0.6
0.2
0.6
3
0.4
2
0.2
10
Dieletric Constant
10
Dieletric Constant
Dielectric Constant
Dielectric Constant
(a) Teflon
(b) Lexan
1
0.6
0.4
0.8
0.8
0.6
0.3
0.4
0.2
0.2
0.1
Loss Factor
Loss Factor
0.5
0.6
0.4
0.3
0.4
0.2
0.2
0.1
10
Dieletric Constant
Dielectric
Constant
10
Dieletric Constant
Dielectric
Constant
(c) Bakelite
(d) Portland cement concrete
Figure 6-6: Error surfaces of the dielectric measurements collected from Teflon, Lexan,
Bakelite, and Portland cement concrete
225
6.6
log (SSE)
log (SSE)
6.5
5.5
6.2
5.8
5
5.4
4.5
0.05
0.1
0.15
0.2
0.05
Loss Factor
0.1
0.15
0.2
Loss Factor
(a) Teflon
(b) Lexan
2.6
2.2
2.1
log(SSE)
log
(SEE)
log (SSE)
2.15
2.05
2.4
2.2
1.95
1.9
0.2
0.4
0.6
0.8
0.2
0.4
0.6
0.8
Loss Factor
Loss Factor
(c) Bakelite
(d) Portland cement concrete
Figure 6-7: Error surfaces of the dielectric measurements collected from Teflon, Lexan,
Bakelite, and Portland cement concrete
226
6.3.3
Validation of the Methodology
In order to evaluate the predictions from the proposed methodology, experimental

measurements of transmission coefficients were conducted for several materials whose
dielectric properties are known from the literature. These materials are Teflon, Lexan,
and Bakelite. In addition, Portland cement concrete and GFRP specimens were also
measured as examples of construction materials.
Sample Description and Experimental Configuration
For the measurements, slab-type material specimens of 12in-by-12in (305mm-by305mm) cross-section with varying thicknesses were used. The selected width and
height of the specimens meet the requirement of radar measurements in the far-field
condition. Table 6.5 shows the various specimen thicknesses for the materials used.
The concrete sample was manufactured with a cement/sand/aggregate mix ratio of
1:2.25:3.2 by weight. The water-to-cement ratio (w/c) was 0.60. Portland cement of
Type I was used. The uniaxial compression strength of the concrete was 3480 psi (24
MPa) at 28 days. The GFRP sample was manufactured by extruding a 12in-by-12in
square from a unidirectional fiber woven sheet, which was then saturated with an
epoxy thermoset polymer. The GFRP sample was cured for 7 days before testing.
The experimental configuration is shown in Figures 6-2 and 6-4. The network anaTable 6.5: Thicknesses of the specimens
Material
Thickness (mm)
Teflon
Lexan
Bakelite
Portland cement concrete
50
GFRP
1.5
lyzer was a Hewlett Packard Model 8510C that was operated in step frequency mode.
At each frequency the network analyzer performs a measurement of the complex
227
transmission coefficient.
Transmission measurements were collected from X-band through Ku-band (8 to
18 GHz) with a frequency step of 12.5 MHz. A typical set of measurements consisted
of amplitudes (dB), phase angles (deg), and their corresponding frequency. Raw
data was then calibrated using free-range measurements, which were conducted in
the absence of the specimen, to obtain the TDOA information. Far-field test conditions were ensured during measurements. The far-field condition is required to
ensure that the wave front is approximately plane, which is directly related to the
theoretical methodology for dielectric property characterization. Far-field conditions
also minimize complex wave behavior in near-field between the horn antennas and
the specimen. Considering the highest frequency of 18 GHz for the proposed experiments, it was calculated that the specimen should be placed at least 1.8 in. (45.72
mm) away from the horn antennas, and the minimum area of the slab specimens
should be greater than 25 in2 (161.29 cm2 ) for adequate illumination to satisfy the
far-field condition. The applied experimental set-up meets these requirements. Direct coupling between the horn antennas is also eliminated in this experimental set-up
due to the use of the network analyzer. The unwanted coupling between antennas
is measured in free-range and stored in the phase angle information collected by the
network analyzer. This information is then used in the calibration of all consequent
measurements.
Estimation of the dielectric constant using TDOA

The TDOA technique was used to estimate the dielectric constant of materials under
investigation from both X- and Ku- bands. After performing inverse Fourier transformation of the experimental transmission coefficient data, the time differences of
arrival (t) for the materials considered in this study were calculated. These results
are tabulated in Table 6.6. Using the TDOA information, application of Eq.(6.45)
yielded estimates for dielectric constants, which are also tabulated in Table 6.6. For
the case of GFRP the TDOA technique could not provide a reliable result due to the
very small thickness of the specimen.
228
Table 6.6: TDOA measurements and dielectric constants 0r of the specimens
Material
t (ns)
Estimated 0r
Reported 0r
Teflon
0.007
1.79
2.0
Lexan
0.012
2.48
2.77
Bakelite
0.030
5.95
5.0
0.235
5.95
4.447.22
Root-searching results of loss factor

Following the previously described procedure, the loss factor can be found by extracting an error curve using TDOA from the error surface generated by PSI and
SSE criterion. The loss factor corresponding to the minimum SSE on the error curve
is identified as the most appropriate value. Estimated loss factors thus found are
given in Table 6.7 for the tested materials.
Table 6.7: Loss factors 00r of the specimens
Material
Estimated 00r
Teflon
0.04
Lexan
0.04
Bakelite
0.28
0.62
Discussion
From the results, it is found that a unique combination of real and imaginary parts of
the complex permittivity can be found through the proposed methodology for low-loss
materials, such as Teflon, Bakelite, Lexan, and dried/hardened concrete in this study.
The non-uniqueness problem of real and imaginary parts of complex permittivity has
been resolved, and appropriate values were predicted using the integrated methodology. The method has potential for in-situ measurement of dielectric properties for
229
construction materials.
The estimated results of complex permittivity using the proposed methodology
are summarized in Tables 6.6 and 6.7. The estimated dielectric constant values of
Teflon, Lexan, Bakelite, and concrete are close to the reported values or within the
ranges reported in the literature. It is emphasized that the values found using the
proposed methodology for dielectric constants and loss factors are not expected to
exactly correlate with the reported values in the literature because of the differences in
frequency ranges and measurement techniques. Note that in Table 6.7 the loss factor
comparison could not be made because of the lack of information on loss factors in the
literature. In addition, loss factor values are generally more sensitive to measurement
frequency and experimental parameters.
Two main issues in this methodology are addressed as follows.
Effect of selected frequency bands in TDOA TDOA estimates for the dielectric property characterization of the materials presented in this study were
obtained using the entire set of measurements for the X- and Ku-band. For
practical applications of this methodology, the question arises as to what would
be the optimum frequency range to provide a sound estimate for the dielectric constant. Calculations were conducted in 1 GHz, 2 GHz, 3.3 GHz, and 5
GHz intervals covering the range from 8 GHz to 18 GHz to study the effect
of frequency bandwidth. Table 6.8 presents a summary of different estimates
of dielectric constant using TDOA that were calculated in 1 GHz bandwidth.
The convergence of these estimate curves is evaluated by the variance shown in
Figure 6-8. It is observed that estimates calculated from narrow frequency bandwidths may provide poor estimates for dielectric constant. Accurate estimates
using the TDOA technique require the use of wide frequency bandwidths for
uniquely identifying the dielectric constant. Reliable estimation was achieved
when frequency bandwidth used exceeded 3.3 GHz in our measurements. The
accuracy of such estimation is critical to the subsequent determination of the
loss factor.
230
Table 6.8: Estimated 0r by TDOA using different frequency bandwidths (GHz)
Material
8-9
910
1011
1112
1213
1314
1415
1516
1617
1718
Teflon
1.42
5.80
1.89
1.89
2.37
1.18
1.32
2.28
2.28
2.34
Lexan
6.05
6.96
2.25
2.53
2.06
4.53
3.61
2.41
3.23
Bakelite
0.96
14.75 5.38
8.86
8.17
5.45
4.69
9.76
Concrete
4.80
6.03
6.64
5.01
4.66
8.04
6.76
5.99
4.47
5.30
Limitations of the use of TDOA
1. Effective thickness of specimen - From the results for GFRP, it is observed that the thickness of the sample affects the performance of TDOA
technique. Thus, a minimum thickness limitation might apply to the accurate characterization of dielectric constant of construction materials using
the proposed methodology. Our preliminary investigations indicate that
measurements of specimens of 6 mm or larger would provide acceptable
results.
2. Accuracy of the use of simplified wave velocity - From EM wave
theory it is known that the wave (phase) velocity of EM waves is a function
of dielectric properties of materials. For example, for lossy materials, the
phase velocity within the medium is [138]
"
vp =
1
1
=
kR
2
!#1/2
2
+1
1+

(6.48)
where kR is the real part of the complex wavenumber. The first-order

expansion of vp provides its approximation as

1
1
1 2
vp =
1+

8
231
(6.49)
25
Teflon
Lexan
Bakelite
Concrete
20
Variance
15
10
0
1
5
6
7
Frequency Bandwidth
10
Frequency bandwidth (GHz)

Figure 6-8: Convergence of estimates of 0r at different frequency bandwidths using
TDOA
For lossless materials = 0, and the wave velocity becomes

1
c
vp = = 0 0

r r
(6.50)
Eq.(6.50) is used in the TDOA procedure described in this paper because

of its simplicity, and it is considered as a convenient approximation of
1. To study the accuracy of this approximation, a

Eq.(6.49) when

quantitative study was performed. Wave velocities using Eqs.(6.48), (6.49),
and (6.50) are calculated with respect to different values of

ranging

from 0 to 20, shown in Figure 6-9. It is observed that the differences
between Eqs.(6.48) and (6.50) or between Eqs.(6.49) and (6.50) may be
substantial when
is greater than 1. In the range of investigation where

< 0.1, the difference between either Eqs.(6.48) and (6.50) or Eqs.(6.49)

and (6.50) is insignificant as demonstrated in Figure 6-9(b). Hence, the use
232
of Eq.(6.50) in the proposed methodology for low-loss materials is justified.
Normalized Phase Velocity
Normalized Phase Velocity
1.001
Theoretical vp in lossy media
Approximated vp in lossy media
vp in lossless media
0.8
0.6
0.4
0.2
10
15
20
0.999
0.998
0.997
Theoretical vp in lossy media
Approximated vp in lossy media
vp in lossless media
0.996
0.995
/ ( )
0.02
0.04
0.06
0.08
0.1
/ ( )
(a) = 0 ~ 20
(b) = 0 ~ 0.1
Figure 6-9: Normalized phase velocity vs. loss factor 00r
6.4
Modeling Approach for the Dielectric Properties of Materials
In the microwave frequency range the predominant polarization is the dipolar polarization which can be derived from the internal field and the external field of the
dielectric. A geometrical analysis technique proposed by Cole and Cole (1941) [50]
called the Cole-Cole diagram is also introduced for illustrating the dispersion response
in a complex plane. Equivalent circuits of these models are also provided for better
understanding of their physics.
6.4.1
Internal Field Approach versus External Field Approach
Internal Field Approach Debyes Treatment

A systematic approach for the modeling of dipolar polarization for homogeneous and
isotropic liquid and solid dielectrics in microwave/radar frequencies was proposed by
233
Debye [65]. The molar polarization P in liquids is derived and defined as

4N
1M
=
P () =
+2
3
2
1
0 +
3kT 1 + i

(6.51)
where is the electrical permittivity (F/m), M is the molecular weight (kg), is

the density, N is Avogadros constant (N = 6.06 1023 = number of particles in
one mole), 0 is a factor accounting for the distortion effect of the molecule, is the
intrinsic electric moment (e.s.u.=300V/cm), T is the absolute temperature (K), and
k is Boltzmans constant (k = 1.37 1016 ergs).
is also called the molar volume
(cc or cm3 ). The electrical permittivity is found to be [65]
P ()
M
=
1 P ()
M
1+2
(6.52)
Two relationships were also reported by Debye (1929).

4N 0
1 M
=
+ 2
3

s 1 M
2
4N
=
0 +
s + 2
3
3kT
(6.53)
(6.54)
With these relationships, becomes

s

+ i
+2
+ 2
= s
1
s
+ i
s + 2
+ 2
(6.55)
Assuming
=
s + 2
+ 2
(6.56)
then the representation of becomes

=
s
0 i00
s
=
1 + i
s
1+i
234
(6.57)
It yields to
=
s
1 + i
(6.58)
where
s
= 0 ()
2
1+
(s )
00 =
= 00 ()
2
1+
0 = +
(6.59)
(6.60)
and

=
+ 2
s + 2
00
0
(6.61)
is a characteristic parameter in Debyes model. Eqs. (6.59) and (6.60) are called
Debyes dispersion equations.
Debyes treatment on deriving the complex permittivity was based on the internal field with Mosottis hypothesis. Mosotti considered the molecule as a polarizable
system, in which the polar moment in an electric field is set up by the force causing
a displacement of the charge in the material. Debye found that, in his dispersion
equations for ideal liquids, Mosottis hypothesis about the internal field was not totally correct, especially when dense dielectrics are considered. However, the difference
between his theory and experimental results was only a quantitative but qualitative
deviation. Since Debyes model provides a good estimation for dilute solvents in general, it is been widely applied as the paradigm model for gaseous and liquid dielectrics
ever since.
External Field Approach Frohlichs Treatment

Unlike Debyes treatment whose relaxation time is made based on the microscopic or
molecular relaxation process, Frohlich (1949) [90] chose another approach in which
the relaxation time is based on a macroscopic process. The time-dependent relaxation
process of dielectric dispersion can described as the convolution between the electric
235
field and a decay function. For t > u + du,

D(t u) = (t u)E(u)du
(6.62)
where D(t u) is the electric displacement (C/m2 ) or dielectric induction, (t u)

is the decay function describing the gradual decrease of D(t u) and (t u) 0
when t , and E(u) is the electric field (V/m). The decay function (t u)
also contains an instantaneous response component which is assumed to be E(u).
Thus, for u < t < u + du,
D(t u) = E(u) + (u)du
(6.63)
Should another field E(u0 ) be applied at a later time interval between u0 and u0 + du0 ,
the corresponding response is linearly superimposed onto the former one (the principle
of superposition). Consider a general case in which D(t) and E(t) may not vanish for
t < 0. The response D(t) at the time t is given by
Z
E(u)(t u)du
D(t) = E(t) +
(6.64)
Assume that the decay function (t) is an exponential function.

t
(t) e
(6.65)
where is the relaxation time observed in the macroscopic process and

d(t)
1
= (t)
dt
(6.66)
Differentiating the response function provides

dE(t)
1
dD(t)
=
+ (0)E(t)
dt
dt
236
E(u)(t u)du
0
(6.67)
After rearranging, it yields
d
[D(t) E(t)] + [D(t) E(t)] = (0)E(t)
dt
(6.68)
Consider the attained equilibrium under the application of a constant field. It provides
d
(D E) = 0
dt
D = s E
(6.69)
(6.70)
Therefore, we have
(0) = s
(6.71)
Since Eq.(6.65), (t) is found to be

t
s
(t) =
e
(6.72)
With these relationships,
d
(D E) + D E = (s ) E
dt
(6.73)
Consider an alternating field of the form E eit . With

dE
= iE
dt
(6.74)
dD
= i()E
dt
(6.75)
and Eq.(6.1), it yields
Substituting Eq.(6.65) into Eq.(6.64), along with E eit , we have

s
()E = E +
237
iu
e
tu
du
(6.76)
Let t u = x. The limit of the integral is changed from [, t] to [, 0]. Also,

du = dx. Then,
Z
s 0 i(x t)
e
()E = E
Z
s it ix
()E = E +
e
e
0
Z
ix
s
e
E
()E = E +
0
Z
ix
s
() = +
e
x
dx
x
dx
x
dx
x
dx
(6.77)
The response function becomes

() =
s
1 + i
(6.78)
and the real and imaginary parts are

s
0
2 = ()
1 + ( )
(s )
00 =
= 00 ()
1 + ( )2
0 = +
(6.79)
(6.80)
00
. The above equations are Frohlichs dispersion equations derived
0
from the equilibrium of an external field E. Frohlichs equations are also known as
where =
the Debye-Drude equations [107]. Frohlich further generalized the variation of () by
accounting the contribution to the static dielectric constant of the groups of dipoles
having individual relaxation times in a range d near . He also assumed that the
contributions of various groups of dipoles can be linearly superposed and modeled
by a function y( )d . The total contribution to the static dielectric constant is then
determined by
Z
s =
y( )d
0
238
(6.81)
where y( ) is termed the distribution function. Thus, Eqs.(6.79) and (6.80) can be
generalized as
0 () = +
Z
0
00 () =
Z
0
y( )
d
1 + ( )2
(6.82)
y( )
d
1 + ( )2
(6.83)
Cole and Cole (1941) [50] proposed a modified model for liquid and solid dielectrics:
= +
s
1 + (i )1 1
(6.84)
where 1 is an empirical parameter and 0 1 1. Davidson and Cole (1950)

proposed another model also based on their dielectric measurements on liquids and
dielectrics:
= +
s
(1 + i )
(6.85)
where is an empirical parameter and 0 1. A parametric study is conducted

to understand the influence of 1 on 0 and 00 . By assuming s = 2 and = 1 the
variations of 0 and 00 with respect to parameters 1 (0 1 1) and (0 1)
in the Cole-Cole model are shown in Figures 6-10 and 6-11 .
The Cole-Cole and Davidson-Cole models were generalized by Havriliak-Negamis
model (1966)[108]:
= + h
s
i
1 + (i )2
(6.86)
where 2 and are empirical parameters and 0 2 1, 0 1. Havriliak and Negamis model has been also used for materials exhibiting non-Debye-type
relaxations such as ionic conductors and magnetic fluids [282].
Other models such as the Fuoss-Kirkwood model (1941) [91](for giant molecule
systems such as long-chain polymers), the Williams-Watts model (1970) [269] (for general polymer systems), the Jonscher model (1975)[125] (for temperature-dependent
239
Figure 6-10: Behavior of the Cole-Cole model (s = 2, = 1)

relaxation response of polymers), and the Dissado-Hill model (1979) [89] (for general
polymers) are not investigated in this research because (1) they are developed primarily for polymers, and (2) they require the measurement of empirical parameters
used in the model.
Comparison of Internal Field and External Field Approaches

Dispersion equations derived by the internal field approach (Debyes treatment) and
external field approach (Frohlichs treatment) are compared in Table 6.9. Since <
+ 2
s ,
< 1. This suggests that the relaxation time derived from internal fields is
s + 2
greater than the one from external fields.
Many derivative models have been proposed and reported based on experimental
measurements on different dielectrics in the following years. Most of the models basically follow Debyes approach due to the sound physical foundation and the concise
mathematical expression offered by the approach.
240
Figure 6-11: Behavior of the Davidson-Cole model (s = 2, = 1)
However, in Debyes treatment, a prior knowledge on the distribution of internal

fields is needed for determining the molar polarization P (). This is difficult to obtain
in the situation where close-range interactions between two different molecules become
significant. The distribution of internal fields, as well as the relationship between
internal and external fields, tends to be complicated. In addition, the molecular-level
variables considered in Debyes treatment of the complex permittivity are not easy to
obtain experimentally, although the picture depicted in Debyes treatment provides a
clear understanding on how dielectric dispersion occurs associated with the inherent
properties of the materials (molecules).
The external field approach, on the other hand, avoids the difficulty on assuming
the distribution of internal fields. It is adopted in this research since the experimental
radar measurements were conducted in macroscopic level, rather than in molecular
level. While there are many wide-bandwidth dielectric models currently available to
use, this research is interested in the dielectric behavior of the materials in GFRP241
Table 6.9: Comparison of internal field approach and external field approach
Variable
0
00
Internal field
External field
(Debyes model)
(Frohlichs model)
+
s

2
s + 2
1+
+ 2

s

2
s + 2
1+
+ 2
+ 2
s + 2
s + 2
+ 2
00
0
+
s
1 + ( )2
(s )
1 + ( )2
00
0
concrete systems in microwave range (X-band (8 12 GHz) and Ku-band (8 18

GHz). The entire dielectric spectroscopy of the considered materials is not in the
scope of this modeling effort as to capture all dielectric dispersion mechanisms within
the frequency range of interest. Knowing that in the frequency range of 8 18 GHz
only dipolar polarization is most significant, the Havriliak-Negami model is chosen as
the prototype model to capture the monopole (single peak) dispersion of the materials.
6.4.2
Geometrical Analysis
An illustration of (0 , 00 ) in complex domain was proposed by Cole and Cole (1941)
for demonstrating the dispersion behavior of dielectrics, which is thereafter termed
the Cole-Cole diagram. The Cole-Cole diagram is formed by assembling 0 and 00
on a complex plane using frequency as an implicit variable. For dielectrics satisfying
Frohlichs equations their Cole-Cole diagram is a semicircular plot; for those obeying
Cole-Coles and Davidson-Coles models, their Cole-Cole diagrams are the deformed
version of a semicircle, as shown in Figures 6-12 and 6-13 in which s = 2 and
= 1. It is obvious that Cole-Coles model and Davidson-Coles model degrade to
Frohlichs model when 1 1 = 1 and = 1. In addition, the variation of 1 can
242
be depicted by the equation of circle with different radiuses and centers, while the
curves with a changing are no longer circular. Consider the Frohlich model. Since
1- 1=1
1- 1=0.2
0.5
1- 1=0.5
1- 1=0.7
0.4
1- 1=0
"()
0.3
0.2
0.1
-0.1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
'()
Figure 6-12: The Cole-Cole diagram of Cole-Coles model

0max = 0 (0) = s and 0min = () = , the corresponding imaginary parts are both
d00
00
00
diminished. It can be seen that min occurs at = 0, and max occurs at 0 = 0.
d
d00
= 0 can be determined by
d0
d00
d00
=
d0
d
d0
d
1
1 ( )2
=
2
(6.87)
It is evident that 00max occurs at = 1 ( must 0). Knowing

that =

00
1

+

1

s
s
, we have 00 = 0 or 0
=
and 00
=
. The
0

2
governing equation for the semicircles in the Cole-Cole diagram for Frohlichs model
is

2

2
s +
s
2
00
0
+ () =
(6.88)
()
2
2

s
s +
The semicircles are of radius
and centered at
, 0 An equivalent
2
2
circuit to Frohlichs model consisting of two capacitors and one resistor is provided
243
=0
=0.2
=0.5
=0.7
=1
0.5
0.4
"()
0.3
0.2
0.1
-0.1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
'()
Figure 6-13: The Cole-Cole diagram of Davidson-Coles model

00 1
when = 1.
0
Two auxiliary vectors are defined for the semicircles in the Cole-Cole diagram; ~u and
in Figure 6-14. The relaxation time is determined as =
Radius =
" ( )
s
= ( , )
G
u
G
v
( )
'
s +
Figure 6-14: The Cole-Cole diagram and the equivalent circuit of Frohlichs model
~v . From the geometry it is known that |~u + ~v | = s . Since ~u is the difference

between points (0 , 00 ) and ( , 0), u = and v = s . With ~u and ~v , the
244
governing equation for semicircles can be represented by

v u + 2

2
0
2
+
002

=
u+v
2
2
(6.89)
and Frohlichs model is represented by

s
u+v
v
= 1 + i =
=1+

u
u
(6.90)
v
s
= i =
u

(6.91)
Therefore,
As for Cole-Coles model, with i11 = e
i(11 )
2
, it can be rationalized to yield

1 + ( )11 sin 21
0
=
s
1 + ( )2(11 ) + 2 ( )11 sin

( )11 cos 21
00
=
s
1 + ( )2(11 ) + 2 ( )11 sin
1
2
(6.92)
1
2
(6.93)
To eliminate , the above equations are rearranged.

0

s
2(11 )
(
)
1
1
i
= 1 h
2(11 )
11
1
1
2
(
)
+
1
+
(
)
sin
2
2
h
i
11
(11 )
1
(
)
(
)
2
1
i
= 1 h
11
(11 )
1
1
2
( )
+ ( )
+ sin 2
2
"
#
1
sinh x

1
=
2
cosh x + sin 21

cos 21
00
1

=
s
2 cosh x + sin 21
1
2
245
(6.94)
(6.95)
where x = (1 1 ) ln ( ). Therefore,
0 = +
s
sinh x

1
1
2
cosh x + cos
2

1
cos
s
2
00 =
1
2
cosh x + sin
2
(6.96)
(6.97)
It can be shown that

s + 20
1
sinh x =
cos
00
2

2
s
1
1
cosh x =
cos
sin
00
2
2
2
(6.98)
(6.99)
With cosh2 x sinh2 x = 1, the governing equation of the semicircles is

s +
0
2
2

s
(s )2
1
1
00
tan
+ =
sec2
+
2
2
4
2

(6.100)

s
s + s
1
1
,
tan
. with radius
sec
.
Its center is located at
2
2
2
2
2
The ratio between two auxiliary vectors is

v
=
u
(s 0 )2 + 002
(0 )2 + 002
(6.101)
and the relaxation time is determined by

=
v 11
u
(6.102)
The Cole-Cole diagram and the equivalent circuit of the Cole-Cole model are shown
in Figure 6-15. For Davidson-Coles model,
= +
s
(1 + i )
246
(6.103)
Radius =
" ( )
s
2
= ( , )
(1 1 )
G
u
1
2
G
v
s +
' ( )
s
( i )
s
1
Figure 6-15: The Cole-Cole diagram and the equivalent circuit of Cole-Coles model
Following similar rationalization process done to Cole-Coles model, it yields
0 = + (s ) cos y cos (y)
(6.104)
00 = (s ) cos y sin (y)
(6.105)
where y = tan1 ( ). Figure 6-16 shows the behavior of Davidson-Coles model and
its equivalent circuit. The slope on the curve in the Cole-Cole diagram is
d00
= cot (1 + ) y
d0
(6.106)
In the following sections, geometrical analysis is applied for developing dielectric

models of the considered materials.
6.5
Dielectric Properties of Water
Dielectric properties of water have been a major research topic in physics and chemistry ever since the work by H. Cavendish and A. Lavoisier in 1780s. It is because
that, through the dielectric properties of water, the structure and behavior of water
molecules can be better understood. While salient research results can be found in
the literatures, due to different interests from various disciplines [65, 76, 113], the
purpose of this investigation is to develop a dielectric model for water specifically in
247
( i )
0.4
0.35
0.3
0.25
"()
" ( )
0.2
0.15
(1 )
0.1
0.05
2
1
1.2
1.4
1.6
( )
' '()
1.8
Figure 6-16: The Cole-Cole diagram and the equivalent circuit of Davidson-Coles
model
the frequency range of 8 GHz18 GHz. In this dissertation, dielectric modeling of

water is conducted for the water molecules inside a cementitious composite like concrete. Therefore, it is important to know the water molecules reside in concrete and
what they are. The following paragraphs briefly defines the types of water molecules
in concrete according to the microstructure of concrete.
Hydrated concrete is a porous, colloidal material (hydration products) whose pore
structure permits air (gaseous) and water (liquid) molecules to reside inside concrete,
making concrete a multi-phase cementitious composite. Sources of liquid water inside
concrete include (i) extra water remaining from hydration, being trapped in the pore
structure of concrete, and (ii) absorbed water from the environment (e.g., rain water).
Typically, the porous, colloidal structure of concrete is characterized by the undifferentiated amorphous C-S-H (Calcium-Silicate-Hydrate), a hydration product having
a major contribution to the binding of other components of concrete, including fine
(sand) and coarse (gravel) aggregates. C-S-H gel is recognized as a solid made up of
small particles of colloidal dimensions (at least one dimension). These particles are
mainly the hydrates of calcium silicate, which are poorly crystalized. The formation
of pore structure of C-S-H is generally regarded as a function of various factors (e.g.,
duration of hydration, cement composition, cement fineness, water/cement (w/c) ratio, temperature, and admixtures of concrete), and certainly it varies with time during
248
the setting and hardening of concrete.

Since the C-S-H gel (cement paste) is considered the major element binding other
components in concrete, it is also considered covering the surface of other components in concrete. As a result, variation of water contents in other components (fine
and coarse aggregates) is assumed to be constant once the cement paste is hardened/hydrated. Therefore, in this research, it is assumed that the variation of water
content in the C-S-H gel or cement gel results in the change of bulk dielectric properties of concrete. Following this assumption, the next step is to understand how
hardened/hydrated cement paste (hcp) can be modeled and what types of water can
present in the pore structure of hcp.
Three classical models on hardened/hydrated cement paste are reviewed and summarized as follows.
Powers model [209, 206, 208]
1. Cement gel is a rigid substance made up of colloid-size particles, possessing
a characteristic porosity of 28% which is limited to normally cured cement
pastes.
2. Two types of pores are modeled; (i) gel pores (average width = 15
A) and
(ii) capillary pores (bigger pores).
3. Three types of water are modeled; (i) chemical/combined/non-evaporable
water, (ii) gel water, and (iii) capillary water.
4. At later stages of hydration, the specific surface area is independent of the
degree of hydration.
5. Based on experimental water-vapour isotherms. On heating at 23 C, gel
and capillary water can be fully removed. On heating at 105 C, part of
the non-evaporable water can be removed. On heating at 105 C, the nonevaporable water is removed.
Ishais model [120]
249
1. In addition to the chemical water, four types of evaporable water are modeled; (i)intracrystalline(zeolitic) water (thickness less than 4
A), (ii) water
adsorbed on the surface of the crystallites (thickness between 4
A and 8
A),
(iii) water confined between adjacent crystallite surfaces (thickness less
than 8
A), and (iv) capillary water (at a distance greater 10
A 20
A).
2. Capillary water is outside the range of surface forces .
Feldman and Seredas model [84]
1. In addition to the chemical water, two types of evaporable water are modeled; (i) interlayer hydrate water, and (ii) physically-adsorbed water in gel
and capillary pores.
2. Water-vapour isotherms for hardened cement pastes are irreversible. The
BET method is not applicable.
3. Variation of measured Asp by different methods is recognized. The use of
nitrogen for Asp determination results in much lower values of Asp than
using water since the size of nitrogen molecules (4.05
A) is greater than the
one of water molecules (3.25
A).
4. Evaporable water can reside in the interlayer pores of the C-S-H.
While many other hcp models have been proposed since these classical models became
available, essential differences among the models are: (1) classification of the water
held in the C-S-H structure, (2) assumed importance of different types of water in
the properties of the C-S-H structure and cement pastes, and (3) distribution of the
pore sizes and configuration of the pores. In this research, two types of water inside
concrete are considered:
1. Free water Water molecules that are subjected to gravitational field only and
can respond to the application of an external EM field without any delay in
their relaxation time.
2. Bound water Water molecules that are subjected to both gravitational field
and the attraction field (chemical potential for adsorption) from the solid phase
250
of cementitious composites, displaying certain amount of delay in their relaxation time when subjected to the application of an EM field.
The amount of bound water is determined by the product of the specific surface
area (Asp ) of hcp and the thickness of adsorbed, bound water layer/film. Adsorption
isotherms are typically used for determining the specific surface area of a material.
The BET (Brunauer-Emmett-Teller) method is usually applied for hardened cementitious composites [221].
In the dielectric modeling of water, it is also assumed that (1) Only liquid water
is considered. Water vapour and ice are not considered; (2) Water can only reside
in concrete, not in GFRP-epoxy composite; (3) Only the water presenting in the
pore structure of C-S-H is considered for modeling. The water in fine and coarse
aggregates is not considered; (4) Chemical or non-evaporable water has no impacts
on the measured dielectric properties of concrete with different water contents; (5)
Free water and bound water are both evaporable water. These assumptions are
considered reasonable for the dielectric modeling of materials in the frequency range
of 8 GHz to 18 GHz in room temperature. In what follows, dielectric modeling of free
water and bound water is provided.
6.5.1
Free Water
In this research, free water is defined as the water molecules sorely subjected to the
gravitational field externally and the O-H bond energy internally. The O-H bond
energy is half the energy of formation of water molecule since water has two O-H
bonds, whose value is 109.7 kcal/mol at 0 K (-273 C). This bond energy is related
to the relaxation time (rad-sec) of water. Our interest of free water lays in its
frequency-dependent dielectric dispersion.
For the dielectric modeling of free water using Havriliak-Negamis model, parameters to be determined are s , , , and 1 . Reported dielectric properties (complex
permittivity; 0r and r ) in the frequency range of 0.58 GHz and 134.9 GHz are quoted
from the literature and used for obtaining the parameters. These measurements are
251
listed in Table 6.10.

Table 6.10: Measured complex permittivity of water at 20 C
Frequency (GHz) 0r (0 ) r (0 )
0.581
1.741
32
3.651
9.32
9.6253
19.253
23.812
33.93
344
70.23
96.772
134.93
80.3
79.3
77.2
76.3
61.5
63.5
41.5
30.8
23.1
19.5
9.5
8.5
5.64
2.75
7.9
13.1
15.6
31.4
29.1
36.5
35.2
31.2
30
19.97
12.0
10.42
Using Havriliak-Negamis model and the mean-square error (MSE) criterion for
curve fitting, the fitted parameters are obtained at the minimum MSE.
()
71.84
= 9.34 +
1
0
1 + (i9.23)0.93
(6.107)
where the relaxation time =9.23 rad-ps (1 ps = 1 picosecond = 1012 seconds)

when (rad/sec) is used, or =9.23 ps when f (1/sec) is used ( is dimensionless).
Obtained is close to the value (9.26 rad-ps or 9.26 ps) reported by Grant et al.(1957)
[75] in 20 C. This model becomes a Cole-Cole-type model since = 1. Result of curvefitting is in good agreement with measurements as shown in Figure 6-17. Reported
model parameters in Debyes model and Cole-Coles model are also summarized in
1
[75]
[107]
3
[153]
4
[96]
2
252
40
"r
30
20
Grant et al. (1957)
Hasted & Shah (1964)
Grant & Shack (1967)
von Loon & Finsy (1975)
Havriliak-Negami' model
10
10
20
30
40
50
'
60
70
80
90
Figure 6-17: Performance of the Cole-Cole model for free water

Table 6.11 for comparison.
Table 6.11: Fitted parameters in Debyes and Cole-Coles models of free water
T ( C)
s /0
/0
(ps/rad)
f (GHz)
[235]
[20]
[75]
19
19
060
81
81
[220]
[107]
[143]
[82]
20
20
78.5
81
81.836
1.8
1.8
4.5
5.5
5.56.2
5.2
1.8
23.461
9.4
3.87
17.66
11.5
9.4
0
0.09
0.020.007
0
0
0
0
0
0.58
30.8
96.77
323.8
0.53
Source
6.5.2
Bound Water
The interaction between the adsorbed molecules and the surface may be classified into
the following types: (1) Chemical adsorption Or chemisorption. e.g., stearic acid
from benzene solution on metal powders; (2) Hydrogen bonding e.g., long-chain
alcohols from hydrocarbon solution on dry oxide surfaces; (3) Hydrophobic bonding
e.g., acids on polystryrene; and (4) van der Waals forces Long-range attraction.
253
In addition to these forces, capillary condensation also plays an important role in

porous structures.
When water molecules (absorbate) enter the pore structure of hydration products
(absorbent), they tend to be adsorbed onto the surface of the pores once they contact
the surface. This type of physisorption (reversible process) can develop an adsorbed
layer on the surface of the pores. Thickness of the adsorbed layer can grow up to a
threshold value at which the surface attraction is no longer capable of holding any free
water molecules onto the layer. In this dissertation, the bound water is referred to
the water molecules within this layer. Figure 6-18 illustrates the considered sorption
model.
H
solid phase of
concrete
optional
boundary
t
c.d.
region
bound water
3.25
chemical
water
free water
tmax
water molecule
Figure 6-18: Considered sorption model in the pore structure of hcp
Dielectric properties of the bound water subjected to surface attraction are different from the ones of the free water since the attraction constrains the movement of
water molecules. Under the application of external fields (radar signals), the rotation
of bound water molecules is hindered, resulting in the change of individual relaxation
times of bound water molecules. It is believed that such surface attraction, as a
254
binding energy between the molecule and the surface, is important for the dielectric
dispersion of molecules possessing permanent electric dipole moments such as water
and can be modeled by a potential function.
Since the variation of individual relaxation times is described by the distribution
function y( ) and recall Eqs.(6.82) and (6.83), the dielectric properties of the bound
water are described by
0bw () = +
Z
d
1
y( )
1 + (bw )2
(6.108)
bw
y( )
1 + (bw )2
(6.109)
00bw () =
Z
0
where bw is the relaxation time for each molecule layer of bound water, which is
related to the potential barrier determining the strength of attraction.
Relaxation Time and the Reaction Rate of Relaxation

It is obvious that the determination of complex permittivity depends on the knowledge
of the distribution function of relaxation time, y( ). Frohlich (1949) [90](page 93)
had shown that, in dilute solutions, the y( ) of dielectric materials displaying Debyes
dispersion can be expressed by
y( ) = (s )
kT 1
v0
(6.110)
where k is the Boltzmanns constant (or kB , 1.38065041023 J K 1 =8.617343

eVK 1 ), T is the Kelvin temperature ( K), v0 is the differential energy barrier of
the energy barrier H0 measured from a normal (permanent) position of the water
molecule. (H0 + v0 ) is the range of influence to the complex permittivity of the
molecule under investigation.
The relationship between the relaxation time and the sorption process can be
explained by the rate theory in chemistry. Kauzmann (1942) [130] suggested that
the reaction rate of relaxation (or simply the relaxation rate) k (rad/s) of a chemical
255
process can be determined by

kT
dF
1
exp
k= =
h
RT
(6.111)
where h is Planks constant (6.6260691034 Js = 4.1356671015 eVs), dF = dH

SdT is the molar (Helmholtz) free energy, dH is the heat of activation for dipole
relaxation, S is the entropy of activation for dipole relaxation, and R is the gas
constant (8.3144
m3 Pa
Kmole
J
).
Kmole
In porous media like concrete, the Gibbs free
energy is considered instead.

dG = V dp + dA +
j dnj SdT
(6.112)
where V is the volume, p is the pressure, is the surface tension (capillary condensation), A is the interfacial area, j is the chemical potential of component j (kJ/mole
or kcal/mole), nj is the number of moles of component j (One gram-mole of water has
a mass of approximately 18.016 grams, and one ml of water has 3.011022 molecules.)
The chemical potential j is given by

j =
G
nj

(6.113)
T ;p;A
Measurements Conducted by Badmann et al.(1981)

Badmann et al. (1981) [212] conducted a series of experiments on the determination
of chemical potential, , for porous hardened cement paste (hcp) in which capillary
condensation (cc) is significant. The BET C constant was obtained from their measurements on hcp (C3 S=59.33%, C2 S=12.86%, C3 A=16.33%, C4 AF=8.92%) from 1%
to 97% relative humidity (r.h.). The chemical potential cc (t) was given by

t
t
1 exp
cc (t) = 0 exp
h0
h1
(6.114)
),
where t is the thickness of the adsorbed water layer or the bound water layer (A
0 = 5.9 3 C kJ/mole = 18.5 kJ/mole when C = 31 as the experimentally-determined

256
BET C constant, h0 = 1.91

A, 1 = 0.142 kJ/mole for h1 = 26
A. An expression of
vdW accounting for the van der Waals forces from silica surfaces was also proposed
by Badmann et al. (1981) [212].
vdW (t) =
AVm
6 (h + w )3
(6.115)
where A is the Hamaker constant (A < 1.4 1013 erg = 1.4 1023 kJ for the water
on silica surfaces), Vm = 18 cm3 /mole is the molar volume of water at 25 C (298 K),
h is the distance from the center of the first-layer water molecule to the center of the
water molecule at other layers (
A), w is the radius of water molecules (w is taken
as 1.625
A). Figure 6-19 shows the determination of h in the sorption model. The
bound
water
chemical water
Figure 6-19: Jellium distance in the sorption model
chemical potential hcp (shown in Figure 6-20) between the bound water and the pore
structure of hcp is modeled by
hcp (t) = cc (t) + vdW (t)
(6.116)
In Figure 6-20, the decay of hcp (t) is essentially an exponential function as suggested
by other researchers [105]. Additionally, the semi-empirical expression of statistical
thickness of the adsorbed water layer (film) on hcp was also reported by Badmann et
257
cc
10
Potential energy (kJ/mole)
vdW
10
-1
10
-2
10
10
15
20
Thickness (angstrom)
Figure 6-20: Used chemical potential accounting for the formation of the bound water
on the surface of hcp
al. (1981) [212].
#

P
5.9 3 C
h0 ln ln
t = h0 ln
RT
P0
"
(6.117)
where P is the pressure at a given r.h., and P0 is the pressure at r.h.=0%. This
expression was obtained from the measurements (shown in Table 6.12). The performance of Eq.(6.117) is illustrated in Figure 6-21. In Figure 6-21, samples of C2 S
and C3 S were prepared as non-porous specimen and hcp as porous specimen. The
maximum average thickness of the adsorbed water layer on hcp was reported to be
around 18
A. The maximum thickness of adsorbed water layer on SiO2 measured by
Harkins and Jura (1944) [105] was 12.8
A when
P
P0
= 80% at which Eq.(6.117) pre-
dicts 12.22
A. T.C. Powers (1965) [207] suggested the maximum thickness on cement
paste at r.h.=100% to be 13
A. Considering the similarity the chemical composition
between SiO2 and hcp, it is believed that Eq.(6.117) can be used to generally describe
the adsorption of water molecules on hardened hcp.
258
Table 6.12: Averaged statistical thickness, t (

A), of adsorbed water layer on hcp
)
t(A
0.92 1.33 1.78 2.28 2.64 2.94 3.24 3.55 3.83 4.06 4.32
P/P0 (%) 1
t(
A)
4.53
2.5
10
15
20
25
30
35
40
45
4.79
5.17
5.37
5.68
6.08
6.61
7.27
8.20
9.87
P/P0 (%) 50
55
60
65
70
75
80
85
90
95
20
Data on C2S and C3S by Badmann et al.(1981)
Statistical thickness, t (A)
18
Model for C2S and C3S (BET C = 31)

Model for hcp (BET C = 31)
16
14
12
10
8
6
4
2
0
0.2
0.4
0.6
0.8
P/P0
Figure 6-21: Measurements and model prediction of non-porous and porous specimens
of hydration products
Proposed Dielectric Model for Bound Water
The purpose of referring the measurements conducted by Badmann et al.(1981) [212]

in this modeling work is to develop an expression for the complex permittivity of
bound water in hcp based on reported experimental measurements made on hcp.
Considering an adiabatic process (dT = 0) for unit mole of water molecules (nj = 1),
in which no surface change and pressure change occur (dA = 0 and dp = 0), we have
G = dG. Recall the reaction rate equation, Eq.(6.111). The formation energy of
259
free water molecules can be calculated from the measured relaxation time, w .

w (T )kT
w (T ) = RT ln
h

(6.118)
where h is Plancks constant, k is Boltzmanns constant, T is the Kelvin temperature,

and R is the gas constant. Table 6.13 lists the reported measurements of the relaxation
time of free water molecules in the temperature range from 0 C to 60 C (or 273 K to
333 K) [106]. With this relation, the calculated bonding potential, w (T ), is provided
Table 6.13: Relaxation time, w (ps), of free water molecules
T( C)
10
20
30
40
50
60
w (ps)
17.7
12.6
9.2
7.1
5.7
4.8
3.9
in Figure 6-22. This bonding potential macroscopically accounts for the multiple
Bonding potential, , (kJ/mole)
10.5
10
9.5
9
270
280
290
300
310
320
330
340
Absolute temperature (K)
Figure 6-22: Calculated bonding potential of free water molecules

interactions between one free water molecule and its neighbor molecules when they
all subject to the application of an external field. Within the adsorbed water layer on
260
hcp, relation of water molecules is believed to be hindered by the attraction potential

from hcp (hcp ). Equivalently, the relaxation time is prolonged due to the presence of
hcp . Outside the adsorbed water layer, the relaxation time of bound water molecules
should reduce to the one of free water molecules. The general form of the bonding
potential is given by
(T, t) = w (T ) + hcp (T, t)
(6.119)
in which the thickness of adsorbed water layers, t, mainly depends on temperature,

T , and pressure ratio,
P
.
P0
Therefore, the relaxation time of bound water can be
calculated by

(T, t)
h
exp
bw (T, t) =
kT
RT
(6.120)
This result is illustrated in Figure 6-23 for T = 293 K or 20 C. Note that, in this
model, the relaxation time of water molecules outside the statistical thickness of the
bound water layer, t, converges to the one of free water molecules. Therefore, the
boundary condition of bw is satisfied in this model. Substitute Eq. (6.110) into
4
x 10
150
1.8
Relaxation time (ps)
Relaxation time (ps)
1.6
1.4
1.2
1
0.8
0.6
100
50
0.4
0.2
0
10
15
20
Thickness (A)
10
15
Thickness (A)
Figure 6-23: Calculated relaxation time distribution over the bound water later
261
Eqs.(6.108) and (6.109). We have

(s ) kT

ln

2
v
0
1 + (bw ) o

(s )bw kT

00bw () =
ln

0
1 + (bw )2 vo
0bw () = +
(6.121)
(6.122)
In the above equations, v0 represents the differential energy barrier of the energy barrier H0 within the adsorbed water layer. This differential energy barrier is considered
to be attributed to the variation in the hcp (t), which is experimentally determined.
Equivalently, v0 is numerically calculated from Figure 6-20. It is noteworthy to point
out that the value of should not exceed the relaxation time at the center of adsorbed
water molecules in the first layer. This constitutes the upper bound of bw . Also, bw
should not less than the relaxation time of free water molecules, w . Thus, the above
equations now become

(s ) kT
bw (T, t = w )
= +
ln
w (T )
1 + (bw )2 vo

(s )bw kT
bw (T, t = w )
00
bw () =
ln
w (T )
1 + (bw )2 vo
0bw ()
(6.123)
(6.124)
Eqs.(6.123) and (6.124) are the proposed dielectric model for determining the complex
permittivity of bound water adsorbed on hardened hcp.
An information needed for determining the contribution of bound water in the

dielectric properties of hcp is the volumetric fraction of bound water in hcp, which
can be experimentally determined by the BET method. This can be done either by
(1) measuring the weight change of specimens or by (2) multiplying the surface area
of hcp by the thickness of bound water (adsorbed water molecules). The volume
computed from both approaches should converge to each other if the measurements
are conducted at perfect precision. However, since the formation of hcp (or concrete)
structure depends on many factors (e.g., w/c ratio, curing temperature, degree of
hydration) and the variability in measured surface areas using different adsorbates
(Nitrogen at 77 K and liquid water at 293 K, typically), obtaining a general equation
262
capable of estimating the amount of adsorbed water molecules on the surface of hcp
is still challenging. While the derivation of such equation is beyond the scope of this
research, a simplified model is proposed, with several assumptions, for approximately
estimating the amount (volume) of bound water.
The measurements using Nitrogen adsorption by Feldman (1972) [83] on the specific surface area (Ssp ) of hcp specimens with different w/c ratios are adopted. It
is assumed that (1)ordinary Portland cement (Type I) used and no admixtures, (2)
stabilized hydration, and (3) regular curing process and temperature. The reported
Ssp values of hcp are listed in Table 6.14. The curve-fitted model for Ssp (m2 /g) is
Table 6.14: Specific surface area of hcp [83]
W/c ratio
0.4
0.5
0.6
0.8
1.0
Ssp
30
55
51
57
57
Ssp = 9.2986 + 18.1437 tan
0.4
0.05

(6.125)
where is the w/c ratio. The main reason for using Nitrogen (N2 ) as the adsorbate
is due to its sensitivity to the change of w/c ratio in the Ssp of hcp determined by
the BET method. Knowing that the Ssp found by N2 molecules is typically much less
than the value found by water molecules (H2 O), the Ssp (N2 ) is indeed the specific
surface area since it does not include the area between hcp gels/layers/pores. Not
only because of the well-known incapability of N2 of penetrating into hcp gel pores, but
also because of the low thermal activation state when using N2 at 77 K. Performance
of the model is shown in Figure 6-24. This model can be further improved with
additional measurements.
6.6
Dielectric Properties of GFRP
Glass fiber reinforced polymer (GFRP) consists of glass fabric and epoxy resin; glass
fabric is for providing tensile strength of the assembly and epoxy resin for molding purpose. Mechanical properties, such as Youngs modulus, Poisson ratio, tensile strength,
263
Specific surface area (m 2/g)
60
50
40
30
20
10
0
0.2
N2 adsorption by Feldman (1972)

Fitted model
0.3
0.4
0.5
0.6
w/c ratio
0.7
0.8
0.9
Figure 6-24: Performance of the curve-fitted model for the specific surface area of hcp
by nitrogen adsorption
shear strength, and flexural strength of GFRP, are typically concerned among its
other properties in civil engineering applications. Durability is also crucial for the
long-term performance of GFRP, such as thermal coefficient, thermal conductivity,
and glass transition temperature. Usually, these properties are laboratory-tested using ASTM (American Society for Testing and Materials) methods and provided by
manufacturers/suppliers.
However, dielectric properties of GFRP have yet become required information in

the property description of GFRP for its use in strengthening of concrete structures.
The manufacturer/supplier of GFRP and epoxy (Fyfe Co. LLC) in this research
did not perform standard measurements of GFRP over a range of frequencies. To
overcome this difficulty, reported dielectric measurements of the GFRP and epoxy
similar to the ones used in this research are adopted for modeling, which are taken
from the experimental work done by von Hippel (1954) [113] in the Laboratory for
Insulation Research at MIT.
264
6.6.1
Epoxy Resin
Chemical molecules are classified by the functional groups they contain. An epoxide
is a molecule containing the epoxide group (one oxygen and two carbon molecules) as
part of its structure. An epoxy resin system is formed when other types of molecules
are added to the epoxide to formulate a thermosetting system. The formation of epoxy
resin systems undergoes a curing reaction to achieve a rigid state, in which curing
agents are also included for required curing rate in practice. Sometimes diluents/fillers
are used for improving the viscosity of uncured epoxy resin systems. Dielectric properties of an epoxy resin system depend on their chemical composition or functional
groups as the resultant macromolecular system.
Complex permittivity of the araldite adhesive Type I Natural manufactured by
r S Epoxy
Chiba-Geigy Ltd.(Japan) whose composition is similar to the used Tyfo
was measured in the frequency range 100 Hz10 GHz in room temperature (25 C)
[113]. Four dielectric models have been tested for their applicability on the epoxy;
Debye model, Cole-Cole model, Davidson-Cole model, and Havriliak-Negami model.
It is found that
1. The Debye model cannot describe the dispersion behavior of this epoxy resin
system simply because the Cole-Cole diagram of the epoxy is not a semicircle.
2. The Davidson-Cole model can describe the dispersion behavior of this epoxy
resin system but with relatively high error comparing to the Cole-Cole model.
3. The Havriliak-Negami model can describe the dispersion behavior of this epoxy
resin system but the coefficient is almost unity, suggesting the use of the
Cole-Cole model instead.
The obtained Cole-Cole model for epoxy is
()
1.06
= 2.96 +
0
1 + (i99.47)0.93
(6.126)
where the unit of constant 99.47 is rad-ns (1 ns = 1 nanosecond = 109 seconds). Its
performance is shown in Figure 6-25.
265
0.14
0.12
0.1
r"
0.08
0.06
0.04
von Hippel (1954)
Model
0.02
0
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
3.9
'
r
Figure 6-25: Performance of the Cole-Cole model for epoxy
6.6.2
E-glass Fabric
r SEH-51A (Fyfe Co. LLC) was used in

A unidirectional glass fabric system Tyfo
the GFRP-retrofitted concrete specimens. Dielectric measurements of this GFRP
system (using E-glass; for electrical applications) are not available from the material
supplier. Similar dielectric measurements on E-glass fabric (Owens-Corning Corp.)
were reported in the literature [113], and they are used for modeling the dielectric
properties of E-glass fabric in the frequency range of 8 GHz to 18 GHz.
Recall the dispersion equations of the Cole-Cole model.
0 = +
s
sinh(1 1 )x

1
1
2
cosh(1 1 )x + cos
2

1
cosh
s
2
00 =
1
2
cosh(1 1 )x + sin
2
266
(6.127)
(6.128)
where x = ln ( ). The relationship between 0 and 00 is derived to be

1 0
s
cos

2
2
00
(6.129)

=

s
1
1
s +
0
sinh (1 1 ) x + cos
sin

2
2
2
2

To perform the curve-fitting as a parametric system identification, a model similar to

Eq.(6.129) is obtained.
00 =
0.94850
+ 1.0391
0
2 #
6.3
0.4406 0.4599 455
+ 0.5578 1.04140
6.3
"
(6.130)
With these parameters the fitted model has a standard error of 0.004242 and a correlation coefficient of 0.9855 with the measured data. The expressions of 0r and r
as a function of frequency are also determined.
0r = 4.2027 0.022 [log(f + 107.74) 1676.45]
(6.131)
r = 0.004732 [log(f 25, 903, 094) 13.72]
(6.132)
where the frequency f is in hertz. The performance of the model is shown in Figures
6-26 and 6-27. Figure 6-27 illustrates the performance of Eqs.(6.131) and (6.132).
It is seen that, although the performance of the model for the glass fabric in lower
frequency ranges is relatively poor, the model provides reasonable estimation in the
high frequency range (GHz) which is the frequency range of interest. Therefore, the
use of this model is considered feasible in the frequency range of 8 GHz to 18 GHz.
6.6.3
GFRP Layer/Sheet
The unidirectional, one-layered GFRP sheet consisting of glass fabric and epoxy is
illustrated in Figure 6-28. Such material is generally considered as a dielectric mixture
since both glass fabric and epoxy are dielectric. Knowing that the mixing ratio of
epoxy to glass fabric is 0.8:1 by mass and densities of epoxy and glass fabric are 0.0011
267
0.1
von Hippel (1954)
Model
0.08
"r
0.06
0.04
0.02
0
6.05
6.1
6.15
6.2
6.25
'
6.3
6.35
6.4
6.45
6.5
Figure 6-26: Performance of the Cole-Cole model for E-glass fabric
kg/cm3 and 0.00255 kg/cm3 [152], the volumetric ratio of epoxy to glass fabric in the
GFRP layer is
Ve : Vg =
me mg
0.8
1
:
=
:
= 1.82 : 1
e g
0.0011 0.0025
(6.133)
where V denotes volume, W denotes weight, and for density. The volumetric
fractions of epoxy and glass fabric are
ve : vg =
1.82
1
:
= 0.645 : 0.355
2.82 2.82
(6.134)
In order to determine the dielectric properties (namely, complex permittivity) of such

dielectric mixture, six two-phase (two-component) dielectric mixing models/laws are
considered [256], in which is the effective or the overall complex permittivity of
the mixture, h is the complex permittivity of host dielectric, and i is the complex
permittivity of inclusion dielectric. In the configuration as shown in Figure 6-28,
epoxy is the host dielectric and GFRP the inclusion dielectric.
268
6.5
0.08
von Hippel (1954)
Model
von Hippel (1954)

Model
0.07
6.4
0.06
0.05
"r
'r
6.3
6.2
0.04
0.03
6.1
0.02
6
0.01
5.9
0
10
10
10
0
0
10
15
10
10
10
log(f) (Hz)
10
10
15
10
log(f) (Hz)
Figure 6-27: Performance of the Cole-Cole model for E-glass fabric Real part and
imaginary part
Epoxy resin
E-glass fabric
Figure 6-28: Unidirectional GFRP layer

1. Maxwell (1891) [174] Spherical inclusion (Maxwell-Garnett formula)
= h +
3vi h
i + 2h
vi
i h
(6.135)
2. Rayleigh (1892) [242] Spherical inclusion (Average t-matrix approximation)

=

i h
1 + 2vi
h
i + 2h
i h
1 vi
i + 2h
(6.136)
3. Wiener (1912) [268] Arbitrary shape inclusion; u = 2h for spherical, u = 2i
269
for disk-like inclusion, u =
i 3h
for needle-like inclusion (used).
2
i h
u
i + 2h
i h
1 vi
i + u
h + v i
=
(6.137)
4. Lichtenecker (1926) [150] Arbitrary shape inclusion

= evi ln i + vh ln h
(6.138)
5. B
ottcher (1945) [21] Spherical inclusion
=
C 2 + 8i h C
4
(6.139)
where C = (1 3vi )i + (3vi 2)h .

6. Looyenga (1965) [155] Spherical inclusion
3
= (vi 3 i + vh 3 h )
(6.140)
In order to evaluate their performance, reported GFRP-epoxy dielectric measurements are needed. Unfortunately, extremely limited amount of dielectric measurements made on E-glass GFRP-epoxy was reported. Only one publication is found
in the literature to be comparable. Seo et al. (2004) [119] measured the complex
permittivity of a E-glass GFRP and similar epoxy sheet (UGN150, SK Chemical,
South Korea) in the frequency range of 8.2 GHz to 12.4 GHz. The measured 0r was
5.004 at 10 GHz and 4.986 at 10.4 GHz. The case of fiber orientation at = 90
in their work corresponds to the configuration in this research. Predicted 0r values
of the GFRP-epoxy system (vi = 0.33) are provided in Figure 6-29. Absolute errors
are calculated for the measured 0r at 10 GHz and 10.4 GHz, as shown in Table 6.16.
While all six models provide acceptable estimation on 0r in a reasonable range of
error, their performance on r is in general much poorer than on 0r . Reported r
270
Table 6.15: Comparison of two GFRP-epoxy systems

Parameter
Seo et al. (2004)
This research
Density epoxy
Density GF RP
Thickness of GFRP-epoxy layer
Vepoxy : VGF RP
vi
10.01 lb/gal
21.20 lb/gal
0.33 cm
0.67 : 0.33
0.3308
9.17 lb/gal
21.28 lb/gal
(0.250.4) cm
0.645 : 0.355
0.355
4.94
0.05
0.045
4.92
0.04
4.9
Maxwell (1981)
Rayleigh (1982)
Wiener (1912)
Lichtenecher (1926)
Bottcher (1945)
Looyenga (1965)
0.035
4.88
r"
r'
0.03
0.025
4.86
0.02
4.84
0.015
4.82
4.8
7.5
0.01
8
8.5
9.5
10 10.5
Frequency (GHz)
11
11.5
12
12.5
0.005
10
Frequency (GHz)
11
12
Figure 6-29: Performance of six mixing models for GFRP-epoxy

values were 0.1112 at 10GHz and 0.1111 at 10.4 GHz. On the other hand, predicted
r values fall in the range of 0.0070.042, all in underestimation. This is attributed
to the difference in the used material (mostly epoxy).
However, it is found that, among these models, only Wieners mixing model displays same descending behavior of r as reported by Seo et al. (2004) [119] when frequency increases. Wieners model also provides the least error on r , compared with
other mixing models. In addition, the needle-like inclusion assumption in Wieners
model is physically analogous to the GFRP inclusion in a epoxy resin system. Therefore, Wieners mixing model is applied for modeling the GFRP-epoxy system in this
research.
i h i 3h
i + 2h
2
i h
1 vi
i 3h
i +
2
h + vi
=
271
(6.141)
Table 6.16: Performance of six mixing models for a GFRP-epoxy system

Model
Maxwell (1811)
Rayleigh (1892)
Wiener (1912)
Lichtenecher (1926)
B
ottcher (1945)
Looyenga (1965)
6.7
Predicted
(10GHz)
4.8602
4.8602
4.8574
4.8221
4.9154
4.9064
0r
Error (%) Predicted 0r

(10.4GHz)
2.80
4.4950
2.80
4.4950
2.86
4.3592
3.57
4.4520
1.70
4.5444
1.88
4.5406
Error (%)
9.85
9.85
12.57
10.71
8.86
8.93
Dielectric Properties of Concrete
Dielectric modeling of multi-phased, heterogeneous composites like concrete is a challenging task. Concrete, as a cementitious composite, is a mixture of
Portland cement hydration products Including calcium silicate hydrate (C-S-H or C3 S2 H8 or 3CaO 2SiO2 8H2 O), calcium hydroxide (CH or
3 H32 or 3CaOAl2 O3 3CaOSO3 32H2 O), and monoCaO H2 O), ettringite (C6 AS
H12 or 3CaOAl2 O3 CaOSO3 12H2 O). ASTM specsulfoaluminate (C3 A CS
ifications such as C150 define the properties of Portland cement.
Water ASTM C94 requires tests of mixing water when its quality is questionable.
Aggregates Including fine aggregate (sand) and coarse aggregate gravels).
ASTM C33 and C294 describe the properties of acceptable aggregates.
Chemical admixtures are also used in practice as a processing addition to aid in
manufacturing and handling of concrete or as a functional addition to modify the
properties of concrete. In this research work, chemical admixtures are not considered
in the dielectric modeling, nor used in the manufacturing of concrete specimens. In
addition to ASTM specifications, other building codes such as ACI 318M also provide
requirements on the quality of the components used in concrete.
Three approaches may be applied to model the dielectric properties of concrete.
272
1. Phenomenological approach Describe the complex permittivity of concrete

as a homogeneous composite by considering one or two major parameters (such
as frequency) only. This can lead to only the global/macroscopic description
on the dielectric dielectric dispersion of concrete.
2. Physical approach Describe the complex permittivities of the components
of concrete individually using the causality relation (or the Kramers-Kronig
relation), and assemble them by the dielectric mixture theory. This approach
can explain the dielectric dispersion of concrete from a microscopic perspective,
although the composition of concrete must be fully understood.
3. Statistical-physical approach Describe the complex permittivities of the
components of concrete individually using the causality relation, and assemble them by statistical distribution functions. This approach can explain he
dielectric dispersion of concrete from a microscopic perspective, while significant amount of experimental data must be provided in order to obtain reliable
statistical functions.
Phenomenological approach has been widely used for engineering applications
since it avoids the direct difficulties on dealing with the microscopic structure of concrete. In many cases, the complex permittivity of concrete is treated as a function
of frequency and the water/cement ratio. Physical approach requires a full knowledge on the composition of concrete at the state of investigation, which is not easy
to acquire considering the spatial variability of concretes composition. Microanalysis methods such as SEM (scanning electron microscopy), NMR (nuclear magnetic
resonance), and X-ray diffraction techniques are typically used for determining the
micro-structure of concrete. At the present time, comprehensive experimental data
on the micro-structure of concrete specimens made of various water/cement ratios
and other manufacturing factors are not reported in literature. Statistical-physical
approach even demands more measurements than the physical approach to result
in reliable distribution functions with sufficient confidence level. It is obvious that
experimental measurements on the complex permittivity of concrete are key to the
273
success of dielectric modeling on concrete.

In the context of this doctoral research, the purpose of this modeling work is to
estimate/predict the complex permittivity of concrete in order to model the concrete
in numerical simulation and to improve the resolutions in the reconstructed images
for condition assessment. In this development, two major challenges arises.
1. Incomplete information regarding the spatial composition of concrete The microscopic structure of concrete is mainly determined by the
water/cement ratio, degree of hydration, and the conditions (temperature and
moisture) under which the concrete is cured. It is difficult to completely describe the components (hydration products, pores, moisture, fine and coarse
aggregates) of concrete and their spatial distributions in concrete even if all the
information during the manufacturing of concrete becomes available. In addition, shape of the pores in concrete is also important since it determines the
distributions of air and water molecules in concrete.
2. Lack of the full dielectric dispersion spectra for the components in
concrete Even if the composition of concrete can be fully aware, dielectric
dispersion spectra of the components in concrete, such as hydration products
(e.g.,C-S-H) and aggregates, are still needed in order to calculate the bulk complex permittivity of concrete either physically or statistically. This information
is also not complete in literature.
These constraints limit the development of a full-spectrum physical model of concrete. As the result of it, dielectric models of concrete and cementitious composites
(cement paste and cement mortar) can only be developed based on specimens with
certain parameters (e.g., frequency, water/cement ratio), on which some dielectric
measurements have been reported.
In view of the limited amount of dielectric measurements of concrete in the frequency range of or nearby 8 GHz18 GHz, dielectric modeling is conducted on anhydrous (oven-dried) hydrated cement paste (hcp) in this research.
274
6.7.1
Determination of Volumetric Fractions
In the process of mixing two or more ingredients/components into one mixture, it

is the volumetric fraction of each component pivotal in determining the properties
of the mixture. One major difference distinguishing concrete as a hydrate is the
formation of a porous structure with high surface area. This porous structure is the
remaining space left by the evaporated mixing water during the hydration process.
Such evaporation can be achieved either by internal thermal gradient originated from
the hydration heat (room-temperature drying) or by artificial heating (oven drying,
defined by ASTM C29).
To what extent the weight loss of hydrated cementitious composites achieve regards the classification of mixing water in the hydration process. Typically, the mixing
water can be categorized into the following two types. This classification is based on
the research works by [209], Copeland, Kantro, and Verbeck (1960) [148], Ishai (1965)
[120], Feldman and Sereda (1970) [85], Feldman (1972) [83], and [238]. Definitions in
recent texts [179, 223] are essentially based on the findings found in the early work.
1. Non-evaporable water The water that participates the hydration reactions
and absorbed by the hydration products is considered non-evaporable.
2. Evaporable water The water that participates in the casting of cementitious
composites for workability purpose but not being absorbed by the hydration
products after the hydration reactions stabilized is considered evaporable. This
includes:
Free water The water residing in the pore structure not subjected to
the surface traction of the hydration products, whose volume is denoted
by Vf w . This type of water can be removed when the composite is heated
up to 105 C. Free water is also termed capillary water.
Bound water The water residing in the pore structure and subjected
to the surface traction of the hydration products, whose volume is denoted
by Vbw . This type of water can be removed when the composite is heated
275
up to 105 C. Bound water is also termed gel water.

Chemical water The water underneath the surface of hydration products, attached by a strong chemical potential, whose volume is denoted by
Vcw . This type of water can be removed when the composite is heated up
to 1010 C. Chemical water is also termed combined water.
Since the chemical water is embedded inside the hydration products, it is considered
that the pore structure in cementitious composites is formed by the liberation of the
free and bound water. The oven-drying scheme used in the specimen preparation
of most experiments in literature can achieve 105 C, suggesting the removal of both
the free and bound water in their measurements. The removal of the chemical water
requires a heating temperature up to 1000 C which is not a situation commonly
encountered in practice. This scheme is illustrated in Figure 6-18.
Considering two extreme cases (fully-saturated and oven-dried specimens), the
volumetric fractions of cement paste, cement mortar, and concrete are defined in
Table 6.17 in which vs denotes the volumetric fraction of the solid phase, vg is the
volumetric fraction of the gaseous phase, and vl is the volumetric fraction of the liquid
ms
mhcp
phase, all in the multi-phased composite. In Table 6.17,
=
for cement paste
s
hcp
ms
mhcp
mf a
(purely hydration cement products or hcp),
= vhcp
+ vf a
for cement
s
hcp
f a
ms
mhcp
mf a
mca
mortar, and
= vhcp
+ vf a
+ vca
for concrete, where vhcp , vf a , and
s
hcp
f a
ca
vca denote the volumetric fractions of cement paste (hcp), fine aggregate (fa), and
coarse aggregate (ca), respectively.
As indicated in Table 6.17, the mass of free and bound water, mf w + mbw , is the
differential mass between the fully-saturated mass and the oven-dried mass of the
specimen, while the total amount of water determined by the water/cement ratio is
mf w + mbw + mcw . The mass measurement is the ratio between the weight, W , and
gravitational constant, g (9.81
m
).
s2
The density of the solid phase of hydrated cement paste (hcp) depends on the
structure of hcp and ,thus, is a function of the w/c ratio and other factors. Feldman
(1972) [83] measured the variation of density of cement paste, hcp , with respect to
276
Table 6.17: Volumetric fractions of cementitious composites

Vol. frac.
Fully-Saturated
Oven-dried
ms
s
ms mf w + mbw + mcw
+
s
w
ms
s
ms mcw
+
s
w
Gaseous, vg
mf w + mbw
w
ms mcw
+
s
w
Liquid, vl
mf w + mbw
w
ms mf w + mbw + mcw
+
s
w
Solid, vs
the w/c ratio , as shown in Table 6.18. Other researchers suggested an average value
of 2.5 g/cm3 [238](page 70). The density of water at room temperature is 1 g/cm3 .
For normal weight concrete the average density of fine and coarse aggregate is about
1.5 g/cm3 . However, maximum density can be achieved when the volumetric ratio
between fine and coarse aggregates is taken to be
vf a
vca
2
3
[223], which is around 1.75
g/cm3 .
Table 6.18: Density of cement paste specimens [83]
6.7.2
w/c ratio
0.4
0.5
0.6
0.8
1.0
Density (g/cm3 )
2.19
2.28
2.30
2.29
Dielectric Modeling of Oven-Dried Hydrated Cement

Paste
Cement paste is made by mixing cement and water at a given water/cement ratio.
It consists only the hydration products. Reported dielectric measurements used in
277
the dielectric modeling on cement paste specimens in the frequency range between 8
GHz and 18 GHz are by Hasted and Shah (1964) [107], De Loor (1961) [154], and
Wittmann and Schlude (1975) [271]. Other measurements made on cement pastes
outside the frequency range of interest are not referred here.
Hydrated cement paste (hcp) is the fundamental structure of other cementitious
composites such as cement mortar and concrete. When hcp is oven-dried, the free and
bound water presenting in the capillary and gel pores of hcp is considered all liberated,
resulting in a two-phase (solid and gaseous) composite with a maximum void ratio or
(vg )max and a minimum moisture content or (vl )min = 0. Such state also determines
the volumetric fraction of the solid phase of hcp, vs . Later, the moisture ingress from
external environment will decrease the void ratio and increase the moisture content,
but the volume of solid phase is considered unchanged during the drying-and-wetting
process. In other words, only vg and vl vary in hcp. vs is considered a constant
once the age of the material reaches 28 days. In brief, the water/cement ratio is the
fundamental parameter determining vs and the structure of hcp.
Our interest here is to investigate how the dielectric properties of hcp change with
respect to not only the measurement frequency but also the water/cement ratio of
hcp. To simplify the influence due to many factors (e.g.,moisture content, different
polarizations), oven-dried hcp is first considered. For oven-dried hcp, generally the
higher the w/c ratio, the lower the dielectric constant. This is because that the
increase of the w/c ratio produces more air voids (0r = 1) in hcp (0r = 5 15),
resulting in lower bulk 0r for the porous medium. This observation is confirmed by
dielectric measurements on oven-dried (hydrous) hcp of different w/c ratios reported
by several researchers in the literature, as shown in Figure 6-30. As can be seen in
Figure 6-30, the product of the w/c ratio and the dielectric constant 0r at high
frequency range (GHz range) is approximately constant when there is no moisture
present in the specimen. It is postulated that the dielectric constant at extremely
high frequency range can be represented by
=
( + C1 ) C2
278
(6.142)
where parameters , C1 , and C2 are experimentally determined. Following this assumption, the term (s inf ) is represented to be
s =
( + C3 ) C4
(6.143)
A modified Debyes model is thus proposed for characterizing the dielectric properties
5
de Loor (1961), 3GHz
Hasted & Shah (1964), 3GHz
Wittman & Schlude (1975), 3GHz
( x ' )
r
0.3
0.32
0.34
0.36
w/c ratio,
0.38
0.4
Figure 6-30: Relationship between the w/c ratio and product of the w/c ratio and
dielectric constant of oven-dried cement paste specimens
of oven-dried hcp.
2
( + C3 ) C4
(, ) =
+
( + C1 ) C2
1 + (i )
(6.144)
In this model, the relation time is also a function of the w/c ratio , accounting for
the change of material structure. The curve-fitting results using experimental data
on oven-dried hcp in the frequency range of 3 GHz24 GHz and the w/c ratio range
0.250.4 are shown in Figure 6-31, with determined values of the parameters listed
in Table 6.19.
279
Table 6.19: Parameters in the oven-dried hcp model

Coef.
C1
C2
C3
C4
Value
0.76
859 ns
1.625
0.55
0.4
0.3
0.5
6.7.3
Challenges in the dielectric modeling of concrete
As generally considered a poorly crystalized mineral, concrete has an amorphous

structure that poses a challenge in describing its spatial composition with accuracy.
Not only such amorphous structure is a traditional problem when the mechanical
properties and fracture of concrete are concerned, but also it creates a problem when
the dielectric properties of concrete are to be modeled. As previously mentioned
in this chapter, the full spectrum dielectric models for the components in concrete
including hydrated cement paste (hcp), fine aggregate (sand), and coarse aggregate
(gravels) also need to be available for the use and development of a dielectric mixing
law on concrete.
Factors that are influential to the measured dielectric properties of concrete are
summarized as follows.
Manufacturing factors Factors determining the manufacturing of concrete
need to be incorporated into the modeling process since they affect the spatial composition of concrete, including the w/c ratio, mixing proportion,
cement properties (cement composition, fineness of the cement), the use of
admixtures, and curing conditions (curing temperature, curing rate, curing
method). These factors are the reason why predicting the formation of concrete
is still a challenging task at present time. Some researchers applied numerical
tools to model the chemical reactions involved in the hydration process with
preliminary results [23], while a general model with sound predictability and
ease for application is still lacking.
Measurement factors Even if the spatial composition of concrete is completely known, the measured dielectric properties of concrete can still vary at
280
different measurement frequencies and temperatures using different measurement techniques. The frequency-dependence and temperature-dependence
are the characteristics of dielectric dispersion as the work by P. Debye (1929) [65]
suggested around 80 years ago, accounting for the combined effect of various
polarizations at different scales (see previous sections on dielectric dispersion
in this chapter). Variation in the measured dielectric properties due to the
selection of measurement technique originates from the difference in the interpretation of dielectric properties. In some techniques additional polarization
may be involved (e.g., interface polarization in contact techniques). This is why
the dielectric measurements at low frequencies can not be directly used for the
modeling in high frequencies and same with temperature.
In view of the above arguments, a comprehensive experimental program is apparently
needed to obtain a complete description of the dielectric dispersion of concrete as a
basis for developing a full-spectrum physical model accounting for the heterogeneity
in concrete. To the best knowledge of the author, such complete description is not
available at present time. However, with increasing appreciation on the usefulness
and importance of dielectric properties as a mutual interest in the communities of civil
engineering, material sciences and engineering, chemistry, and physics, it is expected
that more experimental measurements will be made and reported in the near future.
6.8
Summary
Research findings are summarized as follows.

From the theoretical perspective, in the frequency range of 8 GHz to 18 GHz, the
real part of complex permittivity (dielectric constant) is the combined result of
the dipolar, atomic, and electronic polarizations, while the imaginary part (loss
factor) is only affected by the dipolar polarization. This suggests the feasibility
of using single-pole dispersion models such as Debyes model for the modeling
work in this specific frequency range.
281
An integrated methodology for determining the unique combination of complex permittivity of materials is developed. This methodology is based on
transmission-only, coherent, wide-bandwidth free-space measurements as a timedomain method, integrating the time difference of arrival (TDOA) measurement
of plate specimens and a root-searching optimization scheme using parametric
system identification (PSI) and the error sum of squares (SSE) criterion for the
unique determination of complex permittivity [36].
Fundamental difference in the dielectric modeling based on internal and external fields is described. The geometrical analysis technique used for constructing
dielectric models for the materials considered in this research is explained. Several representations (analytical equation, the Cole-Cole diagram, and equivalent
circuit) of dielectric dispersion models are provided as the basis for constructing dielectric models from measurements. Dielectric modeling is conducted for
the use in the frequency range of 8 GHz to 18 GHz and room temperature
(20 C25 C).
A Cole-Cole type model for the dielectric properties of free water is proposed.
As for bound water, the relaxation time function of bound water molecules is
derived based on the statistical thickness and bonding potential of adsorbed
water layer in hcp by Badmann et al. (1981) [212] and the reaction rate theory.
Complex permittivity of bound water molecules at different levels within the
adsorbed water layer is calculated by this model.
Dielectric models of epoxy resin and E-glass fabric are developed using the
dielectric measurements reported in the literature [113]. Model parameters were
determined by geometric analysis using the Cole-Cole diagram. GFRP layer, as
a dielectric mixture of epoxy resin and E-glass fabric, is modeled by Wieners
model, out of six two-phase dielectric mixing laws considered in this chapter.
Dielectric properties of oven-dried hydrated cement paste (hcp) are determined
by a modified Debyes model incorporating the water/cement ratio as a major
282
material parameter. The challenges of dielectric modeling on concrete originate

from the combined effect of manufacturing and measurement factors in concrete.
A physically-sound dielectric model for concrete must be validated by sufficient
amount of experimental data to account for all possible combinations of those
factors.
In this chapter, dielectric modeling of materials including water, GFRP, and ovendried hydrated cement paste is performed. The knowledge of dielectric properties of
materials is advantageous for better modeling materials in numerical simulation (e.g.,
Chapter 3). In the investigation of dielectric modeling of multi-phase cementitious
composites such as concrete, complexities and challenges are addressed, as well as
demonstrated by the variation of radar measurements on a plain concrete cylinder
shown in Chapter 4 Laboratory Radar Measurements. It is also indicated
in Chapter 5 Image Reconstruction that dielectric properties are important for
improving the resolution of reconstructed images when the defects deep from the
surface are to be located.
283
0.8
"r
'r
7
6
5
3GHz / H&S (1964)

9GHz
24GHz
hcp model: w/c=0.28
0.6
0.4
10
15
20
25
0.2
30
10
"r
'r
0.5
10
15
20
25
30
10
"r
25
30
15
20
25
30
3 GHz / H&S (1964)

9 GHz
hcp model: w/c=0.325
7
'r
20
3GHz / de Loor (1961)

3.75GHz
7.45GHz
9.37GHz
hcp model: w/c=0.31
15
0.5
6
5
10
15
20
25
30
10
"r
'r
20
25
30
3 GHz / H&S (1964)

9 GHz
24 GHz
hcp model: w/c=0.34
7
6
15
0.5
5
4
10
15
20
25
30
0.8
"r
'r
10
15
20
25
30
3 GHz / H&S (1964)

10 GHz / W&S (1975)
hcp model: w/c=0.4
0.6
0.4
0.2
10
15
20
Frequency (GHz)
25
30
10
15
20
Frequency (GHz)
25
30
Figure 6-31: Curve-fitting results of the oven-dried hcp in the frequency range of 3
GHz to 24 GHz
284
Chapter 7
Condition Assessment of
GFRP-concrete Systems FAR
NDT
A good engineer should posses a scientists brain, and a good scientist
should own an engineers hands.
In this chapter, a non-contact, far-field radar NDT method capable of detecting

defects and damages that may attribute to either GFRP delamination, or GFRP
debonding from concrete substrate, or sizeable air pockets trapped between GFRP
sheets and concrete substrate during manufacturing in near-surface region of concrete
structures wrapped and/or bonded with GFRP sheets/plates (Chapter 4 Laboratory Radar Measurements), or the combination of above, using monostatic inverse
synthetic aperture radar (ISAR) measurements and tomographic image reconstruction algorithms (Chapter 5 Image Reconstruction) is described. High-frequency
continuous waves (time harmonic) in the frequency range of 8 GHz to 18 GHz are
used as incident signals. One or more electronic signals, in terms of amplitude and
phase, reflected from the GFRP-retrofitted concrete structure are collected by radar
antenna in the far-field region, which represent the electromagnetic response of the
285
structure.
The proposed NDT capability consists of three components; a physical inspection component allowing distant assessment of civil infrastructures, a numerical
processing component transforming reflection response to spatial profile of the structure, and a pattern recognition component quantitatively evaluating the presence
of damage.
In the following sections, technical details of three components in the FAR NDT
are addressed. Validation of the methodology is conducted using the far-field ISAR
measurements of laboratory GFRP-confined concrete specimens.
7.1
Components of FAR NDT
The developed FAR NDT technique consists of a far-field monostatic ISAR inspection
with two measurement schemes, and a data processing algorithm. Instrumentation of
the technique mainly consists of an airborne horn antenna, a signal generator, a signal
modulator, and an analyzer. In principle, modulated radar signals or EM waves are
designed and generated by the signal generator, modulated by the modulator, and
transmitted by the radar antenna (a horn antenna was used in the experimental
validation). Assessment procedure and the information obtained in each component
are described in Figure 7-1.
This technique is essentially characterized by the following features:
1. Normal and oblique incidence inspection schemes - The signal transmitter and receiver can operate in two inspection schemes for measuring the reflection response of the target structure; normal and oblique incidence schemes, as
shown in Figure 7-2. Distant (near-field and far-field) measurements conducted
in these schemes construct the ISAR frequency-angle data, also as shown in
Figure 7-2.
2. Data processing algorithm - The measured ISAR frequency-angle data are
processed by tomographic reconstruction methods. The spatial profile of the
286
Component
Procedure
Information
Physical
Inspection
Far-field ISAR
Measurements
Frequency-angle
Measurements
Numerical
Processing
Image
Reconstruction
Range Cross-range
Images (spatial profile)
Pattern
Recognition
Damage
Assessment
Feature-Extracted
Images with
Quantitative Indices
Figure 7-1: Overview of the FAR NDT

structure is characterized by the range-cross-range imagery, and is used for
condition assessment.
3. Progressive image focusing - Image resolution is progressively improved by
assembling all subaperture images in this technique. The spatial profile of the
structure can be first reconstructed by using the data in one subaperture for
preliminary evaluation. The resolution of imagery can be enhanced finishing all
planned frequency bandwidth and the range of azimuth angle. This computational focusing scheme enables the technique to be used for different purposes,
such as preliminary inspection and detailed inspection.
The frequency-angle data is processed by an image reconstruction algorithm which
is based on tomographic reconstruction methods to generate range-cross-range imagery for visualization. Spatial profile of the inspected structure is characterized by
the range-cross-range imagery in which geometric features of the structure and unseen
near-surface defects are revealed by scattering signals. Image resolution is in general
proportional to (a) the frequency bandwidth and (b) the angular range of incident
angle.
Presence of anomalies (defects, damages, and rebars) are detected and represented
by scattering signals in the reconstructed images (Figure 7-3). Locations of anomalies
287
GFRPretrofitted
concrete
cylinder
z
y
x
fname: CYLAD1FV, HH Pol., max = -3.59 dBsm
12
fname: CYLAD1BH, HH Pol., max = -6.41 dBsm

12
11.5
11.5
-2
11
10.5
-4
10
-6
9.5
y-z plane
Frequency (GHz)
x-y plane
Frequency (GHz)
11
-5
10.5
10
-10
9.5
9
-15
-8
8.5
8.5
8
-30
-20
-10
10
20
30
-10
8
-30
-20
-10
10
20
(deg)
(deg)
(a) Normal incidence
(b) Oblique incidence
Figure 7-2: Normal and oblique incidence inspection schemes of the FAR NDT technique
can also be determined in the images. In Figure 7-3 (a), the presence of an artificial
defect is illustrated by a scattering signal centered at the location where the defect is
placed. In Figure 7-3 (b), rebars are discovered by the appearance of two scattering
signals.
288
30
-20
(a) Defect detection Oblique incidence inspection
(b) Rebar detection Normal incidence inspection
Figure 7-3: Defect and rebar detection
7.2
Physical Inspection Far-field ISAR Measurements
For the far-field monostatic ISAR inspection as described in Chapter 4 Laboratory

Radar Measurements, an airborne horn antenna is configured in normal incidence
(specular dominant, Figure 7-4) or oblique incidence (specular recessive, Figure 7-5)
measurement scheme. Coherent reflection responses are collected at different frequencies and azimuth angles in ISAR mode to form the frequency-angle data. The
frequency range of 8 GHz to 18 GHz is chosen for its relevancy and abilities in assessing the unseen damages and defects hidden behind the GFRP sheets in the interface
region between GFRP sheets and concrete substrate. The standoff or inspection distance, depending on the aperture size of the antenna, can be practically as far as 10
meters and more in the chosen measurement frequency range, enabling a real dis289
tant inspection capability. Figure 7-6 shows the far-field distance calculated in the
frequency range of 8 GHz to 18 GHz using two sizes of antenna aperture.
Antenna
The
GFRP-concrete
structure
int/2
=0
int/2
r
GFRP-concrete
structure
( = int/2 int/2)
(a) Top view
(b) Front view
Figure 7-4: Specular dominant circumstance
Antenna
GFRP-concrete
structure
int/2
=0
int/2
r
GFRP-concrete
structure
( = int/2 int/2)
(a) Front view
(b) Top view
Figure 7-5: Specular recessive circumstance

inspection procedure of the technique is described in Figure 7-7.
290
Far-field distance, dff (m)
20
15
D2 = 0.4 m
10
D1 = 0.2 m
10
12
14
Frequency (GHz)
16
18
Figure 7-6: Computed far-field distances at various frequencies and two antenna
apertures
291
Installation of
the inspection
apparatus
Defect detection
or
rebar detection?
Rebar detection
Normal incidence
ISAR measurements
Defect detection
Oblique incidence
ISAR measurements
Image reconstruction
YES
Improve the
image resolution?
YES
NO
Condition
assessment
Figure 7-7: Inspection procedure of the FAR NDT technique
292
The inspection is conducted by deploying a radar antenna, sending radar signals by

the antenna, and collecting the reflected radar signals (responses, measurements) by
the same antenna. The radar antenna can be placed beyond the far-field distance from
the target structure. Steady-state responses are measured at particular frequency and
azimuth angle. Incident frequency is shifted from the starting frequency (8GHz) to
the ending frequency (18GHz) at a frequency increment of 0.2GHz. Azimuth angle
is shifted from the starting angle (sta or sta ) to the ending angle (end or end )
after the required frequency bandwidth is explored at each azimuth angle. Since the
radar can operate in two inspection schemes, the possible trajectory of the radar
can be described as a sub-domain of a sphere as shown in Figure 7-2. Two sets
of far-field ISAR measurements collected from a GFRP-retrofitted concrete cylinder
using linearly HH-polarized signals are also provided in Figure 7-2 as examples of the
frequency-angle data. The developed technique allows convenient inspections from
distance for highway and cross-river bridge columns. Figure 7-8 illustrates several
scenarios for bridge inspection.
Radar
unit
Radar
unit
(a) Beam inspection
(b) Column inspection
Figure 7-8: Bridge inspection Beam and column
293
7.3
Numerical Processing Image Reconstruction
In-depth profiles or range-cross-range images of the target structure are generated by

the image reconstruction method as described in Chapter 5 Image Reconstruction, using the far-field ISAR measurements. The used tomographic reconstruction
method is implemented by a time-domain backprojection algorithm for the efficiency
and flexibility offered by the algorithm. Each pixel in the backprojection images
is reconstructed by coherently integrating the far-field ISAR measurements over the
ranges of incident frequency and azimuth angle. Processing efficiency and inspection
flexibility are provided by using the progressive image focusing scheme. This scheme
enables the improvement of image resolution by increasing the amount of measurements, rather than sacrificing inspection convenience by shortening the inspection
range. With this flexibility in signal processing, different purposes of inspection can
be chosen with respect to desired resolution.
In this section, the physical meaning of scattering signals in the reconstructed
backprojection images is first explained. The progressive image focusing scheme of
the FAR NDT technique is described with two types of integration for improving
damage detectability. Damage detectability using backprojection images is defined
as the ability defect signals can be separated/distinguished from background signals
in the images.
7.3.1
Physical Meaning of the Scattering Signals in the Images
In the monostatic mode of radar operation the scattering signals in backprojection

images are proportional to the magnitude of far-field ISAR measurements. The backprojection processing of ISAR measurements distinguishes the contributions of various
scatterers in the physical coordinate system and reconstructs these contributions as
the scattering signals in the backprojection image. Therefore, the shape and magnitude of a scattering signal in backprojection images indicate the influence of a
scatterer (defect) in the structure. In other words, the presence of defects or damages
294
triggers the scattering effect and is revealed by the backprojection processing.

The backprojection images contains two types of scattering signals; defect scattering signal (or defect signal) and background scattering signal (or background signal).
Defect scattering signals are due to the presence of defects, while background scattering signals are attributed to the direction reflection from the surface of the structure,
and therefore, are related to the geometry of the structure. Background scattering signals can be identified providing the information about the geometry of the structure.
This is usually the case in field inspection. After excluding background scattering signals, the remaining scattering signals are believed related to the presence (location,
size, and orientation) of defects.
7.3.2
Progressive Image Focusing
As a result of the sub-aperture and sub-band processing nature of the backprojection

algorithm, sub-aperture and/or sub-band images can be rendered before the far-field
ISAR measurements are completed. Summation of sub-aperture and/or sub-band
images can be conducted in part or in whole to progressively focus the reconstructed
image. Coherent summation/integration is conducted by time-shifting and spacealigning the signal collected at each aperture position for every pixel in the image.
Complete reconstruction is accomplished after the integration is done throughout the
entire data plane.
During this summation process, advantages are offered for the convenience of field
inspection because:
1. Processing efficiency is provided by dividing the entire data plane (Figure 72) into several sub-domains and by performing one-dimensional (1D) inverse
Fourier transformation (IFT) for each sub-domain in the data plane.
2. Sub-aperture and sub-band images can be used for the purpose of preliminary
inspection which can be rapidly completed. This is especially advantageous for
the inspection of civil infrastructures. Inspection flexibility is offered by the
option of rendering images at sub-aperture or sub-band levels.
295
3. The convergence of image pattern can be used to understand the features of

the structure (background signals). With this knowledge, better identification
of defect signals can be achieved.
Two types of integration, frequency and angular, are described in the following sections, using the oblique incidence (specular recessive) ISAR frequency-angle data of
the specimen AD1 as an example.
Frequency Integration
Integration of the far-field ISAR frequency-angle data over certain frequency bandwidth can be conducted at constant center frequency (Figure 7-9 (a)) or at shifting
center frequency (Figure 7-9 (b)). Improvement of image resolutions using both confmax
fmax
fc
fc
fmin
fmin
Bmin
Bmax
(a) Constant center frequency
Bmin
Bmax
(b) Shifting center frequency
Figure 7-9: Two types of frequency bandwidth integration

stant center frequency and shifting center frequency schemes are provided in Figure
7-10 for the images produced using HH polarized radar signals, in which Bmin =
0.2 GHz and Bmax = 4 GHz. Image resolutions, range and cross-range, are better
improved when shifting center frequency scheme is used. Formulae are provided in
Table 7.1 for estimating image resolutions (range and cross-range) through the backprojection processing in this development. In Table 7.1, c is the speed of light in free
space (c = 3 108 m/s), B is the bandwidth (GHz), and c is the wavelength of
296
Range
Cross-range
0.8
Resolution (m)
Resolution (m)
0.6
0.4
0.2
Range
Cross-range
0.8
0.6
0.4
0.2
2
Bandwidth (GHz)
(a) Constant center frequency
2
Bandwidth (GHz)
(b) Shifting center frequency
Figure 7-10: Improvement of image resolutions Progressive image focusing Frequency integration, HH polarization
c
) (m). Range resolution is accurately estimated in general,
fc
while the empirical constants in Table 7.1 are associated with various degrees of error
center frequency (c =
as shown in Figure 7-11. The approximation error of cross-range resolution using the
empirical formulae is less than 3 cm when the bandwidth is greater than 1 GHz.
Table 7.1: Image resolution formulae
Resolution
Constant
center
frequency
Shifting
center
frequency
Range
c
2B
c
2B
Cross-range
5.72
c
B
c
B
Figure 7-12 shows the performance of bandwidth integration using the frequencydata of the specimen AD1. In this case, shifting center frequency is adopted. The
center frequency ranges from 8.2 GHz (bandwidth = 0.4 GHz) to 9.8 GHz (bandwidth
= 3.6 GHz). It is found that the contrast of defect signals with background signals
297
0.25
Shifting fc
Constant fc
0.2
Error (m)
0.15
0.1
0.05
0.5
1.5
2
2.5
Bandwidth (GHz)
3.5
Figure 7-11: Prediction error of cross-range resolution formulae

improves significantly when the used bandwidth increases. The presence of defect
is gradually revealed in the backprojection images with B > 1.6 GHz (Figure 7-12
(d), B = 1.6 GHz). The example in Figure 7-12 demonstrates the effectiveness of
frequency integration on improving image resolutions for damage detection.
Angular Integration
Unlike the bandwidth integration, integration of the far-field ISAR frequency-angle
data over certain angular range may not guarantee the improvement of damage detectability due to the angular-dependent detectability of defects. For instance, in
the specular dominant circumstance, background signals cover defect signals in the
reflection response. On the other hand, the weakening/recessing of background signals reveals defect signals in the specular recessive circumstance. Averaging images
in these circumstances will not improve the damage detectability in the specular
dominant circumstance to a significant extent, but will reduce the one in the specular recessive circumstance dramatically. Therefore, angular integration should be
conducted by incorporating images only at incident angles in the specular recessive
298
circumstance. These angles are considered effective on detecting near-surface defects

in the GFRP-wrapped concrete columns.
Another issue is the variation of cross-range resolution, xr , with respect to bandwidth and angular support. Both constant and shifting center frequency schemes are
investigated. Figures 7-13(a) and (b) show the variation of xr (B, ) in the angular
range of [30, 30] ( or deg.) and in the bandwidth range of B [0.2, 4] (GHz). In
these figures, xr is improved with the increase of bandwidth B as expected. Figures
7-13(c) and (d) show the increase of required angular support for an increasing B. In
order to achieve the maximum xr at a given combination of fc and B, a minimum
amount of angular measurements is needed, which is the size of angular support.
The fundamental principle behind the effectiveness of detecting near-surface defects in the GFRP-concrete systems is the scattering effect at oblique incidence. Such
effectiveness is evaluated by the magnitude of scattered signals, which is also a function of incident angle (or azimuth angle). Figures 7-14 and 7-15 show the comparison
of reconstructed images of the intact(without defect) and damaged (with defect) surfaces of the specimen AD1 using full bandwidth signals (8GHz to 12GHz).
7.4
Image Resolutions and Damage Detectability
In the FAR NDT technique, it is the reconstructed backprojection images that are
used as the basis for damage detection. Therefore, the image resolutions (range and
cross-range) are crucial for the damage detectability of the technique and for revealing
the presence of near-surface defects by reconstructing scattering signals.
However, it is noteworthy to point out the following:
1. Near-surface defects can be detected only when the scattering signals can be
physically reconstructed in a backprojection image. This is associated with
the penetration depth of far-field radar signals, which is related to the incident
frequency and the dielectric properties of the medium; higher frequencies and
lossy media result in poor penetration and vice versa.
299
2. While the image resolutions can be estimated by the formulas provided in

Chapter 5 Image Reconstruction, it is believed that the angular-dependent
nature of the scattering signals due to near-surface defects plays a critical role
in the damage detectability of the technique. Image resolutions may be related
to the damage detectability only when (1) effective incident angles are used,
and (2) sufficient frequency bandwidth is provided.
3. When effective incident angles are used with sufficient frequency bandwidth in
the reconstructed images, image resolutions represent the capability of the images for damage detection. This capability is (1) to distinguish defect scattering
signals (due to near-surface defects) from background scattering signals (due to
the edges or geometrical features of the structure) and (2) to distinguish two
adjacent defect scattering signals.
With the above discussions, it is clear that the damage detectability of the FAR NDT
technique should be interpreted in conjunction with the following:
The ability of far-field radar signals to penetrate/reach near-surface defects
(penetration depth)
The condition of effective incident angles
The condition of sufficient frequency bandwidth
For the practical use and field applications of the FAR NDT technique, it is realized
that quantitative indices are usually more useful than qualitative guidance. Therefore,
quantitative interpretations on the backprojection images are believed to be of great
potential for practical use and, therefore, to be addressed in the following sections.
7.5
Pattern Recognition Damage Detection
Damage detection for the structural assessment of GFRP-wrapped concrete systems is

conducted by interpreting scattering signals in the reconstructed images. To demonstrate this, reconstructed images of the intact side (without defect) and the damaged
300
side (with defect) of the specimen AD1 (Figure 7-16 (a) and (b)) are rendered. Images reconstructed from the ISAR measurements (Figure 7-16 (c)) of the specimen
AD2 are generated for demonstrating the effectiveness of damage detection. In this
research, the maximum amplitude (local index) and the pattern of backprojection
images (global index) are used for quantifying the presence of defects. While the
maximum amplitude of a backprojection image locally indicates the significance (size,
angular sensitivity) of scattering signals (defect or background), the morphological
pattern of a backprojection image captures the global feature of the image. Knowing
that the backprojection image of a damaged structure contains defect signals in addition to background signals, the image pattern should be distinguishable from the
one of an intact structure. This difference is globally characterized by mathematical
morphology using a quantitative index (Eulers number) as a damage indicator. In
what follows, these two approaches are to be explained.
7.5.1
Local Index - Maximum Amplitude
Since the presence of defects leads to the appearance of scattering signals in the
backprojection images, no defect signals should be expected in the image of an intact
structure after excluding background signals. However, in some circumstances such
as the specular dominant case as shown in Figure 7-4, background signals become
dominant and cover the defect signal. This is the case when low signal-to-noise ratio
(SNR) is encountered, and the maximum amplitude is associated with background
signals. On the other hand, in the specular recessive case as shown in Figure 7-5,
high SNR can be expected due to the alleviation of background signals (noise), and
the maximum amplitude is likely to be associated with defect signals. The maximum
amplitude in the backprojection images is defined by

i
Imax
= max I i (x, y) |(x, y) s
(7.1)

d
Imax
= max I d (x, y) |(x, y) s
(7.2)
or
301
i
d
where Imax
and Imax
are the maximum amplitude in the backprojection images of
intact structure, I i (x, y), and of damaged structure, I d (x, y), respectively. s is the
physical domain of the specimen in the backprojection images defined by (x, y). The
difference between I i (x, y) and I d (x, y) is the defect signal.
The images of the specimen AD1 are used as an example. In the use of maximum
amplitude as a damage indicator, identifying background signals is important (in
order to exclude them), and it is also important to know how background signals affects the total response. Maximum amplitudes are extracted from the backprojection
images in two scenarios; (i) including background signals (blind test), and (ii) excluding background signals (a priori test). Evaluating a backprojection image without
excluding background signals is a blind test, from which the overall maximum amplitude is obtained. Excluding background signals requires a priori knowledge about
the structure. Figures 7-17 (a) and (b) show the maximum amplitudes of backprojection images produced from the intact and damaged sides of the specimen AD1 in
both scenarios using HH polarized signals. In this example, it is observed that the
defect signal is detectable in two angular regions, [30 , 7 ] and [7 , 30 ],
in blind test scenario in Figure 7-17(a). Two similar regions are also found in Figure
7-17(b) where background signals are excluded, but with better detectability due to
the significant separation between the curves. In the angular region (7 , 7 ),
specular effects are dominant, and the difference is not distinguishable.
Differential amplitude between maximum amplitude curves is calculated for both
blind test and a priori test scenarios, as shown in Figure 7-18. The differential
amplitude, A, is determined by
A =
Ad A i
Ai
(7.3)
where Ad represents the damage curve and Ai the intact curve in Figure 7-17. In
Figure 7-18, A is shown in percentage, whose pattern demonstrates the sensitivity
and effectiveness of incident angle on damage assessment. The greater the difference
is, the more effective the chosen incident angle is. Figure 7-18 also indicates that,
302
in the example presented in this section, the optimal incident angle is approximately
15 deviating from perpendicular (normal) incidence. Findings obtained from the
example presented in this section include:
1. Effectiveness of incident angle on damage assessment is affected by the measurement scheme. In the specular recessive scheme as shown in Figure 7-4,
background signals dominate the total response, making defect signals undetectable. This is the case of (7 , 7 ) in Figures 7-17(a) and (b).
2. With a priori knowledge about background signals, detectability can be greatly
improved.
3. The shape of the maximum amplitude curve of the intact side indicates the
pattern of background signals.
4. Detectability of defects is sensitive to the selection of incident angles. A defect
or damage is more detectable in some ranges of incident angle than in other
ranges.
7.5.2
Global Index - Mathematical Morphology
Pattern recognition of the backprojection images is conducted by extracting the features (edges) from the images and by evaluating the features quantitatively using
mathematical morphology. Global features (e.g., shape of scattering signals) are characterized and quantitatively evaluated using techniques in mathematical morphology
[Shirai 1987; Marchand-Maillet and Sharaiha 2000; Nixon and Aguado 2002]. In this
research, the backprojection images are evaluated by mathematical morphology in
the following two steps:
1. Feature extraction - The backprojection images are rendered with continuous
response levels, in which both background and defect signals are involved. In
order to extract the characteristic shape of a backprojection image, the image
is first transformed into a binary image based on a threshold value nthv . Two
morphological operations, erosion and dilation, are subsequently applied to the
303
binary image to obtain a feature-extracted version of the original backprojection

image. These morphological operations are defined by
K (I) = {
r|Kr I(x, y)}
(7.4)
V (I) = {
r|Vr I(x, y) 6= }
(7.5)
where I(x, y) is the backprojection image, K is the erosion operation functioning with the erosion structure K, Kr is the eroded set operating at position r,
V is the dilation operator functioning with the dilation structure V , Vr is the
dilated set operating at r, and is the empty set. An eight-node element is
adopted for both erosion and dilation structures, as shown in Figure 7-19. The
feature extraction operation on I(x, y) is performed on the binary version of
I(x, y) in this research, denoted by IBW (x, y|nthv ). The operation is defined by
I (x, y|nthv ) = V [K [IBW (x, y|nthv )]]
(7.6)
where I (x, y|nthv ) is the feature-extracted binary image characterized by a

threshold value nthv related to the level of the extracted edge in the image.
2. Feature quantification - A quantitative index used in this research to globally
characterize I (x, y|nthv ) is Eulers number, nE . The variation of nE with respect
to the incident angle is investigated. For each I (x, y|nthv ) obtained at a given
incident angle , nE is defined by
nE (|nthv ) = nobj (|nthv ) nhol (|nthv )
(7.7)
where nobj (|nthv ) is the number of objects in I (x, y|nthv ), and nhol (|nthv ) is
the number of holes within the objects in I (x, y|nthv ). With a fixed value of
nthv , nE () can be obtained.
The presence of damages introduces additional defect signals into backprojection
images globally, and changes the maximum amplitude locally. Logically, the presence
304
of scatterers (defect or background) leads to an increasing nobj . The value of nE is

subsequently altered. Given same nthv and same inspection domain s , the fluctuation
of defect scattering signals will create more holes than objects, thus, resulting in
small nE (). The purpose of using mathematical morphology is to quantify such
change. Additionally, in view of the angular sensitivity of defect signals, it is believed
that damage assessment based on single measurement (or image) is unlikely reliable.
Multiple images (more information) are needed to confirm the speculation on one
suspicious image. For this reason, an averaging (low-pass) filter is applied to the
nE () curve, which is defined by
int /2
nfE
() =
X
=int /2
nE ()
L
(7.8)
where nfE () is the filtered nE () curve, and L is the length of the filter. The purpose of this filter is to remove local fluctuations from the original nE () curve in
order to (i) avoid false alarm at local level and (ii) obtain a globally consistent result. Additionally, the length of the filter suggests the required amount of angular
measurements. The length of the filter also relies on the resolution of the image.
For high resolution images, small L is expected. The backprojection images used for
morphological processing are shown in Figures 7-20 and 7-21. In these figures, the
physical location of the specimen is indicated by a solid-line rectangle which is the
inspection domain s . I i (x, y)
In applying mathematical morphology for damage detection, the value of nthv must
be determined. The value of nthv is decided when the variation of Eulers number nE is
at its critical stage. When small values of nthv are chosen, all or most of the signals are
preserved, leading to a feature-extracted image dominated by low amplitude signals.
When large values of nthv are chosen, I (x, y|nthv ) will be dominated by high amplitude
signals. Consequently, the computed Eulers number nE in these two extreme cases
does not represent/reflect the main feature of the image. To avoid these misleading
circumstances, the pattern of nE is first investigated as shown in Figures 7-22 and 723 in which nthv = 1 corresponds to the maximum amplitude of I (x, y|nthv ). Figure
305
7-22 shows the variation of nE with the indication of critical nthv = 0.81 for both HH
and VV polarized signals in the intact-side images. Same pattern is also found in the
damaged-side images as shown in Figure 7-23 with a critical value of nthv = 0.73.
Figure 7-22(a) is produced from Figure 7-20(a) and Figure 7-22(b) from Figure
7-21(a). Figure 7-23(a) is produced from Figure 7-20(b) and Figure 7-23(b) from Figure 7-21(b). Figure 7-24 shows the reconstructed backprojection images I (x, y) and
their feature-extracted version I (x, y|nthv ) of the specimen AD1 using HH polarized
signals at full bandwidth. With the selection of critical nthv ,I (x, y|nthv ) captures the
main feature of the original I (x, y). The Eulers number of the intact-side image
I (x, y|nthv = 0.81) is nthv = 1 . For the damaged-side image I (x, y|nthv = 0.73),
nthv = 2.The feature-extracted images using VV polarized signals also provide similar result and are not repeatedly shown here. Following the same procedure described
above, I (x, y|nthv ) can be produced for other incident angles, resulting in the intactside and damaged-side curves nE () of the specimen AD1 as shown in Figure 7-25.
The nE () curves in Figure 7-25 demonstrate the sensitivity and effectiveness of incident angle with respect to the damage indication using Eulers number. Since the
scattering due to defects is angle-dependent, evaluating the structure using images at
several incident angles is needed. This leads to the application of an averaging filter
to obtain nfE () curves. Findings obtained in the use of mathematical morphology
for damage detection are summarized in the following.
1. Figure 7-26 shows the nfE () curves using a filter length of L = 3. In Figure 7-26,
the filtering produces a clear separation between the intact-side and damagedside nE () curves, except in the (10 , 10 ) region. The nE () values of
the intact-side images are in general greater than the ones of the damaged-side
images since the presence of defect signals creates more holes in the images,
resulting in smaller values of nE ().
2. The filter length L is related to the required amount of angular measurements
for achieving a globally consistent assessment. Knowing that specular effects
are dominant in the angular range of (7 , 7 ) as found in the use of
306
maximum amplitude for damage detection, separation between the nE () curves

of intact and damaged images is required outside the range of (7 , 7 ).
L is subsequently determined in order to achieve this goal. In the example
presented in this section, at least three angular measurements at an interval of
1 (3 of angular range) are needed for each comparison between the images of
intact and damaged structures.
3. In the case of the absence of background signals in backprojection images, variation of nobj and nhol can be used instead.
7.6
Summary
In this chapter, a non-contact, far-field radar NDT method capable of detecting (1)
GFRP (glass fiber reinforced polymer) delamination, (2) GFRP debonding from concrete substrate, and (3) sizeable air pockets trapped between GFRP sheets and concrete substrate during manufacturing in near-surface region of concrete and masonry
structures wrapped and/or bonded with GFRP sheets/plates, using monostatic inverse synthetic aperture radar (ISAR) measurements and tomographic image reconstruction algorithms is described.
The proposed NDT capability consists of three components; a physical inspection component allowing distant assessment of civil infrastructures, a numerical
processing component transforming reflection response to spatial profile of the structure, and a pattern recognition component quantitatively evaluating the presence
of damage.
In physical inspection, far-field ISAR measurements of the structure under investigation are collected at different frequencies and inspection (azimuth) angles. Steadystate reflection responses (amplitude and phase) of the structure are collected at each
frequency and angle, forming the frequency-angle measurements. Frequency-angle
measurements are processed by an image reconstruction algorithm based on tomographic reconstruction methods to render the transformed in-depth profiles (range307
cross-range images) of the structure. Finally, a pattern recognition technique based on

mathematical morphology analyzes the range-cross-range images with a quantitative
index for condition assessment. Information regarding the presence and location of
anomaly is obtained from the result of condition assessment and can be used for routine maintenance, as well as for special purpose safety assessment, of GFRP-retrofitted
concrete structures including bridges, buildings, and walls, as well as masonry walls
strengthened by GFRP sheets. This information is also beneficial to the management of civil infrastructures in the way resources for repair and rehabilitation can be
effectively allocated.
308
0.8
0.8
0.8
0.6
0.6
0.6
0.4
0.4
0.4
Cross-range (m)
Cross-range (m)
0.2
0
-0.2
-0.4
Cross-range (m)
0.2
0
-0.2
-0.4
-0.4
-0.6
-0.6
-0.8
-0.8
-0.8
-1
-1
-0.5
0.5
-1
-1
-0.5
Range (m)
0.5
-1
(b) fc = 8.4GHz, B = 0.8GHz
0.8
0.6
0.6
0.6
0.4
0.4
0.4
-0.4
Cross-range (m)
0.8
Cross-range (m)
0.8
-0.2
0.2
0
-0.2
-0.4
0
-0.2
-0.4
-0.6
-0.6
-0.8
-0.8
-0.8
0.5
-1
-1
-0.5
Range (m)
0.5
-1
-1
(e) fc = 9.0GHz, B = 2.0GHz

0.8
0.8
0.6
0.6
0.6
0.4
0.4
0.4
-0.2
-0.4
Cross-range (m)
0.8
Cross-range (m)
0.2
0
-0.2
-0.4
0
-0.2
-0.4
-0.6
-0.6
-0.8
-0.8
-0.8
0.5
-1
-1
-0.5
0.5
Range (m)
Range (m)
(g) fc = 9.4GHz, B = 2.8GHz
(h) fc = 9.6GHz, B = 3.2GHz
-1
-1
-0.5
0.5
Range (m)
(i) fc = 9.8GHz, B = 3.6GHz
Figure 7-12: Progressive image focusing Frequency integration using shifting center
frequency, HH polarization, = 15
309
0.2
-0.6
-0.5
0.5
(f) fc = 9.2GHz, B = 2.4GHz
0.2
Range (m)
-1
-1
-0.5
Range (m)
(d) fc = 8.8GHz, B = 1.6GHz
0.2
-0.6
-0.5
0.5
(c) fc = 8.6GHz, B = 1.2GHz

1
0.2
Range (m)
-1
-1
-0.5
Range (m)
(a) fc = 8.2GHz, B = 0.4 GHz
Cross-range (m)
0
-0.2
-0.6
-1
Cross-range (m)
0.2
(a) Constant fc xr ( B, )
(b) Shifting fc xr ( B, )
0.35
0.35
B = 1 GHz
B = 2 GHz
B = 4 GHz
0.25
0.2
0.15
0.1
0.05
0
-30
B = 1 GHz
B = 2 GHz
B = 4 GHz
0.3
Cross-range resolution (m)
Cross-range resolution (m)
0.3
0.25
0.2
0.15
0.1
0.05
-20
-10
0
10
Incident angle, (deg.)
20
0
-30
30
(c) Constant fc xr ( B [1, 2, 4] , )
-20
-10
0
10
20
(d) Shifting fc xr ( B [1, 2, 4] , )
Figure 7-13: Improvement of image resolutions progressive image focusing Angular

integration, HH polarization
310
30
0.5
0.4
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0
-0.1
-0.2
Cross-range (m)
0.5
Cross-range (m)
Cross-range (m)
0.5
0.1
0
-0.1
-0.2
0.1
0
-0.1
-0.2
-0.3
-0.3
-0.3
-0.4
-0.4
-0.4
-0.5
-0.5
-0.5
-0.5
0.5
Range (m)
(a-1) 30
0.5
-0.5
-0.5
0.5
Range (m)
Range (m)
(a-2) 20
(a-3) 10
0.5
0.5
0.4
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0
-0.1
-0.2
Cross-range (m)
0.5
Cross-range (m)
Cross-range (m)
(a) Intact surface
0.1
0
-0.1
-0.2
0.1
0
-0.1
-0.2
-0.3
-0.3
-0.3
-0.4
-0.4
-0.4
-0.5
-0.5
Range (m)
(b-1) 30
0.5
-0.5
-0.5
Range (m)
(b-2) 20
0.5
-0.5
-0.5
Range (m)
(b-3) 10
(b) Damaged surface

Figure 7-14: Comparison of images of the intact and damaged surfaces of the specimen
AD1 at different incident angles (30 10 )
311
0.5
0.5
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0
-0.1
-0.2
Cross-range (m)
0.5
0.4
Cross-range (m)
Cross-range (m)
0.5
0.1
0
-0.1
-0.2
0.1
0
-0.1
-0.2
-0.3
-0.3
-0.3
-0.4
-0.4
-0.4
-0.5
-0.5
-0.5
-0.5
0.5
Range (m)
-0.5
-0.5
0.5
Range (m)
(a-4) -10
0.5
Range (m)
(a-5) -20
(a-6) -30
0.5
0.5
0.4
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0
-0.1
-0.2
Cross-range (m)
0.5
Cross-range (m)
Cross-range (m)
(a) Intact surface
0.1
0
-0.1
-0.2
0.1
0
-0.1
-0.2
-0.3
-0.3
-0.3
-0.4
-0.4
-0.4
-0.5
-0.5
0.5
-0.5
-0.5
0.5
-0.5
-0.5
Range (m)
Range (m)
Range (m)
(b-4) -10
(b-5) -20
(b-6) -30
(b) Damaged surface

Figure 7-15: Comparison of images of the intact and damaged surfaces of the specimen
AD1 at different incident angles (10 30 )
312
0.5
= 30
30
30
0
30
= 30
= 30
(a) Specimen AD1
Intact side
(b) Specimen AD1

Damaged side
(c) Specimen AD2

Damaged side
-15
-15
-20
-20
-25
-25
Max. amplitude (dBsm)
Max. amplitude (dBsm)
Figure 7-16: Description of the used far-field ISAR measurements and specimens for
damage detection
-30
-35
-40
-45
-50
Damaged
Intact
-55
-60
-30
-20
-10
0
10
20
-30
-35
-40
-45
-50
Damaged
Intact
-55
30
(a) Blind test
-60
-30
-20
-10
0
10
20
(b) A priori test
Figure 7-17: Maximum amplitudes of the backprojection images of the specimen AD1
Full bandwidth (8GHz12GHz), HH polarization
313
30
12
Blind test
A priori test
Differential amplitude (%)
10
0
-30
-20
-10
0
10
20
30
Figure 7-18: Differential maximum amplitudes of the backprojection images of the

specimen AD1 Full bandwidth (8GHz12GHz), HH polarization
r (xj, yj )
x
Figure 7-19: An eight-node element for morphological operations
314
(dBsm)
(dBsm)
-25
0.3
-25
0.3
-30
-30
0.2
0.2
-35
-35
-40
-45
0
-50
-55
-0.1
Cross-range (m)
Cross-range (m)
-40
0.1
0.1
-45
0
-50
-55
-0.1
-60
-60
-65
-0.2
-65
-0.2
-70
-70
-0.3
-0.3
-0.3
-0.2
-0.1
0
0.1
Range (m)
0.2
0.3
-75
-0.3
-0.2
(a) Intact side
-0.1
0
0.1
Range (m)
0.2
0.3
-75
(b) Damaged side
Figure 7-20: Backprojection images of the specimen AD1 HH polarization, = 15
(dBsm)
(dBsm)
-25
0.3
-25
0.3
-30
0.2
-30
0.2
-35
-35
-45
0
-50
-55
-0.1
-40
Cross-range (m)
Cross-range (m)
-40
0.1
0.1
-45
0
-50
-55
-0.1
-60
-65
-0.2
-60
-65
-0.2
-70
-0.3
-70
-0.3
-0.3
-0.2
-0.1
0
0.1
Range (m)
0.2
0.3
-75
(a) Intact side
-0.3
-0.2
-0.1
0
0.1
Range (m)
0.2
0.3
(b) Damaged side
Figure 7-21: Backprojection images of the specimen AD1 VV polarization, = 15
315
-75
20
Euler's number, n
Euler's number, n
15
10
15
10
0.2
0.4
0.6
Threshold value, n
0.8
0.2
thv
0.4
0.6
Threshold value, n
0.8
thv
(b) VV polarization
(a) HH polarization
Figure 7-22: Variation of nE with respect to nthv of the intact-side of the specimen
AD1
30
30
25
Euler's number, n
Euler's number, n
25
20
15
10
20
15
10
5
0
0
0.2
0.4
0.6
Threshold value, n
0.8
thv
(a) HH polarization
0.2
0.4
0.6
Threshold value, n
0.8
thv
(b) VV polarization
Figure 7-23: Variation of nE with respect to nthv of the damaged-side of the specimen
AD1
316
(a) Intact-side images nthv = 0.81
(b) Damaged-side images nthv = 0.73
Figure 7-24: Backprojection images and their feature-extracted version of the specimen AD1 (HH polarization, full bandwidth, = 15 )
2
AD1intact
1.5
AD1damaged
Euler' number
0.5
0
-0.5
-1
-1.5
-2
-2.5
-3
-30
-20
-10
0
10
Inspection angle (deg.)
20
30
Figure 7-25: Original nE () curves of the intact and damaged surfaces of the specimen
AD1 (HH polarization, full bandwidth)
317
1.5
AD1intact
AD1damaged
Averaged Euler' number
0.5
-0.5
-1
-1.5
-2
-30
-20
-10
0
10
Inspection angle, (deg.)
20
30
Figure 7-26: Filtered nE () curves of the intact and damaged surfaces of the specimen
AD1 (HH polarization, full bandwidth, L = 3)
318
Chapter 8
Conclusions
In this thesis, a novel NDT technique termed FAR NDT using far-field monostatic
radar measurements for the distant inspection of GFRP-retrofitted concrete cylinders
is developed. In particular, the investigation of dielectric modeling of construction
materials, application of the FDTD method to the far-field response of dielectric
cylinders with local cavity, and processing of far-field radar measurements into spatial
images using the fast backprojection algorithm are presented.
In what follows, obtained research findings are first concluded. The contributions
of this research are summarized, and possible future work is suggested.
The research findings from this work include:
Current NDT techniques and their applicability on GFRP-concrete systems are
reviewed in this research. A distant NDT technique that is capable of providing
in-depth information of GFRP-concrete systems in field conditions is lacking.
There is a need to develop such technique for the condition assessment (safety
inspection) and proper maintenance of GFRP-concrete systems and potentially
for the damage detection of concrete and reinforced concrete structures.
Numerical simulation
1. FDTD simulation can demonstrate the multiple reflections within and the
scattering response of dielectric media, providing an accurate description
to material properties and geometry.
319
2. The simulated scattering response of a concrete cylinder (dielectric) with

four rebars indicates strong reflections not only in the line-of-sight direction
of the radar but also in other directions. This is the reason why far-field
monostatic ISAR normal incidence measurements can be used for detecting
rebars.
3. FDTD simulation can be used as a tool for understanding the pattern of
scattering signals due to the presence of near-surface defects in dielectric
media. The knowledge learned from FDTD simulation is beneficial for
planning an effective experimental program and for identifying key system
design parameters including measurement scheme, measurement frequency,
and the range of incident angles.
Measurement scheme
1. From laboratory radar measurements, far-field monostatic ISAR oblique
incidence scheme can be conclusively used for detecting near-surface defects
such as GFRP debonding in GFRP-retrofitted concrete cylinders, while the
normal incidence scheme cannot provide a conclusive indication.
2. The magnitude of far-field monostatic ISAR normal incidence measurements is sensitive to the geometry, surface roughness, and dielectric properties of the target structure. Variation of these properties in normal incidence measurements can obstruct the detection of near-surface defects.
Measurement frequency
1. Continuous wave radar signals in the frequency range of 8 GHz to 18
GHz are found to be effective for detecting near-surface defects in GFRPretrofitted concrete cylinders.
2. The use of higher frequency signals covers more angular regions than the
use of lower frequency signals, in which defect signals appear.
Range of incident angles
320
1. In the reconstructed images of the specimen AD1, two angular regions,

[10 25 ] and [10 20 ], are found to be effective for detecting the
presence of a cubic-like defect. Detectability can be further improved if
the background signal can be reduced/removed. Incident angles outside
the range [30 30 ] are not effective for damage detection.
2. The result by mathematical morphology reveals the observation in laboratory radar measurements on the multiple angular regions for damage
detection. The angular region [7 8 ] is found not effective for detecting near-surface defects.
1. The backprojection processing of monostatic ISAR oblique incidence measurements is capable of visualizing and locating the presence of nearsurface defects; with a bandwidth [8GHz12GHz], a 1.5-by-1.5-by-1
(cubic-like) near-surface defect can be detected, and a 3-by-3-by-0.2
(delamination-like) near-surface defect can be detected with detailed features (edges).
2. The use of a shifting center frequency provides better convergence of crossrange resolution than the use of a constant center frequency.
3. Maximum amplitude of the reconstructed images can be used as an index
for damage detection.
4. In the image reconstruction algorithm, dielectric properties of the target
structure (GFRP-concrete cylinder) are not necessary; radar signals are
assumed propagating in free space. This is the reason why the backprojection processing can provide satisfactory results for detecting near-surface
defects; changes in the phase velocity of radar signals are insignificant when
only a small portion of propagation distance in dielectric media (non-free
space) is involved.
Dielectric properties and modeling
321
1. Dielectric properties are the key information for the correct modeling of
materials in numerical simulation, as well as for improving the accuracy of
backprojection processing for detecting deep defects.
2. In the proposed methodology for determining dielectric properties of materials, the estimation of TDOA is crucial; thickness of the specimen must
be greater than 0.236 in. (6 mm) for a reliable estimate of TDOA.
3. In the frequency range of 8 GHz to 18 GHz, signal-pole models can be
used for water, epoxy resin, and E-glass fabric considered as homogeneous
media. Dipolar polarization is the dominant dispersion mechanism in this
frequency range.
4. Wieners dielectric mixing law is applicable for GFRP (epoxy resin and
E-glass fabric).
5. Dielectric properties of cementitious (porous) composites such as concrete
depend on not only the volumetric fractions of each component but also
the distribution of each component. At the present time, lack of a complete
dielectric database for cementitious composites is a major obstacle for developing a physically-sound dielectric model for multi-phase cementitious
composites.
Contributions of this research are summarized in the following:
1. A distant NDT capability (FAR NDT) utilizing far-field monostatic ISAR measurements for the in-depth inspection of GFRP-concrete structures is developed
for the safety assessment of the structures [280]. This capability is applicable
for field applications such as highway bridge piers, cross-river bridge piers and
cross-valley bridge piers strengthened with GFRP composites.
2. The feasibility of using the backprojection algorithm for reconstructing the spatial profile of multi-layered dielectric systems is demonstrated. The representation of near-surface defects in the backprojection images is provided.
322
3. The use of FDTD simulation as a predictive tool to estimate/explain the reflection response of dielectric media with local features (near-surface defects and
rebars) is demonstrated.
4. An integrated methodology for the determination of unique combination of
complex permittivity using transmission-only, coherent, wide-bandwidth freespace measurements is developed [36].
5. Dielectric models of materials including water, epoxy resin, E-glass fabric,
GFRP, and oven-dried hydrated cement paste are developed for their use in
the frequency range of 8 GHz to 18 GHz.
Upon the research findings based on preliminary laboratory testing and imaging
results and addressed in this research, suggested future work is summarized in the
following.
1. Field performance of the FAR NDT technique Upon the excellent
performance of the FAR NDT technique in a laboratory environment, it would
be of interest to know how the technique performs on real structures in field
conditions. Not only the background electromagnetic noises (stationary and/or
non-stationary) could present, but also the operational constraints in the field
could lead to the need for further optimization of the technique. These issues
cannot be fully understood without conducting field measurements using the
technique. Consequently, the development of a portable phototype system of
the FAR NDT technique is needed, as well as the development of denoising
strategy using finite support (e.g., wavelet transforms) and infinite support
(e.g., Fourier transforms) basis functions.
2. Dielectric modeling of multi-phase cementitious composites The dielectric modeling work presented in this dissertation is based on limited dielectric measurements reported in the literature [113]. Further experiments on the
cyclic absorption-and-desorption (of water molecules) behavior of cement and
323
concrete specimens are advantageous for better understanding the microstructure of cementitious composites and for the improvement of the proposed dielectric models in this research.
3. Image reconstruction Further investigation of the backprojection images
on the features of defect signals with respect to the properties (e.g., size, shape,
and orientation) of the defect can lead to a better interpretation of the images.
Additionally, modification to the image reconstruction algorithm by considering
the influence of dielectric properties is advantageous for improving the accuracy
of locating defects.
324
Appendix A
Phase Velocity of Love Waves in A
Layer Underlain by A Half Space
Medium
This appendix derives the analytical expression of the phase velocity of Love waves
in a layer underlain by a solid half space, which is mentioned in Chapter 2 Literature
Review.
A.E.H. Love (1911) [156] discovered the existence of one type of surface waves
(different from the Rayleigh wave) that can be hospitalized/trapped within layered
media when he investigated the propagation data of seismic waves. Ewing, Jardetzky,
and Press (1957) [272] followed the same approach provided by Love and derived the
period of Love waves in a thin layer underlain by a solid half space, as shown in Figure
A-1. In Figure A-1, a planar layer (Medium 2) with thickness of H is placed above
a half space medium (Medium 1). Only the solid medium is considered here. The
incident wave propagates from a source located in a distance far from the boundary
between Media 1 and 2, resulting in the case of plane wave incidence at the boundary.
After impinging the boundary, the transmitted wave bounces back-and-forth within
Medium 2, being trapped within the layer and forming Love waves.
325
z= - H
E2, 2, 2
Medium 2
z= 0
x
E1, 1, 1
ki
Medium 1
Incident wave
Figure A-1: A layer underlain a solid half space
326
Consider the period equation by Ewing, Jardetzky, and Press (1957) (Sec. 4-5,
p.210) as follows.
s
2
c
1
vs2
G2
s
tan k
1 H =
2
G1
c
1
vs1

(A.1)
is the wavenumber of Love waves, = 2f is the radian frequency,

vL
vs
1 =
L is the phase velocity of Love waves, f is the temporal frequency in Hertz,
r
2
Gi
c
is the shear wave velocity, and
1 , c is the speed of light, vsi =
vs1
i
r
Ei
Gi =
is the modulus of rigidity, i is Poissons ratio, is the density, and
2(1 + i )
H is the thickness of the layer. Knowing that
where k =
tan x = x +
x3 2x5 17x7
+
+
,
3
15
315
(A.2)
the first-two-term approximation is

x3
tan x
=x+
3
s
(A.3)
2
G2
c
1 and C2 =
vs1
G1
the approximated tangent function becomes
Letting C1 = 1 H = H

tan
C1
vL

=
C1
vL
1
+
3
C1
vL

1
2 s 2
c
c
/
1,
vs2
vs1
3
= C2
(A.4)
After re-arrangement, an inhomogeneous cubic equation is yielded.

C2
vL 3 3 vL 2 3 = 0
C1
C1
3 C1
= vL 3
vL 2
=0
C2
3 C2
3
(A.5)
Using the method by Scipione del Ferro (1465-1526) and assuming vs1 < c < vs2 , the
327
discriminant equation for determining the characteristic of roots is

4
=
3
C1
C2
2
"
3 C1
2
27
3 C2 27
C1
C2
3 #
>0
(A.6)
suggesting there is one real root and a pair of conjugate roots. The inhomogeneous
cubic equation is in the form of
a vL3 + b vL2 + c vL + d = 0
(A.7)
with coefficients
a=1
C1
b =
C2
c=0
3
C1
d=
3 C2
(A.8)
The parameters defined in del Ferros method are

C1 1
a3
=
3
C2 3

2 a3 9ab
1
C1
q =c+
=
2 + 9
27
27
C2
"
2
1/2 #1/3
q
p3
q
+
+
u=
2
4
27
p=b
(A.9)
(A.10)
(A.11)
Therefore, the phase velocity of Love waves in this model is

vL =
p
a
p
1
u =
u
3u
3
3u
3
328
(A.12)
where the coefficient
C1
C2
as
2
2
c
c

1/2
1
1
C1
G1
E1 (1 + 2 )
vs1
vs1
=H
"
"
2 #1/2 = H E2 (1 + 1 )
2 #1/2
C2
G2
c
c
1
1
vs2
vs2
1/2

1/2

21 (1 + 1 )c2
22 (1 + 2 )c2
E1 (1 + 2 )
1 1
(A.13)
=H
E2 (1 + 1 )
E1
E2

Since
C1
C1
=
(H, E1 , E2 , 1 , 2 , 1 , 2 , c)
C2
C2
p = p (, H, E1 , E2 , 1 , 2 , 1 , 2 , c)
q = q (, H, E1 , E2 , 1 , 2 , 1 , 2 , c)
u = u(p, q) = u (, H, E1 , E2 , 1 , 2 , 1 , 2 , c)
are determined by material properties (E1 , E2 , 1 , 2 , 1 , 2 ), geometry (layer thickness
H), and incident wave property (frequency ), vL can be fully determined once these
properties are provided.
329
330
Appendix B
Analytical Approach to Several
Plane Wave Incidence Problems
This appendix addresses the analytical approach to several electromagnetic scattering
problems in which plane wave incidence is considered. In the FAR NDT technique,
reflection response of the target structure is used for image reconstruction. The evaluation of reflected radar signals from a GFRP-wrapped concrete column is in nature
an electromagnetic scattering problem in which the use of far-field measurements
indicates the incidence of plane waves.
The purpose of this appendix is to evaluate the reflected/scattered fields in two
plane wave incidence problems. The default geometry of the target structure/scatterer
in the problems is circular, and the target scatterer is a non-conductive dielectric. Definition of reflection coefficient and reflectivity is first provided. One two-dimensional
scattering problem and one three-dimensional scattering problem are studied.
B.1
Reflection Coefficient and Reflectivity
Reflection responses or signals can be evaluated by the reflection coefficient (field

intensity) or reflectivity (power intensity). Consider a two-layer model with one
infinite boundary subjected to the impinging of a TE polarized plane wave traveling
toward the boundary, as shown in Figure B-1 . The dielectric properties of regions 0
331
and 1 are characterized by (0 , 0 ) and (1 , 1 ), respectively. The plane wave impinges
the boundary at an incident angle i , creating a reflected wave and a transmitted wave.
The incident fields are considered as [138]
Region 0
e0, m0
Hi
Hr
Ei
Incidence
Er
ki
qi
qr
Reflection
Hi
Hr
Ei
kr
kr
Er
ki
x
Transmission
Region 1
e1, m1
Ht
qt
Et
kt
Ht
Et
kt
: Incidence plane
Figure B-1: A two-dimensional two-layer model with infinite boundary (TE waves)
Ei = yE0 eikix x ikiz z

i = 1 Ei
H
i0
kix
kiz
= x
E0 eikix x ikiz z + z
E0 eikix x ikiz z
0
0
(B.1)
(B.2)
indicating an electric field propagating in (

x,
z ) direction with a +
y component,
and a magnetic field propagating in (
x,
z ) direction with +
x and +
z components.
The reflected fields are
Er = yR01 E0 eikrx x + ikrz z
krx
krz
r =
H
x
R01 E0 eikrx x + ikrz z + z
R01 E0 eikrx x + ikrz z
0
0
332
(B.3)
(B.4)
The transmitted fields are

Et = yT01 E0 eiktx x iktz z
t = x ktz T01 E0 eiktx x + iktz z z ktx T01 E0 eiktx x + iktz z
H
0
0
(B.5)
(B.6)
x ) at the boundary (z = 0) must be

The continuity of tangential fields (Ey and H
satisfied. Therefore, it provides two conditions for two unknowns.
Ey : Eiy + Ery = Ety = 1 + R01 = T01
x : Hix + Hrx = Htx = kiz krz R = ktz T
H
0
0
1
(B.7)
(B.8)
With the two conditions, the Fresnels reflection and transmission coefficients
of TE waves can be determined as follows.
ktz 0
kiz 1
=
ktz 0
1+
kiz 1
2
=
ktz 0
1+
kiz 1
1
T E
R
01
TE
T01
(B.9)
(B.10)
y and Ex at z = 0. Following
For TM waves the continuity conditions are held on H
same approach, the Fresnels reflection and transmission coefficients of TM
waves are obtained.
T M
R
01
TM
T01
ktz + kix 0
1
kiz + kix 1
=
ktz + kix 0
+1
kiz + kix 1
2
=
ktz + kix 0
+1
kiz + kix 1
(B.11)
(B.12)
Both TE and TM wave modes are illustrated in Figure B-2, along with the incident
wave vector ki . The power reflection coefficient or the reflectivity is defined to be
333
(TM)
Ei
ki
Hi
Hi
q0
Ei
(TE)
y
f0
x
Figure B-2: TE and TM waves and incident wave vector

T E 2
TE

r01
= R01
(B.13)
The power transmission coefficient or the transmissibility is then defined to be

TE
tT01E = 1 r01
=
0 ktx T E 2
T
1 kix 01
(B.14)
Notice that ktx and kix are the x component of wave vectors in Regions 1 and 0,
respectively. It is the reflectivity and the transmissibility of a target that are measured
by radar in laboratory or in field.
B.2
A Two-dimensional Three-layer Model
The influence of the presence of a thin layer to the reflectivity of layered systems can
be first evaluated by a two-dimensional three-layer model as shown in Figure B-3.
The reflection coefficient of the three-layer model is determined following the same
approach mentioned for the two-layer model, using the transmitted wave in Region 2
as the incident wave for Region 3. The general reflection coefficient for the three-layer
334
Hi
Region 0
e0, m0
1st reflection
2nd reflection
Ei
ki
qi
3rd
qr
reflection
x
Region 1
e1, m1
Interface
coefficients
R01, T01
R10, T10
qt
R12, T12
R21, T21
Region 2
e2, m2
Figure B-3: A two-dimensional three-layer model (TE waves)
model is [47]
2ik1z d1
01 = R01 + T01 R12 T10 e
R
1 R10 R12 e2ik1z d1
(B.15)
where
TE waves:
TE
Rmn
=
n kmz m knz
n kmz + m knz
(B.16)
TM
=
Rmn
n kmz m knz
n kmz + m knz
(B.17)
TM waves:
Wave vectors are determined by

kmz = kz cos m
(B.18)
knz = kz cos n
(B.19)
335
where m and n are determined by Snells law:

sin m
=
sin n
m m
n n
(B.20)
Reflection coefficients of multi-layer models can be derived by the propagation matrix approach [138], and by the transmission line theory in which the impedance of
materials is used [188].
B.3
A Three-dimensional Infinite Dielectric Cylinder

Model
In this section, the scattering problem of an infinite dielectric cylinder of radius a impinged by a plane wave is considered. The problem is a three-dimensional one since
the wave vector of incident waves is not required to be perpendicular to the axis of
the cylinder (Figure B-4). This problem characterizes the scattering of far-field radar
signals (plane waves) from GFRP-wrapped concrete columns. Derivation of the scattered fields in this problem is performed using the Hertzian potentials for solving the
Helmholtz wave equations in cylindrical coordinates. Unknown coefficients are determined by matching the boundary condition at the surface of the dielectric cylinder as
the result of continuation of tangential fields at the interface. For such problems, TE
and TM wave modes are usually coupled and needed in the representation of scattered
fields, except in some special cases (e.g., a PEC cylinder impinged by plane waves).
First, the TE and TM incident waves are considered in the cylindrical coordinates
(, , z).
TE waves:
Hzi = H0 sin 0 eiki r
Hi = H0 cos 0 cos( 0 )eiki r
Hi = H0 cos 0 sin( 0 )eiki r

336
(B.21)
(B.22)
(B.23)
ki
q0
y
x
f0
y
x
Figure B-4: A three-dimensional infinite dielectric cylinder impinged by plane waves
TM waves:
Ezi = H0 sin 0 eiki r
Ei = E0 cos 0 cos( 0 )eiki r
Ei = E0 cos 0 sin( 0 )eiki r
(B.24)
(B.25)
(B.26)
where ki = k(sin 0 cos 0 x + sin 0 sin 0 y + cos 0 z) is the wave vector of incident
waves, r = cos
x + sin
y + z z is the observation position vector, and ki r =
kz cos 0 + k sin 0 cos( 0 ).
Substituting these fields into Maxwells curl equations yields the relationship between
337
electromagnetic fields and Hertzian potentials.

2
2
Hzi =
+ ki TziE
z 2
= A k 2 sin2 eiki r

i i

2
2
Ezi =
+ ki TziM
z 2
= B k 2 sin2 eiki r
(B.27)
(B.28)
i i
(B.29)
(B.30)
where
H0
sin 0
E0
Bi = 2
k sin 0
Ai =
k2
(B.31)
(B.32)
The Hertzian potentials are then expanded in Fourier series with respect to for the
mathematical convenience of satisfying the boundary condition at the surface of the
cylinder.
TziE
TziM =
X
n=
Ai eikz cos 0 Jn (k sin 0 )ein( 0 + /2)
(B.33)
Bi eikz cos 0 Jn (k sin 0 )ein( 0 + /2)
(B.34)
n=
where Jn (k sin 0 ) is the Bessel function with argument (k sin 0 ). The scattered
fields are written in similar form with unknown coefficients Asn and Bsn .
TzsE
TzsM =
X
n=
Asn eiki z cos 0 Hn(2) (ki sin 0 )ein( 0 + /2)
(B.35)
Bsn eiki z cos 0 Hn(2) (ki sin 0 )ein( 0 + /2)
(B.36)
n=
338
(2)
where where Hn (ki sin 0 ) is the Hankel function of the second kind with argument
(ki sin 0 ). The transmitted fields inside the cylinder are
TztE =
TztM =
X
n=
Atn eikt z cos 0 Jn (kt sin 0 )ein( 0 + /2)
(B.37)
Btn eikt z cos 0 Jn (kt sin 0 )ein( 0 + /2)
(B.38)
n=
Unknown coefficients Asn , Atn , Bsn , and Btn are determined by the boundary condition which is the continuity of tangential fields (z and components) at = a. The
continuity conditions are
Ez :

(k 2 sin2 0 ) Bi Jn (ka sin 0 ) + Bsn Hn(2) (ki a sin 0 ) = kt2 sin2 t Btn Jn (kt a sin t )
(B.39)
Hz :

(k 2 sin2 0 ) Ai Jn (ka sin 0 ) + Asn Hn(2) (ki a sin 0 ) = kt2 sin2 t Atn Jn (kt a sin t ) (B.40)
E :

ki n cos 0
Bi Jn (ka sin 0 ) + Bsn Hn(2) (ki a sin 0 )
a h
i
0
+iki sin 0 Ai Jn0 (ki a sin 0 ) + Asn Hn(2) (ki a sin 0 )

=
ki n cos 0
[Btn Jn (kt a sin t )] + i sin t [Atn Jn0 (kt a sin t )]
a
(B.41)
H :

ki n cos 0
Ai Jn (ki a sin 0 ) + Asn Hn(2) (ki a sin 0 )
a
i
h
0
+iki sin 0 Ai Jn0 (ki a sin 0 ) + Asn Hn(2) (ki a sin 0 )

=
ki n cos 0
[Btn Jn (kt a sin t )] + i sin t [Atn Jn0 (kt a sin t )]
a
339
(B.42)
where t is determined by Snells law, which is

t = cos
ki
cos 0
kt

(B.43)
With these four conditions, the four unknowns can be evaluated completely. Considering only non-magnetic materials provides
k t = t t = t 0 = 0 0 r = k 0 r
Furthermore, choosing 0 =
Asn
(B.44)
pi
pi
leads to t =
and therefore,
2
2
E0 0
E0 r Jn0 (kt a)
J (ka)
Jn (k0 a)
ki n
kt Jn (kt a)
= 2
ki r Jn0 (kt a) (
0
Hn 2)(ki a) ki Hn(2) (ki a)
kt Jn (kt a)
(B.45)
kt2 Jn (kt a)
=
E0 Jn (ki a) + ki2 Hn(2) (ki a)Asn
(B.46)
r H0 Jn0 (kt a)
H0 0
Jn (ki a)
Jn (ki a)
ki
kt Jn (kt a)
=
ki2 Jn0 (kt a) (2)
0
Hn ki Hn(2) (ki a)
kt Jn (kt a)
(B.47)
Atn
Bsn
Btn =
kt2 Jn (kt a)
(2)
H0 Jn (ki a) + ki2 Hn (ki a)Bsn
340
(B.48)
The scattered fields outside the cylinder are evaluated.
Es =
Es
Asn
n=
(B.49)
Hzs = k02 TzsE
(B.50)
(2)
H (k)ejn(0 +/2)
n
(2)
Hn (k) jn(0 +/2)

Bsn
e
n=
n
Bsn
n=
(2)
H (k)ejn(0 +/2)
n
(B.52)
j
Asn Hn(2) (k)(jn)ejn(0 +/2)
n=
+
Hs = j
X
n=
X
n=
Bsn
(2)
Hn (k)ejn(0 +/2)
Asn
(2)
H (k)ejn(0 +/2)
n
1 X
Bsn Hn(2) (k)(jn)ejn(0 +/2)
+
n=
B.4
(B.51)
1 X
Asn Hn(2) (k)(jn)ejn(0 +/2)
=
n=
+j
Hs =
Ezs = k02 TzsM
(B.53)
(B.54)
Summary
This appendix describes the analytical investigation of two scattering problems in

which a circular scatterer is considered. It is known that the analytical form of
scattered fields from a three-dimensional scatterer of finite size and with localized
anomaly may not be available. Mathematical difficulty arises in finding a relevant
special function to analytically describe the scattered fields. Approximation methods
such as perturbation analysis [22] and numerical methods such as FDTD [251] are
usually applied instead.
341
342
Bibliography
[1] M. Shahawy A. Mirmiran and H. El Echary. Acoustic emission monitoring of
hybrid frp-concrete columns. ASCE J. Engng. Mech., 125(8):899905, 1999.
[2] J.D. Shane N.S. Berke T.O. Mason A. Schiessl, W.J. Weiss and S.P. Shah.
Assessing the moisture profile of drying concrete using impedance spectroscopy.
Concrete Sci. Engng., 2:10616, 2000.
[3] S.G. Millard A. Shaari and J.H. Bungey. Modelling the propagation of a radar
signal through concrete as a low-pass filter. NDT&E Intl., 37:23742, 2004.
[4] M.K. Lim A.G. Davis and C.G. Petersen. Rapid and economical evaluation of
concrete tunnel linings with impulse response and impulse radar non-destructive
methods. NDT&E Intl., 38:1816, 2005.
[5] I. L. Al-Qadi, O. A. Hazim, and S. M. Riad. Dielectric properties of portland cement concrete at low radio frequencies. ASCE J. Mater. Civil Engng.,
7(3):1928, 1995.
[6] I.L. Al-Qadi and S. Lahouar. Measuring layer thickness with gpr theory to
practice. Constr. Build. Mater., 19:76372, 2005.
[7] B.P. Anderson and J.J. Sacchini. A comparison of stepped continuous-wave
radar imaging algorithms. IEEE, pages 3038, 1994.
[8] K. Arunachlam, V. R. Melapudi, L. Upda, and S. S. Upda. Microwave ndt
of cement-based materials using far-field reflection coefficients. NDT&E Intl.,
39(7):58593, 2006.
343
[9] C. Au and O. Buyukozturk. Peel and shear fracture characterization of debonding in frp plated concrete affected by moisture. ASCE J. Compos. Constr.,
10(1):3547, 2006.
[10] J.F. Aurand. A tem-horn antenna with dielectric lens for fast impulse response.
In Proc. of the 3rd Intl. Conf. Ultra-Wideband, Short-Pulse Electromagnetics,
volume 95-2382C, pages 18, Albuquerque, NM, 1996. Sandia Natl. Lab.
[11] R. Zoughi J. Myers B. Akuthota, D. Hughes and A. Nanni. Near-field microwave
detection of disbond in carbon fiber reinforced polymer composites used for
strengthening cement-based structures and disbond repair verification. ASCE
J. Mater. Civil Engng., 16(6):5406, 2004.
[12] L. Krahenbuhl B. Sareni and A. Beroual. Complex effective permittivity of a
lossy composite material. J. Appl. Phys., 80(15):45605, 1996.
[13] D.D. Rolander M. Moore B.A. Graybeal, B.M. Phares and G. Washer. Visual
inspection of highway bridges. J. Nondestruct. Eval., 21(3):6783, 2003.
[14] D.L. Balageas and A. Deom. Fundamentals of infrared thermography. In NonIntrusive Measurement Techniques, pages 226. Von Karman Institute, 1994.
[15] C.A. Balanis. Advanced Engineering Electromagnetics. Wiley, New York, NY,
1989.
[16] C.A. Balanis. Antenna Theory: Analysis and Design. John Wiley, Hoboken,
NJ, 2005.
[17] C.L. Barnes and J.-F. Trottier. Ground-penetrating radar for network-level
concrete deck repair management. NDT&E Intl., 37:23742, 2000.
[18] A. Bayliss and E. Turkel. Radiation boundary conditions for wave-like equations. Comm. Pure. Appl. Math., 23:70725, 1980.
344
[19] J.M. Chicharrom A. Bayon and F. Salazar. Measurement of field-dependence

elastic modolus and magnetomechanical coupling factor by optical heterodyne
interferometry. J. Magnetism Magnetic Mater., 202:46572, 1999.
[20] G. Baz. Uber verlustmessungen an dipolflussigkeiten und festen technischen
isolierstoffen bei zentimeterwellen. Physik Zeits, 40.
[21] C.J.F. B
ottcher. Theory of Electric Polarisation. Elsevier, New York, NY, 1952.
[22] C.M. Bender and S.A. Orszag. Advanced Mathematical Methods for Scientists
and Engineers. McGraw-Hill, New York, NY, 1978.
[23] D.P. Bentz. CEMHYD3D: A Three-Dimensional Cement Hydration and Microstructure Development Modeling Package, volume NISTIR 7232. National
Institute of Standards and Technology, Gaithersburg, MD, 2005.
[24] J.P. Berenger. A perfectly matched layer for the absorption of electromagnetic
waves. J. Computational Phys., 114:185200, 1994.
[25] H. Berger. Neutron Radiography: Methods, Capabilities, and Applications. Elsevier, Amsterdam, The Netherlands, 1969.
[26] J.M. Berthelot and J.M. Robert. Modeling concrete damage by acoustic emission. J. Acoustic Emission, 6(1):4360, 1987.
[27] L. Binda, G. Lenzi, and A. Saisi. Nde of masonry structures: use of radar tests
for the characterization of stone masonries. NDT&E Intl., 31(6):4119, 1998.
[28] A.R. Blythe. Electrical Properties of Polymers. Cambridge University Press,
Cambridge, UK, 1979.
[29] K. J. Bois, A. D. Benally, and R. Zoughi. Microwave near-field reflection properties analysis of concrete for material content determination. IEEE T. Instrum.
Meas., 49(1):4955, 2000.
345
[30] K.J. Bois and R. Zoughi. A decision process implementation for microwave
near-field characterization of concrete constituent makeup. Subsurf. Sensing
Tech. Appl., 2(4):36376, 2001.
[31] M. Born and E. Wolf. Principles of Optics Electromagnetic Theory of Propagation, Interference and Diffraction of Light. Pergamon Press, Oxford, UK,
2002.
[32] D.E. Bray and R.K. Stanley. Nondestructive Evaluation: A Tool for Design,
Manufacturing, and Service. McGraw-Hill, New York, NY, 1989.
[33] J.R. Brown and H.R. Hamilton III. Nde of reinforced concrete strengthened
with fiber-reinforced polymer composites using infrared thermoraphy. In Proc.
InfraMation 2003, pages 14954, 2003.
[34] O. Buyukozturk and B. Hearing. Failure behavior of precracked concrete beams
retrofitted with frp. ASCE J. Compos. Constr., 2(3):13844, 1998.
[35] O. Buyukozturk and H. C. Rhim. Radar measurements of concrete for nondestructive evaluation of dams. In Proc. of SPIE, volume 2457, pages 16874.
SPIE, 1995.
[36] O. Buyukozturk and T.-Y. Yu. Understanding and assessment of debonding
failures in frp-concrete systems. In Proc. of Seventh Intl. Congress on Advances
in Civil Engng., Istanbul, Turkey, 2006. Yildiz Tech. Univ. Press.
[37] Ch. Maierhofer H.-J. Hohberger C. Colla, M. Krause and H. Sommer. Combination of ndt techniques for site investigation of non-ballasted railway tracks.
NDT&E Intl., 35:95105, 2002.
[38] D. McCann C. Colla and M.C. Forde. Radar testing of a masonry composite
structure with sand and water backfill. ASCE J. Bridge Engng., 6(4):26270,
2001.
346
[39] D. McCann C. Colla, P.C. Das and M.C. Forde. Sonic, electromagnetic and
impulse radar investigation of stone masonry bridges. NDT&E Intl., 30(4):249
54, 1997.
[40] J. Cariou C. Fauchard, X. Derobert and Ph. Cote. Gpr performance for thickness calibration on road test sites. NDT&E Intl., 36:6775, 2003.
[41] F. Caputo and G. Giudice. Photoelastic coating method approach to the study
of stress distribution in composite materials. Fibre Sci. Tech., 18:25564, 1983.
[42] R.R. Carlton and A.M. Adler. Principles of Radiography Imaging. Delmar
Publishers, Albany, NY, 1996.
[43] L. Cartz. Nondestructive Testing. ASM International, Materials Park, OH,
1995.
[44] V.L. Brown E. Cosenza J.F. Davalos J.J. Lesko A. Machida S.H. Rizkalla
C.E. Bakis1, L.C. Bank and T.C. Triantafillou. Fiber-reinforced polymer composites for construction state-of-the-art review. ASCE J. Compos. for Constr.,
6(2):7387, 2000.
[45] M.F. Petrou C.G. Papakonstantinou and K. Harries. Fatigue behavior of rc
beams strengthened with gfrp sheets. ASCE J. Compos. Constr., 5(4):24653,
2001.
[46] P. Chen and Y. Wang. Multicriterion compton backscatter imaging. In Proc.
Sci. Meas. Technol., volume 143, pages 35761. IEE, 1996.
[47] W.C. Chew. Waves and Fields in Inhomogeneous Media. IEEE Press, Piscataway, NJ, 1995.
[48] J.S. Dahele C.M. Alabaster and R.A. Badcock. Damage and discontinuity detection in planar composites using millimeter wave permittivity measurement.
Mater. Eval., February:12943, 2006.
347
[49] National Research Council (CNR). Guidelines for the Design and Construction of Externally Bonded FRP Systems for Strengthening Existing Structures,
volume DT 201. CNR, Rome, Italy, 2005.
[50] K.S. Cole and R.H. Cole. Dispersion and absorption in dielectrics i. alternating
current characteristics. J. Chem. Phys., 9:34151, 1941.
[51] K.S. Cole and R.H. Cole. Dispersion and absorption in dielectrics ii. direct
current characteristics. J. Chem. Phys., 10:98105, 1942.
[52] R.E. Collin. Foundations for Microwave Engineering. IEEE Press, New York,
NY, 2001.
[53] American Concrete Institute (ACI) 440 Committee. Guide for the Design
and Construction Externally Bonded FRP Systems for Strengthening Concrete
Structures. Emerging Technology Series. ACI, Framington Hills, MI, 2002.
[54] A.H. Compton and S.K. Allison. X-rays in Theory and Experiment. D. Van
Nostrand Company, New York, NY, 1935.
[55] Mitsubishi Chemical Corp.
Design Guide REPLARK System.
Master
Builders, Inc., Tokyo, Japan, 2000.

[56] J. Nadauld C.P. Pantelides and L. Cercone. Repair of cracked aluminum overhead sign structures with glass fiber reinforced polymer composites. ASCE J.
Mater. Civil Engng., 7(2):11826, 2003.
[57] A. Migliori C.R. Farrar, T.W. Darling and W.E. Baker. Microwave interferometers for non-contact vibration measurements in large structures. Mech. Sys.
Signal Process., 13(2):24351, 1999.
[58] D.E. Wahl C.V. Jakowatz Jr. and P.H. Eichel. Spotlight-Mode Synthetic Aperture Radar: A Signal Processing Approach.
Boston, MA, 1996.
348
Kluwer Academic Publishers,
[59] P. Palanichamy D.K. Bhattacharya C.V. Subramanian, M. Thavasimuthu and

B. Raj. Evaluation of bond integrity in sandwiched structures by dry couplant
ultrasonic technique. NDT&E Intl., 24:2931, 1991.
[60] K. Maser W. Weedon D. Huston, J. Qiong and C. Adam. Gima ground penetrating radar system for monitoring concrete bridge decks. ASCE J. Transport.
Engng., 126(3):25762, 2000.
[61] X. Decoopman D. Mercier, J. Lesage and D. Chicot. Eddy currents and hardness
testing for evaluation of steel decarburizing. NDT&E Intl., 39:65260, 2006.
[62] H.J. Dagher and M. Bragdon. Advanced frp-wood composites in bridge applications. In J.F. Orofino, editor, Structures 2001-A Structural Engineering
Odyssey, pages 52635, New York, NY, 2001. ASCE.
[63] D.W. Davidson and R.H. Cole. Dielectric relaxation in glycerine. J. Chem.
Phys., 18:1417, 1950.
[64] N.C. Davidson and S.B. Chase. Radar tomography of bridge decks. In Proc. of
SPIE, volume 3400, pages 2506. SPIE, 1998.
[65] P. Debye. Polar Molecules. Chemical Catalog Company, New York, NY, 1929.
[66] M.C. Decreton and M.S. Ramachandraiah. Nondestructive measurement of
complex permittivity for dielectric slabs. IEEE T. Microw. Theory, 52(1):1077
80, 1975.
[67] M.D. Desai and W.K. jenkins. Convolution backprojection image reconstruction for spotlight modesynthetic aperture radar. IEEE T. Image Processing,
1(4):50517, 1992.
[68] A.J. Devaney. A fast backprojection algorithm for ultrasound tomography.
IEEE T. Ultrasonic, Ferroelectrics, and Freq. Control, UFFC-34(3):33040,
1987.
349
[69] D.J. Gunton D.J. Daniels and H.F. Scott. Introduction to subsurface radar. In
IEE Proc. Pt. F, volume 135, pages 278320. IEE, 1988.
[70] Y. Camel T.M. Antonsen D.L. Gershon, J.P. Calame and R.M. Hutchen. Openended coaxial probe for high-temperature and broad-band dielectric measurements. IEEE T. Microw. Theory, 47(9):16408, 1999.
[71] F. Dunn and W.J. Fry. Precision calibration of ultrasonic fields by thermoelectric probes. T. IRS Professional Group Ultrasonic Engng., 5(1):5965, 1957.
[72] N. Dasgupta N. Somadevan E. Masad, V.K. Jandhyala and N. Shashidhar.
Characterization of air void distribution in asphalt mixes using x-ray computed
tomography. ASCE J. Mater. Civil Engng., 14(2):1229, 2002.
[73] M.-F. Wong V.Fouad-Hanna H. Baudrand E. Richalot, M. Bonilla and J. Wiart.
Electromagnetic propagation into reinforced-concrete walls.
IEEE T. Mi-
crowave Theory Tech., 48(3):35766, 2000.

[74] S.M. Gubanski E. Tuncer and B. Nettelbald. Dielectric relaxation in dielectric
mixtures: application of the finite element method and its comparison with
dielectric mixture formulas. J. Appl. Phys., 89(12):8092100, 2001.
[75] T.J. Buchanan E.H. Grant and H.F. Cook. Dielectric behavior of water at
microwave frequencies. J. Chem. Phys., 26(1):15661, 1957.
[76] D.S. Eisenberg and W. Kauzmann. The Structure and Properties of Water.
Oxford University Press, New York, NY, 1969.
[77] S.R. Elliot. A theory of a.c. conduction in chalcogenide glasses. Philos. Mag.,
36(6):1291304, 1977.
[78] B. Engquist and A. Majda. Absorbing boundary conditions for numerical simulation of waves. Proc. Natl. Acad. Sci., 74(5):17656, 1977.
[79] D.J. Epstein. Magnetic Loss Phenomena in Ferromagnetic Materials. MIT
Laboratory for Insulation Research, Cambridge, MA, 1962.
350
[80] A.D. Tommaso F. Bastianini and G. Pascale. Ultrasonic non-destructive assessment of bonding defects in composite structural strengthenings. Compos.
Struct., 53(4):4637, 2001.
[81] M. Faraday. Experimental Researches in Electricity. Dover, New York, NY,
1965.
[82] E.G. Farr and C.A. Frost. Impulse Propagation Measuremnets of the Dielectric
Properties of Water, Dry Sand, Moist Sand, and Concrete. Farr Research, Inc.,
Albuquerque, NM, 1997.
[83] R.F. Feldman. Density and porosity of hydrated portland cement. Cement
Technol., 3(1):514, 1972.
[84] R.F. Feldman and P.J. Sereda. A model for hydrated portland cement paste as
deduced from sorption length change and mechanical properties. Mater. Struct.,
1(6):50920, 1968.
[85] R.F. Feldman and P.J. Sereda. A new model for hydrated portland cement
paste and its practical implications. Can. Engng. J., 53(8/9):539, 1970.
[86] M.E. Fitzpatrick and A. Lodini. Analysis of Residual Stress by Diffraction Using
Neutron and Synchrotron Radiation. Taylor and Francis, London, UK, 2003.
[87] Intelligent Sensing for Innovative Structures (ISIS) Canada. Strengthening Reinforcing Concrete Structures with Externally-bonded Fiber Reinforced Polymers,
volume 4 of Design Manual. ISIS Canada, Winnipeg, MB, Canada, 2001.
[88] M. C. Forde, D. M. McCann, M. R. Clark, K. J. Broughton, P. J. Fenning, and
A. Brown. Radar measurement of bridge scour. NDT&E Intl., 32(8):48192,
1999.
[89] G. Frenning. Dielectric-response function determined by regular signular-point
analysis. Phys. Rev. B, 65:245117:18, 2002.
351
[90] H. Frohlich. Theory of Dielectrics Dielectric Constant and Dielectroc Loss.

Oxford University Press, Oxford, UK, 1949.
[91] R.M. Fuoss and J.G. Kirkwood. Electrical properties of soilds. viii. dipole moments in polyvinyl chloride-diphenyl systems. J. Am. Chem. Soc., 63:38594,
1941.
[92] K.N. Lee G. Chen and S.N. Tewari. The concept and development of emission
spectroscopy for the non-destructive evaluation (nde) of the failure of thermal
barrier coatings. NDT&E Intl., 40:26570, 2007.
[93] M. Buttram F. Zutarven-D. Brown G. Loubriel, J. Aurand and W. Helgeson.
Final Report of LDRD Project: Electromagnetic Impulse Radar for Detection
of Underground Structures. Sandia National Laboratories, Alburquerque, NM,
1998.
[94] C. Nockenmann G.-R. Tillack and C. Bellon. X-ray modeling for industrial
applications. NDT&E Intl., 33:4818, 2000.
[95] A. Giannopoulos. GprMax2D/3D Users Manual. (http://www.gprmax.org/),
Edinburgh, UK, 2nd edition, 2005.
[96] E.H. Grant and R. Shack. Complex permittivity measurements at 8.6 mm
wavelength over the temperature range 1-60 . Brit. J. Appl. Phys., 18:180714,
1967.
[97] R. Gregory. Composite repair inspection by laser shearography. Insight NonDestructive Testing and Condition Monitoring, 45(3):1835, 2003.
[98] M. Chabaat H. Hadjab and J.-Fr. Thimus. Use of scanning electron microscope
and non-local isotropic damage model to investigate fracture process zone in
notched concrete beams. Exp. Mech., 47(4):47384, 2007.
[99] D. Schaurich K. Tulla H. Kaariainen, M. Rudolph and H. Wiggenhauser.
Moisture measurements in building materials with microwaves. NDT&E Intl.,
34:38994, 2001.
352
[100] H. Jager H. Phillipp, T. Neger and J. woisetschlager. Optical tomography of

phase objects by holographic interferometry. Measurement, 10(4):17081, 1992.
[101] P. Lee-Sullivan H.A. Jama, E.M.A. Hussein.
Detection of debonding in
composite-aluminum joints using gamma-ray compton scattering.
NDT&E
Intl., 31(2):99103, 1998.

[102] D.J. Hagemaier. Fundamentals of Eddy Current Testing. ASNT, Columbus,
OH, 1990.
[103] R. Halmshaw. Non-destructive Testing. Edward Arnold, London, UK, 1987.
[104] R.W. Hamming. Numerical Methods for Scientists and Engineers. Dover, New
York, NY, 2nd edition, 1973.
[105] W.D. Harkins and G. Jura. Surface of solids. x. extension of the attractive
energy of a solid into an adjacent liquid or film, the decrease of energy with
distance, and the thickness of films. J. Am. Chem. Soc., 66:91927, 1944.
[106] J.B. Hasted. Aqueous Dielectrics. Chapman&Hall, London, UK, 1973.
[107] J.B. Hasted and M.A. Shah. Microwave absorption by water in building materials. Brit. J. Appl. Phys., 15:82536, 1964.
[108] S. Havriliak and S.J. Havrilliak. Comparison of dieletric theories that explicitly
include viscoelastic parameters. J. Polym. Sci. Pt. B Pol. Phys., 33:224552,
1995.
[109] G. Hegde and A. Asundi. Performance analysis of all-fiber polarimetric strain
sensor for composites structural health monitoring. NDT&E Intl., 39:3207,
2006.
[110] A.A. Abdel-Shafy I.M. Elewa H.I. Shafeek, E.S. Gadelmawla. Assessment of
welding defects for gas pipeline radiographs using computer vision. NDT&E
Intl., 37:2919, 2004.
353
[111] R.L. Higdon. Numerical absorbing boundary conditions for the wave equation.
Math. Computation, 49:6590, 1987.
[112] R.W. Higgs. Sonic signature analysis for arc furnance diagnostics and control.
In IEEE Ultrasonic Symposium, pages 65363, 1974.
[113] A.R. Von Hippel. Dielectric Materials and Applications; Papers by Twenty-Two
Contributors. Wiley, New York, NY, 1954.
[114] W.M.W.A. Majid H.M.A. Al-Matterneh, D.K. Ghodgaonkar. Determination of
compressive strength of concrete using free-space reflection measurements in the
frequency of 8-12.5 ghz. In IEEE Proc. APMC2001, volume 2, pages 67982,
2001.
[115] R. Holland. Threde: A free-field emp coupling and scattering code. IEEE T.
Nuclear Sci., NS-24(6):241621, 1977.
[116] M.T. Hutchings. Introduction to the Characterization of Residual Stress by
Neutron Diffraction. Taylor and Francis, Boca Raton, FL, 2005.
[117] R.H. Haddad I.L. Al-Qadi and S.M. Riad. Detection of chlorides in concrete
using low radio frequencies. ASCE J. Mater. Civil Engng., 9(1):2934, 1997.
[118] H. Inada. Backscatterred short pulse response of surface waves from dielectric
spheres. Appl. Optics, 13(8):192833, 1974.
[119] W.S. Chin I.S. Seo and D.G. Lee. Characterization of electromagnetic properties
of polymeric composite materials with free space method. Compos. Struct.,
66:53342, 2004.
[120] O. Ishai. The time-dependent deformation behavior of cement paste, mortar
and concrete. Proc. Conf. The Structure of Concrete and Its Behaviour Under
Load, pages 34564, 1965.
[121] A. Ishimaru. Electromagnetic Wave Propagation, Radiation, and Scattering.
Prentice Hall, Upper Saddle River, NJ, 1991.
354
[122] J.H. Lienhard IV and J.H. Lienhard V. A Heat Transfer Textbook. Philogiston
Press, Cambridge, MA, 3rd edition, 2003.
[123] B.F. Riddle R.T. Johnk P. Kabos C.L. Holloway R.G. Geyer J. Baker-Jarvis,
M.D. Janezic and C.A. Grosvenor. Measuring the Permittivity and Permeability
of Lossy Materials: Solids, Liquids, Metals, Building Materials, and NegativeIndex Materials, volume Technical Note 1536. National Institute of Standards
and Technology, Boulder, CO, 2004.
[124] Peter S. Ray J.J. Stephens and T.W. Kitterman. Far-field impulse response verification of selected high-frequency optics backscattering analogs. Appl. Optics,
14(9):216976, 1975.
[125] A.K. Jonscher. A new model of dielectric loss in polymers. Colloid Polym. Sci.,
253(3):23150, 1975.
[126] A.K. Jonscher. Dielectric Relaxation in Solids. Chelsea Dielectrics Press, London, UK, 1983.
[127] M. Le R.A. Yarbrough G. Gogoshin J.R. Claycomb, A. Brazdeikis and J.H.
Miller. Nondestructive testing of pem fuel cells. IEEE T. Appl. Supercond.,
13(2):2114, 2003.
[128] T. Shoji M. Sato K. Ogawa, D. Minkov and H. Hashimoto. Nde of degradation
of thermal barrier coating by means of impedance spectroscopy. NDT&E Intl.,
32:17785, 1999.
[129] A.C. Kak and M. Slaney. Principles of Computerized Tomographic Imaging.
Society for Industrial and Applied Mathematics, Philadelphia, PA, 1999.
[130] W. Kauzmann. Dielectric relaxation as a chemical rate process. Rev. Mod.
Phys., 14.
[131] M. Kildemo. New spark-test device for material characterization. Nuclear Instruments Mater. Phys. Res. A, 530:596606, 2004.
355
[132] T.N. Salonikios K.K. Antoniades and A.J. Kappos. Tests on seismically damaged reinforced concrete walls repaired and strengthened using fiber-reinforced
polymer. ASCE J. Compos. Constr., 9(3):23646, 2005.
[133] T.J. Wipf K.L. Rens and F. Wayne. Review of nondestructive evaluation
techniques of civil infrastructure. ASCE J. Perform. Constructed Facilities,
11(4):15260, 1997.
[134] G. Klysz, J.-P. Balayssac, and S. Laurens. Spectral analysis of radar surface
waves for non-destructive evaluation of cover concrete. NDT&E Intl., 37(3):221
7, 2004.
[135] R.J. Knight and A. Nur. The dielectric constant of sandstone, 60 khz to 4 mhz.
Geophys., 52(5):64454, 1987.
[136] A.C. Knoll. Ultrasonic holography techniques for localizing and imaging solid
objects. IEEE T. Robotics Automation, 7(4):44967, 1991.
[137] E. F. Knott, J. F. Shaeffer, and M. T. Tuley. Radar Cross Section. SciTech
Publishing, Rayleigh, NC, 2004.
[138] J. A. Kong. Electromagnetic Wave Theory. EMW Publishing, Cambridge, MA,
2000.
[139] O.B. Lile K.R. Backe and S.K. Lyomov. Characterizing curing cement slurries
by electrical conductivity. SPE Drill. Completion, Dec.:2017, 2001.
[140] H.A. Kramers. In Atti. Congr. Internaz. Fisici., volume 53, pages 539, Como,
Italy, 1927.
[141] A. Kraszewski. Microwave aquametry: An effective tool for nondestructive
moisture sensing. Subsurf. Sensing Tech. Appl., 2(4):34762, 2001.
[142] J.D. Kraus and R.J. Marthefka. Antennas for All Applications. McGraw-Hill,
Boston, MA, 3rd edition, 2002.
356
[143] K.S. Kunz and R.J. Luebbers. The Finite Difference Time Domain Method for
Electromagnetics. Boca Rotan, Boston, MA, 1993.
[144] C. Tiraboschi L. Binda, A. Saisi. Application of sonic tests to the diagnosis of
damaged and repaired structures. NDT&E Intl., 34:12338, 2001.
[145] L. Lam and J.G. Teng. Strength of rc cantilever slabs bonded with gfrp strips.
ASCE J. Compos. Constr., 5(4):2217, 2001.
[146] J.B. Lambert and E.P. Mazzola. Nuclear Magnetic Resonance Spectroscopy: An
Introduction to Principles, Applications, and Experimental Methods. Pearson,
Upper Saddle River, NJ, 2004.
[147] L.D. Landau and E.M. Lifshitz. Electrodynamics of Continuous Media. Pergamon, Oxford, UK, 1984.
[148] D.L. Kantro L.E. Copeland and G.J. Verbeck. Chemistry of hydration of portland cement. In Proc. 4th Intl. Symp. Chem. Cement, volume 1, pages 42968,
Washington, D.C., 1960.
[149] J. Li and C. Liu. Noncontact detection of air voids under glass epoxy jackets
using a microwave system. Subsurf. Sensing Tech. Appl., 2(4):41123, 2001.
[150] K. Lichtenecker. Die dielektrizitatskonstante naturlicher und kunstlicher mischkorper. Phys. Z., 27:115158, 1926.
[151] J.D. Livingston. Electronic Properties of Engineering Materials. Wiley, New
York, NY, 1999.
[152] Fyfe Co. LLC. Design Manual for the Tyfo Fibrwrap System. Fyfe Co. LLC,
San Diego, CA, 1999.
[153] R. Van Loon and R. Finsy. The precise microwave permittivity measurmenet
of liquids using a multipoint technique and curve-fitting procedure. J. Phys. D:
Appl. Phys., 8:123243, 1975.
357
[154] G.P. De Loor. The effect of moisture on the dielectric constant of hardened
portland cement paste. Appl. Sci. Res. B, 9:297308, 1961.
[155] H. Looyenga. Dielectric constants of mixtures. Physica, 321:4016, 1965.
[156] A.E.H. Love. Some Problems of Geodynamics: Being an essay to which the
Adams prize in the University of Cambridge was adjudged in 1911. Cambridge
University Press, Cambridge, UK, 1911.
[157] P. Chou M. Chang and H. Lee. Tomographic microwave imaging for nondestructive evaluation and object recognition of civil structures and materials. In
Proc. ASILOMAR-29, pages 10616, 1996.
[158] M. Gordon M. Clark and M. Forde. Issues over high-speed non-invasive monitoring of railway trackbed. NDT&E Intl., 37:1319, 2004.
[159] D. Fratta M. Daigle and L.B. Wang. Ultrasonic and x-ray tomographic imaging
of highly contrasting inclusions in concrete specimens. In Geo-Frontiers Site
Characterization and Modeling. ASCE, 2005.
[160] E.D. Sotelino M.-H. Teng and W.-F. Chen. Performance evaluation of reinforced
concrete bridge columns wrapped with fiber reinforced polymers. ASCE J.
Compos. Constr., 7(2):8392, 2003.
[161] H. Ogi M. Hirao and H. Yasui. Contactless measurement of bolt axial stress
using a shear-wave electromagnetic acoustic transducer. NDT&E Intl., 34:179
83, 2001.
[162] G. Bourse J.W. Xu M. Lematre, Y. Benemehrez and M. Ourak. Determination
of elastic parameters in isotropic plates by using acoustic microscopy measurements and an optimization method. NDT&E Intl., 35:493502, 2002.
[163] B. Graybeal D. Rolander M. Moore, B. Pharnes and G. Washer. Reliability
of Visual Inspection for Highway Bridges: Final Report, volume RD-01-020.
FHWA, McLean, VA, 2001.
358
[164] A. Delahaza C.G. Santos M. Moore B. Graybeal M. Scott, A. Rezaizadeh and

G. Washer. A comparison of nondestructive evaluation methods for bridge deck
assessment. NDT&E Intl., 36:24555, 2003.
[165] A.L. Salazar M.A. Omar, B.H. Gharaibeth and K. Saito. Infrared thermography (irt) and ultraviolet fluorescence (uvf) for the nondetsructive evaluation of
ballast tanks coated surfaces. NDT&E Intl., 40:6270, 2007.
[166] N.J. Carino M.A. Starnes and E.A. Kausel. Preliminary thermography studies for quality control of concrete structures strengthened with fiber-reinforced
polymer composites. ASCE J. Mater. Civil Engng., 15(3):26673, 2003.
[167] C. Maierhofer, A. Brink, M. Rollig, and H. Wiggenhauser. Detection of shallow
voids in concrete structrues with impulse thermography and radar. NDT&E
Intl., 36(4):25763, 2003.
[168] C. Maierhofer, M. Krause, and H. Wiggenhauser. Non-destructive investigation
of slucies using radar and ultrasonic impulse echo. NDT&E Intl., 31(6):4217,
1998.
[169] C. Maierhofer and S. Leipold.
Radar investigation of masonry structures.
NDT&E Intl., 34(2):13947, 2001.

[170] J. Makar and R. Desnoyers. Magnetic field techniques for the inspection of steel
under concrete cover. NDT&E Intl., 34:44556, 2001.
[171] L.E. Malvern.
Introduction To the Mechanics of A Continuous Medium.
Prentice-Hall, Englewood Cliffs, NJ, 1969.

[172] D.T. Markus and T.S. Lee. A method of direct corona imaging on a dielectric
transparency. IEEE T. Industry Appl., 32(4):8326, 1996.
[173] K.R. Maser.
Condition assessment of transportation infrastructure using
ground-penetrating radar. ASCE J. Infrastruct. Sys., 2(2):94101, 1996.

359
[174] J.C. Maxwell. A Treatise on Electricity and Magnetism, Vol. I, II. Dover
Publication, New York, NY, 1954.
[175] D.M. McCann and M.C. Forde. Review of ndt methods in the assessment of
concrete and masonry structures. NDT&E Intl., 34:7184, 2001.
[176] W.J. McCarter. The Application of Impedance Spectroscopy to Cementitious
Systems. Number BTR029. Solartron Limited, Hampshire, UK, 1999.
[177] J. McCorkle and M. Rofheart. An order n2 logn backprojection algorithm for
focusing wide-angle wide-bandwidth arbitrary-motion synthetic aperture radar.
In Proc. of SPIE, volume 2747, pages 2536, Orlando, FL, 1996. SPIE.
[178] P.I. Tiemsin M.D. Deshpande, C.J. Reddy and R. Cravey. A new approach to
estimate complex permittivity of dielectric materials at microwave frequencies
using waveguide measurements. IEEE T. Microw. Theory, 45(3):35965, 1997.
[179] P.K. Mehta and P.J.M. Monteiro. Concrete Structure, Properties, and Materials. Prentice-Hall, Englewood Cliffs, NJ, 1993.
[180] D.L. Mensa. High Resolution Radar Cross-Section Imaging. Artech House,
Boston, MA, 1999.
[181] S. G. Millard, J. H. Bungey, C. Thomas, M. N. Soutos, M. R. Shaw, and
A. Patterson. Assessing bridge pier scour by radar. NDT&E Intl., 31(4):2518,
1998.
[182] A. Mirmiran and S. Philip. Comparison of acoustic emission activity in steelreinforced and frp-reinforced concrete beams. Constr. Build. Mater., 14:299
310, 2000.
[183] A. Mirmiran and Y. Wei. Damage assessment of frp-encased concrete using
ultrasonic pulse velocity. ASCE J. Engng. Mech., 127(2):12635, 2001.
[184] T. Moreno and O.C. Lundstrom. Microwave power measurement. In Proc. of
IRE, volume 35, pages 5148, 1947.
360
[185] Y.J. Kim M.Q. Feng, F. De Flaviis.
Microwaves for damage detection of
fiber reinforced polymer-wrapped concrete structures. ASCE J. Engng. Mech.,

128(2):17283, 2002.
[186] G. Mur. Absorbing boundary conditions for the finite-difference approximation
ofthe time-domain electromagnetic field equations. IEEE T. Electromagnetic
Compatibility, 23:37782, 1981.
[187] R. Murayam and K. Misumi. Development of a non-contact stress measurement
system during tensile testing using the lectromagnetic acoustic transducer for
a lamb wave. NDT&E Intl., 39:299303, 2006.
acek. Microwave Measurements of Complex Permittivity by
[188] J. Musil and F. Z
Free Space Methods and Their Applications. Elsevier, Amsterdam, The Netherlands, 1986.
[189] M. Fatemi M.W. Urban, G.T. Silva and J.F. Greenleaf. Multifrequency vibroacoustography. IEEE T. Medical Imaging, 25(10):128495, 2006.
[190] R. Zoughi N. Qaddoumi and G.W. Carriveau. Microwave detection and depth
determination of disbonds in low-permittivity and low-loss sandwich composites. Res. Nondestr. Eval., 8:5163, 1996.
[191] J. Nadakuduti, G. Chen, and R. Zoughi. Semiempirical electromagnetic modeling of crack detection and sizing in cement-based materials using near-field
microwave methods. IEEE T. Instrum. Meas., 55(2):58897, 2006.
[192] M.S.A. Hardy N.C. Davidson and M.C. Forde. Bridge scour assessment by
impulse radar. In IEE Colloquium on Radar and Microwave Tech. for NDE,
pages 8/18. IEE, 1995.
[193] National Cooperative Highway Research Program (NCHRP). Bonded Repair
and Retrofit of Concrete Structures Using FRP Composites, volume 514. Transportation Research Board (TRB), Washington, D.C., 2004.
361
[194] S. Nilsson and L.-E. Andersson. Application of fast backprojection technqiues

for some inverse problems of synthetic aperture radar. In Proc. of SPIE, volume
3370, pages 6272, Orlando, FL, 1998. SPIE.
[195] E. Nyfors. Industrial microwave sensors a review. Subsurf. Sensing Tech.
Appl., 1(1):2343, 2000.
[196] O. Gunes O. Buyukozturk and E. Karaca. Progress in understanding debonding problems in reinforced concrete and steel members strengthened using frp
composites. Constr. Build. Mater., 18(1):919, 2004.
[197] S. Alampalli O. Hag-Elsafi and J. Kunin. In-service evaluation of a reinforced
concrete t-beam bridge frp strengthening system. Compos. Struct., 64:17988,
2004.
[198] U.S. Army Corps of Engineers. Engineering and Design of Composite Materials
for Civil Engineering Structures, volume ETL 1110-2-548. Department of the
Army, Washington, D.C., 1987.
[199] K. ONeil. Radar detection of near-surface buried metallic reflectors in wet soil.
In IEEE Intl. Geosci. Remote Sensing Symp., pages 128890, 1997.
[200] R.D. Owen. Portable linear accelerators for x-ray and electron-beam applications in civil engineering. NDT&E Intl., 31(6):4019, 1998.
[201] M.N. Ozisik and H.R.B. Orlande. Inverse Heat Transfer: Fundamentals and
Applications. Taylor & Francis, New York, NY, 2000.
[202] N. Swamy P. Mukhopadhyaya and C. Lynsdale. Optimizing structural respone
of beams strengthened with gfrp plates. ASCE J. Compos. Constr., 2(2):8795,
2000.
[203] S.X. Pan and A.C. Kak. A computational study of reconstruction algorithms
for diffraction tomography: interpolation versus filtered backpropagation. IEEE
T. Acoust. Speech Signal Process., ASSP-31(5):126275, 1983.
362
[204] C.J. Cronin P.B. Leeds-Harrison, R.K. Fry and J.E. Gregory. A technique for
the non-destructive monitoring of subsurface drains. J. Agric. Engng. Res.,
28:47984, 1983.
[205] M. Pieraccini, G. Luzi, D. Mecatti, M. Fratini, L. Noferini, and L. Carssimi.
Remote sensing of building structural displacements using a microwave interferometer with image capability. NDT&E Intl., 37(9):54550, 2004.
[206] T.C. Powers. Structure and physical properties of hardened portland cement
paste. J. Am. Ceram. Soc., 41(1):16, 1958.
[207] T.C. Powers. Mechanisms of shrinkage and reversible creep of hardened cement
paste. In Proc. Intl. Conf. The Structure of Concrete and Its Behaviour under
Load, pages 31944, London, UK, 1965. Cement Concrete Assoc.
[208] T.C. Powers. The Properties of Fresh Concrete. John Wiley&Sons, New York,
NY, 1968.
[209] T.C. Powers and T.L. Brownyard. Studies of the physical properties of hardened
portland cement paste. J. Am. Concrete Inst., 43:10132, 1947.
[210] J.G. Parzych P.V. Mtenga and R. Limerick. Quality assurance of frp retrofit
using infrared thermography. New York, NY, 2001. ASCE.
[211] N. Qaddoumi, E. Ranu, J. D. McColskey, R. Mirshahi, and R. Zoughi. Microwave detection of stress-induced fatigue cracks in steel and potential for
crack opening determination. Res. Nondestr. Eval., 12:83103, 2000.
[212] N. Stockhausen R. Badmann and M.J. Setzer. The statistical thickness and the
chemical potential of adsorbed water films. J. Colloid Interf. Sci., 82(2):53442,
1981.
[213] W.D. Dover R. Clark and L.J. Bond. The effect of crack closure on the reliability
of ndt predictions of crack size. NDT Intl., 20(5):26975, 1987.
363
[214] K. Friedrichs R. Courant and H. Lewy. On the partial difference equations of

mathemttical physics. IBM Journal, March:21534, 1967.
[215] J. Lai R. Zoughi and K. Munoz. A brief review of microwave testing of stratified composite structures: A comparison between plane wave and near field
approaches. Mater. Eval., February:1717, 2002.
[216] A. Galgaro R.G. Francese and A. Grespan. Venice channel side-wall assessment
with gpr technique a case study. NDT&E Intl., 56:3140, 2004.
[217] H. C. Rhim and O. Buyukozturk. Wideband microwave imaging of concrete for
nondestructive testing. ASCE J. Struct. Engng., 126(12):14517, 2000.
[218] M.A. Richter and J.W. Wagner. Experimental evaluation of hinge phenomenon
in notched three point bend bars using laser speckle metrology. Insight NonDestructive Testing and Condition Monitoring, 30(6):81926, 1998.
[219] M.P. Robert. High Energy Density Microwaves. American Institute of Physics,
New York, NY, 1998.
[220] R.C. Miller R.W. Rampolla and C.P. Smyth. Microwave absorption and molecular structure in liquids. xxv. measurements of dielectric constant and loss at
3.1 mm wavelength by an interferometric method. J. Chem. Phys., 30.
[221] P.H. Emmett S. Brunauer and E. Teller. Adsorption of gases in multimolecular
layers. J. Am. Chem. Soc., 60:30919, 1938.
[222] T. Uchimoto T. Takagi S. Konoplyuk, T. Abe and M. Kurosawa. Characterization of ductile cast iron by eddy current method. NDT&E Intl., 38:6236,
2005.
[223] J.F. Young S. Mindess and D. Darwin. Concrete. Prentice-Hall, Upper Saddle
River, NJ, 2003.
364
[224] E. Gallaher K.E. Kurtis S. Peer, J.T. Case and R. Zoughi. Microwave reflection
and dielectric properties of mortar subjected to compression force and cyclically exposed to water and sodium chloride solution. IEEE T. Instrum. Meas.,
52(1):1118, 2003.
[225] M. Johnson R. Sexsmith S. Ross, A. Boyd and N. Banthia. Potential retrofit
methods for concrete channel beam bridges with glass fiber reinforced polymer.
ASCE J. Bridge Engng., 9(1):6674, 2004.
[226] D. De Zutter S. Van Damme, A. Franchois and L. Taerwe. Nondestructive
determination of the steel fiber content in concrete slabs with an open-ended
coaxial probe. IEEE T. Geosci. Remote Sensing, 42(11):25112521, 2004.
[227] H. Saadatmanesh and M.R. Ehasni. Fiber composite plates can strengthen
beams. ACI Concrete Intl., pages 6571, 1990.
[228] H. Saadatmanesh and M.R. Ehsani. Application of fiber-composites in civil
engineering. In J.F. Orofino, editor, Structural Materials, pages 52635, New
York, NY, 1989. ASCE.
[229] T. Saarenketo and T. Scullion. Road evaluation with ground penetrating radar.
J. App. Geophys., 43, pages =.
[230] J.S. Sandhu and R.E. Thomas. Acoustographic nondestructive evaluation. In
IEEE Ultrasonic Symposium, pages 10536, 1988.
[231] M. Sansalone and W.B. Streett. Impact-echo: Non-destructive evaluation of
concrete and masonry. Bullbrier Press, Ithaca, NY, 1997.
[232] A.M. Abdel-Ghaffer M. Higazy R.O. Claus S.F. Masri, M.S. Agbabian and M.J.
de Vries. Experimental study of embedded fiber-optic strain gauges in concrete
structures. ASCE J. Engng. Mech., 120(8):1696717, 1994.
[233] M. R. Shaw, S. G. Millard, T. C. K. Molyneaux, M. J. Taylor, and J. H. Bungey.
Location of steel reinforcement in concrete using ground penetrating radar and
neutral networks. NDT&E Intl., 38(3):20312, 2005.
365
[234] M. Shinozuka, R. Ghanem, B. Houshmand, and B. Mansouri. Damage detection

in urban areas by sar imagery. ASCE J. Engng. Mech., 126(7):76977, 2000.
[235] K.E. Slevogt. Dispersion und absorption elektrischer wellen in alckoholen und
wabrigen losungen. Ann. Phys. (Leipzig), 36.
[236] C.P. Smyth. Dielectric Behavior and Structur Dielectric Constant and Loss,
Dipole Moment and Molecular Structure. McGraw-Hill, New York, NY, 1955.
[237] U.C. Hasar S.N. Kharkovsky, F. Akay and C. C.D. Atis. Measurement and
monitoring of microwave reflection and transmission properties of cement-based
specimens. IEEE T. Instrum. Measure., 51(6):12108, 2002.
[238] I. Soroka. Portland Cement Paste and Concrete. MacMillan Press, London,
UK, 1979.
[239] M. Soumekh. A system model and inversion for synthetic aperture radar imaging. IEEE T. on Imaging Processing, 1(1):6476, 1992.
[240] M. Soumekh. Synthetic aperture radar signal processing with MATLAB algorithms. Wiley, New York, NY, 1999.
[241] W.C. Stone. Non-Line-of-Sight (NLS) Construction Metrology, volume NISTIR
5825. National Institute of Standards and Technology, Gaithersburg, MD, 1996.
[242] Lord Rayleigh (J.W. Strutt). On the influence of obstacles arranged in rectangular order upon the properties of a medium. Philos. Mag., 32:481502, 1892.
[243] W.L. Stutzman and G.A. Thiele. Antenna Theory and Design. John Wiley &
Sons, New York, NY, 2nd edition, 1998.
[244] M. Sucher and J. Fox. Handbook of Microwave Measurements. Polytechnic
Press, New York, NY, 1963.
[245] A.P. Sutton. Interfaces in Crystalline Materials. Oxford, Clarendon, UK, 1996.
366
[246] T. Masliwec T. Chung, C.R. Carter and D.G. Manning. Impulse radar evaluation of concrete, asphalt and waterproofing membrane. IEEE T. Aero. Electronic Sys., 30(2):40415, 1994.
[247] K.I. Maslov T. Kundu, M. Ehsani and D. Guo. C-scan and l-scan generated
images of the concrete/gfrp composite interface. NDT&E Intl., 32:619, 1999.
[248] L. Achraii-A. Mamouni T. Larsi, D. Glay and Y. Leroy. Microwave methods and
systems for nondestructive control. Subsurf. Sensing Tech. Appl., 1(1):14160,
2000.
[249] H. Toyoda T. Nagaishi, H. Kugai and H. Itozaki. Ndt of high speed fine particles
by high tc squid. IEEE T. Appl. Supercond., 7(2):28869, 1997.
[250] O. Manfroni T. Stratford, G. Pascale and B. Bonfiglioli. Shear strengthening
masonry panels with sheet glass-fiber reinforced polymer. ASCE J. Compos.
Constr., 8(5):43443, 2004.
[251] A. Taflove.
Computational Electrodynamics The Finite-Difference Time-
Domain Method. Artech House, Norwood, MA, 1995.

[252] A. Taflove and M.E. Brodwin. Numerical solution of steady-state electromagnetic scattering problems using the time-dependent maxwells equations. IEEE
T. Microwave Theory Tech., MTI-23(8):62330, 1975.
[253] Master Builders Technologies. MBrace Composite Strengthening System Engineering Design Guidelines. Master Builders, Inc., Cleveland, OH, 1998.
[254] P. Theophelakes. (private communication). Division of Advanced Space Systems
and Concepts, MIT Lincoln Laboratory, 2007.
[255] S.P. Timonshenko and J.N. Goodier. Theory of Elasticity. McGraw-Hill, New
York, NY, 1970.
[256] W.R. Tinga.
Mixture laws and microwave-material interactions.
(Progress In Electromagnetics Research), 6:140, 1992.

367
PIER
[257] D.J. Titman. Applications of thermography in non-destructive testing of structures. NDT&E Intl., 34:14954, 2001.
[258] H. Toutanji and P. Balaguru. Durability characteristics of concrete columns
wrapped with frp tow sheets. ASCE J. Mater. Civil Engng., 10(1):527, 2000.
[259] E.T. Selig T.R. Sussmann and J.P. Hyslip. Railway track condition indicators
from ground penetrating radar. NDT&E Intl., 36:15767, 2003.
[260] T.C. Triantafillou and N. Plevris. Strengthening of rc beams with epoxy-bonded
fibre-composite materials. Mater. Struct., 25(4):20111, 1992.
[261] L. Tsang, J. A. Kong, and K.-H. Ding. Scattering of Electromagnetic Waves
Theories and Applications. John Wiley & Sons, New York, 2000.
[262] H.V.S. GangaRao P. Klinkhachorn U.B. Halabe, H. Al-Qennah and E.S.
Sazonov. Nondestructive evaluation of frp composite bridge components using infrared thermography. Rev. Quant. NDE, 21:13039, 2002.
[263] C. Korhonen V. Janoo and M. Hovan. Measurement of water content in portland
cement concrete. ASCE J. Transport. Engng., 125(3):2459, 1999.
[264] B.A. Shaw V. Moorthy and P. Hopkins. Magnetic barkhausen emission technique for detecting the overstressing during bending fatigue in case-carburised
en36 steel. IEEE T. Industry Appl., 38:15966, 2005.
[265] A. VanderLugt. Optical Signal Processing. John Wiley, Hoboken, NJ, 2005.
[266] S.R. Venkatesh and M.A. Dahleh. On system identification of complex systems
from finite data. IEEE T. Automat. Contr., 46(2):23557, 2001.
[267] I.A. Viktorov. Rayleigh and Lamb Waves: Physical Theory and Applications.
Plenum Press, New York, NY, 1967.
[268] O. Wiener. Die theorie des mischkorpers fur das feld der staionaren stromung.
Abh. Math. Phys. Kl. Sachs Akad. Wiss. Leipzig, 32:509604, 1912.
368
[269] G. Williams and D.C. Watts. Non-symmetrical dielectric relaxation behavior

arising from a simple empirical decay function. Trans. Faraday Soc., 66:805,
1970.
[270] J.G. Wilson and H.W. Whittington. Variations in the electrical properties of
concrete with change in frequency. IEE Proc. Pt. A, 137(5):24654, 1990.
[271] F.H. Wittmann and F. Schlude. Microwave absorption of hardened cement
paste. Cement Concrete Res., 5:6371, 1975.
[272] W.S. Jardetzky W.M. Ewing and F. Press. Elastic Waves in Layered Media.
McGraw-Hill, New York, NY, 1957.
[273] A. Mamoumi X. Derobert and Y. Leroy. Feasibility of nondestructive testing of concrete by microwave radiometry. Microwave Optical Technol. Letters,
14(1):1720, 1997.
[274] C. Aubagnae X. Derobert and O. Abraham. Comparison of ndt techniques on
a post-tensioned beam before its autopsy. NDT&E Intl., 35:5418, 2002.
[275] T. Jackson X. Zhang and E. Lafond. Stiffness properties and stiffness orientation
distribution for various paper grades by non-contact laser ultrasonics. NDT&E
Intl., 39:594601, 2006.
[276] H. Takashima Y. Hatsukade, N. Kasai and A. Ishiyama. Non-contact squidndt method using a ferrite core for carbon-fiber composites. Supercond. Sci.
Technol., 15:172832, 2002.
[277] K.S. Yee.
Numerical solution of initial boundary value problems involing
maxwells equations in isotropic media. IEEE T. Antennas and Propagation,

AP-14(3):3027, 1966.
[278] A. F. Yegulalp. Fast backprojection algorithm for synthetic aperture radar. In
Proc. of IEEE Radar Conf., pages 605. IEEE, 1999.
369
[279] F. De Flaviis Y.J. Kim, L. Jofre and M.Q. Feng. Microwave subsurface imaging
technology for damage detection. ASCE J. Engng. Mech., 130(7):85866, 2004.
[280] T.-Y. Yu and O. Buyukozturk. A far-field radar ndt technique for detecting
debonding in gfrp-retrofitted concrete structures. NDT&E Intl., 4:1024, 2008.
[281] T.H. Loutas Y.Z. Pappas, A. Kontsos and V. Kostopoulos. On the characterization of continuous fibres fracture by quantifying acoustic emission and
acousto-ultrasonic waveforms. NDT&E Intl., 37:389401, 2004.
[282] R. Zorn. Applicability of distribution functions for the havriliak-negami spectral
function. J. Polym. Sci. Pt. B Pol. Phys., 37:104344, 1999.
[283] R. Zoughi and S. Bakhtiari. Microwave nondestructive detection and evaluation
of disbonding and delamination in layered-dielectric-slabs. IEEE T. Instrum.
Measure., 39(6):105963, 1990.
[284] X.Q. Shi Z.W. Zhang and K.W. Wong. Measurements of deformations and
strains of a si-epoxy-frp structure during thermal testing. NDT&E Intl., 38:719
25, 2005.
370

Condition Assess of GFRP-Retrofitted Concrete Cylinders Using Electromagnetic Waves (2008) - Thesis PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Condition Assess of GFRP-Retrofitted Concrete Cylinders Using Electromagnetic Waves (2008) - Thesis PDF

Hochgeladen von

Copyright:

Verfügbare Formate

Condition Assessment of GFRP-Retrofitted

Concrete Cylinders Using Electromagnetic Waves

Condition Assessment of GFRP-Retrofitted Concrete

sessment of GFRP-retrofitted reinforced concrete and other reinforced concrete civil

I should like to take this opportunity to express my gratitude to Professor Michael

Organization of the Dissertation . . . . . . . . . . . . . . . . . . . . .

Nondestructive Testing (NDT) Techniques . . . . . . . . . . . . . . .

Magnetic and Electrical Methods . . . . . . . . . . . . . . . .

Microwave and Radar Methods . . . . . . . . . . . . . . . . .

Maxwells Curl Equations and Linearly Polarized EM Waves . . . . .

Finite Difference Time Domain Solution and Yees Algorithm . . . . .

Absorbing Boundary Condition Perfectly Matched Layer . . . . . .

Stability Criteria in Discretization . . . . . . . . . . . . . . . . . . . .

Two-Dimensional and Three-Dimensional Simulations . . . . . . . . .

Validation of the Code . . . . . . . . . . . . . . . . . . . . . .

Actual Far-Field Simulation . . . . . . . . . . . . . . . . . . .

Damage Detection in Normal Incidence . . . . . . . . . . . . .

Damage Detection in Oblique Incidence . . . . . . . . . . . . .

Effect of Defect Width in Normal Incidence . . . . . . . . . .

Effect of Defect Depth in Normal Incidence . . . . . . . . . . .

Rebar Detection in Normal Incidence . . . . . . . . . . . . . .

2D and 3D Responses . . . . . . . . . . . . . . . . . . . . . . 107

4 Laboratory Radar Measurements

Experimental Program . . . . . . . . . . . . . . . . . . . . . . . . . . 112

Manufacturing of the Specimens . . . . . . . . . . . . . . . . . . . . . 116

Experimental Configuration and Parameters . . . . . . . . . . . . . . 118

Monostatic ISAR Normal Incidence Scheme . . . . . . . . . . 120

Monostatic ISAR Oblique Incidence Scheme . . . . . . . . . . 121

Calibration of Laboratory Radar Measurements . . . . . . . . . . . . 121

PEC Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . 122

Lossy Dielectric Specimen and Its Optical Model . . . . . . . 124

Frequency-Angle Measurements . . . . . . . . . . . . . . . . . . . . . 127

Monostatic ISAR Normal Incidence Scheme . . . . . . . . . . 133

Monostatic ISAR Oblique Incidence Scheme . . . . . . . . . . 152

Single Scattering and Synthetic Aperture Radar . . . . . . . . . . . . 160

Inverse Synthetic Aperture Radar . . . . . . . . . . . . . . . . . . . . 172

Backprojection Algorithms . . . . . . . . . . . . . . . . . . . . . . . . 175

Range Compression . . . . . . . . . . . . . . . . . . . . . . . . 175

Backprojection Processing . . . . . . . . . . . . . . . . . . . . 178

Support Band Analysis Method of Stationary Phase . . . . . 179

Advantages of Backprojection Algorithms . . . . . . . . . . . 187

Implementation and Coding Procedure . . . . . . . . . . . . . . . . . 187

Effects of Aperture Size and Bandwidth

Aperture Size . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

6 Dielectric Modeling of GFRP-concrete Systems

Definition and Physics of Dielectric Properties . . . . . . . . . 202

Dielectric Spectroscopy and Dielectric Dispersion . . . . . . . 211

Energy Storage and Dissipation Mechanisms . . . . . . . . . . 212

Dielectric Properties in Microwave and Radar NDT . . . . . . 216

Approaches for the Determination of Dielectric Properties . . . . . . 216

Integrated Methodology for the Determination of Dielectric Properties 219

Time Difference of Arrival (TDOA) . . . . . . . . . . . . . . . 220

Root-searching Optimization Scheme . . . . . . . . . . . . . . 222

Validation of the Methodology . . . . . . . . . . . . . . . . . . 227

Modeling Approach for the Dielectric Properties of Materials . . . . . 233

Internal Field Approach versus External Field Approach . . . 233

Geometrical Analysis . . . . . . . . . . . . . . . . . . . . . . . 242

Dielectric Properties of Water . . . . . . . . . . . . . . . . . . . . . . 247

Bound Water . . . . . . . . . . . . . . . . . . . . . . . . . . . 253

Dielectric Properties of GFRP . . . . . . . . . . . . . . . . . . . . . . 263

Epoxy Resin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

E-glass Fabric . . . . . . . . . . . . . . . . . . . . . . . . . . . 266

GFRP Layer/Sheet . . . . . . . . . . . . . . . . . . . . . . . . 267

6-11 Behavior of the Davidson-Cole model (s = 2, = 1) . . . . . . . . . 241

TDOA measurements and dielectric constants 0r of the specimens . . 229

Loss factors 00r of the specimens . . . . . . . . . . . . . . . . . . . . . 229