Guided Waves Introduction 1

GUIDED WAVE BASED INTEGRATED STRUCTURAL HEALTH MONITORING
AND NONDESTRUCTIVE EVALUATION
by
Zhenhua Tian
Bachelor of Science
North China Electric Power University, 2008
Submitted in Partial Fulfillment of the Requirements
For the Degree of Doctor of Philosophy in
Mechanical Engineering
College of Engineering and Computing
University of South Carolina
2015
Accepted by:
Lingyu Yu, Major Professor
Victor Giurgiutiu, Committee Member
Yuh J. Chao, Committee Member
Paul Ziehl, Committee Member
Cara A.C. Leckey, Committee Member
Lacy Ford, Senior Vice Provost and Dean of Graduate Studies

ProQuest Number: 10008935
All rights reserved
INFORMATION TO ALL USERS

The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
ProQuest 10008935
Published by ProQuest LLC (2016). Copyright of the Dissertation is held by the Author.
All rights reserved.

This work is protected against unauthorized copying under Title 17, United States Code
Microform Edition © ProQuest LLC.
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
© Copyright by Zhenhua Tian, 2015
All Rights Reserved
ii
ACKNOWLEDGEMENTS
I would like to express my deepest gratitude to my advisor, Dr. Lingyu Yu, for her
dedicated guidance, thoughtful advice, tremendous support and willingness to allow me
to explore research areas I am truly passionate about. It is an honor to be her first Ph.D.
student and follow her as a role model of successful professor and independent thinker.
Her enthusiasm in research has motivated me throughout my Ph.D. journey and will
continuously encourage me in my future career.
I would like to give special thanks to my committee members. I am lucky to have
Dr. Victor Giurgiutiu as an excellent example to follow, to learn from his outstanding
work and insightful suggestions in Friday meetings. I also gained valuable guidance from
Dr. Cara A.C. Leckey, Dr. Yuh J. Chao and Dr. Paul Ziehl, who have enriched my
knowledge and broadened my horizon.
My gratitude is also extended to my dear colleagues in VSHM and LAMSS, who
have contributed immensely to my professional and personal life in the past four years.
The collaboration and friendship among us enlightened me to explore and achieve more.
The great conversations with Dr. Jingjing Bao, Dr. Bin Lin, Dr. Yanfeng Shen and
Banibrata Poddar, always inspired spark in me.
Lastly, I want to thank my parents for their unconditional love and support, and
Dr. Hong Xu from NCEPU for his unending encouragement. My friends in Columbia are
also appreciated for giving me the unforgettable memories. I would like to thank my
girlfriend Xue Bai, for her understanding, quiet patience and belief in me.
iii
ABSTRACT
Damage detection and health monitoring are critical for ensuring the structural safety in
various fields, such as aerospace, civil and nuclear engineering. Structural health
monitoring (SHM) performs online nondestructive evaluation (NDE) and can predict the
structural remaining life through appropriate diagnosis and prognosis technologies.
Among various SHM/NDE technologies, guided ultrasonic waves have shown great
potential for fast and large area SHM/NDE, due to their sensitivity to small defects and
capability to propagate long distances. Recent advances in guided wave based SHM/NDE
technologies have demonstrated the feasibility of detecting damage in simple structures
such as metallic plates and pipes. However, there remain many challenging tasks for
quantifying damage, especially for damage quantification in complex structures such as
laminated composites and honeycomb sandwich structures. Moreover, guided wave
propagations in complex structures, and wave interactions with various types of defects
such as crack, delamination and debonding damage, need to be investigated.
The objective of this dissertation research is to develop guided wave based
integrated SHM and NDE methodologies for damage detection and quantification in
complex structures. This objective is achieved through guided wave modeling, optimized
sensor and sensing system development, and quantitative and visualized damage
diagnoses. Moreover, the developed SHM/NDE methodologies are used for various
damage detection and health monitoring applications.
iv
This dissertation is organized in two major parts. Part I focuses on the
development of integrated SHM/NDE damage diagnosis methodologies. A non-contact
laser vibrometry sensing system is optimized to acquire high spatial resolution wavefields
of guided waves. The guided wavefields in terms of time and space dimensions contain a
wealth of information regarding guided wave propagations in structures and wave
interactions with structural discontinuities. To extract informative wave signatures from
the time-space wavefields and characterize the complex wave propagation and interaction
phenomenon, guided wavefield analysis methods, including frequency-wavenumber
analysis, wavefield decomposition and space-frequency-wavenumber analysis, are
investigated. Using these analysis methods, the multi-modal and dispersive guided waves
can be resolved, and the complex wave propagation and interaction can be interpreted
and analyzed in time, space, frequency, and wavenumber multi-domains.
In Part I, a hierarchical damage diagnosis methodology is also developed for
quantitative and visualized damage detection. The hierarchical methodology
systematically combines phased array imaging and wavefield based imaging to achieve
efficient and precise damage detection and quantification. The generic phased array
imaging is developed based on classic delay-and-sum principle and works for both
isotropic and anisotropic materials. Using the phased array imaging, an intensity scanning
image of the structure is generated to efficiently visualize and locate the damage zone.
Then the wavefield based imaging methods such as filter reconstruction imaging and
spatial wavenumber imaging are performed to precisely quantify the damage size, shape
and depth.
v
In Part II, the developed methodologies are applied to five different SHM/NDE
applications: (1) gas accumulation detection and quantification in water loaded structures,
(2) crack damage detection and quantification in isotropic plates, (3) thickness loss
evaluation in isotropic plates, (4) delamination damage detection and quantification in
composite laminates, (5) debonding detection and quantification in honeycomb sandwich
structures.
This dissertation research will initiate sensing and diagnosis methodologies that
provide rapid noncontact inspection of damage and diagnosis of structural health. In the
long run, it contributes to the development of advanced sensor and sensing technologies
based on guided waves, and to providing on-demand health information at component or
subsystem level for the safety and reliability of the structure.
vi
TABLE OF CONTENTS
ACKNOWLEDGEMENTS........................................................................................................ iii
ABSTRACT .......................................................................................................................... iv
LIST OF TABLES ....................................................................................................................x
LIST OF FIGURES ................................................................................................................. xi
CHAPTER 1 INTRODUCTION ...............................................................................................1
1.1. BACKGROUND .....................................................................................................1
1.2. MOTIVATION.......................................................................................................3
1.3. OBJECTIVES ........................................................................................................8
CHAPTER 2 STATE OF THE ART ........................................................................................10
2.1. GUIDED WAVE PROPAGATION STUDY ................................................................10
2.2. GUIDED WAVE EXCITATION AND SENSING .........................................................18
2.3. GUIDED WAVEFIELD DATA ANALYSIS AND DAMAGE DIAGNOSIS

METHODOLOGIES ..............................................................................................26
PART I: INTEGRATED SHM/NDE DIAGNOSIS METHODOLOGIES .......................................38
CHAPTER 3 ANALYTICAL AND NUMERICAL STUDIES OF GUIDED WAVES .......................38
3.1. DISPERSION AND MODE SHAPE DERIVATIONS ....................................................38
3.2. WAVE PROPAGATION AND INTERACTION SIMULATIONS ....................................44
CHAPTER 4 WAVEFIELD ACQUISITION AND INTRINSIC CHARACTERIZATION ...................53
4.1. HYBRID PZT-SLDV SENSING SYSTEM FOR WAVEFIELD ACQUISITION ..............53
4.2. FREQUENCY-WAVENUMBER ANALYSIS .............................................................57
vii
4.3. GUIDED WAVEFIELD DECOMPOSITION ...............................................................59
4.4. SPACE-FREQUENCY-WAVENUMBER ANALYSIS ..................................................61
CHAPTER 5 QUANTITATIVE AND VISUALIZED DAMAGE DIAGNOSIS ................................64
5.1. GUIDED WAVE PHASED ARRAYS IMAGING .........................................................64
5.2. FILTER RECONSTRUCTION IMAGING ..................................................................85
5.3. SPATIAL WAVENUMBER IMAGING .....................................................................87
PART II: SHM/NDE APPLICATIONS ..................................................................................89
CHAPTER 6 GUIDED WAVES IN WATER LOADED STRUCTURES AND WATER LEVEL

DETECTION................................................................................................................89
6.1. GUIDED WAVE WATER LEVEL DETECTION .........................................................89
6.2. GAS ACCUMULATION DETECTION IN NUCLEAR COOLANT PIPES .........................96
CHAPTER 7 CRACK DETECTION AND QUANTIFICATION IN ISOTROPIC PLATES ...............103
7.1. MOTIVATION AND INTRODUCTION ..................................................................103
7.2. WAVE INTERACTION WITH CRACK DAMAGE ....................................................104
7.3. CRACK DETECTION AND QUANTIFICATION ......................................................106
CHAPTER 8 THICKNESS LOSS EVALUATION IN ISOTROPIC PLATES ................................112
8.2. WAVE INTERACTION WITH MATERIAL LOSS DAMAGE ......................................112
8.3. DISPERSION CURVE REGRESSION FOR MATERIAL THICKNESS QUANTIFICATION116
8.4. EVALUATION OF SIMULATED PITTING CORROSION DAMAGE ............................119
CHAPTER 9 DELAMINATION DETECTION AND QUANTIFICATION IN COMPOSITE

LAMINATES .............................................................................................................121
9.2. WAVE INTERACTION WITH DELAMINATION DAMAGE ......................................123
9.3. DETECTION AND QUANTIFICATION OF SIMULATED DELAMINATION DAMAGE ..128
viii
9.4. DETECTION AND QUANTIFICATION OF IMPACT-INDUCED DELAMINATION
DAMAGE ..........................................................................................................133
CHAPTER 10 DEBONDING DETECTION AND QUANTIFICATION IN HONEYCOMB SANDWICH

STRUCTURES ...........................................................................................................144
10.2. GUIDED WAVES IN HONEYCOMB SANDWICH STRUCTURES...............................145
10.3. WAVE INTERACTION WITH DEBONDING DAMAGE ............................................150
10.4. DEBONDING DETECTION AND QUANTIFICATION ..............................................157
CHAPTER 11 CONCLUDING REMARKS AND FUTURE WORK ...........................................158
11.1. RESEARCH CONCLUSIONS ...............................................................................160
11.2. MAJOR CONTRIBUTIONS ..................................................................................168
11.3. RECOMMENDATION FOR FUTURE WORK ..........................................................172
REFERENCES .....................................................................................................................175
ix
LIST OF TABLES
Table 3.1 Wave modes in EFIT results when incident waves are A0 mode ................. 48
Table 3.2 Wave modes in EFIT results when incident waves are S0 mode .................. 49
Table 3.3 Geometry parameters of the finite element model. (Units: mm) .................. 50
Table 5.1 Material properties of a single ply. ............................................................... 71
Table 5.2 Configurations of three rectangular arrays for beamforming study.............. 71
Table 5.3 BF(k) characteristics of the three rectangular arrays .................................... 74
Table 5.4 Full width of the mainlobe at one-half peak value........................................ 77
Table 9.1 Material properties of the IM7/8552 lamina. .............................................. 125
Table 10.1 Geometry parameters of the honeycomb sandwich structure (units: mm). 147
x
LIST OF FIGURES
Figure 1.1 The comparison between human nervous system and SHM system. ..........2
Figure 1.2 Guided waves in different waveguides ........................................................3
Figure 1.3 Dispersion curves of a 1 mm thickness aluminum plate (2024-T3). ...........4
Figure 1.4 Complex structure examples ........................................................................5
Figure 1.5 Examples of various defects ........................................................................5
Figure 1.6 An illustration of an 8×8 piezoelectric transducers phased array ................6
Figure 1.7 Guided wave waveforms..............................................................................6
Figure 2.1 Sketches of multilayered structures ...........................................................11
Figure 2.2 Illustration of structures with complex cross sections. ..............................13
Figure 2.3 Examples of the SAFE method for solving the dispersion problems of
periodical structures ...................................................................................13
Figure 2.4 An example of the analytical simulation of guided waves generated by

a piezo-disc ................................................................................................14
Figure 2.5 An example of FEM simulation of guided waves in a honeycomb

sandwich ....................................................................................................17
Figure 2.6 Examples of EFIT simulations...................................................................18
Figure 2.7 Examples of various transducers ...............................................................19
Figure 2.8 PWAS interaction with a structure ............................................................21
Figure 2.9 PWAS tuning curves of a round PWAS (7mm diameter, 0.2mm
thickness) bonded on an aluminum 2024-T3 plate (1mm thick) ...............22
Figure 2.10 Schematics of PWAS applications.............................................................22
Figure 2.11 A schematic of SLDV principle (BS: beam splitter). ................................23
xi
Figure 2.12 Visualization of wave interaction with a fatigue crack by using laser
vibrometry ..................................................................................................25
Figure 2.13 The experimental wavefields in a structure (fabricated from two 0.8
mm thick aluminum plates that were epoxy-bonded together) ..................26
Figure 2.14 Amplitude profile of a guided wavefield for delamination detection ........28
Figure 2.15 Source removal for delamination detection ...............................................28
Figure 2.16 Standing wave filtering for delamination detection ...................................28
Figure 2.17 Sparse array imaging results for the detection of a through hole with a
diameter of 3.2 mm ....................................................................................31
Figure 2.18 Sparse array imaging results for the detection of a 6 mm diameter
through hole ...............................................................................................32
Figure 2.19 Baseline-free imaging ................................................................................32
Figure 2.20 Sparse array imaging for crack detection...................................................32
Figure 2.21 Comparisons between omnidirectional guided waves and focused

guided waves ..............................................................................................33
Figure 2.22 EUSR detection results ..............................................................................34
Figure 2.23 2-D planar arrays........................................................................................35
Figure 3.1 Sketches for: (a) free plate and (b) plate with one side in water................39
Figure 3.2 Wavenumber dispersion curves of a free steel plate with 1.2 mm
thickness .....................................................................................................39
Figure 3.3 Wavenumber dispersion curves of a 1.2 mm thick steel plate with one
side in water ...............................................................................................41
Figure 3.4 Theoretical wave mode shapes of the quasi-Scholte mode .......................41
Figure 3.5 SAFE model of a CFRP composite plate with [02/902]s layup ..................42
Figure 3.6 Dispersion curves of guided waves propagating in 90° direction in the
[02/902]s CFRP composite plate .................................................................43
Figure 3.7 Wavenumber and group velocity curves at 300 kHz for guided waves
in the [02/902]s CFRP composite plate .......................................................43
xii
Figure 3.8 Mode shapes of guided waves at 300 kHz propagating in 90° direction
in the [02/902]s CFRP composite plate .......................................................43
Figure 3.9 Configuration of EFIT simulation for the study of guided waves in a
water loaded plate ......................................................................................45
Figure 3.10 vz velocity results of EFIT simulation at 100 kHz excitation ....................45
Figure 3.11 vz velocity results of EFIT simulation at 600 kHz excitation ....................46
Figure 3.12 The EFIT simulation configuration for the study of wave interaction
with delamination damage in a CFRP plate with [02/902]s layup. .............46
Figure 3.13 EFIT results at 300 kHz for A0 mode interaction with delamination
damage .......................................................................................................48
Figure 3.14 EFIT results at 300 kHz for S0 mode interaction with delamination
damage .......................................................................................................49
Figure 3.15 The finite element model of the honeycomb sandwich structure with a
surface-bonded PZT. ..................................................................................50
Figure 3.16 Simulation results of displacements in the z direction uz ..........................50
Figure 4.1 A schematic of the PZT-SLDV sensing.....................................................55
Figure 4.2 Crack detection by using the PZT-SLDV sensing .....................................55
Figure 4.3 Delamination detection by using the PZT-SLDV sensing .........................57
Figure 4.4 Frequency-wavenumber analysis on guided waves in a 6.35 mm thick

T304 steel plate ..........................................................................................58
Figure 4.5 Guided wave decomposition by using frequency-wavenumber filtering ..60
Figure 4.6 An example of the window function. .........................................................63
Figure 4.7 An example of the space-frequency-wavenumber analysis .......................63
Figure 5.1 Schematics of geometric relations of guided waves in composite plates ..66
Figure 5.2 Schematic of a P×Q rectangular array .......................................................70
Figure 5.3 Wavenumber curve k(γ) of the 90 kHz A0 mode in the [0/45/90/-45]s
CFRP composite laminate..........................................................................71
Figure 5.4 BF(k) images for the array (1) ...................................................................74
xiii
Figure 5.5 BF(k) images when the phase delays (-0.02, 0.68)∙pp,q is applied.............74
Figure 5.6 BF(θ) plots for array (1) .............................................................................77
Figure 5.7 BF(θ) plots when the beamsteering directions are θS =0°, 45°, 90° and
135° ............................................................................................................77
Figure 5.8 Experimental setup for multi-defects detection on a CFRP composite

laminate ......................................................................................................80
Figure 5.9 SLDV measurements. ................................................................................82
Figure 5.10 Beamforming and imaging of array (2) .....................................................84
Figure 5.11 Array imaging results .................................................................................84
Figure 5.12 Filter reconstruction imaging for delamination detection in a CFRP

composite plate. .........................................................................................86
Figure 5.13 Spatial wavenumber imaging for delamination detection in a CFPR

plate ............................................................................................................88
Figure 6.1 The PZT-SLDV experiment on a water tank .............................................91
Figure 6.2 Experimental results...................................................................................92
Figure 6.3 Frequency-wavenumber analysis results ...................................................92
Figure 6.4 An illustration of the pitch-catch configuration. ........................................95
Figure 6.5 Water influence on wave propagation at 100 kHz .....................................95
Figure 6.6 The experimental setup ..............................................................................97
Figure 6.7 Experimental results...................................................................................97
Figure 6.8 Frequency-wavenumber analysis ...............................................................98
Figure 6.9 An illustration of the pitch-catch sensing on a small scale pipe. .............100
Figure 6.10 Experimental results for the small scale pipe ..........................................100
Figure 6.11 Test on a tilted large scale pipe ................................................................102
Figure 6.12 Experimental results for the tilted large scale pipe ..................................102
xiv
Figure 7.1 Experimental setup for crack detection in a 1 mm thick aluminum
plate ..........................................................................................................104
Figure 7.2 Wavefield images measured by the SLDV at 360 kHz excitation on a 1
mm thick aluminum plate with a TT crack ..............................................105
Figure 7.3 Filter reconstruction imaging with experimental data in an aluminum

plate with a TT crack ...............................................................................107
Figure 7.4 Spatial Wavenumber imaging with experimental data in an aluminum

plate with a TT crack. ..............................................................................107
Figure 7.5 The filter reconstruction imaging results at 360 kHz...............................110
Figure 7.6 Amplitudes along the cracks extracted from the filter reconstruction
imaging results. Broken, solid and dotted lines are for TT, 3X and hTT
cracks, respectively ..................................................................................110
Figure 7.7 Spatial wavenumber imaging results at 360kHz ......................................111
Figure 7.8 Wavenumber distributions along the crack. Broken, solid and dotted
lines are for TT, 3X and hTT cracks, respectively. .................................111
Figure 8.1. Experimental setup...................................................................................113
Figure 8.2. Chirp excitation........................................................................................114
Figure 8.3 Spectral fields (real parts) for the plate with 0.75 mm depth material
loss damage ..............................................................................................115
Figure 8.4 Spatial wavenumber images for the plate with 0.75 mm depth material
loss damage ..............................................................................................115
Figure 8.5 A0 mode wavenumber versus plate thickness and wave frequency. ........116
Figure 8.6 Dispersion curve regression result at the location (0, 60) mm, when the
damage depth is 0.75 mm. .......................................................................118
Figure 8.7 The thickness loss quantification results..................................................118
Figure 8.8. Experimental setup...................................................................................119
Figure 8.9 Spectral fields (real parts) for the plate with simulated pitting corrosion
damage .....................................................................................................119
Figure 8.10. Imaging result of the plate with simulated pitting corrosion damage. .....120
xv
Figure 9.1 The scanning laser vibrometry test ..........................................................126
Figure 9.2 Time-space wavefields of waves propagating along the y axis ...............126
Figure 9.3 SLDV wavefield snapshots at 15 µs ........................................................127
Figure 9.4 Wavenumber spectra at 300 kHz excitation frequency ...........................128
Figure 9.5 The filtering process in wavenumber domain ..........................................130
Figure 9.6 The filter reconstruction imaging results .................................................130
Figure 9.7 Spatial wavenumber image of the Teflon delaminated plate ...................132
Figure 9.8 Dispersion curves of A0 modes propagating towards the 90° direction ..133
Figure 9.9 Quasi-static indentation for delamination damage growth ......................135
Figure 9.10 The 26 ply laminate after the quasi-static indentation test ......................135
Figure 9.11 The SLDV phased array sensing layout for detecting and imaging the
impact-induced delamination damage. ....................................................136
Figure 9.12. SLDV measurements ...............................................................................137
Figure 9.13 Phased array imaging results for detecting and imaging the impact-
induced delamination damage..................................................................138
Figure 9.14 Experimental setup for quantifying the impact-induced delamination. ...139
Figure 9.15 SLDV wavefield images at 45 µs for 100 kHz excitation. ......................140
Figure 9.16 SLDV wavefield images at 40 µs for 500 kHz excitation .......................140
Figure 9.17 Filter reconstruction imaging at the low frequency 100 kHz ..................141
Figure 9.18 Filter reconstruction imaging at the high frequency 500 kHz .................142
Figure 9.19 Spatial wavenumber imaging result in the frequency band 480~520
kHz. ..........................................................................................................143
Figure 10.1 A schematic of the PZT-SLDV sensing setup. ........................................146
Figure 10.2 Wavefields measured from the experiment .............................................147
Figure 10.3 Wavenumber spectra ................................................................................149
xvi
Figure 10.4 The sensing layout of the PZT-SLDV experiment on a honeycomb
sandwich with debonding damage in the center of the sandwich. ...........150
Figure 10.5 Experimental results at the high frequency 100 kHz for the pristine
(left column) and damaged (right column) sandwich structures .............151
Figure 10.6 Wavenumber spectra at 100 kHz: (a) for the pristine sandwich and (b)
for the sandwich with debonding damage. ..............................................151
Figure 10.7 Wave decomposition for the pristine sandwich .......................................153
Figure 10.8 Wave decomposition of the waves in the sandwich with debonding
damage .....................................................................................................154
Figure 10.9 Experimental results at the low frequency 40 kHz for the pristine (left
column) and damaged (right column) sandwich structures .....................155
Figure 10.10 Wavenumber analysis results at 40 kHz ..................................................156
Figure 10.11 Debonding detection and quantification ..................................................157
xvii
CHAPTER 1
INTRODUCTION
1.1. BACKGROUND
Structural health monitoring (SHM) assesses the state of structural health, and through
appropriate data processing and interpretation, may predict the remaining life the
structure [1]. Over the past two decades, SHM has been experiencing spectacular
progress in various areas such as aerospace, marine, civil, and nuclear engineering [1-3].
An SHM system behaves like the nervous system (as shown in Figure 1.1) with the
particular purpose of monitoring structural behavior in quasi-real time, indicating the
health state of structure, and “triggering” safety and remedial actions [4]. In the SHM
system, there are four important elements: data collection, data processing, information
processing to define the current state, and the decision making and dissemination [4]. The
SHM system can perform in two main ways: passive and active [1, 4]. The passive SHM
is concerned with measuring various operational parameters (such as acoustic emissions,
vibration levels, stresses in critical locations, etc.) by using receivers and then inferring
the health state of structure [1]. In contrast, the active SHM is concerned with directly
assessing the health state of structure by using both actuators and receivers [1].
The health state of the structure is indicated by the presence and extent of
structural damage. The damage can be changes to material properties, geometric
parameters, and boundary conditions, which adversely affect the present and future safety
of structures [2]. Generally, the identification of the damage state involves the following
1
five aspects: (1) the existence of damage, (2) the locations of damage, (3) the types of
damage, (4) the severity of the damage, and (5) the remaining service life of the structure
[5].
Human nervous system SHM system
?
Sensor network (nerves)
Indication
Diagnosis
?
Actions
Figure 1.1 The comparison between a human nervous system and an SHM system [4].
Various means, such as ultrasonic waves, structure vibration, electromechanical
impedance, eddy current, micro waves, etc., can be used for NDE and SHM [1, 4, 6].
Particularly, this dissertation study focuses on the guided ultrasonic waves, which are
elastic waves propagating along waveguides [7]. Different structural components, such as
plate-like structures and pipes can serve as waveguides [7-11]. In these waveguides, the
guided waves travel in different forms. For example, Rayleigh waves travel along a free
surface (Figure 1.2a), symmetric Lamb waves travel along a plate (Figure 1.2b), torsional
guided waves travel in a cylinder (Figure 1.2c), etc. Other cases, like a solid-water
interface, a solid-solid interface, etc., can also serve as the waveguides [7-11].
Compared to bulk waves, guided waves have the advantage of long propagation
distance with less energy loss. Moreover, guided waves have high sensitivity to small
defects, such as crack, corrosion, delamination, etc. [6, 8]. These merits enable the guided
wave SHM of large areas from small sensing areas. In the past two decades, various
2
guided wave SHM technologies have been developed for the purpose of heath monitoring,
such as phased arrays [12-21], sparse arrays [22, 23], time-reversal method [24-26],
guided wave tomography [27, 28] and full-wavefield imaging [29-32]. These advances in
guided SHM technologies have demonstrated the feasibility of health monitoring by
using guided waves.
Semi-infinite space
(a) (b) (c)
Figure 1.2 Guided waves in different waveguides: (a) Rayleigh waves along a free
surface of a semi-infinite space solid medium; (b) symmetric Lamb waves in a plate; (c)
torsional guided wave in a cylinder.
1.2. MOTIVATION
Although the advances in guided wave technologies have demonstrated the feasibility of
guided wave SHM, there remain many challenges for applications of guided wave SHM,
especially for structures with complex geometries and subjected to multiple damage
when serving in harsh environment conditions.
The followings summarize the challenges:
(1) Guided wave propagation and interaction mechanisms
 Guided waves have multimodal and dispersive features [8]. For example, in the
dispersion curves of a 1 mm thickness aluminum plate (Figure 1.3), there are at
least two wave modes, A0 and S0, existing simultaneously even at low frequencies.
When the product of frequency and plate half-thickness goes higher, more wave
3
modes (such as A1 and S1) appear. Moreover, the dispersive nature shows the
guided wave speed changes with respect to frequency.
S0
Wave number k (rad/mm)
Group velocity (m/s)

S1
S2
A1
A0
A0
S0 S1 S2 A2 A1 A2
(a) Frequency (kHz) (b) Frequency (kHz)
Figure 1.3 Dispersion curves of a 1 mm thickness aluminum plate (2024-T3): (a)

frequency-wavenumber dispersion curves, and (b) group velocity dispersion curves.
 The dispersion curves are important features of guided waves and play an
important role for guided wave SHM. The dispersion curves can be used to
determine the wave propagation speed, identify the guided wave mode, and
develop appropriate data analysis algorithms for damage detection [1]. For simple
structures, such as isotropic plates and pipes, the dispersion curves can be easily
obtained by solving the guide wave characteristic equations which have been fully
developed in many textbooks and articles [1, 8-11, 33]. However, for complex
structures, such as multilayer structures, arbitrary cross section structures,
periodical structures, etc., it is more challenge to derive their dispersion curves
and understand how the guided waves propagate in these complex structures.
 Structures can have various types of defects, such as cracks, delaminations and
corrosions (Figure 1.5). In order to develop effective methods to detect defects, it
is necessary to understand the mechanism of how defects affect the guided wave
propagation.
4
(a) (b) (c)
Figure 1.4 Complex structure examples: (a) multilayer structure; (b) arbitrary cross
section structure; (c) periodical structure.
Fatigue Crack Delamination Corrosion
(a) (b) (c)
Figure 1.5 Examples of various defects: (a) fatigue crack; (b) X-Ray CT image of
delamination damage [34]; (c) corrosion damage near the rivets [35].
(2) Guided wave excitation and sensing.
 The traditional guided wave sensing system may require a large amount of
sensors, such as the 8×8 phased array in Figure 1.6, resulting in great sensor
installation, wiring and weight penalties.
 Traditional guided wave transducers, such as piezoelectric transducers in Figure
1.6, may have large sensor profiles, which limit the spatial sampling resolution of
the sensing.
 Guided wave sensors may work in harsh environment, such as high temperature
and radiation, which may affect the sensor properties. For example, in nuclear
applications, the sensors may work in the high temperature and radiation
environment.
 The accessible area for inspection may be limited, and the surface bounded sensor
may not be allowed in real applications.
5
Figure 1.6 An illustration of an 8×8 piezoelectric transducers phased array [17].
Velocity (mm/s)
Velocity (mm/s)
(a) Time (μs) (b) Time (μs)
Figure 1.7 Guided wave waveforms: (a) a waveform which contains 7 wave modes (A0,
A1, A2, A3, S0, S1 and S2); (b) a waveform which contains both incident and reflection
waves (incident A0 and S0 modes, and reflection A0 and S0 modes)
(3) Data interpretation and damage diagnosis methodologies
 Due to the dispersive and multimodal natures of guided waves, the data
interpretation of guided wave signal becomes very challenging. For example, in
Figure 1.7a, a guided waveform, there are seven different wave modes (A0, A1, A2,
A3, S0, S1 and S2). When multiple dispersive wave modes coexist and interact
with each other, the wave interpretation and analysis become very difficult. It is
desired that different modes can be separated for individual analysis and the
dispersion effect can be mitigated.
 Structures can have various types of defects, such as cracks, delaminations and
corrosions. When guided waves interact with these defects, refection, scattering
and mode conversion waves may present. These damage induced waves make the
interpretation of guided wave signal very difficult. For example, the guided
waveform in Figure 1.7b, contains incident A0 and S0 modes generated from the
6
actuator, and reflection A0 and S0 modes induced by the damage. Since these
different components coexist and interact with each other, the interoperation of
wave signal becomes very difficult. To analyze the damage affections, such as the
damage reflection and scattering coefficients, the damage induced waves are
needed. Moreover, for damage detection, the damage induced waves which
contain the damage information are needed. Therefore, it is desired that the
damage induced waves can be extracted for wave analysis and damage detection.
 Various damage detection algorithms have been developed, such as phased arrays
[12-21], sparse arrays [22, 23], time-reversal method [24-26], guided wave
tomography [27, 28] and full-wavefield imaging [29-32]. Although recent
advances in SHM technology have demonstrated the feasibility of detecting
damage, quantifying the damage based on sensory data still remains a challenging
task. It is challenging to quantify the location, size and shape of the damage.
Moreover, the challenges will be accentuated when the damage diagnosis
methods are applied on complex structures.
These challenges involve researches and studies in multidisciplinary areas. Hence,
this dissertation research aims to address the above challenges and move guided wave
based SHM and NDE toward field applications. This dissertation focuses on the
excitation and sensing technologies, data interpretation and damage diagnosis
methodologies. Wave propagation and interaction are also studied for the purpose of
optimization of damage detection and quantification.
7
1.3. OBJECTIVES
Considering the challenges summarized in section 1.2, the following detailed research
objectives are targeted:
(1) Both the analytical modeling and numerical modeling of guided waves will be
investigated, in order to acquire preliminary understanding of the guided waves in
the complex structures, to obtain theoretical data for comparison to experimental
results, as well as to provide abundant data for parametric study;
(2) Guided wave excitation and sensing systems will be investigated and optimized in
order to reduce the sensor installation, generate high energy guided waves, and
acquire the high resolution guided wavefield;
(3) Advanced data analysis approaches will be developed to characterize the
multimodal and dispersive guided waves in complex structures, analyze the wave
interaction with damage, and extract wave characteristic parameters (such as
dispersion, frequency, and wavenumber contents);
(4) Quantitative structural integrity assessment methodologies will be developed for
SHM\NDE in complex structures, in order to quantify the damage location, size,
and detailed geometry features;
(5) The methodology for wave propagation and interaction analysis in complex
structures will be developed. The framework of how to use the fundamental
knowledge of guided wave modeling, excitation and sensing, and data analysis
methods for the analysis wave propagation and interaction problems will be
established. Case studies on various representative complex structures will be
performed: (i) guided waves in water loaded structures and water level detection; (ii)
8
crack detection and quantification in isotropic plates; (iii) damage thickness
mapping in isotropic plates; (iv) delamination detection and quantification in
composite laminates; (v) debonding detection and quantification in honeycomb
sandwich structures.
9
CHAPTER 2
STATE OF THE ART
Guided waves have been widely used for NDE/SHM in the past couple of decades [1, 8,
36, 37]. Significant works have been done in guided wave modeling, guided wave
generation and sensing, and damage detection. This section presents the state of the art in
these research areas with particular emphasis on guided waves in complex structures.
2.1. GUIDED WAVE PROPAGATION STUDY
Modeling can serve as an efficient and cost effective way to assist the guided wave study
and the development of diagnostic algorithms before conducting labor and/or time
consuming experiments [1, 8, 36, 37]. Modeling is also helpful in investigating scenarios
that are difficult to set up experimentally. With modeling, important guided wave features
such as dispersions and wave mode shapes can be acquired. Moreover, wave
propagations in complex structures and wave interactions with structural discontinuities
can be easily simulated [1, 8, 36, 37].
2.1.1. Dispersion curve and mode shape derivations
Dispersion curves and mode shapes are important features of guided waves. They can be
used to determine the wave propagation speed, identify the guided wave mode, and assist
the development of appropriate data analysis algorithms for damage detection [1]. In
order to obtain the dispersion relations and mode shapes, analytical and numerical
methods have been developed [1, 8-10, 38-42].
10
2.1.1.1. Analytical methods
The analytical methods are based on closed-form solutions of guided wave characteristic
equations [1, 8-10, 33]. They are very convenient to solve the dispersion relations for
simple geometry structures, such as single layer plates and pipes. The characteristic
equations and solutions for these simple structures have been fully developed as given in
many textbooks and articles [1, 8-11, 33].
(b)
(a) (c)
Figure 2.1 Sketches of multilayered structures: (a) a laminated composite [1]; (b) an
adhesive joint[43]; (c) a multilayered pipe [44].
For multilayered structures such as laminated composites (Figure 2.1a), adhesive
joints (Figure 2.1b), and pipes with coating layers (Figure 2.1c), their dispersion curves
are usually obtained analytically using matrix methods [33]. The matrix methods are
developed in matrix formulations by combining the dynamic behaviors of the continuum
within each layer [33]. Two matrix methods are widely used, the transfer matrix method
(TMM) and the global matrix method (GMM) [33]. The TMM works by condensing the
multilayered system into a single layer system relating the boundary conditions at the
first interface to the boundary conditions at the last interface [33]. In this process, the
equations for the intermediate interfaces are eliminated so that all the layers of the plate
11
are described solely in terms of the external boundary conditions. The advantage of the
TMM is that the condensed matrix is small and therefore it needs less computation time
[33]. A drawback of the transfer matrix method is the dispersion solution becomes
unstable for large frequency-thickness product [33].
The idea of GMM is to directly assemble a single matrix which represents the
complete system [33]. The GMM is based on satisfying the boundary conditions at each
interface [33]. The solution is carried out on the full matrix, addressing all of the
equations concurrently. The advantage of GMM is that it is robust and stable for any
frequency-thickness product. The disadvantage is that the global matrix might be large
and therefore the method needs more computational effort.
2.1.1.2. Numerical methods
For arbitrary cross section structures, such as the structures in Figure 2.2, it is difficult to
use analytical methods to solve the guided wave dispersion problems [40]. For these
cases, numerical methods emerge and show their advantages [38-41]. One promising
numerical method is the semi-analytical finite element (SAFE) method that has also been
referred to as spectral or waveguide finite element method [38-41].
For the structure with a uniform arbitrary cross section (Figure 2.2a), the SAFE
method simply requires the finite element discretization of the cross-section, and assumes
harmonic motion along the wave propagation direction [38]. By this means, the guided
wave dispersion problem can be casted into an eigenvalue problem which can be easily
solved. For example, Fan and Lowe used SAFE to solve the guided wave dispersion
problem of a welded joint which was considered as an uniform arbitrary cross section
12
structure [45]. Bartoli et al. used SAFE method to solve the dispersion problem of a rail
[38].
Besides the structure with a uniform arbitrary cross section, the SAFE method has
also been used to solve the dispersion problem for the periodic structure that has
changing cross sections [39, 42]. Figure 2.3 gives two examples of using the SAFE
method to solve the dispersion problems for 1-D and 2-D periodical structures [39, 42].
The SAFE method simply requires the finite element discretization of a unit cell, and
assumes periodic boundary conditions on the unit cell [39, 42].
x3
Changing cross
section along x3
Uniform cross Unit cell
section along x3
(a) (b)
Figure 2.2 Illustration of structures with complex cross sections : (a) with a uniform
arbitrary cross-section, and (b) with periodically changing cross-sections [40].
1D periodical Unit cell 2D periodical Unit cell
(a) (b)
Figure 2.3 Examples of the SAFE method for solving the dispersion problems of
periodical structures: (a) 1-D periodical structure [39]; (b) 2-D periodical structure[42].
13
2.1.2. Wave propagation simulations
Guided wave simulations can provide guided wave propagation and interaction details,
such as how guided waves propagate in the structure and how guided waves interact with
defects [1, 37, 46]. To simulate guided waves, there are various methods, such as
analytical method [47, 48], finite element method (FEM) [49-51], elastodynamic finite
integration technique (EFIT) [52-54], finite difference method (FDM) [55-58], spectral
element method (SEM) [59, 60], local interaction simulation approach (LISA) [61, 62],
distributed point source method (DPSM) [63], hybrid global-local method (HGL) [64-66],
boundary element method (BEM) [67-69], etc. The analytical method, FEM and EFIT
will be reviewed as followed.
(a) (b)
Figure 2.4 An example of the analytical simulation of guided waves generated by a

piezo-disc: (a) simulation configuration; (b) simulated guided waves [47].
2.1.2.1. Analytical method
The analytical method is based on closed-form solutions of guided wave transfer
functions. To get the closed-form solutions, guided wave equations are usually solved by
using Fourier transform [1, 47, 48]. Once the closed-form solutions of the guided wave
transfer functions are obtained, the guided wave propagation can be simulated efficiently.
14
Figure 2.4 shows a simulation example of guided waves actuated by a piezo-disc [47].
Although the analytical method is computational efficient, this method is restricted to
structures with simple geometries. When the structure geometry is complex, it is difficult
to get the closed-form solutions of structure transfer functions.
2.1.2.2. Numerical methods
Numerical methods begin with the mathematical models represented by Partial
Differential Equations (PDEs) which govern the wave mechanics [49-51]. To solve the
PDEs, there are various numerical methods, such as FDM, EFIT, FEM, etc. [49-58].
These methods have the advantage of handling complicated guided wave propagation and
interaction problems, such as wave propagations in complex geometry structures, wave
interactions with various defects, nonlinear guided wave propagation, etc. [48, 70-72].
However, the numerical methods have inherent errors induced by the time and space
discretization (or approximation). The success of numerical methods strongly depends on
the minimization of approximation errors, which requires sufficient time and space
discretization. In addition, the numerical methods require large computational resources,
such as computation time and computer memory.
2.1.2.2.1. Finite element method (FEM)
FEM, which is a powerful numerical technique to solve problems governed by PDEs over
complex domains, has severed as a primary numerical tool to simulate the elastic wave
propagation [49-51]. In the FEM, the governing PDEs are transformed to weak
formulations, which are integral representations of the governing PDEs. The analysis
domain is discretized into a collection of finite elements. The FEM then approximates the
15
solutions to PDEs over finite elements in the analysis domain. The details of FEM are
available in many text books for example [49-51].
Nowadays, with off the shelf commercial FEM software products such as
ABAQUS, ANSYS, and COMSOL, it is very convenient to use the FEM for the guided
wave simulation [70-72]. These commercial software products have many advantages,
such as user friendly model construction tools, various element types in the element
library, robust meshing tools, efficient FEM solvers, strong post processing and analysis
tools, etc. With these merits, many researchers used the commercial FEM products to
study various guided wave problems [70-73]. For example, Song et al. studied guided
wave propagations in honeycomb sandwiches by using ANSYS [72]. Figure 2.5a shows
their finite element model of the honeycomb sandwich. In their model, they took the
advantage of ANSYS element library to model the piezoelectric transducers using
SOLID5 element, hexagonal honeycomb core using SHELL63 element, and skin panels
using SOLID45 element. With efficient solvers and strong post processing tools, they
were able to solve the wave propagation problem, obtain detailed waveforms, and
visualize guided waves (Figure 2.5a) for further analyses and studies. Pistone et al.
simulated guided waves in an immersed plate by using ANSYS [71]. In their finite
element model, they used the PLANE42 and FLUID29 elements of the ANSYS element
library to model the isotropic plate and surrounding water, respectively. By using the
ANSYS post processing tools, they can easily visualize the wave interactions among the
leaky waves, bulk waves and quasi-Scholte waves.
16
(a) (b)
Figure 2.5 An example of FEM simulation of guided waves in a honeycomb sandwich:

(a) the FEM model and (b) the simulated guided waves [72].
2.1.2.2.2. Elastodynamic finite integration technique (EFIT)
EFIT is another prominent numerical method to simulate guided waves. The EFIT is
based on the finite integration technique (FIT) which integrates PDEs over a control
volume and then approximates the integrals [52]. For guided wave simulations, the EFIT
have following advantages: (1) the EFIT naturally requires staggered spatial and temporal
grids which lead to stability, (2) the boundary conditions are easily incorporated in the
EFIT model, and (3) the mathematical analysis is straight-forward and leads to equations
that can be easily implemented in any programming language [53].
In the past two decades, many researchers used EFIT to study various guided
waves related problems, such as guided waves in structures with complex geometries,
and guided wave interaction with damage [53, 54, 74-79]. Leckey et al. studied guided
wave interactions with defects, such as the material loss defect and microcracks using the
EFIT [53, 54]. With the EFIT simulation, they can directly visualize how the guided
waves interact with defects (Figure 2.6). Rudd et al. simulated the guided waves in pipe
bends [77]. With the EFIT simulation, they directly showed energy concentrations at pipe
bends. Schubert simulated wave propagation and scattering in a concrete structure by
17
using the EFIT [79]. The structure details such as sand aggregates and structural porosity
were successfully modeled with the EFIT.
(a) (b)
Figure 2.6 Examples of EFIT simulations: (a) guided wave interaction with a material
loss damage in an isotropic plate [53]; (b) guided wave interactions with microcracks in a
composite plate [54]
2.2. GUIDED WAVE EXCITATION AND SENSING
For guided wave excitation and sensing, various transducers have been used, such as
piezoelectric wafer active sensors (PWAS) (Figure 2.7a), comb transducer (Figure 2.7b),
macro fiber composites (MFC) (Figure 2.7c), wedge transducers (Figure 2.7d), fiber
optics (Figure 2.7e), electromagnetic acoustic transducers (EMAT) (Figure 2.7f), air-
coupled transducers (Figure 2.7g) and laser devices (Figure 2.7h) [80-94]. Among these
transducers, the low profile PWAS are widely used for guided wave excitation and
sensing [1]. PWAS are small and light, and suitable for integration into host structures
(surface-mounting or embedding in composites) without significant intrusion, serving as
good candidates for built-in transducers. Moreover, PWAS can serve several purposes,
such as high-bandwidth strain sensors and exciters, resonators, and embedded modal
sensors [1]. Recently, the laser devices, such as the high power pulse laser and the laser
Doppler vibrometer, have emerged for non-contact guided wave applications [94]. The
pulse laser can excite high energy wide band guided waves based on either the thermal
elastic effect or structure surface oblation. The laser Doppler vibrometer can measure the
18
velocity or displacement at the sensing point on structural surface in the direction of the
laser beam based on the Doppler effect. Their non-contact and remote sensing natures
have attracted a lot of attention.
(a) (b)
(c) (d)
Permanent
Magnetic field
magnet
Pancake
(e) (f) coil
Doppler effect
(g) (h)
Figure 2.7 Examples of various transducers: (a) PWAS [1]; (b) comb transducer [84,
92]; (c) MFC [90]; (d) wedge transducers [93]; (e) FBG [95]; (f) EMAT transducer [80];
(g) air-coupled transducer [96]; (h) laser transducers [97].
2.2.1. PWAS guided wave transducers
PWAS work on the piezoelectric principle and couple the electrical and mechanical
effects (mechanical strain Sij , mechanical stress Tkl , electrical field Ek , and electrical
displacement D j ), as described in the piezoelectric constitutive equations [1]
19
E
Sij sijkl
= Tkl + d kij Ek
(2.1)
D j d jklTkl + ε Tjk Ek
=
E
where sijkl is the mechanical compliance of the material measured at zero electric field
(E=0), ε Tjk is the dielectric permittivity measured at zero mechanical stress (T=0), and
d kij represents the piezoelectric coupling effect. PWAS can excite and receive guided
waves through in-plane motion coupling through the d31 piezoelectric coupling effect [1].
Compared to conventional ultrasonic probes, PWAS transducers have the following
fundamental differences [1]: (1) PWAS achieve Lamb wave excitation and sensing
through surface “pinching” (in-plane strains), while conventional ultrasonic probes work
through surface “tapping” (normal stress); (2) PWAS are strongly coupled with the
structure and follow the structural dynamics, while conventional ultrasonic probes are
relatively free from the structure and follow their own dynamics; (3) PWAS are non-
resonant wideband devices, while conventional ultrasonic probes are narrow-band
resonators.
Many researchers have contributed to the development, modeling, and
applications of PWAS to NDE/SHM [1, 88, 98-101]. Giurgiutiu studied the coupling
effect between PWAS and the host structure, and proposed a pin force model [101]. The
surface bonded PWAS (Figure 2.8 a) can couple their in-plane motion with the surface of
the host structure through the bonding layer. When the bonding layer is very thin (1μm),
the shear stress transmitted to the structure is only at the ends of PWAS (Figure 2.8 b).
With the ideal bonding assumption, the transmitted shear stress can be considered using a
pin force model which has all loads confined to an infinitesimal region at the PWAS edge
[101].
20
An important characteristic of PWAS, which distinguishes them from
conventional ultrasonic transducers, is their capability of tuning into various guided wave
modes. The tuning effect of PWAS will benefit the selective actuation and sensing of
guided wave modes. Giurgiutiu and Santoni et al. did comprehensive studies of the
PWAS tuning effect through theoretical analyses and experimental tests [100, 101].
Figure 2.9 shows theoretical and experimental tuning curves of a round PWAS (7 mm
diameter, 0.2 mm thickness) bonded on an aluminum 2024-T3 plate (1 mm thickness).
The theoretical tuning curves agree well with the experimental tuning curves, and show
mode amplitudes vary with excitation frequencies. At 300 kHz, the amplitude of the S0
mode is dominated, while the A0 mode hits zero. Therefore, the 300 kHz is a “sweet spot”
for single S0 mode excitation. With the embedded guided wave excitation and sensing
abilities, PWAS have be used for various SHM applications [1, 17, 88, 98-101], such as
(1) active sensing of far-field damage using pitch-catch (Figure 2.10 a), pulse-echo
(Figure 2.10 b) and phased array (Figure 2.10 e) methods, (2) active sensing of near field
damage using E/M impedance method (Figure 2.10 f) and thickness sensing mode
(Figure 2.10 c), and (3) passive sensing of acoustic emission (AE) events (Figure 2.10 d).
(a) (b)
Figure 2.8 PWAS interaction with a structure: (a) a model of layer interaction between
PWAS and structure; (b) variation of shear-lag transfer mechanism with bond thickness
for a APC-850 PWAS attached to a thin-wall aluminum through a bond layer[17].
21
(a) (b)
Figure 2.9 PWAS tuning curves of a round PWAS (7mm diameter, 0.2mm thickness)
bonded on an aluminum 2024-T3 plate (1mm thick): (a) theoretical results, and (b)
experimental results [17].
Pitch-catch PWAS phased array
(a)
Pulse-echo
(e)
(b)
Thickness mode E/M impedance
(c)
Impact/AE detection
(d) (f)
Figure 2.10 Schematics of PWAS applications: (a) pitch-catch sensing; (b) pulse-echo
sensing; (c) thickness sensing mode; (d) impact/AE detection; (e) PWAS phased array; (f)
E/M impedance method[1, 17, 88, 98-101].
22
2.2.2. SLDV high spatial resolution guided wave sensor
The scanning laser Doppler vibrometer (SLDV) can make accurate surface
velocity/displacement measurements of an object based on Doppler effect on light waves.
Figure 2.11 shows a schematic of the laser vibrometry principle [102]. The laser beam is
split by a beam splitter (BS1) into a reference beam and a measurement beam. After
focusing, the measurement beam is directed to the surface of the vibrating object through
vertical and horizontal scanning mirrors. The reflection light from the vibrating object
exhibits a frequency shift due to Doppler effect in response to the motion at the scanning
point on the vibrating object. The SLDV receives the reflection light and compares its
frequency against the reference beam and generates a voltage corresponding to the
frequency shift. This voltage signal is correlated to the surface velocity.
BS1 BS2 Lens Vertical

Laser source scanning mirror
Measurement beam
Reference beam
Bragg Cell
Horizontal
BS3 scanning mirror
Mirror Frequency shift by Doppler
effect upon reflection
Vibrating
Detector object
Figure 2.11 A schematic of SLDV principle (BS: beam splitter).
In recent years, the SLDV have been used as a non-contact means for guided
wave sensing and wavefield visualization. The visualization can provide deeper
understanding of guided waves in various structures and their interactions with structural
23
discontinuities. Many researchers have used laser vibrometry for wave propagation
visualization, damage detection, model validation, etc. [29, 30, 103-110]. Staszewski et al.
studied laser vibrometry for guided wave measurements and visualized wave interactions
with fatigue cracks by using a 3-D SLDV for the purpose of fatigue crack detection [104].
As shown in their laser vibrometry results in Figure 2.12, the 3-D laser vibrometry
sensing provides a series of images which visualize the wave propagation and interaction.
In addition, 3-D SLDV provides 3-D wave components (X, Y and Z components), which
contain a wealth of information to assist the analysis of such complicated wave
propagation and interaction. Ostachowicz et al. used a 3-D laser scanning vibrometer to
measure wavefields of Lamb waves and to understand the interactions between the waves
and structural discontinuities such as notch and holes [108]. Swenson et al. compared the
1-D and 3-D SLDV measurements of Lamb waves that were excited by a piezoelectric
transducer [106]. They showed the 3-D SLDV can provide both in- and out-of-plane
wave components, which are especially important for the study of both symmetric and
anti-symmetric wave modes. However, the 1-D SLDV can only measure the motion
along the laser beam. When the test specimens are placed normal to the laser beam, the 1-
D SLDV can measure the out-of-plane wave components which are well-suited for
studying anti-symmetric Lamb wave modes. Due to the large cost of the 3-D SLDV, most
researchers use the 1-D SLDV instead. Sohn et al. used a 1-D SLDV to visualize the
wave interaction with delamination damage in a CFRP composite plate, and observed the
“standing waves” trapped at the delamination damage [103]. Ruzzene [30] and Michaels
et al. [29, 105] measured full wavefield with a 1-D SLDV to detect delamination and
crack damage. They also compared the wavefield measured by the 1-D SLDV with the
24
wavefield measured by the traditional scanning air-coupled ultrasonic transducer (SAUT)
[29], as shown in Figure 2.13. The comparison shows the SAUT measurement has a
better signal-to-noise ratio (SNR), while the SLDV result exhibits less ringing and is
more broadband. The laser vibrometry also provides a means to validate guided wave
simulation. Olson et al. verified the FEM simulations of guided waves in aluminum
plates by using laser vibrometry [107]. Yu et al. verified out-of-plane EFIT simulation
results of guided waves in aluminum plates by using 1-D SLDV measurements [111].
Various wave features of EFIT simulations such as wavefield, group velocity and wave
amplitude, were verified by the SLDV measurements.
(a) (b)
(c) (d)
Figure 2.12 Visualization of wave interaction with a fatigue crack by using laser
vibrometry: (a) photos of the fatigue crack, (b) in-plane X-components, (c) in-plane Y-
components, and (d) out-of-plane Z-components [104].
25
(a) (b)
Figure 2.13 The experimental wavefields in a structure (fabricated from two 0.8 mm
thick aluminum plates that were epoxy-bonded together): (a) the wavefield at 45μs
measured by SAUT, and (b) the wavefield at 45μs measured by SLDV [29]. The
excitation was a wideband pulse with 40Hz repetition rate.
2.3. GUIDED WAVEFIELD DATA ANALYSIS AND DAMAGE DIAGNOSIS METHODOLOGIES
2.3.1. Guided wavefield analysis
Guided wavefields are multi-dimensional wave signals of guided waves in terms of time
(t) and space (x, y) dimensions. The SLDV can provide rapid and precise measurements
of guided wavefields with very high spatial resolutions to allow for multi-dimensional
analyses of guided waves. In contrast, guided wavefields acquired with densely
distributed sensors, such as wedge transducers and wafer type piezoelectric sensors, have
low spatial resolutions limited by sensor sizes.
Compared to the traditional 1-D waveform data with only the time argument, the
guided wavefields acquired from the SLDV extend to the time-space domain and include
additional spatial information to allow for obtaining more details of guided waves.
Depending on the number of spatial argument, the guided wavefields can be 2-D time-
space wavefields v(t, x) representing 1-D wave propagation or 3-D time-space wavefields
v(t, x, y) representing 2-D wave propagation. Hence, the guided wavefields include a
wealth of information regarding the wave propagation and interaction in both time and
space dimensions. Various methods have been used to analyze the guided wavefield, such
26
as wavefield amplitude profile, multi-dimensional Fourier transform, frequency-
wavenumber filtering, standing wave filtering, local wavenumber analysis, etc. [29-32,
103, 112-114]. Among them, the frequency-wavenumber analysis through multi-
dimensional Fourier transform provides information in terms of frequency-wavenumber
components, enabling wave mode identification, isolation, and damage detection, etc.
Staszewski et al. used amplitude profiles of guided wavefields to detect
delamination damage [112]. As illustrated in Figure 2.14, an amplitude profile directly
shows both the location and size of the delamination damage. Alleyne and Cawley used
2D FT to transform a guide wavefield into frequency-wavenumber domain where they
obtained detailed dispersion information of multi-modal Lamb waves [113]. Hayashi and
Kawashima used a filtering technique in frequency-wavenumber domain and successfully
separated A0 and S0 modes [114]. Ruzzene presented a damage visualization technique
by filtering the damage reflection waves in frequency-wavenumber domain and
demonstrated an application on crack damage visualization [30]. Michaels et al.
presented source removal and wave mode separation techniques by using frequency-
wavenumber filtering and demonstrated applications on delamination and crack detection
[29]. As illustrated in Figure 2.15, after source removal, delamination induced scattering
waves were highlighted. Sohn et al. proposed a “standing wave filter” which can isolate
standing wave components in a given wavefield for delamination detection in composites
[103]. As shown in Figure 2.16, after filtering, the standing waves at the delamination
were isolated indicating the location of the delamination. Rogge and Leckey presented a
local wavenumber analysis to process guided wavefields and demonstrated their method
can be used to quantify the depth and size of delamination damage [32]. Flynn et al.
27
presented a structural imaging method, which used frequency-wavenumber filtering and
local wavenumber estimation, and showed their method was able to detect wall-thinning
damage in an aluminum plate and a steel pipe, and delamination damage in a composite
structure [31].
Figure 2.14 Amplitude profile of a guided wavefield for delamination detection [112].
Delamination
(a) (b)
Figure 2.15 Source removal for delamination detection: (a) for a composite plate without
damage, and (b) for a composite plate with delamination damage [29].
(a) (b)
Figure 2.16 Standing wave filtering for delamination detection: (a) the wavefield before
filtering; (b) the wavefield after filtering [103].
28
2.3.2. Guided wave array imaging
In NDE/SHM applications, various guided wave array imaging methods have been
developed. In general, they can be categorized into two groups, the sparse array imaging
where sensors are distributed over the structure being inspected [23-26, 115], and the
phased array imaging where sensors are physically placed close to each other [12-21].
Both methods can visually show the damage location [12-21, 23, 115].
2.3.2.1. Guided wave sparse arrays
Guided wave sparse arrays employ sparsely distributed sensors over the region of interest
to inspect and monitor the structure [115]. The idea of the sparse array approach is to
minimize the number of required sensors and optimize the array configuration for
practical applications. The sparse array sensors work in the round-robin pattern, in which
each sensor can act in turn as the wave actuator to send interrogation waves and the rest
as the receivers to record diagnostic wave signals [115]. By using sparse array imaging
methods, such as delay-and-sum imaging, minimum variance ultrasonic imaging, and
time-reversal imaging, scanning images of the structure being inspected can be
constructed from the diagnostic wave signals. The scanning images can visually show the
damage location and may indicate the size [115-117].
In recent years, guided wave sparse arrays have been studied by many researchers
from various perspectives, such as array optimization, array imaging algorithm, and
sparse array applications. The guided wave sparse arrays for SHM were initially
proposed by Wang et al. [117]. In their approach, a synthetic time-reversal imaging
algorithm was used to process sparse array data to obtain an intensity image for locating
and sizing damage. Later, Michaels and Michaels presented an image fusion method to
29
process sparse array data for damage localization [118]. Their method used time shift
averaging to process the differential signals filtered at multiple frequencies, resulting in
multiple images of the component being inspected. Then, images generated at different
frequencies were fused to improve the damage localization accuracy and reduce the
background noise (Figure 2.17). Later, Michaels presented an in-situ sparse array for
damage detection, localization and characterization in plates (Figure 2.18a) [115]. This
approach used differential signals between diagnostic signals and baselines, and the
delay-and-sum algorithm to construct intensity images. Hall and Michaels proposed a
minimum variance distortionless response (MVDR) imaging whereby weighting
coefficients were adaptively computed at each pixel location, and showed the MVDR can
significantly improve the image quality compared to the delay-and-sum imaging (Figure
2.18b) [116]. The time-reversal, delay-and-sum, and MVDR methods all rely upon
baseline subtraction to separate scattered signals from direct arrivals and geometrical
reflections. However, in real applications, baseline data may not be available, or the
baseline data may have been recorded under mismatched environmental or operational
conditions. For such situations, Lee et al. developed a baseline-free imaging method
which was based upon estimating the source waveform and then adaptively removing the
direct arrival from each received signal [21]. As shown in Figure 2.19a, the imaging
result with raw received signals showed only the source. After adaptive source removal,
the imaging result (Figure 2.19b) with residual signals clearly showed the scatterer.
The sparse arrays have been used for various applications. Clarke et al. studied
the damage detection capability of a sparse array on a complex structure under varying
thermal conditions [23]. By using the signals after temperature compensation, they could
30
get good localization of holes with a 5-mm diameter or above in a complex structure. Yu
and Leckey explored the sparse array approach for crack detection in aluminum plates
[119]. With the sparse array, the crack tips could be clearly detected (Figure 2.20a). Su et
al. used sparse arrays to detect delamination damage in composite laminates [120]. Cho
and Lissenden used a sparse array on a multi-fastener joint for fatigue crack growth
monitoring [121]. Compared to phased arrays which use backscattering waves for
damage detection, the spare arrays can also take advantage of the so-called shadowing
effect or forward scattering waves, which are partially blocked by the damage [122].
y (mm)
y (mm)
(a) x (mm) (b) x (mm)

y (mm)
y (mm)
(c) x (mm) (d) x (mm)
Figure 2.17 Sparse array imaging results for the detection of a through hole with a
diameter of 3.2 mm: (a) result at 140 kHz, (b) result at 230 kHz, (c) result at 320 kHz,
and (d) the fused result. The “+” represents the location of the through hole[118].
31
y (mm)
y (mm)
(a) x (mm) (b) x (mm)
Figure 2.18 Sparse array imaging results for the detection of a 6 mm diameter through
hole: (a) the result of delay-and-sum imaging (10 dB color scale) [115], and (b) the result
of MVDR imaging (20 dB color scale) [116]. The open circles represent the transducers,
and the “+” represents the through hole.
y (mm)
y (mm)
(a) x (mm) (b) x (mm)
Figure 2.19 Baseline-free imaging: (a) delay-and-sum imaging of the source constructed
from the raw data, and (b) delay-and-sum imaging of the scatter constructed from signals
with direct arrivals removed via adaptive source removal. The open circles represent the
receivers [21].
Crack tips
300
200 (309,255)
y (mm)
(328,243)
100
0 100 200 300 400

x (mm)
Figure 2.20 Sparse array imaging for crack detection [119].
32
2.3.2.2. Guided wave phased arrays
The guided wave phased array employs sensors physically placed in a compact format in
which the relative phases of sensor signals are controlled in such a way that guided waves
(Figure 2.16 a) are focused into a wave beam in a desired direction (Figure 2.16 b) [1, 17,
123]. By steering the wave beam to all directions, the phased arrays can inspect and
monitor a large area of a structure. In NDE/SHM applications, the guided wave phased
arrays show multiple advantages such as reinforced wave energy in the beam, efficient
and flexible control of the beam direction, improved signal-to-noise-ratio, large area
inspection through a small sensing area, and promising damage detection results [12, 17,
124].
200 200
100 100
y (mm)
y (mm)
0 0
-100 -100
-200 -200
-200 -100 0 100 200 -200 -100 0 100 200
(a) x (mm) (b) x (mm)
Figure 2.21 Comparisons between omnidirectional guided waves and focused guided
waves: (a) omnidirectional guided waves, (b) the focused guided waves in 45° direction.
Guided wave phased arrays have been studied intensively [12-19, 125-130].
Giurgiutiu and Bao developed an embedded-ultrasonic structural radar (EUSR) with 1-D
linear phased arrays for in-situ NDE/SHM of thin-wall metallic structures [12]. They
showed the capability of 1-D linear phased array in detecting both broadside and offside
cracks in metallic plates (Figure 2.22). The 1-D linear arrays have shown that they can
successfully detect damage by performing large area scanning and diagnosis. However,
33
the 1-D linear arrays encounter some major shortcomings, such as (a) the degradations of
beamforming properties at angles close to 0° and 180° directions, and (b) the half-plane
mirror effect, which does not allow discriminating between a target placed above the
array and a target placed below the array (Figure 2.22c) [1].
y (m)
(a)
y (m)
x (m)
y (m)
(c) x (m)
(b) x (m)
Figure 2.22 EUSR detection results: (a) detection of a broadside crack, (b) detection of
an offside crack, (c) detection result showing the half-plane mirror effect [12, 17].
To overcome the shortcomings of 1-D linear arrays, 2-D planar arrays have been
adopted, which not only provide more control and optimization parameters for improving
the array performance, but also are able to perform full-range 360° beam steering and
reduce the mirror effect [1, 16, 17]. Wilcox studied omni-directional guided wave phased
arrays with 2-D circular configurations (Figure 2.23 (d) and (e)) for the rapid inspection
of large areas of aluminum plates [16]. In his work, phased arrays were constructed by
electromagnetic acoustic transducers. The measured time domain wave signals were
converted to wavenumber domain where the dispersion effect was compensated. After
the dispersion compensation, the phased addition algorithm in wavenumber domain was
used to achieve beam steering in different steering directions and to construct an omni-
34
direction, B-scan image. Yu and Giurgiutiu studied the 2-D phased arrays design and
implementation for damage detection in aluminum plates [17]. Various 2-D phased array
configurations were studied and compared, such as cross-shaped array, rectangular grid
array, rectangular ring array, circular ring array, and concentric circular array in Figure
2.23 (a)-(e), respectively. Yoo et al. developed a 2-D spiral phased array (Figure 2.23 (f))
by using piezoelectric-paint for SHM of thin isotropic panels [19]. Ambrozinski et al.
developed an efficient tool for the designing of 2-D phased arrays for isotropic plates
[131]. Koduru et al. proposed a 2-D phased annular array transducer (Figure 2.23 (f))
that can generate mode controlled omnidirectional guided waves in isotropic plate-like
structures by independently controlling the amplitude and phase of each array element
[132]. Senesi et al. designed a 2-D periodic frequency-steerable array (Figure 2.23 (g))
for SHM of aluminum plates. The array was composed of PZT discs in a quadrilateral
periodic topology. The beamsteering direction of the array can be controlled by the
excitation frequency [20].
x2 x2 x2
d1 d1 x2
d2
mth
d2
d2
d1 1
0
x1 x1 x1 x1
(a) (b) (c) (d)

x2 x2
x2 x2
mth x2
d1
1
x1
0 x2 d2
x1 x1 x1 x1
th
n ring
(e) (f) (g) (h)
Figure 2.23 2-D planar arrays: (a) cross-shaped array [17], (b) rectangular grid array
[17], (c) rectangular ring array [17], (d) circular ring array [16], (e) concentric circular
array [16], (f) spiral array [19], (g) annular array [132], and (h) quadrilateral periodic
array [20].
35
Although most research works focus on phased arrays for isotropic materials,
some researchers have investigated the phased arrays for composite materials [14]. Yan
and Rose studied beamsteering of linear arrays in composite plates [133]. They found the
traditional beamsteering technique for isotropic materials might fail in composite
materials due to the anisotropic behavior of the composite plates. Hence, they chose a
quasi-isotropic wave mode for beamsteering, which can suppress the influence of the
anisotropic behavior. Rajagopalan et al. adopted an array of a single transmitter and
multiple receivers (STMR) to locate a defect (hole) in a composite plate [134]. In their
array imaging method, they used a weakly anisotropic wave mode and assumed that the
phase and group velocity directions coincide locally. Later, Vishnuvardhan et al. used the
STMR array to detect impact induced delamination damage in a quasi-isotropic
composite plate. Leleux et al. used ultrasonic phased array probes for long range
detection of defects in composite plates [129]. Their method was limited to wave modes
having phase and group velocities oriented in the same direction, where the skew angle
was zero. Purekar and Pines investigated the capability of 1-D linear phased arrays in
detecting delamination damage in a cross-ply composite plate [14]. They showed the
array can detect damage at 0° direction, where the phase and group velocities had the
same orientation. Osterc et al. investigated the beamsteering of 1-D linear arrays in
composite laminates [135]. In their study, the exact phase velocity curve was used to
develop a beamforming algorithm that accounted for non-omnidirectional guided waves
in anisotropic materials.
Compared to the array beamforming in isotropic plates, the array beamforming in
anisotropic composite plates are more challenge. In anisotropic composite plates, guided
36
wave parameters such as wavenumber, phase velocity and group velocity are
directionally dependent, due to the directionally dependent physical properties of
composite materials [11, 136, 137]. Moreover, the guided waves have energy skewness
that the group velocity direction is not always parallel to the phase velocity direction.
Last but not least, the wave fronts of guided waves are no longer circular shapes. These
properties of guided waves in composite plates have significant influences on the phased
array beamforming. The traditional beamforming technique for isotropic materials might
fail in composite materials.
37
PART I: INTEGRATED SHM/NDE DIAGNOSIS METHODOLOGIES
CHAPTER 3
ANALYTICAL AND NUMERICAL STUDIES OF GUIDED WAVES
Analytical and numerical studies can provide a preliminary understanding of wave
propagations in complex structures and wave interactions with structural discontinuities,
theoretical data for assisting experimental work, and abundant information for parametric
studies. This chapter presents both analytical and numerical studies of guided waves in
complex structures. The GMM and the SAFE methods are used to obtain wave dispersion
curves and mode shapes, while the EFIT and the FEM methods are adopted to simulate
wave propagation and interaction.
3.1. DISPERSION AND MODE SHAPE DERIVATIONS
3.1.1. Water loaded plates
The classic problem of guided wave propagation in a traction-free homogeneous and
isotropic plate has been thoroughly studied and can be found in many references such as
[1, 8, 33, 138]. In a plate with traction-free boundary conditions (Figure 3.1 a), the
characteristic equation can be derived and expressed as [1, 8, 33]:
k S2 − ξ 2 k S2 − ξ 2 −2 k S ξ 2k S ξ
2k Lξ −2k Lξ k S2 − ξ 2 k S2 − ξ 2
k S2 − ξ 2 2k S ξ
( k S2 − ξ 2 ) g L −2 k S ξ g S =0 (3.2)
gL gS
−2k Lξ k S2 − ξ 2
2k Lξ g L ( k S2 − ξ 2 ) g S
gL gS
where,
38
ω2 ω2 2π
2
k=
L − ξ 2 , k=
2
S −ξ2, ξ = ,
c 2
L c 2
S λ
2 µ (1 − ν ) µ
cL = ,c = , g L = eik d , g S = eik d , d = 2h
L S
ρ (1 − 2ν ) S ρ
h, ω, ξ and λ are half plate thickness, circular frequency, wavenumber and wavelength,
respectively. cL and cS are longitudinal and transverse bulk wave velocities, respectively.
ρ, μ and ν are density, shear modulus and Poisson’s ratio, respectively. Solutions of
Eq.(3.2) can be found in many classic references such as [1, 8, 33, 138]. Figure 3.2 plots
frequency-wavenumber dispersion curves of a steel plate with 1.2 mm thickness. At
frequencies lower than 1000 kHz, there are fundamental anti-symmetric A0 and
symmetric S0 modes.
σ zzt = 0 σ zzt = 0
 t  t
x τ xz = 0 x τ xz = 0
Plate z Plate
d
z
d
σ zzb = 0 Semi-infinite half space water uzzb = uzzwater

 b  b
τ xz = 0 σ zz = σ zz
water
(a) (b)
 b
τ xz = 0
Figure 3.1 Sketches for: (a) free plate and (b) plate with one side in water.
wavenumber k (rad/mm)
A0
Real part of
S0
Frequency (kHz)
Figure 3.2 Wavenumber dispersion curves of a free steel plate with 1.2 mm thickness.
(Note only the real parts are given. The imaginary parts are zero.)
39
When one side of a plate is immersed in water (plate-half space water waveguide,
as shown in Figure 3.1 b), the top surface of the plate maintains traction-free boundary
conditions. On the bottom surface of the plate, wave energy can transmit into water since
the normal displacement uzz and stress σzz are continuous at the plate-water interface. The
shear stress τxz remains zero at the interface, since water does not support the shear stress.
The characteristic equation for the immersed plate can be assembled and expressed as,
k S2 − ξ 2 k S2 − ξ 2 −2 k S ξ 2k S ξ 0
2k Lξ −2k Lξ k S2 − ξ 2 k S2 − ξ 2 0
k S2 − ξ 2 2k S ξ ω2ρw
(k − ξ ) g L
2
S
2
−2 k S ξ g S
gL gS µ
=0 (3.3)
−2k Lξ k S2 − ξ 2
2k Lξ g L (k − ξ ) g S
2
S
2
0
gL gS
−k L −ξ
kL gL −ξ g S γ
gL gS
ω2 λw
γ
where, =2
− ξ 2 and cLw = . ρw, λw and cLw are density, bulk stiffness, and bulk
c 2
Lw ρw
wave velocity of water, respectively. By solving Eq. (3.3), the dispersion curves for the
1.2 mm thick steel plate with one side in water can be obtained. Due to the energy
leakage, the wavenumber roots of Eq. (3.3) have complex values with the real parts
indicating wave components and the imaginary parts giving the extent of energy
attenuation.
Figure 3.3 a plots the real parts of wavenumber roots for the immersed plate. The
fundamental symmetric and antisymmetric guided wave modes are present and are
notated as A 0WP and S0WP to distinguish them from the A0 and S0 modes in the free plate.
Furthermore, there is an additional mode that exhibits the same dispersive behavior as the
quasi-Scholte mode discovered in a two-side immersed plate [139], which is dispersive at
40
low frequencies and asymptotes to the non-dispersive Scholte mode at high frequencies
[139]. The one-side immersed steel plate in this study shows its capability to support the
quasi-Scholte mode as well.
Figure 3.4 gives mode shapes of the quasi-Scholte mode at various frequencies.
At 150 kHz, the mode shape in Figure 3.4 a shows large in-plane and out-of-plane
displacements in the plate. When frequency increases, both in-plane and out-of-plane
displacements in the plate decrease, as shown in Figure 3.4 b and c at 300 kHz and 600
kHz. At the high frequency of 600 kHz, the displacement is dominated by the in-plane
displacement in the water, and the displacement in the plate is almost zero.
Quasi-Scholte A0WP
Imaginary part of
Real part of
A0WP
S0WP
S0WP
Quasi-
Scholte
Figure 3.3 Wavenumber dispersion curves of a 1.2 mm thick steel plate with one side
in water: (a) real parts, (b) imaginary parts.
Plate
Water
z (mm)
z (mm)
z (mm)
(a) Normalized displacement (b) Normalized displacement (c) Normalized displacement
Figure 3.4 Theoretical wave mode shapes of the quasi-Scholte mode: (a) at 150 kHz, (b)
at 300 kHz, and (c) at 600 kHz.
41
3.1.2. Composite laminates
Composite laminates are multilayer structures. Knowing their dispersion curves is
important for selecting the appropriate wave mode for NDE/SHM in composites. The
SAFE method has shown its effectiveness and efficiency for deriving dispersion curves
and mode shapes [38]. For the development of NDE/SHM in composites, the SAFE
method is used to obtain dispersion curves and mode shapes of guided waves in CFRP
composite laminates. Figure 3.5 a gives a schematic of an 8-ply CFRP composite plate
with [02/902]s layup. In the SAFE model, the cross section of the composite plate is
discretized using 3 node quadratic elements, as illustrated in Figure 3.5 b. In each ply,
two elements are used for finite element approximation. By solving the discretized model
with the SAFE method, dispersion curves and mode shapes can be derived.
ux1
Ply 1 E#1 N#1
Ply 1 uy1
E#2 uz1
ux2
….....
N#2
uy2
[02/902]S uz2
layup
E#N-1 ux3
Ply 8 #N
N#3
Ply 8 E uy3
(a) (b) uz3
Figure 3.5 SAFE model of a CFRP composite plate with [02/902]s layup: (a) an
illustration of the composite plate, (b) an illustration of the finite element discretization.
Figure 3.6 plots the dispersion curves of guided waves propagating in 90°
direction in the composite plate. Figure 3.7 plots wavenumbers and group velocities in all
directions at 300 kHz. It is seen that both wavenumber and group velocity depend on
wave propagation direction in the composite plate.
Figure 3.8 gives the mode shapes for A0, S0 and SH0 modes at 300 kHz
propagating in 90° direction in the composite plate. For the A0 mode, the out-of-plane
42
displacement uz is dominated. For the S0 mode, the in-plane displacement uy is dominated.
For the SH0 mode, the in-plane displacement ux is dominated.
S0 S2
Group velocity (m/s)

A0
SH1 SH2 SH3 SH0

SH0 A0
S0 A1 S1 S2 A2 SH1 A1 S1 SH2 A2 SH3
Figure 3.6 Dispersion curves of guided waves propagating in 90° direction in the
[02/902]s CFRP composite plate: (a) frequency-wavenumber dispersion curves; (b) group
velocity dispersion curves.
(rad/mm) (m/s)
θ (°)
θ (°)
(a) (b)
Figure 3.7 Wavenumber and group velocity curves at 300 kHz for guided waves in the
[02/902]s CFRP composite plate: (a) wavenumber curves; (b) group velocity curves. The
solid, broken and dotted lines are for the A0, S0 and SH0 modes.
z (mm)
z (mm)
z (mm)
uy
ux
uz
(a) Normalized displacement (b) Normalized displacement (c) Normalized displacement
Figure 3.8 Mode shapes of guided waves at 300 kHz propagating in 90° direction in the
[02/902]s CFRP composite plate: (a) A0 mode, (b) S0 mode and (c) SH0 mode.
43
3.2. WAVE PROPAGATION AND INTERACTION SIMULATIONS
To simulate guided wave propagation and interaction, both the EFIT and the FEM
methods are used. With straightforward equations, EFIT computational codes are
developed using Matlab programming language. The EFIT method is used to simulate
guided waves in a water loaded plate and guided wave interaction with delamination
damage in a laminated composite plate. In addition, the commercial FEM software
ANSYS is adopted to simulate guided waves in a honeycomb sandwich structure.
3.2.1. EFIT for water loaded plates
Figure 3.9 plots the configuration of EFIT simulation for the study of guided waves in a
water loaded plate. A T304 steel plate with 1.2 mm thickness has one surface in water. A
PZT is installed on the top surface of the plate for guided wave generation. In the EFIT
model, the PZT excitation is simulated as pin forces at the edges of the PZT. By using the
EFIT method, guided waves in the water loaded plate are simulated.
Figure 3.10 plots EFIT simulation results (out-of-plane velocity vz) at 100 kHz
excitation. The simulation results show three wave components: quasi-Scholte waves,
leaky S0 waves and pressure waves (P-waves). The quasi-Scholte waves are present at the
plate-water interface and propagate along the interface. The leaky S0 waves in water are
induced by the energy leakage of plate S0 waves. In addition, the leaky S0 waves have
plane wave fronts and propagate faster compared to the P-waves which have circular
wave fronts.
44
PZT
x
O
Plate
1.2
z Water
Symmetric
100
boundary
Unit: mm
300
Figure 3.9 Configuration of EFIT simulation for the study of guided waves in a water
loaded plate.
Quasi-Scholte waves
0 1
P-waves
Normal. Amp.
0.5
Leaky S0 waves
z (mm)
50 0
-0.5
100 -1
0 50 100 150 200 250 300
(a) x (mm)
Quasi-Scholte waves
0 1
Normal. Amp.
P-waves 0.5
z (mm)
Leaky S0 waves
50 0
-0.5
100 -1
0 50 100 150 200 250 300
(b) x (mm)
Figure 3.10 vz velocity results of EFIT simulation at 100 kHz excitation: (a) at 40 μs and
(b) at 60 μs.
Figure 3.11 plots EFIT simulation results (out-of-plane velocity vz) at 600 kHz
excitation. The simulation results show three wave components: leaky A0 waves, leaky S0
waves, and P-waves. Compared to the results at the lower frequency 100 kHz, at the
higher frequency 600 kHz, quasi-Scholte waves at the plate-water interface disappear,
while leaky A0 waves present, which are induced by the energy leakage of plate A0 waves.
45
0 1
P-waves
Normal. Amp.
0.5
Leaky S0
z (mm)
50 0
Leaky A0
-0.5
100 -1
0 50 100 150 200 250 300
(a) x (mm)
Leaky A0
0 1
P-waves
Normal. Amp.
0.5
z (mm)
50 0
Leaky S0
-0.5
100 -1
0 50 100 150 200 250 300
(b) x (mm)
Figure 3.11 vz velocity results of EFIT simulation at 600 kHz excitation: (a) at 40 μs and
(b) at 60 μs.
3.2.2. EFIT for composite laminates
EFIT simulations are implemented to investigate guided wave interaction with
delamination damage. Figure 3.12 gives the simulation configuration of a CFRP plate
with [02/902]s layup. Under the second ply there is a 40 mm length delamination defect
with its center at x=100 mm. In the EFIT simulation, the delamination is modeled by
using stress free boundary conditions. Two PZT elements on the top and bottom surfaces
of the plate are used for wave generation and mode control. When the two PZT elements
are excited in-phase, symmetric modes will be generated. When the two PZT elements
are excited out-of-phase, anti-symmetric modes will be generated.
PZT #1 Delamination
x
O 0.1068
z
Symmetric
boundary
Unit: mm
PZT #2 200
Figure 3.12 The EFIT simulation configuration for the study of wave interaction with
delamination damage in a CFRP plate with [02/902]s layup.
46
Figure 3.13 plots in-plane vx and out-of-plane vz velocity components of the EFIT
simulation results when two PZT elements are excited out-of-phase and only an incident
A0 mode is actuated. Tabe 3.1 lists all the wave modes, including the incident A0 mode
and other modes induced by delamination damage observed from the simulation results.
These wave modes have been verified by comparing their wavelengths and group
velocities with their theoretical values.
At 40 μs, both the vx and vz components show only an incident A0 mode (incident
#1 in Figure 3.13 a and c). Since delamination damage separates the full [02/902]s
laminate to two laminates with different layups, one above the delamination with [0]2
layup and the other below the delamination with [(90)4/02] layup, it is expected that
guided waves are different in these two laminates.
As expected, simulation results at 70 μs show that the incident A0 mode separates
to different waves above and below the delamination damage. The in-plane vx plot shows
weak A0 (#3 in Figure 3.13 b) and strong S0 (#4 in Figure 3.13 b) modes supported by the
[0]2 laminate above the delamination. Below the delamination, an A0 mode (#5 in Figure
3.13 b) supported by the [(90)4/02] laminate is observed.
The out-of-plane vz plot at 70 μs shows an A0 mode (#3 in Figure 3.13 d) in the
[0]2 laminate above the delamination. Compared to the in-plane vx plot, the S0 mode in
the [0]2 laminate is not observed in the out-of-plane vz plot. The reason is that the out-of-
plane motion of the S0 mode is too weak to be observed from the image in Figure 3.13 d.
Below the delamination, an A0 mode (#5 in Figure 3.13 d) supported by the [(90)4/02]
laminate is observed, which agrees well with the A0 mode (#5 in Figure 3.13 b) observed
in the in-plane vx plot.
47
0 1 0 1
Normal. Amp.
Normal. Amp.
#3 #4
0.2 0.5 0.2 0.5
z (mm)
z (mm)
0.4 0 0.4 0
0.6 Incident #1 -0.5 0.6 -0.5
#5
0.8 -1 0.8 -1
0 50 100 150 200 0 50 100 150 200
(a) x (mm) (b) x (mm)
0 1 0 1
Normal. Amp.
Normal. Amp.
#3
0.2 0.5 0.2 0.5
z (mm)
z (mm)
0.4 0 0.4 0
l
0.6 Incident #1 0.6 -0.5

-0.5 #5
0.8 0.8 -1
N
-1
0 50 100 150 200 0 50 100 150 200
(c) x (mm) (d) x (mm)
Figure 3.13 EFIT results at 300 kHz for A0 mode interaction with delamination damage:
(a) and (b) are in-plane vx velocity plots at 40 μs and 70 μs; (c) and (d) are out-of-plane vz
velocity plots at 40 μs and 70 μs.
Table 3.1 Wave modes in EFIT results when incident waves are A0 mode
Number Wave Components
#1 A0 in the [02/902]s laminate
#3 A0 mode in the [0]2 laminate above the delamination
#4 S0 mode in the [0]2 laminate above the delamination
#5 A0 mode in the [(90)4/02] laminate below the delamination
Figure 3.14 plots in-plane vx and out-of-plane vz velocity components of EFIT
simulation results, when two PZT elements are excited in-phase and only an incident S0
mode is actuated. Table 3.3.2 lists all the wave modes, including the incident S0 mode
and other modes induced by delamination damage observed from the simulation results.
At 10 μs, an incident S0 mode (incident #2) is observed from the in-plane vx simulation
result in Figure 3.14 a, while the out-of-plane vz motion of the S0 mode is too weak to be
observed from the out-of-plane vz simulation result in Figure 3.14 c.
At delamination damage, the incident S0 mode interacts with the damage and is
separated to different waves above and below the damage, as shown in Figure 3.14 b. The
in-plane vx plot at 20 μs shows three strong components: an S0 mode (#4 in Figure 3.14 b)
mode supported by the [0]2 laminate above the delamination, an S0 mode (#6 in Figure
48
3.14 b) supported by the [(90)4/02] laminate below the delamination, and a transmitted S0
mode (transmitted #2 in Figure 3.14 b) in the full [02/902]s laminate. Note that there is no
A0 mode induced by the delamination observed in the in-plane vx plot (Figure 3.14 b),
since its in-plane motion is too weak.
The out-of-plane vz plot at 20 μs shows an A0 mode (#3 in Figure 3.14 d) in the
[0]2 laminate above the delamination and an A0 mode (#5 in Figure 3.14 d) in the
[(90)4/02] laminate below the delamination. Note that there is no S0 mode induced by the
delamination observed in the out-of-plane vz plot (Figure 3.14 d), since its out-of-plane
motion is too weak.
#4 Transmitted #2
0 1 0 1
Normal. Amp.
Normal. Amp.
0.2 0.5 0.2 0.5
z (mm)
z (mm)
0.4 Incident #2 0 0.4 0

0.6 -0.5 0.6 -0.5
#6
0.8 -1 0.8 -1
0 50 100 150 200 0 50 100 150 200
(a) x (mm) (b) x (mm)
#3
0 1 0 1
Normal. Amp.
Normal. Amp.
0.2 0.5 0.2 0.5
z (mm)
z (mm)
0.4 0 0.4 #5 0
l
0.6 -0.5 0.6 -0.5

0.8 0.8
N
-1 -1
0 50 100 150 200 0 50 100 150 200
(c) x (mm) (d) x (mm)
Figure 3.14 EFIT results at 300 kHz for S0 mode interaction with delamination damage:
(a) and (b) are in-plane vx velocity plots at 10 μs and 20 μs ; (c) and (d) are out-of-plane
vz velocity plots at 10 μs and 20 μs.
Table 3.2 Wave modes in EFIT results when incident waves are S0 mode
Number Wave modes
#2 S0 in the [02/902]s laminate
#3 A0 mode in the [0]2 laminate above the delamination
#4 S0 mode in the [0]2 laminate above the delamination
#5 A0 mode in the [(90)4/02] laminate below the delamination
#6 S0 mode in the [(90)4/02] laminate below the delamination
49
3.2.3. FEM for honeycomb sandwich structures
To simulate guided waves in honeycomb sandwich structures, the commercial FEM
software, ANSYS/Multiphysics 11.0 is used. The honeycomb sandwich investigated in
this study is composed of two aluminum skin plates and one aluminum honeycomb core.
The skin plates and honeycomb core have the same material properties, Young’s modulus
70 GPa, Poisson’s ratio 0.33, and density 2700 kg/m3.
PZT
Plates Core
Figure 3.15 The finite element model of the honeycomb sandwich structure with a
surface-bonded PZT.
Table 3.3 Geometry parameters of the finite element model. (Units: mm)
Skin panels Honeycomb core PZT actuator
Length Width Thickness Cell size Wall thickness Height Diameter Thickness
135 135 1 6.35 0.055 12.7 7 0.2
15 kHz 50 kHz 100 kHz
(a) (c) (d)
Figure 3.16 Simulation results of displacements in the z direction uz: (a) top and front
views at 15 kHz excitation, (b) top and front views at 50 kHz excitation, (c) top and front
views at 100 kHz excitation.
50
Figure 3.15 plots the finite element model of the honeycomb sandwich with a
surface bonded PZT. The geometry parameters of the model are listed in Table 3.3.3. The
SOLID45 elements are used to build the two skin plates. The SHELL63 elements are
used to build the honeycomb core that has hexagonal cells. The coupled field SOLID5
elements are selected to construct the finite element model of the PZT. Due to symmetry
of the problem, only a quarter-volume of the honeycomb sandwich is modeled.
Symmetric boundary conditions are applied on all the nodes on the two symmetric planes
(O-X-Z plane and O-Y-Z plane). The PZT excitations are 3 count tone burst signals with
center frequencies selected at 15kHz, 50kHz and 100kHz.
Transient analysis is used to simulate guided waves in the honeycomb sandwich.
In order to accurately simulate the guided waves that propagate in the honeycomb
sandwich, sizes of elements need to be small enough such that at least ten elements per
wavelength along the wave propagation direction [72]. In this study, the sizes of elements
are in the range 0.5 mm ~ 1.0 mm. Moreover, the integration time step should be small
enough for the stability of the transient analysis [72]. In this study, the time step is 0.2 μs.
Figure 3.16 gives the simulation results of displacements in the z direction uz at
15 kHz, 50 kHz and 100 kHz excitation frequencies. At the low frequency 15 kHz, the
simulation result in Figure 3.16 a shows that global guided waves propagate in the entire
sandwich. These global guided waves have large wavelengths and elliptical wave fronts.
In this case, the entire sandwich can be considered as a solid plate-like waveguide which
supports the global guided waves. At 50 kHz, although the global guided waves can still
be observed, their wave fronts become closer to circular shapes other than elliptical
shapes. At the high frequency 100 kHz, the global guided waves disappear. There are
51
guided waves propagating in skin plates with circular wave fronts. Moreover, the 100
kHz simulation result shows complex wave interactions, which could be induced by the
wave interaction with the honeycomb core.
52
CHAPTER 4
WAVEFIELD ACQUISITION AND INTRINSIC CHARACTERIZATION
Wavefields of guided waves in terms of time and space dimensions contain a wealth of
information regarding guided wave propagations in structures and interactions with
structural discontinuities. In this chapter, a wavefield sensing system is developed and
optimized to acquire high spatial resolution guided wavefields. To extract informative
wave signatures from the wavefields for wave characterization, wavefield analysis
methods, including frequency-wavenumber analysis, wavefield decomposition and space-
frequency-wavenumber analysis are presented.
4.1. HYBRID PZT-SLDV SENSING SYSTEM FOR WAVEFIELD ACQUISITION
The hybrid PZT-SLDV sensing system takes the advantages of PZT guided wave
generation and the SLDV non-contact high spatial resolution wavefield sensing. On one
hand, the low profile PZT is a commonly used guided wave actuator, which works on
piezoelectric principles and converts electric energy into mechanical energy and vice
versa [1]. When a PZT is bonded on a structure surface, it can couple its in-plane motion
with the structure through the bonding layer and generate guided waves [1]. The PZT
guided wave actuation and tuning mechanisms have been well established and proved
already [1, 88, 98-101]. In this dissertation study, PZT actuators (low profile: 5~10 mm
53
dimension and 0.2~0.5mm thickness) made of the piezo material APC 850 or 851 1 are
used.
On the other hand, the SLDV is a rapid, high spatial resolution and non-contact
means for guided wave sensing and visualization [29, 30, 140]. In this study, the Polytec
2
PSV-400-M2 scanning laser Doppler vibrometer is used. The particle
displacement/velocity in the direction of the laser beam is measured based on the Doppler
effect. The SLDV can direct the laser beam to various locations for measurements, by
employing two moving scanning mirrors driven by galvanometric actuators. For the
Polytec PSV-400-M2 used in this study, angular resolutions of the two scanning mirrors
are less than 0.002°.
Figure 4.1 gives a schematic of the PZT-SLDV sensing system. The PZT actuator
is surface bonded on the structure to generate guided waves. The excitation signal, such
as a signal of Hanning window modulated 3-cycle sine waves, is generated by an
arbitrary waveform generator and then amplified through a voltage amplifier. The SLDV
is used to measure velocities or displacements of guided waves at user-defined scanning
points. The SLDV acquires a waveform of guided waves (velocity or displacement as a
function of time) at a scanning point and then moves to the next point to acquire the next
waveform. The scan can be performed over user-defined scanning points in various
configurations, such as linear distribution for 1-D scanning and rectangular grid
distribution for 2-D scanning.
1
APC851 http://www.americanpiezo.com/apc-materials/choosing.html
2
Polytec PSV-400-M2 http://www.polytec.com/us/products/vibration-sensors/scanning-vibrometers/
54
Test specimen
Scanning grid Response DAQ & signal Full
processing wavefield
Crack SLDV head

Trigger
Voltage Excitation Function

PZT signal
amplifier generator
Figure 4.1 A schematic of the PZT-SLDV sensing.
Scanning line 3
250 Transmitted
2
Velocity (mm/s)
Distance (mm)
200
1
Crack
150 0
Incident Reflected -1
100
y
-2
x 50
PZT -3
0 50 100 150
(a) (b) Time (μs)
Transmitted
Scanning area
Reflected
Velocity (mm/s)
Crack
Time t (μs)
y
Incident
PZT x
y (mm)
(c) (d) x (mm)
Figure 4.2 Crack detection by using the PZT-SLDV sensing: (a) and (b) are the
scanning configuration and the measured time-space wavefield for the 1-D scanning; (c)
and (d) are the scanning configuration and the measured time-space wavefield for the 2-D
scanning.
The hybrid PZT-SLDV sensing system can be used to visualize guided wave
interactions with structural discontinuities. Moreover, the sensing system provides an
immediate means of damage detection. Figure 4.2 gives an example of using the hybrid
PZT-SLDV sensing system for crack detection in a 1 mm thick aluminum plate. Figure
4.2 a and c gives the configurations of 1-D scanning with points along a line and 2-D
scanning with points in a scanning area. The guided waves are generated from a PZT
55
(APC 850: 0.2 mm thickness and 7 mm diameter) bonded on the plate surface. The
excitation signal is a 3-cycle tone burst with the frequency of 120 kHz and amplitude of
40 V. Figure 4.2 b and Figure 4.2 d plot the time-space wavefields measured from the 1-
D scanning and the 2-D scanning. Both wavefields show three wave components:
incident, reflected and transmitted waves. When incident waves arrive at the crack
damage, reflected and transmitted waves occur. From the guided wavefields, the complex
wave propagation and interaction are visualized. Moreover, the crack location can be
qualitatively indicated from the observed wave behaviors.
Figure 4.3 gives another example of applying the hybrid PZT-SLDV sensing
system for delamination detection. The test specimen is an 8-ply IM7/8552 composite
plate with [02/902]s layup. Under the second ply, there is a 20 mm × 20 mm Teflon film
embedded during the manufacturing process to create delamination damage. Figure 4.3 a
and c give 1-D and 2-D scanning configurations. Guided waves are generated from a PZT
patch (APC 850: 0.2 mm thickness and 7 mm diameter) bonded on the plate surface. The
excitation signal is a 3-cycle tone burst with the frequency of 300 kHz and amplitude of
40 V. Figure 4.3 b plots the time-space wavefield measured from the 1-D scanning,
showing that guided waves interact with the delamination. Guided waves propagating
back and forth between the top (y=20 mm) and bottom (y=40 mm) boundaries of the
delamination, can be distinctively seen, exhibiting the “trapped wave” phenomenon [103,
141-143]. Meanwhile, the wavefield image at 15 μs (in Figure 4.3 d) acquired by the 2-D
scanning, shows wave interaction at the delamination region, as waves are propagating
between delamination boundaries (denoted by the dotted box). Therefore, with the PZT-
SLDV sensing, the complex wave interaction with delamination damage is visualized.
56
60 6
Scanning line
50 4
Velocity (mm/s)
Delamination 40 2
y (mm)
0
30
-2
y 20
x Wave interaction with -4
PZT delamination damage
10
-6
0 10 20 30
(a) (b) Time (μs)
60 6
Wave interaction with
50 4
delamination damage
Scanning area
2
40
y (mm)
0
30
-2
y 20
x -4
PZT 10
-6
-20 -10 0 10 20
(c) (d) x (mm)
Figure 4.3 Delamination detection by using the PZT-SLDV sensing: (a) and (b) are the
scanning configuration and the measured time-space wavefield for the 1-D scanning; (c)
and (d) are the scanning configuration and the measured time-space wavefield for the 2-D
scanning.
4.2. FREQUENCY-WAVENUMBER ANALYSIS
Although guided wavefields in time-space domain can visually show guided wave
propagation and interaction, the detailed wave signatures such as wave modes,
frequencies and wavenumbers are not available. To achieve these detailed wave
signatures for further analysis, the guided wavefields in time-space domain can be
converted to the representation in frequency-wavenumber domain by using the three
dimensional (3-D) Fourier Transform (FT), given as [29, 30, 113, 140, 142, 144]
∞ ∞
V ( f , k ) =  [ v (t , x ) ] = ∫ v (t , x )e − (
i 2π ft − k ⋅x )
−∞ −∞ ∫ dtdx (4.4)
where the space vector x and the wavenumber vector k are defined as (x, y) and (kx, ky).
v(t, x) is the time-space wavefield in terms of time t and space vector x. V(f, k) is the
57
resulting frequency-wavenumber representation or “spectrum” in terms of frequency f
and wavenumber vector k. As the frequency is the counter part of time, the wavenumber
is the counter part of space.
With the spectrum in frequency-wavenumber domain, the frequency,
wavenumber, wave mode and wave dispersion can be readily achieved and analyzed.
Figure 4.4 gives an example of frequency-wavenumber analysis on guided waves
generated by a PZT actuator on a 6.35 mm thick T304 steel plate. The excitation signal is
a broadband pulse with 0.8 μs duration and 10 V amplitude. The measured time-space
wavefield is plotted in Figure 4.4 a, which shows complex wave propagation and
interaction. From the wavefield in time-space domain, it is hard to identify the wave
mode, frequency, and wavenumber. However, these components can be easily achieved
in frequency wavenumber domain. As plotted in Figure 4.4 b, the frequency-wavenumber
spectrum reveals the frequency, wavenumber and wave mode. Within 0 to 800 kHz, in
the frequency-wavenumber spectrum, six wave modes including A0, A1, A2, S0, S1 and S2
are discernible and identifiable. This information is otherwise unavailable in the time-
space wavefield.
140 1
Wave number kx (rad/mm)
120
Normalized amplitude
0.5
x position (mm)
Velocity (mm/s)
100
80
0 A1
60 S1
40 -0.5
A0
20
S0 S2 A2
-1
0 10 20 30 40 50 60 70
(a) Time (μs) (b) Frequency (kHz)
Figure 4.4 Frequency-wavenumber analysis on guided waves in a 6.35 mm thick T304

steel plate: (a) time-space wavefield, and (b) frequency-wavenumber spectrum.
58
4.3. GUIDED WAVEFIELD DECOMPOSITION
By using FT, time-space wavefields are transformed to frequency-wavenumber spectra,
where different wave modes and wave components can be easily identified. Therefore,
any wave component of interest can be extracted for further analysis, through a filtering
process in the frequency-wavenumber domain. Similar to the digital filtering [145], the
filtering process in frequency-wavenumber domain is mathematically expressed as the
product between the original frequency-wavenumber spectrum V(f, k) and the filter
function F(f, k) as:
VF ( f , k ) = V ( f , k ) F ( f , k ) (4.5)
where VF(f, k) is the filtered spectrum, which keeps the desired frequency-wavenumber
component. By transforming the filtered spectrum VF(f, k) back into time-space domain
through inverse 3-D FT, the filtered wavefield vF(t, x) can be obtained as
1 ∞ ∞
−1
vF (t , x ) =
= [VF ( f , k )] ∫ ∫ VF ( f , k )e
j ( 2π ft − k ⋅x )
dfdk (4.6)
(2π ) 2 −∞ −∞
With the frequency-wavenumber filtering, guided waves which have multiple wave
modes that propagate in various directions can be decomposed into individual modes. To
obtain a desired wave mode, the band pass filter is designed to keep the frequency-
wavenumber component of the wave mode,
F ( f , k ) = w f ( f ) wk ( f , k ) (4.7)
where w f ( f ) is a raised-cosine filter in frequency domain to extract the frequency
component of the desired wave mode. wk ( f , k ) is a 3-D filter centered at the theoretical
wavenumber value k(f) of the desired wave mode to be extracted, as:
59

Frequency-wavenumber
for S2 mode
S2

140 0.15
120 0.1
Velocity (mm/s)
S2
Velocity (mm/s)
100
x position (mm)
S2 0.05
80
0
60
-0.05
40
-0.1
(d) Time (μs)
20
0 10 20 30 40 50 60 70
(c) Time (μs)
Figure 4.5 Guided wave decomposition by using frequency-wavenumber filtering: (a)

frequency-wavenumber filter for the S2 mode; (b) filtered frequency-wavenumber
spectrum; (c) reconstructed time-space wavefield of the S2 mode; (d) reconstructed
waveform of the S2 mode at x=40 mm.
   D ( f , k ) 
0.5 1 + cos  2π k   if Dk ( f , k ) ≤ DWk / 2
wk ( f , k ) =    DWk   (4.8)

 0 otherwise
with
Dk ( f , k=
) k − k( f ) (4.9)
where DWk is the wavenumber bandwidth.
Figure 4.5 shows an example of guided wave decomposition. The original time-
space wavefield in Figure 4.4 a, contains six wave modes including A0, A1, A2, S0, S1 and
S2. To extract the S2 mode, a band pass filter (Figure 4.5a) is designed, which only keeps
the frequency-wavenumber component of S2 mode and rejects everything else. Through
60
the filtering process in frequency-wavenumber domain, the S2 mode within the frequency
range 500~750 kHz and the wavenumber range 0~1 rad/mm is extracted (Figure 4.5b).
Then the extracted frequency-wavenumber spectrum of the S2 mode is reconstructed back
into time-space domain. Figure 4.5 c and d plot the reconstructed wavefield and the
waveform at x =40 mm, which only contain the S2 mode.
4.4. SPACE-FREQUENCY-WAVENUMBER ANALYSIS
The wavefield analysis in frequency-wavenumber domain can reveal wave propagation
characteristics that cannot be explicitly seen in time-space domain. However, after the FT,
spatial information of guided wavefields is lost. It will be beneficial if the spatial
information can be retained. To retain the spatial information, the space-frequency-
wavenumber analysis is used.
The space-frequency-wavenumber analysis is a straightforward extension of the
short time FT [146] to multi-dimensional problems. It breaks down a time-space
wavefield into small segments over the spatial dimension before the Fourier
transformation. Hence, it is also called short space FT. The small segment is acquired by
multiplying the wavefield with a window function which is non-zero for only a short
period in space while constant over the entire time dimension. When the window slides
along the space dimension, a set of windowed wavefield segments is generated. The FT is
applied to these segments, resulting in a set of frequency-wavenumber spectra that are
indexed by window locations such that the spatial information is retained.
Mathematically, the short space FT is given as
∞ ∞
S ( x, f , k )
= ∫ ∫
−∞ −∞
u(t , x )W * (t , x − x )e − j (2π ft −k⋅x ) dtdx (4.10)
61
where x is the retained spatial vector ( x , y ) of the windowed segments. W(t, x) is the
window function that keeps constant in time and varies in space. For example, a Hanning
function can be used to construct the window W(t, x), given as:
   x 
0.5 1+ cos  2π   if x ≤ Dx 2
W (t , x ) =    Dx   (4.11)

 0 otherwise
where Dx is the window length in space domain. Figure 4.6 gives an example of the
window function W(t, x). Note that in this example, only the dimensions of time t and
space x are included for plotting. As shown in Figure 4.6, the window function keeps
constant in the time dimension and varies in the space dimension.
Since the short space FT retains the spatial information, the relation between
space and wavenumber can be investigated. Figure 4.7 gives an example of using the
short space FT to analyze a guided wavefield and show how wavenumber components
vary in space. The time-space wavefield in Figure 4.7 a shows a forward propagating A0
mode. By applying a window function located at x = 25 mm, a small segment of the
wavefield is isolated, as plotted in Figure 4.7 b with its frequency-wavenumber spectrum
in Figure 4.7 c. In the short space FT process, the window slides through the wavefield
from 0 to 270 mm with a step size of 0.5 mm resulting in a set of windowed wavefield
segments. All the segments are then transformed to frequency-wavenumber domain and
stacked up, generating a space-frequency-wavenumber representation. Figure 4.7 d
presents the space-wavenumber spectrum at a selected frequency 120 kHz. The spectrum
is centered about the wavenumber 0.73 rad/mm and long the entire spatial extant. When
the wave propagates away from the source (increasing of x) the amplitude in the space-
wavenumber spectrum drops gradually.
62
Amplitude
x (mm)
t (μs)
Figure 4.6 An example of the window function.
Incident A0
Velocity (mm/s)
Velocity (mm/s)
x (mm)
x (mm)
Incident A0

Incident A0
Incident A0
x (mm)
(c) Frequency (kHz) (d) Wave number (rad/mm)
Figure 4.7 An example of the space-frequency-wavenumber analysis: (a) the original

time-space wavefield; (b) windowed wavefield; (c) frequency-wavenumber spectrum of
the windowed wavefield; (d) space-wavenumber spectrum at a selected frequency 120
kHz.
63
CHAPTER 5
QUANTITATIVE AND VISUALIZED DAMAGE DIAGNOSIS
This chapter presents quantitative and visualized damage diagnosis methods for
NDE/SHM. The generic formation of phased array imaging for anisotropic composite
laminates is developed, based on classic delay-and-sum principle. The phased array
imaging can generate an intensity image of the composite been inspected to visualize and
locate the damage in the composite. To further quantify the damage size, wavefield based
imaging methods such as filter reconstruction imaging and spatial wavenumber imaging
are used. The systemic combination of the phased array imaging and wavefield based
imaging provides a hierarchical damage diagnosis, which can efficiently and effectively
locate the damage and quantify the damage size.
5.1. GUIDED WAVE PHASED ARRAYS IMAGING
Guided wave phased arrays employs sensors physically placed in a compact format in
which the relative phases of sensor signals are controlled such that guided waves are
focused into a beam in a desired direction [1]. By steering the focused beam to all
directions, the phased array can inspect a large area of a structure to detect damage. In
NDE/SHM applications, the guided wave phased array has advantages, such as
reinforced wave energy in the beam, efficient and flexible control of the beam direction,
improved signal-to-noise-ratio, large area inspection through a small sensing area, and
promising damage detection results [12, 17, 124].
64
Compared to the array beamforming in isotropic plates, the array beamforming in
anisotropic composite plates are more challenging. In anisotropic composite plates,
guided wave parameters such as wavenumber, phase velocity and group velocity are
directionally dependent due to the directionally dependent physical properties of
composite materials [11, 136, 137]. Moreover, the guided waves have energy skewness
and the group velocity direction is not always parallel to the phase velocity direction.
Last but not least, the wave fronts of guided waves are no longer circular shapes. These
properties of guided waves in composite plates have significant influences on the phased
array beamforming. The traditional beamforming technique for isotropic materials might
fail in composite materials.
In this section, the guided wave beamforming in anisotropic composite plates is
investigated. Based on the classic delay-and-sum principle, the formation of
beamforming in anisotropic composite plates is developed, in which the directional
dependent guided wave parameters are considered. This formulation is valid not only for
the case that phase and group velocities have the same orientation, but also for the
general case that the two velocities are in different directions. Moreover, the array
characterization and optimization for beamforming in composite plates are investigated
with demonstrations of 2-D rectangular grid arrays. For the proof of concept,
experimental tests are performed by using rectangular grid arrays made of scanning
points of a non-contact SLDV. The experimental results show that multiple defects in
various directions can be successfully detected.
65
5.1.1. Beamforming in composite laminates
5.1.1.1. Guided waves in composite laminates
In anisotropic composite laminates, guided wave parameters such as wavenumber, phase
velocity and group velocity are directionally dependent, due to the directionally
dependent physical properties of composite materials [11, 136, 137]. Figure 5.1 a plots
the wavenumber curve k(γ), and shows the geometry relation between the wavenumber
vector k which is perpendicular to the wave front and the group velocity vector cg which
is orthogonal to the wavenumber curve [11, 136, 137]. The angle of wavenumber vector
k is γ, referred to as wavenumber angle. The angle of group velocity vector cg is θ,
referred to as group velocity angle (or energy propagation angle). The angle β between k
and cg is referred to as skew angle with β = γ − θ. When the skew angle β is not zero, cg is
not parallel to k, and hence the wave energy propagation direction is not perpendicular to
the wave front.
Wa
ve
fro
cg nt Propagation direction
ky
β of wave energy
Ta
x
ng
en
θ
t
k
Wa
ve
fro
𝛾𝛾 nt
o β
kx y
𝛾𝛾
β θ
𝛾𝛾 Source at pm
(a) (b) θ
Source at O x
Figure 5.1 Schematics of geometric relations of guided waves in composite plates: (a)
the geometric relation between the wavenumber vector k and the group velocity vector cg,
and (b) the geometric relation of guided waves (in the far field) generated by sources at
different locations.
66
When a source located at the coordinate origin O generates a guided wave with
frequency ω and wavenumber k , the wave arriving at a location x that is far away from
the source (Figure 5.1 b) can be expressed as [1, 8, 11]
u(t , x ) = Ae j(ωt −k⋅x ) (5.12)
where A is the amplitude. With the geometric relation illustrated in Figure 5.1 b, we
have
k ⋅ x = k x cos β =
k (γ ) x cos β (5.13)
And hence,
j ωt − k ( γ ) x cos β 
u(t , x ) = Ae (5.14)
For a source located at pm, the generated wave at the location x is
j[ωt − k ⋅( x − pm )] j ωt − k ( γ ) x − pm cos β 

um (t , x ) Ae
= = Ae (5.15)
5.1.1.2. Delay-and-sum beamforming
Consider an array with M identical elements located at {p m } (m=0, 2, 3, …M-1) which
are physically close to each other. The phase center is set as the origin O of the Cartesian
∑
M −1
coordinate system, i.e., 1
M m =0
p m = 0 . Each element serves as a guided wave source.
When all elements generate a guided wave with frequency ω and wavenumber vector k at
the same time, the total output of the array at the location x can be derived as
z (t , x ) = ∑ Ae j[ωt −k⋅( x −pm )] =u(t , x )∑

M −1 M −1
=m 0=m 0
e jk⋅pm (5.16)
It is seen from Eq. (5.16) the synthetic wave is an amplification of the wave emitting
from a source at the Origin and the amplification is controlled by the exponential
components in Eq. (5.16), which can be maximized when the exponent become zero. The
exponent becomes zero when a delay Δm is applied to result in:
67
j ( k ⋅pm −∆ m )
e = e j0 (5.17)
The delay Δm relates to the mth element position and the wavenumber vector k. By
judicially selecting Δm, the total output Z(t, x) of the array can be maximized at a desired
direction, i.e., generating a directional “beam”. In principle, the delay Δm can be
implemented through time delay in the time domain or phase shift in the frequency
domain [1, 16, 17, 144]. In our study, the latter is adopted. In addition to delay, weighting
factors wm can also be applied to the output of individual delayed wave to improve the
output. In a summary, the beamforming can be represented as
z (t , x) = u (t , x)∑ m =0 wm e
M −1 j ( k ⋅p m −∆ m )
(5.18)
Assume the array output is directed toward a desired direction θS and the corresponding
delay is denoted as Δm(θS). We introduce the beamforming factor [1, 17], BF, which is
given as
1 j  k ⋅pm −∆ m (θ S ) 
∑
M −1
BF = wm e  
(5.19)
M m =0
And Eq. (5.18) can be re-written as
z (t , x ) =⋅
M u(t , x ) ⋅ BF (5.20)
To maximize the BF in the desired direction θS, the delay Δm(θS) should be selected to
result in zero exponent i.e.,
∆ m (θ S ) =
k ⋅ pm (5.21)
The wavenumber vector k here depends on the wave frequency ω and wavenumber angle
γ S that corresponds to the wave energy steering angle θS, i.e., k can be expressed as
k (ω , γ S ) . Based on the geometry relation that γ S = θ S + β S , Eq. (5.21) can be written as
∆ m (θ =
S) k (ω,θ S + β S ) ⋅ p m (5.22)
68
where βS is referred to as the skew angle between the wavenumber angle γ S and the wave
energy steering angle θS. With the phase delay, it can be seen that the total output z(t, x)
∑
M −1
is m =0
wm times reinforcement of the individual wavefield u(t, x) in the θS direction.
The weighting factor wm can be used to further enhance the synthetic beam shape and
reduce sidelobe levels [1, 144].
By substituting Eq. (5.22) into Eq.(5.19), the beamforming factor becomes,
1
∑ wm e j[k −k (ω ,θS + β S )]⋅pm
M −1
BF ( k | wm , p m , M ,θ S ) = (5.23)
M m =0
The beamforming factor as given in Eq. (5.23) can be interpreted as a function of
wavenumber vector k, denoted as, BF ( k ) in the wavenumber domain. wm , p m , M and
θ S are the parameters of the wavenumber function BF ( k ) . For 2D wave propagation,
BF ( k ) evaluates the beamforming result at any wavenumber vector k in the kx-ky
wavenumber plane.
The beamforming can also be represented as a function of the wave energy
propagation angle θ as,
1
∑ wm e j[k (ω ,θ + β )−k (ω ,θS + β S )]⋅pm
M −1
BF (θ | wm , p m , M ,θ S ) = (5.24)
M m =0
where k(ω, θ+β) is the wavenumber dispersion relation of guided waves in composite
laminates. The BF(θ) evaluates the beamforming output w.r.t. the wave energy
propagation angle θ for the guided waves with the wavenumber dispersion relation k(ω,
θ+β). Hence it can directly indicate the phased array’s directional beamforming pattern.
69
5.1.2. Implementation and characterization of rectangular arrays
Phased arrays can be configured with various arrangements, such as linear arrays[12],
rectangular arrays[17], and spiral arrays[19]. In this work, rectangular arrays with array
elements being uniformly arranged on rectangular grids are investigated. Figure 5.2
illustrates a P×Q rectangular array (total of M= P×Q elements) with its phase center
coinciding with the coordinate origin O. The coordinates of the (p, q)th (p = 0, 1, 2, …P-1
and q = 0, 1, 2, …Q-1) element of the array are
 P −1 Q −1 
p p ,q =
( p − )d x ,( q − )d y  (5.25)
 2 2 
where dx and dy are array spacings in x and y directions. The array spans in x and y
directions are
D=
x ( P − 1)d x and D=
y (Q − 1)d y (5.26)
Q-1
O x
dy
0 P-1
dx
Figure 5.2 Schematic of a P×Q rectangular array
For the beamforming study in composite material, an 8-ply [0/45/90/-45]s layup
CFRP composite plate with 2.54 mm thickness (material properties in Table 5.5.1) is
used. The A0 Lamb mode at 120 kHz (wavenumber curve k(γ) is plotted in Figure 5.3) is
used for the beamforming, with the maximum wavenumber components in kx and ky
directions being kx,max = 0.55 rad/mm and ky,max= 0.68 rad/mm, resulting in the minimum
wavelengths 𝜆𝜆x,min= 2π/kx,max = 11.4 mm and 𝜆𝜆y,min = 2π/ky,max = 9.2 mm. To study the
70
beamforming, three array configurations listed in Table 5.5.2 are considered. Note (a)
arrays (1) and (2) have the same spacings, while the spans in array (2) are twice of those
in array (1); (b) Arrays (2) and (3) have the same spans, while the spacings in array (3)
are twice of those in array (2). In the following subsections, both beamforming factor
BF(k) in Eq. (5.23) and beam pattern BF(θ) in Eq. (5.24) will be used to evaluate the
beamforming of each array.
Ta cg
1 ng β
en
t
θ
0.5
k Wa
ve
fro
ky (rad/mm)
𝛾𝛾 nt
0
-0.5
-1
-1 -0.5 0 0.5 1
kx (rad/mm)
Figure 5.3 Wavenumber curve k(γ) of the 90 kHz A0 mode in the [0/45/90/-45]s CFRP
composite laminate.
Table 5.1 Material properties of a single ply.

ρ E1 E2 E3 G12 G13 G23 Thickness
𝜈𝜈 12 𝜈𝜈 13 𝜈𝜈 23
(kg/m3) (GPa) (GPa) (GPa) (GPa) (GPa) (GPa) (mm)
1577.8 129.6 8.7 8.7 4.8 4.8 2.9 0.30 0.30 0.34 0.3175
Table 5.2 Configurations of three rectangular arrays for beamforming study

Number of elements Spacing Span
P×Q d (mm) D (mm)
Array (1) 11×11 dx,1= dy,1=𝜆𝜆y,min/4=2.3 Dx,1= Dy,1=23
Array (2) 21×21 dx,2= dy,2=𝜆𝜆y,min/4=2.3 Dx,2= Dy,2=46
Array (3) 11×11 dx,3= dy,3=𝜆𝜆y,min/2=4.6 Dx,3= Dy,3=46a
71
5.1.2.1. BF(k) characterization
The array performance is studied by the beamforming factor BF(k) in Eq.(5.23). By
substituting the array coordinates in Eq. (5.25) into Eq. (5.23), we can derive the BF(k)
equation for rectangular arrays,
 P −1 Q −1 
1 j[ k − k ( ω ,θ S + β S )]⋅ ( p − ) d x ,( q − )d y 
∑ ∑ p ,q
P −1 Q −1
BF ( k | w p ,q , d x , d y , P, Q,θ S ) = w e  2 2 
(5.27)
PQ =p 0=q 0
By substituting the array spans in Eq.(5.26) into Eq.(5.27), the BF(k) becomes
D Dy  Dx Dy 
j[ k − k ( ω ,θ S + β S )]⋅ pd x − ,qd y −
1 x

2  (5.28)
∑ ∑
dy 2
BF ( k | w p ,q , d x , d y , Dx , D y ,θ S ) = dx
w p , q e 
 Dx   Dy  =p 0=q 0
d + 1 
 d + 1 
 x  y 
This equation shows the beamforming factor BF(k) is a function of k defined up to the
parameters w p ,q , d x , d y , Dx , D y and θ S .
Figure 5.4 a graphically plots the BF(k) of array (1) as an intensity image in the
kx-ky wavenumber plane, when phase delays are not applied. Four intensified spots at (0,
0), (0, 4ky,max), (4ky,max, 0) and (4ky,max, 4ky,max) are present in the given region. The sizes
of intensified spots are evaluated by using the full widths of the intensified spots at one-
half peak value (denoted as FWHMx and FWHMy for kx and ky directions) [144], which
are listed in Table 5.5.3. The intensified spots represent local maxima of BF(k). When the
array generates guided waves of wavenumbers at these local maxima, the array’s output
will be optimized. Note that Figure 5.4a only plots the BF(k) in the wavenumber range -
1.0 rad/mm ≤ kx ≤ 3.7 rad/mm and -1.0 rad/mm ≤ ky ≤ 3.7 rad/mm. Indeed the intensified
spots present in a periodical pattern within the entire wavenumber domain. The periodical
pattern can be perceived from the expression of the beamforming factor BF(k) in
72
Eq.(5.28). The periods Kx and Ky of the periodical pattern in kx and ky directions are listed
in Table 5.5.3.
The wavenumber curve k(𝛾𝛾) of the 120 kHz A0 mode in the subject composite
plate is also plotted in Figure 5.4a. It can be seen that no maximum of BF(k) falls on the
wavenumber curve. This signifies that if the array generates the 120 kHz A0 mode
without delaying, the array will not have maximized output. Hence, array’s output can be
maximized by applying phase delays given in Eq. (5.22), i.e., applying phase delays that
relocate a maximum of BF(k) on the wavenumber curve. Figure 5.4b plots the BF(k) of
the array (1) with the phase delays (-0.02, 0.68)∙pp,q. The delay is selected such that the
local maxima located at (0, 0) before the delay is now moved to (-0.02, 0.68) rad/mm on
the wavenumber curve.
Figure 5.5 a and b plot the BF(k) images of arrays (2) and (3), when the phase
delays (-0.02, 0.68)∙pp,q are applied. Their FWHM values are listed in Table 5.5.3, in
addition to those of array (1). FWHM values of arrays (2) and (3) (with the same spans
Dx,2= Dy,2= Dx,3= Dy,3=46 mm) are the same and smaller than those of array (1) (with
smaller spans Dx,1= Dy,1=23 mm). This means larger array spans give smaller size
intensified spots and better wavenumber resolutions.
73
Kx,1= 4ky,max Kx,1= 4ky,max
3 3
Ky,1= 4ky,max
2 2
Ky,1=4ky,max
Amplitude
ky (rad/mm)
ky (rad/mm)
Amplitude
Intensified Intensified spot at
spot at (0, 0) (-0.02, 0.68)rad/mm on k(γ)
1 1
0 0
k(γ) k(γ)
-1 -1
-1 0 1 2 3 -1 0 1 2 3
(a) kx (rad/mm) (b) kx (rad/mm)
Figure 5.4 BF(k) images for the array (1): (a) without phase delay, and (b) with the
phase delay (0, ky,max)∙pp,q . The solid white curve is the wavenumber curve k(𝛾𝛾) of the
120 kHz A0 mode in the [0/45/90/-45]s CFRP composite plate.
Kx,2= 4ky,max 3
3 Kx,3=2ky,max
Ky,2= 4ky,max
2 2
Ky,3=2ky,max
ky (rad/mm)
Amplitude
Amplitude
ky (rad/mm)
Intensified spot at Intensified spot at

(-0.02, 0.68)rad/mm on k(γ) (-0.02, 0.68)rad/mm on k(γ)
1 1
k(γ)
0 k(γ) 0
Intensified spot at
(-0.02, -0.68)rad/mm on k(γ)
-1 -1
-1 0 1 2 3 -1 0 1 2 3
Figure 5.5 BF(k) images when the phase delays (-0.02, 0.68)∙pp,q is applied: (a) for
array (2), and (b) for array (3). The solid white curve is the wavenumber curve k(𝛾𝛾) of the
120 kHz A0 mode in the [0/45/90/-45]s CFRP composite plate.
Table 5.3 BF(k) characteristics of the three rectangular arrays

Wavenumber periods, Full widths of intensified spots at the
K (rad/mm) one-half peak value, FWHM (rad/mm)
Array (1) Kx,1= Ky,1=2π/dx,1=2π/dy,1=4ky,min=2.72 FWHMx,1= FWHMy,1=0.30
74
The wavenumber periods of arrays (1), (2) and (3) are also compared in Table
5.5.3. Wavenumber periods of arrays (1) and (2) (with the same array spacings dx,1= dy,1=
dx,2= dy,2= 2.3 mm) are the same and larger than those of array (3) (with larger array
spacings dx,3= dy,3= 4.6 mm). This means smaller array spacings generate larger
wavenumber periods. The wavenumber periods can affect the beamforming performance.
For illustration, for the array (3) which has the array spacings dx,3= dy,3= 4.6 mm, the
BF(k) image in Figure 5.5b shows there are two intensified spots at (-0.02, 0.68)rad/mm
and (-0.02, -0.68)rad/mm on the wavenumber curve k(𝛾𝛾) of the 120 kHz A0 mode. This
means if the array (3) generates the 120 kHz A0 mode with the phase delays (-0.02,
0.68)∙pp,q, the synthetic waves generated from the array will have two intensified
components: (i) waves with the wavenumber (-0.02, 0.68)rad/mm, and (ii) waves with the
wavenumber (-0.02, -0.68)rad/mm. The component (ii) is not desired and should be
avoided, since it gives misleading beamforming result. In order to avoid the undesired
component, the wavenumber periods should satisfy Kx> 2kx,max and Ky> 2ky,max, i.e., the
array spacings should satisfy dx< 𝜆𝜆x,min/2 and dy< 𝜆𝜆y,min/2.
5.1.2.2. BF(θ) characterization
The beamforming can also be studied by the directional beam pattern BF(θ) given in
Eq.(5.24). By substituting the array coordinates in Eq.(5.25) into Eq. (5.24), we can
derive the BF(θ) equation for rectangular arrays,
 P −1 Q −1 
1 j[ k ( ω ,θ + β ) − k ( ω ,θ S + β S )]⋅ ( p − ) d x ,( q − )d y 
∑ ∑
P −1 Q −1
BF (θ | w p ,q , d x , d y , P, Q,θ S ) = w p ,q e  2 2 
(5.29)
PQ =p 0=q 0
By substituting the array spans in Eq.(5.26) into Eq.(5.29), the BF(θ) becomes,
75
D Dy  Dx Dy 
1 j[ k ( ω ,θ + β ) − k ( ω ,θ S + β S )]⋅ pd x − ,qd y −
(5.30)
x

∑ ∑
dy 2 2
BF (θ | w p ,q , d x , d y , Dx , D y ,θ S ) = dx
w p , q e  
 Dx   Dy  = p 0= q 0
 d + 1  d + 1
 x  y 
This equation shows the beam pattern BF(θ) is a function of θ defined up to the
parameters w p ,q , d x , d y , Dx , D y and θ S .
Figure 5.6a plots the directional beam pattern BF(θ) of array (1) using the 120
kHz A0 mode without phase delays. The amplitude is seen low at all directions. In order
to steer and maximize BF(θ) to a desired direction θS, phase delays determined from Eq.
(5.22) are applied. Figure 5.6 b plots beamsteering results toward 0°, 45°, 90° and 135°
directions, by applying delays at (0.55, -0.02)∙pp,q, (0.37, 0.51)∙pp,q, (-0.02, 0.68)∙pp,q and
(-0.37,0.54)∙pp,q, respectively.
To evaluate the beamforming quality, the full width of the mainlobe at one-half
peak value (FWHM) (as illustrated in Figure 5.6 b) is used [144]. The smaller FWHM
will give better directional resolution of beamforming. The FWHM results of array (1) are
listed in Table 5.5.4. By comparing the FWHM results of the four steering angles, it is
found that the FWHM of 0° has the smallest value. Hence, the beamforming toward 0°
has the best resolution.
76
90 Mainlobe
90
120 60 120 60
150 30 150 30
FWHM
θ (°)
θ (°)
180 0 180 0
0.2 0.2
0.4 0.4
θS = 0°
0.6 0.6
210 0.8 330 210 0.8 330 θS = 45°
1 1 θS = 90°
240 300 240 300 θS = 135°
(a) 270 (b) 270
Figure 5.6 BF(θ) plots for array (1): (a) without beamsteering, and (b) with
beamsteering in θS =0°, 45°, 90° and 135°.
Mainlobe 90
90
120 60 120 60
150 30 150 30
θ (°)
θ (°)
180 0 180 0
0.2 0.2
0.4 0.4
0.6 θS = 0° 0.6 θS = 0°
210 0.8 330 210 0.8 330
θS = 45° Grating θS = 45°
1 1
θS = 90° lobe θS = 90°
240 300 240 300
270 θS = 135° 270 θS = 135°
(a) (b)
Figure 5.7 BF(θ) plots when the beamsteering directions are θS =0°, 45°, 90° and 135°:
(a) for array (2), and (b) for array (3).
Table 5.4 Full width of the mainlobe at one-half peak value
θS =0° θS =45° θS =90° θS =135°

Array (1) 14° 31° 35° 35°
Array (2) 7° 16° 18° 18°
Array (3) 7° 16° 18° 18°
77
Figure 5.7 a and Figure 5.7 b plot the directional beam patterns of arrays (2) and
(3), when the beamsteering directions are 0°, 45°, 90° and 135°. For 0°, 45° and 135°
beamsteering directions, the beam patterns of array (2) are nearly the same as those of
array (3). However, for the 90° beamsteering direction, the beam pattern of array (3) in
Figure 5.7 b (dotted line) shows an additional lobe at 265° with nearly the same
amplitude as the mainlobe. This additional lobe in 265° is called grating lobe caused by
spatial aliasing [1, 144]. The grating lobe is not desired and should be avoided, since it
gives misleading beamforming result. In addition, the presence of grating lobe can also
be seen from the BF(k) image in Figure 5.5 b, which shows two intensified spots on the
wavenumber curve k(𝛾𝛾), one at (-0.02, 0.68) rad/mm and the other at (-0.02, -0.68)
rad/mm.
The FWHM results of arrays (1), (2) and (3) are compared in Table 5.5.4. FWHM
values of arrays (2) and (3) are the same and both smaller than those of array (1). This
means array (2) and (3) gives better angular resolution than array (1). The better angular
resolution found here is inconsistent with the better wavenumber resolution, which is
found in the comparison among BF(k) images of arrays (1), (2) and (3).
5.1.3. SLDV guided wave phased array
In this subsection, beamforming algorithms are implemented using rectangular grid
arrays made of scanning points of a non-contact SLDV. For the proof of concept,
experimental tests are performed on a CFRP composite laminate with four surface
bonded quartz rods as defects. By using the SLDV phased array, the four defects are
successfully detected.
78
5.1.3.1. SLDV phased array setup
Figure 5.8 a and c show a photo and a schematic of the experimental setup for the SLDV
phased array sensing. The test specimen is an 8-ply [0/45/90/-45]s layup CFRP composite
plate with 2.54 mm thickness (material properties in Table 5.5.1). Four defects are
simulated by four quartz rods (Q1, Q2, Q3 and Q4) bonded on the front side of the plate,
which are the same distance 100 mm away from the plate center and in different angles
0°, 45°, 90° and 135°, as shown in Figure 5.8 b. A PZT wafer (APC 851: 7 mm diameter,
0.2 mm thickness) is bonded on the plate to generate guided waves. The center of the
PZT is set as the coordinate origin. A 3-cycle toneburst with the frequency 90 kHz is
generated from the function generator (model: Agilent 33522B), amplified to 30V by a
voltage amplifier (model: Krohn-Hite 1506), and applied to the PZT wafer for excitation.
A SLDV (model: Polytec PSV-400-M2) is used to acquire the velocity wavefield of
guided waves over the scanning area on the back side of the plate, a 45 mm ×45 mm
square with the center at the coordinate origin. The horizontal and vertical spatial
resolutions of the scanning are 0.1 mm. Based on Doppler Effect, the SLDV can measure
the velocity wavefield v(t, x) (as a function of both time t and space x) of guided waves
over the scanning area. The laser beam is set perpendicular to the plate such that the out-
of-plane motion of guided waves is measured. The phased array is then constructed using
selected SLDV scanning points at {p m } (m=0, 1, 2, …M-1), whose phase center satisfies
∑
M −1
1
M m =0
pm = 0 .
79
Test plate
Oscilloscope
Q3
Q4
Q2
Function
generator
Amplifier PZT Q1
Laser head (b)
(a)
Q3
Q2
Q4
y
Incident Reflected
mth scanning point at pm

Scanning θ
area O Q1
PZT x
(c)
Figure 5.8 Experimental setup for multi-defects detection on a CFRP composite

laminate: (a) photo of test setup, (b) front side of the test specimen, and (c) schematic of
the sensing layout.
5.1.3.2. Array beamsteering and imaging
From the time-space wavefield v(t, x) acquired by the SLDV, the time signal of the mth
array point at location pm can be found as vm(t)=v(t,pm). Its frequency spectrum can be
derived through the Fourier transform, as
∞
= [ vm ( t ) ]
Vm (ω ) = ∫ −∞
vm (t )e − jωt dt (5.31)
80
By using the frequency spectrum Vm(ω), we can derive a synthetic frequency-space
representation Z(ω, x),
j  −ϕ ( ω ,x ) −∆ m ( ω ,x ) 
Z (ω, x ) = ∑ m =0 wmVm (ω )e
M −1
 
(5.32)
where
∆ m (ω, x ) = k (ω, γ ) ⋅ p m , and ϕ (ω, x ) =

−2k (ω, γ ) ⋅ x (5.33)
∆m(ω, x) is the phase delay applied to the mth array point for beamsteering. φ(ω, x)
represents the spatial phase shift. Since the guided waves received from the array
undergone spatial phase shift φ(ω, x) as the guided waves travelled from the PZT to the
damage and then back to the array, −φ(ω, x) is apply to Eq. (5.32) in order to compensate
the spatial phase shift φ(ω, x). In Eq. (5.33), k(ω,γ) is the wavenumber vector at the
frequency ω and wavenumber angle γ, which is obtained from wavenumber dispersion
curve. The wavenumber angle γ is determined from the geometry relation γ= θ + β ,
where θ is the energy propagation angle and equals to the angle of vector x, i.e., θ = ∠x ,
and β is the skew angle corresponding to the energy propagation angle θ.
Using inverse Fourier transform, the synthetic frequency-space representation
Z(ω, x) can be transformed to the time-space domain, as
1 ∞
z (t , x ) =  −1 [ Z (ω , x )] = ∫ Z (ω , x )e jωt d ω (5.34)
2π −∞
where z(t, x) represents the synthetic time-space wavefield with beamsteering. In order to
generate an image of the plate being inspected by the phased array, using the synthetic
time-space wavefield z(t, x), the pixel value at location x is defined as:
I ( x ) = z (t = 0, x ) (5.35)
81
5.1.3.3. Experimental results
5.1.3.3.1. SLDV measurements
Figure 5.9 gives the SLDV measurements, when the four quartz rods have the same
diameter 8 mm and height 10 mm, i.e., four defects with the same size 8 mm. At 30 μs,
the wavefield in Figure 5.9 a shows incident waves with non-circular wavefronts
generated from the PZT wafer. At 145 μs, the wavefield in Figure 5.9 b shows guided
waves reflected by the four defects. Figure 5.9 c and d plot wavenumber spectra achieved
by frequency-wavenumber analysis on wavefields of incident and reflected waves[29,
140, 142]. Both spectra show components on the wavenumber curve of A0 mode,
indicating that the incident and reflected waves are A0 mode.
t = 30 μs t = 145 μs
0.2
40 40
4
20 20 0.1
2
Velocity (mm/s)
Velocity (mm/s)
y (mm)
y (mm)
0 0 0 0
-20 -2 -20 -0.1
-4
-40 -40
-0.2
-40 -20 0 20 40 -40 -20 0 20 40
(a) x (mm) (b) x (mm)
1 0.07
1 1
0.06
0.8
0.5 0.5 0.05
ky (rad/mm)
ky (rad/mm)
0.6
0 0 0.04
0.4 0.03
-0.5 -0.5
0.02
-1 0.2 -1
0.01
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
(c) kx (rad/mm) (d) kx (rad/mm)
Figure 5.9 SLDV measurements: (a) and (b) are wavefields at 30μs and 145 μs,
showing the incident and reflected waves; (c) and (d) are wavenumber spectra of incident
and reflected waves at 120 kHz. The solid white line is the wavenumber curve of the A0
mode at 120 kHz.
82
5.1.3.3.2. Array imaging results
21×21 SLDV scanning points are chosen to construct the array (2) whose configuration is
listed in Table 5.5.2. By using the beamsteering method, a synthetic time-space wavefield
z(t, x) is constructed. The wavefield images at t =80 μs, 40 μs and 0 μs (Figure 5.10 a, b
and c) show four beamformed waves in 0°, 45°, 90° and 135° directions. With the
decrease of time, the beamformed waves are gradually getting closer to the four defects.
At 0 μs, the waves are at the four defects. An intensity image of the plate is constructed
using the synthetic wavefield z(t=0, x) and Eq. (5.35). Figure 5.10 d plots the array
imaging result, clearly showing four intensified areas which indicate the four defects.
Figure 5.11 a and b give the imaging results of arrays (1) and (3) whose array
configurations are listed in Table 5.5.2. By comparing the imaging results of arrays (1),
(2) and (3) in Figure 5.11 a, 10 d and 11 b, it can be found that arrays (2) and (3) give
four intensified areas (indicating the four defects) with smaller sizes than array (1). It
means the imaging results of arrays (2) and (3) have better resolution than array (1). This
is in consistent with the BF(θ) comparison in Figure 5.7, which shows that BF(θ) plots of
arrays (2) and (3) have smaller FWHM and better angular resolution than the array (1).
Although arrays (2) and (3) nearly give the same imaging results at the four
defects, the imaging result of array (3) shows an additional intensified area around the
location (-10, -100) mm. This additional area is induced by the spatial aliasing, and in
consistent with the BF(θ) plot of the array (3) (in Figure 5.7 b) which shows a grating
lobe at 265° direction.
83
t = 80 μs t = 40 μs
1 1
100 100
0.5 0.5
50 50
y (mm)
y (mm)
0 0 0 0
-50 -50
-0.5 -0.5
-100 -100
-1 -1
-100 -50 0 50 100 -100 -50 0 50 100
(a) x (mm) (b) x (mm)
t = 0 μs Imaging result of array (2)

1 1
100 100 0.8
Normalized Amplitude
0.5
50 50
0.6
y (mm)
y (mm)
0 0 0
0.4
-50 -50
-0.5
-100 0.2
-100
-1 0
-100 -50 0 50 100 -100 -50 0 50 100
(c) x (mm) (d) x (mm)
Figure 5.10 Beamforming and imaging of array (2): (a), (b) and (c) are synthetic
wavefields z(t, x) at 80 μs, 40 μs and 0 μs, (d) is the imaging result.
Imaging result of array (1) Imaging result of array (3)

1 1
100 100 0.8

0.8
50 50
0.6 0.6
y (mm)
y (mm)
0 0
0.4 0.4
-50 -50
-100 0.2 0.2

-100
0 0
-100 -50 0 50 100 -100 -50 0 50 100
(a) x (mm) (b) x (mm)
Figure 5.11 Array imaging results: (a) for array (1), and (b) for array (3).
84
5.2. FILTER RECONSTRUCTION IMAGING
The filter reconstruction imaging is based on the frequency-wavenumber filtering process
(presented in Chapter 4) by extracting damage induced frequency-wavenumber
components and reconstructing them to the time-space wavefield vF(t, x) which only
contains damage induced waves. Then, the cumulative wave energy of the damage
induced waves can be derived [103, 147] for the purpose of damage detection and
imaging, as:
t 1 2
E F ( x, t ) = ∫ vF (τ , x )dτ (5.36)
0 2
where EF(x, t) represents the cumulative wave energy of the damage induced waves.
Figure 5.12 gives an example of filter reconstruction imaging for the detection of
delamination damage in a CFRP composite. The details of this example can be found in
[148]. As shown in the wavefield (Figure 5.12a) measured by a SLDV, there are damage
induced waves in the delamination area. Accordingly, the wavenumber spectrum (Figure
5.12b) shows additional wavenumber components. A “band-pass” wavenumber filter
(Figure 5.12c) is designed to retain only the additional wavenumber components and
remove the rest. The filtering is then implemented by multiplying the band-pass filter
with the spectrum. The filtered spectrum in Figure 5.12d only contains the delamination
induced additional wavenumber values. Meanwhile, the reconstructed wavefield in
Figure 5.12e only has the delamination induced waves which are confined in the
delamination area. The resulting cumulative energy map in Figure 5.12f shows a high
energy concentration area which indicates the delamination location and size.
85
60 6 Delamination induced
Delamination wavenumber components0
area 4 4
Normalized amplitude (dB)

50
-5
A0
Velocity (mm/s)
2 2
40 -10
ky (rad/mm)
y (mm)
0
30 0 S0 -15
-2
20 -2 -20
-4
-25
10 -4
-6 -30
-20 -10 0 10 20 -4 -2 0 2 4
(a) x (mm) (b) kx (rad/mm)
1 0
4 4

-5
0.8
2 2 -10
Amplitude
ky (rad/mm)
ky (rad/mm)
0.6
0 0 -15
0.4
-2 -2 -20
0.2 -25
-4 -4
0 -30
-4 -2 0 2 4 -4 -2 0 2 4
60 6 60 1
50 4
50 0.8
Velocity (mm/s)
40 2
40
0.6
y (mm)
y (mm)
0
30 30
0.4
-2
20 20
(e)
-4 0.2
10 10
-6 0
-20 -10 0 10 20 -20 -10 0 10 20
(e) x (mm) (f) x (mm)
Figure 5.12 Filter reconstruction imaging for delamination detection in a CFRP

composite plate: (a) the wavefield at 15 μs on a CFRP plate with delamination damage,
(b) the wavenumber spectrum of the wavefield showing delamination induced
wavenumber components, (c) a band pass filter in wavenumber domain, (d) the filtered
wavenumber spectrum, (e) the reconstructed wavefield at 15 μs, (f) the accumulated
energy map in the time range 0 to 50 μs. The red dotted box indicates the actual
delamination region.
86
5.3. SPATIAL WAVENUMBER IMAGING
The space-frequency-wavenumber analysis (presented in Chapter 4) can generate the
space-frequency-wavenumber representation S ( x , f , k ) . From this representation, spatial
wavenumber functions can be derived for damage detection and imaging. In the
following, two different methods to derive spatial wavenumber functions are provided.
In the first method (weighted sum method), the spatial wavenumber function at
the frequency f0 is defined through a weighted sum of k [32], i.e.,
∑  S ( x, f , k ) k 
0
k sw ( x, f ) = k
(5.37)
0
∑ S ( x, f , k )
k
0
This method considers the contribution of each wavenumber k weighted by its
amplitude, and generates a spatial function of effective wavenumber k sw ( x , f 0 ) . In
practice, when two or more wave modes exist, the function k sw ( x , f 0 ) is influenced by the
contribution of all the present wave modes.
In the second method (local maximum method), the spatial wavenumber function
is defined by finding the wavenumber where the representation S ( x , f 0 , k ) has the
largest amplitude [149], i.e.,
k sw ( x , f 0 ) = k sw ( x , f 0 ) (5.38)
k sw ( x , f 0 ) = arg max S ( x , f 0 , k ) (5.39)

k
This method only considers the wavenumber where the largest S ( x , f 0 , k ) appears. In
practice, when two or more wave modes exist, the function k sw ( x , f 0 ) is only influenced
by the wave mode with the largest amplitude.
87
The two methods above generate wavenumber functions at a selected frequency f0.
Considering that no real signal has a singular frequency component but will have some
finite bandwidth ∆f, the spatial wavenumber function is defined as:
1 N
k sw ( x ) =
N
∑k
i =1
sw ( x, fi ) (5.40)
where fi (i=1,2,3…N) is a series of frequency samples over the frequency bandwidth ∆f.
The spatial wavenumber function contains the local wavenumber information, and
can be used damage detection. Figure 5.13 gives an example of the spatial wavenumber
method for delamination detection in a CFRP plate. The wavefield in Figure 5.13 a shows
complex wave interactions in the delamination area. By using the spatial wavenumber
method, the wavenumber information of the wavefield is unveiled. As shown in in Figure
5.13 b, the spatial wavenumber image indicates an area of approximately 20 mm × 20
mm which has larger wavenumber values than the rest of the plate. This area matches
well with the delamination damage (indicated by the red dotted box in Figure 5.13 b).
Obviously, the delamination affects wavenumbers of waves that propagate in the
delamination region. The spatial wavenumber method shows its potential for detecting
and quantifying delamination damage. The details of this example can be found in [148].
60 3.5 60 3.5
Wavenumber (rad/mm)
50 50
Velocity (mm/s)
3 3
40 40
y (mm)
y (mm)
30 30
2.5 2.5
20 20
10 2 10 2
-20 -10 0 10 20 -20 -10 0 10 20

(a) x (mm) (b) x (mm)
Figure 5.13 Spatial wavenumber imaging for delamination detection in a CFPR plate: (a)
a wavefield at 45 μs showing wave interactions with delamination damage, and (b) the
spatial wavenumber image indicating the delamination damage.
88
PART II: SHM/NDE APPLICATIONS
CHAPTER 6
GUIDED WAVES IN WATER LOADED STRUCTURES
AND WATER LEVEL DETECTION
6.1. GUIDED WAVE WATER LEVEL DETECTION
6.1.1. Motivation and introduction
Guided waves have been used for damage detection in water loaded structures, such as
ship hulls and submarines [150-155]. Nevertheless, using guided waves for evaluating
structural surface condition change due to the presence of water is much less discussed.
When a free plate is immersed in water, guided waves in the plate will be affected. The
out-of-plane motions of the guided waves in the plate will transmit into water through the
plate-water interface [8]. The surrounding water provides a way for guided wave energy
to leak outwards from the plate known as leaky guided waves [8]. Compared to guided
waves in a free plate, the guided waves in a water loaded plate have different behaviors,
such as mode shapes, wave speeds and attenuations [11, 63, 139, 156-159]. Moreover,
there is another wave mode reported by Cegla et al. known as quasi-Scholte mode present
in the water loaded plate [139, 159]. They studied the quasi-Scholte mode in a plate with
both sides immersed in infinite water, where the plate is considered as a symmetrically
loaded waveguide [139]. It was found that the quasi-Scholte mode is dispersive at low
frequencies and asymptotes to the non-dispersive Scholte mode behavior at high
frequencies [139]. Also at low frequencies energy of the quasi-Scholte mode
89
predominantly flows in the plate, while at high frequencies most of the energy is in the
water [139].
In this study, both simulations and experiments are performed to further
investigate guided waves in one-side water loaded plates. The numerical simulations are
performed with the EFIT method. The experimental investigations are enabled by the
PZT-SLDV sensing. Both results confirm the existence of quasi-Scholte waves in one-
side water loaded plates. In addition, the influence of water amount on guided waves is
investigated. It is seen that the water amount affects the wave propagation time, which
can be potentially used for the estimation of water level in a container.
6.1.2. Modeling of guided waves in one-side water loaded plates
To investigate the characteristics of guided waves in one-side water loaded plates, the
modeling techniques presented in Chapter 3 are used. Using the GMM method, guided
wave dispersion curves and mode shapes in a one-side water loaded plate are derived,
and using the EFIT numerical method, the wave propagation is simulated. The related
model setups and results are presented in Chapter 3.
6.1.3. Experimental characterizations
A test tank (Figure 6.1 a) has been designed with the front wall as the test plate (a 1.2 mm
thick T304 steel plate). When the tank is empty, the front wall is considered as a free
plate; with water in the tank, it is then considered as a water loaded plate. A PZT wafer
(APC 850: 7 mm×7 mm dimensions, 0.2 mm thickness) is bonded close to the bottom of
the front wall as the wave actuator. A coordinate x is defined with the origin O being set
at the center of the PZT and aligned vertically upward. Figure 6.1b gives the overall
90
experimental setup using the hybrid PZT-SLDV sensing. In the test, the laser beam is
normal to the plate such that only the out-of-plane wave motion is measured.
Water tank Test plate

Water tank
Laser head
Scan Oscilloscope
points
Function
generator
PZT actuator
(a) (b)
Figure 6.1 The PZT-SLDV experiment on a water tank: (a) an illustration of the water
tank, and (b) the overall test setup.
A pulse signal with 0.5 μs duration and 10V amplitude is sent to the PZT from an
arbitrary function generator (model: Agilent 33500B) to excite guided waves. Waves at
multiple scanning points along a vertical line started from the PZT actuator, as indicated
in Figure 6.1a, are measured by the SLDV. The spatial resolution of the scanning is 0.3
mm. Two situations, a free plate (from an empty tank) and a fully loaded plate (from a
full tank), are investigated. Figure 6.2 a and b plot waveforms measured at 40 mm away
from the PZT, and Figure 6.2 c and d plot time-space wavefields in the free and fully
loaded plates, respectively. Discernible differences can be observed from both the
waveforms and the wavefields of the two cases. However, it is difficult to identify the
guided wave modes. It is also hard to qualitatively and quantitatively analyze the
differences between the two cases.
91
Velocity (mm/s)
Velocity (mm/s)
Velocity (mm/s)
Velocity (mm/s)
x (mm)
x (mm)
(c) Time (μs) (d) Time (μs)
Figure 6.2 Experimental results: (a) and (b) are received waveforms at 40 mm away
from the PZT in free and water loaded plates; (c) and (d) are time-space wavefields
measured in free and water loaded plates.
2.5 1 2.5 1
Quasi-Scholte
Wave number (rad/mm)
2 0.8 2 0.8
A0 A0WP
1.5 0.6 1.5 0.6
1 0.4 1 0.4
S0 S0WP
0.5 0.2 0.5 0.2
0 0 0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
2.5 1 Quasi-Scholte 1
Quasi-Scholte 1.5
2 A0
0.5 0.5
A0
1.5 1
0 0
1
S0 0.5
S0
-0.5 -0.5
0.5
0 -1 0 -1
0 200 400 600 800 1000 0 100 200 300 400
(c) Frequency (kHz) (d) Frequency (kHz)
Figure 6.3 Frequency-wavenumber analysis results: (a) for the free plate, (b) for the
water loaded plate, (c) spectrum difference obtained by subtracting the spectrum of the
free plate from the spectrum of the water loaded plate, and (d) zoom-in plot of figure (c).
92
Using frequency-wavenumber analysis, time-space wavefields for free and fully
loaded plates are transformed into frequency-wavenumber spectra, as presented in Figure
6.3 a and b. The wavenumber dispersion curves solved from the GMM are also plotted in
Figure 6.3, in order to identify the wave modes in the spectra. By comparing the spectra
with the wavenumber curves, each wave mode can be identified. In the free plate
spectrum in Figure 6.3 a, A0 and S0 modes are discernibly identified. Note that since the
out-of-plane motion of A0 mode is much stronger than that of S0 mode and the SLDV
only measures the out-of-plane motion, the amplitude of A0 mode in the spectrum is
larger than that of S0 mode since [140]. Particularly, at frequencies below 500 kHz, the S0
mode is barely seen due to the small out-of-plane motion in the low frequency range.
In the spectrum of the water loaded plate (Figure 6.3b), three wave modes
including A 0WP , S0WP and quasi-Scholte modes are identified. Compared to the spectrum
of the free plate, the spectrum of the water loaded plate shows that the amplitude of S0WP
mode remains nearly the same as S0 mode, while the amplitude of A 0WP mode
significantly decreases due to the energy leakage. Particularly, at frequencies below 500
kHz, the A 0WP mode nearly disappears. Meanwhile, the quasi-Scholte mode shows up in
the frequency range up to 300 kHz, which is consistent with what has been reported in
[139]. Moreover, it is seen that the quasi-Scholte mode has different wavenumbers
compared to the A0 mode.
To clearly show the difference, the free plate spectrum (Figure 6.3 a) is subtracted
from the immersed plate spectrum (Figure 6.3 b). The subtraction result and a zoomed-in
plot are given in Figure 6.3c and d. It is noticed in Figure 6.3c that S0 and S0WP modes
93
have almost been canceled out after the subtraction since S0 and S0WP modes are nearly
the same in both the free and immersed plates. At low frequencies, both quasi-Scholte
and A0 modes are remained. The difference plot confirms that the quasi-Scholte mode in
the immersed plate has different wavenumbers compared to the A0 mode in the free plate.
6.1.4. Wave signature and water level correlation
Figure 6.4 plots the pitch-catch sensing setup on the same water tank given in Figure 6.1,
to investigate the influence of water level on guided waves. The guided waves are
generated from a PZT with a 3 count tone burst signal at 100 kHz. The waves at 140 mm
away from the PZT are measured by an SLDV receiver. Wave signals are acquired at
different water levels dW from 5 to 135 mm with a 10 mm step.
As shown in the setup, the wave propagation path dT-R consists of the water path
dW in the immersed plate portion, and the dry path dT-R‒dW in the free plate portion. The
guided waves leave the actuator as the quasi-Scholte mode in the immersed portion dW,
then convert to the A0 mode in the free portion dT-R‒dW. Therefore, the total wave
traveling time tT-R can be expressed as,
dW d T − R − dW
tT=
−R + (6.41)
cQS c A0
where cQS and cA0 are group velocities of quasi-Scholte and A0 modes (at 100 kHz,
cQS=1811 m/s and cA0=1895 m/s). Using the case dW=0 as the baseline, the change of
traveling time at any water level dW w.r.t. the baseline can be derived as,
1 1
∆tT =
−R dW ( − ) (6.42)
cQS c A0
94
Eq.(6.42) shows a linear relation between the change of traveling time ∆tT-R and the water
level dW. This linear relation is plotted in Figure 6.5 a (the solid line). Thus, once the time
difference ∆tT-R is known, the water level can be determined based on the linear relation.
Figure 6.5 b presents the baseline signal at dW=0 (solid line) and the signal at dW =
95 mm (broken line) measured by the SLDV. The change of traveling time ∆tT-R can be
estimated by calculating the time difference between the signal at dW=95mm and the
baseline. For all the water levels from 5 to 135 mm, their ∆tT-R are plotted in Figure 6.5 a
(triangle markers), in comparison to the theoretical prediction (the solid line). It is seen
that the experimental data agrees very well with theoretical prediction, which validates
the linear relationship between the change of traveling time and the water level. Hence,
this sensing methodology can be potentially used for water level detection in a container.
Receiver: SLDV
Dry
path
dT-R
Water dW
path
Actuator: PZT
Figure 6.4 An illustration of the pitch-catch configuration.
4
Experimental
Theoretical dW =0
Time change ∆tT-R (μs)
Velocity (mm/s)
3
dW =95 mm
0 (b) Time (μs)

0 50 100 150
(a) Water level dW (mm)
Figure 6.5 Water influence on wave propagation at 100 kHz: (a) the relation between
the water level and the change of traveling time ∆tT-R, and (b) received waveforms when
water level dW = 0 (solid line) and dW = 95 mm (broken line).
95
6.2. GAS ACCUMULATION DETECTION IN NUCLEAR COOLANT PIPES
6.2.1. Motivation and introduction
The U.S. Nuclear Regulatory Commission (NRC) issued Generic Letter 2008-01 due to
the safety concern associated with gas accumulation events in emergency core cooling
pipes of nuclear power plants [160]. Since the gas accumulation may critically damage
pipes, pumps, and valves and affect the safety operation of nuclear power plants, the gas
accumulation needs to be detected as well as quantified. In a pipe with water inside, when
a large amount of gas presents and accumulates, there will be a gas-water boundary
between the gas and the water. Therefore, the location of the gas-water boundary can be
used as an indication of gas accumulation.
Based on the study of guided wave water level detection in the last section, the
gas accumulation detection using guided waves for nuclear coolant pipes are investigated
in this section. Guided waves in an empty pipe and a water loaded pipe are acquired
through experiments, and analyzed using the frequency wavenumber analysis. Analysis
results show that guided wave characteristics such as wavenumbers and wave speeds are
different between empty and water loaded pipes. A gas accumulation detection method is
developed and verified through laboratory proof-of-concept tests.
6.2.2. Guided waves in water loaded pipes
When water presents in a pipe, guided waves in the pipe will be affected. In order to
understand how the water affects the guided waves, experiments using PZT-SLDV
sensing are performed. Figure 6.6 a plots a schematic of the experimental setup. A small
scale test pipe with 1.56 mm wall thickness and 114.4 mm outer diameter made of
stainless steel (T304) is used. A PZT actuator (APC 850: 5 mm × 5 mm dimensions and
96
0.2 mm thickness) is surface mounted on the pipe to generate guided waves. The
excitation signal is a pulse with 2.5 μs duration and 60 V amplitude. An SLDV is used to
measure the guided waves at various scanning points along a vertical line with a spatial
resolution of 0.3 mm.
Empty pipe Full pipe

Water pipe
Measurement DAQ & signal
processing
R
SLDV Scanning Scaning
Trigger
points points
Voltage T
Excitation Function
PZT amplifier signal generator
PZT PZT
(a) (b) (c)
Figure 6.6 The experimental setup: (a) a schematic of the setup, (b) illustration of an
empty pipe, (c) illustration of a full pipe.
1 1
Velocity (mm/s)
Velocity (mm/s)
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 20 40 60 80 0 20 40 60 80
1.5 1.5
80 80
1 1
Velocity (mm/s)
Velocity (mm/s)
0.5 0.5
60 60
x (mm)
x (mm)
0 0
40 40
-0.5 -0.5
-1 -1
20 20
-1.5 -1.5
0 20 40 60 80 0 20 40 60 80
(c) Time (μs) (d) Time (μs)
Figure 6.7 Experimental results: (a) and (b) are received waveforms at 90 mm away
from the PZT in the empty and full pipes; (c) and (d) are time-space wavefields measured
in the empty and full pipes.
97
1.5 1.5
Wave number (rad/mm) 1 1

0.8
0.8
1 1
0.6 0.6
0.4 0.4
0.5 0.5
0.2 0.2
0 0
0 0
0 100 200 300 400 500 0 100 200 300 400 500
1.5 1
0.5
1 Full pipe
0.5
Empty pipe -0.5
-1
0
0 100 200 300 400 500
(c) Frequency (kHz)
Figure 6.8 Frequency-wavenumber analysis: (a) and (b) are frequency-wavenumber

spectra for empty and full pipes; (c) the spectrum difference obtained by subtracting the
spectrum of the empty pipe from the spectrum of the full pipe.
Figure 6.7 a and b plot waveforms at 90 mm away from the actuator for empty
and full pipes. Figure 6.7 c and d plot time-space wavefields for both cases. Compared to
guided waves in the empty pipe, guided waves in the full pipe travel slower with smaller
amplitudes. The time-space wavefields are further processed using frequency-
wavenumber analysis. The frequency-wavenumber spectra for both cases are given in
Figure 6.8 a and b. The two spectra show components with large amplitudes at
frequencies lower than 250 kHz. To show the difference between the two spectra, the
empty pipe spectrum in Figure 6.8 a is subtracted from the full pipe spectrum in Figure
98
6.8 b. The subtraction result is plotted in Figure 6.8 c, in which the red and the blue
component are for guided waves in full and empty pipes, respectively. Figure 6.8 c shows
guided waves in the full pipe have larger wavenumbers than those in the empty pipe.
6.2.3. Gas accumulation detection
6.2.3.1. Test on a small scale pipe
Figure 6.9 plots the pitch-catch sensing setup on the small scale pipe, to investigate the
influence of water level on guided waves. Guided waves are generated from a PZT. The
excitation is a three-cycle toneburst with 30 V amplitude and 100 kHz frequency. The
guided waves at 90 mm away from the PZT are measured by an SLDV receiver. In order
to investigate how water levels affect guided waves, the guided waves are measured at
different water levels dW from 0 to 90 mm with a step of 10 mm.
As shown in the pitch-catch sensing configuration in Figure 6.9, the wave
propagation path (dT-R) from the actuator to the receiver consists of water path dW and dry
path dT-R‒dW. That’s to say, the total wave traveling time tT-R can be expressed as
dW d T − R − dW
tT − R
= + (6.43)
cwaterpipe cemptypipe
where cwaterpipe and cemptypipe are group velocities of guided waves in the full pipe and the
empty pipe, respectively. Thus, using the empty pipe when dW = 0 as the baseline, the
change of traveling time at any water level of dW w.r.t. the baseline can be derived as
1 1
∆tT − R dW (
= − ) (6.44)
cwaterpipe cemptypipe
Eq. (6.44) shows a linear relation between the change of traveling time ∆tT-R and the
water level dW. This linear relation is plotted in Figure 6.10 a (solid line).
99
Figure 6.10 b plots guided wave signals measured at water levels dW=0, 50 and 90
mm. From these waveforms, it is found that the wave traveling time increases and the
wave amplitude decreases with the increase of water level. The time differences ∆tT-R for
different water levels are plotted in Figure 6.10 a (triangle markers). The experimental
results show a linear relation between the change of traveling time and the water level,
and agree well with the theoretical prediction. Thus, by measuring the traveling time of
guided waves, the water amount (or gas amount) in the pipe can be estimated.
Receiver: SLDV
Dry
R path
dT-R
Water
dW
T path
Actuator:
PZT
Figure 6.9 An illustration of the pitch-catch sensing on a small scale pipe.
Experimental 50 mm
3
Theoretical
1 90 mm
0 mm
Velocity (mm/s)
2 0.5
0
1 -0.5
-1
0 0 20 40 60 80 100
(b) Time (μs)
0 20 40 60 80
Figure 6.10 Experimental results for the small scale pipe: (a) the relation between the
water level and change of traveling time ∆tT-R, (b) received waveforms when water levels
are 0 mm, 50 mm and 90 mm.
100
6.2.3.2. Test on a tilted full scale pipe
Beside the test on the small scale pipe, an experimental test is performed on a tilted full
scale pipe. Figure 6.11 a gives the overall test setup. The test pipe (material: T304 steel,
wall thickness: 7.1 mm, outer diameter: 168.3 mm) is tilted and supported by a wood
frame. Figure 6.11 b plots the pitch-catch sensing configuration. Two PZT sensors (APC
850: 7 mm×7 mm dimension and 0.2 mm thickness) are mounted on the pipe outer
surface with 300 mm apart along the axial direction of the pipe. The lower PZT is used
the actuator and the upper PZT is used as the receiver. The excitation signal is a 3 count
toneburst with 30 kHz frequency and amplitude 25 V amplitude. The guided waves are
measured at different water levels dW from 0 to 300 mm with a step of 30 mm.
Figure 6.12 b plots guided wave signals measured at water levels dW=0,
150 and 300 mm. These signals show that the wave traveling time increases and the wave
amplitude decreases with the increase of water level. This trend is the same as what is
observed in the small scale pipe. The time differences ∆tT-R for different water levels are
plotted in Figure 6.12 b (triangle markers), in comparison with the theoretical
prediction (solid line). The experimental results show a linear relation between the
change of traveling time and the water level, and agree well with the theoretical
prediction. Thus, by measuring the traveling time of guided waves, the location of the
gas-water boundary can be detected. Hence, the water amount (or gas amount) in the pipe
can be estimated. This confirms the presented guided wave sensing method is able to
detect and quantify the gas accumulation.
101
Oscilloscope
Dry
Water pipe path
Water Receiver:
PZT
Function path dW
generator
dT-R
Amplifier Actuator:
PZT
(a) (a)
Figure 6.11 Test on a tilted large scale pipe: (a) photo of the test setup, (b) schematic of
the pitch-catch sensing configuration with PZT transducers.
3
Experimental 150 mm
2.5 Theoretical
300 mm
2 0 mm
Velocity (mm/s)
2
1.5 0
1
-2
0.5
0 50 100 150 200 250 300
0 (b) Time (μs)
0 100 200 300
Figure 6.12 Experimental results for the tilted large scale pipe: (a) the relation between
the water level and change of traveling time ∆tT-R, (b) received waveforms when water
levels are 0 mm, 150 mm and 300 mm.
102
CHAPTER 7
CRACK DETECTION AND QUANTIFICATION IN ISOTROPIC PLATES
7.1. MOTIVATION AND INTRODUCTION
Among various crack detection methods, Lamb wave based NDE/SHM has shown great
potential for crack detection in plate-like structures, due to the sensitivity of Lamb waves
to a variety of defects and their capability of traveling long distances in the plate-like
structures [29, 161]. Intensive studies of Lamb wave based NDE/SHM have been
presented in literature about using Lamb waves to detect crack damage in isotropic plates,
as well as to quantify the crack size [12, 19-24, 80, 117, 121, 122, 124, 131, 162-178].
Ihn and Chang used a built-in piezoelectric sensor/actuator network to detect and monitor
crack growth [179]. Tua et al. used PZT transducers on a square grid to detect through,
half-through, and concealed cracks in aluminum plates [167]. Lu et al. used Lamb wave
scattering to quantitatively assess the through-thickness crack size [180]. Yu and Leckey
used Lamb wave based sparse arrays to detect crack damage [165]. Chen et al. developed
a load-differential imaging method to process Lamb wave signals received from a sparse
array, for the purpose of crack detection and localization [122].
In literature, many Lamb wave based damage detection approaches use
measurements from distributive sensors, and damage estimation is performed by
processing data acquired from the point-wise locations of the sensors. Methods based on
such point-wise measurements have several limitations due to the sparsity of data, such as
the limitation in accurate determination of damage size, and the variation in detectability
103
of damage which relates to relative locations of sensors w.r.t. the damage [165]. These
limitations have led to recent advancements in the acquisition and analysis of full
wavefields [104, 109, 110, 161, 181, 182].
The work presented in this chapter focuses on crack detection and quantitative
evaluation using PZT-SLDV sensing and wavefield analysis. The intention of this work
is to demonstrate contactless wavefield measurement and wavefield processing methods
for detection and quantification of crack damage. In particular, this work presents two
robust damage imaging methods, filter reconstruction and spatial wavenumber imaging,
in order to detect crack damage and quantify the crack length, depth and width.
610 mm
Oscilloscope Test plate
PZT
y
Crack
510 mm
Amplifier Laser head
PZT x
Function
generator
(a) (b)
Figure 7.1 Experimental setup for crack detection in a 1 mm thick aluminum plate: (a)
overall test setup, and (b) schematic of the test plate.
7.2. WAVE INTERACTION WITH CRACK DAMAGE
Figure 7.1 gives the overall experimental setup. The test specimen is a 2024-T3
aluminum plate (dimensions: 610 mm × 510 mm × 1 mm) with a through-thickness (TT)
crack. The crack is made by electric discharge machining (EDM) and has a length of 31
mm. A round shaped PZT (APC 850: 7 mm diameter and 0.2 mm thickness) is used to
104
generate Lamb waves and the center of the PZT is set as the origin of the x-y coordinate
system. The excitation signal is a 3 count toneburst with 360 kHz frequency and 40 V
amplitude. An SLDV (Polytec PSV-400-M2) is used to measure Lamb waves. Note that
the laser beam is set normal to the plate, in order to acquire the out-of-plane velocity (vz)
component. Lamb waves in a 190 mm × 110 mm scanning area are measured, with a
spatial sampling interval Δd of 0.5 mm. Note the sampling interval Δd is selected by
meeting the sampling theorem requirement, i.e., Δd≤λ/2, where λ is the wavelength of
the interrogating wave mode [145]. Additionally, experience has shown that a smaller Δd
will result in a better spatial resolution. In our test, the wavelength of the interrogating A0
mode is 4.6 mm. Hence, the 0.5 mm sampling interval satisfies the sampling theorem
while providing a good spatial resolution.
Figure 7.2 a and b plot wavefield images acquired by the SLDV at 36 μs and 65
μs, respectively. The wavefield images clearly capture incident A0 and incident S0 modes
as well as their interactions with the crack. At 36 μs, the incident S0 mode has passed the
crack, and the incident A0 has not yet arrived. The wavefield shows a reflection S0 mode
and a scattering A0 mode generated from S0 to A0 mode conversion. At 65 μs, the A0
mode has passed the crack and a reflection A0 mode is generated.
36 µs 65 µs
1 1
Incident S0 Incident A0
100 100
Incident A0 0.5 0.5
80 80 Reflection A
0
y (mm)
y (mm)
60 0 60 0
40 40
Scattering A0 -0.5 -0.5
20 20
Reflection S0
-1 -1
50 100 150 200 50 100 150 200
(a) x (mm) (b) x (mm)
Figure 7.2 Wavefield images measured by the SLDV at 360 kHz excitation on a 1 mm
thick aluminum plate with a TT crack: (a) at 36 μs, and (b) at 65 μs.
105
7.3. CRACK DETECTION AND QUANTIFICATION
7.3.1. Crack detection and quantification with experimental data
For the purpose of crack detection and quantification, the filter reconstruction imaging is
applied to the time-space wavefield measured by the SLDV. Figure 7.3 c plots the
wavenumber spectrum at 360 kHz with theoretical wavenumber curves of A0 and S0
modes. The wavenumber spectrum shows strong components on the curves of A0 and S0
modes, representing the wavenumber components A0 and S0 modes. Besides the A0 and
S0 modes, there are other wavenumber components (diagonally distributed components)
in the spectrum. The wavenumber filtering is then applied to keep only the diagonal
wavenumbers by using a wavenumber filter as plotted in Figure 7.3 a and b. Figure 7.3 d
plots the reconstructed image after wavenumber filtering, showing a high amplitude area
representing the detected crack damage. This area agrees well with the actual crack.
The reconstructed image confirms that the diagonal wavenumbers represent
waves that gather closely around the structural discontinuity, i.e. the crack. In another
words, the diagonal wavenumbers are caused by abrupt changes of waves around the
crack. The filter reconstruction imaging therefore provides a useful means of mapping the
structural discontinuity, and yields the location and size of the discontinuity. From the
imaging result in Figure 7.3 d, we can estimate the location (crack center at (132.6 mm,
56.1 mm) with relative errors (1.22%, 1.58%)) and length of the crack (28.1 mm with a
relative error 9.35%).
106
1
2
0.8
1 1
ky (rad/mm)
Amplitude
Amplitude
0.6
0 0.5
0.4
-1 0
0.2 0 0.5 1 1.5 2 2.5
-2 (b) |k| (rad/mm)
0
-2 -1 0 1 2
(a) kx (rad/mm)
0
2
-10 1
1 100
-20 0.5
ky (rad/mm)
80
y (mm)
0 -30
60 0
-1 -40 40
-0.5
-50 20
-2
-1
-60 50 100 150 200
-2 -1 0 1 2 x (mm)
(c) kx (rad/mm) (d)
Figure 7.3 Filter reconstruction imaging with experimental data in an aluminum plate
with a TT crack: (a) high-pass wavenumber filter, (b) illustration of the filter along the kx
axis, (c) the frequency-wavenumber spectrum at 360 kHz, and (d) the reconstructed
imaging of the plate indicating the presence of the crack damage. Solid and broken lines
in figures a and c are theoretical wavenumber curves of S0 and A0 modes.
Wavenumber (rad/mm)
100 2.4
80
2.2
y (mm)
60
2
40
20 1.8
50 100 150 200

x (mm)
Figure 7.4 Spatial Wavenumber imaging with experimental data in an aluminum plate
with a TT crack.
107
The spatial wavenumber imaging is also applied to the wavefield measured by the
SLDV. Figure 7.4 gives the imaging result over the scan area at the excitation frequency
360 kHz. The crack damage is clearly indicated by a region with large wavenumbers.
This means the wavenumber increases due to the influence of the structural discontinuity.
From the imaging result in Figure 7.4, we can estimate the location (crack center at
(131.9 mm, 56.8 mm) with relative errors (0.69%, 0.35%)) and length of the crack (29.2
mm with a relative error 5.81%). Thus, the spatial wavenumber imaging provides another
method to detect crack damage and quantify the crack location and size.
7.3.2. Crack detection and quantification with simulation data
Modeling can serve as an efficient and cost effective way to assist the development of
diagnostic algorithms [1, 8, 36, 37]. In this study, the EFIT modeling technique is used to
simulate the cases of (i) a plate with a through-thickness (TT) crack, (ii) a plate with the
same crack as in case (i), except with three times the crack width (3X), and (iii) a plate
with the same crack as in case (i) except with the crack extending only half-way through
the plate thickness (hTT). Then the simulation data generated from EFIT is used to
demonstrate the quantifications of crack length, depth and width using the filter
reconstruction imaging and spatial wavenumber imaging. The layout of the simulation is
given in Figure 7.1 b, which is the same as the layout of the experiment. The plate in
simulation is the same as the plate in experiment. A round shaped PZT (7 mm in diameter
and 0.2 mm thick) is used for generating Lamb waves and its location is set as the origin
of the x-y coordinate system. In order to closely approximate the PZT actuation, the EFIT
source excitation is implemented as in-plane displacements occurring at edges of the PZT
actuator (assuming a ring-type excitation source) [47]. The excitation signal is a 3 count
108
toneburst with a frequency of 360 kHz. Three cases (TT, 3X and hTT cracks) are
simulated using EFIT method. The centers of cracks are at the coordinates (131mm,
57mm), with the same distance from the crack center to the actuator in all three cases
(143 mm). More details of simulations can be found in reference [183].The time-space
wavefields generated from simulations are then processed by using the two imaging
methods for crack quantification.
The results of filter reconstruction imaging at 360 kHz are given in Figure 7.5.
The reconstructed images show locations and lengths of cracks. It is also noted that
amplitudes in the reconstructed images are different among the three cracks. For direct
comparison, the amplitudes along these cracks are plotted in Figure 7.6. It shows that the
3X crack has the largest amplitude, the TT crack has the medium amplitude, and the hTT
crack has the smallest amplitude. The comparison in Figure 7.6 indicates that the more
significant discontinuity leads to the higher amplitude in the imaging results. Hence, the
filter reconstruction imaging method can detect crack damage, quantify the crack size,
and show the crack severity.
The results of spatial wavenumber imaging at 360 kHz are given in Figure 7.7.
Cracks are represented as larger wavenumbers in contrast to wavenumbers at all other
locations. It is observed that the spatial wavenumber images are different for the three
cracks. For comparison, the wavenumber values along the three cracks are plotted in
Figure 7.8. The 3X crack has the largest wavenumber value, the TT crack has the
medium value, and the hTT crack has the smallest value. As the filter reconstruction
imaging, the spatial wavenumber imaging can also detect crack damage, quantify the
crack size, and show the crack severity.
109
1
100
0.5
80
y (mm)
60 0
40
-0.5
20
-1
50 100 150 200
(a) x (mm)
100
0.5
80
y (mm)
60 0
40
-0.5
20
-1
50 100 150 200
(b) x (mm)
100
0.5
80
y (mm)
60 0
40
-0.5
20
-1
50 100 150 200
(c) x (mm)
Figure 7.5 The filter reconstruction imaging results at 360 kHz: (a) for the TT crack
plate, (b) for the 3X crack plate, and (c) for the hTT crack plate.
1.2
0.9
0.6
TT
0.3 3X
hTT
0
-10 -5 0 5 10
Distance to the crack center (mm)
Figure 7.6 Amplitudes along the cracks extracted from the filter reconstruction imaging
results. Broken, solid and dotted lines are for TT, 3X and hTT cracks, respectively.
110
Wavenumber (rad/mm)
100 2.4
80
2.2
y (mm)
60
2
40
20 1.8
50 100 150 200

(a) x (mm)
Wavenumber (rad/mm)
100 2.4
80
2.2
y (mm)
60
2
40
20 1.8
50 100 150 200

(b) x (mm)
Wavenumber (rad/mm)
100 2.4
80
2.2
y (mm)
60
2
40
20 1.8
50 100 150 200

(c) x (mm)
Figure 7.7 Spatial wavenumber imaging results at 360kHz: (a) for the TT crack plate;
(b) for the 3X crack plate; (c) for the hTT crack plate.
3
Wavenumber (rad/mm)
2.5
2
TT
1.5 3X
hTT
1
-10 -5 0 5 10
Distance to the crack center (mm)
Figure 7.8 Wavenumber distributions along the crack. Broken, solid and dotted lines
are for TT, 3X and hTT cracks, respectively.
111
CHAPTER 8
THICKNESS LOSS EVALUATION IN ISOTROPIC PLATES
Intensive studies have been presented in the literature about using Lamb waves to
determine the presence of damage in plates as well as quantify the damage location and
size [29-32, 103, 104, 106, 107, 109, 110, 113, 114, 140, 142, 143, 148, 149, 161, 184-
193]. However, applications of Lamb waves for quantifications of material thickness and
damage depth are much less discussed. The work presented in this chapter focuses on
material thickness and damage depth quantifications by using Lamb waves.
Recently, wavefield analysis methods have been used for the quantification of
delamination depth by using correlations with dispersion curves [32, 149, 194, 195]. In
this study, in order to quantify the material thickness as well as the damage depth in
isotropic plates, a dispersion curve regression method is developed and verified through
experiments. It is shown that the dispersion curve regression method can quantify the
material thickness and damage depth in isotropic plates. Moreover, this technique can be
potentially used for detection and quantification of corrosion damage.
8.2. WAVE INTERACTION WITH MATERIAL LOSS DAMAGE
The test specimen is a 610×610×3.2 (unit: mm) aluminum 6061-T6 plate with material
loss damage. The damage is created by removing a portion of the plate material using a
milling machine. By removing different portions of the material in the same area, damage
with different depths can be created. Figure 8.1 a gives a photo of damage with a
112
diameter of 25.4 mm and a depth of 0.5 mm. Four different cases are studied, when the
damage depths are 0.25 mm, 0.50 mm, 0.75 mm and 1.00 mm. Figure 8.1 a gives the
overall experimental setup. Figure 8.1 c is a schematic of the PZT-SLDV sensing layout.
A PZT (APC 851 type with diameter 7 mm diameter and 0.2 mm thickness) is installed
on the plate surface to excite Lamb waves. The center of the PZT is set as the coordinate
origin. The center of damage is at (0, 60) mm. A SLDV is used to acquire the time-space
wavefields v(t, x) in the scanning area. Once the wavefield is acquired, it is transformed
to the spectral field V(f, x) through FT.
Oscilloscope
Test plate
Function
generator
Amplifier
(a)
(b) Laser head
80
Scanning area
25.4
Material
loss
80
damage
y
20
x
PZT actuator
(c) Unit: mm
Figure 8.1. Experimental setup: (a) material loss damage with a diameter of 25.4 mm
and a depth of 0.5 mm on the front side of the test specimen, (b) photo of the setup, and
(c) schematic of the sensing layout.
113
20
Amplitude (V)
0
-20
0 10 20 30 40 50
(a) Time (μs)
Normalized
amplitude
0.5
0
0 200 400 600 800 1000 1200
(b) Frequency (kHz)
Figure 8.2. Chirp excitation: (a) the waveform, and (b) the frequency spectrum
The PZT excitation is a chirp signal with wide band frequency components.
Figure 8.2 a and b plot the waveform and the frequency spectrum of the chirp excitation.
In the frequency range 300~800 kHz, the chirp signal has stable dominant magnitudes.
According to Lamb wave dispersion curves, for the chirp excitation over 300 to 800 kHz,
both the fundamental A0 and S0 Lamb modes can be generated. Since the A0 mode has
smaller wavelengths and is more sensitive to thickness variations than the S0 mode, the
study focuses will be on the application of A0 mode for the quantification of material
thickness and damage depth.
The A0 mode is extracted from the spectra field V(f, x) through the frequency-
wavenumber filtering technique. Figure 8.3 a and b plot real parts of the A0 mode spectral
fields at 300 kHz and 500 kHz when the damage depth is 0.75 mm. The spectral fields
show complex wave propagation and interaction. At the location of the damage,
discontinues of wave fronts can be observed. For further characterization, the spatial
wavenumber analysis is applied, with the results given in Figure 8.4 a and b. The spatial
wavenumber images show an area with larger wavenumber values. The larger
114
wavenumber values are induced by the thickness reduction at the material loss damage,
for the reason that the wavenumber increases with the decrease of material thickness (as
indicated in Figure 8.5). In addition, the location and size of the area with larger
wavenumbers agree well with those of the actual material loss damage. Although the
damage location and size can be determined from the spatial wavenumber images in
Figure 8.4 a and b, the damage depth is still unknown.
100 1 100 1
80 0.5 80 0.5
y (mm)
y (mm)
60 0 60 0
40 -0.5 40 -0.5
20 -1 20 -1
-40 -20 0 20 40 -40 -20 0 20 40
(b) x (mm) (b) x (mm)
Figure 8.3 Spectral fields (real parts) for the plate with 0.75 mm depth material loss
damage: (a) at 300 kHz, and (b) at 500 kHz.
100 100
0.92 1.34
Wavenumber (rad/mm)
Wavenumber (rad/mm)
80 80 1.32
0.9
1.3
y (mm)
y (mm)
60 0.88 60
1.28
0.86 1.26
40 40
0.84 1.24
20 20 1.22
-40 -20 0 20 40 -40 -20 0 20 40
(b) x (mm) (b) x (mm)
Figure 8.4 Spatial wavenumber images for the plate with 0.75 mm depth material loss
115
2
Wavenumber (rad/mm)
1.8
Wavenumber (rad/mm)
1.6
1.4
1.2
1
0.8
0.6
Thickness (mm)
Frequency (kHz)
Figure 8.5 A0 mode wavenumber versus plate thickness and wave frequency.
8.3. DISPERSION CURVE REGRESSION FOR MATERIAL THICKNESS QUANTIFICATION
The spatial wavenumber result k sw ( x , f ) can be used for damage imaging and
quantification. In addition, the wavenumber information can be used for evaluating the
material thickness, since the wavenumber is correlated to the plate thickness. Thus, the
plate thickness can be quantified by comparing the spatial wavenumber k sw ( x , f ) to the
theoretical dispersion curve k ( d , f ) , as
N
d ( x ) arg min ∑ [ k sw ( x , f i ) − k ( d , f i )]
2
= (6.45)
d
i =1
where fi (i=1,2,3…N) represents a series of frequency samples over a frequency
bandwidth ∆f, and d ( x ) is the material thickness determined through a least square
regression process. To evaluated the performance of the regression in Eq. (6.45), the
standard wavenumber deviation is calculated, as
1 N
∑[k ( x , f i ) − k ( d ( x ), f i ) ]
2
=σ k (x) sw (6.46)
N i =1
116
Moreover, the standard deviation for thickness evaluation can be expressed as
σ k (x) N
2δ d
σ d (x) =
N
∑ k (d ( x ) + δ d , f ) − k (d ( x ) + δ d , f )
i =1
(6.47)
i i
where δd is a small thickness perturbation.
Figure 8.6 gives a dispersion curve regression result at the location (0, 60) mm,
when the damage depth is 0.75 mm. In Figure 8.6, the experimental data is from the
spatial wavenumber k sw ( x , f ) . Through the regression, the dispersion curve of A0 mode
that fits the experimental data can be found, and the material thickness can be determined.
The solid line in Figure 8.6 is the dispersion curve obtained from the regression. The
dotted and broken lines represent the upper and lower boundaries of the 95% confident
region.
With the dispersion curve regression method, the material thickness can be
quantified. In addition, the thickness loss at the damage can also be found. Figure 8.7 a, b,
c and d plot the thickness loss images for the four cases when the depths of defects are
0.25 mm, 0.50 mm, 0.75 mm and 1.00 mm, respectively. These four images clearly show
the locations, sizes and depths of the four defects.
117
2
1.8
Wavenumber (rad/mm)
1.6
1.4
1.2 Experimental data

Regression curve k(f)
1 k(f)+2k𝜎𝜎
k(f)-2k𝜎𝜎
300 400 500 600 700 800
Frequency (kHz)
Figure 8.6 Dispersion curve regression result at the location (0, 60) mm, when the
damage depth is 0.75 mm. The experimental data is from the spatial wavenumber
k sw ( x , f ) . The solid line is the dispersion curve obtained from the least square regression.
The dotted and broken lines represent the upper and lower boundaries of the 95%
confident region.
100 1.2 100 1.2
1 1
Thickness loss (mm)
Thickness loss (mm)

80 80
0.8 0.8
y (mm)
y (mm)
60 0.6 60 0.6
0.4 0.4
40 40
0.2 0.2
20 0 20 0
-40 -20 0 20 40 -40 -20 0 20 40
(a) x (mm) (b) x (mm)
100 1.2 100 1.2
1 1
Thickness loss (mm)
Thickness loss (mm)
80 80
0.8 0.8
y (mm)
y (mm)
60 0.6 60 0.6
0.4 0.4
40 40
0.2 0.2
20 0 20 0
-40 -20 0 20 40 -40 -20 0 20 40
(c) x (mm) (d) x (mm)
Figure 8.7 The thickness loss quantification results: (a), (b), (c) and (d) are for the
plates with 0.25 mm, 0.50 mm, 0.75 mm and 1.00 mm depth damage, respectively.
118
8.4. EVALUATION OF SIMULATED PITTING CORROSION DAMAGE
In this subsection, the developed material loss evaluation method is used to detect and
quantity pitting corrosion damage. Figure 8.8 a plots the sensing layout. The pitting
corrosion damage is simulated by a nine-hole cluster on a 3 × 3 grid with the center at (0,
60) mm. Figure 8.8 b gives a photo of the simulated pitting corrosion. The holes are half
ball shaped with the diameter of 2 mm, which are created by a ball-end end mill. The
spatial intervals in x and y directions are both 4 mm.
80
Scanning area
2
Simulated
pitting
80
corrosion
damage
4
y
4
20
Unit: mm
x (b)
PZT actuator
(a) Unit: mm
Figure 8.8. Experimental setup: (a) schematic of the sensing layout, and (b) photo of
pitting corrosion damage simulated by a nine-hole cluster.
100 1 100 1
80 0.5 80 0.5
y (mm)
y (mm)
60 0 60 0
40 -0.5 40 -0.5
20 -1 20 -1
-40 -20 0 20 40 -40 -20 0 20 40
(b) x (mm) (b) x (mm)
Figure 8.9 Spectral fields (real parts) for the plate with simulated pitting corrosion
119
Figure 8.9 a and b plot real parts of spectral fields measured at 300 kHz and 500
kHz. It is hard to see any spectral field discontinuity induced by the damage, since the
influence of the damage is so weak. The dispersion curve regression method is applied to
the spectral field V(f, x) with the result plotted in Figure 8.10. The imaging result clearly
shows an area with larger thickness loss than the rest. The location and size of this area
agree well with those of the simulated pitting corrosion damage (the nine-hole cluster).
Therefore, the presented method has the capability of detecting pitting corrosion damage
and quantifying its size. Note that although the imaging result in Figure 8.10 shows the
nine-hole cluster, the 2 mm diameter small holes are not distinguished in space. One
possible reason of this effect is that the wavelengths of incident waves are too large to
distinguish the holes with such small intervals in the nine-hole cluster.
100 1.2
1
Thickness loss (mm)
80
0.8
y (mm)
60 0.6
0.4
40
0.2
20 0
-40 -20 0 20 40
(b) x (mm)
Figure 8.10. Imaging result of the plate with simulated pitting corrosion damage.
120
CHAPTER 9
DELAMINATION DETECTION AND QUANTIFICATION IN COMPOSITE LAMINATES
Unexpected damage can occur in aerospace composites due to impact events or material
stress during off-nominal loading events. In particular, laminated composites are
susceptible to delamination damage due to weak transverse tensile and interlaminar shear
strengths [191]. Developments of reliable and quantitative techniques to detect
delamination damage in laminated composites are imperative for safe and functional
optimally-designed next-generation composite structures. Delamination damage can
occur largely internally in composites, while damage indications are barely visible on
composite surfaces to the naked eye [196]. Such hidden delamination damage must be
detected and evaluated before it becomes critical. The problem of how best to detect such
damage in laminated composites is still being solved, and a solution to this problem will
be imperative for safety-critical composites. Current composite “safety margin” design
comes with a weight penalty and could be mitigated by efficient and reliable NDE/SHM
methods to rapidly and accurately locate and quantify the damage in composite materials.
Among various NDE/SHM methods, ultrasonic guided waves have proven
effective for damage detection in plate-like metallic and composite structures due to their
ability to inspect large areas while maintaining sensitivities to small defects in the
structures [1, 8, 36, 197-199]. Advances in guided wave based NDE/SHM technologies
over the last decade have demonstrated feasibilities of detecting and locating damage in
121
composite structures [29, 32, 74, 136, 184, 193, 200-208]. Despite these advances, the
state-of-the-art is still insufficient in regards to a quantitative understanding of the
mechanism of wave interaction with delamination damage in composites. In addition, the
state-of-the-art is insufficient for evaluating delamination damage details that are
necessary to feed into damage progression models. Last but not least, most delamination
detection methods focus on locating the delamination in plane, without quantifying the
delamination depth.
This chapter presents studies of guided wave interaction with delamination
damage in laminated composites, and delamination detection and quantification. EFIT
simulations and PZT-SLDV experiments are performed to investigate guided wave
interaction with delamination damage. In order to understand complex wave interactions
that occur when incident waves encounter delamination damage, guided wavefields and
wavenumber spectra are analyzed. These analyses are applied to a composite laminate
plate containing a simulated (Teflon insert) single delamination and to a plate with
impact-induced delamination damage (multiple ply layer delamination damage).
To detect and quantify impact-induced delamination damage, this chapter presents
a combined phased array and wavefield based imaging methodology. First, the phased
array imaging is used to generate an intensity image of the composite being inspected to
visualize and locate delamination damage in the composite. Once the delamination
damage is detected, wavefield based imaging methods such as filter reconstruction
imaging and spatial wavenumber imaging are performed to further quantify the damage.
The systemic combination of phased array imaging and wavefield based imaging
122
provides a hierarchical damage diagnosis, which can efficiently and effectively locate the
damage, and quantify the damage size and depth.
9.2. WAVE INTERACTION WITH DELAMINATION DAMAGE
Compared to guided waves in isotropic materials, guided waves in composite materials
present additional complexities, especially when guided waves interact with delamination
damage [192]. When guided waves propagate in delaminated composites, multiple
reflections induced by delamination damage can occur, as shown in references [103, 141,
143, 209, 210]. Hence, a considerable amount of wave energy can be “trapped” above,
below, and between delaminated regions until eventually the energy dissipates. Ramadas
et al. studied the interaction of the anti-symmetric A0 mode with symmetric delamination
damage [210]. They identified multiple reflections in the delamination region. The
trapped energy phenomenon of guided waves in delaminated composites has also been
studied experimentally through the use of wavefield images obtained from laser
vibrometry tests [103, 141]. It has been found that upon entering the delaminated region,
the incident waves will be split above and below the delamination, propagating
independently through upper and lower laminates (for a single delamination scenario). As
waves propagate towards the edge of delamination damage (exit region), a portion of
wave energy is reflected backwards from the far edge of the delamination. The reflected
waves in the damage region then pass back above (or below) the delaminated area and
experience reflections at the original entrance to the delamination region. The process is
repeated until the energy is dissipated, as found in the simulation work reported by
Hayashi and Kawashima [209]. Some portion of the wave energy exits the delaminated
region into the pristine region, re-forming the guided wave modes that exist in the
123
pristine plate [74]. Glushkov et al. also observed similar wave energy trapping and
localization in a strip delamination from wavefields measured by laser vibrometry [143].
Guided wave energy trapping in delamination damage has recently been quantitatively
studied [149]. As expected from the guided wave theory, the trapped waves have
different wavenumbers compared to those traveling in the undamaged area. Therefore, it
is expected that by analyzing the trapped waves, the delamination dimensions can be
quantified. Rogge and Leckey demonstrated that wavenumber analysis can be used to
determine the approximate depth and size of near-surface delamination damage through a
localized wavenumber field imaging method [32]. In this work, both EFIT simulations
and PZT-SLDV experiments are performed to investigate the wave interaction with
delamination damage.
9.2.1. Simulations of guided waves in composites
The EFIT simulations are implemented to investigate the guided wave interaction with
delamination damage. The guided wave interaction with delamination damage in a
[02/902]s layup CFRP composite plate is simulated. Moreover, interactions of A0 and S0
modes with delamination damage are simulated and analyzed. The simulation results
show that when incident waves interact with the delamination damage, mode conversion,
reflection and transmission waves appear. Moreover, waves in the delamination area
show different wave lengths and wave speeds compared to those in the rest area. The
details of the EFIT model setup and simulation results can be found in Chapter 3.
9.2.2. Experimental characterizations using the PZT-SLDV sensing
Experiments are performed using the PZT-SLDV sensing. Two 380 mm×380 mm 8-ply
IM7/8552 composite laminate plates (material properties in Table 9.9.1) with the same
124
layup [02/902]S are provided by NASA Langley Research Center (LaRC). One plate is the
pristine, while the other contains delamination damage created by inserting a 20 mm × 20
mm Teflon film under the second ply during manufacturing, as shown in Figure 9.1.
The overall experimental setup is shown in Figure 9.1 c. Guided waves are
actuated with a surface bonded 0.2 mm thick and 7 mm round PZT patch. The PZT
excitation signal is a 3-cycle toneburst with 300 kHz frequency and 40 V amplitude. The
out-of-plane wavefield data are acquired using the SLDV. The spatial resolution of the
SLDV scan grid is set at 0.2 mm.
Figure 9.2 plots time-space wavefields of guided waves propagating along the y-
axis (1-D wave propagation). For the pristine case in Figure 9.2 a, a faster S0 mode and a
slower A0 mode can clearly be observed. For the delaminated case in Figure 9.2 b, guided
waves propagating back and forth between the top (y=20 mm) and bottom (y=40 mm)
boundaries of the delamination can be distinctively seen, exhibiting the “trapped wave”
phenomenon [141-143].
Figure 9.3 a and b plot snapshots of wavefields at the time 15 μs in the scan area
(2-D wave propagation) in the pristine and delaminated plates. In the snapshot of the
pristine plate, two wave modes, a faster S0 mode and a slower A0 mode, are observed. In
the snapshot of the delaminated plate, interactions at the delamination region are
observed as waves propagating between delamination boundaries (denoted by the dotted
box), after the incident S0 mode has passed the delamination region.
Table 9.1 Material properties of the IM7/8552 lamina.

ρ E1 E2 E3 G12 G13 G23 Thickness
ν12 ν13 v23
(kg/m3) (GPa) (GPa) (GPa) (GPa) (GPa) (GPa) (mm)
1570 171.4 9.08 9.08 5.29 5.29 2.80 0.32 0.32 0.34 0.1068
125
40
Scan area
20 Ply 1
The upper 2 plies:
[0]2 layup
Teflon Teflon insert
55
20
insert
[02/902]S layer
layup The lower 6 plies:
y [904/02] layup
20
Ply 8
(b)
x
Unit: mm PZT actuator
(a)
Oscilloscope Test plate
Laser head
Function
(c) generator Amplifier
Figure 9.1 The scanning laser vibrometry test: (a) the sensing layout, (b) the side view
of the [02/902]S composite plate with a Teflon insert, and (c) the experimental setup.
60 6 60 6
Top
50 4 50 boundary 4
Velocity (mm/s)
Velocity (mm/s)
40 2 40 2
y (mm)
y (mm)
0 0
30 30
S0 -2 delamination damage -2
20 20
-4
10
S0 Bottom-4
10 A0 A0
-6 boundary
-6
0 10 20 30 0 10 20 30
Figure 9.2 Time-space wavefields of waves propagating along the y axis: (a) for the
pristine plate, and (b) for the plate with a Teflon insert. The two solid black lines at 20
mm and 40 mm are bottom and top boundaries of the delamination area.
126
60 6 60 6
Top
50 4 50 boundary 4
Velocity (mm/s)
Velocity (mm/s)
2 2
40 S0 40
y (mm)
y (mm)
0 0
30 30
-2 -2
20 20
-4 -4
A0 Bottom
10 10
boundary
-6 -6
-20 -10 0 10 20 -20 -10 0 10 20
(a) x (mm) (b) x (mm)
Figure 9.3 SLDV wavefield snapshots at 15 µs: (a) for the pristine plate, and (b) for the
plate with a Teflon insert. The red dotted box indicates the actual Teflon insert.
9.2.3. Frequency-wavenumber analysis
Complicated localized wave interactions at the Teflon delamination region are observed
in the raw wavefield snapshot. To better understand the wave interactions occurring at the
delamination region, time-space wavefields are converted to frequency-wavenumber
representations by the 3-D FT. Figure 9.4 a and b plot the wavenumber spectra at 300
kHz achieved by the 3-D FT for the pristine and delaminated cases, respectively. To
identify wave mode information, wavenumber spectra are compared to theoretical
wavenumber curves which are plotted as dotted (A0) and solid (S0) lines. The pristine
plate wavenumber spectrum (Figure 9.4 a) shows components, which match well with the
theoretical wavenumber curves of A0 and S0 modes in the [02/902]S composite.
In the Teflon delaminated plate, the delamination damage splits the originally
well bonded full laminate into two parts (above and below the delamination). Since the
two separated material portions have different layups and thicknesses, compared to the
original pristine laminate, the guided waves in the delaminated region will have modified
propagation characteristics (such as modified dispersion and wavenumbers).
127
A0 in [0]2
0 laminate 0
4

4

-5 -5
A0 A0
2 -10 2 -10
ky (rad/mm)
ky (rad/mm)
S0 in [0]2
0 S0 -15 0 S0 laminate -15
-2 -20 -2 -20
-25 -25
-4 -4
-30 -30
-4 -2 0 2 4 -4 -2 0 2 4
Figure 9.4 Wavenumber spectra at 300 kHz excitation frequency: (a) for the pristine
plate, and (b) for the plate with a Teflon insert. The dotted and solid lines are theoretical
wavenumber curves for A0 and S0 modes.
As expected, the spectrum of the delaminated case (Figure 9.4 b) shows
wavenumber components in addition to the original A0 and S0 modes. The additional
wavenumber components include both positive and negative counterparts, which
correspond to forward and backward propagating waves, respectively. The additional
wavenumber components are compared to wavenumber curves of the material above the
delamination, equivalent to a two ply [0]2 laminate, since only the waves in the top plate
surface were measured in the experiment. The comparison in Figure 9.4b shows the
additional wavenumber values match well with the theoretical curves of A0 and S0 modes
in the two ply [0]2 laminate.
9.3. DETECTION AND QUANTIFICATION OF SIMULATED DELAMINATION DAMAGE
9.3.1. Filter reconstruction imaging
From the frequency-wavenumber representation, additional wavenumber components are
identified and believed to be related to the trapped waves in the delamination region. To
confirm the correlation between the additional wavenumbers and the trapped waves, a
filtering technique is applied to acquire and study only the additional wavenumbers in the
128
spectrum given in Figure 9.4 b, and then to reconstruct them back to the time-space
domain.
To retain the additional wavenumbers that are believed to be correlated with the
delamination (as indicated in Figure 9.4 b), a “band-pass” wavenumber filter is designed,
as shown in Figure 9.5 a. The filter is then implemented by multiplying the band-pass
filter with the spectrum V(f, k). The resulted filtered spectrum in Figure 9.5 b only
contains the delamination induced additional wavenumbers.
To understand what waves the filtered spectrum VF(f, k) represents, the
corresponding time-space representation vF(t, x) is reconstructed by transforming the
filtered spectrum VF(f, k) back to the time-space domain. Figure 9.6 a, b and c are filtered
wavefields at 5 μs, 15 μs and 25 μs, respectively. It is seen that the additional
wavenumbers correspond to waves that are well confined in the delaminated area.
With the filtered wavefield vF(t, x), we can therefore generate an intensity map or
image of the cumulative wave energy that is trapped in the delamination area. The
resulting energy map of the delaminated plate is given in Figure 9.6 d showing high
energy concentrations at the delaminated region due to the wave trapping phenomenon.
The energy map achieved from the filter reconstruction imaging can locate and roughly
image the size and shape of the delamination damage in the laminated composite.
129
1
Amplitude
0.5
0
0 1 2 3 4 5
kx (rad/mm)
1 0
4 4

0.8 -5
2 2 -10
Amplitude
ky (rad/mm)
ky (rad/mm)
0.6
0 0 -15
0.4
-2 -2 -20
0.2
-25
-4 -4
0 -30
-4 -2 0 2 4 -4 -2 0 2 4
Figure 9.5 The filtering process in wavenumber domain: (a) band-pass wavenumber
filter; (b) filtered spectrum for the plate with a Teflon insert. The dotted line is the
theoretical wavenumber curve of the A0 mode in the [0]2 laminate above the delamination.
60 6 60 6
50 4 50 4
Velocity (mm/s)
Velocity (mm/s)
2 40 2
40
y (mm)
y (mm)
0 0
30 30
-2 -2
20 20
-4 -4
10 10
-6 -6
-20 -10 0 10 20 -20 -10 0 10 20
(a) x (mm) (b) x (mm)
60 6 60 1
4
50 50 0.8
Velocity (mm/s)
2
40 40
0.6
y (mm)
y (mm)
0
30 30
0.4
-2
20 20
-4 0.2
10 10
-6 0
-20 -10 0 10 20 -20 -10 0 10 20
(c) x (mm) (d) x (mm)
Figure 9.6 The filter reconstruction imaging results: (a) at 5 μs, (b) at 15 μs, (c) at 25 μs,
and (d) energy map. The red dotted box indicates the actual Teflon insert.
130
9.3.2. Spatial wavenumber imaging
The spatial wavenumber analysis is also applied for delamination detection and
quantification. Figure 9.7 gives the spatial wavenumber image calculated from the
wavefield data of the delaminated plate at the excitation frequency 300 kHz. The spatial
wavenumber image indicates an area approximately 20 mm × 20 mm in size which has
significantly larger wavenumber values than the rest of the plate. This region matches
well with the Teflon insert delamination location, shape, and size (indicated by the red
dotted box in Figure 9.7). Clearly, the delamination affects wavenumbers of waves that
interacts in the delamination region. The spatial wavenumber imaging shows its potential
for detecting and quantifying the delamination damage.
Compared to the filter reconstruction imaging, the spatial wavenumber imaging is
more robust. For the filter reconstruction imaging, wavenumber components need to be
identified and selected from the wavenumber spectrum for the filtering process. However,
if the excitation frequency and/or plate thickness increases, more wave modes are present
and propagate in the plate. It can become difficult to identify and isolate specific
wavenumber values in the spectrum of the damaged plate for the filter reconstruction
imaging. In contrast, the spatial wavenumber imaging can directly generate a spatial map
of wavenumber variations, and the delamination damage is directly observable as a
region with altered wavenumbers.
131
60 3.5
Wavenumber (rad/mm)
50
3
40
y (mm)
30
2.5
20
2
10
-20 -10 0 10 20
x (mm)
Figure 9.7 Spatial wavenumber image of the Teflon delaminated plate. The red dotted
box indicates the actual Teflon insert.
9.3.3. Delamination depth quantification
It has been shown that the simple delamination (such as the Teflon insert in the
experimental specimen) that splits the originally bonded laminate into two parts will
result in waves propagating above and below the delamination. Since the materials above
and below the delamination have different layups and thicknesses compared to the
original pristine laminate, waves measured at an outer plate surface region above (or
below) the delamination will have modified propagation characteristics and
wavenumbers (which will depend on the number of plies in the split parts). Hence, the
delamination depth can be estimated by relating the wavenumbers to the number of plies
above (or below) the delamination.
In this study, the plies from the surface where waves are measured are numbered
as ply 1, 2… etc., as shown in Figure 9.1 b. A set of theoretical dispersion curves is
generated for laminates with different plies. Figure 9.8 a plots dispersion curves of A0
modes in laminates of 1 to 8 plies. For example, a 5 ply laminate would consist of the
uppermost 5 plies, [0/0/90/90/0]. Figure 9.8 b plots a theoretical wavenumber curve at
300 kHz for laminates of 1 to 8 plies. The theoretical wavenumber curve shows a relation
132
of wavenumber versus the number of plies. Therefore, if we know the wavenumber at the
delamination damage, by fitting the delamination induced wavenumber to the theoretical
wavenumber curve, the number of plies above the delamination (delamination depth) can
be determined. For the spatial wavenumber image in Figure 9.7, the wavenumber is
around 3.5 rad/mm in the delamination area. By comparing the wavenumber 3.5 rad/mm
to the theoretical wavenumber curve in Figure 9.8 b, we can find the delamination
damage is under the second ply.

(a) Frequency (kHz) (c) Number of plies
Figure 9.8 Dispersion curves of A0 modes propagating towards the 90° direction: (a)
dispersion curves for laminates of 1 to 8 plies; (b) wavenumbers at 300 kHz vs. the
number of plies. (Note: the number of plies is counted from the top lamina).
9.4. DETECTION AND QUANTIFICATION OF IMPACT-INDUCED DELAMINATION DAMAGE
Although wavefield based imaging methods can give promising delamination detection
and quantification results, these methods are based on full wavefields that usually take a
long time to acquire since the SLDV needs to measure guided waves at a large number of
points in the scanning area. Moreover, the wavefield based imaging methods only detect
and quantify damage in the scanning area.
In contrast, the SLDV phased array imaging is a more efficient way to quickly
inspect the specimen. The phased array imaging only needs to acquire guided waves in a
133
small scanning area. Moreover, the phased array imaging can detect damage outside the
scanning area. Therefore, the SLDV phased array imaging is first used as a quick means
to inspect the composite plate and to locate the delamination damage. Once the
delamination damage is detected, wavefield based imaging methods are performed to
further quantify the damage. The systemic combination of phased array imaging and
wavefield based imaging provides a hierarchical damage diagnosis, which can efficiently
and effectively locate the damage, and further quantify the damage. In this section, the
combined phased array and wavefield based imaging methodology is used to detect and
quantify impact-induced delamination damage in a composite plate.
9.4.1. Impact-induced delamination damage
The test specimen is a 26 ply IM7/8552 composite laminate (material properties listed in
Table 9.9.1) with a quasi-isotropic layup of [(02/452/-452)2/90]S and dimensions 381 mm
× 381 mm provided by LaRC. The delamination damage is created by a quasi-static
indentation, which has been shown by other researchers to be an ideal method for
controlled growth of impact-induced damage in composite laminates [32, 211]. Figure
9.9 a gives the quasi-static indentation setup. The composite plate is loaded in a
mechanical testing system using a 50.8 mm diameter indenter at a constant displacement
rate of 1.27 mm/min. The displacement and load are monitored during the indentation
test and the composite is unloaded when the data shows a sudden drop in the load (due to
damage growth), as shown in Figure 9.9 b. A photo of the composite plate after the
indentation test is shown in Figure 9.10 a. In the zoomed in view, Figure 9.10 b, a very
slight indentation can be observed on the surface with no other damage.
134
Following the quasi-static indentation test, an ultrasound C-scan is conducted at
LaRC to quantify the delamination damage in the composite plate. The ultrasound data
are collected in an immersion tank using a 10 MHz transducer and a scan step size of
0.25 mm. The resulting C-scan image in Figure 9.10 c shows multiple delaminations
within a 32.3 mm × 22.6 mm damage zone, occurring at various depths (ply layers)
through the plate thickness. The figure notes the ply depths at which the delaminations
occur, determined through time of flight measurements from the ultrasound data.
50.8 mm diameter Displacement rate Unloading

indenter 1.27 mm/min
Force (kN)
Loading
Damage
26 plies growth
composite plate
(a) (b) Displacement (mm)
Figure 9.9 Quasi-static indentation for delamination damage growth: (a) the indentation
setup, and (b) the curve of indentation force versus displacement.
1
5 14
16 0.8
10-11
7
Cscan intensity
0.6
22.6 mm
9
0.4
9
7 0.2
5
26.1 mm 16 14-15
0
(a) (b) (c)
32.3 mm
Figure 9.10 The 26 ply laminate after the quasi-static indentation test: (a) a photo of the
specimen, (b) a zoomed in view of the indentation, and (c) an ultrasonic C-scan image
showing internal delamination damage generated by the indentation test. The red solid
lines in figure (c) represent boundaries of delaminations. The numbers in figure (c)
represent delamination depths. For example, the number 5 represents the delamination is
under the fifth ply.
135
9.4.2. Delamination detection using SLDV phased arrays
9.4.2.1. Experimental setup
Figure 9.11 plots a schematic of the SLDV phased array sensing layout for the impact
induced delamination detection. A PZT wafer (7 mm diameter, 0.2 mm thickness) is
bonded on the plate to generate guided waves. The center of the PZT is set as the
coordinate origin. The center of the delamination is at the location (0 mm, 100 mm). A 3-
cycle toneburst with the frequency 150 kHz and amplitude 30V is applied to the PZT
wafer for excitation. The time-space wavefield in the scanning area, a 45 mm ×45 mm
square with the center at the coordinate origin, is measured by the SLDV. The horizontal
and vertical spatial resolutions of the scanning are 0.1 mm. At 30μs, the measured
wavefield in Figure 9.12 a shows incident waves generated from the PZT wafer. At 140
μs, the wavefield in Figure 9.12 b shows guided waves reflected by the delamination
damage. The wavenumber spectra in Figure 9.12 c and d reveal wavenumber components
on the wavenumber curve of A0 mode, indicating that the incident and reflected waves
are A0 mode.
Delamination
damage
processing
y
Scanning
area O
x SLDV Trigger
PZT
amplifier signal generator
[(02/452/-452)2/90]S laminate
Figure 9.11 The SLDV phased array sensing layout for detecting and imaging the
impact-induced delamination damage.
136
9.4.2.2. Array imaging results
31×31 points with the intervals dx =dy =2 mm, and the center at the origin are chosen
from the SLDV scanning area to generate a grid array. Applying array imaging
algorithms presented in Chapter 5, a beamformed wavefield z(t, x) is obtained. The
wavefield images at t =80 μs, 40 μs and 0 μs (Figure 9.13 a, b and c) show beamformed
waves propagate in the 90° direction. At the time t = 80 μs, waves propagate from the
phase array to the delamination damage. At the time t = 0 μs, waves arrive at the
delamination damage. The imaging result in Figure 9.13 d reveals a damage area round
the location (-1.1, 103.5) mm with the error (-1.1, 3.5) mm.
t = 30 μs t = 140 μs
40 4 40 0.06
0.04
20 2 20
Velocity (mm/s)
Velocity (mm/s)
0.02
y (mm)
y (mm)
0 0 0 0
-0.02
-20 -2 -20
-0.04
-40 -4 -40 -0.06

-40 -20 0 20 40 -40 -20 0 20 40
(a) x (mm) (b) x (mm)
1 0.05
1 1
0.8 0.04
0.5 0.5
ky (rad/mm)
ky (rad/mm)
0.03
0.6
0 0
0.02
-0.5 0.4 -0.5
0.01
-1 0.2 -1
0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
Figure 9.12. SLDV measurements: (a) and (b) are wavefields at 30μs and 140 μs,
showing the incident and reflected waves; (c) and (d) are wavenumber spectra of incident
and reflected waves. The solid white lines in (c) and (d) are dispersion curves of the A0
mode in the full layup laminate.
137
t = 80 μs t = 50 μs
15 15
100 10 100 10
Velocity (mm/s)
Velocity (mm/s)
50 5 50 5
y (mm)
y (mm)
0 0 0 0
-50 -5 -50 -5
-100 -10 -100 -10
-15 -15
-100 -50 0 50 100 -100 -50 0 50 100
(a) x (mm) (b) x (mm)
t = 0 μs Detection result
15 1
100 10 100 0.9
0.8
Velocity (mm/s)
50 5 50
0.7
y (mm)
y (mm)
0 0 0
0.6
-50 -5 -50
0.5
-100 -10 -100 0.4
-15 0.3
-100 -50 0 50 100 -100 -50 0 50 100
(c) x (mm) (d) x (mm)
Figure 9.13 Phased array imaging results for detecting and imaging the impact-induced
delamination damage: (a), (b), (c) are beamformed wavefields at 80, 40 and 0 μs, and (d)
is the delamination detection result.
9.4.3. Delamination quantification using wavefield based imaging
Once the delamination damage is detected by the efficient SLDV phased array imaging, a
refined scan will be performed in the damage region in order to perform the wavefield
based imaging for damage quantification.
9.4.3.1. Experimental setup
Figure 9.14 plot a schematic of the experimental setup for the quantification of the
impact-induced delamination damage. A 0.2 mm thick 7 mm round PZT is installed on
the plate surface to actuate the guided waves. The center of the PZT is set as the origin.
138
The wavefields at two different excitation frequencies (low frequency 100 kHz and high
frequency 500 kHz) are measured using the SLDV. Figure 9.15 a and b plot wavefields
measured at the 100 kHz excitation frequency for the pristine and damaged plates. The
pristine wavefield shows a single incident wave mode with a wavelength of 11.5 mm.
The damaged wavefield shows strong wave interactions at the damage, compared to the
pristine case. Figure 9.16 a and b plot wavefields at the 500 kHz excitation frequency for
the pristine and damaged plates. Compared to the low frequency wavefield at 100 kHz in
Figure 9.15 a, the pristine wavefield at 500 kHz in Figure 9.16 a is much more complex.
For example, Figure 9.16 a shows various wave modes with different wavelengths, while
the wavefield at 100 kHz only has one incident wave mode. Figure 9.16 b plots the 500
kHz wavefield for the damaged case. The wavefield shows strong wave interactions at
the damage region. Moreover, the interaction pattern at 500 kHz in Figure 9.16 b is more
complex than that at 100 kHz in Figure 9.15 b.
40 mm
Scanning area
Delamination
55 mm
damage
PZT actuator x
Figure 9.14 Experimental setup for quantifying the impact-induced delamination.
139
90 90
0.6 0.6
80 80
0.4 0.4
70
Velocity (mm/s)
70
Velocity (mm/s)
0.2 0.2
60 60
y (mm)
y (mm)
0 0
50 50
-0.2 -0.2
40 40
-0.4 Wave -0.4
30 30 interaction
-0.6 -0.6
20 20
-20 -10 0 10 20 -20 -10 0 10 20
(a) x (mm) (b) x (mm)
Figure 9.15 SLDV wavefield images at 45 µs for 100 kHz excitation: (a) for the pristine
plate, and (b) for the plate with damage.
90 90
0.4 0.4
80 80
Velocity (mm/s)
70
Velocity (mm/s)
70 0.2 0.2
60 60
y (mm)
y (mm)
0 0
50 50
40 -0.2 40 -0.2
Wave
30 30 interaction
-0.4 -0.4
20 20
-20 -10 0 10 20 -20 -10 0 10 20
(a) x (mm) (b) x (mm)
Figure 9.16 SLDV wavefield images at 40 µs for 500 kHz excitation: (a) for the pristine
plate, and (b) for the plate with damage.
9.4.3.2. Filter reconstruction imaging
To quantify the impact-induced delamination damage, the filter reconstruction imaging
method is applied to wavefields measured from experiments. Figure 9.17 shows the filter
reconstruction imaging process and result at the low frequency 100 kHz. In the pristine
wavenumber spectrum Figure 9.17 a, there is only the incident A0 wave mode, while
additional wavenumber components present in the damaged plate spectrum Figure 9.17 b.
The wavenumbers corresponding to the A0 mode is removed using a band-stop filter. The
resulted filtered spectrum is plotted in Figure 9.17 c. The energy map of the filtered
components is given in Figure 9.17 d. The delamination damage shows up as an area with
high energy which matches well with the C-scan result (contour plot).
140
4 0 4 0

2 -10 2 -10
A0 A0
ky (rad/mm)
ky (rad/mm)
0 -20 0 -20
-2 -30 -2 -30
-4 -40 -4 -40
-4 -2 0 2 4 -4 -2 0 2 4
90 1
4 0
80
0.8
2 -10 70
ky (rad/mm)
0.6
60
y (mm)
0 -20 50 0.4
40
-2 -30 0.2
30
-4 -40 20 0
-4 -2 0 2 4 -20 -10 0 10 20
(c) kx (rad/mm) (d) x (mm)
Figure 9.17 Filter reconstruction imaging at the low frequency 100 kHz: (a)
wavenumber spectrum for the pristine plate, (b) wavenumber spectrum for the damaged
plate, (c) filtered wavenumber spectrum, and (d) energy map. The red solid lines in figure
(d) represent boundaries extracted from the C-scan image.
Figure 9.18 shows the filter reconstruction imaging process and result at the high
frequency 500 kHz. Compared to the pristine spectrum at the low frequency 100 kHz, the
pristine spectrum at 500 kHz in Figure 9.18 a contains more wave modes, including the
A0, A1, S0 S1 and S2 modes. Moreover, the identification and isolation of additional
wavenumbers in the spectrum of the damaged plate become more difficult than the low
frequency case. A high-pass filter is applied to extract the additional wavenumbers that
are larger than the A0 mode. The filtering result in wavenumber domain is shown in
Figure 9.18 c. The energy map of the filtered components is constructed and plotted in
Figure 9.18 d. Compared to the low frequency 100 kHz, the energy map at 500 kHz
141
matches with a smaller portion of the C-scan result. It is expected that the portion of
damage that does not show up in Figure 9.18 d may coincide with the delamination
induced additional wavenumbers appearing in Figure 9.18 b that are cut out by the
chosen filter.
4 0 4 0

2 -10 2 -10
ky (rad/mm)
ky (rad/mm)
A0 A0
0 -20 0 -20
S0 S0
A1 A1
-2 -30 -2 -30
S1 S1
S2 S2
-4 -40 -4 -40
-4 -2 0 2 4 -4 -2 0 2 4
90 1
4 0
80
Wavenumber ky (rad/mm)
0.8
2 -10 70
60 0.6
y (mm)
0 -20
50 0.4
40
-2 -30 0.2
30
-4 -40 20 0
-4 -2 0 2 4 -20 -10 0 10 20
(c) Wavenumber kx (rad/mm) (d) x (mm)
Figure 9.18 Filter reconstruction imaging at the high frequency 500 kHz: (a)
wavenumber spectrum for the pristine plate, (b) wavenumber spectrum for the damaged
plate, (c) filtered wavenumber spectrum, and (d) energy map. The red solid lines in figure
(d) represent the boundaries extracted from the C-scan image.
9.4.3.3. Spatial wavenumber imaging
At the high frequency 500 kHz, the spatial wavenumber imaging method is applied to
quantify the impact-induced delamination damage. Figure 9.19 plots the spatial
wavenumber imaging result in the frequency band 480~520 kHz. The imaging result
shows a region with increased wavenumbers, indicating the delamination damage. The
142
increased wavenumber region is observed to fall within the boundaries of the ultrasonic
C-scan result (plotted as red solid lines in Figure 9.19).
In Figure 9.19, the detected delamination size in the wavenumber field is
approximately 30 mm × 20 mm, which agrees well with the C-scan result (32.3 mm ×
22.6 mm). The delamination shape in the spatial wavenumber image matches closely
with the uppermost portion of the delamination damage in the C-scan image. It is
expected, as reported by other researchers, that the deeper delamination damage is more
difficult to detect using wavefield based imaging methods [32]. The spatial wavenumber
result shows that the delaminations closer to the SLDV scanned composite surface are
easier to detect and result in relatively higher wavenumber modes in the damage region.
For example, in the close surface delamination region under the fifth ply, the spatial
wavenumber image gives large wavenumbers of nearly 3.4 rad/mm.
90 3.4
Wavenumber (rad/mm)
80 3.2
70 3
60
y (mm)
2.8
50
2.6
40
2.4
30
2.2
20
-20 -10 0 10 20
x (mm)
Figure 9.19 Spatial wavenumber imaging result in the frequency band 480~520 kHz.
143
CHAPTER 10
DEBONDING DETECTION AND QUANTIFICATION
IN HONEYCOMB SANDWICH STRUCTURES
Honeycomb sandwich structures have been increasingly used for aerospace, marine and
automotive applications, since they have attractive features such as high
strength/stiffness-to-weight ratio, effective acoustic insulation, noise control and
vibration damping [212-215]. However, due to the weak shear strength of the bonding
layer, honeycomb sandwich structures are susceptible to debonding damage along the
skin-core interface, especially under an intensive or repeating loading on the honeycomb
core [72, 203, 216-218]. Since the debonding damage occurs internally between the skin
panel and the honeycomb core, it is not visible to the naked eye [196]. The hidden
debonding damage must be detected and evaluated, since it affects the integrity and
safety of the whole structure.
In recent years, researchers investigated debonding detection in honeycomb
sandwich structures by using guided waves [72, 188, 203, 216-221]. However, there are
still many difficulties and challenges. Honeycomb sandwich structures have more
complicated guided waves, compared to those in the traditional simple structures like
metallic plates. In addition, honeycomb sandwich structures can have debonding damage
between the skin panel and the complex geometry honeycomb core. Effects of debonding
damage on guided waves are less studied in the literature.
144
This chapter presents the fundamental studies of guided waves in honeycomb
sandwich structures, and the debonding detection and quantification using wavefield
based imaging methods. Both finite element simulations as well as laser vibrometry
experiments are performed to investigate guided waves in honeycomb sandwich
structures. Guided wavefields at different frequencies are acquired and compared. To
further understand the guided wave propagation fundamentals, the guided wavefields are
transformed to wavenumber spectra. In the wavenumber domain, the wavenumber
components of guided waves in honeycomb sandwich structures are investigated.
Moreover, the guided wave interaction with debonding damage is investigated by using
guided wavefields and wavenumber spectra. Last, wavefield based imaging methods are
adopted to detect and quantify the debonding damage.
10.2. GUIDED WAVES IN HONEYCOMB SANDWICH STRUCTURES
10.2.1. Simulations of guided waves by using FEM
To understand how guided waves propagate in honeycomb sandwich structures, the
commercial FEM software, ANSYS/Multiphysics 11.0, is used to simulate the guided
waves. A finite element model of a honeycomb sandwich panel is built. The model setup
is presented in Chapter 3. Guided waves in the sandwich at different frequencies, 15 kHz,
50 kHz, and 100 kHz, are simulated and compared. The simulation results show that at
the low frequency 15 kHz, there are global guided waves in the entire sandwich plate,
which have large wavelengths and elliptical wave fronts. At the higher 50 kHz, although
global guided waves can still be observed, their wave fronts become closer to circular
shapes other than elliptical shapes. At the highest frequency 100 kHz, global guided
waves disappear, while guided waves in the skin plates with circular wave fronts are
145
observed. Moreover, the 100 kHz simulation result shows complex wave interactions
with the honeycomb core. The details of simulation results can be found in Chapter 3.
10.2.2. Experimental characterizations by using the PZT-SLDV sensing
Laser vibrometry tests are performed to investigate guided waves in honeycomb
sandwich structures. Figure 10.1 gives a schematic of the experimental setup. The
honeycomb sandwich (dimensions: 305 mm ×305 mm × 15 mm) is composed of top and
bottom aluminum alloy skins (T3003 aluminum alloy plates) and one hexagonal-celled ¼
inch aluminum core. The geometry parameters are listed in Table 10.10.1. Guided waves
are excited by a surface bonded PZT actuator (APC 851: 7 mm diameter, 0.2 mm
thickness) located at the coordinate origin. The SLDV is used to acquire the time-space
wavefield over the scanning area. The scanning is performed in a 135 mm × 135 mm
scanning area with a spatial resolution of 0.2 mm. At each point, the measurement is
averaged 30 times to improve the signal to noise ratio. Besides the guided waves in
honeycomb sandwich, the guided waves in a single free top skin plate are measured for
comparison purpose. Hence, the influence of the aluminum core on the wave propagation
can be studied.

Scanning area
processing
SLDV Trigger
y
PZT amplifier signal generator
x
Figure 10.1 A schematic of the PZT-SLDV sensing setup.
146
Table 10.1 Geometry parameters of the honeycomb sandwich structure (units: mm).
Skin panels Honeycomb core
Length Width Thickness Cell size Wall thickness Height
135 135 1 6.35 0.055 12.7
0.3 1
120 15 kHz 120 15 kHz
0.2
Velocity (mm/s)
Velocity (mm/s)
0.5
0.1
80
y (mm)
80
y (mm)
0 0
-0.1
40 40
-0.5
-0.2
0 -0.3 0 -1
0 40 80 120 0 40 80 120
(a) x (mm) (d) x (mm)
1 1
120 50 kHz 120 50 kHz
Velocity (mm/s)
Velocity (mm/s)
0.5 0.5
80 80
y (mm)
y (mm)
0 0
40 40
-0.5 -0.5
0 -1 0 -1
0 40 80 120 0 40 80 120
(b) x (mm) (e) x (mm)
1 1
120 100 kHz 120 100 kHz
Velocity (mm/s)
Velocity (mm/s)
0.5 0.5
80 80
y (mm)
y (mm)
0 0
40 40
-0.5 -0.5
0 -1 0 -1
0 40 80 Speckle
120 0 40 80 120
(c) x (mm) patterns (f) x (mm)
Figure 10.2 Wavefields measured from the experiment: (a), (b) and (c) are wavefields in
the honeycomb sandwich at 15 kHz, 50 kHz and 100 kHz; (d), (e) and (f) are wavefields
in the single skin plate at 15 kHz, 50 kHz and 100 kHz.
Figure 10.2 a, b and c are experimental results of the honeycomb sandwich at
three different excitation frequencies 15 kHz, 50 kHz and 100 kHz, respectively; and
147
Figure 10.2 d, e and f are the results of the single skin plate. At the low frequency 15 kHz,
guided waves in the honeycomb sandwich have elliptical wave fronts and larger
wavelengths, compared to those in the single skin plate. When the frequency goes higher
from 15 kHz to 100 kHz, the wave fronts in the honeycomb sandwich are gradually
changing from elliptical fronts to circular fronts, as the wavelengths in the honeycomb
sandwich are getting closer to the wavelengths in the single skin plate. At the highest
frequency 100 kHz, the wave fronts in the honeycomb sandwich become circular, and the
wavelength in the honeycomb sandwich is nearly the same as the wavelength in a single
plate. Moreover, at 100 kHz, the wavefield in Figure 10.2 c shows speckle patterns which
might be induced by the wave interaction with the inside honeycomb core.
10.2.3. Frequency-wavenumber analysis
The frequency-wavenumber analysis is applied to the experimental guided wavefields in
order to investigate the wavenumber information. For the single skin plate, wavenumber
spectra in Figure 10.3 d, e and f show circular ring-shaped wavenumber bands which
agree well with theoretical wavenumber curves of the plate A0 mode. Compared to the
single skin plate, the honeycomb sandwich gives different wavenumber spectra, as shown
in Figure 10.3 a, b and c. The spectrum at 15 kHz shows an elliptical ring-shaped
wavenumber band, which has smaller wavenumbers than the plate A0 mode (represented
by the solid line). With the increase of frequency from 15 kHz to 100 kHz, the
wavenumber band expands and the wavenumber values increase. Moreover, the
elliptical-shaped wavenumber band is gradually getting closer to the wavenumber curve
of the plate A0 mode. The comparisons show that guided waves in the honeycomb
sandwich have different wavenumbers compared to the single skin plate, and with the
148
increase of wave frequency, the wavenumber in the honeycomb sandwich is gradually
asymptotic to that in the single skin plate.
1 1 1 1
15 kHz 15 kHz
0.8 0.8
0.5 0.5
ky (rad/mm)
ky (rad/mm)
0.6 0.6
0 0
0.4 0.4
-0.5 -0.5
0.2 0.2
-1 0 -1 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
(a) kx (rad/mm) (d) kx (rad/mm)
1 1 1 1
50 kHz 50 kHz
0.8 0.8
0.5 0.5
ky (rad/mm)
ky (rad/mm)
0.6 0.6
0 0
0.4 0.4
-0.5 -0.5
0.2 0.2
-1 0 -1 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
(b) kx (rad/mm) (e) kx (rad/mm)
1 1 1 1
100 kHz 100 kHz
0.8 0.8
0.5 0.5
ky (rad/mm)
ky (rad/mm)
0.6 0.6
0 0
0.4 0.4
-0.5 -0.5
0.2 0.2
-1 0 -1 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
(c) kx (rad/mm) (f) kx (rad/mm)
Figure 10.3 Wavenumber spectra: (a), (b) and (c) are spectra at 15 kHz, 50 kHz and 100
kHz for the honeycomb sandwich; (d), (e), and (f) are spectra at 15 kHz, 50 kHz and 100
kHz for the single skin plate. The solid line represents the theoretical wavenumber curves
of the A0 mode in the single skin plate.
149
10.3. WAVE INTERACTION WITH DEBONDING DAMAGE
10.3.1. Experimental setup
The influences of debonding damage on guided waves in honeycomb sandwich structures
are investigated. Two sandwiches with the same dimensions 305 mm × 305 mm × 15 mm
are used. One is the pristine, while the other contains debonding damage with the size of
58 mm × 58 mm, which is created by removing the adhesive during the manufacturing
process. Figure 10.4 gives the experimental sensing layout. The scanning is performed in
a 150 mm × 150 mm scanning area with a spatial resolution of 0.2 mm. Two frequencies,
the high frequency 100 kHz and the low frequency 40 kHz, are tested.
10.3.2. Wave interaction with debonding damage at 100 kHz
Figure 10.5 plots the SLDV measurements at the high frequency 100 kHz for the pristine
and damaged sandwiches. The results of both the pristine and damaged sandwiches show
waves generated from the PZT propagating forward. Moreover, there are speckle patterns
left behind the wave fronts, which are induced by wave interactions with the honeycomb
core. Compared to the pristine sandwich, the damaged sandwich has less wave
interactions with the honeycomb core: (i) the wave fronts are smoother and (ii) the
speckle patterns have lower amplitudes in the damage sandwich.
Scanning area
y
x
PZT
Debonding damage
Figure 10.4 The sensing layout of the PZT-SLDV experiment on a honeycomb sandwich
with debonding damage in the center of the sandwich.
150
60 30 μs 3 60 30 μs 3
2
Velocity (mm/s)
2
Velocity (mm/s)
30 30
1 1
y (mm)
y (mm)
0 0 0 0
-1 -1
-30 -30
-2 -2
-60 -60 -3
-3
0 50 100 150 0 50 100 150
(a) x (mm) (d) x (mm)
Smoother
Wave fronts wave fronts
60 55 μs 3 60 55 μs 3
2 2
Velocity (mm/s)
Velocity (mm/s)
30 30
1 1
y (mm)
y (mm)
0 0 0 0
-1 -1
-30 -30
-2 -2
-60 -3 -60 -3
0 50 100 150 0 50 100 150
(b) x (mm) (e) x (mm)
Speckle Speckle patterns with

patterns lower amplitudes
60 80 μs 3 60 80 μs 3
2 2
Velocity (mm/s)
Velocity (mm/s)
30 30
1 1
y (mm)
y (mm)
0 0 0 0
-1 -1
-30 -30
-2 -2
-60 -3 -60 -3
0 50 100 150 0 50 100 150
(c) x (mm) (f) x (mm)
Figure 10.5 Experimental results at the high frequency 100 kHz for the pristine (left
column) and damaged (right column) sandwich structures: (a) and (d) at 30 μs; (b) and (e)
at 55 μs; (c) and (f) at 80μs.
1 1 1 1
0.8 0.8
0.5 0.5
ky (rad/mm)
ky (rad/mm)
0.6 0.6
0 0
0.4 0.4
-0.5 -0.5
0.2 0.2
-1 0 -1 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
Figure 10.6 Wavenumber spectra at 100 kHz: (a) for the pristine sandwich and (b) for
the sandwich with debonding damage.
151
For further analysis, the wavefields are transformed to the frequency-wavenumber
spectra. For the pristine sandwich, the spectrum is plotted in Figure 10.6 a, which shows
four distinct wavenumber components, the right, left, top and bottom components. In
order to know why these components appear and what waves do they represent, the
wavenumber spectrum in Figure 10.6 a is decomposed into individual components, as
plotted in Figure 10.7 a, b, c and d. By using the inverse Fourier transform, the
wavefields that correspond to the decomposed spectra in Figure 10.7 a, b, c and d are
constructed, as plotted in Figure 10.7 e, f, g and h, respectively. The constructed
wavefields show waves propagate forward, backward, upward and downward. The
forward waves in Figure 10.7 e are generated from the PZT actuator and gradually
propagate away from the actuator. The backward waves in Figure 10.7 f are the reflection
waves induced by the honeycomb core. Moreover, the upward and downward waves in
Figure 10.7 g and h are the scattering waves induced by the honeycomb core.
Figure 10.6 b plots the wavenumber spectrum for the honeycomb sandwich with
debonding damage. Figure 10.8 a, b, c and d are the decomposed wavenumber spectra.
Figure 10.8 e, f, g and h are the decomposed wavefields. These decomposed wavefields
are compared to the wavefields for the pristine sandwich in Figure 10.7 e, f, g and h. It
can be found that the amplitudes of backward waves (in Figure 10.8 f) are lower in the
sandwich with debonding damage. The possible reason is that in the ideal debonding
region the honeycomb core is fully detached from the skin plate and cannot induce any
reflection wave.
152
1 1
60 80 μs 3
0.8 2
Velocity (mm/s)
0.5
30
ky (rad/mm) 0.6
1
y (mm)
0 0 0
0.4
-1
-30
-0.5 -2
0.2
-60 -3
-1 0
-1 -0.5 0 0.5 1 0 50 100 150
(a) kx (rad/mm) (e) x (mm)
1 1
60 80 μs 3
0.8 2
Velocity (mm/s)
0.5
30
ky (rad/mm)
1
0.6
y (mm)
0 0 0
0.4
-1
-30
-0.5 -2
0.2
-60 -3
-1 0
-1 -0.5 0 0.5 1 0 50 100 150
(b) kx (rad/mm) (f) x (mm)
1 1
60 80 μs 3
0.8 2
Velocity (mm/s)
0.5 30
1
ky (rad/mm)
0.6
y (mm)
0 0 0
0.4 -1
-30
-0.5 -2
0.2
-60 -3
-1 0
-1 -0.5 0 0.5 1 0 50 100 150
(c) kx (rad/mm) (g) x (mm)
1 1
60 80 μs 3
0.8 2
Velocity (mm/s)
0.5
30
ky (rad/mm)
0.6 1
y (mm)
0 0 0
0.4
-1
-30
-0.5 -2
0.2
-60 -3
-1 0
-1 -0.5 0 0.5 1 0 50 100 150
(d) kx (rad/mm) (h) x (mm)
Figure 10.7 Wave decomposition for the pristine sandwich: (a), (b), (c) and (d) are
decomposed right, left, top and bottom parts of the wavenumber spectrum at 100 kHz,
respectively; (e), (f), (g) and (h) are decomposed wavefields at 80 μs, which correspond
to the spectra in (a), (b), (c) and (d), respectively.
153
1 1
60 80 μs 3
0.8 2
Velocity (mm/s)
0.5
ky (rad/mm) 30
1
0.6
y (mm)
0 0 0
0.4
-1
-30
-0.5 -2
0.2
-60 -3
-1 0
-1 -0.5 0 0.5 1 0 50 100 150
(a) kx (rad/mm) (e) x (mm)
1 1
60 80 μs 3
0.8 2
Velocity (mm/s)
0.5
30
ky (rad/mm)
1
0.6
y (mm)
0 0 0
0.4
-1
-30
-0.5 -2
0.2
-60 -3
-1 0
-1 -0.5 0 0.5 1 0 50 100 150
(b) kx (rad/mm) (f) x (mm)
1 1
60 80 μs 3
0.8 2
Velocity (mm/s)
0.5
30
1
ky (rad/mm)
0.6
y (mm)
0 0 0
0.4 -1
-30
-0.5 -2
0.2
-60 -3
-1 0
-1 -0.5 0 0.5 1 0 50 100 150
(c) kx (rad/mm) (g) x (mm)
1 1
60 80 μs 3
0.8 2
Velocity (mm/s)
0.5
30
ky (rad/mm)
0.6 1
y (mm)
0 0 0
0.4
-1
-30
-0.5 -2
0.2
-60 -3
-1 0
-1 -0.5 0 0.5 1 0 50 100 150
(d) kx (rad/mm) (h) x (mm)
Figure 10.8 Wave decomposition of the waves in the sandwich with debonding damage:
(a), (b), (c) and (d) are decomposed right, left, top and bottom parts of the wavenumber
spectrum at 100 kHz, respectively; (e), (f), (g) and (h) are decomposed wavefields at 80
μs which corresponds to the spectra in (a), (b), (c) and (d), respectively.
154
10.3.3. Wave interaction with debonding damage at 40 kHz
Figure 10.9 plots the experimental measurements at the low frequency 40 kHz for the
pristine and damaged sandwich structures. Compared to the waves in the pristine
sandwich, the waves in the damaged sandwich show strong interactions with the
debonding damage. Moreover, at 160 μs after passed the damage area, the waves in the
damaged sandwich (Figure 10.9 f) show higher amplitudes than the waves in the pristine
sandwich (Figure 10.9 c).
60 60 μs 3 60 60 μs 3
2 2
Velocity (mm/s)
Velocity (mm/s)
30 30
1 1
y (mm)
y (mm)
0 0 0 0
-1 -1
-30 -30
-2 -2
-60 -3 -60 -3
0 50 100 150 0 50 100 150
(a) x (mm) (d) x (mm)
the debonding damage
60 110 μs 3 60 110 μs 3
2 2
Velocity (mm/s)
Velocity (mm/s)
30 30
1 1
y (mm)
y (mm)
0 0 0 0
-1 -1
-30 -30
-2 -2
-60 -3 -60 -3
0 50 100 150 0 50 100 150
(b) x (mm) (e) x (mm)
Higher amplitudes
than those of the
60 160 μs 3 60 160 μs 3
pristine sandwich
2 2
Velocity (mm/s)
Velocity (mm/s)
30 30
1 1
y (mm)
y (mm)
0 0 0 0
-1 -1
-30 -30
-2 -2
-60 -3 -60 -3
0 50 100 150 0 50 100 150
(c) x (mm) (f) x (mm)
Figure 10.9 Experimental results at the low frequency 40 kHz for the pristine (left
column) and damaged (right column) sandwich structures: (a) and (d) at 60 μs; (b) and (e)
at 110 μs; (c) and (f) at 160 μs.
155
Figure 10.10 a and b are wavenumber spectra for the pristine and damaged
honeycomb sandwiches. The wavenumber spectrum of the damaged sandwich in Figure
10.10 b shows an additional wavenumber component, compared to the spectrum of the
pristine sandwich. To identify the waves of the additional wavenumber component, this
additional component is extracted by using the frequency-wavenumber filtering (in
Figure 10.10 c), and then reconstructed to the wavefield (in Figure 10.10 d). The
reconstructed wavefield shows waves inside the debonding area. This means the
additional wavenumber component is related to the waves inside the debonding area.
Therefore, the additional wavenumber component can be used for debonding detection.
1 1 1 Additional wavenumber
1
component
0.8 0.8
0.5 0.5
ky (rad/mm)
ky (rad/mm)
0.6 0.6
0 0
0.4 0.4
-0.5 -0.5
0.2 0.2
-1 0 -1 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
1 1
60 110 μs 3
0.8 2
Velocity (mm/s)
0.5
30
ky (rad/mm)
1
0.6
y (mm)
0 0 0
0.4 -1
-30
-0.5 -2
0.2
-60 -3
-1 0
-1 -0.5 0 0.5 1 0 50 100 150
(c) kx (rad/mm) (d) x (mm)
Figure 10.10 Wavenumber analysis results at 40 kHz: (a) wavenumber spectrum for the
pristine sandwich, (b) wavenumber spectrum for the sandwich with debonding damage,
(c) filtered wavenumber spectrum, and (d) reconstructed wavefield at 110 μs that
corresponds to the spectrum in (c). The dotted box represents the area of actual
debonding damage.
156
10.4. DEBONDING DETECTION AND QUANTIFICATION
The analysis of wave interaction with debonding damage at 40 kHz show that additional
wavenumber component which relates to the waves in the debonding area appears in the
wavenumber spectrum. To detect and quantify the debonding damage, the filter
reconstruction imaging method is applied. The energy map of the filtered waves is shown
in Figure 10.11 a. The energy map shows a high energy area which matches well with the
debonding damage (the dotted box in Figure 10.11 a).
Besides the filter reconstruction imaging, the spatial wavenumber imaging is
applied. Figure 10.11 b plots the result of spatial wavenumber imaging in the frequency
band 35~45 kHz. The imaging result shows an area with increased wavenumber values,
indicating the debonding damage. The increased wavenumber area agrees well with the
debonding damage (the dotted box in Figure 10.11 b).
1 0.42
Wavenumber (rad/mm)
60 60
0.8 0.4
30 30 0.38
0.6
y (mm)
y (mm)
0 0 0.36
0.4
-30 -30 0.34
0.2 0.32
-60 -60
0 0.3
0 50 100 150 0 50 100 150
(a) x (mm) (b) x (mm)
Figure 10.11 Debonding detection and quantification: (a) the result of filtering
reconstruction imaging, and (b) the result of spatial wavenumber imaging. The dotted box
represents the area of actual debonding damage.
157
CHAPTER 11
CONCLUDING REMARKS AND FUTURE WORK
This dissertation research develops guided wave based integrated SHM/NDE
methodologies for damage detection and quantification in complex structures. The
integrated SHM/NDE methodologies provide non-contact guided wave sensing for wave
visualization and high spatial resolution wavefield acquisition, intrinsic wavefield
analysis and characterization methods for the interpretation of complex wave propagation
and interaction phenomenon, and hierarchical damage diagnosis for quantitative and
visualized damage detection.
Part I focuses on the development of integrated SHM/NDE methodologies. First,
a non-contact laser vibrometry sensing system is optimized to visualize the guided wave
propagation and interaction and to acquire high spatial resolution wavefields of guided
waves. Second, to extract informative wave signatures from the wavefields and
characterize complex wave propagation and interaction phenomenon, guided wavefield
analysis methods, including frequency-wavenumber analysis, wavefield decomposition
and space-frequency-wavenumber analysis are investigated. Using these analysis
methods, multi-modal and dispersive guided waves can be resolved and complex wave
propagation and interaction can be interpreted and analyzed in time, space, frequency and
wavenumber multi-domains. Third, a hierarchical damage diagnosis methodology is
developed for quantitative and visualized damage detection. The hierarchical
158
methodology systematically combines phased array imaging and wavefield based
imaging to achieve efficient and precise damage detection and quantification.
In Part II, the developed methodologies are applied to five different SHM/NDE
applications: (1) gas accumulation detection and quantification in water loaded structures,
(2) crack damage detection and quantification in isotropic plates, (3) thickness loss
evaluation in isotropic plates, (4) delamination damage detection and quantification in
composite laminates, (5) debonding detection and quantification in honeycomb sandwich
structures.
This dissertation research will initiate sensing and diagnosis methodologies that
provide rapid non-contact inspection of damage and diagnosis of structural health.
Moreover, this research provides intrinsic guided wave analysis methods which can be
used to analyze multi-modal wave propagation and interaction in complex structures.
These analysis methods will benefit researchers in understanding complex wave
propagation and interaction phenomenon and acquiring intrinsic wave signatures. Last
but not least, this dissertation research provides efficient and precise integrated
SHM/NDE methodologies for quantitative and visualized damage detection in complex
structures. In the long run, this research will contribute to the development of advanced
sensor and sensing technologies based on guided waves, and to providing on-demand
health information at component or subsystem level for the safety and reliability of
structures.
159
11.1. RESEARCH CONCLUSIONS
11.1.1. Analytical and numerical studies of guided waves
In Chapter 3, both analytical and numerical studies of guided waves are performed to
provide preliminary understanding of guided waves in complex structures, theoretical
data for assisting experimental work, and abundant information for parametric studies.
Both analytical and numerical methods are used to investigate guided waves in different
complex structures and wave interactions with defects or changes of boundary conditions.
Guided waves in a plate with one side in water are investigated. By using the
GMM, dispersion curves and mode shapes of guided waves in the plate with one side in
water are derived. Compared to dispersion curves of a free plate, those of the plate in
water have an additional wave mode, the quasi-Scholte mode, which is dispersive at low
frequencies and asymptotes to the non-dispersive Scholte mode at high frequencies.
Moreover, EFIT simulations are performed to simulate guided waves in the plate with
one side in water. The simulation results confirm the appearance of the quasi-Scholte
mode. The simulation results also show leak guided waves in water induced by the
energy leakage of guided waves in the plate.
Guided waves in composite laminates are investigated. By using the SAFE
method, dispersion curves and mode shapes of guided waves in a composite laminate are
derived. From dispersion plots, it is found that both wavenumber and group velocity
depend on wave propagation direction. Moreover, EFIT simulations are performed to
investigate guided wave interaction with delamination damage. Simulation results show
that once arrived to the delamination damage, incident guided waves separate to different
waves propagating in the materials above and below the delamination damage. These
160
waves above and below the delamination have different wavenumbers and wave speeds
compared to the incident waves in the full laminate.
Guided waves in honeycomb sandwich structures are simulated by using FEM.
The finite element model is constructed of a hexagonal aluminum core and two
aluminum skins. FEM results show that at low frequency 15 kHz, the entire sandwich
serves as a waveguide, in which global guided waves propagate with large wavelengths
and elliptical wave fronts. With the increase of wave frequency from 15 kHz to 100 kHz,
wavelengths gradually decrease and wave fronts gradually change from elliptical wave
fronts to circular wave fronts. Moreover, global guided waves gradually disappear which
means it becomes difficult for the global guided waves to propagate in the entire
sandwich. At high frequency 100 kHz, the global guided waves completely disappear,
and most of wave energy propagates in the skin plate in the form of Lamb waves. These
Lamb waves strongly interact with the honeycomb core, generating complex wave
interactions.
11.1.2. Wavefield acquisition and intrinsic characterization
In Chapter 4, a laser vibrometry sensing system is optimized to achieve rapid, high spatial
resolution and non-contact sensing for guided wave visualization and wavefield
measurements. The optimized laser vibrometry sensing system can be used to visualize
guided wave propagation and interaction. Moreover, the sensing system provides an
immediate means of damage evaluation. Experiments have been performed to
demonstrate the sensing system for crack detection in an aluminum plate and
delamination detection a composite plate. For the crack detection, the experimental
measurement visualizes incident waves, crack reflection waves and transmission waves.
161
The crack location can be qualitatively identified from the measurement. For
delamination detection, the experimental measurement visualizes complex wave
interactions at the delamination damage. Guided waves propagating back and forth
between the delamination boundaries are observed, exhibiting the “trapped wave”
phenomenon. The location of delamination damage can also be qualitatively seen from
the experimental measurement.
Guided wavefields in terms of time and space dimensions contain a wealth of
information regarding guided wave propagations in structures and wave interactions with
structural discontinuities. To extract informative wave signatures from the time-space
wavefields and characterize the complex wave propagation and interaction phenomenon,
guided wavefield analysis methods, including frequency-wavenumber analysis, wavefield
decomposition and space-frequency-wavenumber analysis, are investigated. Using these
analysis methods, multi-modal and dispersive guided waves can be resolved, and
complex wave propagation and interaction can be interpreted and analyzed in time, space,
frequency and wavenumber multi-domains.
11.1.3. Quantitative and visualized damage diagnosis
In Chapter 5, a hierarchical damage diagnosis methodology is developed for quantitative
and visualized damage detection. The hierarchical methodology systematically combines
phased array imaging and wavefield based imaging to achieve efficient and precise
damage detection and quantification. Using the phased array imaging, an intensity
scanning image of the structure is generated to efficiently visualize and locate the damage
zone. Then the wavefield based imaging is applied to precisely quantify the damage size,
shape and depth.
162
The generic phased array imaging is developed based on the classic delay-and-
sum principal, in which directional dependent guided wave parameters are considered.
This formulation is valid not only for the case that phase and group velocities have the
same orientation, but also for the general case that the two velocities are in different
directions. Therefore, this generic phased array imaging works for both isotropic and
anisotropic materials. For the proof of concept, an experimental test is performed using
the generic phased array imaging to detect multiple defects in an anisotropic composite
laminate. In the test, the phased array is constructed with scanning points of a non-contact
SLDV. Experimental results show that multiple defects in different directions w.r.t. the
array center can be successfully detected.
Two wavefield based imaging methods, the filter reconstruction imaging and the
spatial wavenumber imaging, are investigated to precisely quantify the damage size,
shape and depth. The filter reconstruction imaging is based on the frequency-
wavenumber filtering process by extracting damage induced frequency-wavenumber
components and reconstructing them to a time-space wavefield which only contains
damage induced waves. The filter reconstruction imaging provides a rapid technique to
locate the damage by showing the damage induced waves.
The spatial wavenumber imaging calculates the wavenumber value at each spatial
location and generates a spatial wavenumber image of the structure, in which the damage
is indicated by different wavenumbers compared to the area without damage. Moreover,
the wavenumber information can be used to quantify the damage depth, since the damage
depth or material thickness affects the wavenumber value. Last but not least, compared
to the filter reconstruction imaging, the spatial wavenumber imaging is more robust, since
163
it works for the cases when multiple wave modes exist and when damage induced waves
are difficult to be extracted using the filter reconstruction imaging.
11.1.4. Guided waves in water loaded structures and water level detection
In Chapter 6, guided waves in water loaded structures are investigated through
simulations, laser vibrometry experiments, and wavefield analyses. Moreover, water level
detection and gas accumulation quantification methodologies are developed.
Water effects on guided wave propagation are investigated. Results of both
simulations and experiments confirm energy leakage through the solid water interface
and the existence of quasi-Scholte waves, when a plate is loaded by water. Due to energy
leakage, wave energy in the plate decreases when the plate is immersed in water.
Moreover, at low frequencies guided waves travel in the water loaded plate as quasi-
Scholte waves, while guided waves travel in the free plate as Lamb waves.
The influence of water level on wave propagation is investigated through a pitch-
catch sensing test on a plate with a portion in water. The test result shows that wave
traveling time in the pitch-catch path is determined by lengths of water path (the portion
of plate in water) and dry path (the portion of plate out of water). Moreover, the wave
traveling time increases with the increase of water path. A linear relationship between the
change of traveling time and the length of water path is confirmed. Thus, by measuring
the traveling time of guided waves, the water level (length of water path) can be
estimated. This sensing methodology can potentially be applied for water level detection
and gas accumulation detection.
164
11.1.5. Crack detection and quantification in isotropic plates
In Chapter 7, guided wave interaction with crack damage is investigated through laser
vibrometry experiments and wavefield analyses. Moreover, the work presented in this
chapter demonstrates contactless wavefield measurements using laser vibrometry sensing
and crack quantification using both filter reconstruction imaging and spatial wavenumber
imaging. These two methods are able to detect crack location, quantify crack length, and
provide information about crack severity.
The cases investigated in this chapter involve scenarios where two Lamb wave
modes are generated. It is expected that the presented methods are also applicable to
scenarios with more than two Lamb wave modes, particularly the filter reconstructed
imaging method. It is also expected that the presented methods could be applied to the
scenario of multiple cracks in various orientations with respect to incident waves.
11.1.6. Thickness loss evaluation in isotropic plates
In Chapter 8, guided wave interaction with thickness loss damage is investigated through
laser vibrometry experiments. Moreover, a dispersion curve regression method is
developed to quantify material thickness and damage depth in isotropic plates.
In this work, a wide band chirp signal is used to generate guided waves with
various wavelengths in a wide frequency band. Interactions of these guided waves with
damage include information at various wavelengths which can benefit the precision of
damage quantification. The guided wavefield is measured by using the non-contact laser
vibrometry sensing. Once the wavefield is acquired, it is transformed to the spectral field,
which is a function of frequency and space.
165
For wave characterization, damage detection and quantification, the spatial
wavenumber imaging is applied first. The imaging result shows that the wavenumber
becomes larger at the material loss damage. In order to quantify material thickness as
well as damage depth, the dispersion curve regression is applied. These methods have
been applied to detect and quantify material loss damage in an aluminum plate. The
location, size and depth of the damage are successfully detected and quantified. Overall
the method outlined in this chapter provides an effective means for detection and
quantification of material loss damage. This method can potentially be used for detection
and quantification of corrosion damage.
11.1.7. Delamination detection and quantification in composite laminates
Chapter 9 presents studies on guided wave interaction with delamination damage and
delamination detection and quantification techniques for laminated composite plates. The
wavefield data analysis confirms the wave trapping phenomenon. Quantitative
interpretation of wave behaviors in the delamination region is demonstrated. Among the
wavenumber analysis methods studied, the frequency-wavenumber analysis shows that
the frequency-wavenumber spectrum can indicate the presence of additional
wavenumbers created by the damage. The reconstructed wavefield images created by
inverting only these additional wavenumber components demonstrate that the additional
wavenumbers correlate to trapped waves in the delamination region.
The phased array imaging and wavefield based imaging methods are applied to
detect and quantify impact-induced delamination damage in a CFRP composite plate. The
delamination damage is created by a quasi-static indentation test, and is barely visible
even with careful visual inspection. By using the phased array imaging, the delamination
166
damage is detected. To further quantify the delamination damage, the wavefield based
imaging methods are applied. The quantification result provides the location, size and
shape of the delamination damage.
11.1.8. Debonding detection and quantification in honeycomb sandwich structures
In Chapter 10, guided wave propagation in a honeycomb sandwich structure and wave
interaction with skin-core debonding damage are investigated. Moreover, the non-contact
laser vibrometry sensing and wavefield based imaging are performed for detection and
quantification of debonding damage.
The honeycomb sandwich used in this study is composed of two aluminum skin
plates and a hexagonal aluminum core. To investigate the guided wave propagation in the
honeycomb sandwich, FEM is used to simulate the wave propagation. Moreover, laser
vibrometry experiments are performed. The simulation results agree well with the
experimental results.
From results of simulations and experiments, it is found that global guided waves
with elliptical wave fronts propagate in the entire sandwich at low frequency 15 kHz. It
means that, the entire sandwich can be considered as a waveguide which supports global
guided waves at this low frequency. With the increase of wave frequency from 15 kHz to
100 kHz, wave fronts gradually change from elliptical wave fronts to circular wave fronts.
Moreover, global guided waves gradually disappear which means it becomes difficult for
the global guided waves to propagate in the entire sandwich. At the high frequency 100
kHz, global guided waves completely disappear, and most of the wave energy propagates
in the skin plate in the form of Lamb waves. These Lamb waves strongly interact with the
honeycomb core generating complex wave interactions.
167
For further interpretation, guided wavefields acquired from experiments are
transferred to wavenumber spectra. At 15 kHz, the wavenumber spectrum of the
honeycomb sandwich shows an elliptical ring-shaped wavenumber band which has
smaller wavenumbers than the plate A0 mode. With the increase of frequency from 15
kHz to 100 kHz, the wavenumber value in the honeycomb sandwich becomes larger, and
the elliptical-shaped wavenumber band gradually gets closer to the plate A0 mode. The
wavenumber comparison shows that guided waves in the honeycomb sandwich have
different wavenumbers compared to those in the single plate, and the wavenumber is
gradually asymptotic to that of the plate wave mode with the increase of wave frequency.
Interactions of global guided waves with debonding damage are investigated by
using non-contact laser vibrometry test and wavefield analysis. Wavefield measurements
show strong interactions with debonding damage. In wavenumber domain, additional
wavenumbers induced by the debonding damage are observed. The reconstructed
wavefield images created by inverting only these additional wavenumber components
demonstrate that the additional wavenumbers correlate to the waves in the debonding
area. Moreover, the result of spatial wavenumber imaging shows an area with increased
wavenumber values, indicating the debonding damage.
11.2. MAJOR CONTRIBUTIONS
This dissertation research provides the following contributions to the state of the art:
(1) A non-contact laser vibrometry sensing system is optimized to acquire a high
spatial resolution guided wavefield. It also enables non-contact inspection of
damage and diagnosis of structural health.
168
(2) Multi-modal and dispersive guided waves are able to be decomposed into
separated wave components through guided wave decomposition. Moreover, all
separated wave components can provide abundant information of wave
interaction with damage and the contributions of these components will lead to
robust damage detection and quantification.
(3) Complex wave propagation and interaction can be characterized and their
signatures can be achieved through wavefield analysis methods. Moreover, these
analysis methods unveil information of guided waves in time, space, frequency
and wavenumber multi-domains.
(4) The guided wave decomposition and wavefield analysis methods enable the use
of wideband guided wave signals for damage diagnosis. Compared to the guided
waves generated from the traditional narrowband tone burst excitation, those
generated from the wideband short-duration pulse or wideband chirp excitation
can contain waves of various modes and wavelengths. The use these multi-modal
and multi-wavelength guided waves will lead to robust and multi-scale damage
detection and quantification.
(5) A hierarchical damage diagnosis methodology is developed for quantitative and
visualized damage detection. This hierarchical methodology systematically
combines phased array imaging and wavefield based imaging, which achieves
efficient and precise damage detection and quantification.
(6) The generic phased array beamforming is developed, in which directional
dependent guided wave parameters are considered. This generic beamforming
method is valid not only for the case that phase and group velocities have the
169
same orientation, but also for the general case that the two velocities are in
different directions. Moreover, this generic method works for both isotropic and
anisotropic materials. Last but not least, it has been shown that this method can
detect multiple defects in different directions w.r.t. the array center in an
anisotropic composite plate.
(7) The wavefield based imaging method that integrates both the filter reconstruction
imaging and spatial wavenumber imaging are developed. It takes advantages of
both methods and lead to a more efficient and robust imaging method for damage
visualization and quantification.
(8) Water effects on guided wave propagation are investigated. This study confirms
the energy leakage through solid water interface and the existence of quasi-
Scholte waves when a plate has one side in water. Moreover, a linear relationship
between the change of traveling time and the length of water path is confirmed.
By measuring the traveling time of guided waves, the length of water path can be
determined. This sensing methodology can potentially be applied for water level
detection and gas accumulation detection in closed containers.
(9) Guided wave interaction with crack damage is studied. This study identifies
additional wavenumber components in wavenumber domain, which correlate to
the crack induced wavefield discontinuities. Moreover, this study shows that
wavefield based imaging methods are able to detect the crack location, quantify
the crack length, and provide information about the crack severity.
(10) A dispersion curve regression method is developed to quantify material thickness
and damage depth in isotropic plates. The location, size and depth of material loss
170
damage can be detected and quantified. This method can potentially be used for
detection and quantification of corrosion damage.
(11) Guided wave interaction with delamination damage is investigated. This study not
only confirms the wave trapping at delamination damage, but also quantitatively
interprets the trapped waves via various wavenumber analysis methods. In
wavenumber domain, additional wavenumbers are observed, which correlate to
trapped waves in the delamination region. Moreover, the developed phased array
imaging successfully detects impact-induced delamination damage. Last but not
least, the wavefield based imaging detects and quantifies the location, size and
shape of the delamination damage.
(12) The guided wave propagation in a complex geometry honeycomb sandwich is
investigated. The study confirms global guided waves in the full sandwich at low
frequencies and strong wave interaction with the inside core at high frequencies.
Moreover, the wavenumber distribution of guided waves is determined through
experiments, which is an elliptical wavenumber band.
The guided wave interaction with debonding damage is investigated. The
study shows mode conversion at debonding damage. Moreover, in wavenumber
domain, additional wavenumbers are observed which correlate to waves in the
debonding area. Last but not least, the study shows that wavefield based imaging
methods successfully detect and quantify the location and size of hidden
debonding damage.
171
11.3. RECOMMENDATION FOR FUTURE WORK
This dissertation research focuses on guided wave based integrated SHM/NDE. Non-
contact sensing, intrinsic wave analysis and advanced damage diagnosis methods are
used to analyze complex wave propagation and interaction, and to detect and quantify
structural damage. The following tasks are recommended to be undertaken for further
studies:
(1) In Chapter 3, guided waves are generated by surface bonded PZT transducers.
The non-contact guided wave excitation by using non-contact transducers, such as
air-coupled transducers and pulse-laser transducers, should be investigated.
Moreover, an integration of these wave excitation methods and laser vibrometry
sensing will lead to a fully non-contact guided wave excitation and sensing
system.
(2) In Chapter 4, wavefield analysis methods are presented to analyze waves
propagating along a 1-D straight line and waves propagating in a 2-D surface.
Wavefield analysis methods that can be used to analyze waves propagating along
curves such as spiral curves and elliptical curves should be investigated.
Moreover, analysis methods for analyzing waves in a 3-D space should be
investigated. It is expected that by using these analysis methods, the wave
propagation and interaction at various structural depths can be characterized and
interpreted.
(3) In Chapter 5, the generic beamforming considers directional dependent guided
wave parameters and works for both isotropic and anisotropic materials. However,
the roles of weighting factors are not investigated. In the future, selections of
172
weighting factors should be investigated to further improve the beamforming
results. Moreover, the wavefield based imaging methods in Chapter 5 are based
on the interpretation of wavenumber information. Other methods such as machine
learning and statistical based methods can also be investigated to achieve robust
damage detection and quantification.
(4) In Chapter 6, the water level detection and gas accumulation detection methods
only use one pitch-catch sensing path to determine the location of water surface in
the path. It is recommended that sensor arrays should be designed to fully
quantify the shape of water surface. Moreover, it is also interesting to investigate
sensing methods that can be used to detect both water surface and structural
damage.
(5) In Chapter 7, the wavefield based imaging methods are used to detect and
quantify cracks of different widths and depths. It is also interesting to investigate
cases of multiple cracks, cracks with different lengths, and cracks in different
orientations.
(6) In Chapter 8, the material loss defects created by using a milling machine only
have simple geometries. The detection and quantification of material loss defects
such as realistic corrosion damage and erosion damage should be studied.
(7) In Chapter 9, a single impact induced defect is detected and quantified. Future
research can be performed to detect and quantified multiple defects of different
types.
(8) In Chapter 10, wave propagation and interaction in honeycomb sandwich
structures are investigated by using FEM simulations and laser vibrometry
173
experiments. Theoretical work should be performed to achieve wave dispersion
curves and modes shapes, and to improve the understanding of waves in
honeycomb sandwich structures. Moreover, the effects of core size, core
geometry and core material on guided wave propagation can also be investigated.
174
REFERENCES
[1] Giurgiutiu, V., Structural Health Monitoring with Piezoelectric Wafer Active
Sensors. Boston, MA, USA: Academic Press, 2008.
[2] Worden, K., Farrar, C. R., Manson, G., and Park, G., "The Fundamental Axioms
of Structural Health Monitoring," Proceedings of the Royal Society a-
Mathematical Physical and Engineering Sciences, vol. 463, pp. 1639-1664, 2007.
[3] Rose, J. L., "The Upcoming Revolution in Ultrasonic Guided Waves,"
Preceedings of SPIE, vol. 7983, p. 798302, March 06, 2011 2011.
[4] Speckmann, H. and Henrich, R., "Structural Health Monitoring (Shm)-Overview
on Technologies under Development," in 16th world conference on NDT,
Montreal, Canada, 2004, p. 563.
[5] Rytter, A., "Vibration Based Inspection of Civil Engineering Structures," Ph. D.,
Department of Building Technology and Structural Engineering, Aalborg
University, Aalborg University, Denmark, 1993.
[6] Rose, J. L., "The Upcoming Revolution in Ultrasonic Guided Waves," presented
at the Nondestructive Characterization for Composite Materials, Aerospace
Engineering, Civil Infrastructure, and Homeland Security 2011, San Diego, CA,
2011.
[7] Miklowitz, J., Theory of Elastic Waves and Waveguides. North Holland, 1978.
[8] Rose, J. L., Ultrasonic Waves in Solid Media. Cambridge: Cambridge University
Press, 1999.
[9] Viktorov, I. A., Rayleigh and Lamb Waves: Plenum Press, 1967.
[10] Graff, K. F., Wave Motions in Solids. Mineola, NY, USA: Dover Publications,
Inc., 1975.
[11] Nayfeh, A. H., Wave Propagation in Layered Anisotropic Media. Amsterdam,
The Netherlands: Elsevier, 1995.
[12] Giurgiutiu, V. and Bao, J., "Embedded-Ultrasonics Structural Radar for in Situ
Structural Health Monitoring of Thin-Wall Structures," Structural Health
Monitoring-an International Journal, vol. 3, pp. 121-140, Jun 2004.
[13] Kim, D. and Philen, M., "Guided Wave Beamsteering Using Mfc Phased Arrays
for Structural Health Monitoring: Analysis and Experiment," Journal of
Intelligent Material Systems and Structures, vol. 21, pp. 1011-1024, Jul 2010.
175
[14] Purekar, A. S. and Pines, D. J., "Damage Detection in Thin Composite Laminates
Using Piezoelectric Phased Sensor Arrays and Guided Lamb Wave
Interrogation," Journal of Intelligent Material Systems and Structures, vol. 21, pp.
995-1010, 2010.
[15] Ruzzene, M., Xu, B., Lee, S. J., Michaels, T. E., and Michaels, J. E., "Damage
Visualization Via Beamforming of Frequency-Wavenumber Filtered Wavefield
Data," in Health Monitoring of Structural and Biological Systems 2010, 2010, p.
76500L.
[16] Wilcox, P. D., "Omni-Directional Guided Wave Transducer Arrays for the Rapid
Inspection of Large Areas of Plate Structures," Ieee Transactions on Ultrasonics
Ferroelectrics and Frequency Control, vol. 50, pp. 699-709, Jun 2003.
[17] Yu, L. and Giurgiutiu, V., "In Situ 2-D Piezoelectric Wafer Active Sensors Arrays
for Guided Wave Damage Detection," Ultrasonics, vol. 48, pp. 117-134, Apr
2008.
[18] Romanoni, M., Gonella, S., Apetre, N., and Ruzzene, M., "Two-Dimensional
Periodic Actuators for Frequency-Based Beam Steering," Smart Materials &
Structures, vol. 18, p. 125023, Dec 2009.
[19] Yoo, B., Purekar, A. S., Zhang, Y., and Pines, D. J., "Piezoelectric-Paint-Based
Two-Dimensional Phased Sensor Arrays for Structural Health Monitoring of Thin
Panels," Smart Materials & Structures, vol. 19, p. 075017, 2010.
[20] Senesi, M., Xu, B., and Ruzzene, M., "Experimental Characterization of Periodic
Frequency-Steerable Arrays for Structural Health Monitoring," Smart Materials
& Structures, vol. 19, p. 055026, 2010.
[21] Lee, S. J., Gandhi, N., Hall, J. S., Michaels, J. E., Xu, B., Michaels, J. E., and
Ruzzene, M., "Baseline-Free Guided Wave Imaging Via Adaptive Source
Removal," Structural Health Monitoring, vol. 11, pp. 472-481, 2012.
[22] Michaels, J. E., "Detection, Localization and Characterization of Damage in
Plates with an in Situ Array of Spatially Distributed Ultrasonic Sensors," Smart
Materials & Structures, vol. 17, p. 035035, Jun 2008.
[23] Clarke, T., Cawley, P., Wilcox, P. D., and Croxford, A. J., "Evaluation of the
Damage Detection Capability of a Sparse-Array Guided-Wave Shm System
Applied to a Complex Structure under Varying Thermal Conditions," Ieee
Transactions on Ultrasonics Ferroelectrics and Frequency Control, vol. 56, pp.
2666-2678, Dec 2009.
[24] Zhu, R., Huang, G. L., and Yuan, F. G., "Fast Damage Imaging Using the Time-
Reversal Technique in the Frequency-Wavenumber Domain," Smart Materials
and Structures, vol. 22, p. 075028, Jul 2013.
[25] Park, H. W., Kim, S. B., and Sohn, H., "Understanding a Time Reversal Process
in Lamb Wave Propagation," Wave Motion, vol. 46, pp. 451-467, Nov 2009.
176
[26] Sohn, H., park, H. W., Law, K. H., and Farrar, C. R., "Damage Detection in
Composite Plates by Using an Enhanced Time Reversal Method," Journal of
Aerospace Engineering, vol. 20, pp. 141-151, 2007.
[27] Hay, T. R., Royer, R. L., Gao, H. D., Zhao, X., and Rose, J. L., "A Comparison of
Embedded Sensor Lamb Wave Ultrasonic Tomography Approaches for Material
Loss Detection," Smart Materials & Structures, vol. 15, pp. 946-951, Aug 2006.
[28] Leonard, K. R. and Hinders, M. K., "Lamb Wave Tomography of Pipe-Like
Structures," Ultrasonics, vol. 43, pp. 574-583, Jun 2005.
[29] Michaels, T. E., Michaels, J. E., and Ruzzene, M., "Frequency-Wavenumber
Domain Analysis of Guided Wavefields," Ultrasonics, vol. 51, pp. 452-466, 2011.
[30] Ruzzene, M., "Frequency-Wavenumber Domain Filtering for Improved Damage
Visualization," Smart Materials and Structures, vol. 16, pp. 2116-2129, 2007.
[31] Flynn, E. B., Chong, S. Y., Jarmer, G. J., and Lee, J. R., "Structural Imaging
through Local Wavenumber Estimation of Guided Waves," Ndt & E International,
vol. 59, pp. 1-10, 2013.
[32] Rogge, M. D. and Leckey, C. A. C., "Characterization of Impact Damage in
Composite Laminates Using Guided Wavefield Imaging and Local Wavenumber
Domain Analysis," Ultrasonics, vol. 53, pp. 1217-1226, 2013.
[33] Lowe, M. J. S., "Matrix Techniques for Modeling Ultrasonic Waves in
Multilayered Media," Ieee Transactions on Ultrasonics Ferroelectrics and
Frequency Control, vol. 42, pp. 525-542, 1995.
[34] Rogge, M. D. and Johnston, P. H., "Wavenumber Imaging for Damage Detection
and Measurement," Review of Progress in Quantitative Nondestructive
Evaluation, vol. 31, pp. 761-768, 2012.
[35] Roberge, P. R., Corrosion Inspection and Monitoring. Hoboken, New Jersey:
John Wiley & Sons, Inc., 2007.
[36] Staszewski, W. J., Boller, C., and Tomlinson, G. R., Health Monitoring of
Aerospace Structures. Chichester: John Wiley & Sons, 2004.
[37] Boller, C., Chang, F. K., and Fujino, Y., Encyclopedia of Structural Health
Monitoring. new york: wiley, 2009.
[38] Bartoli, I., Marzani, A., di Scalea, F. L., and Viola, E., "Modeling Wave
Propagation in Damped Waveguides of Arbitrary Cross-Section," Journal of
Sound and Vibration, vol. 295, pp. 685-707, Aug 22 2006.
[39] Maess, M., Herrmann, J., and Gaul, L., "Finite Element Analysis of Guided
Waves in Fluid-Filled Corrugated Pipes," Journal of the Acoustical Society of
America, vol. 121, pp. 1313-1323, Mar 2007.
177
[40] Fan, Z., "Applications of Guided Wave Propagation on Waveguides with
Irregular Cross-Section," Doctor of Philosophy, Department of Mechanical
Engineering, Imperial College London, London, 2010.
[41] Predoi, M. V., Castaings, M., Hosten, B., and Bacon, C., "Wave Propagation
Along Transversely Periodic Structures," Journal of the Acoustical Society of
America, vol. 121, pp. 1935-1944, Apr 2007.
[42] Ruzzene, M., Scarpa, F., and Soranna, F., "Wave Beaming Effects in Two-
Dimensional Cellular Structures," Smart Materials & Structures, vol. 12, pp. 363-
372, 2003.
[43] di Scalea, F. L., Rizzo, P., and Marzani, A., "Propagation of Ultrasonic Guided
Waves in Lap-Shear Adhesive Joints: Case of Incident a(0) Lamb Wave," Journal
of the Acoustical Society of America, vol. 115, pp. 146-156, Jan 2004.
[44] Barshinger, J. N. and Rose, J. L., "Guided Wave Propagation in an Elastic Hollow
Cylinder Coated with a Viscoelastic Material," Ieee Transactions on Ultrasonics
Ferroelectrics and Frequency Control, vol. 51, pp. 1547-1556, 2004.
[45] Fan, Z. and Lowe, M. J. S., "Elastic Waves Guided by a Welded Joint in a Plate,"
Proceedings of the Royal Society A, vol. 465, pp. 2053-2067, 2009.
[46] Gopalakrishnan, S., Ruzzene, M., and Hanagud, S., Computational Techniques
for Structural Health Monitoring. London: Springer, 2011.
[47] Raghavan, A. and Cesnik, C. E. S., "Finite-Dimensional Piezoelectric Transducer
Modeling for Guided Wave Based Structural Health Monitoring," Smart
Materials & Structures, vol. 14, pp. 1448-1461, Dec 2005.
[48] Shen, Y. and Giurgiutiu, V., "Predictive Modeling of Nonlinear Wave
Propagation for Structural Health Monitoring with Piezoelectric Wafer Active
Sensors," Journal of Intelligent Material Systems and Structures, vol. 25, pp. 506-
520, 2014.
[49] Reddy, J. N., An Introduction to the Finite Element Method (Third Edition). New
York: McGraw-Hill, 1985.
[50] Cook, R. D., Malkus, D. S., Plesha, M. E., and Witt, R. J., Concepts and
Applications of Finite Element Analysis (4th Edition). New York: Wiley, 2001.
[51] Bathe, K. J., Finite Element Procedures. Englewood Cliffs: Printice Hall, 1997.
[52] Fellinger, P. and Langenberg, K. J., "Numerical Techniques for Elastic Wave
Propagation and Scattering," Proceedings of IUTAM Symposium, pp. 81-86, 1990.
[53] Leckey, C., Rogge, M. D., Miller, C., and Hinders, M., "Multiple-Mode Lamb
Wave Scattering Simulations Using 3d Elastodynamic Finite Integration
Technique," Ultrasonics, vol. 52, pp. 193-207, 2012.
[54] Leckey, C. A. C., Rogge, M. D., and Parker, F. R., "Microcracking in Composite
Laminates: Simulation of Crack-Induced Ultrasound Attenuation," in 39th Annual
178
Review of Progress in Quantitative Nondestructive Evaluation Conference,
Denver, CO, 2012, pp. 947-954.
[55] Lui, Q., Schoen, E., Daube, F., Randall, C., Liu, H., and Lee, P., "Large-Scale 3d
Finite-Difference Simulation of Elastic Wave Propagation in Borehole
Environments," Journal of Acoustic Society of America, vol. 96, pp. 3337-3337,
1994.
[56] Leutenegger, T. and Dual, J., "Detection of Defects in Cylindrical Structures
Using a Time Reverse Method and a Finite-Difference Approach," Ultrasonics,
vol. 40, pp. 721-725, 2002.
[57] Kelly, K., "Numerical Study of Love Wave Propagation," Geophysics, vol. 48, pp.
833-853, 1983.
[58] Virieux, J., "Wave Propagation in Heterogeneous Media : Velocity-Stress Finite-
Difference Method," Geophysics, vol. 49, pp. 1933-1957, 1984.
[59] Peng, H. K., Ye, L., Meng, G. A., Mustapha, S., and Li, F. C., "Concise Analysis
of Wave Propagation Using the Spectral Element Method and Identification of
Delamination in Cf/Ep Composite Beams," Smart Materials & Structures, vol. 19,
p. 085018, Aug 2010.
[60] Kim, Y., Ha, S., and Chang, F. K., "Time-Domain Spectral Element Method for
Built-in Piezoelectric-Actuator-Induced Lamb Wave Propagation Analysis," Aiaa
Journal, vol. 46, pp. 591-600, Mar 2008.
[61] Sundararaman, S. and Adams, D. E., "Modeling Guided Waves for Damage
Identification in Isotropic and Orthotropic Plates Using a Local Interaction
Simulation Approach," Journal of Vibration and Acoustics-Transactions of the
Asme, vol. 130, pp. -, Aug 2008.
[62] Lee, B. C. and Staszewski, W. J., "Lamb Wave Propagation Modelling for
Damage Detection: Ii. Damage Monitoring Strategy," Smart Materials &
Structures, vol. 16, pp. 260-274, Apr 2007.
[63] Banerjee, S. and Kundu, T., "Ultrasonic Field Modeling in Plates Immersed in
Fluid," International Journal of Solids and Structures, vol. 44, pp. 6013-6029,
Sep 2007.
[64] Gresil, M. and Giurgiutiu, V., "Time-Domain Hybrid Global-Local Concept for
Guided-Wave Propagation with Piezoelectric Wafer Active Sensor," Journal of
Intelligent Material Systems and Structures, vol. 24, pp. 1897-1911, 2013.
[65] Mal, A. and Zensheu, C., "A Semi-Numerical Method for Elastic Wave Scattering
Calculations," Geophysical Journal International, vol. 143, pp. 328-34, 11/ 2000.
[66] Srivastava, A. and di Scalea, F. L., "Quantitative Structural Health Monitoring by
Ultrasonic Guided Waves," Journal of Engineering Mechanics-Asce, vol. 136, pp.
937-944, Aug 2010.
179
[67] Mansur, W. J. and Brebbia, C. A., "Numerical Implementation of the Boundary
Element Method for Two Dimensional Transient Scalar Wave Propagation
Problems," Applied Mathematical Modeling, vol. 6, pp. 299-306, 1982.
[68] Zhu, W. and Rose, J. L., "Lamb Wave Generation and Reception with Time-
Delay Periodic Linear Arrays: A Bem Simulation and Experimental Study," Ieee
654-663, 1999.
[69] Aimi, A., Gazzola, S., and Guardasoni, C., "Energetic Boundary Element Method
Analysis of Wave Propagation in 2d Multilayered Media," Mathematical Methods
in the Applied Sciences, vol. 35, pp. 1140-1160, 2011.
[70] Soni, S., Das, S., and Chattopadhyay, A., "Simulation of Damage-Features in a
Lug Joint Using Guided Waves," Journal of Intelligent Material Systems and
Structures, vol. 20, pp. 1451-1464, Aug 2009.
[71] Pistone, E., Li, K., and Pizzo, P., "Noncontact Monitoring of Immersed Plates by
Means of Laser-Induced Ultrasounds," Structural Health Monitoring, vol. 12, pp.
549-565, 2013.
[72] Song, F., Huang, G. L., and Hudson, K., "Guided Wave Propagation in
Honeycomb Sandwich Structures Using a Piezoelectric Actuator/Sensor System,"
Smart Materials & Structures, vol. 18, pp. -, Dec 2009.
[73] Liu, Y., Lissenden, C. J., and Rose, J. L., "Cumulative Second Harmonics in
Weakly Nonlinear Plates and Shells," in Health Monitoring of Structural and
Biological Systems 2013, 2013.
[74] Leckey, C. A. C., Rogge, M. D., and Parker, R., "Guided Waves in Anisotropic
and Quasi-Isotropic Aerospace Composites: Three-Dimensional Simulation and
Experiment," Ultrasonics, vol. 54, pp. 385-394, 2014.
[75] Schubert, F., Peiffer, A., and Kohler, B., "The Elastodynamic Finite Integration
Technique for Waves in Cylindrical Geometries," Journal of the Acoustical
Society of America, vol. 104, pp. 2604-2614, 1998.
[76] Bingham, J. and Hinders, M., "3d Elastodynamic Finite Integration Technique
Simulation of Guided Waves in Extended Built-up Structures Containing Flaws,"
Journal of Computational Acoustics, vol. 18, pp. 1-28, 2010.
[77] Rudd, K. E., Leonard, K. R., Bingham, J. P., and Hinders, M. K., "Simulation of
Guided Waves in Complex Piping Geometries Using the Elastodynamic Finite
Integration Technique," Journal of the Acoustical Society of America, vol. 121, pp.
1449-1458, Mar 2007.
[78] Halkjaer, S., "Elastic Wave Propagation in Anisotropic Inhomogeneous
Materials," Ph.D. Thesis, Institute of Mathematical Modelling, Technical
University of Denmark, Denmark, 2000.
180
[79] Schubert, F., "Numerical Time-Domain Modeling of Linear and Nonlinear
Ultrasonic Wave Propagation Using Finite Integration Techniques - Theory and
Applications," Ultrasonics, vol. 42, pp. 221-229, Apr 2004.
[80] Wilcox, P. D., Lowe, M., and Cawley, P., "Omnidirectional Guided Wave
Inspection of Large Metallic Plate Structures Using an Emat Array," Ieee
653-665, Apr 2005.
[81] Murayama, R. and Mizutani, K., "Conventional Electromagnetic Acoustic
Transducers Development for Optimum Lamb Wave Modes," Ultrasonics, vol.
40, pp. 491-495, 2002.
[82] Ann, A. K., Kim, H. M., Jhang, K. Y., Park, I. K., Kwag, N. G., and Lee, C. M.,
"Detection of Laser Excited Lamb Wave Using Air Coupled Transducer and
Wave Mode Identification Using Wavelet Transform," Review of Progress in
Quantitative Nondestructive Evaluation, vol. 760, pp. 1018-1025, 2004.
[83] Garcia-Hernanadez, M. J., Chavez, J. A., Yanez, Y., Kichou, H. B., Prego-Borges,
J. L., Salazar, J., Turo, A., and de Espinosa, F. M., "Ultrasonic Lamb Wave Nde
System Using an Air Coupled Concave Array Transducer," IEEE Ultrasonics
Symposium, 2004 IEEE International Ultrasonics, Ferroelectronics, and
Frequency Control Joint 50th Anniversary Conference, vol. 2, pp. 1282-1285,
2004.
[84] Rose, J. L., "A Comb Transducer for Mode Control in Guided Wave Nde,"
Ultrasonics, vol. 36, pp. 163-169, 1998.
[85] Wevers, M., Fransens, F., and Leuven, K. U., "Ultrasonic Lamb Wave Inspection
of Aircraft Components Using Integrated Optical Fibre Sensing Technology,"
ECNDT 2006, vol. Mo.2.7.1, 2006.
[86] Li, F. C., Murayama, H., Kageyama, K., and Shirai, T., "Guided Wave and
Damage Detection in Composite Laminates Using Different Fiber Optic Sensors,"
Sensors, vol. 9, pp. 4005-4021, May 2009.
[87] Dupont, M., Osmont, R., Gouyon, R., and Balageas, D. L., "Permanent
Monitoring of Damage Impacts by a Piezoelectric Sensor Based Integrated
System," Structural Health Monitoring, pp. 561-570, 2000.
[88] Yu, L. and Giurgiutiu, V., "Piezoelectric Wafer Active Sensors," in Encyclopedia
of Structural Health Monitoring vol. 3, C. Boller, F. K. Chang, and Y. Fujino,
Eds., ed: John Wiley & Sons, Ltd., 2009, pp. 1013-1027.
[89] Williams, R. B., Park, G., Inman, D. J., and Wilkie, W. K., "An Overview of
Composite Actuators with Piezoceramic Fibers," presented at the 20th
International Modal Analysis Conference, Los Angeles, CA, USA, 2002.
[90] Raghavan, A. and Cesnik, C. E. S., "Review of Guided-Wave Strucutral Health
Monitoring," The Shock and Vibration Digest, vol. 39, pp. 91-114, 2007.
181
[91] di Scalea, F. L., Matt, H., Bartoli, I., Coccia, S., Park, G., and Farrar, C., "Health
Monitoring of Uav Wing Skin-to-Spar Joints Using Guided Waves and Macro
Fiber Composite Transducers," Journal of Intelligent Material Systems and
Structures, vol. 18, pp. 373-388, Apr 2007.
[92] Rose, J. L., "A Baseline and Vision of Ultrasonic Guided Wave Inspection
Potential," Journal of Pressure Vessel Technology, vol. 124, pp. 273-282, 2002.
[93] Culshaw, B., Pierce, S. G., and Staszewski, W. J., "Condition Monitoring in
Composite Materials: An Integrated Systems Approach," Proceedings of the
Institution of Mechanical Engineers, Journal of Systems and Control Engineering,
Part I, vol. 212, pp. 189-202, 1998.
[94] Sohn, H., "Noncontact Laser Sensing Technology for Structural Health
Monitoring and Nondestructive Testing," in Bioinspiration, Biomimetics, and
Bioreplication 2014, San Diego, California, USA, 2014, p. 90550W.
[95] Rajic, N., Davis, C., and Thomson, A., "Acoustic-Wave-Mode Separation Using a
Distributed Bragg Grating Sensor," Smart Materials & Structures, vol. 18, pp. -,
Dec 2009.
[96] Ke, W., Castaings, M., and Bacon, C., "3d Finite Element Simulations of an Air-
Coupled Ultrasonic Ndt System," Ndt & E International, vol. 42, pp. 524-533,
2009.
[97] Lee, H. and Sohn, H., "Damage Detection in Nuclear Power Plant Pipes Using
Laser-Based Mechanical Impedance," Review of Progress in Quantitative
Nondestructive Evaluation, 2014.
[98] Giurgiutiu, V., Zagrai, A., and Bao, J. J., "Damage Identification in Aging
Aircraft Structures with Piezoelectric Wafer Active Sensors," Journal of
Intelligent Material Systems and Structures, vol. 15, pp. 673-687, Sep-Oct 2004.
[99] Giurgiutiu, V., Xu, B. L., and Liu, W. P., "Development and Testing of High-
Temperature Piezoelectric Wafer Active Sensors for Extreme Environments,"
Structural Health Monitoring-an International Journal, vol. 9, pp. 513-525, Nov
2010.
[100] Santoni, G. B., Yu, L., Xu, B., and Giurgiutiu, V., "Lamb Wave-Mode Tuning of
Piezoelectric Wafer Active Sensors for Structural Health Monitoring," Journal of
Vibration and Acoustics-Transactions of the Asme, vol. 129, pp. 752-762, Dec
2007.
[101] Giurgiutiu, V., "Tuned Lamb Wave Excitation and Detection with Piezoelectric
Wafer Active Sensors for Structural Health Monitoring," Journal of Intelligent
Material Systems and Structures, vol. 16, pp. 291-305, Apr 2005.
[102] Drain, L. E., The Laser Doppler Concepts. New York, NY: John Wiley Sons,
1980.
[103] Sohn, H., Dutta, D., Yang, J. Y., DeSimio, M., Olson, S., and Swenson, E.,
"Automated Detection of Delamination and Disbond from Wavefield Images
182
Obtained Using a Scanning Laser Vibrometer," Smart Materials & Structures, vol.
20, p. 045017, 2011.
[104] Staszewski, W. J., Lee, B. C., and Traynor, R., "Fatigue Crack Detection in
Metallic Structures with Lamb Waves and 3d Laser Vibrometry," Measurement
Science & Technology, vol. 18, pp. 727-739, Mar 2007.
[105] Michaels, T. E., Ruzzene, M., and Michaels, J. E., "Incident Wave Removal
through Frequency-Wavenumber Filtering of Full Wavefield Data," Review of
Progress in Quantitative Nondestructive Evaluation, vol. 1096, pp. 604-611, 2009.
[106] Swenson, E. D., Sohn, H., Olson, S. E., and DeSimio, M. P., "A Comparison of
1d and 3d Laser Vibrometry Measurements of Lamb Waves," Proceedings of
SPIE, vol. 7650, p. 765003, 2010.
[107] Olson, S. E., DeSimio, M. P., Matthew, J. D., Swenson, E. D., and Sohn, H.,
"Computational Lamb Wave Model Validation Using 1d and 3d Laser
Vibrometer Measurement," Proc. SPIE 7650, Health Monitoring of Structural
and Biological Systems 2010, vol. 7650, p. 76500M, 2010.
[108] Ostachowicz, W., Wanddowski, T., and Malinowski, P., "Damage Detection
Using Laser Vibrometry," in 2nd International Symposium on NDT in Aerospace
2010, 2010.
[109] Kudela, P., Radzienski, M., and Ostachowicz, W., "Identification of Cracks in
Thin-Walled Structures by Means of Wavenumber Filtering," Mechanical
Systems and Signal Processing, vol. 50-51, pp. 456-466, Jan 2015.
[110] Ostachowicz, W., Radzienski, M., and Kudela, P., "50th Anniversary Article:
Comparison Studies of Full Wavefield Signal Processing for Crack Detection,"
Strain, vol. 50, pp. 275-291, 2014.
[111] Yu, L., Leckey, C., and Tian, Z., "Study on Crack Scattering in Aluminum Plates
with Lamb Wave Frequency-Wavenumber Analysis," Smart Materials &
Structures, vol. 22, pp. 1-12, 2013.
[112] Staszewski, W. J., Mahzan, S., and Traynor, R., "Health Monitoring of Aerospace
Composite Structures - Active and Passive Approach," Composites Science and
Technology, vol. 69, pp. 1678-1685, Sep 2009.
[113] Alleyne, D. N. and Cawley, P., "A Two Dimensional Fourier Transform Method
for the Measurement of Propagating Multimode Signals," Journal of Acoustic
Society of America, vol. 89, pp. 1159-1168, 1991.
[114] Hayashi, T. and Kawashima, K., "Single Mode Extraction from Multiple Modes
of Lamb Wave and Its Application to Defect Detection," JSME International
Journal Series A, Solid Mechanics and Material Engineering, vol. 46, pp. 620-
626, 2003.
[115] Michaels, J. E., "Detection, Localization and Characterization of Damage in
Plates with an in Situ Array of Spatially Distributed Ultrasonic Sensors," Smart
Materials & Structures, vol. 17, pp. -, Jun 2008.
183
[116] Hall, J. S. and Michaels, J. E., "Minimum Variance Ultrasonic Imaging Applied
to an in Situ Sparse Guided Wave Array," Ieee Transactions on Ultrasonics
Ferroelectrics and Frequency Control, vol. 57, pp. 2311-2323, Oct 2010.
[117] Wang, C. H., Rose, J. T., and Chang, F. K., "A Synthetic Time-Reversal Imaging
Method for Structural Health Monitoring," Smart Materials & Structures, vol. 13,
pp. 415-423, Apr 2004.
[118] Michaels, J. E. and Michaels, T. E., "Guided Wave Signal Processing and Image
Fusion for in Situ Damage Localization in Plates," Wave Motion, vol. 44, pp. 482-
492, Jun 2007.
[119] Yu, L. and Leckey, C., "Lamb Wave-Based Quantitative Crack Detection Using a
Focusing Array Algorithm," Journal of Intelligent Material Systems and
Structures, vol. 29, pp. 1138-1152, 2013.
[120] Su, Z. Q., Cheng, L., Wang, X. M., Yu, L., and Zhou, C., "Predicting
Delamination of Composite Laminates Using an Imaging Approach," Smart
Materials & Structures, vol. 18, pp. -, Jul 2009.
[121] Cho, H. J. and Lissenden, C. J., "Structural Health Monitoring of Fatigue Crack
Growth in Plate Structures with Ultrasonic Guided Waves," Structural Health
Monitoring-an International Journal, vol. 11, pp. 393-404, Jul 2012.
[122] Chen, X., Michaels, J. E., Lee, S. J., and Michaels, T. E., "Load-Differential
Imaging for Detection and Localization of Fatigue Cracks Using Lamb Waves,"
Ndt & E International, vol. 51, pp. 142-149, Oct 2012.
[123] Wilcox, P. D., "Guided-Wave Array Methods," in Encyclopedia of Structural
Health Monitoring vol. 1, C. Boller, F. K. Chang, and Y. Fujino, Eds., ed: John
Wiley & Sons, Ltd., 2009, pp. 333-352.
[124] Purekar, A. S., Pines, D. J., Sundararaman, S., and Adams, D. E., "Directional
Piezoelectric Phased Array Filters for Detecting Damage in Isotropic Plates,"
Smart Materials & Structures, vol. 13, pp. 838-850, Aug 2004.
[125] Ambrozinski, L., Stepinski, T., and Uhl, T., "2d Aperture Synthesis for Lamb
Wave Imaging Using Co-Arrays," presented at the Health Monitoring of
Structural and Biological Systems 2014, San Diego, CA, 2014.
[126] de Marchi, L., Moll, J., and Marzani, A., "A Sparsity Promoting Algorithm for
Time of Flight Estimation in Guided Waves-Based Shm," in EWSHM-7th
European Workshop on Structural Health Monitoring, Nantes, France, 2014, pp.
583-590.
[127] Engholm, M. and Stepinski, T., "Direction of Arrival Estimation of Lamb Waves
Using Circular Arrays," Structural Health Monitoring-an International Journal,
vol. 10, pp. 467-480, Sep 2011.
[128] Fromme, P., Wilcox, P. D., Lowe, M. J. S., and Canvley, P., "On the
Development and Testing of a Guided Ultrasonic Wave Array for Structural
184
Integrity Monitoring," Ieee Transactions on Ultrasonics Ferroelectrics and
Frequency Control, vol. 53, pp. 777-785, Apr 2006.
[129] Leleux, A., Micheau, P., and Castaings, M., "Long Range Detection of Defects in
Composite Plates Using Lamb Waves Generated and Detected by Ultrasonic
Phased Array Probes," Journal of Nondestructive Evaluation, vol. 32, pp. 200-214,
Jun 2013.
[130] Malinowski, P., Wandowski, T., Trendafilova, I., and Ostachowicz, W., "A
Phased Array-Based Method for Damage Detection and Localization in Thin
Plates," Structural Health Monitoring-an International Journal, vol. 8, pp. 5-15,
Jan 2009.
[131] Ambrozinski, L., Stepinski, T., and Uhl, T., "Efficient Tool for Designing 2d
Phased Arrays in Lamb Waves Imaging of Isotropic Structures," Journal of
Intelligent Material Systems and Structures, p. online, 2014.
[132] Koduru, J. P., Momeni, S., and Rose, J. L., "Phased Annular Array Transducers
for Omnidirectional Guided Wave Mode Control in Isotropic Plate Like
Structures," Smart Materials and Structures, vol. 22, p. 8, Dec 2013.
[133] Yan, F. and Rose, J. L., "Guided Wave Phased Array Beam Steering in
Composite Plates - Art. No. 65320g," Health Monitoring of Structural and
Biological Systems 2007, vol. 6532, pp. G5320-G5320, 2007.
[134] Rajagopalan, J., Multila, K. H. M. O. A., Balasubramaniam, K., and
Krishnamurthy, C. V., "A Phase Reconstruction Algorithm for Lamb Wave Based
Structural Health Monitoring of Anisotropic Multilayered Composite Plates,"
Journal of the Acoustical Society of America, vol. 119, pp. 872-878, Feb 2006.
[135] Osterc, P., Kim, D., and Yoo, B., "Phased Array Beamsteering in Composite
Laminates for Guided Wave Structural Health Monitoring," presented at the
ASME 2013 Conference on Smart Materials, Adaptive Structures and Intelligent
Systems, Snowbird, Utah, USA, 2013.
[136] Wang, L. and Yuan, F. G., "Group Velocity and Characteristic Wave Curves of
Lamb Waves in Composites: Modeling and Experiments," Composites Science
and Technology, vol. 67, pp. 1370-1384, Jun 2007.
[137] Glushkov, E., Glushkova, N., and Eremin, A., "Group Velocity of Cylindrical
Guided Waves in Anisotropic Laminate Composites," Journal of Acoustic Society
of America, vol. 135, pp. 148-154, 2014.
[138] Lamb, H., "On Waves in an Elastic Plate," Proceedings of Royal Society, A:
Mathematical, Physical and Engineering Sciences, vol. 93, pp. 114-128, 1917.
[139] Cegla, F. B., Cawley, P., and Lowe, M. J. S., "Material Property Measurement
Using the Quasi-Scholte Mode - a Waveguide Sensor," Journal of the Acoustical
Society of America, vol. 117, pp. 1098-1107, Mar 2005.
185
[140] Yu, L. and Tian, Z., "Lamb Wave Structural Health Monitoring Using a Hybrid
Pzt-Laser Vibrometer Approach," Structural Health Monitoring, vol. 12, pp. 469-
483, 2013.
[141] Sohn, H., Dutta, D., Yang, J. Y., Park, H. J., DeSimio, M. P., Olson, S. E., and
Swenson, E. D., "Delamination Detection in Composites through Guided Wave
Field Image Processing," Composites Science and Technology, vol. 71, pp. 1250-
1256, 2011.
[142] Tian, Z. and Yu, L., "Lamb Wave Frequency-Wavenumber Analysis and
Decomposition," Journal of Intelligent Material Systems and Structures, vol. 25,
pp. 1107-1123, 2014.
[143] Glushkov, E., Glushakova, N., Golub, M. V., Moll, J., and Fritzen, C. P., "Wave
Energy Trapping and Localization in a Plate with a Delamination," Smart
Materials & Structures, vol. 21, p. 125001, 2012.
[144] Johnson, D. H. and Dudgeon, D. E., Array Signal Processing: Concepts and
Techniques. Upper Saddle River, New Jersey, USA: Prentice-Hall Inc., 1993.
[145] Oppenheim, A. V., Willsky, A. S., and Hamid, S., Signals and Systems (2nd
Edition). New Jersey, USA: Prentice Hall, 1997.
[146] Cohen, L., Time Frequency Analysis: Theory and Applications. Upper Saddle
River, New Jersey, USA: Prentice-Hall Inc., 1994.
[147] Leckey, C. A. C. and Seebo, J. P., "Guided Wave Energy Trapping to Detect
Hidden Multilayer Delamination Damage," AIP Conference Proceedings, vol.
1650, pp. 1162-1169, 2015.
[148] Tian, Z., Leckey, C. A. C., Seebo, J. P., and Yu, L., "Guided Wave Delamination
Detection and Quantification with Wavefield Data Analysis," presented at the
ASME 2014 Conference on Smart Material, Adaptive Structures and Intelligent
Systems, Newport, Rhode Island, 2014.
[149] Tian, Z., Yu, L., and Leckey, C. A. C., "Delamination Detection and
Quantification on Laminated Composite Structures with Lamb Waves and
Wavenumber Analysis," Journal of Intelligent Material Systems and Structures,
2014.
[150] Na, W. B. and Kundu, T., "Underwater Pipeline Inspection Using Guided
Waves," Journal of Pressure Vessel Technology, vol. 124, pp. 196-200, 2002.
[151] Bingham, J., Hinders, M., and Friedman, A., "Lamb Wave Detection of Limpet
Mines on Ship Hulls," Ultrasonics, vol. 49, pp. 706-722, Dec 2009.
[152] Chen, J. G., Su, Z. Q., and Cheng, L., "Identification of Corrosion Damage in
Submerged Structures Using Fundamental Anti-Symmetric Lamb Waves," Smart
Materials & Structures, vol. 19, p. 015004, Jan 2010.
186
[153] Rizzo, P., Han, J. G., and Ni, X. L., "Structural Health Monitoring of Immersed
Structures by Means of Guided Ultrasonic Waves," Journal of Intelligent
Material Systems and Structures, vol. 21, pp. 1397-1407, Sep 2010.
[154] Pistone, E., Li, K. Y., and Rizzo, P., "Noncontact Monitoring of Immersed Plates
by Means of Laser-Induced Ultrasounds," Structural Health Monitoring-an
International Journal, vol. 12, pp. 549-565, Sep 2013.
[155] Koduru, J. P. and Rose, J. L., "Mode Controlled Guided Wave Tomography
Using Annular Array Transducers for Shm of Water Loaded Plate Like
Structures," Smart Materials and Structures, vol. 22, p. 10, Dec 2013.
[156] Chimenti, D. E. and Nayfeh, A. H., "Leaky Lamb Waves in Fibrous Composite
Laminates," Journal of Applied Physics, vol. 58, pp. 4531-4538, 1985.
[157] Wu, J. and Zhu, Z., "The Propagation of Lamb Waves in a Plate Bordered with
Layers of a Liquid," Journal of Acoustic Society of America, vol. 91, pp. 861-867,
1992.
[158] Shuvalov, A. L., Poncelet, O., and Deschamps, M., "Analysis of the Dispersion
Spectrum of Fluid-Loaded Anisotropic Plates: Leaky-Wave Branches," Journal of
Sound and Vibration, vol. 296, pp. 494-517, 2006.
[159] Cegla, F. B., Cawley, P., and Lowe, M. J. S., "Fluid Bulk Velocity and
Attenuation Measurements in Non-Newtonian Liquids Using a Dipstick Sensor,"
Measurement Science & Technology, vol. 17, pp. 264-274, Feb 2006.
[160] "Managing Gas Accumulation in Emergency Core Cooling, Decay Heat Removal,
and Containment Spray Systems," NRC Generic Letter 2008-01, p.
ML072910759, January 11 2008.
[161] Michaels, J. E., Dawson, A. J., Michaels, J. E., and Ruzzene, M., "Approaches to
Hybrid Shm and Nde of Composite Aerospace Structures," in Health Monitoring
of Structural and Biological Systems 2014, San Diego, CA, USA, 2014.
[162] Su, Z. Q., Wang, X. M., Chen, Z. P., and Ye, L., "A Hierarchical Data Fusion
Scheme for Identifying Multi-Damage in Composite Structures with a Built-in
Sensor Network," Smart Materials & Structures, vol. 16, pp. 2067-2079, Dec
2007.
[163] Kim, S. B. and Sohn, H., "Instantaneous Reference-Free Crack Detection Based
on Polarization Characteristics of Piezoelectric Materials," Smart Materials &
Structures, vol. 16, pp. 2375-2387, Dec 2007.
[164] Ihn, J. B. and Chang, F. K., "Pitch-Catch Active Sensing Methods in Structural
Health Monitoring for Aircraft Structures," Structural Health Monitoring-an
International Journal, vol. 7, pp. 5-19, Mar 2008.
[165] Yu, L. and Leckey, C., "Lamb Wave-Based Quantitative Crack Detection Using a
Focusing Array Algorithm," vol. 29, pp. 1138-1152, 2013.
[166] Kundu, T., "Acoustic Source Localization," Ultrasonics, vol. 54, pp. 24-38, 2014.
187
[167] Tua, P. S., Quek, S. T., and Wang, Q., "Detection of Cracks in Plates Using
Piezo-Actuated Lamb Waves," Smart Materials & Structures, vol. 13, pp. 643-
660, Aug 2004.
[168] Bagheri, A., Li, K. Y., and Rizzo, P., "Reference-Free Damage Detection by
Means of Wavelet Transform and Empirical Mode Decomposition Applied to
Lamb Waves," Journal of Intelligent Material Systems and Structures, vol. 24, pp.
194-208, Jan 2013.
[169] Hall, J. S., McKeon, P., Satyanarayan, L., Michaels, J. E., Declercq, N. F., and
Berthelot, Y. H., "Minimum Variance Guided Wave Imaging in a Quasi-Isotropic
Composite Plate," Smart Materials and Structures, vol. 20, p. 8, Feb 2011.
[170] Zhao, X. L., Gao, H. D., Zhang, G. F., Ayhan, B., Yan, F., Kwan, C., and Rose, J.
L., "Active Health Monitoring of an Aircraft Wing with Embedded Piezoelectric
Sensor/Actuator Network: I. Defect Detection, Localization and Growth
Monitoring," Smart Materials & Structures, vol. 16, pp. 1208-1217, Aug 2007.
[171] Levine, R. M. and Michaels, J. E., "Model-Based Imaging of Damage with Lamb
Waves Via Sparse Reconstruction," Journal of Acoustic Society of America, vol.
133, pp. 1525-1534, 2013.
[172] Castaings, M., Le Clezio, E., and Hosten, B., "Modal Decomposition Method for
Modeling the Interaction of Lamb Waves with Cracks," Journal of the Acoustical
Society of America, vol. 112, pp. 2567-2582, Dec 2002.
[173] An, Y.-K., Kwon, Y., and Sohn, H., "Noncontact Laser Ultrasonic Crack
Detection for Plates with Additional Structural Complexities," Structural Health
Monitoring, vol. 12, pp. 522-538, 2013.
[174] Cammarata, M., Rizzo, P., Dutta, D., and Sohn, H., "Application of Principal
Component Analysis and Wavelet Transform to Fatigue Crack Detection in
Waveguides," Smart Structures and Systems, vol. 6, pp. 349-362, May 2010.
[175] Betz, D. C., Staszewski, W. J., Thursby, G., and Culshaw, B., "Multi-Functional
Fibre Bragg Grating Sensors for Fatigue Crack Detection in Metallic Structures,"
Proceedings of the Institution of Mechanical Engineers Part G-Journal of
Aerospace Engineering, vol. 220, pp. 453-461, Oct 2006.
[176] Lu, Y., Ye, L., and Su, Z. Q., "Crack Identification in Aluminium Plates Using
Lamb Wave Signals of a Pzt Sensor Network," Smart Materials & Structures, vol.
15, pp. 839-849, Jun 2006.
[177] Renshaw, J., Holland, S. D., and Thompson, R. B., "Measurement of Crack
Opening Stresses and Crack Closure Stress Profiles from Heat Generation in
Vibrating Cracks," Applied Physics Letters, vol. 93, pp. -, Aug 25 2008.
[178] Rizzo, P., Cammarata, M., Dutta, D., Sohn, H., and Harries, K., "An
Unsupervised Learning Algorithm for Fatigue Crack Detection in Waveguides,"
Smart Materials & Structures, vol. 18, pp. -, Feb 2009.
188
[179] Ihn, J. B. and Chang, F. K., "Detection and Monitoring of Hidden Fatigue Crack
Growth Using a Built-in Piezoelectric Sensor/Actuator Network: I. Diagnostics,"
Smart Materials & Structures, vol. 13, pp. 609-620, Jun 2004.
[180] Lu, Y., Ye, L., Su, Z. Q., and Yang, C. H., "Quantitative Assessment of through-
Thickness Crack Size Based on Lamb Wave Scattering in Aluminium Plates," Ndt
& E International, vol. 41, pp. 59-68, Jan 2008.
[181] An, Y. K., Kim, J. M., and Sohn, H., "Laser Lock-in Thermography for Detection
of Surface-Breaking Fatigue Cracks on Uncoated Steel Structures," Ndt & E
International, vol. 65, pp. 54-63, Jul 2014.
[182] An, Y.-K., Park, B., and Sohn, H., "Complete Noncontact Laser Ultrasonic
Imaging for Automated Crack Visualization in a Plate," Smart Materials &
Structures, vol. 22, p. 025022, 2013.
[183] Yu, L., Tian, Z., and Leckey, C. A. C., "Crack Imaging and Quantification in
Aluminum Plates with Guided Wave Wavenumber Analysis Methods,"
Ultrasonics, 2015.
[184] Castaings, M. and Hosten, B., "Ultrasonic Guided Waves for Health Monitoring
of High-Pressure Composite Tanks," Ndt & E International, vol. 41, pp. 648-655,
Dec 2008.
[185] Tian, Z., Yu, L., and Leckey, C. A. C., "Guided Wave Propagation Study on
Laminated Composites by Frequency-Wavenumber Technique," in
Nondestructive Characterization for Composite Materials, Aerospace
Engineering, Civil Infrastructure, and Homeland Security 2014, San Diego,
California, USA, 2014, p. 90631O.
[186] Yu, L., Cara, L., and Tian, Z., "Study on Crack Scattering in Aluminum Plates
with Lamb Wave Frequency-Wavenumber Analysis," Smart Materials &
Structures, vol. 22, p. 065019, 2013.
[187] Tian, Z., Leckey, C., Rogge, M. D., and Yu, L., "Crack Detection with Lamb
Wave Wavenumber Analysis," presented at the Health Monitoring of Structural
and Biological Systems 2013, San Diego, CA, 2013.
[188] Tian, Z., Huang, G. L., and Yu, L., "Study of Guided Wave Propagation in
Honeycomb Sandwich Structures," presented at the ASME 2014 Conference on
Smart Materials, Adaptive Structures and Intelligent Systems, Newport, Rhode
Island, USA, 2014.
[189] Tian, Z. and Yu, L., "Lamb Wave Propagation Study Using Frequency-
Wavenumber Analysis," ASME 2012 Conference on Smart Material, Adaptive
Structures and Intelligent Systems, vol. SMASIS2012-8013, 2012.
[190] An, Y. K., Park, B., and Sohn, H., "Complete Noncontact Laser Ultrasonic
Imaging for Automated Crack Visualization in a Plate," Smart Materials and
Structures, vol. 22, Feb 2013.
189
[191] Ip, K. H. and Mai, Y. W., "Delamination Detection in Smart Composite Beams
Using Lamb Waves," Smart Materials & Structures, vol. 13, pp. 544-551, Jun
2004.
[192] Ricci, F., Mal, A. K., Monaco, E., Maio, L., Boffa, N. D., Di Palma, M., and
Lecce, L., "Guided Waves in Layered Plate with Delaminations," presented at the
Proceedings of 7th European Workshop on Structural Health Monitoring, La Cite,
Nantes, France, 2014.
[193] Salamone, S., Bartoli, I., di Scalea, F. L., and Coccia, S., "Guided-Wave Health
Monitoring of Aircraft Composite Panels under Changing Temperature," Journal
of Intelligent Material Systems and Structures, vol. 20, pp. 1079-1090, Jun 2009.
[194] Mesnil, O., Leckey, C. A. C., and Ruzzene, M., "Instantaneous and Local
Wavenumber Estimations for Damage Quantification in Composites," Structural
Health Monitoring, vol. 14, pp. 193-204, 2014.
[195] Juarez, P. D. and Leckey, C. A. C., "Multi-Frequency Local Wavenumber
Analysis and Ply Correlation of Delamination Damage," Ultrasonics, vol. 62, pp.
56-65, 2015.
[196] Staszewski, W. J., Mahzan, S., and Traynor, R., "Health Monitoring of Aerospace
Composite Structures-Active and Passive Approach," Composites Science and
Technology, vol. 69, pp. 1678-1685, 2009.
[197] Alleyne, D. N. and Cawley, P., "The Interaction of Lamb Waves with Defects,"
Ieee Transactions on Ultrasonics Ferroelectrics and Frequency Control, vol. 39,
pp. 381-397, 1992.
[198] Boller, C., "Next Generation Structural Health Monitoring and Its Integration into
Aircraft Design," International Journal of Systems Science, vol. 31, pp. 1333-
1349, 2000.
[199] Ong, W. H. and Chiu, W. K., "Redirection of Lamb Waves for Structural Health
Monitoring," Smart Materials Research, vol. 2012, p. 718686, 2012.
[200] Kundu, T., Das, S., Martin, S. A., and Jata, K. V., "Locating Point of Impact in
Anisotropic Fiber Reinforced Composite Plates," Ultrasonics, vol. 48, pp. 193-
201, Jul 2008.
[201] Su, Z. Q., Cheng, L., Wang, X. M., Yu, L., and Zhou, C., "Predicting
Delamination of Composite Laminates Using an Imaging Approach," Smart
Materials & Structures, vol. 18, p. 074002, Jul 2009.
[202] Salas, K. I. and Cesnik, C. E. S., "Guided Wave Structural Health Monitoring
Using Clover Transducers in Composite Materials," Smart Materials & Structures,
vol. 19, p. 015014, Jan 2010.
[203] Song, F., Huang, G. L., and Hu, G. K., "Online Guided Wave-Based Debonding
Detection in Honeycomb Sandwich Structures," Aiaa Journal, vol. 50, pp. 284-
293, 2012.
190
[204] Giurgiutiu, V. and Soutis, C., "Enhanced Composites Integrity through Structural
Health Monitoring," Applied Composite Materials, vol. 19, pp. 813-829, 2012.
[205] Guo, N. and Cawley, P., "The Interaction of Lamb Waves with Delaminations in
Composite Laminates," Journal of Acoustic Society of America, vol. 94, pp. 2240-
2246, 1993.
[206] Petculescu, G., Krishnaswamy, S., and Achenbach, J. D., "Group Delay
Measurements Using Modally Selective Lamb Wave Transducers for Detection
and Sizing of Delaminations in Composites," Smart Materials & Structures, vol.
17, p. 015007, Feb 2008.
[207] Purekar, A. S. and Pines, D. J., "Damage Detection in Thin Composite Laminates
Using Piezoelectric Phased Sensor Arrays and Guided Lamb Wave
Interrogation," Journal of Intelligent Material Systems and Structures, vol. 21, pp.
995-1010, Jul 2010.
[208] Yeum, C. M., Sohn, H., Ihn, J. B., and Lim, H. J., "Instantaneous Delamination
Detection in a Composite Plate Using a Dual Piezoelectric Transducer Network,"
Composite Structures, vol. 94, pp. 3490-3499, 2012.
[209] Hayashi, T. and Kawashima, K., "Multiple Reflections of Lamb Waves at a
Delamination," Ultrasonics, vol. 40, pp. 193-197, 2002.
[210] Ramadas, C., Balasubramaniam, K., Joshi, M., and Krishnamurthy, C. V.,
"Interaction of the Primary Anti-Symmetric Lamb Mode (A0) with Symmetric
Delaminations: Numerical and Experimental Studies," Smart Materials &
Structures, vol. 18, p. 085011, Aug 2009.
[211] Williams, G., Trask, R., and Bond, I., "A Self-Healing Carbon Fibre Reinforced
Polymer for Aerospace Applications," Composites Part a-Applied Science and
Manufacturing, vol. 38, pp. 1525-1632, 2007.
[212] Fatemi, J. and Lemmen, M., "Effective Thermal/Mechanical Properties of
Honeycomb Core Panels for Hot Structure Applications," Journal of Spacecraft
and Rockets, vol. 46, pp. 514-525, 2009.
[213] T., B., Honeycomb Technology: Materials, Design, Manufacturing, Applications
and Testing: Springer, 1997.
[214] Zenkert, D., The Handbook of Sandwich Construction. London: EMAS
Publishing, 1997.
[215] Vinson, J. R., The Behavior of Sandwich Structures of Isotropic and Composite
Materials. Lancaster: Technomic Publishing Company, 1999.
[216] Minakuchi, S., Okabe, Y., and Takeda, N., "Real-Time Detection of Debonding
between Honeycomb Core and Facesheet Using a Small-Diameter Fbg Sensor
Embedded in Adhesive Layer," Journal of Sandwich Structures & Materials, vol.
9, pp. 9-33, 2007.
191
[217] Takeda, N., Minakuchi, S., and Okabe, Y., "Smart Composite Sandwich
Structures for Future Aerospace Application -Damage Detection and Suppression:
A Review," Journal of Solid Mechanics and Materials Engineering, vol. 1, pp. 3-
17, 2007.
[218] Goswami, S. and Becker, W., "Analysis of Debonding Fracture in a Sandwich
Plate with Hexagonal Core," Composite Structures, vol. 49, pp. 385-392, 8// 2000.
[219] Hosseini, S. M., Willberg, C., Kharaghani, A., and Gabbert, U., "Characterization
of the Guided Wave Propagation in Simplified Foam Honeycomb and Hollow
Sphere Structures," Composite: Part B, vol. 56, pp. 553-566, 2014.
[220] Chakraborty, N., Rathod, V. T., Mahapatra, D. R., and Gopalakrishnan, S.,
"Guided Wave Based Detection of Damage in Honeycomb Core Sandwich
Structures," Ndt & E International, vol. 49, pp. 27-33, Jul 2012.
[221] Hou, Y., Neville, R., Scarpa, F., Remillat, C., Gu, B., and Ruzzene, M., "Graded
Conventional-Auxetic Kirigami Sandwich Structures: Flatwise Compression and
Edgewise Loading," Composites Part B: Engineering, vol. 59, pp. 33-42, 2014.
192

Guided Waves Introduction 1

Uploaded by

Copyright:

Available Formats

Guided Waves Introduction 1

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Guided Waves Introduction 1

Uploaded by

Copyright:

Available Formats

GUIDED WAVE BASED INTEGRATED STRUCTURAL HEALTH MONITORING

AND NONDESTRUCTIVE EVALUATION

Submitted in Partial Fulfillment of the Requirements

For the Degree of Doctor of Philosophy in

College of Engineering and Computing

University of South Carolina

Lingyu Yu, Major Professor

Victor Giurgiutiu, Committee Member

Yuh J. Chao, Committee Member

Paul Ziehl, Committee Member

Cara A.C. Leckey, Committee Member

Lacy Ford, Senior Vice Provost and Dean of Graduate Studies

All rights reserved

INFORMATION TO ALL USERS

All rights reserved.

dedicated guidance, thoughtful advice, tremendous support and willingness to allow me

continuously encourage me in my future career.

I would like to give special thanks to my committee members. I am lucky to have

knowledge and broadened my horizon.

My gratitude is also extended to my dear colleagues in VSHM and LAMSS, who

Banibrata Poddar, always inspired spark in me.

structural remaining life through appropriate diagnosis and prognosis technologies.

technologies have demonstrated the feasibility of detecting damage in simple structures

quantifying damage, especially for damage quantification in complex structures such as

laminated composites and honeycomb sandwich structures. Moreover, guided wave

such as crack, delamination and debonding damage, need to be investigated.

The objective of this dissertation research is to develop guided wave based

damage detection and health monitoring applications.

development of integrated SHM/NDE damage diagnosis methodologies. A non-contact

wealth of information regarding guided wave propagations in structures and wave

interactions with structural discontinuities. To extract informative wave signatures from

phenomenon, guided wavefield analysis methods, including frequency-wavenumber

analysis, wavefield decomposition and space-frequency-wavenumber analysis, are

and analyzed in time, space, frequency, and wavenumber multi-domains.

In Part I, a hierarchical damage diagnosis methodology is also developed for

quantitative and visualized damage detection. The hierarchical methodology

evaluation in isotropic plates, (4) delamination damage detection and quantification in

composite laminates, (5) debonding detection and quantification in honeycomb sandwich

based on guided waves, and to providing on-demand health information at component or

subsystem level for the safety and reliability of the structure.

LIST OF TABLES ....................................................................................................................x

LIST OF FIGURES ................................................................................................................. xi

CHAPTER 1 INTRODUCTION ...............................................................................................1

1.1. BACKGROUND .....................................................................................................1

1.3. OBJECTIVES ........................................................................................................8

CHAPTER 2 STATE OF THE ART ........................................................................................10

2.1. GUIDED WAVE PROPAGATION STUDY ................................................................10

2.2. GUIDED WAVE EXCITATION AND SENSING .........................................................18

2.3. GUIDED WAVEFIELD DATA ANALYSIS AND DAMAGE DIAGNOSIS

PART I: INTEGRATED SHM/NDE DIAGNOSIS METHODOLOGIES .......................................38

CHAPTER 3 ANALYTICAL AND NUMERICAL STUDIES OF GUIDED WAVES .......................38

3.1. DISPERSION AND MODE SHAPE DERIVATIONS ....................................................38

3.2. WAVE PROPAGATION AND INTERACTION SIMULATIONS ....................................44

CHAPTER 4 WAVEFIELD ACQUISITION AND INTRINSIC CHARACTERIZATION ...................53

4.1. HYBRID PZT-SLDV SENSING SYSTEM FOR WAVEFIELD ACQUISITION ..............53

4.2. FREQUENCY-WAVENUMBER ANALYSIS .............................................................57

4.4. SPACE-FREQUENCY-WAVENUMBER ANALYSIS ..................................................61

CHAPTER 5 QUANTITATIVE AND VISUALIZED DAMAGE DIAGNOSIS ................................64

5.1. GUIDED WAVE PHASED ARRAYS IMAGING .........................................................64

5.2. FILTER RECONSTRUCTION IMAGING ..................................................................85