modeling of delamination in carbon/epoxy composite...
TRANSCRIPT
MODELING OF DELAMINATION IN CARBON/EPOXY COMPOSITE LAMINATES
UNDER FOUR POINT BENDING FOR DAMAGE DETECTION AND SENSOR
PLACEMENT OPTIMIZATION
by
STEPHEN ABOAGYE ADU
SAMIT ROY, COMMITTEE CHAIR
MARK BARKEY
EDWARD SAZONOV
A THESIS
Submitted in partial fulfillment of the requirements
for the degree of Master of Science
in the Department of Aerospace Engineering and Mechanics
in the Graduate School of
The University of Alabama
TUSCALOOSA, ALABAMA
2013
Copyright Stephen Aboagye Adu 2013
ALL RIGHTS RESERVED
ii
ABSTRACT
Laminated carbon fiber-reinforced polymer composites (CFRPs) possess very high
specific strength and stiffness and this has accounted for their wide use in structural applications,
most especially in the aerospace industry, where the trade-off between weight and strength is
critical. Even though they possess much larger strength ratio as compared to metals like
aluminum and lithium, damage in the metals mentioned is rather localized. However, CFRPs
generate complex damage zones at stress concentration, with damage progression in the form of
matrix cracking, delamination and fiber fracture or fiber/matrix de-bonding.
This thesis is aimed at performing; stiffness degradation analysis on composite coupons,
containing embedded delamination using the Four-Point Bend Test. The Lamb wave-based
approach as a structural health monitoring (SHM) technique is used for damage detection in the
composite coupons.
Tests were carried-out on unidirectional composite coupons, obtained from panels
manufactured with pre-existing defect in the form of embedded delamination in a laminate of
stacking sequence [06/904/06]T. Composite coupons were obtained from panels, fabricated using
vacuum assisted resin transfer molding (VARTM), a liquid composite molding (LCM) process.
The discontinuity in the laminate structure due to the de-bonding of the middle plies caused by
the insertion of a 0.3 mm thick wax, in-between the middle four (4) ninety degree (90o) plies, is
detected using lamb waves generated by surface mounted piezoelectric (PZT) actuators. From
iii
the surface mounted piezoelectric sensors, response for both undamaged (coupon with no defect)
and damaged (delaminated coupon) is obtained.
A numerical study of the embedded crack propagation in the composite coupon under
four-point and three-point bending was carried out using FEM. Model validation was then
carried out comparing the numerical results with the experimental. Here, surface-to-surface
contact property was used to model the composite coupon under simply supported boundary
conditions. Theoretically calculated bending stiffness’s and maximum deflection were compared
with that of the experimental case and the numerical. After the FEA model was properly
benchmarked with test data and exact solution, data obtained from the FEM model were used for
sensor placement optimization.
iv
DEDICATION
This thesis is dedicated to everyone who prayed, assisted and guided me through the
trials of creating this manuscript. In particular, my family and close friends who stood by me
throughout the time taken to complete this thesis work.
v
LIST OF ABBREVIATIONS
ASTM American society for testing and materials
CFRP Carbon fiber reinforced polymer
CLT Classical lamination theory
CZM Cohesive zone method
EPFM Elastic-plastic fracture mechanics
FEA Finite element analysis
FEM Finite element methods
IM7 Intermediate modulus 7
LEFM Linear elastic fracture mechanics
MTS Mechanical testing services
PMC Polymer matrix composites
PZT Lead zirconate titanate
RTM Resin transfer molding
SHM Structural health monitoring
SNOBFIT Stable noisy optimization by branch and fit
SPO Sensor placement optimization
VARTM Vacuum assisted resin transfer molding
VCCT Virtual crack closure technique
vi
ACKNOWLEDGEMENTS
My sincerest gratitude first and foremost, goes to God Almighty for bringing me this far
in life, to my committee members without who’s guidance this manuscript could not have been
made and then to my family and friends.
I am most indebted to my advisor Dr. Samit Roy, for giving me the opportunity to be part
of the NASA Structural Health Monitoring (SHM) team and for his excellent technical guidance,
encouragement, and academic advice. I would also like to acknowledge my thesis committee
members, Dr. Mark Barkey and Dr. Edward Sazonov for their time, valuable advice and
feedback. Their unsurpassed knowledge in their respective fields made it possible for this thesis
to be completed.
I also appreciate the knowledge and time investment made by Dr. Susan Burkett and Dr.
Sushma Kotru. Their lab and technical guidance played a major role with regard to the structural
health monitoring section of this thesis. The friendly and cheerful nature of the entire SHM
group; Nicholas Hayes, Donald Tyler Paul, Ben Fairbee, Kanchan Trivedi, Michelle Spiegel,
Akshata Nagabhushana, Muhammed Farooq and Hao Zheng made this research work possible. I
am very grateful for the group’s assistance and guidance at the initial stages of the composite
fabrication process, experimental and numerical testing. Special thanks to Priyank Upadhyaya
for his valuable input in the numerical modeling.
vii
This thesis would not have been possible without the support from the United States
National Aeronautics and Space Administration, grant number: NNX10AJ14G; Technical
monitor, Dr. Curtis E. Banks, NASA MSFC. Their financial support is very much appreciated.
I would like to acknowledge the financial, academic and technical support of The
University of Alabama, Tuscaloosa and its staff, particularly in the award of a Graduate
Teaching Assistantship and giving me an opportunity to realize my dream.
My final thanks go to my family, especially my mum for her prayers and support.
viii
CONTENTS
ABSTRACT ................................................................................................ ii
DEDICATION ........................................................................................... iv
LIST OF ABBREVIATIONS ......................................................................v
ACKNOWLEDGEMENTS ....................................................................... vi
LIST OF TABLES .......................................................................................x
LIST OF FIGURES ................................................................................... xi
CHAPTER 1: INTRODUCTION ................................................................1
CHAPTER 2: LITERATURE REVIEW .....................................................5
2.1 Introduction ............................................................................................5
2.2 Composite materials...............................................................................5
2.3 Damaged in laminated composites ........................................................8
2.4 Modeling damage in laminated composites .........................................15
2.5 Four-point bend test .............................................................................16
2.6 Modeling techniques used in delamination..........................................18
CHAPTER 3: FABRICATION OF COMPOSITE LAMINATE ..............22
CHAPTER 4: STRUCTURAL HEALTH MONITORING ......................27
4.1 Damage detection.................................................................................28
4.2 Lamb wave methods ............................................................................29
4.3 Sensing methods ..................................................................................29
ix
4.4 Piezoelectric transducers and principle of operation ...........................30
4.5 Actuation parameters ...........................................................................30
4.6 Delamination detection ........................................................................31
4.7 Test for damage detection ....................................................................34
4.8 Experimental analysis of damage evolution in delaminated coupon ...36
CHAPTER 5: FINITE ELEMENT ANALYSIS .......................................41
CHAPTER 6: FINITE ELEMENT MODEL FOR OPTIMIZATION ......50
6.1 Interfaces between MATLAB and Abaqus ..........................................50
6.2 Embedded delamination test ................................................................51
CHAPTER 7: CONCLUSIONS AND FUTURE WORK .........................55
7.1 Conclusion ...........................................................................................55
7.2 Suggestions for future work .................................................................57
REFERENCES ..........................................................................................59
APPENDIX A: LabVIEW program...........................................................66
x
LIST OF TABLES
Table 4.1: Comparison of Slope (Structural stiffness) of the Load-Deflection data obtained
from the Four-Point Bend Test of tested specimens of dimension (10 in. x 1 in. x 0.14 in).........38
Table 4.2 Comparison of Slope (Structural stiffness) of the Load-Deflection data obtained
from the Four-Point Bend Test of tested specimens of dimension (254mm x25.4mm x3.6mm)
(Presented in SI units)…………………………............................................................................38
Table 4.3 Comparison of the peak load and peak stress of the tested specimens..........................39
Table 4.4 Comparison of the peak load and peak stress of the tested specimens (Presented in
SI units)..........................................................................................................................................39
Table 5.1 Material properties of IM7/PET15 composite used for numerical simulation
(Presented in SI units)....................................................................................................................41
Table 5.2: Comparison of the maximum deflections, for damaged and undamaged specimens
of stacking sequence [06/904/06]T under three-point and four-point load conditions..................49
xi
LIST OF FIGURES
Figure 2.1 Schematic of modes of failure in CFRP laminates [5].................................................14
Figure 2.2 Schematic of laminate configuration (16 plies) with wax insert in-between the
middle ply......................................................................................................................................14
Figure 2.3 Schematic of a beam cross-section under four-point bending.....................................17
Figure 3.1 Schematic of the VARTM Process [65].......................................................................22
Figure 3.2 Photograph showing wax on resin inlet side and cladded aluminum sheet on the
vacuum outlet side of the laid carbon fiber plies...........................................................................24
Figure 3.3 VARTM materials stacking sequence (release fabric, carbon fibers, release fabric,
distribution mesh) and taping process............................................................................................25
Figure 3.4, (a) Vacuum Pulled on system and (b) Vacuum Gauge on the Resin Catch reading
about 95kPa....................................................................................................................................25
Figure 3.5 Photograph of setup being de-bulked, showing inlet-outlet lines, resin catch and
vacuum pump.................................................................................................................................26
Figure 3.6 Photograph of VARTM manufactured laminate [06/904/06]T with embedded
delamination...................................................................................................................................26
Figure 4.1 A typical PZT excitation input signal for the testing scheme used in this study.........31
Figure 4.2 Data acquisition system using PZT actuators and sensors...........................................33
Figure 4.3 Pictorial representation of composite coupon with delamination (wax insert)
demonstrating the actuator and sensor placement relative to delamination..................................35
Figure 4.4 Experimental results comparing the Lamb wave response of a [06/904/06]T
orientation undamaged coupon and a coupon with embedded delamination in the middle
four plies........................................................................................................................................35
xii
Figure 4.5 Photographs of the Four-Point Bend Fixtures mounted in the MTS QTest 25
machine..........................................................................................................................................36
Figure 4.6 Photographs showing damaged specimen, (a) under load; (b) Delamination
Extension in specimen; (c) and (d) Transition from Matrix Crack, Delamination to Fiber
Fracture..........................................................................................................................................37
Figure 4.7 Crack initiation and evolution in the undamaged coupon............................................37
Figure 4.8 Schematic of the support and load spans of the Four-Point Bend Fixtures.................38
Figure 4.9 Load (F)-Deflection plots for Damaged and Undamaged specimen under Four-Point
Bending .........................................................................................................................................40
Figure 5.1 Illustration of damaged coupon with partial delamination in-between two
sub-laminates (1) and (2), with stacking sequence of [06/902]T and [902/06]T respectively...........43
Figure 5.2 FE model of a composite coupon with stacking sequence of [06/904/06]T
Representing the loading and boundary conditions for Four-Point Bending................................43
Figure 5.3 Abaqus results showing the deflection contour plot and legend for undamaged
composite coupon undergoing Four-Point Bending in millimeters...............................................44
Figure 5.4 (a) and (b), FE model showing top and bottom sub-laminates of stacking sequence
[06/902]T and [902/06]T respectively.............................................................................................45
Figure 5.5 (a) and (b), FE model showing constraint and interaction property for damaged
composite coupon..........................................................................................................................45
Figure 5.6 Abaqus results showing the deflection contour plot and legend for damaged
composite coupon undergoing Four-Point Bending in millimeters...............................................46
Figure 5.7 FE model of undamaged composite coupon with stacking sequence of
[06/904/06]T representing the load and boundary conditions for Three-Point Bending.............47
Figure 5.8 Abaqus results showing the deflection contour plot and legend for undamaged
composite coupon undergoing Three-Point Bending in millimeters.............................................47
Figure 5.9 (a) and (b), FE model showing constraint and interaction property for damaged
composite coupon..........................................................................................................................48
Figure 5.10 Abaqus results showing the deflection contour plot and legend for damaged
composite coupon undergoing Three-Point Bending in millimeters.............................................48
xiii
Figure 6.1 A flow chart demonstrating MATLAB, Abaqus interface used for generating data
set for sensor placement optimization (where count is the iteration number)...............................51
Figure 6.2 A FE representation of longitudinal strain component contour plot observed on the
top surface of a composite coupon with embedded delamination (in the middle) and subjected
to four point bending......................................................................................................................52
Figure 6.3 A Flow chart demonstrating the sensor placement optimization process....................53
Figure 6.4 Displacement surface plot comparison between damaged and undamaged coupon....54
Figure 6.5 Initial and final placement of sensors after completion of optimization algorithm
(dots represent sensors)..................................................................................................................54
1
CHAPTER 1
INTRODUCTION
Due to their low density, high specific strength and modulus to weight ratios that are markedly
superior to those of metallic materials, carbon fiber reinforced materials are being substituted for
metals in many weight critical components in aerospace, automotive and offshore structures.
However, their use in aerospace structures is still dependent on many difficulties in prediction of
damage tolerance and this has led to over-conservative designs, not fully realizing the promised
economic performance benefits. Delamination, or interfacial cracking between composite layers,
is undoubtedly one of the more common and critical types of damage in laminated fiber-
reinforced composites due to their relatively weak inter-laminar strengths. It is caused by out-of-
plane shear and normal stresses, which may arise under various circumstances such as, impact
loads, free edge effects, ply drop offs, or transverse loading and may be divided into
delamination onset and delamination growth in a laminated composite structure.
For the structural integrity of composite materials to be maintained, delamination as a
failure mode needs to be well understood, since it is difficult to detect under inspection due the
fact that delamination usually occur between plies internal to the laminate.
Tragic incidents that have occurred over the years in structures built with composite
materials have also prompted a significant amount of research in the area of damage detection in
composite materials using an array of actuator-sensor combinations placed at strategic locations.
Structural health monitoring (SHM) therefore has become a vital tool to help engineers improve
2
safety and maintainability of critical structures. Here, a variety of sensing technologies with an
embedded controller to capture, log, and analyze real-time data are designed to reliably monitor
and test the health and performance of structures as seen in the case of the Boeing 787 aircraft
health monitoring plan [1].
SHM implementation over the years have helped save unnecessary maintenance
downtime, money and reduced the chances of catastrophic failures. Evolving from non-
destructive evaluation (NDE) techniques in which sensors and actuators are external to a
structure and the sensing probes can be adjusted and manipulated to extract the maximum signal,
tests are conducted at scheduled intervals. Unlike NDE techniques, examples of which include,
laser ultrasound, laser shearography, infrared thermography and X-ray radiography, in SHM, the
locations of sensors and actuators are fixed and remain on the structure continuously and are
therefore not completely unobtrusive. This integrated sensing scheme allows for continuous
monitoring and provides important information on the damage evolution; the structure does not
have to be taken out of service during structural health testing.
The SHM process involves implementation evaluation, data acquisition and damage
feature extraction, and classification. Implementation evaluation includes studying the economic
and human life safety advantages, possible damage types, operational and environmental
conditions, cost and maintenance needs and the time requirement for deploying the SHM system.
The data acquisition portion of the process involved a decision about the number and type of
sensors, placement of sensors, type of actuation, acquisition hardware, and signal transmission
hardware and data storage. A signal processing step is typically performed after data acquisition
to cleanse and remove any noise from the data. In feature extraction, damage sensitive features
such as the natural frequency of vibration, strain, time of flight, displacement, and modal shape,
3
are separated out while also compensating for any environmental effects. As a final step, a
statistical analysis of acquired data leads to a classification of the structure as damaged or
undamaged [2].
The four-point bend test has been used over the years to carry out various tests, ranging
from determination of flexural properties and bending stiffness’s of all kinds of materials, to the
investigation and evaluation of the different modes of failure in engineering materials. Unlike the
three point bend test, the four-point bending method allows for uniform load distribution
between the two loading fixtures. Following the testing conventions specified in the American
Society for Testing and Materials (ASTM D6272-00), transverse vertical loads are applied to
horizontal beams such that, a constant bending moment results between the two inner load
locations. With the embedded delamination located mid-way along the length of the coupon and
across the width of the specimen, the four-point bend test serve as an appropriate engineering
procedure for the analysis of stiffness degradation in the composite coupon. Here the specimen
lies on a span and stress is uniformly distributed between the loading fixtures. The horizontal
distance between the two support fixtures and the vertical depth which a specimen can be
displaced are dependent on material type and thickness. The load span to support can also be a
ratio of 1 to 2, or 1 to 3.
Historical data for the experiment was obtained for analyses from data acquisition
software and hardware used with the machine.
In this thesis, a Lamb wave-based approach for damage detection in composite coupon
with embedded delamination is described. A surface mounted piezoelectric actuator was used to
generate Lamb waves in a composite coupon and the response was measured using a surface
mounted piezoelectric sensor. This aspect of the research involves optimization of testing
4
parameters, sensor location and the number of sensors. Stiffness degradation under four-point
bend test was also investigated using a universal testing machine (MTS QTest/25) with a pre-
written program (Testworks QT 2.03), and data acquisition software.
The thesis is divided into seven chapters. Chapter 2 is focused on reviewing the available
research in the areas relevant to this project. An introduction to composite materials and a
discussion of damage in laminated composites is provided, followed by a review of analysis
methods employed to study failure, in particular delamination, in laminates. The four-point bend
test (FPBT), as a failure analysis tool in composite materials is also discussed. Some modeling
techniques used in the analysis of delamination is also presented in this chapter. The processing
of composite laminates, with and without delamination is presented in Chapter 3.
A brief background on SHM using different sensors is presented in chapter 4. The damage
detection system used to detect embedded delamination in a composite coupon is also detailed in
this chapter. Mechanical property testing and delamination propagation, in CFRP coupons under
four-point bending using the MTS machine is presented, and corresponding experimental results
are analyzed for both damaged and undamaged coupons. Chapter 5 describes the finite element
simulation of delamination in composite coupons, using the commercially available finite
element code Abaqus. Chapter 6 has details on the sensor optimization process. Here the time
domain information along with finite element analysis was used to characterize the damage. A
brief description of interfacing Abaqus and optimization algorithms that can be used for sensor
placement optimization is provided. Chapter 7, the final chapter, focuses on presenting the major
conclusions deduced from the work carried out in the project. Some suggestions for future
research are presented in this chapter.
5
CHAPTER 2
LITERATURE REVIEW
2.1 Introduction
The main objective of this thesis work is damage detection, testing and sensor optimization
analysis in unidirectional intermediate modulus (IM7/epoxy) CFRP laminate coupon with an
embedded delamination using Lamb waves and the study of stiffness degradation in CFRP
laminate coupon subjected to four-point bending. However, before addressing this problem, it is
necessary to provide an introduction to the laminated composite materials and their failure
mechanisms with focus on delamination.
An introduction to composite materials and a discussion of damage in laminated
composites is first introduced in this chapter, followed by a review of the analysis methods
employed to study their failure, in particular delamination. The four-point bend test as a failure
analysis tool in laminated composites is discussed and finally modeling techniques used in the
analysis of delamination and then damage propagation in CFRP is explained.
2.2 Composite materials
The rise in demand of structural materials, with high specific strength, high specific
stiffness, low density, corrosion resistance, wear resistance, fatigue resistance and design
flexibility has led to the development of composite materials. Since it is difficult to achieve all
these desired properties within a single material, it has become necessary, to combine two or
6
more materials on a macroscopic scale to form a useful third that would exhibit the best qualities
of the individual components and hence the desired properties for structural applications.
Composites therefore, are types of materials that consist of more than one constituent, where
constituent materials are proportionally distributed and geometry is controlled to obtain optimum
material properties [3]. It normally consists of a matrix and reinforcement. The matrix is used to:
transfer loads into and out of fiber
remove abrasion between fibers
transfer internal loads between fibers in short-fiber composites
provide a barrier against adverse environment
effect in-plane shear and inter-laminar shear properties of composites.
The reinforcement which may be fibrous or particulate provides the composite with high
stiffness and strength properties [3]. The material used in this thesis work consists of SC-780
epoxy matrix, reinforced with plies of unidirectional IM7 carbon fiber with the total stacking
sequence of [06/904/06]T.
2.2.1 Classification of composite materials
Composite materials are classified into two categories, fiber-reinforced composites and
particle-reinforced composites. They are further re-classified based on the kind of matrix
material used in their formation and this leads to four different categories of composites namely;
polymer matrix composites (PMCs), metal-matrix composites (MMCs), ceramic-matrix
composites (CMCs), and carbon-carbon composites (CCCs) [3].
Over the years, polymer matrix composites have become dominant, due their good in-
plane stiffness and strength, low density, ease of fabricability, relatively low cost, corrosion
resistance, fatigue resistance, low coefficient of thermal expansion, creep and creep fracture
7
resistance, relatively mature technology and excellent in-service experience compared to MMCs
and CMCs [3]. However, CMCs have special applications in the space industry where they serve
as thermal protection materials in the nose cap and the leading edges of the wing of space
shuttles as well as being applied in rocket nozzles and gas turbines. MMCs have found limited
use due to problems in controlling the chemical reaction between the fibers and the molten
metals at high processing temperatures required for such composites [3].
Polymer matrix composites (PMCs) consist of polymer (resin) matrix combined with
dispersed fibrous reinforcements and are very popular due to their low cost and simple
fabrication methods [4]. According to the reinforcement material, PMCs are grouped as;
Glass fiber reinforced polymers
Carbon fiber reinforced polymer composites
Kevlar (aramid) fiber reinforced polymers
They are used for manufacturing applications such as load-bearing aerospace structures, boat
bodies, automotive parts, radio controlled vehicles and bullet-proof vests and their properties are
determined by;
Properties of the fibers
Orientation of the fibers
Concentration of the fibers
Properties of the matrix
They are typically anisotropic materials and have low stiffness in the transverse fiber direction
where their properties are usually matrix-dominated. For example, the tensile strength and
modulus of a unidirectional oriented fiber-reinforced polymer are maximum when these
properties are measured in the longitudinal direction of fibers, but at any other angle, these
8
properties are lower [3]. Lower mechanical properties of fiber-reinforced polymer composites in
directions other than the fiber direction and their generally anisotropic nature goes a long way to
have an effect on all aspects of their design and analysis, and failure modes, making isotropic
analytical methods inapplicable for their design analysis [5].
The design of a fiber-reinforced composite structure is considerably more difficult than
that of a metal structure, principally due to the difference in its properties in different directions.
However, its non-isotropic nature creates a unique opportunity of tailoring its properties
according to the design requirement and this coupled with their high specific strength and
stiffness have attracted many researchers over the years [3]. Compared with monolithic
materials, the damage initiation and propagation behavior of fiber-enforced composites laminates
is a distinct, complicated and a progressive phenomenon. This usually starts with matrix
cracking, which evolve to delamination between the layers and then to fiber breakage. This
adversely degrades the mechanical properties of the structures. It is therefore critical that the
nature and effect of damage in these kinds of materials is investigated for better and safer
structural applications.
2.3 Damage in laminated composites
CFRPs generate complex damage zones at stress concentration, with damage progression
in the form of, matrix cracking, delamination and fiber fracture or fiber/matrix de-bonding.
Research on the failure mechanism of unidirectional composites started in close relation to the
utilization of such composite materials in aerospace industries. Countless efforts have been made
in this area and substantial breakthroughs have been made with constant improvements of
9
existing theories still on-going, to lead the way towards a better understanding of the essence
behind the various failure behaviors of lamina/laminate [6].
Since fiber reinforced composites are comprised of two materials with distinct
mechanical and thermal properties, if geometric inhomogeneity is taken into account, fiber and
matrix play different roles under different loading conditions. Fibers are longitudinally
continuous and contribute more to the development of longitudinal modulus and strength,
indicating that the fiber supports almost the entire longitudinal tensile load applied to a ply and
compressive load without consideration of fiber buckling [6]. The matrix on the other hand,
binds fibers together and contributes more to the development of transverse strength and
modulus [3].
With the longitudinal modulus of laminated composites being always greater than the
transverse modulus, minimizing the transverse stresses of laminated composite materials
therefore is a critical design consideration. Load transfer between adjacent layers in a fiber
reinforced laminate takes place by means of inter-laminar stresses which develop due to
mismatch in Poisson’s ratios and coefficients of mutual influence between various layers [3].
Generally, tensile normal stresses applied to an isotropic material cause’s elongation in
the direction of the applied stresses and contraction in the two transverse directions, a more
complex response however, is experienced in laminated composites where a combination of
extensional and shear deformation may be produced. To minimize extension-shear coupling
therefore laminates are designed to be balanced and symmetric about the mid-plane with
examples being unidirectional and cross-ply laminates. In this thesis, a carbon fiber reinforced
symmetric composite laminate was manufactured using 16 plies of unidirectional IM7 carbon
fibers with damage, in the form of delamination, embedded during manufacturing of the
10
laminate. Damage mechanisms experienced by such laminates and some previous research
finding in this area are discussed in the next section.
2.3.1 Damage mechanisms in laminated composites
The mechanical separation of engineering materials is normally categorized as ductile or
brittle fractures. While ductile fractures absorb more energy, brittle fractures absorb little energy
and are generally characterized by fracture with smooth surfaces [7]. Laminated polymer matrix
composites however, often exhibit a mixture of ductile and brittle failure processes such as
delamination or inter-laminar fracture, matrix-fiber debonding and fiber breaking [8]. Various
models have been proposed in an effort to simulate in-plane and out-of-plane failure mechanisms
and describe different load distributions associated with damage accumulation [9-12]. In-plane
failure is mostly due to material property degradation and theories such as the classical
lamination theory have been used to determine in-plane stress state and the stresses/strains that
can be used to determine the failure criteria for the various failure mechanisms at locations of
interest in a material. The laminated composite material can be accepted for use in structural
application when its actual strength envelope is in agreement with what is predicted by the
failure criterion.
When external and thermal loadings are applied to carbon fiber reinforced composite
laminates, non-uniform micro-stresses at the constituent level develop due to material and
geometric inhomogeneity arising from the inclusion of fibers. Any point within the composite
belongs to one of three regions namely, the fiber, the matrix, or the fiber-matrix interface. This
leads to the initiation of ply failure which can have dissimilar failure mechanisms based on the
location of critical points [6]. Failure in the form of transverse matrix cracking normally occur in
90o
plies, delamination typically occur between 0o and 90
o plies, longitudinal matrix cracking
11
takes place in the fiber direction of 0o plies and fiber fracture in the 0
o plies. All the modes of
failure mentioned above are in-plane with the exception of delamination, which is an out-of-
plane damage mode which evolves with inter-laminar stresses [13].
Matrix cracking as an intra-laminar laminated composite failure mode, can adversely
affect the integrity of the composites by initiating other damage developments such as
delamination and fiber fracture, which leads to stiffness reduction of the composite laminate
[14]. For structural integrity of laminated composites to be maintained therefore, insight into the
matrix cracking phenomenon in composite laminates is required. Over the years, a lot of research
has gone in search of better understanding, and prediction of transverse cracking behavior and its
effects on the modulus reduction of laminated composites [11, 15-24]. Shear-lag analysis was
used by Highsmith and Reifsnider [11] for predicting modulus reduction of cross-ply laminates
due to matrix cracking. A modified version of this was adopted by Lim and Hong [15] where the
concept of inter-laminar shear layer and an energy criterion was used to predict the progressive
matrix cracking and stiffness reduction in cross-ply laminates. Smith and Ogin [22] also
presented a one-dimensional analysis of a cross-ply laminate with embedded transverse crack,
under flexural load, where bending theory was used in conjunction with shear-lag analysis to
calculate the degradation of longitudinal modulus of a cracked transverse ply from which the
laminate flexural modulus is to be determined.
Fiber fracture, the mechanical separation of fibers in fiber reinforced composite laminates
owing to the application of stress can be categorized as ductile or brittle fractures. Brittle
fractures absorb little energy and are mostly characterized by fracture with flat surfaces, while
ductile fractures absorb more energy [25]. It represents the ultimate mode of failure of the
composite since most of the load carrying capabilities of the laminate are achieved by the fibers.
12
A fracture in the fibers therefore leads to a severe decrease in modulus of the fibers which is
detectable by conventional failure interrogation. It is usually preceded by the occurrence of other
modes of failure such as matrix cracking and delamination. To predict fiber fracture as the
ultimate failure mode of the fiber therefore, models representing the progressive development of
all modes of damage, from transverse matrix cracking and delamination as well as interactions
between the two need to be modeled. The type of materials, the orientation of the laminate plies,
geometry and type of loading play a major part in the complexity of the interaction between the
different damage modes.
High inter-laminar stresses accounts for delamination, a failure mechanism uniquely
characteristic of composite laminates. The existence of these inter-laminar stresses leads to
delamination around free edges of composite laminates at locations around the edges, around a
hole, or at the ends of tubular configuration. Delamination therefore causes a reduction in
stiffness and strength, which leads to premature failure in laminated composites and therefore
needs to be critically considered in specimen design [26].
The influence of delamination on the structural behavior of fiber-reinforced composite
laminates is well known and has been studied extensively in the literature. This kind of damage
is usually embedded within a composite structure, which makes it invisible to the naked eye and
hence difficult to detect under inspection. Due to the complexity of this kind of damage, a lot of
research still needs to be carried out to overcome the issue of overly conservative design in the
application of fiber reinforced composite laminates. A review of relevant literature on work
carried out by a number of researchers in the area of delamination analysis of CFRP is discussed
in the following paragraph.
13
Wang et al. [27] considered two types of delamination, namely, delamination initiating
from the intersection of the transverse cracks and the free edges of the laminate, and
delamination initiating from the intersection of transverse cracks and longitudinal splits in the 0o
layers. Here three dimensional finite element analyses was carried out on CFRP cross-ply
laminates to estimate the energy release during delamination growth which led to results which
agreed qualitatively with the experimental. Analytical methods with a capability of predicting the
stress transfer between 0o and 90
o plies in a [0/90/0] cross-ply laminate containing transverse
cracks were developed by McCartney [24]. Wimmer et al. [28] used a combination of a first ply
failure criterion and a fracture mechanics approach where FEM was used to simulate
delamination in a structure without initial delamination and flaws. To predict the delamination
onset, the failure criterion developed by Puck was used, and starting delamination was created
based on the results obtained from the Puck criterion. It was built on a physically based,
phenomenological model which is suitable for long fiber reinforced polymers. Here difference
between fiber failure and the other several modes of matrix dominated failure was established,
with delamination being one of them. Delamination strengths were estimated by slightly
reducing the transverse ply strengths. Stresses in each Gauss point was evaluated together with
the spatial distribution of the risk of failure and failure mode. Virtual growth of the starting
delamination was simulated using virtual crack closure technique (VCCT), where delamination
growth was modeled based on the Griffith crack growth criterion assuming linear elastic fracture
mechanics (LEFM) which can be applied only when a starting crack exists. LEFM deals with the
occurrence of brittle fracture due to crack growth in the absence of noticeable plastic
deformation at the crack tip [29]. Results obtained were in agreement with simulation using
cohesive zone method (CZM), where cohesive element is used to describe the cohesive forces,
14
which occur when material elements are being pulled apart. In CZM, the cohesive elements open
to simulate crack initiation and growth. The procedure allowed for the determination of the
critical size and location of a starting delamination and provided information about the
delamination growth stability. Figure 2.1 below is a schematic of a cross-ply laminate showing
the various failure modes in a composite laminate.
Figure 2.1 Schematic of modes of failure in CFRP laminates [5]
In this thesis however, an embedded delamination was artificially inserted at the mid-plane
between the middle four 90o plies of a composite laminate of stacking sequence [06/904/06]T
made up of unidirectional IM7-SC 780 carbon/epoxy composite laminate using a 0.3 mm thick
wax insert. Delaminated and undamaged coupons of dimension (254 mm x 3.5 mm) cut-out from
the same manufactured panel were analyzed. Here the four-point and three-point bend tests were
employed to investigate stiffness degradation of composite coupons containing embedded
delamination. Figure 2.2 below is a schematic of the laminate cross-sectional configuration with
0.3 mm thick wax insert.
Figure 2.2 Schematic of laminate configuration (16 plies) with Wax insert in-between the middle
Ply
15
2.4 Modeling damage in laminated composites
This section reviews some of the methods that have been employed over the years in the
design and analysis of damage in laminated composites.
2.4.1 Fracture mechanics
Failures normally occur for many reasons, including defects in materials, inadequacies in
design, uncertainties in loading or environment, and deficiencies in construction or maintenance.
Design against fracture has a technology of its own, and this has become very active in areas of
current research.
Fracture mechanics therefore, deals with mechanics of solids containing planes of
displacement discontinuities (cracks), with special attention to their growth. The nature of the
manufacturing process of laminated polymer composites provides innumerable sites for the
initiation of all kinds of defects in their interior and surfaces. Not all these defects however
become unstable or grow under service conditions. To determine if a flaw is stable or will
propagate as a crack and cause failure requires the use of fracture mechanics, which is divided
into two sections as discussed next.
2.4.1.1 Linear elastic fracture mechanics
Linear elastic fracture mechanics (LEFM) is the basic theory of fracture that deals with
sharp cracks in linearly elastic bodies. It is applicable to any material as long as the material is
elastic except in a vanishingly small region at the crack tip (assumption of small scale yielding),
brittle or quasi-brittle fracture, and stable or unstable crack growth. Griffith began his pioneering
studies of fracture in the years just prior to 1920 where he considered how it might be used in
developing a fundamental approach in predicting fracture strengths. He employed an energy-
balance approach to investigate fracture in very brittle materials, specifically glass rods [25]. It
16
was however realized that, when the materials exhibits more ductility, consideration of the
surface energy alone failed to provide an accurate model for fracture. This deficiency was later
remedied independently, by Irwin and Orowan [30]. They suggested that a majority of the
released strain energy in ductile materials was absorbed not by creating new surfaces, but by
energy dissipation due to plastic flow in the material near the crack tip. They also suggested that
catastrophic fracture occurred when the strain energy was released at a rate sufficient to satisfy
the needs of all these energy “sinks,” and denoted this critical strain energy release rate by the
parameter Gc. Therefore LEFM generally gives an accurate criterion for the establishment of
catastrophic failure in brittle materials but runs into a limitation for brittle materials having blunt
notches, but no cracks, since a finite initial crack size is required for its application. Another
limitation for ductile materials occurs, where the size of the non-linear zone due to plasticity is
not negligible. To overcome these limitations therefore required non-linear fracture mechanics
and is discussed next.
2.4.1.2 Elastic-plastic fracture mechanics
In elastic-plastic fracture mechanics (EPFM), the crack tip undergoes significant plasticity.
Usually applicable in engineering materials that exhibit some nonlinear elastic and inelastic
behavior under operating conditions that involves large loads. It applies to elastic-rate-
independent plastic materials generally in the range of large-scale plastic deformation.
2.5 Four-point bend test
The four-point bend test together with three-point bend test, remain the two most used
methods for the determination of flexural properties of laminates. Flexural tests are usually used
to determine the mechanical properties of resins and laminated fiber composite materials owing
17
to their simplicity in terms of procedure, instrumentation and equipment requirement [31]. In the
four-point bend method usually a flat rectangular beam specimen is simply supported close to its
ends and with two loads placed symmetrically between the supports, giving four-point bending
load condition as shown in Figure 2.3 below. Here the bending moments increase linearly from
zero at the supports to a maximum at the loading points and are constant between these points, as
shown in Figure 2.3. The shear force and inter-laminar shear stress are zero between the loading
points, so that, that portion of the beam specimen is subjected to a pure bending moment. The
American Society for Testing and Materials (ASTM) specifications allows a series of different
span-to-thickness (S/h) ratios (16:1, 32:1, 40:1 and 60:1) in four-point bending [31], and also
specifies the appropriate specimen width, length, support spans for four-point bending and rate
of crosshead motion for the full range of specimen thicknesses and span-to-thickness ratios
allowed.
Figure 2.3 Schematic of a beam cross-section under four-point bending
Over the years, fracture characterization of all kinds of bonded joints under the opening
(Mode I dominant fracture) has been extensively studied and documented. The four-point bend
test has also been used to investigate the flexural modulus of various engineering materials as
well as different modes of delamination and mixed-mode fractures. More recently, Mode II
fracture toughness of different end-notched flexural (ENF) specimen has been investigated using
the four-point bend test. Martin and Davidson [31] first proposed the use of four-point bending to
analyze Mode II fracture toughness of end-notched flexure specimen (4-ENF) in (1999).
18
Schuecker and Davidson (2000) [32]; Vinciquerrqa and Davidson 2000 [33]; Yoshihara 2004,
2008 [34]; de Morais and de Moura 2006 [35]; and Thouless 2009 [36], have all studied fracture
characterization of 4-ENF with focus on specimens made of symmetric sub-layers with identical
material. Qiao et al. [37] used it to analyze mixed-mode fracture of hybrid material bonded
interfaces where a combined analytical and experimental approach was used to characterize the
mixed-mode fracture. The closed-form solution of the compliance and energy release rate (ERR)
was also obtained. The solutions predicted by the conventional beam theory (CBT) and finite-
element analysis (FEA) were in strong agreement with the analytical solution for both the
compliance and ERR of the mixed mode fracture specimens. It was established that the crack-tip
deformation played an important role in accurately characterizing the mixed-mode fracture
toughness of hybrid material bonded interfaces under four-point bending load. In this thesis
however, the four-point bending was used to analyze flexural stiffness degradation for a
composite laminate with an embedded delamination.
2.6 Modeling techniques used in delamination analysis
Progressive failure in composite structures have been modeled over the years with
different models ranging from, the fairly simple material property/stiffness degradation methods
(MPDM) to much more sophisticated MPDM based on continuum damage mechanics (CDM)
and fracture mechanics. The idea of MPDM is that the damaged material’s post-initial failure
behavior is modeled by reduced stiffness, an example of which is given by Tan et al. [38–40]
who proposed a two-dimensional (2D) progressive damage model for laminates containing
central holes subjected to in-plane tensile or compressive loading. Continuum damage mechanics
(CDM), a more sophisticated form of the material stiffness degradation scheme based on the use
19
of an internal state variable has also been used to model general damage as well as delamination,
an example is given by Barbero et al. [41]. MPDM schemes are often implemented in FE codes.
In commercial FE software, the implementation is often achieved through user-defined
subroutines such as UMAT for Abaqus [42–46], Ansys [47] or LS-Dyna [48]. Because damage
in composites is three dimensional (3D) in nature with strong and complex interactions between
various damage modes, it is therefore generally modeled with (3D) elements. Delamination
propagation in composite structures, for example is a 3D phenomenon because, delamination
frequently propagates in a non-self-similar fashion and may extend into other plies and propagate
along another ply interface [49]. The presence of micro-cracks often interferes with the direction
and shape of delamination growth which leads to higher computational time and resource
requirement. Analyses based on 2D plate or shell elements [50, 51, 52, 53, 54, 55–57] are still
used in some studies to save computational effort, although Robbins and Reddy [58] recently
showed that first-order shear deformation elements do not give acceptable results and that
higher-order or 3D elements should be used.
2.6.1 Cohesive zone method
Among various damage models, the cohesive-zone model (CZM) is the most commonly
used to investigate interfacial cracks since it handles both delamination onset and growth. The
CZM approach is based on the assumption that, a cohesive zone or softening plasticity develops
near the crack tip and links the microstructural failure mechanisms to the continuum fields
governing bulk deformations. Thus, a CZM is characterized by the properties of the bulk
material, the crack initiation condition, and the crack evolution function. Cohesive damage zone
models relate traction to displacement jumps at an interface where a crack may occur. Damage
initiation is related to the interfacial strength, i.e., the maximum traction on the traction-
20
displacement jump relation. When the area under the traction-displacement jump relation is
equal to the fracture toughness, the traction is reduced to zero, and new crack surfaces are
formed [59-61]. The CZM however, is numerically expensive and requires fine meshes in order
to represent the damage zone adequately.
2.6.2 Finite element method
The simulation of delamination using the finite element method (FEM) is normally
performed using the Virtual Crack Closure Technique (VCCT) [62], or decohesion finite
elements [63-64]. The VCCT is based on the assumption that the energy released during de-
lamination propagation equals the work required to close the crack back to its original position.
Based on this assumption, the single-mode components of the energy release rate are computed
from the nodal forces and nodal relative displacements [62]. Delamination growth is predicted
when a combination of the components of the energy release rate is equal to a critical value.
There are some difficulties when using the VCCT in the simulation of progressive delamination.
The calculation of fracture parameters requires nodal variable and topological information from
the nodes ahead and behind the crack front. Such calculations are tedious to perform and may
require re-meshing for crack propagation. Decohesion finite elements have been used to
overcome some of the above difficulties by being able to predict both the onset and non-self-
similar propagation of delamination. However, the simulation of progressive delamination using
decohesion elements poses numerical difficulties related with the proper definition of the
stiffness of the cohesive layer, the requirement of extremely refined meshes, and the
convergence difficulties associated with problems involving softening constitutive models.
Turon et al, in their paper: (An Engineering Solution for using Coarse Meshes in the Simulation
of Delamination with Cohesive Zone Models) addressed the proper definition of interface
21
stiffness and proposed a novel procedure which allowed the use of coarse meshes in the
simulation of delamination.
22
CHAPTER 3
FABRICATION OF COMPOSITE LAMINATE
Manufacturing of thermoset composite laminates can be accomplished in one of many
ways including, resin transfer molding (RTM), compression molding, and filament winding. In
this thesis work, vacuum assisted resin transfer molding (VARTM), a variation of RTM is used
in the fabrication of CFRP composite panels. The VARTM process, which is a liquid composite
molding process (LCM), involves the injection of premixed liquid thermoset resin into dry fiber
preform in a closed mold, where the liquid coats the fibers, fills spaces between fibers, expels air
and finally as it cures transforms into matrix [3]. The VARTM process has an advantage of low
tooling costs, low pressure requirements compared to compression molding and has improved
quality and fiber wet-out compared to hand layup process. An idealized schematic of the
VARTM setup is shown in Figure 3.1.
Figure 3.1 Schematic of the VARTM Process [65]
Sized unidirectional IM7 carbon fibers were used for the fabrication of the composite
panels. IM7 fibers possess high tensile and shear strength. The VARTM process involved the
23
surface preparation of a one-sided hard surface mold, fiber layup, vacuum creation, resin
preparation, infusion of resin and curing. The fabricated panel used in the thesis work had ply
stacking sequence of [06/904/06]T with and without an wax insert in-between the middle four 90o
plies to simulate damaged (delaminated) and undamaged coupons respectively.
The one-sided hard surface mold in the form of a steel plate was sanded and cleaned to
remove any contaminants that may interfere with the fabrication of the laminate. The plate was
then cleaned using acetone and a release agent (Freekote) to prevent the laminate from adhering
to the plate after resin cure. To further aid in the release of the laminate after curing, a release
film of dimension 330.2 mm x 330.2 mm was placed on the mold surface before laying up the
fabric. Each ply of IM7 carbon fabric was cut into a dimension of 304.8 mm x 304.8 mm square
that form the 16 plies. Six plies were laid in the 0o direction, followed by two plies in the 90
o
direction. To embed the delamination, a 0.03 mm thick, aluminum insert of dimension 50.8 mm,
cladded in peel ply was centered on top of the eight plies at the vacuum outlet section of the
panel and a cast wax of 0.3 mm thickness of dimension of 101.6 mm x 50.8 mm, on the resin
inlet section as shown in Figure (3.2). The reason for using the two different methods above was
to investigate which one of the two would give the desired stiffness reduction outcome. The
middle section of the panel had no form of defect, to enable coupons to be obtained for the
undamaged (section of panel with no delamination - control sample) and damaged cases
respectively, from the same panel. This would allow for proper analysis and comparison of test
results.
24
Figure 3.2 Photograph showing wax on resin inlet side and cladded aluminum sheet on the
vacuum outlet side of the laid carbon fiber plies
The remaining two, 90o plies and six, 0
o plies were then placed on top of the inserts in that
order and orientation, making the laminate symmetric about the mid-plane. A second release
fabric, followed by a distribution mesh, both of dimension 330.2 mm x 330.2 mm was then
placed above the stack of fibers. The low flow resistance distribution mesh, helped control the
resin flow and promoted good fiber wet-out by ensuring uniform distribution of resin. The entire
setup was vacuum sealed with the aid of tacky tape, leaving out inlet and outlet point for resin
infusion and excess resin collection respectively. Vacuum pressure was monitored with the aid of
a pressure gauge placed at the outlet which was set at 95 kPa. The vacuum was mainly used for
resin infusion and assisted in consolidating the fibers to achieve the desired fiber volume fraction
[66]. The system was then allowed to debulk for an hour to remove air pockets and promote
adhesion of fibers to resin. Figure 3.5 below shows the system being debulked, and the inlet and
outlet connections.
SC-780 matrix, a two part epoxy, mixed in a ratio of resin to hardener, 4:1 by volume, was
used as the matrix material. After the resin infusion, the setup was left for an additional 1.5 hours
to ensure uniform resin distribution and fiber wet-out by eliminating dry spots within the fiber
stack. Investigations into the flow pattern and fill time of resin distribution using finite element
25
models [67, 68] has proved that the release film placed between the carbon fiber fabric and
distribution mesh reduced the resin fill time. The vacuum was removed and the setup was green
cured overnight at room temperature. The process was then followed by oven curing for 6 hours
at 71oC, and then post-curing for an additional 2 hours at 104
oC to ensure complete resin cross-
linking. The final trimmed-out panel was cut into 254 mm (10 in.) x 25.4 mm (1 in) coupons for
damage detection tests and 25.4 mm (1 in.) x 25.4 mm (1 in.) pieces for moisture and acid
digestion testing. The fiber volume fraction of the fabricated composite panel, obtained using
acid digestion experiment, was found to be in the range of 49-50%.
Figure 3.3 VARTM materials stacking sequence (release fabric, carbon fibers, release fabric,
distribution mesh) and taping process
Figure 3.4, (a) Vacuum Pulled on system and (b) Vacuum Gauge on the Resin Catch reading
about 95kPa
26
Figure 3.5 Photograph of setup being de-bulked, showing inlet-outlet lines, resin catch and
vacuum pump
Figure 3.6 Photograph of VARTM manufactured laminate [06/904/06]T with embedded
delamination
27
CHAPTER 4
STRUCTURAL HEALTH MONITORING
The SHM process involves implementation evaluation, data acquisition and damage
feature extraction, and classification. Implementation evaluation includes studying the economic
and human life safety advantages, possible damage types, operational and environmental
conditions, cost and maintenance needs and the time requirement for deploying the SHM system.
The data acquisition portion of the process involved a decision about the number and type of
sensors, placement of sensors, type of actuation, acquisition hardware, and signal transmission
hardware and data storage. A signal processing step is typically performed after data acquisition
to cleanse and remove any noise from the data. In feature extraction, damage sensitive features
such as the natural frequency of vibration, strain, time of flight, displacement, and modal shape,
are extracted from sensor data while also compensating for any environmental effects. As a final
step, a statistical analysis of acquired data leads to a classification of the structure as damaged or
undamaged [2]. The Boeing 787 for example, has a health monitoring plan in place which
provides real-time data which enables the prediction of remaining useful life of the structure by
detecting the existence of damage, evaluating its nature and severity and localizing the damage
[69]. The SHM process therefore involves limited human intervention with testing structural
elements not being required to be taken out of service for its implementation compared to other
NDE techniques such as laser ultrasound, laser shearography, infrared thermography and X-ray
radiography.
28
4.1 Damage detection
Damage detection can be done in one of two forms, the active or passive approach.
Damage detection by the active approach involves the use of actuators such as, piezoelectric
patches to cause perturbations in a structure, while disturbances in the passive approach are
caused by ambient vibrations, damage, or impact [70]. However, in both instances, structural
response is obtained by surface mounted or embedded sensors. Kessler et al. [71] used both
approaches, with the effectiveness of the combination being confirmed later on by Mal et al.
[72]. Passive SHM can be in the form of impact measurements using foil gages or piezoelectric
sensors [72, 73] with examples of active approach being the Lamb wave technique and
impedance method.
In this thesis, damage, in the form of embedded mid-plane delamination is detected using
the Lamb wave technique. The Lamb wave technique has shown great promise among the other
damage detection methods and is capable of detecting existence of damage, its severity, type,
location and size. They were first used in damage detection by Worlton [74].
Lamb waves are elastic in nature and are used for both internal and external damage
detection due to their ability to travel through the thicknesses of plates. They respond to different
types of damage with changes in flight time, wave attenuation, phase information and
propagation characteristics [75] and are therefore very useful in investigating structural integrity.
The difference in amplitude for the damaged and undamaged structure, gives information on the
severity of the damage with the difference in time of flight providing information on the damage
location [76]. They are elastic waves confined to a plate surface and can travel long distances. A
surface mounted piezoelectric actuator was used to generate Lamb waves in a composite coupon
29
and the response was measured using a surface mounted piezoelectric sensor. This aspect of the
research involved the optimization of sensor location and the number of sensors.
4.2 Lamb wave methods
Damage detection using Lamb waves involves two primary modes of operation namely;
pulse echo and pitch catch [77]. The pitch catch method involves signal transmitted by an
actuator from one end of a structure being received by a sensor placed at the opposite end.
Damage detection here relies on changes in amplitude ratio and time of flight of the transmitted
signal and can cover long distances but requires a dense network of sensors [78]. The pulse echo
uses either the same transducer or a sensor co-located with an actuator to transmit and receive
signals; with damage detection relying on reflections or echoes from any discontinuities such as
cracks or delamination in a structure.
4.3 Sensing methods
Different kinds of sensors ranging from fiber optics, fiber Bragg grating, strain gages and
magnetostrictive sensor have been used to detect various types of structural damages. However,
piezoelectric transducers have an advantage of light weight and high sensitivity to small amounts
of strain, and can generate relatively high amplitude output signals compared to the others. Also,
lead zirconate titanate (PZT) has a major advantage in its ease of use in different detection
techniques such as acoustic emission, ultrasonic Lamb waves, and strain-based methods [71].
The ability of the same PZT transducer to act as either an actuator or a sensor at any time allows
for baseline-free detection and gives the flexibility to monitor proper sensor operations. Given
30
these advantages, the piezoelectric transducers coupled with Lamb waves were used for
delamination detection in composite laminate coupon in this thesis work.
4.4 Piezoelectric transducers and principle of operation
PZT actuators are normally mounted on the surfaces of composite materials to generate
Lamb waves [79]. They operate on both principles of piezoelectric and inverse piezoelectric
effect. Due to the crystalline property of piezoelectric materials, electric pulses are able to be
converted into mechanical vibrations in accordance to the inverse piezoelectric effect.
Piezoelectric sensors, on the other hand, convert mechanical vibrations into electrical charge by
the piezoelectric effect.
Examples of piezoelectric materials include lead metaniobate, PZT and polyvinylidene
fluoride. The PZT is the preferred choice when using the Lamb wave detection technique
because of its ability to withstand high operating temperatures (365oC) [78].
4.5 Actuation parameters
Piezoelectric actuators are excited by the application of voltage. To produce strong Lamb waves
with larger signal to noise ratio requires high voltage, which leads to high consumption of power
[80]. SHM systems however, have lower power requirement with an optimum voltage range of
5-10V. Higher energy is associated signals made up of large number of peaks, which
compensates for any attenuation due to damping in long distance signal propagation. Increasing
the number of peaks however, increases the pulse width and in turn, the interference [81, 82].
The use of low frequency for composite damage detection has been studied by Diamanti and
Soutis [83]. Smaller number of modes can be excited using lower frequencies which make them
31
more distinguishable. The first order anti-symmetric (A0) mode has a smaller wavelength for a
given frequency compared to the first order symmetric (S0) mode making A0 waves more
sensitive to small levels of damage. S0 modes also undergo higher attenuation during propagation
[76, 82]. Based on these characteristics, a five cycle tone burst of 1 kHz center frequency and
amplitude of 6 V to generate A0 mode Lamb wave was considered in the testing protocol. The
excitation input signal is as shown below in Figure 4.1.
Figure 4.1 A typical PZT excitation input signal for the testing scheme used in this study
4.6 Delamination detection
In this thesis, delamination, a form of inter-laminar damage, prevalent in composite
laminates is detected using the Lamb wave technique. Inter-laminar stresses lead to delamination
in composites at different locations such as, around the edges and holes, which causes a
reduction in the modulus and strength of the laminate leading to premature failure [23]. The
Lamb wave based approach with PZT actuators and sensors has been used by Soni et al. [82] to
study changes in amplitude ratio due to fatigue crack. In this thesis however, the Lamb wave
32
based approach was used to study delamination in composite materials. A surface mounted PZT
sensor was used to measure the response of Lamb waves generated using a surface mounted PZT
actuator. The response of Lamb waves to embedded delamination in a composite coupon
containing 16 plies of IM7 carbon fiber was studied.
4.6.1 Experimental analysis of delamination detection
Composite coupons of dimension 254 mm x 25.4 mm machined from 304.8 mm x 304.8 mm
panels were used for the test. Variation in the modulus of elasticity and width was minimized by
cutting each set of test coupon from a single piece of composite panel. Tests were carried out on
both the undamaged (coupons with no delamination) and damaged (delaminated coupons)
obtained from the same panel. A Lamb wave based approach was used for damage detection. A
surface mounted PZT actuator was used to excite Lamb waves in the composite coupon and the
response, received by surface mounted PZT sensor. The response from an undamaged composite
coupon was used as a reference baseline.
4.6.2 Data acquisition setup
The composite coupon was fixed at one end with a PZT actuator surface-mounted on the
free end. An AC signal applied to the terminals of a PZT actuator puts the coupon into vibrations
by the inverse piezoelectric effect. A Hanning modulated five cycle tone burst signal at a center
frequency of 1 kHz was used for Lamb wave excitation. The actuation signal was amplified by a
factor of 20 using a piezo amplifier before sending to the actuator. The Lamb waves, generated
by the actuator were received by the sensor located 60 mm from the clamped end. The sensor
generates a voltage equivalent to the induced strain by the direct piezoelectric effect. The sensor
was connected to a data acquisition (DAQ) card through a connector box with a sampling rate of
100 kHz. National instrument LabView software was used to synchronize the actuation initiation
33
and sensor data collection (DAQ). Real time sensor data were collected for further analysis. The
experimental setup is shown in Figure 4.2 below.
Figure 4.2 Data acquisition system using PZT actuators and sensors
Actuation input was generated using an Agilent 33210A arbitrary waveform generator. A
National Instruments LabVIEW program was developed using the Agilent 33XXX driver and
VIs (virtual instruments) to control the waveform generator through the general purpose interface
bus (GPIB) port. A GPIB to USB cable connects the waveform generator to the computer. The
arbitrary waveform generation feature was utilized to generate a Hanning burst. The frequency,
amplitude, burst count and burst rate of actuation input were user configurable in the LabVIEW
program.
The output of the waveform generator was fed to a piezo amplifier AA 301HS that had an
amplification factor of 20. The amplitude of the waveform generator was set to 6 Vp-p to generate
input actuation amplitude of 120 Vp-p after amplification. This was within a safe margin for the
PZT actuator that had an operating range of 180 Vp-p.
34
Data from the PZT sensors were read into a LabVIEW program using the NI DAQmx driver
software. The number of sensors to import, and the number of samples to read were configurable
in the LabVIEW program. The program displayed a real time waveform of voltage versus time
for all of the sensors for rapid data analysis. The sensor data was stored in text files for further
analysis. The experimental system was found to generate the desired amplitude and shape A0
mode Lamb waves in a composite coupon required for this study.
4.7 Test for damage detection
4.6.1 Simulated damage using an embedded longitudinal delamination
Tests were performed on 254 mm x 25.4 mm coupons, machined from [06/904/06]T
IM7/SC780 composite laminate to enable both damaged (delaminated) and undamaged coupons
to be obtained from the same panel. The fabrication process outlined in chapter 3 was followed.
Embedded delamination in between the mid-plane of the [06/904/06]T laminate extending through
the width, was used to simulate damage. . The response from an undamaged composite coupon
was used to provide a baseline for reference. The coupon was clamped at one end with a PZT
actuator disk of center frequency 7.3 kHz measured 12.7 mm in diameter and was 0.38 mm
thick. A Hanning modulated five cycle tone burst signal at a center frequency of 1 kHz was used
for Lamb wave excitation. The Lamb waves generated by the actuator were received by a PZT
sensor located 60 mm from the clamped end as shown in Figure 4.3.
35
Figure 4.3 Pictorial representation of composite coupon with delamination (wax insert)
demonstrating the actuator and sensor placement relative to delamination
Figure 4.4 below shows the comparison of the Lamb wave response of an undamaged
composite coupon and a coupon with embedded delamination. The difference in amplitude ratio
of the incident wave between the undamaged and damaged coupon on the average, was found to
be 0.05% due to the embedded delamination in the damaged coupon.
Figure 4.4 Experimental results comparing the Lamb wave response of a,
[06/904/06]T orientation, undamaged coupon and a coupon with embedded delamination in the
middle four plies
36
4.8 Experimental analysis of damage evolution in delaminated coupon
The four-point bend test as a failure characterization tool was employed to study crack
growth analysis on composite coupons containing embedded delamination. With the embedded
delamination located mid-way along the length of the coupon and spanning the entire width of
the specimen, the four-point bend test served as an appropriate engineering procedure for the
analysis of flexural stiffness degradation in the composite coupon. Variation in the modulus of
elasticity and width was minimized by cutting each set of test coupon from a single piece of
composite panel. The MTS QTest/25 machine with fixtures of 152.4 mm (6 in) supporting span,
and a 76.2 mm (3 in) loading span as shown in Figure 4.5 below was used for the analysis.
Figure 4.5 Photographs of the Four-Point Bend Fixtures mounted in the MTS QTest 25 machine
During the testing process to investigate crack evolution, a transverse vertical load was
applied with a crosshead speed of 2 in/min (5.08 mm/min), which initially led to matrix cracking
for the damaged specimen, then to delamination extension followed by the final state of fiber
fracture. This served as basis for investigating the various modes of failure in the composite
37
coupons. Shown below in Figure 4.6, are pictures of the specimen under loading, going through
the damage evolution process in the MTS QTest/25 machine.
Figure 4.6 Photograph showing damaged specimen, (a) under load; (b) Delamination Extension
in specimen; (c) and (d) Transition from Matrix Crack, Delamination to Fiber Fracture
Damage evolution for the undamaged specimen, as shown in Figure 4.7 below, started
with matrix cracking, then delamination at the (0o/90
o) ply interface where there is a change in
stacking sequence. The four-point bend specimens then fractured in a brittle manner with across
the width-fractures, which appeared to have initiated under the central loading rollers causing not
only width-wise fracture but also inter-laminar failures between these central rollers.
Figure 4.7 Crack initiation and evolution in undamaged coupon
From the configurable program used with the Testworks QT 2.03, which enabled all the
coupon dimensions to be configured, historical data for the experiment was obtained for
38
analyses. Tabulated below are the load-deflection data obtained from the test. Slopes were
obtained from linear portions of the load-deflection plots and represented the structural stiffness
of the composite coupons given as , where F represented the load and δ the
deflection at load application points as shown on the schematic below.
Figure 4.8 Schematic of the support and load spans of the Four-Point Bend Fixtures
Table 4.1: Comparison of Slope (Structural stiffness) of the Load-Deflection data obtained from
the Four-Point Bend Test of tested specimens of dimension (10 in. x 1 in. x 0.14 in)
SPECIMEN
TYPE
UNDAMAGED
1
UNDAMAGED
2
DAMAGED
1
DAMAGED
2
SLOPE
(lbs./in) 1406 1423 896 1146
Table 4.2: Comparison of Slope (Structural stiffness) of the Load-Deflection data obtained from
the Four-Point Bend Test of tested specimens of dimension
(254 mm x 25.4 mm x 3.6 mm) (Presented in SI units)
SPECIMEN
TYPE
UNDAMAGED
1
UNDAMAGED
2
DAMAGED
1
DAMAGED
2
SLOPE (k N/m) 246 249 157 200
39
Table 4.3: Comparison of the peak load and peak stress of the tested specimens
SPECIMEN
NUM. TYPE
PEAK LOAD
(lbs)
PEAK STRESS
(ksi)
1 UNDAMAGED 360.1 36.7
2 UNDAMAGED 365.2 37.3
MEAN 362.65 37
STDEV 2.55 0.3
3 DAMAGED 230.0 23.6
4 DAMAGED 236.5 24.1
MEAN 233.3 23.85
STDEV 3.25 0.25
% CHANGE -36 -36
Table 4.4: Comparison of the peak load and peak stress of the tested specimens
(Presented in SI units)
SPECIMEN
NUM. TYPE PEAK LOAD (N)
PEAK STRESS
(GPa)
1 UNDAMAGED 1609 0.248
2 UNDAMAGED 1624.6 0.251
MEAN
1616.8 0.2495
STDEV
7.8 0.0015
3 DAMAGED 1023.0 0.158
4 DAMAGED 1052.0 0.162
MEAN
1037.5 0.16
STDEV
14.5 0.002
% CHANGE
-36 -36
Figure 4.9 shows a plot of the load against deflection for damaged and undamaged
specimens under the four-point bending. The undamaged specimens exhibited a greater bending
stiffness than the damaged specimens as shown by the slope of the load-deflection curve, due to
40
the effect of the embedded delamination in the damaged specimen. Also, the difference in peak
load and peak stress gave information on the stiffness degradation in the damaged specimen as a
result of the embedded delamination with the average peak load for the undamaged specimen
being higher than that of the damaged.
Figure 4.9 Load (F)-Deflection plots for Damaged and Undamaged specimen under Four-Point
Bending
41
CHAPTER 5
FINITE ELEMENT ANALYSIS
Due to their ability to model complex shapes and loading conditions, finite element
methods have become widely used for analyzing damage in different kinds of materials under
different loading and boundary conditions. In this present study, a two dimensional finite
element analysis of delamination in IM7-SC780 carbon/epoxy coupons of dimension
254 mm x 3.5 mm under four-point bending (FPBT) was carried out using Abaqus version 6.11-
1 commercial FEM software. Four node bilinear plane stress quadrilateral elements (CPS4R)
were used to discretize both damaged (simulated embedded delamination) and undamaged
coupons. The two dimensional shell element models for the finite element analysis were
developed using material properties of the composite coupons obtained from mechanical testing
of the specimen. These properties as tabulated in Table 5.1 were used for the numerical
simulation.
Table 5.1: Material properties of IM7/PET15 composite used for numerical simulation
(Presented in SI units)
E11 E22 E33 G12 G13 G23 Ѵ12 Ѵ21 Ѵ23
GPa GPa GPa GPa GPa GPa
113.8 5.24 5.24 31.78 31.78 1.081 0.3 0.014 0.42
The bending stiffness for the 16 plies composite laminate was obtained from the Classical
Lamination Theory (CLT) using the equations (5.1 – 5.9) below. The stiffness matrices of the 0o
and 90o plies of the laminate are given by:
42
(5.1)
(5.2)
(5.3)
(5.4)
And the angle-invariant stiffness properties of the laminate , and were calculated from:
(5.5)
(5.6)
(5.7)
From Equations (5.5-5.7), the reduced stiffness matrix, was calculated from:
(5.8)
Where is the ply orientation angle of the laminate [3]. The theoretical bending stiffness for
the laminate with stacking sequence [06/904/06]T was calculated using Equation (5.9) below;
(5.9)
Where N represents the total number of laminas in the laminate and is the distance from
the mid-plane to the top of the lamina and is the distance from the mid-plane to the bottom
of the lamina.
Figure 5.1 shows a schematic of the damaged composite with partial delamination in-
between the two sub-laminates, under four-point bending.
43
Figure 5.1 Illustration of damaged coupon with partial delamination in-between two sub-
laminates (1) and (2), with stacking sequence of T and T respectively
The four-point bend test of undamaged specimen was modeled using 2D shell element as
shown in Figure 5.2 below. It was modeled as a four-layered composite laminate corresponding
to the stacking sequence of T and T respectively and meshed using 2400, four
node linear quadrilateral elements (CPS4R). Both ends were fixed. For numerical analysis of the
undamaged coupon under four-point bending a uniform pressure load of 40 N/mm2 was applied
to the model.
Figure 5.2 FE model of a composite coupon with stacking sequence of
[06/904/06]T representing the loading and boundary conditions for Four-Point Bending
The value of the maximum displacement for the four-point bend test was then analyzed
numerically using Abaqus. Below is a plot of how displacement changes along the length of the
specimens for both cases of damaged and undamaged coupons.
44
Figure 5.3 Abaqus results showing the deflection contour plot and legend for undamaged
composite coupon undergoing Four-Point Bending in millimeters
The damaged specimen (delaminated specimen) was modeled using two sub-laminates of
2D shell elements with one below and one above the plane of delamination. The top and bottom
sub-laminates had a stacking sequence of [06/902]T and [902/06]T respectively as shown in Figure
5.4. The sub-laminates were modeled by finite elements. The undamaged portion of the sub-
laminates was modeled using a contact element such that, the nodes corresponding to the top and
bottom sub-laminates had contact property with their rotational degrees of freedom (DOF) tied
together. Delamination was explicitly modeled, but with master-slave nodes defined for contact,
thus preventing interpenetration of contact surfaces.
The damaged portion (mid-rectangular section shown in Figure 5.5(b)) was modeled such
that the sub-laminates had just surface-to-surface property, simulating the wax insert used
experimentally. This ensured continuity of displacements and rotation at the plane of
delamination. Figures 5.4 - 5.5 below represent the model for the damaged specimen.
45
Figure 5.4 (a) and (b), FE model showing top and bottom sub-laminates of stacking sequence
[06/902]T and [902/06]T respectively
Figure 5.5 (a) and (b), FE model showing constraint and interaction property for damaged
composite coupon
46
Figure 5.6 Abaqus results showing the deflection contour plot and legend for damaged composite
coupon undergoing Four-Point Bending in millimeters
From the analytical results it was realized that the location of the delamination at the mid-
section of the composite coupon under four-point bending, showed limited difference in the
maximum deflection between the damaged and undamaged cases, as can be verified from the
Abaqus results, even though experimental results showed otherwise. This was later attributed to
the pure bending in-between the two loading points which led to zero shear forces and inter-
laminar shear stresses being experienced at the mid-section of the composite coupon in both
damaged and undamaged coupons under four-point bending. This situation could be remedied by
relocating the delamination to one end of the specimen or subjecting the damaged specimen to
three-point bending.
Based on this observation, finite element analysis of the composite coupon under three-
point bending was then carried out. The dimensions for both damaged and undamaged specimens
were the same as that used for the four-point bending, with the only differences being the load
and load application point. With the same material properties, finite element analysis of three-
point bending for damaged (delaminated) and undamaged composite coupons were carried out
using a mid-point uniform pressure load of 40 N/mm2. Figure 5.7 shows the load and boundary
conditions of the finite element model for the undamaged case.
47
Figure 5.7 FE model of undamaged composite coupon with stacking sequence of
[06/904/06]T representing the load and boundary conditions for Three-Point Bending
The finite element model was meshed using 2400 elements and below is the deflection contour
plot and legend showing the maximum deflection of the undamaged coupon.
Figure 5.8 Abaqus results showing the deflection contour plot and legend for undamaged
composite coupon undergoing Three-Point Bending in millimeters
The delaminated specimen was modeled similar to that used for the four-point bending with the
extreme ends constrained and tied, and the delaminated mid-section with surface-to-surface
interaction property as shown in Figure 5.9.
(a)
48
(b)
Figure 5.9 (a) and (b), FE model showing constraint and interaction property for damaged
composite coupon
1200 elements each for the top and bottom sub-laminates were used to discretize the model,
below is the deflection contour plot and legend showing the maximum deflection of the damaged
coupon under three-point bending.
Figure 5.10 Abaqus results showing the deflection contour plot and legend for damaged
composite coupon undergoing Three-Point Bending in millimeters
Analytical results for the maximum deflection of the damaged specimen undergoing three-
point bending showed a significant increase in displacement (28%) compared to that of the
undamaged case. This was due to the presence of delamination and the effect of shear forces and
inter-laminar shear stresses between the sub-laminates of the composite coupon as a result of the
location of the applied uniform pressure load.
The comparison of the numerical solutions for both cases of damaged and undamaged
composite coupons under three-point and four-point bending is as shown in Table 5.2.
49
Table 5.2: Comparison of the maximum deflections, for damaged and undamaged specimens of
stacking sequence [06/904/06]T under three-point and four-point load conditions
Four-point bending Pressure load
(N/mm2)
Numerical
Solution
Maximum
deflection
(mm)
% difference
Undamaged specimen
40
Abaqus-CPS4R
(2400 elements) 7.41
0.80 Damaged specimen Abaqus-CPS4R
(2400 elements) 7.47
Three-point bending
Undamaged specimen
40
Abaqus-CPS4R
(2400 elements) 7.30
28.22 Damaged specimen Abaqus-CPS4R
(2400 elements) 9.36
50
CHAPTER 6
FINITE ELEMENT MODEL FOR OPTIMIZATION
6.1 Interfaces between MATLAB and Abaqus
Sensing optimization problems usually seek to find a set of sensors that provides the best
possible sensing quality at the minimum possible cost. Genetic and evolutionary algorithms [84]
and stable noisy optimization by branch and fit (SNOBFIT) [85] have both been used
successfully for sensor placement optimization (SPO). Even though large sensor density
generally implies easy damage detection, the negative effect of cost, power and weight cannot be
overemphasized and therefore the need for optimal sensor locations and density. A four step
sensor optimization method was proposed by Guratzsch et al. [86], where a finite element
analysis was performed by modeling the plate thickness and material properties; Young’s
modulus, Poisson’s ratio and density, as a Gaussian random field. Each node of the finite
element model was considered for a probable sensor location and the SNOBFIT algorithm was
used for SPO.
Using an approach similar to what is outlined above; SPO was achieved for damaged and
undamaged finite element models developed using Abaqus. Data obtained from reports
generated on each particular model in Abaqus, was saved as a DAT file and called using
MATLAB to achieve the desired number of data sets for particular damaged and undamaged
cases. Each Abaqus run was called after values of material properties for both the damaged and
undamaged cases were varied within experimentally determined standard deviation to regenerate
51
a new input files. A suitable feature set (strain, displacement, and stress) was written to the report
files. The data was recorded only for the top surface nodes for the undamaged model and the also
the top surface nodes of the top layer sub-laminate for the damaged case, with the aim to
optimize sensor placement only on the top surface of the specimen. Embedding the sensors or
placing sensors on the sides or bottom of the coupon were not considered in this research project.
The process flow is presented in Figure 6.1 below.
Figure 6.1 A flow chart demonstrating MATLAB, Abaqus interface used for generating data set
for sensor placement optimization (where count is the iteration number)
6.2 Embedded delamination test
The delaminated coupon of dimensions 254 mm x 25.4mm x 3.6 mm was modeled using
two sub-laminates of 3D solid element, with one below and one above the plane of delamination.
52
The top and bottom sub-laminates had a stacking sequence of [06/902]T and [902/06]T
respectively, and the geometry was discretized using 38304 linear brick elements of type
C3D8R. Damage in the form of delamination was introduced into the middle section of the two
sub-laminate using wax insert. The coupon was simply supported by a four-point bend fixture of
supporting span 152.4 mm (6 in) and loading span of 76.2 mm (3 in). The coupon was subjected
to a uniform pressure load at the loading roller sections modeled as an area, with dimensions of
25.4 mm (1 in) x 3 mm (.12 in). For this test, 60 simulations were run, 30 simulations of the
undamaged model and 30 simulations of the damaged model with varying material properties.
The longitudinal strain contour plot for the damaged case is shown in Figure 6.2. The static strain
values measured from 30 undamaged simulation cases at the entire surface nodes was used to
train the classifier.
Figure 6.2 A FE representation of longitudinal strain component contour plot observed on the
top surface of a composite coupon with embedded delamination (in the middle) and subjected to
four-point bending
The aim of the experiment was to find the optimal sensor placement given the damage at a
specific location. The process flow is as summarized below in Figure 6.3.
53
Figure 6.3 A flow chart demonstrating the sensor placement optimization process
Damage classification was performed using artificial neural network (ANN) and sensor
optimization carried out using the evolutionary strategy. A sample result has been added in this
section for illustration. This result came from the work led by Muhammad Farooq [87]. The
surface plot of one of the feature vectors (displacement) used in the sensor optimization
procedure is as shown in Figure 6.4; the initial and final placement of sensors proposed by the
optimization algorithm is as shown in Figure 6.5. The sensors failed to move during the
optimization process with possible reasons being over-fitting of the classifier to the given data
set since we had close to 20% damage in this case. Also the use of multiple sensors in this
particular case may have accounted for the inability of the sensors to move due to its dependence
on the initial sensor locations with respect to the damage and level of noise.
54
Figure 6.4 Displacement surface plot comparison between damaged and undamaged coupon
Figure 6.5 Initial and final placement of sensors after completion of optimization algorithm
(dots represent sensors)
55
CHAPTER 7
CONCLUSIONS AND FUTURE WORK
7.1 Conclusion
With their enormous advantages in the manufacturing sector taken into consideration,
damage analysis and its comprehension in CFRP composites has become very critical as a way
to sustain its current and prospective applications, as well as its extension into other areas of
manufacturing. The search for a reliable structural health monitoring system with its added
economic advantage of enabling maintenance of engineering structures to be carried out based
on prevailing condition of a structure, rather than on estimated time has become paramount. This
has become necessary especially in large and expensive structures. This thesis therefore makes a
useful contribution to the detection and analysis of delamination in CFRP laminated composites
by analyzing stiffness degradation using four-point and three point bend test. Here, numerical
modeling was combined with an experimental study using a Lamb wave-based approach to
detect delamination in composite coupons. Damaged and undamaged test specimens were
fabricated using a VARTM process and subjected to four-point bend test to analyze delamination
onset and evolution. Numerical analysis of delaminated and un-delaminated composite coupons,
were also carried out using three point bending. Surface mounted PZT actuators and PZT sensors
were used for actuation of Lamb waves in the composite materials and in sensing the response of
the structure. The actuation parameters and the choice of transducers were made based on a
review of relevant literature. National Instruments LabVIEW software was used to synchronize
56
the actuation initiation and response measurement. Embedded delamination using a surface
mounted PZT actuator and a surface mounted sensor was experimentally demonstrated.
Using the MTS QTest equipment, and with the pre-written program Testworks QT 2.03,
the four-point bend test experimental analysis was carried out. This analysis was supported by
finite element models developed using Abaqus version 6.11-1, commercial FEM software.
Numerical results obtained from finite element modeling for both cases of damaged and
undamaged specimens under four-point bend test produced insignificant difference in maximum
deflection due to absence of shear stresses at the delaminated section of the specimen. Further
numerical analysis was then carried out for both cases of damaged and undamaged specimen
using three-point bend test. The effect of shear stresses this time led to differences in maximum
deflection for the two cases. Time constraints however made it impossible to verify this
experimentally.
From the experimental results, it was found out that the A0 mode Lamb waves carried
sufficient information to detect an embedded delamination in a composite coupon. The
amplitude ratio of the incident wave for the undamaged coupon was found to be higher than the
damaged coupon due to the embedded delamination.
A detailed description of a damage detection system, results from finite element models of
four-point and three-point bend test for damaged and undamaged composite coupon, and a
review of the limitations and capabilities of anti-symmetric mode Lamb waves in embedded
delamination detection were discussed in this thesis. This information can serve as a good
reference for four- point and three-point bend tests modeling and help develop a reliable
structural health monitoring system for delamination.
57
7.2 Suggestions for future work
The presence of stress concentration zones during the performance of four-point bend test
has a critical influence on the total accuracy of the experimental results. The loading rollers
introduce stress concentration under four-point loading. To mitigate this effect and enable more
accurate results to be obtained in future experiments, the rollers need to be lined with a soft
plastic or rubbery material or even the steel rollers would have to be replaced completely with
plastic ones altogether to avoid influencing damage on the compressive side of the specimen.
Strain gages could also be mounted underneath the mid-section of the specimen to give a more
accurate strain reading of the specimen. Experimental error can also be reduced by carrying out
effective calibration of the testing equipment. Also, three-point bend test for the damaged and
undamaged composite coupons could be carried out experimentally and results compared with
the exact and numerical solutions. This will help provide useful information on the behavior of
CFRP’s.
To efficiently classify the damage (delamination), the wavelength of actuation frequency
needs to be on the order of, or comparable to the damage length. Thus the actuation frequency
plays a very important role in generating the desired mode Lamb waves and also in deciding the
level of damage that can be detected. More careful selection of frequency must therefore be
carried out in future experiments.
The VARTM manufacturing process could also be improved by placing the resin/epoxy
mixture in a vacuum chamber to rid it of air bubbles before infusion into the fibers to ensure
improved fiber wet out, reduce void content to the barest minimum and obtain a more uniform
specimen width. Also more innovative ways to embed delamination in carbon/epoxy composite
laminates should also be explored to assist in obtaining the desired effect without damaging the
58
material properties of the laminate. This thesis was focused primarily on detecting delamination,
which in this case was similar to a horizontal crack. Transverse crack damage, which is also
common in composite materials can also be investigated by similar fabrication process and SHM
system setup in future works.
Abaqus 11-1 software should be explored further to improve upon composite coupon
models, to give a more accurate representation of the surface irregularities of the coupons.
This thesis presented a case study showing the result of sensor placement optimization for a
composite coupon with embedded delamination. The percentage of damage and number of
sensors can be reduced and the same algorithm re-applied to check its effectiveness. Also, other
optimization algorithms such as the genetic algorithm can be explored to find out which will
produce the desired results in a case similar to what was undertaken in this thesis work. Tests
considered strain and displacement values as the feature vector set. The existing setup can be
extended to perform dynamic tests and extract frequency domain as well as time domain features
for damage classification and sensor placement optimization.
59
REFERENCES
[1] R. Waugh, “Not just a load of hot air: Dream becomes reality as Boeing’s new carbon-fiber
787 Dreamliner heralds a new age of air travel,” 2011.
[2] H. Sohn, C. R. Farrar, F. M. Hemez, and J. J. Czarnecki, A Review of Structural Health
Monitoring Literature 1996-2001, 2002.
[3] P. K. Mallick, “Fiber Reinforced Composites: materials, manufacturing, and design,” Third
Edition
[4] http://www.substech.com/dokuwiki/doku.php?id=polymer_matrix_composites_introduction
[5] Z. R. Khokhar, Finite-element analysis of delamination in CFRP laminates: effect of material
randomness, Doctoral Thesis-Loughborough University. (April, 2010).
[6] Sung Kyu Ha, Kyo Kook Jin and Yuanchen Huang. Micro-Mechanics of Failure (MMF) for
Continuous Fiber Reinforced Composites, DOI: 10.1177/0021998308093911
Journal of Composite Materials 2008 42: 1873 originally published online 17 July 2008
[7] M. S. Sham Prasad, C.S. Venkatesh, T. Jayaraju. Experimental Methods of Determining
Fracture Toughness of Fiber Reinforced Polymer Composites under Various Loading
Conditions. Journal of Minerals & Materials Characterization & Engineering, Vol. 10, No.13,
pp.1263-1275, 2011.
[8] C. K. H. Dharan. (1978). Fracture mechanics of composite materials. Journal of engineering
materials and technology. 100: 223-247.
[9] K.Y. Chang, S. Liu, and F. K. Chang. Damage tolerance of laminated composites containing
an open hole and subjected to tensile loadings. Journal of Composite Materials, (1991), 25: pp.
274-301.
[10] L.R. Dharani and F.S. Ji, Effect of transverse cracking on stiffness reduction of cross-ply
laminates. Applied Composite Materials, (1995), 2(4): pp. 217-231.
[11] A.L. Highsmith and K.L. Reifsnider, Stiffness reduction mechanisms in composite
laminates ASTM digital library, (1982), 775: pp. 103-117.
60
[12] Y. Qingda and B. Cox, Cohesive models for damage evolution in laminated composites.
International Journal of Fracture, (2005), 133(2): pp. 107-137.
[13] S. Ogiharaa and N. Takedab, Interaction between transverse cracks and delamination during
damage progress in CFRP cross-ply laminates. Composites Science and Technology, (1995),
54(4): pp. 395-404.
[14] K.W. Garrett and J.E. Bailey, Multiple transverse fracture in 90° cross-ply laminates of a
glass fibre-reinforced polyester. Journal Of Materials Science, (1977), 12(1): pp. 157-168.
[15] S.G. Lim and C.S. Hong, Prediction of transverse cracking and stiffness reduction in cross-
ply laminated composites. Journal of Composite Materials, (1989), 23(7): pp. 695-713.
[16] A. Parvizi, K.W. Garrett, and J.E. Bailey, Constrained cracking in glass fiber reinforced
Epoxy cross-ply laminates. Journal of Materials Science, (1978), 13(1): pp. 195-201.
[17] D. Flaggs, Prediction of tensile matrix failure in composite laminates. Journal of Composite
Materials, (1985), 19(1): pp. 29-50.
[18] S.C. Tan and R.J. Nuismer, A theory of progressive matrix cracking in composite laminates.
Journal of Composite Materials, (1989), 23(10): pp. 1029-1047.
[19] Z. Hashin, Analysis of cracked laminates: a variational approach. Journal of Mechanics of
Materials, (1985), 4: pp. 121-136.
[20] R. Talreja, Modeling of damage development in composites using internal variables
concepts. Damage Mechanics in Composites, ASME, (1987), 12.
[21] L. Boniface, P.A. Smith, M.G. Bader, and A.H. Rezaifard, Transverse Ply Cracking in
Cross-Ply CFRP Laminates--Initiation or Propagation Controlled? Journal of Composite
Materials, (1997), 31(11): pp. 1080-1112.
[22] P.A. Smith and S.L. Ogin, On transverse matrix cracking in cross-ply laminates loaded in
simple bending. Composites A, (1999), 30(8): pp. 1003-1008.
[23] Robert M. Jones, Mechanics of Composite Materials-second edition
[24] L.N. McCartney, Theory of stress transfer in a 0°-90°-0° cross-ply laminate containing a
parallel array of transverse cracks. Journal of the Mechanics and Physics of Solids, (1992),
40(1): pp. 27-68.
[25] A.A. Griffith, “The Phenomena of Rupture and Flow in Solids.” Philosophical
Transactions of the Royal Society London (1921), A(221): pp. 163-198.
61
[26] T.E. Tay, G. Liu, V.B.C. Tan, X.S. Sun and D.C. Pham, Progressive Failure Analysis of
Composites, Journal of Composite Materials 2008 42: 1921 originally published online 17 July
2008, DOI: 10.1177/0021998308093912
[27] A.S.D. Wang, N.N. Kishore, and C.A. Li, Crack development in graphite-epoxy cross-ply
laminates under uniaxial tension. Composites Science and Technology, (1985), 24(1): pp. 1-31.
[28] G. Wimmer, C. Schuecker, H.E. Pettermann, Numerical simulation of delamination onset
and growth in laminated composites. Austrian Aeronautics Research (AAR) / Network for
Materials and Engineering (2006)
[29] V.Z. Parton, Fracture Mechanics-From theory to practice. Gordon and Breach Publisher,
Philadelphia, (1992).
[30] G. R. Irwin, Analysis of stresses and strains near the end of a crack traversing a plate.
Journal of Applied Mechanics, A.S.M.E., New York, (1957).
[31] Martin, R. H., Davidson, B.D. (1999). “Mode II fracture toughness evaluation using four
point bend, end notched flexure test.” Plast., Rubber Compos., 28(8) 401-406.
[32] Shuecker, C., and Davidson, B.D. (2000). “Evaluation of the accuracy of the four-point
bend end-notched flexure test for mode II delamination toughness determination.” Compos. Sci.
Technol., 60,2137-2146
[33] Vinciquerra, A. J., Davidson, D.B. (2004). “ Effect of crack length measurement technique
and data reduction procedures on the perceived toughness from four-point bend end-notched
flexure tests., J. Reinf. Plast. Compos., 23, 1051-1062.
[34] Yoshihara, H. (2004). “Mode II R-curve of wood measured by 4-ENF test.” Eng. Frac.
Mech., 71, 2065-2077.Yoshihara, H. (2008). “Theoretical analysis of 4-ENF test for mode II
fracturing in wood by finite element method.” Eng. Frac. Mech., 75, 290-296.
[35] de Morais, A. B., and de Moura, M.F.S.F. (2006). “Evaluation of initiation criteria used in
inter-laminar fracture tests.” Eng. Fract. Mech., 73, 2264-2276.
[36] Thouless, M.D. (2009). ‘’ The effects of transverse shear on the delamination of edge-notch
flexure and 3-point bend geometries.” Composites, Part B, 40, 305-312
[37] P. Qiao, P.E., F. ASCE; F. Chen, C. S. Hamey, and J. Wang. “Mixed-mode fracture of
hybrid material bonded interfaces under four-point bending.”
[38] Tan, S.C. and Nuismer, R.J. (1989). A Theory for Progressive Matrix Cracking in
Composite Laminates, Journal of Composite Materials, 23(10): 1029–1047.
62
[39] Tan, S.C. and Perez, J. (1993). Progressive Failure of Laminated Composites with a Hole
under Compressive Loading, Journal of Reinforced Plastics and Composites, 12(10): 1043–
1057.
[40] Tan, S.C. (1991). A Progressive Failure Model for Composite Laminates Containing
Openings, Journal of Composite Materials, 25(5): 556–577.
[41] Barbero, E.J. and Lonetti, P. (2002). An Inelastic Damage Model for Fiber Reinforced
Laminates, Journal of Composite Materials, 36(8): 941–962.
[42] Xiao, Y. and Ishikawa, T. (2002). Bearing Failure in Bolted Composite Joints: Analytical
Tools Development, Advanced Composite Materials, 11(4): 375–391.
[43] Chen, J., Crisfield, M., Kinloch, A.J., Busso, E.P., Matthews, F.L. and Qiu, Y. (1999).
Predicting Progressive Delamination of Composite Material Specimens via Interface Elements,
Mechanics of Advanced Materials and Structures, 6(4): 301–317.
[44] McCarthy, C.T., McCarthy, M.A. and Lawlor, V.P. (2005). Progressive Damage Analysis
of Multi-Bolt Composite Joints with Variable Bolt-Hole Clearances, Composites Part B. 36(4):
290–305. 1962 T. E. TAY ET AL. Downloaded from jcm.sagepub.com at UNIV ALABAMA
LIBRARY/SERIALS on June 13, 2012
[45] Basu, S., Waas, A.M. and Ambur, D.R. (2007). Prediction of Progressive Failure in
Multidirectional Composite Laminated Panels, International Journal of Solids and Structures,
44(9): 2648–2676.
[46] Goswami, S. (2005). A Finite Element Investigation on Progressive Failure Analysis of
Composite Bolted Joints Under Thermal Environment, Journal of Reinforced Plastics and
Composites, 24(2): 161–171.
[47] Riccio, A. (2005). Effects of Geometrical and Material Features on Damage Onset and
Propagation in Single-lap Bolted Composite Joints under Tensile Load: Part II – Numerical
Studies, Journal of Composite Materials, 39(23): 2091–2112.
[48] McGregor, C.J., Vaziri, R., Poursartip, A. and Xiao, X. (2007). Simulation of Progressive
Damage Development in Braided Composite Tubes under Axial Compression, Composites Part
A: Applied Science and Manufacturing, 38(11): 2247–2259.
[49] Tao, J. and Sun, C.T. (1998). Influence of Ply Orientation on Delamination in Composite
Laminates, Journal of Composite Materials, 32(21): 1933–1947.
[50] Laurin, F., Carre` re, N. and Maire, J.F. (2007). A Multiscale Progressive Failure Approach
for Composite Laminates Based on Thermodynamical Viscoelastic and Damage Models,
Composites Part A: Applied Science and Manufacturing, 38(1): 198–209.
[51] Reddy, Y.S.N., Moorthy, C.M.D. and Reddy, J.N. (1995). Non-Linear Progressive Failure
63
Analysis of Laminated Composite Plates, International Journal of Non-Linear Mechanics, 30(5):
629–649.
[52] Pal, P. and Ray, C. (2002). Progressive Failure Analysis of Laminated Composite Plates by
Finite Element Method, Journal of Reinforced Plastics and Composites, 21(16): 1505–1513.
[53] Xiao, Y. and Ishikawa, T. (2002). Bearing Failure in Bolted Composite Joints: Analytical
Tools Development, Advanced Composite Materials, 11(4): 375–391.
[54] Basu, S., Waas, A.M. and Ambur, D.R. (2007). Prediction of Progressive Failure in
Multidirectional Composite Laminated Panels, International Journal of Solids and Structures,
44(9): 2648–2676.
[55] Kweon, J.H., Shin, S.Y. and Choi, J.H. (2007). A Two-Dimensional Progressive Failure
Analysis of Pinned Joints in Unidirectional-Fabric Laminated Composites, Journal of Composite
Materials, 41(17): 2083–2104.
[56] Wolford, G.F. and Hyer, M.W. (2005). Failure Initiation and Progression in Internally-
Pressurized Elliptical Composite Cylinders, Mechanics of Advanced Materials and Structures,
12(6): 437–455.
[57] Reddy, Y.S. and Reddy, J.N. (1993). Three-Dimensional Finite Element Progressive Failure
Analysis of Composite Laminates under Axial Extension, Journal of Composites Technology
and Research, 15(2): 73–87.
[58] Robbins, D.H., Reddy, J.N. and Rostam-Abadi, F. (2005). Layerwise Modeling of
Progressive Damage in Fiber-Reinforced Composite Laminates, International Journal of
Mechanics and Materials in Design, 2(3): 19–36.
[59] Dugdale, D.S., Yielding of steel sheets containing slits. Journal of the Mechanics and
Physics of Solids 8, 100-104. 1960.
[60] Barenblatt, G., The mathematical theory of equilibrium cracks in brittle fracture. Advances
in Applied Mechanics, 7, 55-129. 1962.
[61] Hillerborg, A., Modeer, M., Petersson, P.E., Analysis of crack formation and crack growth
in concrete by means of fracture mechanics and finite elements Cement and Concrete Research.
6, 773-782. 1976.)
[62] Krueger, R., The virtual crack closure technique: history, approach and applications.
NASA/Contractor Report-2002-211628, 2002.
[63] Turon, A., Camanho, P.P., Costa, J. Davila, C.G., An interface damage model for the
simulation of delamination under variable-mode ratio in composite materials. NASA/Technical
Memorandum 213277, 2004.
64
[64] Turon, A., Camanho, P.P., Costa, J. Davila, C.G., A damage model for the simulation of
delamination in advanced composites under variable-mode loading. Mechanics of Materials,
submitted, 2005.
[65] Jitendra S. Tate, Adekunle T. Akinola, and Dmitri Kabakov, “Bio-based Nanocomposites:
An Alternative to Traditional Composites,” The Journal of Technology Studies, vol. 35 No. 1,
fall 2009.
[66] X. Song, “Vacuum assisted resin transfer molding (VARTM): Model development and
verification.” Ph.D. 3147766, Virginia Polytechnic Institute and State University,
United States – Virginia, 2003.
[67] X. Sun, S. Li, and L. J. Lee, “Mold filling analysis in vacuum-assisted resin transfer
molding. Part I: SCRIMP based on a high-permeable medium,” Polymer Composites, vol. 19,
pp. 807-817, 1998.
[68] R. Chen, C. Dong, Z. Liang, C. Zhang, and B. Wang, “Flow modeling and simulation for
vacuum assisted resin transfer molding process with the equivalent permeability method,”
Polymer Composites, vol. 25, pp. 146-164,2004.
[69] R. Waugh, “Not just a load of hot air: Dream becomes reality as Boeing’s new carbon-fiber
787 Dreamliner heralds a new age of air travel,” 2011.
[70] W. J. Stsaszewski, S. Mahzan, and R. Traynor, “Health monitoring of aerospace composite
structures – Active and passive approach,” Composites Science and Technology, vol. 69, pp.
1678-1685, 2009.
[71] S. S. Kessler and S. M. Spearing, “Design of a piezoelectric-based structural health
monitoring system for damage detection in composite materials,” in Smart Structures and
Materials 2002: Smart Structures and Integrated Systems, 18-21 March 2002, USA, 2002, PP.
86-96.
[72] A. K. Mal, S. Banerjee, and F. Ricci, “Automated structural health monitoring system using
acoustic emission and modal data,” in Health Monitoring and Smart Nondestructive Evaluation
of Structural and Biological Systems III, March 15, 2004 – March 17, 2004, San Diego, CA,
United states, 2004, pp. 1-10.
[73] M. A. Maseras-Gutierrez, W. J. Staszewski, M. S. Found, and K. Worden, “Detection of
impacts in composite materials using piezoceramic sensors and neural networks,” in Smart
Structures and Materials 1998: Smart Structures and Integrated Systems, February 3, 1998 –
February 3, 1998, San Diego, CA, United states, 1998, pp. 491-497.
[74] D. C. Worlton, “Experimental confirmation of lamb waves at megacycle frequencies,”
Journal of Applied Physics, vol. 32, pp. 967-971, 1961.
65
[75] R. Jha, I. Kim, and D. Widana-Gamage, “Investigation of Incident Lamb Wave Parameters
on Detection of Composite Delamination,” ASME Conference Proceedings, vol. 2010, pp. 675-
683, 2010.
[76] K. Diamanti, J. M. Hodgkinson, and C. Soutis, “Detection of Low-velocity Impact Damage
in Composite Plates using Lamb Waves,” Structural Health Monitoring, Vol. 3, pp. 33-41, 2004.
[77] A. J Croxford, P. D. Wilcox, B. W. Drinkwater, and G. Konstantinidis, “Strategies for
guided-wave structural health monitoring,” Proceedings of the Royal Society A: Mathematical,
Physical and Engineering Science, vol. 463, pp. 2961-2981, 2007.
[78] A. Raghavan and C. E. S. Cesnik, “Review of guided-wave structural health monitoring,”
Shock and Vibration Digest, vol. 39, pp. 91-114, 2007.
[79] V. Giurgiutiu, “Lamb Wave Generation with Piezoelectric Wafer Active Sensors for
Structural Health Monitoring,” in Smart Structures and Materials 2003: Smart Structures and
Integrated Systems, March 3, 2003 – March 6, 2003, San Diego, CA, United states, 2003, pp.
111-122.
[80] J. R. Wait, G. Park, H. Sohn, and C. R. Farrar, “Plate damage identification using wave
propagation and impedance methods,” in Health Monitoring and Smart Nondestructive
Evaluation of Structural and Biological Systems III, March 15, 2004 – March 17, 2004, San
Diego, CA, United states, 2004, pp. 53-65.
[81] A. B. Thien, “Pipeline structural health monitoring using macro-fiber composite active
sensors,” University of Cincinnati, 2006
[82] S. Soni, S. Das, and A. Chattopadhyay, “Simulation of damage-features in a lug joint using
guided waves,” Journal of Intelligent Materials Systems and Structures, vol. 20, pp. 1451-64,
2009.
[83] K. Diamanti and C. Soutis, “Structural health monitoring techniques for aircraft composite
structures,” Progress in Aerospace Science, vol. 46, pp. 342-352, 2010.
[84] H. Gao and J. L. Rose, “Ultrasonic sensor placement optimization in structural health
monitoring using evolutionary strategy” in American Institute of Physics, 2006.
[85] W. Huyer and A. Neumaier, “SNOBFIT – Stable noisy optimization by branch and fit,”
ACM Transactions on Mathematical Software, vol. 35, 2008.
[86] R. F. Guratzsch, S. Mahadevan, C. L. Pettit, and M. M. Derriso, “Sensor placement
optimization for shm systems under uncertainty,” 46th
AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics and Materials Conference, April 18, 2005 – April 21, 2005, Austin, TX,
United states, 2005, pp. 5182-5188.
[87] “NASA technical report,” The University of Alabama, 2012.
66
APPENDIX A: LabVIEW program
LabVIEW program developed by my project colleague A. Nagabhushana, and used to write
arbitrary waveform generator, to generate Hanning waveform input.
LabVIEW program developed to read from DAQ