PRABHUGOUD MOHANRAJ. Damage Assessment in Composites using
Fiber Bragg Grating Sensors. (Under the direction of Assistant Professor Kara J. Peters).
This dissertation develops a methodology to assess damage in composites using
fiber Bragg grating (FBG) strain sensors. First, a strain-transfer model using the finite
element (FE) method is developed to simulate the response of an embedded FBG to the
applied loading. This FE model is also able to calculate birefringence in the FBG due to
applied transverse loads. The model is validated considering the two-dimensional problem
of diametrical compression of polarization-maintaining fibers. A modified T-matrix model
is then formulated to simulate the response of an embedded FBG due to an applied axial
strain field. The response of FBGs surface mounted on PMMA and two-dimensional woven
composites subjected to multiple low velocity impacts is experimentally investigated. The
complex spectral response is related to the residual strains after impact in the PMMA
specimens and the surface strain to failure in the two-dimensional woven composites. The
feasibility of using FBGs to measure internal strain in woven composites during damage
by
Mohanraj Prabhugoud
A dissertation submitted to the Graduate Faculty of North Carolina State University
in partial satisfaction of the requirements for the Degree of
Doctor of Philosophy
Mechanical Engineering
Raleigh
2005
Approved By:
Dr. Eric C. Klang Dr. Tasnim Hassan
Dr. Kara J. Peters Dr. Mohammed A. Zikry
Biography
MOHANRAJ PRABHUGOUD was born in Bangalore, India. He was brought up in
Ban-galore, where he attended Cambridge School for his secondary education. He majored in
Science at S. Nijalingappa College. He opted for a career in Engineering and attended M.
S. Ramaiah Institute of Technology, Bangalore, from September 1995 to September 1999.
He graduated with a Bachelor of Technology degee in Mechanical Engineering.
In January 2001 he moved to the United States for his graduate studies at North
Carolina State University, Raleigh, NC. He obtained a Master of Science in Mechanical
Engineering in 2002. In Spring 2003, he started working towards his Ph.D. in Mechanical
Acknowledgements
I would like to thank Dr. Kara Peters for her support, encouragement and invaluable
guidance throughout the course of this work. She was one of the first people to formally
introduce me to the wonderful world of fiber optics and has since motivated me to pursue
a research career in this field.
I am extremely thankful for the financial support provided by National Science
Foundation, which enabled me to pursue this doctoral program. I would like to thank
Advanced Composites for donating the composite twill and Corning Inc., for donating the
gratings used for the experiments.
I would like to specially thank Dr. Mohammed Zikry for his valuable inputs for
the plexiglass experiments. I would like to thank James D Pearson for being available at all
times (with untiring enthusiasm) and helping me out with the experiments especially using
the drop tower.
I would like to thank my committee members, Dr. Eric Klang, and Dr. Tasnim
Hassan for serving on my committee. I would like to extend my thanks to Sharon Kiesel,
Mike Sitar, Luke Davis and Sarah Wilson for their help with the experiments. I would also
like to thank Rufus (Skip) L. Richardson and Mike Breedlove for their help in machining
components required for conducting the experiments.
Graduate study at NCSU has been a wonderful experience, during which I made
a number of dear friends. I would like to thank all my friends for their support and help.
I would like to thank my family members for their moral support and encouragement. My
Contents
List of Figures vii
List of Tables xiv
1 Introduction 1
2 Background 3
2.1 FIBER BRAGG GRATINGS . . . 3
2.2 APPLICATION OF BRAGG GRATING STRAIN SENSORS TO COM-POSITES . . . 6
2.3 MODELING OF EMBEDDED BRAGG GRATING STRAIN SENSORS . 7 2.4 MOTIVATION . . . 11
3 Finite Element Model for Embedded Fiber Bragg Grating Sensor 12 3.1 FINITE ELEMENT FORMULATION . . . 13
3.1.1 Overview . . . 14
3.1.2 Calculation of indices of refraction for an element . . . 15
3.1.3 Calculation of propagation constants for a sensor segment . . . 19
3.1.4 Calculation of the sensor response . . . 25
3.2 NUMERICAL EXAMPLES . . . 25
3.2.1 Circular core, step-index fiber . . . 26
3.2.2 PM fibers . . . 28
3.3 SUMMARY . . . 42
4 Modified Transfer Matrix Formulation for Bragg Grating Strain Sensors 45 4.1 INTRODUCTION . . . 45
4.2 REVIEW OF TWO-MODE COUPLING IN BRAGG GRATINGS . . . 46
4.2.1 Coupled mode theory . . . 47
4.2.2 Transfer matrix approximation . . . 49
4.3 T-MATRIX FORMULATION BASED ON MODIFIED PERIOD FUNCTION 50 4.3.1 Uniform gratings . . . 50
4.3.3 Calculation of applied strain . . . 56
4.3.4 Discussion . . . 58
4.4 EXPERIMENTAL VALIDATION . . . 58
4.5 APPLICATION TO PM FIBER . . . 59
4.5.1 Additional comments . . . 61
4.5.2 Elliptical SAP fiber . . . 62
4.6 SUMMARY . . . 64
5 Experimental Investigation of FBG Strain Sensors 68 5.1 INTRODUCTION . . . 68
5.2 BENCHMARK PMMA STUDIES . . . 69
5.2.1 Results . . . 71
5.2.2 Discussion . . . 82
5.3 WOVEN COMPOSITES . . . 83
5.3.1 Surface Mounted Sensors . . . 84
5.3.2 Embedded Sensors . . . 98
5.3.3 Discussion . . . 102
5.4 SUMMARY . . . 104
6 Conclusions 113
References 115
Appendices 122
A Experimental Data for PMMA Specimens 123
List of Figures
2.1 Classification of optical fibers. . . 4
2.2 FBG written into a PM fiber, subjected to multi-axis loading. . . 9
2.3 Typical optical fiber cross-sections: (a) circular core, step-index (no
birefrin-gence), (b) elliptical core, step-index (geometrical birefrinbirefrin-gence), (c) bow-tie
fiber with pre-stressed regions (geometrical and residual stress birefringence). 10
3.1 Schematic of the procedure for calculation of FBG spectral response for a
sensor embedded in a host material system. . . 16
3.2 Discretization of optical fiber into FBG sensor elements. Also shown are
definition of local polarization axesp andq, global polarization axes, X and
Y, and local principle strain axes, 1−2−3. z is the direction of propagation
along the optical fiber. . . 17
3.3 Definition of triangular element with nodal coordinates. . . 23
3.4 Global index ellipse showing variation ofneff for a given mode with rotation
of global axes X −Y. X0 −Y0 corresponds to global propagation axes of
optical fiber. . . 24
3.5 Cross-section of circular core, step-index fiber with applied transverse load.
Core size is exaggerated to show dimensions. . . 27
3.6 Variation of effective index of refraction with normalized frequency for
cir-cular cross-section, step-index optical fiber. Circles represent exact solution. Triangles represent results of coarse mesh simulations and squares represent
results of fine mesh (270 core elements, 452 cladding elements) simulations. 28
3.7 Variation of Bragg wavelength with applied diametrical load for FBG in
circular core, step-index optical fiber. Squares represent result of center
strain approximation. FE result is plotted as a solid line. . . 29
3.8 Geometry of PM fiber types considered: (a) elliptical core fiber; (b) D-fiber;
(c) elliptical core SAP fiber; (d) Bow-Tie fiber; (e) Panda fiber. The slow
and fast axes are also indicated. All dimensions shown inµm. . . 30
3.9 Variation of Bragg wavelength with applied load for FBG in elliptical core
fiber of figure 3.8 (a). Results from both the finite element and center strain
3.10 Variation of Bragg wavelength with applied load for FBG in D-fiber of fig-ure 3.8 (b). Results from both the finite element and center strain solutions
are plotted. (a) γ = 0◦; (b)γ = 90◦. . . 34
3.11 Variation of Bragg wavelength with applied load for FBG in elliptical core
SAP fiber of figure 3.8 (c) with outer diameter d = 125 µm. Results from
both the finite element and center strain solutions are plotted. (a) γ = 0◦;
(b)γ = 90◦. . . 35
3.12 Variation of Bragg wavelength with applied load for FBG in elliptical core
SAP fiber of figure 3.8 (c) with outer diameter d = 80 µm. Results from
both the finite element and center strain solutions are plotted. (a) γ = 0◦;
(b)γ = 90◦; (c) γ= 36◦; and (d)γ = 72◦. . . 36
3.13 Variation of Bragg wavelength with applied load for FBG in bow-tie fiber of figure 3.8 (d). Results from both the finite element and center strain solutions
are plotted. (a) γ = 0◦; (b)γ = 90◦; (c)γ = 36◦; and (d)γ = 72◦. . . 37
3.14 Variation of Bragg wavelength with applied load for FBG in panda fiber of figure 3.8 (e). Results from both the finite element and center strain solutions
are plotted. (a) γ = 0◦; (b)γ = 90◦; (c)γ = 36◦; and (d)γ = 72◦. . . 39
3.15 Variation of the orientation of global index ellipse with applied load for FBG in panda fiber of figure 3.8 (e). Results from both the finite element and
center strain solutions are plotted. (a) γ = 0◦; (b)γ = 90◦; (c)γ = 36◦; and
(d)γ = 72◦. . . 40
3.16 Variation of Bragg wavelength with applied load for multiple orientations of
applied loading angle,γ, relative to geometrical axes of Panda fiber. Results
from both the finite element and center strain solutions are plotted. . . 41
3.17 Variation of Bragg wavelength with applied temperature (a) elliptical core
fiber (d = 125 µm); (b) bow-tie fiber; (c) panda fiber. Results for each axis
calculated using the center strain approximation (CSA) are also plotted. . . 43
4.1 Response of a uniform grating subjected to an applied (a) linear strain field.
(b) Reflected spectra due to the applied linear strain field along the grating length: dashed line is Runge-Kutta simulation of the coupled mode
equa-tions; solid line is T-matrix simulation using Λ(z); dotted line is T-matrix
simulation using ˜Λ(z). . . 52
4.2 Response of a uniform grating subjected to an applied (a) quadratic strain
field. (b) Reflected spectra due to the applied quadratic strain field along the grating length: dashed line is Runge-Kutta simulation of the coupled
mode equations; solid line is T-matrix simulation using Λ(z); dotted line is
T-matrix simulation using ˜Λ(z). . . 53
4.3 Response of a uniform grating subjected to an applied (a) linear strain field
with high strain gradient. (b) Reflected spectra due to the applied linear strain field along the grating length: dashed line is Runge-Kutta simulation
of the coupled mode equations; solid line is T-matrix simulation using Λ(z);
4.4 Response of a uniform grating subjected to an applied (a) exponentially varying strain field with high strain gradient. (b) Reflected spectra due to the applied exponentially varying strain field along the grating length: dashed line is Runge-Kutta simulation of the coupled mode equations; solid line is
T-matrix simulation using Λ(z); dotted line is T-matrix simulation using ˜Λ(z). 55
4.5 Response of a chirped grating: (a) ²0(z) = 1000×10−6 and ²(z) = (1×
10−10/mm)z; (b) ²0(z) = 1000×10−6 and ²(z) = (1×10−16/mm2)z2; and
(c) ²0(z) = (1×10−10/mm)z and ²(z) = (1×10−16/mm2)z2. Runge-Kutta
simulations are plotted as dashed lines and T-matrix simulations using ˜Λ(z)
are plotted as dotted lines. . . 57
4.6 Reflection spectra for a grating surface mounted on an aluminum diaphragm.
(a) Experimentally measured spectra (from Chang and Vohra [84]), (b) Spec-tra simulated including sSpec-train gradient in the effective period formulation, and (c) Spectra simulated without strain gradient in the effective period
for-mulation. . . 60
4.7 Cross-section of PM fiber with applied transverse load. Core size is
exagger-ated to show dimensions. . . 63
4.8 Azimuthal variation of global effective refractive index forP = 10 N/mm and
γ = 54◦ (squares). Ellipse oriented at θ with major and minor axes nmaxeff
and nmin
eff is plotted as a solid line. . . 65
4.9 (a) Variation of Bragg wavelength with applied diametrical load atγ = 36◦
for PM fiber. Squares represent result of center strain approximation, while FE results are plotted as solid line. (b) Variation of the orientation of global
index ellipse with the applied load. The variation ofθis plotted over a greater
range ofP to demonstrate asymptotic behavior. . . 66
4.10 (a) Variation of Bragg wavelength with applied diametrical load atγ = 72◦
for PM fiber. Squares represent result of center strain approximation, while FE results are plotted as solid line. (b) Variation of the orientation of global
index ellipse with the applied load. The variation ofθis plotted over a greater
range ofP to demonstrate asymptotic behavior. . . 66
4.11 (a) Linear variation of applied diametrical load along the length of the FBG. FBG spectra obtained before (dashed line) and after (solid line) application
of transverse load at (b) γ = 36◦ and (c)γ = 72◦. . . 67
5.1 (a) PMMA specimen with surface mounted FBG. (b) Schematic
representa-tion of PMMA specimen. Also indicated are the locarepresenta-tion of the optical fibers
and FBGs for each specimen and the impact point. All dimensions in mm. 74
5.2 Instrumentation for FBG sensor. . . 75
5.3 Drop tower for impact loading of the specimens . . . 75
5.4 Variation of strain with strike # for FBG surface mounted on specimen A58. 76
5.5 Variation of strain with strike # for FBG surface mounted on specimen A59. 76
5.6 Variation of strain with strike # for FBG surface mounted on specimen A69. 77
5.7 Variation of strain with strike # for FBG surface mounted on specimen A71. 77
5.8 Variation of strain with strike # for FBG surface mounted on specimen A73. 78
5.10 Variation of strain with strike # for FBG surface mounted on specimen A79. 79 5.11 Measured transmitted spectra of a FBG surface mounted on specimen A81
at two different load levels. (a) 300 lbs, maximum load; (b) 300 lbs, zero
load; (c) 1000 lbs, maximum load; (d) 1000 lbs, zero load . . . 80
5.12 Variation of strain with load for an FBG surface mounted on specimen A80. 81
5.13 Variation of strain with load for an FBG surface mounted on specimen A81. 82
5.14 Variation of the maximum residual strain with the radius for 90◦ and 45◦
orientations. Solid and dashed lines are the quadratic fit for 90◦ and 45◦
orientations respectively. . . 83
5.15 Curing cycle for the fabrication of 2D woven composites. . . 86
5.16 Photograph of two-dimensional woven composite specimen (a) before and (b)
after impact. . . 88
5.17 Cross-section of two-dimensional woven composite specimen under a optical
microscope (a) before and (b) after impact. . . 89
5.18 Instrumentation for continuous FBG sensing. Also shown is the 5 Hz
inter-rogator from Micron Optics. . . 90
5.19 Schematic representation of specimen C5 along with optical fiber and impact
location. All dimensions in mm. . . 90
5.20 Global measurements for specimen C5. (a) Variation of peak acceleration
with strike #. (b) Variation of dissipated energy with strike #. . . 91
5.21 Variation of strain with strike # for an FBG surface mounted on specimen C5. Strains computed from both the reflection and transmission spectra are
plotted. . . 91
5.22 Schematic representation of composite specimens. Also shown are the lo-cation of FBGs for C6,C17,C18 and C15 along with impact lolo-cation. All
dimensions in mm. . . 92
5.23 Global measurements for C6. (a) Variation of peak acceleration with strike
#. (b) Variation of dissipated energy with strike #. . . 92
5.24 Variation of strain with strike # for an FBG surface mounted on C6. Strains
computed from both the reflection and transmission spectra are plotted. . . 93
5.25 Global measurements for C17. (a) Variation of peak acceleration with strike
#. (b) Variation of dissipated energy with strike #. . . 93
5.26 Variation of strain with strike # for an FBG surface mounted on C17. Strains
computed from the reflection spectra of (a) End 1 and (b) End 2 are plotted. 94
5.27 Measured reflection spectra from End 1 for a FBG surface mounted on
spec-imen C17 after strikes #: (a) 11, (b) 14, (c) 17, and (d) 20. . . 95
5.28 Measured reflection spectra from End 2 for a FBG surface mounted on
spec-imen C17 after strikes #: (a) 11, (b) 14, (c) 16, and (d) 18. . . 96
5.29 Global measurements for specimen C18. (a) Variation of peak acceleration
with strike #. (b) Variation of dissipated energy with strike #. . . 97
5.30 Variation of strain with strike # for an FBG surface mounted on specimen C18. Strains computed from both the reflection and transmission spectra are
5.31 Measured spectral response of FBG surface mounted on specimen specimen
C18 after strikes #: (a) 6, (b) 9, (c) 10, (d) 14, (e) 15, (f) 16, and (g) 20. . 105
5.32 Global measurements for C15. (a) Variation of peak acceleration with strike
#. (b) Variation of dissipated energy with strike #. . . 106
5.33 Cross-section of two dimensional woven composite specimen embedded with
optical fiber under optical microscope. . . 106
5.34 Schematic representation of specimen C49 along with optical fibers and
im-pact location. All dimensions in mm. . . 107
5.35 Schematic representation of specimen C50 along with optical fibers and
im-pact location. All dimensions in mm. . . 107
5.36 Measured spectra for FBG-A embedded in specimen C49 (a) after embedding,
after strikes #: (b) 1, (c) 6, and (d) 8. . . 108
5.37 Measured spectra for FBG-B embedded in specimen C49 (a) after embedding,
after strikes #: (b) 1, (c) 8, and (d) 13. . . 109
5.38 Measured spectra for FBG-C embedded in specimen C50 (a) after embedding,
after strikes #: (b) 1, (c) 4, and (d) 7. . . 110
5.39 Measured spectra for FBG-D embedded in specimen C50 (a) after embedding,
after strikes #: (b) 1, (c) 4, and (d) 8. . . 111
5.40 Cross-section of two dimensional woven composite specimens under optical microscope: (a) specimen C29, (b) specimen C30, (c) specimen C32, and (d)
specimen C34. . . 112
A.1 Measured spectral response of FBG surface mounted on A58 for impact
strikes 1-5. . . 124
A.2 Measured spectral response of FBG surface mounted on A59 for impact
strikes 1-7. . . 125
A.3 Measured spectral response of FBG surface mounted on A59 for impact
strikes 8-15. . . 126
A.4 Measured spectral response of FBG surface mounted on A59 for impact
strikes 16-19. . . 127
A.5 Measured spectral response of FBG surface mounted on A69 for impact
strikes 1-6. . . 128
A.6 Measured spectral response of FBG surface mounted on A69 for impact
strikes 7-14. . . 129
A.7 Measured spectral response of FBG surface mounted on A69 for impact
strikes 15-23. . . 130
A.8 Measured spectral response of FBG surface mounted on A69 for impact
strikes 23-27. . . 131
A.9 Measured spectral response of FBG surface mounted on A69 at the end of
impact loading. . . 132
A.10 Measured spectral response of FBG surface mounted on A71 for impact
strikes 1-4. . . 133
A.11 Measured spectral response of FBG surface mounted on A73 for impact
A.12 Measured spectral response of FBG surface mounted on A73 for impact
strikes 8-16. . . 135
A.13 Measured spectral response of FBG surface mounted on A73 for impact
strikes 17-25. . . 136
A.14 Measured spectral response of FBG surface mounted on A73 for impact
strikes 26-34. . . 137
A.15 Measured spectral response of FBG surface mounted on A73 for impact
strikes 35-36. . . 138
A.16 Measured spectral response of FBG surface mounted on A75 for impact
strikes 1-4. . . 139
A.17 Measured spectral response of FBG surface mounted on A79 for impact
strikes 1-7. . . 140
A.18 Measured spectral response of FBG surface mounted on A79 for impact
strikes 8-16. . . 141
A.19 Measured spectral response of FBG surface mounted on A79 for impact
strikes 17-20. . . 142
A.20 PMMA specimens surface mounted with FBGs. . . 143
A.21 Measured spectral response of FBG surface mounted on A80 for a static
loading from 0-700 lbs. . . 144
A.22 Measured spectral response of FBG surface mounted on A80 for a static
loading from 800-1300 lbs. . . 145
A.23 Measured spectral response of FBG surface mounted on A81 for a static loading from 0-400L lbs. Notation L and U stand for loading and unloading
respectively. . . 146
A.24 Measured spectral response of FBG surface mounted on A81 for a static load-ing from 400U-800U lbs. Notation L and U stand for loadload-ing and unloadload-ing
respectively. . . 147
A.25 Measured spectral response of FBG surface mounted on A81 for a static load-ing from 900L-1100U lbs. Notation L and U stand for loadload-ing and unloadload-ing
respectively. . . 148
B.1 Dimensions of the mold used to fabricate the composite specimens. All
di-mensions in inches. . . 150
B.2 Progression of damage in a typical two-dimensional woven composite specimen.151 B.3 Measured spectral response of FBG surface mounted on C17 for impact
strikes 1-7. . . 152
B.4 Measured spectral response of FBG surface mounted on C17 for impact
strikes 8-13. . . 153
B.5 Measured spectral response of FBG surface mounted on C17 for impact
strikes 14-17 . . . 154
B.6 Measured spectral response of FBG surface mounted on C17 for impact
strikes 18-20. . . 155
B.7 Measured spectral response of FBG surface mounted on C18 for impact
B.8 Measured spectral response of FBG surface mounted on C18 for impact
strikes 8-16. . . 157
B.9 Measured spectral response of FBG surface mounted on C18 for impact
strikes 17-22. . . 158
B.10 Measured spectral response of FBG embedded in C15 for impact strikes 1-7. 159 B.11 Measured spectral response of FBG embedded in C15 for impact strikes 8-14. 160 B.12 Measured spectral response of FBG-A embedded in C49 for impact strikes 1-4.161 B.13 Measured spectral response of FBG-A embedded in C49 for impact strikes
5-13. . . 162
B.14 Measured spectral response of FBG-B embedded in C49 for impact strikes 1-4.163 B.15 Measured spectral response of FBG-B embedded in C49 for impact strikes
5-13. . . 164
B.16 Measured spectral response of FBG-B embedded in C49 for impact strikes
14-18. . . 165
B.17 Measured spectral response of FBG-C embedded in C50 for impact strikes 1-4.166 B.18 Measured spectral response of FBG-C embedded in C50 for impact strikes
5-12. . . 167
B.19 Measured spectral response of FBG-C embedded in C50 for impact strikes
13-16. . . 168
B.20 Measured spectral response of FBG-C embedded in C50 for impact strikes
17-24. . . 169
B.21 Measured spectral response of FBG-D embedded in C50 for impact strikes 1-4.170 B.22 Measured spectral response of FBG-D embedded in C50 for impact strikes
5-10. . . 171
B.23 Measured spectral response of FBG-D embedded in C50 for impact strikes
11-18. . . 172
B.24 Measured spectral response of FBG-D embedded in C50 for impact strikes
List of Tables
3.1 Parameters of scalar wave equation for fundamental LP modes. . . 20
3.2 Parameters of circular core, step-index fiber. . . 26
3.3 Material Properties of Modeled Optical Fibers. . . 31
3.4 Number of Elements Meshed for PM Fiber Types. . . 32
3.5 Comparison of Sensitivity to Applied Load for PM Fiber Types. . . 42
3.6 Comparison of Sensitivity to Applied Temperature for PM Fiber Types. . . 42
4.1 Parameters of PM fiber. . . 64
5.1 Configurations of PMMA Specimens . . . 70
5.2 Configurations of 2D woven composite specimens . . . 84
5.3 Configurations of 2D woven composite embedded with polyimide coated fibers without FBGs. . . 100
Chapter 1
Introduction
The goal of a damage assessment system is to measure failure initiation and
pro-gression in a structure. From this one estimates the remaining life time of the damaged
structure. This subsystem therefore forms an integral part of long-term structural health
monitoring system. The concept of damage assessment is particularly important for fiber
reinforced composites subjected to low velocity impacts due to the different failure
mech-anisms involved. The interaction of these various mechmech-anisms during the progression of
damage poses a challenge to the assessment system.
The first step towards building an effective damage assessment system is the choice
of the sensor(s). Due to its ability to measure strain distributions as well as with
multiplex-ing capabilities fiber Bragg gratmultiplex-ing (FBG) sensor is an effective tool for a damage assessment
system. In addition, FBGs can be embedded into many host structures to measure internal
damage. After selecting the sensor, the next step is to model the strain transfer from the
host material to the sensor and the sensor response due to the physical change in the host
structure incorporating this transfer. One of the challenges posed by a damage assessment
system is to be able to distinguish between sensor failures and host material failures from
the complex sensor response.
This dissertation eliminates the need for a reduced strain transfer model by using
a finite element (FE) formulation to model the response of an embedded FBG sensor.
A linear strain-optic effect is applied to model the local opto-mechanical response of the
FBG sensors are surfaced mounted and embedded in two-dimensional woven composites.
The developed FE model for the sensor response and the analysis of experimental data is
important towards (i) better understanding the sensor response to be able to distinguish
between sensor failure and host material failure and (ii) integrating into existing FE models
for structural behavior.
This thesis is organized as follows: Chapter 2 presents a literature review of existing
models for strain transfer and motivations for developing the FE formulation. Chapter 3
presents the FE formulation for the complex strain transfer using a linear strain-optic
effect. The model is applied to a two-dimensional problem of diametrical compression of
FBGs written in PM fibers for validation. Chapter 4 presents the spectral response of
the FBG sensor using a modified T-matrix model including the effects of strain gradients.
Chapter 5 presents experimental results and analysis of the FBG sensor response when
surface mounted and embedded in two-dimensional woven composites. We also present
the results of FBG’s surface mounted on PMMA specimens. Finally Chapter 6 presents
Chapter 2
Background
2.1
FIBER BRAGG GRATINGS
Figure 2.1 shows a classification of optical fibers based on the diameter of the core
into single mode or multi mode fibers. For small core diameters (4-10µm) the fiber supports
only the first mode. The diameter of the claddding is typically 125µm. Single mode fibers
are further classified into circular core and polarization maintaining fibers. In polarization
maintaining fibers the first linearly polarized mode, LP01, is split into two modes each
having a specific propagation constant. Thus two polarization axes are established. PM
fibers are further classified as low birefringent and high birefringent fibers based on whether
the birefringence is due to geometry or a stress applying part (SAP) [18].
The first fiber optic sensor was introduced by Butter and Hocker in 1978 [1]. They
used an interferometric technique to measure the strain on the surface of a cantilever beam.
They related the change in phase of a lightwave propagating through the optical fiber, ∆φ,
to the axial strain applied to the fiber by
∆ ={β−1
2βn
2[(1−ν)p
12−νp11]}²L (2.1)
whereβ =nk0 is the propagation constant,nis the effective refractive index,k0 = 2π/λis
the wavenumber, λ is the wavelength of light, L is the sensing length of the optical fiber,
Optical fibers
Multi-mode Single-mode
Circular core PM fiber
Low birefringent High birefringent Elliptical core
D-fiber
Panda fiber
Figure 2.1: Classification of optical fibers.
strain-optic effect one arrives at the Butter and Hocker formulation. The coefficients p11
andp12are the strain optic coefficients for the optical fiber. Later Hocker, showed that the
sensitivity to temperature is higher than the pressure sensitivity [2].
In 1988, Bertholds and Dandliker developed a technique to experimentally
deter-mine the strain optic coefficients for an isotropic, mechanically homogeneous optical fiber
[3]. They applied axial strain to the fiber and used (2.1) to compute the difference in strain
optic coefficients [5]. They applied torsional strain to the fiber and related the change in
polarization of light to the difference in strain optic coefficients [4]. Using these two
inde-pendent measurements they computed the individual strain optic coefficients. The values of
strain-optic coefficients computed arep11= 0.113 andp12= 0.252. These values have been
universally accepted and widely used, including in this thesis. The main disadvantages
to applying with the interferometric techniques to measure strain are the high vibration
sensitivity and the need to count fringes which becomes laborious at high strains.
The formation of permanent gratings in an optical fiber was first demonstrated
by Hill et al. in 1978 at the Canadian Communications Research Centre (CRC) [8]. This
achievement was an outgrowth of research on the nonlinear properties of germania-doped
silica fibers. Currently, these gratings are used extensively as filters for telecommunication
purposes and as strain sensors. A Bragg grating is a periodic modulation of the core index
of refraction along a segment of optical fiber. It is formed usually by means of exposure to
an interference pattern of intense ultraviolet light (' 245 nm). The optical back-reflected
occurs at the Bragg wavelength λB, which can be related to the effective core refractive
index, n, and the period of the index modulation, Λ, by the relation [6]:
λB= 2nΛ (2.2)
This Bragg wavelength will shift with changes to either n or Λ, thus measuring the
wave-length of this narrow-band spectrum will determine the strain or temperature to which the
optical fiber is subjected. The application of these gratings to measure axial strain was
demonstrated by Measures in 1992 [6]. Measures formulated the change in Bragg
wave-length, ∆λB, to the applied axial strain,², by the relation
∆λB
λB
= (1−pe)² (2.3)
wherepe= 0.22 is the effective photoelastic constant. This linear relation has been widely
used to measure axial strain. Since then, Bragg gratings have been used for many
applica-tions including aircraft health monitoring [19], and spacecraft [12].
Bragg gratings can be classified based on their length as short period (≤500 nm)
and long period gratings. From the modeling point of view, in a short period grating the
coupling is between a forward propagating core mode and a backward propagating core
mode. In a long period grating the coupling is between a forward propagating core mode
and a forward propagating cladding mode [20, 21]. For a review of the formation, modeling
and applications of fiber Bragg gratings see Kashyap [9] and Othonoset al. [10]. For more
specific reviews, Hillet al. from CRC present an excellent review on the formation of Bragg
gratings in [7]; Kerseyet al. from the Naval Research Laboratory review the intrumentation
for Bragg gratings sensors in [17]; and Erdogan reviews the modeling of Bragg grating sensor
response in [13].
To model the Bragg grating sensor response, Erdogan first assumes light to be an
electromagnetic wave propagating through a weakly guiding waveguide (optical fiber) [13].
One notes that this approximation is valid for all optical fibers with a difference in core and
cladding refractive index of less than 2%. Using the above approximation, one solves the
wave equation to obtain the guided electromagnetic fields in the optical fiber. An important
point to note is that the exact solution to the wave equation exists in terms of modified
Bessel functions due to the geometrical symmetry of the fiber. The electromagnetic fields
are defined in terms of modified Bessel function of first kind, and in the cladding in terms
as a perturbation to the effective refractive index,nof the guided mode(s) as [14]:
δneff =δneff(z){1 +νcos[2π
Λz+φ(z)]} (2.4)
whereδneff is the “dc”index change spatially averaged over a grating period, ν is the fringe
visibility of the index change, andφ(z) describes the grating chirp. This perturbation causes
the forward propagating core mode to couple to the backward propagating core mode. This
coupling is maximum at the Bragg wavelength where in energy from the forward propagating
core mode is transfered to the backward propagating core mode. The coupling between the
modes is well described by the coupled-mode theory [15, 16]. The spectral response of the
Bragg grating sensor can then be derived using coupled-mode theory [13]. This modeling
approach allows one to consider the response of the Bragg grating to more general strain
distributions than the constant axial strain assumed in (2.3).
2.2
APPLICATION OF BRAGG GRATING STRAIN
SEN-SORS TO COMPOSITES
McKenzie et al. used optical fiber sensors to study the feasibility of optical fiber
sensors for health monitoring of bonded repair systems [25]. They surface mounted FBGs
to an aluminum-boron epoxy patch to monitor fatigue crack growth and used the finite
element method to determine the optimal placement of sensors. This work demonstrates
that FBGs can be used to detect crack growth. The problems encountered were debonding
of the fibers and a distortion in the response of FBGs due to strong strain gradients.
Murukeshan et al. used fiber Bragg gratings for cure monitoring of composites
[26]. They considered different composite material systems embedded with FBGs. The
results demonstrated the use of FBGs to measure residual strain during curing and also the
linear response of the sensor to bending. An average strain was computed from the spectral
response of the FBG, however an attempt to analyze sensor response was not made.
Takedaet al. embedded Bragg gratings in Carbon Fiber Reinforced Plastic (CFRP)
laminates to detect transverse cracks [27]. They performed tensile tests on laminates
em-bedded with FBG’s to induce transverse cracks. They performed both theoretical and
fiber coating on strain transfer from the laminate to the embedded optical fiber by applying
the widely used shear lag theory for the strain transfer to numerically simulate the spectral
response of the embedded grating [31]. The authors also demonstrated the use of small
diameter fiber Bragg grating sensors to detect transverse crack and delamination in CFRP
laminates [33, 34].
Kuang et al. embedded Bragg gratings in advanced composite materials and fiber
metal laminates (FML) [28]. They measured the residual stress during fabrication and
attributed the split in the Bragg reflection peak to transverse loading. Afterwards they
subjected the FML’s to tensile loading and observed an increase in intensity of one Bragg
peak with respect to one another. They called this “wavelength hopping”phenomenon.
This wavelength hopping phenomenon could be due to stress relief during loading. One
observes high strain gradients induced during loading from the spectral response. However,
a detailed analysis is lacking in [28]. They also studied the spectral response of FBG’s,
embedded in FML’s, subjected to low velocity impacts [29, 32]. They observed spectrum
distortion during impact loading and also a drop in intensity. They measured the S-shape
trend in the shift in Bragg peak with strike number. A similar observation was made by
Guemes and Men´endez [30]. Kuang et al. also demonstrated the use of FBG’s, embedded
in FML’s, to monitor damage induced during cyclic loading [35]. Fiber Bragg gratings have
also been used to measure internal strain in textile composites [36, 37, 38].
2.3
MODELING OF EMBEDDED BRAGG GRATING STRAIN
SENSORS
As mentioned above, optical fiber Bragg grating (FBG) sensors have been widely
embedded in composite material systems for the measurement of curing stresses,
interlam-ina stresses, delaminterlam-ination, crack growth, and other phenomena [22]. Their unique ability to
be embedded within fiber-reinforced composites with a minimum perturbation to the
sur-rounding host material makes them attractive for the above applications. One particular
field of current interest is long-term health monitoring of composite structures, for
exam-ple aircraft or FRP reinforced concrete structures. Therefore, a clear understanding of the
A variety of models have been developed for the stress transfer in isotropic
ma-terials embedded with optical fibers [39, 40, 41]. The goal of such models is to calculate
the strain in the optical fiber, and hence the sensor response, due to loading applied to the
host material system. Koll´ar and Van Steenkiste developed a model of the strain transfer
between a laminated composite and an optical fiber embedded between two laminae [42].
The optical fiber is treated as an elliptical inclusion in the composite laminate cross-section
and the average fiber strain due to the surrounding strain in the laminate is calculated
analytically. Later, Prabhugoud and Peters modeled a unidirectional composite by
apply-ing a combination of the finite element method and optimal shear-lag theory [43]. While
applying varying levels of detail to the host structure, each of these models considers the
optical fiber as an isotropic or orthotropic, homogeneous fiber with constant strain across
the cross-section of the fiber. Although this assumption produces excellent results for the
mechanical response of embedded fibers [44], it is not evident that the same is true for the
optical response of the FBG due to the optical non-homogeneities of the optical fiber.
Once the strain field applied to the optical fiber is known, the shift in Bragg
wavelength of the FBG, ∆λB, is calculated as [6],
∆λB = ∆(2Λneff) = 2Λ∆neff + 2neff∆Λ (2.5)
where Λ is the period of the grating and ∆neffis the change in effective refractive index of the
fiber cross-section due to the applied strain. For example, in the case of an unconstrained
fiber subjected to an axial load,
∆λB=λB²z
h
1−n
2 eff
2 {p11−ν(p11+p12)}
i
(2.6)
where ²z is the axial strain in the fiber, p11 and p12 are photoelastic constants for silica,
and ν is the Poisson’s ratio of the fiber [1].
Although the term ∆Λ in (2.5) can be directly related to the resulting axial strain
in the FBG, the term ∆neff is more difficult to calculate. This change is due both to
geometrical changes in the cross-section and the strain-optic effect. In addition, induced
birefringence in the fiber separates the single Bragg reflection peak into two peaks [45].
This birefringence is highly dependent on the index of refraction and material distribution
throughout the fiber cross-section. Examples of typical optical fiber cross-sections are shown
in figure 2.3. The circular and elliptical step-index cross-sections shown in figure 2.3(a)
P
3
P
1
P
2
Figure 2.2: FBG written into a PM fiber, subjected to multi-axis loading.
ratio νcore = νclad. Optically, they are non-homogeneous with two index of refraction
regions: ncore in the core and nclad in the cladding. Whereas the circular fiber propagates
light at one propagation constant, β1, per mode, the elliptical fiber propagates modes
along two orthogonal axes with two different propagation constants, β1 and β2. The third
optical fiber, referred to as the polarization maintaining (PM) fiber, shown in figure 2.3(c),
includes regions of a separate material (and is therefore both mechanically and optically
non-homogeneous) called the stress applying part (SAP). The purpose of these regions is
to provide stress to the fiber core as the fiber is drawn, inducing residual birefringence due
to the thermal expansion mismatch of the silica and SAP. Several other configurations for
the PM fiber exist other than the bow-tie form shown in figure 2.3(c).
To calculate ∆neff due to an applied strain field, Kimet al. considered the fibers
shown in figure 2.3 to be optically and mechanically homogeneous with isotropic or
transver-sly isotropic material properties [46]. Based on the model of an inclusion in a composite
laminate, the remote strains were analytically linked to the principle strains in the fiber
core and ∆β in the direction of propagation calculated. As in later models, the
assump-tion is made that most of the energy of the fundamental mode propagating in the fiber is
contained in the core, therefore the principle strains at the center of the fiber are sufficient
to estimate ∆neff. In addition, none of the above models account for the birefringence due
(a) (b) (c) n , E
clad clad
n , Ecore core
n , E
clad clad
n , Ecore core
n , E
clad clad
n
Ecorecore n E
SAP
SAP
Figure 2.3: Typical optical fiber cross-sections: (a) circular core, step-index (no birefrin-gence), (b) elliptical core, step-index (geometrical birefrinbirefrin-gence), (c) bow-tie fiber with pre-stressed regions (geometrical and residual stress birefringence).
the strain-optic effect. Sirkis related the change in Bragg wavelengths, ∆λB,1 and ∆λB,2,
to the principle strains at the center of the core,²1,²2, and ²3, as [47]
∆λB,1 λB,1
=²1−n
2
o
2 (p11²2+p12²3+p12²1) (2.7)
∆λB,2 λB,2
=²1−n
2
o
2 (p12²2+p11²3+p12²1)
where ²2 and ²3 are in the plane of the fiber cross-section, ²1 is in the axial direction and
no is the effective refractive index of the fundamental mode. Wagreichet al. demonstrated
the linear dependence of ∆λB,1 and ∆λB,2 on the applied load for a circular core fiber
(figure 2.3(a)) under diametrical compression [45]. Lawrence et al. [48] modeled the
me-chanical non-homogeneities in a PM fiber with elliptical SAP using a finite element analysis
to calculate the strains at the center of the fiber due to applied transverse loading. Their
main goal was to calculate the calibration matrix for the transverse sensitivity of a FBG
sensor in a PM fiber. Later Bosiaet al. [49] also modeled the mechanical non-homogenieties
in a bowtie type PM fiber using finite element analysis to calculate the principle strains at
the center of the core and hence the shift in Bragg wavelength due to applied transverse
loading using (2.7). Both the experimental and numerical studies of [48] and [49]
demon-strated that for a PM fiber the shift in Bragg wavelength is nonlinear with transverse load
for certain loading angles. Gafsi and El-Sherif expanded the center strain formulation to
include variations of refractive indices along the axis of the fiber by introducing (2.7) into
the coupled mode equations describing the spectral response of the FBG [50]. However,
and theoretical sensitivity to applied strain.
2.4
MOTIVATION
As described in the above section there exists handful of models to simulate the
response of a FBG subjected to arbitrary loading. One should note that all the models use
the center strain approximation (CSA). None of the models described in the above section
can therefore account for the variation of the refractive index in the cross-section of the
fiber due to the applied strain.
The goal of this thesis is to present a methodology to calculate the response of a
FBG subjected to arbitrary loading as shown in figure 2.2. We formulate a finite element
(FE) model to calculate the birefringence effect due to the applied transverse load. The
linear strain-optic law is assumed to calculate the change in refractive index due to the
ap-plied transverse load. One can integrate this model into existing FE models for structural
behavior. We will consider a diametrical compression of PM fibers to validate the model.
As described in the above section PM fibers are both mechanically and optically
nonho-mogeneous. We will also formulate a modified T-matrix model to simulate the response of
FBG subjected to longitudinal strain. This model considers the effect of strain gradients
in the response and is computationally efficient. Afterwards, the response of FBGs surface
mounted on PMMA and two-dimensional woven composites subjected to low velocity
im-pacts are analyzed. Finally, FBGs are embedded in two-dimensional woven composites to
Chapter 3
Finite Element Model for
Embedded Fiber Bragg Grating
Sensor
The goal of this chapter is to derive a finite element (FE) formulation to predict
the optical response of an embedded FBG sensor as a function of the loading applied to
the host structure. The formulation incorporates both the mechanical and optical
non-homogeneities of the optical fiber. Firstly, the FE fomulation calculates the change in index
of refraction distribution throughout the cross-section of the fiber due to the resulting
mechanical stresses. From the updated index of refraction distribution, the propagation
constants of the fundamental modes, as well as the propagation axes, are obtained. Previous
work by Huang has approached a similar problem for planar waveguides analytically [51].
As can be observed from [51], analytical solutions are only obtainable for a few loading
conditions. In the current formulation, the propagation constants are then introduced into
a discretized version of the coupled mode equations to determine the spectral response of
the FBG.
systems for which extensive FE modeling has already been performed. Examples include
fiber reinforced composites and concrete structures. The current model also allows one to
accurately calculate the sensitivity of the FBG to transverse strains and is applicable to
various fiber types including bowtie and panda PM fibers. The formulation also includes
the effect of rotating polarization axes due to significant strain amplitudes especially when
the fiber is embedded near a stress concentration or failure location.
3.1
FINITE ELEMENT FORMULATION
The propagation of a given guided mode through an optical fiber can be
char-acterized through the mode distribution in the cross-section of the optical fiber and the
propagation constant, β, for a given frequency. The mode propagation constant is related
to the effective index of refraction, neff, for the particular mode through
β = 2π
λ neff (3.1)
whereλis the propagating wavelength [52]. Exact solutions for the propagation
character-istics of optical fibers, obtained by solving wave equations, are limited to relatively simple
geometries (e.g., circular or elliptical cross-sections) with an axisymmetric index of
refrac-tion distriburefrac-tion in the core. Thus, to calculate the propagarefrac-tion characteristics of an optical
fiber with an arbitrary cross-sectional shape or arbitrary variation of refractive index in the
core, cladding, and SAP, one needs to adopt a numerical method such as the finite element
method [53]. Current finite element methods for optical fiber waveguides can be classified
into vector methods and scalar methods.
Vector finite element methods are applicable to all values of refractive index
differ-ence between the core and the cladding. The main disadvantages of these methods are the
large computational effort required and the appearance of spurious modes in the solution.
The spurious modes can be eliminated using a penalty approach [53]. Different variations
of the vector formulation are based on the components of the electric field,E~, or magnetic
field, H~, considered. For example, in the formulation of Yeh et al., the axial components
of E~ and H~ field are considered [54]. All other components are then expressed in terms of
these axial components using Maxwell’s equations. A minimizing functional is obtained by
interface. In the work of Koshiba, the minimizing functional is obtained from the complete ~
E orH~ field satisfying the vector wave equation [53]. Different approaches have also been
proposed to address the open boundary problem for example, applying the FE formulation
to the core and appropriate boundary condition to the core-cladding interface. This reduces
the number of elements required in the cladding to obtain an accurate solution [55, 56].
Scalar finite element methods, on the other hand, are only applicable to weakly
guiding fibers, i.e. for which the variation of the refractive index is negligible over a distance
of one wavelength [52]. However, such an assumption is reasonable for most fibers into which
FBGs are written, within their elastic strain limit. The advantages of a scalar method are
that no spurious solutions appear (since only linearly polarized modes are captured) and
only one component of the E~ or H~ field is considered, reducing the size of the required
system of equations to solve for the propagation constant. For this reason, in this chapter
we derive a sensor element based on a scalar formulation without imposing the assumption
of axisymmetry.
3.1.1 Overview
In the current analysis, the prediction of the FBG spectral response is performed
through the following steps (see figure 3.1):
• The surrounding host composite material and optical fiber sensor are meshed using
a commercial FE package (e.g., ANSYS for the current work). The chosen sensor
mesh is shown in figure 3.2, where the fiber is divided into segments of length ∆z in
the axial direction and each cross-section is meshed using 2D plane stress triangular
elements. An example element is also shown in figure 3.2. For the purpose of later
calculations, this “3D propagation”element is characterized by its stiffness properties,
indices of refraction, and length, ∆z. Axis 1 is along the propagation direction and
coincides with the global propagation axis,z. Axes pand q are the local optical axes
of propagation.
• Using the thermo-mechanical FE model, the nodal displacements are obtained due to
the external applied loads. From the nodal displacements, strain components in each
element are also calculated.
due to the applied strain field is calculated in local optical axes using a linear
strain-optic law (see section 3.1.2). The updated indices of refractions are then transformed
from the local axes to the global structural axes.
• For each sensor segment, the propagation constant for the optical fiber about the
global structural axis is calculated using the optical FE formulation including the
updated index of refraction distribution and the nodal displacements.
• The propagation constant/effective index of refraction at two other angles with respect
to the global structural axis is calculated. From these values the maximum and
mininum propagation constants are calculated for the cross-section as well as the
global optical axes corresponding to these extrema.
• The FBG spectral response is calculated from the local axial strain, effective indices
of refraction, and curvature of each segment using the modified T-matrix method (see
section 3.1.4).
The details of some of these calculations are given below.
3.1.2 Calculation of indices of refraction for an element
Each element is assumed to be optically isotropic with an index of refraction in
the unstressed state ofnoe. The displacement field vector for light propagating through the
element in the 1-direction is given as [46],
{D}=Ap{sp}sin
·
ωt− 2πn
p e
λ x1
¸
+Aq{sq}sin
·
ωt−2πn
q e
λ x1
¸
(3.2)
wheresp andsqare orthogonal unit vectors in the 2-3 plane in the direction of the principle
optical axes, npe and nqe are the element index of refractions about these axes, ω is the
angular frequency of the wave, and Ap and Aq are the amplitudes of the displacement
vector components. We can write the wave equation for this displacement field vector as
[46],
{s} ×({s} ×[B]{D}) + 1 (ne)2{D}
Structural Model (geometry, material properties, loading)
Nodal Displacements
Element Indices of Refraction and Local Optical Axes
ANSYS
Transfer Matrix Formulation Finite Element Formulation
Segment Global Optical Axes and Principle Propagation Constants
Strain - Optic Law
Segment Propagation Constants at Orientations α, β, and γ
FBG Spectral Response Global Index Ellipse
Figure 3.1: Schematic of the procedure for calculation of FBG spectral response for a sensor embedded in a host material system.
where ne = npe or nqe, {s} is the unit vector in the propagation direction, and [B] is the
material dielectric impermeability tensor,
[B] =
B1 B6 B5
B6 B2 B4
B5 B4 B3
(3.4)
Writing{D} in terms of its components {D}= (0, D2, D3), {s}= (1,0,0), and evaluating
(3.3) yields the matrix equation,
B2−1/n2e B4
B4 B3−1/n2e
D2 D3
= 0 (3.5)
The non-trivial solutions to (3.5) npe and nqe are,
1 (np,qe )2
= (B2+B3)±
p
(B2−B3)2+ 4B42
direc
tion of pr
opaga
tion
∆ z
q
1
p
2 3
X
Z
Y
Figure 3.2: Discretization of optical fiber into FBG sensor elements. Also shown are
def-inition of local polarization axes p and q, global polarization axes, X and Y, and local
principle strain axes, 1−2−3. z is the direction of propagation along the optical fiber.
These solutions correspond to the indices of refraction about the principle optical axes p
and q in the 2−3 plane shown in figure 3.2. These axes will be determined later. For an
optically isotropic material, B1 = B2 = B3 = 1/(noe)2, B4 = B5 = B6 = 0. Therefore,
npe =nqe=noe.
Once strain is applied to the element, the dielectric impermeability tensor change
is defined by the linear strain-optic equation,
∆Bi =
6
X
j=1
pij²j (3.7)
where [p] is the strain-optic tensor and the compact notation is used for the strain
compo-nents (²1 =²11, ²2 =²22, ²3 =²33, ²4 =γ23, ²5 =γ31, ²6 =γ12) [46]. Expanding the solution
of (3.6), writing Bi =Bio+ ∆Bi, and applying the isotropic properties to Bio, we find,
1 (np,qe )2
= 1
(no e)
+(∆B2+ ∆B3)
2 ±
1 2
q
For an optically isotopic material, the strain-optic tensor, [p], reduces to,
[p] =
p11 p12 p12 0 0 0 p12 p11 p12 0 0 0 p12 p12 p11 0 0 0 0 0 0
(p11−p22)/2 0 0 0 0 0 0
(p11−p22)/2 0 0 0 0 0 0
(p11−p22)/2
(3.9)
Substituting (3.9) into (3.7) into (3.8) yieldsnpe andnqe for an element in the stressed state.
1 (np,qe )2
= 1
(no e)2
+p12²1+ (p11+2 p12)(²2+²3)
±(p11−p12)
2
q
(²2−²3)2+²24 (3.10)
Kim et al. considered the same formulation for the optical fiber as a single
ho-mogeneous element and derived a linearized form of (3.10) to calculate the sensitivity of
the FBG to transverse strain [46]. To model a polarization maintaining fiber (such as
fig-ure 2.3(c)) they considered the fiber to be initially optically orthotropic. This approach
produces identical results to the center strain approximation of (2.7).
For FBG sensor problems including thermal loading, (3.6) can be expanded to
include a linear thermo-optic effect [46],
∆Bi=Wi∆T+pij(²j−αj∆T) (3.11)
where{α}are the coefficients of thermal expansion of the sensor in the local optical
coordi-nates. For an isotropic sensor (α1=α2 =α3 =α and α4=α5 =α6 = 0). The coefficients
Wi, defined as,
Wi=
Ã
∂Bi ∂T
!
σ=const.
(3.12)
are measured during iso-stress conditions. For an optically isotropic material, the non-zero
coefficients are thus evaluated as,
W1 =W2=W3 = ∂
∂T
Ã
1 (no
e)2
!
=− 2
(no e)3
Ã
∂noe ∂T
!
(3.13)
A typical value for the thermooptic coefficient (∂no
e)/(∂T) for silica is given by Kimet al.
(3.10)) would thus be modified to
1 (np,qeff)2 =
1 (no
e)2
+p12²1+(p11+2 p12)(²2+²3)−(n2o
e)3
Ã
∂noe ∂T
!
± (p11−p12)
2
q
²24+ (²2−²3)2 (3.14)
The angle of orientation of the element principle optical axes are identical to the
principle strain directions [57]. Although the index of refraction is not a true tensor quantity,
it can be represented by an ellipse in the 2-3 plane with the major and minor axes of length
npe and nqe in the principle strain directions [57]. Therefore to calculatene about the global
axesX and Y, we find
nXe = n
p enqe
p
(npecosψe)2+ (nqesinψe)2
nYe = n
p enqe
p
(npesinψe)2+ (nqecosψe)2
(3.15)
whereψe is the angle required to rotate thep axis to theX axis.
3.1.3 Calculation of propagation constants for a sensor segment
Once the index of refraction for each element is known about both the X and Y
axes, the propagation constantsβmax andβmin and the orientation of the principle optical
axes for the complete cross-section must be calculated. The propagation characteristics for
linearly polarized (LP) modes propagating through a waveguide of arbitrary cross-section
and arbitrary variation of refractive index are determined by solving the scalar wave
equa-tion over the cross-secequa-tion of the fiber, here defined as the region Ω,
∂ ∂x
h
pzx∂x∂ (pxΦ(x, y))
i
+ ∂
∂y
h
pzy∂y∂ (pyΦ(x, y))
i
+
(qk20−β2)Φ(x, y) = 0 (3.16)
where the field Φ(x, y) and the coefficients px, pzx,py, pzy, and q are defined in table 3.1
for the fundamental LPx01 andLPy01 modes.
The scalar wave equation defined in (3.16) is derived as follows:
Maxwell’s equations for source free, time harmonic fields are [52]
Table 3.1: Parameters of scalar wave equation for fundamental LP modes.
Mode Φ px pzx py pzy q
LPx01 Ex n2x 1/n2z 1 1 n2x
LPy01 Ey 1 1 n2y 1/n2z n2y
∇ ×H~ =−jω²0[²]E~ (3.18)
∇ ·D~ =²0∇ ·([²r]E~) = 0 (3.19)
∇ ·H~ = 0 (3.20)
where E~(x, y, z) = E~(x, y)e−jβz and H~(x, y, z) = H~(x, y)e−jβz are the electric field and
magnetic fields respectively, ²0 and µ0 are the free space permittivity and permeability
constants, and [²r] is the material permittivity tensor given by
[²r] =
n2x 0
0 0
n2y 0
0
0
n2z
(3.21)
Taking the curl of (3.17) and substituting (3.18) we obtain
∂2Ey
∂x∂y −
∂2Ex
∂y2 −
∂2Ex
∂z2 +
∂2Ez
∂x∂z =k
2
0n2xEx (3.22)
∂2Ex
∂x∂y−
∂2Ey
∂x2 −
∂2Ey
∂z2 +
∂2Ez
∂y∂z =k
2
0n2yEy (3.23)
∂2Ex
∂x∂z −
∂2Ez
∂x2 −
∂2Ez
∂y2 +
∂2Ey
∂y∂z =k
2
0n2zEz (3.24)
wherek0= 2π/λ. Noting that for theLPx01 mode, Ey = 0 and using (3.19) yields
Ez =− j
βn2z ∂ ∂x(n
2
xEx) (3.25)
Substituting (3.25) into (3.22) we obtain
∂ ∂x
h 1
n2z ∂ ∂x(n
2
xEx)
i
+∂
2E
x
∂y2 +k
2
0n2xEx−β2Ex= 0 (3.26)
Similarly for the LPy01 mode,Ex = 0 and
∂ ∂y
h 1
n2z ∂ ∂y(n
2
yEy)
i
+ ∂
2E
y
∂x2 +k
2
Thus, the scalar wave equation allows us to solve for the fields Ex and Ey independently.
Although, the scalar wave equation derived above is similar to the one derived by Koshiba
[53], here we include the gradient of the refractive index in the formulation.
The functional for (3.16) is given by,
F = Z Z Ω δΦ n ∂ ∂x h
pzx(x, y)∂x∂ (px(x, y)Φ(x, y))
i
+
∂ ∂y
h
pzy(x, y)∂y∂ (py(x, y)Φ(x, y))
i
+ (q(x, y)k20−β2)Φ(x, y)
o
dxdy (3.28)
Taking the first variation of (3.28) and reducing, one obtains
δF =
Z Z
Ω
h
pzx ∂ ∂x(pxΦ)
∂
∂x(δΦ) +pzy ∂ ∂y(pyΦ)
∂ ∂y(δΦ)
+(β2−qk02)(δΦ)Φ
i dxdy − Z Γ n δΦ h
pzx∂x∂ (pxΦ) +pzy∂y∂ (pyΦ)
io
dΓ = 0 (3.29)
where Γ is the boundary of the region Ω. Since we are only concerned with propagated
modes, i.e. modes that are fully contained in the optical fiber, we apply the boundary
condition Φ = 0 on Γ. Discretizing the region into triangular elements as shown in figure 3.2
and noting that the coefficients px,pzx,py,pzy, and q are constant for each element, (3.29)
reduces to X e Z Z Ωe h
pezxpex∂Φ ∂x
∂
∂x(δΦ) +p e zypey
∂Φ
∂y ∂ ∂y(δΦ)
+(β2−qek20)(δΦ)Φ
i
dxdy= 0 (3.30)
where the superscript e refers to the values for a given element. We expand Φ in each
element,
Φe= [N1 N2 N3]
Φ1 Φ2 Φ3
={N}T{Φ} (3.31)
whereN1,N2, andN3 are element shape functions and Φ1, Φ2, and Φ3 are nodal values of
Φ. Substituting (3.31) into (3.30), we obtain the global matrix equation,