title:

author:
publisher:
isbn10 | asin:
print isbn13:
ebook isbn13:
language:
subject
publication date:
lcc:
ddc:
subject:

Fiber Optic Sensors for Construction Materials

and Bridges : Proceedings of the International
Workshop On Fiber Optic Sensors for
Construction Materials and Bridges, May 3-6,
1998
Ansari, Farhad.
Taylor & Francis Routledge
1566766710
9781566766715
9780585216072
English
Optical fiber detectors--Congresses, Building
materials--Testing--Congresses, Bridges--testing-Congresses, Strains and stresses-Measurement--Congresses.
1998
TA1815.I59 1998eb
624/.028/7
Optical fiber detectors--Congresses, Building
materials--Testing--Congresses, Bridges--testing-Congresses, Strains and stresses-Measurement--Congresses.

Page iii

Fiber Optic Sensors for Construction Materials and

Bridges
Proceedings of the International Workshop on Fiber Optic
Sensors for Construction Materials and Bridges May 36, 1998
Sponsored by the National Science Foundation

Edited by
Farhad Ansari

Page iv

Fiber Optic Sensors for Construction Materials and Bridges

a TECHNOMIC publication
Published in the Western Hemisphere by
Technomic Publishing Company, Inc.
851 New Holland Avenue, Box 3535
Lancaster, Pennsylvania 17604 U.S.A.
Distributed in the Rest of the World by
Technomic Publishing AG
Missionsstrasse 44
CH-4055 Basel, Switzerland
Copyright 1998 by Technomic Publishing Company, Inc.
All rights reserved
No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by an means, electronic, mechanical, photocopying,
recording, or otherwise, without the prior written permission of the publisher.
Printed in the United States of America
10 9 8 7 6 5 4 2 1
Main entry under title:
Fiber Optic Sensors for Construction Materials and Bridges
A Technomic Publishing Company book
Bibliography: p.
Includes index p. 263
Library of Congress Catalog Card No. 98-60482
ISBN No. 1-56676-671-0

Page v

Table of Contents
Foreword

ix

Preface

xi

Chapter 1: Research and Development Programs

Federal Highway Administration Research Program
in Fiber Optics for the Infrastructure
R. A. Livingston
Summary of New Jersey Department of
Transportation's Need for New Fiber Optic Tools
N. Vitillo

263

Chapter 2: State-of-the-Art
Existing Technologies for Condition Monitoring of
Construction Materials and Bridges
R. E. Green, Jr

17

Monitoring Systems and Civil EngineeringSome

Possibilities for Fibre Optic Sensors
B. Culshaw

29

Using Bragg Grating Sensor Systems in

Construction Materials and Bridges: Perspectives
and Challenges
J. S. Sirkis

44

Chapter 3: Distributed and Multiplexed Sensors

Developments in Optical Techniques for Point and
Distributed Sensing in Large Structures
65
V. Lecoeuche, N. E. Fisher, C. N. Pannell, D. J.
Webb and D. A. Jackson

Page vi

Distributed and Chemical Fiber Optic Sensing and

Installation in Bridges
D. R. Huston and P. L. Fuhr

79

Technology and Applications of Distributed Optical

Fibre Sensors
89
A. H. Hartog
Fiber Optic White Light Distributed Sensor for
Condition Monitoring of Civil Structures
Z. Chen, A. Mendez, Q. Li and F. Ansari

101

Chapter 4: Condition Monitoring of Bridges

Reliability of Optical Fibers and Bragg Grating
Sensors for Bridge Monitoring
U. Sennhauser, R. Bronnimann, P. Mauron and
Ph. M. Nellen

117

A Recent Experience in Bridge Strain Monitoring

with Fiber Grating Sensors
R. Maaskant, A. T. Alavie and R. M. Measures

129

Pulse-Echo Fiber Optic Sensor for Nondestructive

Evaluation of Concrete Bridges
136
X. Chen and F. Ansari
Preliminary Results on the Monitoring of an InService Bridge Using a 32-Channel Fiber Bragg
Grating Sensor System
S. T. Vohra, C. C. Chang, B. A. Danver, B.
Althouse, M. A. Davis and R. Idriss

148

Monitoring and Evaluation of an Interstate

Highway Bridge Using a Network of Optical Fiber
Sensors
159
R. L. Idriss, K. R. White, J. W. Pate, S. T. Vohra,
C. C. Chang, B. A. Danver and M. A. Davis
Distributed Multiaxis Fiber Grating Strain Sensor
Applications for Bridges

E. Udd, W. Schulz, J. Seim, G. McGill and H. M.

168
Laylor
Chapter 5: Sensor Material Interaction and Reliability
Lifetime and Reliability of Embedded Optical
Sensor Fibers
Ph. M. Nellen, A. Frank, P. Mauron and U.
Sennhauser

183

Page vii

Evaluation of Adhesion Behavior of Optical Fibers

for Sensors Embedded in Cementitious Materials 194
W. R. Habel, E. Schulz, G. Kalinka and A. Bismarck
Chapter 6: Sensor Transduction Mechanism and Signal
Recovery
Intensity Fiber Optic Sensors for Civil
Infrastructures
F. Casciati, S. Merlo and G. Zonta

209

A New Signal Recovery Scheme for Fiber Bragg

Grating Sensors
F. Farahi, E. V. Diatzikis, L. A. Ferreira and J. L.
Santos

219

Elastica Fiber Optic Sensors for Structural

Monitoring
K. H. Wanser, K. F. Voss and K. R. Francis

231

Liquid Core Optical Fibers for Crack Detection and

Repairs in Concrete Matrices
243
C. Dry
A Dual Core Forward Time Division Multiplexing
Optical Fiber for Weight-in-Motion Sensing
R. B. Malla, N. W. Garrick, A. Sen and P. Dua
Author Index

251
267

Page ix

Foreword
This book is a compilation of the proceedings of a workshop with focus on fiber
optic sensors for construction materials and bridges. This is the first time an
international forum has been asked to specifically concentrate on a highly
interdisciplinary topic regarding infrastructure. Infrastructure is the
encompassing framework of our daily lives. Bridges, tunnels, seaports, and
highways comprise a major national investment. For this reason, we have
placed emphasis in finding more advanced materials and methods for design of
stronger and more durable structures. Certainly, technology has not stood still.
We also need to find ways to protect this investment. While research is needed
to enhance the attributes of our construction materials, there is also an
urgency for methodologies that could identify structural anomalies, presence of
cracks, onset of failure, estimate the extent of degradation, and locate the
damaged zone for repair. At this point in time, there is no one method that can
provide this information in a practical manner.
It seems that optical fiber sensors have been successfully employed in the
aeronautics and defense industry. It is this widespread utilization of sensors in
other disciplines that has brought about the recent surge in fiber optic
research within the civil engineering discipline. Albeit, revitalization of our
public works infrastructure system requires advancement through innovative
technologies. Employment of distributed sensing systems such as fiber optics in
bridges will result in superior condition monitoring capabilities and life cycle
cost savings. In view of these benefits, it is surprising to see that the civil
infrastructure system has not benefited from this new technology, especially in
bridge monitoring activities. Current research and development activities are
fragmented, dispersed, and not fully devoted to construction materials.
Moreover, these research activities are not coordinated, contain duplications
and are not comprehensive. Coordinated research and development work is
needed to jump-start the transition from military and aerospace to civil
engineering applications.
This workshop is designed to accomplish development of a coordinated re-

Page x

search program in order to guarantee that federal funds are used efficiently. A
coordinated research program sets a well-defined research agenda and
provides recommendations for sustained budget support. A comprehensive
research plan will increase the accountability of research teas, increase their
visibility, and improve their ability to directly involve the industry in R&D
process. This workshop has brought together an interdisciplinary forum of
engineers and scientists in order to formulate a five-year sequential research
program for implementation of fiber optic sensor technology in construction
materials and bridges. I am encouraged to see the great enthusiasm by the
international community. The prime directive is to develop fiber optic sensors
for civil infrastructure systems. In five years from now, the success of this
workshop will be evaluated based on the quality of research products.
DR. JOHN B. SCALZI
CIVIL & MECHANICAL SYSTEMS
NATIONAL SCIENCE FOUNDATION

Page xi

Preface
Fiber optic sensors have the potential to provide true real-time condition
monitoring capabilities for construction materials and bridges in the twenty-first
century. In this regard, the application of these new technologies will have a
significant impact on health, and efficiency of the public works infrastructure
system. However, full adaptation of the fiber optic sensor technology to
construction materials requires research at all levels. This would involve
addressing issues at the structural as well as the material levels. Experience
has shown that the technologies already developed for applications in
aeronautics cannot be directly applied for condition monitoring of bridges.
Physical characteristics and mechanical properties of construction materials, as
well as construction processes involved for embeddment and adhesion of
optical fibers to structural elements are different from those involved in the
aeronautics industry. Heterogeneity and alkalinity of concrete as a host
material poses stringent requirements on the durability of an embedded optical
fiber. Interface mechanics and stress transfer mechanism from the concrete to
fiber and interpretation of signal from such interactions create further
complications.
Other issues correspond to the durability of sensors under adverse processes
involved during the construction of bridges. New construction procedures need
to be developed for placement of such sensory systems. This needs to be done
at the early stages of sensor design process in order to develop practical
systems for use in conjunction with construction materials. Signal interpretation
and data analysis for condition monitoring activities need to be accomplished in
real-time and in a very simplified manner for use by bridge inspectors. Tests
and standards need to be developed as to the measurement parameters,
sensitivity, and geometrical configurations of the fiber optic sensory system.
A coordinated research program should address all of these important issues
relevant to full scale applications in bridges. As discussed in the foreword
section of this book, the primary goal of the workshop is to develop a
coordinated plan for implementation of decisive research programs specifically
designed for

Page xii

applications in civil infrastructure systems. Since this book is being published

just ahead of the workshop, it is difficult to predict the exact outcomes of this
effort. It is planned to establish a comprehensive research program that would
span over the next five years. Four separate technical committees will oversee
the technical activities of the workshop. These committees meet concurrently
during the workshop to discuss application issues and lay out a coordinated
plan for implementation of needed research. Some of the technical issues that
will be discussed for the development of the research plan pertain to the
following topics:
1. Identification of fiber optic sensor types materials and instrumentation
suitable for civil construction materials use, and development of a consistent
database of previous and current projects
2. Development of experimental procedures to determine the capabilities of
optical fiber sensors in detection and measurement of damage and mechanical
perturbations in construction materials
3. Determination of the sensor/material interface mechanics for embedded and
adhered optical fibers through experimental and analytical studies
4. Development of methodologies for signal processing, and interpretation of
real-time data acquired from sensors
5. Determination of sensor integrity under structural loads, construction
processes, and exposure to chemical and ambient conditions
6. Development of experimental procedures for distributed measurements of
cracking, deformations, and strains in bridge structures
7. Development of calibration and measurement methodologies and
standardization of the developed methodologies through working with
appropriate organizations (NIST, ASTM)
8. Implementation of field demonstration projects through full-scale tests of
bridges
This book is organized into six chapters according to the subject matter of the
papers. The subject matter of some of the papers would qualify them for
insertion into more than one section. All the papers included in this book were
peer reviewed and accepted after revisions. The chapters in this book are

arranged in the following order:

1. Research and Development Programs
2. State-of-the-Art
3. Distributed and Multiplexed Sensors
4. Condition Monitoring of Bridges
5. Sensor Material Interaction and Reliability
6. Sensor Transduction Mechanism and Signal Recovery
Development of a comprehensive program in a highly interdisciplinary area
requires careful planning in order to assure that the objectives are
appropriately

Page xiii

prioritized, and they are implemented within the stated time period. It is for
these reasons that the formulation of the coordinated program will require the
work of an interdisciplinary forum of engineers and scientists in order to assist
the National Science Foundation (NSF) in prioritization of research. I am
indebted to my colleagues, members of the advisory and technical committees
of this workshop for their diligent work and for taking the responsibility in
developing the national research plan. The technical committee members
comprise an international group of experts in the field. The advisory panel
members consist of government research program managers and experts in
the field and their responsibility is to coordinate the activities of the four
technical committees. The advisory panel's responsibility is to assure that the
research plans are steered towards civil infrastructure applications. It is my
pleasure to list the names and affiliations of the members of the advisory and
technical committees:
Advisory Panel
Dr. John Scalzi (National Science Foundation)
Dr. Richard Livingston (Federal Highway Administration)
Dr. Inam Jawed (Transportation Research Board)
Mr. Nicholas Vitillo (New Jersey Department of Transportation)
Prof. Farhad Ansari (New Jersey Institute of Technology)
ASCE Representative: Ms. Patricia Brown
Committee-1: Sensor Characteristics &
Transduction Mechanism
Prof. Carolyn Dry (Univ. Illinois)

USA

Prof. Faramarz
Farahi

(Univ. North Carolina,

Charlotte)

USA

Prof. David
Jackson

(Univ. Kent)

UK

Prof. Ramash
Malla

(Univ. Connecticut)

USA

Prof. George
Sigel
Dr. Francis
Sladen

(Rutgers Univ.)

USA

(Photon Kinetics)

USA

Committee-2: Sensor Embeddment & Durability

Issues
Mr. Wolfgang
Habel

(BAM)

Germany

Prof. Dryver
Huston

(Univ. Vermont)

USA

Prof. Iwona Jasiuk (Georgia Tech.)

USA

Prof. Ali Maher

(Rutgers Univ.)

USA

Prof. Philip
Perdikaris

(Case West. Res.

Univ.)

USA

Dr. Urs
Sennhauser

(EMPA)

Switzerland

Prof. James Sirkis (U. Maryland)

USA

Mr. Whitten Schulz(Blue Rd. Res.)

USA

Page xiv

Committee-3: Distributed Measurements &

Multiplexing Technologies
Prof. Fabio Casciati

(Univ. Pavia)

Italy

Prof. Brian Culshaw (Univ. Strathclyde)

UK

Prof. Nabil Grace

(Lawrence Tech.)

USA

Dr. Arthur Hartog

(Photon Kinetics)

UK

Prof. Rolla Idriss

(New Mexico State

Univ.)

USA

Dr. Alan Kersey

(CiDRA Corp.)

USA

Dr. Robert Maaskant (Electro Photonics)

Canada

Prof. Raymond
Measures

(Univ. Toronto

Canada

Mr. Eric Udd

(Blue Rd. Res.)

USA

Committee-4: Standards & Specifications

Dr. Tino Alavie

(Electro Photonics)

Canada

Prof. Robert Green,

(Johns Hopkins Univ.) USA
Jr.
Dr. Alexis Mendez

(ABB Electric Syst.

Tech.)

USA

Dr. Kent Rochford

(NIST)

USA

Prof. Keith Wanser (California State Univ.) USA

Dr. Sandeep Vohra (Naval Res. Lab.)

USA

A post-workshop report will be developed. The report will encompass a

synthesis of the technical committee reports and outline the direction of the
national research program. In conclusion, the financial support provided by the
National Science Foundation, the Federal Highway Administration, and the New

Jersey Department of Transportation is greatly acknowledged.

FARHAD ANSARI
SMART SENSORS & NDT LABORATORY
SPRING 1998

Page 1

Chapter 1
Research and Development Programs

Page 3

Federal Highway Administration Research Program in Fiber Optics for the

Infrastructure
R. A. Livingston
Abstract
The Federal Highway Administration (FHWA) is conducting an research
program to explore the use of fiber optic sensors to a variety of highway
applications both in structures and in pavements. Possibilities range from
monitoring stresses in bridge beams during fabrication and transport to the
construction site to testing of the design rating of the bridge to monitoring the
deterioration of the structure over time (structural health monitoring). Under
an interagency agreement with the Naval Research Laboratory, the
applications of the Bragg grating type fiber optic sensors are being evaluated.
In another project, the use of fiber optics in truck weigh-in-motion (WIM)
systems is being investigated. Future research includes sensor development,
applications to pavements and to monitoring of chemistry.
Introduction
The magnitude of the need for monitoring in the infrastructure can be
indicated by a few statistics. The official National Highway System consists of
260,000 km of roadway, out of a total public mileage of roughly 6.3 million km.
Approximately one-half of this is rated as poor to fair[1]. There are about
576,000 bridges, of which 187,000 are rated as deficient[2].
The responsibility for managing these infrastructure assets falls on the state
and local governments, rather than the Federal government. However, the
Federal Highway Administration (FHWA) provides assistance, among other
ways, by conducting a research and development program[3]. The research
program concerned with fiber optic sensors is located at the Turner Fairbank
Highway Research Center in McLean, and has been funded on the order of
$300,000 per year. At present the FHWA fiber optics program consists of three
major projects, described in more detail below.
The Federal Highway Administration's fiber optics research program focuses on
Richard A. Livingston, Exploratory Research Team, Federal Highway Administration,
HNR-2, 6300 Georgetown Pike, McLean VA 22101.

Page 4

evaluating the sensors for applications in the highway infrastructure. In

addition to applications in structures such as bridges, the research also covers
uses in pavements and in soils. Also, besides field applications, fiber optic
sensors may be applied in the laboratory for such purposes as materials
characterization. For example, they could be used to monitor the shrinkage
during curing of concrete test specimens.
The FHWA research, which has been conducted under an interagency
agreement with the Naval Research Laboratory, has focused on fiber optic
strain gauges based on Bragg gratings. This technology permits multiple
gauges, as many as 100, to be put on the same fiber[4]. This greatly simplifies
the installation of the gauges and reduces the cabling requirement, as well as
reducing the cost per sensor. Also, since the Bragg grating system operates by
sensing the wavelength of light reflected back from the grating, it can be more
reliable and repeatable than other types of fiber optic strain gauges that use
other measurement methods such as interferometry.
The possibility of installing large numbers of sensors suggests a new approach
to structural monitoring, which typically has involved networks on the order of
10ndash;20 strain gauges. The amount of data obtained from such sparse
networks has generally proved insufficient for detecting early signs of
deterioration. Livingston has estimated that for several different applications,
networks on the order of one thousand sensors, or one kilosensor, may be
required[5]. An extensive network for the very largest structures could
conceivably reach megasensor numbers. Such a dense array of sensors would
enable the use of advanced computational methods of data analysis. At the
same time, the high rate of data flow create poses considerable challenges for
the data acquisition and management systems.
The application of these fiber optics sensors networks to all 576,000 bridges in
the nation's infrastructure may not be justified from a cost-benefit perspective.
It is more likely that they will be installed on only certain categories of bridges.
One category would be existing bridges that have severe deterioration. Another
would be major bridges that are critical links in urban transportation systems,
especially in seismically active areas, where shutting down even for a few hours
for post-incident damage assessment would create major traffic disruption. A
third category would be bridges made of novel materials such as polymer
composites, for which there is little information on performance in actual

structures.
At this point, it is not possible to perform such a cost-benefit analysis. The
costs of mass producing and installing large numbers of Bragg grating fiber
optics has not been established. Moreover, the optimum systems architecture,
which involves tradeoffs among number of sensors per fiber, number of fibers
and associated photonic instrumentation has not been determined[5]. On the
other side of the equation, the benefits of the data acquired in terms of
reduced construction or maintenance costs has not been established either.
Consequently the FHWA research program seeks to provide an information
base for an effective cost-benefit analysis. Issues under investigation include
overall monitoring strategy, systems architecture, installation procedures,
temperature compensation, durability, compatibility with other materials,
sensor standardization and data analysis.

Page 5

Applications of Fiber Optics

Fiber optic systems can be used in infrastructure applications for a wide range
of purposes that involve different time and spatial scales. One set of
applications primarily concerns the construction period. The sensors could be
used to monitor the prestress forces applied to the steel strands in precast
concrete components. This would be used initially to verify that the correct
stresses were applied. Subsequently it could measure the development length,
defined as the distance along the strand in which the bond between the steel
and the concrete is eventually lost[6]. Another application would be to monitor
the strains produced in the concrete by shrinkage during the initial hardening
process. Finally, the system could be used once the structure is completed to
verify the design calculations. This consists of placing known loads on the
bridge and observing the resulting deflections[7].
Applications for the post-construction period include continuous monitoring for
incident detection. These events such as the fracture of a member or a severe
impact such as the collision of a barge with a bridge pier, that would require
immediate response.
Instead of detecting a discrete event, the fiber optic monitoring system can be
used to sense a change in the mechanical properties that implies deterioration
of the materials, such as the aging of asphalt or the microcracking of concrete
or in the effective length of members caused by cracking or by the freezing of
bearings due to corrosion[8]. This approach, known as ldquo;health
monitoringrdquo;, includes sensing a change in the amplitude, frequency or
damping of transient vibrations that may be excited by either random traffic or
wind loadings, or by forced loading devices.
Another application could be weigh-in-motion (WIM) measurements of
vehicles. Although WIM systems have the potential for use as enforcement
tools to detect vehicles that exceed legal weight limitations[9], actual WIM
systems have been used for the less demanding task of obtaining spectra of
loads. These data can analyzed to estimate the probability that the maximum
load rating of the bridge is being exceeded[10] or to calculate the fatigue
cycles experienced by the structure, which can be used to predict remaining
service life[11].
These applications all involve the measurement of strain. Temperature sensing

is also of interest, both for temperature compensation of strain measurement,

but in addition for such topics as the prediction of ice formation or for curing of
concrete.
Another area is chemical sensing. For instance, the detection of chlorides in
reinforced concrete is important for the prediction of corrosion of the steel.
Also, measurement of the variation in pH would be useful for monitoring the
curing of concrete.
Sensor Requirements
The phenomena that could be monitored range over many orders of magnitude
depending on the application and the structure being monitored. Strains range
from 5 microstrain for vehicle loadings up to a few percent for soil movements.
As shown in Fig. 1, the time scales cover 16 orders of magnitude in frequency.

Page 6

Figure 1:
Timescales for Infrastructure Monitoring.

The ultimate time scale is the service life of the bridge, which can exceed 100
years. On the scale of years to decades are the deterioration processes that
include alkali-aggregate reactions in concrete or corrosion of steel, etc. On the
next scale, less than a year, occurs seasonal and diurnal cycles associated with
climatic variables. These are important because the natural frequency of the
structure can vary significantly with temperature, masking other effects[12].
Conversely, changes in the natural frequency with temperature can be used to
diagnose deterioration, such as the sticking of bearings[13]. Traffic loadings
can also show cycles on these time scales as well as on a weekly basis, as a
result of differences in driving patterns between workdays and weekends.
The time distribution of loadings imposed by individual vehicles depends on
vehicle spacing and speeds. On roads in rural areas, where there may be as
few as one vehicle per hour, this amounts to a frequency of recurrence on the
order of 10-4 Hz. On the other hand, for closely spaced traffic, traveling at the
legal speed limit 105 km/hr (65 mph), the frequency could approach 1 Hz.
The structure itself will vibrate at fundamental frequencies typically in the
range of 1ndash;20 Hz depending the span length, stiffness of span and on
type of design, e.g. truss or suspension.[14] However, higher order modes
with frequencies up 40 Hz are also generated and may be essential for
characterizing structural condition[15]. Forces associated with moving vehicles
have frequencies in the ranges of 1ndash;4 Hz for body/suspension motion and
8ndash;15 Hz for axle hop[16]. Higher frequencies, on the order of 10 kHz,
may be applied for evaluation of pavement dynamics[17]. Finally acoustic

emissions associated with the propagation of fatigue cracks fall in the range of
20 kHz to 1 MHZ[18].
In the design of the data acquisition system the sampling rate must be
matched to

Page 7

the timescale of interest. For example, since modal analysis can involve
frequencies up to 40 Hz, the minimum sampling rate should then be 80 Hz,
according to the Nyquist criteria[19]. However, this criteria concerns the
determination of frequencies at a single sensor. For correlations among
sensors, phase information is also important, and hence the sampling rate
would need to be higher.
Sensor Placement
The specification of the spacing of sensors along the fiber, and the placement
of the fibers around the structure, are critical factors in the design of the
monitoring program. This has the effect of determining the total number of
sensors. These factors in turn depend on the specific objectives of the
monitoring program.
For load rating measurements which involve monitoring the deflection of
girders, the conventional approach has been to locate strain gauges along the
bottom flange at the midpoint and at either quarter or third points, so that the
spacing between measurement points would between 5ndash;15 meters for
spans in the 20ndash;60 m range. In contrast, installing a single optic fiber
containing 100 sensors over the same range of lengths would provide strain
measurements on 30 cm to 60 cm centers. This higher spatial resolution would
improve the accuracy of the observed deflection curve. In the measurement of
bridge dynamics, it would increase the number of vibrational modes that could
be detected.
Bridge health monitoring operates by sensing natural frequencies of vibration.
For an idealized simple beam, the maximum amplitude of the fundamental
vibration mode would be at the midpoint, and for the second mode, at the
quarter points, and so on[20]. However, in the complicated 3-dimensional
geometry of a real bridge, a more advanced analysis would be required
involving the use of a finite element model.
Currently, Bragg grating fiber optics are produced on essentially a custom
basis. In each case the customer specifies the length, sensor spacings and
total number per fiber. For widespread routine use, it will probably be
necessary to standardize on a limited set of values for these parameters. This
would simplify the specification of embeddable sensor systems and would also
reduce production costs.

Data Acquisition Considerations

As noted above, the scale of the networks that could be created with Bragg
grating fiber optic sensors can produce tremendous quantities of data that
would have to be managed. For example a sensor sampled at 100 Hz
continuously over the lifetime of a structure, say 100 years, would produce 320
billion data points.
This brute force approach of bulk storage of raw data could avoided through
appropriate signal processing. For example, in WIM monitoring there may be
long stretches of dead time in which no vehicles pass by. This could be
handled by analog triggering circuits which turn the system on only when a
vehicle is detected. After the data are collected and digitized, digital data
compression methods such as those used in audio CD recording could be used
to minimize storage requirements.

Page 8

For a large structure, the transmission of the digital electronic signals is itself a
concern. At a sampling rate of 100 Hz and an analog to digital resolution of 16
bits, a one-kilosensor system would have a digital data rate of 1.3 Mbits per
second (Mbps). Long runs of cable are prone to damage and susceptible to
electromagnetic interference. Use of radio telemetry on a local network is
under serious consideration[21]. Another possibility would be to convert the
digital electronic signal back into an digital optical one to permit the use of
fiber optics as a communications system. Among the advantages would greater
bandwidth and insensitivity to electromagnetic interference. Furthermore, as
the technology of optical computing advances, it may ultimately be possible to
eliminate the electronic digital data processing step and go to a completely
optical system[22].
After all these data are acquired, they must be analyzed. Simply visualizing the
output from a kilosensor array, where each point can have multiple attributes
(2 or three-dimensional location coordinates, strain, acceleration, damping,
power spectra, temperature, etc) can be a daunting task. This suggests that
advanced data exploration or mining software and pattern recognition methods
such as neural nets should be considered[23].
FHWA Research Projects
Naval Research Laboratory
This is an interagency collaboration with the Optical Sciences Division of the
Naval Research Laboratory for the application of Bragg grating fiber optic
sensors to the infrastructure. This collaboration covers sensor development,
systems design and field testing. The initial phases of this program have
concerned the measurement of strains in portland cement concrete and steel
structures.
The first phase was tests of prototype instrumentation in concrete beams and
deck panels under test at the Turner Fairbank Highway Research Center[4].
The sensors were embedded in the concrete structures, which were then
loaded to failure. A total of 35 sensors was embedded in various beams and
deck panels. This demonstrated that the sensors were rugged enough to
survive the process of pouring and compacting the concrete. They showed
strain sensitivities comparable to, or better than, conventional strain gauges.
The sensors remained intact after the concrete itself cracked.

In the next phase, a sensor system with a 32 sensor capacity was designed
and built. This was then used in field tests at New Mexico State University,
described below. One outcome of this part of the research will be the
development of a set of preliminary guidelines for installation of Bragg grating
fiber optic sensors on existing bridges. This is intended to assist other
organizations such as state departments of transportation that seek to install
these monitoring systems on their own structures.
Currently, in collaboration with the Swiss Federal Institute of Technology,
Bragg grating sensors are being installed on the Viaduct de Vaux, a steel
bridge under construction in Switzerland. An initial set of 32 sensors will be
used to monitor the strains in the steel girders during erection. Subsequently, a
second set of 32 will be

Page 9

embedded in the concrete deck, using Bragg grating sensor in casings

designed by the Swiss.
Planning is now underway to install a system of 128 sensors on the Woodrow
Wilson Bridge in Washington DC. In addition to setting a new world's record
for number of sensors, this phase will investigate the issues of installing and
operating a very large scale network under heavy traffic conditions. Monitoring
systems for several other bridges are under consideration.
Another phase of this project concerns the development of strain gauges for
use in pavements and in soil. It has been difficult to monitor strains in these
friable and heterogeneous materials using conventional gauges. To date,
several gauges have been built using conventional designs of the gauge
housings (plate, diaphragm and ringuralite), but with Bragg grating sensors
replacing the conventional foil sensing elements[24]. The gauges are
scheduled to be evaluated under load in test pavement sections with a heavy
vehicle simulator at the US Army Corps of Engineers Cold Region Research and
Engineering Laboratory at Hanover, NH.
Finally, research has also been carried out on the problem of the temperature
dependence of fiber optic sensors. A method has been developed using
Brillouin scattering[25].
New Mexico State University
The purpose of this project is to investigate practical issues in the full scale
application and regular operation of fiber optics by civil engineers rather than
optical scientists. This project is co-funded with the National Science
Foundation. The first stage involved tests on a full scale concrete bridge deck
in the laboratory[26]. This was followed by the installation of sensors on an
existing steel bridge on Interstate 10 in Las Cruces, New Mexico, in
combination with some conventional strain gauges. The current number of
installed fiber optic sensors is 64, which ties the world record established by
ISIS Canada on a concrete bridge at Headlingley, Manitoba[27]. This research
at Las Cruces has demonstrated that fiber optics can effectively replace
conventional strain gauges in field situations. This system will continue to
operate for several more years in order to investigate long term performance
issues.
State Pooled Fund Study of Truck Weigh in Motion

Several state departments of transportation have pooled research funds to

conduct a study of the application of fiber optics to the specific problem of
weighing trucks in motion. This study will evaluate a variety of fiber optic
methods have been proposed for this purpose. At the time of writing, a
contract had not yet been awarded for this project.

Page 10

Future FHWA Research Directions

The scope of the fiber optic sensor research program depends on the future
levels of funding. This in turn depends upon the reauthorization of the
Department of Transportation's surface transportation program. At the time of
writing the passage of this legislation has been postponed.
Concerning full scale bridge installations, such as I-10 bridge in Las Cruces, NM
it is anticipated that responsibility for this type of project will be transferred
from the Exploratory Research program to other programs within FWHA that
deal with technology transfer, and to the state departments of transportation.
With the emergence of commercial vendors of Bragg grating fiber optic
monitoring systems, and the publication in the near future by FHWA of
guidelines for the application of these systems to bridge monitoring, the
installation of fiber optic sensors should be regarded as a part of the normal
design and construction process, rather than a specialized research project.
Research on fiber optic strain gauges for use in pavements and soils will
continue to be a priority. It will focus on improved housings for the sensors
that effectively transfer pressures and strains from the surrounding soil or
pavement with minimum disruption.
Research is also planned on the sensors themselves and associated photonic
components. This includes different grating configurations such as long period
or chirped gratings and distributed sensing. Additional work on temperature
sensing and compensation is also needed. Chemical sensing especially pH and
chlorides would also be a possibility. Development of photonic components
includes light sources with a broader wavelength range and higher bandwidth
opto-electronic convertors
Options for data transmission and management research include advanced
wireless systems and all optical systems. Finally computer methods for data
visualization and algorithms for structural health assessment need to be
developed.
Conclusions
Fiber optic strain gauges based on Bragg gratings have proven to be effective
replacements for conventional strain gauges for highway applications. They
create opportunities for monitoring on time and spatial scales that could not be

previously achieved in civil engineering. The potential for a very large number
of sensors also requires careful consideration of the design of the individual
components of the system and in their integration. The data management and
interpretation aspects must also be taken into account in the monitoring
strategy. The Federal Highway Administration's research focuses on resolving
these issues so that fiber optics can enter widespread use by state
departments of transportation.

Page 11

References
1. US Department of Transportation, 1995, 1995 Status of the Nation's Surface
Transportation System: Condition and Performance, FHWA-PL-96-007,
Washington DC, US Department of Transportation, pp 120ndash;135.
2. Chase, S.B. and G.L. Washer, 1997. ldquo;Nondestructive Evaluation for
Bridge Management in the Next Centuryrdquo;, Public Roads,
61(1):16ndash;25.
3. Federal Highway Administration, 1996. 1996 Research and Technology
Program Highlights, McLean VA, Turner-Fairbank Highway Research Center, 41
pg.
4. Davis, M.A., D.G Bellemore and A.D. Kersey, 1997. ldquo;Distributed Fiber
Bragg Grating Strain Sensing in Reinforced Concrete Structural
Componentsrdquo;, in Journal of Cement amp; Concrete Composites, 19(1):
5. Livingston, R.A. 1996. ldquo;Embeddable Sensor Monitoring Strategies for
the Infrastructurerdquo;, in Nondestructive Evaluation of Utilities and Pipelines,
M. Prager and R.M.Tilley, eds., Bellingham, WA: SPIE, pp. 246ndash;267.
6. Lane, S.N., 1995. ldquo;Development Length of Prestressing Strand in
Bridge Membersrdquo;, in Proceedings 4th Int. Bridge Engineering
Conference, Washington DC: Transportation Research Board, pp.
161ndash;168
7. A.G. Lichtenstein, 1994. Bridge Rating through Nondestructive Load Testing, NCHRP 12ndash;28(13)A, Transportation Research Board, Washington
DC.
8. Alampalli, S., G. Gu and E.W. Dillon, 1995. ldquo;On the Use of Measured
Vibrations for Detecting Bridge Damagerdquo;, Fourth International Bridge
Conference, Vol. 1, Washington DC: Transportation Research Board, pp.
125ndash;137.
9. FHWA, 1996. Highways, Code of Federal Regulations 23, Parts
657ndash;658, Washington DC: U.S. Government Printing Office.
10. Sarak, V.K. and A.S. Nowak, 1996. ldquo;Verification of Load Carrying
Capacity of an Old Bridgerdquo; in Proceedings 3rd Conference on
Nondestructive Evaluation of Civil Structures and Materials, M.P. Schuller and

D.P. Woodham eds., Boulder CO: Atkinson-Noland Associates, pp

431ndash;440.
11. Fisher, J.W, 1984. Fatigue and Fracture in Bridge Steels, NY: John Wiley
and Sons.
12. Roberts, G.P. and W.S. Atkins, 1995. ldquo;Recent Advances in Long Span
Bridge Dynamic Monitoringrdquo; Proceedings 6th International Conference on
Structural Faults and Repairs, M.C. Forde ed., Edinburgh: Engineering
Technics Press, pp. 61ndash;66.

Page 12

13. DeWolf, J.T., P.E. Coon and P.N. O'Leary, 1995. ldquo;Continuous
Monitoring of Bridge Structuresrdquo; in Extending the Lifespan of Structures,
IABSE Vol 73/2, Zurich: International Association for Bridge and Structural
Engineering, pp. 935ndash;940.
14. Burdet, O.L. and S. Corthay, ldquo;Dynamic Load Testing of Swiss
Bridgesrdquo; 1995. in Extending the Lifespan of Structures, IABSE Vol 73/2,
Zurich: International Association for Bridge and Structural Engineering, pp.
1123ndash;1128.
15. Aktan, A.E., V. Dalal, A. Helmicki, V. Hunt, M. Lenett, N. Catbas and A.
Levi, 1996. ldquo;Objective Bridge Condition Assessment for
Serviceabilityrdquo;, in Proceedings 3rd Conference on Nondestructive
Evaluation of Civil Structures and Materials, M.P. Schuller and D.P. Woodham
eds., Boulder, CO: Atkinson-Noland Associates, pp.183ndash;197.
16. Heywood, R.J, 1995. ldquo;Are Road-Friendly Suspensions Bridge-Friendly?
OECD Divinerdquo;, in Proceedings 4th Int. Bridge Engineering Conference,
Washington DC: Transportation Research Board, pp. 281ndash;295.
17. Martin c ek G., 1994. Dynamics of Pavement Structures, London: Eamp;FN
Spon.
18. Miller, R.K. and P. McIntire, eds., 1987. Nondestructive Testing Handbook,
Volume Five: Acoustic Emission Testing, Columbus, OH: American Society for
Nondestructive Testing.
19. Proakis, J.G and D.G. Manolakis, 1996. Digital Signal Processing: Principles,
Algorithms and Applications, Upper Saddle River, NJ: Prentice-Hall.
20. Mazurek, D.F., S.R. Jordan, D.J. Palazzetti and G.S. Robertson, 1992.
ldquo;Damage Detectability in Bridge Structures by Vibrational Analysisrdquo;,
in Proceedings Nondestructive Evaluation of Civil Structures and Materials, B.A.
Suprenant, J.L. Nolan and M.P. Schuller eds., Boulder CO: Atkinson-Noland
Associates, pp. 181ndash;194.
21. Maser, K., R. Egri, A. Lichtenstein and S. Chase, 1996. ldquo;Development
of a Wireless Global Bridge Evaluation and Monitoring Systemrdquo;, in
Structural Materials Technology- An NDT Conference, P.E. Hartbower and P.J.
Stolarski, eds., Lancaster PA: Technomic Publishing Co., pp.245ndash;251.

22. Eden, M. and L. Yaroslavsky, 1996. Fundamentals of Digital Optics, Boston:

Birkhuml;auser
23. Wasserman, P.D 1989. Neural Computing: Theory and Practice, NY: Van
Nostrand Reinhold.
24. Chang, C-C. and A.D. Kersey, 1997. ldquo;Development of Fiber Bragg
Grating Sensor Based Load Transducersrdquo; in 12th International
Conference on Optical Fiber Sensors: Technical Digest, G. Day and A. D.
Kersey, eds., Washington, DC: Optical Society of America, pp. 174ndash;177.

Page 13

25. Davis, M.A. and A.D. Kersey, 1996. ldquo;Separating the Temperature and
Strain Effects on Fiber Bragg Grating Sensors Using Stimulated Brillouin
Scatteringrdquo; in 1996 SPIE Symposium on Smart Structures amp;
Materials, K. Murphy and D. Huston eds., Bellingham WA:SPIE, Vol. 2718.
26. Davis, M.A., D.G. Bellemore, A.D. Kersey, M.A. Putnam, E.J. Friebele, R.L.
Idriss and M. Kodinduma, 1996. ldquo;High Sensor-Count Bragg Grating
Instrumentation System for Large-Scale Structural Monitoring
Applicationsrdquo;, 1996 SPIE Symposium on Smart Structures amp; Material,
K. Murphy and D. Huston eds., Bellingham WA:SPIE, Vol. 2718, pp
303ndash;310.
27. Rizkalla, S. and G.Tadros, 1994. ldquo;First Smart Bridge in Canadardquo;,
ACI Concrete International, Vol. 16(6) pp.42ndash;44.

Page 15

Chapter 2
State-of-the-Art

Page 17

Existing Technologies for Condition Monitoring of Construction Materials and

Bridges
R. E. Green, JR.
Abstract
This paper gives a brief overview of the some of the current nondestructive
evaluation (NDE) methods used for inspection of civil structures in the United
States particularly bridges. Several examples are given of instrumentation
developments over the recent past which have resulted in improvements in
nondestructive inspection of civil structures. Examples also are given of
techniques either used in other applications or those emerging from research
laboratories which afford the potential to greatly improve present NDE
technologies for inspecting civil structures. Finally, suggestions are made for
advanced NDE systems which when optimally developed will increase the
safety of civil structures.
Introduction
The role of nondestructive evaluation has changed dramatically in the United
States. Although historically nondestructive testing techniques have been used
almost exclusively for detection of macroscopic defects (mostly cracks) in
structures after they have been in service, it has become increasingly evident
that it is both practical and cost effective to expand the role of nondestructive
evaluation (NDE) to include all aspects of materials production and structural
fabrication. The premature failure of structures and devices dramatically shows
that our traditional approaches must be drastically modified if we are to be able
to ensure that structural components are to perform safely throughout their
expected life. Moreover, with the ever increasing requirements being placed on
structural components, the necessity for increasing the educational level of
nondestructive evaluation personnel and for transitioning new developments in
basic research into practical applications has become of vital importance.*
*Robert E. Green, Jr., The Johns Hopkins University, Center for Nondestructive
Evaluation, 3400 N. Charles St., 102 Maryland Hall, Baltimore, MD 21218

Page 18

Current Status of NDE for Bridges

As of June 1988, there were 577,710 inventoried and classified bridges in the
United States, of which 135,826 were classified as structurally deficient -meaning that they were closed or required immediate rehabilitation. Civil
structures including bridges have not been given high priority for application of
advanced NDE techniques as compared with aircraft and aerospace structures.
In fact visual inspection approximately once every two years is the primary
technique used for routine inspection of bridges. Conventional NDE is only used
as an adjunct to visual inspection and very few advanced NDE techniques are
used. Moreover, although in recent years there has been an increase in NDE
techniques for construction materials such as concrete and wood, the funding
for these endeavors is minuscule compared with that spent on advanced
materials such as special metal alloys and composites.
Steel Bridges
For steel bridges cracks are the prime defect needing detection and sizing,
with special emphasis on welds. A method for monitoring the behavior of
cracks under service conditions is needed. Corrosion of steel cables of
suspension bridges is also a problem. Conventional radiography has been the
method of choice for detection of volumetric defects in steel bridges. Other
conventional NDE methods such as liquid penetrant, magnetic particle, eddy
current, ultrasonics, and acoustic emission are being used for crack detection
and monitoring. All of these conventional NDE techniques are used as an
adjunct to visual inspection.
Concrete Bridges
For concrete bridges corrosion of steel reinforcing bars is the main problem
needing detection, monitoring, and assessment. The fact that the reinforcing
bars are encased in concrete make inspection particularly difficult. The global
techniques for inspection of concrete bridges are load testing and modal
analysis, although the results of modal analysis techniques have proven rather
limited in capability. The local techniques for inspection of concrete bridges are
ultrasonics, impact echo, magnetic, electrical resistivity, corrosion potential,
infrared thermography, ground-penetrating radar, radiography, and acoustic
emission.

Page 19

Timber Bridges
The main cause of damage in timber bridges is attack by fungi, insects, and
marine borers. Visual inspection is the major technique used to inspect timber
bridges. There is currently no advanced nondestructive evaluation methodology
widely used for assessing the quality or the performance of timber bridges.
However, sonic and spectral stress-wave and acoustic emission techniques
appear to be the NDE methods showing most promise for timber bridge
inspection.
Composite Bridges
Although there are only a few composite bridges, the type of defects expected
are the same as those in other composite structures. There are no NDE
techniques developed specifically for composite bridges, however, there are
numerous relatively sophisticated NDE techniques for monitoring degradation
of composite materials. These techniques have been primarily developed for
the aircraft and aerospace industries. Planning for NDE of composite bridge
structures should begin now; such structures afford outstanding opportunities
for embedded sensors.
Goal of NDE for Bridges
The overall goals of NDE for civil structures are: develop inexpensive, rapid,
automated inspection techniques; develop techniques that provide quantitative
flaw size information not subject to operator interpretation; develop multi-mode
multi-sensor NDE techniques which will yield complimentary information;
develop self-monitoring systems including sensors and actuators; plan for NDE
as part of the design process for new bridges; integrate the sensitivity of
inspection methods and severity of the discontinuity into realistic accept/reject
criteria; develop standardization and codes of practice; create continuing
education programs for engineers involved in bridge construction and
maintenance.
Nondestructive Techniques for Civil Structures
Civil structures have been examined historically by one of the big five
nondestructive testing techniques, namely: ET (eddy current), MT (magnetic
testing), PT (penetrant testing), RT (radiographic testing), and UT (ultrasonic
testing). Recently the list of techniques have expanded to include those listed

in Table I. In addition, semi-destructive tests, which could equally well be

labeled semi-nondestructive, have been developed for concrete as listed in
Table II.

Page 20
TABLE I CURRENTLY USED NONDESTRUCTIVE
EVALUATION TECHNIQUES FOR CIVIL STRUCTURES
Acoustic Emission
Magnetic Particle & Flux Leakage
Barkhausen Noise
Eddy Current
Electrical
Gamma, X-radiography & Tomography
Holography
Impact Echo & Rebound Hammer
Infrared Thermography
Microwave Absorption
Neutron Radiography & Scattering
Nuclear Magnetic Resonance
Visual & Optical Interferometry & Shearography
Liquid Penetrants
Impulse Radar
Ultrasonics
X-Ray Diffraction
Resonant Ultrasound Spectroscopy
Modal & Vibrational Analysis
TABLE II SEMI-DESTRUCTIVE AND DESTRUCTIVE TESTS
FOR CONCRETE (Property determined indicated in
parentheses)
Pull-out (Indirect Shear, Tensile Strength)
Pull-off (Tensile Strength)
Break-off (Flexural Strength)
Windson Probe (Penetration Resistance)
Tescon Probe (Stress-strain Relation)
Cores (Strength)
Maturity Method (Relationship Between Time And
Temperature, Temperature Probe)
Permeability Testing (Chloride Ions, Electrical, Gas
Permeability)

Page 21

Examples of Recent NDE Applications to Civil Structures

Ultrasonic Crack Detection
Ultrasonic pulse-echo techniques have been used more often than other
nondestructive evaluation techniques for detection of cracks in metal
structures. A major reason for this choice of inspection is the relatively portable
nature of ultrasonic inspection equipment. However, even modern solid state
devices usually require one hand to hold the electronic unit and the other hand
to scan the transducer. This often makes it very difficult for the inspector to
maintain a safe and comfortable position on the test structure. An improved
ultrasonic unit designed to eliminate this problem is an ultrasonic system
consisting of an electronic backpack and a hand-held transducer probe. Two
separate transducer channels operate alternately. One transmits compressional
(longitudinal) pulses of ultrasonic energy in a direction perpendicular to the
inspection surface. When the ultrasonic energy reflected back to the
transducer acting as a receiver is above a threshold level, illumination of a
green light on the probe indicates that sufficient energy has been coupled into
the inspection surface to permit reliable inspection.
A more sophisticated ultrasonic inspection system has been developed for
automatic imaging of cracks in metal structures. This system consists of two
major components linked by a communications cable, containing power and
ultrasonic, audio, and video signals. One component carried by the inspector
consists of a special hand-scanner unit, transducer, and a video display linked
to the processor. The special portable hand-scanning unit can be mounted on
the structure to be inspected and monitors the position of the ultrasonic
transducer continuously as the operator hand scans the inspection area. The
other major component, located at a data analysis station, consists of a
computer system with a digitizer, graphics display, data analysis software and
disk storage, and an ultrasonic mainframe system. Two operators, one at the
inspection site and the other at the data analysis location, operate the system
together. The scanner operator is provided with visual feedback via the
television monitor and verbal feedback from the processor operator through a
headset.
The feature of separation of a relatively unskilled inspector and highly trained
data analyzer, who are in continuous communication, can be optimally used in
other nondestructive evaluation applications. Addition of a number of

inspectors in communication with a single data analysis center could optimize

inspection of large civil structures without the need for highly skilled inspectors.

Page 22

Acoustic Emission Monitor

A unique acoustic emission monitoring system developed for flaw detection in
steel highway bridges incorporated a digital memory microcircuit with source
isolation to admit signals only from a preselected area. The entire system
consisted of amplifiers, source isolation circuits, digital memories, and memory
programming controls in a compact housing weighing only 5 pounds. A
separate power supply provided power from internal rechargeable batteries.
Two separate acoustic emission parameters were stored in solid state digital
memories, which could be removed from the on-site recorder and taken to a
data analysis center for readout. The digital memories could be reused by
erasing with ultraviolet light.
It is suggested that such systems using fiber optic sensors could be located
strategically on bridge structures to monitor crack initiation and growth. Similar
systems could be economically placed on a variety of constructed facilities and
the digital memory microcircuits collected and replaced with others in a regular
scheduled fashion by relatively unskilled people and brought into a central
analysis facility for readout, signal assessment, and erasing. A slightly more
expensive installation would have a radio transmitter as an integral part of each
monitor, with detected acoustic emission signals periodically broadcast to a
recording and analysis instrumentation van.
Coherent Laser Radar
Another proposed novel method of detecting acoustic emissions events from
bridge and other constructed facilities during mechanical loading is use
application of a coherent laser radar system. For military applications a
coherent laser radar system has been developed to detect and characterize
transient acoustic waves from optical reflective surfaces at long ranges (greater
than 6km). This system has been used to detect the vibrational signatures of
ships at sea and aircraft in the air and to identify them uniquely. A properly
designed laser radar system should be able to obtain reliable acoustic emission
measurements or vibrational modes by probing load-bearing structures at
relatively large standoff distances.
Holographic Interferometry and Laser Speckle
Increasingly, NDE techniques developed for nonstructural applications have
come to be of direct use in testing and inspecting large structures. Example

are holographic interferometry and laser speckle. Holographic interferometry

consists of exposing a photographic recording device to a monochromatic
coherent laser beam which has been split into two components by a beam
splitter. One of the split beams illuminates the surface of the object to be
tested and is then reflected to

Page 23

a television camera, while the other split beam, called the reference beam,
travels directly to the television camera. The resulting interaction of the two
beams results in an interference pattern on the television camera monitor. If
the test object is then subjected to a stress such that finite displacements
occur, the interference pattern will change in such a fashion as to indicate the
nature of the displacements.
Laser speckle uses two symmetric laser beams to illuminate the surface of the
object to be tested. The interaction of the two laser beams with the object
surface roughness causes laser light ''speckles" over the area illuminated by the
two beams. By proper recording of this speckle pattern full body motion can be
distinguished from in-plane deformations caused by imposed stresses. By
making similar recordings at selected intervals, the state of deformation of the
surface of the object can be determined.
Similarly holographic interferometry and speckle patterns can be recorded
using microwave radiation sources rather than optical ones. The major
difference in the resulting speckle patterns is that, since microwave
wavelengths are larger than optical wavelengths, larger surface displacements
are observed with microwaves.
Dual Wavelength Infrared Imaging
A major problem with conventional single wavelength infrared imaging is that
often the differences in emissivity of structures to be examined cause more
alterations in detected signals than true thermal signatures. A dual wavelength
infrared imaging system has been developed and used to inspect concrete
highways and bridge decks for subsurface defects. Because of the dual
wavelength nature of this system it was able to distinguish temperature
differences at locations of delamination from emissivity differences due to oil
stains, sand, gravel, metal parts, and surface roughness.
Magnetic Flux Leakage Perturbation
Several robotic magnetic flux leakage perturbation systems for inspection of
steel cables, beams and reinforcing bars in concrete are under development.
One suspension bridge cable inspection system rides on the cable. When a
suspender rope element is approached the magnets are shunted and the
leading bridging adapter caliper and magnetic array are opened, passed over
the suspender rope band and the bridging caliper closed. The unit is then

moved forward until the trailing bridging adapter caliper contacts the
suspender rope element, is opened, moved over the suspender robe band and
closed. Upon completing passage over the band, the magnetic array is closed
over the cable and perturbation measurements made. To install or remove the
system from the cable, both calipers and magnetic array are opened fully. A
more recent system under development also has a self

Page 24

clamping/driving unit that can scan the underside of a concrete member along
its length and a control/data recording and analysis system on the ground. The
system holds a set of two permanent magnets and an array of ten Hall probes.
The system will be controlled remotely using radio telemetry.
Magnetic Detection of Corrosion Damage
Several electrochemical techniques have been developed to monitoring the
rate of corrosion of reinforced concrete bridges. A novel nondestructive method
for remote non-contact detection and evaluation of corrosion in metallic
materials has been developed. Measurements on sections of underground gas
pipelines, under varied conditions of coatings and soil coverings, have been
made using Electrochemical Impedance Spectroscopy (EIS) instrumentation
with magnetometer sensing of pipe and earth current. The methodology
permits survey of underground pipelines from surface measurements, locates
corroding regions in base pipelines, determines defective regions in coated
pipelines including defect size and corrosion rate. The system will also prove
extremely useful for other applications such as corrosion evaluation of steel
reinforcing bars in concrete structures.
Magnetic Tagging
Westinghouse investigators have developed a magnetic tagging method for
construction materials. The method consists of the addition of a small amount
of tiny magnetic ceramic particles to a material during manufacture so that
subsequent interrogation with electromagnetic instrumentation produces a
signature related to material integrity. This technique has been demonstrated
in the laboratory of process control and in-service inspection of concrete,
asphalt, polymer piping, structural textiles, adhesives and geotechnical clay.
Computer Assisted Tomography
Computer assisted x-ray tomography has become a standard feature at all
advanced medical institutions and both x-ray and neutron tomography has
made inroads into the nondestructive evaluation community. The
expensiveness of the techniques somewhat limits present usage, but future
developments in detectors, computers, and reconstruction algorithms portends
well for reduction of inspection cost and more use for civil structures in the
future. Computer assisted tomography permits retrieval of three dimensional
information about the exact locations and dimensions of the internal features of

the test object. The applications of fall into two classes: inspection and
characterization. Inspection techniques require object

Page 25

recognition and dimensioning. Characterization applications require

quantitative analysis. Examples include the density/atomic number gradients in
high explosives to the radioactive assay of waste containers. The major
application of tomography to civil structures is that of imaging of cracks in steel
and corrosion of reinforcing bars in concrete.
Telemetry Systems for Self-Monitoring of Bridges
Infrastructure inspection has either been made by human inspectors using
hand operated instruments or with sensors installed on structures and
connected by wires to computers and recorders located on or under the
structure. Beginning in the missile and space programs, where it is now garden
variety type of data handling, to electric utilities, steel mills, coal mines, and
nuclear power plants, radio telemetry brings measurement data from sensors
to computers, recorders, plotters, and analyzers in a most convenient manner.
In one adaptation, data is collected from sensors and stored in a microcircuit
packaged with the sensor. This data is read out remotely on demand either
with a laser or electromagnetic probe. In another adaptation, short range radio
transmitters collect data from local sensors and transmit it by a long range
radio relay to a central monitoring station. Both systems are able to
simultaneously accommodate several different kinds of sensors including strain
gauges, accelerometers, seismometers, clinometers, geophones, linear voltage
differential transformers, eddy current coils, acoustic emission and ultrasonic
transducers and provide power to those that require it. The application of
modern fiber optic sensors coupled with radio telemetry will make outstanding
nondestructive self-monitoring systems for bridges and other civil structures.
The range of a radio link is limited by the strength of the signal radiated by the
transmitter toward the receiver and by the sensitivity of that receiver. A 10microwatt output will transmit data easily one hundred feet with a bandwidth
of 100 KHz. The wider the bandwidth, the more the effect from noise, and
therefore the more transmitting power required for an acceptable signal.
Because of space limitations on the structure to be monitored, the transmitting
station often must be small, possibly doughnut size, but sometimes no bigger
than a pea. It also must be self-sufficient, carrying it's own power or perhaps
receiving it by laser or microwave. At the receiving station, there are usually no
space restrictions in accommodating large antennas, sensitive radio tuners and
recorders, and an ample power supply.

The functional requirements for industrial radio telemetry are much different
from those in space telemetry. Distances are much shorter, a matter of a few
feet to a few hundred yards; signal power can be radiated directly from the
transmitter circuitry or from an antenna as simple as an inch or two of wire.
Quantities can be measured one or two at a time, rather than requiring an
enormous amount of information to be transmitted at once. This enables
simpler circuitry at both the

Page 26

transmitting and receiving ends. Environment plays the most critical role in
industrial telemetry. Industrial telemeters must operate repeatedly without
adjustment and calibration. It is almost mandatory that they be completely
encapsulated to be impervious to humidity and water, and often to many other
chemical fluids and fumes. Short range radio transmitters can collect data from
local sensors (strain gauges, accelerometers, seismometers, clinometers,
acoustic emission transducers, and fiber optic sensors) and transmit via a long
range radio relay to a land based building or vehicle.
Other Recent Developments:
The Federal Highway Administration has a number of projects underway in
various phases of completion to further augment NDE of civil structures.
Among the projects are: Ground-penetrating radar imaging for bridge deck
inspection; Global bridge monitoring with wirelsss transponders; Bridge
deflection measurement using a precision differential global positioning system;
Bridge overload measurement and monitoring using TRIP steel sensors;
Combined ultrasonic and magnetic analyzer for cracks; Passive fatigue load
measurement device; Fatigue load measurement using electromagnetic
acoustic transducers; Eddy current detection of weld cracks; Crack detection
using alternation current field measurement; Impact-echo system for detection
of voids in post-tensioning ducts; Embedded corrosion microsensor; Bridge
substructure evaluation using forced vibration response; Cable-stay force
measurement using laser vibrometers; Microwave detection and quantification
of fatigue cracks.
In addition, a most important and very timely development is the construction
of a new Nondestructive Evaluation Validation Center at the Turner-Fairbank
Highway Research Center in McLean, Virginia.
Fiber Optic & Other Microsensors
Fiber optic and other microsensors have found many applications in the field of
nondestructive evaluation as well in the area of smart materials for civil
infrastructure applications. Fiber optic sensors generally operate based on one
of two methods. In the first method light travels out of the fiber, interacts with
the medium to be sensed, and travels back into the fiber for analysis. In the
second method the effect to be sensed causes alteration of the light which
remains entirely inside the fiber. Such sensors have been used to monitor

temperature, position, rotation, acceleration, strain, pressure, flow, vibrations,

acoustics, viscosity, chemical change, electric fields, magnetic fields, current,
etc.
A fiber optic system has been developed for nondestructive measurement of air
content in fresh concrete. Fiber optic Bragg grating systems have been used
for

Page 27

remote deformation monitoring of the critical section of structural concrete

elements and a multi-girder steel bridge. Fiber optics sensors can be used to
detect failure of a material structure, depth of carbonation, alkali reaction,
chloride penetration and crack opening displacement in concrete structures. An
optimum fiber optic sensor would be small, have low weight, require low
power, possess environmental ruggedness and immunity to electromagnetic
interference, require low power and be of minimal cost. Structures could be
constructed using fiber optic systems that not only network all occupants to
essential data services, but serve to monitor the structural health of the
structure.
Proposed Bridge Monitoring Systems
Now is the time to start developing smart, efficient, low cost nondestructive
evaluation systems for inspection of construction materials, bridges, and other
civil structures. It is hoped that as a result of this International Workshop on
Fiber Optic Sensors for Construction Materials and Bridges plans will be
developed and funds provided to design and construct smart NDE monitoring
systems which multiplex sensors, particularly fiber optic ones, possessing
different detection capabilities with either local data recording and subsequent
transmission or instantaneous data transmission using radio telemetry to a
central monitoring location.
References
1. Nondestructive Testing Handbook (Second Edition), 19831996, American
Society for Nondestructive Testing, 1711 Arlingate Lane, Columbus, OH.
2. Green, R.E. Jr. et al. (Eds.), Nondestructive Characterization of Materials,
Volumes I-VIII, 19841998, Center for Nondestructive Evaluation, The Johns
Hopkins University, Baltimore, MD.
3. Malhorta, V.M. (Ed.), 1984, In Situ/Nondestructive Testing of Concrete,
Publication SP-82, American Concrete Institute, Detroit, MI.
4. Manning, D.G., 1985, Detecting Defects and Deterioration in Highway
Structures, NCHRP Synthesis of Highway Practive, Transportation Research
Board Report 118.
5. Aghabian, M.S. and S.F. Masri (Eds.), 1988, Proceedings of the
International Workshop on Nondestructive Evaluation for Performance of Civil

Structures, Department of Civil Engineering, University of Southern California,

Los Angeles, CA.

Page 28

6. dos Reis, H.L.M., 1988, Nondestructive Testing and Evaluation for

Manufacturing and Construction, Hemisphere Publishing Corp., New York, NY.
7. Lew, H.S. (Ed.), 1988, Nondestructive Testing, American Concrete Institute,
Detroit, MI.
8. Barton, J.R, C.M. Teller, and S.A. Suhler, 1989, Design, Development and
Fabricate a Prototype Nondestructive Inspection and Monitoring System for
Structural Cables and Strands of Suspension Bridges, Federal Highway
Administration Report No. FHWA/RD-89-158.
9. Singh, G.P. 1989, Advanced Bridge Inspection Methods: Applications and
Guidelines, Federal Highway Administration Report No. TS-89-017.
10. Suprenant, B.A, et al., Conferences 19901996, Proceedings of Conference
on Nondestructive Evaluation of Civil Structures and Materials, Atkinson-Noland
& Associates, 2619 Spruce Street, Boulder, CO.
11. Malhotra, V.M. and N.J. Carino, 1991, Handbook on Nondestructive
Testing of Concrete, CRC Press, Boca Raton, FL.
12. Ansari, F and S. Sture (Eds.), 1992, Nondestructive Testing of Concrete
Elements and Structures, American Society of Civil Engineers, 345 East 47th
Street, New York, NY.
13. Streit, R.E., A.E. Holt, and D.E. Bray (Eds.), 1992, Proceedings of ASME
Topical Conference on NDE's Role in Concurrent Engineering, American Society
of MechanicalEngineers, New York, NY, Session 4, Paper 5.
14. Proceedings Conference on Nondestructive Evaluation of Bridges, 1993,
FHWA-RD-93-040A, DOT-VNTSC-FHWA-93-1.
15. Weyers, R.E, G.S. Brown, J.Duke, I.L. Al-Qadi, and S.M. Riad (Eds.), 1993,
Workshop on Engineering Needs for Non-Invasive Evaluation of Transportation
Structures, The Center for Infrastructure Assessment and Management,
Virginia Polytechnic Institute and State University, Blacksburg, VA.
16. Scancella, R.J. and M.E. Callahan (Eds.), 1994, Structural Materials
Technology, An NDT Conference, Atlantic City, NJ.
17. Chase, S.B and G. Washer, July/August 1997, Nondestructive Evaluation
for Bridge Management in the Next Century, Public Roads, pp. 1625.

Page 29

Monitoring Systems and Civil EngineeringSome Possibilities for Fibre Optic

Sensors
B. Culshaw
Abstract
Fibre optic sensing systems undoubtably have a role to play in future civil
engineering structures and indeed this conference is dedicated to exploring
that role.
This paper presents an input to this debate initially by exploring the particular
features of fibre optic sensors in general which may stimulate their exploitation.
The paper then describes some applications based predominantly on first hand
experience. Two generic applications are analysed. The first is long gauge
length strain measurement involving lengths of 10s of centimetres or more.
This particular measurement function is only available using optical fibre
technology. The second examines distributed moisture ingress measurement
whereby moisture level as a function of position may be plotted along an
optical fibre sensor over lengths of up to a few kilometres. In this case there
are a few electrical instruments which use time domain reflectometry which
offer similar capability but over a range which is restricted to metres - so again
optical fibre technology has something unique to offer.
Despite the technological edge there remain some barriers to exploitation. Of
these the most important is that the measurements must be recognised as
necessary by the eventual user. Thereafter fibre optics must compete on cost
and technical performance with alternative technologies. Here the immense
progress in fibre optic communication systems has already established one
level of confidence and thereafter all that is required is successful application
trials.
Brian Culshaw, University of Strathclyde, Department of Electronic & Electrical
Engineering Royal College Building, 204 George Street, Glasgow G1 1XW, UK

Page 30

1. Introduction
Structural measurements in civil engineering have historically almost exclusively
been limited to defining a construction process. The traditional methods of
surveying, checking structural orientations, dimensional monitoring and similar
functions have been in operation for generations. The philosophy has been that
if the structure has been built to code then it has been built to last. However in
the past decade or so this philosophy has been increasingly questioned as
vastly expensive structures, particularly bridges, have become structurally
distressed causing at least considerable social inconvenience and occasionally
resulting in catastrophic failures [1]. Whilst it is possibly presumptuous to
generalise, most of these structural failures are due to one from or other of
corrosion effects (including erosion and scouring) with some contributions from
foundation movements and occasionally from questionable construction
procedures in the first instance. This then prompts the obvious question - could
these considerable problems have been avoided had they been detected earlier
and if so what kind of monitoring systems would alert the appropriate evasive
action.

Figure 1:
Who decides to build a bridge???

Whilst the obvious answer to the first question may be "yes" a rapid more
careful analysis quickly demonstrates that the answer to the question depends
critically upon who you ask. The "best value conflict" (Figure 1) is particularly
evident in the civil engineering context where, with the exception of dams,
there are no legislative requirements to install continuous monitoring
instrumentation systems. Major infrastructure investment, such as bridges, is in
the final reckoning approved by (non-technical usually) politicians whose
principal objective is to please their current electorate by providing maximum

obvious benefit for minimum investment. Consequently there has to be the

maximum amount of

Page 31

bridge for the money and no frills apart from the visual. Our politician will have
retired or be out of office before any problems will arise. The contractor needs
to maximise his profit, the consultant who designs the bridge will only get the
job if his designs please both the contractor and the politician. Meanwhile the
paying public, unaccustomed to even contemplating the situation twenty years
down the line, remain impressed by all the visual amenity for their tax dollar.
The construction process is monitored on behalf of the politicians by the
municipal works department whose direction is simply to ensure that the
structure is built to appropriate safety codes and that the normal precautions
of the day are observed. Nowhere in this discussion is the long term reliability
of the structure ever hinted upon - but after all it is made of concrete which,
since it is derived from natural rock, must of course last for every.
Only a gradual change in the political, economic and social framework can then
call this basic process into question - and the experimental evidence in a
decaying civil infrastructure has dramatically pointed out the need for
comprehensive continuous structural assessment. Build and operate contracts
are beginning to be introduced (though for relatively restricted periods, say 20
years, during which structural reliability can usually be assured). The social
costs of bridge closures and load limitations are beginning to be recognised
though recent experience in London, England, has indicated that closing
bridges can often decrease traffic congestion by deflecting the road user
elsewhere. Additionally there have been several horror stories of bridge
collapse in major cities [2] and all these factors have reopened the debate and
in particular have stimulated public awareness. So perhaps it is beneficial to
make measurements after all.
There are hundreds - probably thousands - of techniques for making
measurements, all specialised, all niche oriented and almost all in the domain
of the agile small company rather than the lumbering multinational. The civil
engineer - assuming he is persuaded that measurements are desirable - is
simply interested in making the measurement and the technology used is
irrelevant provided that it works and is reliable. Fibre optics, which is the
subject of this meeting, must therefore take its place beside the thousands of
competing technologies and recognise its benefits and problems. Most
measuring systems involve monitoring physical parameter fields principally
strain and temperature and these are functions for which there are already well
established tried and proven techniques available, principally based upon

electrical measurement and transmission though also involving some good old
fashioned hydraulic and mechanical systems (which work very well). Fibre optic
sensing is an alien concept in a world of poured concrete and pile drivers and
so in use must be compatible with this world. It must do something different
and potentially very beneficial when compared to the competition. It is
therefore critical that the potential user fully understands the benefits of the
distributed and quasi distributed sensing and in-line multiplexed architectures
which characterise fibre optic systems. This is particularly true at the data
interpretation stage in a culture which, if it measures at all, looks at strain
gauges, vibrating wire gauges and thermocouples attached to the outside of
whatever is of interest.

Page 32

There are also other benefits of fibre optics. The sensing system can be
installed inside as well as attached to the structure of interest. It can be and
has been designed to withstand the rigors of the installation process and of onsite handling. There is usually no accompanying wiring or power supply to
corrode and/or pick up electromagnetic interference and very large scale
systems can be passively addressed from a single interrogation point. These
are very real benefits but must be stacked against the questions of
unfamiliarity not only with the sensing technology but also with the
measurement functions which are offered. The perceived cost in both the
sensing system itself and the training time for use and installation and the
continuing debate over how to interpret all the data which is produced
especially if this data is in a relatively unfamiliar measurement format also
inhibit the application of fibre optic technology. But the technical benefits are
real, the enthusiasm of the fibre optic sensor community to see something
happen is real and many of the initial trials in field installations are civil
engineering laboratories have produced convincing and promising results.
2. Some Reflections on Measurement Architectures
The fibre optic sensor systems which feature most strongly in the civil
engineering context use the fibre itself in a long line to sense the parameters of
interest as a function of position along this line. This can be realised in three
broad architectural formats namely point multiplexed, distributed and quasi
distributed arrays the features of which are shown schematically in Figure 2[3].

Figure 2:
Illustrating the outputs for (a) multiplexed
point sensor (b) quasi-distributed array and
(c) distributed sensor with short pulse width

Page 33

The last two of these measurement architectures are relatively unfamiliar

within the user community which is traditionally reliant upon point
measurements of parameters fields. It is important to appreciate that these
three measurement functions perform different tasks and a simple way to
visualise this difference is shown in Figure 3. Here we represent a beam with a
crack at a particular point subjected to a longitudinal load. Strain gauges at
points A, B C etc (and also point multiplexed systems) will give the same
readings in the presence or absence of the crack. However a quasi distributed
system which measures the distances between points A, B, C etc will
immediately highlight the presence of the fault. These observations clearly
apply to static loading conditions. The story for dynamic loading conditions will
be different and the influence on the mode shape of crack size would require
more accurate modelling though again the modified dynamic strain in the
section including the fault will be more obvious using the quasi distributed
approach. The distinction between quasi distributed and distributed
measurements is more subtle but essentially evolves around the speed of
response. Fully distributed systems, involving for example Raman and Brillouin
scatter [4], have been designed to monitor very slowly varying quantities and
have total system interrogation times typically of the order of minutes. This
precludes their operation as dynamic detectors without considerable
modifications to receiver design and operational software. Quasi distributed
systems interrogate a known number of points and are typically more rapid to
respond. Functionally the two should be identical: it is technological detail
which in fact imposes the distinctions.

Figure 3:
A simple indication of the functions of sensor architectures.

In the civil engineering context there are also several monitoring functions
which are relevant including:

Page 34

- Positional mapping of a structure with respect to its environment which is the

domain of established surveying techniques and to which fibre optics has
probably little to contribute.
- Positional mapping of the relative positions of nodes within or upon the
structure, a function to which fibre optics could contribute.
- Interrogation of local phenomena such as strain distributions and corrosion
processes, again a function to which fibre optics could contribute.
- Damage and impact detection through external agents such as shipping,
earthquake, landslide, etc, again a technology to which fibre optics could
contribute.
Each of the combinations of application sector and measurement architecture
offers its own idiosyncratic benefits and compromises. Many of the possibilities
have been evaluated using optical fibre technology. The ubiquitous fibre Bragg
grating [5] enables the realisation of point multiplexed strain gauge (and
thermometer) arrays and similarly the EFPI [6] (extrinsic Fabry Perot
interferometer) technology fulfils a similar measurement function. Distributed
measurements based upon induced microbend loss have achieved some
success for both strain monitoring and moisture ingress monitoring and quasi
distributed strain measuring systems (effectively point to point distance
measurement systems) have been extensively evaluated. Fully distributed
systems based upon non-linear scattering processes have yet to be seriously
evaluated in the civil engineering context though do offer some possibilities for
measurements over very long distances in relatively stable environments. Many
of these techniques are explored in accompanying papers in this volume - in
the remainder of this contribution we shall examine in more detail the
possibilities for distributed chemical measurement and the relatively mature
technology for quasi distributed point to point distance measurement systems.
3. Distributed and Quasi Distributed Architectures - Some Examples
Physical Measurements Based on Quasi Distributed Architectures
There are numerous techniques available for the precision measurement of
individual section lengths in a quasi distributed sensor architecture. These
include the use of straightforward ultra short pulse OTDR [7], microwave
subcarrier phase measurements [8] and several white light interferometer

schemes using either separate fibres or two modes within a fibre as a reference
[9,10] and signal path. These techniques are all capable of better than 100 m
long term resolution and accuracy and sub-micrometre short term resolution.
For the short pulse and microwave subcarrier systems these resolutions are
also independent of the

Page 35

interaction length. Only in dual mode interference systems does the interaction
length influence measurement resolution.
There are also numerous good reasons why these quasi distributed
architectures are appropriate. Figure 3 has indicated the general principle of
one of these - namely that the quasi distributed architecture is sensitive to
changes in structural properties anywhere along the length of the sensor
element. There is one other important applications reason in that in some types
of civil engineering structure the strain transfer process between the optical
fibre measurement element and the structure itself can fluctuate radically
along the length of the fibre. This is particularly true in situations where it is
not possible to ensure that the mechanical contact between the structure and
the measurement fibre is exactly uniform throughout its length. In this case
local strain measurements can be totally misleading indicating either high or
low values whilst the average value over a length compared to the fluctuation
period is significantly more meaningful.

Figure 4:
Photograph of trial sensor packages for ground anchor tests

The applications also depend upon the availability of suitable packaging to

ensure both fibre protection and efficient strain transfer. Some examples of
single fibre packaged length measuring sensor elements are shown in Figure 4.
These sensors were designed for laboratory evaluation of a system to monitor
the length changes of ground anchor bars when subjected to tensile stress and
were based upon a quasi static microwave subcarrier approach with, in this
particular system, a length resolution and medium term accuracy of a few tens
of micrometres [11]. The results of one such test on this system is shown in
Figure 5. This figure also compares the results for a surface mounted strain

gauge attached to the ground anchor bolt. The strain gauge indicates a slightly
greater elongation corresponding to the strain relief anticipated within the
polymer based packaging through which the optical fibre instrument was
interfaced to the bolt itself.

Page 36

Figure 5
Results from microwave sub carrier sensor

A simple system of this nature only measures optical fibre delay and this is of
course dependent not only on strain but also on temperature. In all cases
therefore some form of temperature referencing is required since in very broad
terms a change in temperature of 1C is equivalent to a change in strain of 10
microstrains and this is in general true for all optical delay based strain
measuring systems within an optical fibre, regardless of the delay monitoring
technique. It is however a relatively straightforward operation to correct
temperature fluctuations and Figure 6 shows some corrected measurements
using the same simple microwave bridge approach and demonstrating that
very good agreement can be obtained between strain (or strictly total length)
measurements over a wide range of temperatures.

Figure 6:
Temperature Compensated Strain Measurement

Page 37

The results quoted above all apply to quasi static measurements in effective
bandwidths of a very small fraction of 1 Hz. One of the principal positive
features of this microwave subcarrier based interrogation system is that it is
simple to configure in a dynamic mode with bandwidths of 10s or 100s of Hertz
which easily accommodate most of the vibrational signatures applicable to civil
engineering structures. In this mode submicron short term resolution is
relatively straightforward to implement corresponding to sub-microstrain
sensitivity on typical gauge lengths of the order of a few metres. It is
interesting that whilst this sensitivity has been demonstrated in the laboratory
its implications in the civil engineering context and the potential offered by
these dynamic measurements have yet to be fully exploited.

Figure 7
Basic features of the SOFO system

The SOFO system shown schematically in Figure 7 uses white light

interferometry to monitor the differential path between a reference and
strained fibre mounted within a single tube. Strain is transferred to the
measuring fibre via the two end clamps on the tube structure. The gauge
length can vary from a metre or less up to approximately 50 metres and the
measurement accuracy and resolution is better than 10 micrometres over a
period of years. In principle several (perhaps up to 6) sensor lengths can be
mounted in series though the total path differences must lie within the
capability of the white light interferometer and additionally each sensor within
this sequence must operate within a specific optical path difference range.
This particular system is probably the most extensively tested of all fibre optic
based civil engineering instrumentation [12]. It has been installed particularly
in bridges and dams primarily in Switzerland and Austria and a considerable

amount of field experience has been accumulated. Its uses have included
monitoring reconstruction processes (for example in Figure 8) and it is capable
of observing

Page 38

long term shift with excellent accuracy. The principal issue in its practical
realisation has been the design of the sensor element. The feedback from early
field experience has now produced sensors which are capable of surviving the
rigours of installation in vibratory compacted concrete and can be handled on
site without excessive precautions. This particular system is a very convincing
demonstration that fibre optics does indeed have something to offer within the
civil engineering industry.

Figure 8:
Some typical results from measurements with the SOFO system

Distributed Chemical Measurements Using Optical Fibres

Corrosion is a major factor in the gradual deterioration of the civil infrastructure
and a necessary, though not always sufficient, condition for corrosion to take
place is water ingress. Water also plays a pivotal role in many construction
processes, for example in determining the properties of mixes of grout and
concrete and in ascertaining the stability of foundations. Further its presence
or absence can be used to, for example, confirm the thoroughness or otherwise
of the grout injection process. Water penetration through cracks, holes and
seepage is extremely important and therefore detecting the presence or
absence of water could be a useful start in monitoring the integrity of a civil
engineering structure.
Since water penetration can occur at any time and any place then by
implication large area monitoring is essential and so a distributed architecture
would be preferred. There are a few electrical systems which approach this
requirement and indeed use electrical time domain reflectometry as the
interrogation mechanism. However these are all limited in range to a few
metres and are therefore restricted in general application. An optical fibre
system which

Page 39

is capable of both detecting the presence or absence of water and indicating

regions of high humidity (and also discriminating between the two) is shown in
Figure 9 [13]. This system relies upon a mechanically induced microbend loss
which is monitored using an optical time domain reflectometer as indicated in
Figure 10. The slope of the OTDR trace is a direct indicator of the moisture
level and reaches a maximum when the cable system is immersed. The cable
can be used in lengths of up a few kilometres and will locate the position of a
wet zone to within a few metres using conventional optoelectronic
instrumentation systems.

Figure 9:
Basic features of the hydrogel based microbend inducing
distributed water detection cable

The sensor assembly has been cabled and evaluated in a number of civil
engineering trial systems. Typical results for one such trial is shown in Figure
11. Here the sensor was inserted in a 12 metre duct which was filled with a
grout of the same type used in post-tensioned concrete construction
processes. The sensor then monitored the drying process in the grout. The
measurements demonstrated that the grout within the central region of the
duct dried very significantly more slowly than that at the ends but that the
drying process could be accelerated by removing small sections of the duct wall
and exposing the grout within (which had long since set hard) to the air
outside. Whilst these results are, with hindsight, unsurprising they did provide
a new insight upon this particular construction process. The same sensor
system has also been used to evaluate the diffusion characteristics of moisture
in soils and for example has demonstrated that the equilibrium process in a
small volume (less than 1 cubic metre) of certain types of soils may take
several weeks to establish since this diffusion process can be extremely slow.
Perhaps the most interesting feature of this sensor is that the mechanically
induced microbend can be arranged to respond to a wide variety of chemical

Page 40

Figure 10
Representative OTDR trace for microbend sensor
High. low etc refer to value of imposed measurand

Figure 11:
Moisture level measurements on a 12m. grouted duct
showing the drying process over a period of weeks

stimuli. The first experiments have focused upon the detection of water since
this is undeniably a most important agent in the deterioration process of the
civil infrastructure. However the process can also be sensitised to pH changes
(making it more selective to determine for example corrosion in reinforcing
bars) or can be made sensitive to liquid hydrocarbons thereby enabling on-site
continuous leakage monitoring in fuel storage stations and pipelines. Whilst
there are certainly other

Page 41

distributed techniques for chemical sensing these other techniques all rely
upon direct interaction between the light in an optical fibre and the species to
be measured and are therefore vulnerable to surface effects and environmental
poisoning of intermediate chemicals. These chemical to mechanical systems in
contrast protect the light within the fibre. Since the basic chemicals have been
demonstrated to be highly stable in other applications (most notably when
used in in vivo drug control implants), the selectivity and moreso the long term
stability of the intermediate chemistry is relatively secure.
4. Data Handling
There is relatively little to say on this subject at least within the context of this
paper. However the general issue of data handling in large scale structural
monitoring systems is critical to their success. There is simply no virtue
whatsoever in collecting data and not being able to interpret it. Consequently
the interaction between the data acquisition system and the data analysis
through a suitable structural model is extremely important and must be
optimised [14,15]. The important information is whether or not the structure
concerned is in distress or potentially in distress and what to do about it if this
turns out to be indeed the case. In the context of this paper I believe it is
important to mention this as a critical issue and also to indicate that some of
the research work with which we have been involved has led along this
direction. Data interpretation is a critical feature in sensor system design.
5. Conclusions
Fibre optic sensors certainly have a role to play in the future of instrumentation
in civil engineering. Some basic sensing technologies have been demonstrated
and many more are emerging from laboratories worldwide. What is required in
the future is more system demonstrators, more confidence building exercises in
a real engineering environment and greater appreciation within both user and
technology communities of what the technology can offer and how it may be
best exploited.
6. References
1. Dunker, K. and Rabbat, B. S., March 1993. ''Why American's Bridges are
Crumbling" Scientific American, 6670.
2. For example, on 21 October 1994, a 48m section of a bridge over the Han

river, Seoul, Korea sheared neatly and floated after falling in the water.

Page 42

3. Culshaw, B. 1996. Smart Structures and Materials, Chapter 3, Artech

House, Norwood Ma.
4. Horiguchi, T., in Culshaw, B. and Dakin, J. P. (eds) 1997. Optical Fibre
Sensors, Vol IV, Artech House, Norwood, Ma.
5. Ferdinand, P. et al, October 1997. Application of Bragg grating Sensors in
Europe, Proc OFS(12), Williamsburg, Virginia, 1419, October Society of
America, Washington D.C. [This is one of several papers on Fibre Bragg
Gratings in this proceedings - they are extensively covered in all current
conferences on optical fibre sensors.]
6. Bhatia, V. et al, October 1994. Application of "absolute" fibre optic sensors
in smart materials and structures, Proc OFS(10), Glasgow Scotland, Proc SPIE,
Bellingham, Washington, Vol. 2360, 171.
7. Garside, B in Culshaw, B. and Dakin, J. P. (eds) 1997. Optical Fibre Sensors
Vol III, Artech House, Norwood Ma.
8. Noharet, B. et al, October 1994. Microwave Sub-carrier optical fibre sensor,
Proc 2nd European Conference on Smart Structures and Materials, Glasgow,
SPIE Vol 2361, 236239.
9. Inaudi, D., Elamari, A and Vurpillot, S, October 1994. Low coherence
interferometry for the monitoring of civil engineering structures, Proc 2nd
European Conference on Smart Structures and Materials, Glasgow, SPIE Vol
2361, 216219.
10. Thursby, G. J., Michie, W. C., Walsh, D., Culshaw, B., Sept 1995.
Simultaneous recovery of strain and temperature fields by the use of two
moded polarimetry with an in-line mode/splitter analyser, Optics Letters 20,
19191921
11. Michie, W. C. et al, February 1996. Optical fibre sensors for the monitoring
of structures, (OSMOS), Proc Smart Sensing, Processing and Instrumentation,
San Diego, Proc SPIE Bellingham, Washing, Vol 2718, 385397.
12. Inaudi, D., October 1997. Field testing and application of fibre optic
displacement sensors in civil structures, Proc OFS(12), Williamsburg, Virginia,
596599 (Optical Society of America).
13. Michie, W. C. et al, October 1997. Fibre optic sensors for distributed water

ingress detection and humidity measurement, Proc OFS(12), Williamsburg,

Virginia, 634637 (Optical Society of America).

Page 43

14. Staszewski, W. J. et al, Jul 1997. Wavelet signal processing for enhanced
Lamb-wave defect detection in composite plates using optical fiber detection,
36(07), 18771888.
15. Sne Rezeki, S. M. et al, 1997. Self diagnosis and self calibration strategies
for distributed intelligent sensor systems, Proc SPIE, Vol 3042, 344351, San
Diego.

Page 44

Using Bragg Grating Sensor Systems in Construction Materials and Bridges:

Perspectives and Challenges
J. S. Sirkis
Abstract
This paper examines the challenges of using arrays of multiplexed Bragg
grating sensors to monitor the health of large civil engineering structures made
from construction materials. Issues addressed include the effects of strain
gradients and non-uniform strain states on sensor performance, proper sensor
configuration and gage length selection, Bragg grating reliability, the trade-off
between numbers of multiplexed sensors and measurement bandwidth, and
others. Those characteristics that presently meet the requirements of large
structure are described, and those requiring further development are
identified.
Introduction
Social and economic pressure is growing to provide on-line health monitoring
capability to large civil engineering structures in order to guarantee safety and
to optimize maintenance [1,2]. This pressure is driven by the increasing
deterioration of the world's infrastructure. For example, 236,000 of the
approximately 576,000 bridges in the United States are considered potentially
unsafe [3]. The degradation of many of these bridges results from overuse,
chemical attack, inadequate maintenance, or mistakes in planning and
construction [4]. Visual inspection is the most prevalent approach to
monitoring structural well-being, but budgetary and man-power mean that the
inspections are infrequent and often cursory. More quantitative monitoring
methods in use by inspectors include vibrating strings, mechanical
extensometers, triangulation, etc. These methods often require special
operators, generally offer limited accuracy, and the high cost of obtaining data
often limits the frequency in which data is obtained. Fiber optic sensors offer a
viable and promising option for cost effective, remote and on-demand "health
monitoring" function. Once specific class of fiber optic sensors Bragg grating
sensors show particular promise.
James S. Sirkis, Smart Materials and Structures Research Center, University of
Maryland, College Park, MD 20742

Page 45

Fiber optic sensors have been embedded into a wide variety of concrete
structures [5] and bonded and/or welded to an equally impressive number of
large steel structures [6]. Examples include highway viaducts, damns, bridges,
nuclear power plants and others [5,7,8]. Interestingly enough, some of the
earliest examples of fiber optic sensors used with civil engineering structures
involved Bragg gratings (the Calgary Bridge, for example [9]), but other sensor
types have dominated the community of late. Long gage length, low finesse
Fabry-Perot sensors are the undisputed champion of in-service demonstrations
of optical fiber sensor technology. This is due in large part to the
commercialization efforts by Inaudi and coworkers [5,1012]. Now that low cost
(on a per sensor basis) Bragg grating readout instrumentation is becoming
readily available, Bragg gratings are finding greater use in civil engineering
applications [1316].
The two critical qualities that are making Bragg grating sensors popular for civil
engineering structures are multiplexing and self-referencing. Multiplexing
enables many grating sensors to be interrogated using common optoelectronic
instrumentation. By sharing the processing optics and electronics, multiplexing
reduces the cost per sensor, reduces the overall weight, and enhances the
robustness of the system. On the other hand, self-referencing refers to the fact
that measurements can be made relative to the time the sensor is
manufactured, and are not interrupted if the electronic instrumentation is
turned off. In an analogy to resistance strain gage technology, Bragg grating
sensors do not require "bridge balancing." This feature is invaluable in
applications involving measurement over days, months, or years, as is
commonly required for dams, bridges, buildings, etc. Bragg grating sensors
have potential drawbacks as well. First of all, they are intrinsic devices, and are
therefore susceptible to high thermal sensitivity and birefringence. Bragg
gratings also have a unique, but not always desirable, response to gradients in
strain and temperature along the grating length. In addition, Bragg gratings
share a series of important technical requirements with other fiber optic
sensors intended for use with construction materials. These include:
Long term stability
Adequate sensitivity and measurement bandwidth
Packaging suitable for construction sites

Reliable and repeatable embedding/attachment procedures

Immunity to high alkalinity and humidity
Durable over the intend life time of the structure
These and other issues will be discussed in the sections that follow.
Bragg Grating Sensor Background
In-line fiber Bragg gratings were developed as narrow band optical filters for
the telecommunications industry [17]. Recent television commercials by the
major telephone companies describe the capability of transmitting as many as
16 telephone calls on the same optical fiber link. This is made possible with
wavelength division

Page 46

multiplexing using Bragg grating or equivalent spectral filtering technology.

This intrinsic multiplexing capability also renders in-fiber Bragg gratings one of
the most promising distributed sensor technologies for large structural systems
commonly built by the construction industry.

Figure 1.
General Concept of Bragg Grating
Fabrication.

Bragg gratings are fabricated by exposing germanium doped optical fibers to a

periodic intensity profile produced by coherent interference. The refractive
index of the optical fiber core changes where the intensity is brightest to
produce a periodic refractive index profile [18,19], as illustrated in Figure 1.
The pitch of the grating, L, is controlled during the manufacturing process, but
is typically around 0.5m. Accordingly, there are ~20,000 periodicities in a
typical 1cm long grating. The amplitude of the periodicity is only the order of
0.1% to 0.01% of the original refractive index, such that the Bragg grating
appears to be an ordinary optical fiber to the human eye [18,19]. The
operating principles of ideal Bragg grating sensors are illustrated in Figure 1,
where light from either a broad band or a tunable laser source interacts with
the grating. The cumulative effect of optical scattering from the many
refractive index periodicities is to reflect a single wavelength, called the Bragg
wavelength. The Bragg wavelength is related to the grating pitch, L, and the
mean refractive index of the core, n, by l B = 2Ln. Both the fiber refractive
index and the grating pitch vary with changes in strain (e zz) and temperature
(DT), such that the Bragg wavelength shifts left or right in wavelength space in
response to applied thermal-mechanical fields. For a Bragg grating sensor
bonded to the surface of a structure, the strain and temperature are related to
the change in the Bragg wavelength by

where as and af are the coefficients of thermal expansion of the structural

material and fiber, respectively, and z is the thermal-optic coefficient, and Pe is
the strain-optic coefficient [20]. This equation relates the measured changes in
wavelength to strain and temperature. However, it is important to remember
that this expression is absolutely restricted to situations where the strain is
uniform along the grating and where there is no stress-induced birefringence.
Non-uniform strain distributions and/or stress-induced birefringence easily
occur when Bragg gratings are embedded in concrete and when the sensor is
bonded to a structure near high stress concentrators like cracks, holes, bolts,
etc.
One important issue that can be gleaned from Eq. (1) is the strong
dependence the Bragg wavelength has on temperature. Consider the case
where the mechanical strain is zero, but the temperature change is non-zero.
If one is unaware of the temperature change, then the changes in the Bragg
wavelength will be

Page 47

incorrectly interpreted as the effects of strain, i.e. e zz = Dl B/(l B Pe) = [(as af) + z/Pe]DT. This is strain called "thermal apparent strain" and is the source
significant consternation in many practical applications. For standard Bragg
grating sensors bonded to a steel structure, the coefficients in Eq. (1)in front
of the strain is 0.79 and the coefficient in front to the temperature is 17.79
e/C. From these values, it is clear that 1C produces a thermal apparent
strain of 22.5 e. This high level of thermal apparent strain is unprecedented in
resistance strain gage applications, and must be accounted for through dummy
gages, or more advanced dual-parameter fiber optic sensor transducer designs
(discussed later in the paper).
Multiplexing
Multiplexing is by far the most critical advantage offered by Bragg grating
sensor technology. The potential benefits of this multiplexing are so farreaching, that they make the high thermal apparent strain issue noted above
less problematic in comparison. Sensor multiplexing is most often accomplished
by producing a fiber with a sequence of spatially separated gratings, each with
different grating pitches, Li, i = 1, 2, 3,, n as illustrated by the schematic of,
and actual spectrum from, eight serialized gratings provided in Figure 2. The
output of the multiplexed sensors is processed through wavelength selective
instrumentation such as a tunable optical bandpass filter [21,22]. In this case,
the reflected spectrum will contain a series of peaks, each associated with a
different Bragg wavelength given by l Bi = 2nLi, where l Bi and Li are the
Bragg wavelength and pitch of the ith grating, respectively. For example, the
measurement field at grating 2 in Figure 2 is uniquely encoded as a
perturbation of the corresponding Bragg wavelength, l B2. It is important to
understand that Bragg gratings can only reflect light that is present at the
Bragg wavelength. Therefore the spectral characteristics of the source are
critical in determining the operation characteristics of a multiplexed sensor
system. As an example of this, consider a typical strain sensing system that
uses a superluminescent diode operating at 850 nm range with a bandwidth of
35nm as the source. Assuming that no sensor will experience more than
10,000e, the central wavelength of each Bragg grating may vary up to 8.5nm
without overlapping each other. Consequently, the entire 35nm bandwidth can
only accommodate five independent strain sensors. Smaller strain ranges
and/or larger source bandwidths enable more gratings along one fiber to be
use, but the limit is ~20 gratings for most practical applications [23,24].

Figure 2.
Schematic Showing Serially Multiplexed Bragg Grating Sensors.

Page 48

Many efforts have recently been reported to increase the number of sensors in
the network. These techniques include a hybrid time and wavelength division
technique [25], the use of a mechanical optical fiber switch to address FBGs
along different fiber channels in turn [26], and Fourier transform processing
using an array of identical gratings [27]. While effective in increasing the
numbers of sensors that can be multiplexed to as many as 100, the sample
rate of each sensor is reduced in these techniques to ~1Hz and below. This
type of bandwidth is acceptable for many civil engineering applications, but
cannot accommodate more dynamic loads such as those caused by fast moving
trucks and trains [13,14], hydroelectric generators [28], earthquakes, wind
gusts or projectile impacts[68]. In addition, 60 to 100 gratings is an
insufficient number of sensors to provide full coverage of a complex, fault
critical structures like dams or bridges. It is possible to improve this sample
rate by simply employing separate detectors and processing electronics for
FBGs written into different fiber channels [29]. However, this approach is not
cost effective because the instrumentation weight, size, cost and complexity
inevitably rises in proportion to the number of sensors in the network.
Recently, Chen et al. introduced a variation of Askins et al.'s [23] Technique
called "digital spatial and wavelength domain multiplexing technique" [24,30].
This technique uses a diffraction grating with two-dimensional CCD array
technology to produce many spectrometers operating in parallel. In this way,
arrays of gratings in different fibers can be interrogated simultaneously. The
preliminary experimental system was capable of interrogating at least 175
FBGs, and the spectral resolution of 1.2pm was achieved over a range of
50nm. This was done with a CCD imager without random address capability,
which limited the sample rate to 25 Hz. Even so, this technology promises to
interrogate more sensors with greater bandwidth than is possible with any
current day technology.
Self-Referencing
Self-referencing enables accurate, long-term measurements to be made in
spite of the fact that the grating instrumentation does not remain on during
the life of the structure. Even though the interrupt immunity made possible
through self-referencing is an intrinsic property of Bragg grating sensors, it is
only available when not removed by the signal processing used in the Bragg
read-out instrumentation. For example, the interferometric Bragg grating read-

out out technique developed by Kersey et al. [31] is a highly sensitive

wavelength discriminator with relatively high frequency bandwidth capability,
but it is only capable of monitoring changes in wavelength and not the
wavelength itself. Therefore this technique trades self-referencing for sensitivity
and bandwidth. Even when signal processing techniques are used that
preserve self-referencing (like the scanning filter techniques [21,32,33]), one
must exercise care to provide accurate calibration wavelength references to
compensate for drift in the instrumentation [14,34]. In addition, the Bragg
wavelength is known to change in a time scale of years if not properly
annealed [35]. This drift can lead to erroneous strain readings, particularly if
the measurements are made over years. Therefore,

Page 49

one must pay attention to grating manufacturing and selection if they are
intended for long term measurement applications. It is worth noting that other
properties of Bragg gratings change with time as well. This issue is described in
more detailed in the section devoted to grating reliability.
Strain Gradients and Transverse Strains

Figure 3.
Illustration Showing the Effects Strain Gradients
Can Have on the Reflected Spectrum of a
Bragg Grating.

Care was taken in the preceding sections to examine those cases where the
strain field or temperature were uniform of the grating length. This was done in
order to clarify the physical nature of grating transduction mechanisms without
overly confusing the relevant issues. When gradients and/or nonuniform loads
are applied to Bragg gratings, the concept of a single Bragg wavelength
becomes obscured figuratively and literally. Consider Figure 3, which shows a
grating with a strain gradient along its length. If the strain distribution is
approximated in a piecewise constant manner, then each section of the grating
corresponding to constant step in the strain distribution will have a slightly
different spectral shift. In essence, there is now a series of cascaded gratings,
each with slightly different Bragg wavelengths. The cumulative effect of this
series of new Bragg gratings is a series of overlapping Bragg wavelengths.
These types of strain distributions can completely smear out the spectrum of a
Bragg grating to the point where no spectral feature associated with the
grating can be identified [36,37]. This behavior is illustrated by the spectrum
shown in Figure 4. This spectrum results from a 1cm long Bragg grating
embedded in a Kevlar epoxy composite subjected to a combination of bending
and contact stresses. This loading resulted in the strain distribution becoming
strongly non-uniform in all dimensions, which had the effect of obscuring all
resemblance of the normally expected grating spectrum. Notice that this
spectrum now has a spectral bandwidth of ~3nm, where it originally had a
bandwidth of 0.3nm. It is worth noting that several researchers are trying to
take advantage of this smeared spectrum effect to measure strain and

temperature gradients [3840]. Unequal transverse stress can also pose

problems for data interpretation because this loading case leads to a bifurcated
Bragg condition (two peaks instead of one) caused by stress-induced
birefringence [37].

Page 50

Figure 4.
Spectrum produced by a Bragg
grating subjected to strain
gradients along its length.

The effects of non-ideal strain distributions make it difficult to make meaningful

fiber measurements in both embedded and surface mounted applications.
Stress concentrations and the resulting complex strain fields associated with
material and geometric inhomogeneaties (aggregate and rebar in concrete, for
example) can yield tremendously complex spectra that will confuse all
automated Bragg grating read-out systems. The same thing can result in
surface mounted applications when the sensor debonds, or it is placed in a
location of high strain gradients. All of this discussion suggests that continued
development is required of Bragg grating strain transducers that keep the
strain distribution in the grating uniform even when the structural strain
distribution is not. In addition, optical methods that distinguish gradient effects
from nonuniform transverse strain effects should be developed further to
provide diagnostic tools for understanding why complex spectra are produced
[37].
Gage Length
It was noted in the preceding paragraph that Bragg grating sensors operate
best when no strain gradients are present. This suggests that one should use
gratings with lengths much smaller than the characteristic length of the
expected gradients. In steel structures, this requirement degenerates into an
exercise in engineering judgment when selecting sensor locations. In concrete
structures, the issue is much more complex because rebar and aggregate are
distributed through out the structure in an almost random fashion. As a result,
avoiding localized strain gradients is not always possible, particularly when the
sensors are embedded. A wide variety of techniques have been develop to deal
with strain gradient problems. One way of extending the gage length of Bragg
grating sensors is to bond them directly to rebar, and then covering them with

strain relieving coatings [41]. Another way is to attach gratings to 10cm to

20cm long metallic transducing rods with welded end anchors [7,8]. Another
approach is to embed gratings in precast bars of Portland cement (no
aggregate), and then embed the cured cement bar in the concrete structure
[14]. In another philosophical approach altogether, Bragg gratings can be
used as reflectors in long gage length Fabry-Perot or laser sensors [15].
Such modifications are not required for all applications, particularly when the
sensors are bonded to the surface of large steel structures. In this instance,
the issue of sensor density is always relevant. Localized cracks cannot be
identified without sufficient sensor density [68,69], but more sensors always
means higher costs. To illustrate the cost-spatial resolution trade-offs, consider
the following thought experiment. Assuming that a 5.0 cm spatial resolution is
required, then 20

Page 51

individual sensors would be required every meter, meaning that 2,000 such
sensors are required to measure the strain distribution over a 100m length of
structure. Aside from the implementational impracticability of such a
proposition with current technology, current market prices for traditional sidewritten in-fiber Bragg gratings dictate that the sensor costs alone for this
sensor array would exceed $200,000.
Temperature Compensation
The fact that Bragg grating sensors possess unacceptably high thermal
apparent strain sensitivity has been noted several times in the preceding
sections [3]. This thermal apparent strain sensitivity can be minimized using
dummy gages or using dual-parameter sensor designs capable of
simultaneously measuring strain and temperature. The dummy gage approach
follows the resistance strain gage example by using one Bragg grating sensor
bonded to a stress-free segment of the structural material [12,42]. This stressfree segment of material is located adjacent to the strain measurement location
such that both the active and dummy Bragg grating sensors experience
identical temperature fields. In theory, the dummy gage will produce a signal
proportional to thermal effects only, and therefore can be subtracted from the
active strain measurement gage. For success, the bonding processes for the
dummy and active gages should be identical and the dummy active gages
should experience temperature fields that are as close to identical as is
possible. Otherwise, strain errors on the order for 50e to 100e often occur.
These requirements suggest that using the dummy gage approach with
structurally embedded grating sensors may be impractical because of the
requirement for a stress free piece of material means that the dummy gage will
most likely not be embedded. This, in turn, suggests that it will be difficult to
guarantee that the dummy gage will experience the same temperature field
and/or bonding conditions as the active sensing Bragg grating.
In dual-parameter sensing approaches to temperature compensation, a fiber
optic sensor is devised to measure strain and temperature at the same time.
The temperature is used with knowledge of the relevant thermal expansion and
thermo-optic coefficients to remove all thermal apparent strain contributions
from the strain measurements. While the optics and/or sensor design of dualparameter sensors are often more complicated than using a dummy gage
approach, dual parameter sensors are more practical for large civil engineering

structures where it can be impractical to distribute stress free pieces of

construction material through out the structure. The dual parameter sensors
using Bragg gratings that have thus far been investigated can be grouped into
the following categories:
Two superimposed Bragg gratings with different wavelengths [43,44]
Cascaded or superimposed Bragg grating and long period grating [4547]
Cascaded Bragg grating and extrinsic Fabry-Perot sensor [4850]
Two cascaded gratings with different wavelengths and with one grating not
glued to the structure [51]
Grating rossettes [52]

Page 52

The primary concern when developing dual parameter sensors is to design a

sensor that yields two independent equations for the unknown strain and
temperature. Unfortunately, the vast majority of dual-parameter sensors thus
far developed have been tested by heating them and stretching them between
two fixed points. This type of testing, calibration, and subsequent evaluation
omits significant contributions from the structure's thermal expansion to the
overall thermal sensitivity of the sensor (see Eq. 1). The systems of equations
that must be solved for strain and temperature are often marginally
independent when the structural thermal expansion is omitted, only to become
dependent (no unique solution) when the expansion is added. In addition, the
accuracy of the strain and temperature separation will be different for different
structural materials because of the dependence on the structural thermal
expansion. Fortunately, the accuracy estimates of all fiber optic dual-parameter
sensors can be derived from published test data [53]. Suffice it to say that of
the Dual-parameter sensors involving Bragg gratings described in the literature
so far, only three have been sufficiently tested to fully assess their accuracy
when bonded to or embedded within structural materials. These are (1)
cascaded Bragg grating/in-line fiber etalon (ILFE) [4950], (2) cascaded Bragg
grating/extrinsic Fabry-Perot interferometer [48], and (3) two serialized Bragg
gratings with different wavelengths in which one grating is glued to the
structure and the other is not [51]. While the ILFE and EFPI concepts severely
limit the numbers of sensor that can be multiplexed, they are the only dual
parameter sensors involving gratings that have been demonstrated in
embedded applications. The serialize arrangement involving glued and unglued
gratings seems to be the most practical approach in that it provides reliable
strain and temperature measurements, while at the same time preserving
multiplexibility. The only drawback to this approach is that is not amenable to
embedded applications without further transducer design to decouple the
temperature sensor from the structure. A recent entry into the temperate
compensated grating arena is the sixty degree rosette [52], which separates
the temperature from the three sensors. This technique has been properly
tested and shows promise, but is limited to surface mounted applications.
It seems that there is no clear winning technical approach when it comes to
temperature compensation, and that further work is still needed to determined
the best technique. Any technique that is developed must provide accurate
measurements, preserve multiplexibility, and be capable of operating while

embedded or surface mounted.

Reliability
Any fiber optic sensor system intended for use in construction cites or in
locations where high levels of human traffic are anticipated must be
ruggedized to the point where it is irrelevant if the construction, maintenance
personal, or pedestrians are aware of the monitoring system. Otherwise
construction and maintenance costs will be negatively impacted or the sensors
will be irrevocably

Page 53

damaged [7]. To understand the importance of this, consider the extreme

example of reinforced concrete structures. These structures are first reinforced,
formed, then the concrete is pumped and vibrated. All are clearly agressive
procedures and point towards the need for high mechanical reliability in the
sensor system. In addition, concrete places rather unique demands on
embedded optical fiber because of high alkalinity and because of the large
thermal gradients that can occur during the initial curing process. There are
also issues of optical reliability to contend with as well. Any optical affect (and
there are several) that obscures the grating characteristics or limits the optical
power budget must also be properly defeated. Fortunately, the
telecommunications systems for which Bragg gratings were originally
developed are expected to last upwards to 35 years, so it should not be
surprising to find that Bragg gratings have been subjected to a battery of
humidity, temperature and strain cycling tests. Many of the
telecommunications results translate directly to construction material
applications, but there are a few instances when civil engineering applications
place unique demands on gratings in particular, and optical fibers in general.
Mechanical Reliability
Any distributed optical fiber sensing system used with large, hard to construct
structures faces challenges with regard to reliability and survivability. As will be
explained below, these issues can be addressed with proper attention to fiber
and component selection. Optical fibers are made from amorphous (glassy)
quartz, which is an intrinsically low fracture toughness material. Low fracture
toughness often implies low mechanical strength. However, optical fibers have
very high mechanical strength, approaching a factor of three higher than the
strongest carbon steels. These apparently contradictory statements (low
fracture toughness and high strength) arise from the fact that optical fiber
manufacturing conditions are so well controlled that the initial surface flaws are
on the atomic scale. When the fiber is properly coated, these very small flaws
rarely exceed the critical length required for unstable crack growth. The
highest strength fiber (proof tested above 200 ksi) is fabricated in Class 1
laminar flow hoods and is immediately coated with a high toughness materials
to protect the fiber from handling-induced surface flaws. Examples of using
high strength optical fiber in harsh operating conditions include pay-out fiber in
anti-tank missiles and in transatlantic communications cables [54,55]. All of
this discussion suggests that optical fiber failure due to axial stresses can be

mitigated by selecting high strength optical fiber.

Choosing a fiber that has high axial strength does not fully protect it from other
important failure mechanisms like crushing, stress corrosion, and surface
oxidation. There is always a potential for impact loads from accidental impacts
(e.g., dropped tools and other similar episodes). Fortunately, a vast array of
small diameter fiber optic cable structures have been developed by the
telecommunications industry to protect against these types of impacts. This
cables usually include a combination of a ~3mm diameter hard polymer
coating extruded over standard coated optical fiber. The fiber assembly is then
usually encased in a structurally reinforced loose fitting tertiary tube [54,55].
Using cable structures can

Page 54

sufficiently protect the optical fibers the sensor system, but the grating sensors
must be exposed to a certain degree in order for the strain to be transferred
from the structure to the optical fibers. Fortunately, the 3mm diameter hard
coating offered by some manufacturers provides sufficient stiffness and
adhesion to provide good strain transfer, while at the same time protecting the
grating form transverse loads. In situations where using a hard coating cannot
be used, the grating must be protected using some other means.
Using high strength optical fiber mitigates many of the mechanical reliability
concerns associated with using optical fiber sensors systems. This assertion is
invalid if the Bragg gratings form weak links in the optical fiber. There is some
evidence to suggest that the optical aspects of writing Bragg gratings have an
effect on the grating strength [56,57]. This data suggests that the degradation
in the fiber strength is on the order of 50%, and is due to the Bragg grating
writing process when a high energy pulsed excimer lasers are used. This
conclusion is not universally accepted, as there have also been evidence that
suggest just the opposite is true, particularly when the gratings are made at
the time of fiber manufacture [58]. It is clear, however, the process of
stripping the protective coating before writing the grating and then recoating it
afterwards can significantly reduce strength if not done properly. Highest
strength gratings are usually achieved when the grating is stripped using
chemical techniques and carefully recoated. Commercial high strength Bragg
gratings made this way are available with proof strains approaching 7%, which
translates to maximum operational strains on the order of 2.5% to 3%.
Moisture absorption is known to break down the covalent bonds in the
amorphous silica dioxide optical fibers [59]. This breakdown is known as stress
corrosion because it is accelerated in the presence of non-zero stress states.
Environments in building construction will subject optical fiber sensor systems
to moisture, and therefore steps may be required to prevent moisture
absorption by the optical fiber. Certain locations in dams, for example,
experience a constant state of 90 percent relative humidity. Fortunately, stress
corrosion is well known to the optical fiber community and has been effectively
defeated through the use of hermetic coatings, such as gold and carbon. Offthe-shelf hermetic coatings are offered by many fiber vendors [59]. One
popular hermetic coating incorporates a 50 nm amorphous carbon hermetic
coating. This is done by depositing a carbon outer layer onto the fiber preform
before manufacturing the fiber. This coating cannot be stripped; therefore,

writing gratings through the carbon coating is unlikely to be successful.

Hermetic seals are also produced using metallic coatings (aluminum or gold are
common). Stripping can be accomplished using an Aqua Regia acid stripping
technique; however, recoating over the gratings is difficult and has not been
demonstrated in the open literature. New polymer hermetic coatings are being
developed that will soon make it possible to recoat gratings. This will enable
metal hermetic coatings to be stripped, gratings to be written, and then
recoated with a polymer hermetic coating. It is worth pointing out that
gratings can be made while manufacturing the fiber, and then immediately
coated with hermetic materials. This would yield hermetically coated gratings
without resorting to stripping and

Page 55

recoating. However, commercial gratings are not yet manufactured on the

draw tower, so we are left to explore recoating options.
The issue of coating selection is more complex when concrete is the used as
the construction material because of its high alkalinity. Mixing cement with
water leads to a hydration reaction, which produces calcium hydroxide as one
of its by-products. It is this calcium hydroxide that is responsible for the high
alkalinity of concrete. Coatings capable of fully protecting optical fibers from
high alkali have not yet been identified, although there are ongoing tests
designed to do so [7]. Acrylate, polyimide, fluorine thermoplastic, and Tefzel
coatings are among those that have been investigated, and all experienced
some degradation (cracks, crazing, etc.) as a result of alkalines and hydration
[60,61]. The issues here are two-fold. The coating must not only prevent
surface flaws and what absorption into the fiber, it must also provide sufficient
strain transfer [61]. As is often the case, these requirements can sometimes be
competing. The tests so far conducted rule out only acrylate coatings, but most
of the data published so far covers relatively short exposure times (on the order
of months) and does not include hermetic coatings. Clearly, further work in this
regard is need if high reliability fiber optic sensors systems intended for use the
concrete are to be developed.
Optical Relaibility

Figure 5.
Change in Bragg Wavelength as a
Function of Repeat Temperature
Cycles Between 20C and 427C Every
Four Hours.

It is well known that the reflectivity of Bragg Gratings reduces with time when
the grating is placed at elevated temperatures [6264]. It is not entirely clear
why the grating properties experience this relaxation, but it is thought to be
related to the diffusion of undisolved hydrogen in the fiber core.
Experimentation has found that the grating properties depend on such factors

as the (1) residual concentration of the hydrogen at the time of exposure, (2)
amount of core hydrogen depletion caused by UV exposure, (3) time, and (4)
temperature. The general trend is that the reflectivity, bandwidth, and Bragg
wavelength will initially change very rapidly when exposed to elevated
temperatures. This initial period is followed by a long, but sustained, decay to
an asymptotic value. In order to achieve thermal stability at higher
temperatures, the gratings must be annealed at temperatures exceeding the
intended operating temperature. Annealing the grating relaxes the portions of
the grating properties that would change over the lifetime of the grating, thus
leaving only the very stable portion of the grating properties. For example,
gratings intended for use at 427C are typically annealed at 650C. If used at
temperatures exceeding 650C, these gratings will experience some degree of
drift in the grating properties, which could then be falsely interpreted as effects
of

Page 56

strain. Humidity is also a concern because it cannot only lead to stress

corrosion, but it can also alter the absorption characteristics of the fiber in
certain spectral regions. Environmental tests intended to investigate how
temperature and humidity effect the optical properties of gratings revealed
limited influence on grating characteristics [6567]. No change in grating
properties were observed after exposure to 1000 hours at a temperature of
85C at 85% relative humidity, nor were any changes in grating properties
observed after 1000 thermal cycles from - 40C to 85C or 512 cycles from
21C to 427C (Figure 5). Mechanical testing done to date also indicates no
change in grating properties, even after 1.4 million strain cycles from 0 to
2,500e
Summary
This paper has reviewed some of the important characteristics of Bragg grating
sensors, and how these characteristics can impact the implementation of Bragg
grating-based health monitoring systems in large civil enineering systems.
Much more work is still needed, particularly in the following areas:
Increasing the numbers of sensor that can be multiplexed with a single Bragg
grating read-out instrument without sacrificing measurement bandwidth.
Develop transducer designs that will prevent strain gradients and/or transverse
strains from reaching the Bragg grating, even if these strain states are present
in the structure.
Develop remote optical diagnostic techniques to understand why inaccessible
gratings become inoperable.
Develop rational approaches for balancing the trade-offs between sensor costs
and greater measurement fidelity afforded by using more sensors.
Reduce Bragg grating sensor costs. Keep in mind that the sensor costs most
often quoted do not apply to gratings with specialized coatings (hermetic,
etc.).
Develop new concepts for extending grating gage length without sacrificing
multilpexibility, accuracy, or sensitivity.
Develop reliable and accurate temperature compensation techniques for Bragg
grating sensors, especially for applications where sensors are embedded. These

sensor designs should have limited impact on sensor cost and multiplexibility.
Develop and test hermetically coated Bragg gratings.
Investigate coatings that protect Bragg gratings from high alkaline
environments.

Page 57

References
1. K. F. Dunker and B. G. Rabbat, ''Why America's Bridges are Crumbling,"
Scientific American, March, pp. 6672, 1993.
2. S. C. Liu, K. P. Chong, and M. P. Singh, "Civil Infrastructure Systems
Research: Hazard Mitigation and Intelligent Materials Systems," Journal of
Smart Materials and Structures, 3, pp. A69A74, 1994.
3. K.R. White, J. Minor, and K. N. Derucher, BRIDGE MAINTENANCE,
INSPECTION AND EVALUATION, Marcel Dekker, New York, 1992.
4. K. Lou, G. Yaniv, and D. Hardtmann, G. Ma, and B. Zimmermann, "Fiber
Optic Strain Monitoring of Bridge Column Retrofitted With Composite Jacket
Under Flexural Loads," Proc. Smart Systems for Bridges, Structures, and
Highways, SPIE Vol. 2446, pp. 1624, 1995.
5. D. Inaudi, "Field Testing and Application of Fiber Optic Displacement
Sensors in Civil Structures," Proc. 12th International Conference in Optical
Fiber Sensors, Williamsburg Va., pp. 596599, 1998.
6. W. Lee, J.Lee, C. Henderson, H. F. Taylor, R. James, C. E. Lee, V. Swenson,
W. N. Gibler, R. A. Atkins, and W. G. Gemeiner, "Railroad Bridge
Instrumentation With Fiber Optic Sensors," Proc. 12th International Conference
in Optical Fiber Sensors, Williamsburg Va., pp. 412415, 1998.
7. W. Habel and B. Hillemeier, "Results in Monitoring and Assessment of
Damages in Large Steel and Concrete Structures by Means of Fiber Optic
Sensors," Proc. Smart Systems for Bridges, Structures, and Highways, SPIE
Vol. 2446, pp. 2536, 1995.
8. V. Dewynter-Mary, S. Rougeault, P. Ferdinand, D. Chauvel, E. Toppani, M.
Leygonie, B. Jarret, and P. Fenaux, "Concrete Measurements and Crack
Detection with Surface Mounted and Embedded Bragg Gratings," Proc. 12th
International Conference in Optical Fiber Sensors, Williamsburg Va., pp.
600603, 1997.
9. R. M. Measures, A. T. Alavie, R. Maaskant, M. Ohn, S. Karr, and S. Huang,
"Structurally Integrated Grating Laser Sensing System for a Carbon Fiber
Reinforced Concrete Highway Bridge," Journal of Smart Materials and
Structures, 4, pp. 2030, 1995.

10. D. Inaudi, N. Casanova, P. Kronenberg, and S. Vurpillot, "Railway Bridge

Monitoring During Construction and Sliding," Proc. Smart Systems for Bridges,
Structures, and Highways, SPIE Vol. 3043, pp. 5864, 1997.
11. S. Vurpillot, N. Casanov, D. Inaudi, and P. Kronenberg, "Bridge Spatial
Displacement Monitoring With 100 Fiber Optic Sensors Deformations: Sensors
Network and Preliminary Results," Proc. Smart Systems for Bridges, Structures,
and Highways, SPIE Vol. 3043, pp. 5157, 1997.
12. Kronenberg, P., Casanova, N., Inaudi, D., and Vurpillot, S., "Dam
Monitoring With Fiber Optics Deformation Sensors," Proc. Smart Systems for
Bridges, Structures, and Highways, SPIE Vol. 3043, pp. 211, 1997.
13. M. A. Davis, A. D. Kersey, T. A. Berkoff, R. T. Jones, R. L. Idriss, and M.
Kodinduma, "Dynamic Strain Monitoring of an In-Use Interstate Bridge Using

Page 58

Fiber Bragg Grating Sensors," Proc. Smart Systems for Bridges, Structures,
and Highways, SPIE Vol. 3043, pp. 8795, 1997.
14. P.M. Nellen, P. Anderegg, R. Bronnimann, and U. Sennhauser, "Application
of Fiber Optical and Resistance Strain Gauges for Long-term surveillance of Civil
Engineering Structures," Proc. Smart Systems for Bridges, Structures, and
Highways, SPIE Vol. 3043, pp. 7786, 1997.
15. A. T. Alavie, S. E. Karr, A.Onthos, and R. M. Measures, "Fiber Laser
Sensing Array," Proc. Smart Sensing, Processing and Instrumentation, SPIE
Vol. 1918, 4, pp. 308318, 1993.
16. C. I. Merzbacher, A. D. Kersey, and E. J. Friebele, "Fiber Optic Sensors in
Concrete Structures: A Review," Journal of Smart Materials and Structures, 5,
pp. 196208, 1996.
17. P. B. Pal, FUNDAMENTALS OF FIBER OPTICS IN TELECOMMUNCATIONS
AND SENSOR SYSTEMS, John Wiley and Sons, N.Y., 1992.
18. G. R. Meltz, W. W. Morey, and W. H. Glen, "Formation of Bragg Gratings in
Optical Fibers by a Transverse Holographic Method," Optics Letters, 14, pp.
823825, 1989.
19. Hill, K.O., Malo, B., Bilodeau, and Johnson, D.C., "Photosensitivity in
Optical Fibers," Ann. Rev. Mater. Sci., 125, 1993.
20. Sirkis, J.S., "A Unified Approach to Phase-Strain-Temperature Models for
Smart Structure Interferometric Optical Fiber Sensors: Part I-Development,"
Optical Engineering, 32(4), pp. 752761, 1993.
21. A. D. Kersey, T. A. Berkoff, and W. W. Morey, "Multiplexed Fiber Bragg
Grating Strain-Sensor System With a Fiber Fabry-Perot Wavelength Filter,"
Optics Letters, 18(16), pp. 13701372, 1993.
22. M. A. Davis, D. G. Bellemore, T. A. Berkoff, and A. D. Kersey, "Design and
Performance of a Fiber Bragg Grating Distributed Strain System," Proc. Smart
Systems for Bridges, Structures, and Highways, SPIE Vol. 2446, pp. 227235,
1995.
23. C. G. Askins, M. A. Putnam, E. J. Friebele, "Instrumentation for
Interrogating Many-element Fiber Bragg Grating Arrays", SPIE Vol. 2444, pp.
257266, 1995.

24. S. Chen, Y. Hu, L. Zhang, I. Bennion, "Digital Wavelength and Spatial

Domain Multiplexing of Bragg Grating Optical Fiber Sensors," Proc. 11th
International Conference in Optical Fiber Sensors, Sapporo (Japan), pp.
100103, May 1996.
25. T. A. Berkoff, M. A. Davis, D. G. Bellemore, A. D. Kersey, G. M. Williams, M.
A. Putnam, "Hybrid Time and Wavelength Division Multiplexed Fiber Bragg
Grating Sensor Array," Proc. of Smart Structures and Materials Conference,
SPIE Vol. 2444, pp.288294, 1995.
26. Davis, M. A., Bellemore, D. G., Putnam, M. A., and Kersey, A. D., "A 60
Element Fiber Bragg Grating Sensor System", Proc. 11th International
Conference in Optical Fiber Sensors, OFS-11, Sapporo (Japan), pp. 100103,
May 1996.
27. L. Melvin, et al. "Integrated Vehicle Health Monitoring (IVHM) for Space
Vehicles", Proceedings of International Workshop on Structure Health
Monitoring, pp. 705714, 1997

Page 59

28. P. L. Fuhr, D. R. Huston, T. P. Ambrose, and D. A. Barker, "Embedded

Sensor Results From the Winooski One Dam," Proc. Smart Structures and
Materials Conf., SPIE Vol. 2191, pp. 446456, 1994.
29. Y. J. Rao, A. B. L. Ribeiro, D. A. Jackson, L. Zhang, and I. Bennion,
"Simultaneous Spatial, Time and Wavelength Division Multiplexed In-fiber
Grating Sensing Network", Optics Communications, 125, pp. 5358, 1995.
30. Y. Hu and S. Chen, "Multiplexing Bragg Gratings Using Combined
Wavelength and Spatial Division Techniques with Digital Resolution
Enhancement", Electronic Letters, 33, pp. 19731975, 1997.
31. A. D. Kersey, T. A. Berkoff, and W. W. Morey, "High Resolution Fiber
Grating Based Strain Sensor With Interferometric Wavelength Shift Detection,"
Electronics Letters, 29, pp. 112113, 1993.
32. H. Geiger, M. G. Xu, J. P. Dakin, N. C. Eaton, and P. J. Chivers, "Progress
on Grating Interrogation Schemes Using a Tunable Filter," Proc. 11th
International Conference in Optical Fiber Sensors, Sapporo (Japan), pp.
376379, May 1996.
33. L. Bjerkan, D. R. Hjelme and K. Johannessen, "Bragg Grating Sensor
Demodulation Scheme Using a Semiconductor Laser for Measuring Slamming
Forces of Marine Vehicles," Proc. 11th International Conference in Optical Fiber
Sensors, Sapporo (Japan), pp. 236239, May 1996.
34. C. Someda et al., "A Stable Accurate Highly Multiplex Fiber Bragg Grating
Interrogation System," Proc. of Electroottica, May 12, 1998.
35. S. Kannan, Z. Y. Guo, and P. J. Lemaire, "Thermal stability analysis of UVinduced fiber Bragg gratings," J. Lightwave Tech., Vol. 15, No. 8, pp.
14781483, August 1997.
36. R. B. Wagreich, W. A. Atia, H. Singh, and J. S. Sirkis, "Effects of Diametric
Load on Fibre Bragg Gratings Fabricated in Low Birefringence Fibre,"
Electronics Letters, Vol. 32, No. 13, pp. 12231224, 1996.
37. R. Wagreich and J. S. Sirkis, "Distinquishing Fiber Bragg Grating Strain
Effects," Proc. 12th International Conference in Optical Fiber Sensors,
Williamsburg Va., pp. 2024, 1997.
38. M. LeBlanc, S. Huang, and R. M. Measures, "Fiber optic Bragg intra-grating

strain gradient sensing," Smart Sensing, Processing, and Instrumentation,

SPIE Vol. 2444, San Diego California, pp. 136147, 1995.
39. S. Huang, M. M. Ohm, M. LeBlanc, R. Lee, and R. M. Measures, "Fiber
optic intra-grating distributed strain sensor," Distributed and Multiplexed Fiber
Optic Sensors IV, SPIE Vol. 2294, pp. 8192, San Diego California, 1994.
40. R. M. Measures, S. Huang, M. LeBlanc, M. Ohm, and A. T. Alavie, "Bragg
Intra-Grating Sensing - Implications for Smart Structures," Smart Sensing,
Processing, and Instrumentation, SPIE Vol. 2191, pp. 436445, Orlando Florida,
1994.
41. J. D. Prohaska, E. Snitzer, B. Chen, M. H. Maher, E. G. Nawy and W. W.
Morey, "Fiber Optic Bragg Grating Strain Sensor in Large Scale Concrete
Structures," Proc. Fiber Optic Smart Structures and Skins, SPIE Vol. 1798, pp.
286-194, 1992.

Page 60

42. J. W Dally, W. Riley, and J. S. Sirkis,, HANDBOOK ON EXPERIMENTAL

MECHANICS, Chapter 2, (A. S. Kobayashi, Ed.)The Society of Experimental
Mechanics, Bethel, VCF Publishing, N.Y., 1993.
43. M. G. Xu, J. Archanbault, L. Reekie, and J. P. Dankin, "Discrimination
Between Strain and Temperature Effects Using Dual-Wavelength Fiber Grating
Sensors," Electronics Letters, Vol. 30, No. 13, pp. 10851087, 1994.
44. E. Udd, C. M. Lawrence, and D. V. Nelson, "Development of a Three-Axis
Strain and Temperature Fiber Optic Grating Sensor," Proc. SPIE, San Diego,
1997.
45. S. E. Kanellopoulos, V. A. Handerek, and A. J. Rogers, "Simultaneous
Strain and Temperature Sensing with Photogenerated In-Fiber Gratings,"
Optics Letters, Vol. 20, No. 3, 1995.
46. H. Patrick, G. M. Williams, and A. D. Kersey, "Strain/Temperature
Discrimination Using Combined Fiber Bragg Grating and Long Period Grating
Sensors," OFS'96 Tu4-2, pp. 9699, 1996.
47. V. Bhatia, D. Campbell, T. D'Alberto, D. Sherr, K. A. Murphy, and R. O.
Claus, "Simultaneous Strain and Temperature Sensing Using Long-Period
Gratings," Proc. SPIE, San Diego, 1997.
48. G. F. Femando, T. Liu, Y. Rao, D. A. Jackson, L. Zhang, and I. Bennion,
"Multiplexed Fiber Bragg Grating Sensor and Extrinsic Fabry-Perot Sensor
System for Simultaneous Strain and Temperature Measurement," Proc. SPIE,
San Diego, 1997.
49. X. D. Jin, J. S. Sirkis, and J. K. Chung, "Optical Fiber Sensor for
Simultaneous Measurement of Strain and Temperature," Proc. SPIE, San
Diego, 1997.
50. H. Singh and J. S. Sirkis, "Temperature and Strain Measurement by
Combining ILFE and Bragg Grating Optical Fiber Sensors," Experimental
Mechanics, 37 (6), pp. 414419, 1997.
51. Y-L. Lo, "Using In-Fiber Bragg-Grating Sensor for Measuring Axial Strain
and Temperature Simultaneously on the Surface of Structures," to appear in
Optical Engineering, 1998.
52. S. Magne, S. Rougeault, M. Vilela, and P. Ferdinand, "Strain-of-Strain

Evaluation With Bragg Grating Rosettes: Application to Discrimination Between

Strain and Temperature Effects in Fiber Sensors," Applied Optics, 36 (36), pp.
94379447, 1997.
53. X. Jin, J. S. Sirkis, and P. Siva, "A Review of Fiber Optic Strain/Temperature
Sensors," in preparation.
54. J. Hayes, FIBER OPTICS TECHNICIANS MANUAL, Delmar Publishers,
Albany, N.Y., 1996.
55. S. Ungar, FIBRE OPTICS: Theory and Applications, John Wiley and Sons,
N. Y., 1989.
56. V. Varelas, H. G. Limberger, R. P. Salathe' and C. Kotrotsios, "UV-induced
mechanical degradation of optical fibres," Electronics Letters, Vol. 33, No. 9,
pp. 804806, 1997.
57. H. G. Linberger, D. Varelas, and R. P. Salathe', "Mechanical Reliability of
UV Irradiated Fibers: Application to Bragg Grating Fabrication," Proc. of Bragg
Gratings, Photosensitivity, and Poling in Glass Fibers a Waveguides:

Page 61

Application and Fundamentals, OSA Technical Digest Series Vol. 17,

Williamsburg, Va., pp. 4648, 1997.
58. C. G. Askins, M. A. Putnam, H. J. Patrick, and E. J. Friebele, "Fibre Stength
Unaffected by On-Line Writing of Single-Pulse Bragg Gratings," Electronics
Letters, 1997.
59. C. R. Kurkjian and D. Inniss, "Understanding Mechanical Properties of
Lightguides: A Commentary," Optical Engineering, 30(6) pp. 681689, 1991.
60. W. Habel, M. Hopcke, F. Basedau, an H. Polster, "The Influence of
Concrete and Alkaline Solutions on Different Surfaces of Optical Fibers
Sensors," 2nd European Conference on Smart Structures and Materials, IOP,
Bristol, pp. 168171, 1994.
61. C. K. Y. Leung and D. Darmawangsa, "Interfacial Changes for Optical
Fibers in A Cementitious Environment" submitted to the Journal of Lightwave
Technology, 1997.
62. T. Erdogan, et al., "Decay of Ultraviolet-induced Fiber Bragg Gratings,"
Journal of Applied Physics, 76, pp. 7380, 1994.
63. H. Patrick, S. L. Gilbert, A. Lidgard, and M. D. Gallagher, "Annealing of
Bragg Gratings in Hydrogen Loaded Optical Fiber," Journal of Applied Physics,
78 (5), pp. 29402945, 1995.
64. D. L. Williams and R. P. Smith, "Accelerated Lifetime Tests on UV Written
Intra-core Gratings in Boron Germania Codoped Silica Fibre," Electronics
Letters, 31 (24) pp. 21202121. 1995.
65. W. W. Morey, G. Ball, H. Singh, "Applications of Fiber Grating Sensors,"
SPIE Vol. 2839, Fiber Optic and Laser Sensors XIV, pp. 27, 1996.
66. 3M Fiber Bragg Grating Application Note, "The Mechanical and Optical
Reliability of Fiber Bragg Gratings," Feb. 1996.
67. U. Sennhauser, R. Bronnimann, and P. M. Nellon, "Reliability Modelling and
Testing of Optical Fiber Bragg Sensors for Strain Measurement," SPIE Vol.
2839, Fiber Optic and Laser Sensors XIV, pp. 264-75, 1996.
68. F. Ansari and R. K. Navalurkar, "Kinematics of Crack Formation in
Cementitious Composites by Fiber Optics," ASCE Journal of Engineering
Mechanics, 119 (5), pp. 10481061, 1993.

69. C. K. Y. Leung, N. Elvin, N. Olson, T. F. Morse, and h-f He, "Optical Fiber
Crack Sensors for Concrete Structures," SPIE Proc. on Smart Sensing,
Processing and Instruementation.

Page 63

Chapter 3
Distributed and Multiplexed Sensors

Page 65

Developments in Optical Techniques for Point and Distributed Sensing in Large

Structures
V. Lecoeuche,
N. E. Fisher,
C. N. Pannell,
D. J. Webb
and D. A. Jackson
Abstract
In this presentation, recent research at the University of Kent is reported on
the development of optical fibre based sensors capable of making temperature
and strain measurements with resolutions compatible with those required for
structural integrity measurements of bridges and similar structures. Initial
results are also reported on the development of an acoustic sensor based on an
'in-fibre Bragg grating', this sensor has a linear frequency response up to 2MHz
with potential for use with existing non destructive techniques to determine
the location of defects or cracks in concrete.
1 Introduction
The recent development of the in-Fibre Bragg Grating Sensor (FBGS) with its
unique ability to detect temperature and strain and the ease with which large
numbers of such sensors, deployed remotely, can be multiplexed, has lead to
the possibility of monitoring the structural integrity of major capital items such
as buildings and bridges. However, as bridges are extremely large structures,
the optimum strategy for sensor deployment and choice of sensor type is not
straightforward. Given the wide range of measurement requirements, for
example in monitoring the daily changes of static strain, it is necessary to allow
for thermally induced strains. Other questions such as how many measurement
points and how to attach the fibre sensors to the bridge are also crucial points
which need to be addressed. Also valuable information about the 'state of
health' of the bridge can be obtained from dynamic measurements using
ultrasonic techniques. Such sensors with flat frequency responses could have a
dual sensory role; a) for monitoring traffic flow (low frequency response) and
b) detecting the presence of 'cracks' or new 'defects' in the bridge.
V. Lecoeuche, N.E. Fisher, D.J. Webb, C.N. Pannell and D.A. Jackson, Applied Optics
Group, School of Physical Sciences, The University, Canterbury, Kent, CT2 7NR, UK

Page 66

The Applied Optics Group (AOG) at the University of Kent has been developing
a variety of optical fibre sensing techniques for diverse applications including
monitoring large structures such as bridges.
In this paper, we discuss the following sensing techniques and their potential
application for monitoring the structural integrity of bridges:
(i) Brillouin distributed sensor for temperature and strain sensing,
(ii) Point acoustic sensors, for non-destructive sensing and possible impact
detection.
2 Brillouin Distributed Sensor for Temperature and Strain Sensing
A truly distributed sensor seems the most reasonable means of addressing
applications where a huge number of measurement points over a very large
structure are required. Stimulated Brillouin Scattering is a unique parametric
interaction which offers a simultaneous sensitivity to temperature and strain.
The Brillouin process couples, through an acoustic wave, two counterpropagating light beams, which are frequency shifted by an amount dependent
on the optical and elastic properties of the medium. A Brillouin based
distributed sensor makes use of the temperature [1] or strain [2] dependence
of the Brillouin shift. In practice two beams, namely the pump and the Stokes
waves, are launched into both ends of the fibre. The measurement is derived
from the acquisition of the transmitted pump or Stokes signal (referred to as
the Brillouin loss or gain method respectively) as a function of the pump/Stokes
frequency shift. This reveals a Lorentzian profile typical of the Brillouin
interaction, the center frequency of which is a linear function of the
temperature and strain parameters. Positional information, which requires that
at least one of the beams be pulsed, is obtained through a standard time delay
analysis. The spatial resolution of the sensor is fixed by the duration of the
pulse, but is limited to about 1 meter by the finite response time of the
Brillouin interaction.
Since it was first proposed in 1989, the method has been the subject of
numerous improvements. We would like firstly to review in this paper the main
problems which have been addressed, and demonstrate that the Brillouin
based distributed sensor now provides an efficient means for monitoring
temperature and strain in large structures. Considerable efforts may however
be required in order to realize a cost effective and viable in field device. In this

context, we are currently investigating various configurations including a

Brillouin laser source, which will be described in the final parts of this section.
2.1 Choice Between the Brillouin Gain or Loss Technique
Three distinct techniques can be implemented in order to obtain a distributed
signal through the Stimulated Brillouin Scattering interaction:
(i) With a pulsed pump and a CW Stokes wave: the Brillouin gain method.
A pulsed pump provides gain to a CW Stokes wave while propagating in the
sensing fibre, the time dependent intensity of the transmitted Stokes therefore

Page 67

provides the distributed signal carrying the information on the Brillouin

interaction over the whole sensing length. However, the pump pulse is rapidly
depleted and the signal arising from the most distant location fades. The
longest sensing length reported so far using this method was 22 km [3].
(ii) With a CW pump and a pulsed Stokes wave: the Brillouin loss method. Here
the signal of interest is the loss induced by the Stokes pulse on the CW pump,
observable in the time dependent evolution of the transmitted pump intensity.
Since the Stokes pulse is amplified while it propagates, the signal is stronger at
the end of the fibre (most distant from the pulsed source). We have
demonstrated a resolution of 1C or 25 e and a spatial resolution of 12 meters
over a 50 km sensing length using this approach [4]. In this set-up, the
working wavelength was 1.3 m; even greater range can be expected in the
1.5 m transmission window.
(iii) With both pump and Stokes waves being pulsed.
Either the gain of the Stokes pulse or the loss of the pump can be monitored.
The highest resolution and accuracy can be achieved that way since the
measured signal only reflects the interaction of the pulses at their crossing
point. It was recently demonstrated how a strain applied over 80 cm of fibre
could be easily observed with that method [5]. Sensing lengths of a few
kilometers have been reported so far [6], the major drawback being that signal
acquisition for the whole sensing length takes much longer than with methods
(i) and (ii) (the scaling factor is L/DL where L is the sensing length and DL the
spatial resolution).
2.2 Signal Fading due to Polarization and Source of Systematic Errors
One of the initial problems which had to be addressed concerns the
polarization sensitivity of the Brillouin gain. Polarization maintaining fibre can
be used for relatively short sensing lengths (a few kilometers at most).
However a cost-effective sensor should employ a standard telecommunications
fibre. In that case, the polarization state changes randomly during the
propagation; furthermore, temperature or strain conditions influence this
evolution which therefore drifts in time. The fine details of the polarization
properties of the Brillouin interaction are well described in reference [7].
Obtaining a maximum signal from every part of the sensing fibre requires at
least two measurements with controlled polarization states for both the pump

and Stokes waves [4]. A recently proposed method involves averaging 3

measurements with states of polarization building up an orthogonal trihedron
on the Poincar sphere [8], ensuring an average maximum Brillouin gain
identical in every part of the sensor and equal to the 2/3 of the value for
parallel linear states.
Systematic errors on the measurement may arise if the temperature or strain
changes on a length scale shorter than the sensor resolution [9]. The
Lorentzian profile of the Brillouin gain may then enlarge and even develop into
a multi-peak structure. In those cases a centroid calculation would then give a
better estimate of the average temperature or strain than a Lorentzian fitting
routine.
The loss (or gain when the Brillouin loss approach is used) that the pump
(Stokes in the case of Brillouin loss) pulse undergoes before attaining a sensing

Page 68

point may also be a cause for systematic errors in the measurement. Since this
effect is much more important for a frequency shift matching the average
temperature or strain condition in the whole fibre, the effect alters the shape of
the gain profile, especially for the locations most distant from the pulsed
source. Geinitz et al [10] have recently proposed a numerical method to
calculate the contribution to the spectrum in order to correct the acquired
data. It should be noted that method (iii) described above completely avoids
this problem.
2.3 Single Ended Operation
A Brillouin interaction requires a pump and a Stokes wave propagating in
opposite directions in the sensing fibre. In the simplest configuration, the two
beams are injected at both ends of the sensing fibre. Although this method is
appropriate for custom-built sensor systems, it is much more difficult to use the
same arrangement for fibres which are already in situ. In this case it is more
desirable to launch both signals into the same end of the sensing fibre and
monitor the signal at the same end. This was first realized by Horiguchi et al
[11] and is also used in [6]. The backward propagating waves are simply
obtained from the Fresnel reflection at the end of the fibre. We suggested in
reference [12] the use of a Faraday rotating mirror spliced at the far end of the
sensor. This ensures a strong returned signal and most importantly eliminates
the need for active control of the state of polarization [13].
2.4 Advantages in the use of a Brillouin Laser Source
The major practical difficulty in realizing a Brillouin sensor is the requirement
for two narrow linewidth laser sources with a 12 GHz frequency difference
controlled to about 1 MHz. Several schemes have already been proposed in
order to synthesize this dual frequency source from a single one. The approach
taken in [14] involved a multi-pass fibre loop including a frequency shifter and
an amplifier to overcome the cavity losses. Ultra-fast electrooptic modulators
have also been used to produce the frequency shift, either using a carrier
frequency at half of the Brillouin shift [6] (note that this configuration is
restricted to method (iii) described above), or equal to the Brillouin shift [15,
16]. The solution that we proposed involves a Brillouin fibre ring laser [17],
which has the advantage of having the required frequency shift with respect to
the pump source, and can be designed to produce a highly stable train of
pulses. This laser could be implemented either in methods (ii) or (iii) described

above and avoids the use of ultra-fast electrooptic devices. Furthermore, we

have shown that for alarm applications, where it is only required to detect an
unsafe level of temperature or strain, the control of the frequency shift is not
required. In this case, the data acquisition is much faster (less than 3 seconds
for a 15 km sensing length) and the experimental arrangement very simple. A
spatial resolution of 8 meters has been demonstrated so far with this approach,
but the sensor can provide signals even when shorter lengths are subject to
heat or strain, and we expect a 1 meter sensitivity with minor changes in the
arrangement.

Page 69

Figure 1:
Experimental setup. PC: Polarization Controller, DC: Directional Coupler
(1: 55/45%, 2-5: 87/13%), D: Detector, AOM: Acousto-Optic Modulator,
EOM: Electro-Optic Modulator, OI: Optical Isolator, PZT: Piezoelectric.

Figure 2:
Distributed gain profile of 50 meters of
the sensing fibre.

An experimental set-up suitable for actual measurement of temperature or

strain is depicted in figure 1. The details of the Brillouin fibre ring laser and the
gating circuit are described in [17]. We have simply modified the arrangement
to allow for the frequency tuning of the Stokes wave; this is obtained through a
phase modulation applied during a short period, synchronised to the
occurrence of a pulse out of the Brillouin laser. Since only the first sideband
should interact with the pump over the whole tuning range, the frequency of
the modulation has to range around 200 MHz at least. In order to compensate
this mean shift, the pump launched in the sensing fibre was shifted by 200
MHz using an acoustooptic modulator. We demonstrated the efficiency of the
technique with a 700 meter long sensing fibre. Two lengths of 10 meters in the
middle of the sensor were subjected to 0 me and 74C, and 1.4 me at ambient

temperature. Another 10 meters of loose fibre at ambient temperature

separated them. The result of the frequency scanning for the region of interest
is

Page 70

depicted in figure 2. Two spikes can be easily identified in the foreground (or
left front) of this 3D plot, characterized by a higher Brillouin shift. The one on
the left corresponding to the heated fibre, the other corresponding to the fibre
under strain.
Although this preliminary result demonstrates the feasibility of a sensor
incorporating a Brillouin laser, it also highlights the impossibility of
distinguishing between temperature and strain with a single measurement. As
already reported in [15], we have observed a slight decrease of the width of
the Brillouin gain curve with the temperature and no significant change with
strain, but the variations were too small and this measurement not precise
enough to allow an estimation of temperature and strain independently.
Furthermore, any strain or temperature inhomogeneity over a sensing length
shorter than the spatial resolution would artificially enlarge the gain curve
width, which does not therefore constitute a reliable parameter. In a previous
paper [18], we proposed an arrangement where the sensing fibre follows a
double path in the structure to be monitored, attached to the structure one
way, and thus subjected to both temperature and strain, and loose on the
return path, measuring the temperature only. Although the method is highly
reliable, the requirement of a double path may reduce the applicability of the
system. We will describe in the next section a completely different way to
separate the temperature and strain contributions, which has been developed
very recently.
2.5 Spontaneously Generated Brillouin Scattering: A means to Separate the
Temperature and Strain Contributions
Recent work has shown that another aspect of Brillouin scattering could be
advantageously used in a distributed sensor. When propagating in a medium,
light undergoes spontaneous scattering processes associated with thermally
excited phonons. The number of phonons and therefore the amount of
backscattered light, depend on the temperature. Launching a light pulse, and
monitoring the intensity of the backscattered Stokes or anti-Stokes waves,
allows the measurement of the temperature independently of the strain
condition. However this process is much less efficient than Rayleigh scattering,
and the Brillouin signal must somehow be extracted through a frequency
filtering method. A ''wide-band" (compared to the Brillouin gain width) filtering
provided by a Mach-Zehnder interferometer was used in [19,20], the method

permitting the measurement of the temperature parameter uniquely. On the

other hand, analyzing the backscattered light with a tunable narrow-band filter
provides more information, giving both the Brillouin shift (dependent on the
temperature and strain) and the level of backscattered light (a function of just
the temperature). The first demonstration of the possibility of measuring
independently the temperature and strain with this method is reported in
reference [21], and uses a high finesse Fabry-Perot filter. The authors obtained
a 4C and 100 e accuracy and 10 meter spatial resolution over 500 meters.
Another way to selectively filter the radiation is to use a coherent detection
technique [16], which gives better signal to noise ratio than can be obtained
with optical filtering. However it requires two light source with a tunable
frequency shift, as is the case for the sensors described above based on
Stimulated Brillouin Scattering; a Brillouin laser source might then be useful in
this configuration as well. In the spontaneous Brillouin scattering sensor, a
large number of signal

Page 71

averages is required to obtain adequate signal to noise ratio, ranging from 216
to 222 for the various detection techniques described above. We are currently
developing a "wide-band" heterodyne detection scheme incorporating a
Brillouin fibre ring laser source. It should combine the advantage of a relatively
large signal reflecting the total backscattered Brillouin light (as in the set-up
described in ref. [17]), with the high-sensitivity of a coherent detection
technique; less signal averaging should be required and we expect more
precise measurements. A hybrid system could then be implemented, using the
pump/probe method described in the first paragraphs to measure the Brillouin
shift, and the other to measure the level of spontaneously backscattered light.
2.6 Brillouin Distributed Sensors - Concluding Remarks
We have described the various techniques which have been developed over
the past ten years to monitor large structures with a Brillouin based distributed
sensor. Measurement of temperature and/or strain with an accuracy of 14C
and 25100 e respectively, and a spatial resolution of 1100 meters over a
sensing length of 150 km are the typical state-of-the-art capacities of this type
of sensor. We have shown that a Brillouin laser source could be implemented in
any of the possible configurations. It would provide a cheap alternative to the
previously proposed arrangements and increase the range of applicability of
the system.
3. Point Acoustic Sensors, for Non-Destructive Sensing and Possible Impact
Detection
Over the last decade, a considerable amount of interest has been shown in the
concept of the "smart structure" in which the body of the structure is
embedded within an array of sensors, actuators and microprocessors. One of
the main benefits of such technology is the ability of these sensors to measure
in situ the effects of temperature, pressure and strain. This allows an improved
ability to monitor the structural health of, for example, bridges and satisfies the
need for more accurate strain measurements within laminated carbon
composite structures such as those used in the aerospace industry.
The standard method to measure these strains still involves the use of electrical
resistance gauges. In the case of composite materials, these are attached to
the surface of the structure and then the internal strains determined using plyby-ply laminate theory [22]. However, focus in research is now shifting to the

use of embedded optic fibre sensors to determine internal strains directly. The
techniques are varied and include optical time domain reflectometry,
interferometric techniques and polarimetric sensing [23].
It is well known that in-fibre Bragg gratings may also be used to determine
internal strains. Their low cost, the ease with which they may be multiplexed
and their potential for simultaneous measurement of temperature, make them
an attractive alternative to other optical schemes. However, rather than
addressing low frequency detection using gratings (which has been well
demonstrated), here we discuss the use of in-fibre Bragg gratings to detect
and measure up to MHz

Page 72

(ultrasound) frequencies. In materials, such high frequencies may for instance

be due to impact damage. They can also be due to acoustic emission and it is
the monitoring of this which may be of particular importance in evaluating the
health of structures: many materials under stress release energy in the form of
(surface) elastic waves. The occurrences of these transient frequency bursts
gives an indication of potential catastrophic failure of the material or structure
and the waveform of the bursts and their frequency spectra (0.1- to 1.0-MHz
range) may be characteristic of the mechanism causing them e.g. twinning,
dislocation motion or cracking [24]. Conventional detection using piezoelectric
transducers has severe limitations. Primarily because of their size (12.5mm
25mm) they can disturb the transients being measured and compromise the
integrity of the material in which they are embedded. In addition, they are
susceptible to electromagnetic interference, signal distortion and a reduced
sensitivity that is due to the electrical-loading effects of the transducer leads.
Hence, optical techniques for monitoring these transients offer obvious
advantages. Such sensors could also be used to determine the location of
cracks etc. if the structure was excited by a transducer driven in either CW or
pulse mode.
For the following preliminary experiments, we simply wish to demonstrate the
feasibility of using gratings to detect and measure these high frequency
vibrations under controlled conditions. To this end, the arrangement we chose
was to position the grating at the focal spot of a focused continuous-wave
ultrasound transducer (normal to the acoustic propagation direction), in water.
The source of the ultrasound has a well-studied and predictable acoustic field
distribution with which we may analyse the grating's response.

Figure 3.
Experimental arrangement. PM = phase modulator. PC = polarisation
controller. DC = directional coupler.

Page 73

3.1 Experiment
The arrangement used to interrogate the grating is shown in figure 3. This
utilised a ramped lithium niobate phase modulator (accurately set to produce a
2p peak-to-peak phase excursion) to frequency shift the light in one arm of an
unbalanced Mach-Zehnder interferometer (MZ), thus permitting the use of
heterodyne signal processing [25]. Light from a pigtailed superluminescent
diode (Superlum, Moscow) giving an output power of 1mW centred at 824nm
with a bandwidth of 42nm was launched into the unbalanced MZ; hence a
channelled spectrum was created at the interferometer's outputs which was
incident on the grating. Incorporated in one arm of the MZ was the phase
modulator. The other arm contained a variable air gap which allowed the
optical path difference (OPD) between the two arms to be adjusted. Provided
that the OPD between the MZ's arms is longer than the source coherence
length and shorter than the effective coherence length of the back-reflected
light from the grating, interference signals are observed at the detector which
can be expressed as

Here, lB is the wavelength of the reflected light from the modulated grating,
w is the angular frequency of the ramp modulation, A is proportional to the
grating reflectivity, V is the visibility of the signals (dependent on the grating
bandwidth and the polarisation properties of the system), F = 2p.OPD/lB and
f(t) is a random phase drift term. A sinusoidal strain-induced change in lB
from the grating, dlB, induces a change in phase shift in equation 1, given by

where w is the angular frequency of the ultrasound incident on the grating.

Hence, from equation 1 (which can be expressed in terms of Bessel functions),
strain induced changes in lB induce a corresponding phase modulation of the
electrical carrier produced by the phase modulator, which we measured by
determining the amplitudes of the upper and lower side-band frequency
components observed on a radio frequency spectrum analyser.
The phase modulator was ramped at (and hence generated a carrier signal at)
10MHz. The grating we used had a nominal Bragg wavelength of 820nm, a
bandwidth of 0.2nm, a reflectivity of 80% and a length of 5mm. The

transducer was driven at its resonant frequency of 1.911MHz and generated a

maximum acoustic power of 5W in a focal spot of radius 1 2mm. Hence, we
calculate the pressure fields in the neighbourhood of the grating used in these
experiments to be of the order 1020 Atmospheres (Atm).

Page 74

3.2 Results and Discussion

Figure 4.
Spectrum analyser trace.

3.2.1 Grating with length 5mm

In our preliminary experiments we observed two striking anomalies in the
response of the system to the ultrasound: consider figure 4, which shows a
spectrum analyser trace recorded in a typical experiment. Firstly, note that the
upper and lower sideband frequency components (the first order Bessel
functions) are asymmetric. And secondly, note the existence of a large
homodyne signal at 1.911MHz which was up to several dB's greater than the
side-band magnitudes.

Figure 5.
Side-band power (normalised by carrier power) as a
function of longitudinal position for 5mm grating.

Now consider figure 5 in which we scanned the focal spot of the acoustical field

along the fibre/grating and recorded one of the side-band powers (normalised
by its corresponding carrier signal power) with displacement. Note the multiple
"peaks" and "troughs" in the system response which are observed over a
distance that is much greater than the grating length. This too is another
unexpected result since the radius of the focal spot is only about 1mm.
However, consider the average distance between the peak responses which is
1.475mm. If we hypothesise that the acoustic coupling from the ultrasonic
field to the optical fibre leads to the formation of stationary waves in the fibre,
this value leads to an experimental acoustic wavelength of 2.95mm which is
close to the predicted value of 3.087mm for compressional waves at 1.911MHz
in fused quartz. We repeated

Page 75

these experiments, but now driving the transducer at 1.6MHz and found an
experimental value of 3.76mm which is again close to the predicted value of
3.68mm for compressional waves. Using the same system, we then modulated
the grating with low frequency (100 Hz) sound waves in air and high
frequency (76kHz) sound waves in water. In both cases, the system response
was as originally expected, with symmetric side-bands and no homodyne signal
observed.
From these experiments we conclude the following. Compressional standing
waves are set up by the ultrasound in the fibre (although the acrylic jackets
which are on either side of the grating and which are, in the case of figure 5,
spaced about 1cm apart, tend to attenuate the acoustic modes). Since these
waves must only partially modulate the grating (as their wavelength is less
than the length of the grating), this means that the grating is subject to a nonuniform strain and so leads to regions of the grating acting as spectral filters
for the back-reflected light from other regions of the grating. This gives rise to
an amplitude modulation (the homodyne signal) and, as we show in [26], the
amplitude modulation in turn gives rise to the asymmetric side-bands. We
finally note that the wavelength of the lower frequency sound waves is greater
than the length of the grating. Hence in this case, the grating is now subject
to a more uniform strain and so none of the anomalies in the system response
were observed.
3.2.2 Grating with length 1mm

Figure 6.
Side-band power (normalised by carrier power) as a
function of longitudinal position for jacketed 1mm grating.

Based on our hypothesis, it is apparent that for the grating to operate correctly

in response to the MHz acoustic field, the grating length should be made
smaller than the acoustic wavelength in fused quartz. In order to demonstrate
this, we took a standard 5mm grating and gradually removed small pieces of it
from one end until approximately only 1mm of grating was remaining. As each
piece was removed, we recorded the system response to the ultrasonic field
using the shortened grating, and noted a dramatic decrease in the homodyne
signal along with more symmetric side-band magnitudes.
However because of the results of figure 5, it is apparent that this shortened
grating on its own cannot be used as a high frequency point probe since it will
exhibit insufficient longitudinal resolution. In order to obtain an improved
performance we must first desensitise nearly all the fibre to the acoustical field.
This we did by jacketing the fibre with PVC sleeving

Page 76

(diameter < 1mm) such that only the 1mm grating at the end of the fibre was
exposed to the field. The results of a scan of the acoustic focal spot along the
fibre is shown in figure 6 and, as may be seen, this data compares favourably
with the diameter of the main diffraction maximum of the transducer.

Figure 7.
Side-band power (normalised by carrier power) as a
function of acoustical power incident on the jacketed 1mm grating.

We finally show in figure 7 the detected power of one of the side-bands

(normalised by its corresponding carrier signal) as a function of acoustical
power incident on the grating. It is clear that the system response is linear and
(for this probe) we determined a noise limited pressure resolution of 1.8
10-2 Atm / Hz.
3.3 Point Acoustic Sensors - Concluding Remarks
Thus far we have shown that a Bragg grating may function effectively as an
ultrasonic probe with mm resolution if: (a) the grating length is less than the
acoustic wavelength in fused quartz and (b) the grating/fibre is suitably
desensitised.
Three final points should however be noted. Firstly, by shortening the grating
we have dramatically reduced its reflectivity (as well as increased its
bandwidth). In our case we found a reduction in the back-reflected light
intensity of well over 100. Gratings of lengths less than 1mm but with 90%
reflectivity can be manufactured and so we anticipate a greatly improved
pressure resolution using such gratings. Secondly, more sophisticated
techniques (entailing multiple coatings) for desensitising optical fibres have
been reported [27]. Such techniques may well have importance when

embedding these gratings in structures. Thirdly, high frequency impact

excitations and acoustic emission are generate transient frequency bursts.
Hence, although the probe we describe responds well to continuous-wave
single frequency fields, it is an obvious important next step to test its response
to these transient bursts.

Page 77

4. Summary
The modes of operation and performances of two different optical fibre sensors
currently being investigated for applications such as structural monitoring have
been discussed, their performances are summarized below;
1) A fully distributed sensor based on Brillouin scattering, this sensor offers
extreme sensing range (in excess of 50 km with resolutions of 1C or 25e).
2) An acoustic point sensor based on a very short FBG with a linear frequency
and amplitude response up to 2MHz.
The authors gratefully acknowledge the support of the UK Engineering and
Physical Sciences Research Council.
4. References
1 D. Culverhouse, F. Farahi, C.N. Pannell, D.A. Jackson, "Potential of

stimulated Brillouin scattering as sensing mechanism of distributed

temperature sensor", Electron. Lett. 25, 913 (1989).
2 T. Horriguchi, T. Kurashima, M. Tateda, "Tensile strain dependence of

Brillouin frequency shift in silica optical fibers", IEEE Photon. Technol. Lett. 1,
107 (1989).
3 X. Bao, D.J. Webb, D.A. Jackson, "22-km distributed temperature sensor

using Brillouin gain in an optical fiber", Opt. Lett. 18, 552 (1993).
4 X. Bao, J. Dhliwayo, N. Heron, D.J. Webb, D.A. Jackson, "Experimental and

theoretical studies on a distributed temperature sensor based on Brillouin

scattering", J. Lightwave Technol. 13, 1340 (1995).
5 A. Fellay, L. Thevenaz, M. Facchini, M. Nikles, P. Robert, "Distributed sensing

using stimulated Brillouin scattering: towards ultimate resolution", Proceedings

of OFS 1997, p. 324.
6 M. Nikles, L. Thevenaz, P.A. Robert, "Simple distributed fiber sensor based on

Brillouin gain spectrum analysis", Opt. Lett. 21, 758 (1996).

7 M.O. van Deventer and A.J. Boot, "Polarization properties of stimulated

Brilouin scattering in single mode fibers", J. Lightwave Technol. 12, 585

(1994).
8 S. Jetschke, U. Ropke, E. Geinitz, "Averaging of polarization modulations in a

distributed Brillouin fiber sensor system", Proceedings of OFS 1997, p. 528.

9 X. Bao, D.J. Webb, D.A. Jackson, "Temperature non-uniformity in distributed

temperature sensors", El. Lett. 29, 976 (1993).

10 E. Geinitz, S. Jetschke, U. Ropke, S. Schroter, R. Willsch, "Improvement of

distributed Brillouin sensing by compensation for systematic errors",

Proceedings of OFS 1997, p. 328.
11 T. Horiguchi, T. Kurashima and Y. Koyamada, "A new technique to shift

lightwave frequency for distributed fiber-optic sensing", in Distribute and

Multiplexed Fiber Optic Sensors II, John P. Dakin, Alan D. Kersey, Editors,
Proc. SPIE 1797, 1830 (1992).
12 N.A. Heron, D.J. Webb and D.A. Jackson, "Single ended Brillouin interaction

based distributed temperature sensor using a Faraday rotating mirror," in

Distribute and Multiplexed Fiber Optic Sensors VI, Alan D. Kersey, John P.
Dakin, Editors, Proc. SPIE 2838, 124128 (1996).

Page 78
13 M. Martinelli, "A universal compensator for polarization changes induced by

birefringence on a retracing beam", Opt. Comm. 72, 341 (1989).

14 K. Shimizu, T. Horiguchi, Y. Koyamada, T. Kurashima, "Coherent

selfheterodyne Brillouin OTDR for measurement of Brillouin frequency shift

distribution in optical fibers", J. Lightwave Technol. 12, 730 (1994).
15 M. Nikles, L. Thevenaz, P.A. Robert, "Brillouin gain spectrum

characterization in single-mode optical fibers", J. Lightwave Technol. 15, 1842

(1997).
16 H. Izumita, T. Sato, M. Tateda, Y. Koyamada, "Brillouin OTDR employing

optical frequency shifter using side-band generation technique with high-speed

LN Phase Modulator", IEEE Photon. Technol. Lett 8, 1674 (1996).
17 V. Lecoeuche, D.J. Webb, C.N. Pannell, D.A. Jackson, "A simple and

efficient technique for an offset frequency shifter for Brillouin based distributed
fiber sensing", Proceedings of OFS 1997, p. 332.
18 X. Bao, D.J. Webb, D.A. Jackson, "Combined distributed temperature and

strain sensor based on Brillouin loss in an optical fiber", Opt. Lett. 16, 141
(1994).
19 K. De Souza, G.P. Lees, P.C. Wait, T.P Newson, "Diode-pumped Landau-

Placzek based distributed temperature sensor utilizing an all-fiber MachZehnder interferometer", El. Lett. 32, 2174 (1996).
20 G.P. Lees, P. Wait, T.P. Newson, "Novel optical fiber distributed temperature

sensor based on the Landau-Placzek ratio", Proceedings of OFS 1997, p. 340.

21 T.R. Parker, M. Farhadiroushan, V.A. Handerek, A.J. Rogers, "A fully

distributed simultaneous strain and temperature sensor using spontaneous

Brillouin backscatter", IEEE Photon. Tech. Lett. 9, 979 (1997).
22 Tsai S W and Hahn H T 1980 Introduction to Composite Materials

(Lancaster, PA: Technomic)

23 Rao Y-J and Jackson D A 1996 Recent progress in fibre optic low-coherence

interferometry Meas. Sci. and Techn. 7 981999

24 Kaiser J 1950 Investigation of Acoustic Emission in Tensile Testing Ph.D

thesis Technische Hochscule Munich Germany

25 Jackson D A, Kersey A D, Corke M and Jones J D C 1982 Pseudo-

heterodyne detection scheme for optical interferometer Electron. Lett. 18 1081

26 Fisher N E et al 1997 Response of in-fibre Bragg gratings to focused

ultrasonic fields Proceedings of Optical Fiber Sensors Conf. Williamsburg

Virginia USA.
27 Lagakos N et al 1985 Desensitization of the ultrasonic response of single-

mode fibers J. Lightwave Techn. 5 10361039

Page 79

Distributed and Chemical Fiber Optic Sensing and Installation in Bridges

D. R. HustonN
and P. L. Fuhr
Abstract
Fiber optic sensors may prove to be an effective technology for the monitoring
of bridges. The application of distributed and chemical chloride fiber optic
sensors to bridges is discussed. The underlying theory and the results from
some laboratory and field tests, are presented. Certain issues that arise in the
field installation of fiber optic sensors are also presented.
Introduction
Bridges can be very large and complicated structures. They are often placed in
harsh environments and are subject to a wide variety of external loads
including traffic, wind, earthquake, temperature gradients, impacts, scour and
chemical attack. Monitoring the response and state of health of a bridge as it is
subject to this loading is becoming a practical reality, primarily due to the
recent advances in sensing and data processing techniques. Fiber optic sensors
are a promising technology that is competing for use in the bridge-monitoring
arena. One of the primary advantages of fiber optic sensors over that of other
technologies is the versatility of the methods that enable the measurement of
quantities that are difficult to measure with other sensor types. Two fiber optic
sensing techniques that may fall into this category are distributed sensing and
chemical sensing.
Distributed Sensing
The load and state parameters of a bridge that are of interest are often
distributed over the span of the bridge. Developing distributed sensing systems
that are capable of measuring and transducing these physical parameters over
an extended area, such as a bridge deck, would be desirable. A description of
the theory of operation and the results of some simple demonstrations of
distributed fiber optic sensors follows.
Dryver R. Huston, Department of Mechanical Engineering
Peter L. Fuhr, Department of Electrical Engineering
University of Vermont, Burlington, VT 05405-0156

Page 80

The analysis is concerned with long lineal sensors. In principle these sensors
can be based on virtually any physical mechanism that acts to transduce a
signal that is propagating through a sensor with a lineal geometry, e.g.
electrical, optical or hydraulic distributed sensors. However, in this context the
focus will be on distributed fiber optic sensors. The sensors are capable of
modifying the signal as a function of position, Figure 1. The modulation of
signals as they pass through this sensor can described by the following
approach. Let x(i) be the position of sensor segment i, Dti the time at which
the signal leaves segment i, Dti the time that it takes for segment i to operate
on the signal, {Pj(i)} the parameter field at segment i at time ti-Dti/2 that
causes a sensor reading, {yk(i)} the characteristics of the signal leaving
segment i at time ti, Si the signal leaving segment i, and Gi the operation of
the sensing action of segment i on the signal. at time ti. Since each sensing
segment of the sensor acts on the signal sequentially as it passes through the
sensor the following operational representation of the signal modulation can be
formed:
and

or

As a first example this scheme can be applied to an intensity modulating

sensor, such as a multimode optical fiber subjected to distributed microbending
[1]. Assume that the fractional transmission per unit length of segment i
depends only on the parameter P, e.g. the local radius of curvature, at that
location so that Ti = Ti(P). The operation of segment i would be

where D1i is the length of segment i. In terms of the fractional loss fi(P) we
have The intensity, I, leaving the sensor is then related to the input according
to

letting D1i d1i we obtain

and

Page 81

If f(1) is sufficiently small, then I simplifies to the linear expression

As a second example consider a phase modulating sensor where the phase

angle of the input signal is modulated. Such a situation occurs in distributed
interferometric fiber optic sensors where

and

where
so that

Since f is the parameter of interest we have

If we want Dfi in terms of a phase change density then

or in integral form

Page 82

Since a sensor can be distributed, this opens up the possibility of distributing

the sensor in such a manner that it is sensitive to certain spatial distributions of
loads, and insensitive to others. This technique has been shown to be useful in
laboratory tests in which the magnitude and location of a load was identified in
a linear 1-dimensional structure [2] and a planar 2-dimensional structure [3].
The technique has also been applied to vehicle classification [4].
Chemical and Chloride Sensing
The corrosion of concrete reinforcing bars has been estimated to cause $150
billion in damage per year in the United States [5]. Of this figure, a significant
amount is due to the corrosion of bridge decks. Chlorides were indicated as a
possible source of rebar corrosion as early as 1920 [6]. However, it was not
until the 1950's with the common use of roadway deicing salts that the effects
of chloride-induced corrosion damage became an issue for bridge engineers. It
is believed that chloride ions cause a differential electrical potential around the
rebars, which promotes corrosion. Chlorides can come into contact with the
rebars by being present in the original concrete mix [7], by diffusing through
pores in the concrete cover [8], and through cracks. Typical levels of chloride
concentration range from about 0.1% up to 3% by weight.
One method of preventing chloride ion attack on bridge decks is to place an
impermeable membrane between the concrete deck and an asphalt overlay. It
is desirable to have a method of detecting chloride penetration in order to
detect a breach of the membrane. One potentially realizable detection scheme
is to use a sensor that is embedded in the concrete [9].
There are many methods of determining the concentration level of a specific
chemical in a material. A whole class of these methods is based on measuring
the change in the optical properties that occur in reagents in the presence of
certain chemical species. Optical changes due to chemical changes can be
manifested in a variety of manners including: 1. Color change, such as that
which occurs in titration; 2. Activation or bonding of a fluorescent chemical
agent; and 3. The precipitation of insoluble solids. All of these methods have
been examined in this study as possible chloride detection mechanisms. The
advantages of optical chemical detection are the potential for fabricating a
system that has a sensitivity that is highly-specific, and the ability of remote
sensing through a fiber-optic interconnection [10].

The requirements of an embedded chloride sensor are: 1. Sensitivity: The

chloride sensor must be able to detect weight concentrations that range from
0.1 to 3%. This sensor would preferably be able to transducer variable
concentration levels and be reversible. However, an ON/OFF-type reporting as
certain concentration thresholds are exceeded would be acceptable. 2.
Selectivity: The ideal chloride sensor would be one that is sensitive to only the
presence of chlorides and not to any other chemical species. 3. Embeddabilty:
The sensor must be able to be placed in the concrete as the deck is cast. This
requires that the sensor have a geometry that is

Page 83

suitable for placement in the deck along with the reinforcing bars. The sensor
must be rugged enough to withstand the casting and curing process. It is also
required that the connecting cables be dressed in such a way that they are not
damaged by the formwork and form removal. 4. Long-Term Reliability: The
sensor should last up to 20 or 30 years. False positive and/or negative readings
should be minimized. 5. Accessibility: Since it is unreasonable to expect all of
the sensors to survive for over 20 years, it would be desirable to have the
sensor geometry configured so that it can be extracted and repaired, or
replaced, if necessary.
Four different methods have been identified for use as possible chloride optical
sensors. The first technique was to use fluorescein in a titration process with
silver nitrate [11]. In this method, a solution containing and unknown amount
of sodium chloride and dichlorofluorescein is initially yellow-green. As silver
nitrate is added to the solution, it changes color to pink, and then to a milky
precipitate. This color change can be detected by a spectrometer.
Unfortunately, the color change is not long lasting because the silver chloride
precipitates and the color change fades. The second method is to exploit the
precipitation reaction of silver chloride and to measure the presence of
precipitates by amount of blocking or occlusion of an optical gap. Broadband
light is shown through the gap and the intensity loss is a function of the
amount of precipitate, and hence, the amount of chloride. The third method is
a technique by which silver chromate is used instead of silver nitrate. [12] The
silver chromate turns from red to white silver chloride. The fourth method uses
a commercially-available proprietary chemical indicator paper. This paper
changes color from yellow to brown in the presence of chlorides. A simple fiber
optic sensor can be constructed by which white light is transmitted through
the paper and is measured with an optical spectrometer.
Figure 2 shows a prototype design of an embeddable and accessible fiber optic
chloride sensor. The design is for the sensor to mount flush with the bottom of
the bridge deck. The current prototype design is somewhat large for a bridge
deck. Further miniaturization of the radial dimension is probably required for
practical usage.
Waterbury Bridge
As a testbed for the application of fiber optic sensor technology to bridges, a
67 m steel truss bridge spanning the Winooski River in Waterbury, Vermont

(USA) was made available during redecking by the Vermont Agency of

Transportation.
The bridge project was to rehabilitate the existing truss and place a new deck,
new substructure and new sidewalk with associated roadway and channel
work. The entire project length is 153 m. This bridge had been subjected to
the application of large amounts of chloride-based deicing agents that resulted
in a significant amount of corrosion-induced failure of the bridge's rebar deck.
While the overall amount of deicing agent used is to be significantly reduced
from prior years' levels, primarily due to pollution concerns, it was deemed
prudent by the State of Vermont Agency of Transportation officials to monitor
the chloride penetration into the deck. Additionally

Page 84

this truss structure has been anecdotally known to vibrate. It was decided to
monitor the bridge's vibrational performance as well as monitor strain in the
deck rebars.
Placement of the sensors, following the construction schedule, occurred in
October 1997. Following placement, the sensors were encased in concrete.
The concrete was vibrated and left to cure for 28 days. After this time the
access boxes to the sensors' leads were excavated. When appropriate or
necessary, pigtailed, connectorized leads were spliced onto the bridge sensors'
leads.
Thirty-six (36) fiber optic chloride sensors, of the type shown in Figure 2, were
embedded at various points along the bridge. Access to these sensors is from
the bottom of the bridge deck.
Eight (8) Bragg strain sensors, commercially available units provided by 3M
Specialty Optical Fibers, were placed at structural points of classically maximum
strain along the bridge. Access to these sensors is obtained from panels along
the outside of the bridge deck. Two (2) multimode fiber optic vibration sensors
stretch along the deck with access found at an outside bridge deck panel. The
multimode fibers can also serve as a crude continuity check (these are cabled
fibers with strength members - certainly not optimal for structural integrity
measurements). A single electrical strain gauge is position near one of the
Bragg strain sensors. The types and locations of the embedded fiber optic
sensors are shown in Figure 3.
As of December 1997, the efforts have been primarily on determining if the
sensors survived the installation, concrete burial (including being subjected to
the vibrator), curing cycle (pH> 12), and excavation of the leads. This has
proven to be a tedious process requiring multiple attempts at field splicing and
connectorization in a suboptimal environment.
A characteristic strain measurement obtained from an embedded Bragg grating
sensor is shown in Figure 4. In this figure measurements obtained with the
bridge basically unloaded are compared with initial baseline measurements
taken after attachment to the deck's rebar matrix but prior to the concrete
pour. Comparison of these measurements with controlled environment
laboratory based measurements, allows for the variation in ambient
temperature to be removed from the readings with an end result of the bridge

deck being subject to a strain of approximately 50 strain.

Chloride penetration measurements are to begin in early 1998. It is anticipated
that snow loading conditions and other conditional effects will be monitored
throughout this winter and on into the future. This suite of sensors should be
able to provide transportation officials and structural engineers with detailed
performance and health information not currently available through other
means.

Page 85

Practical Issues-Installation and Repair

Experience gained in placing sensors into the Waterbury bridge and other
structures has brought to the forefront some key issues that must be
addressed for the successful usage of fiber optic sensors on bridge monitoring
projects. These are:
1. Construction site environment - The typical construction site is a very harsh
environment with lots of vigorous activity and the use of heavy equipment. The
fiber optic sensors must be designed to be rugged enough so as to withstand
routine mishaps. It is also quite useful to have sensor personnel on site during
all phases of the operation in which the sensors are left exposed and
vulnerable.
2. Ingress and egress from concrete formwork - If the sensors are to be
embedded in concrete, then means must be provided for the cables that
connect to the embedded sensor be placed so that they are easily accessed
after the forms are removed. This requires careful coordination with the
engineers and the construction crews.
3. Construction site timing - The timing of events on construction sites can be
quite chaotic. The installation techniques for the fiber optic sensors should be
rapid and produce minimal disruption to the construction crew.
4. Avoidance of long cable runs - The difficulty of placing sensors into the
internal workings of a complicated bridge structure, particularly if it is a mesh
of reinforcing bars, seems to go up exponentially with the length of the cable
that is directly attached to the sensor. These difficulties can be avoided by
either creating multiple points for the attachment of optical instruments,
splicing, or, even better, a distributed spliced fiber optic network.
5. Sensor access - Bridges can last centuries. It may not be realistic to expect
that the original fiber optic sensors to last this long. Therefore, a very desirable
feature of embedded fiber optic sensors is that they be placed so that they can
be accessed and repaired. If this is not possible, then the sensors should be
either very robust or expendable.
Conclusions
Fiber optic sensors offer a considerable amount of versatility that may make
them competitive alternatives for bridge monitoring applications. The

successful application of fiber optic sensors will require the development of

additional technology, such as on site networking, and the exploitation of the
wide variety of possible sensing techniques.
Acknowledgements
The authors wish to thank the State of Vermont Agency of Transportation, the
U.S. National Science Foundation, the New England Transportation
Consortium, 3M Specialty Optical Fibers, Corning Glass, and Siecor for their
partial support of and generous component donations to this project.

Page 86

References
1. Udd, E., Fiber Optic Sensors, Wiley-Interscience, New York, 1991.
2. Spillman Jr. WB, and Huston DR. ''Scaling and Antenna Gain in Integrating
Fiber Optic Sensors," IEEE Jnl. of Lightwave Technology, Vol. 13, No. 7, pp.
12221230, July 1995.
3. Spillman Jr. WB, and Huston D. "Detection, location and characterization of
point perturbations over a two dimensional area using two spatially weighted
distributed fiber optic sensors" presented at the SPIE Smart Structures and
Materials Conference, 3042-14, March 1997, San Diego CA.
4. Huston DR, Spillman Jr. WB, Claus RO, and Ayra V. "Vehicle Classification by
Pattern Matching Gage Sensors," SPIE 2718A-27 Smart Structures and
Materials Conference, San Diego, Feb. 1996.
5. Fasullo, E.J. (1992). "Infrastructure: The Battlefield of Corrosion." Corrosion
Forms and Control for Infrastructure, ASTM STP 1137, V Chaker, Ed.,
Philadelphia, PA.
6. Hime, W.G. (1994). (May) "Chloride-Caused Corrosion of Steel in Concrete:
a New Historical Perspective." Concrete International, pp. 5462.
7. Ali, M.G., Dannish, S.A., and A1-Hussaini, A. (1996). (July) "Strength and
Durability of Concrete Structures in Bahrain." Concrete International, pp. 3946.
8. Johansen, V., Goltermann, P., and Thaulow, N. (1995). (July)"Chloride
Transport in Concrete." pp. 4344, Concrete International.
9. Fuhr, P.L., and Huston, D.R., "Embedded and Surface Attached Fiber Optic
Corrosion Sensors for Civil Structures." Proc. Engineering Solns. to Industrial
Corrosion Problems, Sandeflord, Norway, June 1993.
10. Boisde, G., Harmer, A. (1996). Chemical and Biochemical Sensing with
Optical Fibers and Waveguides, Artech House, Boston, MA.
11. McPadden, A. (1995). "Fiber Optic Corrosion and Chloride Sensors." MS
Thesis, Dept. of Electrical Engineering, University of Vermont, Burlington, VT.
12. Cosentino, P., Grossman, B., Shieh, C., Doi, S., Xi, H., and Erbland, P.
(1995). "Fiber Optic Chloride Sensor Development." ASCE Jnl. of Geotechnical
Eng., pp. 610617.

Page 87

Figure 1.
Geometry of a distributed fiber optic sensor.

Figure 2.
External and internal view of the sensors which
are ready to be embedded into a concrete bridge
deck.

Page 88

Figure 3.
A suite of fiber optic sensors have been embedded into the Waterbury Vermont TH2/TH4
Bridge #31.

Figure 4.
Preliminary measurements taken from a single embedded
strain sensor.

Page 89

Technology and Applications of Distributed Optical Fibre Sensors

A. H. Hartog
Abstract:
The paper reviews the principles of operation, design and application of
distributed fibre optic sensors. Examples are drawn primarily from the widelyused Raman-based distributed temperatures but other technologies and
applications in civil engineering are also considered.
Introduction
Research into the use of optical fibres as transducer elements can be traced
back at least 20 years. It has given rise to a plethora of devices in the
laboratory, but rather fewer have found practical commercial applications. The
primary reasons for this lack of acceptance are the existence of alternative
transducers which in many cases provide adequate solutions, together with the
usually high cost of optical fibre sensors.
However, certain fibre-optic sensors are uniquely placed to address previously
unsolved measurement problems. Examples include measurements in very high
electro-magnetic fields, such as in power transformers and energy cables,
where fibre-optic temperature and current sensors have found a niche owing
to their immunity to EMI. The fibre-optic gyroscope has also proved to be a
versatile device, capable of being designed either to meet stringent cost
targets with performance adequate for in-car inertial navigation, or
alternatively, to exceed the sensitivity, drift and scale factor stability of the best
mechanical or laser gyroscopes.
This paper addresses a class of sensors, namely, that of distributed sensors
[1], which although unique to optical fibres presents similarities to techniques
such as OTDR[2], Radar, Lidar or Ultrasonic reflectometry. Specifically, the
operating principles and the applications of distributed temperature sensors
(DTS) are discussed. Extension to the needs of civil engineering will also be
explored. Whereas similar techniques of time domain reflectometry have also
been proposed to multiplex arrays of point-sensors, only truly distributed
sensors are addressed here.
The principal characteristics of distributed sensors are that they provide a
complete measurand profile of the fibre which forms the sensor. Thus typically,

a
A.H. Hartog, York Sensors Ltd. York House,
Premier Way, Abbey Park, Romsey SO51 9AQ, Hants, UK

Page 90

single instrument connected to an optical fibre several km long will provide

thousands of temperature readings at points spaced as closely as 1m apart
along the entire length of the fibre. In addition, the instrumentation
interrogates fibres of the same designs as are used for telecommunications;
these fibres are thus inexpensive and manufactured in large volumes to very
high standards of optical transmission, dimensional accuracy and mechanical
strength. Furthermore, in comparison with connecting many single-point
sensors to acquisition equipment, the bulk and cost of the interconnections are
considerably reduced. As a result, the cost per sensing point of an installed
distributed sensor can be significantly lower than may be achieved with
conventional technology. This is particularly the case in applications such as
plant monitoring in hazardous areas where special precautions must be
adopted with electrical transducers to ensure intrinsic safety, whereas in
optical transducers, the risk of causing the ignition of a flammable or explosive
atmosphere may be addressed by ensuring that the power entering the fibre
does not exceed a predetermined limit.
A further benefit, especially applicable to civil engineering where embedded
fibres cannot be inspected, is that provided the transmission of the fibre
remains adequate, then the user may be confident that the sensor is operating
correctly, i.e. there is a measure of self-testing in these systems.
Principles of Operation
Most commercially-available distributed temperature sensors are based on the
principles of Raman Optical Time-Domain Reflectometry [3, 4] in which a short
pulse light is launched into the sensing fibre. From the backscattered light
spectrum, a very weak, but temperature-sensitive band (the so-called antiStokes Raman band) is selected. The time dependence of the latter's intensity
is a direct measure of the fibre temperature at a position determined by the
round-trip propagation time. The efficiency of the process used, spontaneous
Raman scattering, is independent of the probe pulse intensity which simplifies
considerably the signal processing required, compared with sensors using nonlinear effects, such as stimulated processes [5] or the Kerr-effect [6]. The basic
optical arrangement is illustrated in Fig. 1 below.

Figure 1:
Basic arrangement of a Raman Distributed Temperature Sensor

Page 91

In practice, a number of normalisation steps are required to eliminate the

effects on the measured value of losses in the fibre, and of variation in fibre
characteristics. Thus in practical systems a second wavelength (the Stokes
Raman band) is frequently used to provide a reference. In multimode systems,
a double-end measurement, where the data is collected successively from each
end of the sensing fibre, provides a supplementary referencing method and the
combination of the dual wavelength and double-ended measurements provides
a very robust determination of the temperature, which has proved to be
resilient to changes in fibre or connector losses.
The optical frequency domain method, in which a variable-frequency but
continuous-wave modulation (rather than a pulse modulation) has also been
proposed [7] and indeed commercialised. In principle, the two methods are
equivalent although they differ in the details of their implementation and in the
applications for which their performance is optimised. The frequency domain
approach suffers from the shot-noise induced by strong near-end signals
swamping the weaker far-end signals. Likewise, in the case of reflective breaks,
the strong reflection which can be gated out in time-domain reflectometry,
overlap in time with the backscatter signals from other locations in the fibre
and this places very stringent demands on the linearity of the entire electronics
chain up to and including the point at which the signals are digitised.
The criteria by which a distributed sensor is measured include its spatial
resolution, i.e. its ability to distinguish two closely-spaced events along the
fibre. Also of interest is the range, i.e. the length of fibre which can be
addressed by a single instrument; together, the spatial resolution and the
range determine the equivalent number of points which the sensor is capable
of addressing. Most commercially available systems are capable of addressing
at least 300 points and the highest performing systems, up to 10000 points.
Clearly, the sampling rate of the digitising electronics must be sufficient
faithfully to replicate, in real time or by, interleaving the backscatter signals
received. Further performance criteria include the resolution of the measurand,
which is usually limited by signal-to-noise ratio at the receiver although some
suppliers have chosen to define this performance parameter in relation to the
resolution of their display. Finally, in systems where intensive digital averaging
is normally relied on to improve the signal quality, the measurement time is an
important measure of performance, since it determines the response time to a
change in the measurand.

The spatial resolution depends on the duration of the probe pulse, the
bandwidth of the fibre (at all relevant wavelengths), the bandwidth of the
receiver and associated electronics as well as the sampling rate of the digitising
system. For constant probe power, as the spatial resolution is increased the
signal-to-noise ratio at the output of the receiver degrades. Thus, enhancing
the spatial resolution usually results in a degraded resolution of the measurand
(owing to a worse signal-to-noise ratio), a longer measurement time (owing to
longer measurement times) or a shorter range to reduce the cumulative losses
experienced by the signals.
In order to avoid these trade-offs, the designer will attempt to launch high
optical powers into the fibre, to improve the internal optical efficiency and the
noise performance of the receiver as well as providing efficient digital
averaging. In

Page 92

practice the power which can be launched into the fibre is limited by non-linear
effects, the first of which is usually stimulated Raman scattering, which limits
launch powers to a few tens of W peak power. In principle, the frequencydomain approach allows more average power to be launched, a feature which,
however, has not resulted in superior performance, for the reasons outlined
above.
To date, most of the systems available commercially operate on multimode
fibre, usually of the 50/125 graded-index type, although some of the
equipment has been designed to operate on the larger 200/280 fibre type. In
general, the preferred operating wavelength increases with the desired range.
Thus short distance systems operating at 850 or 900nm based on
semiconductor laser sources, are available with ranges of up to 13km; spatial
resolutions of 13m are quoted. York Sensors' most widely installed product is
based on a diode-pumped solid-state laser operating at 1064nm and achieves
a spatial resolution of around 1m over up to 10km of fibre. For fibres lengths
beyond 10km, the wavelength of the source is preferably increased to 1300nm
or 1550nm. York Sensors has supplied some 120 DTS systems to date, most of
these based on 1064nm sources.
As the operating wavelength is increased, a number of adverse factors affect
the near-end performance of the system. Thus, increasing wavelengths result
in reduced scattering factors which reduce the initial signal, but over long fibre
distances may be compensated by reduced transmission losses. Moreover, the
performance of the detectors degrades markedly beyond about 1000nm and
therefore long-wavelength systems are unattractive for distances below about
10km. A further factor to be considered in the design of Raman DTS systems is
that multimode fibres are seldom optimised, in their bandwidth for operation
beyond 1300nm. At 1550nm, bandwidth product is not usually specified by
fibre manufacturers and thus operation at this wavelength on multimode fibre
is something of a lottery.
Single-Mode Systems
In view of requirements for increasingly long range distributed sensors, e.g. for
sub-sea power cables or pipeline monitoring, interest has shifted towards
Raman DTS systems operating on single mode fibres (SMF). The ability to
measure temperature on SMF is an important step in gaining acceptance of the
technology, since extensive networks of these fibres are already in place for

communication purposes, especially alongside, or within, energy cables.

The measurement of temperature distributions in SMF is especially difficult
because of the low levels of power which can be launched before the onset of
non-linear effects; the permissible launch power is typically one tenth of that
which may be used in multimode fibres. In addition the design of this type of
fibre, and the fact that they normally operate at much longer wavelengths than
multimode DTS systems, results in a further decrease in the signal level
available. Thus careful design of the optics for minimal transmission loss,
optimisation of the receiver noise performance, and a highly efficient digital
averaging are required. In compensation, however, the losses in SMF at the
preferred telecommunications wavelength of 1550nm are typically 2025% of
those found in multimode, graded index fibres operated at 1064nm and, at a
distance of around 30km, the theoretical performance of single mode systems
is similar to that of those based on multimode fibre.

Page 93

Figure 2:
Optical arrangement for a single mode DTS system

Owing to the very accurate referencing of all propagation losses which is

possible in this configuration, the need for double-ended measurements has
been eliminated. In particular, losses arising from non-linear effects in the fibre
can be fully corrected. The power level in the fibre can thus be optimised for
maximum backscatter signal at the remote end.
The optical arrangement used in single-mode DTS systems is shown in Fig. 2.
Without such signal processing schemes, the maximum power is dictated by
the errors caused, e.g. in the anti-Stokes/Stokes ratiometric approach, by
transfer of power to the Stokes wavelength. The resulting increase in allowable
launch power is equivalent to at least 5 km of measured cable.
The ability confidently to measure from one end only of the fibre, together with
the achievement of 30km range (with 10m resolution for an averaging time of
600s), has allowed very long cables (up to 60km), especially sub-sea cables, to
be monitored over their entire length. This total range is achieved by
interrogating the fibre from separate instruments sited at each of the shore
stations. The first such sub-sea single-mode system to be operated on an
energy cable in commercial operation was brought into service in 1996
between the Island of Penang and the Malaysian mainland. Results on a single
mode land-based cable of similar length in the Netherlands are shown in Fig. 3.
In order to illustrate the performance of the single-mode systems, a detail of
the remote end of a 30km length of fibre is shown in Fig. 4. In this case, it was
arranged for the temperature of successive 100m sections of fibre, separated
by further 100m sections at room temperature to vary in sections of 100m from
0C to 60C. In this case, the measurement time was 600s and the spatial
resolution was set to 10m. The spatial resolution is apparent in the sharp

edges separating the sections and the temperature resolution is well below 1C
rms.

Page 94

Figure 3:
Distributed temperature measurement on a live energy cable using single mode fibre
integrated in the cable.

Figure 4:
Thermal profile induced on a 30km single mode fibre in the laboratory and recorded
using a single mode distributed temperature sensor.

Applications of Distributed Temperature Sensing.

Distributed temperature sensing has found applications in a large number of
fields, including electrical energy distribution [8], plant monitoring [9],
geological measurements, fire detection, environmental monitoring and the
construction industry. Although York Sensors has established applications in
each of these areas, only the first three are addressed here.
Energy Distribution
The dielectric nature of the optical fibre sensor, together with its elongated
shape naturally lends itself to the monitoring of energy cables and
transformers, where high electro-magnetic fields preclude the use of electrical
sensors. In addition, both cables and transformers require the measurement of
many points which again has encouraged the development of these
applications for distributed temperature sensors. Turning specifically towards
the energy cable application, which has

Page 95

Figure 5:
Temperature distribution on an urban energy prior to operation.

attracted some 40% of York Sensors installations to date, the main problem to
be solved is to detect, locate and quantify localised hot-spots. Such surveys
allow the thermal environment of the cable to be determined before the latter
is energised; the maximum capacity of the cable can then be predicted more
accurately than using the existing methods based on cable rating models,
which use parameters such as soil conditions.
One example of the thermal distribution of an energy cable is shown in Fig. 5,
prior to application of the electrical load to the cable; in this case, the fibre was
looped back at the far end of the cable, and the resulting symmetry in the
thermal distribution is evident. Despite the cable not being under load, a
number of temperature features are readily apparent and have been traced
back to the conditions of the cable; in particular, the hot spot at some 600m
from either end has been attributed to the crossing of the route of the cable by
buried steam conduits. The soil conditions play an important part in
determining the thermal profile along the cable. The profile in Fig. 4 was
recorded in the winter under low-load conditions and it shows a number of
features which were found to be related to the cable routing, including
crossing of a waterway, deep passage under a major road junction and other
landmarks readily identified from the known cable routing.
The industry is moving now towards dynamic cable rating systems where the
capacity of each circuit is calculated from recent thermal history together with
the known response of the cable to a temporary overload. In this way, the
effect of overloading the cable for a short time (e.g. 1/2 hour) can be
predicted. If these peak capacities within the network are sufficient to handle
extreme peak loads, then the utility can drive their existing cables longer and
thus delay the purchase of expensive new transmission capacity. The savings
due to delayed investment (even for a single year) are usually far higher than

the cost of the monitoring equipment. The knowledge of peak capacity is also
used for contingency planning and network controllers can deal with the
effects of damage to the network by temporarily overloading their cables.
Since the lifetime of the cable is dependent on the overloads it has experienced
a continuous log of the temperature of the cable may also be used to predict
the life of the cable and thus allow maintenance and replacement to be
planned.

Page 96

As a result of these benefits, many utilities are now specifying that their cables,
especially those at or above 130kV should contain fibres for thermal profile
monitoring. Of course, in general, a newly installed cable is operated well
below its rating; the benefit to the utility of being able to operate the cable
close to its rating may therefore not accrue immediately. However, it is almost
impossible to retrofit fibres once the cable is laid and this explains the
increasing proportion of cables within integrated optical fibre. Thermal
monitoring on new cables has however frequently allowed their owners to gain
a thermal profile prior to, and immediately after, energising. These cable
surveys frequently reveal unanticipated hot-spots which can be used to refine
the rating model for the cable or, in some cases, can be rectified prior to full
operation of the cable.
Plant Monitoring
DTS systems have been used in a number of applications where they are
required to detect the formation of hot-spots on the surface of process vessels,
such as reformers in petrochemical plants. In this case the operators are
concerned with the failure of the ceramic lining or, in some designs, of the
burner driving the process. Vessels for hydrocarbon reforming are a typical
example of such an application where a 100mm steel wall, typically lined with
200300mm of refractory bricks contains an oxygen-rich, high pressure process
running with burner temperatures in excess of 2000C. In order to monitor the
surface of the vessel, a fibre is wrapped in a helical pattern around its
circumference, with a typical pitch of 100mm. A typical resulting thermal
profile is shown in Fig. 6, where very high spatial frequency temperature
variations, with peak excursions approaching 100C pk-pk, may readily be seen
and are thought to be caused by imperfections, including cracks in the
refractory lining.
Plant monitoring is a demanding application owing to the need to detect small,
but fast developing hot spots against a thermal background which is usually
very non-uniform. The monitoring system is required to deal with high data
rates. It must also be able to make decisions based on noisy data and
communicate directly with the process control systems. This application has
thus resulted in the design of very high performance systems incorporating
advanced signal processing and able to reduce the temperature information to
the essential elements, e.g. a single alarm signal or a few relay outputs.

The high temperatures at which these processes operate has required the
development of high temperature fibre cables. The most commonly used
design consists of a polyimide-coated fibre built into a small stainless steel
tube. This arrangement can be designed to operate above 300C for extended
periods.
Further engineering work has been required on the installation and fixing
methods for the fibre to ensure good thermal contact, robust adhesion and
mechanical protection from impact damage whilst minimising the fibre loss and
the thermal response time.

Page 97

Figure 6:
Measured surface temperature distribution of a secondary reformer.

Geological Applications.
DTS systems have been used in a variety of geological (usually down-hole)
applications, where thermal profiles provide valuable information. The
applications explored to date include geophysical measurements aimed at
improving predictions of volcanic and seismic activity.
Underground surveys in preparation for waste disposal have also used thermal
profiles generated by the DTS to determine the local rock structure. Fig. 7
illustrates one such survey, where the central, dome-like feature is the
underground section.
The most actively developed application is in the field of oil and gas
exploration, especially for the monitoring of the effect of steam injection in
enhanced oil recovery. Many of the wells where steam injection is used are only
marginally cost sensitive and a better management of the energy use in this
type of oil field can substantially improve the profitability, allow a greater
proportion of the reserves to be extracted and reduce the CO2 emissions
resulting from the extraction activity. One example of a data series obtained on
a horizontal well during steam injection is shown in Fig. 8. The lowest curve
shown was obtained prior to steam injection. With increasing time from the
start of the injection process, the temperature profiles

Figure 7:
Thermal profile in a salt-rock formation

Page 98

Figure 8:
Family of thermal profiles recorded during steam injection in an horizontal well

are seen to rise, which provides the field operator with a wealth of information,
including rate of penetration of the steam, the position of the heat front, the
temperature at the point where the oil is extracted. Using this type of
information, especially if collected over an entire field, the recovery efficiency
can be improved dramatically, which will contribute to increasing the
economically recoverable reserves in heavy oil or tar deposits.
Civil Engineering
Possibly the earliest application of distributed sensing technology in civil
engineering was the monitoring of concrete curing in major projects, where the
exothermic reaction can to accelerating curing rates and insufficient strength
of the finished product. Trials were conducted in Japan, e.g. at the Miyagase
Dam in 1990 in which fibres were embedded in the surface prior to further
layers of concrete being poured. The objective here is to monitor the cure rate
(in practice, to detect that the peak temperature has passed) before the next
load of concrete is allowed to be applied.
The more substantial applications in civil engineering must involve the sensing
of strain as well as temperature. As related technique to the Raman OTDR
method, involving Brillouin backscatter has been developed in recent years, in
particular by researchers at NTT, but since then taken up in other laboratories.
In essence the Brillouin method involves the measurement of the frequency
difference between the probe pulse and that of Brillouin signal. The frequency
measurement is achieved by determining the Brillouin at each point along the
fibre for a range of settings of the frequency difference between two counterpropagating optical signals [10]. The Brillouin shift is sensitive to strain and
also to temperature; thus if the temperature can be determined independently,

the fibre strain may be established. The fibre temperature distribution may be
determined using the Raman approach -now feasible in single-mode fibres.
Alternatively, in the so-called Landau-Placzek ratio method [11], the intensity
of the Brillouin spontaneous backscatter

Page 99

signal may be compared with the Rayleigh backscatter signal, the ratio of these
signals being directly related to temperature and having only a weak strain
dependence.
An alternative, single-ended method has also been proposed, using a
heterodyne detection arrangement in which the probe pulse is frequencyshifted prior to launching into the sensing fibre [12]. A non-frequency shifted
portion of the source output is used to provide a local oscillator.
The Brillouin method has several significant advantages, including its use of
standard telecommunication-type single-mode fibre; there is no need for the
fibre to be made from special materials or to contain embedded features, such
as in-fibre Bragg gratings. This results in lower-cost fibre, which is likely to be
stronger than more highly-processed fibre.
Technologically, the Brillouin method requires far higher control of the
frequency of the sources and filtering elements; however, this component
technology is rapidly becoming available owing to the development of dense
WDM networks in telecommunications.
In the case of the Brillouin methods, the sensing fibre is also easy to install
since all sections are equivalent in sensitivity and a truly continuous sensor is
obtained, as in the case of Raman-based devices, but unlike those sensors
incorporating Bragg gratings or miniature Fabry-Prot talons. However,
because of their distributed nature, Brillouin-based sensors may be more
difficult to apply where truly single-point measurements are required- e.g.
across a crack. Nevertheless, where an entire strain profile is required, the
Brillouin method is promises to provide moderately accurate (of order 0.01%
extension) strain values, with resolutions variously quoted from 1 to 10m over
very long (>20km) fibre lengths.
The Brillouin technique has been reported primarily in applications where the
strain within an optical cable is determined. Methods for embedding long
lengths of fibre inside a structure, e.g. a concrete construction, without
causing excess loss and transfer strain effectively along the entire length of the
embedded section is a serious challenge to be overcome before the method
can used widely within the civil engineering field. Even where the fibre is
sensing strain within steel cables in major construction projects, the issue of
transferring the strain to the fibre will require significant engineering work.

Conclusion
Distributed temperature sensing, using spontaneous Raman scattering, is
finding viable application in three main areas, namely, electrical power
transmission, plant monitoring and geological measurements. The number of
systems installed by York Sensors is well above 120 units.
The technology is being driven forward by increasingly challenging demands of
the users and recently systems based on single mode fibres have been
introduced.
Further technical challenges are to be found in the use of the technology to
new fields where the applications engineering is a significant factor.
Clear progress is being made towards the development of distributed strain
sensors which are especially appropriate in the long-term monitoring of major
civil engineering projects.

Page 100

References
[1] Hartog, A.H., 1983, ''Distributed Temperature Sensor Based on Liquid-Core
Optical Fibres" J. Lightwave Technol., LT-1:498509.
[2] Barnoski, M.K. and Jensen, S.M., 1976, "Fiber-Waveguides: a Novel
Technique for Investigating Attenuation Characteristics", Appl. Opt.,
15:21122115.
[3] Dakin, J.P. et al., 1985, "Distributed anti-Stokes Raman Thermometry"
Proc. 3rd Int. Conf. on Opt. Fibre Sensors, San Diego, February (post-deadline
paper)
[4] Hartog, A.H. et al., 1985, "Distributed Temperature Sensing in Solid Core
Fibres" Electron. Lett., 21:1061-3.
[5] Farries, M.C and Rogers, A.J, 1984, Distributed Sensing Using Stimulated
Raman Interaction in an Optical Fibre, Proc. 2nd Int. Conf. on Optical Fibre
Sensors, Stuttgart, paper 4.5, pp121-32.
[6] Dakin, J.P., 1987, "Distributed Fibre Temperature Sensor using the Optical
Kerr Effect", Proc. SPIE, 798, Fibre Optics Sensors II, pp149156.
[7] Nakayama, J. et al., 1987, "Optical Fiber Fault Locator by the Step
Frequency Method", Appl. Opt., 26:440-3.
[8] Hartog, A.H., 1995, "Distributed fibre-optic temperature sensors:
technology and applications in the power industry", Power engineering, June
issue.
[9] Hartog, A. H. 1995, "Fibre-optic temperature sensors monitor Wakamatsu"
Modern Power Systems, February issue, pp2528.
[10] Horiguchi, T. et al., 1990, "A Technique to Measure Distributed Strain in
Optical Fibers", IEEE Photonics Technology Letters, 2:352-4.
[11] Lees, G.P. et al., 1998, "Advances in Optical Fiber Distributed
Temperature Sensing using the Landau-Placzek Ratio", IEEE Photonics
Technology Letters, to appear in January issue.
[12] Shimizu, K. et al., 1994, "Coherent Self-Heterodyne Brillouin OTDR for
measurement of the Brillouin Frequency Shift Distribution in Optical Fibers", J.
Lightwave Technol., 12:730-6.

Page 101

Fiber Optic White Light Distributed Sensor for Condition Monitoring of Civil
Structures
Z. Chen,
A. Mendez,
Q. Li and
F. Ansari
Abstract
Condition Monitoring of civil structures is best accomplished by way of
distributed sensors. Sensors capable of making distributed measurements allow
for monitoring the entire structure. Optical fiber sensors are especially very
attractive for this purpose, since they are geometrically versatile and they can
be readily integrated within various types of structures and materials. The
research presented here describes the development of a new optical fiber
sensor system for measurement of structural strains based on white light
interferometry. Individual sensors are linked through simple connectors, and an
optical switch provides for multiplexing of strain signals from various locations
in the structure. Redundant Bragg grating type fiber optic sensors as well as
strain gauges were employed for comparison and verification of strain signals
as measured by the new system. The system provides capability for distributed
sensing of strains in large structures.
Introduction
In contrast to the existing nondestructive evaluation techniques, optical fibers
are able to detect minute variations in structural conditions through remote
measurements. It is possible to monitor the initiation and progress of various
mechanical or environmentally induced perturbations in concrete elements by
way of fully integrated optical fiber sensors. The full potential of optical fiber
sensors is fully realized in distributed and multiplexed applications. In these
configurations, information about the state of strain and or damage within
various sections of a structure is acquired simultaneously and in real-time [1].
Visiting Scholar, Smart Sensors & NDT Laboratory, New Jersey Institute of
Technology, Newark, NJ 07102.
Group leader, Fiber Optics Laboratory, ABB Electric System Technology, Raleigh, NC
27606. Post Doctoral Fellow, Smart Sensors & NDT Laboratory, New Jersey Institute of
Technology, Newark, NJ 07102.

Professor, Smart Sensors & NDT Laboratory, New Jersey Institute of Technology,
Newark, NJ 07102.

Page 102

Multiplexed sensors are usually constructed by combining a number of

individual sensors for measurement of perturbations over a large structure.
Theoretically, it is possible to use optical switching and other innovative ideas
for this purpose. A promising technique is based on wavelength division
multiplexing by using Bragg gratings optical fiber sensors [2]. In using this
technique, a broad band light source (white light containing a broad spectrum)
is employed for scanning a number of Bragg grating type sensors in series
and/or in parallel. The reflected wavelength of each Bragg is slightly different
from the other. In this way, wavelength shifts of individual sensors are
recognized and detected which is then related to the magnitude of strain at
specific sensor locations.
Distributed sensors make full use of optical fibers, in that each element of the
optical fiber is used for both measurement and data transmission purposes [3].
The purpose for making measurements by distributed or multiplexed optical
fibers is to determine locations and intensity of strains along the entire length
of fiber. These sensors are most appropriate for large structure applications,
due to their multi-point measurement capabilities. A distributed sensor permits
measurement of a desired parameter as a function of length along the fiber.
The most widely employed distributed sensing technique is based on
measurement of propagation time delays of light traveling in the fiber based on
the strain-induced change in the transmission of light. An optical time domain
reflectometer (OTDR) is used for this purpose. A from a number of partial
reflectors along the fiber length are recovered from the same fiber end. By
using this concept, it is possible to determine the location and magnitude of
the strain field within different sections of the structure [4].
The research presented here describes the development of a new multiplexed
optical fiber sensor that provides distributed measurement capability for
structures. The optical fiber sensor is based on white light interferometry. The
white light sensor is highly accurate and very sensitive since it uses
interferometry for measurement of strains. The system is more efficient than
the previously developed distributed sensors in terms of signal conditioning
instrumentation and cost. An optical switch provides for multiplexing of strain
signals from various locations in the structure. The experimental program
included testing of a specimen instrumented with the white light sensor
system. For comparison a number of redundant Bragg grating type fiber optic
sensors as well as strain gauges were employed for verification of strain signals

as measured by the new system.

Fiber Optic White Light Sensor
Michelson interferometers have been extensively employed for characterization
of laser and light emitting diodes (LED) in terms of coherence length [5].
Coherence length of a light source pertains to the ability of the lightwave to
retain a stable phase difference in time. Lasers are capable of producing single
wavelength emissions possessing long coherent lengths. On the other hand
LED's produce white light of low coherence. A white light source such as an
LED produces broad band emissions. In a Michelson white light interferometer
arrangement, light from an LED is split into two beams (Fig. 1). One of the
beams travels a fixed distance (reference beam). The

Page 103

path length of the other beam is allowed to change (sensing arm). Then the
beams are recombined, and if the path length of the variable beam of light is
made equal to that of the reference beam, then an interference pattern similar
to that given in Fig.2 is generated. The difference in path length over which
the interference pattern resides is used for the determination of coherence
length. This same technique can be utilized as a powerful tool for
measurement of deformations and strains. The fiber optic based Michelson
white light interferometer is comprised of an LED, a fiber optic coupler for
separating and recombining the light, two optical fiber arms with partially
reflective surfaces, and a scanning mirror mounted on a piezoelectric
actuator/stepper motor positioning system (Fig. 1). The length of the fiber
optic sensor arm is fixed, and the reference arm is made slightly shorter than
the sensing arm (about 1~2 mm). Separate beams of light travel through both
the reference and the sensing arms of the optical fiber. The partially reflective
surfaces define the gauge length of the sensor. The reflective surfaces in both
of the optical fibers guide the light back to a detector by way of a coupler. The
sensing arm is adhered to the material under test. In the unloaded position the
mirror scans a short distance in front of the reference arm. Once the sum of
the scanned distance plus the length of the reference arm equals that of the
sensing arm white light fringes similar to those shown in Fig.4 will appear. The
zero order fringe which is approximately in the center of fringe pattern and has
the highest amplitude and corresponds to the exact match in the optical path
lengths of these two beams.

Figure 1.
Fiber optic white light interferometric strain sensor system.

If the sensing arm is deformed, the mirror in front of the reference arm has to
once again scan the distance in front of the reference arm in order to match
the two optical paths corresponding to the reference and the sensing arms of

the fiber after deformation. Once the optical paths are matched a new white
light fringe pattern

Page 104

develops. This procedure can be repeated for locating the new white light
fringe patterns for subsequent straining of the sensing arm. As shown in Fig.3,
the distance between the zero-order fringe patterns for the undeformed and
deformed positions gives the amount of the optical path shift within the 2Lgauge length. White light interferometery is a very powerful technique, and it
provides for high-resolution measurement of deformations. The sensitivity of
measurements depends on the resolution of the scanning mirror. For this
reason, in the present study experiments, a high-resolution stepper motor (1
micron step intervals) was employed for coarse scanning of the distance, and
the piezoelectric actuator (sub-micron resolution) was used for fine tuning the
location of the fringe pattern. This experimental arrangement provided for a
resolution of 0.01 microns for the measured mirror displacements. The
procedure can be repeated for measurement of successive deformations by
way of automation.
The displacement of the mirror is expressed by the change in optical path, Dx,
shown in Fig.3. Dx is employed for measurement of the strain induced in the
sensing arm. It is expressed as (6):

where n is the refractive index, Dx is the combined displacement of the mirror

and the piezoelectric actuator. The first term DL in Eq. (1) represents the
physical change of optical fiber length produced by the strain. It is directly
related to the axial strain, e, through the expression:

The second term, the change in optical path due to change in the refractive
index of the fiber is given by:

Thus, we have:

where,

represents the effective refractive index of the glass

fiber core. The optical properties of the fiber core are the wavelength

Page 105

l = 1300 nm, and the refractive index, n = 1.46. The mechanical properties
are given in
Reference [6] and are the Poisson's ratio, v = 0.25, and the photo-elastic
constants p11 0.12, and p12 0.27. By using this information, the effective
index can be calculated as neff 1.19. Hence, the strain induced in the optical
fiber within the 2L gauge length is computed from the fringe shift, Dx by:

Figure 2.
White light fringe pattern.

Page 106

Figure 3.
Typical shift in white light interference fringe due to straining of the optical fiber.

Sensor Structure Interaction

Optical fibers are either structurally embedded during production of elements
or attached to the surface of the test piece by way of adherents. The ability of
an optical fiber sensor to monitor strain distribution in a structural material
depends on the bonding characteristics between the test material and the
optical fiber. The strain field transferred from the structure to the optical fiber
sensor generates changes in the characteristics of the light signal transmitted
by the glass core of the optical fiber. Transduction of this signal will provide
means for measurement of strain. However, the strain acquired directly from
the output of the optical signal is not entirely representative of the strain
induced in the structure. This is due to the fact that the optical fiber core is
protected by a layer of coating (polymeric or silicon based) which is more
compliant than the glass core of the optical fiber. The strain transfer
characteristics of optical fibers depend on the mechanical properties of the
fiber protective coating, the glass core, and the gauge length or the length of
optical fiber in contact with the structural material. In almost all cases, the
modulus of elasticity of the protective coating is considerably lower than the
one for the core. Therefore, the strain experienced by the structure is not
wholly transferred to the core, and some of it is lost through the interface
shear transfer within the protective coating.

Page 107

Sensor Calibration
The objective for calibration was to develop the relationship between the actual
structural strains and those measured by the optical fiber sensor. In a previous
study [7] a theoretical relationship was established for this purpose. Complete
details pertaining to the formulation of theoretical relationships and
experimental procedures can be found in reference [7]. A brief description of
the procedures is given here for completeness. A tapered cantilever beam was
employed for calibration of the optical fiber sensor. Due to the tapered width
geometry of the beam, end loads create a constant strain field along the beam
length. The optical fiber as well as a strain gauge was employed for
measurement the strain in the beam. By making the redundant measurement
it was possible to evaluate the level of strain loss due to the coating of the
optical fiber.
The experimental program encompassed testing of fibers ranging in length
from 30 to 180 mm. Each sensor series corresponded to a different gauge
length. Three replicate sensors were fabricated for each series in order to
examine the repeatability of measurements. As shown in Fig.4, experiments
and data acquisition was controlled by a computer through a general purpose
interface bus (GPIB). The controller card was employed in providing commands
for successive application of load to the cantilever beam, and to drive the
mirror mounted stepper motor in the white light interferometer system.
Command programs for the GPIB controller were written in C language. The
command program included routines for detection of white light fringes and
measurement of Dx. The cantilever beam end was successively deflected at its
end, and the measured strains were acquired by a data acquisition board and
displayed in real-time. Two sets of data corresponding to the optical fiber and
the strain gauge are acquired at any instant. The computer controlled the
movement of the stepper stage and the piezoelectric actuator in order to follow
the shift in the white light fringe pattern. Experimental results shown in Fig.5
indicate that the fiber optic measured strains pertain to a fraction of strains
induced in the structural material as measured by the strain gauge. The gauge
length dependency of the measurements are clearly expressed by these
results. According to these experimental results, the linear relationship
between the strains measured by the strain gauge, e sg , and the
corresponding strain ( ) measured by the optical fiber over the gauge length
2L, can be expressed as:

where, a is the constant of proportionality between the two measurements and

it is a function of the optical fiber gauge length, L, and its mechanical
properties, k. Assuming that the strains measured by the strain gauge system
correspond to the actual structural strains (within 37% experimental error).
These formulations lead to the determination of as:

Page 108

Figure 4.
Experimental setup.

Figure 5.
Experimental results.

Page 109

Multiplexing Methodology
Multiplexed sensor systems are developed through fusion of several sensors.
They can operate in series or in parallel, and they are employed for discrete or
distributed measurement of strains in structures. The multiplexed system
described here was constructed by coupling of several white light sensors in a
distributed fashion. In other words, the system consists of a number of sensors
attached to each other in series. As shown in Fig. 8, the sensing arm of the
sensor consists of several different gauge lengths of the white light sensor.

Figure 6.
Multiplexed white light optical fiber sensor system.

An optical switch was used for transmission of optical signal into every sensor.
A coupler was employed for dividing the optical signal for transmission into the
sensing and the reference arms of the sensor. The reference arms of individual
sensors were connected to individual channels of the optical switch and they
varied in length according to the gauge lengths of the individual sensing arms.
A motorized mirror stage was used for balancing of the optical paths for all the
sensors according to the white light sensing principles discussed in the
previous sections. The sensing arm was constructed by combining N sensors in
a series fashion with gauge lengths l1,l2,,lN respectively (Fig.7). Once the
sensors deform the deformed length of the

Page 110

N sensors change to
and strains become
The
gauge lengths were measured considering the total length of the fiber from the
light source (LED) to sensor i as:
and deformations as:
The individual sensor strains were evaluated in the following manner:

Experimental Investigation
A four-sensor system was fabricated for use in the experimental program. A
stepped-width tensile specimen was instrumented with the distributed white
light sensor. The specimen geometry and dimensions are shown in Figure 8.
The purpose for the stepped width geometry was to induce distinct levels of
strain within the four sections of the tensile specimen. In this way it would be
possible to examine the capability of the multiplexed system for discernment of
individual strain states within the four sections of the specimen.
The specimen was further instrumented with two other types of redundant
sensors for comparison with strain measurements by the white light system.
The redundant sensors consisted of the conventional strain gauges as well as
the fiber optic Bragg grating system. The experimental arrangement is
schematically depicted in Fig. 9. The specimen was held horizontally in
between two supports and tensile loads were applied through a weigh and
lever assembly. Weigh blocks were used for gradual loading of the specimen. A
multi-channel data acquisition board was employed for recording of strain data.
Replicate experiments were performed in order to examine the repeatability of
strain measurements. Strains were also evaluated through simple elastic
analysis of the tensile specimen. Comparison of the experimental results and
analytical calculations are given in Figure 10. Results shown in Figure 10
correspond to the strain distribution within various sections of the specimen. As
shown in this Figure, strains as measured by the various techniques are in very

good agreement.

Page 111

Figure 7.
Optical gauge lengths and deformations for the white light distributed sensor.

Conclusions
Development of a new optical fiber sensor system for measurement of
structural strains based on white light interferometry was described. The
system uses an optical switch in order to provide multiplexing capabilities for
strain measurements from various locations in the structure. A stepped-width
tensile specimen was employed for evaluation of the sensor system. The
specimen was instrumented with three different types of sensors that included
white light and Bragg grating type fiber optic sensors as well as strain gauges.
Comparison of experimental results indicated very close agreement among the
measured strains. The white light system can be configured with numerous
sensors for large structural applications. Gauge length of the individual sensors
within the system can be designed according to the structural requirements.
This feature is especially very attractive for large structural applications.

Page 112

Figure 8.
Specimen geometry and dimensions.

Figure 9.
Experimental setup and the instrumented tensile specimen.

Page 113

Figure 10
Comparison of strain measurements along the length of the specimen.

Acknowledgements
This work was made possible by a grant from the National Science Foundation,
grant number CMS-9402671, to the New Jersey Institute of Technology.
References
1. F. Ansari, 1997. ''State-of-the-art in the Applications of Fiber Optic Sensors
to Cementitious Composites," Cement & Concrete Composites, 19(1), pp. 319.
2. A.D. Kersey, W.W. Morey, 1993. "Multiplexed Bragg Grating Fiber-Laser
Strain Sensor System with Mode-Locked Interrogation," Electronic Lett. 29, pp.
112118.
3. F. Ansari, 1997. "Theory and Applications of Integrated Fiber Optic Sensors
in Concrete Structures," in Intelligent Civil Engineering Materials and
Structures, Edited by F. Ansari, A. Maji, and C. Leung, ASCE, SP, New York,
pp. 228.

Page 114

4. B.D. Zimmerman, R.O. Claus, 1993. "Spatially Multiplexed Optical Fiber Time
Domain Sensors for Civil engineering Applications," in Applications of Fiber
Optic Sensors in Engineering Mechanics, ASCE, SP Ed., F. Ansari, New York,
pp. 280-87.
5. D.A. Jackson, 1993. "Engineering Applications of Fiber Optic Sensors," in
Applications of Fiber Optic Sensors in Engineering Mechanics, ASCE, SP Ed., F.
Ansari, New York, pp. 216.
6. Butter, C.D., Hocker, G.B, 1978. "Fiber Optic Strain Gauge," Applied Optics,
17(8), pp. 28672868.
7. F. Ansari, and Y. Libo, 1998. "Mechanics of Bond and Interface Shear
Transfer in Optical Fiber Sensors," ASCE, J. Engrg. Mechanics, 124(4).

Page 115

Chapter 4
Condition Monitoring of Bridges

Page 117

Reliability of Optical Fibers and Bragg Grating Sensors for Bridge Monitoring
U. Sennhauser,
R. Bronnimann,
P. Mauron
and PH. M. Nellen
Abstract
We report on Bragg grating sensor applications on a bridge and reliability
modeling including laboratory tests. Durability and robustness of fiber
components is essential for their use in optical communication networks and
sensor applications. Breaking strength data of accelerated aging of fibers for
telecom and sensor applications were fitted to a lifetime model and show that
high acceleration factors can be achieved with elevated temperature and high
humidity. On the other hand the out door application at the Stork Bridge in
Winterthur delivers valuable data for reliability modeling, and shows stable
operation for more than twenty months.
Introduction
Adequate strength and structural integrity have to be maintained during
lifetime of safety critical structures like bridges. State of the art testing
methods usually include removable detection systems recording strains,
displacements, frequencies, amplitudes and damping factors for well defined
static or dynamic load states. With these methods residual stress in prestressed
elements cannot be determined non-destructively. Long term monitoring with
permanently installed detection systems can avoid this problem, but is very
demanding with respect to high reliability and negligible drift over expected
lifetime of sensors and read-out systems. Furthermore costs of the nonremovable parts compared to reinstallation costs have to be considered.
Fiber optic sensors based on fused silica glass are intrinsically very stable, if
they are protected and treated properly. Bragg gratings are particularly
attractive as direct strain detectors due to their adequate accuracy,
multiplexing capability and potentially low cost for high volume production.
Several methods of strain
U. Sennhauser, R. Brnnimann, P. Mauron, Ph. M. Nellen, EMPA, Swiss Federal
Laboratories for Materials Testing and Research, Ueberlandstrasse 129, CH-8600
Duebendorf, Switzerland

Page 118

detection with Bragg grating sensors have been developed and applied for
monitoring of civil structures [1 5]. Durability of optical fiber sensors includes
parameters of glass, cladding, coating, protective sealing, Bragg grating
production and environmental conditions as temperature, humidity and
mechanical stress. For strain sensors coating and protection have to allow
proper strain transfer and fulfil adhesion requirements. Although mechanical
strength and fatigue through stressenhanced crack corrosion has been
investigated [6,7], there is no full understanding of the microscopic aging
mechanism and lifetime predictions are inaccurate. Extensive studies have
been performed demonstrating optical stability of Bragg gratings even at
elevated temperature [8,9].
In this paper we report on reliability and durability investigations of optical
fibers and Bragg grating sensors for long term strain monitoring. Results will be
given of the first 20 months of strain measurements of the carbon fiber
reinforced polymer stay cables of the Stork Bridge (figure 1) in Switzerland.
Data is compared to results obtained with electrical resistance strain gages.
Environmental data as temperature and humidity inside the protective tube is
collected as basis for laboratory testing.

Figure 1:
The Stork Bridge in Winterthur

Reliability and Durability Modeling and Testing

Reliability and durability of fiber optic sensors are strongly dependent on type
of optical fiber and coating, method of Bragg grating production including deand recoating process, time structure and energy of UV-irradiation, mechanical
stress and environmental conditions like temperature, humidity, pH-value and
type and mobility of ions. The strong coupling of several of these parameters
requires strict control of testing conditions to obtain consistent data as required

for modeling and comparison to data from other work.

Page 119

Optical Properties
The core of fibers used for Bragg grating production is doped with Germanium
or loaded with hydrogen to make it sensitive to UV-irradiation producing small
periodic changes of refractive index n of 10-3 to 10-5.
Microscopic structure and physical properties of Bragg gratings are not yet fully
understood. Aging models have to be used for estimating decay time based on
accelerated tests. Models are phenomenological with the parameter values
depending on fiber type and grating production procedures as well as on
wavelength and range of reflectivity.
The decay of grating strength can be approximated with a power law, which
can be explained by a sum of (exponential) Arrhenius decays. A distribution of
activation energies with a central value of 2.8 eV and a FWHM of 1.6 eV has
been reported for germanium-doped silica fiber [9]. Since the population
density in activation energy space is not constant during aging due to different
decay times, data not corrected for pre-annealing or not considering density of
states for different fiber types are not consistent.
For commonly used production techniques decay has been shown to be small
up to 300 C and for some fibers up to 600 C after a burn-in phase of a
fraction of an hour [10]. If properly preannealed, decay of reflectivity over an
expected lifetime of 20100 years is negligible for most applications in civil
engineering where temperature is below 80 C. Since the strength of gratings
is given by a change of index of refraction of 10-3 to 10-5, substantial decay
may cause an offset drift in strain sensors which are typically sensitive to 1
mm/m.
Mechanical Strength
Mechanical strength and fatigue of optical fibers can be modeled with fracture
mechanics of brittle material. Surface flaws stemming from fiber or Bragg
grating production are growing by stress-enhanced crack corrosion mechanism.
Although permeation of vapor and ions through fiber coating is directly
influencing lifetime, it is usually not considered in these models. This is limiting
comparability of test with different coatings.
By aging fibers under several climatic and mechanical stress conditions before
strength testing the acceleration factors for degradation of mechanical

strengths can be determined phenomenologically and activation energies can

be extracted. Due to large number of parameters, which may even be
correlated, performance of very time consuming comprehensive test programs
is necessary to increase accuracy of high acceleration factors as required for
predicting the very long lifetimes.
In our laboratory a comprehensive test program is in progress to improve
accuracy of accelerated aging models and reduce test time by extending range
of climatic conditions to 125C and 85% relative humidity. Fracture strength of
fibers and gratings are measured with a dynamic strength testing apparatus
according to standard IEC 793-1-3. To evaluate data a two parameter Weibull
distribution was assumed; it was found to provide a good model for fiber
fracture strength

Page 120

[equation(1)]. The cumulative fracture probability F is a function of the

dynamic shape and scaling parameters md and s0.
Figure 2 shows Weibull plots, cumulated fracture probability F versus fracture
stress, s, of standard telecommunication fibers aged at low mechanical stress,
85% relative humidity and at various temperatures of 60 C to 150 C. Fiber
strength is reduced and the distribution is braodened with a dual mode
behavior at 150 C. Shape parameters (md-value) of Weibull distributions of
more than 100 for new fibers are reduced to less then 10 after aging (Figure
3).

Figure 2:
Weibull plots of standard telecom fibers after 5 days aging. The
conditions were: coils ~320 mm, 85% rH, temperature range
from 60 C to 150C. After aging the fibers were room-conditioned
overnight before measurements.

Figure 3.
Reduction of md-value during aging (log-scale). Smaller md-values
correspond to broader fracture strength distributions.

Page 121

Strength degradation of fiber with time has been analyzed with a combined fit
using an empirical model [equation (2)] first proposed in [11] (Figure 4).

Time constant a is assumed to have an Arrhenius behavior, i.e., a Boltzmann

factor describes its temperature dependence and one single degradation
process of activation energy EA is responsible for strength degradation. The
values bs0, and a0 are model parameters, k is the Boltzmann constant.
Activation energies fitted are in the range of 0.8 to 1 eV for investigated fiber
types. They agree well with the value published in [12]. Residual strength for
aging at room temperature has been estimated by extrapolation with an
activation energy of 0.82 eV(Figure 4). The dotted lines indicate the lifetime
uncertainty caused by an uncertainty of 0.07 eV of the activation energy.
Aging time is reduced by a factor of 3000 by increasing temperature from 25C
to 125C. However at elevated temperature phase transition of materials may
change the effective activation energies or other parameters of both glass or
coating. We found that a temperature of up to 125C was appropriate for
accelerated aging of fibers with acrylic coating, although most showed a
tendency to become yellow above 85C. Aging at 150C, 85% relative humidity
for 5 days degraded coating more severely leading to a corrugated surface
(Figure 5).

Figure 4:
Breaking strength of standard telecommunication fibers after aging
at 60, 85, 110, 125, and 150C and 85% rH. Thin lines: combined
fit of given example (a0=9.8.109, b=0.33, EA=0.82 eV);
bold line: extrapolation to room temperature; dashed line:
extrapolation to 25 C with values given in [12]; dotted lines: EA

confidence interval (other values left constant).

Page 12

Figure 5:
Coating surface of an acrylate-coated fiber after
exposure to 150 C/85% rH for 5 days. Surface of
a non-aged fiber is smooth and not corrugated as
shown here.

Exposing a standard double layer acrylic coating to 23C and 95% relative
humidity results in an absorption of about 1.4% by weight of water (Figure 6).
Dependence of fracture strengths from ambient humidity can vary up to 0.5 GPa
when humidity in laboratory is not kept constant.

Figure 6:
Weight increase due to water absorption of 10 m standard telecom pristine fibe
r exposed to
23 C, 95% rH after storage at 23C, 30% rH (left) and change in fracture strength with tim
at 23 C, 30% rH after storage at 23 C, 95% rH (right).

Mechanical strength and aging behavior of optical fibers are strongly affected by
the Bragg grating production processes. Methods using stripping and recoating
are potentially damaging by introducing additional surface flaws. By surface
melting and reflow during irradiation a periodic structure can be induced at the
surface depending on thermal load. A change of intrinsic material strength and
the material aging coefficient due to modification of microscopical structure must
be expected.
Adequate reproducibility in a grating production process is required to properly

separate the effects of each step on strength and durability. This might explain
inconsistencies in published data. For high confidence level of the reliability
function, all fibers should be proof-tested after recoating.

Page 123

Fiber Sensor Application and Protection

In strain sensor applications not only strength of fiber itself is of importance
but also adhesion of coating to cladding and to surrounding glue and
specimen. In low cost applications exotic solutions for expensive coatings,
Bragg grating production methods or time consuming application procedures
are excluded. In some applications Bragg sensors can be applied in a
compressive state or on already prestressed specimen to reduce long term
tensile stress states of optical fibers and to reduce adhesion problems.
Monitoring of Stork Bridge in Winterthur
The Stork Bridge
In 1996 the old Stork Bridge in Winterthur, Switzerland was replaced by a stay
cable bridge. As a novelty one pair of the twelve cable pairs was made of CFRP
(carbon fiber reinforced polymer) wires instead of the usual steel wires [13].
High specific strength, corrosion resistance and excellent fatigue behavior of
CFRP are its main advantages. The novel design of the cables requires to
monitor their performance.

Figure 7:
Schematic view of the new Stork Bridge in Winterthur, crossing 17 railway tracks

Figure 7 shows a drawing of the bridge, its dimensions and the position of the
CFRP cable pair. Each CFRP cable is equipped with an array of seven Bragg
grating sensors, conventional resistance strain gauges (RSG), temperature and
humidity sensors, and displacement transducers in the anchors of the cables to
monitor the behavior of the cables during construction, under traffic loads, and
daily and seasonal fluctuations.
The two CFRP cables have a length of 35 m and they consist of 241 CFRP

wires each with a diameter of 5 mm, assembled in a hexagonal structure. Only

three of the seven Bragg gratings of the array (BGI, BG4 and BG7) are
adhered to loaded

Page 124

wires positioned at the circumference and spaced by 120 relative to the

center of the cable. In this arrangement the average load as well as the load
gradient can be measured. The other four Bragg gratings (BG2, BG3, BG5 and
BG6) are used as dummies for temperature compensation and for monitoring
of long term drifts.
The dummies were adhered to 30 cm long CFRP wires of the same material
which were attached to loaded wires in a way that they were insensitive to the
load on the cable. BG5 and BG6 were adhered in a pre-strained state. The
strain on these dummies is about 2500 mm/m, more than double the strain
experienced by the active gratings. Creep or debonding of the fiber coating or
the epoxy adhesive due to environmental influences and consequently failure
of the monitoring system would result in a decay of the measured strain of
these Bragg gratings.

Figure 8:
Picture of the Stork Bridge during construction. The CFRP cable K43 is mounted.

Figure 8 shows the bridge during the mounting of one of the CFRP cables. The
other CFRP cable is already in place. Shipping, mounting and the construction
work were a major risk for damaging to the sensing system.
Monitoring System
Figure 9 shows the fiber optic part of the monitoring system. It is capable of
unattended operation of several sensing fibers. The radiation of a
superluminescence diode (SLED) is coupled into a fiber optical 2.2 coupler.
SLED, coupler, and current and temperature controller for the SLED are
mounted in a module which is plugged into an electronics rack. The three open
fibers of the coupler are accessible through connectors on the front panel of

the module. The fiber containing the Bragg grating array is connected to one
branch of the coupler. Each Bragg grating reflects a different wavelength. The
reflected radiation passes the coupler again and is guided to a 0.46m imaging
grating spectrometer equipped with an 1024 element CCD-array detector. The
signal of the CCD is digitized and recorded by a computer. The rack contains
three SLED modules to read-out sensing

Page 125

fibers. To couple all the signals into the spectrometer several fibers are
mounted in parallel on a plate and polished at the ends. This device acts as an
entrance slit.
For absolute wavelength calibration the light of a spectral lamp (argon) is
simultaneously coupled into the spectrometer by another fiber. An
incandescence lamp inside the spectrometer allows to calibrate the sensitivity
of the CCD. The SLED modules, spectral lamp and incandescence lamp are
computer controlled.

Figure 9:
Setup to read-out optical fiber Bragg grating arrays

Because the sensing fiber on cable K43 was broken during adhering the fiber
on the CFRP wires between BG 6 and BG 7 both ends are connected to two
different SLED.
Results

Figure 10.
Strain measurements performed during construction and use of the Stork Bridge.

The fiber optical measurement system is already operational for almost two
years and works satisfactory. Results obtained from one of the cables are

shown in

Page 126

figure 10. The load changed strongly during the construction time. The
reference (zero level for the strain measurement) corresponds to an initial load
of 200 kN after installation of the cables on the bridge which itself corresponds
to a strain of 280 mm/m. Results of the measurement with conventional
resistance strain gauges are also given in figure 10 and show a good
agreement.

Figure 11:
Daily variation of strain and temperature in one CFRP cable.

Figure 11 shows the evolution of strain and temperature on one cable. The
load correlates strongly with the temperature as expected: Steel has a large
thermal expansion compared to CFRP. With falling temperature the steel wires
contract, taking up more load. This effect can also be seen in figure 10. The
strain difference from summer to winter is about 200 mm/m.

Figure 12.
Change of relative strain of the dummies.

It can be disastrous if a monitoring system continues to deliver apparently

reasonable data while it is actually defective as it can happen in the case of
debonding. As mentioned above some of the Bragg gratings are mounted on
dummies in a prestrained state. If slippage occurs the measured strain of the

Pag

prestrained dummies would decay. Figure 12 shows the development of the stra
the dummies relative to their average strain. No systematic change could be
observed so far. The long term measurement uncertainty of the complete
measurement chain corresponds to a standard deviation of 8 mm/m calculated w
the correct number of degrees of freedom.

Figure 13:
Plot of relative humidity vs. temperature measured inside the CFRP cable. The diamonds in
the measured date. The observation time is one month.

A major risk for the durability of the sensing system was the possibility of high re
humidity at elevated temperature inside the protective tube[14]. Therefore
temperature and humidity in the cables have been continuously recorded. The
humidity varies between 30% to 60% rel. humidity while the temperature is in th
range from - 3C to 35C (tree hours average over a year). The humidity is rough
anticorrelated with the daily temperature change. In figure 13 the humidity is pl
vs. the temperature. The humidity drops with rising temperatures. At constant
temperature the humidity drifts towards the average. This behavior can be expla
by a buffer effect stabilizing humidity. Probably the epoxy forming the matrix of t
CFRP wires acts as such a buffer. The important point for the reliability of the sen
system is the fact that temperature was below 40C and that the humidity is bel
50% at this temperature most of the time. Knowledge of time dependent
temperature and humidity can be used for accelerated aging of laboratory sampl
and durability modeling.
Conclusion

Critical parts of civil engineering structures require monitoring of their performan

Reliability of non-repairable sensing systems is critical. The application on the Sto
Bridge works stable for almost two years now but it showed also the importance
system redundancy. The collected climatic data delivered valuable input for testi

and modeling. Results of accelerated tests of

Page 128

breaking strength and its modeling are presented. Reliable of useful life time of
fiber optical sensing systems will require additional modeling and testing.
Acknowledgments
The authors wish to thank Charles G. Askins and Martin A. Putnam, Naval
Research Laboratory, Washington DC, for providing Bragg grating arrays. Urs
Meier, EMPA, is acknowledged for supporting the field tests at the Stork Bridge
in Winterthur.
References
1. Askins C. G., T.E. Tsai, G. M. Williams, M. A. Putnam, M. Bashkansky and
E.J. Friebele, 1992. Fiber Bragg reflectors prepared by a single excimer pulse'',
Opt. Lett. 17(11), 833
2. Maashant R., T. Alavie, R. M. Measures, M. Ohn, S. Karr, D. Glennie, C.
Wade, G. Tadros, S. Rizkalla, 1994. Fiber optic Bragg grating sensor network
installed in a concrete road bridge", SPIE Vol. 2191: 457465
3. Merzbacher C. I., A. D. Kersey, E. J. Friebele, 1996. Fiber optic sensors in
concrete: a review", Smart Mater. Struct. 5: 196208
4. Nellen Ph. M., R. Brnnimann, U. Sennhauser, C. G. Askins, M. A Putnam,
1996, Strain measurements on concrete beam and carbon fiber cable with
distributed optical fiber Bragg grating sensors", Opt. Eng. 35(9):25702577
5. Sennhauser U., R. Brnnimann, Ph. M. Nellen, 1996. Reliability modeling
and testing of optical fiber Bragg sensors for strain measurements", SPIE Vol.
2839-07:6475.
6. Kurkjian C. R. and D. Inniss, 1991. Understanding mechanical properties of
lightguides: a commentary", Opt. Eng., 30(6):681689
7. Griffioen W., 1994. Evaluation of optical fiber lifetime models based on the
power law", Opt. Eng., 33(2):488497
8. Morey W. W., G. Meltz, and W. H. Glenn, 1989. Fiber optic Bragg grating
sensors", Proc. SPIE Vol. 1169:98107,
9. Erdogan T., V. Mizrahi, P. J. Lemaire, and D. Monroe, 1994. Decay of
ultraviolet-induced fiber Bragg gratings" J. of Appl. Phys. 76:7380,
10. Poignant H., J. Bayon, J. le Melot, D. Grot, E. Delevaque, M. Monerie,

1995. Thermal behavior of Bragg gratings", 1. European Workshop on Bragg

Grating Reliability, Berne, October 2, pp. 137176
11. France P. W., W. J. Duncan, D. J. Smith, and K. J. Beales, 1983. Strength
and fatique of multicomponent optical glass fibres" J. of Mat. Sci. 18:785793,
12. Kurkjian C. R., and M. J. Matthewson, 1996. Strength degradation of
lightguide fibers in room temperature water" Elec. Lett. 32(8):761762,
13. Meier U. and H. Meier, 1996. CFRP finds use in support for bridge" Modern
Plastics International 4:8385
14. Brnnimann, R., Ph. M. Nellen, and U. Sennhauser, 1998. Application and
Reliability of a Fiber Optical Surveillance System for a Stay Cable Bridge."
Smart Mater. Struct., Special Issue of Fiber Optical Structural Sensing: in press

Page 129

A Recent Experience in Bridge Strain Monitoring with Fiber Grating Sensors

R. Maaskant,
A. T. Alavie
and R. M. Measure
Abstract
A fiber Bragg grating sensing system has been recently installed in a new road
bridge near Winnipeg, Manitoba. It comprises a network of 65 fiber grating
sensors and resident instrumentation which is linked by telephone line to
engineering offices for continuous monitoring. A description of the installation
details serves as a vehicle for the examination of the need for new standards
and techniques to address the particular needs of this technology in structural
monitoring applications.
Introduction
Fiber optic Bragg grating sensors have been utilized in laboratory and field
applications in civil engineering structures for some time now and have
achieved a great deal of success (e.g. [1], [2]). Sales of commercial fiber
grating sensor systems, which have been available since 1993, are undergoing
a transition from exploratory applications to applications generated by current
measurement needs which require the special features offered by this sensing
technology. They are rapidly becoming the fiber optic sensor of choice because
of their unique combination of characteristics such as the following: the sensor
signal is spectrally encoded, the sensor is physically localized and intrinsic to
the optical fiber, and the sensors can be multiplexed serially.
The deployment of fiber optic sensing systems appropriate for widespread field
use has required the development of sensing instruments, sensors and sensor
installation techniques which satisfy the needs of the measurements
community. Many of the necessary concepts and procedures for the
deployment of fiber grating sensors in the field have been formulated by
individual developers of fiber optic sensing systems [3]. These have benefited
greatly from the established standards and techniques adopted by both the
telecommunications industry and those of the
Robert Maaskant, A. Tino Alavie, ElectroPhotonics Corporation, 7941 Jane St., Unit
200, Concord, Ontario, CANADA, L4K 4L6

Raymond M. Measures, Institute for Aerospace Studies, University of Toronto, 4925

Dufferin St., Downsview, Ontario, CANADA, M3H 5T6

Page 130

conventional electronic strain sensor industry. However, fiber optic sensors

possess unique features and are often applied in situations which call for the
establishment of new standards and techniques.
Among these new demands are unified calibration procedures, systematic and
uniform conventions for sensor parameters, and qualified sensor installation
techniques. In the context of civil engineering structures which have life spans
on the order of 50 or 100 years, long term sensor performance and durability
becomes an important issue. Here one is dealing with relatively new
requirements for which precedents, for the most part, are absent. In addition
to the issues surrounding the sensor and the sensing transduction mechanism,
implementation of large networks of sensors in bridge structures, for example,
requires the establishment of cabling and connectorizing techniques to ensure
long term integrity of the sensor network.
In what follows, some of these features are identified in the context of a recent
implementation of a network of fiber optic grating sensors in the Taylor Bridge
near Winnipeg, Manitoba, Canada. This sensing system consists of a total of 65
fiber optic grating sensors, resident on-site instrumentation and a telephone
link to engineering offices for continuous monitoring of the structure under
traffic loads.
The new road bridge incorporates fiber reinforced plastic (FRP) material as
reinforcement and pre-stressing bars in the girders, the bridge deck slab and
the barrier wall [4]. The project was implemented as part of a Canadian
national research and development effort entitled Canadian Network Centre of
Excellence on Intelligent Sensing for Innovative Structures (ISIS).
Bridge and Sensor Network
The Taylor Bridge is a two lane bridge spanning the Assiniboine River and
consists of 5 spans with a total length of 165 metres and a width of 9.6
metres. Each span consists of eight pre-stressed girders and two of the spans
contain FRP pre-stressing and reinforcement material and the network of fiber
optic grating sensors. A portion of the bridge deck and barrier wall is also so
equipped.
Within the north and south spans of the bridge, a total of ten girders contain
fiber grating sensors. Among these girders the FRP products, CFCC [Tokyo
Rope, Japan] and Leadline [Mitsubishi Chemical Corporation, Japan], are used

both as reinforcement and pre-stressing tendons. Two girders of each of these

materials were installed, one of which contains conventional steel stirrup
reinforcements while the other utilizes the FRP stirrup reinforcement (see
figure 1). Also shown is the deck slab region which is reinforced with Leadline
bars and the barrier wall which is reinfored with GFRP C-Bar [Marshall
Industries Composites Inc, USA].
The fiber grating sensor network is distributed over the structure as illustrated
in figure 1. The sensors are mounted on both steel and FRP reinforced girders
as well as in the FRP reinforced deck and barrier wall. The figure also illustrates
sensor positions on pre-stressing strands, stirrups and deck and barrier wall
reinforcement bars. Temperature sensors were installed for the purpose of
compensation for thermal apparent strain. Twenty-two temperature sensors
provide representative temperature measurements among different girders and
at various locations along and through the depth of the girder and deck.

Page 131

Figure 1.
Sensor network layout.

A typical sensor installation is shown in figure 2. Surface preparation

procedures and epoxy compounds common in foil strain gauge bonding
techniques were used. The epoxy used has excellent bonding properties and
provides good cover over the sensor for protection from both mechanical
disturbance and moisture infiltration. As a final protective measure, rubber
compounds both in liquid and sheet forms are used to seal and cover the
installation site.

Page 132

Figure 2.
Sensor installation

The fiber optic grating sensing system installed in the Taylor Bridge is based on
a parallel sensor architecture and hence requires independent fiber leads for
each of the 65 sensors installed. As shown in figure 1, the fiber network
consists of cabled single fiber leads and multi-fiber cables to connect the
sensors to the instrumentation which is located at one end of the bridge
structure.
The individual sensor leads are cabled in a rugged stainless steel tubing which
protects the fiber from crushing and pinching loads as it proceeds, embedded
within the concrete, to the end of the girder. The leads then enter a recessed
terminal box in the upper flange of the girder. The leads pass through the
terminal box and enter a conduit in the bridge deck which leads to a main pull
box recessed in the diaphragm at the pier location. The terminal box at the
end of the girder provides emergency access to the fibers immediately prior to
entering the deck conduit. The main pull box gathers the leads from all of the
girders within that span and houses the fusion spliced connections to the main
multi-fiber cable. The multi-fiber cable leads to the instrument enclosure
located at one end of the structure.
Some practical considerations which led to this arrangement should be noted.
The large number of sensors and, in the case of the south span, the long
distance to the instrument necessitated the use of multi-fiber cable. This cable
is designed to withstand the necessary pulling forces during installation and
also allowed the use of a relatively small conduit within the bridge deck. The
cable installation also influenced the decision to use fusion splices at the main
pull boxes. Although field splicing may be avoided by using pre-connectorized
leads at both ends of the main cable, this creates additional cable installation
difficulties because of the size and awkwardness of the pre-installed
connectors. Larger conduit size and special care during the installation

procedure makes this alternative less attractive than the requirement for the
relatively routine process of field splicing.
It is worthwhile at this juncture to consider the relative merits of the present
parallel architecture with those of a serially structured sensing system within
the context of the constraints of the present project. In the latter case, a
number of grating sensors are written into a single optical fiber at arbitrary
locations and all of these may be accessed by a single channel of a strain
indicator instrument. This has obvious advantages from the point of view of the
complexity of the optical fiber network since the number of leads may be
significantly reduced. However it must

Page 133

be recognized that there are also disadvantages leading from practical

considerations of sensor network reliability issues and installation procedures.
When a number of sensors coexist on a single fiber, the failure of that fiber or
the failure of any of the sensors along that fiber can jeopardize the remaining
sensors, particularly if these are embedded or physically inaccessible. From the
time of fabrication of the serialized grating through to the installation of the
sensor, the weakest points are the gratings themselves. This danger is
somewhat heightened by the awkwardness inherent in the handling and
installation of such a sensor.
In accordance with the project constraints, sensor installation must be
performed in-situ so as not to interfere with the girder and deck casting
process. With intermittent sensors along a single fiber, the ruggedized cable
must be discontinuous to allow for the sensor to be attached to the structure.
This can be somewhat troublesome for handling and installation, particularly in
the case of surface bonded sensors for which the sensor is bounded by the
significantly stiffer and heavier cable. When sensors can be pre-installed on
components within laboratory or controlled environments such difficulties can
be overcome with added care in the installation process. However, when it is
necessary to install the sensors in-situ, the possibility of damage or failure of
the sensors is greatly increased. These difficulties can be alleviated to a some
extent when the gratings are contained within more rigid transducers such as
weldable gauges and concrete embeddable gauges.
The need for sensors on the various reinforcement components such as the
prestressing tendons, the stirrup reinforcement, and the deck reinforcement
bars can also restrict the manner in which serialized gratings are used. To
avoid contorted paths for the fiber optic lead and the inherent dangers
associated with these, the serial sensor configuration must allow for a certain
degree of parallelism. This could take place at the instrument in the form of
multiple serial channels, or at the end of the lead fiber, where the addition of a
1xn coupler would allow multiple sensor leads to find their own way to
disparate sensor locations. The implications of this configuration for the
instrumentation must of course be considered.
Sensor Measurement System
In the Taylor Bridge project, the sensor strains are recorded with the use of a

fiber grating strain indicator (FLS3500R) and a 32 channel multiplexing unit

for quasi-static strain measurements. An independent channel for dynamic
measurement was also provided on the instrument. In addition to
measurements from a multi-channel temperature sensor instrument, the strain
readings are logged by a resident computer which can be accessed by modem
for data logging and downloading. Correction for thermal apparent strain may
be performed either in real time within the FLS3500R instrument, or by postprocessing after the data has been logged. The instruments and the computer
are housed in a heated enclosure mounted in the abutment of the bridge. This
arrangement is illustrated schematically in figure 3.

Page 134

Figure 3.
Schematic of instrumentation and office link

Preliminary results recorded by ISIS staff at the University of Manitoba are

shown in figure 4 for a test loading produced by a slowly moving truck and
trailer. The heavy truck made several passes over the North span, forwards and
backwards, and a fiber grating sensor located on the pre-stressing tendon at
midspan was able to record the strain excursion experienced by the girder. In
the figure 4 one can clearly identify both the truck and the trailer loading as
two distinct peaks in the response curve even though the magnitude of the
strains are quite small. The direction of travel can also be detected by the
relative magnitudes of the peaks since the tractor load is larger than the trailer
axle load. Hence the first event in the figure represents a backward pass and
the subsequent one is a forward pass over the sensor.
Discussion
An extensive network of fiber grating sensors has been successfully deployed in
the Taylor Bridge near Winnipeg, Manitoba. This provides an example of the
practical issues of installation and implementation of such a system for long
term structural monitoring. The fiber grating sensing technology is rapidly
becoming the fiber optic sensor of choice in these applications and as such
would benefit from the attention of the measurement community to address
the absence of standards and

Page 135

standardised techniques for calibration, specification and implementation of

these systems. These issues must also be assessed in the context of serially
multiplexed sensing systems where the multiple sensors coexist on a single
optical fiber. This architecture is very attractive where linear arrays of sensors
are required. However these advantages may in some circumstances be
counterbalanced by practical considerations of reliability and implementation.

Figure 4.
Truck passage data

References
1. Maaskant, R., Alavie, T., Measures, R.M., Ohn, M., Karr, S., Glennie, D.,
Wade, C., Tadros, G. and Rizkalla, S., 1994, "Fiber Optic Bragg Grating Sensor
Network Installed in a Concrete Road Bridge", SPIE Proc., 2191 (53):1318.
2. Davis, M.A., Bellemore, D.G. and Kersey, A.D., "Distributed Fiber Bragg
Grating Strain Sensing in Reinforced Concrete Structural Components", Cement
and Concrete Composites, 19: 4557.
3. Maaskant, R., Alavie, T., Measures, R.M., Tadros, G. and Rizkalla, S.H. and
Guha-Thakurta, A., 1996, "Fiber-Optic Bragg Grating Sensors for Bridge
Monitoring", Cement and Concrete Composites, 19: 2133.
4. Rizkalla, S., Shehata, E., Abdelrahman, A., Tadros, G., and Saltzberg, W.,
1998, "FRP and Optical Sensors for a Highway Bridge in Canada", to be
published in Proceedings of the Structural Engineers World Congress, ASCE,
San Francisco.

Page 136

Pulse-Echo Fiber Optic Sensor for Nondestructive Evaluation of Concrete

Bridges
X. Chen
and F.Ansari
Abstract
A high-resolution fiber optic ultrasonic sensor for distributed detection of flaws
and defects in concrete bridge elements have been developed. Distributed
sensing provides the opportunity for the optical fiber sensor to detect
anomalies throughout the entire structure. Optical fiber sensors can be
embedded in structures in new construction, or they can be adhered to the
surface of existing concrete elements. An ultrasonic transmitter source or an
impact device is used for point-by-point generation of stress waves within the
bridge deck. The optical fiber senses the stress waves, and a one-ended
measurement of the optical signal by an interferometer provides information
about the integrity of the impact location. The higher sensitivity of the optical
fiber allows for detection of weaker signals and therefore scanning capabilities
at larger member depths. This configuration eliminates the need for using a
PZT sensor that can only make localized measurements. PZT sensors need to
be moved around from point-to-point on the surface of the bridge deck, and
require a coupling medium for proper transmission as well as sensing of the
stress waves. Furthermore, fiber optic sensors will prove to be very useful in
applications where access is limited to one side of the structure only.
Introduction
Internal cracks and delaminations comprise some of the major forms of
deterioration in concrete bridge elements. Internal cracks grow and are
frequently initiated in the form of delaminations at the interface between the
reinforcement and the concrete. The most probable causes for the generation
of internal cracks and delaminations are interfacial bond stresses and the
accumulation of corrosion by-products at the interface between the
reinforcement and the concrete. Failure to detect these defects at the early
stages of development results in time consuming and costly repair and
replacement operations. Rapid condition monitoring methodologies need to be
developed to accurately identify the presence of cracks, estimate the extent of
damage, and locate the damaged zone for repair. The ideal nondestructive

testing procedure for bridge decks would provide information about the
existence and location as well as the extent of deterioration, corrosion or
structural damage. Moreover, it is important to perform the condition
monitoring operations in a very rapid manner in order to avoid lengthy closures

Page 137
1of the bridge to the traveling public. At this point in time there is no one

method that can provide this information in a rapid and yet accurate fashion.
Stress Wave propagation techniques such as the impact echo method [1] have
proven to be successful in locating cracks and delamination in concrete.
However, these techniques require high levels of sophistication on the part of
the operator. The method is based on the transmission characteristics of stress
waves at the boundaries of different layers. For example, at a delamination, the
reflection and refraction processes bring about changes to the energy of the
transmitted stress wave that can be picked up by a receiver. In practice, the
technique is time consuming and requires point-by point inspection of the
deck. The stress wave methods show promise as they have successfully
detected delaminations in field testing of concrete decks. However, the
technique does not fare well on asphalt covered decks due to inefficient
coupling of energy between the asphalt and concrete [2]. These drawbacks
have limited the practical usage of the impact echo method by the highway
engineers [3].
Objectives
To avoid long bridge closure times existing point-by-point NDT techniques need
to be replaced by distributed sensors for rapid scanning of concrete decks. The
work described here pertains to the development of a high resolution and
effective sensor for condition monitoring of concrete bridges. An optical fiber
sensor is designed for distributed sensing of stress waves generated by an
ultrasonic source on the surface of the deck. A methodology is developed for
detection of anomalies and defects within concrete elements. Higher
sensitivities achieved by these sensors yield detailed and more reliable flaw
detection capabilities. The optical fiber Sensors are versatile and can be
employed for a variety of applications including in new construction, existing
structures, and hard to access structural elements.
Fiber Optic Acoustic Sensors
A comprehensive review of literature is beyond the scope of the work described
in this article. However, some of the previous work is briefly described here.
Lyamshev et al. [4] developed a one-arm interferometric acoustic sensor by
way of a polarization maintaining optical fiber. Although, their technique did
not produce a very sensitive sensor, yet they were very successful in

development of a very compact hydrophone for under water applications. The

interferometers developed for sensing of acoustic waves are inherently different
from strain sensing interferometric sensors. The acoustic sensors are immune
to thermally induced signal fluctuations. This is due to the fact that the lower
frequency thermal signals can be easily differentiated from the acoustic signals
and removed at the frequency level. Consequently, a number of interferometric
signal modulation techniques were developed which removed the lower
frequency signals from
Xi Chen, Graduate Research Assistant, Smart sensors & NDT Laboratory, Civil &
Environmental Engineering, NJIT, Newark, NJ 07102.
Farhad Ansari, Professor, Smart sensors & NDT Laboratory, Civil & Environmental
Engineering, NJIT, Newark, NJ 07102.

Page 138

the acoustic signal during sensing. Andreyeva et al [5] developed an acoustic

MachZender interferometer based on the above-mentioned signal modulation
techniques. Liu et al. [6] applied the fiber optic acoustic sensor for detection of
flaws in Kevlar-epoxy composite materials. They worked in the frequency range
of 100 KHz to 1 MHz. Their research was very successful in detecting air voids
in the epoxy during the material processing stage of the composite production.
Dorighi et. al. [7] demonstrated the possibility of using a Fabry-Perot
interferometric sensor for detection of acoustic emissions at a frequency of 50
KHz, and ultrasounds at a frequency of 5 MHz. Glennie et. al. [8] was very
successful in developing a robust and very sensitive Michelson interferometer
for detection of ultrasound in a wide range of frequencies. Pierce et al. [9]
employed an interferometric sensor for adhesion and embeddment in carbon
fiber reinforced plastics. Their research was of fundamental nature, since they
demonstrated the superior sensitivity of optical fibers in detection of Lamb
waves.
Sensor Methodology
The basis for the methodology developed in this study is optical detection of
stress waves in concrete by way of an interferometric optical fiber sensor. The
objective is to generate stress waves within the concrete by means of a
transmitter or an impactor, and sense the reflected signals by an optical fiber
so as to detect internal flaws and anomalies. The proposed sensor is based on
the general principles pertaining to the Michelson interferometers. However,
optical sensing of acoustics requires stabilization of signal in order to lock out
the low frequency thermal and ambient strain levels from the interferometer.
This allows for the sensor to pick up the minute disturbances generated on its
periphery by the ultrasonic waves. The interaction between the optical fiber
and the incident ultrasonic wave leads to a change in the phase of light
traveling through the optical fiber (Fig. 1).

Figure 1.
Interaction of stress waves with the optical fiber.

Page 139

The combined effect of the stress wave on the length and the optical fiber's
index of refraction leads to the variation of optical phase in the fiber as:

and

Where Df is the phase change, n is the index of refraction of optical fiber core
and L is the sensing length of the optical fiber.
Stress waves induce a change in length as well as the refractive index in the
optical fiber. Most of the applications in acoustic sensors have dealt with
uniaxial strains in optical fibers, i.e. surface acoustic waves [10, 11, 12, 13,
14]. In the present configuration, stress waves are applied orthogonal with
respect to the axial direction of the fiber. Therefore, radial strains induce the
greatest change to the optical fiber signal. The PZT-ultrasonic pulse transmitter
for use with concrete sends out a pulse train with peak frequency at 54 KHz.
This is due to the fact that concrete is a heterogeneous material and low
frequency signals need to be used in order to avoid attenuation due to
scattering by aggregate particles within the material. In comparison with the
diameter of the optical fiber (0.125 mm), the wavelength of the transmitted
ultrasonic signal is very large ( 80 mm), and therefore the stress waves induce
a uniform radial strain field, e r (t), in the optical fiber. To compute the strain
induced phase change in the optical fiber we consider the change in refractive
index in the plane perpendicular to the optical fiber axis:

where, P11 and P12 are photo-elastic coefficients, e r , and e a are the radial and
axial strains. For fused silica fiber core, P11 =0.121 and P12 =0.27 [15]. In
terms of the Poisson's ratio ,u , equation (3) can be rewritten as:

Combining Equations (2) and (4) yields:

If the radial strain is explicitly expressed in terms of time, e r (t), then

combining Equations (1) and (5) yields:

Page 140

where, Cf is a constant and evaluated as:

which is related to the photo-elastic properties of the optical fiber, its gauge
length, and the wavelength of the light-wave. Equation (6) expresses a linear
relationship between the change in optical phase and the radial strain, e r (t), ,
in the fiber.
Sensor Configuration
The strain-field that must be sensed by the optical fiber due to the stress
waves is much weaker than the mechanical strains. The interferometers
employed for detection of such signals need to operate at their quadrature
point. The system should be able to pick up the dynamic events due to the
stress waves and eliminate the low frequency signals associated with thermal
and mechanical shifts in the strain field. A Michelson interferometer with a low
pass filter in a closed-loop configuration was developed for this purpose.
Michelson interferometer is especially very attractive in practical applications as
it would only require one-ended measurements of the signal. The sensing arm
of the interferometer can be designed to any length and embedded (or
adhered) in the structure. The various components of the sensor are described
in Figure 2.

Figure 2.
The fiber optic interferometer employed for detection of ultrasound.

Page 141

The system is comprised of a frequency stabilized laser source and the isolator
assembly. A fiber coupler is used for splitting the signal for the reference and
sensing arms of the interferometer. The low pass filter and the PZT tube are
employed in the closed-loop system in order to compensate for the low
frequency phase shifts. The system is automated by way of a DAQ interface to
the computer.
Flaw Detection Methodology
Conventional flaw detection techniques based on the stress wave propagation
methods are frequently used for condition monitoring of bridge elements. For
instance, impact-echo and pulse-echo methods are being used for thickness
measurements, and for detection of flaws and their depths. In these
techniques a stress pulse is interjected into the surface of the bridge deck with
a PZT transmitter or an impactor. Flaws and delaminations at interfaces reflect
the transmitted pulse. The reflected waves are either picked up by the same
PZT, which acted as the transmitter or a second PZT receiver, placed near the
transmitter. There are currently two distinct signal-processing procedures
available for interpretation of the reflected signals, namely based on time or
frequency domain analysis of data. In the time domain analysis, the round trip
travel time of the pulse is determined based on the speed of the stress wave in
concrete. This information is used to determine thickness of the slab or depth
of the flaw.
In the frequency analysis method, the multiple reflection of the stress wave
between the top surface of the concrete deck and the flaw is used for
detection as well as determination of depth. Depending on the boundary
conditions and geometry of the structural member an analysis is made and the
information about the flaw depth is acquired. For plate type geometry, the
effect of side boundaries can be neglected. The plate thickness and or the flaw
depth can be evaluated with the simple relationship among the depth, d, Pwave speed, Cp, and the frequency, f, of the resonance established by the
reflections of the impact pulse between the boundaries and discontinuities as
in the following:

For beams, the side boundary effect can not be neglected, and the frequency
pattern of impact pulse response becomes much more complicated. The

transient response of the beam to a stress wave is dominated by the vibration

modes of the cross section provided that the length of the beam is many times
larger than the cross-sectional dimensions. Accordingly, the effect of beam's
cross sectional dimensions on the fundamental frequency, D, need to be taken
into account. Based on the numerical results of reference [16] the relationship
between the aspect ratio D/B of the cross section and can be expressed as
follows:

Page 142

Where Cp is the P-wave speed in concrete, b is the frequency parameter

determined by the aspect ratio D/B, D is the depth, and B is the width of the
cross section. Reference [16] lists the frequency parameter, b for a variety of
beam cross section aspect ratios. For instance, for a beam with an aspect ratio
of 2, i.e. D=2B, b=0.96. For square beam, i.e. D/B=1, the frequency
parameter, b is 0.87. Therefore, for a beam with D=12'', B=6", the
fundamental frequency is:

Modes other than the fundamental frequency given in Equation (10) also
contribute to the waveform peaks in frequency domain. These modes were
numerically evaluated in reference [16]. Equation (8) is used for the
determination of flaw depth both in plate-like as well as in beam type
structures. Equation (8) is also employed for the determination of bridge deck
depth since it is a plate-like structure. However, equation (9) is used for
determination of fundamental frequency of the member taking into account
the boundary effects for beam type elements. This technique requires point-bypoint interrogation of deck surfaces. A grid of points is laid out on the surface
of bridge deck, and the grid spacing depends on the smallest defect that the
operator wishes to detect. In practice, the choice of grid spacing and sample
size is tempered with consideration of cost. In most cases, cost effectiveness
and time constraints do not allow point-by-point survey of the entire concrete
surface. In the study reported here, the point receiver (PZT) is replaced by an
optical fiber (Fig. 3). The optical fiber is either embedded during construction
of new members or adhered to the surface of the deck by epoxy glue. The
optical fiber can be distributed all along the element and this eliminates the
need for point-by-point movement of the receiver along the deck.

Figure 3.

Conventional pulse-echo and fiber optic distributed pulse-echo systems.

Page 143

Experimental Investigation
The feasibility of using a distributed fiber optic ultrasonic receiver in the pulseecho method for detection of flaws in concrete elements was examined
through the experimental program described here. A concrete beam was
constructed in which two plexi-glass plates were intentionally embedded in
order to simulate flaws. Fig.4 illustrates the test setup and specimen
dimensions.

Figure 4.
Concrete beam specimen dimensions, location of intentional flaws, test setup
and the surface adhered optical fiber sensor.

Page 144

The objective was to examine the flaw detection capability of the fiber optic
sensor in the laboratory in a controlled experiment. In frequency domain, the
optical fiber should detect a peak corresponding to the flaw depth of 8-inch
(0.203). As shown in Figure 4, these flaws were embedded within different
sections of the beam (referred to positions 1 and 3). Therefore, the depth and
positions of these flaws were known apriori. Position 2 refers to flaw-free
positions along the beam. An optical fiber was glued to the surface of the
beam, and it was connected to the Michelson interferometer as the sensing
arm of the system. A 54 KHz ultrasonic transmitter was employed for
generation of stress waves in concrete. As a rule, the maximum size of
aggregate in concrete should be smaller than the wavelength of the
transmitter, otherwise the wave energy will attenuate at the aggregate
interfaces. In most concretes, nominal aggregate sizes of 1 to 2 inches (2.5 to
5.0 cm) are common. For that reason low frequency transmitters (54 KHz) are
primarily used in applications involving concrete.
Two PZT transducers were employed for measuring the P-wave speed in
concrete. A P-wave speed of 4234 m/s was recorded and used in subsequent
computations. Equation (8) was used to predict the resonant frequency of the
stress wave reflections between the top surface and the flaw depth of 8 inch
(203.2 mm):
During the experimentation, the PZT transmitter was moved along the beam
and it was used for inducing stress waves at positions 1, 2, and 3 respectively
(Figure 4). A typical time-domain response of the optical fiber sensor is shown
in Figure 5. Frequency spectra for positions 1 through 3 are given in Figure 6.

Figure 5.
Time Domain Response of the fiber optic sensor to the stress wave.

Page 145

Figure 6.
Frequency spectra of the reflected waves at positions 1, 2, and 3 along the beam.

As shown in Figure 7, for positions 1 and 3 the maximum amplitude of the

signal occurred at 10.87 and 10.18 KHz corresponding to flaw depths of 194.8
and 208 mm respectively. These measurements show good agreement with
the actual flaw depth of 204 mm. Other peaks in the spectrum correspond to
the multiple reflections at the boundaries of the specimen. Depending on the
specimen geometry, and flaw size, the frequency peaks due to the specimen
boundary are of lower amplitude than the one from

Page 146

the flaw. In the present case the boundary peaks are an order of magnitude
lower in amplitude than the flaw peak.
Frequency peaks representing the fundamental modes (1st, 2nd,..) of beam
vibrations are also present in the spectrum. For instance, at position 2, the tru
fundamental frequency of the beam is represented by the peak at D=6.67.
This was also established through finite element analysis in reference [16].
Existence of the flaw shifts the fundamental frequency. For the experiments
described here, the fundamental modes of the beam at positions 1 and 3 were
shifted to 6.53, and 6.59 KHz respectively. The shift in frequency also occurs
for higher modes of beam vibrations. The 12.07 and 12.02 KHz peaks in Figure
7 correspond to the second mode of beam vibrations for positions 1 and 3
respectively.
Conclusions
Development and testing of a fiber optic ultrasonic sensor for detection of flaws
and delaminations in concrete elements has been described. The experimental
program pertained to testing of a beam with embedded flaws. A PZT source
created the stress waves in the beam element. The optical fiber sensor was
adhered to the surface of the beam and it was used as an ultrasonic receiver.
The objective was to demonstrate the effectiveness of the optical fiber system
in detection of flaws and anomalies in concrete structures. It was
demonstrated that it is possible to detect the locations of the embedded flaws
through scanning of the beam in the frequency domain. The system is
convenient for use in practical applications as the fiber optic sensor is
distributed over the length of the structure, and therefore there is no need for
point-by-point movement of the sensor.
Acknowledgements
This work was made possible by a grant from the National Science Foundation,
grant number CMS-9402671, to the New Jersey Institute of Technology.
References
1. Y. Lin, M. Sansalone and N.J. Carino, 1990. "Finite Element Studies of the

Impact-Echo Response of Plates Containing Thin Layers and Voids", Journal of

Nondestructive Evaluation, 9 (1), pp. 27.
2. J.A. Lee, S.C., Warfield and J.C. Duke, 1990, "Condition of Asphalt Covered

Bridge Decks Using Impact Echo," Transportation Research Board, SHRP

report, SHRP-S/IR-90-002, Washington, D.C.
3. Transportation Research Board, 1991. "Condition Evaluation of Concrete

Bridges Relative to Reinforcement Corrosion, Vol. 1, National Research Council.

4. L.M. Lyamshev,, Y.Y. Smirnov 1989. "Fiber Optic Sound Level meter," Sov.

Phys. Acoust. 36 (6), pp. 632635.

Page 147
5. Y.I., Andreyeva, N.Y.,Smirnova, 1991. "An Interference Fiber Optic Sound

Sensor," Radiotekhnida, 11, pp.. 9194.

6. K. Liu, S.M. Ferguson, R. Measures 1990. "Fiber Optic Interferometric

Sensor for the Detection of Acoustic Emission Within Composite Materials,"

Optics Letters, 15 (2), pp. 12551257.
7. J. F. Dorighi, S. Krishnaswamy, and J. D. Achenbach, 1995. "Stabilization of

an Embedded Fiber Optic Fabry-Perot Sensor for Ultrasound Detection", IEEE

Transactions on Ultrasonics, Feeroelectrics and Frequency Control, 42 (5),
pp820824.
8. D. Glennie, T. Alavie, K. Liu and R. Measures, 1993. "The applicability of

fiber optic sensors to the detection of surface acoustic waves on metals"., SPIE
Vol.1918 Smart Sensing, Processing and Instrumentation, pp97109.
9. S. G. Pierce, W. R. Philp, A. Gachagan, A. McNab, G. Hayward, and B.

Culshaw, 1996. "Surface-bonded and Embedded Optical Fibers as Ultrasonic

Sensors", Applied Optics, 35 (25), pp. 51915197.
10. T.G. Giallorenzi, J. A. Bucaro, A. D Dandridge, 1982. "Optical Fiber Sensor

Technology", IEEE J. of Quantum Electronics, 18(4), pp626664.

11. R. Hughes and J. Jarzynski, 1980. "Static Pressure Sensitivity Amplification

in Interferometric Fiber-Optic Hydrophones", Applied Optics, 19(1), pp. 98.

12. R. P. Depaula, L. Flax, J.H. Cole and J. A. Bucaro, 1982. "Single Mode Fiber

Ultrasonic Sensor", IEEE J. of Quantum Electronics, QE-18 (4), pp.680.

13. N. Lagakos, T.R. Hickman, P. Ehernfeuchter, J. A. Bucaro, and A

Dandridge, 1990. "Planar Flexible Fiber-Optic Acoustic Sensors," Journal of

Lightwave Technology, 8 (9), pp. 1298.
14. K. A. Murphy, M. F. Gunther, R. O. Claus, T. A. Tran and M. S. Miller, 1993.

"Optical Fiber Sensors for Measurement of Strain and Acoustic Waves", Smart
Sensing, Processing, and Instrumentation, SPIE Vol.1918, pp.110.
15. A. Yariv and P. Yeh, 1984. "Optical Waves in Crystals," John Wiley&Sons,

pp.318.
16. Y. Lin and M. Sansalone, 1992. "Detecting Flaws in Concrete Beams and

Columns Using the Impact-Echo Method", ACI Material Journal, 89 (4),

pp394405.

Page 148

Preliminary Results on the Monitoring of an In-Service Bridge Using a 32Channel Fiber Bragg Grating Sensor System
S. T. Vohra,
C. C. Chang,
B. A. Danver,
B. Althouse,
M. A. Davis
and R. Idriss
Abstract
Design and performance of a 32-channel fiber Bragg grating (FBG) system
used to monitor the dynamic response of an in-service interstate bridge1 are
reported. Fiber Bragg grating sensors were attached on four different
supporting girders in groups of three at various locations along the span of the
bridge. Using an interrogation approach based on the scanning Fabry-Perot
system, the FBG sensors were monitored for various vehicle loading conditions.
The response of the bridge to a typical loading is described both in time and
frequency domains. The preliminary results of the field test reported here
clearly demonstrate that an optical FBG sensor system is well suited for
monitoring both, weigh-in-motion traffic events as well as the dynamic
characteristics of bridge structures.
Introduction
Fiber optic sensors have been proposed for use in numerous field applications,
such as evaluation of strains and stresses in structures. Fiber optic sensors
offer a series of unique advantages over their conventional electronic
counterparts including electromagnetic immunity, light weight, small size, low
transmission loss, and resistance to corrosion. Fiber Bragg grating (FBG)
sensors are particularly attractive for distributed structural sensing since they
offer wavelength encoded operation. FBGs are wavelength-encoded in that the
strain information obtained from the device is contained within the reflected
wavelength components. Another major advantage of wavelength encoded
operation is that it simplifies the process of multiplexing numbers of FBG
sensors on a single strand of optical fiber. The multiplexed devices can be
written at different initial spectral locations that can be correlated to the
physical location of each device along the fiber. Any wavelength shift of a given

grating can be attributed to the localized

S. T. Vohra, C. C. Chang, B. A. Danver and B. Althouse, Naval Research Laboratory,
Code 5670, Washington DC 20375
M. A. Davis, Cidra Corporation, Wallingford, CT 06492.
1 North-bound span, I-10 at University Blvd., Las Cruces, New Mexico.

Page 149

strain at that particular location and a quasi-distributed strain sensor system

can be created with a minimum of optical fibers. Much of the recent work in
the field has focused on the development of a variety of wavelength detection
techniques which provide the capability to demodulate and multiplex several
FBG strain sensors using either the inherent wavelength division addressing
capabilities of FBGs or by a using a hybrid wavelength-time division
multiplexing approach [1].
This report describes the results of a field test carried out using a wavelength
division multiplexed array of 32 FBGs attached at various locations along the
girders of an in-service bridge in New Mexico. The FBG responses are
interrogated using a method based on the scanning fiber Fabry-Perot (FFP)
filter [2]. Preliminary data obtained from the 32 FBGs indicates that the
magnitude of the dynamic strains induced in the bridge by moving vehicles is
under 1000 me and the frequencies excited in the structure are well below 20
Hz.
FBG Interrogation System
One of the more successful techniques for interrogating FBG sensors involves
the use of tunable passband Fabry-Perot (FP) filter for tracking the FBG signal
[2]. In this approach, light reflected back from an array of Bragg grating
sensors is passed through a FP filter which passes one narrowband
wavelength, depending on the spacing between the mirrors in the device. The
narrowband wavelength signal is passed on to the photodetector and
electronically differentiated. This produces a series of pulses for each of the
signal reflected by the FBGs with the zero-crossing of each pulse representing
the peak reflected wavelength from each sensor. The zero-crossings are
recorded by a computer along with the digital value of the ramp voltage
applied to the FP system. By referencing any shifts in the zero-crossings from
each FBG to the voltage applied to the FP filter, the strain in each FBG sensor
can be obtained. The schematic of the electo-optical system used in this study
is shown in Figure 1. The free spectral range (FSR) and the resolution
bandwidth along with the spectral spacing between the FBGs determine the
number of FBG sensors which can be interrogated per one scan of the FP filter.
The choice of 16 bit resolution for the FP ramp, and a FSR of 45 nm produces
a minimum detectable wavelength shift of 0.7 pm, or equivalent strain
resolution of less than 1 me. However, the actual minimum detectable strain is

on the order of ~10 me due to electronic noise in the system. For the field test
reported in this work, a scanning FP system similar to the one just described
was used to interrogate 32 FBG sensors.
As shown in Figure 1, light from two ELEDs is directed through a pair of 22
couplers connected to two FBG sensor arrays. Each FBG array is made up of
sixteen FBG sensors. The reflected light from the FBG arrays is combined with
a third coupler and directed to a single FP filter. The fiber Fabry-Perot filter
used in this work had an FSR of about 45 nm, thus allowing 16 individual
sensors with a bandwidth of ~0.2 nm, spaced by approximately 2.7 nm to be
interrogated per filter scan. This spacing is sufficient to allow strains of about
1300 me for each FBG to

Page 150

Figure 1
The schematic of the FP instrumentation system to interrogate the FBG array

be easily monitored. The FFP filter is capable of achieving scan rates of about
360 Hz and supporting 64 FBGs but in this work the system was set to operate
at a sampling rate of 45 Hz. This was driven by two reasons: (1) only the up
scan was used to prevent the hystersis difference of voltage-to-wavelength
tuning response between the up and down scan, and (2) the FBG sensor
readings were averaged 4 times to provide better resolution. The ELEDs' on/off
was synchronized with the ramp of the filter drive voltage such that only the
FBG array being interrogated by the FP filter was lit at any given time. It
should be pointed out that in monitoring large scale structures, static and
quasi-static strains (< 10 Hz) are of prime interest and in the particular case of
the structure being reported here, sampling rate of 45 Hz was considered
adequate. However, by using different FP filters capable of providing higher
sampling rate or by modifying the approach utilized here, it should be possible
to achieve significantly enhanced sampling rates ( 1 kHz).

Page 151

Spectral and Physical Arrangement of FBG Sensors

The spectral characteristic of the ELED output has a Gaussian-like distribution
[3]. Therefore, in order to maximize the signal-to-noise (SNR) ratio at each FBG
wavelength the reflectivities of the FBGs must be tailored accordingly. In other
words, it would be prudent to have FBGs with high reflectivities have their
Bragg wavelength fall near the wings of the ELED spectrum while FBGs with
lower reflectivities should have their Bragg wavelength fall around the center of
the ELED spectrum. This would ensure, approximately, equal amount of light
being reflected back from all the FBG sensors, thus resulting in reduced
variation in the SNR among various FBG sensors. The FBG sensors used on the
bridge had tailored reflectivities to achieve approximately equal light levels from
all the sensors.
The I-10 bridge over Las Cruces, New Mexico, consists of spans made up of a
concrete deck supported by welded steel plate girders (Figure 2). Each span is
about 36 meters long and FBG sensors were attached at the pier 1/8-span and
1/2-span of girder 1, girder 2, girder 3 and girder 4.

Figure 2
A photograph showing continuous girders along the span of the I-10
bridge. A small portion of the pier is shown in the upper left corner for
perspective. The girders were (arbitrarily) labeled in ascending order
starting from left to right. FBG sensors were placed on girders 1 to 4 at
the pier, 1/8-span and 1/2-span.

Page 152

Figure 3
A schematic of the arrangement of FBG sensors on the girders. Sensors
were attached at 1/8-span, 1-2 span and the pier. FBG sensors were only
placed on girders 1 to 4.

A sketch indicating the sensor locations on the girders is shown in Figure 3.

Three FBGs, ranging from lowest to highest wavelength in a given sub-array,
are surface mounted on the top flange, upper web and bottom flange of the
girder, as shown in Figure 4. In each FBG location, a small area of the girder
was filed smooth and prepared by following the installation tips for the
conventional resistance strain gage and the FBG were bonded using M-Bond
200 adhesive. For the first FBG array (also know as the 1/2-span array), the
temperature reference sensor, T, is installed inside of the FBG instrumentation
(FFP system) located on one of the piers and is used to temperaturecompensate the rest of 31 FBGs.
Preliminary Results of I-10 Bridge Monitoring
The fiber Fabry-Perot (FFP) system for interrogating the arrays was located on
one of the bridge piers and is designed to have a sampling rate of 45 Hz. The
data from the FBG arrays is stored onto a computer hard disk using a threshold
trigger method which allows storage of data only if a certain strain threshold
trigger method which allows storage of data only if a certain strain threshold is
crossed. In the measurements described here, the data collection software was
set

Page 153

Figure 4
Local arrangement of the FBG sensors at each
girder location. Sensors were placed
on the top flange, the web and bottom
flange of the girder.

to store data only if a 20 me loading event is experienced at the FBG sensor

located on the bottom flange of girder 3. This approach allows judicious
storage of data on the computer hard drive and also ensures that only data
with significant loading events are stored. It should be pointed out that the
FFP system is continuously monitoring the arrays but storage of the data
occurs only if the strain threshold is crossed. Each data file includes 32
columns of 1350 data points (equivalent to data recorded in 30 seconds at 45
Hz sampling rate) corresponding to each FBG on the bridge. The program
assigns a 225-points pre-trigger length (equivalent to 5 seconds).
The response of FBG sensors located at half-span, girder 3 of the bridge to a
vehicle load event is shown in Figure 5. Data from FBG sensors located on the
top flange, the web and the bottom flange are depicted in Figure 5. The data
clearly shows that the during a 30-seconds duration, a large strain event (e.g.
a truck) followed by a smaller strain event (e.g. a car) takes place on the
bridge. All three FBG sensors located on girder 3 recorded the data as
expected. The truck, seen as the

Page 154

Figure 5
Responses due to transient loading from FBG sensors located at
(a) the top Flange, (b) the web and (c) the bottom Flange of the
half-span, Girder 3.

Page 155

large peak, produces a peak negative strain of about 40 me at the top flange
location, peak negative strain of about 100 me at the web location and peak
positive strain of about 80 me at the bottom flange location. Modal excitations
of the bridge induced by the transient loading event can be seen (Figure 5c).
A closer look at the data reveals that the strain response at all three locations
changes sign between the time period of 3 to 7 seconds. For instance, in
Figure 5c (bottom flange FBG sensor), negative strain of about 20 me is
observed in the FBG sensor followed by a sharp rise in strain towards the
positive direction, eventually peaking at about 80 me. This can be explained as
follows. The I-10 bridge girder is a continuous structure which undergoes
compressive strain at the bottom flange prior to the truck reaching the sensor
location (half-span, girder 3). The truck travelling at about 65 mph then
reaches the location directly above half-span girder 3, thus inducing a bend in
the beam due to direct loading which results in positive strain at the bottom
flange of the I-beam. This is followed by transient load (i.e. truck) induced
modal excitations in the bridge, which can be seen as ringing in the FBG
sensor response (Figure 5c). A similar argument explains the top flange and
web sensor responses.
Since the top flange sensor is on the opposite side of the neutral axis of the Ibeam with respect to the bottom flange sensor, the sign of measured strain
between the top and bottom flange is reversed for the same loading event at
that location. The web sensor is located significantly closer to the top flange of
the I-beam. The web sensor shows large compressive strain signal during the
actual transit of the truck directly over the sensor location. This is preceded by
a slight positive strain at the web sensor for the same reasons as described
earlier. The second event in Figure 5, near the 23-seconds mark, can be
attributed to a smaller vehicle, a car, passing over that location producing a
peak strain of about 25 me. The bottom flange undergoes tension and the top
flange (above the neutral axis) undergoes compression, resulting in positive
and negative strains at those locations due to the vehicle. The car does not
induce noticeable bridge vibrations. In addition, the results of Figure 5 showing
no slow drifts of the rest FBG temperature reference sensor compensates the
thermal drifts of the rest FBG sensors. Therefore, the system is capable of
performing static or quasi-static measurements.
Power spectral density calculations were performed on the time series data

generated from the FBG sensors installed on the bridge. The times series data
shown in Figure 5 was used to estimate the frequency content. Since the
induced bridge vibrations are significantly smaller in magnitude than the
measured strain due to a vehicle passing directly above sensor location, the
spectra were computed in two parts. Figure 6a shows the power spectra of the
first eight seconds of the time series obtained from the three FBG sensors (top
flange, bottom flange and web) located at girder 3, half-span. The spectra of
Figure 6a clearly show a large and broad strain peak well in the sub-Hertz
regime. Considering the transient nature of the event other analytical
techniques (e.g. wavelet transforms) may be

Page 156

Figure 6
Power spectral density (in arbitrary amplitude units) of the top flange,
bottom flange and the web FBG sensors located at girder 3, half-span (see
Figure 5 for corresponding time series). The top part of the figure, (a)
shows the power spectra of the data corresponding to the first 8 seconds of
the time series while (b) corresponds to the power spectra of the time
series from 8 to 30 seconds. The load induced vibrations at 2.6 Hz, 3.7 Hz,
5 Hz and 8.7 Hz are clearly observed in (b).

Page 157

better suited to analyze such events in the time series. However, the power
spectrum density obtained from the data for time periods beyond 8 seconds
clearly show distinct peaks at 2.6 Hz and beyond (up to 8.7 Hz). The peaks
observed in Figure 6b are due to various vibrational modes being excited in the
bridge due to the transient loading event (i.e. due to the heavy truck passing
over the bridge). No significant vibrations with frequencies above 10 Hz were
observed during the period of observation reported here. Further analysis of
the data is on-going and will be reported in the future.
An interesting observation from the power spectrum density shown in Figure
6b is that the tonal vibration near 2.6 Hz was also observed in sensors located
at girders other than at half-span. In contrast, the higher order tonals (e.g. at
frequencies near 5 and 8.7 Hz) observed at half-span, girder 3 were not
observed at all or were very weakly observed in FBG sensors located on other
girders. For instance, Figure 7 shows power spectra of the bottom flange
sensor located at half-span, girder 1 and girder 4. The peak at 2.6 Hz is clearly
observed in both sensors, indicating that the 2.6 Hz vibration measured in the
bridge must represent some sort of fundamental mode of the given bridge
structure.
Data from all the FBG sensors located in a spatially distributed manner should
allow us to determine which lanes of the bridge are most often being used, the
speed of the vehicle, the load being carried by the vehicle, while the vibrational
spectra obtained from the sensors should allow for determination of the modal
excitations being induced in the bridge. Data from the bridge is being
continuously collected and data analysis on the large data sets being collected
will be reported in the future.

Figure 7
Power spectral density (in arbitrary amplitude units) of the bottom flange
FBG sensor located at half-span girder 1 and girder 4. The data clearly

shows that both sensors measure the vibration at 2.6 Hz, while some of the
higher order vibrations at individual locations tend to vary (see also Figure
6b).

Page 158

Conclusions
We have presented strain measurements performed on an in-operation bridge
(I-10, New Mexico) using fiber Bragg grating sensors. The 32-channel system
showed wide range of peak strains beings induced at various locations on the
girders of the bridge structure. These measurements show that it is possible to
not only distinguish between heavy (e.g. trucks) and light (e.g. cars) transient
loading events occurring on the bridge but it is also possible to determine the
induced bridge vibrations. The measurements also imply that more advanced
data processing techniques (e.g. wavelet analysis or wave propagation
techniques) should be applied to the data obtained from spatially distributed
arrays of sensors. The test conducted so far demonstrate that FBG strain
sensing systems are well suited for infra-structure health monitoring.
Acknowledgments
The support of the Federal Highway Administration is gratefully acknowledged.
References
1. A.D. Kersey, M.A. Davis, H.J. Patrick, M. LeBlanc, K.P. Koo, C.G. Askins,
M.A. Putnam and E.J. Frieble, J. Lightwave Technol., Vol. 15, p. 1442 (1997).
2. A.D. Kersey et.al., Optics Letters, Vol. 18, p. 1370 (1993).

Page 159

Monitoring and Evaluation of an Interstate Highway Bridge Using a Network of

Optical Fiber Sensors
R. L. Idriss,
K. R. White,
J. W. Pate,
S. T. Vohra,
C. C. Chang,
B. A. Danver
and M. A. Davis
Abstract
A multiplexed Bragg grating optical fiber monitoring system is designed and
retrofitted to an in-service interstate highway bridge. The test bridge, the I-10
over University in Las Cruces New Mexico, is a two span continuous multi-girder
steel bridge. A network of optical fiber Bragg grating sensors is used to monitor
the strain at several locations in the bridge, with sensors bonded to the web
and flanges of the girders. Preliminary data obtained from the system is
evaluated.
Introduction
Highway agencies are struggling to cope with the increasing demands on their
highways, and deteriorating bridges are becoming more severe choke points in
the system.
The current state of bridge decks and bridge-related structures in the United
States is a topic of considerable concern due to (a) the deterioration of critical
structural elements, making them unfit for current use; (b) the drastic need for
retrofit of existing structures to bring them up to code; and (c) the need to
increase the load-carrying capacity of damaged or already weakened structures
due to increased traffic demands.
1R. L. Idriss, Associate Professor of Civil Engineering, New Mexico State University,
Box 3CE, Las Cruces, NM 88003.
K. R. White, Professor of Civil Engineering, New Mexico State University, Box 3CE, Las
Cruces NM, 88003.
J. W. Pate Mechanical Systems; Physical Science Laboratory, New Mexico State
University, Box 30002, Las Cruces, NM, 88003.

S.T. Vohra, C.C. Chang and B.A. Danver, Optical Sciences Division. Code 5673. Naval
Research Laboratory, Washington, DC 20375
M. A. Davis, Cidra Corporation, Wallingford, CT 06492

Page 160

In December 1967, there was the tragic collapse of the Silver Bridge between
Point Pleasant, W. Va., and Gallipolis, Ohio, with the loss of 46 lives. A few
more names from the recent past -- Mianus River Bridge, Hatchie River Bridge,
Schoharie Creek Bridge, and the San Francisco-Oakland Bay Bridge --bring to
mind a chain of events: a spectacular and tragic collapse, extensive media
coverage, congressional inquiry and detailed investigation, and a demand for
an accelerated search for answers. In many of these cases, the problems were
foreseen. In some of these cases, improved inspection technology would have
allowed the problems to be spotted in time and repairs implemented
preventing catastrophic failure.
Visual inspection approximately once every two years is still the primary
technique used for routine inspection of bridges. Conventional non destructive
evaluation (NDE) techniques are only used as an adjunct to visual inspection
and very few advanced NDE techniques are used at all. In order to optimize
safe operation of civil structures with a minimum of expense it is necessary to
develop innovative nondestructive testing systems for remote self-monitoring of
highway bridges using state-of-the-art sensors, telemetry, recorders, and
analyzers.
The objective of this multi-disciplinary research project is the development and
implementation of a bridge monitoring and evaluation system. An installed
system can be designed as a versatile tool, to supplement visual bridge
inspection and provide data that can be used to:
Assess the traffic loading on the bridge
Define a baseline behavior for the structure
Evaluate the effects of damage on the capacity and performance
Assess the effectiveness of repairs and maintenance programs.
Check the performance compared to the design assumption
Remotely monitor critical structures, and provide a warning when abnormal
conditions occur.
The data generated can be directly incorporated in the inspection file and into
life cycle management of the structure.
Research Approach

The multi-disciplinary research study is an on-going testing program,

implemented in several phases [1,2,3]. In the current field testing phase of the
project, an optical fiber monitoring system is designed and retrofitted to an inservice interstate highway bridge. A network of 32 sensors is used to measure
the strain at several location in the bridge, with sensors bonded to the top,
bottom flange and web of the girders.
The main objectives of this on-going study are:
Further development and field testing of optical fiber sensors and
instrumentation

Page 161

Data evaluation: traffic monitoring, developing a base line behavior for the
structure, and fatigue evaluation of the bridge.
Bridge Description
The structure is a two span continuous, non composite steel girder bridge (Fig
1, 2 &3). The bridge has a 27 skew. During inspection several fatigue cracks
were discovered at the top and bottom of the web, near the stiffener weld, at
diaphragm to girder connections. These cracks can be attributed to the out of
plane bending stresses in the web. This problem is amplified by the skew
present in the bridge.

Figure 1:
Longitudinal Profile of the Bridge

Figure 2:
Bridge Cross-Section

Page 162

Figure 3:
Overall view of the I-10 bridge.

Fiber Bragg Grating Sensors (FBG)

The Fiber Bragg Grating (FBG) sensor is a segment of an optical fiber that has
been internally modified by exposure to two interfering beams of ultraviolet
light. This creates a small periodic modulation of the refractive index of the
fiber core. Due to the periodic nature of the index perturbation only one optical
frequency will resonate in the structure. The center wavelength of this
resonance condition in a FBG can be expressed as:

where neff is the effective index of refraction of the core and L is the period of
the refractive index modulation. As can be seen by Eq. 1, any change in the
periodicity of the refractive index modulation or the overall index of refraction
will change the Bragg wavelength. Consequently, any temperature or strain
induced effects on the FBG can be determined by the corresponding shift in
the center Bragg wavelength, as illustrated in Figure 4.

Page 163

Figure 4:
Operation of a FBG element as a wavelength-encoded strain sensor.

Bragg gratings can be multiplexed via their inherent wavelength encoded

properties by writing several gratings in-line in a fiber at different nominal
Bragg wavelengths.
The FBG instrumentation used on the I-10 bridge has the capability of
interrogating up to 32 fiber Bragg grating (FBG) sensors at 360 Hz sampling
rate. The minimum detectable strain is in the order of ~10 mstrains with
consideration of noise from the electronic system.
Sensor Layout
Thirty-two optical fiber Bragg grating sensors are used to monitor the bridge.
The sensors were placed on girders G1 through G4. To evaluate the strain
induced by out of plane bending at stiffener ends in the web, sensors were
bonded to the top of the web. Sensors were also attached to the top and
bottom flanges of the girders. Three sections were monitored: the pier, which
is the location of large negative moment, the mid-span, location of large
positive bending moment, as well as the 1/8 span section. A temperature
reference sensor was installed inside of the FBG instrumentation. A second
temperature reference sensor was coiled and attached to the bridge. The
location of the sensors is shown in fig. 5,6,&7.

Page 164

Figure 6:
Schematic of FBG Sensor Installed Locations

Page 165

Figures7 (a) & (b):

Sensor placement at the diaphragm to girder connection.

Page 166

Data Collection
The system was set up to collect data for the passage of every truck. The
bottom flange sensor for girder 3 (G3) at midspan was used as a trigger, with
a threshold of 20 mstrains. For every event, data was collected with a 5
seconds pre-trigger, and 25 seconds post-trigger time.
Sample Data
Data from a sample file is shown in figures 8. In fig.8, the response of the four
girders monitored is shown at midspan during the passage of a truck. The
girder bottom flanges show the same pattern of behavior: induced compression
when the truck is still in the first span at around 5sec followed by peak tension
when the truck passes over midspan at about 7.5 sec on the time scale.
From the data collected detailed information can be generated pertaining to
the structure itself as well as statistical information on the traffic loading on the
bridge. Information such as truck speed, as well as statistical truck loading
weight can be generated. Load patterns and load distribution in the bridge, as
well as fatigue life predictions can be assessed based on the strain data
obtained from the monitoring system.

Figure. 8:
Midspan Bottom Flange Strain response of girders
G1, G2, G3 & G4

Page 167

Conclusions
A bridge monitoring system using a network of 32 FBG sensors has been
successfully retrofitted to an in-service Interstate highway bridge. Preliminary
data collected from the I-10 bridge shows that such a monitoring system can
prove to be a very effective tool for the bridge engineer. Traffic information
such as vehicle weight and distribution, as well as vehicle speed can be
extracted from the strain signatures. Furthermore, a statistical pattern of strain
response for the bridge can be developed. Changes from that baseline
behavior can be monitored and interpreted as possible changes or defects in
the structure. Ultimately, the system can be used to remotely monitor critical
structures, giving the bridge engineer a warning signal if abnormal conditions
should occur. In summary, the monitoring system can provide a profile of the
bridge, with detailed, quantitative information on it's structural behavior, as
well as the vehicles crossing it.
Acknowledgment
This project is a collaborative project between New Mexico State University and
the Naval Research Laboratory. It is supported by the National Science
Foundation grant CMS-9457604, the New Mexico State Highway and
Transportation Department and the Federal Highway Administration.
References
(1) Idriss, R. L. and Kodindouma, M.B '' Design And Implementation of an
Optical Fiber Monitoring System In A full Scale Laboratory bridge" New Mexico
State University, Department of Civil Engineering, Report No. 97-02 to the
Federal Highway Administration and the National Science Foundation. 1997.
(2) Kodindouma, M.B., Idriss, R. L., Kersey, A. D., Davis, M. A., D.G.
Bellemore, Friebele, E.J. and Putman, M.A., "Damage Assessment of a Full
Scale Bridge Using an Optical Fiber Monitoring System," SPIE Symposium on
Smart Structures and Materials, Smart Systems for Bridges Structures and
Highways, 2829 February 1996, San Diego Ca, pp 2719-32.
(3) Davis, M.A., Bellemore, D.G., Berkoff T.A., and Kersey A.D. "Design and
Performance of a Fiber Bragg Grating Distributed Strain Sensor System," SPIE
Symposium on Smart Structures and Materials, March 1995, San Diego Ca,
pp.227235.

Page 168

Distributed Multiaxis Fiber Grating Strain Sensor Applications for Bridges

E. Udd,
W. Schulz,
J. Seim,
G. McGill
and H. M. Laylor
Abstract
Fiber optic gratings have been widely used to measure longitudinal strain and
temperature. By writing fiber gratings onto polarization maintaining optical
fiber with substantial birefringence, transverse strain measurements may be
made. For bridges and other civil structures, this opens up a series of new
opportunities where multiaxis strain sensing may be effectively used for health
monitoring systems.
Introduction
Many of the strain measurements made today are performed using electrical
strain gauges. These devices cost about $20 each and typically have gauge
lengths of 0.5 to 1 cm. The vast majority of these devices are surface mounted
and the procedure for mounting them involves preparing a flat surface, often
with sanding, followed by cleaning, an acid wash, gluing the strain gauge to
the surface and attaching electrical leads. The process with a skilled technician
takes about an hour per gauge, adding perhaps $25 to $30 to the cost of each
sensor. The electrical strain gauges are then attached to electronic
demodulators such as a Wheatstone bridge, which costs anywhere from $300
to over $2000 per channel, depending on the performance and features of the
unit. Issues associated with conventional strain gauges include (1) the strain
gauges falling off or partially detaching from the surface on which they are
mounted, (2) limited temperature ranges, (3) susceptibility to electromagnetic
interference, (4) difficulties associated with embedding them into composite
and metallic materials and (5) Multiplexing difficulties.
Eric Udd, Blue Road Research, 2555 N.E. 205th Avenue, Fairview, OR 97024
Whitten Schulz, 2555 N.E. 205th Avenue, Fairview, OR 97024
John Seim, 2555 N.E. 205th Avenue, Fairview, OR 97024
Galen McGill, Oregon Department of Transportation, Research Unit,
2950 State Street, Salem, OR 97310

H. Martin Laylor, 2950 State Street, Salem, OR 97310

Page 169

Currently fiber optic grating based strain sensors cost from $150 to $500 each
in small quantities and are supported by fiber grating based demodulators
costing about $20,000 that can support 10 to 60 longitudinal strain and
temperature measurements. The price of the fiber gratings is substantially
lower in quantities of 100, on the order of $100 to $200 per item and as
companies start to move toward mass production of these items, their cost
should rapidly approach the $20 target offered by electrical foil gauges. The
reasons for these improvements include advances in preparing the fiber using
such techniques as hydrogen loading as well as modifications to the fiber
design. Strides have also been made in improving the quality of fiber gratings
that are actually made as the fiber is being drawn. While additional work
remains to be done to improve the reflectivity of fiber gratings made in this
manner, there is the real potential of mass-producing fiber gratings while the
fiber is being drawn.
While rapid progress is being made in improving fiber grating technology for
the telecommunication industry, the introduction of fiber grating sensors into
structures is likely to face significant competition from entrenched electrical
strain gauge manufacturers in the near future. What is needed is additional
differentiation from electrical strain gauge capabilities. Transverse strain and
multiaxis single point strain capability are such differentiating items that can be
realized for small cost premiums associated with polarization preserving fiber.
Longitudinal Strain and Temperature
Since the period of the fiber grating is written along its length, environmental
effects that cause elongation of the fiber will change the period and the
wavelength band that the fiber grating reflects and transmits. Two of the
principle environmental effects that cause this change are temperature and
longitudinal strain [1]. Other types of environmental effects can be measured
such as magnetic field or pressure by designing coatings or transducers that
will convert these effects into strain on the fiber.
The usual problem is to separate strain and temperature so that strain can be
accurately measured. One approach [1] to this problem is to put two fiber
gratings in close proximity to one another with one attached to the material on
which strain is to be measured and the other nearby but not attached to the
material so that it floats and is not subject to strain. Another method is to use
fiber gratings in combination with another fiber sensor that measures

temperature only. The Raman scattering based optical time domain

reflectometer systems offered by York and Hitachi offer distributed temperature
sensing along the length of the fiber [2]. The main issue associated with both
of these approaches is that the temperature of the fiber grating making the
strain measurement may be substantially different from that of the spatially
displaced "floating" fiber grating or the temperature indicated by the Raman
scattering based temperature sensor which may have spatial resolution on the
order of a meter or more. To overcome these problems an alternative is to
write two fiber gratings of widely separated wavelengths on top of one another
[3]. By measuring the change in wavelength of both of these gratings

Page 170

with strain and temperature, one can establish two equations in two unknowns
and solve for both quantities [4].
The current approach to writing dual overlaid fiber gratings is to use a phase
mask with two parts, one for each wavelength. Once the fiber is positioned,
first one wavelength is written, then the other. It may be possible to write two
gratings on the fiber with one exposure, reducing the intrinsic cost to that of a
single fiber grating. Initial efforts indicate this may not result in as clean a fiber
grating profile as the individually written fiber grating patterns. Further work
remains on this problem.
To date, the reflectivity of the dual overlaid fiber gratings has been about 80%,
a value chosen for good overall performance and reduction of sidelobes to low
levels.
The intrinsic strength of these dual gratings is approximately the same as a
single fiber grating which 3M Fiber Grating Technologies has demonstrated can
be the same as the original strength of the optical fiber.
One additional consideration with respect to strain and temperature
measurements is the ability of the fiber gratings to withstand long term
operation at elevated temperature. This depends heavily on how the fiber
grating is written and processed. Typically, fiber gratings that are written using
side-imaging techniques either holographically or with phase masks will operate
with very little change up to approximately 400 degrees C. By writing "strong"
gratings that are essentially overexposed and annealed, this temperature range
may be extended to 650 degrees C [1]. Fiber gratings that are exposed to
short bursts of light that cause optical damage to form the fiber grating have
survived temperatures up to about 800 degrees C. One major issue of
operating at elevated temperatures for extended periods of time is that the
glass may soften allowing for the possibility of creep. This may in fact be the
limitation of using silicon dioxide based fiber gratings to measure strain at high
temperature. The temperature issue is important to civil structures when fiber
gratings are to be attached directly to structures using high temperature
processes.
Transverse Strain
One exciting advantage of fiber optic grating sensors is that they can readily be
embedded into a variety of materials to measure internal strain. Example of

application areas would include embeddment into fiberglass structures, carbon

epoxy, thermoplastics, concrete, and metals such as aluminum. When these
fiber gratings are embedded into materials, transverse strain may arise that will
also shift the period of the gratings. An initial design to attempt to solve the
transverse strain measurement problem involved writing four overlaid fiber
gratings, two of which were tilted so that transverse strain would change their
effective period [5]. The idea is to use the four resulting wavelength shifts from
the fiber gratings to generate four equations in four unknowns. A second
approach is to use dual overlaid gratings written onto polarization preserving
optical fiber. In this case, each fiber grating results in two effective fiber
gratings so again there are four equations in four unknowns [6]. Fabrication of
this fiber grating sensor is simplified since the

Page 171

two birefringent axes of the fiber provide automatic alignment of the transverse
sensing axes and only two fiber gratings need to be written along the
longitudinal axis of the optical fiber. It should be noted that polarization
preserving fiber need be used only in the vicinity of the fiber sensor since
information on all of the four effective fiber gratings is wavelength encoded.
One further issue with respect to using the fiber optic grating to measure
transverse strain is the orientation of the fiber in the material or on a surface.
Since many fibers are cylindrically symmetric, it would be necessary to mark
the coating on this type of fiber and establish appropriate procedures for
placement. An alternative approach is to use fibers that are shaped with flat
sides. The D shaped fiber marketed by Andrew Corporation or the flattened
sides of the ATT polarization preserving fiber would be appropriate choices.
These different fiber configurations will result in a different response to
transverse strain and work is in progress in this area. When the fiber is drawn,
it would also be possible to coat the fiber with an asymmetric coating that
would simplify placement of an appropriately oriented fiber.
Figure 1 shows a diagram of a multiaxis fiber grating strain sensor that is
formed by writing dual overlaid fiber gratings on polarization maintaining fiber.
This effectively establishes four gratings whose output wavelengths may be
tracked. The result is four equations in four unknowns allowing the potential for
three axes of strain and temperature to be measured. Recent experiments with
dual overlaid fiber gratings written at 1300 and 1550 nm onto polarization
preserving fiber made by 3M and Fibercore have shown that two axes of
transverse strain can be measured when the longitudinal strain and
temperature are known and that also three axes of strain may be measured
when the temperature is measured independently [7]. Figure 2 shows the
spectral response of a dual overlaid fiber grating when subjected to axial and
transverse loads. With an axial load, the peaks will shift, and with a transverse
load, the peak separation will increase.

Figure 1.
Dual overlaid gratings written onto polarization preserving fiber.

Page 172

Figure 2.
Peak shifts and peak separations due to axial and transverse strains.

It is difficult to measure temperature simultaneously because of the condition

number of matrix associated with temperature for standard polarization
maintaining fiber. Significant improvements can be expected by optimizing the
thermal and dispersion properties of the fiber as well as the characteristics of
the waveguide itself.
Figure 3 shows a system that has been used to track changes in a multiaxis
fiber grating based strain sensor. Edge emitting diodes at 1300 and 1550 nm
are combined using a wavelength division multiplexing element. The combined
light beam is directed into a 50/50 beamsplitter and used to illuminate a dual
overlaid fiber grating written onto polarization maintaining fiber. The reflected
light beam is directed back via the beamsplitter into an Ando 6315A Optical
Spectrometer, which is controlled via a GPIB bus, and a Dell 200 MMX PC. Blue
Road Research has developed automatic spectral peak tracking software for
this set of instrumentation that can be used to characterize the multiaxis fiber
grating strain sensors.

Page 173

Figure 3.
Demodulation system for a multiaxis fiber grating strain sensor.

Figure 4 shows the typical output spectrum of the dual overlaid fiber gratings,
in this case for 1300 and 1550 nm fiber gratings written onto Fibercore
polarization maintaining optical fiber. To characterize the transverse strain
response, Stanford and Sandia collaborated with Blue Road Research on the
design of a transverse test fixture shown in Figure 5. Rotational stages are
used to hold the fiber and rotate it with known angles. Typical response for the
fibers is shown in Figures 6 and 7 as a function of load. These tests were
performed using bare fiber with an outer diameter of 125 microns. Along one
of the polarization axes, the double peak spectral structure moves apart while
along the other axis it moves together. The fiber sensor can be oriented such
that the larger transverse load will cause the peak separation to increase,
thereby eliminating the limitation of the peaks shifting together. A typical
response curve is shown in Figure 8 for both wavelengths.

Page 174

Figure 4.
Reflected spectra from multiaxis grating sensor.

Figure 5.
Transverse strain test fixture for multiaxis fiber grating strain sensor.

Page 175

Figure 6.
Plot showing the peak to peak spectral separation versus transverse loading near 1300 nm
using a multiaxis fiber grating strain sensor.

Figure 7.
Plot showing the peak to peak spectral separation versus transverse loading near 1550 nm
using a multiaxis fiber grating strain sensor.

Page 176

Figure 8.
Transverse strain sensitivity versus angle for 1300 and 1550 nm using a multiaxis fiber
grating strain sensor.

Integrated Strain Measurements

In some cases it is highly desirable to be able to measure strain over large
distances that may be on the order of hundreds of meters or kilometers.
Possible applications would be measuring integrated strain over bridge spans
or strain build up in rocks prior to an earthquake or slide. Fiber gratings can be
used as virtual light sources that can be used to support the operation of
interferometeric sensors for integrated strain measurements. An example is the
Sagnac/Michelson

Page 177

interferometer. It consists of a Sagnac interferometer with fiber gratings placed

in the Sagnac loop [8]. The fiber gratings could be to measure local point
multiaxis strain and temperature, the Michelson portion of the system
measures integrated strain and the combination of the Sagnac and Michelson
interferometers offers distributed sensing of time varying events such as
acoustic waves [9].
Demodulation Methods
The ability of the fiber gratings to effectively measure strain depends on the
wavelength resolution of the demodulator. A modest performance open loop
approach involves taking a fiber coupler that is overcoupled so that it has
strong wavelength dependence to its split ratio [1011]. By monitoring both
legs and taking the ratio between them, the output wavelength may be
determined. A third open loop approach is to use a Mach Zehnder
interferometer at the output [12].
For higher performance applications, closed loop operation and high resolution
demodulators may be required. This may be particularly true of dual overlaid
fiber grating sensors to measure strain and temperature simultaneously or 3
axes of strain plus temperature. In both of the latter situations, accurate
results depend on successfully inverting matrices to solve simultaneous
equations [13]. In terms of wavelength accuracy at 1300 nm, 1000 microstrain
along the longitudinal axis corresponds approximately to 1 nm of wavelength
shift. For a sensor with a 10 microstrain requirement, this would correspond to
measurement accuracy of about 0.01 nm. For transverse strain, a 0.01 nm
resolution of the demodulator corresponds to approximately 30 microstrain or a
three to one decrease in sensitivity along the transverse axis. To achieve these
higher levels of performance, closed loop modulation techniques similar to
those used with the fiber optic gyro can be employed. In this case however,
instead of comparing and modulating phase, the spectrum of a tunable filter
which might be a fiber etalon, acousto-optic tunable filter or other well
controlled device may be used. When the tunable filter is modulated so that its
filter function beats against that of the fiber grating sensor first and second
harmonic signals result. By closing the loop and nulling out the first harmonic
signal, the reference demodulator filter can be locked onto the spectral peak of
the fiber grating sensor, potentially improving sensitivity by two ore more
orders of magnitude [3,5].

Multiaxis Fiber Grating Strain Sensor Applications for Bridge and Civil Structures
When a fiber grating is embedded into a composite structure, it is in real
applications subject to transverse as well as longitudinal strain. For simple fiber
grating systems that only measure longitudinal strain and temperature this
results in errors that may in some cases be quite large, especially in those
cases where transverse strain is one of the principal strain components. By
using the multiaxis fiber grating strain sensor, this problem can be overcome
and fiber gratings can be

Page 178

Figure 9.
Load cell with embedded distributed transverse strain sensing fiber grating sensors.

used in applications where transverse strain is significant and in fact the

important parameter to be measured.
Figure 9 shows a load cell where three in line multiaxis fiber gratings have
been placed in a plate with one of their transverse sensing axes aligned out of
the plane of the plate. This load cell can then be used to measure transverse
strain at key positions on bridges including drawbridge decks and
counterweights as shown in Figure 10. They could also be placed in the
footings of bridges and used to measure changes in the load distribution as
shown in Figure 11. Similar load cells could also be constructed in washers and
used for measuring tension on cables and loading of nut and bolt assemblies as
shown in figure 12.
Other civil structure applications that are not shown would include loading on
support structures for tunnels and mines, foundations of buildings and stress
distribution on retaining walls.
Summary
A summary of the usage of fiber optic gratings to measure multiaxis strain has
been provided. Means to measure transverse and integrated strains using fiber
gratings and temperature compensation has been discussed.

Figure 10.
Fiber grating based load cells monitoring decks and
counterweights in drawbridge applications.

Page 179

Figure 11.
Monitoring of bridge scouring and load distributions
using fiber grating based load cells.

Acknowledgements
Blue Road Research has been awarded a Phase II NASA SBIR contract, ''Single
Point 3-Axis Strain and Temperature Fiber Optic Grating Sensor", NASI-97003.
Dr. Robert Rogowski is the technical program monitor. This support, which is
being used to develop commercial multiaxis fiber grating strain sensor systems,
is greatefully acknowledged. The author would like to thank Jay R. Spingarn
and Tom E. Bennett of Sandia National Labs, Drew Nelson, Alberto Makino and
Craig Lawrence of Stanford University and Andreas Weisshaar of Oregon State
University for helpful discussions and test support.
Portions of the text of this article and all figures are adapted from the Blue
Road Research, Fiber Optic Sensor Workbook, copyright 1993, 1994, 1995,
1996, 1997 and 1998 and used with permission.

Figure 12.
Washer with embedded distributed fiber grating strain sensors to monitor
cable tensions and bolt loading.

Page 180

References
1. W. W. Morey, "Fiber Optic Grating Technology", Proceedings of SPIE, Vol.
2574, p. 22, 1995.
2. J. Dakin, "Distributed Fiber Optic Sensors", in Fiber Optic Smart Structures,
edited by E. Udd, Wiley, 1995.
3. E. Udd and T. E. Clark, "Fiber Grating Sensor Systems for Sensing
Environmental Effects", U.S. Patent 5,380,995, Jan. 10, 1995.
4. M. G. Xu, H. Geiger, and J. P. Dakin, "Multiplexed Point and Stepwise
Continuous Fibre Grating Based Sensors: Practical Sensor for Structural
Monitoring?", Proceedings of SPIE, Vol. 2294, p. 69, 1994.
5. E. Udd and T. E. Clark, "Sensor Systems Employing Fiber Optic Gratings",
U.S. Patent, 5,397,891, Mar. 14, 1995.
6. E. Udd, "Multiparameter Fiber Optic Grating Systems", U.S. Patent
5,591,965, Jan. 7, 1997.
7. C.M. Lawrence, D.V. Nelson, E. Udd, "A Fiber Optic Sensor for Transverse
Strain Measurement", Journal of Experimental Mechanics, Submitted
November, 1997
8. E. Udd, "Sagnac/Michelson Distributed Sensing Systems", U.S.Patent
5,636,021, Jun. 3, 1997.
9. E. Udd, "Sagnac Distributed Sensor Concepts", Proceedings of SPIE, vol.
1586, p. 46, 1991.
10. E. Udd, K. Corona, K. T. Slattery and D. J. Dorr, "Fiber Grating Systems
Used to Measure Strain in Cylindrical Structures", Optical Engineering, Vol. 37,
p.1893, July, 1997.
11. R. P. Kenny, "All-Fibre Grating Sensors Employing Many Cycle Directional
Couplers", Proceedings of SPIE, Vol. 2872, p. 32, 1996.
12. A. D. Kersey and T. A. Berkoff, "Fibre Grating Based Sensing System with
Interferometeric Wavelength Shift Detection", U.S. Patent 5,361,130, Nov. 1,
1994.
13. C. M. Lawrence, D. V. Nelson and E. Udd, "Multiparameter Sensing with
Fiber Bragg Gratings", Proceedings of SPIE, Vol. 2872, p. 24, 1996.

Page 181

Chapter 5
Sensor Material Interaction and Reliability

Page 183

Lifetime and Reliability of Embedded Optical Sensor Fibers

PH. M. Nellen,
A. Frank,
P. MuaronN
and U. Sennhauser
Abstract
Fiber reinforced structural elements equipped with glass fiber sensors are of
potential interest for long term surveillance of bridges, dams, and other civil
engineering applications. We embedded fiber Bragg grating sensors with
various coatings into glass fiber reinforced epoxy material during a continuously
on-going pultrusion process on a large scale production machine. The reliability
of embedded fiber sensors was tested by determining coating properties like
shear modulus and viscosity directly on the fiber and by performing push-out
tests. The latter testing revealed information about interfacial properties like
lateral stress on the embedded fibers, coefficient of friction and adhesion
surface energy. The residual axial strain was measured with the embedded
Bragg gratings and their sensitivity was determined. These investigations are
necessary to identify best-suited coatings to embed and protect fiber sensors
and to obtain reliability data on fiber sensor performance.
Introduction
Distributed fiber optical sensors have an excellent potential for monitoring
structural materials. Especially the fiber's filamentary geometry, their light
weight, and the lack of joule heating make them ideal candidates for low
intrusive embedding in composite structures [1,2].
Discrete multiplexed sensors, based on fiber optic Bragg grating technology
dominate interest for strain sensing. Bragg gratings are produced by specially
modulated UV-light inducing periodic spatial refractive index changes in the
fiber core. The irradiated fiber section acts as a Bragg reflector, i.e., light in a
narrow spectral range around a central wavelength determined by the optical
grating period
Ph. M. Nellen, A. Frank, P. Mauron, U. Sennhauser, EMPA, Swiss Federal
Laboratories for Materials Testing and Research, Ueberlandstrasse 129, CH-8600
Duebendorf, Switzerland

Page 184

is reflected. Applied strain and temperature change the optical grating period
resulting in a wavelength shift [3,4].
The objective of our studies is to develop a process to routinely equip glass
fiber reinforced polymer (GFRP) structural elements with distributed Bragg
grating fiber sensors for use in applications in civil engineering, e.g. long term
surveillance of bridges, dams, etc. Special emphasis is put on easy-to-use
measurement procedures, robustness, and low additional fabrication costs.
Major concerns of civil industry regarding the application of new technologies
are lifetime, reliability, and durability of the components. These can be
demonstrated by real world applications [5,6] but they have to be supported
by experimental laboratory studies based on accelerated aging and fatigue
tests, characterization of interfacial properties between host, coating, and
fiber, as well as proper modeling of the underlying mechanisms for lifetime
estimates.
We report on the embedding of optical fibers in GFRP rods on a full-scale
production pultrusion machine. Different modern fiber coatings were used, and
the fibers contained a total of twenty Bragg grating sensors. System
performance was investigated with respect to survival of fibers and coatings
due to embedding and to sensitivity of embedded sensors to temperature and
strain. Coating properties like shear modulus and viscosity were determined
with a torsion pendulum set-up and interfacial properties like chemical
adhesion, thermal stresses, and mechanical interlocking were studied with a
push-out apparatus built into a SEM.
Performing such investigations before and after exposing the structural
elements to environmental tests (climate tests, fatigue tests) allows to
determine the long term influence of various parameters like humidity,
temperature, quasi-static or dynamic load on the interfacial properties and thus
on the sensitivity and accuracy of the sensing parts.
Embedding of Fibers and Bragg Gratings into GFRP
Materials and Methods
Multi- and single-mode optical fibers, the latter with and without Bragg
gratings, were embedded into GFRP. All fibers were coated. We investigated
different polymeric coatings, namely two standard telecom double-layered

acrylic coatings, an older type (A1) and a recently released type (A2), a single
hard acrylic coating (A3), two polyimide coatings, one directly adhered to the
glass (P1), and one with a thin carbon layer between fiber and coating (P2).
The fiber diameter was 125 mm except for the hard acrylic-coated fibers (85
mm). The thickness of the acrylic coatings was 63 mm, that of the polyimide
coatings was about 15 mm. Some sensors were single shot low reflectivity
Bragg gratings directly produced on a fiber draw tower and coated without any
further manipulation [7], others were commercially available Bragg gratings
which were produced by stripping the fiber section, writing the grating, and recoating the fiber.

Page 185

A full-scale pultrusion machine was used to continuously embed the optical

fibers parallel to the unidirectional reinforcing fibers during the fabrication
process. The maximum curing temperature of the used epoxy resin was 200
C. The rod of 8 mm in diameter was cut to the desired length after curing.
The fiber-equipped GFRP specimens were first inspected visually. Samples were
prepared for stereo optical and electron microscopy investigations. The Bragg
grating sensors were characterized before and after embedding with respect to
their peak wavelength, width, and reflection shape using a direct spectroscopic
approach to measure wavelength changes: a grating spectrometer with CCDarray with center wavelength of 830 nm, range of 30 nm, absolute wavelength
accuracy of 4 pm, and a short term resolution of 1 pm [5].
Three light-coupling concepts were tested to optically access the embedded
Bragg gratings:
a) One end of the rod was thinned down to a small diameter and the
remaining epoxy (including coating) was stripped off the fiber. The protrudent
bare fiber was then spliced to a fiber lead. Though mechanically protected
afterwards the process is very fragile and is not used further.
b) Another method utilized GRIN lens coupling to the polished rod surface
without permanent fixation of a lead fiber. This coupling is easy to handle but
is very sensitive to any vibrations.
c) The most used coupling method was fiber butt coupling (core to core) to the
polished GFRP surface and then bonding the fiber with optical adhesive. After a
difficult search for a suitable adhesive having low shrinkage during curing,
thermal stress problems leading to misaligned fiber cores could be solved.
All three light coupling concepts are time consuming and other methods are
under investigation.
Embedding Results
Both, acrylate and polyimide coated fibers survived the embedding process. No
special care was taken to place the fiber exactly in the middle of the rods. Thus
several fiber parts were found to be close to the surface of the rods. In
principle, this could be easily avoided by a proper mechanical fiber guidance.
Sample inspection after preparation for push-out tests (see Fig. 1 below)

including sawing and surface polishing showed that the acrylic coatings (A1
and A2) were often degraded whereas the polyimide coatings (P1, P2) and the
hard acrylic coating (A3) remained intact. Reasons are the high processing
temperature of 200 C which is well above typical glass transition temperatures
of around 60 C for acrylate coatings (A1, A2) and well below for polyimide
(around 265 C) or the relatively low elastic modulus of acrylic coatings (A1,
A2) which easily can be damaged mechanically during sample preparation.
Figure 1a and 1b show polished surfaces of embedded acrylate coated fibers,
one with a double-layer coating structure, the other with a single-layer coating.

Page 186

Figure 1:
Aspect of embedded fibers under electron microscope with a) double-layer (A2)
and b) single-layer acrylate coating (A3).

The fiber transmission and grating reflectivity was not measured during the
continuous pultrusion process. Afterwards, light attenuation of multi-mode
fibers due to the embedding process was found to be acceptable: i.e., < 0.11
dB / m. Attenuation was negligible for the single-mode fibers due to their
smaller sensitivity to microbending.
Only one out of twenty embedded Bragg gratings was broken during the
embedding process. The reason was identified to be poor polyimide recoating
of the grating region.
In contrast to GFRP (0/90)12 laminate specimen fabricated earlier, where the
shape of the reflection peak of the embedded Bragg gratings splitted [8], the
pultrusion process did not change the reflection peak shapes. However, the
gratings showed a residual axial strain after the embedding. The mean residual
strain after the pultrusion process of A3-coated fiber Bragg gratings was + 150
mm/m, they were strained due to the difference in thermal expansion
coefficients of matrix and fiber. The residual axial strain distribution shown in
Fig. 2 ranges from 100 to + 400 mm/m.

Figure 2:
Residual axial strain distribution of Bragg gratings
with A3-coating.

Page 187

Testing Sensor Reliability of Embedded Fibers

Optical fiber coatings are important to improve fiber strength, fatigue
resistance, and low attenuation [9]. Any coating should protect against
external abrasion and mechanical damage, aging (stress corrosion, crackgrowth), and microbending. Modern standard telecom fibers are coated with
dual-layer acrylate coatings; a low modulus primary coating acting as a cushion
for microbending losses surrounded by a high modulus secondary coating
which dissipates lateral forces along the fiber. To fulfil the requirement of good
strippability the primary coating has a minimum bonding to glass. The
secondary coating protects against various environmental conditions. Therefore
most modern telecom coatings do not fulfil the requirements for embedded
fiber optic strain sensors. Instead of high strippability they should have high
debonding strength and creep free force transfer from matrix to fiber and they
should survive the high curing temperatures of polymers.
Sensor performance critically depends on interfacial properties like debonding
strength, surface bond energy, and frictional forces. By means of a torsion
pendulum, the shear modulus G and viscosity h of the coatings were measured
in-situ directly on the fiber. Push-out tests with a push-out apparatus built into
a SEM were performed allowing inspecting the composites of interest directly
and not an arbitrary model system, which is often the case for pull-out tests.
The results were fitted to a push-out model to obtain interfacial properties like
friction coefficient m, normal stress sn, and interface toughness (surface
energy) G. Some of these physical parameters are expected to change with
time due to different loading schemes like dynamic fatigue tests, constant load
tests or environmental conditions like temperature or chemical influence. The
interfacial adhesion may therefore not only determine lifetime and reliability of
the sensors but may also be of importance in coating optimization.
Coating Characterization with Torsion Pendulum
A fiber torsion pendulum was used to measure shear modulus G and viscosity h
of the coatings directly on the fiber at a temperature of 23 C and at a
frequency of 0.1 Hz [Fig. 3a]. The parameters G and h as a function of
frequency can be used to model the viscoelastic behavior of the coating [10].
The shear moduli were used as input in the push-out model (see below). No
attempts were made to measure over a broad frequency range and to separate
creep and delayed elastic behavior.

The differential equation of the pendulum is given in Eq. 1. The aluminum disc
determines the moment of inertia q. Torsion stiffness T of the pendulum
depends on the shear modulus of fiber Gf and coating Gc and on fiber and
coating radius, rf and rc, respectively. Damping g depends on coating viscosity
hc, damping of fiber gf and aluminum disc gd, and both radii.

Page 188

Figure 3:
Torison pendulum set-up (a) and amplitude damping of coated (A3) and uncoated
fibers of two different lengths (b).

To eliminate damping effects from gf and gd, two measurements were made
with the same fiber, one with the coating the other with the bare fiber with the
coating being chemically stripped. The main error in determining Gf and hc
results from the uncertainties in fiber and coating radii as they enter the
equation raised to the fourth power. For double-layered coatings mainly the
shear modulus of the outer, secondary coating contributes; the inner coating
cannot be characterized with this method. In Fig. 3b amplitude versus time
measurements are shown for coating A3 for two different fiber lengths, 100
mm and 370 mm, each measured first coated and then stripped.
Figure 4 compares the measured values of shear modulus and viscosity of
different fiber coatings. Coatings A1 and A2 have a shear modulus of 2 and 9
times less than A3 and 330 times less than P2. Viscosity is in the same order of
magnitude for all coatings.
Push-Out Tests to Determine Interfacial Properties
We performed extensive push-out testing with the different coated fibers
embedded into the GFRP samples made with the pultrusion process described
above. A schematic drawing of the push-out principle is shown in Fig. 5. A 100
mm tungsten carbide indenter pushes the fiber through the matrix (indenter is
partially visible in Fig. 1a). The indenter velocity was 0.3 mm/s.

Page 189

Figure 4:
Shear modulus and viscosity of coatings A1,
A2, A3, and P2.

Various measured push-out load-displacement curves are shown in Fig. 6. In

push-out tests load increases until debonding starts. The initial linear slope can
be attributed to the overall system compliance. Interface failure initiates in the
region of maximum stress near the surface. Once a crack has formed,
compliance of the specimen will increase due to the increased debonded
length of the fiber and the load-displacement curve will begin to bend and rises
with a decreasing slope to maximum load. The work to propagate the crack tip
may be assumed to be constant but increasing work is required to slip the
increasing debonded length of the fiber against friction. In addition, Poisson
expansion increases load to push out the fiber. At the end of the process the
remaining bonded length debonds instantaneously. Peak load occurs before
this final debonding. It is the load required to slip the entire fiber length and to
debond the final increment. The load drops then to that required to overcome
friction along the remaining embedded but debonded length. If the coefficient
of static friction is sufficiently greater than the coefficient of dynamic friction,
slip-stick behavior can be observed. In case of abrasion the friction will
decrease more than linearly with length, i.e., the friction coefficient may
change during the push-out test.

Figure 5:
Schematic drawing of push-out
experiment.

Page 190

Figure 6:
Push-out curves of embedded fibers with coatings a) P1, b) P2, c) A3, and d) A2.

Analysis of the push-out tests was performed according to the model described
in [11]. It was modified to include the fact that the optical fiber is not directly
embedded into the matrix but is surrounded with a less stiff coating. A simple
shear-lag model was assumed in which matrix and fiber are only compressed
and the coating is sheared [12]. From a load-displacement curve interface
parameters can be determined, viz., friction coefficient m, normal stress at
interface sn, fracture toughness of the interface (surface energy) G, and the
remaining bonded length after peak load in percent n of the total embedded
length.
In Fig. 6a - d the push-out load-displacement curves measured for embedded
fibers with coating types P1, P2, A2, and A3 are shown. The thickness of all
specimen was around 800 mm. Note the different scales for both displacement
and load. Obviously, the push-out curves show large differences between the
various coatings, which can be attributed to the different interfacial properties.
We found that for the tested fibers and matrix system the fiber/coating
interface was weakest, i.e., in all cases the fiber/coating interface debonded
and not the coating/matrix interface.

Page 191
TABLE 1: PUSH-OUT FIT
coating type
P1
G [J/m2]
33
100
s n [N/mm2
m [1]
0.2
n [%]
70

PARAMETERS
P2 A3 A2
200 3.6 100 0.5 0.5 0.5 46 56 -

Also shown in Fig. 6 are the curve fits with the average values of the interfacial
parameters. The push-out data scatters considerably. This may be caused by
neighboring reinforcing fibers influencing local Young's modulus or changing
the effects of Poisson compression and thermally induced radial stresses or
other effects including measurement uncertainty of tests. The fit parameters
are preliminary. Nevertheless, the fits allow to compare the different interfacial
properties that are summarized in Table 1.
In case of coating P2 a very fast push out is observed after maximum load
(curve is interrupted because the measuring rate of 4 Hz was to slow to follow
up). This may lead to high abrasive effects between the intermediate carbon
layer and the coating leading to a varying friction coefficient and normal
asperity friction stress. This assumption was supported by data obtained from a
very small specimen thickness of 260 mm, where the frictional load was found
to be higher than for the 800 mm specimen. Therefore we fitted the curve with
a higher frictional load than measured for the 800 mm specimen which leads to
the deviation seen in the tailing part of the curve.
The double-layered standard telecom coating A2 was found to scatter so
much, that no fitting was done at all. It seems that for embedding purposes
specialty coatings have to be used although their application comes at
somewhat higher price.
Temperature Sensitivity of GFRP Embedded Sensors
In order to test sensor performance we compared the temperature sensitivity
coefficient of dl/dT = 6.8 pm/ C for fiber Bragg grating sensors cited in the
literature [3], to the measured temperature dependence of bare, acrylate
coated (A3-coating), and GFRP-embedded fiber gratings over a temperature
range of - 40 C to + 80 C (see Fig. 7).
The specimens were tested in a climate chamber which was cycled from room
temperature down to - 40 C up to + 80 C and back again. The cooling and

heating rate was slow enough to guarantee thermal equilibrium. We found

temperature sensitivities of dl/dT = 4.9 pm/ C, 5.4 pm/ rdegre;C (T > 0 C)
and 8.5 pm/C (T < 0 C), and 8.8 pm/ C, for the bare, acrylate coated, and
embedded gratings respectively. The sensitivity above and below 0 C of the
coated fiber was different. The reason may be a phase transition within the
coating.

Page 192

Figure 7:
Temperature sensitivities of bare, A3-coated,
and embedded Bragg gratings.

Both, the curve of the bare and the embedded grating showed a slightly
quadratic term of d2l/dT2 = 0.007 pm/C2, and d2l/dT2 = 0.008 pm/C2 over
the investigated temperature range.
This example demonstrates the importance of a careful calibration of
embedded fiber sensors. No doubt, similar tests have to be performed to
compare tensile load sensitivity to theoretical values and to check load transfer
from matrix through coating to the sensing Bragg grating. Finally, such tests
must also be done under simulated environmental service conditions and under
or after accelerated aging.
Conclusions
During a continuously on-going pultrusion process on a large scale production
machine fiber Bragg grating sensors, coated with acrylate and polyimide, were
embedded into glass fiber reinforced epoxy rods. Coating shear modulus and
viscosity as well as the lateral stress on the embedded fibers, coefficient of
friction and adhesion surface energy were determined directly on the system of
interest with a fiber torsion pendulum and push-out tests. The residual axial
strain was measured with the embedded Bragg gratings and their sensitivity
was determined. Clear differences were observed between the coatings.
Further investigations are now necessary to correlate GFRP-embedded fiber
sensor failure modes with their interfacial properties to answer reliability
questions and estimating sensor lifetimes.
Acknowledgements
We would like to thank EMPA Thun, CH, for their help with the push-out

experiments, Naval Research Laboratory, Washington DC, USA, for providing

Page 193

high quality fiber Bragg gratings, the companies Weidmann AG, Rapperswil,
CH, and Sportex GmbH, Neu-Ulm, D, for the opportunity to use their pultrusion
machine, and the Swiss Priority Program Optique II for their financial support.
References
1. Carman G. P. and Sendeckyj, G. P., 1995. ''Review of the Mechanics of
Embedded Optical Sensors", Journal of Composites Technology & Research,
JCTRER, 17(2):183193.
2. Sirkis, J. S. and Dasgupta, A., 1993. "The Role of Local Interaction
Mechanics in Fiber Optic Smart Structures", Journal of Intelligent Material
Systems and Structures, 4:260271
3. Rao, Y.-J., 1997. "In-Fibre Bragg Grating Sensors", Meas. Sci. Technol.,
8:355375.
4. Hill, K. O. and Meltz, G., 1997. "Fiber Bragg Grating Technology
Fundamentals and Overview", Journal of Lightwave Technology,
15(8):12651276.
5. Nellen, P. M., Anderegg, P., Brnnimann, R., and Sennhauser, U., 1997,
"Application of Fiber Optical and Resistance Strain Gauges for Long-Term
Surveillance of Civil Engineering Structures", Conference on Smart Systems for
Bridges, Structures, and Highways, San Diego, SPIE Proceeding, 3043:7786.
6. Brnnimann R., Nellen, P. M., and Sennhauser, U., 1998. "Application and
Reliability of a Fiber Optical Surveillance System for a Stay Cable Bridge",
Smart Mater. Struct., Special Issue on Fiber Optical Structural Sensing: in
press.
7. Askins, C. G., Tsai, T.E., Williams, G. M., Putnam, M. A., Bashkansky, M.,
and Friebele, E. J., 1992. "Fiber Bragg Grating Reflectors Prepared by a Single
Excimer Pulse", Opt. Lett., 17(11):833.
8. Sennhauser, U., Brnnimann, R., and Nellen, P. M., 1996, "Reliability
Modelling and Testing of Optical Fiber Bragg Sensors for Strain
Measurements", Conference on Fiber Optic and Laser Sensors XIV, Denver,
SPIE Proceeding, 2839:6475.
9. Biswas, D. R., 1994. "Optical Fiber Coatings", in Fiber Optics Reliability and
Testing, Critical Reviews of Optical Science and Technology, Paul, D. K., ed.

Bellingham, CR50:6379.
10. Gebizlioglu, O. S., Grimado, P. B., Plitz, I. M., and Frantz, R. A., 1994.
"Mechanical Properties of Fiber Coatings by Dynamic Mechanical Analysis of
Optical Fibers", SPIE, 2290:718.
11. Kerans, R. J., 1991. "Theoretical Analysis of the Fiber Pullout and Pushout
Tests", J. Am. Ceram. Soc., 74(7):15851596.
12. Suhir, E., 1993. "Approximate Evaluation of the Interfacial Shearing Stress
in Cylindrical Double Lap Shear Joints with Application to Dual-Coated Optical
Fibers", Int. J. Solids Structures, 31(23):3261-3284.

Page 194

Evaluation of Adhesion Behavior of Optical Fibers for Sensors Embedded in

Cementitious Materials
W. R. Habel,
E. Schulz,
G. Kalinka
and A. Bismark
Abstract
Fiber-optic strain sensors intended for embeddment in cementitious matrices
should be coated with very thin alkaline-resistant coatings. This is necessary to
achieve a reliable strain transfer from the matrix to the fiber sensor.
In order to clarify possible ways to replace ways to replace rather thick
coatings and to design fiber optic sensor areas for embeddment, modified
optical fibers have been investigated with regard to their chemical compatibility
and to adhesion behavior to the matrix. The surface state of optical fibers
intended for embeddment has been characterized An additional strength
estimation of alkaline attacked fibers was carried out in order to evaluate the
damage rate of the fiber surfaces if they come into contact with cement
pastes. All estimations have been compared with those of bare alkaline
attacked fibers.
The results suggest that fibers coated with very thin organic layers have a
sufficient tensile strength for handling and have remarkable conditions for
adhesion. The real adhesion behavior were estimated. Results of the
fiber/cement bond strength of very thinly coated fibers are presented.
Examples of application are referred to.
Introduction
The development of mortars and concrete products, the quality control during
construction, and the evaluation of the process of condition monitoring of
buildings, all require information about material properties and component
parameters.
Structure-integrated fiber-optic sensors deliver deformation and vibration
information needed for monitoring the state of a structure. They should be
installed during construction in order to achieve in-place quality control of
building materials in the construction phase, as well as the monitoring of

damage in the structure's

Wolfgang R. Habel (to whom correspondence should be addressed), Gerhard
Kalinka, Eckhard Schulz, Federal Institute for Materials Research and Testing (BAM)
Berlin, D-12200 Berlin/Germany Alexander Bismarck, TU Berlin, Inst. f. Technische
Chemie, Str. d. 17. Juni 135, D-10623 Berlin/G W.R.H. conducted this research
partly at the Institute fr Erhaltung und Modernisierung von Bauwerken e.V.
(Institute for Rehabilitation & Modernization of Buildings) at the TU of Berlin

Page 195

service life. Long-gage length sensors installed for measurement of integral

strain in extended concrete sections of bridges, dams, tunnels or towers, can
be coated with thick polymer materials and/or packed into flexible tubings. The
necessary bond to the structural unit is achieved by several fixing points along
the fiber path. This method impedes a reliable sensing of signals occurring
inside the concrete microstructure, e.g. recording crack initiation and
progression or acoustic emissions. In order to make optical fibers sensitive to
these signals, a direct contact between sensor and surrounding matrix material
is required. An interesting application could be the monitoring the high
performance characteristics of high strength concrete, e. g. for bridges,
wastewater treatment plants, foundation piles. For such purposes, the fiber
sensor design has to be optimized for a reliable sensor/matrix interaction. An
intimate embeddment of non-reactive fiber-optic micro strain sensors such as
fiber Fabry-Prot interferometer sensors stimulate deformation measurements
with high resolution at early ages as well as later. On the other hand, if
distributed fiber optic sensing systems, e. g. Brillouin sensors, bonded
throughout the fiber length with the cement matrix, or neural networks or
single fibers, e. g. integrated into glass cloth-reinforced concrete components
or into the contact zone between component layers, were established in
structural monitoring, an embeddment technique such as the afore-mentioned
would enable the determination of harmful effects on long-term integrity earlier
than presently possible. Personnel responsible for safety and bearing capacity
could be warned properly if unexpected damages occurred, for example, in
buildings open to the public or concrete bridges and tunnels for high speed
trains.
Demands On Fiber-Optic Sensors To Be Integrated Into Concrete Structures
Integration of fiber-optic sensors without an extensive packaging in a concrete
material poses a challenge for all engineers involved. An embedded voluminous
sensor produces stress concentrations in the zone to be evaluated and the
stiffness will be altered compared to the plain material. This will be a problem
when shrinkage of still low strength mortar or concrete blocks are to be
observed and controlled [1]. In order to find the optimum sensor design which
is not only practically relevant but also enables reliable strain transfer from the
matrix, two main questions have to be considered:
a) Which minimum protection will suffice in order to guarantee safe-handling

during installation and construction as well as to protect the optical elements

from adverse influences on-site?
b) Which quality of strain transfer, that means, which bond strength can be
achieved for the chosen sensor protection?
Both problems are mutually important. The demand on a reliable strain transfer
requires an unobtrusive protection, whereas the demand on safe-handling
obstructs the transfer of the measured value and thus the performance of fiber
sensors apart from laboratory applications. In order to make full use of the
performance spectrum of fiber sensors we started investigations to create
alternative, very thin coatings for the sensing area of optical fibers. However,
the practical use of such methods will be strongly influenced by applicationtechnological issues at the building site.

Page 196

Sensor Durability
Continuously drawn cheap fibers usable for sensors possess, in general, a
polyacrylate coating system or a polyimide coating. Such coating types have a
limited performance when they are attacked by concrete pore solutions [2].
We have been investigating some relevant questions for the past few years. It
could be found that all polymerically coated fibers and fiber sensors survived
the embeddment procedure, the initial set and the deformation during
hardening. However, it should be noticed that such fiber coatings exposed to
alkaline attacks for more than three years did not show a long-term stability.
They could not prevent a contact between moisture or pore solution and the
glass surface. Only one coating type (fluorinepolymer) seemed to be alkalineresistant, but it lost its dimensional stability.
We believe that the development of alternative alkali-resistant coatings, such
as combined inorganic/organic or metallic ones, will be a matter of economy.
One must reckon with alkaline attacks against the glass surface as long as
cheaper fibers manufactured for telecommunication purposes are used for
sensor applications. However, the latter need not lead to sensor failure if the
fiber sensors are to record elastic strain changes in the pre-cracking case of
concrete. Assuming a tensile failure strain of concrete in the order of 0.01 %
and assuming elastic bonding to the sensor/matrix composite (can be actually
assumed as discussed later), an embedded fiber must endure a strain of < 1
mm/m. This corresponds to an axial stress of < 70 MPa for decoated fibers.
Tension tests carried out with 450 mm long decoated optical fibers stored for
30 days in an artificially produced pore solution (pH = 13,1) corresponding to
that found in concrete proved that decoated fibers still ensure, at least, an
axial stress of 280 MPa. The same fibers embedded in wet aged Portland
cement mortar blocks (w/c ratio = 0.45, aggregate: 0.25 0.5 mm) survived a
storage time of three months and a subsequent bending test until matrix
cracks occurred.
Concluding from these strength measurement results, bare fiber-optic sensor
areas have been also evaluated prior and after embeddment in cementitious
material. In contrast to long sensor fibers, single-piece manufactured fiber
sensors (extrinsic fiber interferometer sensors or photoinduced Bragg grating
sensors) rather enable the application of alternative treatment and protection
methods to control the adhesion between fiber sensors and cementitious

matrix.
In the following, experimental methods for controlling the bonding willingness
of the sensor surface and the evaluation of the adhesion strength are
described.
Pre-Evaluation Of The Adhesion Behavior By Measurement The Wetting
Behavior Of Optical Fibers
In order to get a maximum adhesion physic-chemical properties and chemical
reactions at the interface are relevant. Surface characteristics of fibers
intended for embeddment should be known with respect to the bonding
willingness of the fiber surface. The wettability can be controlled by pretreatment of the sensor surface so that the fiber surface and the fluid matrix
are brought into contact in a molecular range (nm-order). A good wetting
behavior is necessary but not sufficient to guarantee the desired adhesion. In
general, chemical bonds between both components and an adequate cohesive
strength of the matrix material surrounding the sensor determines the success.

Page 197

Methods for Controlling the Wettability

The measuring methods applied for characterization of the wetting-determined
fiber surface behavior is based on macroscopic models. The equilibrium contact
angle q for a liquid drop situated onto an ideally smooth, homogeneous, planar
and nondeformable surface is related to the various interfacial tensions. The
observed contact angle can therefore be used as a direct measure for the
wettability of the fiber sensor.
A reliable and convenient contact angle determination method for drop-on-fiber
systems has been developed on the basis of our computer aided pendant-drop
method [3]. Droplets of a pore solution (extracted from a fresh Portland
cement mortar; age 2 hours) were put on optical fibers by dipping them into
the pore solution. Due to the relative high vapor pressure and the small volume
of such droplets the liquid was vaporized very quickly, and therefore only a
small number of the droplets could be measured.
Using the generalized drop length-height method we obtained - within the
error range - the same contact angle for pore solution on the optical fibers as
found by using the Wilhelmy Method vs. water. We calculated the solid surface
tension of the pore solution from the advancing contact angle qa determined
by immersing the dry optical fibers into the test liquid. An analogous result
followed from a determination of the surface tension using a Du-Noy ring
tensiometer. The measured value for pore solution (g = 72.2 mN/m) is
comparable to distilled water (g = 72.8 mN/m). Therefore, the fiber solid
surface tension can be calculated by contact angle measurements versus
twice-distilled water and diiodomethane. All necessary values are well-defined.
There is still another advantage: the use of water as test liquid does not lead to
problems due to carbonation of the cementitious solution in air.
A more reliable way to determine the contact angle is the gravimetric method the modified Wilhely technique. The advancing and receding angles qa and qr
obtained by this technique can be calculated by the Wilhelmy Equation from
the weight data (mass changes Lm), which were detected during the
immersion and emersion of a fiber into the test liquid:

were g is the acceleration of gravity, U the fiber perimeter and g the surface

tension of the test liquid in use. The effective fiber diameter d was checked by
Scanning Electron Microscopy (SEM). In order to determine the solid surface
tension one has to know the polar and non-polar parts (gp and gd) of the test
liquid's surface tension. This is achieved by estimation of the contact angle
values between the optical fiber and a polar liquid (e. g. water) as well as a
non-polar liquid (e. g. diiodomethane). To calculate the surface tension of the
fibers we used the harmonic-mean method introduced by Wu [4].
Tests carried out with polymerically coated optical fibers by using the drop
length-height method yielded the contact angles for coated fibers (e. g. for
polyimide coating after solvent cleaning: qr = 79,8&rdegree;). The adhesion to
a cementitious matrix is as expected weak. Research was conducted with the
aim to improve adhesion behavior. In order to achieve good adhesion behavior
combined with a protection of

Page 198

the fiber surface we treated chemically decoated surfaces of optical fibers with
different silanes. The basic idea is that a silanol layer increases the
hydrophobicity which protects the fibers against alkaline attacks. We have
chosen two special silanes which are applied as hydrophobic agents in building
protection [5]. It is important to notice that after finishing the condensation
reaction the silanes have a remaining activity due to non-reacted functional
groups. To activate the relatively non-polar glass surfaces, decoated optical
fibers were oxygen-plasma treated to increase the amount of oxygencontaining surface groups onto the fiber surface.
Some effort is required to find out the optimum silane concentration. For this
reason we carried out a considerable number of test series comprising, first,
decoated untreated oxygen-plasma treated optical fibers (modified for 1 min
up to 7.5 min) and, second, those fibers silanized with different silane
concentrations, ranging from 0.1 to 1 mol-% silane for the same time of
silanization.
Results in Surface Characterization
All measurements were carried out, at least, six times on fibers of the same
sample to check the reproducibility. Thus, all contact angle values represent
average values with standard deviations. The contact angle measurements
were performed in an air-conditioned room at a constant temperature of
20&rdegree;C.
Untreated but chemically (in THF) decoated optical fibers have bad wettability
properties relative to water. However, the wettability increases up to an
exposure time in oxygen plasma of 5 min. It decreases again for plasma
treatment longer than 5 min. This behavior could be confirmed by more than 5
test series with 6 identical samples [6] (shown in figure 1). The plasma
modification leads to an increase of the fiber surface polarity and therefore to
an increased wettability versus the polar liquids (water and pore solution).

Figure 1
Contact angles versus water on oxygen plasma treated optical glass fibers
*tetrahydrofurane

Page 199

Table I. shows that the polar component of the surface tension gp enlarges
dramatically due to the oxygen plasma treatment so that the addition of both
parts (gd + gp) results in an increasing fiber surface tension g.
TABLE I. - SURFACE TENSIONS OF OXYGEN PLASMA
TREATED GLASS FIBERS
Plasma treatment gp
gd
g [m/Nm] Polarity Xp
time [min]
[mN/m][mN/m] = gd + gp = gp / g
19.1 25.3
0
44.4 3.0
0.43
1.7
2.5
25.7 17.9
1
43.6 3.3
0.59
3.2
1.0
31.5 15.6
2.5
47.1 1.2
0.67
1.2
0.4
38.8 15.8
5
54.6 7.0
0.71
6.9
0.6
31.4 16.1
7.5
47.5 0.8
0.66
0.8
0.2
TABLE II - ADVANCING AND RECEDING CONTACT ANGLES FOR
SILANIZIED GLASS FIBERS VS. WATER
Silane
concentr.
0.1
0.25
0.75
1,0
[%]
Treatment
qa [] qr [] qa [] qr [] qa [] qr [] qa [] qr []
time
59.3 41.6 67.7 57.5 60.3 51.3 75.0 57.3
1 min
0.8 0.1 2.3 0.7 6.7 5.0 2.4 2.4
62.3 45.6 49.1 54.6 66.3 48.5 70.0 47.4
5 min
7.4 0.8 0.5 0.5 1.8 2.3 4.4 1.1

Table II. shows the contact angles of modified and silanized optical fibers
estimated against water. In comparison to plasma treated but unsilanized glass
fibers the contact angles are enlarged by a huge value due to non-polar
aliphatic groups of the silanole. Only the 1 min-plasma treated silanized optical
fibers (0.1% and 1% silanole solution) did not show this behavior. The 0.1%silanole concentration is probably too low to create a continuously silanized
fiber surface, whereas with a 1%-silanole solution concentration a polysilanole
film on the fiber surface is formed. Bare optical fibers silanized after a 5 min-

oxygen plasma treatment showed for all silanole concentrations the same
contact angles within the error range. However, the contact angle values of
silanized optical glass fibers estimated against water are lower compared to
aliphatic polymers, polysiloxanes and other silanized glass surfaces. This
indicates that the surface is not completely covered with silanole and
unreacted groups of the silanole are left on the surface. In both cases a rest
activity of functional groups bonded to the fiber surface was achieved which
would possibly lead to a chemical reaction between silanized glass fibers and
concrete (see bond strength tests).
Experimental Investigation of the Fiber/Matrix Adhesion Strengh
The intention to produce fiber-reinforced concrete initiated extensive
investigations concerning the bonding mechanism between fibers and matrix
materials. Based on these results, the strain transfer behavior of very thinly
coated fiber-optic sensors

Page 200

has been investigated. Research by Majumdar & Laws [7] and Nanni et al. [8]
in the field of glass fiber-reinforced concrete field will help to understand what
happens if the surface of optical fibers come into contact to cement paste.
Assuming that an optical fiber sensor is embedded without any packaging and
is has completely bonded along the fiber length with the cementitious material,
a shear stress due to loading can arise at the interface. Up to a definite stress
level elastic stress transfer will be the dominant mechanism at the interface.
The elastic shear stress distribution along the optical fiber is constant, with
exception of its ends if the fiber ends inside the matrix. If the sensor is located
in the area along the fiber where the shear stress is constant, it will work
without systematic error due to irregularities in shear stress distribution.
However, a more realistic situation is that the sensor fiber will have a nonconstant shear stress distribution along the sensing field caused by a reduced
bond strength due to irregularities in the sensor surface or deviations in the
matrix microstructure. This can lead to an early debonding of the sensor and
its function will fail. In order to estimate the actual range of deformation which
will not lead to debonding the bond strength values must be known for which
the stress transfer mechanism changes.
Test Methods
Two test methods can be used for the determination of the adhesion level
between optical fibers and cement based matrix materials: the single fiber pullout test and the single fiber push-in test. In contrast to the real life situation,
where the deformation of the concrete material is transferred into the fiber
sensor, in these tests the fibers are deformed under well-defined conditions
and the stress deformation is transferred into the matrix surrounding the
loaded fiber. Nevertheless, as in the real situation a shear stress exceeds a
certain level. This fact gives information about the quality of adhesion between
the components. In order to interpret these experimental results, one has to
take into account that the stress field along the embedded fiber is nonhomogenous as explained above.
Pull-out test [9]
For pull-out tests, a tensile stress is applied to an embedded fiber. For this
purpose, small mortar blocks (13.5 mm 10 mm 5 mm) which contain
individual optical fibers were prepared. The length of the embedded part of the

fibers was about 2.5 mm and the free fiber length was about 10 mm. Before
testing, the samples were cured for 28 days, at least, in moist air (96%RH). In
order to perform the pull-out test, the samples were loaded by a self made test
machine. Each mortar block was attached to a piezo translator which is able to
generate a maximal displacement of 150 mm. The free fiber end was glued
with a cyano-acrylate to the opposite fixture where a force transducer was
integrated. The gage length of the fiber between the mortar block and the
transducer was about 3 mm. The pull-out procedure was controlled by a
computer. The controller unit ensured a linear deformation of the translator.
The pull-out velocity was 0.8 mm/s at the translator. The elongated fiber
induced a force which was detected by the piezo-based force transducer.

Page 201

Push-in test [10]

A thin slice of material is carefully cut out of a mortar block containing an
embedded single fiber. The cutting direction is perpendicular to the fiber
direction. In order to perform the test, a conical indenter needle with a flat top
is positioned accurately on the top of the fiber segment. Then an increasing
pressure force is applied to the fiber in the axial direction which produces an
increasing shear stress at the interface. Debonding starts beyond a certain
load, it proceeds and finally the fiber is completely debonded and could be
pushed through the slice. Figure 2 illustrates the testing arrangement and
shows a typical indenter needle used.

Figure 2
Testing arrangement for push-out tests a) schematic, b) Indenter needle used

The change of the compliance caused by debonding influences the forcedisplacement relation significantly and the beginning of the debonding could
be determined by extrapolation. In order to avoid fiber splitting, the contact
area between fiber and indenter has to be as large as possible. Good results
were obtained by using a conical geometry with a flat top covering about 70%
of the fiber top area. A cone is stiffer than a cylinder and the increasing
diameter of the cone should not be a problem because the displacement
between fiber and matrix is expected to be low. The angle of the cone should
be less than 30. Two surface properties can be determined for an assessment
of the bond strength by using the push-in test: the surface energy G and the
frictional stress in the already debonded area t. For the calculation of these
properties, an extended model was used.
Comparison of the Methods used for Bond Strength Measurements
The main difference between push-in and pull-out test lies in the direction of
the force applied to the fiber. For the push-in test a pressure stress is applied

to a single fiber while for the pull-out experiment the composite is loaded with
a tension stress. For this reason, the push-in test method should be preferred
for the evaluation of the shear stress situation of embedded sensor fibers in
loaded, uncracked matrices. On the other hand, the pull-out method enables
the assessment of the tensile strength of the embedded (and thus alkalineinfluenced) optical sensor fiber.

Page 202

Considering the stress concentrations on the fiber entrance and on its end in
the samples, it is obvious that the debonding loading determined by
micromechanical methods is lower than that of the adhesion based. On
account of these stress distributions the push-in test is predominant as a
sensor technique in the case of detection of the first fiber sensor debonding
phenomenon at component loading. The pull-out method reflects better the
sensor conditions in the vicinity of crack formation and propagation. One
should take into account that a direct comparison of the results of both
methods is problematic because of the different force implementation and
stress transfer in the sample.
Results from Experimental Adhesion Strength Investigations
An almost linear force-displacement relation (Hooke-like behavior) was found in
the beginning of both pull-out and push-in tests (figures 3 to 6). By using the
pull-out tests the onset of the debonding process between fiber and matrix is
hard to detect. However, the end of the debonding process can be found as a
significant drop in the force function. Figure 3 shows that the increased
hydrophobicity of silanized glass fibers (see also contact angles values, table
II.) does not lead to a weaker bond between both components. It should be
noticed, first, that bare optical glass fibers also show considerable adhesion
strength in the cement paste matrix and, second, that no tensile fracture
occurred in the embedded part of the fiber. As could be seen by SEM, bare
optical glass fibers exposed to pore solution for several days have a relatively
smooth surface. Thus, mechanical interlocking should be excluded as the
reason for the increased adhesion strength. Corresponding to the pull-out load
applied to the fiber the tensile strength of decoated fibers embedded in
cement paste is, at least, 125 MPa. This value corresponds to an ultimate
strain of larger than 0.18%. Regarding the tensile failure strain of cement
paste (0.02 0.06%) embedded optical fiber sensors or long-gage length sensor
fibers survive the deformation of uncracked cementitious materials.
Figures 4 to 6 show force-displacement curves obtained by push-in tests. In
the first part of the curve the linear slope is determined by the elastic
deformation behavior of the composite corresponding to Hooke's law. The
following non-linear part shows first debonding events between the fiber and
the cement mortar matrix, reflecting a change of the compliance. The onset of
this behavior clearly detected for polyimide-coated fibers (figure 4) and

determined by extrapolation should be defined as the actual limit of strain

measurement for embedded fiber sensors. Strain transfer begins to fail in this
force range. The comparison of figures 4 to 6 also shows that the elastic strain
transfer range is greater for decoated and considerably greater for silanetreated fibers. Figure 4b shows a clear separation of the polymer coating from
the cementitious matrix. Although polyimide is known as chemically inert
against organic solvents and dilluted inorganic acids and bases, cracks and
disruptions were observed at the coating surface, as described earlier [2]. It is
worthwile mentioning that the push-in test arrangement could not introduce
sufficient force to achieve the debonding stage for decoated and silane-treated
fibers. The increased surface polarity of such optical fibers results in an
enlargement of the attractive interactions at the interface so that a higher
bond strength have been observed.

Figure 3
Pull-out test results for decoated optical fibers embedded in Portland
blast-furnace cement mortar (w/c ratio: 0.68, grain size: 0.08 0.16 mm)
1 - oxygen plasma modified and silanized fiber, 2 and 3 - untreated bare fibers

Figure 4
Push-in test results for a polyimide-coated optical fiber embedded in
Portland cement paste (w/c ratio: 0.3)

Figure 5
Pushin test results for decoated untreatet optical fiber embedded in Portland cem. paste (w/c ra

Page 204

Figure 6
Push-in test results for decoated oxygen plasma treated
and silanized (concentr.:0.5%) optical fiber embedded
in Portland cement paste (w/c ratio: 0.3)

Conclusions
We were looking for ways to achieve a reliable strain transfer throughout the
sensor area of short-gage length fiber-optic micro strain sensors as well as
long-gage length fiber sensors to be embedded in cementitious materials. In
order to pre-evaluate the bonding behavior which can be expected at the
interface glass/cementitious matrix we used a method to characterize the
wettability and for predicting the adhesion behavior of differently treated
fibers. On the other hand, we measured the bond strength of differently
treated and subsequently embedded optical fibers by micromechanical
methods.
The determination of contact angles allows an estimation of the bonding
behavior of well-aimed treated surfaces of optical fibers prepared for
embeddment which could be confirmed by using the Wilhelmy Method. Based
on these investigations, the desired adhesion between fiber-optic materials and
cementitious matrices, depending on the measurement task, can be controlled
by specific physical and/or chemical modification of the glass surface. All
results obtained from micromechanical adhesion tests were consistent with the
results obtained by surface-energetic investigations. It could be ascertained
further that decoated glass fibers embedded in freshly poured cementitious
mortars survived the hardening process and all loading cycles. They have a
higher adhesion strength compared to polymer-coated glass fibers; this
behavior is due to the increased surface polarity of bare optical fibers.
In order to make optical fibers for strain transfer purposes more handable
different micro-layers were coated or deposited onto the fiber surface and their

adhesion behavior has been compared with that of bare fibers and common
polymer (polyacrylate and polyimide) coating surfaces. In this paper, we
reported on optical fibers silanized with two specially chosen silanes of the
group of hydrophobic agents commonly used for building protection. The
increased hydrophobicity of these treated optical fibers lessen the alkaline
attack onto the fiber surface and does not diminish the adhesion behavior.
These properties could be concluded from exem-

Page 205

plaric tensile strength investigations of fibers exposed to pore solution (pH =

13,1) and from fiber pull-out and push-in tests of Portland cement mortars and
pastes. This behavior can be explained by measurement of the fiber's contact
angles as a direct measure of its wettability. The silanized fibers have shown to
have a rest activity of non-reacted functional groups. These groups react with
the cementitious matrix and therefore the increased adhesion strength is due
to chemical bonds between both phases. To prove these results, research is in
progress.
We could clearly show that an oxygen-plasma pre-treatment (< 5 min) of bare
optical fibers to design embeddable sensors is necessary to achieve the
required density of a silane layer. An investigation by SEM confirmed that there
is no possibility to achieve a closed silanole layer without plasma treatment. It
should also be concluded that an activation of thin polymer networks used as
coatings will improve the adhesion behavior considerably.
Based on these results, even bare fiber-optic micro strain sensors were
embedded in different real mortars and they all survived. Such highly
resolvable sensors were used to measure reliably and reproducably the micro
deformation behavior of special grouts and of Portland cement mortar
depending on the w/c-ratio, from the beginning of the hydration reaction and
subsequently over a time range of several weeks [1].
Acknowledgement
The investigations were partially supported by the BMFT of the Federal
Republic of Germany and the Berlin Senate Administration for Science and
Research. This support is greatly appreciated. The authors would also like to
thank Mrs. Egia Ajuriagojeaskoa (TU Berlin) as well as Mrs. Bistritz and Mr.
Schumacher (both at BAM) for their skill in the preparation of the fiber samples
and the test arrangements. Two of the authors (W.R.H. and A. B.) wish to
express their thanks to Prof. B. Hillemeier and Prof J. Springer (both at TU of
Berlin) for their continuous encouragement.
References
1. Habel, W. R., B. Hillemeier, M. Jung; J. Plhn and F. Basedau, 1998. ''NonReactive Measurement of Mortar Deformation at Very Early Ages by Means of
Embedded compliant Fiber-optic Micro Strain Sensors" will be published at the
ASCE '98 Engg. Mechanics Conf., San Diego. May 1998

2. Habel, W. R. and H. Polster, 1995. "The Influence of Cementitious Building

Materials on Polymeric Surfaces of Embedded Optical Fibers for Sensors."
Journal of Ligthwave Technology. IEEE Piscataway, N. J. USA, 13(7):1324-30.
3. Song, B. and J. Springer, 1996. "Determination of Interfacial Tension from
the Profile of a Pendant Drop Using Computer-Aided Image Processing." part
2: Experimental. Journal of Colloid Interface Science 184, 77-91
4. Wu, S. 1982. Polymer Interface and Adhesion, M. Dekker Inc. New York,
Basel.

Page 206

5. Brochure 1993. Application of organofunctional silanes. HLS AG.

6. Egia Ajuriagojeaskoa, E., A. Bismarck, J. Springer and W. R. Habel. 1998
"Surface-energetic investigation of modified and silanized glass fibers".
(Original in French) to be published in Le Journal de Chemie physique.
7. Majumdar, A. J. and V. Laws. 1991. Glass fiber reinforced cement. BSP
Professional Books, London
8. Nanni, A., C. C. Yang, K. Pan, J. Wang, and R. R. Michael, 1991 "Fiber optic
sensors for concrete strain/stress measurement." ACI Mater J., 88, 257264.
9. Hampe, A., G. Kalinka, S. Meretz, and E. Schulz, 1995. "An advanced
equipment for single-fibre pull-out test designed to monitor the fracture
process." Composites 26, 4046.
10. Kalinka, G., A. Leistner, and A. Hampe, 1997. "Characterisation of the
fibre/- matrix interface in reinforced polymers by push-in technique."
Composite Sci. and Techn. 57, 845851.

Page 207

Chapter 6
Sensor Transduction Mechanism and Signal Recovery

Page 209

Intensity Fiber Optic Sensors for Civil Infrastructures

F. Casciati,
S. Merlo
and G. Zonta
Abstract
An optical fiber device for bridge monitoring is presented: it pursues the
sensing of traffic loads, such as the passage of a heavy truck.
In the present proposal an intensity device especially conceived for monitoring
steel bridges is described. The working principle is illustrated and a reference
example is conducted by monitoring the behaviour of a steel beam. A testing
machine applies vertical loads in different sections of the steel beam, thus
simulating the passage of a heavy truck over the bridge. The results are
discussed and compared to the theoretical values. The actual monitoring of an
existing bridge is also discussed.
Introduction
In the civil engineering field, buildings, bridges, dams and large structures in
general represent an enormous financial investment. Embedded optical fiber
sensors [1, 2, 3, 4, 5] could improve the diagnostic potential toward aging and
deterioration of highway bridges. This is recognized to be one of the major
problems which must be presently solved by structural engineers in the United
States, in the European Community and in Japan. A structurally integrated
sensing system could monitor the state of the structure throughout its working
life. It could record deformation, load distribution and environmental
degradation experienced by the structure [4]. Monitoring external loads
applied to steel bridges can provide a very useful information to owners,
operators and designers of the structure. Since comprehensive load
measurements are often quite difficult to obtain, expecially without the help of
modern devices, most of the loads that structures experience are known
imprecisely. Thus, design and rating
Fabio Casciati, Department of Structural Mechanics, University of Pavia, 27100
Pavia, Italy
Sabina Merlo, Department of Electronics, University of Pavia, 27100 Pavia, Italy
Gianluca Zonta, Department of Structural Mechanics, University of Pavia, 27100
Pavia, Italy

Page 210

decisions are made with incomplete and uncertain information. If the external
loads were measured, design and rating procedures could be improved as well
as active structural control could be introduced [6, 7, 8].
A part from dead loads, the action of heavy trucks on bridges and the
distribution of the loads in suspension and cable-stayed bridges are worthwhile
to measure. Such measurements can increase site safety and reduce
construction costs. Fiber optic sensors could possibly be more durable; they
already show less expensive cabling requirements than their conventional
counterparts. Moreover, these sensors are intrinsically immune to electromagnetic interferences (EMI), which is a very interesting feature when they
are applied close to high voltage lines, as in the case of railway bridges.
In order to monitor the load condition (by measuring the induced state of
strain), the authors have designed and implemented a fiber optic intensity
sensor, characterized by robustness and reliability, even when working in the
real structure.
Sensor Design
Since the typical deformation of a bridge is characterized by a large diameter
curve, the intensity sensor has been developed introducing a localized loss,
related to the curvature of the deformed shape. The sensor mechanism is
based on the coupling loss, that is the power loss arising when two optical
fibers (or a fiber and a source) are misaligned. The sensor structure is
illustrated in Fig. 1, and consists of two multimode optical fibers aligned in a Vgroove cut in a proper substrate. The fibers are previously introduced in glass
micro-pipets to increase fiber stiffness, and then glued in the grooves such that
one of them is free to describe an angle, representing the tangent to the
bending shape. When the structure looses its straight shape (under load),
power loss increases, as a fraction of the incident light is no more captured by
the output fiber core. As the deformation increases, the bending induced
optical losses increase.
To implement the sensor in a beam, a simple device was conceived. It shows
the following characteristics [9]:
A thin plate (h = 3mm) of rectangular shape, with V grooves over the upper
surface, is prepared. The selected material for the plate must ensure an
adequate stiffness value and a suitable temperature insensitivity.

The fibers were introduced in glass micro-pipets and there fixed to each other,
using a special glue. The length of each pipet was assumed equal to 4 cm. The
two pipets, with the fibers, were then placed inside the groove on the plate,
facing each other (see Fig. 1). At the end of the alignment operation, the
pipets (and the fibers) are glued into the groove. To increase sensitivity, one
capillary tube was glued along all its length, while the other one was partially
fixed, spreading the glue along a line of length 2.0 cm. This assembling phase
was entirely developed under a microscope.

Page 211

Figure 1.
Outline of the optical device

Optic Relations
The sensor behaviour has been framed in a theoretical model, which
represents the transmission of the light wave through the coupling mechanism.
The theoretical model can be applied to assess the efficiency h, defined as

where Pin is the total optical power carried by the input fiber, and Pout is the
power that is coupled into the output fiber. The model considers multimode
optical fibers with a step-index refractive index profile. This kind of fiber has a
spatially constant numerical aperture N A and also a constant intensity, if
uniform power distribution among the modes is ensured. Under these
conditions, and neglecting losses due to glass-air interfaces, the coupling
efficiency can be calculated as the ratio between the emitting area Ain and the
effective receiving area Aout [10]. For example, if the misalignment can be
modelled with an off-set between the fiber axis, the efficiency is the ratio of the
core-overlap area to the core cross section.
In assesing the area ratio, three different effects are distinguished:
1. tilt angle g of the unbound fiber section;
2. axial fiber off-set, w;
3. spacing between fiber end, z.
In particular, one can relate the second and third effect to the tilt angle g,
which is imposed by the deformation. By considering the designed sensor
structure

Page 212

Figure 2.
Behaviour of the relation h - g: normal line for z0 = 0,
dashed line for z0 = 50 m

under deformation, we obtain that the emitting area is given by:

where:

and rf is the fiber core diameter, while z0 represents the initial spacing
between the fiber ends. The receiving area is evaluated as the overlap area:

where:

(w0 is assumed positive for the deformation detected during the experimental
test). The diagram in Fig. 2 reports the relation between the efficiency
and the value of the tilt angle g for, initially, perfectly aligned fibers (z0 = 0)

and for z0 = 50m.

Page 213

Figure 3.
The global experimental set up

Laboratory Implementation
In the experimental testing of the developed sensor, the flexural behavior of a
steel beam was monitored [9]. For this paper, the tests were repeated by
moving the device to different positions along the beam. The experimental set
up consisted of:
1. A simply supported steel beam.
2. A testing machine to load the structure vertically (Fig. 3).
3. Multimode silica optical fiber (core diameter 50 m, overall diameter 250
m).
4. A LED used as a light source.
5. A photodiode acting as a light detector, with a transimpedence amplifier
acting as a light detector.
6. A personal computer, equipped with a data acquisition board and associated
software.
First the optical fiber sensor was placed on the lower edge of the beam, across
the middle section, so that the fiber interruption was coincident with the
maximum strain section. Global adhesion of the sensor device with the beam
was achieved by spreading a special glue for strain gauges (Rapid Adhesive
X60, Hottinger Baldwin Messtechnik) between the two surfaces.
The testing machine applied vertical loads to the beam in different longitudinal
positions, thus representing the passage of a heavy truck over the bridge. The
primary scope in this phase of the research was to conceive a system which
was able to recognize this kind of load. The device should enable to check the

passage of the truck along the deck, and to locate the load position. For this
purpose, the steel beam was loaded by a vertical load equal to 8 t., applied in

Page 214

Figure 4.
Experimental efficiency for different loads (sensor in section 5)

different sections starting from the right support. Fig. 4 shows the diagram of
the experimental efficiency h, obtained from the data collected during the
experimental testing, when the sensor is in the middle section.
At the moment, the values of h are not immediately comparable to the
theoretical values of h. Actually, h is obtained as

where Pout(def) is the optical power collected during beam deformation, and
Pout(undef) refers to the optical power in the undeformed condition. The
previously defined efficiency h is recovered by dividing the value of h by the
term
, which could be in principle determined during the sensor
assembling procedure. The term K is usually less than 1, because of residual
initial fiber misalignment.
The device was then removed and attached to the beam in a different position,
closer to the left support. The same loading was also performed in this second
case and results are drawn in Fig. 5. One can observe that the highest loss
value was obtained when the load section was nearest to the beam section
where the

Page 215

Figure 5.
Experimental efficiency for different loads (sensor in section 8)

device was applied (in agreement with the influence line theory). Furthermore
Figs. 4 and 5 show that sensor behavior does not depend on the section in
which it is applied, but only on the amplitude of the bending deformation.
Multiplexing
The possibility of demultiplexing (the subdivision of light wave from one
channel into several ones) was also investigated. It allows one to deal with
many sensors, placing each of them along a single fiber. The ability to
multiplex sensors in networks can be advantageous with regard to a number of
system aspects, including reduced component costs, ease of I/O interfacing
and overall system immunity to electromagnetic interference. The development
of efficient multiplexing techniques enhances the competitiveness of fiber
sensors compared with conventional technologies in most application areas.
The need of multiplexing when monitoring large infrastructures, is of
fundamental importance to facilitate the efficient interrogation of more sensors
that may be required to be distributed over a complex smart structure. The
handling of such large flows of sensor information is one of the key areas in the
development of advanced responsive structures.
The sensors can be deployed in either a linear or a star arrangement, as in Fig.
6, or a combination of the two can be used [1, 2]. These are equivalent to
linear and star buses commonly used for communications. With a star
configuration, the sensors can be positioned as desired to monitor a smart
structure,

Page 216

Figure 6.
Linear and star arrangements for multiplexing of intensity
based devices

using different lenghts of fiber delay line between the coupler and the sensors.
The star topology is probably more suitable for the proposed sensor. The
monitoring system is then architectured by a unique optical source, a multimode IXN fiber optic coupler, which divides the optical power into N branches
and N photodetectors. Thus, each photodetector collects the signal coming
from a single sensor placed in a specific bridge section. It is possible to sense
different sections of the structure, collecting information concerning how the
applied load deforms the structure in each one. Processing the output data
provides simple and useful information about the entity and the location of the
applied load.
Field Monitoring
The main features of the optical fiber monitoring system are conceived for a
direct implementation. When dealing with real structures, problems concerning
with environmental aspects arise, owing to harsh weather conditions and huge
dimensions. A steel bridge (Fig. 7) in the county of Pavia has been selected to
apply the designed monitoring system. The theoretical governing relations
allow one to settle the design parameters, such as the length L of free fiber, in
order to calibrate the sensor sensitivity. Such values depend on the curvature
of the deck of the bridge, thus involving the length of the spans, the structural
scheme and the material. When dealing with old bridges, it is usually difficult
and expensive to obtain some of the required information. The sensor design
requires a preliminary field inspection, in which conventional mechanical

transducers are

Page 217

Figure 7.
The bridge under inspection

applied to the structure. Three measurements, referred to different sections,

enable one to know the angular deformation g when the measurand load is
acting. The design of the fiber optic device is then conducted accordingly to
such data. When working in the field, particular attention must be paid to data
transmission. The conceived system consists of a local station equipped with
personal computers enabled to data acquisition. From this station, placed near
the monitored bridge, data are sent to a central operative station by either a
modem (if a telephone line is easily connectable from the local station) or a
radio connection.
Conclusions
A preliminary design of a system for the diagnostic monitoring of the integrity
of bridges is presented. The intent is to detect and locate damage or
degradation in structural components due to heavy loads and to provide this
information quickly and in a form easily understood by the operators. The
architectured optical fiber device performs such a capability by measuring the
power dissipated by a coupling mechanism. The adoption of the monitoring
system to a real structure is also introduced.
Acknowledgements
This paper was supported by a grant from the Italian Research Council (CNR),
contract # 96.05415.ST74, with Prof. E. Benvenuto of the University of Genoa
as national coordinator. The authors also acknowledge the cooperation of the
Civil Protection Office of the Prefecture of the town of Pavia.

Page 218

References
[1] J. A. Buck, 1995. Fundamental of Optical Fibers, John Wiley & Sons, Inc.
[2] J. Dakin, B. Culshaw, (eds.) 1988. Optical Fiber Sensors: Principles and
Components, Vol. 1, Artech House, Boston & London.
[3] D. R. Huston, Smart Civil Structures - An Overview, Proc. SPIE - Int. Soc.
Opt. Eng., 1588, 21
[4] E. Udd, (ed.) 1995. Fiber Optic Smart Structures, John Wiley & Sons, Inc..
[5] A. Wang, K. A. Murphy, 1992. Smart Materials and Structures, pp. 57.
[6] K.P. Chong, S.C. Liu, and J.C. Li, 1990, Intelligent Structures, Elsevier
Applied Sciences, New York.
[7] G.W. Housner, S.F. Masri, F. Casciati and H. Kameda, 1992, Structural
Control and Intelligent Systems, USC Pubblication CE 9210.
[8] Y.K. Wen, 1992. Intelligent Structures 2, Elsevier Applied Sciences, New
York.
[9] F. Casciati, S. Merlo, G.L. Zonta, 1997. Bridge Monitoring by Optical Fiber
Device. Intelligent Renewal of CIS (F. Casciati et al., eds.), World Scientific,
Singapore.
[10] H.G. Unger, 1977. Planar Optical Waveguides and Fibres, Clarendon
Press, Oxford.

Page 219

A New Signal Recovery Scheme for Fiber Bragg Grating Sensors

F. Farahi,
E. V. Diatzikis,
L. A. Ferreira
and J. L. Santos
Abstract
A simple scheme for measuring change in wavelength of fiber Bragg grating for
sensing application is presented. In this scheme an electrical carrier is
produced by using a frequency modulated multimode laser diode to excite the
fiber grating. The change in Bragg wavelength is measured by tracking the
phase of the carrier at the detector output in either an open or a closed loop
scheme. A theoretical analysis of the interrogation technique in terms of
linearity and dynamic range is presented. Experimental data were obtained for
both strain and temperature measurements. Sensitivities of 0.7 e/Hz and
0.05 C/Hz were obtained over a dynamic range of 60 dB.
1. Introduction
Fiber Bragg gratings are simple, low size intrinsic sensing elements which can
be written in silica fibers [1,2] and have all the advantages normally attributed
to fiber sensors. In addition, these devices are inherently self referenced and
can be easily multiplexed [3], which is particularly important considering the
increasing interest in distributed embedded sensing [4,5]. In recent years a
large number of wavelength detection techniques for fiber Bragg grating
sensors have been developed. These make use of interferometric techniques,
wavelength dependent filters, tunable filters, laser/sensor combinations and
others [616]. We have previously presented a pseudoF. Farahi and E. V. Diatzikis, Physics Department, University of North Carolina at
Charlotte, Charlotte, NC 28223
L. A. Ferreira and J. L. Santos, Departamento de Fsica da Faculdade de Cincias da
Universidade do Porto, Rua do Campo Alegre, 687, 4150 Porto, Portugal.

Page 220

heterodyne demodulation technique for fiber Bragg sensors using two matched
gratings, one acting as a sensor and the other for signal processing. In that
scheme, the processing grating was periodically stretched to generate an
electrical carrier at the detector output whose phase was directly related to the
Bragg wavelength difference between the two gratings [17]. However, low
dynamic range, low processing frequency and operation in open loop, were
intrinsically associated with that technique due to limitations in stretching the
processing grating. In other words, in order to produce a carrier, a strain ramp
signal with large amplitude must be applied to the processing grating to
completely sweep the sensing grating over the entire measurement range. This
process imposes practical constraints to the system design.
In this paper, we describe a system where the carrier is generated by
stimulating the sensing grating with a ramp modulated multimode laser diode
light source. It is shown that the phase of such a carrier is linearly dependent
on the difference between the Bragg wavelength and the laser mode
wavelength and is modified by external effects such as strain and temperature
applied to the sensing grating. By locking the system at a specific phase value,
measurements for temperature and strain can be performed in a closed loop
operation. Experimental results are given which demonstrate the applicability
of the proposed demodulation technique.
2. Principle
Consider the optical sensing system in Fig.1a which is illuminated with a
multimode laser diode. If one of the laser modes coincides within the spectrum
of the fiber Bragg grating, reflected light associated with that laser mode will
be detected at the output. The intensity of the signal at the detector is
proportional to the overlap integral of the functions f(l - l 0) and g(l - l B),
representing the spectral characteristics of one mode of the laser and the fiber
Bragg grating, respectively. Assuming both functions are Gaussian, we can
write

and

where P0 is the optical power of the laser mode being reflected, Dl m is the
spectral

Page 221

width at half maximum and l 0 is the central wavelength; R is the grating

maximum reflectivity (0 < R < 1), l B the Bragg wavelength and Dl B the
grating spectral width at half maximum. If the mode spacing of the multimode
laser is much greater than the grating spectral width, then at any time only
light associated with one mode may be affected by the presence of the grating
Therefore, the intensity of the signal at the detector can be written as
(assuming a nominal coupling factor of 1/2):

In general, the spectral width of the grating, Dl B, is much larger than the
spectral width one mode of the laser, Dl m, and the above integral becomes:

where Dl = l 0 - l B is the difference between the fiber grating and the laser
mode central wavelengths. Clearly, this integral is maximum when Dl = 0 and
it will approach zero when Dl is increased and becomes larger than Dl B. We
use this concept to produce a carrier at the detector output (see Figs.1b-e).
By driving the laser with a sawtooth waveform, a time dependent change in
the central wavelength of the laser mode is produced. In this case, Dl in Eq.4
becomes a function of time and therefore the output intensity is time
dependent. The signal in the photodetector may be band-pass filtered to
produce a sinusoidal waveform, i.e.,
where w = 2p/T is the angular frequency at which the laser is modulated, k is
a detection coefficient and A and q are Fourier coefficients, given by

where

Let's assume that the applied ramp signal is perfect, i.e., there is no flyback
associated with it. If the amplitude of modulation is large enough to sweep l 0
over the entire spectrum of the fiber Bragg grating, that is, Dl 0 >> Dl B, the
integrals in (8)

Page 222

can be evaluated for t going from zero to infinity. In that case, the expressions
for the amplitude and phase of the filtered signal become:

Figure 1:
Illustration of the proposed
concept: (a) basic optical configuration; (b)
functions g(l-l B) and f(l-l 0) that is tuned
by a sawtooth wave applied to the laser; (c)
applied waveform; (d) typical signal; (e)
phase shifted signal.

This situation is illustrated in figures 2 (a) - (d). Fig.2a and 2b show a typical
output signal and its corresponding first harmonic respectively. Fig.2c gives the
relation between the first harmonic amplitude and the Bragg wavelength and
Fig.2d gives the relation between the phase and the Bragg wavelength. Since

q is linearly dependent on the Bragg wavelength it must also be a linear

function of temperature and applied strain. The amplitude, as expected, does
not depend on the Bragg wavelength. Another obvious advantage of this
technique arises from the fact that the performance of such a sensor is not
affected by the spectral structure of the sensing fiber Bragg grating. In fact,
the analysis described fir the Gaussian shape would be fully applicable for a
grating with imperfect spectral features. In these calculations we have used
the following values which closely represent our experimental parameters; R =
0.9 Dl B = 0.15 nm, l 0 = 1300 nm, P0 = 2 mW and T= 1 ms. Since the phase
linearly depends on the Bragg

Page 223

wavelength, closed loop and open loop operations are both applicable.
3. Experiments and Results

Figure 2:
a) Output signal for
l B = 1300.3 nm; b) corresponding
first harmonic, c) its amplitude variation,
and d) its phase change when the Bragg
wavelength is modified. We have
assumed Dl 0 = 0.6 nm >> Dl B.

Fig.3 shows the experimental set-up used to demonstrate the proposed signal
processing scheme. A fiber pigtailed multimode laser diode (FUJITSU
FLD130C2LK/352) with the central mode of wavelength 1318 nm was used to
illuminate the sensing fiber Bragg grating through a 3 dB coupler, C1. The
separation between adjacent longitudinal modes of the laser was measured to
be 0.75 nm. This is much greater than Dl B = 0.15 nm thus justifying our

intensity calculation (Eq. 2) to include only the spectral characteristics of one

laser mode only, although we use multimode laser as a light source. The
temperature was tuned and the bias current was adjusted such that one of the
lateral modes coincided with the fiber grating (the current tuning capability of
the laser was 11.3 pm/mA). The laser wavelength was modulated by a
sawtooth signal at 1 kHz. The spectral characteristics of the grating were
analyzed at room temperature, and with no applied axial strain we recorded a
reflectivity of 60%, at l B = 1304.4 nm and dl 0.2 nm. Reflected light from
the grating was sent to the detection unit

Page 224

through coupler C1. Spurious reflection along the optical system were
minimized by applying index matching gel to the fiber ends. Light coming
directly from the laser to the coupler output was used to compensate the effect
of intensity modulation of the light injected into the system. This intensity
modulation comes from the unavoidable residual fraction of the entire laser
spectrum reflected at the fiber ends and by the splices. The effect of this
intensity modulation is to introduce measurement errors which can be
important considering that its strength can be comparable or even higher than
the signal power that comes only from the laser mode reflected by the grating.
Here we should emphasize that this ''broadband" intensity modulation does not
depend on the grating position.

Figure 3:
Experimental configuration.

A typical signal output as seen in the scope is shown in Fig.4. To illustrate the
demodulation principle, Fig.4 also shows the change in the output signal when
strain is applied to the sensing grating and the system is operating in open
loop.
The analysis given in section 2 indicated that there are advantages to
operating the configuration in a close loop format. Therefore, to achieve this,
the detector signal was sent to a lock-in amplifier to measure its phase relative
to the sawtooth wave. The output of the lock-in amplifier was then sent to a
servo circuit to keep the phase at a fixed value. The change in Bragg
wavelength can thus be measured by monitoring the

Page 225

laser bias current. Figure 5 illustrates the effect of the servo in maintaining the
phase constant.

Figure 4:
Sawtooth waveform applied to the laser diode (above) and the
corresponding output carrier for two different values of applied strain
(below).

Figure 5:
Illustration of the effectiveness of the servo operation: lock-in output when
the sensor head is subjected to abrupt temperature and strain variations
(for t < 90 s the servo is off; for t 90 s the servo is on).

To evaluate the system bandwidth, a sinusoidal signal with small amplitude

was added to the laser diode injection current. With the loop closed and
without strain or temperature applied to the grating, the integrator output
(error signal) was then observed for different frequencies. The result is
presented in Fig.6 and indicates a bandwidth of 20 Hz. To some extent, this
value is controllable by the lock-in time constant.

Page 226

Fig.7 shows the system response to applied temperature (7a) and strain (7b).
Data in this figure indicate linear behavior. From the spread of the data and
the bandwidth of the feedback loop, sensitivities of 0.05 C/Hz for
temperature measurements and 0.7 e/Hz for strain measurements were
obtained.

Figure 6:
System's frequency response.

Figure 7:
System response for (a) temperature change and (b) applied strain.

Page 227

Fig.8a shows the system response to strain steps of 30 e, at a period of 20

seconds, during a time interval of 100 seconds. It illustrates the stability of the
system readout. Also, from the signal to noise ratio apparent in the figure, a
static strain sensitivity of 0.5 e/Hz was obtained, in good agreement with the
result derived from Fig.7b. Fig.8b demonstrates the system insensitivity to an
extreme change in optical power. For a fixed strain the optical power was
varied by 2/3 of its initial value with no detectable change in the signal output.
On the other hand, the same system exhibits considerable change to an
applied strain of 30 e. The variation in the system optical power was
achieved by bending the input fiber. AC strain measurement was performed by
applying a sinusoidal strain signal with an amplitude of 0.26 e at a frequency
of 200 Hz to the fiber grating using a piezoelectric fiber stretcher. Since the
frequency of the applied strain is beyond the system bandwidth, the induced
signal was monitored at the photodetector output using an electrical spectrum
analyzer and a sensitivity of 2 0.7 e/Hz was derived.

Figure 8:
Sensor response to (a) strain steps (b) variation in the optical power (in both cases
De = 30 e).

The measurement range of this system is limited primarily by the current range
of the laser diode. A dynamic range of 60 dB was obtained for temperature
and strain measurements. One advantage of using a multimode laser over a
single mode laser is the possibility of switching between adjacent modes in
order to track the fiber Bragg grating while keeping the bias current within a
safe limit. This capability was experimentally demonstrated and is illustrated in
Figs.9a and 9b. Fig.9a shows that the system was initially locked to a laser
mode of wavelength l OP. Then, the laser

Page 228

injection current was increased by an amount that shifts a laser mode by 0.75
nm, the modal separation of the laser used in our experiments. This means the
lower adjacent mode relative to the mode of wavelength l OP approaches the
grating Bragg wavelength, and may be locked to this wavelength by the servo.
Later, the laser current returns to its initial value and the system is locked
again in the mode of wavelength l OP. Fig.9b illustrates the same phenomenon
but now the transition to the upper adjacent mode is considered. The
transitory behavior shown in the edges of the current waveform are due to
temperature effects in the laser spectrum.
Another significant advantage of using a multimode laser diode is that optical
isolation is not required, which is crucial when single mode illumination is
considered. It was found that the lateral modes of the laser spectrum were
rather stable even in the presence of strong back reflection, which has a large
impact in the more central modes. When compared with these strong central
modes, the power of the lateral modes is much smaller. However, because the
laser mode width is much smaller than the grating spectral width, the average
return power reflected by the grating is still comparable with the situation
where the fiber grating is illuminated by a broadband source. Additional
advantages of this technique is its potential to be applied in multiplexed
sensing and the use of low cost light source.

Figure 9:
The system response when switched between the two modes of wavelength l OP
utilized in the experiment and one of its adjacent modes: a) for the lower mode; b)
for the upper mode.

Page 229

4. Conclusion
We have demonstrated a signal processing scheme to measure wavelength
shifts in optical fiber Bragg gratings based on the generation of an electrical
carrier via modulation of a multimode laser diode. A detailed theoretical
analysis for this modulation technique was performed. The scheme has the
advantages associated with these type of sources, such as high power injected
in the system, low cost and relative immunity to back reflected power.
Insensitivity of this system to optical power fluctuations was demonstrated and
an AC strain resolution of 2 e/Hz at 200 Hz was obtained in a closed loop
operation. It was shown that the dynamic measurement range for strain and
temperature (60 dB) can be increased by using more than one of the laser
modes.
Acknowledgments:
L. A. Ferreira acknowledges financial support from "Programa PRAXIS XXI".
This work was in part supported by NSF grant DMI-9413966.
References:
[1] Meltz G., Morey W.W., and W.H. Glenn W.H., 1989 "Formation of Bragg
gratings in optical fiber by a transverse holographic method", Opt. Lett., 14,
823.
[2] Hill K.O., Malo B., Bilodeau F., Johnson D.C, and Albert J., 1993 "Bragg
gratings fabricated in a monomode photosensitive optical fiber by UV exposure
through a phase mask", Appl. Phys. Lett., 62, 1035.
[3] Rao Y.J, Lobo Ribeiro A.B., Jackson D.A, Zhang L., and Bennion I., 1996
"Simultaneous spatial, time and wavelength division multiplexing in-fiber
grating sensing network", Opt. Commun., 125, 53.
[4] Fuhr P.L., Huston D.R., Kajenski P.J., and Ambrose T.P., 1992
"Performance and health monitoring of the Stafford medical building using
embedded sensors", Smart. Mater. Struct., 1, 63.
[5] Measures R.M., Alavie A.T., Maaskant R., Ohn M., Karr S., and Huang S.,
1995 "A structurally integrated Bragg grating laser sensing system for a carbon
fiber prestressed concrete highway bridge", Smart. Mater. Struct., 4, 20.

Page 230

[6] Brady G.P., Hope S., Lobo Ribeiro A.B., Webb D.J., Reekie L., Archambault
J.L, and Jackson D.A., 1994 "Demultiplexing of fiber Bragg temperature and
strain sensors", Opt. Commun., 111, 51.
[7] Zhang Q., Brown D.A., Kung H., Townsend J.E., Chen M., Reinhart L.J.,
and Morse T.F, 1995 "Use of highly overcoupled couplers to detect shifts in
Bragg wavelength", Electron. Lett., 31, 480.
[8] Ferreira L.A., and Santos J.L., 1996 "Demodulation scheme for fiber Bragg
sensors based on source spectral characteristics", Pure Appl. Opt., 5, 257.
[9] Koo K.P., and Kersey A.D., 1995 "Bragg grating-based sensors systems
with interferometric interrogation and wavelength division multiplexing", J.
Lightwave Tech., 13, 1243.
[10] Volanthen M., Geiger H., Xu M.G., and Dakin J.P., 1996 "Simultaneous
monitoring of multiple fiber gratings with a single acousto-optic tunable filter",
Electron. Lett., 32, 1228.
[11] Coroy T., and Measures R.M., 1996 "Active wavelength demodulation of a
Bragg grating fiber optic strain sensor using a quantum well electroabsorption
filtering detector", Electron. Lett., 32, 1811.
[12] Ferreira L.A., Lobo Ribeiro A.B., Santos J.L., and Farahi F., 1996
"Simultaneous measurement of displacement and temperature using a low
finesse cavity and a fiber Bragg grating", Photon. Tech. Lett., 8, 1519.
[13] Hjelme D.R., Bjerkan L., Neegard S., Rambech J.S., and Aarsnes J.V.,
1997 "Application of Bragg grating sensors in the characterization of scaled
marine vehicle models", Appl. Opt., 36, 328.
[14] Flavin D.A., McBride R., and Jones J.D.C., 1997 "Short optical path scan
interferometric interrogation of a fiber Bragg grating embedded in a
composite", Electron. Lett., 33, 319.
[15] Kersey A.D., Davis M.A., Patrick H.J., LeBlanc M., Koo K.P., Askins C.G.,
Putnam M.A., and Friebele E.J., 1997 "Fiber Bragg grating sensors", J.
Lightwave Tech., 15, 1442.
[16] Rao Y.J., 1997 "In-fiber Bragg grating sensors", Meas. Sci. Technol., 8,
355.
[17] Ferreira L.A., Santos J.L., and Farahi F., 1997 "Pseudoheterodyne

demodulation technique for fiber Bragg grating sensors using two matched
gratings", Photon. Tech. Lett., 9, 487.

Page 231

Elastica Fiber Optic Sensors for Structural Monitoring

K. H. Wanser,
K. F. Voss
and K. R. Francis
Abstract
In this paper the operational characteristics of Elastica fiber optic sensors are
presented. In particular the sensor shape function, transmission-displacement
scaling relationship as a function of sensor length, and reflection mode
operation are described. Sensors of all lengths studied are found to exhibit
large regions of linear response. A 5 mm length Elastica exhibits a 16 bit
resolution-limited displacement sensitivity of 6.4 nanometers and a gauge
factor G = 14.5 in transmission. A high-cycle sensor fatigue test at large strain,
the use of these sensors in structural monitoring experiments, and some of
their unique features are also discussed.
Introduction
Recently we have introduced a new class of fiber optic strain-displacement
sensors which are based on precisely controlled nonlinear buckling of optical
fibers and the resulting optical bending loss (not microbending) [1]. These
sensors, named Elastica [2], offer many advantages over other fiber optic
sensors due to their simplicity, low-cost, large strain range, linearity, high
fatigue cycle endurance, high temperature capability, extremely low force
constant, flexibility in mounting method, and low thermal apparent strain. The
all-fiber sensors of this type are capable of detecting displacements ranging
from a few nanometers to several millimeters. In addition, both unidirectional
and omnidirectional responses have been demonstrated, the latter being
particularly useful in the detection of cracks [3]. Elastica fiber sensors can be
interrogated in both reflection and transmission modes, with either continuous
or pulsed light, depending on the desired application. In this paper we
describe some of the novel properties of Elastica fiber sensors which make
them especially suitable for use in structural monitoring applications.
Keith H. Wanser, Department of Physics, California State University Fullerton,
Fullerton CA 92634
Karl F. Voss, vossco, 12716 87th St., Kirkland, WA 98034
Kurt R. Francis, Corning OCA, Garden Grove, CA 92841

Page 232

Shape Function for Uniaxial Elastica

The Elastica fiber sensor introduced in recent work is based on the precisely
controlled bending or buckling of an optical fiber into the shape shown in figure
1a and the resulting optical bending loss. A highly reproducible shape of the
fiber is obtained by application of forces and/or torques at two locations on a
freely suspended length of optical fiber, which achieves precise control of
transition and macro bending losses. The fiber is unconstrained laterally
between the two locations of application of forces and/or torques. In the
uniaxial Elastica sensor configuration shown in figure 1a, the boundary
conditions on the clamped regions are that the lateral displacements and the
slopes are both zero at these locations (A and E in figure 1a), so-called
clamped-clamped or "mortised" boundary conditions. One way to enforce these
boundary conditions is to house the fiber in short, small diameter tubes, which
can then be held in numerous ways to achieve the desired fiber configuration.
There are numerous Elastica shapes that can be implemented for a variety of
fiber sensing configurations, each with unique properties. One of the significant
advantages of working with fiber Elastica is that the fiber shapes are precisely
defined mathematically and accurately repeatable experimentally. Thus the
Elastica are well suited to theoretical modeling of optical effects and
comparison of experiment with theory.

Figure 1.
a.) Schematic of uniaxial type of Elastica fiber sensor. The distance between
the two clamping points A and E is varied. b.) Shape function for various amounts

of compression. Successive increments are 10%.

Page 233

Examples of the shape function y(x) describing the lateral displacement of the
fiber from the axis as a function of the distance x measured from one clamping
point of the Elastica sensor, in units of the straight fiber length Lo, are shown
in Figure 1b. Each successive curve corresponds to an increase in compression
of 10% relative to the straight fiber length Lo, the curve with greatest Xmax
corresponding to a 10% compression. Note that both coordinates of the fiber
are scaled by the straight fiber length Lo, emphasizing that the shapes of the
curves are universal for a given fractional amount of compression or strain,
independent of the straight fiber length. Furthermore, the shapes of the fiber
are independent of the elastic modulus or diameter of the fiber, an important
property that minimizes temperature and fiber coating dependent effects.
Optical Transmission Response
The optical transmission T vs. displacement Dx for an Lo = 10 mm Elastica fiber
sensor employing Corning 50/125 m multimode graded index fiber with a
numerical aperture of 0.20, operating at a wavelength of 840 nm is shown in
figure 2. As the figure shows, there is a flat region of almost no response until
a threshold displacement of DLT = 675 m. At slightly larger displacement
there is a highly linear sensing range between 1 and 2 mm, in fact the sensor
linearity in this region is within 2 10-3, and appears to be limited by
nonlinearity in the micropositioner used to produce the displacement. At larger
displacements, the transmission response is non-linear, and the sensor
transmission begins to saturate. While the useful sensing range shown in the
figure is from 0.6 mm to 3.5 mm displacement, corresponding to a strain of
31%, the Elastica fiber sensor is capable of surviving even larger strains. Note
that the sensor transmission changes by -7.2 dB for a displacement of 3 mm.

Figure 2.

a.) Transmission vs. displacement for Lo = 10 mm Elastica. b.) Linear region

least square fit. Intersection of dashed lines indicates threshold
displacement DLT.

Page 234

The maximum difference in transmission between corresponding compression

and expansion displacements, when cycled over the entire range of 3.5 mm,
was within 3 10-3, demonstrating excellent repeatability and low hysteresis,
in spite of the soft polyacrylate fiber coating used. Even lower values of
hysteresis can be obtained by employing higher modulus and less hysteretic
fiber coatings such as polyimide or gold.
The linear sensing region of the 10 mm Elastica has a slope of dT/dx = -3.325
10-4/m, a gauge factor G = (LG/T)|dT/dx| = 3.93 and a linear sensing
range of 6% strain in either direction about a bias point of 1.5 mm
displacement (assuming a sensor gauge length LG = 8.5 mm). This compares
favorably with conventional resistance strain gauges and fiber Bragg strain
sensors, which have nominal gauge factors of 2 and 0.784 respectively and a
considerably smaller sensing range. With the above slope of the transmission
response, the 16 bit resolution-limited displacement sensitivity of the Lo = 10
mm Elastica fiber sensor is 46 nm in the linear region, which corresponds to a
strain resolution of 5.4 strain for an 8.5 mm sensor gauge length.
The optical transmission T vs. displacement Dx curves of four similar Elastica
fiber sensors of lengths Lo = 5, 10, 15, and 20 mm are shown in figure 3. As
can be seen from the figure, the displacement sensitivity and sensing range
can be controlled by the straight fiber length Lo. The curves in figure 3 exhibit
a similar shape and can be scaled. The key observation is to scale the sensor
displacements by the threshold displacement DLT defined in figure 2. When
this scaling is performed, where the (dimensionless) scaled displacement x is
defined as x = Dx/DLT, all the sensors lie on the same universal curve, as
shown in figure 4. Empirically it is found that for Elastica operating at 840 nm,
the threshold displacement is related to the sensor length by

Figure 3.
Transmission response of four different length Elastica fiber sensors.

Page 235

Figure 4.
Scaling behavior of Elastica fiber sensor response. Same data as figure 3 for all 4
sensors, but with displacement scaled by threshold values.

As can be seen from figure 3, the slope of the Elastica transmission response is
a strong function of the sensor straight length Lo. In figure 5 we show the
sensitivity of an Lo = 5 mm Elastica in transmission. Although the sensing
range of this sensor is limited to about 0.8 mm, corresponding to a maximum
strain of 16%, the sensitivity of dT/dx = - 2.4 10-3/m is a factor of 7.2
higher than a 10 mm length Elastica. The 16 bit resolution-limited
displacement sensitivity is 6.4 nanometers, corresponding to a strain resolution
of 1.3 strain, and the gauge factor is G = 14.5 for a gauge length of 4.8 mm.
This large gauge factor combined with small size makes such short Elastica
suitable for a variety of high sensitivity applications.

Figure 5.
Expanded view of data for Lo = 5 mm Elastica fiber sensor showing linear fit.
Data points spaced apart by 0.25 m are easily resolved with low noise. Sampling
rate was 4 Hz.

Page 236

Elastica Reflection Mode Response

The transmission mode Elastica fiber sensor is very low cost and extremely
simple to use due to the fact that no fiber beamsplitters or other optical
components are required (at least for a single sensor), and the effects of
backreflections from connector interfaces and fiber intrinsic Rayleigh scattering
are insignificant, thus minimizing sources of error. In fact, due to the extremely
low cost of silicon photodetector chips with on-board signal processing
electronics, an extremely low cost method of sensor ''multiplexing" is simply to
run one fiber and photodetector for each sensing channel. Surprisingly, this
approach is lower in cost and easier to implement than more sophisticated
optical and electronic multiplexing techniques. In addition, this method yields
the greatest number of surviving sensor channels in the case of fiber cable
damage arising from structural failures due to various conditions or hazards.
However, in some situations, it is desirable to have only a single fiber going out
to and returning from each sensor. The reasons for this are varied, but a
common one is simply to minimize the sensor size. We have reported on the
use of optical time domain reflectometry (OTDR) to interrogate Elastica fiber
sensors using pulsed light, which has the advantage of making the
measurement lead insensitve [36]. However, as is well known, OTDR's based
on weak Rayleigh backscattering are unable to give accurate measurements at
rates faster than about one per minute, due to the extensive signal averaging
required [7]. This is adequate for many structural monitoring situations,
however it is inadequate for monitoring structural vibration modes, which are
of great interest. Furthermore, although OTDR's are powerful instruments, they
are too expensive for many sensing applications.
In continuous wave (CW) reflection mode operation, the optical signal is much
larger and one can easily measure at 100 Hz rates or faster, which is sufficient
for most structural vibration applications. In CW reflection mode operation, one
must arrange for a mirror to reflect the light back through the Elastica sensor
lead-in fiber. In principle, the mirror can be near or far away from the sensor,
however in practical situations there are significant differences. In order to
realize the advantages of a single lead-in, lead-out fiber, it is desirable to have
the mirror fabricated right at one end of the Elastica sensor (point E in figure
1a, assuming light enters the sensor from the left at point A). We have
developed a low-cost technique for fabricating mirrors in-situ on the end of

fibers that are mounted in small diameter tubes. A 2 2 fiber beamsplitter

arrangement is utilized in order to both send and receive light to and from the
sensor. This causes some complications, as care must be exercised to reduce
sources of backreflections from connectors and fiber ends, which can cause
offsets due to light reaching the detector which has not passed through the
sensor.
In figure 6 we show both the single pass transmission, T1, and double pass
reflection, T2, response of the same Lo = 10 mm Elastica sensor in exactly the
same mounting. The only difference between the two optical setups was
whether or not the output lead of the Elastica was connected directly to the
detector, or inserted in a vial of mercury and the unused return leg of the 2 2
coupler connected to the

Page 237

detector instead. In both cases, the 2 2 beamsplitter was used to input light
to the sensor, and all other optical connections were the same. If the sensor is
strictly reciprocal, one naively expects that the double pass transmission
response T2 is simply the square of the single pass response T1. As can be
seen from the figure, this is not the case, and is evidence for some mode
depopulation taking place in the sensor, as is to be expected on theoretical
grounds [89]. However, the slope of the double pass response is nearly the
same as the slope of the square of the single pass response over most of the
sensing region. The point is that the reflection mode slope sensitivity dT/dx is
about a factor of 1.4 times higher than that for the transmission mode, thus
giving the reflection sensors a slightly higher sensitivity than the transmission
mode sensors. Unfortunately, one has to pay for this with at least a factor of 4
light loss due to the use of the fiber beamsplitter on the common input and
output fiber lead.

Figure 6.
Comparison of transmission and reflection response of an Lo =10 mm Elastica
fiber sensor. T1, single pass transmission, T2, double pass reflection, T12 is the
square of the single pass transmission data.

Large Strain Fatigue Testing of an Elastica Fiber Sensor

Conventional resistance strain gauges have a typical fatigue life specification of
107 cycles at 1500 e (micro strain) [10]. Usually the maximum strain that
can be tolerated by conventional resistance gauges is about 5% or less. Most
fiber optic sensors, such as the EFPI, polarimetric, and fiber Bragg gratings,
can only survive even smaller strains, typically about 2% or less. On the other
hand, typical Elastica fiber optic sensors can sense and survive strains as large
as 30% or more. Since the Elastica fiber optic sensor undergoes high bending

stress at large displacements, a natural question to ask is, how robust are
these sensors and what is their fatigue life?
In order to address the question of fatigue life, a reflection mode Elastica fiber
sensor was installed on a test fixture as shown in figure 7. The mirrored end of
the

Page 238

sensor was mounted on a voice coil shaker, and the other end to a fixed,
motorized translation stage. The translation stage mount was chosen in order
to allow a convenient method to exercise the sensor in situ before, during, and
after the fatigue testing. The shaker had a lock down mechanism to prevent it
from moving when the translation stage was moved.
The sensor straight length L0 was 10 mm and was pre-biased by 1.45 mm for
the fatigue testing, which corresponded to the middle of the linear sensing
region. The fatigue test consisted of displacing the sensor in compression and
extension with the shaker, with an amplitude of 1 mm at 10 Hz,
corresponding to a strain of 11.7%. The sensor transmission was optically
monitored during the fatigue testing in order to look for signs of sensor
degradation or breakage. In order to obtain consistent data, the intensity
signal was averaged for periods of 60 seconds, corresponding to an accuracy of
about 0.17%. Initially, we only wanted to test the sensor to see if it would
survive for 106 cycles at large strain. After the sensor survived the first million
cycles, we decided to continue the test until a total of 2.02 106 cycles, at
which point the test was stopped due to time constraints.
The mere survival of the sensor at such large strains for 2 million cycles is very
encouraging in itself and speaks volumes about the robustness of Elastica fiber
sensors. Furthermore, as can be seen in figure 8, the optical transmission vs.
displacement response of the sensor at the end of the test was nearly identical
to that at the beginning of the test, showing very little signs of sensor
degradation. The small changes in transmission that were observed are
attributed to details of the fabrication of this particular sensor and the testing
procedures. In particular, the lock-down mechanism on the shaker had a small
amount of play in it,

Figure 7.
Close-up of reflection mode Elastica fiber sensor fatigue test mounting. The
mirrored sensor end and a silicon micromachined accelerometer for measuring the
displacement waveform are mounted to the mechanical shaker. Other end of
sensor is mounted to a motorized translation stage.

Page 239

corresponding to a position uncertainty of about 100 m between initial and

final transmission response tests. The slight initial increase in transmission and
small difference in transmission response after the fatigue test is attributed to
either or both: 1.) a small change in mirror reflectivity when the fiber was
stretched slightly beyond its straight length L0, which was difficult to determine
exactly due to the end play in the shaker lock down mechanism. 2.) some
indentation in the soft polyacrylate coating near the clamped end of the fiber
mounted on the translation stage, due to a small amount of lateral play in the
stainless steel holding tube in this particular sensor, (point A in figure 1).
The excellent results of this initial fatigue test, combined with the successful
long-term, high temperature (6 months, 1100F) structural monitoring
experience with the sensors [6] are extremely encouraging. It is clear that
even better results can be obtained by improving our sensor fabrication and
fatigue testing techniques and employing sensors with other coatings, such as
polyimide and gold. As a result, more extensive and systematic testing, and
refinements in the sensors and testing fixture are planned.

Figure 8
Transmission response of a reflection mode Elastica sensor before and after
large strain fatigue test, showing nearly identical response.

Inertial Balance with Elastica Fiber Optic Sensor

Elastica fiber sensors are well suited to a wide variety of demanding structural
monitoring and engineering mechanics measurement situations [1, 6]. In order
to demonstrate their utility in a simple situation on a vibrating "structure", we
decided to upgrade a classic experiment in the introductory physics lab, the
inertial balance. The inertial balance apparatus is shown in figure 9 and
consists of two spring arms attached to a platform which oscillates laterally.

Students determine the period as a function of mass added to the platform by

counting oscillations. This is a tedious and dizzying experience, especially if one
counts 100 oscillations for several of the shorter periods (which are less than
one second). Furthermore, the students have no experimental evidence that
the balance acts like a damped simple harmonic oscillator.

Page 240

Figure 9.
Inertial balance apparatus with Elastica fiber sensor (platform displaced).

Figure 10.
Close up of reflection mode Elastica fiber sensor magnetically mounted on inertial
balance. Sensor brackets are attached to balance arm springs with two thin flat
magnets (white rectangles).

An L0 = 10 mm Elastica fiber sensor was magnetically attached near the fixed

end of one spring arm of the inertial balance, as shown in figures 9 and 10.
This was possible due to the extremely low force constant of Elastica fiber
sensors, (in the range of 0.02 N/mm) which produces negligible loading to the
structure when it is attached to it. This is an important property of Elastica
fiber sensors which ensures excellent strain transfer, even for very soft, lowmodulus and flexible materials, and allows numerous methods of attaching to
the structure of interest [1].

Page 241

The sensor was prebiased to 1.17 mm and calibrated by displacing the balance
platform laterally and measuring the sensor response. It was found that the
sensor response was extremely linear for platform lateral displacements as
large as 4 cm.
Figure 11 shows the damping of the oscillations of the balance with 3 kg mass
added to the platform, when released with an initial 4 cm lateral displacement.
Figure 12 shows the harmonic nature of the oscillations and analysis of the
data reveals an oscillation frequency of 0.51 Hz. Clearly students can learn
much more about damped harmonic motion from such detailed data, and have
a much more enjoyable learning experience in the process. This application is
just one simple example of how the novel properties of Elastica fiber sensors
can be easily and cost effectively employed in structural monitoring and
engineering mechanics measurement applications.

Figure 11.
Ring down of inertial balance with 3 kg mass added to 0.883 kg platform,
as measured with a 10 mm Elastica fiber optic sensor.

Figure 12.

Expanded view of 3 kg added mass data showing 0.51 Hz oscillations of the balance
sampled at 10 Hz.

Page 242

References
1. Voss, K. F. and Wanser, K. H., 1997. "Fiber-Optic Strain-Displacement
Sensor Employing Nonlinear Buckling", Applied Optics, 36, pp. 29442946.
2. Elastica is a trademark of South Coast Spectrum, Laguna Hills, CA 92654.
3. Wanser, K. H. and Voss, K. F., 1994. "Crack Detection using Multimode
Fiber Optical Time Domain Reflectometry", Proc. SPIE 2294, Distributed and
Multiplexed Fiber Optic Sensors IV, pp. 4352.
4. Voss K. F. and Wanser K. H., 1994. "Fiber Sensors for Monitoring Structural
Strain and Cracks", Proc. 2nd European Conference on Smart Structures and
Materials (ECSSM 2), Glasgow, Scotland, pp. 144147.
5. Wanser K. H., Voss, K. F., and Griffiths R. W., 1994. "Distributed Fiber Optic
Sensors for Structural Health and Vibration Monitoring using Optical Time
Domain Reflectometry", Proc. First World Conference on Structural Control
(1WCSC), Pasadena, CA, pp. WA3-3-WA3-12.
6. Wanser, K. H. and Voss, K. F., 1997. "Fiber Optic Strain Monitoring Inside a
Power Plant Boiler", Intelligent Civil Engineering Materials and Structures, F.
Ansari, A. Maji, and C. Leung, eds., American Society of Civil Engineers, New
York, pp. 213228.
7. Wanser, K. H., Haselhuhn, M., Lafond, M., and Williams, J., 1993.
"Distributed Fiber Optic Sensors for Civil Structures using OTDR", Applications
of Fiber Optic Sensors in Engineering Mechanics, F. Ansari, ed., American
Society of Civil Engineers, New York, pp. 303327.
8. Wanser, K. H., Voss, K. F., and Kersey, A. D., 1994. "Novel Fiber Devices
and Sensors based on Multimode Fiber Bragg Gratings", Proc. 10th
International Conference on Optical Fiber Sensors, Glasgow, Scotland, pp.
265268.
9. Wanser, K. H., Voss, K. F., and Kersey, A. D., 1994. "Multimode Fiber Bragg
Gratings for Real Time Structural Monitoring using Optical Time Domain
Reflectometry", Proc. Symposium and Workshop on Time Domain
Reflectometry in Environmental, Infrastructure, and Mining Applications,
Northwestern University, Evanston Illinois, U. S. Bureau of Mines Special
Publication SP 1994, pp. 472483.

10. See for example, The Pressure Strain and Force Handbook, Omega
Engineering, 1996, vol. 29, p. E-7, and Transducer-Class Strain Gages,
Bondable Resistors, Installation Accessories, Measurements Group Inc.,
Catalog TC-116-4, 1995, p. 5.

Page 243

Liquid Core Optical Fibers for Crack Detection and Repairs in Concrete Matrices
C. Dry
Abstract
The need to monitor the internal state of civil structures and materials is great.
Existing instrumentation techniques that mainly rely on magnetism, electricity,
or stress gauges are limited if used for remote measurements in concrete or
composites. They are sensitive to electromagnetic noises and they degrade in
the environment over time. Optical fibers are attractive because they are
immune to electromagnetic interference and are sensitive over long distances.
The combination of the ability to remotely measure crack occurrences in real
time and determine the location and volume of crack damage in the matrix is
unique in the file of optic sensors (or any sensors in general). The combination
of this with crack repair, rebonding of any detached or broken fiber, and
replenishment of liquid core chemicals, when necessary, make this a potentially
powerful sensing and repair tool.
This research is an investigation into the use of adhesive liquid core optical
fibers for the detection of cracks and their location and volume in opaque and
semi-opaque brittle materials. The study combines work based on the concept
of internal adhesive delivery from hollow fibers for repair with nondestructive
fiber optic analysis of crack locations and volume within the same system. The
liquid filled hollow fibers can carry light in the liquid. The cracked fiber that has
released liquid projects diffraction patterns from the meniscuses at the ends of
the liquid. The size relationship of these patterns allow us to determine the
location of the cracks and the amount of liquid lost into the matrix and perhaps
relate it to the volume and location of the cracks in the matrix. One sees into
the hollow fiber by projecting light in from one end. This work is sponsored by
the University of Illinois and the National Science Foundation.
Carolyn Dry, University of Illinois, School of Architecture; Champaign-Urbana, Illinois

Page 244

Significance and Impact on the Field

Swartz[1] developed a clever method to reveal fracture surfaces and to
estimate crack length and fracture toughness. He used a reservoir of dye on
the cracked area in a beam. This method required active addition of the dye to
read results. Koji Otsuka[2] developed an x-ray technique using a contrast
medium to detect fine cracks in reinforced concrete. This is a nondestructive
technique that requires the addition of materials in the field. The result is a
complex pattern of internal cracks shown on x-ray film which are difficult to
read. The thickness of specimens may also make the results difficult to assess.
The approach in our research is to not add dye in the field, but only to take
measurements.
Originally optical fiber crack detection and monitoring by fiber breakage in
polymers was developed by Liu, Ferguson and Measure[3]. The concept is that
cracking in the matrix results in cracking in the optical fibers, causing the light
transmission to be attenuated. The lighted ends of the fibers can be seen
through the transparent polymer matrix. The result is a remote crack detection
system suitable for transparent matrices. This approach does not work to
investigate cracking in opaque matrices.
In our research of self repairing materials, the presence of adhesive-liquid filled
fibers for self-repair inside an non-transparent matrix, such as cement or an
opaque polymer composite, suggested the possibility of using these fibers for
nondestructive internal evaluation of the material as well as for crack repair.
This detection system not only detects of the presence of cracks, but can be
used to determine their location and volume based on liquid lost into the matrix
cracks. We intend to test our ability to use this method for crack location and
assessment in opaque matrices.
Description of Work
The goal was to develop liquid core optical fibers for the detection (and selfrepair) of cracking wherever and whenever damage is generated in cement or
polymer materials or opaque and semi-opaque brittle material by dynamic or
static loading.
In the project entitled "Passive Smart Composites, for Dynamic Control of
Structures, Self-Timed Release of Chemicals for Repair of Crack Damage",
performed as a subcontract with University of Michigan sponsored by NSF and

Strategic Highway Researched Programs (SHRP), we explored fiber composites

designed to release repair chemicals from hollow fibers when loading causes
cracking in the matrix and cracking of the fibers. The released repair chemicals
or adhesives act to bond and seal the cracks and rebond fibers making the
cement stronger in bending and more resilient under further loading.
[4,5,6,7,8]

Page 245

Crack Detection Techniques Using Optical Fiber

Optical fibers can be used for crack detection by directing a laser beam
through a hollow fiber glass tube in an opaque matrix and measuring the light
projection pattern and power at the far end of the sample.
As the laser light beam passing through the tube reaches a meniscus in the
adhesive liquid, it projects a bright field of light on the screen beyond the
sample's end. The diameters of these rings of light can be used to determine
the size and location of the void in the tube where the adhesive has leaked into
the matrix cracks. The size variation of the two discs is an indication of distance
between them and therefore, the volume of liquid lost from the fiber. This is an
exact way to measure how much adhesive leaks out of the glass fiber and is
related to how much cracking is present in the matrix. The absolute size of the
rings is an indication of distance from the laser source to the disc diffraction
pattern. This is a means of determining crack location in the pipettes and is
related to crack location in the matrix. If the crack volume is larger than the
volume of liquid in the glass capillary, then this detection method will be
ineffective, however the cracks of that size should be visible to the naked eye.
As an example, a 100 millimeter capillary pipette with half of its liquid lost from
one end with two meniscuses will project two circles approximately 4 mm and
2 mm in diameter. If the same amount of liquid is lost not from one end, but
from somewhere else along the length of the pipette, the 2 to 1 ratio
relationship in size will remain, but the absolute size will vary with distance of
the voids, and therefore the meniscus, from the projection screen. (See figure
1)

Figure 1.
System for measuring the light diffraction through meniscuses.

A photosensitive optic diode at the sample end detects the power of the

resultant beam and translates that information to a voltmeter. Fibers with

various amounts of liquid adhesive will allow a consequently variable amount of
laser light through to the detector, creating a means of evaluating how much
adhesive remains inside the tube. This allows confirmation of results from the
first method.

Page 246

Test Methods
A small laser is aligned on a wood base with two ''screens". Adhesive filled
pipettes are placed with one end against the laser light aperture. The first
screen has a small hole through which the opposite end of the pipette
protrudes very slightly. The second screen is placed two inches behind the first
screen in order to measure the laser light diffraction pattern emitted from the
end of the pipette. An optic diode, for measuring the power of the laser light, is
attached to the end of the glass pipette. A number of pipettes are filled using a
variable amount of adhesive. The borosilicate glass pipettes are 4 1/8" (10.5
mm) long and have an average interior diameter of 0.8 mm and an average
exterior diameter of 1.6 mm. The adhesive is held by capillary action by sealing
one end of the pipette.
The pipettes are individually measured for their adhesive content and then
tested on the laser apparatus to determine correlation between the varying
quantities of adhesive and the resultant power and diffraction patterns of the
laser light directed through them. The amount of adhesive in the pipettes is
represented by the measurement of air space at the end that was not filled
with adhesive. The laser light power is measured by a volt meter diode placed
on a screen placed 2" (5 mm) away from the end of the pipettes. The
diameters of the circular diffraction patterns projected on this same screen are
also measured. (See figure 2)

Figure 2.
Diffraction pattern projected on screen through adhesive filled pipette.

Page 247

Results of Test on Fiber Optic Methods

The laser light diffraction patterns and beam strength are directly related to
the amount of adhesive in the pipettes. The results of testing show a linear
relationship between adhesive quantities and the resultant laser light power
and diffraction pattern size. This data is represented in the graphs shown in
figures 3 and 4. This is a new idea applied to an old problem urgently needing
a solution: crack location and determination of crack volume.

Figure 3.
Liquid quantity voided from capillary pipette vs. size of laser light
diffraction pattern.

Figure 4.
Liquid quantity voided from capillary pipette vs. Voltage measured

Page 248

X-ray Testing of Opaque Samples

An x-ray technique can also be used to study the matrix, in order to relate of
location and volume of adhesive loss from the pipettes to crack location and
volume in the matrix. A 2" 3" size double dental film is a convenient and
practical film system that works for our application. A simpler "Rigaku" x-ray
machines is used at an x-ray power setting of 35 kV and 20 Ma to give the
best contrast and resolution for our 1" thick concrete samples. The x-ray
machine has a rectangular port of approximately 1/2" x 1/4"; the x-ray beam
widens and becomes weaker further away from the port. In order to obtain a
3'' long image of a 1" wide concrete sample, the samples are moved an
appropriate distance from the x-ray port and a sample holding device slowly
moves the sample vertically across the beam. The sample holding device used
is an electric motor and pulley system that is timed in order to achieve an
appropriate exposure for the film (which is simply attached directly behind the
concrete sample). These x-ray photographs show the concrete matrix as gray,
the steel reinforcing wires as thin white lines, the adhesive tubes as very dark,
and the cracks as black. See figure 5. All of the concrete samples are made
with carefully aligned rows and columns of thin steel wire reinforcement and
adhesive filled glass capillaries so that an x-ray view across them could be
easily read.

Figure 5.
A developed x-ray photograph showing the concrete as gray, piano wire
as the white lines, the adhesive tubes as very dark and the cracks as black.

The contrast between the adhesive in the cracks and the cracks themselves is
often not significant enough to detect any difference with the naked eye.
Therefore, the adhesive is "tagged" with an x-ray sensitive contrast agent
which shows clearly in the x-ray photographs. Powdered tungsten is mixed
with the adhesive; it is white specs in the x-ray photographs. See figure 6.

Page 249

Figure 6.
X-ray photograph of adhesive concrete specimen
in which adhesive containing tungsten (small water
specs") was released from hollow glass fibers.
The small white lines are metal reinforcing wires.

References
1 Swartz, Stuart, "Cracked Surface Revealed by Dye and Its Utility in
Determining Fracture Parameters", Fracture Toughness and Fracture Energy,
Mihashi et al., eds, Balkema, Rotterdam, ISBN9061919886, 1989, p509.
2 Otsuka, Koji, "X-ray Technique with Contrast Medium to Detect Fine Cracks
in Reinforced Concrete", Fracture Toughness and Fracture Energy, Mihashi et
al., eds , Balkema, Rotterdam, ISBN9061919886, 1989, p521.
3 Lui, Ferguson, Measure, "Damage Determining in Composites with
Embedding Fiber Optic Interferometric Sensors", SPIE Proceedings#1170,
1989, p205.
4 C.M. Dry., "Timed Release of Chemicals into Hardened Cementitious Matrices
for Crack Repair, Rebonding Fibers, and Increasing Flexural Toughening,"
Fractural Mechanics, 25th Volume, ASTM STPI 220, Fayel Erdrogan, editor,
1995, pp. 268282.
5 C.M. Dry., "Passive Tuneable Fibers and Matrices," International Journal of
Modern Physics B, Vol. 6, Nos. 15 & 16, Rivers Edge, NJ: World Scientific
Publishing Co., 1992, pp. 27632271.
6 C.M. Dry, "Matrix Cracking Repair and Filling Using Active and Passive Modes
for Smart Timed Release of Chemicals from Fibers into Matrices," Journal of
Smart Materials and Structures, Vol. 3, No. 2, June 1994, pp. 118123.

Page 250

7 C.M. Dry., "Smart Multiphase Composite Materials which Repair Themselves

by a Release of Liquids which become Solids," Smart Structures and Materials
1994: Smart Sensing, Processing and Instrumentation, Vijay K. Varadan,
editor, Proceedings, SPIE 2189, 1994, pp. 6270.
8 C.M. Dry, "Smart Multiphase Composite Materials which Repair Themselves
by a Release of Liquids which become Solids," Smart Structures and Materials
1995: Smart Sensing, Processing and Instrumentation, William B. Spillman, Jr.
editor, Proceedings, SPIE 2444, 1995, pp. 410413.

Page 251

A Dual Core Forward Time Division Multiplexing Optical Fiber for Weigh-inMotion Sensing
R. B. Malla,
N. W. Garrick,
A. Sen
and P. Dua
Abstract
A novel dual core Forward Time Division Multiplexing (FTDM) optical fiber for
determining weight of vehicles in motion is investigated. Several mechanical
and optical experiments have been performed on the test fiber. Results of
these experiments confirm the sensing fiber's ability to measure the magnitude
as well as to determine the location of the load acting on it. It is envisioned
that the system will also be capable of simultaneously obtaining vehicle related
data such as vehicle volume and speed.
Introduction
In the past several decades, both government highway agencies and private
industry have given considerable attention to designing a weigh-in-motion
(WIM) system capable of accurately measuring the weight of moving vehicles.
Earlier vehicle-weighing systems were typically large, permanently installed,
and required vehicles to leave the roadway during the weighing process. In
some applications, where portable scales were used, the enormous size of the
vehicle made the measurement process very complex and tedious.
The benefits of a low cost, and accurate WIM systems to the transportation
industry are indisputable. Vehicles will no longer need to stop at weigh stations
on freeways. Installation and calibration costs will be greatly reduced and so
will law enforcement costs.
Although in recent years WIM systems based on piezoelectric crystals and
other technologies have been investigated and tested in the field, there is
currently no commercial WIM system which uses optical fiber as sensor.
However, it is widely known that the characteristics of optical fibers make it
possible to develop
Ramesh B. Malla*, Norman W. Garrick , and Amlan Sen ,Department of Civil and
Environmental Engineering, U-37, University of Connecticut, Storrs CT 06269 (*

contact author: Tel.(860)-486-3683,E-mail: mallar@engr.uconn.edu)

Puneit Dua , Department of Electrical and Systems Engineering ,University of
Connecticut, Storrs CT 06269

Page 252

sensors with excellent performance. In theory, the physical quantity to be

measured is used to modulate either the intensity or the phase of the light
wave traveling into an optical fiber, which is the output signal used for
measurement.
Optical fibers are sensitive to applied strain and pressure, and could be used
for 'real time' sensing. One system using optical fibers as sensors is based on
the optical time domain reflectometry (OTDR) in which the magnitude of the
mechanical forces is measured by the decrease they cause in the intensity of
the Rayleigh-backscatter from the interrogating light pulses. A serious
shortcoming of this technique is that it sends light out of the fiber on bending
and hence provides an estimate of the load by comparing two very weak and
noisy Rayleigh-backscattered signals.
Other techniques proposed for WIM systems, include those based on light
polarization and interferometry (such as the Alcatel system[1], [2]) and on the
measurement of transmitted light intensity through an optical fiber having a
single core (such as the Oak Ridge system [3], [4] and Virginia Tech system
[5]). These techniques are not suitable for measuring simultaneously loads
acting at different points along a continuous fiber. Furthermore, interferometric
techniques are complicated and require a different interferometer for each
fixed sensing point, separated from other interferometers by a system of fiber
optic couplers. The main disadvantage of polarization technique is that a
polarization change at any point along the fiber affects the polarization state of
the fiber at other sensing points. The technique based on light transmission
along a single core has the disadvantage that it is not able to determine
locations of forces applied along different points of the fiber.3
This paper presents a new fiber technique for direct measurement of the
weight of vehicles in motion by using the Dual core Forward Time Division
Multiplexing (FTDM) fiber. The magnitude and the position of the load can be
measured simultaneously. It also has the advantage of providing an output
signal with a much higher intensity than obtained from the OTDR technique,
thus allowing easier detection and measurement.
Characteristics of the Test Fiber
According to the information provided by the manufacturer the fiber used in
this project was drawn by the Plasma-enhanced Modified Chemical Vapor

Deposition (PCVD) process. The fiber is doped with fluorine and germanium.
The outer plastic jacket is a double layer of UV curable acrylate. The test fiber
has four regions of different refractive indices as shown in Figure 1 [6]. These
regions are single mode central core (A), covered with a cladding region (B) of
lower refractive index, a second, non-contiguous light-guiding region (C )
having longer effective optical path length. A second cladding region (D)
having index of refraction, n4. Figure 2 shows the typical refractive index
profile of the test fiber. The principle by which the fiber acts as a force sensor
is described in the following paragraph.
A train of light pulses having width of the order of sub-nanosecond or shorter is
injected into the central core (A) of the fiber at one end. When a mechanical
force is applied at any point along the fiber, the fiber will be bent and a

Page 253

fraction of the light pulses propagating along the central core (A) is deflected
into the second light guiding region (C). These deflected light signals
propagate along the second region and reach the other end of the fiber some
time after the arrival of the undeflected light pulses as the effective refractive
index, n3, of this region is higher than the refractive index, n1, of the central
core. This time difference between the two light pulses gives the location along
the fiber where the force was applied.
As some of the light escapes into the outer core there is a loss in the intensity
of the light traveling in the inner core. As a result new light pulses are
observed in the outer core. The magnitude of the force can be determined
from the relative intensities of these two light signals or from the decrease of
inner core light and increase of the outer core light. This separation of the
signals in time is called Forward Time Division Multiplexing (FTDM). The
deflected signals in the FTDM fiber are several orders of magnitude stronger
than other scattered light signals like the Rayleigh-backscattered signals.
Laboratory Testing of the Fiber
For the laboratory experiments, two fibers were used, one having an outer
diameter of 125 micrometer (Fiber 1) whose refractive index profile is given in
Figure 2; and the other having an outer diameter of 140 micrometer (Fiber 2)
whose the refractive index profile is given in Figure 3. Mechanical and optical
tests were done on Fiber 1, while the optical experiments were done with Fiber
2.
Mechanical Characteristics
A number of tests were performed on Fiber 1 (125um diameter) to determine
its modulus of elasticity and ultimate tensile strength. It is important to get
some sense of values of these parameters because the fiber will ultimately be
subject to bending from vehicular loads on the highway. These parameter will
shed light onto what magnitude of stresses the fiber can withstand in the
pavement.
The modulus of elasticity of any material is the ratio of the stress that the
material is subject to, to the strain that is developed due to the stress. A higher
modulus of elasticity (E) would mean that the fiber will bend with lower
curvature under a certain loading, while a lower modulus would on the other
hand indicate a higher curvature of bending under the same load. Thus the

bending rigidity of the fiber is important in order to design a mechanical device

that will bend the fiber appropriately so that adequate light is transmitted from
inner core to the outer when a force is applied on the fiber. In addition to
measuring the modulus of elasticity, it is also of importance to get a good
estimate of the ultimate tensile stress that the fiber can withstand.
Five different tensile tests were carried out with the fiber. The first test had the
fiber sandwiched and bonded between two square-steel plates having one
square inch area at each end and then mounted on a 101b Universal Testing
Machine (Instron Machine Model 1011). The five test specimens of gage
length

Page 254

varying between 520 cms were individually subjected to tension by pulling the
two ends away from each other. This test failed as the fiber began to slip
relative to the steel. The second test involved looping and bonding of the fiber
between the end plates. The test was performed on 3 specimens of gage
lengths of 5 and 10 cms. The value of the modulus (E) obtained after
calculations was approximately 70.50 GPa. The third test was carried out on 5
specimens of gage length 5 and 10 cms. In this test the fiber was wrapped
around a steel cylinder 4 inch in diameter and 1/4 inch long at each end. This
test failed too due to the slippage of the fiber and the rotation of the cylinder
from the grip arms of the Instron machine. The fourth test was similar to the
first test but cardboard plates were used instead of the steel plates at the two
ends of the fiber. It was performed on 10 specimens of length 5 and 10 cms.
The value obtained from this test was calculated to be approximately 53.93
GPa. The fifth test was done using an instrument called cathotometer. The
fiber was attached to a rigid support at the upper end and a weight was hung
from its lower end. The stretching of the fiber was measured by the
cathotometer. The test was not conclusive due to difficulty in reading the
minute change in length as well as due to slippage at the two ends. Because of
the problem with bonding and lack of adequate instrumentation available for
testing a single fiber, the results obtained here give only the approximate
range of the value for the fiber's modulus of elasticity.
Optical Characteristics of Fiber
The fiber cross-section was photographed by a NIKON transmission mode
optical microscope. For Fiber 1 (125um diameter) the cross section is shown in
Figure 4. It shows the four different regions of the fiber: i) the inner core, A,
which is the light guiding region at the center; ii) a cladding region, B, around
the inner core; iii) an outer light guiding region, C, around the cladding and,
iv) an outer cladding, D, around the outer core, C.
The refractive index profiles for these two fibers (125um outer diameter and
140 um ) were supplied by the manufacturer. The dimensions of different
regions in the cross section are given in Table I below (also see Figures 2 and
5):
TABLE I-DIAMETERS OF THE VARIOUS REGIONS IN THE FIBER CROSSSECTION
Region
Diameter(um): Fiber 1
Diameter(um): Fiber 2

Inner
Core
Inner
cladding
Outer Core
Outer Cladding

6
18
75
125

7
18
90
140

Page 255

The refractive index values for key locations in the profiles of the two fibers
(125um outer diameter and 140um outer diameter) are given in Table II
below.
TABLE I-DIAMETERS OF THE VARIOUS REGIONS IN THE
FIBER CROSS-SECTION
Region
Diameter(um): Fiber 1Diameter(um): Fiber 2
Inner Core
6
7
Inner cladding
18
18
Outer Core
75
90
Outer Cladding
125
140

Optical Experimental Set Up

In the laboratory experimental set up the interrogating light pulses are
obtained by modulating the laser diode with the help of a pulse generator
(Avtech Model no. AVM-2-C). The pulsed light beam is focused onto the fiber
so that most of the light enters the inner core. A graded index (GRIN) lens was
used to focus the light. At the other end of the fiber a photo-detector is used
to convert the light into electrical energy and transmit the signal to the
oscilloscope. In order to determine the most suitable combination of the laser
diode and the photo-detector various laser diodes in the wavelength range of
670830 nm with output power in the range 515 mW were tried with two
different photo-detectors. One was a Silicon Avalanche Photo Detector
(Newport Model No. 877APD) and the other Ultrafast InGaAs Amplified
detector (Newport Model No. AD-300DC-FC). The experiments were conducted
on both the fibers: Fiber 1 (125um diameter) and Fiber 2 (140um diameter).
The oscilloscopes used in different set ups for observing the output pulses were
HP modular oscilloscope (Model No.54720D) and Tektronix 7104 oscilloscope.
The laboratory set up of the apparatus is shown in Figure 5.
Fiber Bending Experiment
The fiber with 140 um outer diameter was bent at different radii of curvature
and the corresponding light output signals were observed using the HP
54720D oscilloscope. Our initial assessment showed that theoretical expected
behavior was essentially correct with the exception of one glitch. In the ideal
case if all light is launched into the single mode core (region A, Figure 1) alone
we expect to get only one peak as the output signal. One output peak would

suggest that all the light stays in the inner core throughout its journey through
the fiber. However, in our experiment, we observed two peaks instead of one
peak (peaks 1 and 2 in Figure 6, curve without bending).

Page 256

When a bend is introduced in the fiber, the result in Figure 6 shows three
peaks, 1, 2 and 3. This is different from what was originally predicted. Based
on the theory, with bending, two voltage peaks were expected -one for light
traveling through the inner core and a second for light traveling through the
outer core (region C, Figure 1). The peak on the left (peak 1, Figure 6)
corresponds to the light in the inner core, while that on the right (peak 3,
Figure 6) corresponds to light that got into the outer core due to the bending.
One possible reason for the appearance of the third peak (peak 2, Figure 6)
could be that the light from the source is being launched not just into the inner
core but also into the outer core of the fiber. We plan to improve the light
launching conditions in order to eliminate the problem of the additional peak.
Figure 7 is the calibration curve which was developed from this experiment. It
can be used to determine the degree of curvature (and hence, amount of load)
from the change in the magnitude of light at the output end of the fiber. For
example, using the calibration curve in Figure 7 we find that a 10% decrease
in the magnitude of the light in the inner core (i.e. 90% of the original value)
corresponds to a bend with a radius of 36 mm. Conversely, a 50% decrease in
light corresponds to a bend radius of 7 mm. The results in Figure 7 show a very
high correlation between the change in the magnitude of light and the degree
of bending.
On Going Work
The initial tests of fiber bending were accomplished by introducing loops of
various diameters in the unshielded fiber. If the fiber is to be used in a WIM
system a mechanical system needs to be designed which will protect the fiber
while allowing load to be transmitted to the fiber. In addition, the transmitted
load must result in a predictable degree of bending in the fiber. A prototype
was designed and built in order to achieve these objectives. A picture of the
prototype device is shown in Figure 8 and details of the fabricated prototype
are shown in Figure 9.
Using this system, the load is transmitted to the fiber through an assembly
consisting of two corrugated metal plates that are separated by springs. The
load is applied to the top plate and thus causes the plates to move closer. The
corrugated faces result in bending of the fiber - the amount of bending is
related to the magnitude of the load. Testing using this device is in progress.

Conclusions and Future Work

The dual core FTDM fiber has the potential to be a very effective method of
measuring truck weights in motion (magnitude and location simultaneously)
and for collecting traffic data. Since the method of measurement is direct, it
should be more accurate than other WIM systems. We have shown that the
fiber, apart from the launching problem stated above, is showing the
characteristics that it was expected to have from theory. For the study to yield
optimum results it is necessary that: i) maximum light is injected into the inner
single mode core from the laser diode at the source end; ii) the photo detector
is able to receive all of the light from

Page 257

the inner and outer light guiding regions at the output end; iii) the
photodetector and the oscilloscope are fast enough to receive light pulses of
very narrow width (picosecond) in order to have better spatial resolution to
distinguish vehicle wheel loads applied on the fiber within a couple of meters
apart. These requirements impose stringent light launch conditions. Further
work is continuing to refine the light launching conditions.
Acknowledgement
The study has been supported in part by the National Academy of Sciences,
Washington D.C. under the NCHRP-IDEA program (contract no. NCHRP 42).
The findings, conclusions or recommendations either inferred or specifically
expressed herein do not necessarily indicate acceptance by the Academy or by
the Federal Highway Administration. Support from Metrilight, Inc. ,
Southbridge, MA (CEO: Dr. Marcos Klinermann) for providing the test fiber and
comments is gratefully appreciated. Authors also acknowledge the financial
and other support received from the University of Connecticut, in particular the
School of Engineering, Department of Civil and Environmental Engineering,
Photonics Research Center, and the Institute of Materials Science.
References
1. Tardy A., M. Jurczyszyn, J.M. Caussignac, G. Morel, G. Briant, "High
Sensitivity Transducer for Fiber-Optic Pressure Sensing Applied to Dynamic
Mechanical Testing and Vehicle Detection on Roads," Optical Fiber Sensors,
Proceedings in Physics, Vol. 44 (H. Arditty, J. Dakin, and Kersten, eds.),
Springer-Verlag, Heidelberg, 1989, pp 215221.
2. Boby J., S. Teral, J.M. Caussignac, M. Siffert, " Weighing of Vehicles in
Motion using Fiber Optic Sensors," Electrical Communication, 1st Quarter,
Alcatel Alsthom publications, France, 1994, pp 7477.
3. Muhs J.D., J.K. Jordan, M.B. Scudiere, K.W.Tobin Jr., " Results of a Portable
Fiber-Optic Weigh-In-Motion system," SPIE Proceedings Series Volume 1584,
1991, pp 374386.
4. Tobins K.W. Jr., J.D.Muhs, " Algorithm for a Novel Fiber-Optic Weigh-InMotion Sensor System," SPIE Proceedings, Volume 1589: Specialty Fiber Optic
Systems for Mobile Platforms, 1991, pp 102109.
5. Safaai-Jazi A., S.A. Ardekani, M.Medhikani, " A Low-Cost Fiber Optic Weigh-

In-Motion Sensor," SHRP-IDEA/UFR-90-002; SHRP-89-ID03, Strategic

Highway Research Program, NRC, TRB, Washington D.C. 1990, 62 page.
6. Klienerman Marcos, Peter W. Kelleher, " A Distributed Force-Sensing Optical
Fiber Using Forward Time Division Multiplexing," SPIE Proceedings Series
Volume 1586, 1991, pp 6777.

Page 258

Figure 1 :
Diagram of the Cross section of the fiber

Figure 2 :
Refractive Index Profile , Fiber diameter 125 um

Page 259

Figure 3 :
Refractive Index Profile , Fiber diameter 140 um

Figure 3 : Refractive Index Profile, Fiber diameter 140 um

Figure 4 :
Photograph of the fiber cross section, outer diameter
= 0.125mm (0.00492 in), taken with Nikon transmission
microscope with magnification = 504 (UConn Inst.of
Materials Science; August 06, 1997)

Page 260

Figure 5 :
Complete Input and Output System for Laboratory Testing

Figure 6 :
Output Signals from the Fiber (with and without bending)

Page 261

Figure 7 :
Calibration Curve - Degree of Curvature versus
Percentage Light in the inner core

Figure 8 :
Picture of WIM Prototype

Page 262

Figure 9 :
Diagram of WIM Prototype

Page 263

Summary of New Jersey Department of Transportation's Need for New Fiber

Optic Tools
N. Vitillo
In preparation for writing this paper, I interviewed many individuals within the
New Jersey Department of Transportation to assess their need for new sensors
and other tools that would facilitate the completion of this work. The
paragraphs that follow are a summation of these conversations.
When I mentioned that NJDOT would be co-hosting a conference on fiber
optics, the comments were all the same, ''Why would NJDOT host a conference
on telecommunications lines?" It was not until I explained that there are many
other uses for fiber optics than communications that I was able to discuss their
needs. It would appear that like so many of us in the technical field, you have
not taken the time to advertise your accomplishments and educate other as to
the benefits of this technology. The first order of business for those involved in
fiber optics research and implementation should be to spread the word and
inform other professional to the possibilities.
States and other agencies invest a great deal of money and resources in its
infrastructure. The infrastructure are the arteries of its economy. Agencies like
NJDOT, New Jersey tool authorities, Metropolitan Planning Organizations
(MPO), counties, and municipalities are responsible for managing this
infrastructure for the public. These organizations are the primary customers
that need to learn more about this emerging technology. In addition to these
governmental organizations, contractors, suppliers, and private industry can
also benefit from these new tools. New programs such as the National Quality
Initiative (NQI) have refocused FHWA, States and others on the quality of our
materials and construction. By emphasizing attention to details in planning and
testing, the goal of this effort is to improve the performance and longevity of
our infrastructure. In order to achieve these goals, everyone involved will need
new and improved tools.
Traditionally, most of the sensor technology has been aimed at the Materials
side of the house. We have looked to the Federal Highway Administration
(FHWA), and standards-setting organizations such as AASHTO, and ASTM to
develop new test equipment and protocols for our use. This has been a time
consuming effort. In today's world, the number of new products, materials,

and test equipment that

New Jersey Department of Transportation, Research, CN 600, Trenton, New Jersey
08625

Page 264

becomes available each year is staggering. It would be impossible for these

organizations alone to develop, evaluate, and approve these new products. The
task than falls to States, universities, consulting firms, and industry to jointly
develop the next generation of tools.
With the proliferation of fast and affordable computer currently available,
equipment manufactures have coupled this technology to their testing
apparatus to run the test in a precise manner and to collect the test data and
perform a variety of analyses that have been laborious tasks in the past. These
advancements in the laboratory also point to test methods and test elements
that need improvement. For instance, the mechanical Linear Variable
Differential Transformers (LVDT) used in repeated triaxial and resilient modulus
tests could be replace by more accurate measuring devices to improve the
quality of our data. Fiber optics could play a role in this improvement.
We rely on the data from our materials testing to provide data for design or to
assess the quality of our materials or as yard sticks for acceptance. We
sometime assume that the equipment is providing consistent results. Materials
testing facilities needs sensors to monitor their equipment to be sure that the
data is consistent. Fiber optics could provide the red flag that alerts the
operator that the equipment needs recalibration or maintenance.
The needs for new material testing and quality control, acceptance, and
monitoring tools does not stop in the laboratory. The need also extends into
the field were the environment is harsher and the durability of the test
equipment becomes a major issue. There are any number of test procedures
conducting the field that could be improved. For example, we currently
determine the entrained air content of concrete by mechanical means. A new
fiber optic sensor to perform this test would greatly improve our ability to
assess the durability of our concrete mixes. This becomes even more critical
now that may agencies are looking to high performance concretes to extend
their design options and reduce life cycle costs.
Like the Materials side of the house, the Construction corner also needs
improvements. Here the need are even broader. The agencies, contractors,
and material suppliers need new and improved means of performing quality
control and quality acceptance. With today's levels of traffic, agencies can not
afford reduce capacity of their facility by keeping lanes out of service.
Therefore much of our work is performed at night under less than ideal

conditions. We also look to new materials such as very high early strength
concrete mixes to make repairs and reopen the roadway to traffic without
inconveniencing the traveling public or commerce. For these types of
applications, we need multifunctional tools that can measure concrete
temperature or maturity and other parameters such as entrained air content or
water cement ratio or moisture content of aggregates.
More and more emphasis is being place on the contractor or suppliers to
develop and utilize quality control plans to produce and place a variety of
materials on construction projects. In the past the agencies performed the
majority of the quality control/quality acceptance testing and informed the
contractor of the results, often in the form of penalties. In their new role, the
contracting and supply industries are very receptive to finding new tools to
accomplish these new programs.
The agencies, contractors, and suppliers also have needs for other tools that
are not material or quality related. The area of Work Zone Safety and Vehicle
Intrusion are

Page 265

extremely important areas. The United States sees many fatalities of both
agency and contractor personnel due the high numbers of high speed vehicles
passing through our work zones daily. SHRP made some strides in making us
aware of these problems and in developing some tools, but the ultimate
solution has not yet been developed.
A relatively new area to many of us is long term monitoring and remote
sensing, data collection, and retrieval. Engineers have conducted short term
laboratory or field tests that assess the immediate condition or performance of
materials or construction practices. More design procedures are being
formulated that replace empirically derived procedures with those based on
mechanistic principles. These procedures need data to assess both short and
long term failure mechanisms such as fatigue cracking, creep, and permanent
strain or deformation to develop the appropriate models. Many of these
mechanistic models need refinements from their strictly linearly elastic
assumptions to those that incorporate the results of long term environmental
monitoring of ambient, and material temperature gradients, moistures
gradients, moistures gradients that are collected hourly, daily, and yearly for
the Design Period. Bridge designers and maintenance engineers also need the
means of assessing the results of exposure to chlorides used to de-ice bridge
decks. The corrosion of decks, abutments and piers consumes a large portion
of our annual budgets to repair.
The Intermodal Surface Transportation Efficiency Act (ISTEA) passed by
Congress mandated the development and implementation of certain
management systems to monitor the initial and long term performance of
pavements, bridges, congestion, and safety features and monitor traffic. These
systems require large amounts of data to feed their analysis routines. Some
distress data such as cracking, and long term deformation of structures could
be monitored remotely through fiber optic installations. In addition, the traffic
monitoring and congestion management systems need better means of
collecting vehicle weight and classification data with weigh-in-motion sensors
and identifying traffic congesting or accident locations and assessing the
advantages of alternate routes.
On the Maintenance side of the house, we need tools that can assist us in the
annual battle of snow and ice control. The area maintenancemanager needs
the ability to remotely monitor pavement temperature sensors being able to

detect pavement icing and estimate the level of Chloride or other anti-icing
chemicals present on the road surface before accidents can occur.
With any new or emerging technology there is a need for training and
familiarization of all the parties concerned. The fiber optic field is no different.
It was not until I started to work with Dr. Ansari at NJIT that I began to
comprehend the possible uses of fiber optic technology beyond the scope to
telecommunications. I believe that you have a great deal of work to complete
in this very exciting field.