Académique Documents
Professionnel Documents
Culture Documents
author:
publisher:
isbn10 | asin:
print isbn13:
ebook isbn13:
language:
subject
publication date:
lcc:
ddc:
subject:
Page iii
Edited by
Farhad Ansari
Page iv
Page v
Table of Contents
Foreword
ix
Preface
xi
263
Chapter 2: State-of-the-Art
Existing Technologies for Condition Monitoring of
Construction Materials and Bridges
R. E. Green, Jr
17
29
44
Page vi
79
101
117
129
148
183
Page vii
209
219
231
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
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
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
(BAM)
Germany
Prof. Dryver
Huston
(Univ. Vermont)
USA
USA
(Rutgers Univ.)
USA
Prof. Philip
Perdikaris
USA
Dr. Urs
Sennhauser
(EMPA)
Switzerland
USA
USA
Page xiv
(Univ. Pavia)
Italy
UK
(Lawrence Tech.)
USA
(Photon Kinetics)
UK
USA
(CiDRA Corp.)
USA
Canada
Prof. Raymond
Measures
(Univ. Toronto
Canada
USA
(Electro Photonics)
Canada
USA
(NIST)
USA
USA
Page 1
Chapter 1
Research and Development Programs
Page 3
Page 4
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
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.
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
Page 10
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
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.
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
Page 18
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
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
Page 22
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
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
Page 27
Page 28
Page 29
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
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
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
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
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
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
Page 39
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
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
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
Page 46
Figure 1.
General Concept of Bragg Grating
Fabrication.
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-
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
Page 50
Figure 4.
Spectrum produced by a Bragg
grating subjected to strain
gradients along its length.
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
Page 52
Page 53
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,
Page 55
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
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.
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.
Page 59
Page 60
Page 61
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
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
Page 67
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
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.
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
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
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
Page 74
Figure 4.
Spectrum analyser trace.
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.
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
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
Page 78
13 M. Martinelli, "A universal compensator for polarization changes induced by
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
Page 79
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
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
and
Page 81
and
where
so that
or in integral form
Page 82
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
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
Page 85
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
a
A.H. Hartog, York Sensors Ltd. York House,
Premier Way, Abbey Park, Romsey SO51 9AQ, Hants, UK
Page 90
Figure 1:
Basic arrangement of a Raman Distributed Temperature Sensor
Page 91
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
Page 93
Figure 2:
Optical arrangement for a single mode DTS system
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.
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
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):
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,
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.
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:
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
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
Page 120
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).
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
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
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
Page 124
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.
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
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,
Page 129
Page 130
Page 131
Figure 1.
Sensor network layout.
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
Page 134
Figure 3.
Schematic of instrumentation and office link
Page 135
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
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
Page 138
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:
Page 140
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
Page 142
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.
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.
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
Page 147
5. Y.I., Andreyeva, N.Y.,Smirnova, 1991. "An Interference Fiber Optic Sound
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.
"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
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
Page 149
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
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.
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.
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
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
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.
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.
Page 164
Figure 6:
Schematic of FBG Sensor Installed Locations
Page 165
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
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
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
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.
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.
Page 177
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.
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
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
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
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.
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
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
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
Page 195
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
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
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
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
Page 206
Page 207
Chapter 6
Sensor Transduction Mechanism and Signal Recovery
Page 209
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
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)
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
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
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
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
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
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).
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
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
Page 232
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
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.
Page 234
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
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.
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
Figure 8
Transmission response of a reflection mode Elastica sensor before and after
large strain fatigue test, showing nearly identical response.
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).
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
Page 245
Figure 1.
System for measuring the light diffraction through meniscuses.
A photosensitive optic diode at the sample end detects the power of the
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
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
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
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 (*
Page 252
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
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
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.
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-
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 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
Page 264
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.