Professional Documents
Culture Documents
net/publication/228540400
Conference Paper in Proceedings of SPIE - The International Society for Optical Engineering · March 1997
CITATIONS READS
28 1,113
4 authors, including:
Pascal Kronenberg
TFB Romandie SA
27 PUBLICATIONS 526 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Pascal Kronenberg on 23 January 2015.
ABSTRACT
The monitoring of dams represents an important task in the management of hydroelectric systems. Their economic, social
and environmental value imposes to know well the real behavior of the structure and its foundations. This paper shows in
two practical cases the possibility to improve the quality of deformation measurements by an appropriate fiber optics
sensor network.
The first case is a study showing the technical and economical feasibility to install an extended, spatial fiber optics
deformation sensor network to detect the relative deflection of an entire shell dam. At this purpose the theoretic study has
been evaluated on the base of typical load situations with their effective deflections on the Schiffenen dam, a shell-
shaped concrete structure near Fribourg (Switzerland).
The second case concerns the development and realization of two long fiber optics deformation sensors (30 m and 40 m)
anchored in the rock to monitor the displacement of the dam relatively to its underground. These sensors have been
installed in the Emosson shell dam (Switzerland).
Monitoring always has been, due to the economic, social and environmental significance of hydroelectric structures, a
substantial task in appropriate dam management. It is fundamental in order to guarantee the safety of the structure and its
users but also to optimize the exploitation and the maintenance of the dam.
The reasons that can guide to outfit dams with a fiber optics deformation sensor network have to be considered in two
groups. Firstly there are many existing dams which are requiring additional monitoring either to replace an existing
system the measurements of which don’t satisfy any more, or to extend an existing monitoring network. These are
installed to look closer at a particular, initially underestimated structural behavior (e.g. crack and joint opening,
displacement of the dam’s foundation) or simply to increase the redundancy of the monitoring network. Secondly there
are the new or the structurally upgraded (e.g. raising) dams where a monitoring network has to be installed. For these a
fiber optics deformation sensor network can be integrated in the project from the beginning, which is in fact much easier
because the sensors have fewer constraints relative to their installation.
-
P.K. Email: kronenberg@imac.epfl.ch
SPIE Conference on Smart Structures and Materials, 5-6.03.1997, San Diego, USA
Indeed these arguments are valid for any monitoring system. But the common advantage in installing fiber optic sensors
instead of “classical” sensors is linked to the particular characteristics of the monitoring system we have chosen. The
SOFO system1 is insensitive to many environmental influences such as a variation of temperature, strong humidity or
electromagnetic fields. These performances are actually predestined to the hostile environment found in dams.
As varied the possible types of application are, as different can be the design of a fiber optic sensor to monitor them.
Following we will present two typical applications which use both the same measurement system but are implemented
with two different types of sensors.
2.1 Description
The first case is a theoretical study done on the Schiffenen shell-mass dam near Fribourg in Switzerland to show the
feasibility of monitoring the internal radial deflection of the structure with a spatial fiber optics deformation sensor
network [8].
The Schiffenen dam is a 42 m high and 370 m long revolution bodied structure with a constant radius of 200 m at the
axis of the coping. The transversal section is curvy shaped, 7 m (coping) to 13 m (footing) thick and invariable up to the
borders where it is slightly more massive to better distribute the forces to the foundation.
The idea is to get the curvature of the shell by a pair of standard fiber optics deformation sensors placed at different
points. Then, by double integrating this curvature information and introducing the appropriate boundary conditions it is
possible to extract the radial deflection of the dam. In the SPIE paper “Mathematical model for the determination of the
Sup. ARCH
SHELL
Key WALL
Figure 1 – Splitting of the shell structure in order to use the beam models to retrieve its
deflection
vertical displacement from internal horizontal measurements of a bridge” [4] we have seen that this model works well for
linear structures such as beams, but is it also applicable to bi-dimensional shells? Indeed it is, we only have to consider
the shell as a net of perpendicular, one-dimensional arches and walls we analyze separately. The knots of this static
network are serving as points of adjustment to bring the deflections of the two subsystems together (fig. 1). This
technique has already been approved for the pre-dimensioning of shells.
The sensor setup chosen for this application gives only the internal deformation of the dam. In order to link this
deformation to an external reference we use the border conditions received from other monitoring systems such as
inclinometers, geodesy / GPS, pendulums or (fiber optics 2 ) extensometers. In the case of this study the shell has been
1
French acronym for “SURVEILLANCE D’OUVRAGE PAR FIBRE OPTIQUE ” [7]
2
See chapter 3. Emosson Shell Dam
SPIE Conference on Smart Structures and Materials, 5-6.03.1997, San Diego, USA
modeled by one arch-strip at the coping acting as a horizontal, partially fixed ended
beam and by one wall-strip at the key acting as a vertical cantilever beam. As the
two subsystems are solicited just by distributed loads (water, temperature…), we can
simplify the mathematical model by considering each subsystem as only one section
(continuous load, geometrical continuity of the axis and continuous stiffness).
3
Cell = network element that gives one curvature information (in the current application -> 2 sensors / cell)
SPIE Conference on Smart Structures and Materials, 5-6.03.1997, San Diego, USA
switch to one(!) fiber optic cable, which is connected to the reading unit situated near the dam. Using this configuration
the complete structure can be monitored without any manual intervention in the dam.
xi ± dx
v( x) = ∫∫ ρ( x, y, L, ∆L )( x) dx 2
y i ± dy
Monte-Carlo v(x) ± dv
Simulation
Li ± dL
x i: Abscise coordinate of each
cell
y i: Cantilever distance between
∆Lj ±d∆L 2 sensors of a cell
Li: Length of cell (sensor)
v(x) ∆Lj: Deformation of sensor
v(x): Deflection
Figure 4 – Schema of the error simulation to retrieve the standard deviation of v(x)
This simulation has been executed for a given mean deflection by changing the free input values x and L; y is fixed by
the dam’s geometry, ∆L depends on the imposed mean deflection.
The standard deviations dx, dy, dL and d∆L are supposed to be constant, conform to the accuracy each variable can be
measured with.
Standard deviations
dx 50 mm
dy 5 mm
dL 25 mm
d∆L 5 µm + 1%
SPIE Conference on Smart Structures and Materials, 5-6.03.1997, San Diego, USA
The final result, after a few optimizations, has quite an encouraging aspect (fig. 5, 6).
35.0
5.0
v(x) [mm]
x [m]
0 50 100 150 200 250 300 350
-5.0
-15.0
-25.0
Figure 5 – Radial deflection of the coping with its simulated error distribution. A typical deflection
has been imposed.
When we calculate the two subsystems independently we get aware that in the knot of adjustment dv of the arch is much
more important than dv of the wall due to the lower sensor density in the arch. Instead of installing more sensors in the
arch it is much more interesting to take the top-deflection of the wall as a boundary condition for the mid-deflection of
the arch. In this way the arch can profit by the better accuracy of the wall.
-35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 0.0 5.0 10.0
0.64 mm v(x) [mm]
-2
x [m]
0.55 mm
-7
0.47 mm
-12 0.40 mm
-17 0.32 mm
0.25 mm
-22
0.18 mm
-27
0.12 mm
-32
0.06 mm
Schiffenen Dam (CH)
-37
Projected deflection of the key wall 0.02 mm
Error analysis
imposed deflection v(x)
-42 0.00 mm
standard deviation dv
Figure 6 – Radial deflection of the key wall with its simulated error distribution. The top deflection
serves as a boundary condition for the arch.
SPIE Conference on Smart Structures and Materials, 5-6.03.1997, San Diego, USA
Regarding the border conditions of the two systems we have introduced them as they are, e.g. without a special standard
deviation. They are defined by two pendulums on each side of the arch and by an inverted pendulum plus an inclinometer
at the footing of the wall. Only the wall-to-arch adjustment condition follows the standard deviation given by the wall
deflection. The fact that the arch is fixed by 3 boundary conditions makes that the position of the deflection curve is
hyper-determined in its external reference. We can see this by observing the borders where the deflection can’t follow the
fixed 0-standard deviation.
Compared to the standard deviations we used for the input the resulting dv is very satisfying. For the fixed setup dv
equals not more than 1 mm for the whole network, at the adjustment knot even less (dv = 0.7 mm ð dv/v = 3%).
Actually these results show well how a good sensor redundancy can improve the mediocre accuracy on x, y and L to get
a reliable accuracy on the deflection monitoring.
By doing a sensibility study it is possible to see that dy has much more influence on dv than dx or dL. It is therefore very
interesting to have a good accuracy on the measurement of y.
Finally, the optimized network setup that has been retained is indicated on the two plans. There is total sensor quantity of
36 pieces; their length is 3 m for the arch respectively 2 m for the wall, with a higher sensor concentration in the center of
the arch.
2.4 Conclusions
The error simulation shows, that the deflection accuracy can be much improved either by optimizing the hardware
(sensor network topology, sensor length, boundary condition sensors) or simply by adapting the software (calculation
model, definition of boundary conditions). For the retained setup the total installation cost (incl. 36 sensors, optical
switch and a reading unit) is cheaper then an equivalent, classical dam monitoring system. Further, the measurement
campaigns don’t cost anything since the system runs automatically.
It is important to insist that this project is not a real application but only a feasibility study, in which we suppose that all
the installed sensors and the structure work as prognosticated. In a real project there would certainly be a more complex
and denser network installed in order to prevent any incertitude due to sensor failure or unexpected structural behavior.
Concerning the structural behavior it would actually be indispensable to investigate more on structural discontinuities
like the vertical joints between wall blocs or geometrical “impurities” (i.e. overflows, connected tower elements or any
shape related details) by applying finite element modeling in order to adapt the network at these points.
SPIE Conference on Smart Structures and Materials, 5-6.03.1997, San Diego, USA
3.1 Description
This very practical application concerns the development and setup of two long SOFO sensors (30 m and 39 m) that have
to replace mechanic rockmeters anchored in
the underground of the Emosson shell dam in
the Swiss Alps. The Emosson dam is situated
near Martigny just on the borderline to
France at about 1900 m.o.s. It is 180 m high
and at its coping is 550 m long, with a
thickness varying from 10 m (coping) to 35
m (footing) (fig. 7).
4
For long sensors there is a special design required where the loose reference and the stressed measurement fiber have to be separated
to avoid an interaction (friction, mechanical coupling) [2].
SPIE Conference on Smart Structures and Materials, 5-6.03.1997, San Diego, USA
Separating tube
Microtube with
reference fiber
PVC mounting
tube
Microtube with
measurement
fiber
In the dam the sensor can then be (un-)anchored by simply bending this torsion-rigid tube at the borehole-entrance. On
the other end of the measurement zone (in the dam gallery) the upper anchoring piece of the sensor is gliding freely out
of the mounting tube and is attached to the reference plate by a wedging screw. In the upper anchoring piece (d = 8.5
mm) we have also placed the coupler and the necessary splices.
The assembly of the sensor is quite labor-intensive. The length of the sensor makes it quasi imperative to have 2 people
working for a whole day for each sensor. This is actually one of the major problems with long sensors compared to short
ones [5]. On the other hand this task is also quite delicate. The fibers have to be cut very precisely in order not to be out
of range while scanning.
MEASUREMENT Fiber
(pre-stress 0.5%) WEDGING SCREW
Coupler
Monitoring zone
3.178" A/D Filter Ampli
Mobile mirror µ
Photo- LED
Diode 1300nm
Internal PC
When preparing a sensor in the lab it is very important to be aware of the climatic conditions present on the site.
Especially for a dam gallery where the relative humidity goes up to 90% and the temperature is all the year at about 3 to
4 °C this information has to be considered when the dimensions or the materials of a sensor are defined. The shortening
of the mounting tube due to the temperature variation lab -> site for
example can easily attain a few centimeters… and the tube will
disappear in the hole!
To centralize the monitoring of the new sensors, which are each on
one side of the dam, optic
extension cords
have been installed.
Now the two
sensors are ready to
be scanned from
one point without
having to displace
the reading unit.
Using this setup it is
even possible to
monitor the
sensors
automatically
over a longer
period without
any human
intervention.
3.4 Results
As the sensors have only been installed by the end of 1996 the measurement campaign is just about to take off.
Unfortunately, by the deadline of this paper the measurements are too few to make any conclusions about the correctness
of their evolution. In a few months, when we have more measurements, a comparative study will be done between the
fiber optics sensors and neighboring mechanical extensometers in order to verify the results.
The schedule plans to monitor the sensors during a year (one load cycle) at monthly intervals, together with the usual
monitoring campaign of the dam. Then, in addition to this, a more frequent monitoring campaign is scheduled for a
month with automatic measurements every hour.
3.5 Conclusion
We have demonstrated that the realization of a long fiber optics extensometer for underground monitoring is possible,
even if the existing sensor context (borehole, anchoring system, length) can’t be modified. A well-designed sensor can be
completely mounted in the lab and can be installed on the site in less than an hour without having to touch any optical
fiber.
The Emosson project shows in an interesting way that the research and the development of new sensor types not
necessarily have to be done in lab conditions. The collaboration with industrial partners (in this case Emosson Electricity
Inc.) gives us the opportunity to check sensor equipment in a real environment under all the imposed severe constraints
(temperature, humidity, accessibility).
SPIE Conference on Smart Structures and Materials, 5-6.03.1997, San Diego, USA
ACKNOWLEDGMENTS
The authors are indebted to Prof. L. Pflug, Prof. R. Sinniger, M. Pedretti, R. Passera, L. Mouvet, J.-M. Rouiller, R. Delez,
G. Steinmann, X. Rodicio, A. Osa-Wyser, J. Hugon and the whole IMAC and EMOSSON teams for their help and useful
discussion. The Emosson dam project could be realized thanks to the financial support of Emosson Electricity Inc.
The SOFO research program is conducted under the financial aid of the Swiss CTI (Commission pour la Technologie et
l’Innovation) and the Board of the Swiss Federal Institutes of Technology.
REFERENCES
For further information on the SOFO project look at the following WWW home page:
http://imacwww.epfl.ch/projets/sofo/sofoa.htm
[1] “Low-coherence interferometry for the monitoring of underground works”, D. Inaudi, L. Vulliet, L. Pflug, S.
Vurpillot, A. Wyser, 1995 North American Conference on Smart Structures and Materials, San Diego February
1995, Volume 2444, 171-178
[2] “Development and laboratory tests of deformation fiber optic sensors for civil engineering applications”, L. Vuillet,
N. Casanova, D. Inaudi, A. Osa-Wyser, S. Vurpillot, International Symposium on Lasers, Optics and Vision for
Productivity in Manufacturing, Europto Series, Besançon, 10-14 June 1996
[3] “In-line coherence multiplexing of displacement sensors: a fiber optic extensometer”, D. Inaudi, S. Vurpillot , S.
Lloret, Smart Structures and materials, San Diego February 1996, SPIE Volume 2718-28
[4] “Mathematical model for the determination of the vertical displacement from internal horizontal measurements of a
bridge” S. Vurpillot, D. Inaudi, A. Scanno, Smart Structures and materials, San Diego February 1996, SPIE Volume
2719-05
[5] “Embedded and surface mounted fiber optics sensors for civil structural monitoring”, D. Inaudi, N. Casanova, P.
Kronenberg, S. Marazzi, S. Vurpillot; Smart Structures and Materials 97, San Diego, SPIE Vol. 3044
[6] “Development and field test of deformation sensors for concrete embedding”, D. Inaudi, S. Vurpillot, N. Casanova,
A. Osa-Wyser, Smart Structures and Materials 96, San Diego, SPIE Vol. 2721
[7] “Bridge Monitoring by Interferometric Deformation Sensors”, D. Inaudi, S. Vurpillot, N. Casanova; Laser
Optoelectronics and Microphotonics: Fiber Optics Sensors, SPIE, Beijing November 1996
[8] “SOFO: Application à un barrage-voûte en exploitation”, P. Kronenberg; Travail de diplôme 1996, IMAC-DGC /
EPFL
[9] “Smart structures and materials”, Brian Culshaw; Ed.1, Artech House 1996
[10] “Improvement of existing dam monitoring: recommendations and case histories”, Commission Internationale des
Grands Barrages; Paris 1992
[11] “Suggested Methods for Monitoring Rock Mouvments using Borehole Extensometers”, Int. society for
rockmechanics; Oxford a.o., 1978, International Journal of rock mechanics and mining sciences, vol. 15, no. 6,
pp.305-368