Abstract
The degradation of concrete structures is always accompanied by the formation of cracks. If cracks are detected at an early stage before significant
degradation occurs, life of the structure can be extended by the performance of relatively low cost but timely maintenance measures. To detect
and monitor cracking in concrete structures, where the crack locations are often not known in a-priori, a distributed optical crack sensor based
on OTDR measurement of bend loss has been developed. In this paper, the sensing principle will be discussed and the fabrication of sensors for
surface attachment and internal embedment is described. Experimental results are shown to demonstrate that the optical power loss versus crack
opening relation at a particular point of bend loss is independent of the number of bends along the fiber. Then, the monitoring of single and multiple
cracks under quasi-static loading is illustrated, followed by the detection of cracks under restrained dimensional change (caused by shrinkage or
thermal effects). Finally, the sensor performance under cyclic loading is studied. Based on the test results, the potential of the sensor for practical
applications is demonstrated.
© 2006 Elsevier B.V. All rights reserved.
0924-4247/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.sna.2006.09.004
K.T. Wan, C.K.Y. Leung / Sensors and Actuators A 135 (2007) 458–464 459
sensors require knowledge of the exact crack locations [9,10], will both exhibit a sudden drop across the crack. By utilizing
which cannot be predicted for concrete structures due to material the optical time domain reflectometer (OTDR), which measures
inhomogeneity. Other sensors are only able to detect the integral the Rayleigh backscattered signal as a function of time, we can
strain along the gauge length [11,12]. For these sensors, if the locate the crack positions from the time corresponding to the
selected gauge length is too large, it is not possible to differenti- signal loss (distance = time × light velocity in the waveguide)
ate between one widely opened crack and many small harmless and the crack opening from the magnitude of the drop. A typical
ones. If the gauge length is small, many ‘gauges’ are required to plot of the backscattered signal versus time, both before and after
cover the plausible cracking region, and this not be realistic in crack formation, is shown in Fig. 1b. Before any crack forms,
practice. Brown et al. [13] and Oka et al. [14] worked on a new the gradual declination of the backscattered signal is solely due
technique that measured the distributed strain along the optical to attenuation. At the turning points of the fiber, power loss
fiber with Brillouin Optical Time Domain Reflectometer. Cracks may also be observed. To minimize these losses, the curvature
formed along the fiber will be revealed as peaks in the strain ver- at the turning points should be reduced. After the cracks open,
sus time plot. However, as high local strain at a crack may lead sharp drops correspond to the crack locations along the fiber will
to fiber debonding, the measured strain represents an average appear in the plot.
value over a debonded region, the length of which is hard to pre-
dict (because the interfacial properties between optical fiber and 3. Experimental verification of the sensing principle
concrete is hard to characterize). Quantitative information on
crack opening, which is important for durability considerations, As shown in Fig. 1b, it is possible to use a single fiber to
will be difficult to obtain. detect and monitor a number of cracks. For sensor applications,
Considering the limitations of various sensing approaches it is important to ensure that the loss across a crack is inde-
discussed above, Leung et al. [15,16] have developed a novel pendent to losses at other cracks or bent regions (such as the
distributed crack sensor based on optical time domain reflec- fiber turning points). Theoretically speaking, if light in the fiber
tometry. To detect cracks with this sensor, a-priori knowledge propagates in single mode, the energy distribution over the fiber
of crack locations is not required. Also, a single fiber is capable cross-section is similar before and after a bent, although the total
to detect and monitor a number of cracks. In the following sec- energy is reduced. The ratio of each drop (in decibels) should
tions, the sensing principle is first explained. Then, results from merely depend on the bent curvature at that position. In our case,
recent experimental investigations will be described to demon- the bend radii vary from less than 1 mm to infinity within very
strate the practical applicability of the sensing concept. short distance (about half of the crack opening). To check the
validity of the assumption, a simple experiment is conducted and
2. Sensing principle its schematic diagram is shown in Fig. 2. Two epoxy blocks are
made as crack simulators. Before the block is cast, an inclined
The principle of the sensor is illustrated in Fig. 1. In concrete, steel wire with diameter slightly larger than the fiber is placed
a highly inhomogeneous material, it is impossible to predict in the mold. After the epoxy hardens, the steel wire is pulled
exact crack locations from theoretical analysis. However, the out and a hole is left for insertion of optical fiber. The OTDR
crack orientation can be reliably determined. An optical fiber (Opto-Electronics OFM-20) is linked to the crack simulator with
can hence be coupled to the structure in such a way that it is a spool of fiber between the OTDR bulkhead and crack simula-
inclined to the crack (Fig. 1a). Once the crack opens, the optical tor so the very weak backscattered signals from the cracks will
fiber needs to bend to maintain its continuity as shown in the not be overwhelmed by the strong pulse created at the bulkhead
inset of Fig. 1a. The bend will induce loss of light power from connection. In this case, optical fiber 3M FS-SN 4224 is used,
the fiber core to the surroundings. As a result, the transmitted for which the light propagates in single mode at 850 nm wave-
light in the fiber core, as well as the backscattered optical signal, length. The crack simulators are opened horizontally one after
another, with the second crack simulator opens first. Afterwards,
the first crack simulator is opening and then the fiber is bent into
a loop in front of the two crack simulators to induce an addi-
tion loss. After each step, the OTDR scans the optical fiber to
obtain the backscattered light intensity versus distance along the
fiber (the distance is obtained from the calculated value from the
internal program of the OTDR by inputting the refractive index
Fig. 1. Concept of distributed sensing with the novel sensor. Fig. 2. Test configuration to demonstrate sensor consistency.
460 K.T. Wan, C.K.Y. Leung / Sensors and Actuators A 135 (2007) 458–464
the mold is also coated by releasing oil so that the fiber can
slide freely and bend to introduce optical power loss after the
sheet breaks. Before the polyester hardens, fine sand is added.
These particles will sink into the polyester to make the sheet
more brittle. Once a crack in the concrete member intersects
with the sheet, it will crack at once. (From our experience, the
plate can be broken when the crack opening in concrete is about
0.08–0.1 mm.) For applications requiring an externally bonded
plate (e.g., monitoring of flexural cracks at the bottom of a bridge
deck), the preparation procedure described above is sufficient.
For an internally embedded sensor, stone aggregates are also
added to the polyester before it hardens (as shown in Fig. 2).
Fig. 3. Zoomed-In OTDR record of different number of cracks.
These aggregates, which protrude from the surface of the sensor
sheet, will improve the bond when the sheet is embedded inside
of the fiber core). The result is shown in Fig. 3. The y-axis of
concrete. At the location where the fiber is coming out of the
the plot is in log-scale so that the same decibel loss is shown
sheet, a small steel tube is placed around the fiber for protection
as equal drop magnitude. From the test results, it is clear that
and the end of the steel tube is further covered with soft rubber.
the loss at one location does not influence that at other part of
This way, the breakage of fiber at the exit point can be prevented.
the fiber. The small discrepancy (e.g., the variation of loss at
the second crack from 5.6 to 5.8 dB) is mainly due to the signal
4.2. Determination of the relation between optical power
noise. Also, the separation of each drop in the plot matches the
loss and crack opening
physical separation.
To obtain the optical power loss versus crack opening relation,
4. Power loss versus crack opening relation
the sensor sheet is embedded inside a concrete beam at a dis-
tance from the bottom. The flexural test configuration is shown
The power loss versus crack opening relation of the sensor
schematically in Fig. 5. To avoid brittle failure of the concrete
is the key ‘parameter’ for its applications in crack monitoring.
beam, two 6 mm diameter mild steel rods are placed on the ten-
Theoretical model has been derived and verified by experiments
sile side. A pair of notches is cut on the both sides of the beam,
[16]. For practical applications, an effective way to couple the
so that the crack location is prescribed. LVDT’s are installed at
optical fiber to the concrete member needs to be developed. For
both sides of the beam to measure the crack opening at the level
convenient sensor installation under site conditions, and fiber
of the sensor. A number of tests have been conducted using sen-
protection against the alkaline concrete environment, the fiber
sor sheets containing Corning SMF28 (light propagates in single
can be first embedded into a polyester plate to form a sensor
mode at 1550 nm) and 3M FS-SN 4224 (light propagates in sin-
sheet, which can then be coupled to the concrete structure later.
gle mode at 850 nm) fibers running at 30◦ or 45◦ to the crack
The fabrication of the sensor sheet is described next.
opening direction. The optical and mechanical properties of the
two fibers are summarized in Table 1. For single mode propaga-
4.1. Fabrication of sensor sheet
tion, light sources with corresponding wavelength are employed.
During the test, forward optical power across the sensor sheet is
The preparation of the sensor sheet is illustrated in Fig. 4.
measured with power detectors (Anritsu MA9611A for 1550 nm
Round pegs are placed at proper location for holding the fiber in a
and MA9412A for 850 nm wavelength) that are connected to an
pre-determined configuration. With the fiber held tight, polyester
optical power meter (Anritsu ML9001A).
is poured slowly into the mold to form a sensor sheet about one
The measured optical power loss versus crack opening of
and half millimeters in thickness. Before the polyester is added,
the sensors made with SMF28 fibers are given in Fig. 6. Fig. 7
releasing oil is applied on all the internal surfaces of the mold
shows the results for sensors made with 3M fibers. Also shown
and the pegs for easy removal of the sheet. The fiber inside
in the figures are results from a theoretical model that consists
Fig. 5. Testing configuration to obtain the signal loss versus crack opening rela-
Fig. 4. Illustration for the fabrication of a sensor sheet. tion for the sensor sheet.
K.T. Wan, C.K.Y. Leung / Sensors and Actuators A 135 (2007) 458–464 461
Table 1
Properties of optical fibers used in the tests
Characteristic 3M FS-SN-4224 Corning SMF 28
Table 2
Summarized result of the shrinkage crack detection
Day Optical Optical power Crack opening Theoretical crack
power (W) Loss (dB) (mm) opening (mm)
0 1.093 – – –
2 0.835 1.2 0.2 0.15
3 0.870 1.0 0.16 0.14
4 0.798 1.4 0.22 0.16
5 0.855 1.1 0.2 0.15
Fig. 11. Variation of optical power loss, crack opening and loading with time.
464 K.T. Wan, C.K.Y. Leung / Sensors and Actuators A 135 (2007) 458–464