Long-period gratings for selective monitoring of loads on a wind turbine blade L. Glavind,1,2,* S. Buggy,1 J. Canning,3 S. Gao,3,4 K. Cook,3 Y. Luo,5 G. D. Peng,5 B. F. Skipper,6 and M. Kristensen2 1
Technology & Service Solutions, Vestas Wind Systems A/S, Hedeager 42, 8200 Aarhus N, Denmark 2
3
Department of Engineering, Aarhus University, Finlandsgade 22, 8200, Aarhus N, Denmark
interdisciplinary Photonics Laboratories, School of Chemistry, 222 Madsen Building F09, The University of Sydney, Sydney, NSW 2006, Australia 4
Center for Optical and Electromagnetic Research, Zhejiang University, Hangzhou 310058, China
5
Photonics and Optical Communications, The University of New South Wales, Sydney NSW 2052, Australia 6
Aarhus School of Engineering, Aarhus University, Finlandsgade 22, 8200, Aarhus N, Denmark *Corresponding author:
[email protected] Received 2 April 2014; accepted 1 May 2014; posted 15 May 2014 (Doc. ID 208963); published 18 June 2014
An optical fiber sensor based on long-period gratings (LPG) for selective measurements of flap- and edgewise bending of a wind turbine blade is presented. Two consecutive LPGs separated by 40 mm interfere to improve resolution and reduce noise in a D-shaped fiber. The mode profile of the device was characterized experimentally to provide a model describing the mode couplings. The sensor was tested on a wind turbine blade. © 2014 Optical Society of America OCIS codes: (050.2770) Gratings; (060.2370) Fiber optics sensors; (120.3180) Interferometry. http://dx.doi.org/10.1364/AO.53.003993
1. Introduction
Over the past decade, wind turbines have developed rapidly in size and complexity, and now have rotor diameters above 150 m. The wind turbine industry has shown interest in structural monitoring for both the optimization of power production from the wind turbine and the detection of fatigue failure in the field [1]. Optical fiber sensors are an ideal candidate for such applications; they are small and made from silica glass, which has excellent properties compatible with embedding into glass–fiber structures [2], and they have been shown to introduce minimal risk of delamination [3]. Moreover, they are nonconductive and immune to electromagnetic interference,
1559-128X/14/183993-09$15.00/0 © 2014 Optical Society of America
which is of significant importance for any instrumentation in a wind turbine blade because of the high risk of lightning strikes, especially for offshore wind turbines. Fiber-Bragg gratings (FBGs) [4] operating as strain sensors in the blades have been demonstrated [5], and commercial FBG-based systems for structural monitoring of wind turbine blades are available. However, FBGs are sensitive to both flapand edge-wise bending of the wind turbine blade. A sensor which can discriminate between the flap- and edge-wise bending is of particular interest to reduce the complexity of the measurement system. Here, we demonstrate long-period gratings (LPGs) as a selective sensor for optimization of the flap-wise monitoring on static loads of a full-scale wind turbine blade. By inscribing LPGs in a D-shaped fiber [6], the sensor provides bend direction selectivity. The general principle of the interaction between a particle and 20 June 2014 / Vol. 53, No. 18 / APPLIED OPTICS
3993
0.008 0.006 0.004 ∆n
two consecutive oscillatory fields, originally developed by Ramsey [7] and here acting as a Mach– Zehnder (MZ) configuration of two LPGs, has been utilized [8], providing narrow interference resonances for high-resolution detection of the sensing parameter and reduced sensitivity to noise. This approach overcomes some of the resolution problems associated with conventional LPG configurations. In this work, we focus on building a model for the functionality and performance of the sensor and demonstrating the capability with measurements on a full-scale wind turbine blade.
0.002 0.000 -0.002 -5
2. Fabrication of the Sensor Element
5
Fig. 2. Index profile of preform for D-shaped fiber, before milling down to D-shape [10].
2 0
Transmission [dB]
The D-shaped fiber was spliced to SMF-28 fiber for launching and collecting the light, and the LPGs were inscribed in the D-shaped fiber [9] (core: GeO2 ∼ 4.5 mol: %, P2 O5 ∼ 1 mol: %), without hydrogen loading. The D-shaped fiber is single mode in the 1500 nm range, with a MFR of w ∼ 5.4 μm at 1550 nm. The diameter of the flat side of the D-shaped fiber is d ∼ 156 μm, with a distance from the center of the core to the flat side of 8 μm. The index difference of the core and cladding is Δn ∼ 0.00456. Figure 1 shows the cross-section of the D-shaped fiber. The index profile of the preform, before milling down to a D-shape, is shown in Fig. 2. Note the bulge near the core due to different material properties. Two identical ∼10 dB LPGs were written through an amplitude mask (50∶50 duty cycle, Λ − 600 μm) using an ArF exciplex laser (λ 193 nm). The second LPG was written 40 mm from the end of the first grating (60 mm center to center). An unpolarized ASE source was used. The D shape of the fiber promotes birefringence, leading to induced polarization interference ripples and UV-induced phase differences visible during writing. The LPGs are therefore written stronger than 3 dB (which would provide maximum interference visibility in the lossless case), so that they are clearly visible during the fabrication process. Figure 3 shows the spectrum for the transmission of both LPGs, providing narrow fringes near 1560 nm with a spectral separation of ∼6 nm. These
0
Radius [mm]
-2 -4 -6 -8 -10 -12 1500
1520
1540
1560
1580
1600
Wavelength [nm] Fig. 3. Normalized spectrum of MZ configuration in D-shaped fiber (1 nm resolution) [10].
were measured using an erbium doped fiber amplifier and an optical spectrum analyzer (OSA, resolution 1 nm). Additionally a second LPG device was written (50∶50 duty cycle, Λ − 600 μm, 40 mm LPGs), where the fundamental center wavelengths of the two LPGs were spectrally misaligned so that it does not operate as a MZ device. Figure 4 shows the transmission spectrum of the first LPG and Fig. 5 shows the transmission spectrum of the device (both LPGs), with a dip near 1510 and 1530 nm for the two LPGs, respectively. The feature at 1560–1580 nm (Figs. 4 and 5) is subject to future investigation. 3. Mode Profile Characterization and Model of the Sensor
Fig. 1. Cross-section of an illuminated D-shaped fiber. The D-shaped fiber also has an inner cladding deposited and visibly seen around the core [10]. 3994
APPLIED OPTICS / Vol. 53, No. 18 / 20 June 2014
A relatively simple qualitative model to describe the sensor and the resonance wavelengths for the couplings between the modes was created to understand the sensing mechanism in greater detail. The model is based on experiments and describes the couplings between the core and cladding modes to predict the resonance wavelengths in the spectrum, including the ones expected outside our detection equipment
0
Transmission [dB]
-2 -4 -6 -8 -10 -12 1500
1520
1540
1560
1580
1600
Wavelength [nm] Fig. 4. Normalized spectrum of one LPG in the D-shaped fiber (1 nm resolution).
Fig. 6. Core mode in D-shaped fiber at 1494 nm. The black line illustrates where the flat side of the fiber is. Distribution in pixels: 1 pixel ≈0.14 μm, from [10].
2
Transmission [dB]
0 -2 -4 -6 -8 -10 1520
1540
1560
1580
1600
Wavelength [nm] Fig. 5. Normalized spectrum of two LPGs at 1510 and 1530 nm in D-shaped fiber (1 nm resolution) [10].
range. To measure the mode profiles in the D-shaped fiber from Fig. 5, it was cleaved at the end of the second grating and mounted in front of an infrared camera, with a collimation lens (×20 microscope objective) between the cleaved fiber end and the camera [10]. The mode-field patterns, in transmission, were measured when launching a 2 mW tunable laser source into the fibers through the SMF-28 fiber (∼1550 nm, resolution: 0.1 nm step size, 100 kHz linewidth). Figure 6 shows primarily excitation of the core mode at 1494 nm and Fig. 7 shows primarily excitation of the inner cladding mode at 1509 nm (for Figs. 6–9, conversion is 1 pixel ≈0.14 μm). In both figures, the flat side is located at the top (illustrated by the black line) near the edge of the mode profiles. Based on the data in Figs. 6 and 7, the intensity distribution of the modes in the core, inner cladding, and air can be approximated. Figure 8 shows a crosssection of the core mode intensity perpendicular to the flat side. The left side is toward the flat side; note the mode is squeezed on this side, compared to the side away from the flat side. The mode field radius (MFR) is 5.4 μm and the core radius is 4.2 μm. With
Fig. 7. Inner cladding mode in D-shaped fiber at 1509 nm. The black line illustrates where the flat side of the fiber is. Distribution in pixels: 1 pixel ≈0.14 μm, from [10].
100
Normalized intensity
-12 1500
80 60 40 20 0
0
200
400
600
800
Distribution [pixels] Fig. 8. Normalized core mode intensity perpendicular to the flat side. Distribution in pixels: 1 pixel ≈0.14 μm.
the measured profile and the MFR (intensity ∼13.5%), it is estimated that about 80% (2%) of the mode intensity is in the core region and the rest is in the cladding and air. With a distance from MFR 20 June 2014 / Vol. 53, No. 18 / APPLIED OPTICS
3995
Fig. 9. Second cladding mode in D-shaped fiber; the black line illustrates the flat side of the fiber. Distribution in pixels: 1 pixel ≈0.14 μm.
to the flat side of 2 μm, it is estimated that the portion of the light in air is very low, and therefore close to 20% of the rest is in the inner cladding. The portion of light in the core and cladding is less critical than the portion in air due to the much higher refractive index (RI) difference from silica to air than between the core and cladding. Based on the cladding mode profile, it is assumed that all three intensity peaks have about 3%–4% of the intensity in the core (10% total), and close to 90% in the inner cladding (the intensity in the outer cladding region is almost negligible and the index difference to the inner cladding very small and is therefore ignored). It is assumed that the portion of intensity in air, from the mode peak away from the flat side, is negligible. When the bulge at the flat side is taken into consideration (see Fig. 1), and the fact that the cladding mode is not nearly as well confined as the core mode, the portion of the intensity of the cladding mode in air is assumed to be about the same as that of the core mode. As an approximation, the transverse electric (TE)-like mode (polarization) has the E field parallel with the surface of the flat side and the transverse magnetic (TM)-like mode is perpendicular to the surface of the flat side [11]. From Maxwell’s equations and the boundary conditions for TE- and TM-like fields, the change in the normal components of the D field (D εE) across an interface should equal the surface charge density. Here, both materials (silica and air) are dielectric materials without any significant electrical potential difference; thus, the surface charge can be approximated to zero. Therefore, ε1 En1 ε2 En2 . Here, the relative dielectric constant of silica is approximately ε n2 ≈ 2.1, and for air it is 1. Therefore, the TM-like field in air can be approximated as
TM-like−cladding 3996
nair 2 TM-like−air ; ncl
APPLIED OPTICS / Vol. 53, No. 18 / 20 June 2014
(1)
where nair and ncl are the RI of the air and the cladding, 1 and 1.44815 respectively. Therefore, the E field is 2.1 times larger in air than in the cladding, and the intensity is 4.2 times larger for the TM-like mode. The E field decreases exponentially away from the interface (and the exponential tail is very short due to the large index step) [11]. As mentioned, the TE-like mode is considered to have the E field parallel to the surface (intensity distribution in the material is assumed the same). Therefore, there are no changes in the field and intensity over the interface. Hence, the intensity for the TE-like mode in air is, as an approximation, 4 times smaller than that of the TM-like mode. Here, the cladding modes are assumed to be all in the inner cladding, based on the mode profiles in Figs. 5 and 6, the cross-section in Fig. 1, and the profile index in Fig. 2, where it can be found that the distance from the edge of the inner cladding to the core center is about 35 μm. With these assumptions, a parameterization can be done as shown in Table 1. Here, η is the portion of intensity in air. To find the resonance wavelength for the couplings from core to cladding mode, the parts of intensity in each of the three regions are multiplied with the given RI (1.48271, 1.44815, 1, respectively). For the coupling from the TM-likecore mode to TE-likecladding mode, the difference in the effective RI is Δneff 0.8 · 1.45271 0.2 − η · 1.44815 η 1 1 − 0.1 · 1.45271 0.9 − η · 1.44815 η 4 4 ⇕ Δneff 0.0031927 − 0.3361125 · η:
(2)
The effective index difference can be found experimentally using [12] λm neff − nm cl Λ.
(3)
Where neff is the effective index of the core mode, ncl m is the RI of cladding mode of order m, and Λ is the spatial grating period. Using the equation, Δneff is found to be 0.0025 for a dip at 1509 nm and with Λ 600 μm. Therefore, η is found to be ≈0.2%. Thereby, 0.2% of the core mode (and 0.05% of the cladding mode) is in the air for the TM-like– TE-like coupling, in qualitative agreement with
Table 1.
Parameterization of Core and First Cladding Mode
Parameterization of Intensity in the Regions
TM-likecore mode TM-likecore mode TM-likeCladding mode TM-likeCladding mode
Intensity in Core
Intensity in Cladding
Intensity in Air
0.8 0.8 0.1 0.1
0.2 − 1 · η 0.2 − 1∕4 · η 0.9 − 1 · η 0.2 − 1∕4 · η
1·η 1∕4 · η 1·η 1∕4 · η
Table 2.
Parameterization of Second Cladding Mode
Parameterization of Intensity in the Regions
TM-likeCladding mode TM-likeCladding mode
Intensity in Core
Intensity in Cladding
Intensity in Air
0.07 0.07
0.93 − 0.5 · η 0.93–2 · η
0.5 · η 2·η
Transmission [dB]
earlier models [13]. The other couplings are not physically possible with Λ 600 nm, as it requires the RI of air to be negative to be within the spectral range of our detection equipment. For η ∼ 0.2%, the other couplings have resonances in the area 1900– 2300 nm. It has been verified in the lab that there was no LPG for a 300 μm period (which would have indicated that our grating would have been a secondorder grating resonance), and therefore this cannot be a solution for the couplings (e.g., Δneff ≈ 0.005). The RI of air is much lower than the cladding and core RI, and therefore only small changes of the portion of the light intensity in air gives large changes on the resonance wavelength, e.g., a 10% relative change in the portion of light in air corresponds to a change in wavelength of 15 nm for the TM-like– TE-like coupling. For a coupling to a TM-like cladding mode, the sensitivity would be increased due to larger distribution of light in air. During profile measurements, there was a tendency to see an extra cladding mode near 1560 nm on the monitor. As seen in Fig. 9, there is a weak second resonance intensity peak near the core peak (core mode dominating at this wavelength). This mode has positive amplitude on the left side of the core and probably negative amplitude on the right side of the core (asymmetrical field distribution). It was not possible to get clear excitation of this cladding mode, but Fig. 9 indicates a second mode with intensity on the left side, possibly due to interference with the dominant cladding mode. Based on the same assumptions as for the first cladding mode, geometry and mode confinement of such a cladding mode profile, it is roughly estimated to have about 100% more intensity in air and about 30% less intensity in the core, compared to the first cladding mode, and is parameterized as shown in Table 2. For this cladding mode, all fundamental resonances are above the 1500 nm range. However, the TE–TM-like coupling is at about 3000 nm, and therefore it would have a second-order resonance in the 1500 nm range (these are weak compared to the fundamental). The absolute wavelength is difficult to quantify, as only 10% change of the intensity in air corresponds to a 100 nm change in the resonance wavelength. The first cladding mode has a symmetrical field distribution and the second cladding mode has an asymmetrical distribution; therefore, the overlap integral is very small, and hence the coupling between them is negligible.
0 -2 -4 -6 -8 -10 -12 -14 -16 -18 1100
1200
1300
1400
1500
1600
1700
Wavelength [nm] Fig. 10. Normalized broadband spectrum of a LPG in MZ configuration.
The spectrum, of the sensor in Fig. 3 has been investigated from 1100 to 1700 nm, with both tunable lasers and broadband light sources, to find other eventual cladding modes. Figure 10 shows the spectrum using a broadband light source (MenloSystems TC-1550-B). The waveguide cutoff of the multimode is seen where it is predicted at ∼1250 nm. No second cladding mode outside the first resonance at 1560 nm was observed; hence, the resonance dip is very small and not possible to observe or it is out of the detection equipment range. The dips at 1400 nm are due to gas absorptions inside the OSA. If the sensor is recoated before embedding, n would be changed by the RI of the recoating material for the TM-like mode due to the continuity of the D-field. For the TE-like mode, n changes by approximately RI3 due to Maxwell’s equations and the above assumptions. Using this on the two modes for a recoating material with a RI of 1.37, the change for the first cladding mode would be up to 15 nm; for the second cladding mode, the change would be several hundred nanometer. 4. Test on Wind Turbine Blade
The sensor shown in Fig. 3 was embedded between two fiber glass base plates utilizing the material used in the matrix of a wind turbine blade [2,9]. The embedded sensor was tested in the lab and showed sensitivity to bending/curvature of ≈9 × 10−9 m2 in both directions, and the temperature sensitivity were measured to be 4.1 nm∕100°C. The sensor was subject to load tests on an 80 m long, 34 ton prototype wind turbine blade for the Vestas 8 MW V164 wind turbine [14]. The sensor was mounted on the inside of the blade (flat side against flap forward). The sensor was located near the optimum location for maximum curvature of the blade structure (when subject to load). For comparison, an FBG strain sensor was located within 2 m of the LPG sensor (but on opposite side of the neutral axis on the same side of the blade); the FBG was interrogated with a commercial FBG sensing system. The blade was fixed on a test rig, to enable static tests of the blade in all directions, forward and reverse, in both flap- and edge-wise directions (see Fig. 11). A setup with a broadband light source “DenseLight DL-CS5169A” and an “Ibsen I-MON 512 E” interrogator was selected. The interrogator 20 June 2014 / Vol. 53, No. 18 / APPLIED OPTICS
3997
Reverse Flap
Reverse Edge
Forward Flap
Sensor Location
Fig. 11. Illustration of sensor placement on blade, from [14].
provides 512 points over the spectrum (1510– 1595 nm, ∼6 sample points per nm), and is customized for FBG sensing and provides for FBGs a fitted resolution of 0.5 pm. The light source is a superluminescent diode and is polarized; therefore, a polarization control was utilized to optimize the spectrum. The light source has an optical power of 16 mW, and it saturated part of the spectrum outside the dip; however, it was still possible to interrogate the sensor (see Fig. 12). Note that the output from the interrogator was linear and with arbitrary units. The detection equipment was located outside the wind turbine blade. As the sensor is fixed to the blade, the deformation of the blade influences the measurement parameters; hence, the measured parameter is a combination of strain and bending picked up by the sensor. The “noise” was interrogated. The spectra were investigated where the dip was fitted to a 10th order polynomial and the noise was analyzed to lead to a wavelength uncertainty of ∼Δλ 3 pm, and it is expected mostly to be natural oscillation from the blade caused by the test environment. The blade was rotated on the test rig to be in position for load testing in the forward edge direction of the blade. With winches and clamps along the blade, it was loaded to give set points and held for a short time, first at 50% (based on Newton meters, Nm, of maximum designed load). The initial start loads are calculated on the self-weight of the blade and the
weight of the clamps (for a point located 1 m away from the sensor location). As the blade is not symmetrical (the blade is designed to take loads primarily in the forward flap direction), the initial loads are different for different orientations. As the start loads are calculated, a slight offset can be expected compared to the measured data. The spectrum was optimized for the dip near 1552 nm with the polarization control before each test. Figure 13 shows the fitted center wavelength, λ, according to the given loads (55 pm in total spectra shift). Figure 14 shows the strain response from the FBG; note that the first load point is calculated, and is with an offset. However, if comparing the response from the LPG and FBG directly, the offset is negligible. It is seen that near 90% load (−300 με), there is a change in the sensitivity. The change in sensitivity is due to overlap of the fields between the cladding and core mode when the fiber is bent (the intensity in the modes interacts when the geometry is changed due to bending). From 0 to 300 με, the LPG sensitivity is approximately 0.08 pm∕με, and above it is approximately 0.2 pm∕με. Note the spectral change for the LPG and FBG is with opposite
1552.31
Wavelength [nm]
Forward Edge
1552.29
1552.27
1552.25
1552.23 40
60
80
100
120
Load [%] Fig. 13. Forward edge: LPG center wavelength compared to load.
7
x 10
4
100
5
0
4
-100
Strain [ µ ε ]
Transmission [arb. unit]
6
3 2 1 0
-300 -400
1520
1540
1560
1580
Wavelength [nm] Fig. 12. Spectrum with interrogator and light source utilized for measurements on blade, from [14]. 3998
-200
APPLIED OPTICS / Vol. 53, No. 18 / 20 June 2014
-500 40
60
80
100
120
Load [%] Fig. 14. Forward edge: FBG strain compared to load.
1552.5
1551.94
Wavelength [nm]
Wavelength [nm]
1552.3 1551.90
1551.86
1552.1 1551.9
1551.7 1551.5
1551.82 40
60
80
40
100
sign, which is due to material composition and the D shape of the fiber. Figure 15 shows the fitted center wavelength, λ, according to the given loads (−90 pm in total spectral shift) for the reverse edge test. Figure 16 shows the FBG response for the reverse edge test. The wavelengths are moving toward shorter wavelengths, opposite to the forward edge test. Again, there is a small change in the LPG sensitivity near 300 με; however, it is much smaller than for the forward edge test. From 0 to 300 με, the LPG sensitivity is approximately −0.2 pm∕με, and above it is approximately −0.16 pm∕με. Figure 17 shows the fitted center wavelength according to the given loads (833 pm in total spectral shift) for the reverse flap test. Figure 18 is the corresponding FBG response. The wavelengths are moving toward longer wavelengths. The wavelength movement is expected to be much larger than for the edge-wise bending, as the sensor is directionally sensitive and the load in the flap direction is higher. The sensitivity is approximately −0.99 pm∕με (close to the typical 1.2 pm∕με for a FBG, but with opposite sign). Compared to the edge-wise bending, the sensitivity is about a factor of minimum ∼5 larger.
500
Strain [ µ ε ]
400 300 200 100
40
50
60
70
80
100
80
90
100
Load [%] Fig. 16. Reverse edge: FBG strain compared to load.
Fig. 17. Reverse flap: LPG center wavelength compared to load.
200 0
Strain [ µ ε ]
Fig. 15. Reverse edge: LPG center wavelength compared to load.
0
60
Load [%]
Load [%]
-200 -400 -600 -800 -1000 40
60
80
100
Load [%] Fig. 18. Reverse flap: FBG strain compared to load.
Different from the edge-wise test, there is no change in sensitivity (initial start load is not taken into account as it is calculated and gives no offset when comparing the LPG and FBG directly). During the flap forward test, an issue occurred; it is likely that the transport fibers were twisted while rotating the blade to the flap forward position. Therefore, there was a significant drop in transmission power and possibly a change in the polarization. Furthermore, the FBG sensor suffered a critical error just above 95% load where the optical power was lost (due to a mechanical splice failure). However, in a commercial version of the sensor, the fiber will be embedded directly into the blade material, providing high stability. The issue with twisting and splice fragility would not be a problem. Moreover, measuring on the left slope of the dip, the change in wavelength from 19% load to 95% load is −2127 pm, and for the FBG the change is 1546 μe, giving a sensitivity of −1.37 pm∕με. These figures have some uncertainty because of the cabling and twisting issue under test, but compared to the reverse flap test, this indicates that the sensor might be about 37% more sensitive in the flap forward direction. 20 June 2014 / Vol. 53, No. 18 / APPLIED OPTICS
3999
The sensitivity was ∼7 times larger for flap-wise than edge-wise measurements. The directional sensitivity is due to the D-shaped design and offset core. The core expands differently in flap- and edgewise directions, due to this asymmetrical off-center D-shape design, which gives different expansion for strain/bend of the core- and cladding-material and thereby sensitivity as the resonance depends of the difference between the effective indexes. During download of the blade, negligible hysteresis was observed in all tests. 5. Discussion
The sensor demonstrated a suppression of the edgewise load of a factor ∼5 for some directions and curvature; this could be further enhanced, as the sensitivity changes due to mode field overlap. Here, the off-centered core and D shape means that the sensor differs from the typical fiber FBG response, which is isotropic and therefore equally sensitive in all directions. Multiple normal FBGs in a centered core standard fiber would be required to monitor different orientations and achieve similar functionality. A very interesting aspect is the change in sensitivity, shown at the edgewise test. This might have the possibility to either enhance the sensitivity in a given direction or suppress it, for example, with a prebend sensor or with utilization of two sensors with opposite properties (one prebent and the other one straight) to outcompensate temperature or strain. This requires future research and development of the fiber to provide the optimum geometry, core off-center, and mode profile, such that a better suppression is possible. The response of the sensor differs for the forward and reverse directions; this is due to the overlap of the mode profiles and it is also affected by a small difference in the fiber and location on the blade. Moreover, the sensor can be designed such that the overlap between the field of core and cladding mode is optimized, such the selectivity can be further increased. This can be of significant importance to suppress edge wise bending, providing less complex software due to a selective sensor. The asymmetric shape of the fiber is responsible for exciting unusual asymmetric modes, which would not happen in a standard fiber. The sensor has a total spectral range of about 20 nm. Since the D-shaped fiber is single mode from 1250 nm and the transport fiber has a relatively short length (∼100 m) and thereby loss is negligible, the entire telecommunication band from 1260 to 1675 nm (O- to U-Band) can be utilized. Therefore, applications with at least 20 distributed sensors are possible. Depending on the application, between 2 and 16 FBG sensors are usually utilized for blade load sensing [1]. As this LPG sensor is selective to the bending moments, a reduced number can be anticipated, perhaps up to 8 sensors. With optimization of the gratings, integration with typical FBG sensing systems is also possible. 4000
APPLIED OPTICS / Vol. 53, No. 18 / 20 June 2014
6. Conclusion
In summary, a full-scale test of the LPG sensor on a prototype blade for a Vestas V164 wind turbine (80 m, 34 tons) was conducted. The test successfully demonstrated a novel prototype of an embedded LPG sensor which is capable of providing measurements with the suitable sensitivity, directionality, and resolution for this application. Compared to traditional FBG sensors, this type of sensor has great advantages in selectivity to compensate for either edge-wise or flap-wise bending of particular interest for blade monitoring. It also provides the possibility to compensate for strain and temperature effects through careful sensor packaging and design. The Authors thank Vestas Wind Systems A/S and the Australian Research Council (ARC, Grant FT110100116) for funding. Shaorui Gao thanks the China Scholarship Council (CSC) for a scholarship and support under the State Scholarship Fund. The authors also thank the Department of Industry, Innovation, Science and Research (DIISR), Australia, for support in an International Science Linkages (ISL) project (CG130013) and the ARC for two LIEF grants (LE0883038 and LE100100098) that helped to establish the National Fibre Facility at UNSW. References 1. L. Glavind, I. S. Olesen, B. F. Skipper, and M. Kristensen, “Fibre-optical grating sensors for wind turbine blades: a review,” Opt. Eng. 52, 030901 (2013). 2. L. Glavind, S. Buggy, I. Olesen, B. Skipper, J. Canning, K. Cook, and M. Kristensen, “Direct embedding of fiberoptical load sensors into wind turbine blades,” in Advanced Photonics, K. Ewing and M. Ferreira, eds., OSA Technical Digest (online) (Optical Society of America, 2013), paper SM3C.6. 3. K. Krebber, W. Habel, T. Gutmann, and C. Schram, “Fibre Bragg grating sensors for monitoring of wind turbine blades,” Proc. SPIE 5855, 1036–1039 (2005). 4. K. O. Hill, Y. Fujii, D. C. Johnson, and B. S. Kawasaki, “Photosensitivity in optical fibre waveguides—Application to reflection filter fabrication,” Appl. Phys. Lett. 32, 647–649 (1978). 5. K. Schroeder, W. Ecke, J. Apitz, E. Lembke, and G. Lenschow, “A fibre Bragg grating sensor system monitors operational load in a wind turbine rotor blade,” Meas. Sci. Technol. 17, 1167–1172 (2006). 6. D. Zhao, X. Chen, K. Zhou, L. Zhang, I. Bennion, W. N. MacPherson, J. S. Barton, and J. D. C. Jones, “Bend sensors with direction recognition based on long-period gratings written in D-shaped fibre,” Appl. Opt. 43, 5425–5428 (2004). 7. N. F. Ramsey, Molecular Beams (Oxford University, 1956). 8. E. Dianov, S. Vasiliev, A. Kurkov, O. Medvedkov, and V. Protopopov, “In-fibre Mach–Zehnder interferometer based on a pair of long-period gratings,” in 22nd European Conference on Optical Communication (IEEE, 1996), pp. 65–68. 9. L. Glavind, S. Gao, K. Cook, J. Canning, B. Skipper, Y. Luo, G. Peng, and M. Kristensen, “Enhanced resolution of longperiod grating bend sensor,” Proc. SPIE 8924, 892437 (2013). 10. L. Glavind, J. Canning, S. Gao, K. Cook, G. D. Peng, Y. Luo, B. F. Skipper, and M. Kristensen, “Long-period gratings in special geometry fibres for high resolution and selective sensors,” Opt. Eng. (to be published).
11. N. Skivesen, A. Tetu, M. Kristensen, J. Kjems, L. H. Frandsen, and P. I. Borel, “Photonic-crystal waveguide biosensor,” Opt. Express 15, 3169–3176 (2007). 12. V. Bhatia, D. K. Campbell, D. Sherr, T. G. D’Alberto, N. A. Zabaronick, G. A. T. Eyck, K. A. Murphy, and R. O. Claus, “Temperature-insensitive and strain-insensitive long-period grating sensors for smart structures,” Opt. Eng. 36, 1872–1876 (1997).
13. D. Marcuse, F. Ladouceur, and J. Love, “Vector modes of D-shaped fibres,” IEE Proc. Optoelectron. 139, 117–126 (1992). 14. L. Glavind, S. Buggy, B. F. Skipper, J. Canning, S. Gao, K. Cook, Y. Luo, G. Peng, and M. Kristensen, “A longperiod grating sensor system monitoring loads on a wind turbine blade,” submitted to OSA BGPP conference.
20 June 2014 / Vol. 53, No. 18 / APPLIED OPTICS
4001
