Sensors and Actuators A xxx (2006) xxx–xxx

Modal analysis and damage detection by Fiber Bragg grating sensors

P. Capoluongo a , C. Ambrosino a , S. Campopiano a , A. Cutolo a ,
M. Giordano b , I. Bovio c , L. Lecce c , A. Cusano a,∗
a Optoelectronic Division-Engineering Department, University of Sannio, Piazza Garibaldi No. 107, Benevento, Italy
b Institute for Composite and Biomedical Materials, National Research Council (IMCB-CNR), Piazzale E. Fermi No. 1, Portici (Napoli), Italy
c Department of Aeronautical Engineering, University “Federico II” Via Claudio No. 21, Napoli, Italy

Received 9 December 2005; received in revised form 12 April 2006; accepted 17 April 2006

Abstract
Damage detection and localization are critical issues for structural health monitoring. To this aim, many techniques have been proposed relating,
the presence of damage to variations of its dynamic features between the undamaged and damaged states. Since a greater information content is
localized at higher frequencies, sensing systems with adequate bandwidth and resolution are required. Here, Fiber Bragg grating sensors with an
adequate interrogation system are exploited to reveal the presence of damage on a structure. As preliminary step, modal analysis tests in a wide
frequency range are performed in order to verify the performances of these optic devices to retrieve high frequency structural dynamic features.
As reference sensors, laser Doppler vibrometers and accelerometers, representative of the state of the art for this application field were exploited.
Experimental results confirm the excellent performances of Fiber Bragg gratings, not only able to detect damage but also to discriminate between
different damage levels.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Fiber Bragg grating sensors (FBG); Dynamic measurements; Modal analysis; Frequency response function (FRF); Structural health monitoring (SHM);
Damage detection

1. Introduction tures extracted from the measured responses of the structure in

the undamaged and damaged states [4]. Since high frequency
The high complexity and costs of modern structural sys- modes are more effective to detect and localize the presence
tems, combined with their high operational reliability and safety small damage [2], one of the mandatory requirement for SHM
needs have brought to an increasing interest in new approaches surveillance systems is a wide bandwidth, joined to a very low
for structural health monitoring (SHM) and damage analysis intrusivity and, in order to simplify the overall architecture,
from both industrial and academic world. Current corrosion and high multiplexing capability [1–5]. Actually, as sensing systems
crack detection systems are based on acoustic emission moni- for these applications, non-contact systems like laser Doppler
toring and active methods such as ultrasound and modal analysis vibrometers or in situ electric devices like accelerometers are
[1,2]. This latter is an experimental methodology able to retrieve widely utilized.
the dynamic features (resonant frequencies, vibration shapes A valid alternative for SHM applications can be represented
and modal damping ratios) of a mechanical structure from its by fiber optic sensors, and in particular by Fiber Bragg grat-
frequency response function (FRF), i.e. the ratio between the ings (FBGs). Key points of this technology are reduced dimen-
Fourier transforms of the sensors time responses and of the exci- sions combined with the intrinsic capability to measure several
tation signals [3]. Based on this line of arguments, Schultz et al. parameters simultaneously; the high resistance to corrosion and
proposed SHM methods classifying the structures in various fatigue, good compatibility with the most advanced compos-
damage states based on the differences between vibration fea- ite materials exploited in the aeronautic and aerospace field,
immunity to electromagnetic interferences, the wide bandwidth
operation, an excellent multiplexing capability [6–10].
∗ Corresponding author. Tel.: +39 0824 30 58 46; fax: +39 0824 30 58 40. In recent works FBGs attitude to perform modal analysis
E-mail address: [Link]@[Link] (A. Cusano). on real structures for aerospace and aeronautic applications has

0924-4247/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/[Link].2006.04.018

SNA-5326; No. of Pages 10

2 P. Capoluongo et al. / Sensors and Actuators A xxx (2006) xxx–xxx

been demonstrated. Experimental modal analysis tests were car-

ried out on the aluminum mock up of a component for an artificial
satellite. In this case both FBGs and reference accelerome-
ters were bonded on the lower surface of this structure and
optic devices were able to detect resonant frequencies and pla-
nar vibrational shapes for the first two flexural modes, located
at about 72 and 120 Hz, in good agreement with preliminary
numerical simulation and reference sensors outputs [11].
Moreover, FBG sensors performances as embedded sensors
in a real composite structure for aeronautical applications were
Fig. 1. Schematic co-ordinate systems used for strain-optic calculation.
tested: these devices were deployed in the spar of the wing of
a composite materials aircraft model. In this case a complex
strain field, due to their position, was expected. Accelerome- optic matrix, where i = 2, 3 is the polarization axis of the light
ters were exploited as reference sensors. Several modes, also propagating along the optic fiber. For bonded configuration and
with a complex shape, were retrieved. Also in this case a good dynamic tests involving lower strain field amplitudes, it can be
agreement with numerical simulations and results obtained by safely assumed that only the strain field parallel the fiber axis
reference sensors was obtained. Retrieved modes for this test significantly affects the FBG response. This means that the trans-
article belonged to range up to 170 Hz [12,13]. verse strain is uniform (ε2 = ε3 ), and related to the axial one by
In this work the attention is focused on the FBGs capability to the Poisson’s ratio. In this way, by exploiting also the symmetry
be exploited as SHM in situ sensors. In particular, the intent is to properties of the pij for the optic fibers [6], (8) becomes indepen-
verify whether these device can detect resonant modes frequency dent of the polarization of the light and the Bragg wavelength
and amplitude variations in a wide frequency range. To this aim, relative shift is proportional to axial strain:
results of damage detection tests on an ad hoc steel structure are  
reported. Modal frequency and amplitude variations on a series λB n20
= 1 − [p12 − ν(p11 + p12 )] ε1 = Sε ε1 (3)
of resonances belonging to the structure in the undamaged and λB 2
damaged states are described. As previous step, results of high
frequencies (up to kilohertz) modal analysis tests carried out on where ν is the Poisson’s ratio and Sε is FBG strain sensitivity.
the same structure are reported. Typical values of strain–optic matrix elements and Poisson’s
ratio for standard optic fibers are: p11 = 0.12, p12 = 0.27 and
2. Fiber Bragg gratings as sensing elements and their ν = 0.17, while Sε = 0.78 × 10−6 ␮␧−1 [6] (Fig. 1).
application to the modal analysis method For these optic devices the interrogation system adopted in
the modal analysis tests relies on a low cost ratiometric technique
A Fiber Bragg grating is a periodic or semi-periodic perma- based on optical filtering combined with broadband interroga-
nent perturbation of the refractive index of the core of an optic tion. By using a chirped and strongly apodized FBG, an optic
fiber [6]. As effect, when a grating is irradiated with a broadband filter with a response linear in wavelength can be obtained
optical source, a narrow band pass spectrum signal is reflected. [15,16]. The block diagram of this system is shown in Fig. 2,
The central wavelength of this signal, called also “Bragg wave- while the transmitted spectrum of the involved filter are shown
length”, λB , is related to the physical parameters of the grating in Fig. 3.
by the following relationship [14]: With reference to the scheme in Fig. 2, each voltage signal, vT
and vR , converted by the photodetectors is related to the convo-
λB = 2nΛ (1) lution between the sensing grating spectrum and, respectively,
where n is the effective refractive index of the mode propagating the transmission and reflection responses of the optical filter
inside the fiber and Λ is the grating period. Each external cause, involved in. The electronic processing of these two detected sig-
able to modify right hand terms of (1), causes a shift of the Bragg nals consists of taking the ratio of the difference and the sum of
wavelength. In particular, when the FBG is subjected to strain the outputs as follows [15]:
fields, relative variations of λB induced by n or Λ changes can be vT − KvR
obtained by differentiating (1) respect to these two parameters, vn = (5)
vT + KvR
and substituting the relationships obtained applying the strain-
optic theory [6]. If complex strain is experienced by a FBG, shift The vn can be called “normalized voltage”. In this procedure an
in the central wavelength taking into account polarization effects additional coefficient K is introduced in order to compensate the
occurs according to the following expression: asymmetry in the receiving scheme due the additional coupler

 used for the reflected signal vR , the optical losses introduced
n2 
3
λB Λ n by the filter and to equalize the filter slope in transmission and
= + = ε1 − 0 pij εj (2)
λB i Λ n 2 reflection modes. In the case of linearly varying filter, the so gen-
j=1
erated normalized signal is a good measure of the centroid of the
where n0 is the effective refractive index of the unstrained, not sensing grating spectrum, and as a consequence, also of Bragg
birefringent optic fiber, pij is the generic element of the elasto- wavelength [15]. Moreover, it is not dependent on the intensity
 P. Capoluongo et al. / Sensors and Actuators A xxx (2006) xxx–xxx 3

Fig. 2. Schematic of the optoelectronic set-up used for broadband interrogation and optical filtering.

fluctuations but directly related to the Bragg wavelength shift, A fundamental aspect related to the exploitation of this class
which in turn is strongly related to the local applied strain along of sensors to dynamic tests is represented by the application
the grating region, as can be observed in (6): of modal analysis, a methodology able to relate the FBG sen-
  sors responses to the dynamic features of the structure under
n20 investigation. Basic element of the modal analysis approach are
vn = Kvn λB = Kvn λB0 1− [p12 − ν(p11 + p12 )] ε1
2 reported in [3]. In principle a mechanical structure is approxi-
(6) mated with a linear system having as inputs excitation signals
and as outputs related sensors responses. Strain or displacement
where Kvn is a calibration constant and λB0 is the Bragg wave- maps can be retrieved associating output values to the locations
length of the unperturbed grating. By properly eliminating inten- of sensors, disposed on adequate location grids. Mechanical
sity variation effects due source instability, optical losses and parameters can be obtained by considering the features of the
unwanted microbending effects along the optical chain, it is ratio between the Fourier transform of the response of the struc-
possible to maintain the advantageous intensity insensitive oper- ture obtained at the jth output location under an excitation signal
ation of this class of sensors. In light of the passive nature of the applied at the kth input, when all other inputs are set to zero, and
proposed technique, system bandwidth is only limited by the the Fourier transform of this excitation signal. In this way a
electronic circuitry involved in the receiving unit. Such a sys- “frequency responses function” (FRFs) Hjk (ω), with ω angu-
tem is capable of a static resolution of 1 ␮␧ and a dynamic one lar frequency, can be obtained. Arranging these function in a
of 40 n␧/(Hz)1/2 @ 50 KHz [15]. matrix, the total response of the system can be obtained. It can
be shown after simple mathematical passages, that each FRF
in an around of rth vibrational mode resonant pulsation has the
following expression [3]:
j
Φr Φkr
Hjk (ω ≈ ωr ) ≈ (−ω2 + ωr2 )
| − ω2 + ωr2 + j 2ωωr ζr |2
j
Φr Φkr
−j 2ωωr ζr (7)
| − ω + ωr2 + j 2ωωr ζr |2
2

where, with reference to the rth mode, ωr is the resonant pul-

j
sation, ζ r the damping coefficient, Φr , Φkr are, respectively, the
jth and the kth component of the related eigenvector. It can be
assumed that if the structure is lightly damped (i.e. the damping
coefficient tends to zero), it all the components of each eigenvec-
+j
Fig. 3. Spectral response in transmission and reflection of the grating based tor (Φr , Φ+k r ) are real quantities [3], and natural frequencies (or
5 nm (1547.5–1552.5 nm) optical filter. pulsations) can be evaluated from zero crossings of the real part
4 P. Capoluongo et al. / Sensors and Actuators A xxx (2006) xxx–xxx

of (7), or otherwise from corresponding to peaks of its imaginary of FBG sensors, they measure strain components, thus a strain
part. These latter give also another information: the amplitude frequency response function (SFRF), treated in the same way
of the structure response (displacement or strain) at the point as the FRF to retrieve the modal feature of the structure under
where the sensor is located [3]. For the mechanical structures of investigation.
interest, excitations and sensors roles can be inverted as a con-
sequence of the Maxwell’s reciprocity principle [3]: this means 3. Sample description
that only a sensor can be placed, while a mobile excitation sys-
tem can be utilized on the points of the grid. In order to perform preliminary dynamic tests, a sample struc-
In the common practice, as outputs for modal analysis tests ture obtained joining two steel bars was exploited. One bar,
acceleration, velocity or displacement are considered. In the case called “A” in Fig. 4(a) is an AISI 4340 steel sample. It has a
rectangular 2.5 cm × 5 cm base and a height of 90 cm. It is hol-
low and the thickness of every walls is about 2 mm. A second thin
AISI 4340 steel beam, called “B” in Fig. 4(a), was placed perpen-
dicularly to the first one and soldered at its midspan. Dimensions
of this latter are 44 cm × 2.5 cm and 5 mm of thickness. As optic
sensors, a FBG with a 1.2 nm FWHM bandwidth, a central wave-
length of 1550.00 nm and a physical length of 1 mm was bonded
with a cyanacrilate glue at a distance of 34.5 cm from the upper
constrain of the beam, in the midspan of the front face of the “A”
steel beam. Another FBG with a 1.2 nm FWHM passing band, a
central wavelength of 1550.00 nm and a physical length of 1 mm
was glued at the centre of the horizontal beam, soldered to the

Table 1a
Data of resonant frequencies obtained from peaks of the imaginary part of
vibrometer’s FRFs for the retrieved 12 vibrational modes
Mean [Hz] Standard Standard deviation/
deviation [Hz] mean [A.U.]

1 114.460 0.129 1.130E−03

2 135.460 0.129 9.500E−04
3 185.630 0.297 1.600E−03
4 205.260 0.320 1.560E−03
5 316.430 0.176 5.600E−04
6 626.000 1.069 1.710E−03
7 660.000 0.189 2.900E−04
8 727.200 0.368 5.100E−04
9 1011.700 1.265 1.250E−03
10 1072.800 1.699 1.580E−03
11 1274.800 0.922 7.200E−04
12 1502.100 0.387 2.600E−04

Table 1b
Data of resonant frequencies obtained from peaks of the imaginary part of FBG’s
SFRFs for the retrieved 12 vibrational modes
Mean [Hz] Standard Variation (σ/µ) [A.U.]
deviation [Hz]

1 114.051 0.129 1.130E−03

2 135.021 0 0
3 185.296 0.320 1.730E−03
4 204.832 0.254 1.240E−03
5 316.083 0.129 4.100E−04
6 625.264 0.838 1.340E−03
7 659.670 0.372 5.600E−04
8 727.814 0.621 8.500E−04
9 1010.191 1.541 1.530E−03
10 1072.168 1.593 1.490E−03
Fig. 4. A scheme of the mechanical structure under investigation (a and b) a 11 1274.699 1.102 8.600E−04
shot where the exploited piezoelectric element and the beam of the vibrometer 12 1502.501 1.132 7.500E−04
are represented.
 P. Capoluongo et al. / Sensors and Actuators A xxx (2006) xxx–xxx 5

mount and to a greater mass which can represent a perfect bond- 4.2. Experimental test procedure
ing for this light structure. Given the constrains, the horizontal
beam can be considered as fixed—a fixed beam with a strain To perform these tests, 15 excitation points were arranged
maximum at its midspan. in a line, traced on the “A” beam (see Fig. 4(b)) with the first
point located at 10 cm from the upper constrain. The spatial
4. Modal analysis tests spacing of about 5 cm, was sufficient for an efficient mode
retrieving.
4.1. Experimental set-up One excitation point (the sixth) was in correspondence of
the FBG. The approach based on the evaluation of both the
As reference sensor a Polytec PSV-400 laser Doppler vibrom- FRF and SFRF was exploited to retrieve resonant frequencies
eter is utilized to perform one-point measurements of the and vibrational shapes. For each excitation point, five hammer
dynamic displacement component out of the plane. The scan- impacts were performed and five excitation–output pairs sig-
ning laser beam was focused to the same location as the bonded nals are acquired in order to perform adequate averaging on the
FBG grating. It is a costly system, representing the state of the obtained data.
art for modal analysis tests. The interrogation technique for the
exploited FBG is represented by a broadband system with a 4.3. Results and discussion
passive optic filter with a linear range of 7 nm (see Fig. 3).
As mechanical excitation system, an instrumented hammer was In Tables 1a and 1b all the resonant frequencies numeric
exploited. This device can furnish a mechanical impulsive input values obtained, respectively, from vibrometer and FBG are
providing a flat bandwidth excitation up to 3.5 kHz. It can be reported. They are the mean of the resonance frequencies evalu-
easily moved from an excitation point to another one, allowing ated from the peaks of imaginary parts of FRFs (SFRFs) belong-
to retrieve modal shapes. All the signals, excitation and both ing to all the measurement points. For each, sensor small values
sensors responses, were synchronously acquired from a DAQ of standard deviation and of standard deviation–mean ratios
system. indicate a good agreement between all detected resonances.

Fig. 5. Comparison between vibrational shapes belonging to the mode at 316 Hz: Fig. 6. Comparison between vibrational shapes belonging to the mode at 727 Hz:
(a) vibrometer output and (b) FBG output. (a) vibrometer output and (b) FBG output.
6 P. Capoluongo et al. / Sensors and Actuators A xxx (2006) xxx–xxx

Comparing results obtained by the two devices, the revealed

resonances have the same value, up to 1500 Hz.
In Fig. 5(a and b) vibrational shapes of a low frequency mode
(316 Hz) are shown for both sensor. A good matching between
the obtained results can be observed. Another modal shape, at
727 Hz, shows a good agreement between responses obtained
by both sensors (see Fig. 6(a and b)). The last detected mode,
corresponding to a resonance at 1502 Hz is shown too. And
also in this case retrieved vibrational shapes are the same (see
Fig. 7(a and b)). For all the reported modes, the two sensors give
similar results even if they measure different parameters. Any-
way, results obtained in this case from FBG sensors demonstrate
their capability to retrieve modal parameters in a wide frequency
range, and to perform damage detection tests, as described in the
following sections.

5. Damage detection tests

5.1. Experimental set-up

For the damage detection tests, the “B” beam shown in

Fig. 4(a) is considered. Mechanical excitation was given by
the piezoelectric element bonded on the right edge of the beam
(see Fig. 4(b)) in light of its capability to generate an excita-
tion with wider frequency than the instrumented hammer. This
device is fed with a 100–10 kHz frequency chirp having a 100 V

Fig. 7. Comparison between vibrational shapes belonging to the mode at

1502 Hz: (a) vibrometer output and (b) FBG output.

Fig. 8. A scheme of the mechanical structure under investigation where locations Fig. 9. A comparison between the FRFs (a) and the SFRFs (b) obtained for the
of both damages are highlighted. three states (undamaged, one damage, two damages).
 P. Capoluongo et al. / Sensors and Actuators A xxx (2006) xxx–xxx 7

peak to peak amplitude, obtained as output of a power amplifier 5.3. Results and discussion
driven by a function generator. In this case on the rear side of
the “B” beam, in correspondence of the FBG, an accelerometer In Fig. 9(a and b) FRFs and SFRFs for the three states (undam-
was placed as reference sensor. Since this latter has a resonance aged, with one damage, with two damages) are reported. An
at about 7 kHz, the frequency range up to 5.5 kHz was consid- arbitrary offset is added to improve the quality of the represen-
ered. During this test all the signals, excitation and both sensors tation. In Table 2 numerical values of frequency and amplitude
responses, were synchronously acquired from a DAQ system. of the detected resonances for the three states are reported.
For all the parameters shown in this table, percentage varia-
5.2. Experimental test procedure tions on the original value (i.e. related to the undamaged state,
indicated with “UD”) are reported. For the amplitudes, sub-
In this case, FRFs (SFRFs) exploited for modal parameters jected to greater variations, the percentage differences between
evaluations are obtained as an average of eight sweeps per- the first damage, “D1”, and the second one, “D2” are also
formed on the structure. In order to improve the quality of the shown.
obtained SFRFs, responses obtained from the FBG sensor, fil- The first frequency under investigation is located at about
tering algorithms based on the Hankel matrix and on the single 317 Hz (see Fig. 10(a and b)). No frequency change is induced
value decomposition (SVD) of a matrix [17] are used. by the presence of damage (see Table 2), while amplitude vari-
After a first acquisition on the undamaged test sample, clay ations can be observed for both sensors (see Table 2): for the
masses were added in order to simulate the presence of damage. accelerometer a 14.16% difference can be observed at the rise
Two masses are added in successive steps, the first close to the up of the first damage, while the second one causes a 12.39%
sensors and the second far away from their localization (see difference. For the FBG the first damage causes an amplitude dif-
Fig. 8). ference of 21.94%, while the presence of the second one causes

Table 2
Comparison between detected modal frequencies and amplitudes for both sensors
Undamaged (UD) frequency [Hz] 1 Damage (1D) frequency [Hz] D1 − UD [%] 2 Damage (2D) frequency [Hz] D2 − UD [%]

Accelerometer—frequencies
317.19 317.19 0.00 317.19 0.00
625.00 623.44 −0.250 623.44 −0.250
992.19 990.63 −0.157 990.63 −0.157
1060.94 1059.38 −0.147 1057.81 −0.295
2096.88 2081.25 −0.745 2081.25 −0.745
3800.00 3784.38 −0.411 3784.38 −0.411
3821.88
FBG—frequencies
317.19 317.19 0.000 317.19 0.000
625.00 623.44 −0.250 623.44 −0.250
992.19 990.63 −0.157 990.63 −0.157
1062.50 1059.38 −0.294 1057.81 −0.441
2096.88 2081.25 −0.745 2081.25 −0.745
3798.44 3782.81 −0.411 3784.38 −0.370
3820.31

Related frequency Undamaged amplitude 1 Damage (D1) D1 − UD [%] 2 Damage (D2) D2 − UD [%] D2 − D1 (%)
[Hz] (UD) [A.U.] amplitude [A.U.] amplitude [A.U.]

Accelerometer—amplitudes
317.19 0.113 0.129 14.159 0.127 12.389 −1.550
625.00 0.108 0.106 −1.852 0.113 4.630 6.604
992.19 0.24 0.359 49.583 0.328 36.667 −8.635
1060.94 0.227 0.222 −2.203 0.225 −0.881 1.351
2096.88 1.325 1.636 23.472 1.643 24.000 0.428
3800.00 0.774 0.761 −1.680 0.811 4.780 6.570
3821.88 0.742
FBG—amplitudes
317.188 2.780E−03 3.390E−03 21.942 2.650E−03 −4.676 −21.829
625.000 2.000E−03 2.090E−03 4.500 1.990E−03 −0.500 −4.785
992.188 2.410E−03 3.540E−03 46.888 2.800E−03 16.183 −20.904
1062.500 1.810E−03 1.660E−03 −8.287 1.520E−03 −16.022 −8.434
2096.875 6.920E−03 8.130E−03 17.486 7.250E−03 4.769 −10.824
3798.438 3.080E−03 3.210E−03 4.221 2.990E−03 −2.922 −6.854
3820.313 3.220E−03
8 P. Capoluongo et al. / Sensors and Actuators A xxx (2006) xxx–xxx

a decreasing, −4.78%; the difference between the two damaged

states is 21.83%.
Fig. 11(a and b) shows the changes in the natural frequencies
for mode at 625 Hz. After the first damage, the same frequency
change is observed by both sensors, while no modification is
observed at the second one. Amplitudes changes can be observed
for this mode: the accelerometer demonstrates a −1.85% differ-
ence at the first damage and an increasing of 4.63% for the
second one. The difference between the two damaged states is
6.60%. For the same mode, the FBG has a 4.50% difference
for the first damage, and a small difference for the second one
(−0.50%), while the difference between the amplitudes induced
by the two damaged states is 4.79%.
Fig. 12(a and b) shows the changes in the natural frequencies
for two shortly spaced modes at 992 and 1060 Hz. With refer-
ence to the 992 Hz mode, after the first damage, the same percent
frequency shift is observed by both sensors, −0.16%. Also in
this case the second damage is not sensed by both devices. About
modal amplitudes, for the accelerometer, the first damage causes
a strong increasing of 49.58%, respect to the undamaged struc-
ture, while for the second alteration, a difference of 36.67%, is
observed. The optic sensor exhibits a 46.89% difference between
the undamaged and damaged state, similar to the accelerome-
ter’s one, and also a good difference at the rising of the second
damage, 16.18%, can be observed.

Fig. 10. Comparison between the FRFs (a) and SFRFs (b) amplitudes for the
317.9 Hz resonant frequency.

Fig. 11. Comparison between the FRFs (a) and SFRFs (b) amplitudes for the Fig. 12. Comparison between the FRFs (a) and SFRFs (b) amplitudes for the
625.0 Hz resonant frequency. 992.2 and 1060 Hz resonant frequencies.
 P. Capoluongo et al. / Sensors and Actuators A xxx (2006) xxx–xxx 9

6. Conclusions

The aim of this work is to verify FBG sensors capability to

detect the presence of damage in mechanical structure reveal-
ing changes of dynamic features in a wide frequency range. To
achieve this goal, preliminary modal analysis tests are performed
on a structure with high resonances. As result, FBGs can retrieve
modal shapes in good agreement with the reference sensors up
to the frequency of 1.5 kHz.
As successive step, damage detection tests are performed.
Presence of damage is revealed by difference of modal frequen-
cies and amplitudes. For the modes associated to higher peaks of
both FRF and SFRF amplitudes, a good agreement in frequency
shifts is obtained by both sensors. For the most part of the inves-
tigated resonances, the presence of a second damage, localized
further from sensor position was not associated to a frequency
shift. FBG amplitudes, as for the reference sensor, showed dif-
ferences for each one of the considered states. Results obtained
during these tests witnessed the possibility to exploit these optic
devices for SHM applications.

References

[1] S.W. Doebling, C.R. Farrar, M.B. Prime, A summary review of

vibration-based damage identification methods, Shock Vibr. Dig. 30 (2)
(1998) 91–105.
[2] S.W. Doebling, C.R. Farrar, M.B. Prime, D.W. Shevitz, Damage Iden-
tification and Health Monitoring of Structural and Mechanical Systems
from Changes in their Vibration Characteristics: A Literature Review.
Los Alamos National Laboratory, Report LA-13070-MS, May 1996.
[3] W. Heyelen, S. Lammens, P. Sas, Modal Analysis Theory and Testing,
Fig. 13. Comparison between the FRFs (a) and SFRFs (b) amplitudes for the Katholieke Universiteit, Leuven, 1997.
2096 Hz resonant frequency. [4] W.L. Schulz, J.P. Conte, E. Udd, J.M. Seiz, Seisimic damage identifica-
tion using long gage fiber Bragg grating sensors, in: Second Workshop
on Advanced Technologies in Urban Earthquake Disaster Mitigation,
July 2000, Kyoto, Japan, 2000.
[5] I. Bovio, E. Monaco, M. Arnese, L. Lecce, Damage detection and
For both sensors the mode at 1060 Hz, detected at 1062 Hz for health monitoring based on vibration measurements and recognition
FBG sensor, is the only mode able to discriminate in frequency algorithms in real-scale aeronautical structural components, in: Fifth
the presence of the second damage: the first one causes a percent International Conference on Damage Assessment of Structures (DAMAS
change of −0.147% for the accelerometer and −0.294% for the 2003), Southampton, UK, Trans. Tech. Publications Ltd., July, 2003.
[6] R.M. Measures, Structural Monitoring with Fiber Optic Technology,
FBG, as it can be observed in Table 2. The second structural Academic Press, London, 2001.
alteration causes a −0.295% difference for the accelerometer [7] B. Culshaw, J. Dakin, Optical Fiber Sensors: Principle and Components,
and −0.441% for the FBG. Related amplitudes exhibit small Artech House Inc., Norwood, 1988.
differences between the first detected damage (−2.20%) and the [8] R.M. Measures, Smart composite structures with embedded sensors, J.
second one (−0.881%) for the accelerometer. The FBG presents Compos. Eng. 2 (1992) 597–618.
[9] E. Udd, 1-st fiber optic smart structures and skins conference, SPIE 986
a difference of −8.29% for the first damage, and −16.02% (1988).
for the second one. As expected from the literature, higher [10] E. Udd, Fiber Optic Smart Structures, John Wiley and Sons, New York,
order modes are more sensitive to the presence of damaged 1995.
area [2]. [11] A. Cusano, P. Capoluongo, S. Campopiano, C. Ambrosino, M. Gior-
Fig. 13(a and b) shows the changes in the strongest detected dano, M. Caponero, A. Paolozzi, F. Felli, Dynamic Measurements on a
Star Tracker Prototype of AMS Using Fiber Optic Sensors, SPIE–OFS
mode: 2096 Hz. After the first damage, the same frequency shift Meeting, Bruges, Belgium, May, 2005, p. 1031.
is observed by both sensors (−0.745%), while no modification [12] A. Cusano, P. Capoluongo, S. Campopiano, M. Giordano, M. Caponero,
is observed at the second one. A difference of 23.47% can be A. Paolozzi, Experimental modal analysis of an aircraft model wing by
observed in the accelerometer’s peak at the first damage and embedded fiber Bragg grating sensors, IEEE Sens. J. 6 (1) (2006) 67–77.
24.00% at the second one. For the FBG’s resonance peak, a [13] A. Cusano, P. Capoluogo, S. Campopiano, A. Cutolo, M. Giordano, F.
Felli, A. Paolozzi, M. Caponero, Measuring modal parameter variation
difference of 17.49% between the first damage and the second on composite structures by fiber Bragg grating sensors, in: Second Euro-
one can be observed while the second one introduces a 4.79% pean Workshop Structural Health Monitoring, Munich, Germany, July,
difference. 2004, pp. 855–861.
10 P. Capoluongo et al. / Sensors and Actuators A xxx (2006) xxx–xxx

[14] A.D. Kersey, A.M. Davis, H.J. Patrick, M. LeBlanc, K.P. Koo, C.G. cal resonators, diagnostic equipment, and harmonic generator for the infrared
Askins, M.A. Putnam, E.J. Friebele, Fiber grating sensors, J. Lightwave FEL. In 1987, he was with Duke University, Durham, NC. In 1986, he was
Technol. 15 (1997) 1442–1462. appointed professor of quantum electronics, University of Naples, where he
[15] A. Cusano, G. Breglio, M. Giordano, L. Nicolais, A. Cutolo, Multi- became professor of optoelectronics in 1993. At present, he is a full pro-
function fiber optic sensing system for smart applications, IEEE/ASME fessor at the University of Sannio, Benevento, Italy. His research interests
Trans. Mechatronics 9 (1) (2004) 40–50. involve fiber-optic sensors, optoelectronic modulators and switching, laser
[16] A. Cusano, A. Cutolo, J. Nasser, M. Giordano, A. Calabrò, Dynamic beam diagnostics, and non-linear optical devices.
strain measurements by fibre Bragg grating sensor, Sens. Actuators A
110 (1–3) (2003) 276–281. Michele Giordano received the degree in chemi-
[17] K.Y. Sanliturk, O. Cakar, A New Method for Noise Elimination from cal engineering from University Federico II, Naples,
Measured Frequency Response Functions, ICATS Technical Publication, Italy, the engineer’s diploma qualification in 1992,
[Link]/pubs/[Link]. and the PhD degree in materials engineering in 1996.
He joined the CNR in 1996. His main research
Biographies field is in the area of polymer-based composites. His
principal research interests in thermoset-based com-
posites manufacturing technologies include model-
Patrizio Capoluongo was born in Naples, Italy, in ing (resin transfer molding, pultrusion), development
1974. He received the electronic engineering degree (autoclave), and monitoring systems (fiber optic). His
at the University Federico II, Naples, Italy, in Octo- expertise has been developed in the area of damage
ber 2001. He is currently pursuing the PhD degree in dynamics of composite structures via acoustic emission techniques.
information engineering in the Engineering Depart-
ment, University of Sannio, Benevento, Italy. His
main research interests are optoelectronic sensors. Andrea Cusano was born in Caserta, Italy, in 1971.
He received the MS degree in electronic engineer-
ing and the PhD degree from the University Federico
II, Naples, Italy, in 1998 and 2002, respectively. At
Antonello Cutolo was born on November 7, 1955, present, he is a permanent researcher at the University
in Naples, Italy. He received the doctorate degree in of Sannio, Benevento, Italy. His field of interest is in
electronic engineering from University Federico II, the area of fiber-optic sensors and FBGs for applica-
Naples, Italy, in 1978. After serving in the Italian tions in smart processing and SHM of polymer-based
Air Force, he joined the Applied Mathematics and composites. At present, he is a permanent researcher
Physics Laboratory of the Technical University of at the University of Sannio, Benevento, Italy. He is
Denmark (1980–1981). From 1981 to 1983, he was the author and coauthor of more than 100 interna-
with Adone Storage Ring of Frascati, Rome, Italy, to tional publications, including international journals and conferences. His field
build a free electron laser. From 1983 to 1986, he was of interest is in the area of fiber-optic sensors for structural and chemical
with the High Energy Physics Laboratory, Stanford, sensing.
CA, where he was in charge of the broadband opti-