Smart RC elements for long-life monitoring of civil infrastructures


Daniele Zonta*a, Matteo Pozzia, Marco Fortia, Oreste S. Bursia DIMS, University of Trento, via Mesiano 77, 38050 Trento, Italy

A research effort has been launched at the University of Trento, aimed at developing an innovative distributed construction system based on smart prefabricated concrete elements allowing for real-time condition assessment of civil infrastructures. So far, two reduced-scale prototypes have been produced, each consisting of a 0.2×0.3×5.6m RC beam specifically designed for permanent instrumentation with 8 long-gauge Fiber Optics Sensors (FOS) at the lower edge. The sensors employed are Fiber Bragg Grating (FBG) -based and can measure finite displacements both in statics and dynamics. The acquisition module uses a single commercial interrogation unit and a software-controlled optical switch, allowing acquisition of dynamic multi-channel signals from FBG-FOS, with a sample frequency of 625 Hz per channel. The performance of the system underwent validation I n the laboratory. The scope of the experiment was to correlate changes in the dynamic response of the beams with different damage scenarios, using a direct modal strain approach. Each specimen was dynamically characterized in the undamaged state and in various damage conditions, simulating different cracking levels and recurrent deterioration scenarios, including concrete cover spalling and partial corrosion of the reinforcement. The location and the extent of damage are evaluated by calculating damage indices which take account of changes in frequency and in strain-mode-shapes. This paper presents in detail the results of the experiment and demonstrates how the damage distribution detected by the system is fully compatible with the damage extent appraised by inspection. Keywords: smart elements, Fiber Bragg Grating, nonlinear vibration, damage location, strain-mode-shapes

Concrete elements are used extensively in bridge construction due to their very favorable cost / performance ratio: on the basis of information found in the literature1, it can be estimated that more than the 70% of the bridge stock in Western countries is built of Reinforced Concrete (RC) and Precast Reinforced Concrete (PRC), while another 6% includes steel-concrete composite structures. RC is typically affected by aging, thus its performance is subject to degradation. The ability to evaluate the actual condition of deteriorating concrete structures is a key aspect of modern bridge management. Today, the appraisal of the bridge condition is almost exclusively based on visual inspection. The main reason for this is economic: an inspection program costs only a few thousand dollars per bridge per year, while the installation of a suitable permanent monitoring system could be in the hundreds of thousands of dollars. Nevertheless, there are facts suggesting that the future trend of condition assessment will be monitoring-based. First, bridge management philosophy is rapidly moving from a condition-based to a reliability-based approach, and visual inspection usually provides information which is insufficiently detailed for a formal bridge safety assessment: monitoring-based condition assessment can overcome this limitation. Second, sensors and communication technologies are ever more affordable, reliable and small-scaled2: the possibility of embedding sensors directly in a structure provides unexplored opportunities in the design of new bridges. With this vision in mind, a research effort has been launched at the University of Trento, to develop a cost-effective construction system based on smart prefabricated concrete elements for life-long real-time condition monitoring of bridges. The idea is to produce PRC elements embedding a low-cost high-durability sensing system: sensors are expected to be produced in the form of standardized packages, conceived as an integral part of the prefabricated element. Depending on the type of instrument used, a smart element will potentially be capable of self-detecting point to point physical quantities such as: strain and stress distribution; vibration response; cracking location and extension; temperature; moisture; pH; chloride concentration. An open, non-proprietary Internet-based Bridge Management Network will be able to interpret these quantities in terms of condition state and reliability, and will make these
*; phone 39 0461 882537; fax 39 0461 882505;

we can find structural damage qualitatively. transportation. during production. designed to accommodate the sensors during prefabrication.6m RC beam specifically designed for permanent instrumentation with 8 long-gauge Fiber Optic Sensors (FOSs) at the lower edge. Although this prefabrication method allows embedding of any kind of instrument. The measurement is based on recognition of the interference pattern generated by the reflected signals at the two surfaces. the system employed in the case of the two prototypes was based on strain sensors. c). construction and operation. As sensors are activated. SOFO (Surveillance d’Ouvrages par Fibres Optiques) system and Fiber Bragg Grating (FBG).8. all those involved in the production/management process can. Beam dimensions and reinforcement details are shown in Figure 1a. . introduced by Leung et al. and a brief summery is provided at the end. The idea was to develop simplified elements suitable for extensive laboratory testing. two fibers are arranged in a capillary tube in such a way as to form two parallel reflecting surfaces7.4. detail of the anchorages (c). In EFPI sensors. welded to the stirrup reinforcement before casting the concrete (Figure 1b. connect via Internet and check directly the actual condition of the element.2 Sensing System Layout and Performance Fiber Optics (FO) was chosen as the most suitable type of sensing technology.2×0. 2. labeled specimen A and B. and is arranged as follows: in the next Section an overview is given of the two smart beams developed so far. The potentialities of FOSs in civil engineering are well documented in Udd5 and Ansari6. in Section 4 the results of the experiment are presented and discussed. Possibly the most commonly utilized FO sensing techniques are based on measuring concepts such as: Extrinsic FabryPerot Interferometers (EFPI). and this is done using two methods: (1) measuring the permanent deformation of the element. OTDR technique.1 Element Description To date.3×5. SYSTEM DESCRIPTION 2. two identical prototypes have been produced. mainly due to the supposed high durability that renders this type preferable to electrical gauges when permanently embedded in concrete structures. builders and operators.available to those concerned: manufacturers. (2) measuring the vibrational response and applying damage detection techniques based on the changes in the strain-mode-shapes3. Further advantages of FOSs include their small dimensions. axonometric view of the sensor arrangement in the specimen (b). Optical Time Domain Reflectometry (OTDR). the expected future low cost and their insensitivity to electric and magnetic fields. but at the same time comprising all those features that make them complete enough to reproduce the behavior of the full-scale element. at any time and from anywhere. This in turn depends on the amplitude of the cavity. The target sensing system identified for these applications was a multi-channel system capable of measuring finite displacements with a sample frequency of at least 500 Hz per channel. The paper presents details of development and laboratory testing of these prototypes. Section 3 describes in detail the experimental validation of the technology. Sensors are connected to the structure by simple metal supports: these anchorages consist of small drilled steel plates. is based on measurement of the back-scattered power of a light pulse sent into one edge of the fiber as a function of time: as the presence of a crack in the monitored structure alters the reflecting properties of the attached fiber. The scope of the instrumentation was to identify point to point the presence of potential damage. a) b) c) Figure 1. Dimensions and reinforcement details of the instrumented beam (a). The elements have at bottom and side two 20×40mm longitudinal cavities. each of these consisting of a 0. The first step of the research initiated with the production and test of reduced-scale prototypes of smart elements. 2.

An FBG-based system generally includes a broadband source (light emission device). with a 4-channel setup. uses long-gauge sensors which measure the phase difference between the light carried by a fiber in direct contact with the structure and by a reference loose fiber. The control of the switch is software-based. an optical switch and a calibrator. Since. The regions having different refractive indices reflect the beam propagating in a narrow band centered about the Bragg wavelength λbragg =2neff Λ. and can therefore be determined by observing the wavelength shift of the reflected beam. and is of the order of 100 kHz with the current set-up. the system is capable of acquiring data at a sampling rate of 625 Hz per channel. as mentioned. the post-processing algorithm recognizes and discards the rise-time Tr and fall-time Tf branches of the signal due to the switching operation. In detail. Hence. the maximum sampling rate basically depends on the DAQ-card performance. and averages the response of each channel over the corresponding plateau. As the sensors exhibit a linear strain relationship to the wavelength shift within the elastic limit of the fiber.The SOFO system. which needs to be decoupled via software. λref is the reference wavelength and ∆λ is the wavelength shift. it was decided to design an improved interrogation unit for FBG sensors with the required specification using off-the-shelf components. their employment is typically restricted to static measurements. the smart-beam application required a dynamic multiplexed system. As the photodiodes work continuously. A Bragg grating sensor is a segment of the optical fiber in which a periodic modulation of effective refractive index neff with grating pitch Λ has been formed by exposing the core to intense ultraviolet light10. The core of the system is the sensing module: this generates the light beam and analyses the signal reflected by the sensors. Optical components are all provided by AOS GmbH. the voltage output of the sensing module takes the form of stepped signal of the type shown in Figure 3a. PC DAQ UA Sensing Module UB Hammer a) Shaker Calibrator b) Figure 2.5 kHz for the model that has been used. Although FOSs find today widespread use in long-term monitoring of civil structures. and include a sensing module. when the system is in single-channel mode. Switch FBG FOS . which is of the order of 2. developed and commercialized by Smartec SA9. Opt. In multi-channel mode. In this case. Any strain variation of the grating region results in changes in the grating pitch and refractive index. acquisition occurs by multiplexing the optical signal through the optical switch. In multi-channel mode. the sample frequency is limited by the maximum switching frequency. Layout of the system (a) and appearance of the Interrogation Unit (b). A National Instruments DAQ-card mounted on an external PC acquires the voltage signal as well as the signal coming from other external devices. most commercial Interrogation Units allowed either multiplexed but low-frequency measurement (with a scan rate typically lower than 50 Hz) or high-frequency but singlechannel measurements (although it should be noted that more recently Smartec SA has begun to offer a SOFO-based multi-channel dynamic acquisition system11). and also operated through the DAQ-card. the axial strain ε of the grating can be calculated as follows: ε= 1 ∆λ GF λref (1) where GF is the gage factor obtained by specific calibration. converting the peak wavelength data in a double electrical tension value (UA and UB) by using a pair of photodiodes. a set of optical fibers with prewritten grating sensors and an Interrogation Unit with an optical spectrum analyzer. A scheme of the system developed is shown in Figure 2. At the time this research started.

labeled effective length lm.5 9. but may also depend on the stiffness characteristics and the construction technology of each sensor component.5 [V] 1. which is of the order of 1 µε.5 Tr2 2. Voltage signal in multi-channel acquisition (a) and detail of a sensor edge. it is convenient to calibrate lm through a simple static test by comparing the FOS measurement to that of a micrometric gauge. AOS GmbH (b). This constant approximately corresponds to the physical distance between the inner edges of the two bars. and highlights the satisfactory agreement between the curves. 2 Tf2 Ua Ub 1. The sensor basically consists of a 600mm-long protected fiber including a grating.0 ch.5 x 10 2 -3 7. The frequency response ∆X(ω) of the FOS is expected to be related to the && && responses X 1 (ω ) and X 2 (ω ) of the accelerometers according to: && && ω 2 ∆X (ω ) = X 1 (ω ) + X 2 (ω ) (2) Figure 4b compares the strain response obtained through the two measuring methods in the frequency domain.5 1 1. the displacement measure ∆l is linearly related to the strain ε measured at the FBG through a constant. arranged in the experimental setup shown in Figure 4b. fixed at the edges inside two segments of threaded bar.0 7.5 0 1 2 3 4 a) [ms] b) Figure 3. In practice.2. 16 14 12 10 FOS A FOS B FOS C FOS D Acc x 10 -4 10 -4 [m] 8 10 -5 ε 6 4 2 0 FOS -2 -1. 3 switch cycle time -0.5 10 a) ∆l[m] b) [Hz] Figure 4.5 8. as shown in Figure 4a. The precision of the FOS is directly related to the nominal resolution of the Bragg grating. The sensor looks like a flexible wire that can be easily handled and coupled to the monitored structure by bolting its heads to simple metal supports. Static (a) and dynamic (b) calibration of FOS. When the sensor is pre-tensioned.5 1 lm 10 -6 -1 -0. 2.0 ch.0 9. 4 0.0 ch.0 8.3 Long-gauge Sensors Design and Calibration A long-gauge sensor model has been specifically developed in order to use FBG in direct displacement measurements. and provided on one side with an optical connection. The dynamic performance of the FOS has been tested by comparing the response of the fiber to an impulse with that of two piezoelectric accelerometers. 1 0. .5 ch.5 0 0.

80 N (level M). Scheme and overview of the dynamic experiment setup. in state B4. in addition to the 8 FOSs (Figure 6). and 150 N (level H). located at section RB.15 0. Finally. 3. three sweeps were repeated. using an electromagnetic shaker as the excitation source. Scheme and overview of the load test setup. During the dynamic test each beam was freely supported on springs and was instrumented with 19 additional accelerometers arranged vertically at pitch 0.30 Accelerometers A4 A5 A6 A7 A8 A9 A10 A11 A12 A13 A14 A15 A16 A17 A18 Spring A1 A2 A3 A19 LA Shaker 0. A and B. and the first yield of the reinforcing steel (A3) respectively. As for specimen B. as shown in Figure 5.3m. condition states B6 and B7 simulate local corrosion of the reinforcement at section RB and LD respectively.2m-long spall was produced at section LD.65 4. the design serviceability load (A2).2m-long spall of the corner cover. Steel beam Spring 0. B3 corresponds to a 0.Actuator 0. The results of the test on the two prototypes are reported below. Damage states B3 to B5 attempt to reproduce increasing spalling of the concrete cover. with the whole longitudinal reinforcement exposed. LABORATORY VALIDATION The effectiveness of smart-beam technology underwent validation in the laboratory.60 3. 3. The maximum load levels applied during each test are associated with the opening of the first crack in the concrete (A1). as summarized in Table 1.40 Figure 5.30 0. at each damage condition. The scope of the experiment was to correlate changes in the dynamic response of the beams with different damage scenarios. adopting different force amplitude levels. approximately equal to 10 N (level L). In order to detect and characterize the nonlinear behavior of the beam.40 1. an additional 0. damage conditions B1 and B2 reproduce the corresponding states of specimen A. A0 and B0. with a 15% loss of section. using a hydraulic actuator.80 5. Testing techniques include stepped-sine tests. were dynamically characterized in the undamaged state. with a single rebar exposed.60 LB LC LD FOSs RD RC RB RA Figure 6. in state B5 the spall located at section RB was extended to the whole bottom cover. . using a direct modal strain approach. increasing cracking conditions were produced in the beam by means of 3-point static load scheme. simulating different cracking levels as well as recurrent deterioration scenarios. More specifically.1 Test protocol The two specimens. As for specimen A. and in various damage states.

3 Approximate modal extraction The smart-beam system uses self-diagnosis techniques based on the changes in the modal parameters. defined as the spectral ratio of response to the force at the excitation frequency. We propose to approximate the behavior of the beam using an expression for the restoring force of the type: Rk (qk ) = −ωk2 qk 1 − β k qk ( n ) (4) where βk is labeled softening coefficient and n is an exponent that defines the class of the model. We can in the same way achieve strain-FRFs based on FOSs’ response. FOS measurements are already expressed in terms of strain. A comparison of the FRFs obtained at different excitation levels (see for example Figure 8a) reveals small shifts in the resonant peaks. the idea is to define as modal parameters of the system those frequencies ωk. governed by an equation of motion of the type: && & qk + 2ξ k ω k qk − Rk (qk ) = f k (3) where Rk is a nonlinear displacement-dependent elastic restoring force and fk= kφ F. obtaining curves of the type shown in Figure 7b. Therefore. A qualitative comparison of Figure 7a and Figure 7b might give the reader the wrong impression that the FRFs obtained through the accelerometers are more accurate than those of the FOSs. the steady-state response of the specimen can be represented in terms of first order Frequency Response Functions (FRFs). more extensive numerical computations need to be performed in order to estimate the strain distribution starting from the bare accelerometer data. at force level M. However. Figure 7a shows a sample of the FRFs of specimen A obtained by the accelerometers. βk can somehow be seen as a generalized modal parameter. A simple but reasonably acceptable assumption is that the response qk in a given mode approximates to that of a nonlinear SDOF system.2 Dynamic response For each damage condition. We can observe that the lower the response amplitude. the closer the system behavior to linearity. for a nonlinear system. Conversely. However. it is understood that a modal extraction cannot be performed in a strict sense. The problem now is how to derive these parameters from the experimental FRF curves. and the meaning of modal parameters has to be somehow redefined. 3. with F the vector of the exciting force in the physical coordinates. and therefore they can be directly employed in the damage detection process. 3. It can be recognized that this formulation generalizes the classical cubic stiffness model. The model indirectly assumes that the damping and the vibration mode are amplitude independent. First order FRFs obtained by accelerometer 1 (a) and strain-FRFs obtained by FOS D1 (b) for Specimen A at force level M. damping rates ξk and mode shapes kφ that approximate the response of the system when the excitation amplitude is close to zero. As expected. and this is an obvious symptom of system nonlinearity.10 -1 10 -5 10 -6 10 -2 [s 2 kg-1] [kg-1] 10 -7 A0 A1 A2 A3 10 -3 10 -8 A0 A1 A2 A3 0 50 100 150 a) 0 50 100 150 [Hz] b) [Hz] Figure 7. . It can also be observed that for n=0. which can be seen as a special case of Equation 4 with n=2. we must keep in mind that damage location methods are based on changes in the strain mode shapes. which quantifies the degree of nonlinearity of the vibration mode. both of the diagrams show losses in frequency associated with damage increase. the nonlinear model collapses in a linear case.

and for example is equal to 3/4 in the case of cubic stiffness (i. The effective frequency is related to the linear frequency ωk through ωk2.003 0. the assumption underlying Equation 3 is not fulfilled. Experimental and theoretical restoring force R1 at different damage level (b). The expression of the first order FRF in the physical coordinates x. If frequencies are well separated. ω ) = φi kφ j + S k (ω ) ωk2. in the case of coupled frequencies. the value of Sk is small and can be approximated with negligible error using one of the following expressions: S k (ω ) ≅ ∑ φi rφ j ≅ 2 2 r ≠ k ω r . for n=2).012 a) 0 18 20 22 [Hz] 24 26 b) 0 0 [m kg1/2] Figure 8. as in the present case.e.04 0. which generally reads: c= {sin θ sin θ }sin θdθ = π2 ∫ (sin θ ) π ∫ 1 2π n 0 0 k n+2 dθ (7) Coefficient c depends mainly on the class of the model. the values of ωk. eq (qk 0 ) = ωk2 (1 − cβ k qkn0 ) 2π (6) where c is the fundamental harmonic component of the restoring force.eq (qk 0 ) − ω 2 + i 2ξ kωωk 1 (5) where ωk. and can be at least extended to those points where the response amplitude is small enough to be considered linear. including all nonlinear effects.The next step is to find a theoretical formulation for the first order FRF α: the method of harmonic balance. 0. for each damage condition.009 0.06 M L H M L 150 [N kg-1/2] A0 A1 A2 A3 [kg-1] 100 0. . hence. as reported in many textbooks12. and the proposed model does not apply.eq (qk 0 ) − ω 2 + i 2ξ k ωωk (8) where kφi is the i-th component of mode k.006 0. ω ) = ω 2 k . It is worth noting that the resulting expression of α is only apparently analogous to the linear expression. can be written formally as: α ij (qk 0 .02 50 H M L 0. Increasing distortion of the first resonant peak of specimen A with force amplitude (a).eq(qk0) is the effective k-th angular frequency for the given response amplitude qk0.08 H 0. and Sk represents an amplitude dependent residual term accounting for the contribution of the other modes. indeed.13. obtained at different excitation levels. ξk kφ and βk can be extracted by simultaneously fitting the three FRFs.. the harmonic response at a resonance peak is dominated by the corresponding mode. about the k-th resonance peak. Conversely. eq (qr 0 ) − ω + i 2ξ rωωr r ∑ω r≠k r 2 r φi rφ j − ω 2 + i 2ξ rωωr (9) In summary. using the theoretical expression of the first order FRFs given by Equations 8 and 9. its validity is restricted to those values of ω in which the response is clearly dominated by a single mode. provides the approximate formulation: α (qk 0 .

26% 1.84% 2.53 69.50 138.99 55.00 0.78% 2.04 0. 0.01 FEM 3-point fit 5-point fit 0.66 118. restricted to the first three vertical bending modes.0 3.9180.87% 1.94 18.96 135.23 122.41% 1.5 4.07% β3 [m-1/2] 84.48 21. In the case of specimen B. when the pitch between the accelerometers is small.13% 2.Id A0 A1 A2 A3 B0 B1 B2 B3 B4 B5 B6 B7 Damage Condition Undamaged First cracking Serviceability load Reinforcement yielding Undamaged First cracking Serviceability load Spalling of corner cover in RB Spalling of bottom cover in LD Spalling of bottom cover in RB 15% loss of steel section in RB 15% loss of steel section in LD f1 [Hz] 23.60% 2.74% 2.0 42. In any case this results in a loss in spatial resolution.13 20.88% 1.15 22.88% 2.02 0.93% 2.04 0.8 84.0 4. Summary of the modal extraction results.10 141.63% 2.5 2.85 119.8 - f3 [Hz] 150.05 0.18 ξ2 1.0 1.8 - Table 1.5 - f2 [Hz] 71.8 15.53 22.87 120.05 0.49% 3.61% 2.63% β2 [m-1/2] 35.02% 2.2 15.94% 2.24% 4. no apparent change in the mode shape obtained by accelerometers data during damage is observed. while damping increases.57% 3.17 17.41 58. Table 1 shows the results of the extraction.82 70.59 20.47 ξ1 1.0 1.5 2.5 0.5 4.42 68.0 a) [m] b) [m] Figure 9.5 5.42% 2. As far as bending shapes are concerned.5 3.67 57. Approximate curvatures can be calculated from the displacements using a central finite difference scheme and data from three adjacent instruments3: & φ&i = φi−1 − 2φi + φi +1 h2 (10) Unfortunately.89 117.0 2.2 17. obtained by accelerometers.00 0. (a) and by FOSs (b).92 ξ3 1.44% 2.64 17.2 43. we saw that satisfactory fitting of the experimental curve is obtained. strain can be associated with curvature. The modal frequencies fk decrease during the damage process.49% 2.03 0.88% 1.03 0.64% 2. At first sight. These errors can be reduced either by smoothing over the curvature data.49 139.79 61.1 Modal parameters After a preliminary analysis.0 2. 4.15% 2.91 17.62 55. RESULTS 4.18% 1. as expected.56 64.06% 2.2 133.0 4.81% 1.82% 4. The experimental curves are compared with those obtained by a Finite Element Model (FEM).71% 2. First curvature-mode-shape of specimen A in the undamaged condition.56 17.16 18. at each excitation level.0 0.10 129.10 134.25% 1. or by using an alternative scheme for fitting the displacement data.48% β1 [m-1/2] 12.18 55. using a model with exponent n=1/2 and coefficient c=0.5 1.02 [kg-1/2m-2] 1.88% 1. the softening coefficient β has not been evaluated.46 65. with a 3-point and a 5point scheme.0 3. this approach is likely to yield inconsistent results due to the propagation of measurement errors and noise that the double derivation causes (Figure 9a).02% 1.6 48.5 5.01 FEM FOSs [kg-1/2m-2] 0.91% 3.5 3. .2 56.

the location and the extent of damage are evaluated in the beam in each condition state through the calculation of damage indices. εk.5 2.0 0. . This result is consistent with observations in experiments reported previously14.0 4.5 A0 A1 A2 A3 1.5 3.04 0.i indicates the value of the k-th modal curvature shape obtained by accelerometer-based data.04 0.02 0. b) for accelerometer. a significant nonlinearity is evident even in the undamaged conditions. Increases in β are associated with the first two damage steps.0 4. while ∆ indicates the difference with respect to the undamaged state. while the third step shows a reduction in nonlinearity.2m comprising 5 adjacent instruments.01 0. and the spatial resolution corresponds to the length l=0.5 3. As for specimen A the analysis of β shows how there are also changes in the degree of nonlinearity with damage.0. but it is unsuitable for revealing more severe damage conditions.4 More specifically.i ⎜ k⎜ ω ⎝ k ⎞ ⎟ ⎟ ⎠ 2 ⎤ ⎥ ⎥ ⎦ ⎡ ⎤ . the curvature mode shape can be evaluated on the basis of the strain mode shapes extracted from the FOS strain-FRFs.5 5. The details of the method can be found in Zonta et al. considered as the ratio between the strain measured at the calibrated sensor and the distance hy=144mm between the fiber and the neutral axis of the beam.i ⎞ ⎥ i ∑k ⎜ ⎟ ⎜ ⎟ 2 ⎢ ⎣ ⎝ h yω k ⎠ ⎥ ⎦ (11a. however the correlation between damage and nonlinearity is not as close as we might expect. In this case. nonlinearity was not considered in the damage detection procedure. δ = ∆ ⎢ ⎛ ε k .01 [kg-1/2 m-2] [kg-1/2 m-2] A0 A1 A2 A3 1.5 0.05 0. In other words. The first curvature mode shape resulting from FOS data is represented in Figure 10b. ωk the angular frequency.0 2. Figures 11 and 12 show how the index varies along the specimens and during the damage process according to each type of data. Alternatively. it appears that nonlinearity is a good marker for the detection of low levels of cracking.03 0.00 0.0 1. φ"k.02 0. In detail.2 Damage detection and location For the reasons mentioned.0 3. which take account of the changes in frequency and curvature-mode shapes.0 a) [m] b) [m] Figure 10: First curvature-mode-shape of Specimen A obtained by accelerometers (a) and FOSs (b). the indices δi associated to the i-th section of the specimen are defined as: δ i = ∆⎢ ⎢ ⎣ ⎡ ∑ ′ ⎛ φk′.0 3.0 2.00 0.05 0. in this case each point of the diagram results from a single instrument measurement. Fig. However the actual spatial resolution of the strain measurement is related to the interval l=1.5 5.5 4. Indeed. 10a shows the evolution of the first curvature-mode shape obtained through a 5-point least-square parabolic fitting scheme.0 1. a conventional curvature is directly calculated for each section.5 2. Moreover.and FOS-based data. This finding shows that it is not generally true that nonlinearity detection allows identification of damage without the knowledge of a baseline dataset. 4.6 m of the sensor.03 0.5 4. The comparison shows both good agreement between the two sets of data and the effectiveness of the self-diagnosis concept. respectively.i the modal strain obtained by FOSs-based data. In this case the curvature is calculated at each accelerometer position and with a nominal grid pitch of 0.3m.

Damage indices for Specimen A obtained by accelerometers (light line) and FOSs (bold line) compared with the crack pattern observed for each damage condition. .100% 80% 60% δ 40% 20% 0% A1 100% 80% 60% δ 40% 20% 0% A2 100% 80% 60% δ 40% 20% 0% A3 Figure 11.

.100% 80% 60% δ 40% 20% 0% B1 100% 80% 60% δ 40% 20% 0% B3 100% 80% 60% δ 40% 20% 0% B6 100% 80% 60% δ 40% 20% 0% B7 Figure 12. Damage indices for Specimen B obtained by accelerometers (light line) and FOSs (bold line) compared with the visual appearance of each damage condition.

D. Samman. allowing identification of a large number of mode shapes in terms of acceleration.). E. Brennan. Chanb. 5. Dulieu-Barton. Worden (Eds. 2001. Lau. Olson. Baldock. DAMAS 2003 Conf. K. Nonlinearity in Structural Dynamics. J. 2003. 2003.J.J. Balkema. Biswas and M. This justifies addressing the technological development of a system capable of direct strain measurements in dynamics. UK. Morse and Y.). Elvin. PA. 6. Identification and Modelling. Worden (Eds.K. WA.M. pp. “Damage Detection from Changes in Curvature Mode Shapes”. 485-591. Institute of Physics Publishing. 321-332. 199-208. Proc. D. practice and application. 4. Casanova and P. Watanabe.M. Worden and G. 2003. D. “Structural monitoring by curvature analysis using interferometric fiber optic sensors”. Zanon. Zhou and W. 2000. Seattle.. pp. Udd (Ed. T. K. 10. “High-Speed Demodulation of Long-Gauge Fibre Optic Strain Sensors for Dynamic Structural Monitoring”. C. DEStech Publications. pp. of Sound and Vibration.K. He. The experiment aimed at detecting the extent and location of damage in prototype beam elements with an embedded FOS system.5.M. DAMAS 2003 Conf. Leiden. and A. Williams. Lancaster.). 1998. Vurpillot. Holford and K. Jin. 12. Uetikon-Zuerich. Pandey. vibration damage detection techniques are mostly based on strain measurements and make use of few low-frequency modes. for this reason it was not considered in the damage detection process. .). Kronenberg. pp.D. J. M. Dulieu-Barton. Inaudi. 245-246. R. Proc. 2004. 11. Detection.M. “Strain monitoring in composite-strengthened concrete structures using optical fibre sensors. Smart Materials and Structures 7. Utsunomiya (Eds.. 8. Leung. C. J. 1991.” Composites Part B. Bristol. Frangopol and T. S. Vibration acquisitions carried out using accelerometers guarantee good results in a wide frequency range. “Mobile Networking for Smart Dust”. UK. 9.S.M.J.J. Ewins Modal Testing: theory. Tomlinson. F. FOSs exhibit features such as durability and stability that ensure they are better suited than electrical gauges to long-term monitoring of civil structures. 1998. A. 87-94. NL. “Structural health monitoring of smart composite materials by using EFPI and FBG sensors” Sensors and Actuators A103. J. CH. Key Engineering Materials. Conf. M. 133-148. D.R. 330-340. 1991.H. pp. Fiber Optic Sensors for Construction Materials and Bridges. The results demonstrate that a dynamic FO-based technology is feasible and technically effective when compared to the classic acceleration-based approach. Bridge Maintenance. 2. “A Strain-flexibility-based Approach to Damage Location”. Brennan. pp. 2000. The authors wish also to acknowledge Michele Corrà. it is unsuitable for revealing more severe damage conditions. Management and Cost. N. ACKNOWLEDGEMENTS This research has been carried out with the contribution of the Autonomous Province of Trento and the Italian Ministry of Education (MIUR). 14. P. “Damage Assessment in Concrete Beams Using Non-Linear Analysis of Vibration Measurement”.A. Luca Santorum. pp. N. PA. Katz and K. However. Kahn. Holford and K.M. Pister. REFERENCES 1. Inaudi.Y. Lancaster. while it appears that the nonlinear feature can serve at least to detect low levels of cracking. Wiley-Interscience. using vibration-based damage detection methods. Torsten Thiel and the AOS team. 13. 3. D. L.. pp. K. Leng.). However..S. N. Lanaro and P. E. Trans Tech Publications. Proc. The experiment also showed changes in the degree of nonlinearity of the RC elements with damage. Zonta. CH. 245-246. A. McFaddenn and M.F. Neild. J. 2nd European Workshop on Structural Health Monitoring. 2001. Proc. K. Glisic and D. Technomic Publishing Co. Posenato. 145(2). Research Studies Press ltd. on Mobile Computing and Networking. 2004. Asundi. B. 33-45. M. 557-564. of ACM/IEEE Intl. “A novel distributed optical crack sensor for concrete structures” Engineering Fracture Mechanics 65. Safety. Trans Tech Publications. S. Ansari (Ed. 32. Key Engineering Materials. In addition to this. 1999. Fiber Optic Sensors: An Introduction for Engineers and Scientists. 7. CONCLUSIONS The laboratory validation of the effectiveness of a construction system based on RC smart elements has been presented. UetikonZuerich.M.

Sign up to vote on this title
UsefulNot useful