Calibration of Elasto-Magnetic Sensors on In-Service Cable-Stayed Bridges for Stress Monitoring

Transcription

1 sensors Article Calibration Elasto-Magnetic Sensors on In-Service Cable-Stayed Bridges for Stress Monitoring Carlo Cappello 1, * ID, Daniele Zonta 2 ID, Hassan Ait Laasri 3, Branko Glisic 4 Ming Wang 5 1 Department Civil, Environmental Mechanical Engineering, University Trento, Trento, Italy 2 Department Civil Environmental Engineering, University Strathclyde, Glasgow G1 1XQ, UK; daniele.zonta@strath.ac.uk 3 Faculty Sciences, Ibn Zohr University, Agadir, Morocco; hassan.or@hotmail.com 4 Department Civil Environmental Engineering, Princeton University, Princeton, NJ 08544, USA; bglisic@princeton.edu 5 Department Civil Environmental Engineering, Norastern University, Boston, MA 02115, USA; mi.wang@norastern.edu * Correspondence: carlo.cappello@unitn.it; Tel.: Received: 9 January 2018; Accepted: 30 January 2018; Published: 5 February 2018 Abstract: The recent developments in measurement technology have led to installation efficient monitoring systems on many bridges or structures all over world. Nowadays, more more structures have been built instrumented with sensors. However, calibration installation sensors remain challenging tasks. In this paper, we use a case study, Adige Bridge, in order to present a low-cost method for calibration installation elasto-magnetic sensors on cable-stayed bridges. Elasto-magnetic sensors enable monitoring cable stress. The sensor installation took place two years after bridge construction. The calibration was conducted in two phases: one in laboratory or one on site. In laboratory, a sensor was built around a segment cable that was identical to those cable-stayed bridge. Then, sample was subjected to a defined tension force. The sensor response was compared with applied load. Experimental results showed that relationship between load magnetic permeability does not depend on sensor fabrication process except for an fset. The determination this fset required in situ calibration after installation. In order to perform in situ calibration without removing cables from bridge, vibration tests were carried out for estimation cables tensions. At end paper, we show discuss one year data from elasto-magnetic sensors. Calibration results demonstrate simplicity installation se sensors on existing bridges new structures. Keywords: calibration; elasto-magnetic sensors; cable-stayed bridge; structural health monitoring; vibration test 1. Introduction Structural health monitoring (SHM) systems have gained rapid progress with aid advanced technologies in sensing, data communication data analysis. They provide relevant information on structural behavior are particularly important for early damage detection, reliability evaluation residual capacity determination. Unnecessary periodical inspections can be avoided leading to a reduction maintenance costs by using permanent monitoring systems. Nowadays, monitoring systems have been widely accepted used to check structural health condition civil infrastructure. More more large-scale monitoring systems have been designed implemented in existing structures during construction stages. However, application monitoring systems on existing structures can be a challenging costly task. Many se structures are cable-stayed Sensors 2018, 18, 466; doi: /s

2 Sensors 2018, 18, or cable-suspended bridges. In se structures, cables are fundamental components that guarantee overall structural safety. Therefore, permanent monitoring cable stress is great importance for assessing health such bridges. Measuring stress in cables during service life bridges can be extremely challenging if we use traditional methods. Elasto-magnetic (EM) sensors use a new low-cost technology for measuring actual stress in ferromagnetic materials such as that steel wires, strs steel bars. They are considered to be a promising tool owing to heir non-destructive non-contact properties, which also include corrosion resistance long service life [1,2]. The fundamental principle EM sensors is magneto-elastic phenomenon ferromagnetic material. The magnetic properties a ferromagnetic material change under application stress under influence temperature. Thus, it is possible to measure stress applied to a ferromagnetic material by knowing relationship between magnetic permeability stress. This relationship can be derived from calibration process. The first studies about magneto-mechanical effect were made in 19th Century. Later studies discussed oretical background magnetization process in ferromagnetic metals its application to magneto-mechanical effect [3,4]. In recent years, many investigations have been devoted to developing, installing improving performance EM sensors. In 1996, Kvasnica Fabo [5] designed a microcomputer-based instrument for magnetic measurement mechanical stress in low-carbon steel wires. In 1998, Wang et al. introduced concept stress monitoring in cables cable-stayed bridges by using EM sensors [6] confirmed, in 2000, that EM technology can provide adequate accuracy reliability to monitor actual stress tendons cables [7]. The reliability EM sensors in measuring actual stress cables reinforcement bars has been shown by several literature sources [1,8]. Ors have applied EM sensors to determine fatigue state in ferromagnetic steels [9]. In 2003, Cajko discussed a pulse variant method for measuring incremental permeability, which improves information reduces specimen heating caused by traditional techniques [10]. In 2004, numerous experiments were performed by Singh et al. on ferromagnetic steels at different temperatures for different cross-sections, in order to test accuracy EM sensors in corrosion evaluation [11]. Rumiche et al. have reported that EM sensors can possibly be used as a reliable non-destructive tools for detection corrosion in early stages carbon steels [12]. Furrmore, Park et al. have proposed a monitoring technique that enables a cable-climbing robot to detect cross-sectional losses [13]. In 2008, Wang verified capability EM sensing technology for long-term structural health monitoring external tendons a double-box girder bridge [14]. In same year, Tang et al. designed a new EM sensor that performs temperature compensation in a wide temperature range [15]. The use EM sensors for detection creep in ferromagnetic materials was examined by Polar et al., in 2010 [16]. In same year, Cao Wang studied structural effects using data collected by EM sensors [17]. In 2011, Duan et al. developed a smart elasto-magneto-electric sensor for stress monitoring in railway infrastructures [18]. In 2012, Duan et al. proposed a magneto-electric sensing unit to replace secondary coil conventional EM sensors [19]. In this work, we present how we can instrument in-service cable-stayed bridges with an EM sensor network, in order to continuously monitor actual stress its cables guarantee structural reliability. EM sensors are usually prefabricated installed on bridge cables during construction stages [1,20]. In this situation, calibration can be carried out by loading unloading cables while recording temperature. However, in our case, sensors had to be installed two years after bridge construction, owner could not allow cables to be unloaded in order to perform calibration tests. In this contribution, we demonstrate a low-cost calibration installation procedure for monitoring tension stay cables using EM sensors. This procedure is suggested for monitoring stresses in in-service cable-stayed bridges. The monitored structure, presented in 1, is a cable-stayed bridge spanning Adige River, 10 km north town Trento, Italy. It is a statically indeterminate structure, having a composite steel-concrete deck a length 260 m overall supported by 12 stay cables, six per deck side. The deck

3 Sensors 2018, 18, Sensors 2018, 18, cross-section consists four I section steel beams a depth 2 m carrying a 25 cm-thick concrete concrete slab. Theslab. deckthe is anchored deck is anchored to a cableto every a cable 30 m. every The30 bridge m. The tower bridge is made tower is four made pylons; four itpylons; is 45 m it high is 45 m high is located is inlocated center in center bridge. The bridge. stay cables The stay arecables full-locked are full-locked steel cablesteel diameters cables diameters 116 mm mm, mm 128 ymm, are designed y for are operational designed loads for operational between 5000 loads between 8000 kn. Structural redundancy, kn. Structural possible redundancy, relaxation losses possible relaxation as-built losses condition an that as-built are different condition fromthat design are different suggest from that long-term design suggest load redistribution that long-term between load redistribution cables can bebetween expected cables [21]. can be expected [21]. This paper is organized as follows. In Section 2, we introduce physical principle an EM sensor justify need for calibration. Next, we discuss calibration process, which is divided into two phases: one in laboratory or one on site. Section 3 is devoted to laboratory calibration, where a sensor was built around a segment cable, identical to those Adige Bridge, loaded up to 9000 kn. The response sensor at at various load load levels levels is compared is compared with with load load applied applied by by machine. machine. Section Section 4 discusses 4 discusses installation installation process process presents presents in situ calibration, in situ calibration, during which during we carried which out we vibration carried out tests vibration in ordertests to estimate in order to tension estimate force tension actual force bridge actual cables. bridge Next, we cables. assess Next, sensor we assess accuracy sensor present accuracy one year present data one recorded year from data recorded EM sensors from after EM calibration. sensors after Finally, calibration. Section Finally, 5 outlines Section conclusions. 5 outlines conclusions. (c) Adige Adige Bridge: Bridge: plan plan view view with with sensor sensor names names (1TN 6TN, (1TN 6TN, 1BZ 6BZ); 1BZ 6BZ); lateral lateral view; view; (c) cross-section (c) cross-section deck. deck. 2. Sensor Physical Principle 2. Sensor Physical Principle When magnetic field is applied to medium, resulting magnetic flux density is related When a magnetic field H is applied to a medium, resulting magnetic flux density B is related to H, given magnetic permeability μ medium. The value μ measures how easily a to H, given magnetic permeability µ medium. The value µ measures how easily a magnetic magnetic field can traverse a medium. For ferromagnetic materials, this relationship is nonlinear hysteretic. Thus, we normally refer to incremental permeability, which is ratio between incremental changes in two magnitudes:

4 Sensors 2018, 18, field can traverse a medium. For ferromagnetic materials, this relationship is nonlinear hysteretic. Thus, we normally refer to incremental permeability, which is ratio between incremental changes in two magnitudes: µ = B H. (1) The magnetic properties a ferromagnetic material are altered with application stress. The magnetic strain energy E σ is related to stress σ angle θ between direction applied stress magnetization vector, according to equation: E σ = 3 2 λ sσ sin 2 θ, (2) where λ s is bulk magneto-restriction strain induced when sample is magnetized to saturation magnetization [8]. Equation (2) shows that, in order to minimize E σ, θ has to change as σ changes. In or words, rom magnetization field that characterize specimen before magnetic field is applied rotates because variation σ, making a magnetization in anor direction more or less difficult. The measured permeability changes as a result. Hence, cable stress state can be obtained from experimental measurements magnetic permeability. The easiest way to achieve this goal is through principle magnetic induction: by magnetizing material using two solenoids. An EM sensor consists two coils wound round tensioned cable, as depicted in 2a. In order to make a measurement, cable is subjected to a pulsed magnetic field H generated by passing a pulsed current through primary coil. Long measurements are avoided because y would increase temperature sensor consequently cause errors during reading. The changes in flux density B produce an output voltage V ind (t) across secondary coil around cable. The induced voltage V ind (t) allows magnetic properties to be sensed deduced through Faraday s law: [ db(t) V ind (t) = N A f dt ] dh(t) + (A 0 A f )µ 0, (3) dt where N is number turns in secondary coil, µ 0 is magnetic permeability free space A 0 A f are cross-sectional areas secondary coil steel element (i.e., cable), respectively. If induced voltage is integrated in time interval [t 1, t 2 ], time-averaged voltage output on secondary circuit is: or: where: t 2 V = 1 V t 2 t ind (t)dt, (4) 1 t 1 V = 1 [ ( ) ] RC NA A0 f B + 1 µ 0 H, (5) A f B = t 2 t 1 db dt, H = dt t 2 t 1 dh dt, (6) dt R C are resistance capacitance circuit shown in 2c. If we take same measurement without ferromagnetic material, resulting output voltage V 0 is: V 0 = 1 RC NA 0µ 0 H. (7)

5 Sensors 2018, 18, By taking ratio (5) (7), permeability can simply be derived from following equation: V = 1 ( A f B V 0 µ 0 A 0 H + 1 A ) ( f A = µ f r + 1 A ) f, (8) Sensors 2018, 18, 466 A 0 A 0 A where µ r is relative permeability ferromagnetic material: where μr is relative permeability ferromagnetic material: µ r = 1 1 B B µ 0 H. (9) r H. (9) From From Equations Equations (8) (8) (9), we (9), eventually we eventually obtain: obtain: µ r = 1 + ( ) A 0 V V 1. (10) r 1 1 A f V. (10) f V0 0 Equation (10) directly correlates relative permeability µ r to μr steel core to sensor output V. Since output voltage voltage is is temperature temperature stress stress dependent, dependent, sensor sensor must must be calibrated be calibrated in order in order to measure to measure only only stress stress steel rod. steel rod. 0 (c) EM EM sensors: sensors: picture picture an an EM EM sensor; sensor; typical typical hysteresis hysteresis loops loops for for stressed stressed non-stressed non-stressed ferromagnetic ferromagnetic materials; materials; (c) (c) measurement measurement principle. principle. 3. Laboratory Calibration The first calibration phase was carried out in laboratory conditions using segments cables identical to those Adige Bridge. The cables, with an EM sensor attached, were subjected to different values load temperature. The EM sensors used in this project were supplied by Intelligent

6 Sensors 2018, 18, Laboratory Calibration The first calibration phase was carried out in laboratory conditions using segments cables identical to those Adige Bridge. The cables, with an EM sensor attached, were subjected to different values load temperature. The EM sensors used in this project were supplied by Intelligent Instrument Systems Inc (Burr Ridge, IL, USA). In order to install se sensors on specimen, primary coil secondary coil were wrapped around cable by a winding rig. Then, measurements were obtained by recording permeability variations steel core based on voltage induced in secondary coil. This voltage is sensitive to: (1) intensity magnetization; (2) stress applied to cable; (3) cable cross-section; (4) sensor manufacturing process; (5) temperature. Therefore, quantification permeability variations requires investigating effect each variable through calibration process, which must be performed before using se sensors. The scope calibration is to provide laws relating sensor measurements to stress applied to cable to compensate effects or variables. If we take into account effect all parameters in Equation (8), we obtain: µ r (σ, T, H) = 1 + A ( ) 0 V(σ, T, H) 1, (11) A f V 0 where σ is stress applied to cable T is temperature at which measurement was done. First, it is necessary to eliminate dependency sensor measurements on magnetic field H. This process is experimentally completed by finding magnitude excitation current that is necessary to make each sensor work. The relationship between magnetic permeability stress cable is linear stable provided applied field H is equal to an optimum value [7]. In fact, magnetic field H should be high enough to technically saturate cable. The magnetic saturation is important to obtain highest possible sensitivity linearity. However, increasing current generates heat, which worsens performance. Thus, input current needs to be optimized. In order to do so, producer sensors performs a set laboratory tests at various conditions, so that best working range can be determined effects H can be ignored in (11). Next, effects temperature stress on relative permeability were analyzed using different stress temperature levels. This investigation is necessary in order to identify stress-permeability relation compensate influence temperature. Since we are interested in permeability variations related only to stress variations, initial permeability cable at zero stress has to be subtracted from measured one. This operation can simply be accomplished by solving: By using Equation (12), we obtain: µ r(σ, T, T 0 ) = µ r(σ, T, T 0 ) = µ r (σ, T) µ r (0, T 0 ). (12) [ 1 + A ( )] [ 0 V(σ, T) A ( )] 0 V(0, T0 ) 1. (13) A f V 0 A f V 0 Then, µ r(σ, T, T 0 ) = A 0 A f [ V(σ, T) V(0, T0 ) V 0 ], (14) where T 0 is baseline temperature at which V 0 permeability at zero stress are measured.

7 Sensors 2018, 18, The temperature effects can be mamatically excluded using common experimental formula, which expresses steel permeability variation due to temperature deviation from baseline T 0. The formula was experimentally tested by many researchers [22] confirmed in this study: µ r (T) = α(t T 0 ), (15) where α = dµ r /dt is temperature sensitivity or temperature compensation coefficient. By compensating influence temperature on measured permeability, Equation (14) becomes: µ r (σ, T, T 0 ) = µ r(σ, T, T 0 ) + µ r (T) = A 0 A f [ V(σ, T) V(0, T0 ) V 0 ] + α(t T 0 ). (16) This equation relates permeability variation due to stress to voltage output temperature variation. By substituting temperature-compensated relative permeability µ r, force F applied to cable can be estimated by polynomial interpolation: F = C 0 + m n = 1 C n µ n r = C 0 + m n = 1 { } n A0 C n [V(σ, T) V(0, T) + α(t T 0 )V 0 ], (17) A f V 0 where C n are coefficients independent temperature to be derived from calibration process. In order to determine se coefficients, as well as parameters α m, sensor response was measured under different load temperature conditions. The load calibration provided C n n, while temperature calibration provided α. During load calibration, cables, with EM sensors attached, were connected to tensing machine 3a, under constant temperature conditions. Then, specimens were loaded unloaded up to a force 9000 kn, in steps 1000 kn. After each load step, sensor response was measured with at least three voltage readings (as suggested by manufacturer) compared to force applied by machine. 3b shows that temperature sensitivity permeability coefficient slightly depends on load. The permeability for force 7000 kn is an outlier must be ignored. 4a shows relationship between applied load F voltage V, for a cable a diameter 128 mm. Curves 3a 2a are responses first second load cycles, at a temperature 27.2 C. It is clear that relationship is not necessarily linear for every value force. However, as long as force is greater than 4000 kn, relationship can be considered linear. Since load applied to cable Adige Bridge is above 4000 kn, a linear equation was adopted for force estimation. Moreover, laboratory calibration showed that that relationship curves obtained during several load cycles become more similar after second cycle. The temperature-permeability relationship was studied, α was estimated by taking voltage readings at different loads for two different temperatures: 27.2 C 37.7 C. Curves 1a 2a 4a show an example for a cable with a diameter 128 mm. The result indicates that temperature modification does not affect slope force-voltage curve F-V, but changes its position. It appears that influence temperature is to shift curve F-V to a new position. This result is in accordance with Equation (15). A deep examination coefficient α reveals that it is not exactly constant, but changes slightly with stress level: as is shown in 4b, values from C 1 to C 1 have been identified at different load levels. Fortunately, relationship is linear can be taken into account in calibration equation. The temperature sensitivity also depends on cable diameter. Based on previous considerations, we can assume that F-V relationship is linear: F = F 0 + a ((V V 0 ) b (T T 0 )), (18) where a = df/dv is force to voltage slope b = dv/dt is voltage to temperature sensitivity. Equation (18) allows calculation force applied to a cable, given its calibration coefficients a, b,

8 Sensors 2018, 18, voltage response V at temperature T response V 0 to a reference load F 0 temperature T 0. Table 1 shows mean stard deviation coefficients α, a b, identified during laboratory calibration (data third cyclic loading were used). Furr tests were carried out to verify repeatability measurements. 4b shows Sensors 2018, 18, response three different sensors manufactured in laboratory on same cable using same procedure. The curves are similar in slope a, but exhibit very different shifts in position, making As a result laboratory calibration, a relationship type shown by Equation (18) is installation each sensor non-repeatable. Additional tests showed that voltage-to-temperature achieved for each combination cable sensor. However, due to uncertainty in reference load sensitivity dv/dt α are virtually independent manufacturing process. The scatter in F0, each sensor requires an on-site calibration after installation for at least one value tension curve slope a is relatively small can be attributed to a bias error. temperature. 3. View an EM sensor on 116-mm cable during laboratory calibration; permeability 3. View an EM sensor on a 116-mm cable during laboratory calibration; permeability sensitivity (to temperature) for cable diameter 128 mm; coefficient varies with sensitivity α (to temperature) for cable diameter 128 mm; coefficient varies with load level. load level.

9 Sensors 2018, 18, Sensors 2018, 18, Sensors 2018, 18, Hz. This let us identify natural frequencies fn shown in Table 2 with a precision 0.02 Hz. After extracting harmonic series signal, tension was estimated using following expression [23,24,25]: 2 2 2L 2 n F k fn EJ n L, (20) different different load-voltage load-voltage calibration calibration curves curves same same sensor: sensor: (3a) (3a) (2a) (2a) lines lines are for are where different k is load linear cycles mass at temperature cable, T L = 27.2 cable for different load cycles at temperature T = 27.2 C, while length, fn (1a) frequency line was recorded nth at temperature harmonic, E C, while (1a) line was recorded at temperature apparent T = = Young s C; C; force modulus force to to voltage voltage ratio ratio cable for for three steel three different different J sensors, sensors, moment after after temperature temperature inertia compensation. compensation. cable cross-section. More specifically, we fitted relationship between experimental frequency fn harmonic order Table 1. Mean stard deviation (StD) coefficients α, a b identified by Table n using 1. Mean F as a parameter. stard deviation (StD) coefficients α, a b identified by laboratory calibration. Despite laboratory this calibration. method considering only bending effect neglecting effects sag-extensibility, Cable Diameter this approach is ten α (1/ C) used α (1/ in C) practice due a (kn/v) to a its (kn/v) simplicity b (V/ C) speediness. b C) For cable 1TN, effect Cable (mm) Diameter parameter (mm) estimation Mean MeanStD is shown StD in 3 Mean Mean 7b. It StD was StDobserved Mean Mean that StD StDfirst seven modes deviate from curve fitting obtained from or modes. The reason may be attributed to sag-extensibility effects, which mainly affect lower modes It has been proven by many studies that utilization higher modes provides better accuracy [25,26]. Thus, first seven 4. lower Installation, modes were On-Site neglected Calibration in this work. As a result laboratory calibration, Accuracy a relationship Estimation type shown by Equation (18) is While vibration test was being performed, voltage EM sensor installed on achieved After for each laboratory combination calibration, cable 12 EM sensors sensor. were However, manufactured due to on site uncertainty installed reference vibrating cable was recorded toger with its temperature value. Table 3 shows acquired on data. 12 load cables F 0, each Adige sensor Bridge. requires The anprimary on-site calibration secondary aftercoils installation were wound for at least around oneeach valuecable tension Experimentally, it has been found that estimated loads have limited accuracy even when by a winding temperature. many harmonics machine, are as identified: shown in stard 5a. After deviation completing baseline winding cable procedure, tension was temperature kn gages for were attached, electric cables were connected. The two coils were separated by plastic 4. Installation, cable tension On-Site varying Calibration from 4000 kn Accuracy to 7000 Estimation kn. Furrmore, due to uncertainties in shells coefficients protected a b, by half-cylindrical accuracy deteriorates epoxy covers when ( conditions 5b). are different from those calibration. The After in situ laboratory calibration 8 shows calibration, was carried expected 12 EM out stard sensors without were deviation unloading manufactured for different stay oncables. site values The installed goal tension onthis calibration temperature 12 cables was Adige simulated to define, Bridge. using for The each Monte primary cable, Carlo analysis fset secondary F0 [27]. that The coils graph were laboratory shows wound calibration that around most each showed significant cableto by be a dependent source winding machine, error on is installation as shown inaccuracy process. 5a. In baseline order After to completing measurement. estimate reference winding Compared procedure, loads to this, temperature cables uncertainty Adige gages in Bridge force were sensitivity attached, without releasing is not electric critical, cables cables, while were vibration connected. temperature tests The were sensitivity twoperformed. coils were can result Two separated different in an by additional plastic accelerometers shells error were 50 protected kn. applied Similar by near half-cylindrical results each were or found on epoxy each for covers cable all ( cables 5c). 5b). Adige The connection Bridge. was made by fastening a steel shell on cable. Then, accelerometers were glued to an aluminum plate, plate bearing accelerometers was later screwed to shell. The whole system was considered perfectly rigid. The test was carried out for each cable by recording its response to a hammer blow with a rate 500 Hz. 6 shows as an example response cable 1TN ( 1a). Next, frequency spectrums were obtained by implementing a fast Fourier transform (FFT). The frequency spectrum cable 1TN is shown in 7a. Despite peak frequencies being clearly visible in graphs obtained through FFT, natural frequencies were calculated by fitting spectrums ( 7b) with oretical expression: 2 Af n (c) q f, 2 2 (19) n 1 fn f 2 i fn f Primary Primary secondary secondary coil coil EM EM sensors sensors are are wound wound around around cables; cables; two two coils coils are are protected where protected by q is acceleration, by epoxy epoxy covers; covers; (c) fn are (c) installation installation peak frequencies, accelerometers accelerometers for f is forfrequency in in situ situ calibration. calibration. ( independent variable), An are acceleration at peak frequencies, ξ is relative damping i 2 = 1. In (19), parameters An, ξ fn were considered unknown, while sum was carried out up to frequencies about

10 Sensors 2018, 18, Sensors 2018, 18, Hz. This let us identify natural frequencies fn shown in Table 2 with a precision 0.02 Hz. After extracting harmonic series signal, tension was estimated using following The inexpression situ calibration [23,24,25]: was carried out without unloading stay cables. The goal this calibration was to define, for each cable, fset 2 F 0 that2 2L laboratory calibration showed to be 2 n dependent on installation process. FIn korder testimate n EJreference loads cables Adige n L, (20) Bridge without releasing cables, vibration tests were performed. Two different accelerometers were where applied k is linear each mass or oncable, each L cable cable ( length, 5c). fn The connection frequency was made nth harmonic, by fastening E a steel apparent shellyoung s on cable. modulus Then, cable accelerometers steel J were moment glued to aninertia aluminum plate, cable cross-section. plate bearing More specifically, accelerometers we fitted was relationship later screwedbetween to shell. experimental The whole system frequency was considered fn harmonic perfectly rigid. order The n using testf was a carried parameter. out for each cable by recording its response to a hammer blow with a rate 500 Despite Hz. this 6method shows as considering an example only response bending cable effect 1TN ( neglecting 1a). Next, frequency effects spectrums sag-extensibility, were obtained this approach by implementing is ten used ain fast practice Fourier due transform to its simplicity (FFT). Thespeediness. frequency spectrum For cable 1TN, cable 1TN effect is shown parameter in estimation 7a. Despite is shown peak in frequencies 7b. It being was observed clearly visible that in first graphs seven obtained modes deviate through from FFT, curve natural fitting frequencies obtained from were calculated or modes. by fitting The reason spectrums may be ( attributed 7b) with to sag-extensibility oretical expression: effects, which mainly affect lower modes. It has been proven by many studies that utilization higher modes provides better accuracy [25,26]. Thus, first seven lower modes were neglected in A n f q( this f ) work. = 2 While vibration test was being n performed, = 1 fn 2 f 2 2ξ voltage i f n f, EM sensor installed on (19) vibrating cable was recorded toger with its temperature value. Table 3 shows acquired data. where q is acceleration, f n are peak frequencies, f is frequency ( independent variable), Experimentally, it has been found that estimated loads have limited A n are acceleration at peak frequencies, ξ is relative damping i 2 accuracy even when = 1. In (19), parameters many harmonics are identified: stard deviation baseline cable tension was kn A n, ξ f n were considered unknown, while sum was carried out up to frequencies about for cable tension varying from 4000 kn to 7000 kn. Furrmore, due to uncertainties in Hz. This let us identify natural frequencies f n shown in Table 2 with a precision 0.02 Hz. coefficients a b, accuracy deteriorates when conditions are different from those After extracting harmonic series signal, tension was estimated using following calibration. 8 shows expected stard deviation for different values tension expression [23 25]: temperature simulated using Monte Carlo ( analysis ) 2L 2 [27]. ( The πn ) graph shows that most significant 2EJ, source error is inaccuracy F = baseline k measurement. fn 2 Compared to this, uncertainty (20) in force sensitivity is not critical, while temperature n sensitivity L can result in an additional error where 50 kn. ksimilar is linear results mass were found cable, for all L cables length, Adige f n Bridge. frequency nth harmonic, E apparent Young s modulus cable steel J moment inertia cable cross-section. More specifically, we fitted relationship between experimental frequency f n harmonic order n using F as a parameter. Despite this method considering only bending effect neglecting effects sag-extensibility, this approach is ten used in practice due to its simplicity speediness. For cable 1TN, effect parameter estimation is shown in 7b. It was observed that first seven modes deviate from curve fitting obtained from or modes. The reason may be attributed to sag-extensibility effects, which mainly affect lower modes. It has been proven by many studies that utilization higher modes provides better accuracy [25,26]. Thus, first seven lower modes were neglected in (c) this work. While 5. vibration Primary test secondary was being coil performed, EM sensors are wound voltagearound cables; EM sensor installed two coils are on vibrating protected cableby was epoxy recorded covers; toger (c) installation with itsaccelerometers temperature value. for in situ Table calibration. 3 shows acquired data. 6. Data recorded by accelerometers for cable 1TN (see 1a). 6. Data recorded by accelerometers for cable 1TN (see 1a).

11 Sensors 2018, 18, Sensors 2018, 18, Fast Fourier transform signal acquired on cable 1TN comparison 7. Fast Fourier transform signal acquired on cable 1TN comparison between measured frequencies (points) oretical trend (continuous line). between measured frequencies (points) oretical trend (continuous line). Table 2. First 16 frequencies in Hz obtained with FFT. No. 1TN 2TN 3TN 4TN 5TN 6TN 1BZ 2BZ 3BZ 4BZ 5BZ 6BZ Stard deviation load on cable TN estimated by Equation (17) Table 2. First 16 frequencies in Hz obtained with FFT. No. 1TN Table2TN 3. Forces3TN from vibration 4TN tests 5TN voltages 6TN acquired 1BZ 2BZ during 3BZ in situ4bz calibration. 5BZ 6BZ Cable TN TN TN TN TN TN BZ BZ BZ BZ BZ BZ F 4 i (kn) σ 5 F,i (kn) V i (V) σ V,i (V) T i ( C) σ8 T,i ( C) Experimentally, it has been found that estimated loads have limited accuracy even when many harmonics are identified: stard deviation baseline cable tension was kn for cable tension varying from kn to kn Furrmore, due to uncertainties in coefficients a b, accuracy deteriorates when conditions are different from those calibration. 8 shows expected stard deviation for different values tension temperature simulated using Monte

12 Sensors 2018, 18, Carlo analysis [27]. The graph shows that most significant source error is inaccuracy baseline measurement. Compared to this, uncertainty in force sensitivity is not critical, while temperature 7. sensitivity Fast Fourier can result transform in additional signal error acquired 50 on kn. Similar cable results 1TN were found comparison for all cables between Adige Bridge. measured frequencies (points) oretical trend (continuous line) Stard Stard deviation deviation load load on on cable cable 5TN 5TN estimated estimated by by Equation Equation (17). (17). Table 2. First 16 frequencies in Hz obtained with FFT. 5. Monitoring Data No. 1TN 2TN 3TN 4TN 5TN 6TN 1BZ 2BZ 3BZ 4BZ 5BZ 6BZ The in situ calibration force in each cable enabled continuous acquisition data from monitoring system The values temperature force acquired since January are2.889 shown in The3.825 effect temperature on5.644 sensors was removed from7.808 data , while effect temperature on structure was not The figure shows, for each cable, one7.521 value per day, recorded in early morning, at about :00 a.m At this time, temperature whole structure is approximately homogenous. However, average daily temperature changes during year Since structure is statically indeterminate, we expected different temperature effects on different cables 9 due to load redistribution proves that force in cables is severely affected by 10 temperature variations The monitoring data show that force cables anchored near bridge 11 shoulders (1TN, BZ, 6TN BZ) increases with temperature, while force cables anchored near tower (3TN, BZ, 4TN BZ) decreases with temperature The explanation this phenomenon is beyond scope this paper will be made in future work. Neverless, force recorded from stay cables was compatible with behavior predicted during design 16 structure was in range force obtained from vibration tests, which confirms reliability EM sensors.

13 To sum up, EM sensors are a promising, simple affordable tool for stress monitoring steel structures. EM sensor technology enables simple inexpensive installation on in-service bridges, without any change being made to bridge structures. Moreover, it provides adequate accuracy reliability for monitoring actual stress steel cables during entire service life Sensors 2018, 18, civil structures. 9. Calibrated monitoring data. Acknowledgments: The monitoring project presented in this paper was funded by Autonomous Province 6. Conclusions Trento, thanks to L. Martorano, S. Rivis, A. Bertò, M. Pravda, P. Nicolussi Paolaz E. Pedrotti. The authors In this contribution, we begin by providing a description EM sensors for monitoring tension in ferromagnetic cables. Then, using case study Adige Bridge, we show feasibility monitoring force in stay cables. We explain how calibration EM sensors installation on existing cables can be easily performed. One year data from EM sensors installed on Adige Bridge

14 Sensors 2018, 18, is presented at end paper in order to show reliability sensors. Our concluding remarks on use this technology can be summarized as follows. The stability force-to-voltage voltage-to-temperature sensitivity, which are not affected by installation process, assert EM sensors applicability eases installation EM sensors on existing structures. Experimental results showed that two calibration stages are required: first must occur in laboratory conditions, second is to be performed on site. The former is needed in order to define force-to-voltage voltage-to-temperature sensitivity; latter is required to determine fset in voltage, which depends on installation. EM sensors can measure real stress a steel cable even when zero-stress state cable is unknown. Actually, neir laboratory calibration, nor in situ calibration enabled us to measure zero-stress state, because force-to-voltage relationship was non-linear for small values load. However, in order to monitor actual force stay cables such as those Adige Bridge, it is required to measure at least one value real stress corresponding sensor response temperature for each cable. In our case study, in order to obtain tension existing cables, vibration tests were carried out. For Adige Bridge, precision force measurements was better than 200 kn, which is a relatively high value, but acceptable. This precision was mainly due to inaccuracy baseline measurements (second calibration stage). To sum up, EM sensors are a promising, simple affordable tool for stress monitoring steel structures. EM sensor technology enables simple inexpensive installation on in-service bridges, without any change being made to bridge structures. Moreover, it provides adequate accuracy reliability for monitoring actual stress steel cables during entire service life civil structures. Acknowledgments: The monitoring project presented in this paper was funded by Autonomous Province Trento, thanks to L. Martorano, S. Rivis, A. Bertò, M. Pravda, P. Nicolussi Paolaz E. Pedrotti. The authors would also like to thank Riccardo Zonini, University Trento, P. Esposito, former PhD student at University Trento, Y. Zhao J. Yim, Intelligent Instrument System Inc., USA, B. T. Gorriz, Universitat Politecnica de Valencia, Spain, for ir help with instrumenting bridge. Author Contributions: Daniele Zonta conceived research project; Hassan Ait Laasri analyzed data; Branko Glisic revised results manuscript; Ming Wang presented physical principle sensors; Carlo Cappello wrote paper. Conflicts Interest: The authors declare no conflict interest. References 1. Sumitro, S.; Kurokawa, S.; Shimano, K.; Wang, M.L. Monitoring based maintenance utilizing actual stress sensory technology. Smart Mater. Struct. 2005, 14, [CrossRef] 2. Jarosevic, A. Magnetoelastic method stress measurement in steel. Smart Struct. NATO Sci. Ser. 1998, 35, Jiles, D.C.; Arton, D.L. Theory magnetization process in ferromagnets its application to magnetomechanical effect. J. Phys. D Appl. Phys. 1984, 17, [CrossRef] 4. Langman, R. The effect stress on magnetization mild steel at moderate field strengths. IEEE Trans. Magn. 1985, 21, [CrossRef] 5. Kvasnica, B.; Fabo, P. Highly precise noncontact instrumentation for magnetic measurement mechanical stress in low-carbon steel wires. Meas. Sci. Technol. 1996, 7, [CrossRef] 6. Wang, M.L.; Satpathi, D.; Koontz, S.; Jarosevic, A.; Choga, M. Monitoring cable forces using magneto-elastic sensors. In Proceedings 2nd US-China Symposium Workshop on Recent Developments Computational Mechanics in Structural Engineering, Dalian, China, May 1998; pp Wang, L.M.; Chen, Z.L.; Koontz, S.S.; Loyd, G. Magnetoelastic permeability measurements for stress monitoring in steel tendons cables. In Proceedings SPIE Nondestructive Evaluation Highways, Utilities, Pipelines, Newport Beach, CA, USA, 7 9 March 2000; Volume 3995.

15 Sensors 2018, 18, Sumitro, S.; Jarosevic, A.; Wang, M.L. Elasto-magnetic sensor utilization on steel cable stress measurement. In Proceedings First Fib Congress, Concrete Structures in 21th Century, Osaka, Japan, October 2002; pp Grimberg, R.; Leitoiu, S.; Bradu, B.E.; Savin, A.; Andreescu, A. Magnetic sensor used for determination fatigue state in ferromagnetic steels. Sens. Actuators A 2000, A81, [CrossRef] 10. Čajko, F. Pulse elasto-magnetic measurement cylindrical-shaped ferromagnetic specimens. In Proceedings 9th International Workshop on Applied Physics Matter (APCOM 2003), Malá Lučivná, Slovak, June Singh, V.; Lloyd, G.M.; Wang, M.L. Effects temperature corrosion thickness composition on magnetic measurements structural steel wires. NDT E Int. 2004, 37, [CrossRef] 12. Rumiche, F.; Indacochea, J.E.; Wang, M.L. Detection monitoring corrosion in structural carbon steels using electromagnetic sensors. ASME J. Eng. Mater. Technol. 2008, 130, [CrossRef] 13. Park, S.; Kim, J.W.; Lee, J.J.; Lim, J.S. Real-time NDE steel cable using elasto-magnetic sensors installed in a cable climbing robot IAARC In Proceedings 28th International Symposium on Automation Robotics in Construction, Seoul, Korea, 29 June 2 July 2011; pp Wang, M.L. Long term health monitoring post-tensioning box girder bridges. J. Smart Struct. Syst. 2008, 4, [CrossRef] 15. Tang, D.; Huang, S.; Chen, W.; Jiang, J. Study a steel str tension sensor with difference single bypass excitation structure based on magneto-elastic effect. Smart Mater. Struct. 2008, 17, [CrossRef] 16. Polar, A.; Indacochea, J.E.; Wang, M.L. Sensing creep evolution in 410 stainless steel by magnetic measurements. J. Eng. Mater. Technol. 2010, 132, [CrossRef] 17. Cao, Y.; Wang, M.L. Cable stress monitoring for a cable stayed bridge. In Proceedings 5th European Workshop on Structural Health Monitoring, Naples, Italy, 29 June 4 July 2010; pp Duan, Y.; Zhang, R.; Zhao, Y.; Or, S.; Fan, K.; Tang, Z. Smart elasto-magneto-electric (EME) sensors for stress monitoring steel structures in railway infrastructures. Appl. Phys. Eng. 2011, 12, [CrossRef] 19. Duan, Y.; Zhang, R.; Zhao, Y.; Or, S.W.; Fan, K.; Tang, Z. Steel stress monitoring sensor based on elasto-magnetic effect using magneto-electric laminated composite. J. Appl. Phys. 2012, 111, 07E516. [CrossRef] 20. Wang, M.L.; Wang, G.; Zhao, Y. Application EM Stress Sensors in Large Steel Cables Sensing Issues in Civil Structural Health Monitoring; Springer: Dordrecht, The Nerls, 2005; pp Esposito, P. Structural Monitoring Cable Stayed Bridge on Adige River on North Trento Rocchetta. Bachelor s Thesis, University Trento, Trento, Italy, Zhao, Y.; Wang, M.L. Fast EM stress sensors for large steel cables. In Proceedings SPIE 6934, Nondestructive Characterization for Composite Materials, Aerospace Engineering, Civil Infrastructure, Homel Security, San Diego, CA, USA, 8 April 2008; Volume Zui, H. Practical formulas for estimation cable tension by vibration method. J. Struct. Eng. 1996, 122, [CrossRef] 24. Kim, B.H.; Park, T.; Shin, H.; Yoon, T. A ccomparative study tension estimation methods for cable supported bridges. Steel Struct. 2007, 7, Den Hartog, J.P. Mechanical Vibrations; Springer: New York, NY, USA, Kim, B.H.; Park, T. Estimation cable tension force using frequency-based system identification method. J. Sound Vib. 2007, 304, [CrossRef] 27. Robert, C.P.; Casella, G. Monte Carlo Statistical Methods; Springer: New York, NY, USA, by authors. Licensee MDPI, Basel, Switzerl. This article is an open access article distributed under terms conditions Creative Commons Attribution (CC BY) license (