Sensors and Actuators A 151 (2009) 127–132
Contents lists available at ScienceDirect
Sensors and Actuators A: Physical journal homepage: www.elsevier.com/locate/sna
Enhanced FBG sensor-based system performance assessment for monitoring strain along a prestressed CFRP rod in structural monitoring Abdelfateh Kerrouche a,∗ , W.J.O. Boyle a , Tong Sun a , K.T.V. Grattan a , J.W. Schmidt b , B. Täljsten b a b
School of Engineering and Mathematical Sciences, City University, London EC1V 0HB, UK Department of Civil Engineering, Technical University of Denmark, Brovej Building 118, 2800 Kgs., Lyngby, Denmark
a r t i c l e
i n f o
Article history: Received 18 October 2008 Received in revised form 16 February 2009 Accepted 16 February 2009 Available online 9 March 2009 Keywords: Fiber Bragg grating FBG-based sensor system Carbon fiber reinforced polymer CFRP
a b s t r a c t Fiber Bragg grating (FBG) sensor-based systems have been widely used for many engineering applications including most recently a number of applications in structural health monitoring. It is well known that strain and temperature both affect the FBG spectrum which in the interrogation system will be converted to a conventional electronic signal. This procedure provides the means for the FBG-based sensor system to be used for several monitoring applications. The aim of this research is to improve an existing monitoring system which has been used for several field test inspections. A brief description of the existing FBG-based system and the evaluation of the software developed to be compatible with a resolution reaching as high as ±0.15 is presented. The system has been tested under particular conditions where a prestressed CFRP (carbon fiber reinforced polymer) rod to which a FBG sensor is attached is used. The test procedure and results obtained are discussed in some detail. Crown Copyright © 2009 Published by Elsevier B.V. All rights reserved.
1. Introduction Fiber Bragg grating-based sensor systems have been used for structural monitoring for several years and the authors and others have published extensively in this field [1–8]. Those systems have been employed in tests carried out over a long period (up to 1–2 years in some cases) [3] and have showed high-quality performance in laboratory as well as in field tests. From the experience of this prior work, the monitoring systems used to determine the data from the fiber Bragg gratings (FBGs) used have proved to be able to take measurement in harsh environments under both static and dynamic load conditions. The FBGs used are fabricated by creating a periodic modulation of the refractive index along a photosensitive fiber and this process has been discussed in detail elsewhere [9]. The FBGs used in this study were ‘tailor-made’ and manufactured in-house using the phase mask technique, of 10 mm length with reflectivity of more than 95% and a bandwidths of 0.2 nm. Fiber Bragg gratings are used as strain and/or temperature sensors in which variations of those parameters are transformed to a shift of the Bragg wavelength reflected back as a Gaussian profile signal. In general, FBG-based systems employ the wavelength division multiplexing (WDM) technique to interrogate the FBG sensors. The basic principle of this method is to scan over the wavelength of a
∗ Corresponding author. Tel.: +44 785 256 2155; fax: +44 207 040 8568. E-mail address:
[email protected] (A. Kerrouche).
narrow-band light obtained from a broadband source by changing the space of two mirrors of a Fabry-Perot (FP) spectrometer. Each FBG from an array of serially connected sensors reflects back a spectral peak depending on its Bragg wavelength which is determined according to the spacing of those mirrors. In addition to optoelectronic hardware used to detect the signal via a PIN photodiode, the system requires an interrogation system that actuates and controls the scanning of the FP spectrometer and acquires synchronized data from the PIN detectors. In this study, a DSP (digital signal processor) board (Adwin-Gold, Jager GmbH) is employed for this purpose. This DSP employs a synchronous real-time operating system and can be programmed to produce controlled output voltage ramps which are synchronized with incoming A/D data to an accuracy of ∼2.5 ns about 10% of the DSP processor clock speed. This temporal accuracy is essential for sensitive estimation of the position of peaks in the optical spectra obtained from light reflected back from each array of sensors during each scan of the Fabry-Perot filter. The DSP system is accessed via a WindowsTM (Microsoft) DLL library and LabViewTM (National Instruments) is used to interface with the DSP. In the system use by the authors prior to this work [1–3] whole sets of optical spectra are transferred from the DSP to the LabView system and the peak positions estimated using the peak position algorithm from LabView. This was shown to give rise to difficulties in accurate estimation of peak wavelengths. In the improved algorithm, the peak wavelengths are estimated using the mean centroid method by the DSP and only the estimates of peak position are transferred to a LabView data handling programme.
0924-4247/$ – see front matter. Crown Copyright © 2009 Published by Elsevier B.V. All rights reserved. doi:10.1016/j.sna.2009.02.030
128
A. Kerrouche et al. / Sensors and Actuators A 151 (2009) 127–132
Fig. 1. (a) Optical fiber glued to the CFRP rod and (b) schematic of the CRFP rod containing the strain sensors loaded in the pulling machine.
The effect of the modified software is demonstrated through an observation of the difference in performance when the systems considered are applied in field studies to monitor the strain induced in carbon fiber reinforced polymer (CFRP) rods incorporated as a repair to a railway bridge in Frövi Sweden. In this study light passenger trains passing over the bridge provided the loading for the test. The results of an initial study [2] prompted a re-evaluation of the software used in the system and thus the development of the modified software for improved strain estimation. The use of the modified system is reported in its application to a laboratory test which was carried out in DTU (Technical University of Denmark). The FBG sensors were glued to a CFRP rod to measure the strain inside an anchor system attached to it, a device which is very important when using CFRP rods in prestressed concrete applications. FBG sensors used here have the advantage over the use of electrical resistance strain gauges because such a sensor can be placed inside an anchor system with 2–3 mm between the wedges that hold the CFRP, which allows one or even two optical fibers to be glued and then strain measurement to be taken. This is shown schematically in Fig. 1. 2. The FBG-based sensor system 2.1. System description The FBG sensor system architecture has proved to be versatile for use in a wide variety of structural platforms and industrial processing plant. The system used consists of one or more channels of FBG sensor networks illuminated with a broadband light source though a multiplexing device coupled to purpose-detection electronics to receive and interpret the reflected signal from each of the channels of the FBG sensor network. The maximum number of FBG sensors used in each array is 8, determined only in this case by the super-luminescent light emitting diode (SLED) free spectral range and the FBG bandwidth. The basic principle of the monitoring system is to identify the wavelength shift of the reflected spectra from the FBG sensors. The centre wavelength (B ) of the reflected spectrum is defined by the Bragg condition (B = 2ne ), where (ne ) is the effective refractive index of the optical fiber and () is the refractive index of the grating. The shift of the reflected Bragg wavelength is represented in the linear relationship in shown in Eq. (1): B = (˛ + )T + (1 − e )ε B
where ε is the change in strain, T is the change in temperature, ˛ (=0.55 × 10−6 /◦ C) is the fiber linear thermal coefficient, (=8.3 × 10−6 /◦ C) is the thermo-optic coefficient and e (=0.26) the effective photo-elastic coefficient. The Bragg wavelength changes linearly with temperature (∼10 pm/◦ C) and strain (∼1.2 pm/) in the wavelength region around 1500 nm. The compact electronic interrogation system designed and reported by Gebremichael et al. which forms the basis of the system used in this work has been described in detail elsewhere [1–3] but in summary comprises a 40 nm broadband super-luminescent light emitting diode source with an optical power of 20 mW and output wavelength centred at 1550 nm. The SLED diode source is driven by a custom designed and built stabilised current source and feedback controlled thermo-electric controller in order to ensure good stability of its output power. The system uses a Fabry-Perot tuneable filter which scans across the 40 nm free spectral range of the SLED broadband to pass a range of specific wavelengths. In addition, 8 low noise InGaAs PIN photodetectors with 30 dB signal-to-noise ratio (SNR) and a responsivity of 0.85 A/W at 1550 nm were used and connected to 8-analogue channels which are related to a windows PC for data storage, analysis and transfer data via the internet using LabView software as illustrated schematically in Fig. 2. The 8-channel system concept is based on the well-known method for interrogating FBG sensors using WDM. This architecture uses a scanning Fabry-Perot spectrometer which is controlled by a ramp voltage to interrogate the FBG Sensors with a very narrowband light that sweeps over the wavelength of the grating array of the sensor network via 1 × 8 WDM coupler. Each sensor in the array reflects back wavelength as a Gaussian peak profile depending on its strain and temperature. The reflected wavelength is detected
Fig. 2. Schematic of the interrogation system.
A. Kerrouche et al. / Sensors and Actuators A 151 (2009) 127–132
129
using 1 × 2 device coupled with detection electronics to receive the reflected spectra from the FBGs. 2.2. The peak identification software The reflected wavelength from each sensor is detected via a set of low noise photodetectors. The analogue output signals from the detectors are captured via a high speed custom board DSP (Digital Signal Processor – AdwinGold) which has two 12 bit ADC (analog to digital convertor) operating up to 1.2 MHz and two DAC (digital to analog convertors) that operate up to approximately 100 kHz of which one was used to scan the Fabry-Perot filter. The DAC generates a saw-tooth ramp voltage with ±2.5 V (peak-to-peak) and frequency of up to 300 Hz. The DSP board detects the light reflected back from each FBG sensor every scan voltage of the Fabry-Perot cavity. The monitoring system software is designed to determine the applied strain from the FBGs from data obtained from determining the transmission spectral peak and the Gaussian reflection peak. The DSP Software is written in a proprietary real-time multitasking language and up to 10 multitasking programs can be run at the same time. The main function of the DSP software is to execute a program loop, which acquires signals from the detectors in the form of time, resolved wavelength spectra. The software then determined from these spectra the position of the Bragg sensor peaks and stored these wavelength measurements in memory. A subsidiary function of the program is used to determine the peak-heights by identifying the position of the mean centroids. It is called every 10 s according to the system clock at a rate determined by a system variable GLOBAL-DELAY (GD) with a typical value of 400 under control of a hardware timer which is accurate to 25 ns. As shown in Fig. 3, the program starts when the Fabry-Perot cavity is scanned by the ramp voltage. The logging data values (yi ()) are added to an accumulator and are also multiplied by the value of the scantime and added to a second accumulator ( yi ()i ) above a pre-set threshold level until the signal falls below this threshold which indicates the end of the peak. The value of the centroid which
Fig. 3. Flowchart of the system software for one channel.
Fig. 4. Graph of strain vs. time showing the void gap of the reference grating data.
is the peak position is determined by calculating the ratio of two accumulators and then normalise by dividing by the time it takes to sweep through one spectrum
yi i centroid = yi
Finally, grating index indicates which peak relates to which sensor in the array as it is placed with the centroid measurement into the buffer FIFO memory for subsequent downloading to the PC for further analysis and display. 2.3. Field test applications The system described above has been employed for several monitoring applications, and been used in particular for railway bridge monitoring to evaluate structural performance. The results obtained are validated by the civil engineers involved in the work and the devices used have been seen to be robust compared to other sensor systems, such as electrical strain gauges. However, in field tests, the amount of data collected can be huge as static load tests often take many hours and dynamic load tests often take place over many days. This is particularly a problem where bridges or other structures are monitored over a long period during its life time and impacts strongly on the data storage and analysis used. The system was employed in studies on two bridges in Sweden. In the first study [1] a bridge reinforced with CFRP rods was tested to destruction during which very large forces were employed and thus very high strains were recorded (>4000 ). A second study was undertaken at a bridge at Frövi, Sweden [2] where this was an ‘in service bridge’ re-enforced with CFRP rods and tubes to increase its life expectance and performance. Here the bridge was rather stiff with maxima strain from heavy trains of ∼20 and from passenger trains
