IEEE SENSORS JOURNAL, VOL. 8, NO. 7, JULY 2008 1243 Hybrid Fiber-Optic/Electrical Measurement System for Characterization of Railway Traffic and Its Effects on a Short Span Bridge Ricardo Miguel da Costa Marques Pimentel, Maria Cristina Beirão Barbosa, Nuno Manuel Silva Costa, Diogo Rodrigo Ferreira Ribeiro, Luís Alberto de Almeida Ferreira, Francisco Manuel Moita Araújo, and Rui Artur Bártolo Calçada Abstract The characterization of traffic effects on a short span railway bridge in Northern Portugal with a new hybrid platform that allows the simultaneous assessment of signals generated by a sensing network composed of both electrical and fiber Bragg grating-based sensors was demonstrated. A commercial fiber-optic-based train characterization system using a bridge weight-in-motion algorithm was also developed and tested. By using only three fiber Bragg grating sensors the system allows on-motion determination of train speed and weight distribution. Index Terms Fiber Bragg gratings, measurement units, railway bridge, sensors, weight-in-motion. I. INTRODUCTION T HE KNOWLEDGE of the dynamic effects on railway bridges is of major importance for the following reasons: the vibrations induced by the passage of the trains over the bridge originate, in general, displacements or internal stresses in structures greater than those produced when the loading is statically applied; excessive vibrations of the structure may lead to a magnification of fatigue phenomena; the deformations and accelerations of the bridge should be controlled and kept within certain limit values, in order to ensure the stability of the track and of the contact wheel-rail at all times; the accelerations in the vehicles should be limited so that the passengers comfort can be guaranteed [1], [2]. Simultaneously, the ability to determine train characteristics from its velocity to its weight distribution as a real time application is gathering increasing interest, for it allows the responsible entity for the infrastructures to actually know applied loads and to have a better control of the operating companies. One of the used processes to determine traffic characteristics is the bridge weight-in-motion (B-WIM) system [3]. Manuscript received August 2, 2007; revised December 21, 2007; accepted January 1, 2008. Published July 16, 2008 (projected). The associate editor coordinating the review of this paper and approving it for publication was Prof. Jose Lopez-Higuera. R. M. da Costa Marques Pimentel, D. R. F. Ribeiro, and R. A. B. Calçada are with the Department of Civil Engineering, Faculty of Engineering, University of Porto, 4200-465 Porto, Portugal (e-mail: mec05001@fe.up.pt; dribeiro@fe.up.pt; ruiabc@fe.up.pt). M. C. B. Barbosa and N. M. S. Costa are with FiberSensing, 4470-640 Maia, Portugal (e-mail: cristina.barbosa@fibersensing.com; nuno.costa@ fibersensing.com). L. A. de Almeida Ferreira and F. M. M. Araújo are with FiberSensing, 4470-640 Maia, Portugal and also with INESC Porto, 4200-465 Porto, Portugal (e-mail: luis.ferreira@fibersensing.com; francisco.araujo@fibersensing.com). Digital Object Identifier 10.1109/JSEN.2008.926519 In the last few years, the examples of application of fiber Bragg grating sensors for structural health monitoring have become more and more frequent [4], [5]. The explanation for this fact relies not only on the particularly appropriate characteristics associated to these devices for measuring the most relevant structural parameters without revealing a number of disadvantages usually attributed to conventional sensors, but also on the growing range of available sensors based on this technology, as well as several types of measurement units with increasing capabilities. In particular, fiber Bragg sensors are most suited for measuring strain in rail track applications, where EM/RF noise often compromises the performance of electrical strain gages [6], [7]. Nevertheless, for the complete dynamic characterization of a structure like a railway bridge, acceleration measurements are mandatory. These can be performed with standard electrical accelerometers, readily available with a diversity of characteristics. However, measurements gathered with fiber-optic and electrical sensors must be synchronized, in order to provide useful data. In this work, it is shown the applicability of a new measurement unit based on a hybrid platform concept, which allows the simultaneous assessment of signals generated by the fiber Bragg grating sensors used to measure strain in the rail and bridge elements, as well as the electrical accelerometers. In addition, on-motion determination of train compositions weight, as well as its distribution, with a simple dynamical measurement system based on three fiber Bragg grating sensors and an algorithm specifically developed for this purpose was demonstrated. II. BRIDGE DESCRIPTION Canelas Bridge is situated on Km +282.944 of the Linha do Norte of the Portuguese railway network. This piece of the national railway system has recently been modernized to allow faster train circulation the Alfa Pendular train can reach up to 220 km/h. The Canelas Bridge is a Filler Beam Bridge with a concrete slab with embedded steel profiles. The bridge s cross section is composed of two symmetric and independent half decks, each supporting one of the railway tracks. There are six simply supported spans approximately 12 m wide. In Fig. 1, a picture of the bridge and its cross section are presented. 1530-437X/$25.00 2008 IEEE
1244 IEEE SENSORS JOURNAL, VOL. 8, NO. 7, JULY 2008 Fig. 1. Canelas bridge: (a) side view and (b) cross section. Fig. 2. Hybrid measurement unit used in the experimental campaigns. III. MEASUREMENT SYSTEM The measurement unit used on the experimental campaigns is based on the BraggSCOPE technology, which combines a high-power broadband optical source with proven thin-film optical filtering technology in an intelligent approach to perform dynamic measurement of the Bragg wavelength. By integrating proprietary add/drop WDM design, it allows the measurement of up to four sensors connected in series, operating within predefined wavelength bands. This, in combination with optical switching makes this measurement technology a cost effective solution for medium scale dynamic sensing networks. The BraggSCOPE benchtop measurement units are expansible through a built-in PXI subrack, so it is possible to increase the number of electrical and optical switching modules or PXI standard data acquisition cards. This is the so-called hybrid concept, which allows the simultaneous assessment of signals generated by a sensing network composed by both electrical and fiber Bragg grating sensors (Fig. 2). Through the unit touch-screen graphical interface, it is possible for the user to configure optical and electrical sensors in a similar way, as well as acquire, operate and store both types of signals. Strain and displacement sensors based on fiber Bragg gratings were used on the experimental campaigns. For strain measurements, weldable fiber Bragg grating strain gauges were applied. This type of sensor is clearly preferable since it minimizes surface preparation requirements and precludes epoxy curing processes during installation. The spot welding process has no implications on the rail tracks for it consists on creating an electrical current between two close points. The weldable fiber Bragg grating strain gauge [Fig. 3(a)] is based on a capillary stainless steel tube containing the sensing element, which is laser welded to a thin stainless steel base that is used for direct spot welding to the metallic structure. The input/output fibers are protected with standard 900 buffer. For permanent installations, this buffer is further protected by a 3 mm armored cable, while the sensors can be covered with a metallic cap also spot weldable to the structure [Fig. 3(b)] [8]. For the first testing campaign, strain was controlled under the bridge on the midspan of two consecutive spans aligned with both rail [Fig. 4(a)]. Strain was also monitored on the rail with two sensors, being one of them deployed along the neutral axis and the other one on the basis of the rail [Fig. 4(b)], mainly for separating and evaluating the effects of rail track deformation due to total deflection of the deck and to train wheels loading. On the second campaign, aimed to calibrate the B-WIM algorithm, two weldable fiber Bragg grating strain gauges were spot-welded to the basis of the rail in two consecutive spans and another two were positioned under the bridge, on the midspan, aligned with both rails. Vertical displacements were obtained using optical long gauges [Fig. 5(a)]. The long gauges consist on a tensioned invar wire, which strain variations are measured with a fiber Bragg grating and integrated along its length for displacement determination. The tension had to be adjusted so that the wire vibration modes did not coincide with the bridge resonances. Vertical displacements were controlled in two consecutive mid spans and on one bearing. Accelerations were measured with traditional piezoelectric accelerometers [Fig. 5(b)]. IV. EXPERIMENTAL CHARACTERIZATION OF RAILWAY TRAFFIC The characterization of railway traffic was performed through the implementation of a B-WIM algorithm [9]. This algorithm allows the evaluation of the speed of the train, acceleration, longitudinal position of each axle and its respective load.
DA COSTA MARQUES PIMENTEL et al.: HYBRID FIBER-OPTIC/ELECTRICAL MEASUREMENT SYSTEM 1245 Fig. 3. Sensor details: (a) weldable strain gauge and (b) weldable protection cap. Fig. 4. Sensor positioning and installation: (a) cross section positioning of the sensors; (b) detail of sensor position in the rail; (c) installed weldable gauge on the rail; and (d) spot-welding process. Fig. 5. Examples of the installed sensors: (a) invar wire long gauge and (b) piezoelectric accelerometer. It has been demonstrated that is possible to identify the geometry and speed of the train using solely the deformation mea- surements of the bridge [10]. In this case, the accuracy achieved in terms of the position of the axles is not very significant, and it is often only possible to identify one bogie rather than its individual axles. This is due to the fact that the length of the influence line of the bridge is large when compared to the distance between axles and therefore it is not feasible to individualize them from the deformation records of the bridge. In order to overcome this problem, it has been decided that the bridge deformation measurements would not be used and the deformation measurements of the rail would be used instead for the identification of the geometry and speed of the train. Since the influence line of the rail deformation is short, it is possible to clearly identify the peaks corresponding to the passage of each of the train axles. In Fig. 6, the register for the deformation on two different points on the rail (E5 and E7) is presented. The shown event represents the passage of the first three bogies of an Alfa Pendular.
1246 IEEE SENSORS JOURNAL, VOL. 8, NO. 7, JULY 2008 Fig. 6. Deformation in two different points on the rail track. Fig. 8. Alfa Pendular train. Fig. 7. Deformation in one point of the deck. Fig. 9. Speed of the calibration vehicle. Using a third measurement of deformation in one of the bridge s beam (Fig. 7) loading per axle is calculated. The registered data is filtered using a low-pass digital filter to eliminate any dynamic effect on the response and used as an input for Moses Algorithm [3]. Secondly, loading per axle is calculated. This is possible using a third measurement of deformation, this time under the deck (Fig. 7). The registered data is filtered using a low-pass digital filter to eliminate any dynamic effect on the response and used as an input for Moses Algorithm [3]. Fig. 7 shows the deformation at the midspan of one metallic beam of the deck aligned with one of the rails for the whole Alfa Pendular passage (E4). The accuracy of the B-WIM is strongly affected by the influence line. Given the difficulty in obtaining a rigorous estimate of the influence line by theoretical means, its estimate is usually made experimentally. The influence line can be automatically obtained based on a matrix methodology in which results of measurements made for the passage of a calibration vehicle, i.e., a vehicle with known characteristics, are used [11]. In order to obtain a good estimate of the influence line a passage of the Alfa Pendular train (Fig. 8) at reduced speed was used. In Fig. 9 are presented the speeds for the 24 axles of the train showing that the speed has varied slightly along the train, causing difficulties in the calibration process. In Fig. 10, the calibrated influence line is presented. In order to eliminate small oscillations of the influence line due to dynamic or noise components that were impossible to eliminate from the signals, the Fig. 10. Influence line of deformation of the deck. influence line ordinates are adjusted by means of a polynomial function. Some of the results obtained by the application of the algorithm are shown in Fig. 11 corresponding to several passages of urban trains (Fig. 12) at different schedules. It was possible to observe that higher loads per axle occur at peak time (rush hour), which can be explained by the concentration of passengers. V. EXPERIMENTAL EVALUATION OF TRAFFIC EFFECTS The importance of the dynamic effects induced by traffic loads on bridges is usually measured by the so-called dynamic amplification factor, which can be defined as the ratio between the maximum dynamic response of the bridge and the corresponding
DA COSTA MARQUES PIMENTEL et al.: HYBRID FIBER-OPTIC/ELECTRICAL MEASUREMENT SYSTEM 1247 Fig. 11. Position and load of each axle for several passages of urban trains. Fig. 14. Power spectral density estimate and magnitude of the filter frequency response function for the passage of an urban train at a speed of 110 km/h. TABLE I DYNAMIC AMPLIFICATION FACTORS RELATIVE TO STRAINS Fig. 12. Urban train. Fig. 15. Vertical displacement at midspan of the first span (Sensor D4) for the passage of an urban train at a speed of 110 km/h. Fig. 13. Strain records for sensor E4 for the passage of an urban train at a speed of 110 km/h. maximum static response. The maximum dynamic response is directly obtained by taking the maximum value of the measured data. On the other hand, the maximum static response can be calculated by applying a digital low-pass filtering to the measured time series, so as to eliminate the corresponding dynamic components, as referred in the previous section. In Fig. 13, the dynamic response of the bridge is presented, characterized by the deformation at midspan of the first span (strain gauge E4) for the passage of an urban train at the speed of 110 km/h. The static response was obtained by the application of a Chebyshev (type II) filter, of polynomial order 10, with an out-of-band passage attenuation of 40 db. In Fig. 14 is presented the estimate of the power spectral density function of the dynamic response, as well as the amplitude of the frequency response function of the applied filter. A closer inspection of the figure enables to observe the existence of a contribution to the dynamic response corresponding to the first mode shape frequency [12]. It can also be noted that the spectral ordinates associated with the quasi-static component of the response do not appear overlapping those associated with the natural frequencies of the bridge, thus concluding the adequacy of the filter application procedure for the assessment of the dynamic amplification factors. In Table I are indicated the maximum values of the dynamic and static response, as well as values of the dynamic amplification factors, obtained for the passage of the urban train at a speed of 110 km/h and the Alfa Pendular train at the speed of 140 km/h. In Fig. 15 is represented the dynamic response of the bridge characterized by the vertical displacement at midspan of the first span of the deck (D4) for the passage of the urban train. The static response was also obtained by the application of a filter to the dynamic response. In Table II are indicated the maximum values of the dynamic and static response relative to displacements. The
1248 IEEE SENSORS JOURNAL, VOL. 8, NO. 7, JULY 2008 TABLE II DYNAMIC AMPLIFICATION FACTORS RELATIVE TO DISPLACEMENTS Fig. 16. Scheme of the developed system. comparison of Tables I and II enables to conclude that the dynamic amplification factors in terms of strains and displacements are practically identical. The dynamic amplification factors referring to displacements may as well be evaluated based on acceleration readings. The series corresponding to the dynamic component of the displacement can be obtained by the double integration of the acceleration series. The signal trend resulting from the integration of the quasi-static component of the signal can be removed by the application of the high-pass digital filter to the integrated signals. The dynamic amplification coefficient is calculated by means of the expression, where is the maximum value of the series containing only the dynamic component of the response and is the maximum value of the static response. From Table II, it can be concluded that the amplification coefficients defined by this process are higher because, since the maximum values of the series corresponding to the static and dynamic components do not generally occur in the same time instant. VI. DEVELOPED COMMERCIAL SYSTEM The algorithm for in-motion train characterization soon revealed its practical application and commercial interest. The use of fiber-optic-based sensors allowed the definition of a simple scheme to monitor railways, easy to install and price competitive (Fig. 16). The developed tool makes it easier for the infrastructure owners to control and know the loads that are actually being applied on their network. Obvious examples of the advantages of installing such a system are the control over the operating companies and the ability to estimate network loading and assess its impact in terms of necessary repair or reinforcement works along the infrastructure. In order to completely characterize train compositions travelling in two different railway tracks, only eight fiber-optic strain sensors organized in two branches are needed. In each branch, one of the four sensors connected in series works as a trigger starting dynamic data acquisition every time a train is detected. The other three sensors measure strain in two different points of the rail and on the bridge structure. The measurement unit can be stored in a technical closet as far as 20 km from the sensors. The equipment is prepared with an UPS and data transition devices in order to acquire, configure, calculate, store and automatic and continuously transmit data. Wireless video cameras can also be installed and synchronized with the measurement unit. One system installed for monitoring two railway tracks will work as follows: the measurement unit is continuously acquiring sensor signals as well as video signals from the optional video cameras. Nevertheless, these signals are not stored unless the approximation of a train travelling in one of the tracks is detected by one of the trigger sensors. When this happens, the measurement unit starts recording the measured signals from the three strain sensors that allow the determination of the train speed and weight for as long as the train is passing over the bridge. After the passage of the train the measurement unit stops recording and data is processed and stored in an internal database while all signals are being normally acquired. VII. CONCLUSIONS In this paper, the development of a new bridge weight-inmotion algorithm was demonstrated. This work was based on data collected during experimental campaigns aimed to characterize traffic effects on a short span railway bridge in Northern Portugal that used a new hybrid platform that allows the synchronous assessment of signals generated by an electrical/fiber Bragg grating sensing network. The obtained results allowed us to design a competitive complete system for on-motion determination of train speed, acceleration and weight distribution based in only three fiber Bragg grating sensors and a single optical channel measurement unit. The application of the system to the Canelas Bridge during its life cycle will enable the monitoring of the real traffic characteristics in the structure, which is a relevant issue not only for this bridge but also to the adjacent infrastructures located at the same railway line section. Monitoring will additionally provide the management authorities with facilitated control over the operating companies, especially those of freight transport. The monitoring system also allows the assessment of the dynamic effects and useful data for the evaluation of the fatigue behavior of the bridge. ACKNOWLEDGMENT The authors acknowledge the support and collaboration provided by Eng. Ana Isabel Silva of REFER. REFERENCES [1] R. Delgado, R. Calçada, and I. Faria, Bridge-vehicles dynamic interaction: Numerical modelling and practical applications, in Proc. Workshop Bridges for High-Speed Railways, Porto, Portugal, Jun. 3 4, 2004. [2] D. Ribeiro, Dynamic behaviour of bridges under high-speed railway traffic, (in Portuguese) M.Sc. thesis, FEUP, Porto, Portugal, 2004. [3] F. Moses, Weigh-in-motion system using instrumented bridges, ASCE Transp. Eng. J., vol. 105, no. TE3, pp. 233 249, 1979. [4] K. Kincade, Fibre sensors lay groundwork for structural health monitoring, Laser Focus World, vol. 42, no. 2, pp. 63 67, 2005.
DA COSTA MARQUES PIMENTEL et al.: HYBRID FIBER-OPTIC/ELECTRICAL MEASUREMENT SYSTEM 1249 [5] Sensing Issues in Civil Structural Health Monitoring, F. Ansari, Ed. Dordrecht, Holland: Springer, 2005. [6] P. Boffi, R. Bratovich, F. Persia, A. Barberis, M. Martinelli, A. Basso, G. Bucca, A. Nicchio, and M. Bocciolone, Fibre sensor for collector strain monitoring in the pantograph-catenary interaction, in Proc. 18th Int. Conf. Optical Fibre Sensors, Cancun, Oct. 23 27, 2006. [7] C.-Y. Wang, H.-L. Wang, and M.-H. Chen, Structural health monitoring activities of applying optical fibre sensors in Taiwan, in Proc. 18th Int. Conf. Optical Fibre Sensors, Cancun, Mexico, Oct. 23 27, 2006. [8] C. Barbosa, N. Costa, L. A. Ferreira, F. M. Araújo, H. Varum, and A. Costa, Monitoring the new circular pedestrian steel bridge over the São Roque and Botirões channels with weldable fibre-bragg grating sensors, in Proc. 18th Int. Conf. Optical Fibre Sensors, Cancun, Mexico, Oct. 23 27, 2006. [9] R. Pimentel, Characterization of traffic and its effects on short span railway bridges, (in Portuguese) M.Sc. thesis, FEUP, Porto, Portugal, 2007. [10] R. Karoumi, C. J. Wiberg, and A. Liljencrantz, Monitoring traffic loads and dynamic effects using an instrumented railway bridge, Eng. Struct., vol. 27, no. 12, pp. 1813 1819, 2005. [11] M. Quilligan, Bridge weigh-in-motion, development and testing of a 2-Dimensional multi-vehicle algorithm, Licentiate thesis, Royal Inst. Technol., Stockholm, Sweden, 2003. [12] J. Rodrigues, Dynamic behaviour of Canelas Bridge under railway traffic, structural dynamics, in Proc. EURODYN, 2002, pp. 997 1002. Ricardo Miguel da Costa Marques Pimentel received the graduate degree in civil engineering from the Faculty of Engineering, University of Porto, Porto, Portugal, in 2005. He is currently pursuing the M.Sc. degree in civil engineering structures at the University of Porto. In January 2008, he received a Ph.D. scholarship. His main research activities have focused on railway bridges dynamics, experimental analysis of structures, and bridge weigh-in-motion algorithms. Maria Cristina Beirão Barbosa received the degree in civil engineering (structures) from the Faculty of Engineering, University of Porto, Porto, Portugal, in 2004. She is an Applications Engineer with FiberSensing, an INESC Porto spin-off company that develops, manufactures and installs advanced monitoring systems based on fiber-optic sensing technology and addresses markets such as structural health monitoring in civil and geotechnical engineering, aerospace, and energy production and distribution. Previously, she was with the Laboratory for Seismic and Structural Engineering (LESE) with an Investigation Scholarship awarded from the Faculty of Engineering, University of Porto. Nuno Manuel Silva Costa received the technical degree in electronics in 1989 from the Colégio de Gaia, Portugal. Between 1989 and 2000, he was an Electronic Technician with INESC Porto. He is now a Fiber- Optic Technician with FiberSensing, an INESC Porto spin-off company that develops, manufactures and installs advanced monitoring systems based on fiber-optic sensing technology, and that addresses markets such as structural health monitoring in civil and geotechnical engineering, aerospace, and energy production and distribution. Diogo Rodrigo Ferreira Ribeiro received the graduate degree in civil engineering from the Faculty of Engineering, University of Porto, Porto, Portugal, in 2002 and the M.Sc. degree in civil engineering structures from the University of Porto in 2005. He is currently pursuing the Ph.D. degree in the area of the experimental and numerical assessment of the dynamic effects in railway bridges. In 2002 2003, he was a Structural Engineer with the design office of Gabinete de Estruturas e Geotécnia (GEG). In 2007, he became a Lecturer with the Technical Institute of Engineering of Porto. Luís Alberto de Almeida Ferreira graduated in 1991 in applied physics (optics and electronics) and in 1995 received the M.Sc. degree in optoelectronics and lasers (white-light interferometry and signal processing in optical fiber sensors), both from the University of Porto, Porto, Portugal. He received the Ph.D. degree in physics from the University of Porto in 2000 in interrogation of fiber-optic Bragg grating sensors, after developing part of his research work in fiber-optic sensing in the Physics Department, University of North Carolina, Charlotte. He is the currently an Engineering Manager at FiberSensing, an INESC Porto spin-off company that he co-founded, and that develops, manufactures, and installs advanced monitoring systems based on fiber-optic sensing technology, and that addresses markets such as structural health monitoring in civil and geotechnical engineering, aerospace, and energy production and distribution. He is also a Senior Researcher at the Optoelectronics and Electronic Systems Unit of INESC Porto, where he develops his main R&D activity in the areas of fiber-optic sensing and optical communications. He is author/co-author of more than 100 international communications, papers, and patents in the fields of fiber-optic sensing and fiber-optic communications. Francisco Manuel Moita Araújo graduated in 1993 in applied physics (optics and electronics) from the University of Porto, Portugal. He received the Ph.D. degree in physics from the the University of Porto, in 2000 (fiber Bragg gratings). He is Product Development Director with FiberSensing, an INESC Porto spin-off company, developing fiber-optic sensors and systems for different markets, such as structural health monitoring. He is a co-founder of FiberSensing. He is also a Senior Researcher with the Optoelectronics and Electronic Systems Unit of INESC Porto. His main activity research is related with optical communications and fiber-optic sensing. Previous positions included leadership of the Fiber Optic Technologies Unit at MultiWave Networks Portugal, a company developing subsystems for fiber-optic communications, Assistant Professor at the Physics Department of the University of Porto (Faculty of Sciences), and Senior Researcher at the Optoelectronics and Electronic Systems Unit of INESC Porto, were he developed research in the area of fiber-optic technologies from 1993 to 2001. He is author/coauthor of more than 100 international communications, papers, and patents in the fields of fiber-optic sensing and fiber-optic communications. Rui Artur Bártolo Calçada received the graduate degree in civil engineering from the Faculty of Engineering, University of Porto, the M.Sc. degree in civil engineering structures and the Ph.D. degree in civil engineering from the University of Porto, Porto, Portugal, in 1995, in 2003. He started his lecturing career in 1994 as Lecturer at the Engineering School of the University of Minho. In 1996, he joined the Faculty of Engineering of the University of Porto, first as Lecturer and since 2003 as Assistant Professor. His main research activities have focused on road and railway bridges dynamics, numerical and experimental analysis of structures, bridge-vehicle and track-bridge interactions, and vibrations on track, on transition zones, and on the vicinity of high speed rail lines. He is currently responsible for many research projects and involved in the supervision of ten Ph.D. and eight M.Sc. students in these areas. He has published over 80 papers in international journals and conference proceedings.