This paper describes the development and implementation of a long term monitoring system for structural safety assessment and for traffic identification using a B-WIM system in an old railway bridge. The corresponding traffic estimates for a period of two years are presented and are within typical official values provided by the railway administration.
The speed spectrum was calculated using the phase difference of the signals as proposed by Leander et al. [15] and led to accurate speed estimations under low computational time.
It is known that the measurements in the rails tend to have important dynamic effects due to the proximity between the wheels and the strain gauges, while on the bridge these effects are much smaller. However, the strain gauges placed on Trezói Bridge are not sensitive enough to measure separate axles and, in some cases, separate bogies. The methodology proposed in this paper consisted in using each type of measurement in accordance with their specific advantages. Therefore, the strain gauges in the rails outside the bridge were used to estimate the number of axles, axle distances and speed. These variables were then used as input to estimate the axle loads based on the measurements in the bridge. The comparison between the estimation of axle loads using strain measurements on the rails and on the cross-girders allowed to conclude that the difference between estimated and expected axle loads is close to 1% and the difference between loads estimated from rail and cross-girder measurements is inferior to 0.70%. However, due to the local dynamic amplification in the rails for higher speeds, in the proposed methodology the long term load estimation was made using exclusively the measurements in the bridge.
To increase the accuracy of weight estimation, an optimization based on genetic algorithm was implemented. This algorithm was chosen due to its ability to find optimal solutions where local maxima or minima exist and due to the fact that, in general, they don’t need an accurate initial guess. In this paper, the initial guess for the optimization process was obtained from Moses algorithm [1].
The results for axle loads, speed and axle spacing from a period of two years were compared with official values obtained from the railway administration and with the values of the fatigue trains of the EN1991. The estimated values are within the reference value ranges and are coherent with the typical values observed in this bridge.
As for future work, the dynamic amplification factors will be evaluated, an algorithm based on the traffic random variables will be proposed to allow the simulation of real trains and to extrapolate for future traffic. Furthermore, the correlation between traffic characteristics and the fatigue behavior of critical elements will be evaluated. Also, the applicability of the presented procedures to railway bridges supporting two or more railroad lines will be evaluated. This procedure is feasible to be adapted to such bridges but would have to include a method of detecting in which line each train is crossing and methods to separate two or more trains crossing the bridges at the same time. Some work has been done by Javier González Silva and Karoumi [16] in the subject of weight in motion methodologies in a bridge with multiple railway lines.
Acknowledgments
The authors would like to acknowledge: (1) all the supports provided by the Portuguese Foundation for Science and Technology (FCT) to ViBest/FEUP for the development of research in the area of Fatigue Assessment and Continuous Dynamic Monitoring; (2) the support of the Portuguese railway agency REFER and particularly of Eng. Ana Isabel Silva and Eng° José Carlos Clemente; (3) the Ph.D. Scholarship (SFRH/BD/75781/2011) provided by FCT to the first author and (4) the financial support from RFCS (Research Fund for Coal and Steel) in the context of the European Project Fatigue Damage Control and Assessment for Railway Bridges (FADLESS).