Seasonal effects on the stiffness properties of a ballasted railway bridge

• Published in: Engineering structures 2013-12, Vol.57, p.63-72
• Main Authors: Gonzales, Ignacio, Ülker-Kaustell, Mahir, Karoumi, Raid
• Format: Article
• Language: English
• Subjects: Applied sciences; Ballasted track; Bayesian updating; Bridges; Bridges (structures); Buildings. Public works; Climatology and bioclimatics for buildings; Dynamics; Exact sciences and technology; Frost; Järnvägsgruppen - Infrastruktur; Markov-Chain Monte Carlo sampling; Mathematical models; Railway bridges; Railway tracks (foundations and track subgrades); Resonant frequency; Seasonal effects; Seasons; Spreads; Stiffness; Strength of materials (elasticity, plasticity, buckling, etc.); Structural analysis. Stresses; Transportation infrastructure; Variance
• ISSN: 0141-0296; 1873-7323; 1873-7323
• DOI: 10.1016/j.engstruct.2013.09.010

•A considerable increase in the bridge stiffness during the cold season is reported.•Distributions of elastic moduli of soil, ballast and concrete are determined in the cold and the warm season.•The elastic modulus of ballast increases by one order of magnitude during the cold season.•The roller bearings in this simply supported bridge “stick” at very small amplitudes of vibration. In this article it is shown empirically that ballasted bridges in cold climates can exhibit a step-like variation of their natural frequencies as the yearly season changes. The bridge under study was observed to have significantly higher natural frequencies (as much as 35%) during the winter months compared to the summer. This variation was rather discrete in nature and not proportional to temperature. Furthermore the increase in natural frequencies took place only after the temperatures had dropped below 0°C for a number of days. It was thus hypothesized that this change in natural frequencies was due to changes in the stiffness parameters of some materials with the onset of frost. In low temperature conditions not only the mean value of the measured frequencies increased, but also their variance increased considerably. Given the large spread of the measured natural frequencies, the stiffness parameters were assumed to be stochastic variables with an unknown multivariate distribution, rather than fixed values. A Bayesian updating scheme was implemented to determine this distribution from measurements. Data gathered during one annum of monitoring was used in conjunction with a finite element model and a meta model, resulting in an estimation of the relevant stiffness parameters for both the cold and the warm condition.