On Condition Monitoring of a Corroding Steel Truss Bridge – A Case Study

Abstract

A railway steel truss bridge connecting the mainland India with Rameswaram had corrosion issues. This bridge was continuously monitored from August 2020 through December 2022 to ensure the serviceability of the bridge. This paper summarizes the technical learning from the monitoring. The truss bridge was heavily instrumented – 84 out of 92 members were instrumented with linear electrical strain gauges, 20 of the 52 nodes had biaxial accelerometers, and 10 resistance-based temperature sensors. For each train pass, the collected data was analyzed to estimate the axle loads of the train, cross-sectional areas of the members, and certain other parameters from the acceleration time history. While the observed strains could be matched excellently with the strains estimated by simple truss analysis in 70 members, the computed accelerations were the order of magnitude lower for all the acceleration measurements. The displacement profile obtained by double integrating the acceleration with appropriate filtering, as suggested in the literature, did not agree with the predicted displacement even qualitatively. As one would expect, even though there was a steady deterioration of the structure, the commonly used dynamic response parameters – the natural frequency and mode shapes – did not reflect the change statistically. The extreme strains in most of the members showed a steady increase or decrease in values. But the truss being statically determinate, the strains could not capture the loss in the area of members due to corrosion at locations other than where they were located. However, certain frequency and time domain parameters of the acceleration were able to capture the steady deterioration of the truss bridge. Thus, in the case of statically determinate truss bridges, data-driven approaches seem to perform better than physics-based approaches in determining the changes occurring in the bridge.