Condition Monitoring of Steel Truss Bridge Using Acceleration Data

Abstract

This study investigates the efficacy of utilizing data-driven techniques to continuously monitor the condition of steel truss bridges. Rather than relying on computationally intensive modeling methods, this approach utilizes parameters derived from raw acceleration time history data of on-field vibrations caused by regular traffic. Specifically, the study examines the relationship between changes in these parameters and the occurrence of damage in steel truss bridges. To conduct this research, the Pamban Bridge, which is over 100 years old, was fitted with accelerometers at various bottom node points on the bridge. The acceleration data was then parameterized into primary, secondary, and tertiary order parameters based on amplitude, frequency, and duration. These parameters were then analyzed to determine their suitability for detecting damage. This study examines the continuous variations in each parameter over an identical duration from March to July 2021 and March to July 2022, spanning 136 days each and over 1000 train passes. The bridge underwent retrofitting during the intervening period. A linear best-fit line is found for each sensor reading for the considered duration. The slope and intercept of the linear fit are studied. It was found that the change in the intercept values indicated the changes that occurred consistently and reflected the expected trends between sensor locations. To further validate the findings, the observations made through these parameters were compared with the retrofitting data of each member. The parameters, such as root mean square acceleration, arias intensity, characteristic intensity, and cumulative absolute velocity, were consistent and exhibited similar patterns during the observed period. Overall, this approach to condition monitoring is highly efficient and requires only on-field vibrations caused by regular traffic to detect potential damages in steel truss bridges, making it an ideal method for continuous monitoring without operational downtime.