Vibration-Based Damage Detection of Bridges under Varying Temperature Effects Using Time-Series Analysis and Artificial Neural Networks

Structural health monitoring (SHM) has become a very important research area for evaluating the performances of bridges. An important issue with continuous SHM and damage detection of bridges is the effects of temperature variations on the measurement data, which can produce bigger effects in the response than the damage itself. In this study, a sensor-clustering-based time-series analysis method integrated with artificial neural networks (ANNs) was employed for damage detection under temperature variations. The damage features obtained solely using the time-series-based damage-detection algorithm are very effective for damage assessment; however, they yield false positives and negatives when temperature variations are present. Therefore, ANNs were used to compensate the temperature effects on the damage features obtained from time-series analysis. This methodology is applied to a footbridge finite-element model in which 2,000 simulations with temperature effects and damage cases were conducted. Results reveal that the proposed method can successfully determine the existence, location, and relative severity of damage using output-only vibration and temperature data even when temperature variations are present.