Eliminating Temperature Effects in Damage Detection for Civil Infrastructure Using Time Series Analysis and Autoassociative Neural Networks

Abstract

Temperature effects may mask the variation in structural properties or responses due to damage by causing equally or even larger changes in structures, resulting in false positive or false negative detections. These temperature effects should be eliminated during the process of damage detection; however, the complexity of operating civil structures makes it difficult to separate those influences from structural damage using closed form solutions or parametric approaches. In this study, a new damage detection approach based on autoassociative neural networks (AANNs) is proposed to detect the structural damage in bridges by eliminating the temperature effects. First, time series analysis–based damage features extracted from undamaged structure under varying temperature effects only are used to train the AANN. The trained neural networks were then fed by damage features with both damage and temperature effects. The results show that the proposed method can detect and locate the damage by tracking the prediction errors of the AANN under varying temperature effects.