Damage localization for bridges monitored within one cluster based on a spatiotemporal correlation model of strain monitoring data

The bridges monitored within one cluster refer to several medium- and small-span beam bridges with similar structural characteristics located in a continuous elevated corridor. The variation in the strain monitoring data of these bridges comprehensively reflects diverse coupling effects. These complex coupling factors present great challenges for the damage diagnosis of bridges. To address this issue, a damage localization method for bridges monitored within one cluster is proposed based on a spatiotemporal correlation model of strain monitoring data between bridges. First, a deep learning architecture combining a convolutional neural network (CNN) with a long short-term memory (LSTM) network is established, which can reveal the complex time-varying mapping relationship between the strain monitoring data for similar bridges within one cluster to obtain an accurate spatiotemporal correlation model. Second, a strain prediction framework is presented that uses the proposed spatiotemporal correlation model after training. On this basis, the predicted and measured strains can be utilized to calculate a damage localization index that is not affected by complex coupling factors. Then, combined with abnormal diagnosis theory, the proposed index is implemented to accurately localize damage in all bridges within one cluster. Finally, the application of the proposed method to three actual bridges monitored within one cluster demonstrates the accuracy of the spatiotemporal correlation model and the effectiveness of structural damage localization.