Methodology for the damage detection of aging bridges based on multi-data and deep learning

초록

The number of bridges over 30 years old is rapidly increasing worldwide. As a result, the maintenance management and performance evaluation of bridges are of great interest to society and structural health monitoring (SHM) systems are attracting significant attention as a technology to facilitate these actions. SHM systems are a method of collecting and analyzing measurement data from sensors and monitoring structural responses in real time. Most SHM systems collect various types of response data, but only use limited types of data for analysis. Therefore, in this paper, we propose a method combining a deep learning model for bridge deterioration estimation and a damage localization model using multi-data. Both models use a common convolutional neural network, but the purpose of each model is different based on its learning method. We expect that the proposed method will be widely used in the analysis of real multi-data from aging bridges and lead to the automation of maintenance and performance evaluation.