Abstract

Deep learning is an effective way to deal with classification problems. This paper reported a deep-learning algorithm used for bridge damage detection. The continuous bridge deflection based on the fiber optic sensing technology was selected to establish the datasets for supervised learning. With a scale-down bridge model, three damage scenarios and an intact state were simulated. A supervised learning model based on the deep convolutional neural networks (CNN) was proposed. After the training under ten-fold cross-validation, the model accuracy can reach to 96.9% for damage classification. By comparison with other four machine learning methods, the proposed model demonstrated its decent abilities in extracting damage features and distinguishing damage from symmetrical locations.

© 2020 The Author(s)