ABSTRACT
This study introduces a deep learning approach to identify patterns in floating bridges, specifically employing the Long Short-Term Memory (LSTM) algorithm to analyze time-domain data. Wind, wave, and displacement that have been measured in the Bergsøysund bridge between the years 2014 and 2018 were used for the pattern recognition. For the feasibility study, the LSTM-based model was trained to recognize the complicated relation between the variables including the environmental conditions and induced response. After that, the response predicted by the trained model was directly compared to the measured one to evaluate the pattern model. According to this study, the proposed method can effectively recognize the structural pattern of a floating bridge from measured data. In addition, the pattern analysis-based structural health monitoring method can be used for early detection of the structural condition change of the floating bridge.
