ABSTRACT
The structural stiffness of a submerged floating tunnel (SFT) moored by tethers is determined by the stiffness of the tethers. If the tethers are damaged, the stiffness of the entire system changes first, resulting in a change in the structural characteristics, including the natural frequencies. In addition, local deformation of the tunnel segments might occur owing to a decrease in local stiffness. A severe structural safety problem can occur if significant damage occurs at the tethers. Therefore, the structural states of the tethers should be carefully checked using a rational method to ensure the structural safety of the submerged transportation system. Owing to the characteristics of submerged structures, there is a limitation on the applicable sensors to check the damage to the submerged tethers. To solve the challenges in the damage estimation of the SFT tethers, this paper suggests an innovative approach based on a deep-learning algorithm. Based on the convolutional neural network (CNN) algorithm, which is one of the most widely used deep learning algorithms in many fields, a damage detection model is proposed. To train and validate the proposed CNN-based damage detection algorithm, we simulated the structural response data of the SFT models with intact and damaged tethers using hydrodynamic analysis under waves with various incident directions. In the simulation to obtain the structural response data for training and validation, the damage to the tethers is reflected by the decrease in their structural stiffness. The test results clearly indicated that the suggested strategy is very effective at detecting damage to the tethers, although a limited measured response is input to the learned model.
