Abstract: A most dangerous and weak typical cross-section of a river-crossing tunnel in Jiangsu, China, is selected and arranged with fiber Bragg grating sensors. Sequentially, a tunnel structural health monitoring system is established to collect the deformation patterns of underwater shield tunnels in real-time, identifying the health status of tunnel structures. Preprocessing treatments, such as identification, elimination, and completion, are implemented for abnormal data in health monitoring data. Then, association analysis is conducted on monitoring data based on temperature, soil pressure, and water pressure, revealing the influencing patterns of such factors on tunnel structural deformation. Finally, the safety of tunnel structure is evaluated by taking the convergence deformation of segment and steel stress as evaluation indices. The structural safety evaluation results are validated by calculating the structural safety factors. The results reveal the following: (1) PauTa criterion and box plot can identify the abnormal data effectively, and the abnormal data can pass the test of estimation domain method. (2) The grey correlation analysis results of soil pressure and water pressure respectively show that the reinforcement stress is considerably affected by soil pressure, and the deformation of pipe segments is more significantly affected by water pressure. (3) The safety performance of a monitoring cross-section is evaluated by taking the convergence deformation of the segment and the stress of the reinforcement as evaluation indices. The evaluation results show that the structural safety is Level 1 with intact structure. The safety factor of the most dangerous segment structure calculated by the bending moment and axial force of the structure is 4.20, demonstrating an accurate evaluation result.

Key words: underwater shield tunnel; structural health monitoring, fiber Bragg grating sensor, data preprocessing; association analysis, safety evaluation