Structural health monitoring and evaluation method for an immersed tunnel based on deep learning.

The health monitoring of the subsea-immersed tunnels is essential for the early detection of anomalies and the assurance of their long-term operational safety. This research examines sensor data to evaluate variations in critical parameters and their effects on structural integrity. It also compares the efficacy of two deep learning algorithms, Long Short-Term Memory (LSTM) and Transformer, in predicting structural conditions. The findings indicate that the Transformer model exhibits high reliability in forecasting and is particularly adept at handling extended periods and intricate time series data, rendering it especially appropriate for health assessments of immersed tube structures. Three correlation analysis techniques are employed to calculate correlation coefficients, thereby identifying the parameters that exert the most significant influence on structural health. The CRITIC method is utilized to assign weights to these parameters. Subsequently, a structural health evaluation model based on the fusion of multi-source information is proposed, facilitating the assessment of the immersed tube's condition and enabling predictions regarding its status over the subsequent 12 to 24 h. This proactive methodology offers a comprehensive insight into the tunnel health, allowing for implementing preventive measures before the emergence of issues and ultimately reducing maintenance costs. The practical implications of this research are significant, as it provides a robust methodology for the real-time monitoring and evaluation of subsea-immersed tunnels, thereby enhancing their operational safety and reducing maintenance costs.