| Čoahkkáigeassu: | Tunnels are crucial for transportation networks, necessitating regular inspection and structural deterioration evaluation to ensure their operational capacity. Previously, only traditional models have been developed to characterize the overall degradation of tunnels and with data shuffling the models trained ignore the utilization of historical data for future scenarios. To address the two gaps mentioned here, this study focuses on developing a data-driven ensemble learning strategy, covering both traditional and state-of-the-art deep learning-based models for future tunnel overall condition evaluation. Our strategy encompasses the latest data collection, preprocessing, model building on historical data, selection, and ensemble learning for assessing future cases. Hence, our model can accurately predict the overall condition of tunnels to optimize the maintenance and intervention plans. Compared to existing methods, our strategy improves accuracy with a more updated database, robustness against data imbalance, and generalizability to diverse tunnel databases and machine learning models.
|
|---|