Deep Learning Method for Real-time Risk Assessment of Tunnel Engineering

Tunnel engineering is essential to the security and longevity of transportation infrastructure. An effective tunnel is essential for identifying potential risks, such as structural damage, and mitigating these risks to ensure tunnel safety. This research proposes a novel Deep Learning (DL)-based method for real-time risk assessment in tunnel engineering, focusing on the automated detection of structural hazards during tunnel construction. The proposed approach’s core is the Dynamic Sailfish Optimizer-driven Intelligent Faster Region-based Convolutional Neural Network (DSO-INT-Faster R-CNN), to increase danger detection’s precision and quickness in real-time. A diverse dataset consisting of risk assessment for tunnel construction. A data pre-processing technique using min-max normalization and the Principal Component Analysis (PCA) technique is utilized for feature extraction, and the DSO is used to improve optimization and fine-tune feature selection. The DSO-INT-Faster R-CNN model significantly enhances real-time performance by accurately identifying and assessing potential risk factors in tunnels, including construction defects, water seepage, and surface irregularities. This enables more timely and precise maintenance actions, reducing the risk of severe structural failure. The proposed DL method offers a powerful tool for real-time risk assessment in tunnel engineering, providing a foundation for more efficient and proactive tunnel maintenance. The findings indicate significant gains in learning outcomes, with the suggested model greater than conventional techniques in terms of MAE (0.32), MAPE (0.220), [Formula: see text](0.60), RMSE (0.60), minimum error (4.632[Formula: see text]mm), maximum error (0.325[Formula: see text]mm), and Time (10.323[Formula: see text]s). This approach can significantly contribute to improving safety and longevity in tunnel infrastructure.