Abstract:

Reinforced concrete in submarine tunnels is subject to synergistic corrosion driven by chloride ingress and oxygen concentration gradients, which compromises structural durability and safety. However, a long-term, stable oxygen concentration monitoring technique applicable to the concrete interior remains unavailable. This study presents an oxygen sensor specifically designed for the complex marine engineering environment, featuring a multi-electrode array, a triple-layer oxygen-permeable membrane, and a composite gel electrolyte, which collectively enhance stability and environmental adaptability. A potential-step polarization method was employed to establish the mapping relationship between oxygen concentration and current density. Comparative analysis of three electrode configurations identified the optimal arrangement of a stainless-steel working electrode, titanium-mesh reference electrode, and stainless-steel counter electrode. To address the influence of perturbations such as temperature, pH, and salinity, the sensor stability dataset was trained using an AdaBoost ensemble learning framework coupled with SHAP-based interpretability analysis, thereby developing a predictive model linking current density to oxygen content under complex and dynamic environmental conditions. The proposed approach enables real-time oxygen content prediction and provides a technical basis for corrosion risk assessment of reinforced concrete in submarine tunnels.

Key words: submarine tunnel, reinforced concrete, differential oxygen corrosion, machine learning, oxygen sensor