ABSTRACT
Utility tunnels in coastal areas are susceptible to multiple deterioration factors, including ion erosion from groundwater and soil-water pressure, which can lead to material degradation, structural damage, and even failure. Furthermore, in utility tunnels, monitoring the service conditions of external structural walls presents significant challenges, while maintenance operations typically necessitate excavation, resulting in substantial environmental impacts. It is difficult to meet the requirements of dynamic changes in the service state of infrastructures when using traditional optimization methods for maintenance strategies. For instance, multi-objective frameworks employing Genetic Algorithms (GAs) face inherent limitations in addressing environmental uncertainties and conflicting objectives. In contrast, the Deep Q-Learning Network (DQN) method provides a dynamic and adaptive solution. In view of this status, an optimization framework of maintenance strategies based on the DQN algorithm is proposed to meet the functional requirements of utility tunnels throughout their lifecycle while minimizing total economic costs and environmental impacts.
