ARTIFICIAL INTELLIGENCE-BASED DURABILITY ASSESSMENT OF REINFORCED CONCRETE STRUCTURES IN MARINE ENVIRONMENTS: A SYSTEMATIC LITERATURE REVIEW

Abstract

The durability assessment of reinforced concrete structures in marine environments is a critical challenge due to the aggressive action of chloride-induced corrosion and material degradation. This systematic literature review aims to synthesize and critically evaluate the state of the art in artificial intelligence applications for predicting and managing such durability issues. We conducted a comprehensive methodological search across major academic databases, followed by a structured screening process to identify relevant studies that apply machine learning, deep learning, or other AI techniques to durability-related problems. The review is organized around key thematic dimensions: AI-driven prediction of chloride ingress and corrosion initiation, machine learning for concrete mix optimization and durability enhancement, structural health monitoring and service life forecasting, and the prediction of mechanical properties under aggressive exposure. Our results reveal that artificial neural networks and ensemble methods have become dominant for modeling chloride diffusion and corrosion rates, often outperforming traditional empirical models. Conversely, we find that while mix optimization models show high accuracy in predicting compressive strength and permeability, their generalizability to field conditions remains limited. Structural health monitoring studies increasingly integrate sensor data with AI for real-time damage diagnosis, yet long-term validation data are scarce. In conclusion, this review identifies a growing consensus that AI provides powerful tools for durability assessment, but significant gaps persist regarding data scarcity, model interpretability, and the translation of laboratory findings to full-scale marine structures. We therefore highlight the need for standardized benchmark datasets and hybrid models that combine physics-based knowledge with data-driven learning to advance the field.