Corrosion Risk Assessment in Coastal Environments Using Machine Learning-Based Predictive Models.

Abstract

Atmospheric corrosion, especially in coastal environments, presents a major challenge for the long-term durability of metallic and concrete infrastructure due to chloride deposition from marine aerosols. With a significant portion of the global population residing in coastal zones-often associated with intense industrial activity-there is growing demand for accurate and early corrosion prediction methods. Traditional standards for assessing atmospheric corrosivity depend on long-term empirical data, limiting their usefulness during the design stage of infrastructure projects. To address this limitation, this study develops predictive models using machine-learning techniques, namely gradient boosting, support vector machine, and neural networks, to estimate chloride deposition levels based on easily accessible climatic and geographical parameters. Our models were trained on a comprehensive dataset that included variables such as land coverage, wind speed, and orientation. Among the models tested, tree-based algorithms, particularly gradient boosting, provided the highest prediction accuracy (F1 score: 0.8673). This approach not only highlights the most influential environmental variables driving chloride deposition but also offers a scalable and cost-effective solution to support corrosion monitoring and structural life assessment in coastal infrastructure.