Machine learning based corrosion prediction in reinforced concrete structures exposed to marine environment

Abstract

Structural health monitoring is the cornerstone in determining the longevity and the safety of Reinforced Concrete (RC) structures, constructed in harsh marine environments. RC structures in marine environments are highly exposed and experience corrosion, which is a major threat. Numerous methods have been developed to predict the existence of corrosion in RC structures in harsh environments. RC structures in the marine environment deteriorated owing to the high concentration of chloride ions and humidity. Traditional Machine Learning (ML) models for corrosion monitoring fall short when implemented in real-time and produce reduced accuracy and precision. To overcome this challenge, this proposed work introduces a novel cutting-edge approach leveraging the ML approach, namely, Neuro Graph Temporal Fusion Network (NGTFNet), to predict the corrosion in RC structures with an enhanced level of reliability and accuracy. The proposed work employs multimodal images combining thermal, optical, and ultrasonic images to deliver predictive insights into corrosion initiation and progression. The proposed research work marks a significant leap towards the realization of smart, resilient infrastructure, capable of withstanding the challenges of an evolving marine environment. The performance of the proposed method was compared with state-of-the-art methodologies for detecting corrosion in RC structures located near marine environments.