HYBRID ML-ANALYTICAL MODELLING FOR PREDICTING MARINE CORROSION IN COMPLEX ENVIRONMENTS

The study develops a novel hybrid model for predicting marine corrosion by combining a physics-based analytical model with a data-driven machine learning model.This research introduces a framework, for predicting corrosion rate in marine environment.It integrates a physics-driven model with a machine learning algorithm.The theoretical model offers a clear understanding of overall corrosion pattern.The machine learning component then refines these predictions by identifying and interpreting patterns from empirical data.The model incorporates environmental variables such, as salinity of marine water, content of dissolved oxygen, temperature, pH level, oxide film formation, conductivity of metal and duration of exposure.Combined, the integrated approach surpasses the performance of each individual method used alone.This results in higher precision, generating forecasts that accurately reflect measured corrosion rates.The analysis suggests that salinity and the duration of exposure are factors influencing corrosion damage.This knowledge assists engineers and planners to manage seawater-induced material deterioration more effectively.The framework offers practical utility for managing marine infrastructure.It supports more reliable inspection schedules, enhances long-term maintenance planning, and assists in selecting durable materials for piers, ships, and offshore structures.Ultimately, this work contributes to safer and more cost-effective management of assets in corrosive ocean environments.