PREDICTING THE EVOLUTION OF THE CORROSION INTERFACE IN MILD STEEL

Abstract

Pipeline transportation is acknowledged as a cost-effective solution in the oil and gas industry. However, its development is limited by pipeline failure due to corrosion, specifically in countries like Singapore with aggravating meteorological factors. This thesis is part of an overarching project aimed at improving the maintenance of civil infrastructures by using drone images to identify potential areas at risk of causing failure. To address this issue, we first developed models capable of describing changes in the corroded interface depending on specific corrosion parameters. We introduced two different models: the isotropic model and the more realistic anisotropic model, both of which demonstrated a strong correlation with experimental data from existing literature. Using datasets derived from these models, we proceeded to train a machine-learning model that could predict corrosion parameters using historical corrosion data. The ultimate goal is to use these predicted parameters to model the future progression of corrosion in mild steel. In parallel, we also defined a methodology for collecting experimental data for future use as it will help enhance the accuracy of our machine-learning model.