A Framework for Predictive Modeling of Alloy Corrosion Resistance in Marine Environments

Thesis Abstract

Marine environments pose significant challenges to the durability and integrity of metallic alloys due to their aggressive corrosion conditions, which are exacerbated by saline water, fluctuating temperatures, and biological activity. Despite advances in alloy development, there remains a critical need for predictive tools that can reliably assess corrosion resistance and guide material selection for maritime structures, offshore installations, and marine transportation equipment. This study aims to develop an integrated framework for the predictive modeling of alloy corrosion resistance in marine environments, thereby enhancing the reliability of longevity assessments and maintenance planning. The specific objectives are to (1) identify primary factors influencing alloy corrosion resistance through comprehensive literature review and experimental analysis, (2) quantify relationships between key material properties and corrosion behavior using advanced statistical techniques, (3) formulate a predictive model grounded in machine learning algorithms, particularly multiple linear regression and support vector machines, and (4) validate the model through empirical testing on a representative sample of marine-grade alloys. The research adopts a mixed-methods approach, combining quantitative experimental investigations with qualitative analysis. The study population includes fifty-six marine-grade alloy samples, selected via stratified random sampling from commercially available stainless steels, aluminum alloys, and superalloys commonly used in marine applications. Data collection involves corrosion testing protocols such as salt spray testing, electrochemical impedance spectroscopy (EIS), and potentiodynamic polarization (PDP). The parameters measured include corrosion rate, surface roughness, passivation potential, and microstructural characteristics obtained through scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy (EDX). Complementary data on environmental variables—salinity, temperature, pH—are recorded using in situ sensors. Validity and reliability of the measurement instruments are confirmed through calibration, standardization, and repeated testing, ensuring coefficient of variation remains below 5%. Data analysis involves descriptive statistical techniques to characterize corrosion behavior across materials. Inferential analysis employs multiple linear regression and support vector regression to establish relationships between independent variables (material composition, surface treatments, environmental factors) and dependent variables (corrosion rate, passivation stability). Feature selection is performed via principal component analysis (PCA) to optimize model performance. Model accuracy and robustness are evaluated using cross-validation, root mean square error (RMSE), and coefficient of determination (R²). The theoretical foundation is anchored in the Electrochemical and Microstructural Theories of Corrosion, complemented by frameworks derived from the Theory of Material Degradation and Machine Learning models. Expected findings include statistically significant correlations between alloy composition, surface treatment protocols, environmental factors, and corrosion resistance. The predictive model is anticipated to achieve an R² exceeding 0.85, demonstrating high explanatory power. It will enable accurate predictions of corrosion behavior under varying marine conditions, supporting proactive maintenance and material selection decision-making. The study contributes to knowledge by integrating materials science, electrochemistry, and machine learning to create a comprehensive, data-driven predictive framework adaptable across diverse marine contexts. The main conclusion advocates for the adoption of the developed modeling framework by industry stakeholders to mitigate corrosion-related failures, reduce maintenance costs, and extend service life of marine structures. Recommendations include further refinement of the model incorporating real-time sensor data, expansion to include biofouling effects, and tailoring the framework to emerging alloy development trends. Future research directions suggest exploring deep learning approaches and incorporating lifecycle assessment models for holistic corrosion management in marine environments.