A Framework for Predicting Alloy Corrosion Resistance Using Machine Learning Models

Thesis Abstract

Corrosion of metallic alloys poses significant challenges in various industrial applications, leading to material degradation, increased maintenance costs, and potential structural failures. Despite extensive research on corrosion mechanisms, predicting an alloy’s resistance to corrosion remains complex due to the multifactorial nature of corrosion processes influenced by material composition, environmental conditions, and operational parameters. This study aims to develop a comprehensive predictive framework utilizing machine learning models to accurately estimate alloy corrosion resistance, thereby enabling proactive material selection and performance assessment. The specific objectives include identifying relevant features influencing corrosion, evaluating the performance of various supervised machine learning algorithms, and formulating an integrated predictive model suitable for practical deployment. Employing a quantitative research design, this study focuses on a diverse dataset comprising experimental corrosion resistance data of 350 alloy samples obtained from industrial testing facilities and published literature. The population primarily includes nickel-based, stainless steel, and aluminum alloys subjected to corrosive environments such as saline, acidic, and marine conditions. Stratified random sampling ensures proportional representation across alloy categories and environmental conditions. Data collection involves collating parameters such as chemical compositions, microstructural characteristics, environmental exposure variables, and corrosion resistance metrics, primarily weight loss and electrochemical impedance measurements. Data pre-processing involves normalization, missing data imputation, and feature extraction to optimize model input. Instrumentation includes standardized corrosion testing protocols, complemented by digital databases and material property repositories. Model development employs a range of machine learning algorithms, including Random Forest, Support Vector Machines, and Gradient Boosting Machines. These models are trained and validated using k-fold cross-validation strategies to prevent overfitting and ensure robustness. The study applies statistical techniques such as Analysis of Variance (ANOVA) for feature significance testing and metrics like R-squared, Root Mean Square Error (RMSE), and F1 score for model performance evaluation. Additionally, feature importance analysis aids in elucidating the relative influence of predictors, guiding the interpretation of corrosion resistance determinants. Ethical considerations center on data confidentiality, proper attribution of sources, and adherence to standards prescribed by the institutional review board. Expected findings indicate that machine learning models can significantly outperform traditional empirical models in predicting alloy corrosion resistance, with Random Forest and Gradient Boosting Machines anticipated to demonstrate superior accuracy due to their ability to handle high-dimensional data and nonlinear relationships. The study is anticipated to identify key predictors—such as alloying element concentrations, microstructural parameters, and environmental variables—that substantially influence corrosion behavior. The resulting predictive framework will be validated through external datasets and case studies, demonstrating its applicability for industrial corrosion management. Furthermore, the research contributes to theoretical understanding by integrating material science principles with advanced data analytics, extending the conceptual basis of corrosion prediction through a data-driven perspective grounded in the Theory of Material Degradation and Systems Theory. The main conclusion emphasizes that the proposed machine learning-based framework offers a reliable, scalable, and cost-effective approach for predicting alloy corrosion resistance, facilitating informed decision-making in materials engineering and corrosion prevention. Recommendations include integrating the framework into existing corrosion monitoring systems, expanding the dataset to encompass a broader spectrum of alloys and environmental conditions, and enhancing model interpretability to foster wider industrial adoption. The study ultimately advances the field of corrosion science by demonstrating the potential of artificial intelligence techniques to address complex metallurgical challenges, laying the groundwork for further research into real-time predictive analytics and adaptive corrosion management strategies.