Developing AI-Driven Predictive Models for Sustainable Steel Microstructure Optimization

Thesis Abstract

The steel industry faces significant challenges in balancing the demand for high-performance materials with sustainability objectives, particularly concerning the optimization of microstructure to enhance mechanical properties while minimizing environmental impact. Traditional methods of steel microstructure design rely heavily on empirical knowledge and trial-and-error experimentation, which are often time-consuming, costly, and inadequate for achieving optimal and sustainable outcomes. This study aims to develop robust, AI-driven predictive models that can accurately forecast steel microstructure characteristics based on processing parameters, thereby facilitating sustainable microstructure optimization. The specific objectives include identifying critical processing variables influencing microstructural features, constructing machine learning models capable of predicting microstructure attributes, validating these models through experimental data, and deriving guidelines for sustainable steel manufacturing practices. The research adopts a mixed-methods approach within a quantitative dominant paradigm, employing an experimental design complemented by computational modeling. The population comprises steel samples produced under controlled laboratory conditions, with a sample size of 150 steel specimens subjected to different thermal and mechanical processing parameters. Data collection instruments include high-resolution scanning electron microscopy (SEM), X-ray diffraction (XRD), and energy-dispersive X-ray spectroscopy (EDX) for detailed microstructural characterization, alongside processing parameter logs. Data analysis employs advanced supervised machine learning techniques, such as random forest regression, support vector machines (SVM), and artificial neural networks (ANN), alongside feature importance analysis and sensitivity testing to identify the most influential processing variables. The models will be trained and validated using 70% of the dataset, with the remaining 30% allocated for testing, employing k-fold cross-validation to prevent overfitting and ensure generalizability. Model performance will be evaluated through metrics including mean squared error (MSE), R-squared (R²), and receiver operating characteristic (ROC) curves where applicable. Furthermore, the study draws upon the Diffusion Theory of Innovation and the Theory of Planned Behavior to underpin the adoption of AI-based solutions within steel manufacturing contexts, providing a theoretical framework for predictive modeling and sustainable practice integration. The expected results include precise, data-driven insights into how specific thermal and mechanical processing parameters influence microstructure features such as grain size, phase distribution, and crystallographic orientation. It is anticipated that the AI models will demonstrate high predictive accuracy (R² > 0.85), offering a reliable tool for microstructure optimization. This research will contribute significantly to the body of knowledge by demonstrating the feasibility and effectiveness of artificial intelligence techniques in predicting and optimizing steel microstructures for sustainability objectives. It offers a novel integration of metallurgical science with advanced computational intelligence, providing a scalable methodology for industry-wide adoption. The study will also produce a comprehensive set of guidelines for sustainable steel processing, emphasizing environmentally friendly practices such as reduced energy consumption and minimized waste generation. In conclusion, the findings are expected to affirm that AI-driven predictive models can revolutionize microstructure design by reducing development turnaround times, lowering costs, and promoting sustainable manufacturing practices. It is recommended that steel producers incorporate these models into their process control systems to enhance microstructure quality and sustainability. Future research avenues include expanding the models to encompass different steel grades, incorporating real-time sensing data via IoT (Internet of Things) devices, and exploring multi-objective optimization algorithms to balance multiple performance and sustainability criteria. Overall, this study aims to bridge the gap between metallurgical innovation and sustainable manufacturing, facilitating environmentally conscious advancements in steel production through intelligent, predictive technologies.