Development of AI-based Quality Prediction System for Metal Alloy Manufacturing

Thesis Abstract

The quality control of metal alloy manufacturing processes remains a critical challenge due to the complex interplay of process parameters and material properties that influence product consistency and performance. Traditional quality assessment methods often rely on post-production testing, which can be time-consuming, costly, and insufficient for real-time guidance and decision-making. This study aims to develop an AI-based quality prediction system capable of forecasting the quality attributes of metal alloys during manufacturing, thereby enabling proactive process adjustments to enhance product consistency and reduce defects. The specific objectives include identifying key process variables affecting alloy quality, designing machine learning models for accurate prediction, and evaluating the system’s performance in a real manufacturing environment. The research adopts a quantitative, exploratory, and developmental research design grounded in the positivist paradigm, emphasizing empirical measurement and predictive modeling. The population comprises data collected from a sheet metal alloy manufacturing plant with a production capacity of approximately 500 tons per month, involving various process stages such as melting, casting, and heat treatment. A stratified random sampling technique was employed to select a sample of 250 production batches, ensuring representation across different alloy types and production shifts. Data collection involved gathering historical process parameter datasets, including temperature, chemical composition, cooling rate, and mechanical properties, through an automated data acquisition system. Additionally, quality assessment reports detailing tensile strength, hardness, and surface finish were recorded for each batch. Data analysis incorporated preprocessing and feature engineering to extract relevant variables, followed by the application of supervised machine learning algorithms such as multiple linear regression, support vector regression, and random forest regression. The models' predictive accuracy was evaluated using metrics including mean absolute error (MAE), root mean square error (RMSE), and R-squared values, with k-fold cross-validation to ensure robustness. The study also applies the Theory of Planned Behavior to interpret how real-time AI feedback could influence operator decision-making, and the Systems Theory to conceptualize the integration of AI within the manufacturing process as part of an interconnected system. It is anticipated that the developed AI models will demonstrate high predictive accuracy, with R-squared values exceeding 0.85, facilitating reliable quality forecasting during manufacturing. The system is expected to identify critical process variables influencing alloy quality and provide real-time alerts and recommendations for process adjustments. This integration aims to reduce variability, decrease rework costs, and improve overall product quality. The study’s findings are poised to contribute novel insights into the application of machine learning techniques for metallurgical process control, filling gaps identified in prior literature which primarily focused on post-production quality assessment rather than real-time prediction. By implementing a scalable, real-time AI-based system, the research advances knowledge on the interface between manufacturing process control and intelligent systems, promoting Industry 4.0 initiatives in metallurgical engineering. The key contributions include the formulation of an effective predictive model tailored to alloy manufacturing, validation of AI as a decision-support tool, and a framework for integrating data-driven quality control systems into existing production lines. The main conclusion emphasizes that AI-driven predictive analytics can significantly enhance process stability and product quality in metal alloy manufacturing. Recommendations include extending the study to incorporate adaptive learning algorithms for continuous model improvement and exploring the integration of IoT sensors for more granular data collection. Future research should also consider the economic feasibility of deploying such systems at an industrial scale and assess their impact on workforce skills and operational workflows. Overall, this thesis establishes a robust foundation for adopting AI-based solutions to transform quality assurance practices within metallurgical manufacturing environments.