ENHANCED PRODUCTION QUALITY PREDICTION IN COLD ROLLING PROCESSES USING TABTRANSFORMER AND MACHINE LEARNING ALGORITHMS

Abstract

In this study, the impact of production parameters on product quality in cold rolling processes was examined, and the qualitative status of products was predicted using machine learning algorithms. While existing literature focuses on production efficiency, this study stands out by systematically comparing eight different machine learning algorithms: Decision Tree, KNN, Naive Bayes, Logistic Regression, Random Forest, XGBoost, Support Vector Machines, and TabTransformer. The results reveal that TabTransformer, a transformer-based model designed for tabular data, outperforms the other algorithms in terms of accuracy and generalization capability, making significant contributions to the automation of quality control in production processes. Additionally, feature importance analysis provides critical insights into parameter optimization, making this study a valuable addition to the literature on industrial quality prediction.