RESEARCH ON THE APPLICATION OF MACHINE LEARNING AND EXPLAINABLE AI IN QUALITY DATA ANALYSIS FOR MANUFACTURING PROCESS IMPROVEMENT

Abstract

This study proposes a machine-learning-based approach for detecting and analyzing steel surface defects combined with explainable artifcial intelligence (XAI). The Steel Plate Defects dataset is used to train fve ML models: Logistic Regression, Decision Tree, Random Forest, XGBoost, and LightGBM. Experimental results show that boosting models (XGBoost and LightGBM) achieve superior performance with high F1-scores and AUC values across most defect types. SHAP and LIME are applied to interpret model decisions, revealing key features associated with each defect category. The proposed framework enhances transparency and supports intelligent quality control for steel surfaces in Industry 4.0 environments.