Predictive Maintenance of Semiconductor Equipment Using Stacking Classifiers and Explainable AI: A Synthetic Data Approach for Fault Detection and Severity Classification

In the semiconductor manufacturing industry, predictive maintenance is a key strategy for reducing equipment downtime. This paper proposes a machine learning-based predictive model for semiconductor equipment operating conditions, employing a stacked classifier that focuses on fault detection and severity classification. The model leverages synthetic data generation to augment realistic fault data, thereby addressing the issue of failure incidence in industrial environments. Furthermore, this research introduces the artificial intelligence techniques SHAP and LIME to support actionable maintenance strategies. The importance of combining machine learning with industrial practices is highlighted, providing more efficient and reliable predictive maintenance for the semiconductor manufacturing sector.