Sustainable fault detection and process simulation in semiconductor manufacturing using machine learning and life cycle assessment

Abstract

As digital transformation advances across various industries, the growing demand for semiconductor manufacturing calls for developing methodologies that enhance production efficiency while minimizing environmental impact. Although various methodologies have demonstrated significant potential in fault detection and process optimization, existing approaches mainly focus on improving defect reduction without systematically incorporating the impacts on sustainability considerations. This study proposes an integrated framework that synergistically combines fault detection, discrete event simulation (DES), and life cycle assessment (LCA) to address both operational efficiency and sustainability in the photolithography process. A machine learning (ML) model is developed for defect prediction. DES is then used by incorporating an inspection control mechanism informed by the defect prediction results, enabling the removal of defective wafers to reduce unnecessary processing and improve production efficiency, while LCA quantifies the corresponding environmental impact to enable sustainability evaluation aligned with industrial practices. Experiments are conducted to validate the effectiveness of the proposed framework, and the results demonstrate improvements in both operational efficiency and environmental footprint in semiconductor manufacturing.