Knowledge-Enhanced Machine Learning for Process Monitoring

Process manufacturers face mounting pressures from increasing global demand, tightening environmental regulations, and rising operational costs. These challenges demand robust process monitoring, particularly fault detection and diagnosis (FDD), to minimize downtime and waste while maintaining product quality and regulatory compliance. As experienced practitioners retire, preserving and leveraging their expertise becomes vital for effective FDD. The advent of Industry 4.0 has led to the proliferation of manufacturing data that could aid in rapid decision-making using advanced statistical methods such as machine learning. However, current machine learning approaches for FDD struggle to balance scalability with the interpretability needed for operational insights. There is a significant research gap for frameworks that can effectively combine data-driven scalability with domain expertise to maintain model interpretability for operational decision-making.