Automated Paperboard Coating Defect Detection and Predictive Maintenance via Multi-Modal Data Fusion and Deep Reinforcement Learning

**Abstract:** The paperboard industry suffers significant economic losses due to coating defects, often detected late in the production process. This research proposes a novel system – Coating Anomaly Predictive Engine (CAPE) – that leverages multi-modal data fusion (visual, spectroscopic, and process data) coupled with Deep Reinforcement Learning (DRL) to achieve real-time defect detection and predictive maintenance. CAPE achieves a 10-fold improvement in early defect detection and a 25% reduction in downtime. This approach directly addresses production inefficiencies and promotes sustainability within the paperboard manufacturing sector.