<![CDATA[The Industry 4.0 era demands a transformation of machine maintenance systems from reactive to predictive to maintain production efficiency. This research focuses on Flexographic Printing machines in the packaging industry, which are experiencing low performance with an average Overall Equipment Effectiveness (OEE) value of only 49.78%, far below global standards. The objective of this research is to enhance maintenance effectiveness through the development of a Smart Maintenance framework named "FlexoTwin". The research methodology integrates quality diagnostic approaches (Fishbone Diagram & FMEA) with Digital Twin technology. The initial diagnostic stage focused on comprehensive risk mitigation, not only based on the highest Risk Priority Number (RPN) in the Feeder unit (336) but also prioritizing the handling of failure modes with critical Severity levels across all machine components. These findings served as the basis for developing a Machine Learning (Random Forest) model integrated into a web-based dashboard. Test results show that the model is capable of predicting maintenance duration with high accuracy ( =0.83). Furthermore, optimization simulations established a machine health threshold at the 40% level as the trigger point for automatic warnings (Auto-Prediction Trigger). The implementation of "FlexoTwin" has proven capable of providing real-time condition visualization and data-driven recommendations, enabling a transition to a more precise and efficient maintenance strategy compared to manual methods.]]>
