Digital twin system for manufacturing processes based on a multi-layer knowledge graph model

Digital twin technology in the manufacturing process faces challenges like integrating diverse data sources and managing real-time data flow. To address this, we propose a novel three-layer knowledge graph architecture to enhance digital twin modeling for manufacturing processes. This architecture consists of a concept layer that structures key information into a knowledge network, a model layer that aligns digital and physical parameters, and a decision layer that leverages model and real-time data for decision support. Validated in aero-engine blade production, this system integrates multi-source data, enhances predictive analysis and anomaly detection, and supports process control and quality management. Over a 5-month validation period, the maximum contour error precision of the blades improved from 0.073 mm to 0.062 mm, and the product qualification rate increased from 81.3% to 85.2%. This demonstrates the system's robust capability for advancing digital twin utilization in manufacturing, highlighting its potential for future improvements.