Research on the Application of an Offline Device Data Real-time Collection System Based on Visual Recognition and Large Language Model

To address the issue of outdated manufacturing equipment lacking internet connectivity and failing to meet real-time data collection requirements for on-site device management, quality control, and digital systems, a visual recognition and large language model—based offline device data real-time collection system is proposed and designed. The system adopts a five-layer architecture, including hardware collection, OCR recognition, intelligent parsing, data access, and business application layers. Field validation demonstrates its strong stability and accuracy, enabling real-time high-frequency offline device operation data collection and anomaly closed-loop handling at low cost without modifying the original equipment's electrical structure. It provides a reusable engineering paradigm for equipment reuse and intelligent upgrading in the manufacturing industry.