Self-learning Time-varying Digital Twin System for Intelligent Monitoring of Automatic Production Line

Abstract

At present, the automation production line has problems such as insufficient intelligence level. The intelligent monitoring, control and improvement of product quality and efficiency are the key common technologies faced by advanced manufacturing industry. Self-learning time varying digital twin (DT) system for intelligent monitoring is proposed in the paper. In the process of automatic production line processing and workpiece detection, an DT consisting of physical production line layer, edge monitoring layer and cloud evolution layer is built. The DT system realizes self-learning time-varying through active excitation of processing parameter optimization. The workpiece quality is a real-time representation of the tool condition, and the tool wear sensitive features extracted by the deep learning algorithm. Through the two-way drive of time-varying physical and virtual data, the tool wear characterization model can be evaluated, self-learning, updated and verified timely in the light of the actual condition to achieve tool condition monitoring and processing parameter optimization. The prediction model is self-iterative and simplified in the cloud, and the edge side is quickly matched and adaptive. Self-learning time-varying DT system based on self-driving of manufacturing process can adaptively improve the ability of intelligent monitoring.