摘要
Abstract
To promote digital transformation in the field of quality control,three main lines are focused on:"overcoming the fluctuation of human judgment,establishing digital evaluation standards,and unified analysis of massive data",a three-stage,three-agent model is proposed centered on deep reinforcement learning,namely"real-time judgment,hierarchical summarization,and trend prediction",to achieve real-time anomaly detection for in-process products and production processes.Under a hierarchical permission system,a high-precision virtual production line model is built to fully realize the visualization of the production line's operation status,process traceability,and process optimization,thereby promoting consistent quality control and continuous improvement of product quality.Engineering application analysis of the visual intelligent quality control system is conducted for the production process of low-complexity and high-value products.With ROI(Return on Investment)as the measurement standard for the quality control system's benefits,it has been verified that after the effective implementation of this model,the defect detection rate has increased from 75.3%to 96.8%,and the response time has been shortened from 68 s to 8.7 s.The system operates in a coordinated and efficient manner,balancing consistency,real-time performance,and predictability,and significantly reduces quality costs.关键词
质量监督/深度强化学习/可视化/数字化工厂Key words
quality supervision/deep reinforcement learning/visualization/digital factory分类
信息技术与安全科学
