上位机工艺与仪表流程图智能识别方法研究  

Research on Intelligent Recognition Method of Process and Instrumentation Diagram on Upper Computer

作　　者：姜碧波 曲波 张超 JIANG Bibo;QU Bo;ZHANG Chao(CNOOC Enertech Equipment Technology Co.,Ltd.,Tianjin 300452,China)

机构地区：[1]中海油能源发展装备技术有限公司,天津300452

出　　处：《仪表技术》2025年第2期48-51,86,共5页Instrumentation Technology

摘　　要：在控制系统工程项目中,上位机工艺与仪表流程图(P&ID)组态是整体项目的前置关键环节,对项目完整性至关重要。传统组态方式依赖工程师手动绘制,存在工作烦琐、任务量大、易出错等问题。提出一种基于深度学习的P&ID智能识别方法,旨在实现上位机P&ID组态的智能自动生成。该方法采用YOLOv5模型检测图元小目标,利用PP-OCRv3提升文字识别准确性,并通过U-Net模型进行连通域检测,语义分割出连接线,确定图元与主接线的连接关系。实验结果显示,该方法对P&ID的整体识别准确率超过90%,有效应用于图纸元器件识别及连接线对应检测,显著提升工作效率。In control system engineering projects,the configuration of the upper computer process and instrumenta-tion diagram(P&ID)is a crucial prerequisite for the overall project and is essential for project integrity.The traditional configuration method relies on manual drawing by engineers,which has problems such as tedious work,heavy workload,and easy errors.A deep learning based intelligent recognition method is proposed for P&ID,aiming to achieve intelligent automatic generation of P&ID diagram configuration on the upper computer.This method uses the YOLOv5 model to de-tect small targets in graphic elements,utilizes PP-OCRv3 to improve text recognition accuracy,and uses the U-Net mod-el for connected domain detection,semantic segmentation of connection lines,and determination of the connection rela-tionship between graphic elements and main wiring.The experimental results show that the overall recognition accuracy of this method for P&ID exceeds 90%,and it is effectively used for component recognition and corresponding detection of connection lines in drawings,significantly improving work efficiency.

关 键 词：上位机 工艺与仪表流程图 图元识别 文字识别 连接线分割 深度学习 

分 类 号：TH122[机械工程—机械设计及理论]