针对建筑设施人工巡检方式存在效率低下、成本高昂、安全隐患等问题，提出了一种基于软件机器人(Robotic Process Automation， RPA)技术的建筑设施虚拟巡检系统方案。首先，基于BIM数据构建动态的数字孪生建筑模型更新算法，与现场设施运行状态形成实时虚实映射关系；然后，设计多模态深度学习设备缺陷检测网络，对设备缺陷实现自动高精度检测；最后，基于软件机器人技术架构实现异构系统间的无缝集成和智能调度算法的运用。本文研发的一体化软件机器人巡检系统，在实验场景中进行了初步验证。结果表明，基于软件机器人技术架构的虚拟巡检机器人系统在设施缺陷检测精度达到97%以上，优于单模态检测方法；作业效率较传统方式平均提高了62.3%以上；运维人力需求则降低至60%-80%，具有一定的工程应用价值。

Aiming at the problems of low efficiency， high cost， and safety hazards in the manual inspection of building facilities， this study proposes a virtual inspection system scheme based on software robotics (RPA) technology. First， a dynamic digital twin building model and update algorithm based on BIM data are constructed， forming a real-time mapping relationship with the on-site facility operation status. Then， a multi-modal deep learning-based device defect detection network is designed to achieve automatic high-precision defect detection. Finally， the heterogeneous system integration and the application of intelligent scheduling algorithms are realized based on the software robotics technology architecture， and an integrated software robotics inspection system is developed. Preliminary verification was carried out in the experimental scenarios. The results show that the virtual inspection robot system based on the software robotics technology architecture has: the facility defect detection accuracy improved to over 97%， surpassing single-modal detection methods； the operation efficiency is increased by an average of more than 62.3% compared to traditional methods；the maintenance manpower demand is reduced to 60%-80%， demonstrating good engineering application value.

上海市