Theory and method of intelligent safety design based on the Internet Of Things
基于物联网的智能安全设计理论与方法研究

摘要

围绕选煤厂供配电系统接入工业物联网后安全边界扩大、信息侧异常可能传导至物理执行端的问题，结合智能安全供配电工程实践，对物联网条件下的安全设计对象与实施方法进行了重新梳理。研究把人员、设备、环境、管理以及网络/数据纳入同一风险链，以本质安全、纵深防御、失效安全、过程可追溯和故障恢复为约束，形成从危险场景识别、需求分解、多源感知、可信判断到安全联锁、物理隔离和运行复核的设计流程。工程系统采用规则判断与随机森林复验相结合的方式，并通过软闭锁、硬联锁和物理隔离限制高风险操作。现场应用中，身份识别精度达到100%，巡检机器人识别精度不低于96%，系统无故障运行时间不低于99.9%，停送电操作耗时缩短40%以上；非计划停电识别准确率为92.5%，平均响应时间不超过4.2min，次生故障率下降70%。结果显示，物联网智能安全设计的关键并非增加监测设备数量，而是使风险识别、数据校验、决策、执行和恢复形成可验证的闭环。相关思路可为能源与矿山工业智能化系统的安全设计及工程验收提供参考。

Abstract

To address the limitations of conventional safety design when industrial Internet of Things (IIoT) extends safety objects from individual equipment to a coupled human-machine-environment-management-network/data system, a five-domain collaborative theory and a lifecycle design method are proposed based on an intelligent safe power-distribution project in a coal preparation plant. Inherent safety, defense in depth, fail-safe behavior, traceability and resilience are taken as fundamental principles. A design chain of hazard identification, safety-requirement mapping, multi-source sensing, trusted decision-making, safety interlocking, physical isolation, recovery validation and iterative optimization is established. Hierarchical decision-making combining rule engines and machine learning, coordination of software blocking and hardware interlocking, and a dual closed loop between the information and physical domains are introduced. Evaluation dimensions include sensing integrity, diagnostic accuracy, control reliability, response timeliness, environmental adaptability, traceability and system resilience. Engineering results show 100% identity-recognition accuracy, robot-recognition accuracy not lower than 96%, system availability not lower than 99.9%, and more than 40% reduction in power-switching operation time. The accuracy of unplanned-outage identification reached 92.5%, the average response time was within 4.2 min, and secondary faults decreased by 70%. The results indicate that IIoT-based intelligent safety design should shift from adding monitoring devices to integrated risk-data-decision-control-recovery design, with authorization, interlocking, physical isolation and fail-safe degradation implemented together with network connectivity.